Manual of Grammatical Interfaces in Romance 9783110311785

Table of contents :
Preface......Page 5
Acknowledgments......Page 7
Table of contents......Page 9
Grammatical interfaces in Romance languages: An introduction......Page 11
I. Sound and structure......Page 31
1. Surface sound and underlying structure: The phonetics-phonology interface......Page 33
2. Segmental phenomena and their interactions: Evidence for prosodic organization and the architecture of grammar......Page 51
3. Prosodic phonology and its interfaces......Page 85
4. Phonology and morphology in Optimality Theory......Page 115
5. Inflectional verb morphology......Page 159
II. Structure and meaning......Page 195
6. Meaning of words and meaning of sentences......Page 197
7. Morphology and semantics: Aspect and modality......Page 223
8. (In)definiteness, specificity, and differential object marking......Page 251
9. Agreement restrictions and agreement oddities......Page 277
10. Auxiliary selection......Page 305
III. Sound, structure, and meaning......Page 337
11. Subjects, null subjects, and expletives......Page 339
12. Object clitics......Page 373
13. Nominalizations......Page 401
14. Information structure, prosody, and word order......Page 429
15 VP and TP ellipsis: Sentential polarity and information structure......Page 467
16. Existential constructions......Page 497
IV. The role of the interfaces in language acquisition and change......Page 527
17. Acquiring multilingual phonologies (2L1, L2 and L3): Are the difficulties in the interfaces?......Page 529
18. Interfaces with syntax in language acquisition......Page 561
19. The role of the interfaces in syntactic change......Page 597
20. Interfacing interfaces: Quechua and Spanish in the Andes......Page 617
21. Grammaticalization and pragmaticalization......Page 645
22. Changes at the syntax-discourse interface......Page 669
Index......Page 693

Citation preview

Manual of Grammatical Interfaces in Romance MRL 10

Manuals of Romance Linguistics Manuels de linguistique romane Manuali di linguistica romanza Manuales de lingüística románica

Edited by Günter Holtus and Fernando Sánchez Miret

Volume 10

Manual of Grammatical Interfaces in Romance

Edited by Susann Fischer and Christoph Gabriel

ISBN 978-3-11-031178-5 e-ISBN (PDF) 978-3-11-031186-0 e-ISBN (EPUB) 978-3-11-039483-2 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. 6 2016 Walter de Gruyter GmbH, Berlin/Boston Cover image: © Marco2811/fotolia Typesetting: RoyalStandard, Hong Kong Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Manuals of Romance Linguistics The new international handbook series Manuals of Romance Linguistics (MRL) will offer an extensive, systematic and state-of-the-art overview of linguistic research in the entire field of present-day Romance Studies. MRL aims to update and expand the contents of the two major reference works available to date: Lexikon der Romanistischen Linguistik (LRL) (1988–2005, vol. 1–8) and Romanische Sprachgeschichte (RSG) (2003–2008, vol. 1–3). It will also seek to integrate new research trends as well as topics that have not yet been explored systematically. Given that a complete revision of LRL and RSG would not be feasible, at least not in a sensible timeframe, the MRL editors have opted for a modular approach that is much more flexible: The series will include approximately 60 volumes (each comprised of approx. 400–600 pages and 15–30 chapters). Each volume will focus on the most central aspects of its topic in a clear and structured manner. As a series, the volumes will cover the entire field of present-day Romance Linguistics, but they can also be used individually. Given that the work on individual MRL volumes will be nowhere near as time-consuming as that on a major reference work in the style of LRL, it will be much easier to take into account even the most recent trends and developments in linguistic research. MRL’s languages of publication are French, Spanish, Italian, English and, in exceptional cases, Portuguese. Each volume will consistently be written in only one of these languages. In each case, the choice of language will depend on the specific topic. English will be used for topics that are of more general relevance beyond the field of Romance Studies (for example Manual of Language Acquisition or Manual of Romance Languages in the Media). The focus of each volume will be either (1) on one specific language or (2) on one specific research field. Concerning volumes of the first type, each of the Romance languages – including Romance-based creoles – will be discussed in a separate volume. A particularly strong focus will be placed on the smaller languages (linguae minores) that other reference works have not treated extensively. MRL will comprise volumes on Friulian, Corsican, Galician, Vulgar Latin, among others, as well as a Manual of Judaeo-Romance Linguistics and Philology. Volumes of the second type will be devoted to the systematic presentation of all traditional and new fields of Romance Linguistics, with the research methods of Romance Linguistics being discussed in a separate volume. Dynamic new research fields and trends will yet again be of particular interest, because although they have become increasingly important in both research and teaching, older reference works have not dealt with them at all or touched upon them only tangentially. MRL will feature volumes dedicated to research fields such as Grammatical Interfaces, Youth Language Research,

VI

Manuals of Romance Linguistics

Urban Varieties, Computational Linguistics, Neurolinguistics, Sign Languages or Forensic Linguistics. Each volume will offer a structured and informative, easy-to-read overview of the history of research as well as of recent research trends. We are delighted that internationally-renowned colleagues from a variety of Romance-speaking countries and beyond have agreed to collaborate on this series and take on the editorship of individual MRL volumes. Thanks to the expertise of the volume editors responsible for the concept and structure of their volumes, as well as for the selection of suitable authors, MRL will not only summarize the current state of knowledge in Romance Linguistics, but will also present much new information and recent research results. As a whole, the MRL series will present a panorama of the discipline that is both extensive and up-to-date, providing interesting and relevant information and useful orientation for every reader, with detailed coverage of specific topics as well as general overviews of present-day Romance Linguistics. We believe that the series will offer a fresh, innovative approach, suited to adequately map the constant advancement of our discipline. Günter Holtus (Lohra/Göttingen) Fernando Sánchez Miret (Salamanca) July 2016

Acknowledgments Editing a manual is a collective effort. High standards can only be met by drawing on the expertise of various scholars from many different linguistic disciplines, be it as authors of the individual chapters or as reviewers and consultants. The editors would like to express their gratitude to the many individuals who generously offered their time and expertise to improve the quality of the present volume. The reviewers are listed in alphabetical order in the following: Artemis Alexiadou, Elena Anagnostopoulou, Theresa Biberauer, Joanna Błaszczak, Ute Bohnacker, Martin Elsig, Anamaria Fălăuş, Cristina Maria Moreira Flores, Chiara Gianollo, Klaus von Heusinger, Mary Kato, Imme Kuchenbrandt, Tanja Kupisch, Winfried Lechner, Susanne Lohrmann, Mihaela Marchis, Thomas McFadden, Trudel Meisenburg, Fabio Montermini, Andrea Pešková, Florian Schäfer, Horst Simon, Carola Trips, Chiara Truppi, Maria del Mar Vanrell, Tonjes Veenstra, Jorge Vega Vilanova, Xavier Villalba and Marina Zielke. Thanks are also due to the following colleagues who have commented on earlier versions of individual chapters: Mathieu Avanzi (chap. 3), Francesco Maria Ciconte (chap. 16), Cristina Flores (chap. 18), Jonas Granfeldt (chap. 18), Klaus Grübl (chap. 14), Marc-Olivier Hinzelin (chap. 18), Dalina Kallulli (chap. 8), Marie Labelle (chap. 6), Clàudia Pons-Moll (chap. 4), Thomas Scharinger (chap. 14) and Hiyon Yoo (chap. 3). Many thanks go to the following native speakers for discussing the linguistic examples in some of the chapters: Fabián Santiago Vargas (chap. 3); Elena Ciutescu (chap. 6); Roberta D’Alessandro, Susana Barros, Adriana Fasanella, Jordi Fortuny, José Cruz da Ângela, Norma Schifano, Carmen Ríos García and Juan Quintanilla (chap. 11); Júlio Matias, Mario Navarro, Jacopo Torregrossa and Jorge Vega Vilanova (chap. 12); Ariadna Benet, Nicholas Catasso and Benjamin Massot (chap. 14). We also wish to thank Frédéric Aumaître, Sarah Jobus, Birgitta Pees and Liefka Würdemann for their assistance with the final proof reading and the cross-checking of references as well as Kirsten Brock and Derek Fobair for checking the language and for proof-reading some of the chapters. Finally, we would like to express our gratitude to the De Gruyter editorial team, Christine Henschel and Ulrike Krauß, for the constant support we received during the editing process, and to the series editors, Günter Holtus and Fernando Sánchez Miret, who provided us with most valuable feedback from the very first sketch of the volume until the submission of its final version. Susann Fischer (Hamburg) and Christoph Gabriel (Mainz) May 2016

Table of contents Preface V Acknowledgments

VII

Susann Fischer and Christoph Gabriel Grammatical interfaces in Romance languages: An introduction

1

I

Sound and structure

1

José Ignacio Hualde and Ioana Chitoran Surface sound and underlying structure: The phonetics-phonology 23 interface

2

Marina Vigário Segmental phenomena and their interactions: Evidence for prosodic 41 organization and the architecture of grammar

3

Élisabeth Delais-Roussarie Prosodic phonology and its interfaces

4

Eulàlia Bonet and Maria-Rosa Lloret Phonology and morphology in Optimality Theory

5

Sascha Gaglia and Marc-Olivier Hinzelin 149 Inflectional verb morphology

75

105

II Structure and meaning 6

M. Teresa Espinal and Susagna Tubau Meaning of words and meaning of sentences

7

Eva-Maria Remberger Morphology and semantics: Aspect and modality

8

Luis López (In)definiteness, specificity, and differential object marking

9

Roberta D’Alessandro and Diego Pescarini Agreement restrictions and agreement oddities

Jaume Mateu Fontanals 295 10 Auxiliary selection

187

213

267

241

X

Table of contents

III Sound, structure, and meaning 11

Michelle Sheehan Subjects, null subjects, and expletives

329

Susann Fischer and Maria Goldbach 363 12 Object clitics Judith Meinschaefer 391 13 Nominalizations Andreas Dufter and Christoph Gabriel 14 Information structure, prosody, and word order

419

Ana Maria Martins 15 VP and TP ellipsis: Sentential polarity and information structure

457

Delia Bentley and Silvio Cruschina 487 16 Existential constructions

IV The role of the interfaces in language acquisition and change Conxita Lleó 17 Acquiring multilingual phonologies (2L1, L2 and L3): Are the difficulties in the 519 interfaces? Tanja Kupisch and Jason Rothman 18 Interfaces with syntax in language acquisition Esther Rinke 19 The role of the interfaces in syntactic change

551

587

Pieter Muysken and Antje Muntendam 20 Interfacing interfaces: Quechua and Spanish in the Andes Ulrich Detges and Richard Waltereit 21 Grammaticalization and pragmaticalization Kristine Eide 22 Changes at the syntax-discourse interface Index

683

635

659

607

Susann Fischer and Christoph Gabriel

Grammatical interfaces in Romance languages: An introduction Abstract: It has been known for a long time that different components of grammar interact in nontrivial ways. However, in modular theories of grammar, like generative grammar, it has been under debate what the actual extent of interaction is and how we can most appropriately represent this in grammatical theory. Keywords: grammaticalization, Neogrammarians, Prague Linguistic Circle, Government and Binding Theory, Minimalist Program, Derivation by Phase, Optimality Theory

1 Interfaces and the architecture of grammar It is a well-known fact that linguistics started out as a historical discipline. The main focus for many centuries was the investigation of how language, the parts of speech, develop, thus how languages change. Language change in general and especially grammaticalization, the development of new grammatical material from lexical material, involves not only a single component of the grammar, but cuts across components, with contributory changes taking place in phonology, morphology, syntax, and semantics. Antoine Meillet is considered to be the first linguist to use the term “grammaticalization” in his article L’évolution des formes grammaticales and to define what is meant: “le passage d’un mot autonome au rôle d’élément grammatical” (‘the shift of an independent word to the status of a grammatical element’, Meillet 1912/1921, 131). However the idea that this kind of language change affects several components of grammar and that the interfaces between those components are relevant to explaining such changes is far older. Wilhelm von Humboldt in his lecture Über das Entstehen der grammatischen Formen und ihren Einfluß auf die Ideenentwicklung (‘On the genesis of grammatical forms and their influence on the evolution of ideas’, 1825) suggests that the grammatical structure of a language develops in four stages. He classifies these stages according to the strictness or freedom of word order (syntax) and the amount of grammatical material (morphology and phonology) used in expressing meaning (semantics; cf. von Humboldt 1825, 422–423), thus, in the interaction of the different components of grammar. Von Humboldt’s insights were subsequently summarized and further investigated by many prominent linguists, e.g. Franz Bopp, August Schleicher, and Georg von der Gabelentz to name only a few. Von der Gabelentz (1891/1901) is especially pathbreaking because he offers an explanation of why languages undergo changes and how components of grammar interact. He proposes that language change, i.e.

2

Susann Fischer and Christoph Gabriel

grammaticalization, is the result of two competing forces, one being the tendency towards ease of articulation, the other the tendency towards distinctness: lazy pronunciation brings about sound changes that wear down words, therefore distinctions consequently become blurred. To regain distinctiveness, word order (the strict syntactic ordering of words) or new forms take over the approximate function of the old forms. Von der Gabelentz observes that the new forms that have stepped in to take over the function of the old ones are also subject to the same processes (semantic bleaching and phonetic reduction) and will again be replaced, thus proposing that language change is cyclic. Under this view not only do components of grammar interact in nontrivial ways, but words, i.e. the different parts of speech, shift back and forth across components. In the 20th century, Hodge (1970, 3) used the slogan “one man’s morphology is yesterday’s syntax”, which was taken up and reformulated by Givón (1971, 413) as “today’s morphology is yesterday’s syntax”.1 Givón (1979, 109) suggests that the parts of speech are to be seen as being located on clines and as shifting between poles, such as child/adult, creole/standard, unplanned/planned, and pragmatic/syntactic, and furthermore suggests the following path: (1)

discourse > syntax > morphology > morphophonemics > zero

Grammaticalization theorists such as Lehmann (1982/2002), Kuryłowicz (1965/1975), Heine/Reh (1984), and Hopper/Traugott (1993), just to mention a few, have developed these ideas and, at the same time, draw on Meillet’s earlier insights (cf. also Heine/ Kuteva 2002 for an overview). Very much in the spirit of Givón, they refrain from making unnatural and ad hoc divisions between grammatical categories and modules and assume instead gradual transitions within a continuum between the two poles of grammar and lexicon,2 proposing a grammaticalization path. (2)

autonomous word > grammatical word > clitic > affix > Ø

Seen from this angle, grammaticalization may be defined as the diachronic process in the course of which a lexical item develops into a functional one, passing through several steps of development, by this cutting across components of grammar (↗21 Grammaticalization and pragmaticalization). Given that such processes take place

1 This idea also has its precursors in the late 19th century; cf. von der Gabelentz (1891/1901), who states that today’s affixes developed from full words (“Was heute Affixe sind, das waren früher selbständige Wörter”, 255). 20th century scholars like Hodge and Givón, however, refer not only to single parts of speech, but to whole grammatical components such as morphology and syntax. 2 This dichotomy might be seen as problematic, since function words, which are often derived from lexical ones, belong to the lexicon in the same way as their lexical precursors do (cf. Gabriel 2003 for further discussion).

Grammatical interfaces in Romance languages: An introduction

3

at any time and at different speeds, a given functional area can display items of different degrees of grammaticalization at any synchronic stage – on the assumption, of course, that we accept the well-known idealizations linked with the concept of synchrony since de Saussure’s Cours de linguistique générale (1916/2013). An example of such a functional area is depicted in (3), where the various linguistic means that are used in contemporary French to express (spatial, temporal, or abstract) relations between two entities are situated on a four-point grammaticalization scale (roughly along the lines of Lehmann 1985, 46; cf. also Gabriel 2003).

Lexicon [1]

au lieu (de) durant

[2]

pendant sur dans à1 [3]

à2 [4]

Grammar

à la place (de)

(3)

The evolution of a functional item from a lexical one is characterized by certain concomitants, expressed in terms of grammaticalization parameters (cf. Lehmann 1982/2002, 108–159; 1985), such as declining phonetic and semantic integrity (i.e. phonetic erosion and loss of semantic features), a decrease in syntactic scope, and diminishing paradigmatic and syntagmatic variability of the element concerned. Taking this into account, it can be argued that the prepositional locution Fr. à la place de ‘instead of’ is syntactically less fixed and thus closer to the lexicon (stage [1], cf. 3, above) than the expression au lieu de (same meaning, stage [2]), since the former allows PP-internal modifications such as the substitution of the DP complement by a possessive determiner as in à ta place ‘instead of you’, which is ungrammatical in the case of au lieu de (yielding au lieu de toi, but not *à ton lieu). For the very same reason, Fr. durant ‘during’ (stage [2]), which may occur either pre- or postnominally (cf. durant toute sa vie or toute sa vie durant ‘during his/her whole life’), is considered less grammaticalized than pendant (same meaning, stage [3]), which only appears in prenominal position. The preposition Fr. à, finally, can be argued to occupy two different stages in (3): It may be used either as a meaningful item as in Fr. il habite à Lyon ‘he lives in Lyon’, where it may be substituted by other prepositional expressions as in Fr. il habite près de Lyon ‘he lives close to Lyon’ (à1, stage [3]), or as a case marking item as in Fr. il donne un cadeau à son frère ‘he gives a present to his brother’ (à2, stage [4]). Other examples from the history of Romance languages involve the evolution of personal pronouns and future tense marking. While Latin possesses a set of strong pronouns (e.g. Lat. ego1 S G .NO M ‘I’, me1 S G . AC C ‘me’), all Romance languages have developed a set of clitic object3 pronouns (e.g. Fr./Sp. me, It. mi), which need 3 In addition, French and some Northern Italian dialects have developed a series of clitic subject pronouns (e.g. Fr. je1 S G ), which contrast with their strong counterparts (moi1 S G ).

4

Susann Fischer and Christoph Gabriel

to be considered both syntactically and phonologically weaker than their Latin counterparts: They cannot occur in isolation or in a stressed position, and they undergo systematic phonetic reduction processes such as vowel elision, at least in some of the Romance languages, among them French and Catalan (↗12 Object clitics). Loss of syntactic, phonological, and also semantic weight also characterizes the evolution of the Latin full verb HABERE ‘to have’ to the Vulgar Latin auxiliary, which forms part of the analytic construction given in (4b), which gradually replaced the synthetic Latin future form (4a), and subsequently further grammaticalized to the Romance future tense morpheme (4c). (4)

a.

Lat.

cantabo sing-1SG . FUT ‘I will sing’

b.

VLat.

cantare sing- INF

c.

Sp.

cantaré sing-1SG . FUT

habeo have-1SG . PRS

The idea of a close interaction of grammatical components plays an essential role not only in grammaticalization theory, but is also fundamental to much work done by members of the Prague Linguistic Circle: Vilém Mathesius (1929), for example, coined basic notions of information structure such as theme and rheme (↗14 Information structure, prosody, and word order, ↗22 Changes at the syntax-discourse interface) and explained the differences between English and Czech word order by resorting to the interplay between syntax, prosody (placement of sentential stress), and information structure. As he points out, the two languages differ in that Czech preferably marks a constituent as new information by means of word order variation (cf. the postverbal focused subject in 5a), while English mainly uses sentential (or: nuclear) stress (indicated through capitalizing, cf. 5b).

a.

‘Who wrote this letter?’ Tenhle dopis napsal this letter wrote

b.

DAD wrote this letter.

(5)

tatínek. dad

(Mathesius 1961/1975, 85)

Historical and typological approaches to language have been concerned with processes and continuous phenomena that cut across different components of grammar and have investigated the interfaces between these components. It thus seems correct to say that there has never been any doubt concerning the interaction of grammatical components (cf. Fischer 2010 for an extensive discussion on this matter).

Grammatical interfaces in Romance languages: An introduction

5

The first exception to this view, however, could be seen in the Neogrammarians’ (Otto Behagel, Berthold Delbrück, Hermann Paul etc.) approach to language change. They were the first to claim that the sound level is autonomous and independent of grammatical structure – or, to put it more precisely: that phonological rules, commonly referred to as Lautgesetze (‘sound laws’), can be formulated which make no reference to morphology, syntax, and semantics. In addition to the autonomy of the sound level they introduced the principle of the regularity of sound change, i.e. every sound in the same phonetic environment is affected in the same way. However, since the regularity of sound change produces morphological irregularities which interfere with the link between sound and meaning, the Neogrammarians introduced, as the second important process in language change, the process of analogy, in this way explaining morphological changes that followed sound changes and obvious exceptions to their proposed sound laws. A Romance example illustrating the interaction of Lautgesetz and analogy is given in (6). (6)

a.

Lat. A ˈ MARE Sg Pl

1 2 3

ˈ AMO A ˈ MAMUS ˈ AMAS A ˈ MATIS ˈ AMAT ˈ AMANT

b.

OFr. amer Sg Pl aim(e) amons aimes amez aime(t) aiment

c.

ModFr. Sg aime aimes aime

aimer Pl aimons aimez aiment

As can be seen in the Old French paradigm in (6b), Latin [a] regularly undergoes diphthongization ([a] > [aj]) in stressed open syllables before a nasal consonant, while its unstressed counterpart remains unchanged (cf. Anglade 1931, 12, 25–27). The regular application of this stress-dependent Lautgesetz thus yields an allomorphic verbal paradigm in Old French, hence morphological irregularity: While for the 1–3 Sg and 3 Pl forms, the stem is [aȷ ͂m-] (the adjacent nasal consonant triggers nasalization), it is [am-] for the 1 and 2 Pl forms (OFr. [am]ons, [am]ez). In Modern French (cf. 6c), the thematic vowel monophthongizes to [ɛ] (still written ai), and the paradigm is regularized through analogical shift of the 1 and 2 Pl forms (e.g. OFr. [a]mons > ModFr. [ɛ]mons). Thus, even though the phonological level is considered to be autonomous in a Neogrammarian view, the morphological component reacts to the changes in the phonological or phonetic component, and syntax reacts to changes concerning morphology and phonology. Early generative grammar can be seen as the first theory where the different components are considered as autonomous modules, independent of each other. In 1957, Noam Chomsky wrote Syntactic structures, a distillation of his dissertation The logical structure of linguistic theory (1955/1975), and therewith founded generative linguistics. His approach to linguistics has always aimed to understand why languages are structured the way they are and to formulate a grammar that is able to explain all sentences of a particular language, including constructions that hardly

6

Susann Fischer and Christoph Gabriel

occur in naturalistic data, but are considered grammatical nonetheless, as well as those sentences that according to normative grammars are ungrammatical, e.g. examples of double negation like I ain’t no nice guy. In contrast to traditional linguistic approaches, generative grammar can be described as an explanatory theory in the tradition of Galileo and Newton. Instead of merely describing and classifying the different language systems, generative grammar aims at finding the rules that can generate all possible sentences (and de facto nothing but those sentences) of a specific language. In the 1950s, the social sciences were dominated by behaviourism, the school of John B. Watson and Burrhus F. Skinner. Terms like “know” and “think” were considered to be unscientific; terms like “mind” and “innate” were branded as being bad words. Science, it was argued, should concentrate on observable facts, and the relationship between a stimulus and a response is observable. Behaviour was studied by looking at rats pressing bars or dogs salivating when hearing a certain tone, and these were explained by laws of stimulus-response learning. On the basis of these results, Skinner (1957) developed the notion of operant conditioning – claiming that human beings operate on their environment. Chomsky criticized this view, arguing that the interesting aspects of social phenomena, including language, are psychological or cognitive. The fact that these phenomena are not directly observable should not hinder us from investigating them. On the contrary, linguists should concentrate on the study of mental grammar (Chomsky 1959). It is already in those years that he called attention to some fundamental facts about language and language acquisition: First, almost every sentence a person utters or understands is a new combination of words that is uttered for the first time in the history of the universe. Second, any native speaker of a language has an implicit knowledge about his/her language, i.e. s/he knows without a doubt which sentences are acceptable (grammatical) and which are unacceptable (ungrammatical). This holds for both word order (cf. 7a vs. b) and morphophonological phenomena such as cliticization (cf. 8). Third, sentences can be grammatically judged without the need to understand their semantics (cf. the examples in (9), below). And finally, every child learns to speak grammatically well-formed utterances in a fairly short time, without negative evidence and without receiving explicit instruction. (7)

Eng.

a.

Yesterday I went to the doctor.

b. *Yesterday went I to the doctor. (8)

Eng.

a.

Kim is the professor here.

b.

Kim’s the professor here.

c.

Kim is happier than Tim is.

d. *Kim is happier than Tim’s.

Grammatical interfaces in Romance languages: An introduction

7

There is no doubt that any speaker of English knows that (7a) is grammatically correct, while (7b) is not. One could argue that English pupils are taught in school that their mother tongue has a strict subject – verb – object order and therefore the subject always needs to precede the finite verb. However, the sentences in (8) are of a different kind. The subtlety of knowing which auxiliary can be reduced in informal speech is something that is not taught in school. Most of the speakers of English are not even aware of this difference; nevertheless, they know that the auxiliary is can be reduced in (8b), as opposed to (8d). But even more importantly for Chomsky’s argument: The same holds for the sentences in (9). (9)

Eng.

a.

Colorless green ideas sleep furiously. (Chomsky 1957, 15)

b. *Furiously sleep ideas green colorless. Every speaker of English will agree that sentence (9a) is grammatically correct and sentence (9b) is not, despite the fact that they most probably have never heard these sentences and that neither of the two sentences is semantically well formed. The examples in (9) are discussed in Chomsky’s Syntactic structures (1957, 15) and show that syntactic structure can be understood without referring to the semantic level. This discussion was basically the beginning of the generative view for many years that the different components of grammar are autonomous. This idea of strictly separated components of grammar is also fundamental to the classical generative Principles and Parameters (or: Government and Binding) paradigm (Chomsky 1981/ 1993), according to which the linear ordering of the constituents of a given sentence (surface structure) is derived from an underlying representation (deep structure) and subsequently interpreted by semantics on the one hand (so-called Logical Form, LF) and phonetics/phonology on the other (Phonetic Form, PF). This view of the architecture of grammar is expressed by means of the famous T- (or Y‑)model (cf. Chomsky 1981/1993, 17), as illustrated in (10) with the example of the French wh-insitu interrogative Pierre a vu qui? (‘Whom has Peter seen?’). (10)

8

Susann Fischer and Christoph Gabriel

As can be seen in (10), both semantics (LF) and phonetics/phonology (PF) have access to the output of syntax, i.e. they simply process what is delivered by the syntactic component after all movement operations have applied (in the case of a wh-in-situ question, no syntactic movement applies, since the interrogative pronoun qui remains in its clause-final base position). Phonology and semantics only come into play when the computation of syntactic structure is completed. A generative framework that completely dispenses with the separation of grammatical components or modules is Optimality Theory (OT; cf. Prince/Smolensky 1993/ 2004). Within OT, the grammar of a given language is essentially understood as a language-specific hierarchy of violable constraints (rather than rules) according to which the output form is selected from among several competing input forms within an assumed evaluation process. These constraints can refer to all linguistic levels and thus be phonological, syntactic, or pragmatic in nature, among other things. An output form such as Sp. Vendió la casa Pablo ‘It was Pablo who sold the house’ (involving rightmost placement of the focused subject Pablo) is to be interpreted as the result of an evaluation process involving several competing input forms, among them, for example, Pablo vendió la casa ([F S]VO) and Vendió la casa Pablo (VO[F S]), and a constraint ranking with a high-ranked phonological constraint that requires clause-final position of nuclear stress. In such a view, the relationship of phonological and syntactic structure does not involve a “mapping” of essentially different subsystems of grammar (cf. e.g. Selkirk 1984), but combines them in a rather connectionist fashion. However, with the rising interest of generative linguistics in language change and language acquisition, it has been recognized that the different components of grammar interact in nontrivial ways. In fact, one of the most significant insights of recent years has been the understanding that the nature of the interfaces between the individual subsystems of grammar, i.e. Lexicon, (Morphology)4, Syntax, Phonology, and Semantics, is just as important as the mechanisms within the subsystems. Different theoretical frameworks have described and formalized the various interfaces differently and it has been under debate how many interfaces there actually are, what the actual extent of interaction is, and how we can most appropriately represent these in grammatical theory. Although no consensus has yet been reached, it seems to be clear that research regarding the nature of the interfaces is crucial for our understanding of language and the language faculty. In other words, understanding the interfaces shows us the restrictions on the relevance of the theory as a whole. Within the generative paradigm, the view of a strict autonomy of grammatical components has consequently been more and more undermined in the course of 4 The brackets indicate that in the original generative T- (or Y-)model, syntax has an interface with phonology and semantics, but not with morphology (cf. the graph in example 10, above), which simply does not exist as an autonomous component (cf. Jackendoff 1997; 2010).

Grammatical interfaces in Romance languages: An introduction

9

recent theoretical developments towards the so-called Minimalist Program (MP; cf. Chomsky 1995; 2000), according to which the four distinct levels of the classical Principles and Parameters approach (deep structure, surface structure, LF, and PF; cf. 10, repeated in 11a, for convenience) are abandoned in favour of a simplified model that only consists of two levels of representation, namely LF and PF (cf. 11b). (11)

a.

b.

According to minimalist assumptions, it is supposed that at a certain point of the derivation, called “spell-out”, the syntactic structure is delivered to PF, where phonological rules apply (cf. 11b). The interplay between PF (i.e. phonology) and the phrase structure building operations becomes relevant from Chomsky’s (2001) “Derivation by Phase” approach on, according to which it is assumed that the phrase marker is not delivered to PF as a whole, but in smaller chunks within several phases. As a consequence, the domain of the phonological component is not restricted to the application of phonological rules to the output of the whole sentence, but to several chunks that reach PF cyclically, with the result that the two components seem to interact with each other. In addition, head movement such as the raising of the verb from its VP-internal base position to a higher functional head (such as T) is no longer seen as “part of the narrow-syntactic computation but an operation of the phonological component” (Chomsky 2001, 37),5 an idea based on the assumption that the position of the inflected verb form in a given clause has no influence on its semantic interpretation (cf. Fr. Il [V regarde] [Adv souvent] la télé vs. Eng. He [Adv often] [V watches] TV). This allows the interpretation of verb movement as PF-syntax and has been applied to focus-induced word order variation (cf. Erteschik-Shir/Strahov 2004 on Germanic languages; Gabriel 2010 on Spanish, and ↗14 Information structure, prosody, and word order). Nowadays the term “interface” can be understood in different ways: as the connection between linguistics and other disciplines (e.g. philosophy, psychology, sociology), between the language faculty and other aspects of cognitive domains, i.e. external interfaces or nonlinguistic interfaces, specifically with the sensorymotor system at PF (e.g. prosody, phonetics) and with the conceptual-intentional

5 For a different view, cf. Lechner (2005), Matushansky (2006), and Roberts (2010), among many others.

10

Susann Fischer and Christoph Gabriel

system at LF (e.g. pragmatics and discourse function), and, in a narrow interpretation, as the interaction between the core computational systems, i.e. internal interfaces between the subsystems of grammar (cf. the discussion in Ramchand/Reiss 2007 and Sorace 2011).

2 Interfaces in Romance languages: The structure of the volume This volume addresses problems at the cross-sections of one or more internal or external interfaces in a number of Romance languages. The individual chapters are by authors who have been working on specific Romance issues that touch upon more than one subsystem of grammar, i.e. problems that go beyond traditional domains of grammar and that cannot easily be accounted for within a modular conception of the linguistic system where the subsystems do not share information with each other. Some of the chapters seek to highlight the controversy of the ongoing debate and/or provide data from nonstandard languages, varieties of Romance that challenge the modular view even more, while others again define the conditions under which grammaticality at the interfaces can be achieved. The volume is intended to function as a state-of-the-art report on research in the field, but at the same time as a manual of Romance languages with special emphasis given to different linguistic phenomena specific to Romance languages. The volume consists of four main parts, addressing the various interfaces between the components of grammar: In a first step, we concentrate on the interdependencies between sound and the underlying structure (I. Sound and structure). The second part is devoted to the interrelations between semantics and structural form (II. Structure and meaning). The third part investigates the complex interplay between the three areas mentioned so far (III. Sound, structure, and meaning). The last part focuses on the acquisition and the evolution of interface phenomena in Romance languages, where especially external interfaces are of importance (IV. The role of the interfaces in language acquisition and change).

2.1 Sound and structure There is no doubt that sound and structure are interrelated to some extent: The phonetic surface of a language is a concrete and measurable manifestation of a language-specific phonology. Within the grammar of a given language, underlying phonological structures form a subsystem which differs in its categorical inventory and compositional principles from both morphology and syntax, but we know that each subsystem can hardly be described adequately without reference to the others.

Grammatical interfaces in Romance languages: An introduction

11

The chapters of this section are concerned with clarifying in what way the different domains are relevant for the generalizations of the other. In a Neogrammarian sense, phonology has been argued to be autonomous; however, investigations into the field have often argued that certain syntactic information is available to phonology, but no phonological information seems to be available to syntax. Generative syntax during the 1990s, though, did allow so-called “last resort” movements in order to protect phonologically weak elements – like enclitics – from appearing in sentenceinitial position (e.g. Cardinaletti/Roberts 2002), or posited PF filters to eliminate unwanted structures. In chapter 1, Surface sound and underlying structure: The phonetics-phonology interface, José Ignacio Hualde and Ioana Chitoran focus on the interdependencies of underlying phonological representation and phonetic surface form and discuss several consonantal and vocalic processes, such as lenition, fortification, and assimilation as well as reduction and coalescence, with examples from a large array of Romance varieties. The historical dimension is also referred to when appropriate. Chapter 2, Segmental phenomena and their interactions: Evidence for prosodic organization and the architecture of grammar, by Marina Vigário, is devoted to the interaction of segmental phonology with other components of grammar. In a first step, it is shown, based mainly on examples from Iberian varieties, how segmental phonology is closely intertwined with suprasegmental features and prosodic structure; in a second step, its interface with morphology and the lexicon is taken into account. Furthermore, specific segmental phenomena related to frequency effects and the phonological integration of loan words are discussed. In the third chapter, Élisabeth Delais-Roussarie focuses on Prosodic phonology and its interfaces, based on the assumption that, first, prosodic units are partly derived from the morphosyntactic and information structure of a given sentence and that, second, they constitute domains for the application of phonological phenomena. These are studied within the framework of Prosodic Phonology, which essentially accounts for the way phonology interacts with the other components of the grammar. After presenting the main features of Prosodic Phonology, the author explains, based on examples from various Romance languages and varieties, how the syntax-phonology mapping is formalized within this framework, with particular reference to the formation of prosodic phrases, and shows how prosody is constrained by information related to discourse. In chapter 4, Phonology and morphology in Optimality Theory, Eulàlia Bonet and Maria-Rosa Lloret address the intriguing question of whether or not a model such as Optimality Theory (OT), which essentially tackles phenomena at the phonologymorphology interface, should allow intermediate levels of representation. They address this pending issue by discussing a large array of phenomena from Romance varieties that challenge the parallel version of OT in order to contrast the additional mechanisms proposed to maintain parallelism (e.g. output-output and alignment constraints) with the analyses provided within different serial (stratal, derivational,

12

Susann Fischer and Christoph Gabriel

or cyclic) versions of OT. The discussion of parallel and serial versions of OT forms the basis for an in-depth analysis of phonologically conditioned phenomena of allomorph selection in several Romance languages. The fifth and last chapter of the first part, Inflectional verb morphology, by Sascha Gaglia and Marc-Olivier Hinzelin, offers a selection of the most relevant phenomena in inflectional verb morphology, thereby focusing on issues that play a role at the morphology-phonology interface, such as stem allomorphy and syncretism, among other things. In addition, further instances of noncanonical morphology are discussed such as suppletion, periphrases, overabundance, and defectiveness. The authors finally discuss the assumption of an autonomous “morphomic” level, thus addressing the pending question of the autonomy of morphology.

2.2 Structure and meaning Part II, Structure and meaning, is concerned with the interface between the structure of words (morphology), the structure of sentences (syntax), and their meaning (semantics). Questions about how words’ morphological characteristics influence their syntactic distribution and/or semantic interpretation, how certain morphological markings change the semantic interpretation, and how syntactic distribution and specific positions of words/categories change the interpretation of sentences as a whole have been discussed vigorously within linguistics. In the chapters of this section issues are discussed that were at the centre of investigation long before the notion of interfaces and their interactions were introduced. The chapters have in common that they show how structure and interpretation interact and that correct analyses need to involve the subsystems as well as the interfaces. In chapter 6, Meaning of words and meaning of sentences, by Teresa Espinal and Susagna Tubau Muntaña, the question is addressed how the meanings of n-words contribute to the meaning of whole sentences. The chapter presents up-to-date data on the use and distribution of n-words in Spanish, Catalan, French, and Romanian, providing an analysis in terms of ±interpretable features and a semantic operator/ feature, and discussing the interface between syntax and interpretation. The authors argue that the semantic feature ensures that the n-words behave as polarity items, and that the syntactic feature guarantees their occurrence in negative concord structures. In this way they provide an interesting typology of n-words concerning the four Romance languages mentioned above, contributing to our general understanding of the interaction between semantic and syntactic lexical features on the one hand and syntactic operations on the other. Chapter 7, Morphology and semantics: Aspect and modality, by Eva-Maria Remberger, discusses the interpretative interrelations between modality, mood, aspect, and also tense in different Romance languages. The chapter pursues a cartographic approach, assuming that aspect (structural and aspectual verbal meaning, thus including lexical aspect, i.e. Aktionsart) is syntactically encoded within the vP domain – the domain

Grammatical interfaces in Romance languages: An introduction

13

of predication, and, semantically, the domain of the event situation. Modality under this view is encoded in a higher domain, namely the IP/TP, which is in close relation to the CP domain – the domain of contextual/illocutionary anchoring, sentence mood, and the speech situation. The intermediate functional domain connected to both aspect and modality is tense in the IP/TP. Thus this chapter shows how the interplay of aspect, tense, and modality is crucial for the discussion of interface phenomena, i.e. between morphosyntax, semantics, and pragmatics in general, and how this interaction can be accounted for in a cartographic account. Chapter 8, (In)definiteness, specificity, and differential object marking, by Luis López, investigates the interface of syntax and interpretation and provides evidence for the restrictions on semantic interpretation in certain syntactic environments. The chapter shows for a number of Romance languages that exhibit differential object marking (DOM) that in these languages a subset of direct objects is distinguished by means of a morphological marking or by syntactic placement. After pointing out the shortcomings of restricting DOM to, for example, animacy (a semantic feature), it is proposed that the phenomenon can be compared to scrambling in the Germanic languages, i.e. DOM and wide scope of indefinites entail scrambling. More specifically the author proposes that objects marked by DOM are composed by ‘function application’ (after type shifting), while objects without DOM are composed by means of ‘restrict’. Thus, syntactic configurations limit the range of possible semantic interpretations. In chapter 9, Agreement restrictions and agreement oddities, Roberta d’Alessandro and Diego Pescarini address exceptional agreement patterns, i.e. cases where the verb does not agree with the subject, where the morphological marking of the pronouns in clusters does not match their coding, and where the pre- or postnominal modifiers do not agree with the noun in the DP. The phenomenon of so-called mismatches or oddities regarding the Romance languages has been known for a very long time; Marcus Terentius Varro in his De lingua latina (47–49 BC ) already listed environments where the verb could only show a 3p agreement ending. This chapter provides an empirically rich overview by discussing the different restriction phenomena in the standard languages, as well as with respect to some Italian and Spanish varieties, by presenting the theoretical approaches in the literature of the last decades, and by examining which issues can be seen as settled and which ones deserve more consideration. Chapter 10, Auxiliary selection, by Jaume Mateu Fontanals, discusses the different factors to which auxiliary selection seems to be sensitive, including (in)transitivity, argument structure (unaccusativity vs. unergativity), lexical semantics, Aktionsart, tense, modality, clausal aspect, and subject person. The data examined in this chapter shows that an impressive range of variation is attested in the standard and nonstandard variations of Romance. After thoroughly discussing and contrasting the syntactic and semantic approaches that have been proposed so far, the author argues in favour of accounts that operate essentially at the interface between syntax and semantics.

14

Susann Fischer and Christoph Gabriel

2.3 Sound, structure, and meaning Part III, Sound, structure, and meaning, addresses issues that are related to the syntax-phonology interface on the one hand and to the syntax-semantics interface on the other. It seems obvious that all languages can serve the same communicative needs. However the various languages use different linguistic means in order to meet these needs. One could say – and it has been suggested – that the standard and nonstandard variations of Romance languages are just paradigmatic examples of different diachronic stages and parameter settings (Uriagereka 1995), i.e. they only vary in which syntactic and semantic features are phonologically realized and which are not, and that this can account for the differences between the Romance languages. Nevertheless, without disagreeing with this view, we would like to add that an important additional aspect of the differences within and across the different Romance languages concerns the various ways in which the different subsystems interact. Thus, it is important how phonology/phonetics (the phonetic realization of syntactic and semantic features) interacts with syntax (the way structure is built up), but it is also important to understand how discourse function and information structure interact with the linking between syntax and semantics. The contributions of this section discuss in what way the interaction and linking between the different modules can best be captured and explained. In chapter 11, Subjects, null-subjects, and expletives, Michelle Sheehan investigates the interplay of overt phonological (morphological) realization and information structure in licensing null subjects. The author examines the behaviour of a variety of null-subject languages with respect to the two main minimalist approaches, i.e. in simplified terms, pro exists and pro does not exist. The two approaches make very different empirical predictions regarding the status of overt subjects in pre- and postverbal positions as well as the (non)existence of null expletives. On the basis of ten diagnostic tests (adverb placement, wide/narrow scope of preverbal subjects, binding of postverbal subjects, non-referential subjects, floating quantifiers, subject vs. topics, hortative contexts, basic word order, disambiguation, and parasitic gaps), it is shown that the Romance languages seem to fall into at least two groups: type (a) null-subject languages, in which either XP or X-movement can satisfy the EPP, i.e. verbal morphology seems to be pronominal, and type (b) nullsubject languages, in which some XP must always satisfy the EPP and null-subjects are deleted. Chapter 12, Object clitics, by Susann Fischer and Maria Goldbach, gives an overview of the classic criteria used to define clitics and discusses the specific properties shared by Romance object clitics, as well as some synchronic and diachronic differences, concerning for example doubling, climbing, mesocliticization, and opaque clusters, i.e. the person-case-constraint. By presenting some of the most prominent (classic) accounts as well as some recent approaches that have been proposed in order to account for the specific behaviour of Romance clitics, the chapter aims at

Grammatical interfaces in Romance languages: An introduction

15

presenting a state-of-the-art perspective on cliticization and its consequences for the interaction of the subsystems of grammar. In chapter 13, Nominalizations, Judith Meinschaefer provides a detailed overview of deverbal nominalizations across the Romance languages at the interface between morphology, syntax, and semantics. In the simplest cases, nouns can be derived from one-place predicates. In more complex cases, they derive from complex verbs and retain much of the syntactic and semantic complexity of their bases. The contribution focuses on two important questions: First, to what degree is the eventstructural and argument-structural complexity of the verbal domain visible in the nominal domain, and second, is the morphological complexity of deverbal nominalizations a reflex of their event-structural and argument-structural-complexity? With a focus on these two questions, the contribution presents the latest perspective on nominalization and its implications for the interfaces between morphology, syntax, and semantics. Chapter 14, Information structure, prosody, and word order, by Andreas Dufter and Christoph Gabriel, addresses the expression of information structure, based mainly on data from Catalan, French, Italian, Portuguese, Romanian, and Spanish, with special regard given to the syntax-phonology interface. After a critical overview of the basic notions of information structure, information-structurally induced word order variation in simple and complex sentences as well as clefting and dislocation structures and their prosodic realizations are discussed. Furthermore, the authors address the question of how free variation (or: optionality), as primarily occurs at the so-called ‘external’ interfaces, can be accounted for in formal models of grammar. The last part of the chapter is devoted to the role of interfaces in learner and contact varieties and in linguistic change. Chapter 15, VP and TP ellipsis: Sentential polarity and information structure, by Ana Maria Martins, focuses on two types of predicate ellipsis, namely verbal phrase ellipsis and tense phrase ellipsis, and is thus situated at the interface of syntax and discourse function. It is shown that there is a nontrivial correlation between the licensing of predicate-ellipsis and the polarity-encoding system of language-particular grammars, which explains the cross-linguistic variation concerning the availability of this phenomenon. Languages that license verbal phrase ellipsis display polar answering systems where the verb plays an important role (e.g. Portuguese, Galician, and Latin) and where these bare-verb answers constitute an unmarked, pervasive, and early acquired manifestation of the syntax-semantics and syntax-pragmaticsdiscourse interfaces. In languages where tense phrase ellipsis and verbal phrase ellipsis are licensed, these are not in free variation but implement different strategies regarding information structure. In chapter 16, Existential constructions, Delia Bentley and Silvio Cruschina discuss the major properties of existential constructions in a variety of different standard and nonstandard Romance languages. It is suggested that the argument structure and the predication of these constructions are distinct from those of locatives and

16

Susann Fischer and Christoph Gabriel

possessives, which is testified by the morphosyntax of the Romance languages discussed. Under the view presented in this contribution, the noncanonical morphosyntactic properties of existentials, such as the postcopular position of the pivot and in some cases, the pivotʼs inability to control number agreement on the copula, can be explained by referring to aspects at the interfaces of morphosyntax with semantics and pragmatics (discourse).

2.4 The role of the interfaces in language acquisition and change The last part of the volume is concerned with the role external and internal interfaces play in language acquisition, language contact, and language change. Since Platzack’s (2001) paper it has often been shown that linguistic phenomena at the (external) interfaces are especially vulnerable in language acquisition (bilingual and second language acquisition) contexts. This view, known as the interface hypothesis (Sorace/Filiaci 2006), has influenced much research on the syntax-pragmaticsinformation structure interface and on the syntax-phonetics-prosody interface since its original formulation, and has recently been questioned (e.g. Domínguez 2013). Under a generative view language change is directly related to language acquisition, which is why this section consists of contributions on language acquisition and on (contact-induced) language change. All contributions however mainly explore the external interfaces and most of all the variables that contribute to the specific (or non-specific) vulnerability of these interfaces. In chapter 17, Acquiring multilingual phonologies (2L1, L2 and L3): Are the difficulties in the interfaces?, Conxita Lleó discusses various types of bilingual and multilingual phonological acquisition, namely the simultaneous acquisition of two first languages (2L1) and the sequential acquisition or learning of second (L2) or third languages (L3). Special attention is paid to 2L1 acquisition, which either leads to two rather balanced phonological competencies or to a biased relation of a stronger and a weaker competence, as is the case for so-called heritage languages (HL). The data referred to come mainly from Spanish as it is acquired in different bilingual settings, but phonological phenomena in the bilingual acquisition of other Romance languages are also taken into account, thereby addressing both the order and the speed of phonological acquisition as well as the question of how phenomena of (negative and positive) transfer can be represented in formal models of grammar. The author argues that the interface hypothesis is not the most explanatory one and proposes a new approach to bilingual development, based on the insights of Optimality Theory. Chapter 18, Interfaces with syntax in language acquisition, by Tanja Kupisch and Jason Rothman, provides an overview of empirical work on the acquisition of various Romance languages, focusing on the differences between the syntax-semantics and the syntax-discourse interfaces. Since the body of research across different acquisition settings and speaker types is very large, the article concentrates with respect

Grammatical interfaces in Romance languages: An introduction

17

to the syntax-semantics interface on adjectival position (pre- and postnominal) and concerning the syntax-discourse-pragmatics interface on null/overt subject and also article distribution. The authors point out that the association of particular phenomena with specific interfaces is problematic, as well as assume that internal interfaces are less problematic than the external ones. In chapter 19, The role of the interfaces in syntactic change, Esther Rinke is concerned with the role that the two interfaces standardly identified in Chomskyan linguistics – the articulatory-perceptual and the conceptual-intensional – play in the context of diachronic change. The contribution gives an overview of two generative approaches to syntactic change and recent work on the interaction between word order and information structure. Using these discussions as the basis of how the interfaces are implicated in change, the contribution also considers the relevance of acquisition to our understanding of syntactic change. Chapter 20, Interfacing interfaces: Quechua and Spanish in the Andes, by Pieter Muysken and Antje Muntendam, explores the phenomena that take place at different external and internal interfaces due to the contact between Quechua and Andean Spanish in Bolivia, Peru, and Ecuador. Phenomena such as aspect and evidentiality in past tenses, null objects for definite and specific antecedents, and prosody in declarative sentences are studies at the syntax-morphology, syntax-pragmatics, and syntax-prosody interfaces. It is shown that Quechua mainly influences the conceptual organization of Andean Spanish. In contrast, focusing on relexification in Media Lengua, the lexicon-grammar interface is examined, where Spanish lexemes are organized according to the semantics, phonology, and morphosyntax of Quechua. In chapter 21, Grammaticalization and pragmaticalization, Ulrich Detges and Richard Waltereit discuss two diachronic processes, namely grammaticalization and pragmaticalization, that apply at internal and external interfaces. They argue that core grammar as well as discourse markers and modal particles are sedimented residues of argumentative moves designed to solve communicative problems. What these processes of sedimentation have in common is that they are driven by language use, more precisely by routinization. Routinization under this view is an aspect inherent to language use. Furthermore it is an interface phenomenon that affects all modules of grammar and aligns changes occurring within them. Thus, the changes triggered by routinization are driven by factors outside the subsystems of grammar, i.e. factors such as relevance of information, felicity of speech acts, and coherence in discourse – in other words, the external interface between syntax and pragmatics/ information structure. Chapter 22, Changes at the syntax-discourse interface, by Kristine Gunn Eide, finally, outlines the general correlations between information-structural categories such as topic and focus on the one hand and syntactic functions such as subject and object on the other, and discusses how the information-structural level of texts can be disentangled from syntax in historical data. Special attention is given to the

18

Susann Fischer and Christoph Gabriel

change from V2-like structures as they occur in the medieval stages of Romance languages to the different ways of structuring information structure and syntax in the contemporary varieties. The focus is mainly on Portuguese, but examples from other Romance varieties are discussed as well.

3 Closing words In this short overview we have only touched upon some of the facets of and problems concerning the interfaces in grammatical theory, i.e. the interface debate in linguistics. We hope that it has been helpful for those readers who were not already familiar with the phenomena. It is also our hope that this review is a valuable lead-in to a volume that presents a state-of-the-art picture of the ongoing discussion, rendering this discussion clearer for those who are familiar with it, but at the same time introducing it in a user-friendly way to those who have never taken part in the dialogue before. We have selected the contributions to this volume for their relevance in Romance linguistics, and because of their combination of empirical data and theoretical insights. All of the contributors have been active participants in the ongoing debate. Enjoy!

4 References Anglade, Joseph (1931), Grammaire élémentaire de l’ancien français, 4e éd., Paris, Colin. Cardinaletti, Anna/Roberts, Ian (2002), “Clause Structure and X-Second”, in: Guglielmo Cinque (ed.), Functional structure in DP and IP. The cartography of syntactic structures, Volume 1, Oxford, Oxford University Press, 123–166. Chomsky, Noam (1955/1975), The logical structure of linguistic theory, PhD dissertation, Cambridge, MA, MIT. Reprint: New York, Plenum Press, 1975. Chomsky, Noam (1957), Syntactic structures (Ianua linguarum, Series minor 4), ’s‑Gravenhage, Mouton. Chomsky, Noam (1959), “A Review of B. F. Skinner’s Verbal Behavior”, Language 35, 26–58. Chomsky, Noam (1981/1993), Lectures on government and binding, Dordrecht, Foris, 1981. Second edition: Berlin/New York, De Gruyter, 1993. Chomsky, Noam (1995), “Categories and transformations”, in: Noam Chomsky, The Minimalist program, Cambridge, MA, MIT Press, 219−394. Chomsky, Noam (2000), “Minimalist inquiries. The framework”, in: Roger Martin/David Michaels/ Juan Uriagereka (edd.), Step by step. Essays on Minimalist syntax in honor of Howard Lasnik, Cambridge, MA, MIT Press, 89−155. Chomsky, Noam (2001), “Derivation by phase”, in: Michael J. Kenstowicz (ed.), Ken Hale. A life in language, Cambridge, MA, MIT Press, 1−52. Domínguez, Laura (2013), Understanding interfaces. Second language acquisition and first language attrition of Spanish subject realization and word order variation, Amsterdam/Philadelphia, Benjamins.

Grammatical interfaces in Romance languages: An introduction

19

Erteschik-Shir, Noemi/Strahov, Natala (2004), “Focus structure architecture and P-syntax”, Lingua 114, 301−323. Fischer, Susann (2010), Word-order change as a source of grammaticalisation, Amsterdam/Philadelphia, Benjamins. von der Gabelentz, Georg (1891/1901), Die Sprachwissenschaft. Ihre Aufgabe, Methoden und bisherigen Ergebnisse, Leipzig, Weigel, 1891. Second, revised edition: Leipzig, Tauchnitz, 1901. Gabriel, Christoph (2003), “Relational elements in French. A minimalist approach to grammaticalization”, Linguistische Berichte 193, 3–32. Gabriel, Christoph (2010), “On focus, prosody, and word order in Argentinean Spanish. A minimalist OT account”, Revista Virtual de Estudos da Linguagem, Special issue 4 “Optimality-theoretic syntax”, 183−222. Givón, Talmy (1971), “Historical syntax and synchronic morphology. An archaeologist’s field trip”, Chicago Linguistic Society 7, 394–415. Givón, Talmy (1979), “From discourse to syntax. Grammar as a processing strategy”, in: Talmy Givón (ed.), Syntax and semantics, vol. 1: Discourse and syntax, New York, Academic Press, 81–112. Heine, Bernd/Kuteva, Tanja (2002), World lexicon of grammaticalization, Cambridge, Cambridge University Press. Heine, Bernd/Reh, Mechthild (1984), Patterns of grammaticalization in African languages (Arbeiten des Kölner Universalien-Projekts 47), Köln, Institut für Sprachwissenschaft, Universität zu Köln. Hodge, Carlton T. (1970), “The linguistic cycle”, Language Sciences 13, 1–7. Hopper, Paul J./Traugott, Elizabeth C. (1993), Grammaticalization, Cambridge, Cambridge University Press. von Humboldt, Wilhelm (1825), “Über das Entstehen der grammatischen Formen und ihren Einfluß auf die Ideenentwicklung”, in: Abhandlungen der Königlichen Akademie der Wissenschaften zu Berlin aus den Jahren 1822–1823, Berlin, Druckerei der Königlichen Akademie, 401–430. Jackendoff, Ray (1997), The architecture of the language faculty, Cambridge, MA, MIT Press. Jackendoff, Ray (2010), “The parallel architecture and its place in cognitive science. Foundations of language, brain”, in: Bernd Heine/Heiko Narrog (edd.), The Oxford handbook of linguistic analysis, Oxford, Oxford University Press, 645–668. Kuryłowicz, Jerzy (1965/1975), “The evolution of grammatical categories”, Diogenes 51, 55–71. Reprint: Kuryłowicz, Jerzy (1975), Esquisses linguistiques II, München, Fink, 55–71. Lechner, Winfried (2005), “Interpretative effects of head-movement”, Ms., University of Tübingen, http://ling.auf.net/lingbuzz/000178 (06.05.2016). Lehmann, Christian (1982/2002), Thoughts on grammaticalization, München, Lincom, 1982. Second, revised edition (Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 9): Erfurt, Seminar für Sprachwissenschaft der Universität, 2002, http://www.db-thueringen.de/ servlets/DerivateServlet/Derivate-2058/ASSidUE09.pdf (06.05.2016). Lehmann, Christian (1985), “The role of grammaticalization in linguistic typology”, in: Hansjakob Seiler/Gunter Brettschneider (edd.), Language invariants and mental operations. International interdisciplinary conference held at Gummersbach/Cologne, Germany, September 18–23, 1983, Tübingen, Niemeyer, 41–52. Mathesius, Vilém (1929), “Zur Satzperspektive im modernen Englisch”, Archiv für das Studium der neueren Sprachen und Literaturen 29/155, 202–210. Mathesius, Vilém (1961/1975), A functional analysis of present day English on a general linguistic basis, ed. by Josef Vachek, transl. by Libuše Dušková, The Hague, Mouton, 1975. Original Czech edition: Obsahový rozbor současné angličtiny na základě obecně lingvistickém, Prague: ČSAV, 1961. Matushansky, Ora (2006), “Head movement in linguistic theory”, Linguistic Inquiry 37, 69–110.

20

Susann Fischer and Christoph Gabriel

Meillet, Antoine (1912/1921), “L’évolution des formes grammaticales”, Scientia (Rivista di scienza) 12/XXVI, 384–400. Reprint: Meillet, Antoine (1921), Linguistique historique et linguistique générale, Paris, Champion, 130–148. Platzack, Christer (2001), “The vulnerable C-domain”, Brain and Language 77, 354–377. Prince, Alan/Smolensky, Paul (1993/2004), Optimality Theory. Constraint interaction in generative grammar, New Brunswick/Boulder, Rutgers University/University of Colorado, 1993, http:// roa.rutgers.edu/files/537-0802/537-0802-PRINCE-0-0.PDF (06.05.2016). Reprint: Malden, MA, Blackwell, 2004. Ramchand, Gillian/Reiss, Charles (2007), “Introduction”, in: Gillian Ramchand/Charles Reiss (edd.), The Oxford handbook of linguistic interfaces, Oxford, Oxford University Press, 1–14. Roberts, Ian (2010), Agreement and head movement. Clitics, incorporation, and defective goals, Cambridge, MA, MIT Press. de Saussure, Ferdinand (1916/2013), Cours de linguistique générale, Paris, Payot, 1916. Peter Wunderli (ed.), Ferdinand de Saussure, Cours de linguistique générale. Zweisprachige Ausgabe französisch-deutsch mit Einleitung, Anmerkungen und Kommentar, Tübingen, Narr, 2013. Selkirk, Elizabeth O. (1984), Phonology and syntax. The relation between sound and structure, Cambridge, MA, MIT Press. Skinner, Burrhus Frederic (1957), Verbal behavior, New York, Appleton-Century-Crafts. Sorace, Antonella (2011), “Pinning down the concept of ‘interface’ in bilingualism”, Linguistic Approaches to Bilingualism 1, 1–33. Sorace, Antonella/Filiaci, Francesca (2006), “Anaphora resolution in near-native speakers of Italian”, Second Language Research 22, 339–368. Uriagereka, Juan (1995), “Aspects of the syntax of clitic placement in Western Romance”, Linguistic Inquiry 26, 79–123. Varro, Marcus Terentius (1964), De lingua latina, ed. by Georg Goetz, Leipzig, Teubner, 1910. Reprint: Amsterdam, Hakkert, 1964.

I Sound and structure

José Ignacio Hualde and Ioana Chitoran

1 Surface sound and underlying structure: The phonetics-phonology interface Abstract: In this chapter we offer an overview of phenomena at the phoneticsphonology interface in the Romance languages. Processes affecting consonants and vowels are studied separately. Parallels with historical sound changes in the same or in another language are mentioned when relevant. Keywords: allophony, lenition, fortition, assimilation, neutralization, vowel reduction, vowel harmony, coda consonants, palatalization

1 Scope of this chapter This chapter on the phonetics-phonology interface focuses on phenomena that can be considered facts of pronunciation, including both postlexical or phrase-level phonological rules and conventionalized, language-specific, phonetic processes. We are thus excluding from our overview morphophonological (or lexical) phonological alternations, which will be dealt with in a different chapter. An example may be useful in order to clarify the set of phenomena that we are excluding from the scope of our discussion. In the Romance languages the evolution of mid vowels in stressed syllables (under different conditions) has created numerous morphophonological alternations, e.g. Sp. puedo/podemos ‘I/we can’, Fr. bois/buvons ‘I/we drink’, It. buono/bontà ‘good/goodness’, Rom. seară/seri ‘evening/s’, etc. The treatment of this type of alternation is different in different theoretical frameworks. In Generative Phonology, morphemes are given a single invariant form in the underlying representations of all words containing them. Accounting for these alternations is thus an important part of phonological analysis within this framework. Morphophonological rules account for the mapping between allomorphs of the same morpheme in different words. In a volume like the present one, however, these facts clearly belong to the phonology-morphology interface, rather than to the phonology-phonetics interface, and thus they will not be examined further in this chapter. Instead, in this chapter we will study both regular obligatory allophony, such as the spirantization of /b d ɡ/ in Spanish, e.g. /la bodeɡa/ [laβoˈðeɣa] ‘the tavern, shop’ (but not in Brazilian Portuguese [aboˈdeɡa]), and language-specific but optional, variable processes, such as the aspiration of /s/ in Spanish varieties and the spirantization of /p t k/ in Florentine Italian.

24

José Ignacio Hualde and Ioana Chitoran

The contemporary Romance languages differ substantially in their phonology and phonetics, although common processes are also found. Interestingly we often see that certain phenomena are recurrent. We wish to draw special attention to the fact that what is an active process at the phonetics-phonology interface in one language may mirror a completed sound change in another language. For instance, the variable aspiration and deletion of coda /s/ in many Spanish dialects ran its full course centuries ago in French, e.g. Sp. escuela [ehˈkwela] ~ [eˈkwela] ‘school’, cf. Fr. école. This view of the evolution of sound systems, with specific application to Romance, was first presented by Pierre Delattre in a 1946 paper explicitly entitled “Stages of Old French phonetic changes observed in Modern Spanish”. Delattre discusses 31 well-known sound changes of Old French for which an equivalent synchronic stage characterized by phonetic variation can be found in Modern Spanish. We will give just one example here. Glide strengthening in initial position takes place in French, with subsequent reduction to a fricative (cf. also section 2.4 below): Germanic [wadja] ‘wage’ > Gallo-Roman [ɡwaɟja] > Fr. [ɡaʒ]. A similar synchronic variation is encountered in Spanish, where huesos ‘bones’ is frequently pronounced [ˈɡwesos]. We refer the reader to Delattre (1946) for other examples of parallel sound changes in Old French and Modern Spanish. In the remainder of this chapter we classify phonetic/phonological phenomena by the nature of the segments that are affected and by their environment. For each main type of phenomenon we will consider the extent to which synchronically active processes at the phonetics-phonology interface find parallels in completed sound changes in the same or other Romance languages. Phenomena affecting consonants are considered before those that apply to vowels.

2 Consonants Consonants are often weakened in the intervocalic context and are subject to various neutralizations (in place, manner and/or voicing) in the coda. When in contact with front glides and vowels, palatalization is frequent. At the phonetic level, strengthening of consonants in phrase-initial position has been observed in several languages, and this phenomenon may be phonologized as word-initial fortition, although such conventionalized processes are relatively rare.

2.1 Intervocalic lenition The weakening of intervocalic consonants is a common phonological phenomenon that is abundantly attested in the synchronic and diachronic phonology of the Romance languages.

Surface sound and underlying structure: The phonetics-phonology interface

25

The historical voicing (and further lenition) of intervocalic obstruents is one of the main features that serve to separate the Western Romance languages (Portuguese, Spanish, Catalan, French, etc.) from Eastern Romance (Italian, Romanian), cf. (1): (1)

Latin intervocalic /p t k/: Eastern vs. Western Romance Lat. It. Sp. Fr. SAPĒRE sapere saber savoir ‘to know’ VĪTA (M ) vita vida vie ‘life’ AMĪCA (M ) amica amiga amie ‘female friend’

The voicing of intervocalic obstruents, which must have started as an across-the-board postlexical rule of allophony (Weinrich 1958), eventually produced phonological recategorization word-internally in the Western Romance area. Voiced stops were subsequently weakened to different extents in different languages, along a common path of development, e.g. [t] > [d] > [ð] > 0. In French, word-internal intervocalic /d/ and /ɡ/ were eventually lost ([ˈvita] > [ˈvida] > [ˈviðə] > [ˈviə], [aˈmika] > [aˈmiɡa] > [aˈmiɣə] > [aˈmiə]) and the labial (from Latin P, B and V ) remains as /v/, so that there is no active rule of allophony in modern French. In Spanish and Catalan, on the other hand, the spirantization of /b d ɡ/ both inside words and across word boundaries is an essentially obligatory allophonic phenomenon and an important aspect of the phonology of these languages, e.g. Sp. dama [ˈdama] ‘lady’, but la dama [laˈðama] ‘the lady’. Unlike the phonological recategorization of voiceless obstruents in Western Romance, which affected exclusively postvocalic consonants, the allophonic spirantization of /b d ɡ/ applies after vowels, glides and most consonants in Catalan and most Spanish varieties, e.g. Sp. árbol [ˈaɾβol] ‘tree’. There are, nevertheless, Spanish dialects (spoken in parts of Central America and Colombia) where systematic spirantization is limited to the intervocalic context, e.g. cada [ˈkaða] ‘each’ but cardo [ˈkaɾdo] ‘thistle’ (cf. Carrasco/Hualde/Simonet 2012). In Brazilian Portuguese intervocalic voiced stops do not undergo systematic weakening. This is thus a major phonological difference between this language and Spanish: (2) sabe lado amiga

Sp. [ˈsaβe] [ˈlaðo] [aˈmiɣa]

BPor. [ˈsabi] [ˈladu] [aˈmiɡa]

‘s/he knows’ ‘side’ ‘female friend’

In European Portuguese the spirantization of /b d ɡ/ is found as a variable phenomenon in northern and central areas (cf. Mateus/d’Andrade 2000, 11, footnote 2). In the varieties of central and southern Italy, as well as Corsica and Sardinia, where Latin intervocalic /p t k/ did not undergo phonological recategorization,

26

José Ignacio Hualde and Ioana Chitoran

we find that, nevertheless, these consonants are often voiced and sometimes spirantized synchronically, both word-internally and across word boundaries, e.g. ho capito [oɡaˈbido] ‘I have understood’. In a recent study, Hualde/Nadeu (2011), over 50% of all tokens of intervocalic /p t k/ were found to be realized as fully voiced in a corpus of Rome Italian. Sardinian has a more systematic alternation, e.g. [ˈtɛrːa] ‘land’, [saˈðɛrːa] ‘the land’ (cf. e.g. Jones 1997). Less widespread voicing (and spirantization) of intervocalic /p t k/ is also found in some varieties of Spanish. Velar /k/ appears to be especially prone to reduction, e.g. Peninsular Sp. lo que te digo [loɣedeˈðiɣo] ‘that which I tell you’ (cf. Hualde/Simonet/Nadeu 2011; Torreira/ Ernestus 2011). In the case of Spanish, this is a ‘second round of voicing’ (Oftedal 1985) from a historical point of view, although continuity with the first round (without recategorization across word boundaries) cannot be excluded. Interestingly, European Portuguese appears to be undergoing the opposite process: partial devoicing of obstruents (Pape/Jesus 2011). Florentine Italian is well known for showing a different type of lenition of intervocalic /p t k/, where these segments are variably realized as voiceless fricatives, e.g. la tavola [laˈθavola] ‘the table’ (Canepari 1979, 214; Giannelli/Savoia 1978; 1979/1980; Cravens 1984; Marotta 2001; 2008; Sorianello 2001; Villafaña Dalcher 2006; 2008). The Western Romance intervocalic voicing sound change also affected singleton /s/ and other fricatives, e.g. PASSU ( M ) > Por. passo [ˈpasu] ‘step’ vs. CASA ( M ) > Por. casa [ˈkazɐ] ‘house’. As a synchronic variable across-the-board process, voicing of intervocalic fricatives is also found in those languages where stops voice in this context (cf. Torreira/Ernestus 2012 and Hualde/Prieto 2014 for Spanish as well as Nocchi/Schmid 2007 for Southern Italian). In the history of the Romance languages we find other processes of intervocalic weakening. In particular the Latin geminates, which are preserved in Italo-Romance, were systematically simplified in both Western Romance and Balkan Romance (e.g. CUPPA (M ) > Sp. copa, Rom. cupă vs. It. coppa; PECCATU (M ) > Sp. pecado, Rom. păcat vs. It. peccato), but these sound changes did not give rise to any robust synchronic alternations (excluding a few limited morphophonemic processes).1 When sequences of identical consonants arise across morphemes or words, sometimes they are reduced to the duration of a single consonant and sometimes they are preserved as geminates. In Spanish, consonants with full contact between the articulators, such as /l/ and /n/, preserve a single vs. geminate contrast more effectively than approximants and fricatives (e.g. come nueces ‘s/he eats walnuts’ vs. comen nueces ‘they eat walnuts’; inútil ‘hopeless’ vs. innoble; but /des-alaɾ/ ‘to

1 In Spanish and Catalan, geminate /n l r/ had a special evolution. In particular /ll/ and /nn/ became palatals. A related fact is the synchronic alternation that we find in cases like Sp. él ‘he’, ella ‘she’.

Surface sound and underlying structure: The phonetics-phonology interface

27

remove the wings’ ~ /des-salaɾ/ ‘to remove the salt’; sabe solo ‘s/he knows only’ ~ sabes solo ‘you know only’; cf. Hualde 2005, 97–98). In French, schwa deletion may produce geminates, as in là-d(e)dans (cf. Walker 2001, 130). In some Romance languages intervocalic sonorants have also undergone weakening. Thus, in Portuguese intervocalic /n/ and /l/ delete, e.g. LŪNA (M ) > lua ‘moon’, COLŌRE ( M ) > cor ‘colour’, and in Romanian intervocalic /l/ becomes /r/, e.g. SŌLE (M ) > soare ‘sun’. This deletion of intervocalic consonant has produced some morphophonological alternations, e.g. Port. sol ‘sun’ vs. sois ‘suns’ < SŌLES , but again, nowadays we do not seem to find any systematic processes of intervocalic sonorant weakening at the phonology-phonetics interface.

2.2 Lenition/neutralization in the coda A number of phenomena may affect coda consonants, generally resulting in a reduced inventory compared to that found in the onset position. Common processes include place assimilation, cluster simplification, voice assimilation and complete assimilation.

2.2.1 Assimilation and weakening of coda obstruents Whereas, as already mentioned, the Latin geminates underwent simplification in both Western Romance and Balkan Romance, Italian has not only preserved the etymological geminate consonants but also developed new geminates. One important source of new geminates that must have already been present in Late Latin (Loporcaro 2011, 93) is the total assimilation of coda obstruents, e.g. AD CASA (M ) > [aˈkːasa] ‘to the house’, which is the origin of the phenomenon known as raddoppiamento (fono)sintattico (RS) or syntactic doubling in central and southern Italian. The synchronic status of RS is particularly interesting. Nowadays it is simply the case that an arbitrary set of function words trigger the gemination of word-initial consonants, since there is no longer any evidence for the word-final consonant that gave rise to the phenomenon; for example, in these Italian varieties, there is gemination of the word-initial consonant in a casa, but not in la casa. RS also takes place after oxytonic words, as in città [pː]iccola ‘small town’. Interestingly, word-initial consonants in the context of RS are protected from intervocalic weakening processes. Thus in Rome Italian, the word-initial stop may be voiced in la casa ‘the house’, but not in a casa ‘to the house’ or tre case ‘three houses’, where it geminates instead (cf. Loporcaro 1997, 1–2). Thus a seemingly low-level, probabilistic, phonetic phenomenon (intervocalic voicing) is blocked by a lexical feature (i.e. the fact that the preceding word belongs to the class of RS-triggering items). In Catalan, there is obligatory devoicing of word-final consonants, which produces many morphophonological alternations, e.g. amic/amica ‘friendM.SG/F.SG’.

28

José Ignacio Hualde and Ioana Chitoran

The treatment of coda /s/ requires particular attention. The loss of word-final /s/ in Eastern Romance is ultimately responsible for one of the most striking differences between Western and Eastern Romance, plural formation by the addition of /s/ vs. vowel change, e.g. Sp. libro/libros ‘book/books’ vs. It. libro/libri.2 This phenomenon affected only word-final consonants and is thus different from the weakening of /s/ in coda position (both word-internally and word-finally) that historically operated in French and is active in a large number of present-day Spanish varieties, with different degrees of incidence. In French, the process is no longer operative, so that coda /s/ in the lexical items where it has been preserved (e.g. espagnol ‘Spanish’) is no longer subject to weakening. Its only synchronic remnant can be seen in the phenomenon of liaison, whereby resyllabification as a syllable onset has allowed a word-final consonant to survive, as in, for example, les garçons [leɡaʁsɔ͂] ‘the boys’ vs. les enfants [leza͂fa͂] ‘the children’ (↗3 Prosodic phonology and its interfaces; ↗4 Phonology and morphology in Optimality Theory). Besides French liaison, the voicing of resyllabified word-final /s/ is found in both Catalan and Portuguese and was undoubtedly found in Old Spanish (as it still is in Judeo-Spanish). In standard European Portuguese and some Brazilian varieties the treatment of word-final /s/ is particularly complex. Coda /s/ is palatalized to [ ʃ ], but if it resyllabifies it is realized as [z], e.g. queres [ ʃ ] ‘you want’, queres algo [z] ‘you want something’ (Mateus/d’Andrade 2000, 12, 145; Perini 2002, 15). Thus, both the voicing and the place of articulation of word-final sibilants are conditioned by phrase-level phonology in these varieties of Portuguese. In Spanish dialects with aspiration and deletion of coda /s/ there are sometimes other associated phenomena. For instance, geminates may result from the assimilation of /s/ (and other consonants in the coda) to a following consonant as in Eastern Andalusian Sp. isla [ˈilːa] ‘island’, caspa [ˈkapːa] ‘dandruff’ (cf. Gerfen 2002). A recent phenomenon is the metathesis of the aspiration in Western Andalusian Spanish, e.g. costa [ˈkotʰa] ‘coast’ (cf. Torreira 2012; Ruch/Harrington 2014). In Eastern Andalusian, on the other hand, word-final /s/ is almost categorically deleted, but its underlying presence is manifested in the opening of the preceding vowels, e.g. paso [ˈpaso] vs. pasos [ˈpasɔ]; tiene [ˈtjene] ‘s/he has’ vs. tienes [ˈtjenɛ] ‘you have’, with possible vowel assimilation in the word domain (cf. e.g. Hernández-Campoy/ Trudgill 2002; Penny 2000, 122–126). For both the Canary Islands and Central America, it has been reported that after deleted /s/, voiced obstruents are realized as stops, so that, for instance, las vacas [laˈbːaka] ‘the cows’ may contrast with la vaca [laˈβaka] ‘the cow’ (cf. Amastae 1989; Dorta/Herrera 1993). 2 Generally speaking, in Western Romance, the plural of nouns continues the Latin accusative plural (e.g. ROSĀS , MŪRŌS ). The early loss of /‑s/ in these forms in Eastern Romance made them identical to the singular, so that only the nominative could mark the plural (e.g. ROSAE > It. rose ‘roses’, MŪRĪ > It. muri ‘walls’).

Surface sound and underlying structure: The phonetics-phonology interface

29

In Romanian, weakening of coda consonants occurs only in a limited set of environments in spontaneous, casual speech. For instance, final stop clusters tend to be reduced before word-initial stops: a[kt] [d]e ~ a[kd]e ‘act of’, with an unreleased C1 and variable voicing assimilation in the two stops. A more systematic deletion affects the [l] of the masculine singular definite article, which is enclitic: omul de pe stradă ~ omu de pe stradă ‘the man in the street’. In casual speech [l] may delete even before a vowel-initial word: omul a plecat ~ omu a plecat ~ om[w] a plecat ‘the man left’, omul ăsta ~ omu ăsta ‘this man’ (Chitoran et al. 2014). This variation suggests the development of a more systematic, morphophonological change, whereby the marking of definiteness is transferred over to the vowel /u/. This vowel is the masculine singular desinence vowel. It only surfaces in cliticized forms (om-u-l ‘the man’, om-u-l-ui ‘to/of the man’) and in non-cliticized forms ending in a consonant cluster (patru ‘four’). 2.2.2 Coda nasal place neutralization and assimilation Most Romance languages show a smaller number of phonemic oppositions among nasals in the coda than syllable-initially. Spanish has three nasal phonemes /m n ɲ/ in syllable-initial position. In word-final position, on the other hand, there is only /n/, which before a pause is realized as [n] or [ŋ] depending on the dialect, e.g. pan /pan/ [pan] ~ [paŋ] ‘bread’. Nasals assimilate in place to following consonants, both word-internally and across word boundaries: so[m] pocos ‘they are few’, so[ŋ] grandes ‘they are big’. In Catalan, unlike in Spanish, there is no coda neutralization. The labial, alveolar and palatal nasal contrast word-finally, som ‘we are’, són ‘they are’, any /aɲ/ ‘year’, and, in addition, there is a fourth surface contrastive nasal in word-final (but not in syllable-initial) position, velar [ŋ], sang [saŋ] ‘blood’. However, the alveolar nasal assimilates to the place of articulation of a following consonant, producing contextual neutralization with the other nasal phonemes: só[m] petits ‘they are small’ neutralized with som petits ‘we are small’, but só[ŋ] grans ‘they are big’ vs. som grans ‘we are big’. Nasal place assimilation occurs in Romanian, as well. Romanian has two nasal phonemes /m n/. Word- or morpheme-final /n/ variably assimilates in place to a following stop, for example, the nasal in the preposition în [ɨn] ‘in’: î [ŋ] [k]asă ‘in the house’, î[m] [p]arc ‘in the park’. The same type of assimilatory coarticulation takes place when a final cluster is reduced and the nasal becomes adjacent to the following word-initial stop, e.g. sînt prieteni → sî[m] [p]rieteni ‘they are friends’, sînt pe drum → sî[m] [p]e drum ‘I am on the road’, sînt curat → sî[ŋ] [k]urat ‘I am clean’. In French, we find a quite different type of nasal assimilation, whereby oral stops become nasal in contact with a nasal consonant or vowel, e.g. et demie /edəmi/ → [enmi] ‘and a half’, bombe atomique /bɔ̃batomik/ → [bɔ̃matɔmik] ‘atomic bomb’ (Walker 2001, 135).

30

José Ignacio Hualde and Ioana Chitoran

2.2.3 Coda liquids The alveolar lateral /l/ may be “light” (with the dorsum in the position for a front vowel, cf. Proctor 2011), as in Spanish, French and Italian, or “dark” or velarized (with the dorsum in the position for a back vowel), as in Portuguese and Catalan. Relaxation of the apical constriction of a “clear” lateral produces [j] in varieties of Dominican Spanish, e.g. papel [paˈpej], whereas, if the lateral is “dark”, its weakening results in [w], as in Brazilian Portuguese, e.g. Brasil [braˈziw] vs. brasi[l]eiro ‘Brazilian’. In the diachrony of the Romance languages both types of vocalization are abundantly attested, e.g. MULTU (M ) > Por. muito ‘much’ (> Sp. mucho, with subsequent palatalization), ALTERU (M ) > *[awtro] > Por. outro ‘other’, Sp. otro, Fr. autre, CABALLOS > Old Fr. chevals > Fr. chevaux ‘horses’. The neutralization of lateral and rhotic liquids in the coda is also a common phenomenon, found, for instance, in Andalusian and Caribbean Spanish and in Rome Italian. In several Spanish varieties with liquid neutralization, coda liquids may assimilate to certain following consonants, e.g. carne [ˈkanːe] ‘meat’, pulga [ˈpuɡːa] ‘flea’.

2.3 Palatalization The palatalization of consonants in contact with glides and front vowels is widely attested as a sound change in Romance. Whereas Classical Latin lacked palatals and prepalatals altogether, a whole range of these consonants arose in Romance via palatalization phenomena, sometimes later developing into sibilants and other consonants, e.g. DĪCIT /diːkit/ > Rom. zice /ziʧe/ ‘s/he says’, VĪNEA (M ) > Port. vinha, Sp. viña, Cat. vinya, Fr. vigne, It. vigna ‘vine’, all with /ɲ/ (cf. Alkire/Rosen 2010 for a convenient summary of palatalization sound changes from Latin to Romance). Palatalization has its source in coarticulation, and most frequently affects coronals and dorsals, as in the examples just given, but in some Romance languages even labials have palatalized historically, e.g. SAPIAT > Old Fr. /sapʧə/ > [saʃ ] sache ‘(that) s/he know’, LŪPI > Rom. lupi [lupj] ‘wolves’. Universally, palatalization appears to follow a hierarchy of triggers [j] » [i] » [e]. Palatalization triggered by a (fronted) low vowel, as in French, e.g. CANTĀRE > Fr. chanter ‘to sing’, is less common. Similarly, the trigger usually follows, but palatalization by a preceding glide is found in developments like NOCTE (M ) > /noi ̯te/ > Sp. noche ‘night’. Historical palatalization has produced morphophonological alternations in many Romance languages, e.g. Fr. blanc/blanche ‘whiteM . S G / F. S G ’, It. amico/amici ‘friendM . S G / M . P L ’, Rom. stradă/străzi ‘streetS G / P L ’. As an active process of allophony, we find palatalization of /t/ and /d/ before /i/ in Brazilian Portuguese, e.g. tia [ʧia] ‘aunt’, dia [ʤia] ‘day’. The raising of word-final front vowels (e.g. sublime [sublimi]) results in palatalization before orthographic final , e.g. BP parte [ˈpahʧi] ‘part’, verde [ˈvehʤi] ‘green’.

Surface sound and underlying structure: The phonetics-phonology interface

31

In Québec French, /t/ and /d/ are affricated to [ts] and [dz], respectively, before the high front vowels /i/ and /y/, e.g. petit [ptsi] ‘small’, tu [tsy] ‘you’. Recent sociolinguistic studies of urban continental French (cf. Lodge 2004; Fagyal 2010) document a type of palatalization/affrication that occurs before high front vowels and glides, but results in [ʧ, ʤ], as in Brazilian Portuguese, rather than the Quebec French [ts, dz]: y a [ʤy] monde ‘it’s crowded’, [ʧy] m’as [ʤi]t ‘you told me’. This phenomenon has been associated with the speech of working class youth of immigrant descent, but more recently has been reported in political discourse (cf. Trimaille 2008) and broadcast news (cf. Candea/Adda-Decker/Lamel 2013). The latter study, based on a large speech corpus, verifies that affrication occurs predominantly with the voiceless variant, before [j] and [i]. Some frequent examples are: moi[ʧj]é ‘half’, chré[ʧj]en ‘Christian’, poli[ʧi]que ‘politics’, exécu[ʧi]f ‘executive’. This tendency to palatalize dentals in contact with high front vowels in present-day French may be related to the high articulatory setting of its vowel system, as compared to, for instance, Spanish (Torreira/Ernestus 2011). Probably as a universal phenomenon of coarticulation, velars have a more fronted place of articulation in contact with front vowels than in contact with back vowels (car vs. key). The exaggeration of this coarticulation produces distinctively palatalized allophones, like in Chilean Spanish, e.g. mujer /muˈxer/ [muˈçer] ‘woman’. A common unconditioned weakening phenomenon repeatedly attested in the Romance family is the weakening of /ʧ/ to /ʃ/ and /ʤ/ to /ʒ/ (e.g. in the history of both Portuguese and French). Nowadays /ʧ/ is variably or systematically realized as [ ʃ ] in several Spanish varieties (Southern Andalusia, Northern Mexico, Panama, Chile).

2.4 Initial fortition As already mentioned, consonant fortition has been observed to occur at the phonetic level as a correlate of the phrase-initial position (Fougeron/Keating 1997). Interestingly, a correlate of emphatic stress in French is consonant gemination: quel [kː]rétin ‘what an idiot!’, c’est [fː]ormidable ‘it’s great!’ (cf. Walker 2001, 131). The phonologization of phrase-initial fortition (as word initial), however, appears to be rare. One example would be the fortition of rhotics in Spanish, where the trill occurs to the exclusion of the tap in word-initial position. Catalan also underwent fortition of word-initial /l/ (which later evolved to palatal /ʎ/, like word-internal geminate /lː/), and in Asturian/Leonese both word-initial /l/ and /n/ show this sound change, e.g. Cat. llop /ʎop/ ‘wolf’, Ast. llobu /ʎobu/ ‘wolf’, ñome /ɲome/ ‘name’, but this is no longer an active process in these languages. Word- and syllable-initial yod strengthened in Late Latin, generally converging with the results of /ɡ/ before a front vowel, e.g. IUNIU (M ) > Fr. juin, Port. junho, It. giugno ‘June’. In modern Spanish both syllable-initial [j] and [w] undergo optional fortition, e.g. yegua [ˈɟeɣwa] ‘mare’, huevo [ˈɡweβo] ‘egg’.

32

José Ignacio Hualde and Ioana Chitoran

3 Vowels Having considered the main phenomena affecting consonants, in this section we examine other processes that target vowels. We classify these processes in three major types: vowel reduction in unstressed syllables, vowel-to-vowel coarticulation and assimilation (including coalescence and harmony), and vowel epenthesis. As we have done for the consonants, productive synchronic processes will be considered in the light of completed sound changes of the same type.

3.1 Vowel reduction in unstressed syllables An important parameter of variation among the Romance languages is the extent to which unstressed vowels are reduced in their duration and centralized in their quality. Iberian Portuguese and Spanish offer a striking contrast in this respect. Even though these are two very closely related languages and their phonological syllable structure is rather similar, they have radically different rhythms. Whereas in Iberian Spanish differences in duration and vowel quality between stressed and unstressed vowels are relatively small, to the extent that it is difficult to identify systematic ways in which they differ in their quality (Nadeu 2014), in Iberian Portuguese, unstressed vowels are greatly reduced and often dropped, mirroring in many respects completed sound changes in a language like French. Although most Spanish varieties have very little reduction of unstressed vowels, the reduction in duration, devoicing and deletion of unstressed vowels in certain positions is a feature of both Mexican (Lope Blanch 1963) and Andean Spanish (Delforge 2008), e.g. cafecito [kafˈsito] ‘a little coffee’. A different type of reduction, with vowel raising, has been found in some varieties of Judeo-Spanish in contact with Slavic languages, such as that of Bulgaria, where both /a/ and /o/ raise in unstressed syllables (cf. Gabriel/Kireva 2014). This latter phenomenon finds a parallel in Central Catalan, where the seven-vowel system found in stressed syllables is reduced to /i ə u/ in most unstressed syllables, giving rise to many morphophonological alternations. Schwa deletion in French is a very complex phenomenon. It is systematic, which qualifies it as a lexical phenomenon, but it is also subject to phonetic, prosodic, dialectal, idiolectal, sociolinguistic, and stylistic constraints. It has been extensively studied and is still not fully understood. Alternating schwas occur word-internally (semaine [səmɛn] ~ [smɛn] ‘week’) and across word boundaries ( je ne sais pas [ʒənəsepa] ~ [ʒənsepa] ‘I don’t know’). The conclusion at this point is that schwa alternation is lexical and phonological (cf. Bürki/Ernestus/Frauenfelder 2010). Variants without schwa are reported in careful speech as well as in casual speech (Côté/Morrison 2007). Nevertheless, Bürki et al. (2011), based on the analysis of a large spoken corpus, conclude that schwa in connected speech also undergoes phonetic reduction typical of interface phenomena.

Surface sound and underlying structure: The phonetics-phonology interface

33

In contrast with French, in Central Catalan, schwa, which replaces /a ɛ e/ in unstressed syllables (e.g. renta [ˈrentə] ‘s/he washes’, rentar [rənˈta] ‘to wash’), is a stable vowel and is not subject to deletion. A recurrent phenomenon in Romance is the raising of mid vowels in pretonic syllables. This phenomenon has the status of a regular sound change in standard Italian, e.g. SECŪRU (M ) > It. sicuro ‘safeM . S G ’, FENESTRA (M ) > It. finestra ‘window’ (Alkire/Rosen 2010, 80). In Brazilian Portuguese it is a variable rule, where variants with pretonic high vowels are seen as less formal (without being stigmatized), and are subject to a number of phonological conditions and lexical marking, e.g. perigo [e]~[i] ‘danger’, tomate [o]~[u] ‘tomato’, but verdade [e], *[i] ‘truth’ (Perini 2002, 37– 38). In Spanish and closely related dialects the phenomenon is nowadays stigmatized as “rural” and it is mostly found as an assimilatory phenomenon (cf. below), when the stressed vowel is high, e.g. comer [koˈmeɾ] ‘to eat’ vs. comería [kumiˈɾia] ‘I would eat’ (Penny 1978, 86). As for word-final unstressed syllables, the pan-Romance tendency is to have a smaller number of contrasts in this position than elsewhere. Thus Italian only has four word-final unstressed vowels /i e a o/, Spanish three (leaving aside a few exceptions) /e a o/, and Central Catalan also three /i ə u/. In several Spanish varieties, both in Spain and in Latin America, /e/ and /o/ tend to rise in this position. This is a stigmatized “rural” phenomenon in Spanish. In Brazilian Portuguese, on the other hand, this is a regular rule of pronunciation: mid vowels become high in unstressed word-final syllables, e.g. /ɡato(s)/ Sp. [ˈɡato(s)] vs. BP [ˈɡatu(s)] ‘cat(s)’. This raising process feeds palatalization, BP parte [pahʧi] ‘part’. Finally the devoicing of phrase-final (or word-final) vowels has different degrees of incidence or regularity in different Romance languages. Portuguese shows pervasive devoicing and deletion of word-final vowels and differs strikingly in this respect from Spanish. Phrase-final devoicing in French primarily affects high vowels and has been well documented and studied experimentally (cf. Fagyal/Moisset 1999; Smith 2003). It should be noted, however, that phrase-final devoicing is not considered an instance of vowel reduction in French, but rather a prosodic and discourse marker. Meunier/Espesser (2011) found that vowels in word-final syllables are less often reduced than preceding ones, in terms of duration and centralization. Similarly, Torreira/Ernestus (2011) show that in French phrase-medial devoicing shows characteristics of vowel reduction, such as shortening and increased coarticulation.

3.2 Vowel-to-vowel coarticulation 3.2.1 Vowel coalescence Vowel coalescence in Romanian occurs in restricted environments in spontaneous speech. We will consider the example of vowel-final function words, such as the

34

José Ignacio Hualde and Ioana Chitoran

prepositions pe ‘on’ and de ‘of’. Coalescence takes place between [e] and a following stressed [a] or [ʌ], but not with other vowels. The following are some representative examples from a casual style of speech: p[e] [a]sta ~ p[e̯a]sta / p[a]sta ‘on this oneF ’, p[e] [ʌ]sta ~ p[ʌ]sta ‘on this oneM ’. This type of coalescence does not seem to occur across lexical words, or at least it is not as salient. It should be noted that all the connected speech phenomena reported here still need to be studied experimentally, preferably in spontaneous speech corpora. In Spanish, sequences of vowels are reduced in connected speech (cf. e.g. Hualde 2005, 89–94). Unstressed high vowels glide in contact with other vowels, e.g. m[ja]migo ‘my friend’, t[wa]buelo ‘your grandfather’, and sequences containing non-high vowels may be reduced to a single syllable in various ways, including deletion, coalescence and gliding, e.g. te acuerdas [taˈkwerðas] ~ [tjaˈkwerðas] ‘you remember’ (cf. Hualde/Torreira/Simonet 2008 for experimental results). In Portuguese these phenomena of syllable coalescence are perhaps more systematic. Thus, for Brazilian Portuguese Perini (2002, 50) states that when both vowels across a word boundary are unstressed, /i/ and /u/ become glides before a non-identical vowel and in all other cases the first vowel is deleted, as in his examples bule amassado [ˈbuljamaˈsadu] ‘dented coffeepot’, casa enorme ‘huge house’ [ˈkazinɔhmi]. This situation seems to also obtain in European Portuguese (Mateus/d’Andrade 2000, 146; Vigário 2003, 104–114). Within words, the reduction to diphthongs of sequences where an unstressed mid vowel is followed by another vowel is widespread in Latin American Spanish, e.g. pelear [peˈljaɾ] ‘to fight’. The phenomenon is subject to lexical and phonological conditions that are still not well understood. For instance, /ea/ is much more readily reduced to [ja] in, for example, golpeamos ‘we strike’ than in leamos ‘we readS B J V ’.

3.2.2 Vowel assimilation and harmony-type effects A well-known vowel-to-vowel assimilation phenomenon in Italo-Romance is metaphony, which generally involves the raising or diphthongization of stressed vowels under the assimilatory effects of a word-final high vowel, e.g. /verde/ ‘greenS G ’ vs. /virdi/ ‘greenP L ’, /pɛde/ ‘foot’ vs. /pjɛdi/ ‘feet’. Metaphonic phenomena are found in both northern and southern Italo-Romance (cf. e.g. Maiden 1991), but not in standard Italian. Similar phenomena are also found in Asturian and Cantabrian dialects in northern Spain (e.g. in Lena Asturian /ɡata/ ‘she-cat’ vs. /ɡetu/ ‘he-cat’; cf. Neira Martínez 1955; Hualde 1989; 1998). Romanian morphophonology exhibits a process of diphthong/singleton alternation that fits the description of metaphony and has consequently been analyzed as such (cf. Chitoran 2002; Renwick 2012, including a comparative experimental study of Romanian and Italian). Such alternations can be found, for example, in nominal morphology in singular-plural forms (e.g. poartă [ˈpoartʌ] ~ porți [ˈportsj] ‘gate/

Surface sound and underlying structure: The phonetics-phonology interface

35

gates’) and in verb morphology (e.g. treacă [ˈtreakʌ] ‘he/they pass(es)S B J V ’ – treci [ˈtreʧ j] ‘you passI ND ’). The stressed singleton vowel is conditioned by the high vowel in the following syllable, as in the classic case of metaphony, but also by the front vowel /e/: treacă [ˈtreakʌ] ‘he/they pass(es)S B J V ’ ~ trece [ˈtreʧe] ‘s/he passesI N D ’, deasă [ˈdeasʌ] ‘thickF. S G ’ ~ dese [ˈdese] ‘thickF. P L ’. These systematic synchronic alternations are the morphologized outcomes of a sound change involving vowel diphthongization under stress (SERA (M ) > Rom. seară [ˈsearʌ] ‘evening’); cf. Alkire/Rosen (2010) for a detailed account. A related phenomenon, the lowering of stressed mid vowels under the influence of final low vowels, is found in Portuguese, e.g. fam[o]so ‘famousM . S G ’ vs. fam[ɔ]sa ‘famousF. S G ’. In running speech, vowel-to-vowel coarticulation is attested variably in Romanian in forms such as bun[ʌ] ziua ~ bun[e] ziua ‘good day’ (greeting), uit[ʌ]-te ~ uit[e]-te ‘look!’. As these examples show, a final central vowel tends to be fronted before a front vowel in the following word. ‘To look’ is a reflexive verb of the first conjugation (a se uita), and should regularly form the imperative in [ʌ]. However, when the imperative of this verb is used without the reflexive pronoun, the form is exclusively uite! [ˈujte]. This can be attributed to lexicalization of the effects of coarticulation originally induced by the reflexive pronoun te. Vowel-to-vowel coarticulation can interact in Romanian with an effect of labial centralization, which is attested as a sound change, but is not known to be synchronically active. Historically, /e/ after a labial becomes central, unless the following vowel is front (Alkire/Rosen 2010, 258). This sound change explains forms such as *MĒLU > Rom. măr /mʌr/ ‘apple’ (cf. It. mela) vs. *MĒLE > Rom. mere /mere/ ‘apples’. It is not clear what the phonetic basis of labial centralization may be, but casual spontaneous speech seems to exhibit a surprisingly similar tendency. A preliminary experimental study of this phenomenon based on a large corpus is presented in Chitoran et al. (2016). We report here some examples typical of a very casual style of speech, involving the vowel of the preposition pe ‘on’, pe urmă ~ p[ʌ] urmă ‘after that’, pe el ~ [pʌjel] ‘himAC C ’, pe la mama ~ p[ʌ] la mama ‘at mother’s house’, pe seară ~ p[ʌ] seară ‘towards evening’, pe ziuă ~ p[ʌ] ziuă ‘during the day’, pe iarnă ~ p[ʌ] iarnă ‘during the winter’. Notice that vowel centralization in Romanian is heard before both back and front vowels; therefore it cannot be attributed to vowel-to-vowel coarticulation, at least not exclusively. However, vowel backing is attested independently, consistent with the vowel fronting presented earlier. We can see this in the behaviour of the preposition de ‘of’, where no labial is present, e.g. de unde ~ d[ʌ] unde ‘from where’. This is most commonly observed in a very casual pronunciation of the very colloquial phrase da de unde ~ da d[ʌ] unde ‘on the contrary’. Evidence that this centralization is due to vowel backing comes from its absence when the following vowel is front, e.g. de seară, *d[ʌ] seară, as in rochie de seară ‘evening dress’, de zi, *d[ʌ] zi ‘of the day’, de iarnă, *d[ʌ] iarnă ‘of the winter’.

36

José Ignacio Hualde and Ioana Chitoran

More extensive phenomena of vowel harmony, characterizing in some cases whole phonological word domains, have been described for Asturian/Cantabrian Ibero-Romance (cf. e.g. Hualde 1989), for some Italo-Romance varieties (cf. e.g. Nibert 1998; Mascaró 2011), for Valencian Catalan (Jiménez 1998) and, as mentioned above, for Eastern Andalusian/Murcian (cf. Hernández-Campoy/Trudgill 2002). The distribution of lower and higher mid vowels in Southern French is said to be dictated by the so-called loi de position ‘law of position’, i.e. high-mid vowels in open syllables, low-mid vowels in closed syllables. In other varieties of French the distribution is less predictable and depends in part on harmonic rules, e.g. bête [bɛt] ‘animal; silly’, but bêtise [betiz] ‘silliness’, where the vowel of the root raises under the influence of the high vowel /i/ in the following syllable (cf. Walker 2001, 54). Above, in section 3.1, we made reference to the raising of pretonic mid vowels, which is sometimes an assimilatory phenomenon conditioned by the presence of a stressed high vowel. The opposite assimilatory phenomenon, the lowering of pretonic mid vowels, is found in Portuguese. In pretonic syllables there is no contrast between /e/ and /ɛ/ or /o/ and /ɔ/. Normally, mid vowels are pronounced as midhigh [e], [o] and may even be raised to [i], [u]. However, we find pretonic [ɛ] when the stressed vowel is /ɛ/, as in Pelé [pɛˈlɛ], and [ɔ] when the stressed vowel is /ɔ/, as in bolota [bɔˈlɔtɐ] ‘acorn’ (cf. Perini 2002, 37).

3.3 Vowel epenthesis Perhaps the most remarkable phenomenon of vowel epenthesis in Romance is the prosthesis of /e/ before word-initial /sC/ clusters. In Spanish and Catalan this is a synchronically active, obligatory rule of pronunciation, a phonotactic constraint that applies to borrowings exceptionlessly (e.g. Sp. estrés ‘stress’) and is observable in the second-language pronunciation of native speakers of these languages. For Ibero-Romance speakers sequences like [st-] and [est-] are not distinct (for experimental evidence, cf. Hallé et al. 2013), so that, for instance, the English words state and estate may perceptually be homophones. Portuguese has the same neutralization, but in varieties with palatalization of preconsonantal sibilants, deletion of the initial vowel is frequent in colloquial speech, e.g. espaço [ˈʃpasu] ‘space’ (Mateus/ d’Andrade 2000, 43). Vocalic prosthesis before word-initial /sC/ clusters is a phenomenon with a surprisingly long pedigree. French does not have this phonotactic constraint, but evolutions like SPATHA (M ) > Fr. épée ‘sword’, SPATIU (M ) > Fr. espace ‘space’, etc., show that it once did. At some historical point, however, /sC/-initial sequences became acceptable in French and words like statue started being incorporated into the lexicon without adaptation. At most, one can speak of a lexical morphophonological rule in French relating words like espace ‘space’ and spacieux ‘spaciousM . SG ’.

Surface sound and underlying structure: The phonetics-phonology interface

37

A historically related phenomenon functioning as a phonotactic rule in word sequences is found in some conservative varieties of Italian that display alternations like strada ‘road’, in istrada ‘on the road’, scritto ‘written’, per iscritto ‘in writing’ (cf. Sampson 2010).

4 Conclusion We have provided a general overview of phonetic/phonological phenomena in Romance. We have chosen examples that we think are particularly common in the Romance languages or particularly interesting, since obvious reasons of space prevent us from being exhaustive in this chapter. An interesting fact, already noticed by Delattre (1946), is that, often, what appears as a variable process at the phoneticsphonology interface in one Romance language may mirror a completed sound change in another (e.g. syllable or word-final /s/ aspiration).

5 References Alkire, Ti/Rosen, Carol (2010), Romance languages. A historical introduction, Cambridge, Cambridge University Press. Amastae, John (1989), “The intersection of s-aspiration/deletion and spirantization in Honduran Spanish”, Language Variation and Change 1, 169–183. Bürki, Audrey/Ernestus, Mirjam/Frauenfelder, Ulrich H. (2010), “Is there only one ‘fenêtre’ in the phonological lexicon? On-line evidence on the nature of phonological representations of pronunciation variants for French schwa words”, Journal of Memory and Language 62, 421–437. Bürki, Audrey/Fougeron, Cécile/Gendrot, Cédric/Frauenfelder, Ulrich H. (2011), “Phonetic reduction versus phonological deletion of French schwa. Some methodological issues”, Journal of Phonetics 39, 279–288. Candea, Maria/Adda-Decker, Martine/Lamel, Lori (2013), “Recent evolution of non-standard consonantal variants in French broadcast news”, in: Frédéric Bimbot/Christophe Cerisara/Cécile Fougeron/Guillaume Gravier/Lori Lamel/François Pellegrino/Pascal Perrier (edd.), Interspeech 2013. Proceedings of the 14th Annual Conference of the International Speech Communication Association, Lyon, France, August 25–29, 2013, 412–416, http://www.isca-speech.org/archive/ archive_papers/interspeech_2013/i13_0412.pdf (06.05.2016). Canepari, Luciano (1979), Introduzione alla fonetica, Torino, Einaudi. Carrasco, Patricio/Hualde, José Ignacio/Simonet, Miquel (2012), “Dialectal differences in Spanish voiced obstruent allophony. Costa Rican vs. Iberian Spanish”, Phonetica 69, 149–179. Chitoran, Ioana (2002), The phonology of Romanian. A constraint-based approach, Berlin/New York, De Gruyter. Chitoran, Ioana/Vasilescu, Ioana/Vieru, Bianca/Lamel, Lori (2014), “Analyzing linguistic variation in a Romanian speech corpus through ASR errors”, Paper presented at Laboratory Approaches to Romance Phonology (LARP) 7, Aix-en-Provence, 3–5 September 2014. Chitoran, Ioana/Vasilescu, Ioana/Vieru, Bianca/Lamel, Lori (2016), “Connected speech in Romanian. A window into the evolution of a sound system.” Talk presented at the 46th Linguistic Symposium on Romance Languages (LSRL 46), Stonybrook University, 31 March–3 April 2016.

38

José Ignacio Hualde and Ioana Chitoran

Côté, Marie-Hélène/Morrison, Geoffrey S. (2007), “The nature of the schwa/zero alternation in French clitics. Experimental and non-experimental evidence”, Journal of French Language Studies 17, 159–186. Cravens, Thomas (1984), “Intervocalic consonant weakening in a phonetic-based phonology. Foleyan hierarchies and the Gorgia Toscana”, Theoretical Linguistics 11, 269–310. Delattre, Pierre (1946), “Stages of Old French phonetic changes observed in Modern Spanish”, Publications of the Modern Language Association 61, 7–41. Delforge, Ann M. (2008), “Unstressed vowel reduction in Andean Spanish”, in: Laura Colantoni/ Jeffrey Steele (edd.), Selected proceedings of the 3rd conference on laboratory approaches to Spanish phonology, Somerville, Cascadilla, 107–124. Dorta, Josefa/Herrera, Juana (1993), “Experimento sobre la discriminación auditiva de las oclusivas tensas grancanarias”, Estudios de Fonética Experimental 5, 163–188. Fagyal, Zsuzsanna (2010), Accents de banlieue. Aspects prosodiques du français populaire en contact avec les langues de l’immigration, Paris, L’Harmattan. Fagyal, Zsuzsanna/Moisset, Christine (1999), “Sound change and articulatory release. Where and why are high vowels devoiced in Parisian French”, in: John J. Ohala/Yoko Hasegawa/Manjari Ohala/Daniel Glanville/Ashley C. Bailey (edd.), Proceedings of 14th International Congress of Phonetic Sciences, vol. 1, San Francisco, Berkeley, The University of California, 309–312. Fougeron, Cécile/Keating, Patricia A. (1997), “Articulatory strengthening at edges of prosodic domains”, Journal of the Acoustical Society of America 101, 3728–3740. Gabriel, Christoph/Kireva, Elena (2014), “Speech rhythm and vowel raising in Bulgarian JudeoSpanish”, in: Nick Campbell/Dafydd Gibbon/Daniel Hirst (edd.), Proceedings of Speech Prosody 2014, Dublin, Trinity College, 728–732. Gerfen, Chip (2002), “Andalusian codas”, Probus 14, 303–333. Giannelli, Luciano/Savoia, Leonardo M. (1978), “L’indebolimento consonantico in Toscana I”, Rivista Italiana di Dialettologia 2, 25–58. Giannelli, Luciano/Savoia, Leonardo M. (1979/1980), “L’indebolimento consonantico in Toscana II”, Rivista Italiana di Dialettologia 3/4, 39–101. Hallé, Pierre/Seguí, Juan/Domínguez, Alberto/Cuetos, Fernando/Isel, Frédéric/Shen, Weilin (2013), “Do smid and esmid sound the same? A cross-language comparison between Spanish and French listeners”, Paper presented at PPLC (Phonetics, Phonology, and Languages in Contact), 21–23 August 2013, Paris. Hernández-Campoy, Juan M./Trudgill, Peter (2002), “Functional compensation and Southern Peninsular Spanish /s/ loss”, Folia Linguistica Historica 23, 31–57. Hualde, José Ignacio (1989), “Autosegmental and metrical spreading in the vowel-harmony systems of northwestern Spain”, Linguistics 27, 773–805. Hualde, José Ignacio (1998), “Asturian and Cantabrian metaphony”, Rivista di Linguistica 10, 99– 108. Hualde, José Ignacio (2005), The sounds of Spanish, Cambridge, Cambridge University Press. Hualde, José Ignacio/Nadeu, Marianna (2011), “Lenition and phonemic overlap in Rome Italian”, Phonetica 68, 215–242. Hualde, José Ignacio/Prieto, Pilar (2014), “Lenition of intervocalic alveolar fricatives in Catalan and Spanish”, Phonetica 71, 109–127. Hualde, José Ignacio/Simonet, Miquel/Nadeu, Marianna (2011), “Consonant lenition and phonological recategorization”, Laboratory Phonology 2, 301–329. Hualde, José Ignacio/Torreira, Francisco/Simonet, Miquel (2008), “Postlexical contraction of nonhigh vowels in Spanish”, Lingua 118, 1906–1925. Jiménez, Jesús (1998), “Valencian vowel harmony”, Rivista di Linguistica 10, 137–161. Jones, Michael A. (1997), “Sardinian”, in: Martin Maiden/Mair Parry (edd.), The dialects of Italy, London, Routledge, 376–384.

Surface sound and underlying structure: The phonetics-phonology interface

39

Lodge, R. Anthony (2004), A sociolinguistic history of Parisian French, Cambridge, Cambridge University Press. Lope Blanch, Juan (1963), “Sobre las vocales caedizas del español mexicano”, Nueva Revista de Filología Hispánica 17, 1–20. Loporcaro, Michele (1997), L’origine del raddoppiamento fonosintattico. Saggio di fonologia diacronica romanza, Basel/Tübingen, Francke. Loporcaro, Michele (2011), “Syllable, segment and prosody”, in: Martin Maiden/John Charles Smith/ Adam Ledgeway (edd.), The Cambridge history of the Romance languages, vol. 1: Structures, Cambridge, Cambridge University Press, 50–108. Maiden, Martin (1991), Interactive morphonology. Metaphony in Italy, London, Routledge. Marotta, Giovanna (2001), “Non solo spiranti. La ‘gorgia toscana’ nel parlato di Pisa”, L’Italia dialettale 62, 27–60. Marotta, Giovanna (2008), “Lenition in Tuscan Italian (Gorgia Toscana)”, in: Joaquim Brandão de Carvalho/Tobias Scheer/Philippe Ségéral (edd.), Lenition and fortition, Berlin/New York, De Gruyter, 235–272. Mascaró, Joan (2011), “An analysis of stress-dependent harmony in Servigliano”, Probus 23, 21–55. Mateus, Maria Helena/d’Andrade, Ernesto (2000), The phonology of Portuguese, Oxford, Oxford University Press. Meunier, Christine/Espesser, Robert (2011), “Vowel reduction in conversational speech in French. The role of lexical factors”, Journal of Phonetics 39, 271–278. Nadeu, Marianna (2014), “Stress- and speech rate-induced vowel quality variation in Catalan and Spanish”, Journal of Phonetics 46, 1–22. Neira Martínez, Jesús (1955), El habla de Lena, Oviedo, Diputación de Oviedo. Nibert, Holly (1998), “Processes of vowel harmony in the Servigliano dialect of Italian. A comparison of two non-linear proposals for the representation of vowel height”, Probus 10, 67–101. Nocchi, Nadia/Schmid, Stephan (2007), “Lenition of voiceless fricatives in two varieties of southern Italian”, in: Jürgen Trouvain (ed.), Proceedings of 16th International Congress of the Phonetic Sciences, Saarbrücken, Universität des Saarlandes, 1497–1500. Oftedal, Magne (1985), Lenition in Celtic and in Insular Spanish. The second voicing of stops in Gran Canaria, Oslo, Universitetsforlaget. Pape, Daniel/Jesus, Luis M. T. (2011), “Devoicing of phonologically voiced obstruents. Is European Portuguese different from other Romance languages?”, in: Wai-Sum Lee/Eric Zee (edd.), Proceedings of the 17th International Congress of Phonetic Sciences 2011, Hong Kong, China, 1566–1569, httpsː//www.internationalphoneticassociation.org/icphs-proceedings/ICPhS2011/ OnlineProceedings/RegularSession/Pape/Pape.pdf (06.05.2016). Penny, Ralph (1978), Estudio estructural del habla de Tudanca, Tübingen, Niemeyer. Penny, Ralph (2000), Variation and change in Spanish, Cambridge, Cambridge University Press. Perini, Mário (2002), A reference grammar of modern Portuguese, New Haven/London, Yale University Press. Proctor, Michael (2011), “Towards a gestural characterization of liquids. Evidence from Spanish and Russian”, Laboratory Phonology 2, 451–485. Renwick, Margaret Elspeth Lambert (2012), Vowels of Romanian. Historical, phonological and phonetic studies, PhD Dissertation, Cornell University, http://conf.ling.cornell.edu/peggy/ Renwick_2012_Vowels-of-Romanian.pdf (06.05.2016). Ruch, Hanna/Harrington, Jonathan (2014), “Synchronic and diachronic factors in the change from pre-aspiration to post-aspiration in Andalusian Spanish”, Journal of Phonetics 45, 12–25. Sampson, Rodney (2010), Vowel prosthesis in Romance. A diachronic study, Oxford, Oxford University Press. Smith, Caroline (2003), “Vowel devoicing in contemporary French”, Journal of French Language Studies 13, 177–194.

40

José Ignacio Hualde and Ioana Chitoran

Sorianello, Patrizia (2001), “Un’analisi acustica della ‘gorgia’ fiorentina”, L’Italia dialettale 62, 61– 94. Torreira, Francisco (2012), “Investigating the nature of aspirated stops in Western Andalusian Spanish”, Journal of the International Phonetic Association 42, 49–63. Torreira, Francisco/Ernestus, Mirjam (2011), “Realization of voiceless stops and vowels in conversational French and Spanish”, Laboratory Phonology 2, 331–353. Torreira, Francisco/Ernestus, Mirjam (2012), “Weakening of intervocalic /s/ in the Nijmegen Corpus of Casual Spanish”, Phonetica 69, 124–148. Trimaille, Cyril (2008), “Who’s not palatalizing? Trying to understand the status of palatalized variants in French”, Paper presented at the 8th Conference of the HDLS (High Desert Linguistics Society Conference), Albuquerque, NM, 6–8 November 2008. Vigário, Marina (2003), The prosodic word in European Portuguese, Berlin/New York, De Gruyter. Villafaña Dalcher, Christina (2006), Consonant weakening in Florentine Italian. An acoustic study of gradient and variable sound change, PhD dissertation, Georgetown University. Villafaña Dalcher, Christina (2008), “Consonant weakening in Florentine Italian. A cross-disciplinary approach to gradient and variable sound change”, Language variation and change 20, 275– 316. Walker, Douglass (2001), French sound structure, Calgary, University of Calgary Press. Weinrich, Harald (1958), Phonologische Studien zur romanischen Sprachgeschichte, Münster, Aschendorff.

Marina Vigário

2 Segmental phenomena and their interactions: Evidence for prosodic organization and the architecture of grammar Abstract: Various areas of segmental phonology in the Romance languages are revealing about the architecture and functioning of grammar and phonological component. This chapter presents a selection of phenomena from different Romance languages, especially Ibero-Romance, that illustrate main interactions of segmental phonology with other parts of phonology (essentially, suprasegmental features and prosodic structure), as well as with morphology and the lexicon. Specific segmental phenomena related to word frequency and loan words are also discussed. Keywords: segmental process, suprasegmentals, prosodic domain, lexical process, post-lexical process

1 Introduction The diversity of types of segmental phenomena in the Romance languages is revealing about the architecture and functioning of grammar and the phonological component. In this chapter we review selected phenomena from different Romance languages that illustrate the main interactions of segmental phonology with non-segmental phonology and with other parts of grammar and that exhibit disparate properties, pointing to a typology of phonological processes. Although several languages are considered, we will intentionally allude to Portuguese and non-standard dialects data whenever possible. We will start by considering a number of segmental phenomena that are constrained by various types of suprasegmental information (section 2). We will then proceed with an illustration of processes that depend on the location of particular segments within a prosodic domain (section 3). Segmental processes that are not purely phonological, in the sense that they are not just sensitive to phonological information, are exemplified in section 4, where we inspect some of the interactions between segmental phonology and morphology and the lexicon, as well as the phonology of highly frequent words and loanwords. We conclude in section 5 with a summary of the main points addressed in this chapter. Acknowledgement: The author gratefully acknowledges the Portuguese National Funding Agency for Science, Research and Technology (Fundação para a Ciência e Tecnologia de Portugal, FCT, project “Interactive atlas of the prosody of Portuguese”, PTDC/CLE-LIN/119787/2010).

42

Marina Vigário

2 Segmental phenomena and suprasegmental information In many Romance languages, various processes illustrate the interaction between segmental and suprasegmental phonology. For instance, both the presence and the absence of word-level stress and higher levels of prominence, as well as tonal configurations constrain the realization of segments in multiple ways, as we will see in the next subsections.

2.1 Segmental phenomena in word-level stressed environments There are many segmental effects of word-level stress in Romance languages and dialects. The diphthongization of stressed vowels is one such phenomenon, widely found across the Romance area, with varying properties across languages and varieties (cf. Loporcaro 2011a for an overview). Different types of diphthongization affecting stressed vowels occurred early on in the formation of the Romance varieties. In some cases, diphthongization was contextually determined or sensitive to syllable structure, as in French, Florentine or Neapolitan (e.g. Loporcaro 2011a). This type of diphthongization is also active today in varieties of Portuguese spoken in Azores and Madeira and in small areas of mainland Portugal (Martins/ Vitorino 1989; Segura/Saramago 1999; 2001). Examples in (1), taken from Martins/ Vitorino (1989), illustrate the phonological conditioning of the process in Terceira (an island of the archipelago of Azores). The stressed vowel is either palatalized or velarized, depending on the place features of the preceding high vowel or glide, irrespective of any intervening consonants (1b–c). That the process does not apply if pretonic vowel is not high is shown in the examples in (1a). Finally, the data (1c) show that the phenomenon operates even if the conditioning pretonic vowel does not surface due to the (optional) extreme reduction of unstressed vowels. (1)

EPor. (Terceira) a. a casa é baixo

[ɐ ˈkazɐ] [ɛ ˈbaʃu]

‘the house’ ‘(it) is low’

b.

em casa por baixo estão fartas

[ĩ ˈkjazɐ] [puɾ ˈbwaʃ u] [ˈtɐ̃w ̃ ˈfwaɾtɐʃ ]

‘at home’ ‘under’ ‘(theyF) are tired’

c.

vitelo pevides comer

[vˈtjɛlu] [pˈvwidɨʃ ] [kˈmweɾ]

‘calfM ’ ‘kernel’ ‘to eat’

Segmental phenomena and their interactions

43

In other cases, diphthongization depends on stress alone. This is the case of Spanish diphthongization, historically responsible for the widespread alternations found between [je we] and [e o] in stressed and unstressed positions, respectively (but cf. Loporcaro 2011a, 121 for the view that contextual conditioning may have preceded generalized diphthongization in Spanish). Pairs of morphologically related words are very common in Spanish, like puedo/podemos ‘(I/we) can’ and puerta/portero ‘door/doorkeeper’. Today this type of diphthongization can no longer be considered an active phonological rule in Spanish, and it seems more adequate to analyze these alternations assuming a lexical diphthong, that usually reduces (monophthongizes) in stressless position (but cf. the discussion in Albright/Andrade/Hayes 2001 and Eddington 2012). As has been widely noticed, although this type of alternations is very common in Spanish, there are also many exceptions to the reduction of the diphthong in stressless position, as in viejito ‘old man’ or arriesgar ‘to risk’ (Albright/ Andrade/Hayes 2001; Hualde 2005; Eddington 2012). Hualde (2005, 193–200) reports that derived words with specific suffixes are usually associated either with the reduction of the diphthong (as -al, like in dental/diente ‘dental/tooth’) or allow for both options (stressless syllables may exhibit both reduction and the diphthong), and evaluative suffixes do not usually trigger reduction of the diphthongs. Diphthongization triggered by stress alone is also found today in some dialects of European Portuguese (EPor), such as the varieties spoken in Oporto, Braga and other Northern regions of Portugal (e.g. Cintra 1971; Segura/Saramago 2001; Rodrigues 2002). The available descriptions report that only the mid vowels [e o] are affected (cf. the examples in 2, adapted from Segura/Saramago 2001 and Rodrigues 2002). (2)

EPor.

dor nervoso quente inocente

[ˈdwoɾ] [nɨɾˈβwozu] [ˈkjẽt] [inuˈsjẽt]

‘pain’ ‘nervous’ ‘hot’ ‘innocent’

flor jogo peso dizer

[ˈflwoɾ] [ˈʒwoɡu] [ˈpjezu] [diˈzjeɾ]

‘flower’ ‘game’ ‘weight’ ‘to say’

Rodrigues (2002) claims that the phenomenon found in the Northern dialects of Portuguese is best analyzed as an instance of spreading of the vocalic features from the nucleus of the stressed syllable onto the preceding onset position: when the nucleus is labial, the preceding consonant acquires a labial secondary articulation, and when it is palatal, the consonants’ secondary articulation is palatal as well. It is unclear how this analysis relates to accounts of diphthongization in other Romance languages, such as Chitoran’s (2002), who treats diphthongization in Romanian as an instance of lowering of stressed vowels, and who highlights the role of a constraint pressing vowels to lower when stressed. Interestingly enough, Rodrigues (2002) mentions that in the production data she collected the phenomenon is not found when a stressed [o] is part of a falling diphthong. In the spirit of Chitoran’s approach to the phenomenon, this may be seen as a consequence of the fact that

44

Marina Vigário

the constraint that forces mid stressed vowels to surface as diphthongs is already satisfied when a basic falling diphthong is present. In the Portuguese dialects where it occurs, this type of diphthongization is an active, optional, regular process, and there are no signs of contextual segmental conditioning. Furthermore, preliminary observations of data collected in Oporto region (Ermesinde and Gião), from the ongoing project Interactive atlas of the prosody of Portuguese (Frota 2010–2015), suggest that not only word stress but also higher-level prominence may play a role in the definition of the context that favors diphthongization, as it seems that the phenomenon is found mainly in syllables bearing phrasal-level prominence (possibly, intonational phrase prominence). The examples in (3), as produced by two speakers of Ermesinde (Oporto), illustrate this. (3) EPor. a. Não vem, o teu avô? ‘Isn’t he coming, your grandpa?’

av[wo]

b. O João era mesmo ganancioso, [. . .] m[e]smo . . . ganaci[wo]so ‘João was really greedy, [. . .]’ To the best of our knowledge the prosodic conditions on this phenomenon remain unstudied to date. This kind of data may contribute to the understanding of possible sources for the genesis of diphthongization in Romance and in other languages as well, as aspects of the phonetics and phonology of syllables in intonational phrase nuclear position (e.g. final lengthening and pitch accent assignment) may, at least in some cases, play or have played a crucial role.1 A phenomenon that may relate to diphthongization is the lengthening of stressed vowels. In most if not all Romance languages longer duration is a major phonetic exponent of word stress, and more so in sentence nuclear position (e.g. Delgado-Martins 1977; Frota 2000 for EPor; Recasens 1986 for Catalan; Farnetani/ Kori 1990 for Italian; Ortega-Llebaria/Prieto 2007 for Spanish). Although in most of these languages lengthening has not phonologized in a certain sense (e.g. speakers are not aware of it and it is not further restricted by particular phonological conditions), phonologization has happened in Italian. Here, vowel lengthening depends not only on the presence of stress, but also on syllable structure, as it affects vowels in non-final open syllables only, as in p[ˈaː]pero ‘duck’ and tartar[ˈuː]ga ‘turtle’ vs. carib[ˈuː] ‘caribou’ (Nespor/Vogel 1986/2007, 131).2 Like other Romance languages, such as Catalan, the Standard variety of EPor lacks diphthongization of either sort, but exhibits other processes that depend on 1 Cf. Meyers (2012) for an account of word and syllable final devoicing (a phonological process also found in many languages) as the phonologization of phonetic properties that are usually present in utterance final position, with generalization to smaller prosodic domains. 2 According to Gabriel/Kireva (2014a), a similar kind of lengthening is also found in Argentinean Spanish, which emerged via transfer from Italian to (the contact variety) Argentinean Spanish.

Segmental phenomena and their interactions

45

word-level stress, including /e/ centralization before palatals, illustrated in (4a) (Mateus 1975/1982, 34–35; Vigário 2003, 78–82), [j] insertion to break a hiatus when the first vowel (V1) is a /e/ in stressed position, as exemplified in (4b) (Mateus 1975/ 1982, 38–39; Vigário 2003, 78–82), or the so-called (verbal) vowel harmony (VH), a process that targets verbal root vowels in stressed position, which harmonize in height with the thematic vowel (TV) when it is followed by another vowel and deletes, as illustrated in (4c) (cf. the details and accounts within autosegmental and feature geometry framework in Wetzels 1991; Mateus/d’Andrade 2000, 81–86 and Matzenauer/Miranda 2005). (4) EPor. a. Stressed /e/ before palatal /e/ in unstressed position (vowel reduction) telha [ˈtɐʎɐ] telhado [tɨˈʎadu] ‘tile’ ‘roof’ b. Stressed hiatus passeio [pɐˈsɐju] ‘(I) walk’

unstressed hiatus (V1 semivocalization) passear [pɐˈsjaɾ] ‘to walk’

c. Stressed root vowel, TV deletion, VH applies devo [ˈdevu] (dev e + u) ‘(I) should’ deva [ˈdevɐ] (dev e + a) ‘(he) shouldSBJV.PRS’

Stressed root vowel, no TV deletion, VH does not apply deves [ˈdɛvɨʃ ] (dev e + s) ‘(you) should’ devem [ˈdɛvɐ̃ȷ ]̃ (dev e + m) ‘(they) should’

In some cases, not only the presence of word stress, but also a specific stress pattern may correlate with particular realizations of segments. There are three processes identified in Portuguese where segments’ quality depends on types of feet. One of these processes is Spondaic Lowering, which is a quite general rule (with some exceptions) that consists in the lowering of stressed mid vowels in words with penultimate stress ending in closed syllable, e.g. d[ˈɔ]cil ‘gentle’, but d[ˈo]ce ‘sweet’ (cf. Wetzels 1992; 2006/2007). It is worth mentioning that this rule, which is shared by both the Brazilian and the European varieties of Portuguese, may be responsible for the emergence of a low nasal round vowel in EPor, a vowel that is found nowhere else in the sound system of the language. In fact, its context is so restricted that, as far as we know, phonologists have failed to notice that [ɔ̃] may occur in EPor, in the words ontem ‘yesterday’ and anteontem ‘the day before yesterday’ (there is individual variation, but for the speakers who have this vowel, it is obligatory). The second rule, Dactylic Lowering, has the same result, and affects stressed vowels in proparoxytone words, e.g. esquel[ˈɛ]tico ‘skeletal’, but esquel[ˈe]to ‘skeleton’ (cf. Wetzels 2006/2007, 21–22). This is reminiscent to a process also found in Catalan whereby a mid vowel (/e/ or /o/) in stem final position is lowered (i.e. Cat. carb[ˈo] → carb[ˈɔ]nic, esf [ˈe]ra → esf [ˈɛ]ric, introduct[ˈo]r – introduct[ˈɔ]ri etc., cf. Mascaró 1976; 2002).

46

Marina Vigário

Finally, the third phenomenon that may be observed only in words with a particular stress pattern applies nearly without exception in EPor. In this case, the lowering rule targets stressless non-high vowels, when they appear in paroxytone words, in word-final syllables closed by consonants other than -s, as in repórt[ɛ]r ‘journalist’ and séni[ɔ]r ‘senior’ (Vigário 2003, 85–89). Notice that in the latter case, reference to the unstressed status of the vowel, the (right edge of the) prosodic word and syllable composition is sufficient to describe the context of the rule and hence, here, reference to the foot seems unnecessary, synchronically. One may wonder why a particular stress pattern should impact on segments’ quality. As pointed out by Wetzels (1992), both Spondaic and Dactylic Lowering apply in contexts of exceptional stress distribution. And an exceptional stress pattern is also involved in the third type of lowering, which affects posttonic vowels in paroxyton words. In fact, as we will see further below, lowering of non-high stressless vowels in EPor seems to relate to exceptions to the application of more or less general phonological rules in other cases as well. Word stress may affect the realization of segments inside words, as in the cases mentioned above, but it can also span word sequences. This is the case of raddoppiamento sintattico, a well-known phenomenon whereby a word-initial consonant is lengthened if preceded by a word ending in a stressed vowel and followed by a non-nasal sonorant (e.g. Nespor/Vogel 1986/2007; Repetti 1991; Loporcaro 1997; Passino 2013, and references therein). The examples in (5), taken from Nespor/Vogel (1986/2007, 167), illustrate the application of this process. (5) It. a. Avrà t[ː]rovato il pescecane. ‘He must have found the shark.’ b. La gabbia è g[ː]ià c[ː]aduta. ‘The cage has already fallen.’ c. È appena passato con tre c[ː]ani. ‘He has just passed by with three dogs.’ Similarly, the realization of stressed elements may be affected by phonological material in adjacent words, as in the case of the progressive assimilatory diphthongization found in Terceira, mentioned above. As shown by the examples in (6), taken from Martins/Vitorino (1989, 333–334), the preceding high vowel or glide that will define the exact articulation of the glide to emerge and the target stressed syllable may belong to different words. (6) EPor. a. para ir dando > [pɐ i ˈðjɐ̃du]

‘in order to be giving’

>

[tɐ̃ j

ˈmjũtuʃ ]

‘(s/he) has many’

c. com os pés

>

[kuʃ

ˈpwɛʃ ]

‘with the feet’

d. sete escudos

> [ˈsɛt ʃˈkjuðuʃ ] ‘seven escudos’ (former Portuguese currency)

b. tem muitos

Segmental phenomena and their interactions

47

2.2 Segmental phenomena in unstressed environments So far, we have seen examples of segmental processes that target word stressed positions. However, Romance languages also exhibit many segmental processes that crucially operate in unstressed environments. These include vowel reduction, semivocalization, vowel deletion and vowel merger. Vowel reduction/neutralization in stressless positions is responsible for many alternations in Romance languages vocalic systems. For example, in several Italian dialects the vowels /ɛ/ and /ɔ/ are only found in stressed position, as neutralization to [e] and [o] applies in unstressed position (e.g. Saltarelli 1970; Nespor/Vogel 1986/ 2007; Krämer 2009, and references therein). A partially similar pattern obtains in Brazilian Portuguese (BPor), where a distinction must be made between pretonic and posttonic positions (cf. 7; Câmara Jr. 1970; Wetzels 1992; Bisol/Magalhães 2004). (7) Stressless vowel system in BPor Pretonic position Posttonic position non-final [i] [u] [i] [u] [e] [o] [e] — [a] [a]

Posttonic final [i] [u] [a]

Vowel reduction in other Romance languages is however more extreme. This is the case of Catalan, with some variation across different dialects (e.g. Bonet/Lloret 1998, 32–39; Wheeler 2005, 52–77), and especially, of EPor, where non-high palatal vowels not only centralize but also raise, becoming [ɨ] instead of [ɐ] or [ə], as in Catalan. Specifically, in EPor, a tonic system with the vowels /i e ɛ a u o ɔ/ (and in specific contexts also [ɐ]) is reduced to [i ɨ ɐ u], both in pretonic and in posttonic position (in verbs, /i/ is also reduced to [ɨ] word-finally, but in non-verbs there are a number of words with exceptional stress pattern showing final [i], such as táxi ‘taxi’ or júri ‘jury’). Wetzels (1992) analyzes vowel neutralization in unstressed position in BPor within the framework of autosegmental phonology (Goldsmith 1976) and feature geometry à la Clements (1991). According to Wetzels, neutralization involves the delinking of aperture tiers (in the case of PB, [+open 3], which is under the Aperture node, together with [+open 1] and [+open 2]. Posttonically, labial vowels also undergo delinking of [+open 2], thus surfacing more reduced. For the neutralization of unstressed vowels in EPor, the same general approach to vowel reduction as delinking of [open] tiers may be adopted, but delinking is nevertheless in general more extreme (cf. also Bisol/Magalhães 2004 and Miglio 2005 for different proposals within Optimality Theory).3 3 Vowel reduction is also present in some varieties of Spanish, presumably due to language contact, as in Mexican Spanish (Lope Blanch 1963), Andean Spanish (Delforge 2008), and Bulgarian JudeoSpanish (Gabriel/Kireva 2014b).

48

Marina Vigário

There are several remarkable issues related to vowel reduction in EPor. One of these refers to the contexts where VR does not apply, as there are quite a number of contexts where the absence of (full) VR is regular (Mateus/d’Andrade 2000, 134– 136; Vigário 2003, 67–72, 85–88, 92–99). These depend essentially on the prosodic position of stressless vowels (e.g. stressless vowels that are prosodic word initial, as well as in syllables closed by lateral or glide, and at the end of prosodic words that are non-final in a morphological compound, cf. Vigário 2003 and section 4.3, below, for details). The absence of VR may also be unpredictable, dependent of specific lexical items. As pointed out in Vigário (2003, 68–69), exceptions to VR are a property of words rather than morphemes, since the same morpheme may behave regularly with respect to VR or exhibit exceptional behavior, as the stem vowel in velhote [vɛˈʎɔt] ‘old man’ vs. velhice [vɨˈʎis] ‘oldness’ (cf. velho [ˈvɛʎu] ‘old’). True exceptions to VR always involve lowering in EPor or at least basic low vowels, as it seems that there are no cases of mid vowels /e o/ involved in true exceptions to VR. Thus, (i) although we cannot say that stressless vowels always undergo VR, we may state that in EPor the mid vowels [e o] are not found in unstressed position (unless in specific prosodic positions, where VR regularly does not apply, like in word initial position, as in ocupar ‘to occupy’, or when the vowel is part of an underlying diphthong, as in louvar ‘to pray’, or part of a syllable closed by lateral, as in moldar ‘to shape’); and (ii) exceptions to VR (which always surface as low), just like regular VR application (where raising affects mid and low vowels) implies the neutralization of mid and low vowels. As far as we know, these generalizations have not been noticed in previous work on EPor phonology. Two other common processes apply in the Romance languages only in unstressed environments: V1 semivocalization and vowel deletion. The first one consists of the gliding of the first of two vowels when the V1 is high, within and across words (e.g. Bisol 1992; 2003 for BPor; Hualde 1999 for Spanish; Frota 2000 for EPor; Cabré/Prieto 2005 for Catalan; Chitoran/Hualde 2007 for Romanian and several other Romance languages). The second process only applies regularly across words, although in particular morphological environments it is also found inside words (e.g. in verbs, the theme vowel deletes when it is followed by another vowel, as in am a + o > amo ‘I love’; a similar analysis is also proposed in Bermúdez-Otero 2006 for stem final vowels in derivational contexts in Spanish). One of the most interesting aspects of V1 semivocalization, vowel deletion, and also vowel merger, is the way these processes may be constrained by word and/or phrase-level prominence, as we will see in the next section.

2.3 Effects of word- and phrase-level prominence on segmental phenomena Segmental phenomena that target unstressed vowels are often sensitive to word- and phrase-level prominence.

Segmental phenomena and their interactions

49

For example, in languages like Spanish and Romanian and within the word, distance relative to stress has been reported to affect the probability of V1 semivocalization in pretonic positions (Hualde 1999; Simonet 2005; Chitoran/Hualde 2007). Similarly, in Catalan word initial high vowels have been shown not to semivocalize if word stress closely follows (i.e. if it is no more than two syllables away, cf. Cabré/ Prieto 2004). Still inside words, in languages like BPor, EPor and Catalan high vowels followed by another vowel in posttonic position obligatorily surface as glides. In this group of languages, while semivocalization is in general optional pretonically, it is obligatory in postonic position, as in família ‘family’ (d’Andrade/Viana 1994; Frota 2000; Mateus/d’Andrade 2000; Cabré/Prieto 2004; Simonet 2005).4 When V1 and V2 belong to different words, phrase-level prominence may also determine whether semivocalization may apply or not. For instance, according to Frota (2000, 83–95) in EPor V2 stress blocks semivocalization under stress clash conditions, as in (8), where both Word 1 and Word 2 bear phonological phrase (ϕ) level prominence, and the two stressed syllables would be strictly adjacent if semivocalization applied (here, and where relevant in this chapter, capitalizing signals word-stressed syllables). (8)

EPor.

(O dançaRIno)ϕ (ama)ϕ (a bailaRIna RUssa)ϕ ‘The dancer loves the Russian chorus girl.’

*dançarin[w]ama

Like semivocalization, higher-level prominence may also constrain the application of vowel deletion across words. In her experimental work, involving acoustic analysis of read sentences created to control for various segmental and prosodic conditions, Frota (2000) found that different rhythmic configurations yield distinct results, depending on levels of stress under clash (namely phonological phrase and intonational phrase level prominence) and the location of phrasal heads (heads on the right or on the left of clashing stresses result in different deletion possibilities). Back vowel deletion, for instance, has the same general properties as semivocalization, except that it is blocked under more restrictive conditions of stress clash. This is illustrated in (9), where V1 is preceded by a vowel bearing phonological phrase 4 In EPor rising diphthongs usually result from the application of an optional process of semivocalization, and according to the results in Chitoran/Hualde (2007), who experimentally compared French, Spanish, Romanian, EPor and BPor in this respect, EPor speakers show a systematic tendency for producing hiatus rather than rising diphthongs. However, there are some regular instances of obligatory rising diphthongs, as in the sequence /jɔn/, surfacing as [jɔn] and [jun] in stressed and unstressed position, respectively (e.g. nacioNAL ‘national’, accioNAR ‘to activate’, aCCIOnas ‘(you) activate’). To our knowledge, this fact was not previously reported in the literature on EPor. Chitoran/Hualde (2007) propose that Portuguese is different from other Romance languages that show a greater tendency to resolving hiatus via V1 gliding, because it lacks the historical sources that created important amounts of rising diphthongs in other Romance languages (especially Spanish), which act as “attractors” to diphthongization.

50

Marina Vigário

level prominence: in (9a) deletion is impossible because deletion would result in a stress clash at the level of the phonological phrase; in (9b) deletion optionally applies, since although V2 bears prosodic word (PW) level stress, it is not the head of its phonological phrase (adapted from Frota 2000, 87–88). (9) EPor. a. (o dançaRIno)ϕ (Ama)ϕ (a bailaRIna RUssa)ϕ b. (o bailaRIno)ϕ (ANda SEMpre)ϕ (de limuSIne PREta)ϕ

i.

s s s s (o dançaRIno)ϕ (Ama)ϕ (a bailaRIna RUssa)ϕ

*dançarin

ama

okbailarin

anda

ϕ-level prominence PW-level prominence

s w s ϕ-level prominence s s s PW-level prominence ii. (o bailaRIno)ϕ (ANda SEMpre)ϕ (de limuSIne PREta)ϕ

V1 deletion sensitivity to stress clash configurations and levels of phrasal prominence is attested, with some relevant differences, in other Romance languages as well, like Galician (Fernández Rei 2002, 130–151, 180–183), BPor (Tenani 2002, 193–195), and Catalan (Cabré/Prieto 2005). Notice that blocking of segmental rules is one of various possible ways of avoiding stress clash, which may also be achieved via stress retraction (as in Northern varieties of Italian, cf. Nespor/Vogel 1986/2007, section 6.3, and in BPor, as shown below), stress demotion (restricted to lower prosodic domains, according to Nespor/ Vogel 1989), and beat insertion, namely via vowel lengthening. The latter strategy operates in EPor when two clashing stresses are within the same phonological phrase, according to the experimental work conducted by Frota (2000). A sentence like (10), taken from Frota (2000, 128), illustrates the context where segmental lengthening is found in EPor as a way to avoid adjacent clashing stresses.5 (10)

EPor.

(O café[ː] LUso)ϕ contém cevada de boa qualidade. ‘The Lusitanian coffee contains barley of good quality.’

When across-words hiatus are formed of similar vowels, Romance languages may show a variety of different solutions. Without entering in too much detail, we highlight the following facts: (i) in EPor, hiatus formed by two unstressed central mid vowels [ɐ] usually merge in a single low [a], while word stress on V2 blocks vowel merger (Frota 2000, 82); (ii) in BPor the same vowels become a single one without concomitant vowel quality; here, both an analysis of vowel merger or of V1 deletion yields the same result; (iii) in very specific cases, stress on V1 does not block vowel

5 Unlike the large majority of segmental phenomena surveyed in this chapter, this type of lengthening is only observable with acoustic measurements, as it is beyond speakers awareness.

Segmental phenomena and their interactions

51

merger (or V1 deletion) in BPor; an explanation for this is that stress shift applies in these cases as a strategy for stress clash resolution, and hence V1 is in fact unstressed, as shown in (11) (cf. Bisol 2003, 190). (11)

BPor.

soFÁ aZUL > sòfazúl

‘blue sofa’

Importantly, there is some indication that variation across (Romance) languages in hiatus tolerance may be explained, among other things, by the availability of different strategies to avoid stress clashes (cf. Frota 2000, 92–94 for EPor; Fernández Rei 2002, 141–142 for Galician). In general, only V1 may delete. However, some languages allow V2 deletion under particular conditions (Casali 1997). An interesting question that may be raised here is why some Romance languages allow V2 deletion (like Galician and Catalan, cf. Fernández Rei 2002, 143–151, and Cabré/Prieto 2005, respectively), while in others V2 deletion is strictly forbidden, regardless of the properties of the vowel sequences (as in EPor, cf. Frota 2000, 88–89). Casali (1997) and Cabré/Prieto (2005) analyze this type of deletion within OT framework: constraint rankings account for different possible outcomes, including variable rankings of position-sensitive faithfulness constraints that ensure that word initial segments are preserved or, by contrast, may be deleted. We should emphasize that stress effects in hiatus resolution phenomena across words are not restricted to clashing configurations. For example, a less common process of [j] insertion to break a hiatus across words is active in Northern varieties of EPor (cf. Lopo 1895; Segura 2013; Oliveira et al. 2014). Insertion applies between non-high central vowels [a ɐ ã ɐ̃], when V2 bears word-level stress, as illustrated in (12a). In a recent study based on speech collected in several Northern areas, Oliveira et al. (2014) found that higher-levels of prominence in V2 may also promote glide insertion. This can be illustrated by the sequence in (12b) where in six occurrences of the same phrase produced by several speakers in a map task, four show glide insertion in Ana_Alves (where Alves bears phonological phrase prominence), while none shows insertion in Moda_Alves (where Alves bears word- but not phrase-level prominence).6 (12)

EPor.

a.

a ÁRvore é amarela ‘the tree is yellow’

a[j]ÁRvore

b.

MOda Ana ALves (. . .) MOda_Ana; Ana[j]ALves ‘Moda Ana Alves’ (name of store)

6 In this map task two subjects of the same variety interact. Each of the subjects has a map with landmarks and a road. One of the subjects is asked to provide the directions for the other to reach a designated point in the map. The two maps are not exactly identical, and the mismatches create the motivation for naturalistic-like dialogs. This type of task is especially suitable to trigger spontaneous-like renditions of various sentence types, contrastive focus and the occurrence of specific words or expressions.

52

Marina Vigário

Importantly, V1 may be stressless and not preceded by another stressed syllable, as in (12a). Hence, in this case glide insertion seems to be a strategy for preserving the integrity of word initial stressed V2, especially favored when V2 is the head of the prosodic domain containing both vowels. We may note that [j] insertion to break a hiatus is also found in some dialects of Galician, but the conditions for the process are very different (cf. Fernández Rei 2002, 246–253; Colina 1997a, section 3.2). Here, distinct vowels may be involved, stress on V2 is not required, and insertion seems to be active across words only when clitics are part of the word sequence.

2.4 Interactions between segmental and tonal phenomena Besides word stress and phrasal prominence, particular intonational contours may also impact on the realization of segments. For example, in EPor both schwa epenthesis at the right-edge of oxyton words ending in sonorant consonant (e.g. azul ‘blue’, mar ‘sea’ surfacing as [ɐˈzulɨ] and [ˈmaɾɨ], respectively) and blocking of nonback vowel deletion (brilhante ‘brilliant’, surfacing as [bɾiˈʎɐ̃tɨ], instead of [bɾiˈʎɐ̃t]), are common in word-final position of non-final intonational phrases, as well as at the right edge of yes-no questions and vocatives with calling contours (cf. Frota 2002; 2014). According to Frota (2014), what these various cases of word-final schwa realization have in common is the presence of complex tonal events (namely a pitch accent and a boundary tone) that require enough segmental space for their realization (in the case of intonational phrases ending in oxyton words, both the nuclear pitch accent and the boundary tone(s) would be realized on a single syllable, the final stressed one). Hence, schwa insertion or schwa deletion blocking ensures the realization of the segmental material required for tonal anchoring.7 We may notice that other strategies exist in Romance languages to deal with complex tunes-text accommodation, such as tonal truncation (as in Southern varieties of Italian, like Bari and Palermo Italian; cf. Grice et al. 2005), tonal compression (e.g. anticipation of tonal targets under tonal crowding coexists with tonal truncation in Neapolitan, cf. D’Imperio 2006) and final vowel lengthening, a possibility also available in EPor (Frota 2002; 2014) and in Logudorese Sardinian. In the latter language, post-tonic syllables of vocatives may be truncated and originate a monosyllable, in which case lengthening creates the space that is required for the realization of the vocative L+H* L* L% melody (cf. Vanrell et al. 2015, and references therein).8 7 Other reasons may, nevertheless, account for vowel insertion in similar contexts, since according to Cruz (2013), IP final epenthesis in a Southern dialect of European Portuguese, spoken in Alentejo, is also found in non-tonal crowding contexts, like in declarative utterance final position. 8 Recent investigation by Frota et al. (2016) has revealed variation in the strategies to deal with tonal crowding in varieties of Portuguese, some favoring the preservation of the segmental string and consequently exhibiting tone truncation and some preferring to maintain tones and changing the segmentals instead.

Segmental phenomena and their interactions

53

2.5 Summary To sum up, in this section we have reviewed a number of phonological processes that apply in Romance languages showing some of the many interactions between segments and suprasegmental properties. In the following section we will exemplify how prosodic organization may also impact on the way segmental phenomena operate.

3 Segmental phenomena and prosodic structure It is well-known that segmental processes may also be constrained by prosodic structure, that is, some rules target specific segments or strings of segments in particular positions within specific prosodic domains. This also means that segmental phenomena may cue prosodic phrasing and are therefore one of the sources of evidence for the existence of a prosodic structure that is distinct in nature from morphological and syntactic structures (e.g. Selkirk 1984; Nespor/Vogel 1986/2007, and the synopsis in Vigário to appear a, b). In this section we will review work on some of the segmental processes that apply with reference to each of the prosodic domains in Romance languages, starting from the syllable, the lower constituent of the prosodic tree. Following a common trend in language, Romance languages allow for a restricted set of consonants in syllable final position, which usually includes coronal fricatives, nasal and lateral consonants and rhotics. There are several processes that affect consonants in this position. One such process is fricative voicing assimilation (e.g. O[ʒ] bon[ʒ] me[ ʃ ]tre[ ʃ ] ‘the good teachers’ (Portuguese), el[z] meu[z] mapes ‘my maps’ (Catalan)), which is found in all major linguistic areas of Romance, and is active word-internally as well as across words (e.g. Frota 2000, 53–74; Tenani 2002, 128– 135; Fernández Rei 2002, 57–87; Hualde 2005, 159–161; notice that from a phonetic point of view fricative voicing may show some variation, as noticed in Frota 2000, 54 for EPor, and Garcia 2013 for Colombian Spanish). The phonology of syllable-final laterals and nasal consonants varies more in the Romance space. For instance, word-internal nasal consonants surface as such in syllable-final position in most languages, but assimilate the point of articulation of the following consonant (e.g. Harris 1969; Hualde 2005 for Spanish; Krämer 2009 for Italian; Bonet/Lloret 1998 for Catalan); furthermore, point of articulation assimilation may also apply across words, e.g. so[m p]etits ‘(they) are small’, so[ŋ k]uatre ‘(they) are four’ in Majorcan Catalan (Prieto 2004, 254; cf. also Nespor/Vogel 1986/ 2007, 211–213 for Spanish). In Portuguese, by contrast, word-internal coda nasals nasalize the preceding vowel and delete (e.g. campo [ˈkɐ̃pu] ‘field’, respectively). d’Andrade/Kihm (1988) propose that in EPor a nasal in this context is an autosegment, i.e. a nasal feature that is not segmentally anchored, which associates to

54

Marina Vigário

the preceding vocalic nucleus, yielding a nasal vowel. As these authors notice, nasalization is not, however, a purely phonological process in the sense that it is also conditioned by morphological and lexical factors (cf. also Vigário 2003, 74–78 and Sampson 1999 for an overview of the evolution of nasal segments in the Romance languages). In the case of alveolar laterals, similarly to English and unlike most other Romance languages (cf. Bullock 1995 for an overview), coronal laterals also show a particular realization in syllable-final position both in BPor and EPor. The two varieties do not treat the lateral in similar ways, however, as in BPor coda laterals surface as a velar glide [w], whereas in EPor the lateral is velarized, that is, it is realized with a velar secondary articulation (l-velarization is also found in Galician, although in more restricted contexts, cf. Álvarez/Xove 2002, 41; cf. also Mateus/ d’Andrade 2000, 137–141 for an analysis of the phonology of syllable-final consonants in Portuguese within the framework of autosegmental phonology and feature geometry). Note that only in EPor velarization behaves like a purely prosodic phenomenon, in the sense that it applies across-the-board, whenever its structural description is met, including in domains larger than the word. In BPor, by contrast, [w] obtains even when across word resyllabification removes /l/ from coda position (cf. 13a). /l/-velarization, but not nasalization in EPor or /l/-gliding in BPor, operates within a larger prosodic domain in EPor (the intonational phrase), as illustrated in (13b), where /l/ is velarized in Miguel, because there is an intonational phrase boundary that blocks resyllabification, unlike in mel ‘honey’, where due to resyllabification the lateral becomes syllable initial and hence the conditions for velarization are not met. (13)

a.

BPor.

Quanto a Migue[w], ele achou o me[w] agradáve[w].

b.

EPor.

Quanto a Migue[ɫ], achou o me[l] agradáve[ɫ]. ‘As to Miguel, (he) found the honey nice.’

Notice that there are also processes that affect segments in other syllabic positions, as in the case of r-strengthening in Spanish (Harris 1983; Nespor/Vogel 1986/2007, 81; Hualde 2005, 181–184) and in Portuguese (Mateus/d’Andrade 2000, 15–16; Vigário 2003, 89–91). Word-internal r-strengthening applies in these languages when /ɾ/ is preceded by a heterosyllabic consonant, in which case it surfaces as a trill, as in honra ‘honor’ in both languages (with a further change in point of articulation, in the case of Portuguese). Let us now turn to segmental processes that apply with reference to the prosodic word (PW). We shall illustrate PW-bound segmental rules with a PW-limit deletion rule in EPor and a process of vowel harmony bound by the PW domain in a variety of Piedmontese (spoken in the North of Italy).

Segmental phenomena and their interactions

55

In EPor, schwas (corresponding to realizations of underlying /e/ and /ɛ/, as well as, more rarely, /i/ in unstressed position) regularly delete in prosodic word final position (except in the contexts where tonal crowding presses segmental material to emerge, as mentioned in section 2.4, above, and in particular metric configurations within a prosodic compound, as we will see further below). This process is very informative of the way clitics and other types of morphosyntactic objects are organized within prosodic words and prosodic word groups (Vigário 2003, 163–165; 2010). The data in (14) illustrate the main facts (‘x’ signals marked, infrequent realizations). Both (14a) and (14b) show the (nearly) obligatory application of [ɨ] deletion in PW-final position, whether or not a following word starts with a vowel. In (14c) [ɨ] deletion does not operate because the prefix re- ([ʁɨ]) is not a PW (the very presence of the reduced vowel [ɨ] shows the unstressed status of this prefix, which therefore cannot be a PW independent of its base). In (14d) encliticization of weak pronominal clitics to the preceding verb yields incorporation of the clitic into the PW that contains the verb; hence the verb-final vowel is no longer PW-final and deletion is impossible. In these cases, the hiatus is post-tonic and therefore gliding is obligatory (like in faMÍlia ‘family’ – cf. section 2.3, above), a fact that further supports the incorporation of enclitics into the preceding PW. The pattern in (14e) is similar to that in (14a–b), because clitic incorporation into the preceding PW causes clitic-final vowels to become PW-final, which is the context for [ɨ] deletion. Finally, examples (14f–g) show that clitics other than post-verbal weak pronouns do not cliticize to the preceding PW, and hence in (14f) the last vowel of that word deletes. Clitics in this position attach to the following PW instead, so that the final vowel of the clitic can no longer undergo PW-final vowel deletion (a reduction process that affects very frequent words, to which we return in subsection 4.3, below, accounts for optional deletion in this case). (14) Por. a. BEbe.

0 / x[ɨ]

PEle. ‘skin’

0 / x[ɨ]

b. BEbe aGOra! ‘DrinkI M P now!’

0 / x[j] / x[ɨ]

PEle ALva ‘white skin’

0 / x[j] / x[ɨ]

c. reavaliAR ‘reevaluate’

*0 / [i] / [j]

reutiliZAR ‘reuse’

*0 / [i] / [j]

d. BEbe-a! ‘Drink itF !’

*0 / [j]

PEde-o! ‘Ask it!’

*0 / [j]

e. VIU-me ONtem. 0 / x[j] / x[ɨ] ‘(S/he) saw-me yesterday’

PEço-te aGOra. ‘(I) beg-you now.’

0 / x[j] / x[ɨ]

f. PEde o LIvro. 0 / x[j] / x[ɨ] ‘(S/he) asks for the book’

SÓ HenRIque o DIsse. 0 / x[j] / x[ɨ] ‘Only Henrique said it.’

‘drinkI M P ’

GOSto de obserVAR. g. JÁ te ofereCI. 0 / [j] / x[ɨ] ‘(I) have already offered (it) to you.’ ‘(I) like watching.’

0 / [j] / x[ɨ]

56

Marina Vigário

The second phenomenon we will consider here is vowel harmony in Piverone (a dialect of Piedmontese). According to Loporcaro (2000), vowel harmony in this dialect is a PW-bound process affecting word-final non-low vowels, which harmonize in height to the stressed vowel. This is illustrated in (15), taken from Loporcaro (2000, 164): non-low vowels surface as [e o] if preceded by a stressed low or mid-low vowel, and as [i u] if preceded by a stressed high or mid-high vowel (15a). The examples in (15b) show that enclitics incorporate into the preceding PW, since they pattern like PW-final elements. (15)

Piedmontese (Piverone) a. [ˈpɛre] ‘stones’ [ˈkanto] ‘(they) sing’ b.

[ˈda-me] [ˈmat-lo]

‘(s/he) gives me’ ‘(s/he) puts it’

[ˈbryti] [ˈtʃitu]

‘uglyF. P L ’ ‘silent’

[ˈmus-mi] [ˈpij-lu]

‘(s/he) shows me’ ‘(s/he) takes it’

Notice that processes like these may be crucial to determine how particular words (like clitics) and morphemes (like affixes) are prosodized (e.g. Peperkamp 1997; Vigário 2003). Another example of a process that has been used to this end is raddoppiamento sintattico (RS), mentioned in subsection 1.2 and section 3. Peperkamp (1997, 71–72) offers it as a diagnostic to the prosodic status of monosyllabic prefixes: Because monosyllabic prefixes do not trigger RS (as illustrated in 16a), they are argued not to bear word-level stress or form autonomous prosodic words (as tré in 16b). (16)

It.

a.

pre[ɡ]réci

‘pre-GreekM . P L ’

b.

tré [ɡː]réci

‘three Greeks’

(Other) segmental phenomena that apply with reference to the PW have been reported, for instance, in Nespor/Vogel (1986/2007) and Peperkamp (1997) for several dialects of Italian (but cf. Loporcaro 2000 for a critical view), Loporcaro (2000) for Algherese, Romanesco and Friulian, three Romance varieties spoken in Italy, Bisol (2000; 2004) and Schwindt (2008) for BPor, and Vigário (2003) for EPor. The next level of prosodic hierarchy that constrains segmental processes is the Prosodic Word Group (Vigário 2010), or, in other approaches, the Clitic Group or the Composite Group (Nespor/Vogel 1986/2007 and Vogel 2009, respectively). The behavior of word-final schwas in EPor illustrates the relevance of the PWG for the realization of segments. As we have seen above, schwas are usually deleted in prosodic word-final position. However, deletion is blocked within PWGs, when the target vowel is followed by a vowel bearing PWG prominence (Vigário 2010). This is illustrated in (17), where both the compound in (17a) and the abbreviation in (17b) are formed of more than one PW. The vowel that starts the rightmost PW bears PWG

Segmental phenomena and their interactions

57

prominence, blocking final e-deletion in the previous PW, and deletion applies between the first and the second PW of the abbreviation because the second PW is not the head of PWG. Word-final e-deletion in EPor may thus be understood as a domain-limit process (it applies at the right-edge of PW) that operates within a larger domain (the PWG). (17)

EPor.

grande-área RFM (erre-efe-eme)

[ˈɡɾɐ̃ ˈdjaɾjɐ] [ˈɛˈʀɛˈfjɛm]

‘penalty area’ (name of radio station)

Let us now consider segmental processes at the level of the phonological phrase. According to Féry (2004, 170–173), in French the gradient processes of obstruent voicing assimilation (e.g. bec de gaz ‘gas tap’ /kd/ → [ɡd]) and nasal-obstruent simplification (e.g. dinde de Noël ‘Christmas turkey’ /ɛ̃dd/ → [ɛ̃nd]) apply within the phonological phrase and are blocked across phonological phrase boundaries (at least when narrow focus is involved). In Romance languages, however, most commonly only the processes that are sensitive to word stress and phrasal prominence are conditioned by the phonological phrase. For example, raddoppiamento sintattico has been argued to apply within, but not across the phonological phrase (Nespor/ Vogel 1986/2007; notice nevertheless that according to Marotta 2011, the process may apply in higher levels as well). Similarly, in EPor vowel lengthening is a strategy for stress clash resolution available within, but not across phonological phrases, as we have seen before. Furthermore, sandhi processes that originate syllable loss, like those referred to in subsection 2.3, are also usually constrained by stress clash configurations involving phonological phrase prominence (Frota 2000; 2014). The Intonational Phrase, by contrast, is the domain for many other types of segmental processes that apply across words. Very often, they involve resyllabification, a phenomenon that is shared by all Romance languages, implicating syllable restructuring across words. Unlike in Germanic languages, where the domain of resyllabification is the prosodic word (e.g. Booij 1995, for Dutch), in the Romance languages it is the intonational phrase (e.g. Nespor/Vogel 1986/2007 for several languages, Frota 2000, 60–62 for EPor; Féry 2004 for French), or even a higher domain, such as the utterance. In fact, in her experimental work on Galician, Fernández Rei (2002) finds that sandhi processes such as fricative voicing and vowel deletion may apply across IP boundaries with no intervening pause, within the Utterance (with some interspeaker variation). This is illustrated in (18), taken from Fernández Rei (2002, 81–82): Fricative voicing optionally applies in (18a), where each sentence may be phrased within a single Utterance, but not in (18b), where both sentences form two distinct Utterances. (18)

Gal.

a.

Xá son maiorciños. Deixaos ir. ‘(They) are older now. Let them go’

b.

Comprou dous iates. Dáme un cigarro. ‘(He) bought two yachts. Give me a cigarette.’

58

Marina Vigário

Furthermore, using a methodology similar to that of Frota (2000) and Fernández Rei (2002), Tenani (2002) finds that in BPor the processes involved in resyllabification (fricative voicing, tapping, vowel degemination, deletion and diphthongization, and syllable degemination) may in fact be unbound, as speakers resyllabify not only within and across IPs, but also across Utterances that do not qualify for restructuring into a single Utterance (with no intervening pause), as in (19) (from Tenani 2002, 178). (19)

Por.

O Pedro comprou pêssego. Alegaram falta de provas. ‘Pedro bought peaches. They have claimed lack of evidence.’ pêsse[ɡw a]legaram pêsse[ɡ a]legaram

What ultimately accounts for the variation across (Romance) languages in the prosodic domain for resyllabification is certainly an interesting, though not very much investigated topic (but cf. Kleinhenz 1997 for some suggestions). There are a number of much debated issues related to resyllabification other than its domain of application. For instance, a well-established fact seems to be the necessity of distinguishing between what one may consider basic syllabification, resyllabification in the lexicon and postlexical resyllabification. Notice that in OT accounts these distinctions may be accomplished without reference to the separation between lexical and post-lexical phonology, but they converge in the need of distinguishing between different types of processes related to the syllabification of segmental material (cf. e.g. Peperkamp 1997; Colina 1997b; Face 2002 for Spanish; Schwindt 2008 for BPor; Peperkamp 1997 and Cardinaletti/Repetti 2009 for Italian dialects). Among the most challenging questions posed by resyllabification is the status of lexical and post-lexical syllables and word boundaries. For example, it is well known that languages are usually affected by constraints that ban certain segments from word-initial position. Thus, like in other Romance languages, in EPor a number of segments cannot appear prosodic word initially, namely [ɨ ɾ ɲ ʎ] (from Vigário 2003, 159). However, resyllabification apparently results in words starting with forbidden segments. One approach to the issue suggested in Vigário (2003, 160) is to admit that phonotactic restrictions of this sort apply only at the lexical level, before word combination and resyllabification (cf. also Peperkamp 1997, 27–30 and references therein). Alternatively, Cardinaletti/Repetti (2009) propose that there are two syllable representations, which are not subject to the exact same requirements: word-level syllables, embedded under the prosodic word, and phrase-level syllables. For example, in a sequence like /l/ + /ɛ/ > [lɛ] ‘he is’ (from the Northern Italian dialect of Donceto), only [ɛ] is syllabified at the prosodic word-level, but at the phonological phrase level [lɛ] forms a (phrase-level) syllable. Duplicating syllable representations in this way is argued, for instance, to account for asymmetries found

Segmental phenomena and their interactions

59

in several Romance languages between syllables that are possible at the word-level and marked or impossible resyllabifications involving prosodic words or clitic-host combinations. This is illustrated by the Spanish examples in (20), where resyllabification may apply only if it does not yield a complex onset or coda (although these are legal in the language as in, e.g., pueblo [pwe.βlo] ‘village’). (20)

Sp.

a.

club elegante

[klu.βe.le.ɣan.te]

‘elegant club’

b.

club lindo

[kluβ.lin.do], *[klu.βlin.do]

‘pretty club’

Under this approach, phrasal syllable boundaries and prosodic word boundaries do not have to match, and hence resyllabification (i.e. the formation of phrase-level syllables) does not imply prosodic word boundaries restructuring. Within the Romance space, French liaison presents a number of specific problems that have long attracted the attention of phonologists, in particular the origin of resyllabified consonants (cf. the literature review in Tranel 1995; Bybee 2001; Ngyen et al. 2009). In general it is assumed that when consonants are lexically anchored, word-final consonants always surface and no alternations emerge (as in seize ‘sixteen’). However, in the case of prenominal adjectives and when closed lexical classed are involved, either allomorphy (petit chat [pətiʃa] ‘little cat’, but petit ami [pətitami] ‘little friend’; les chats [leʃa] ‘the cats’, les amis [lezami] ‘the friends’) or floating consonants (requiring association to segmental and syllabic tiers in order to be realized, as when the following word starts in a vowel) account for the observed alternations. Evidence from speech perception studies in French suggest that wordfinal fixed and liaison consonants (e.g. seize élèves ‘sixteen pupils’ vs. des élèves ‘pupils’) have a different phonological status (cf. Ngyen et al. 2009, and the references therein for other work on the perception of liaison consonants in French).9

4 Non-general segmental phenomena: subphonological grammars or cophonologies and the lexicon Many of the processes mentioned in the previous sections are considered purely phonological, in the sense that they depend on phonological constraints alone. However, like in other languages, in Romance languages innumerous segmental

9 It is well-known furthermore that in French the so-called h aspiré words block liaison and enchaînement, despite the fact that they start phonetically with a vowel. We refer to Gabriel/Meisenburg (2009) for a review of the immense literature on this matter and for a recent analysis of the phenomenon within Optimality Theory.

60

Marina Vigário

phenomena only apply in smaller areas of the lexicon or in particular morphosyntactic contexts. Additionally, in many cases specific lexical items or morphemes must be somehow lexically specified in order to be exempted from the application of a rather general process or, on the contrary, so as to exhibit a specific phonological behavior. In different frameworks, distinct theoretical apparatus have been devised to handle this type of phonological facts (cf. e.g. Inkelas/Orgun/Zoll 1997). Here, we will focus on various types of processes rather than on specific theoretical approaches to exceptions in phonology. We will briefly consider three distinct types of processes, which differ at least in their origin.

4.1 Processes that refer to morphological and lexical information Ever since the early days of generative phonology, processes referring to morphological information and/or with lexical exceptions have been widely documented in the Romance languages space (e.g. Harris 1969; Saltarelli 1970; Mateus 1975/1982; d’Andrade 1977). These rules are not purely phonological in the sense that for their application it is not sufficient that the phonological description of the rule is met, as they are restricted to certain morphological environments, they may have exceptions, and/or they are not sensitive to post-lexical information. Inflectional and derivational environments, as well as verb-clitic and clitic-clitic combinations, display an array of such segmental phenomena. Well-known illustrative examples from Portuguese of this type of phenomena are the realization of plurals ending in laterals and the realization of nasal feature (Morales-Front/Holt 1997; Vigário 2003, 74–76), and several processes of regular verbal inflection (Mateus 1975/1982; Mateus/d’Andrade 2000), the so-called processes of metaphony (in nouns and adjectives) and vowel harmony (in verbs) (Matzenauer/ Miranda 2005 and references therein). Among the processes with similar properties mentioned in the previous sections from other Romance languages are vowel/ diphthong alternations in Spanish (cf. Hualde 2005, 193–198; Eddington 2012, among others) and diphthongization in Romanian, which also involves the interaction of morphological and phonological constraints (Chitoran 2002), to mention but a couple of these processes in Romance. In some approaches, the properties of phonological rules are seen to point to a particular organization of grammar. Furthermore, identifying the locus in grammar that is relevant for a given phonological process or constraint may in fact have further implications for the phonological analysis. For example, as we have seen above, sound patterns may signal prosodic structure. However, this is true only if phonological units are prosodized in the same point in grammar where the particular phonological processes or relevant generalizations apply. We shall illustrate this with a couple of processes in different Romance languages involving verb-clitic combinations.

Segmental phenomena and their interactions

61

In her proposal on the prosodization of clitics in various dialects of Italian, Peperkamp (1997) relies on the word stress patterns displayed by host-clitic combinations. For example, the fact that enclitics do not affect stress placement in Standard Italian, as in PORtamelo ‘Bring it to me!’ (bringI M P-me-it), is seen to show that post-verbal clitics do not incorporate into the host prosodic word in this language variety. However, Loporcaro (2000) and Vigário (2003, 333) point out that if stress placement in Standard Italian is a lexical process and pronominal clitics are syntactic words (and not affixes) which combine with their hosts post-lexically, the mere fact that verb-clitic combinations are not present at the lexical component of phonology accounts for the non-application of lexical phonological processes. Loporcaro (2000) further shows that, like in Standard Italian, in Algherese (a Catalan dialect spoken in Sardinia), pronominal clitics do not affect stress placement either. However, post-lexical processes, such as vowel epenthesis, reveal that in this dialect enclitics are indeed integrated within the host prosodic word. EPor data also show the same need for separating lexical from post-lexical phenomena when considering the prosodization of clitics (Vigário 2003, 162–164). For example, /e/ obligatorily centralizes in stressed position followed by a palatal high segment (coda fricatives are assumed to be lexically underspecified for place features and thus do not trigger the process, e.g. mesmo ‘even’ or vespa ‘wasp’). Although it has a few exceptions when /e/ is followed by fricatives (mexo ‘(I) touch’, rejo ‘(I) govern’), the rule is exceptionless when sonorants are involved (telha ‘tile’, tenho ‘(I) have’, areia ‘sand’, lei ‘law’). The fact that it has exceptions is a symptom that this is a lexical process. Crucially, when a stressed /e/ is followed by a sonorant palatal segment that belongs to an enclitic, it never centralizes. Notice that evidence from regular, postlexial phonological processes clearly shows that pronominal enclitics incorporate into the prosodic word of their host, as we have seen previously.10 Not only segmental and suprasegmental processes, but also other important phonological generalizations, such as the Three Syllables Stress Window, as well as phonotactic constraints, show the importance of distinguishing between lexical and 10 Enclitics in the varieties of Romance may interact with stress location in various ways (cf. e.g. Peperkamp 1997, 176–178; Ordóñez/Repetti 2006, among many others, and references therein). In most standard varieties, clitics do not affect word stress location. In some varieties, however, enclitics may bear stress, while the stress in the host is also maintained, as in Neapolitan (e.g. CÓNtaTÍle ‘tellIMP-youREFL-itF’). In addition, enclitics may also be totally integrated in the host prosodic word, contribute to the computation of stress location, and eventually bear the main stress of host-clitic combination, as in Lucanian, Gascon, Majorcan Catalan or Cheso Aragonese (cf. VÍnne ‘sellIMP’, vinnemMÍle ‘sellIMPme-it’, in Lucanian). In some cases, variation is also found in the same variety, as in Argentinean Spanish, where forms like ¡Preguntáselo! ‘Ask it to him!’ may alternate with ¡Preguntaseló!, in colloquial, mainly emphatic speech (cf. Gabriel/Rinke 2010; Colantoni/Cuervo 2013). The variation found in the way enclitics interact with prosodic word stress points to different types of prosodic integration of the clitic into the host PW and/or to differences in the point in grammar where the host-clitic combination obtains (at the lexical component or postlexically), and both facts may correlate with varying degrees of grammaticalization of the host-clitic combination.

62

Marina Vigário

post-lexical phonology. It is well-known that there is a universal tendency for stress to fall within the three initial or final syllables of a phonological domain (e.g. van der Hulst 1996). On the basis of EPor data, Vigário (2003, 67) argues that this generalization can only be maintained if it is assumed that it operates at the lexical component. In fact, the incorporation of pronominal enclitics into the preceding verbal host may result in (extended) prosodic words with stress on the fourth or even the fifth to the last syllable of the word (21). (21)

EPor.

a.

abandonáramo-la [ɐ.bɐ̃.du.ˈna.ɾɐ.mu.lɐ] ‘(we) had abandon her’

b.

oferecíamo-no-lo [o.fɾɨ.ˈsi.ɐ.mu.nu.lu] ‘we used to offer it to ourselves’

In the same line, Loporcaro (2011b) observes that only post-lexical rules, such as schwa epenthesis, may originate prosodic words with stress before the antepenultimate syllable in Romance languages.11 Phonotactic restrictions in EPor banning prosodic word initial [ɲ ʎ ɾ] are also to be observed only in lexical phonology, since proclitic adjunction to prosodic words ( já lhe ofereci > já lhofereci ‘(I) have already offered-him’) and resyllabification (venho aqui > venhaqui ‘(I) come here’) may originate prosodic words starting with these forbidden segments (but cf. Cardinaletti/Repetti 2009 for an alternative account, as mentioned above). To conclude this section, we would like to point out that in some cases, it is difficult to determine whether a given pattern of segmental distribution is best analyzed as resulting from the application of phonological rules that operate in very restricted environments or via constraints ordering reflecting preference for a given allomorph over allomorphic competitors (cf. for example the discussion in Mascaró 2007 and Nevins 2011). Alternations may involve affixes, clitics and clitichost combinations, and other types of words as well. Examples of such alternations in Romance, analyzed from many different perspectives are copious, e.g. Italian pronominal clitics (Nespor/Vogel 1986/2007; Peperkamp 1997, and references therein) and inflected prepositions (Napoli/Nevins 1987), Catalan clitics (Bonet/Lloret 2005) and definite masculine article (Mascaró 2007), EPor pronominal clitics (Vigário 2003) and the plural morpheme in the nominal system (Mateus 1975/1982; Morales-Front/Holt 1997), French liaison allomorphy (Zwicky 1985; Tranel 1996; Perlmutter 1998), and the feminine definite article in Spanish (Zwicky 1985; Harris 1987).

11 It is well known that the third person plural morpheme in the present tenses may also cause stress to fall before the antepenultimate syllable of the word in Standard Italian (e.g. teléfonano ‘(they) phone’, cf. Peperkamp 1997, 194, and the references therein).

Segmental phenomena and their interactions

63

4.2 Frequency effects on the realization of segments Word frequency may also affect segmental realization in several ways (e.g. Bybee/ Hopper 2001). This topic has attracted less attention in the realm of studies on the segmental phonology of Romance languages, although frequency is very often reported to affect the application of particular rules. Very frequent words seem to favor the application of idiosyncratic (reduction) processes. Vigário (2003, 303–309) notes that many reductions affecting highly frequent words in EPor result in the avoidance of marked phonological patterns, e.g. marked syllable structures, such as those with complex onsets (22a), complex nuclei (22b), complex rhymes (22c), and empty onsets (22c) as well as marked clitic formats, such as disyllabic clitics (cf. 22d). (22)

a.

CGV > CV CCV > CV VG > V

de arte com a grande em ao

[ˈdjaɾt] > [ˈdaɾt] [kwɐ] > [kɐ] [ˈɡɾɐ̃d] > [ˈɡɐ̃dɐ] [ɐ̃ ȷ ]̃ > [ẽ] [aw] > [ɔ]

‘of art’ ‘with theF ’ ‘big’ ‘in’ ‘to-theM ’

b.

CVC > CV CVGC > CVG

mesmo pois

[ˈmeʒmu] > [ˈmemu] [ˈpojʃ ] > [ˈpoj]

‘really’ ‘as’

c.

VCV > CV

avô avó até

[ɐˈvo] > [ˈvo] [ɐˈvɔ] > [ˈvɔ] [ɐˈtɛ] > [ˈtɛ]

‘grandpa’ ‘grandma’ ‘even’

d.

disyllabic > monosyllabic clitics para [pɐɾɐ] > [pɾɐ] > [pɐ] pelo [pelu] > [plu]

‘for’ ‘by-theM ’

In some cases, it seems that reductions of highly frequent words lexicalize, and both reduced and unreduced allomorphs coexist. However, it is not the case that these phenomena spread into other areas of the lexicon and eventually generalize as a pure phonological rule. Additionally, very frequent combinations of words also seem especially prone to lexicalize (e.g. Napoli/Nevins 1987 for Italian; Bybee 2001 for French; Vigário 2003, 317 for EPor). Finally, high frequency may also result in the preservation of irregularity (Bybee/Hopper 2001). For example, Bybee (2001) argues that in French, grammatical words like les ‘theP L ’ occurred frequently in positions where their final consonants were prevocalic (as in les enfants ‘theP L children’) and this is why they have allomorphs that exhibit the maintenance of the word final consonant in liaison contexts, whereas other words (like bois ‘forest’) completely lost it.

4.3 Loanword phonology Loanwords may pose a number of challenges to native phonological grammar. Only rather recently this area of phonology has been given some attention in Romance, in

64

Marina Vigário

particular in languages where language-contact creates large-scale borrowings. This is the case of Eastern Catalan, a language very much exposed to Spanish borrowings. According to Cabré (2009), loanwords in Eastern Catalan have triggered a new phonology, especially patent in stressless vowel system. For Mascaró (2002, 110– 113) loanwords exhibit lexically specified exceptions to vowel reduction, while according to Cabré (2009), in addition to vowel reduction blockage, vowel harmony is also involved, consisting of the long distance assimilation of stressed mid vowels to a following [+ATR] mid vowel. As shown in examples in (23), taken from Cabré (2009, 268), instead of reducing to schwa, stressless vowels surface as [+ATR] (cf. 23a), and mid vowels are pronounced as close mid when followed by close mid vowels (cf. 23b). (23)

Cat.

a.

N[e]pal

Versall[e]s

C[o]lgat[e]

b.

p[e]st[o]

B[o]st[o]n

[o]sl[o]

Cabré (2009) proposes that these specific phonological patterns are used to identify loans within the lexicon. Like Eastern Catalan, EPor loanwords also tend to show non-native phonological behavior with respect to vowel reduction (VR). In this language, not only stressless vowels of borrowed words often escape VR but in addition, in these cases, stressless vowels usually surface as low. As we have seen in section 2, in EPor non-high stressless vowels not affected by VR are exceptional, and exceptions to VR that surface as low are by no means exclusive to loanwords, like lexical exceptions to VR that surface as low (cf. 24a), in several prosodic positions non-high stressless vowels are also realized as low [ɛ a ɔ], instead of [ɨ ɐ u], respectively), namely in prosodic word final positions in syllables closed by consonants different from ‑s (cf. 24b), at the right edge of prosodic words that are non-final within prosodic word groups, namely in morphological compounds (cf. 24c), or in truncated words (cf. 24d) (cf. Vigário 2003, 67–73). Interestingly enough, like in EPor, the absence of VR and [+ATR] realization in compounds and truncated forms are also found in Eastern Catalan (Cabré 2009, 273). (24)

EPor.

a.

r[ɛ]tórica ‘rhetoric’

el[ɛ]ctricidade ‘electricity’

pr[ɔ]curar ‘(to) search’

b.

tór[a]x ‘chest’

abdóm[ɛ]n ‘abdomen’

Vít[ɔ]r ‘Victor’

c.

mon[ɔ]-acentual ‘monoaccentual’

cin[ɛ]-radiografia ‘cine-radiography’

sóci[ɔ]-demográfico ‘socio-demographic’

d.

exp[ɔ] ‘expo’

eur[ɔ] ‘euro’

fot[ɔ] ‘photo’

Segmental phenomena and their interactions

65

A constraint avoiding mid vowels in word-internal stressless positions could be seen to be responsible for this outcome in EPor, which would also be active in other areas of the lexicon where vowel reduction exceptionally does not apply. However, a piece of evidence suggests that lowering is not specific of stressless vowels. As we have seen in section 2, above, lowering also affects stressed vowels in Portuguese in words with marked stress patterns. Hence in this case, a constraint requiring nonhigh vowels to surface as low seems to actually signal exceptional, non-regular (in some cases, non-native) phonology (very much along the lines of Wetzels’ 1992 suggestions on Spondaic Lowering and Dactylic Lowering cited above). Importantly, EPor also offers evidence suggesting a division of labor between lowering and blocking of vowel reductions. In fact, in the cases illustrated in (24b–d), unreduced vowels in stressless positions also signal prosodic word right-edges. That what is relevant for cuing prosodic edges is the blocking of VR is shown by the realization of the vowels at the left-edge of prosodic words, which regularly escape full VR. Crucially, here lowering does not apply and hence in this position unstressed mid vowels may regularly surface in EPor (e.g. elegante ‘elegant’, where the initial vowel may be realized as [e] or [i], but not [ɨ] or [ɛ]). Pons-Moll (2012) also investigates the contexts for underapplication of vowel reduction in Majorcan Catalan. Besides some striking similarities with Eastern Catalan and EPor, her data show important commonalities between loanwords and learned words as well as other areas of the lexicon (Pons-Moll 2012 for the details). Loanword phonology has also been investigated in Italian. For example Repetti (2012) examines vowel epenthesis as a repair strategy to avoid marked or impossible structures in consonant-final loanwords in Italian, a language known for the scarcity of word final consonants (e.g. stop [ˈstɔppe], [ˈstɔppə] or [ˈstɔppə]). Here, the focus is on the phonology of the epenthetic vowel. Repetti shows that the vowels inserted in this context are phonologically inert (e.g. they do not interact with word stress), are influenced by phonetic and morphological factors, and are distinct in quality from the unmarked epenthetic vowel found in non-final position in the language, which is [i] (e.g. in [i]Svizzera ‘in Switzerland’). A feature common to the different phenomena seen above is that, whereas loanword adaptations occur as means of dealing with marked or illegal patterns in L1, often they still display marked or specific phonology.

5 Conclusion In this chapter we have examined grammatical factors that may affect the realization of segments in Romance languages. We have seen that the presence and absence of word-level stress and higher levels of prominence constrain the realization of segments in various ways. For example, diphthongization in many Romance languages, vowel lengthening in Italian and vowel harmony in Portuguese affect vowels

66

Marina Vigário

in stressed position. Some segmental rules depend on particular stress patterns, like Spondaic Lowering and Dactylic Lowering in EPor and BPor. The interaction between word stress and segments realization may be observed inside words, as in the cases above, but it can also span words, as in the case of raddoppiamento sintattico in Italian, of progressive assimilatory diphthongization in the Portuguese dialect of Terceira (Azores) and of [j]-insertion to break a sequence of central vowels in Northern dialects of EPor. Segmental phenomena may also be sensitive to higherlevel prominence. This has been reported for several languages, including Catalan, Galician and Portuguese. Like word stress and higher-level prominence, the lack of stress may condition phonological processes as well. Most Romance languages exhibit neutralization of vowels as a result of varying degrees of reduction in stressless positions (e.g. Italian, BPor, Catalan, EPor). Similarly, semivocalization and vowel deletion as strategies to break hiatus are found across the Romance space (e.g. BPor, Catalan, EPor, Romanian, Spanish). In the latter case, processes are further conditioned, in varying ways, by the presence of word and higher-levels of prominence, since vowel deletion or semivocalization originating stress clash are often avoided. Other sources of language variation may relate to the ranking of faithfulness constraints in particular languages. For example, while in Catalan or BPor hiatus V2 is allowed to delete and semivocalize in some prosodic configurations, in general only V1 is the target of these processes. Besides word stress and phrasal prominence, also particular intonational contours may impact on segments realization. Schwa insertion or blocking of schwa deletion ensures the realization of the segmental material required for tonal anchoring in EPor, and a similar effect is obtained through lengthening in Logudorese Sardinian. Segmental processes may further be constrained by prosodic structure. Romance languages exhibit different types of processes that depend on the position of segments in every domain of the prosodic hierarchy. Most often, it seems that segmental phenomena in Romance languages are not bound by the phonological phrase, whereas segmental phenomena constrained by the position in the syllable, in the prosodic word, in the prosodic word group and in the intonational phrase are copious. Nevertheless, hiatus resolution processes (namely, vowel deletion and V1 semivocalization), which apply within the intonational phrase or a higher domain, are sensitive to phonological phrase prominence, i.e. they are blocked under stress-clash configurations (despite some variation in the definition of what creates stress-clash configurations in these languages; evidence for this type of sensitivity is found in languages like BPor, EPor, Galician, and Spanish). We have seen that segmental phenomena may signal prosodic structure in Romance languages. However, this is usually only the case of purely phonological processes. In fact, the realization of segments may also be constrained by nonphonological information. Segmental processes that depend on morphological or lexical information are abundant throughout Romance and have long been reported

Segmental phenomena and their interactions

67

and analyzed under various perspectives. Specific phonology is also found in two other areas of the lexicon: loanwords and highly frequent words. Although these also seem domains for phonological generalizations, they appear to have attracted less attention in Romance. These several types of phenomena raise important research questions: (i) are segmental alternations that are not entirely regular best analyzed as resulting from the application of irregular phonological rules or constraint orderings, or from lexically listed allomorphy, instead, with concomitant ranking of constraints imposing a particular allomorph selection over allomorph competitors? (ii) What are the conditions for lexicalization and generalizations across the lexicon? (iii) Under what conditions loanwords resist full integration into the native language phonology and what effects non-integration may have on the phonological system as whole? Whatever model is adopted to account for segmental phenomena, it is clear that phonological processes have different properties. What is at stake is not the specific type of phenomena, but rather how general they are. While regular processes only refer to phonological information and are often optional and sensitive to conditions that obtain from word combinations, there are several types of less general phonological phenomena. Some are obligatory and word-bound, in which case they may be sensitive to morphological information and may have exceptions. Others are sensitive to rather superficial, phonetic information, including speech rate, appearance in prosodic positions that exhibit particular phonetics, such as phonetic lengthening or articulatory strengthening, and word frequency. The former type seems to be especially prone to also be sensitive to lexical items’ origin, as loanwords more often escape lexical, non-general phonological processes than pure phonological rules. The former also necessarily involves lexicalization, while the latter sometimes result in lexicalized, categorical alternations, but in other cases what we find seems to be phonetic, gradual, non-lexicalized alternations. Most Romance languages exhibit remarkable similarities in their segmental phonology, certainly partially because of their common Latin origin. For this reason, it is especially interesting to identify areas of divergence and to investigate the conditions for the emergence of phonological variation. Ultimately, we hope to have contributed in this chapter to show the fruitfulness of the comparative approach to Romance phonology.

6 References Albright, Adam/Andrade, Argelia/Hayes, Bruce (2001), “Segmental environments of Spanish diphthongization”, in: Adam Albright/Taehong Cho (edd.), UCLA Working Papers in Linguistics 7 (Papers in Phonology 5), 117–151. Álvarez, Rosario/Xove, Xosé (2002), Gramática da lingua galega, Vigo, Editorial Galaxia. d’Andrade, Ernesto (1977), Aspects de la phonologie (générative) du portugais, Lisboa, INIC.

68

Marina Vigário

d’Andrade, Ernesto/Kihm, Alain (1988), “Fonologia auto-segmental e nasais em português”, in: Actas do III Encontro Nacional da Associação Portuguesa de Linguística, Lisboa, APL/Colibri, 51–60. d’Andrade, Ernesto/Viana, Maria do Céu (1994), “Sinérese, diérese e estrutura silábica”, in: Actas do IX Encontro Nacional da Associação Portuguesa de Linguística, Lisboa, APL/Colibri, 31–42. Bermúdez-Otero, Ricardo (2006), “Morphological structure and phonological domains in Spanish denominal derivation”, in: Fernando Martínez-Gil/Sonia Colina (edd.), Optimality-theoretic studies in Spanish phonology, Amsterdam/Philadelphia, Benjamins, 278–311. Bisol, Leda (1992), “Sândi vocálico externo. Degeminação e elisão”, in: M. Bernardete Abaurre/Leo Wetzels (edd.), Cadernos de Estudos Lingüísticos 23, 83–101. Bisol, Leda (2000), “O clítico e o seu status prosódico”, Revista de Estudos de Linguagem UFMG 9, 5–30. Bisol, Leda (2003), “Sandhi in Brazilian Portuguese”, Probus 15, 177–200. Bisol, Leda (2004), “Mattoso Câmara Jr. e a palavra prosódica”, DELTA 20, 59–70. Bisol, Leda/Magalhães, José Sueli A. (2004), “Redução vocálica no Português Brasileiro. Avaliação via restrições”, Revista da Abralin (UnB Brasília) 3, 195–216. Bonet, Eulàlia/Lloret, Maria-Rosa (1998), Fonologia catalana, Barcelona, Ariel. Bonet, Eulàlia/Lloret, Maria-Rosa (2005), “More on alignment as an alternative to domains. The syllabification of Catalan clitics”, Probus 17, 37–78. Booij, Geert (1995), The phonology of Dutch, Oxford, Clarendon Press. Bullock, Barbara E. (1995), “Separating the root node. On coda velarization in Romance”, Cahiers linguistiques d’Ottawa 23, 45–66. Bybee, Joan (2001), “Frequency effects on French liaison”, in: Joan Bybee/Paul Hopper (edd.), Frequency and the emergence of linguistic structure, Amsterdam/Philadelphia, Benjamins, 337–359. Bybee, Joan/Hopper, Paul (2001), “Introduction to frequency and the emergence of linguistic structure”, in: Joan Bybee/Paul Hopper (edd.), Frequency and the emergence of linguistic structure, Amsterdam/Philadelphia, Benjamins, 1–24. Cabré, Teresa (2009), “Vowel reduction and vowel harmony in Eastern Catalan loanword phonology”, in: Marina Vigário/Sónia Frota/M. João Freitas (edd.), Phonetics and phonology. Interactions and interrelations, Amsterdam/Philadelphia, Benjamins, 267–285. Cabré, Teresa/Prieto, Pilar (2004), “Prosodic and analogical effects in lexical glide formation in Catalan”, Probus 16, 113–150. Cabré, Teresa/Prieto, Pilar (2005), “Positional and metrical prominence effects on vowel sandhi in Catalan”, in: Marina Vigário/Sónia Frota/M. João Freitas (edd.), Prosodies. With special reference to Iberian languages, Berlin/New York, De Gruyter, 123–157. Câmara Jr., Joaquim Mattoso (1970), Estrutura da língua portuguesa, Petrópolis, Vozes. Cardinaletti, Anna/Repetti, Lori (2009), “Phrase-level and word-level syllables. Resyllabification and prosodization of clitics”, in: Janet Grijzenhout/Barış Kabak (edd.), Phonological domains. Universals and deviations, Berlin/New York, De Gruyter, 79–104. Casali, Roderic F. (1997), “Vowel elision in hiatus contexts. Which vowel goes?”, Language 73, 493– 533. Chitoran, Ioana (2002), “The phonology and morphology of Romanian diphthongization”, Probus 14, 205–246. Chitoran, Ioana/Hualde, José Ignacio (2007), “From hiatus to diphthong. The evolution of vowel sequences in Romance”, Phonology 24, 37–75. Cintra, Luís F. Lindley (1971), “Nova proposta de classificação dos dialectos galego-portugueses”, Boletim de Filologia 22, 81–116.

Segmental phenomena and their interactions

69

Clements, George N. (1991), “Place of articulation in consonants and vowels. A unified theory”, Working papers of the Cornell phonetics laboratory 5, 77–123. Colantoni, Laura/Cuervo, María Cristina (2013), “Clíticos acentuados”, in: Laura Colantoni/Celeste Rodríguez Louro (edd.), Perspectivas teóricas y experimentales sobre el español de la Argentina, Frankfurt/Madrid, Vervuert, 143–157. Colina, Sonia (1997a), “Epenthesis and deletion in Galician. An optimality-theoretic approach”, in: Fernando Martínez-Gil/Alfonso Morales-Front (edd.), Issues in the phonology and morphology of the major Iberian languages, Washington, Georgetown University Press, 235–267. Colina, Sonia (1997b), “Identity constraints and Spanish resyllabification”, Lingua 103, 1–23. Cruz, Marisa (2013), Prosodic variation in EP. Phrasing, intonation and rhythm in Southern varieties, PhD dissertation, Universidade de Lisboa. Delforge, Ann Marie (2008), “Unstressed vowel reduction in Andean Spanish”, in: Laura Colantoni/ Jeffrey Steele (edd.), Selected proceedings of the 3rd conference on laboratory approaches to Spanish phonology, Somerville, Cascadilla, 107–124. Delgado-Martins, Raquel (1977), Aspects de l’accent en portugais. Voyelles toniques et atones, PhD dissertation, Université de Strasbourg. D’Imperio, Mariapaola (2006), “Italian intonation. An overview and some questions”, Probus 14, 37– 69. Eddington, David (2012), “Morphophonological alternations”, in: José Ignacio Hualde/Antxon Olarrea/ Erin O’Rourke (edd.), The handbook of Hispanic linguistics, Malden, Wiley-Blackwell, 193–208. Face, Timothy L. (2002), “Re-examining Spanish ‘resyllabification’”, in: Teresa Satterfield/Christina Tortora/Diana Cresti (edd.), Current issues in Romance languages. Selected papers from the 29th linguistic symposium on Romance languages (LSRL), Amsterdam/Philadelphia, Benjamins, 81– 94. Farnetani, Edda/Kori, Shiro (1990), “Rhythmic duration in Italian noun phrases. A study on vowel durations”, Phonetica 47, 50–65. Fernández Rei, Elisa (2002), Regras fonolóxicas e regras precompiladas de alomorfia sintagmática. Dominios prosódicos en galego, PhD dissertation, Universidade de Santiago de Compostela. Féry, Caroline (2004), “Gradient prosodic correlates in phrasing in French”, in: Trudel Meisenburg/ Maria Selig (edd.), Nouveaux départs en phonologie. Les conceptions sub- et suprasegmentales, Tübingen, Narr, 161–180. Frota, Sónia (2000), Prosody and focus in European Portuguese. Phonological phrasing and intonation, New York, Garland. Frota, Sónia (2002), “Nuclear falls and rises in European Portuguese. A phonological analysis of declarative and question intonation”, Probus 14, 113–146. Frota, Sónia (Coord.) (2010–2015), Interactive atlas of the prosody of Portuguese, http://labfon. letras.ulisboa.pt/InAPoP/ (06.05.2016). Frota, Sónia (2014), “The intonational phonology of European Portuguese”, in: Sun-Ah Jun (ed.), Prosodic typology II, Oxford, Oxford University Press, 6–42. Frota, Sónia/Cruz, Marisa/Castelo, Joelma/Barros, Nádia/Crespo-Sendra, Verònica/Vigário, Marina (2016), “Tune or text? Tunetext accommodation strategies in Portuguese”, in: Jon Barnes/ Alejna Brugos/Stefanie Shattuck-Hufnagel/Nanette Veilleux (edd.), Proceedings of Speech Prosody 2016, Boston, Questrom School of Business, 1129–1133. Gabriel, Christoph/Kireva, Elena (2014a), “Prosodic transfer in learner and contact varieties. Speech rhythm and intonation of Buenos Aires Spanish and L2 Castilian Spanish produced by Italian native speakers”, Studies in Second Language Acquisition 36, 257–281. Gabriel, Christoph/Kireva, Elena (2014b), “Speech rhythm and vowel raising in Bulgarian JudeoSpanish”, in: Nick Campbell/Dafydd Gibbon/Daniel Hirst (edd.), Proceedings of Speech Prosody 2014, Dublin, Trinity College, 728–732.

70

Marina Vigário

Gabriel, Christoph/Meisenburg, Trudel (2009), “Silent onsets? An optimality-theoretic approach to French ‘h aspiré’ words”, in: Frank Kügler/Caroline Féry/Ruben van de Vijver (edd.), Variation and gradience in phonetics and phonology, Berlin/New York, De Gruyter, 163–184. Gabriel, Christoph/Rinke, Esther (2010), “Information packaging and the rise of clitic-doubling in the history of Spanish”, in: Gisella Ferraresi/Rosemarie Lühr (edd.), Diachronic studies on information structure. Language acquisition and change, Berlin/New York, De Gruyter, 63–86. Garcia, Alison (2013), Allophonic variation in the Spanish sibilant fricative, PhD dissertation, University of Wisconsin-Milwaukee. Goldsmith, John (1976), Autosegmental phonology, Bloomington, IULC. Grice, Martine/D’Imperio, Mariapaola/Savino, Michelina/Avesani, Cinzia (2005), “Strategies for intonation labelling across varieties of Italian”, in: Sun-Ah Jun (ed.), Prosodic typology. The phonology of intonation and phrasing, Oxford, Oxford University Press, 362–389. Harris, James W. (1969), Spanish phonology, Cambridge, MA, MIT Press. Harris, James W. (1983), Syllable structure and stress in Spanish, Cambridge, MA, MIT Press. Harris, James W. (1987), “Disagreement rules, referral rules, and the Spanish feminine article ‘el’”, Journal of Linguistics 23, 177–183. Hayes, Bruce (1990), “Precompiled phrasal phonology”, in: Sharon Inkelas/Draga Zec (edd.), The phonology-syntax connection, Chicago, University of Chicago Press, 85–108. Hualde, José Ignacio (1999), “La silabificación en español”, in: Rafael A. Núñez-Cedeño/Alfonso Morales-Front (edd.), Fonología generativa contemporánea de la lengua española, Washington, Georgetown University Press, 170–188. Hualde, José Ignacio (2005), The sounds of Spanish, Cambridge, Cambridge University Press. van der Hulst, Harry (1996), “Separating primary and secondary accent”, in: Rob Goedemans/Harry van der Hulst/Ellen van Zanten (edd.), Stress patterns of the world. Part I: Background, The Hague, Holland Academic Graphics, 1–25. Inkelas, Sharon/Orgun, Orhan/Zoll, Cheryl (1997), “The implications of lexical exceptions for the nature of grammar”, in: Iggy Roca (ed.), Constraints and derivations in phonology, Oxford, Clarendon, 393–418. Kleinhenz, Ursula (1997), “Domain typology at the phonology-syntax interface”, in: Gabriela Matos/ Matilde Miguel/Inés Duarte/Isabel Faria (edd.), Interfaces in linguistic theory, Lisbon, APL/ Colibri, 201–220. Krämer, Martin (2009), The phonology of Italian, Oxford, Oxford University Press. Lope Blanch, Juan (1963), “Sobre las vocales caedizas del español mexicano”, Nueva Revista de Filología Hispánica 17, 1–20. Lopo, Joaquim de Castro (1895), “Linguagem Popular de Valpaços”, Revista Lusitana 3, 325–329. Loporcaro, Michele (1997), “Lengthening and raddoppiamento fonosintattico”, in: Martin Maiden/ Mair Parry (edd.), The dialects of Italy, London, Routledge, 41–51. Loporcaro, Michele (2000), “Stress stability under cliticization and the prosodic status of Romance clitics”, in: Lori Repetti (ed.), Phonological theory and the dialects of Italy, Amsterdam/ Philadelphia, Benjamins, 137–168. Loporcaro, Michele (2011a), “Phonological Processes”, in: Martin Maiden/John Charles Smith/Adam Ledgeway (edd.), The Cambridge history of the Romance languages, vol. 1: Structures, Cambridge, Cambridge University Press, 109–154. Loporcaro, Michele (2011b), “Syllable, segment and prosody”, in: Martin Maiden/John Charles Smith/Adam Ledgeway (edd.), The Cambridge history of the Romance languages, vol. 1: Structures, Cambridge, Cambridge University Press, 50–108. Marotta, Giovanna (2011), “Raddoppiamento sintattico”, Enciclopedia dell’italiano Treccani, vol. 2, Roma, Istituto dell’Enciclopedia Italiana, 1214–1216.

Segmental phenomena and their interactions

71

Martins, Ana Maria/Vitorino, Gabriela (1989), “Palatalisation et vélarisation conditionnées de la voyelle tonique dans certains dialectes portugais. Évolutions identiques dans l’espace roman”, in: Yvonne Johannot (ed.), Espaces romans. Études de dialectologie et de géolinguistique offertes à Gaston Tuaillon, vol. 2, Grenoble, Ellug/Université Stendhal, 330–356. Mascaró, Joan (1976), Catalan phonology and the phonological cycle, PhD dissertation, Massachusetts Institute of Technology. Mascaró, Joan (2002), “El sistema vocàlic. Reducció vocàlica”, in: Joan Solà/Maria-Rosa Lloret/Joan Mascaró/Manuel Pérez Saldanya (edd.), Gramàtica del català contemporani, vol. 1, Barcelona, Empúries, 89–123. Mascaró, Joan (2007), “External allomorphy and lexical representations”, Linguistic Inquiry 38, 715– 735. Mateus, Maria Helena Mira (1975/1982), Aspectos da fonologia portuguesa, Lisboa, Livraria Sá da Costa, 1975. Second, revised edition: Lisboa, Inst. Nacional de Investigação Científica, 1982. Mateus, Maria Helena Mira/d’Andrade, Ernesto (2000), The phonology of Portuguese, Oxford, Oxford University Press. Matzenauer, Carmen/Miranda, Ana Ruth (2005), “Nominal metaphony and vocalic harmony in Brazilian Portuguese. A constraint-based approach”, in: Sónia Frota/Marina Vigário/M. João Freitas (edd.), Prosodies. With special reference to Iberian languages, Berlin/New York, De Gruyter, 339–360. Meyers, Scott (2012), “Final devoicing. Production and perception studies”, in: Toni Borowsky/Shigeto Kawahara/Takahito Shinya/Mariko Sugahara (edd.), Prosody matters. Essays in honor of Elisabeth Selkirk, Sheffield/Bristol, Equinox Press, 148–180. Miglio, Viola G. (2005), Markedness and faithfulness in vowel systems, New York, Routledge. Morales-Front, Alfonso/Holt, D. Eric (1997), “The interplay of morphology, prosody and faithfulness in Portuguese pluralization”, in: Fernando Martínez-Gil/Alfonso Morales-Front (edd.), Issues in the phonology and morphology of the major Iberian languages, Washington, Georgetown University Press, 393–439. Napoli, Donna Jo/Nevins, Joel (1987), “Inflected prepositions in Italian”, Phonology Yearbook 4, 195– 209. Nespor, Marina/Vogel, Irene (1986/2007), Prosodic phonology, Dordrecht, Foris (Berlin/New York, De Gruyter, 2007). Nespor, Marina/Vogel, Irene (1989), “On clashes and lapses”, Phonology 6, 69–116. Nevins, Andrew (2011), “Phonologically conditioned allomorph selection”, in: Colin J. Ewen/Elizabeth Hume/Mark van Oostendorp/Keren Rice (edd.), The companion to phonology, Malden, WileyBlackwell, 2357–2382. Ngyen, Noël/Wauquier-Gravelines, Sophie/Lancia, Leonardo/Tuller, Betty (2009), “Detection of liaison consonants in speech processing in French. Experimental data and theoretical implications”, in: Pilar Prieto/Joan Mascaró/Maria-Josep Solé (edd.), Segmental and prosodic issues in Romance phonology, Amsterdam/Philadelphia, Benjamins, 3–23. Oliveira, Pedro/Paulino, Nuno/Cruz, Marisa/Vigário, Marina (2014), “Onde ainda([j])há o fenómeno. Contributo para o estudo da inserção de glide entre vogais centrais”, in: António Moreno/ Fátima Silva/Isabel Falé/Isabel Pereira/João Veloso (edd.), Textos selecionados do XXIX Encontro Nacional da Associação Portuguesa de Linguística, Porto, APL, 419–436. Ordóñez, Francisco/Repetti, Lori (2006), “Stressed enclitics?”, in: Jean-Pierre Montreuil (ed.), New analyses on Romance linguistics, vol. 2: Phonetics, phonology and dialectology. Selected papers from the 35th LSRL 35, Austin, Texas, February 2005, Amsterdam/Philadelphia, Benjamins, 167– 181. Ortega-Llebaria, Marta/Prieto, Pilar (2007), “Disentangling stress from accent in Spanish. Production patterns of the stress contrast in deaccented syllables”, in: Pilar Prieto/Joan Mascaró/ Maria-Josep Solé (edd.), Segmental and prosodic issues in Romance phonology, Amsterdam/ Philadelphia, Benjamins, 155–175.

72

Marina Vigário

Passino, Diana (2013), “A unified account of consonant gemination in external sandhi in Italian. Raddoppiamento Sintattico and related phenomena”, The Linguistic Review 30, 313–346. Peperkamp, Sharon (1997), Prosodic words, The Hague, Holland Academic Graphics. Perlmutter, David (1998), “Interfaces. Explanations of allomorphy and the architecture of grammars”, in: Steven G. Lapointe/Diane K. Brentari/Patrick M. Farrell (edd.), Morphology and its relation to phonology and syntax, Stanford, CSLI, 307–338. Pons-Moll, Clàudia (2012), “Loanword phonology, lexical exceptions, morphologically driven underapplication and the nature of positionally biased constraints”, Catalan Journal of Linguistics 11, 128–166. Prieto, Pilar (2004), Fonètica i fonologia. Els sons del català, Barcelona, Edicions de la Universitat Oberta de Catalunya. Recasens, Daniel (1986), Estudis de fonètica experimental del català oriental central, Barcelona, Publicacions de l’Abadia de Montserrat. Repetti, Lori (1991), “A moraic analysis of raddoppiamento fonosintattico”, Rivista di Linguistica 3, 307–330. Repetti, Lori (2012), “Consonant-final loanwords and epenthetic vowels in Italian”, Catalan Journal of Linguistics 11, 167–188. Rodrigues, Celeste (2002), “Questões de espraiamento em PE”, in: Maria Helena Mira Mateus/ Anabela Gonçalves/Clara Nunes Correia (edd.), Actas do XVII Encontro Nacional da Associação Portuguesa de Linguística, Lisboa, APL, 419–432. Saltarelli, Mario (1970), A phonology of Italian in a generative grammar, The Hague, Mouton. Sampson, Rodney (1999), Nasal vowel evolution in Romance, Oxford, Oxford University Press. Schwindt, Luiz Carlos (2008), “Revisitando o estatuto prosódico e morfológico de palavras prefixadas do PB numa perspectiva de restrições”, Alfa 52, 391–404. Segura, M. Luísa (2013), “Variedades dialectais do Português Europeu”, in: Eduardo Raposo/ Fernanda B. Nascimento/M. Antónia Mota/Luísa Seguro/Amália Mendes (edd.), Gramática do Português, vol. 1, Lisboa, Fundação Calouste Gulbenkian/Centro de Linguística da Universidade de Lisboa, 85–142. Segura, M. Luísa/Saramago, João (1999), “Açores e Madeira. Autonomia e coesão dialectais”, in: Isabel Hub Faria (ed.), Lindley Cintra. Homenagem ao Homem, ao Mestre e ao Cidadão, Lisboa, Edições Cosmos/Faculdade de Letras da Universidade de Lisboa, 707–738. Segura, Luísa/Saramago, João (2001), “Variedades dialectais portuguesas”, in: Caminhos do Português. Exposição comemorativa do ano europeu das línguas (catálogo), Lisboa, Biblioteca Nacional, 221–237. Selkirk, Elizabeth O. (1984), Phonology and syntax. The relation between sound and structure, Cambridge, MA, MIT Press. Simonet, Miquel (2005), “Prosody and syllabification intuitions of [CiV] sequences in Spanish and Catalan”, in: Sónia Frota/Marina Vigário/M. João Freitas (edd.), Prosodies. With special reference to Iberian languages, Berlin/New York, De Gruyter, 247–267. Tenani, Luciani (2002), Domínios prosódicos no português, PhD dissertation, Universidade Estadual de Campinas. Tranel, Bernard (1995), “Current issues in French phonology. Liaison and position theories”, in: John Goldsmith (ed.), The handbook of phonological theory, Cambridge, MA, Blackwell, 798–816. Tranel, Bernard (1996), “French liaison and elision revisited. A unified account within Optimality Theory”, in: Claudia Parodi/Carlos Quicoli/Mario Saltarelli/María Luisa Zubizarreta (edd.), Aspects of Romance linguistics, Washington, Georgetown University Press, 433–455. Vanrell, Maria del Mar/Ballone, Francesc/Schirru, Carlo/Prieto, Pilar (2015), “Sardinian intonational phonology. Logudorese and Campidanese varieties”, in: Sónia Frota/Pilar Prieto (edd.), Intonation in Romance, Oxford, Oxford University Press, 317–349.

Segmental phenomena and their interactions

73

Vigário, Marina (2003), The prosodic word in European Portuguese, Berlin/New York, De Gruyter. Vigário, Marina (2010), “Prosodic structure between the Prosodic Word and the Phonological Phrase. Recursive nodes or an independent domain?”, The Linguistic Review 27, 485–530. Vigário, Marina (to appear a), “Prosodic hierarchy”, in: Stefan J. Schierholz/Herbert Ernst Wiegand (edd.), Wörterbücher zur Sprach- und Kommunikationswissenschaft (WSK) Online, vol. 4: Phonetik und Phonologie, edd. T. Alan Hall/Bernd Pompino-Marschall, Berlin/New York, De Gruyter, http://www.degruyter.com/view/db/wsk?format=ONMO (06.05.2016). Vigário, Marina (to appear b), “Prosodic phonology”, in: Stefan J. Schierholz/Herbert Ernst Wiegand (edd.), Wörterbücher zur Sprach- und Kommunikationswissenschaft (WSK) Online, vol. 4: Phonetik und Phonologie, edd. T. Alan Hall/Bernd Pompino-Marschall, Berlin/New York, De Gruyter, http://www.degruyter.com/view/db/wsk?format=ONMO (06.05.2016). Vogel, Irene (2009), “The status of the Clitic Group”, in: Janet Grijzenhout/Barış Kabak (edd.), Phonological domains. Universals and deviations, Berlin/New York, De Gruyter, 15–46. Wetzels, Leo (1991), “Harmonização vocálica, truncamento, abaixamento e neutralização no sistema verbal do português. Uma análise auto-segmental”, Cadernos de Estudos Lingüísticos 21, 25–58. Wetzels, Leo (1992), “Mid vowel neutralization in Brazilian Portuguese”, Cadernos de Estudos Lingüísticos 23, 19–55. Wetzels, Leo (2006/2007), “Primary stress in Brazilian Portuguese and the quantity parameter”, Journal of Portuguese Linguistics 5/6, 9–58. Wheeler, Max W. (2005), The phonology of Catalan, Oxford, Oxford University Press. Zwicky, Arnold (1985), “Rules of allomorphy and phonology-syntax interactions”, Journal of Linguistics 21, 431–436.

Élisabeth Delais-Roussarie

3 Prosodic phonology and its interfaces Abstract: It is well established that the speech flow is organized into a prosodic constituent structure that differs from the morphosyntactic and information structure. Indeed, the prosodic units have two major characteristics: (i) they are partly derived from the morphosyntactic and information structure of the sentence to which they should give access, and (ii) they constitute domains for the application of phonological phenomena. They are studied in all their complexity within Prosodic Phonology, a framework that accounts for the way phonology interacts with the other components of the grammar. The aim of this chapter is threefold: (i) presenting the main features of Prosodic Phonology, this framework being crucial to study the way prosody interfaces with the other grammatical components; (ii) explaining how the syntaxphonology mapping has been formalized within this framework, in particular to account for prosodic phrase formation; and (iii) showing how prosody is constrained by information related to discourse. Keywords: prosodic structure, intonation, accentuation, syntax-phonology interface, prosodic phrasing

1 Introduction When one listens to someone talking or when one speaks, the speech flow is not a mere sequence of sounds. It is structured: the segments and syllables – i.e. the elements that compose the phonological representation associated with the linguistic elements of a sentence – are grouped into units of higher rank, which in turn can also come together to form larger ones. These units are organized into a hierarchical structure, the ‘prosodic structure’. The way an utterance is prosodically structured is often crucial for a good interpretation. Prosodic phrasing reflects for instance the attachment of syntactic adjuncts. In the French example (1), the adjunct la semaine dernière ‘last week’ may either be syntactically dependent from the first clause les enfants sont partis en vacances ‘the children went on holiday’ or from the second one j’ai gardé leur chat ‘I took care of their cat’. Prosodic phrasing allows distinguishing the two cases. In (1a), a prosodic break (#), which is realized by a pause and a continuation rise (H*H%) on the syllable ‑nière [njɛʁ], occurs at the end of the adjunct la semaine dernière, and the utterance is interpreted as ‘it is last week Acknowledgement: The author gratefully acknowledges the French National Research Agency/ General Commissariat for Investment (Agence nationale de la recherché, ANR/Commissariat général à l’investissement, CGI, Laboratoire d’excellence “Empirical Foundations of Language” Labex EFL, Université Paris Diderot).

76

Élisabeth Delais-Roussarie

that the children went on holiday’. By contrast, in (1b), the pause and the continuation rise (H*H%) are associated with the syllable ‑cances [kɑ͂s], and the sentence means that ‘the children went on holiday, and that I took care of their cat last week’. (1) Fr. Les enfants sont partis en vacances – la semaine dernière – j’ai gardé leur chat. ‘The children went on holiday – last week – I took care of their cat.’ a. Les enfants sont partis en vacances la semaine dernière # H* H*H% j’ai gardé leur chat. b. Les enfants sont partis en vacances # H*H% la semaine dernière, j’ai gardé leur chat. H* (H-)

Figure 1a: Waveform, spectrogram, and F0 trace for example (1a)

Figure 1b: Waveform, spectrogram, and F0 trace for example (1b)

Prosodic phrasing often helps resolving syntactic ambiguity, as for PP attachment (cf. example 2 for Spanish). In (2a), the adjunct con much gusto ‘with great pleasure’

Prosodic phonology and its interfaces

77

is syntactically related to the first clause Los chicos se fueron a la playa ‘the children went to the beach’, whereas it is related to the second clause in (2b). The location of the pause, which is associated to a major prosodic boundary, is crucial. (2) Sp. a. Los chicos se fueron a la playa con mucho gusto # su madre se quedó en casa. ‘The children went to the beach with great pleasure, their mother stayed home’ b. Los chicos se fueron a la playa # con mucho gusto su madre se quedó en casa. ‘The children went to the beach. With great pleasure, their mother stayed home.’ Apart from revealing the morphosyntactic structure, prosodic phrasing often reflects information structure. In many languages, it has been argued that a prosodic boundary aligns with the right edge of the focus constituent (cf. e.g. Kanerva 1990 for Chichewa;1 Jun 1993 for Korean, and section 4). The segmentation into prosodic units is brought into light by the realization of a wide array of phonetic and phonological phenomena such as the realization of a pause (cf. Fig. 1a and 1b), the occurrence of an accent, or even the application of phonological processes that modify the segmental representation of the sentence. In European Portuguese (cf. Frota 2014), a word-final fricative is realized as a [z] when followed by a word-initial vowel within an intonational phrase (IP), but it is realized as a [ ʃ ] in IP final position; cf. the contrast between (3a) and (3b). (3) EPor. a. (a[z] aluna[z] obtiveram boa[z] avaliaçõe[ ʃ ])IP ‘The students have got good marks.’ b. (a[z] aluna[ ʃ ])IP (até onde sabemo[ ʃ ])IP (obtiveram boa[z] avaliaçõe[ ʃ ])IP ‘The students, as far as we know, have got good marks.’ (Frota 2014, 14) The aim of this chapter is to show that prosody is an interface phenomenon: prosodic as well as tonal and metrical patterns associated with an utterance are partly determined by morphosyntactic and information structure. The term ‘prosody’ refers here to three distinct phonological components: prosodic structure or phrasing, tonal patterns and metrical patterns (cf. e.g. Ladd 2008). Even if these elements may be highly intertwined (in particular in French, cf. Post 2011), they will be considered separately here, special attention being given to prosodic phrasing since the formation of prosodic phrases is constrained by morphosyntactic and informational features. The chapter is organized as follows. In section 2, the major characteristics and the theoretical assumptions on prosodic structure are presented. Section 3 focuses 1 Note that the results obtained in experimental study on Chichewa (Downing/Pompino-Marschall 2013) argue against the correlation between focus and phrasing.

78

Élisabeth Delais-Roussarie

on the syntax-phonology interface, and more precisely on the mapping rules that account for prosodic phrase construction. The interface between prosody and discourse is addressed in section 4. Several examples taken from various Romance languages will be given throughout the chapter.

2 Prosodic structure: theoretical assumptions and phonological characteristics As already mentioned, among the three prosodic components, phrasing constitutes a crucial element in the interface between phonology and the other grammatical components. Indeed, the prosodic structure is the only structure to which phonology has access, the prosodic phrases being domains for the application of phonetic and phonological processes (cf. e.g. Selkirk 1981; 1984; 1986). The prosodic events, be they metrical or intonational, are thus analyzed with respect to prosodic phrases (cf. e.g. Pierrehumbert/Beckman 1988; Jun 1993).2 In this section, prosodic structure will be presented. We will insist on phenomena usually used as cue for prosodic phrasing.

2.1 Prosodic Structure Theory and the prosodic hierarchy In Prosodic Structure Theory, the phonological representation of a sentence consists of a hierarchically organized structure distinct from the morphosyntactic structure. The units of the prosodic hierarchy are considered as domains for the application of phonological phenomena, be they segmental, accentual or intonational. However, a distinction can be made between two approaches (cf. Frota 2012): in some studies (Selkirk 1981; 1984; 1986; Nespor/Vogel 1986/2007, among others), the prosodic units are defined according to their relation to the morphosyntactic structure, whereas in some others studies (Pierrehumbert/Beckman 1988; Jun 1993; Jun/Fougeron 2000; Jun 2005, among others), intonation plays a crucial role in the definition of the various units. The prosodic hierarchy is presented in (4).

2 Metrical patterns were sometimes analyzed with a grid without any reference to prosodic structure and to specific phrases (cf. e.g. Prince 1983; Selkirk 1984; Delais-Roussarie 2000). However it has been shown that an alternative analysis can be achieved by referring to prosodic structure (cf. e.g. Nespor/Vogel 1989).

Prosodic phonology and its interfaces

79

(4) Prosodic Hierarchy Intonational Phrase Phonological Phrase / Major phrase3 or Intermediate Phrase Phonological Phrase / Minor phrase or Accentual Phrase Prosodic Word Foot Syllable Despite the differences, which are partly visible in the names given to the various units, the hierarchy shares at least three main features: (i) Prosodic structure is distinct from syntactic structure. (ii) Prosodic constituents are hierarchically organized. (iii) No bracketing paradox may occur, i.e. the right edge of a phrase always aligns with the right edge of a lower level constituent. In addition, there may be some variation regarding the number of levels that are necessary above the word level (cf. Frota 2012 for discussion).4 Note also that some authors have used an integrated view, in which phonological processes affecting segments, intonation and prominence are all taken into account to define the various prosodic constituents (cf. Jun 1993; Frota 2000). Such an approach has the advantage of reinforcing the validity of a given level of structuring by avoiding circularity. Among the prosodic constituents, the syllable and the foot are phonological by nature (the foot, for instance, is defined according to metrical criteria). By contrast, the constituents above the word level are mostly defined relatively to the morphosyntactic and information structure, without being isomorphic to it (cf. section 3 and Selkirk 1986; Truckenbrodt 1999, among others). As for the internal structure of the prosodic hierarchy, it was argued that it should be strictly layered and thus conformed to the Strict Layer Hypothesis (Selkirk 1981; 1986; Nespor/Vogel 1986/2007). Since this hypothesis embodied four distinct principles, it has been factored out in Optimality Theory (5), cf. Selkirk (1995a). Four constraints relative to the well-formedness of the prosodic structure have been formulated (Selkirk 1995a).

3 According to Selkirk (1986), the Minor and Major Phrase are distinguished depending on the mapping parameters used to derive them. But it is not really clear whether both level of structuring are expected to occur in a single language. In French, for instance, what is called ‘phonological phrase’ by Post (2000a) is mostly equivalent to a Minor Phrase. 4 Some additional units have been added to the basic hierarchy. On a language specific basis, the mora has been inserted in Japanese (cf. Pierrehumbert/Beckman 1988). Above the word level, the name and number of constituents also vary: the Clitic Group is argued for by Nespor/Vogel (1986/ 2007); constituents such as the accentual phrase and the intermediate phrase are referred to in replacement of the Clitic Group and the Phonological Phrase (cf. Jun/Fougeron 2000 for French and Nibert 1999 for Spanish).

80

Élisabeth Delais-Roussarie

(5) Constraints on Prosodic Structure (Selkirk 1995a, 443) No constituent Ci dominates a constituent Cj, j>i. LAYEREDNESS : Example: No syllable (σ) dominates a foot (Σ). HEADEDNESS :

Any constituent Ci must dominate a constituent of level Ci-1, except if Ci is a syllable (σ). Example: A phonological word (ω) must dominate a foot (Σ).

EXHAUSTIVITY:

No constituent Ci immediately dominates a constituent Cj, jB}), in this case {o>lo}. The faithfulness constraint P RIORITY (“Respect lexical priority (ordering) of allomorphs”; Bonet/Lloret/Mascaró 2007, 906; Mascaró 2007, 726) penalizes the insertion of the

Phonology and morphology in Optimality Theory

115

second allomorph (as in 13) unless markedness conditions force its appearance, which then surfaces as a TETU effect; cf. (14), below.8 The high ranking of the markedness sonority constraint penalizing heterosyllabic contacts between consonants and vowels (*C.V) ensures that, even though Galician admits onsetless syllables in certain contexts, cf. (13), intersyllabically they must be syllabified with a following vowel, cf. (14). The alignment constraint enforcing that morphological word edges coincide with syllable boundaries (A LIGN -L(MW,σ) (“The left edge of a morphological word (MW) must coincide with the left edge of a syllable (σ)”; Kikuchi 2006, 44) penalizes resyllabification between words; cf. its effects for allomorph selection in (14). (13)

Based on the fact that sequences of a continuant and a stop that arise from pronominal encliticization are maintained (e.g. visita[ɾ.m]e ‘to visit me’, vémo[s.t]e ‘we see you’), Kikuchi proposes that the deletion of -r and -s is due to the ranking of the markedness OCP constraint penalizing adjacent continuant consonants in prosodically close domains (OCP[+cont]) above the anti-deletion faithfulness constraint M AX-C (14b); deletion in morpheme internal position (merlo [ˈmeɾ.lo], *[ˈme.lo] ‘blackbird’) is banned by ranking the OCP constraint below I-C ONTIGUITY, which is the constraint that “rules out deletion of elements internal to the input string” – or “No Skipping” (McCarthy/Prince 1995, 371; cf. also Kenstowicz 1994). However, as noticed by Nevins (2011, 2365, 2373), in order to handle opacity the lateral allomorph has to be chosen at an intermediate stage before OCP[+cont] is decisive during the evaluation, with OCP[+cont] crucially ranked below M AX-C (14a). As shown in (15), a single step evaluation with the ranking proposed in Kikuchi (2006, 47) will lead to an ungrammatical result.

8 The classic example that illustrates the role of P RIORITY in unnatural allomorph choice is the definite article distribution displayed by Haitian creole, where the a allomorph appears after a stem ending in a vowel (papa-a ‘father-the’), but la appears after a stem ending in a consonant (liv-la ‘book-the’). This anti-markedness allomorph distribution is accounted for with the ordered set {a>la} and P RIORITY, benefitting a (as in [pa.pa.a]) unless right alignment of the stem with the syllable or *C.V causes the selection of la (as in [liv.la] vs. *[li.va], *[liv.a]) (Bonet/Lloret/Mascaró 2007, section 2). In our view, the role of the constraint B REVITY in i/gli and il/lo selection in Italian (cf. section 2.2) to benefit i and il, respectively, can be handled with the use of the ordered sets {i>ʎi} and {il>lo} and P RIORITY replacing B REVITY in the ranking.

116

Eulàlia Bonet and Maria-Rosa Lloret

(14) Two-step derivation for /beɾ {o>lo} neno/: [ˈbe.lo.ˈne.no]

(15)

A parallel derivation for /beɾ {o>lo} neno/: [ˈbe.lo.ˈne.no]

The problem raised by Galician can also not be handled in serial OT models based on morphologically determined levels, like Stratal OT (cf. e.g. Bermúdez-Otero 2013), because the opacity issue arises within the same level or stratum; serial models like Harmonic Serialism (cf. e.g. McCarthy 2000), which allow only one change at a time in each evaluation step, would handle them adequately only if a different constraint ranking were allowed at different steps, an option usually not assumed in Harmonic Serialism. Note, however, that even with the concurrent use of serialism, alignment constraints and lexically ordered allomorphs, the analysis would fail to capture the following fact: the onsetless article allomorph is chosen after a word ending in -/n/, which resyllabifies as an (alveolar) onset with the article (e.g. comen o caldo /komen o kaldo/: [[ˈko.me.no]PW [ˈkal.do]PW] ‘they eat the broth’). Under the presumed ranking, A LIGN -L(MW,σ) will always promote the second allomorph (*[[ˈko.men.lo]PW

Phonology and morphology in Optimality Theory

117

[ˈkal.do]PW]). Following work by Álvarez Blanco (1983), Kikuchi (2006, 47) suggests the possibility that a third set of nasal allomorphs come into play (i.e. no(s), na(s)), which will induce coalescence of the adjacent nasals (/komen1n2o kaldo/: [[ˈko.me.n12o]PW [ˈkal.do]PW], where the subscript digits in [n12] indicate coalescence of the sequence /n1n2/).

3 Asymmetric surface relations and constraints on alignment In this section, in addition to further exemplifying PCA and discussing the limits of abstractness when positing inputs as well as the apparent need for serialism, we introduce alternative parallel ways to deal with these phenomena focusing on which elements can stand in correspondence. We first present the phenomenon of diphthongization in Spanish, which illustrates a case of opaque interaction that can be handled in serial terms as well as within the tenets of parallelism (section 3.1). We then draw attention to a case of overapplication of epenthesis in Catalan cliticization, where some effects derived from the phonology-morphology interaction are explained in the parallel model, with the use of alignment constraints, but cannot be captured in serial terms (section 3.2).

3.1 Diphthongization in Spanish The stress-driven alternation that affects mid-vowels in Spanish illustrates a wellknown paradox of cyclicity. Generally, pure vowels appear in unstressed position while diphthongs appear in stressed position (e.g. c[o]ntar ‘to tell’ – c[ˈwe]nto ‘tale/I tell’, n[e]gó ‘s/he denied’ – n[ˈje]go ‘I deny’), and it seems that diphthongization depends on stress position, but stress in turn is sensitive to diphthongization (i.e. syllable weight). The phenomenon is lexically idiosyncratic because the alternation coexists with non-alternating pure vowels (cf. m[o]ntó ‘s/he mounted’ – m[ˈo]nto ‘I mount’, p[e]gó ‘s/he hit (past)’ – p[ˈe]ga ‘s/he hits’) and non-alternating diphthongs (cf. frec[we]ntó ‘s/he frequented’ – frec[ˈwe]nta ‘s/he frequents’, v[je]nés ‘Viennese’ – V[ˈje]na ‘Vienna’). In the alternating cases, the presence of diphthongs in unstressed positions of certain derived words but not in others (c[we]ntecito ‘tale (diminutive)’ vs. c[o]ntable ‘tellable’) also raises the issue of locality in cyclic application. Traditional generative analyses derive the alternation from a unique underlying representation with the use of diacritic marks, empty skeletal slots and specific cyclic rule application (e.g. Harris 1969; 1985; Halle/Harris/Vergnaud 1991). Harris (1985), and along the same lines Halle/Harris/Vergnaud (1991), proposes an abstract

118

Eulàlia Bonet and Maria-Rosa Lloret

representation containing single segmental units followed by an empty skeletal position (/oX/, /eX/). In stressed position, the skeletal slot is filled through the derivation by means of word-level ordered rules: /oX/, /eX/ turn into ‘oV̯ ’, ‘eV̯ ’ through diphthongization in stressed syllables, which then become ‘oe̯’, ‘ee̯’ by e-default insertion; ‘oe̯’, ‘ee̯’ are later adapted as ‘o̯e’, ‘e̯e’ due to adjustment in nuclearity (on the assumption that sonority prefers rising complex nuclei), and they finally surface as [we], [je] via glide formation. In unstressed position, the diphthongization rule does not apply; the skeletal slot remains empty and hence is eliminated at the end of the derivation. The presence of unexpected unstressed diphthongs in certain derived forms is captured through a different underlying morphological composition of words 〚c[o]nt-a-ble〛vs.〚 ( 〚c[we]nt〛‑ecito〛) in Harris (1969), which Halle/Harris/Vergnaud (1991) reanalyze as a difference in the kind of affixes the words contain:9 cyclic stem-level affixes (such as denominal ‑ble) or non-cyclic word-level affixes (such as evaluative ‑(ec)ito). Under this view, stem-level affixed words do not display diphthongization effects of first-cycle stress assignment because the diphthongization rule applies at the word level, after stress has shifted to the stem-level suffix 〚 ( 〚〚c[o]nt-[ˈa]〛SL-bl-e〛SL〛WL). At the word level, though, diphthongization is extrinsically ordered before stress reassignment; hence, the diphthongization effect of first-cycle stress assignment surfaces 〚 ( 〚c[we]nt〛SL-ec[ˈi]to〛WL ). Bermúdez-Otero (2006; 2013), in line with the observations made by, for example, Eddington (1996) and Albright/Andrade/Hayes (2001), argues that this cyclic approach, as well as its recasting in Distributed Morphology (Embick 2013), requires excessively powerful phonological devices that crucially subvert the concept of cyclic domain. He proposes instead a phonologically driven allomorphic approach within Stratal OT, with the use of allomorphy and presuming a specific morphological structure of words. Under his view, nominals and verbs have a stem formative meaningless morph added to the root to satisfy a morphomic constraint on stem well-formedness (a ‘morphome’ in terms of Aronoff 1994). Stem formatives (SF) include nominal word-markers (i.e. o-stems, a-stems, e/{e,Ø}-stems: cuent[o] ‘tale’, mes[a] ‘table’, immun[e] – immun[e]s ‘immune (singular – plural)’ / panØ – pan[e]s ‘bread(s)’), as well as verbal theme vowels (as in cont[ˈa]r ‘to tell’, cont[ˈa]ble ‘tellable’, respond[ˈe]r ‘to answer’, respond[ˈi]a ‘s/he answered’). The root plus the stem formative forms the inner stem; verbal inflected forms and most derivation are built at the stem level (16a), while evaluative derivation (e.g. cuentecit[o] ‘tale (diminutive)’) is formed at the word level (16b).

9 In these examples, hollow brackets notate morphological constituents and cyclic domains and, as usual, for phonetic transcriptions ordinary square brackets are used.

Phonology and morphology in Optimality Theory

(16)

a.

119

Stem level (SL): Underived noun 〚Root – SF〛SL 〚cuent-o〛SL

b.

Verb inflection

〚〚Root – SF〛SL X〛SL 〚〚cont-a〛SL-r〛SL

Most derivation

〚〚Root – SF〛SL X – SF〛SL 〚〚cont-a〛SL-bl-e〛SL

Word level (WL): Evaluative 〚〚Root – SF〛SL X – SF〛WL 〚〚cuent-o〛SL-ecit-o〛WL

The key feature of Bermúdez-Otero’s analysis for diphthongization is that diphthongal allomorphy is a property of stems rather than of roots, which is a language-particular fact that must be encoded in the lexical entries of stems (Bermúdez-Otero 2013, 72). This implies that all root allomorphs are present at the stem level, whereas at the word level the only allomorph available is the stem-level output, which functions as the input of the word level. Additionally, the analysis requires (non-iterative) vowel deletion of unstressed stem-final vowels before suffixes beginning with another vowel, -ist-a〛‘taleregardless of its morphological affiliation (cf.〚cuent-o〛‘tale’ –〚 〚cuent-o〛 -ón〛‘cheeky’ –〚 〚respond-í〛 -a〛‘s/he teller’,〚 〚respond-e〛 -r〛‘to answer’ –〚 〚respond-e〛 answered’) (Bermúdez-Otero 2006, 280, section 2.1; 2013, 38–39). With these premises in mind, diphthongization turns out to be an instance of phonologically driven allomorph selection by output optimization. The lexicon supplies two listed allomorphs for alternating items (one containing a pure vowel and the other containing a diphthong). Both allomorphs are inserted at the stem level, insofar as this instruction is encoded for the lexical entries of stems. Hence, the outputs satisfy the faithfulness I DENTITY constraint whether they contain a diphthong (from the diphthongal input allomorph) or a pure vowel (from the pure vowel input allomorph; cf. examples 18–20). During evaluation, the diphthongs are preferred in stressed syllables (as a sonority effect of the constraint *P EAK Foot/e,o, which penalizes the pure vowels [e, o] in the peak node, i.e. the head, of a foot) on the assumption that diphthongs are more sonorous than pure vowels and hence are better suited as the head of a foot, i.e. the stressed syllable (Kenstowicz 1997, 162) (18), while pure mid-vowels occur elsewhere (driven by a context-free markedness constraint against complex nuclei: *C OMPLEX N UC ; cf. examples 19–20). For wordlevel affixes, however, the only input available is the nominal diphthong stem allomorph; therefore, I DENTITY discards the candidate without a diphthong (21). Bermúdez-Otero insightfully illustrates the analysis with the pair enc[o]ntrón ‘abrupt meeting’ (20) (stem-level derivation from the inner verbal stem〚{enkwentɾ, enkontɾ}-a〛; cf. enc[ˈwe]ntra ‘s/he meets’ in (18b) and enc[o]ntrar ‘to meet’ in 19) and

120

Eulàlia Bonet and Maria-Rosa Lloret

enc[we]ntrón ‘meeting (augmentative)’ (21) (word-level derivation from the nominal stem〚eŋˈkwentɾ-o〛; cf. enc[ˈwe]ntro ‘meeting’ in 18a). For our purposes, we omit the analysis of stem-final vowel deletion and {e,Ø} stem-formative selection. (17)

Stem level (SL): Underived noun 〚{enkwentɾ,enkontɾ}-o〛SL [eŋˈkwentɾo]

(cf. 18a)

〚{enkwentɾ,enkontɾ}-a〛SL [eŋˈkwentɾa]

(cf. 18b)

〚〚{eŋˈkwentɾa,eŋˈkontɾa}〛SL-ɾ〛SL [eŋkonˈtɾaɾ]

(cf. 19)

〚〚{eŋˈkwentɾa,eŋˈkontɾa}〛SL-on-{e,Ø}〛SL [eŋkonˈtɾon]

(cf. 20)

Inflected verb

Deverbal

Word level (WL): Evaluative 〚〚eŋˈkwentɾ-o〛SL -on-{e,Ø}〛WL [eŋkwenˈtɾon]

(cf. 21)

(18) a. encuentro ‘meeting (noun)’

b. encuentra ‘s/he meets’

(19) encontrar ‘to meet’

(20) encontrón ‘abrupt meeting (deverbal)’10

10 In (20) *COMPLEX N UC is the decisive constraint because the two candidates fair evenly with respect to *P EAK Foot/e,o. The ranking *P EAK Foot/e,o » COMPLEX N UC is proven in the tableaux in (18).

Phonology and morphology in Optimality Theory

(21)

121

encuentrón ‘meeting (augmentative)’

The apparent dual behaviour of some denominal (non-evaluative) suffixes (e.g. ‑ista in c[we]ntista ‘taleteller’ – c[ˈwe]nto ‘tale’ vs. conc[e]rtista ‘concertist’ – conc[ˈje]rto ‘concert’) is accounted for by admitting that, although historically descended from the same root, some nominals have ended up having a single stem (with a diphthong) rather than two as verbal stems have (Bermúdez-Otero 2013, 78, 84). An alternative parallel analysis of diphthongization in Spanish is possible with output-output (OO) asymmetric correspondences (cf. e.g. McCarthy/Prince 1994; 1995; Benua 1995; 1997) and Kager’s (1999b, 282) specific notion of ‘base’ (cf. 22), based on Kager (1999a). We replicate Lloret/Mascaró’s (2006) analysis of the phenomenon of depalatalization in Spanish.11 (22) Definition of ‘base’ a. The base is a free-standing output form – a word. b. The base contains a subset of the grammatical features of the derived form. According to (22a), the base must always be an output itself, an existing word. According to (22b), the base must be compositionally related to the affixed word in a morphological and semantic sense, and must be in a proper subset relation

11 Depalatalization in Spanish (i.e. the non-occurrence of palatal nasals and laterals in word-final position) provided a classic argument for cyclic application within derivational phonology, as exemplified by the famous triplet desdé[n] ‘disdain’ – desde[n]es ‘disdains’ – desde[ɲ]es ‘you disdain (subjunctive)’ (Harris 1983). For historical reasons there are few cases with alternations in traditional words that provide evidence for a synchronic phenomenon of depalatalization. For this reason, some scholars claim that they are lexical remnants that should be treated in terms of allomorphy (e.g. Pensado 1997; Harris 1999; Eddington 2012), while others, from data drawn from old and recent loan adaptation, provide evidence for maintaining productive depalatalization (e.g. Lloret/Mascaró 2006). Within the former view, Bermúdez-Otero (2006) considers there not to be a synchronic relation between the nominal and the verbal stem of such items and hence assumes that nominal stems have a root ending in a coronal (/desden-{e,Ø}/SL in desdé[n], desde[n]es) whereas verbal stems contain a palatal-final root (/desdeɲ-a/SL in desde[ɲ]es). Alternatively, on the assumption that all forms derive from single palatal inputs, the alternation has been captured, in the parallel view, through OO correspondence relations, based on asymmetric (base-dependent) OO relations (cf. Lloret/Mascaró 2006, who refine Baković’s 1998; 2001 analysis) or symmetric OO relations (cf. Pons-Moll 2012, section 3.1.2, within the Optimal Paradigms model developed in McCarthy 2005). Kikuchi (1999) proposes instead a parallel OT analysis based on the Sympathy model proposed in McCarthy (1999), which uses additional machinery to enable the use of certain candidate outputs as inputs to mimic the reference to intermediate forms.

122

Eulàlia Bonet and Maria-Rosa Lloret

with it. The morphological relations of a plural with respect to its singular base form and a diminutive with respect to its non-diminutive base form satisfy this subset relation. First, the number category does not change features but just adds the feature [PLURAL] in plurals and the plural is always formed over the shape of its singular base. The situation is different in masculine/feminine pairs, since even with the use of a single privative feature [FEMININE ] the masculine form is never a proper subset of the semantic features of the feminine form and the feminine is not always formed over the shape of its masculine counterpart (sol-o/a ‘aloneM / F ’). Evidence for the asymmetry between number and gender is found, for instance, in Greenberg (1963) and in Harris (1992); cf. Lloret/Mascaró (2006, 88) and Bermúdez-Otero (2013, section 2.4.2) for the specific case of Spanish. Furthermore, diminutives contain all morphosemantic features of their corresponding non-diminutive forms and have as base the free-standing non-diminutive word, as proven, among other facts, by allomorph selection: in general, ‑ecit in monosyllabic words but ‑(c)it in polysyllabic words, as in sol ‘sun’ – solecito (diminutive) vs. solo ‘alone (masculine singular)’ – solito (diminutive) (cf. e.g. Jaeggli 1980). The base identity constraint targeting the nuclei (I DENT B ASE (N UC )) together with the markedness constraints mentioned above (*P EAK Foot/e,o and *C OMPLEX N UC ) will do the job with the ranking given in (23). The tableaux in (24–27) illustrate the evaluation. (23) I DENT B ASE (N UC ) » *P EAK F O O T /e,o » *C OMPLEX N UC (24) Sp. encuentro ‘meeting (noun)’

(25) Sp. encontrar ‘to meet’

(26) Sp. encontrón ‘abrupt meeting (deverbal)’

Phonology and morphology in Optimality Theory

123

(27) Sp. encuentrón ‘meeting (augmentative)’

Under Kager’s (1999b) contained notion of ‘base’ presented in (22), the OO constraints only capture a restricted set of relations – as Stratal OT does – that adequately holds for plurals and evaluatives (including superlatives), and hence the criticism of cyclic views to OO approaches for the allowance of unrestricted access to the global environment (cf. e.g. Bermúdez-Otero 2006; 2013) does not hold true. A more intriguing presence of the diphthongal allomorph in unstressed position is the conjugation I verbs with the prefix a- ([a. . .ˈa-ɾ]), which are causatives derived from nouns and adjectives that Bermúdez-Otero (2013, 61) limits to change-ofstate verbs, as in av[je]jar ‘to make old’ vs. env[e]jecer ‘to become old’ (cf. v[ˈje]jo ‘old’ – v[e]jez ‘oldness’). There are not, however, many such cases and, in turn, other parallel denominal [a. . .ˈa-ɾ] derivations do not present diphthongal allomorphs in this unstressed position (e.g. as[e]rrar ‘cut with a saw’), although they exhibit the regular alternating pattern elsewhere (cf. s[ˈje]rra ‘saw’, diminutive s[je]rrecita – s[e]rrería ‘sawmill’). All in all, one cannot but conclude that the diphthongal [a. . .ˈa-ɾ] forms are better treated as instances of lexical idiosyncracy. In fact, as demonstrated in Eddington (1996; 2012) and Albright/Andrade/Hayes (2001), diphthongization shows more variation than expected in traditional words as well as in loans and nonce words, depending on the morphological and the phonological environments. The relevant morphological context is the type of affix, especially, as seen, in more productive affixation, namely: (i) diminutives, augmentatives and superlatives as well as the causative [a. . .á-r] construction derived from nouns and adjectives are typically associated with diphthongs; (ii) the nominal affixes ‑ero, ‑al and ‑(i)dad are less likely to occur with diphthongs (buñ[ˈwe]lo ‘fritter’ – buñ[o]lero ‘fritter maker’, but c[ˈwe]nto ‘tale’ – c[we]ntero ‘taleteller’); but (iii) the nominal affixes -oso and -ista do not show a significant preference for either diphthongs or pure vowels (c[ˈwe]nto – c[we]ntista ‘taleteller’ vs. conc[ˈje]rto ‘concert’ – conc[e]rtista ‘concertist’) (Eddington 1996). Albright/Andrade/Hayes (2001) suggest that in verbs conjugation class might have some influence too. As for the phonological environment, a decisive factor is the shape of the root in environments specific to front or back vowels (e.g. the [X__ɾɾ] context favours the presence of diphthongs in e roots but not in o roots: c[e]rrar ‘to close’ – c[ˈje]rro ‘I close’ vs. b[o]rrar ‘to erase’ – b[ˈo]rro ‘I erase’ and also in nonce words d[e]rrar – d[ˈje]rro vs. n[o]rrar – n[ˈo]rro) (Albright/Andrade/Hayes 2001). In order to capture the gradient productivity of diphthongization, Eddington (1996) proposes a treatment of the phenomenon within the tenets of the lexicon-based approach (e.g. Bybee 1985), while Albright/Andrade/Hayes (2001) model the data with a learning algorithm that

124

Eulàlia Bonet and Maria-Rosa Lloret

predicts stochastic behaviour by rule pattern association. As noted by BermúdezOtero (2013, 64), the observation that native speakers have statistically based intuitions on diphthongization is compatible with the aforementioned OT analyses. On the whole, both the serial stratal analysis and the parallel paradigmatic one manage to hold up well, though the two approaches are conceptually different with respect to whether or not they use intermediate levels of representation.

3.2 Overapplication of epenthesis in Catalan cliticization The phonological behaviour of pronominal clitics in Catalan demonstrates that serial analyses cannot account for some apparent domain effects (Bonet/Lloret 2005). The facts are as follows. In the Catalan variety spoken in Barcelona, some pronominal clitics are underlyingly asyllabic (e.g. 1st person singular /m/, partitive /n/) while others have an underlying vowel (e.g. feminine /ə/ in 3rd person /l-ə/) (cf. e.g. Wheeler 1979; 2005; Viaplana 1980; Mascaró 1986; Bonet/Lloret 2003; 2005). In proclisis, asyllabic clitics surface with an initial epenthetic schwa (underlined in the examples) before a verb starting with a consonant for syllabic reasons (28a), but the epenthetic vowel appears after the clitic when the asyllabic clitic follows a verb ending in a consonant (28b). In combinations of more than one clitic, though, a schwa is always inserted between a clitic ending in a consonant and a clitic beginning with a consonant (29a), even when a licit consonantal contact would arise without epenthesis (29b) or when the surface form of the single clitic would solve the problem (29c). (28)

Cat.

a.

b.

(29)

Cat.

em tira /m#tiɾə/: ‘s/he throws (to) me’ cf. m’imita /m#imitə/: ‘s/he imitates me’

[əm.ˈtiɾə], *[mə.ˈtiɾə]

tirem-ne /tiɾɛm#n/: ‘let’s throw some’ cf. tira’m tiɾə#m/: ‘throw (to) me’

[ti.ˈɾɛm.nə], *[ti.ˈɾɛ.mən]

[mi.ˈmi.tə]

[ˈti.ɾəm]

a.

tira-me’n /tiɾə#m#n/: ‘throw some to me’

[ˈti.ɾə.mən]

b.

tira-me-la /tiɾə#m#lə/: ‘throw itF to me’ cf. fem-la /fɛm#lə/: ‘let us do itF ’

[ˈti.ɾə.mə.lə], *[ˈti.ɾəm.lə]

c.

[ˈfɛm.lə], *[ˈfɛ.mə.lə]

me la tira /m#lə#tiɾə/: [mə.lə.ˈti.ɾə], *[əm.lə.ˈti.ɾə] ‘s/he throws itF to me’ (cf. 28a)

Phonology and morphology in Optimality Theory

125

In parallel approaches to OT, alignment constraints have often been used to account for the position of clitics in the utterance. For our purposes, the constraint A LIGN (CL /VB ) will ensure adjacency in the contact of the clitics and the verb in proclisis and enclisis (30) (cf. Colina 1995; Jiménez/Todolí 1995; Jiménez 1997; Bonet/ Lloret 2002; 2003; 2005; Wheeler 2005). (30)

A LIGN (CL /VB ):

Align the left/right edge of a pronominal clitic with the right/left edge of a verb. (Bonet/Lloret 2005, 1308)

With single clitics, the peripheral effect of epenthesis is captured by ranking A LIGN (CL /VB ) below σ-S TRUC (a cover constraint for syllable well-formedness) and above *C ODA , O NSET and D EP-V (32); but for epenthesis not to overapply in the presence of licit codas, F INAL C (“Every Prosodic Word is consonant-final”; McCarthy/ Prince 1994, 357) has to be ranked between A LIGN (CL /VB ) and *C ODA (33).12 The fact that epenthesis is inserted in contexts in which it is not needed shows that D EP-V must be ranked very low (34). (31) σ-S TRUC » A LIGN ( CL / VB ) » F INAL C » *C ODA , O NSET » D EP-V (32)

(33)

12 In this variety of Catalan, the clitic (a function word) together with its host (a lexical word) constitute a prosodic word. Hence, F INAL C is violated when the clitic group, as a whole, ends in a vowel, as in all output candidates in (32a) or as in (33a) in *[ˈti.ɾə.mə] but not in [ˈti.ɾəm].

126

Eulàlia Bonet and Maria-Rosa Lloret

(34)

In serial OT approaches, the work done by morphological alignment constraints in parallel approaches should be captured by the organization in cycles or strata and the possibility of constraint re-ranking at different steps of the evaluation. In these approaches, the faithfulness constraint O-C ONTIGUITY (which is the constraint that rules out insertion of elements internal to the input string – or “No Intrusion”; McCarthy/Prince 1995, 371; cf. also Kenstowicz 1994) has scope over the domain of each stratum, regardless of the internal morphological composition of that domain. Therefore, in the step that includes both the clitic and the verb O-C ONTIGUITY does the job of A LIGN ( CL / VB ) in parallel analyses, in so far as it penalizes the insertion of material between the adjacent string set up by this cycle or stratum. Assuming a strata-based analysis, with the structure (Cl Cl (Verb)) and ((Verb) Cl Cl) at the clitic group stratum, the ranking that better accounts for the cases belonging to this stratum is the one given in (35), where O-C ONTIGUITY occupies a lower position than its parallel A LIGN ( CL / VB ) counterpart in the ranking in (31). We now can account for some instances of overapplication (cf. 36a–b), but critically cannot explain peripherality of epenthesis with single proclitics as well (cf. 37a–b). (35) Clitic group stratum: σ-S TRUC » F INAL C » *C ODA , O NSET » O-C ONT, D EP-V

(36)

Phonology and morphology in Optimality Theory

127

(37)

In Bonet/Lloret (2005) other serial analyses in terms of strict cyclicity, (cl (cl (Verb))) and (((Verb) Cl) Cl), and adjacent independent domains, ((Cl Cl) (Verb)) and ((Verb) (Cl Cl)), are discussed and proved to also be unable to account for these data.

4 Symmetric surface relations In this section we illustrate, with data from insular Catalan, output-to-output relations for which no specific base (i.e. no leading form) can be identified.

4.1 Underapplication of vowel epenthesis in insular Catalan Insular Catalan, i.e. the varieties spoken on the Balearic Islands and in the city of Alghero on Sardinia, differs from all other varieties in having no inflectional affix for the 1st person singular present indicative (1PI): pas ‘I pass’, cant ‘I sing’. Null affixation is also seen in regular masculine singular nominals in all Catalan varieties (pas ‘step’, cant ‘song’). However, while 1PI tolerates final consonants that are not permitted elsewhere in the language (e.g. clusters violating the sonority sequencing principle: filtr ‘I filter’, ensofr ‘I sulfurate’), parallel nominal forms always surface with the final default vowel, [ə] in Balearic Catalan, [a] in Alghero Catalan ( filtr[ə] ‘filter’, sofr[ə] ‘sulfur’ in Balearic; filtr[a], sofr[a] in Alghero Catalan), which is considered to be epenthetic (/filtɾ/, /sofɾ/) (e.g. Mascaró 1978; Wheeler 1979; 2005; Lloret 2002; 2004a). Pre-OT approaches (Mascaró 1983; Dols 1993; Dols/Wheeler 1996) as well as some OT studies (Serra 1996; Dols 2000) base their analyses on the observation that the illicit consonantal endings of 1PI are possible onsets and hence relate their interpretation to this syllabic position. However, among other problems, onsetrelated analyses cannot offer a straightforward explanation for the overwhelming majority of coda phenomena that take place in these verbal forms, such as wordfinal obstruent devoicing (aca[p] ‘I finish’ vs. aca[b]a ‘s/he finishes’; o[pɾ] ‘I open’ vs. o[bɾ]ir ‘to open’) (Lloret 2003; 2004b). Under the assumption that these endings are codas, underapplication of epenthesis in 1PI is explained in terms of OO

128

Eulàlia Bonet and Maria-Rosa Lloret

paradigmatic correspondences, either as a uniformity (analogical) effect (Lloret 2004a; Wheeler 2005, 269–275) or as a contrast (homophony-avoidance) effect (Pons-Moll 2007 and references therein). The uniformity view put forward in Lloret (2004a) bases the analysis on the notion of “Optimal Paradigms” (OP, cf. McCarthy 2005), whose function is to control the correspondence relation between the output stems of the inflected forms of an inflectional paradigm, where no clear base can be identified as attractor. OP establishes an OO symmetrical correspondence relation between each potential stem allomorph (marked with ʻ]ʼ in the tableaux below), and a set of OP intraparadigmatic faithfulness constraints governs stem allomorphy. In insular Catalan, the ranking of OP-D EP-V (penalizing members with inserted vowels) above the sonority sequencing principle (SSP) and (IO-)D EP-V rules out epenthesis in 1PI; in turn, the addition of the epenthetic vowel throughout the paradigm to satisfy uniformity in stems is penalized by the highly ranked *H IATUS (cf. 38, realizations are from Majorcan Catalan; epenthetic vowels appear underlined). (Arguments against treating the inserted schwa as part of the inflection are presented in Lloret 2004b). In the following tableaux, paradigms appear in angle brackets. (38)

Nouns, with a paradigm of two inflected forms (), undergo epenthesis because it levels the paradigms in the other direction (39): the candidate with epenthesis in both forms wins because all members of the paradigm need a vowel to satisfy the sonority constraint. The OP approach, hence, is able to correlate the phonologically different behaviour of verbs and nouns to the fundamental difference in length in their respective paradigms. (39)

Phonology and morphology in Optimality Theory

129

In turn, the contrast view builds the analysis upon the notion of ‘paradigmatic contrast’ (PC) (Kenstowicz 2002), whose function is to avoid identical phonetic forms in a paradigm. According to Pons-Moll’s (2007) analysis of Balearic Catalan, PC blocks epenthesis to avoid homophony between the 1st and 3rd person singular present indicative of conjugation I verbs, because the 3rd person displays an unstressed inflectional -a (-[ə]) morph that would coincide with the epenthetic schwa in 1PI (cf. 40a), where we only include the 1st and 3rd person singular for illustration).13 Notice, however, that homophony itself is not a fatal problem, since other tenses show identical 1st and 3rd person singular in their paradigms (e.g. filtri ‘I, s/he filter (subjunctive)’). In these cases, though, the endings are input inflectional morphs in both forms (e.g. present subjunctive ‑/i/ suffix), which are preserved by high-ranked faithfulness constraints protecting input morphs – and especially single-segment affixes – such as the general constraint R EALIZE M(ORPHEME ), interpreted in the spirit of PARSE M ORPH (“A morph must be realized in the output”; Akinlabi 1996, 247) in (40b). According to this analysis, epenthesis is required in both members of the nominal paradigm because PC is not decisive here and hence SSP must be satisfied (40c). (40)

All the examples given so far belong to conjugation I verbs. The few verbs of conjugations II and III that have illicit consonantal endings (obr- ‘open’, umpl- ‘fill’, corr- ‘run’) show variation. In this case, the facts from Alghero Catalan favour the

13 An anonymous reviewer mentions that analogy is assumed to have been the driving force for the levelling in Old French between 3PL , with a final schwa, and 1PL , which originally had no final schwa but became homophonous with 3PI . An important difference between the French case and the insular Catalan case is that the former involves morphological material (exponents of a given morpheme) while the latter involves a phonological phenomenon (insertion or not of an epenthetic vowel).

130

Eulàlia Bonet and Maria-Rosa Lloret

OP proposal: in conjugation III, PI exhibits regular inflectional -i morphs except in 1PI (e.g. ), yet epenthesis in 1PI is banned. The data from most varieties of Balearic Catalan, in which all three singular PI forms lack a vocalic suffix, favour the OP approach too (e.g. ), although some varieties fit the contrast view better (i.e. ). The different behaviour between verbal forms, which allow final clusters with increasing sonority, and nominal forms, which surface with an epenthetic vowel in the same phonological context, cannot be dealt with in serial models of OT, which do not resort to OO constraints. They must assume instead that nominals do not have final epenthesis and that insular Catalan allows codas with increasing sonority.

4.2 Vowel reduction in Majorcan Catalan All dialects of Catalan have vowel reduction in unstressed position. In Majorcan Catalan (MC) /ɔ/ reduces to [o], and the non-high unrounded vowels /a/, /ɛ/, /e/ and /ə/ reduce to [ə] (cf. Mascaró 2002 for examples and discussion). However, in a complex set of cases [e] is found in unstressed position. Before addressing these exceptions to vowel reduction, let us see how vowel reduction can be accounted for within OT. Most analyses of vowel reduction in Catalan (Wheeler 2005; Lloret/Jiménez 2008; Pons-Moll 2011) are based on Crosswhite (1999; 2004). Vowel reduction is the result of the competition between prominence-related constraints and faithfulness to input vowel features. The combination of an accentual prominence scale and a vocalic prominence scale gives rise to the constraint ranking in (41) (Crosswhite 2004, (17)), where *–S TR is a shorthand for *U NSTRESSED. Under this ranking the vowel [ə] is the optimal vowel in unstressed position. (41) *–S TR a » *–S TR ɛ, ɔ » *–S TR e, o » *–S TR i, u » *–S TR ə As mentioned above, a fairly large number of words surface with unstressed [e]. Although it is not easy to find a systematic distribution for this exceptional presence of [e], some tendencies can be observed. As pointed out in Bibiloni (1998), the intervening factors are morphological relations, phonetic context, and Spanish L2 interference. An analysis of these factors is given in Pons-Moll (2011; cf. also references therein). Leaving aside the L2 interference, Pons-Moll (2011) proposes an OT account of the following two facts: (i) in nominal derivation (nouns and adjectives), [e] appears in the initial syllable of productive derivatives for which the base contains a syllable-initial stressed [ˈɛ] or [ˈe]; (ii) in verbal inflection, [e] appears in the initial syllable when the verbal paradigm contains, in the initial syllable, forms with [ˈe] (not with [ˈɛ], which always alternates with schwa). (42) provides examples

Phonology and morphology in Optimality Theory

131

in which one of the derivationally or inflectionally related words has a stressed vowel in the initial syllable. (42)

a.

Nominal derivatives: p[ˈe]ix ‘fish’, p[e]ixet ‘fish (dim.)’, but: p[ə]ixater ‘fisherman’ t[ˈɛ]rra ‘earth’, t[e]rreta ‘earth (dim.)’, but: t[ə]rrestre ‘terrestrial’

b.

Verbal inflection: p[ˈe]gues ‘(you) hit’, etc., p[e]gam ‘(we) hit’, p[e]garé ‘(I) will hit’, etc. cf.: x[ˈɛ]rres ‘(you) talk’, etc., but: x[ə]rram ‘(we) talk’, etc.

When the stem contains a stressed vowel in non-initial position, vowel reduction applies, as expected, when the vowel is unstressed (cf. pap[ˈe]r ‘paper’ and pap[ə]ret ‘small paper’, or cont[ˈe]sta ‘(s/he) answers’ and cont[ə]stam ‘(we) answer’). Pons-Moll (2011) resorts to McCarthy’s symmetric OP model to account for underapplication of vowel reduction in verbal inflection. For a verb like pegar ‘to hit’ OP constraints force all forms to end up having the same vowel in the first syllable, and this vowel is [e] instead of [ə], in both stressed and unstressed syllables, because the ranking proposed favours displaying [e] even in unstressed positions over having [ˈə] in stressed positions. Nominal derivatives are subject to asymmetric (base-dependent) OO constraints: a derivative like t[e]rreta ‘earth (dim.)’ has a surface [e] because it resembles the stressed [ˈɛ] of the base noun t[ˈɛ]rra ‘earth’ for the relevant features. The OO constraints proposed by Pons-Moll (2011) incorporate three additional notions within the same constraint: (i) reference to the initial syllable of the stem (the position in which underapplication of vowel reduction is found), (ii) reference to either paradigms or subparadigms, where the term ‘subparadigm’ is applied to productive derivation, and (iii) reference to a particular feature. For example, a form like *t[ə]rreta, with productive derivation, violates the constraint OO-S UBPAR I DENT I NITIAL S YLL S TEM (post) because the vowel in the initial syllable of the stem contains the feature [+posterior], while the base form t[ˈɛ]rra contains the feature [–posterior]; the grammatical form t[e]rreta does not violate this constraint. Contrariwise, a form like pap[ə]ret does not violate the constraint because the relevant vowel is not in the initial syllable of the stem. Turning to verbal forms, there are similar OP constraints, but in this case no reference to subparadigms is encoded. The fact that in verbal inflection underapplication is found only when the stressed vowel is [ˈe], while in derivation it is also found when the stressed vowel is [ˈɛ], is determined by the higher ranking of the constraint OPI DENT I NITIAL S YLL S TEM (ATR) ([ɛ] and [ə] being considered [–ATR]) and the lower ranking of OO-S UBPAR I DENT I NITIAL S YLL S TEM (ATR). This proposal accounts for most of the data but it does raise some questions, mostly related to the notion of ‘subparadigm’ applied to derivational morphology. With respect to inflectional morphology, subparadigms can easily be defined by

132

Eulàlia Bonet and Maria-Rosa Lloret

referring to some inflectional feature or category (like [±plural]), but it is much more difficult to relate the concept to degrees of productivity (cf. also sections 3.1 and 4.1). Ohannesian/Pons (2009) compare and discuss the two types of subparadigmatic relations (i.e. inflectional and derivational) and propose, for the derivational type, a set of universally ranked Paradigm Cohesion constraints, but it is difficult to foresee how these constraints would interact with the ones proposed in Pons-Moll (2011). Another question is what can count as a productive suffix, independently of regular vowel reduction or underapplication of vowel reduction. To give an example of the difficulties that arise, according to Bibiloni (1998), a derived word like ventall ‘fan’ is pronounced with a stressless unreduced [e], as in its base v[ˈe]nt ‘wind’, in spite of the low productivity of the suffix -all, while a word like p[ə]drera ‘quarry’, with a base p[ˈe]dra ‘stone’, has regular vowel reduction in spite of the high productivity of the suffix -era. The only cases where underapplication of vowel reduction seems to be systematic is evaluative morphology. It could easily be argued that evaluative morphemes, and more especially diminutives, have a different structure than other word-building suffixes (cf., among many others, De Belder/Faust/Lampitelli 2014 and references therein), and their particular phonological behaviour could be a consequence of this difference (on gradient productivity effects in Spanish, cf. also section 3.1). A further issue to consider is to what extent the application of the notion of subparadigms to derivation can be reduced to the notion of ‘lexically indexed constraints’ (Pater 2000, among others). Under this type of approach, some constraints have a general version, let’s say CG, but also a restricted one, a lexically indexed constraint, CL, which applies only to a specified set of lexical items, XL, CL always being ranked higher than CG. Typically these lexically indexed constraints are said to be faithfulness constraints. In the case at hand one could imagine indexed constraints like I DENT I NITIAL S YLL S TEM (post)L (instead of OO-S UBPAR I DENT I NITIAL S YLLS TEM (post)), which would be IO constraints. Finally, it remains to be studied whether an OT serial analysis of these facts would be able to provide better insights on this type of phenomenon.

5 Resorting to blending of existing forms? In this section we review three phenomena that have been accounted for in several papers by Steriade by resorting to the notion of ‘lexical conservatism’. The first of them, section 5.1, concerns the French bel/beau allomorphy that was discussed in section 2.1. Here we review the analysis put forward in Steriade (1999a; 2001) and also the counteranalysis suggested in Bermúdez-Otero (to appear). In section 5.2 we sketch the analysis that Steriade puts forward for the Latin perfect (Steriade 2012) and for Romanian derivation (Steriade 2008). Finally, in section 5.3 we address the allomorphy found in imperatives with enclitics in Balearic dialects of Catalan.

Phonology and morphology in Optimality Theory

133

5.1 Bel/beau allomorphy in French again Steriade (1999a; 2001), inspired by Perlmutter (1998)’s work on ‘lexical sourcing’, brings into play the notion of ‘lexical conservatism’ in order to restrain the number of inputs by limiting the set of candidates to pre-existing output forms that share semantic and morphosyntactic properties. For the French case, the OO correspondence relation that controls the use of a consonant-final form is given in (43). (43) L EX-C]: The absolute final C in the target allomorph of morpheme μ has a correspondent C’ in some listed allomorph of μ and is featurally identical to C’. (Steriade 2001, 7) The paradigm of adjectives with a single listed stem allomorph, either with a final vowel (e.g. [ʒɔli]) or with a final consonant (e.g. [kɛl]), yields marked syllabifications, with hiatus (e.g. [ʒɔli] abbé) or with a closed syllable (e.g. [kɛl] mari), because the creation of an unprecedented form through the insertion or loss of phonological material is penalized first (cf. e.g. 44). However, the paradigm of adjectives with two listed stem allomorphs (e.g. [bɛl] and [bo]) can satisfy *H IATUS without the creation of phonologically novel forms by simply resorting to the use of the consonantfinal listed allomorph (cf. 45). (44)

Listed allomorphs: [ʒɔli]

(45)

Listed allomorphs: [bɛl], [bo]

As mentioned in section 2.1, in favour of the lexically listed output-stem approach, Steriade notices the fact that for many French speakers some liaison forms in the masculine contexts do not completely coincide with the output of the citation feminine form, but show the stem vowel of the masculine and the liaison consonant of the feminine; cf. [sot] éléphant in (46). (46)

Fr.

[so] mari ‘silly husband’ [sɔt] femme ‘silly woman’

[sɔt] ~ [sot] éléphant ‘silly elephantM ’ [sɔt] éléphante ‘silly elephantF ’

Split-base formations such as [sot] alternating with [sɔt] reveal that while the feminine consonant is always used to satisfy L EX-C], the masculine vowel may be used to partially encode the grammatical gender of the adjective in order to satisfy a lexical conservative constraint targeting the stressed vowel of the stem, which signals gender (L EX-‘V(gender)). A global condition on lexical conservatism of stressed syllables (Lex-‘σ) ranked above or below L EX-‘V(gender) leads to a pure conservative solution ([sɔt] éléphant) or to a blend solution ([sot] éléphant), respectively.

134

Eulàlia Bonet and Maria-Rosa Lloret

Bermúdez-Otero (to appear) acknowledges the influence of independently existing (listed) output forms but argues against Steriade’s view and in favour of a serial approach. One of his main arguments is that Steriade has to resort to a specific constraint on salience to allow split bases in examples like [sot] éléphant ‘silly elephant (masculine)’ but to prevent blending in cases like [bel] abbé ‘beautiful abbot’, for which, through reranking of the relevant constraints, one should expect variation with *[bol] abbé, with the final [l] found in the feminine, a form that is never found. Another drawback is related to the resyllabification facts that were pointed out in (3) and (4) in section 2.1: while adjectives like petit(e) ‘small’ allow resyllabification in right dislocation, suppletive adjectives like [bɛl]–[bo] ‘nice’ do not. Steriade has to resort to specific constraints to account for this different behaviour. According to Bermúdez-Otero these problems do not arise in his Stratal OT analysis, which relies on the underlying form that the learner would posit for each item, which in turn gives rise to different surface allomorphs. Among the underlying representations he posits, following the basic aspects of the analysis in Wetzels (2002), are the ones that appear slightly adapted below. (47a) corresponds to invariable adjectives (a vowel-final adjective would have the same structure). The lexical item in (47b) (as well as the top one in 47c) has a floating segment. This segment can receive a skeletal slot either by docking to the next syllable in liaison environments or, when the item is feminine, by association to the feature [+fem] (Wetzels 1986). The item in (47c) has two allomorphs. When agreement takes place with a [+fem] noun, the lower allomorph is chosen; since the last consonant has an X slot, the condition in (48) is satisfied. When the item is not feminine the upper allomorph will be selected. The ‘S’ symbol that appears in (47c) represents allomorph selection prior to phonological evaluation. Finally (47d) also has two allomorphs, both of them without floating segments. The upper one is incompatible with feminine adjectives, which must obligatorily select the lower allomorph. The ‘P’ symbol indicates that when this incompatibility does not arise (that is, with masculine adjectives), the two allomorphs are available and the decision is left to the phonology. (47)

Phonology and morphology in Optimality Theory

(48)

Feminine suffix:

135

[+fem] ↕ X

The fact that each lexical item can have a different behaviour with respect to liaison follows here from the lexical idiosyncracies of each lexical item and therefore does not present the problem that Steriade’s LEX constraints would face. The differences in realization in right dislocation environments also mentioned above follow here from the underlying representations in (47) and the condition in (48): resyllabification is only possible with consonants that do not have a skeletal slot at the word level. Finally, under the stratal approach, (apparent) blending is only possible if there is a single underlying representation and the differences in the output forms can be attributed to regular phonological processes. An ungrammatical split base formation *[bol] in sequences like *[bol] abbé (instead of the grammatical *[bɛl] abbé) cannot arise because there is no phonological process of [l]-insertion in French. However, a liaison output form [sot] ‘silly’, which differs from the citation forms [so] (masculine) and [sɔt] (feminine), can be derived because at the word level, mid vowels are generally realized as mid-low in closed syllables and as mid-high in open syllables (loi de position ‘law of position’). Under this view it follows that ‘blending’ will never arise with suppletive allomorphs. The approach sketched here raises at least a couple of issues. A minor one is that feminine adjectives all end up having a skeletal slot but through two different mechanisms: either the skeletal slot is present in the underlying representation or it is added by association to the feature [+fem]. The other, more important, worry is related to the different treatment given to the underlying representations in (47c) and (47d). In (47c) the presence of the feature [+fem] in one of the allomorphs is said to prevent the other allomorph from being selected in feminine environments, but in (47d) the presence of the feature [–fem] does not prevent the other allomorph from being selected in masculine environments. In spite of the labels ‘S’ and ‘P’, this different interpretation of the representations does not follow from anything; it’s a mere stipulation.

5.2 Allomorphy in the Latin perfect and Romanian derivation Mester (1994) brought up an interesting case of allomorphy found in the Latin perfect in an influential paper on prosodic minimality and maximality, when OT was not yet fully developed. He takes into consideration verbs of the Latin conjugation II and proposes that the distribution of the allomorphs -u and -s in the perfect is prosodically driven, -u being the allomorph chosen when no conflict arises. For verbs like monēre ‘warn (present infinitive)’, the 1st person singular perfect monuī is

136

Eulàlia Bonet and Maria-Rosa Lloret

fully parsed with the default allomorph: [mo nu].14 However, a verb like augēre ‘enlargeIN F. P R S ’, must build the 1st person singular perfect with the more marked allomorph s, auksī, because it is the only one that allows perfect parsing: [auk] (cf. *[au]gu, with a trapped syllable).15 Embick (2010) argues that Mester’s global proposal wrongly predicts that when other suffixes are considered allomorph selection should vacillate depending on the prosodic structure of the whole word. For instance, one would expect to find *auguimus for the 1st plural perfect because it can be fully parsed into feet, [au][gui], while this is not the case for the grammatical form augsimus, [aug]si, which contains a trapped foot. Steriade (2012) addresses the allomorphy of the Latin perfect more broadly, taking into account all conjugations (not only the second one), all tenses with perfect aspect, and all types of allomorphy, not just the ‑u/‑s alternation. Among other things, she observes that, for any given verb, all perfect forms have a similar perfect stem (while there is a lot of variation across verbs). According to her, this similarity has to do with syllable count, not necessarily with segmental identity. Steriade’s crucial point is that although these phonological similarities are tied to morphosyntactically related forms, with one serving as base for the other, one is not contained in the other. This point can be illustrated with the relation between the verbal perfect and the perfect participle. (49) Lat.

a. b.

1SG perfect [scrip-s]-ī [hab-u]-ī

perfect participle [scrip-t]-us *scripitus [hab-it]-us *haptus

‘write’ ‘have’

In (49) the stem, enclosed in square brackets, ends up having the same number of syllables in both the perfect and the perfect participle. Different allomorphs are used but the allomorph chosen does not alter the syllable count. Notice that [scrip-s]-ī is not contained in [scrip-t]-us or vice versa. The constraint that controls syllable count is M AX V (P ERFECT ), which requires the perfect participle to have a vowel if the perfect stem contains one. Steriade also argues that the building of perfect forms is in turn influenced by non-perfect (infectum) forms. The complete direction of influence is infectum → perfect verbal forms → perfect participle. The fact that, as shown, there is no containment relation between forms excludes the possibility of a cyclic analysis or an analysis based on output-output correspondence of the sort argued for in Benua (1997). Steriade suggests that the selection of a base is related to type frequency. In the Latin case, the infectum has 16 different categories (combinations of mood, tense etc.), the perfect verbal forms six, and the perfect participle only two (participle and supine). A form with more categories can influence a form with fewer categories and 14 Square brackets indicate the edges of feet, while angle brackets mark extrametrical syllables. 15 In the examples from Latin the orthography is adapted phonologically when relevant.

Phonology and morphology in Optimality Theory

137

act as a base for that one. She further argues that this conception can be extended to the base-derivative relation, because the base also appears in the derivative, but not the other way around. Steriade (2008) discusses a case from Romanian to argue that inflection can determine phonological properties of derivation in ways similar to what we saw for Latin. Romanian has a productive phonological process of palatalization by which velar stops palatalize, only in derived environments, before a non-back vowel (K-Palatalization). In (50a) the presence of a palatal vocoid in the plural triggers palatalization, while in (50b) the selection of a non-palatal inflectional suffix does not trigger it: (50)

Rom.

a.

stângă [ˈstɨnɡ-ʌ] – stângi [ˈstɨndʒ-i ̯]

‘leftS G – P L ’

b.

foc [ˈfok] – focuri [ˈfok-uri ̯]

‘fireS G – P L ’

Steriade (2008) shows that the presence or absence of palatalization in inflection influences derivation in palatalizing contexts. If palatalized and non-palatalized roots alternate in inflection, the behaviour in derivation is as expected: for the adjective stângă ‘left’, which surfaces with a palatalized consonant in the plural, cf. (50a), in derivation the suffix ‑ist [‑ˈist] triggers palatalization while the suffix ‑aci [‑ˈatʃ ] doesn’t, cf. (51a). However, for the root meaning ‘fire’, in (50b), there is underapplication of palatalization in derivation: a velar consonant surfaces before a non-back vowel, cf. (51b). (51)

Rom.

a.

stângist [stɨnˈdʒ-ist] stângaci [stɨnˈg-atʃ ]

‘leftist’ ‘lefty’

b.

fochist [foˈk-ist], *[foˈtʃ-ist]

‘locomotive engineer’

The analysis proposed relies on the notion of derived lexicon; the generation of morphologically complex items is done in different passes through the grammar. In a first pass, inflected forms are generated; these are stored in the derived lexicon and are taken into account in the generation of morphologically derived words, through L EX P constraints. The relevant L EX P constraint for the cases at hand is I DENT L EX [αF] (where F stands for ‘feature’), defined below: (52)

I DENT L E X [αF]:

For any segment s in a subconstituent C of an expression under evaluation, if s is [αF] then s has an [αF] correspondent in a listed allomorph of C.

This constraint rules out candidates like *[foˈtʃ‑ist], because no listed allomorph has a palatal consonant. An additional point made by Steriade (2008) is that what counts for the presence or absence of palatalization in derivatives is not the potential

138

Eulàlia Bonet and Maria-Rosa Lloret

capacity of palatalizing but the actual presence of palatalized consonants in inflected bases. A proper name like Franco [ˈfrank-o] has a derivative Franchist [franˈk-ist] ‘Franco supporter’ (*[franˈtʃ-ist]), without palatalization, because the plural of Franco does not exist; that means that the plural is not stored in the derived lexicon and therefore cannot serve as a base for derivatives. The facts analysed in Steriade (2008) pose a challenge for theories based on the cycle, such as Lexical Phonology and Morphology (Kiparsky 1982) and OT versions of it (like Kiparsky 2000), because a plural is not a subconstituent of a derived word and therefore should not be available when derivational suffixes are attached to the root. For a reinterpretation of the facts and an analysis in Stratal Phonology that crucially resorts to thematic elements, cf. Bermúdez-Otero (to appear).

5.3 Imperatives with enclitics in Catalan In Catalan, the 2nd person imperative of verbs of conjugations II and III does not have any overt inflectional morphology; it also lacks a theme vowel (except for conjugation IIIa verbs, which have an -eix increment). In most cases these verbal forms are a bare stem and end in a consonant or a glide. However, when pronominal enclitics are added, extra verbal material appears, which we will refer to as ‘accretion’. The accretion can be a single vowel or a longer sequence, depending on the dialect and the verb. The examples below, from Bonet/Torres-Tamarit (2010), illustrate some of the accretions (underlined) in two insular varieties of Catalan: Formenteran and Majorcan. In these two varieties enclisis also causes stress displacement.16 (53)

Cat.

prometre ‘to promise’ (conjugation II) bullir ‘to boil’ (conjugation III) in isolation with enclitics a. Majorcan [pɾoˈmət] [pɾomətəˈli] [ˈbuʎ] [buˈʎil] b.

Formenteran

[pɾuˈmət] [ˈbuʎ]

[pɾuməˈtəli] [buʎiˈɣəl]

‘promise (to him/her)!ʼ ‘boil(itM )!’ ‘promise(to him/her)!’ ‘boil(itM )!’

Bonet/Torres-Tamarit (2010; 2011) analyse these cases as the effect of a phonological constraint requiring a prosodic head foot that outranks OO constraints. The phonological constraints are slightly different in each variety, Majorcan requiring an iamb 16 Central Catalan also has an accretion in enclisis, which is always realized as a schwa. We do not discuss it in this section because it differs in two significant ways from the other two varieties: (i) the accretion does not appear systematically in enclisis and shows some idiolectal variation; (ii) enclisis does not cause stress shift. Cf. Bonet/Torres-Tamarit (2011) for a description and analysis of the data.

Phonology and morphology in Optimality Theory

139

and Formenteran, a moraic trochee. These two constraints, which are needed independently to account for stress shift, appear in (54). (54)

a.

b.

Majorcan Cat. I AMB ]:

Assign one violation mark for any V+CL sequence that lacks an iamb aligned at the right edge.

Formenteran Cat. Assign one violation mark for any V+CL sequence that μT ROCHEE ]: lacks a moraic trochee aligned at the right edge.

These phonological constraints would be violated if the accretion were not present (cf. Formenteran [pɾumə(ˈtəli)], with a well-formed moraic trochee, vs. *[pɾuˈmətli], which would have an ill-formed foot). Regular epenthesis is also ruled out by A LIGN ( CL / VB ), requiring adjacency between the verb and clitics; cf. section 3.2 and specifically (30). In Formenteran Catalan the verbal root plus the accretion coincides exactly with the inflectional stem found in the first and second person plural imperative. Adopting the notion of lexical conservatism in Steriade (1999a) and later work, Bonet/ Torres-Tamarit (2010) propose two C ORR L E X constraints. One of them, (55a), allows a correspondence relation to be established between the inflectional stem of the cliticized imperative and other listed forms of the imperative alone, but not with verbal forms belonging to other tenses. (55b) is a more specific version of the constraint that penalizes candidates which, in addition, have different person and number (φ) features: (55)

a.

C ORR L E X I NFL S TEM Imp (C ORR L EX I): Assign one violation mark for any inflectional stem of a pre-clitic imperative that does not have a correspondent in the inflectional stem of an imperative form (the base).

b.

C ORR L E X I NFL S TEM Imp-φ (C ORR L EX I-φ): Assign one violation mark for any inflectional stem of a pre-clitic imperative that does not have a correspondent in the inflectional stem of an imperative form with the same φ-features (the base).

The tableau in (56) illustrates the proposal. The relevant listed inflectional stems include those belonging to the imperative. The two first candidates have a correspondent in one of the two forms found in the imperative, but the last one has a correspondent in a different tense (which could be, for instance, the imperfective indicative). For this reason this last candidate violates the constraint C ORR L E X I NFL S TEM Imp. The only candidate that does not violate any of the C ORR L E X constraints violates the highly ranked phonological constraint; therefore it is ruled out.

140

Eulàlia Bonet and Maria-Rosa Lloret

(56) Formenteran Cat. bull-la ‘boil itF !’ /buʎ#lə/: [buʎiˈɣə#lə] Listed output inflectional stems: [ˈbuʎi], [buʎiˈɣəj (m,w)]

The tableau in (56) does not include any candidates that are an unfaithful copy of one of the inflectional candidates. Such a candidate (for instance bu(ˈʎij lə)) would violate an OO-M AX constraint because two of the segments of the listed base ([buʎiˈɣəj (m,w)]) are not present in the candidate. In addition, the winning candidate has an inflectional stem that contains three segments more than the input; it violates IO-D EP. OO faithfulness constraints (OO-FAITH ) have a high ranking in Formenteran Catalan and are always satisfied (the relative ranking being OO-FAITH » IO-D EP ). However, the opposite relative ranking in Majorcan (IO-D EP » OO-FAITH ) is crucial for the selection of an accretion that is taken from other forms of the same tense, but which contains fewer segments. This point is illustrated in (57). All the forms of the imperative except for the 2nd person singular contain a velar segment [ɣ], but this segment is not present in the accretion of the winning candidate. (57) Majorcan Cat. resol-li ‘solve for him/her!’ /rəzɔl#li/: [rəzolə#ˈli] Some listed inflectional stems: [rəzˈɔli], [rəzolˈɣəj (m,w)]

The phenomenon described here is not easy to account for in a serial model of OT. One aspect to take into account is that in some cases the accretion can hardly be identified with a single morpheme. This is the case of the Formenteran sequence [iˈɣə], whose derivation was illustrated in (56). Another even more relevant aspect is that the appropriateness of the accretion can only be evaluated once it has been inserted and feet have been built. There is no justification for the presence of the accretion before the incorporation of clitics.

6 Conclusion In this chapter we have presented several phenomena from Romance languages that lie at the core of the phonology-morphology interface. We have discussed different OT analyses of these phenomena, focusing on a central debate in the theory, namely

Phonology and morphology in Optimality Theory

141

whether the interaction between the two components proceeds in a parallel or a serial fashion. Parallel accounts need to resort, in many cases, to output-output constraints to explain some of these interactions. Among serial OT models, two of them have been widely explored in the literature. Stratal OT incorporates seriality through morphologically ordered levels, but within each level evaluation is done in a parallel fashion; different levels can have different constraint rankings. Harmonic Serialism, instead, recovers the one-change-at-a-time procedure of classic generative phonology but using ranked constraints instead of rules, with a fixed ranking throughout the whole derivation. Some of the phenomena that have been reviewed here can be treated equally well within both the serial and the parallel views, while for other phenomena one view or the other seems better suited to handle the facts. It is difficult, therefore, to find a single approach that could account for all the phenomena in a satisfactory fashion and close the debate. More work needs to be done in the area of Romance linguistics to tilt the scales, and more attention should be paid to frequency effects, briefly discussed at the end of section 3.1. These effects are explored in models like stochastic Optimality Theory (Boersma/Hayes 2001, and others), in which a mechanism is proposed to compute the probability of outputs through the assignation of a numerical value to each constraint.

7 References Akinlabi, Akinbiyi (1996), “Featural affixation”, Journal of Linguistics 32, 239–289. Albright, Adam/Andrade, Argelia/Hayes, Bruce (2001), “Segmental environments of Spanish diphthongization”, in: Adam Albright/Taehong Cho (edd.), UCLA Working Papers in Linguistics 7. Papers in phonology 5, 117–151. Álvarez Blanco, Rosario (1983), “O artigo en galego. Morfoloxía”, Verba. Anuario galego de filoloxía 10, 169–182. Aronoff, Mark (1994), Morphology by itself. Stems and inflectional classes, Cambridge, MA, MIT Press. Baković, Eric J. (1998), “Spanish codas and overapplication”, in: Armin Schwegler/Bernard Tranel/ Myriam Uribe-Etxebarria (edd.), Romance linguistics. Theoretical perspectives (Selected papers from the 27th linguistic symposium on Romance languages, 1997), Amsterdam/Philadelphia, Benjamins, 13–23. Baković, Eric J. (2001), “Nasal place neutralization in Spanish”, in: Michelle Minnick Fox/Alexander Williams/Elsi Kaiser (edd.), U. Penn working papers in linguistics 7.1: Proceedings of the 24th annual Penn linguistics colloquium, Philadelphia, PA, PWPL, 1–13. Benua, Laura (1995), “Identity effects in morphological truncation”, in: Jill Beckman/Laura Walsh Dickey/Suzanne Urbanczyk (edd.), University of Massachusetts occasional papers in linguistics 18. Papers in Optimality Theory, Amherst, MA, Graduate Linguistic Student Association, 77–136, http://roa.rutgers.edu/files/74-0000/74-0000-BENUA-0-0.PDF (06.05.2016). Benua, Laura (1997), Transderivational identity. Phonological relations between words, PhD dissertation, University of Massachusetts, Amherst (published as Phonological relations between words, New York, Garland, 2000).

142

Eulàlia Bonet and Maria-Rosa Lloret

Bermúdez-Otero, Ricardo (2006), “Morphological structure and phonological domains in Spanish denominal derivation”, in: Fernando Martínez-Gil/Sonia Colina (edd.), Optimality-theoretic studies in Spanish phonology, Amsterdam/Philadelphia, Benjamins, 278–311. Bermúdez-Otero, Ricardo (2013), “The Spanish lexicon stores stems with theme vowels, not roots with inflectional class features”, Probus 25, 3–103. Bermúdez-Otero, Ricardo (to appear), “Stratal phonology. Arguments for cyclic containment, morphological implications”, in: Andrew Spencer (ed.), The handbook of morphology, second edition, Oxford, Blackwell. Bertinetto, Pier Marco/Loporcaro, Michele (2005), “The sound pattern of Standard Italian, as compared with the varieties spoken in Florence, Milan and Rome”, Journal of the International Phonetic Association 35, 131–151. Bibiloni, Gabriel (1998), “La e àtona en el català de Mallorca”, in: Josep Massot i Muntaner (ed.), Estudis de llengua i literatura en honor de Joan Veny II, Barcelona, Universitat de Barcelona/ Publicacions de l’Abadia de Montserrat, 533–539. Boersma, Paul/Hayes, Bruce (2001), “Empirical tests of the gradual learning algorithm”, Linguistic Inquiry 32, 45–86. Bonet, Eulàlia/Lloret, Maria-Rosa (2002), “OCP effects in Catalan cliticization”, Catalan Journal of Linguistics 1, 19–39. Bonet, Eulàlia/Lloret, Maria-Rosa (2003), “More on alignment as an alternative to domains. The syllabification of Catalan clitics”, Probus 17, 37–78. Bonet, Eulàlia/Lloret, Maria-Rosa (2005), “Against serial evaluation in Optimality Theory”, Lingua 115, 1303–1323. Bonet, Eulàlia/Lloret, Maria-Rosa/Mascaró, Joan (2003), “Phonology-morphology conflicts in gender allomorphy. A unified approach”, paper presented at the 26th Generative Linguistics in the Old World Colloquium, Lund University. Bonet, Eulàlia/Lloret, Maria-Rosa/Mascaró, Joan (2007), “Allomorph selection and lexical preferences”, Lingua 117, 903–927. Bonet, Eulàlia/Torres-Tamarit, Francesc (2010), “Allomorphy in pre-clitic imperatives in Formenteran Catalan. An output-based analysis”, in: Sonia Colina/Antxon Olarrea/Ana Maria Carvalho (edd.), Romance Linguistics 2009. Selected papers from the 39th Linguistic Symposium on Romance Languages (LSRL), Tucson, Arizona, March 2009, Amsterdam/Philadelphia, Benjamins, 337–351. Bonet, Eulàlia/Torres-Tamarit, Francesc (2011), “Les formes d’imperatiu seguides de clític. Un cas de conservadorisme lèxic”, in: Maria-Rosa Lloret/Clàudia Pons (edd.), Noves aproximacions a la fonologia i la morfologia del català (Col·lecció Symposia Philologica), Alacant, Institut Interuniversitari de Filologia Valenciana, 37–61. Burzio, Luigi (1996), “Surface constraints versus underlying representations”, in: Jacques Durand/ Bernard Laks (edd.), Current trends in phonology. Models and methods, Salford, European Studies Research Institute, 123–141. Burzio, Luigi (2002), “Surface-to-surface morphology. When your representations turn into constraints”, in: Paul Boucher (ed.), Many morphologies, Somerville, MA, Cascadilla Press, 142–177. Burzio, Luigi (2005), “Sources of paradigm uniformity”, in: Laura J. Downing/T. Alan Hall/Renate Raffelsiefen (edd.), Paradigms in phonological theory, Oxford, Oxford University Press, 65–106. Bybee, Joan (1985), Morphology, Amsterdam/Philadelphia, Benjamins. Carstairs, Andrew (1987), Allomorphy in inflexion, London, Croom Helm. Carstairs, Andrew (1988), “Some implications of phonologically-conditioned suppletion”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of morphology 1988, Dordrecht, Foris, 67–94. Chierchia, Gennaro (1986), “Length, syllabification, and the phonological cycle in Italian”, Journal of Italian Linguistics 8, 5–34.

Phonology and morphology in Optimality Theory

143

Colina, Sonia (1995), A constraint-based analysis of syllabification in Spanish, Catalan and Galician, PhD dissertation, University of Illinois, Urbana-Champaign. Côté, Marie-Hélène (2011), “French liaison”, in: Marc van Oostendorp/Colin J. Ewen/Elizabeth Hume/ Keren Rice (edd.), The Blackwell companion to phonology, vol. V: Phonology across languages, Chichester, Wiley-Blackwell, 2685–2710. Crosswhite, Katherine (1999), Vowel reduction in Optimality Theory, PhD dissertation, UCLA. [Published in 2001, New York/London, Routledge]. Crosswhite, Katherine (2004), “Vowel reduction”, in: Bruce Hayes/Robert Kirchner/Donca Steriade (edd.), Phonetically based phonology, Cambridge, Cambridge University Press, 191–231. Davis, Stuart (1990), “Italian onset structure and the distribution of ‘il’ and ‘lo’”, Linguistics 28, 43– 55. De Belder, Marijke/Faust, Noam/Lampitelli, Nicola (2014), “On a low and high diminutive. Evidence from Italian and Hebrew”, in: Artemis Alexiadou/Hagit Borer/Florian Schäfer (edd.), The syntax of roots and the roots of syntax, Oxford, Oxford University Press, 149–163. Del Gobbo, Francesca (2001), “An OT account of the distribution of articles in Italian”, in: Wind Cowles/Linda Godson/Andy Hickl/Alicia Muñoz/Paola Nieddu/Todd O’Bryan/Heidi Tonomura (edd.), Proceedings of the 5th annual Southwest workshop on Optimality Theory (SWOT V), University of California, San Diego, May 15–16, 1999 (Linguistic notes from La Jolla, vol. 20), La Jolla, CA, Department of Linguistics, University of California, San Diego, 1–18. Dols, Nicolau (1993), “La silabificación del catalán de Mallorca”, Hispanorama 65, 22–26. Dols, Nicolau (2000), “Teoria fonològica i sil·labificació. El cas del català de Mallorca”, PhD dissertation, Palma, Universitat de les Illes Balears. Dols, Nicolau/Wheeler, Max W. (1996), “El consonantisme final mallorquí i el ‘llicenciament d’obertures’”, Caplletra. Revista internacional de filologia 19, 51–63. Eddington, David (1996), “Diphthongization in Spanish derivational morphology. An empirical investigation”, Hispanic Linguistics 8, 1–13. Eddington, David (2012), “Morphophonological alternations”, in: José-Ignacio Hualde/Antxon Olarrea/ Erin O’Rourke (edd.), Handbook of Hispanic linguistics, Cambridge, Wiley-Blackwell, 193–208. Embick, David (2010), Localism versus globalism in morphology and phonology, Cambridge, MA, MIT Press. Embick, David (2013), “Contextual conditions on stem alternations. Illustrations from the Spanish conjugation”, in: Irene Franco/Sara Lusini/Andres Saab (edd.), Romance languages and linguistic theory 2010. Selected papers from “Going Romance”, Leiden, 2010, Amsterdam/Philadelphia, Benjamins, 21–40. Greenberg, Joseph H. (1963), “Some universals of grammar with particular reference to the order of meaningful elements”, in: Joseph Greenberg (ed.), Universals of language, Cambridge, MA, MIT Press, 73–113. Halle, Morris/Harris, James W./Vergnaud, Jean-Roger (1991), “A reexamination of the stress erasure convention and Spanish stress”, Linguistic Inquiry 22, 141–159. Hargus, Sharon (1998), “Prosodically conditioned allomorphy in Optimality Theory. Brevity in Witsuwit’en”, Ms., University of Washington. Harris, James W. (1969), Spanish phonology, Cambridge, MA, MIT Press. Harris, James W. (1983), Syllable structures and stress in Spanish. An autosegmental analysis, Cambridge, MA, MIT Press. Harris, James W. (1985), “Spanish diphthongization and stress. A paradox resolved”, Phonology Yearbook 2, 31–45. Harris, James W. (1992), “The form classes of Spanish substantives”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of Morphology 1991, Dordrecht, Kluwer, 65–88.

144

Eulàlia Bonet and Maria-Rosa Lloret

Harris, James W. (1999), “Nasal depalatalization ‘no’, morphological wellformedness ‘sí’; the structure of Spanish word classes”, in: Karlos Arregi/Benjamin Bruening/Cornelia Krause/Vivian Lin (edd.), MIT working papers in linguistics 33. Papers on morphology and syntax, cycle one, Cambridge, MA, MIT, 47–82. Jaeggli, Osvaldo A. (1980), “Spanish diminutives”, in: Frank H. Nuessel (ed.), Contemporary studies in Romance languages, Bloomington, IN, Indiana University Linguistics Club, 142–158. Jiménez, Jesús (1997), “L’estructura sil·làbica del dialecte valencià”, PhD dissertation, Universitat de València. Jiménez, Jesús/Todolí, Júlia (1995), “La forma dels pronoms clítics catalans. Condicions sil·làbiques i alineament morfològic”, in: Carlos Martín Vide (ed.), Lenguajes naturales y lenguajes formales XI, Barcelona, PPU, 429–437. Kager, René (1999a), “Surface opacity of metrical structure in Optimality Theory”, in: Ben Hermans/ Marc van Oostendorp (edd.), The derivational residue in phonological Optimality Theory, Amsterdam/Philadelphia, Benjamins, 207–245. Kager, René (1999b), Optimality Theory, Cambridge, Cambridge University Press. Kenstowicz, Michael (1994), “Syllabification in Chukchee. A constraint-based analysis”, in: Alice Davison/Nicole Maier/Glaucia Silva/Wan Su Yan (edd.), Proceedings of the formal linguistics society of Mid America 4, Iowa City, University of Iowa, 160–181. Kenstowicz, Michael (1997), “Quality-sensitive stress”, Rivista di Linguistica 9, 157–187. Kenstowicz, Michael (2002), “Paradigmatic uniformity and contrast”, in: Anikó Csirmaz/Zhiqiang Li/ Andrew Nevins/Olga Vaysman/Michael Wagner (edd.), MIT working papers in linguistics 42. Phonological answers (and their corresponding questions), Cambridge, MA, MITWPL, 141–164. (Republished in: Laura J. Downing/T. Allan Hall/Renate Raffelsiefen (edd.), Paradigms in phonological theory, Oxford, Oxford University Press, 2005, 145–169). Kikuchi, Seiichiro (1999), “Opacity and transparency in Spanish plurals. A sympathetic approach”, On’in Kenkyu [Phonological Studies] 2, 61–68. Kikuchi, Seiichiro (2006), “On Galician definite article allomorphy”, On’in Kenkyu [Phonological Studies] 9, 41–48. Kiparsky, Paul (1982), “From cyclic phonology to lexical phonology”, in: Harry van der Hulst/Norval Smith (edd.), The structure of phonological representations, part I, Dordrecht, Foris, 131–175. Kiparsky, Paul (2000), “Opacity and Cyclicity”, The Linguistic Review 17, 351–367. Lloret, Maria-Rosa (2002), “Estructura sil·làbica”, in: Joan Solà/Maria-Rosa Lloret/Joan Mascaró/ Manuel Pérez Saldanya (edd.), Gramàtica del català contemporani, vol. 1, Barcelona, Empúries, 195–249. Lloret, Maria-Rosa (2003), “Variació condicionada. Aspectes qualitatius i quantitatius”, Catalan Review. International Journal of Catalan Culture 17, 107–122. Lloret, Maria-Rosa (2004a), “The phonological role of paradigms. The case of insular Catalan”, in: Julie Auger/Clancy Clements/Barbara Vance (edd.), Contemporary approaches to Romance linguistics, Amsterdam/Philadelphia, Benjamins, 275–297. Lloret, Maria-Rosa (2004b), “Efectes col·laterals de la ‘desinència zero’ en la flexió verbal algueresa”, in: Miscel·lània Joan Veny, vol. 5 (Estudis de llengua i literatura catalanes 49), Barcelona, Abadia de Montserrat, 233–266. Lloret, Maria-Rosa/Jiménez, Jesús (2008), “Marcatge posicional i prominència en el vocalisme àton”, Caplletra. Revista internacional de filologia 45, 55–91, http://www.raco.cat/index.php/Caplletra/ article/view/276800/364747 (06.05.2016). Lloret, Maria-Rosa/Mascaró, Joan (2006), “Depalatalization in Spanish revised”, in: Fernando MartínezGil/Sonia Colina (edd.), Optimality-theoretic studies in Spanish phonology, Amsterdam/ Philadelphia, Benjamins, 74–98.

Phonology and morphology in Optimality Theory

145

Marotta, Giovanna (1993), “Selezione dell’articolo e sillaba in italiano. Un’interazione totale?”, Studi di grammatica italiana 15, 255–293. Mascaró, Joan (1978), Catalan phonology and the phonological cycle, Bloomington, IN, Indiana University Linguistics Club. Mascaró, Joan (1983), “Three Spanish variations and a Majorcan theme”, Ms., Universitat Autònoma de Barcelona. Mascaró, Joan (1986), Morfologia, Barcelona, Enciclopèdia Catalana. Mascaró, Joan (1996), “External allomorphy and contractions in Romance”, Probus 8, 181–205. Mascaró, Joan (2002), “El sistema vocàlic. Reducció vocàlica”, in: Joan Solà/Maria-Rosa Lloret/Joan Mascaró/Manuel Pérez Saldanya (edd.), Gramàtica del català contemporani, vol. 1, Barcelona, Empúries, 89–123. Mascaró, Joan (2007), “External allomorphy and lexical representation”, Linguistic Inquiry 38, 715– 735. McCarthy, John J. (1993), “The parallel advantage. Containment, consistency and alignment”, Talk presented at Rutgers Optimality workshop 1, Rutgers University, New Brunswick, NJ. McCarthy, John J. (1999), “Sympathy and phonological opacity”, Phonology 16, 331–399. McCarthy, John J. (2000), “Harmonic serialism and parallelism”, in: Masako Hirotani/Andries Coetzee/ Nancy Hall (edd.), Proceedings of the North East linguistic society 30, Amherst, MA, Graduate Linguistic Student Association, 501–524. McCarthy, John J. (2005), “Optimal paradigms”, in: Laura J. Downing/T. Allan Hall/ Renate Raffelsiefen (edd.), Paradigms in phonological theory, Oxford, Oxford University Press, 170–210. McCarthy, John J./Prince, Alan (1993), “Generalized alignment”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of Morphology, Dordrecht, Kluwer, 79–153. McCarthy, John J./Prince, Alan (1994), “The emergence of the unmarked. Optimality in prosodic morphology”, in: Mercè González (ed.), Proceedings of the North East linguistic society, Amherst, MA, Graduate Linguistic Student Association, 333–379. McCarthy, John J./Prince, Alan (1995), “Faithfulness and reduplicative identity”, in: Jill Beckman/ Laura Walsh Dickey/Suzanne Urbanczyk (edd.), University of Massachusetts occasional papers in linguistics 18. Papers in Optimality Theory, Amherst, MA, Graduate Linguistic Student Association, 249–384, http://roa.rutgers.edu/files/60-0000/60-0000-MCCARTHY-0-0.PDF (06.05.2016). McCrary, Kristie Marie (2004), Reassessing the role of the syllable in Italian phonology. An experimental study of consonant cluster syllabification, definite article allomorphy and segment duration, PhD dissertation, University of California, Los Angeles. Mester, Armin (1994), “The quantitative trochee in Latin”, Natural language and linguistic theory 12, 1–61. Morelli, Frida (1999), The phonotactics and phonology of obstruent clusters in Optimality Theory, PhD dissertation, University of Maryland, College Park. Nevins, Andrew (2011), “Phonologically conditioned allomorph selection”, in: Marc van Oostendorp/ Colin J. Ewen/Elizabeth Hume/Keren Rice (edd.), The Blackwell companion to phonology, vol. 4: Phonological interfaces, Chichester, Wiley-Blackwell, 2357–2382. Ohannesian, Maria/Pons, Clàudia (2009), “Shattering paradigms. An attempt to formalize pressures within subparadigms”, in: Gilles Boyé/Nabil Hathout/Fabio Montermini (edd.), Proceedings of Décembrettes 6, Somerville, MA, Cascadilla Proceedings Project, 76–94. Paster, Mary (2006), Phonological conditions on affixation, PhD dissertation, University of California, Berkeley, http://pages.pomona.edu/~mp034747/Paster_dissertation.pdf (06.05.2016). Pater, Joe (2000), “Non-uniformity in English secondary stress. The role of ranked and lexically specific constraints”, Phonology 17, 237–274. Pensado, Carmen (1997), “On the Spanish depalatalization /ɲ/ and /ʎ/ in rhymes”, in: Fernando Martínez-Gil/Alfonso Morales (edd.), Issues in the phonology and morphology of the major Iberian languages, Washington, Georgetown University Press, 595–618.

146

Eulàlia Bonet and Maria-Rosa Lloret

Perlmutter, David (1998), “Interfaces. Explanation of allomorphy and the architecture of grammars”, in: Steven Lapointe/Diane Brentari/Patrick M. Farrell (edd.), Morphology and its relation to phonology and syntax, Stanford, CSLI Publications, 307–338. Piera, Carlos (1985), “On the representation of higher order complex words”, in: Larry D. King/ Catherine A. Male (edd.), Selected papers from the XIIIth Linguistic Symposium on Romance Languages, Chapel Hill, N.C., 24–26 March 1983, Amsterdam/Philadelphia, Benjamins, 287– 313. Pons-Moll, Clàudia (2007), “The true mystery of the world is the visible, not the invisible. Some reflections on the verbal morphophonology of Balearic Catalan”, Acta Linguistica Hungarica 54, 295–339. Pons-Moll, Clàudia (2011), “Underapplication of vowel reduction to schwa in Majorcan Catalan productive derivation and verbal inflection”, in: Janine Berns/Haike Jacobs/Tobias Scheer (edd.), Romance languages and linguistic theory 2009. Selected papers from “Going Romance”, Nice 2009, Amsterdam/Philadelphia, Benjamins, 273–289. Pons-Moll, Clàudia (2012), “When diachrony meets synchrony. Phonological change, phonological variation and Optimal Paradigms”, in: Gunther De Vogelaer/Guido Seiler (edd.), The dialect laboratory. Dialects as a testing ground for theories of language change, Amsterdam/Philadelphia, Benjamins, 197–225. Prince, Alan/Smolensky, Paul (1993/2004), Optimality Theory. Constraint interaction in generative grammar, New Brunswick/Boulder, Rutgers University/University of Colorado, 1993, http:// roa.rutgers.edu/files/537-0802/537-0802-PRINCE-0-0.PDF (06.05.2016). Reprint: Malden, MA, Blackwell, 2004. Repetti, Lori (1991), “A moraic analysis of raddoppiamento fonosintattico”, Rivista di Linguistica 3, 307–330. Serra, Pep (1996), La fonologia prosòdica del català, PhD dissertation, Universitat de Girona. Steriade, Donca (1999a), “Lexical conservatism in French adjectival liaison”, in: Jean-Marc Authier/ Barbara E. Bullock/Lisa A. Reed (edd.), Formal perspectives in Romance linguistics. Selected papers from the 28th linguistic symposium on Romance languages, Amsterdam/Philadelphia, Benjamins, 243–270. Steriade, Donca (1999b), “Alternatives to syllable-based accounts of consonantal phonotactics”, in: Osamu Fujimura/Brian D. Joseph/Bohumil Palek (edd.), Proceedings of the 1998 linguistics and phonetics conference, Prague, The Karolinum Press, 205–242, http://www.linguistics.ucla.edu/ people/steriade/papers/Alternatives_to_Syllables.pdf (06.05.2016). Steriade, Donca (2001), “Lexical conservatism and the notion base of affixation”, Ms. University of California, Los Angeles, http://www.linguistics.ucla.edu/people/steriade/papers/ LexicalConservatism.pdf (06.05.2016). Steriade, Donca (2008), “A pseudo-cyclic effect in Romanian morphophonology”, in: Asaf Bachrach/ Andrew Nevins (edd.), Inflectional identity, Oxford, Oxford University Press, 313–359. Steriade, Donca (2009 [2001]), “The phonology of perceptibility effects. The P-map and its consequences for constraint organization”, in: Kristin Hanson/Sharon Inkelas (edd.), The nature of the word. Studies in honor of Paul Kiparsky, Cambridge, MA, MIT Press, 151–179. Steriade, Donca (2012), “The cycle without containment: Latin perfect stems”, Ms. Cambridge, MA, MIT. Tranel, Bernard (1981), Concreteness in generative phonology. Evidence from French, Berkeley/Los Angeles, The University of California Press. Tranel, Bernard (1990), “On suppletion and French liaison”, Probus 2, 169–208. Tranel, Bernard (1996a), “Exceptionality in Optimality Theory and final consonants in French”, in: Karen Zagona (ed.), Grammatical theory and Romance languages. Selected papers from the 25th linguistic symposium on Romance languages (LSRL XXV), 2–4 March 1995, Amsterdam/ Philadelphia, Benjamins, 275–291.

Phonology and morphology in Optimality Theory

147

Tranel, Bernard (1996b), “French liaison and elision revisited. A unified account within Optimality Theory”, in: Claudia Parodi/Carlos Quicoli/Mario Saltarelli/María Luisa Zubizarreta (edd.), Aspects of Romance linguistics. Selected papers from the Linguistic Symposium on Romance Languages XXIV, March 10–3, 1994, Washington, D.C., Georgetown University Press, 433–455. Tranel, Bernard (1999), Suppletion and OT. On the issue of the syntax/phonology interaction, in: Emily Curtis/James Lyle/Gabriel Webster (edd.), The proceedings of the sixteenth West coast conference on formal linguistics (WCCFL 16), Stanford, CSLI Publications, 415–429. Viaplana, Joaquim (1980), “Algunes consideracions sobre les formes pronominals clítiques del barceloní”, Anuario de Filología 6, 459–483. Vogel, Irene (1977), The syllable in phonological theory with special reference to Italian, PhD dissertation, Stanford University. Wetzels, Leo (1986), “The timing of latent consonants in Modern French”, in: Carol Neidle/Rafael Núñez-Cedeño (edd.), Studies in Romance languages, Dordrecht, Foris, 283–317. Wetzels, Leo (2002), “Les adjectifs pré-nominaux du français. Formes longues et formes féminines”, in: Haike Jacobs/Leo Wetzels (edd.), Liber amicorum Bernard Bichakjian, Maastricht, Shaker, 273–306. Wheeler, Max W. (1979), Phonology of Catalan, Oxford, Blackwell. Wheeler, Max W. (2005), The phonology of Catalan, Oxford, Oxford University Press. Wiltshire, Caroline/Maranzana, Elisa (1999), “Geminates and clusters in Italian and Piedmontese. A case for OT ranking”, in: Jean-Marc Authier/Barbara E. Bullock/Lisa A. Reed (edd.), Formal perspectives on Romance linguistics. Selected papers from the 28th linguistic symposium on Romance languages (LSRL XXVIII), University Park, 16–19 April 1998, Amsterdam/Philadelphia, Benjamins, 289–303. Wolf, Matthew (2008), Optimal interleaving. Serial phonology-morphology interaction in a constraintbased model, PhD dissertation, Amherst, University of Massachusetts.

Sascha Gaglia and Marc-Olivier Hinzelin

5 Inflectional verb morphology Abstract: In this chapter, we shall present a selection of the most relevant phenomena in inflectional verb morphology, i.e. where a debate about their nature has significantly evolved: With respect to the morphology-phonology interface, i.e. morphophonology, we shall basically highlight stem allomorphy. Throughout this chapter, syncretism will play a major role as a prime example and recurrent theme, which elucidates the variety of interfaces pertaining to the inflection of verbs – as all important interfaces with morphology are involved: phonology, syntax, and semantics/ pragmatics. Other instances of non-canonical morphology discussed include suppletion, periphrases, overabundance, and defectiveness. Lastly, we shall address the notion of the autonomy of morphology, i.e. the assumption of an autonomous morphomic level. Keywords: defectiveness, grammaticalization, high-mid alternation, metaphony, palatalization, velar insertion, morphophonology, overabundance, periphrasis, suppletion, syncretism

1 Introduction The notion of interfaces relies crucially on the assumption of a modular structure of the mind in general and of the human language capacity in particular (cf. Chomsky 1980, 40–46, 59–65, 89–92). An important question that has to be discussed beforehand is what actually constitutes an interface phenomenon. The answer to this fundamental question is utterly theory-dependent: First, only in a theoretical framework which assumes a modular structure is this question meaningful at all.1 Second, if such a framework (e.g. generative grammar) is adopted, the answer depends largely on the internal division into modules, e.g. if one assumes that morphology is an autonomous module and, beyond that, possesses a morphomic level in the sense of Aronoff (1994), many phenomena that are possibly at the interface may be explained by invoking morphology alone, without any other module at the interface (cf. section 5). If, on the other hand, the existence of a separate morphology module is denied altogether, then, in consequence, there cannot exist any interface with it.2 If a theory 1 The non-modularity assumption is shared by frameworks like Cognitive Linguistics, Construction Grammar or Functional Grammar. For this reason, interfaces do not exist in these frameworks and they will not be discussed in this chapter. 2 In many generative approaches, syntax and morphology are lumped together as ‘grammar’ or ‘morphosyntax’, cf. section 3.

150

Sascha Gaglia and Marc-Olivier Hinzelin

assumes a separate morphology module in the first place, then this will be situated almost always at the center of the linguistic model and thus has direct and important ties to all other modules.3 These considerations will constitute the starting point for the present chapter. The morphology-phonology interface is the most obvious and the first one to be assessed in linguistic research historically. Trubetzkoy coined the term morphophonology (“morphonologie” in Trubetzkoy 1929) and provides one of the earliest and best-known examples of an explicit description of the inflectional morphology interface (cf. also Dressler 1985 and the debate on morphophonology in the contributions in Singh/Desrochers 1996). As mentioned above, many morphological phenomena are situated at several interfaces at the same time; syncretism has phonological, syntactic, and semantic/ pragmatic angles which are more or less prevalent depending on the case at hand. Furthermore, a notion like ‘morphosyntactic property’ can only be interpreted at several interfaces simultaneously: A property like ‘person/number’ is morphological and corresponds directly with syntax but has a pragmatic-semantic interpretation. Morphological features may thus be used by the syntax for agreement purposes (cf. section 3.1) and are decoded in semantics and pragmatics. Hence a phenomenon crucially invoking ‘morphosyntactic properties’ is per se at the interfaces with syntax (cf. section 3) and semantics/pragmatics (cf. section 4). Choosing the most relevant interface to house the discussion of the phenomenon proves to be a challenging decision. Weighing the different contributions from each module against each other turns out to be a rather intricate task, thus the final decision often remains quite an arbitrary one.4 Lastly, any account of interfaces with morphology inevitably needs to address the concept of ‘grammaticalization’: Linguistic entities participating in a grammaticalization process are always, by definition, at (several) interfaces. More so, they always end up in the morphology module following the “cline of grammaticality” (cf. Hopper/Traugott 2003, 7, 99–100) or “grammaticalization scale” (cf. Lehmann 1985, 44–45, 47). In the present chapter, a structuralist perspective (item-and-arrangement; cf. Hockett 1954) is sometimes taken as the basis for a segmentation into morphemes. A typical Romance verb form may thus be segmented into the following parts: The stem with the root (r) (or an already morphologically complex form in the case of

3 The modules themselves may be even subdivided further and ‘traditional’ interfaces may be perceived as separate modules themselves, cf. Zwicky (1983) for this “high modularity” approach in which he divides morphology into three separate modules: word formation, allomorphy, and morphophonemics (Zwicky 1983, 199, 206). 4 Although a sophisticated analysis of grammatical features along the lines of Kibort (2011) may eventually provide a definitive answer inside a framework adopted.

Inflectional verb morphology

151

derived verbs) and a theme vowel (thV) identifying the conjugation class as well as the desinence consisting of a tense/mood (TM) and a person/number suffix (PN):5 (1)

Sp.

habl r stem

+a + thV

+ ba + mos + TM + PN + desinence

(1PL IMPF. IND hablar ‘speak’)

2 Interfaces with phonology In this section, some of the most important interface analyses of morphophonological phenomena occurring within the conjugation systems of Romance languages will be presented and discussed. Therefore, we shall consider studies which assume at least a bi-modular process, i.e. involving inflectional morphology and phonology, even if these studies do not specifically refer to the concept of interfaces. The decisive criterion is, thus, the combination of inflectional features, principles, and/ or structures with phonological ones.

2.1 Morphophonology of segments We shall discuss vocalic phenomena (diphthongization, metaphony, mid-high alternation) and consonantal phenomena (palatalization, velar insertion) conveying stem allomorphy. Diphthongization of Lat. Ĕ and Ŏ occurs in most of the Romance languages (cf. Sánchez Miret 1998, among others). Regarding standard Italian, Pirrelli (2000, 87–88) claims that diphthongization in inflectional paradigms is purely morphological since it is lexically restricted and its distribution within verb paradigms is not regular; e.g., the paradigm dolere ‘hurt’ (2a) exhibits diphthongization in 2SG and 3SG of the present indicative.6 This would be phonologically predictable, since it affects 5 This applies to almost all Romance languages and Latin with the exception of French where no theme vowels are found (except in -ir-verbs) and the Latin 3rd (or consonant) conjugation without a theme vowel: This class is still found in Catalan, Occitan, French, Italian, Sardinian, Romansh, and Romanian but not in Spanish and Portuguese where it has fused with the Latin 2nd conjugation (Lausberg 21972, 184–186). Moreover, French has lost many person/number suffixes which are often preserved in spelling though. 6 From an autonomous morphological perspective (cf. section 5), standard Italian exhibits a fusion of two morphomic patterns within some paradigms: Velar insertion (cf. below) in the 1 SG is a reflex of the so-called U-pattern and blocks diphthongization which adheres the N-pattern (cf. Maiden 2005, 162–163, for fusion of morphomic patterns in some Romance varieties; an Italian verb completely adhering to the N-pattern is morire, compare (2b)). Instead, Spanish shows diphthongization in the 1SG of doler, i.e. duelo. Hence, no fusion has occurred in this verb in Spanish (other verbs like venir ‘come’ show a fusion of the morphomic patterns for diphthongization, the N-pattern, and for velar insertion, the L-pattern).

152

Sascha Gaglia and Marc-Olivier Hinzelin

only stressed open syllables, which is confirmed by the paradigm of morire (2b), exhibiting diphthongization also in the 1SG . However, for muovere ‘move’ (2c), diphthongization is observable in the stem throughout the PRS . IND and, therefore, also in unstressed syllables of the 1PL and 2PL (muoviamo, muovete). In these cells, stress is arrhizotonic, i.e. it occurs to the right of the root. The verb trovare ‘find’ (2d), opposingly, does not exhibit diphthongs anywhere in its paradigm: (2)

It.

a.

dolgo, duoli, duole, dogliamo, dolete, dolgono

b.

muoio, muori, muore, moriamo, morite, muoiono

c.

muovo, muovi, muove, muoviamo, muovete, muovono

d.

trovo, trovi, trova, troviamo, trovate, trovano

Diphthongization is undoubtfully morphologized in Romance languages (cf. Sánchez Miret 1998, among others). Nonetheless, the phenomenon has been analysed by claiming a contribution of the phonological component (cf. Sluyters 1990; Vogel 1993), especially within the framework of Lexical Phonology (cf. Kiparsky 1982; 1985). These analyses refer to Italian diphthongization as lexically restricted, but (morpho)phonologically predictable “once it is established that a verb does exhibit this phenomenon” (Vogel 1993, 226).7 E.g., Sluyters (1990) treats the diphthongization of /ɛ/ and /ɔ/ in Standard Italian as a cyclic rule (cf. also Sluyters 1988 for his cyclic approach to metaphony below) and, therefore, as the result of a complex morphophonological process which depends on the stressed syllable of a binary foot. European Spanish shows the following patterns for diphthongization. (3)

Sp.

a.

pienso, piensas, piensa, pensamos, pensáis, piensan (INF pensar ‘think’)

b.

duermo, duermes, duerme, dormimos, dormís, duermen (INF dormir ‘sleep’)

Harris (1978) gives a rule-based, cyclic analysis settled within Lexical Phonology (cf. also his numerous works on this issue cited therein), where he analyses diphthongization and high-mid vowel raising (cf. below) in Spanish verbs assuming the underlying diacritics [D] for the former and [HM] for the latter. These phonological diacritics are associated with the corresponding stems (i.e. root + theme vowel) to allow the processes phonologically to occur depending on their morphological contexts. In Harris’s attempts to explain stem variation, surface forms are derived from one single underlying stem representation. 7 Vogel (1993, 226) does not give a detailed analysis of diphthongization with respect to the interaction of phonology and morphology. She refers to it to corroborate her claims about velar insertion (cf. below).

Inflectional verb morphology

153

Hooper (1976) deals with stem variation, including diphthongization, by ‘morphophonemic’ rules, i.e. alternating surface forms are built upon underlying representations which are already marked in the lexicon for the alternation, i.e. containing a diphthong and a monophthong. E.g., following Hooper (1976, 159), the Spanish verb mentir ‘lie’ has the underlying representation /m{je/e/i}nt-/. The surface representation is then determined by the given morphological contexts.8 Hooper (1976, 45) refuses the cyclic proposal by Harris (1969). She argues that surface forms which exhibit diphthongs also in an unstressed position, e.g. adiestramos9 (‘drill/train-1PL PRS . IND ’), cannot be analysed in cyclic terms. From our point of view and with respect to the highly idiosyncratic behaviour of the phenomenon, Pirrelli’s (2000) paradigmatic account of Italian mentioned above is a feasible way to handle diphthongization since it treats diphthongization as autonomously morphological (cf. section 5). Instead, interface accounts like Harris (1978) and (to a lesser degree) Hooper (1976) obscure the strong morpho-lexical character of the phenomenon by referring to disjunct morphological contexts serving as triggers for morphophonological/-phonemic diphthongization. Moreover, Harris’s (1978) account is problematic because of the stipulation of a huge battery of ordered rules for deriving diphthongization from one underlying representation (cf. also the criticisms in Bybee/Pardo 1981, 942 who doubt the rule-based character of morphophonological/-phonemic alternations).10 Metaphony generally means raising and/or diphthongization of stressed vowels, originally triggered phonologically (or phonetically) by word-final high vowels, i.e. -U and -I (cf. Maiden 1991, among others). Like diphthongization, the phenomenon is also widespread in Romance languages and all non-high vowels can undergo metaphony in principle; although the raising of stressed -A- is scarce compared to the metaphony of mid vowels. In verb paradigms, the 2SG PRS . IND is historically affected by - I .11 One of the most studied metaphonic systems is the Neapolitan type, exhibiting raising of high mid vowels (i.e. etymological -Ē -/-Ō -) and diphthongization of low mid vowels (-Ĕ -/-Ŏ -); cf. the following examples for the paradigms corresponding to Italian mettere ‘put’ (4a), pensare ‘think’ (4b), rompere ‘destroy/damage’ (4c), muovere ‘move’ (4d): 8 Instead, a ‘morphophonological’ rule (e.g. in the sense of Lexical Phonology, cf. above) derives a surface form from an underlying representation by manipulating the underlying segment phonologically. 9 The stressed syllable is underlined. 10 An interesting account on Italian diphthongization has more recently been presented by van der Veer (2006) within the framework of Optimality Theory (OT; ↗4 Phonology and morphology in Optimality Theory). 11 Etymologically, the 2SG inflectional ending ‑I was only present in Latin 3rd/4th conjugation class verbs. Its presence outside these classes in Italo-Romance can either be attributed to phonological change or to analogy (analogical spreading to the 1st conjugation class is undoubtful, in any case). An in-depth discussion is provided by Maiden (1996).

154

(4)

Sascha Gaglia and Marc-Olivier Hinzelin

Neapolitan Metaphony (2SG )

No metaphony (3SG )

Etymological stressed V

a.

[ˈmittə]

[ˈmettə]

-Ē -

b.

[ˈpjensə]

[ˈpɛnsə]

-Ĕ -

c.

[ˈrumbə]

[ˈrombə]

-Ō -

d.

[ˈmwovə]

[ˈmɔvə]

-Ŏ -

In modern Romance varieties, metaphony is morphologized (cf. Tuttle 1985; Maiden 1991; Fanciullo 1988; 1994, among others).12 However, this does not imply that phonology is not at issue ‒ as can easily be deduced from the literature: many investigations on Romance metaphony, especially those in a generative context, concentrate on the phonological component of the phenomenon, claiming the features [(±)high] or [(±)ATR ]13/[(±)tense], or a combination thereof, to trigger it (cf. Frigeni 2002; Calabrese 2011 and his earlier articles on the issue; Gaglia 2012, among others). For constraining the phonological rule, morphological conditions are stated, e.g. by Calabrese (1985), among others. With respect to the architecture of grammar, these studies refer more or less explicitly to metaphony as an interface phenomenon. Sluyters (1988), for example, claims a cyclic rule (cf. also Sluyter’s approach on dipthongization, discussed above), following the principles of Lexical Phonology (cf. Kiparsky 1982; 1985), where metaphony can only occur phonologically across morpheme boundaries, i.e. if the triggering high vowel is an inflectional suffix.14 Similarly, Gaglia (2012) analyses metaphony with respect to verbs in a Campanian variety. He claims that metaphony in verb paradigms is morphophonological because the phenomenon is phonologically predictable and metaphonic stems do not extend analogically, i.e. metaphony occurs only in 2SG .15 However, due to the high degree of exceptions within nouns and adjectives, metaphony cannot be analysed as being purely phonological. Because of lexical restrictions, the author assumes a lexically marked diacritic for paradigms where metaphony is instantiated. An interaction between morphology and phonetics is proposed by Maiden (1991) who defines the phonetic alternations to be exploited by morphology especially concerning so called ‘hypermetaphony’, i.e. the contrast of low-mid and high vowels. For a critical overview of the most prominent analyses on metaphony, cf. especially Calabrese (2011). 12 An exception is Logudorese Sardinian, where metaphony is still phonological according to Loporcaro (2011, 130). 13 ATR = advanced tongue root. 14 Fanciullo (1994) sees a rather direct connection between alternating phonemes and their morphological environment and adopts, hence, an approach involving not a morphophonological but rather a morphophonemic procedure for metaphony in nouns (for a definition of ‘morphophonological’ vs. ‘morphophonemic’ cf. above with respect to Hooper 1976 and note 8, above). 15 For nouns and adjectives in the given dialect, Gaglia (2009; 2012) shows that metaphony can also be the result of morpho-lexical selection.

Inflectional verb morphology

155

A minor phenomenon, but to which considerable attention has been paid, is what Harris (1978) terms high-mid alternation in Spanish (cf. also Hooper 1976; Bybee/Pardo 1981). The phenomenon consists of alternations of high mid and high vowels in 3rd conjugation stems, e.g. servimos ‘serve-1PL PRS . IND ’ vs. sirvamos ‘serve-1PL PRS . SBJV ’. Some verbs show a threeway alternation because of diphthongization, i.e. -i-, -e- and -ie- (hirvamos ‘boil-1PL PRS . SBJV ’, hervimos ‘boil-1PL PRS . IND ’, hiervo ‘boil-1SG PRS . IND ’) and -u-, -o- and -ue- (muramos ‘die-1PL PRS . SBJV ’, morimos ‘die-1PL PRS . IND ’, muero ‘die-1PL PRS . IND ’; cf. Harris 1978, 44 who counts approx. 25 verbs of this type). Harris explains high-mid alternation as the raising of the underlying high-mid vowel via rule (cf. Harris 1978, 45). But like diphthongization (cf. above), high-mid vowel raising cannot be explained in pure phonological terms. According to Harris, the lexical diacritic [HM ] (= ‘high-mid’) associated with the target vowel triggers the alternation for verbs of the 3rd conjugation (given as [3conj] in the conditioning context; cf. above for Harris’s account on diphthongization and Hooper 1976, whose diphthongization analysis is consistent with her analysis of high-mid alternation in the same book). Our criticism of Harris’s (1978) proposal has already been spelled out regarding diphthongization where we have objected to the number of ordered phonological rules which Harris stipulates. Moreover, by Harris’s postulation of disjunct morphological contexts, the morpho-lexical character of the phenomena investigated is obscured (cf. also Bybee/Pardo 1981 for criticisms). Romance palatalization is a well-studied phenomenon from a diachronic as well as a synchronic perspective (cf. Väänänen 1963, §§95–100; Dressler 1985, 168– 181; Krämer 2009, 56–84; Loporcaro 2011, 143–150; Maiden 2011, to name just a few).16 Regarding verb inflection in modern Italian, two types of palatalization were especially investigated within interface accounts, i.e. [ɡ]/[k] alternating with the palatal affricate [(d)ʤ]/[(t)ʧ ] (leggere ‘read’ (5a), indurre ‘induce’ (5b)) and [sk] alternating with [ ʃʃ ] in paradigms displaying the augment (e.g. conoscere ‘know’, 5c). In the 1st conjugation, palatalization never occurs (e.g. pagare ‘pay’, 5d):17

16 Historically, the segmental contexts for palatalization were [i], affecting [ɡ]/[k] in all Romance varieties but Sardinian, as well as [e] with the exception of Sardinian and Vegliote; the outcomes were [ts]/[s], [ʧ ]/[ʃ ], [dz]/[z], [ʤ]/[ʒ]. In Gallo- and Raeto-Romance [ɡ]/[k] were affected by [a] (outcomes: [ʧ ]/[ ʃ ], [ʤ/ʒ]). The glide [j] affected all consonants in all Romance areas, leading to various outcomes (cf. the overview in Vincent 1988a, 40 also for examples). 17 Palatalization is historically attributed to the Latin 3rd (Italian 2nd) and 4th (Italian 3rd) conjugation (cf. Dressler 1985, 176).

156

(5)

Sascha Gaglia and Marc-Olivier Hinzelin

It.

a.

le[ɡɡ]o, le[dʤ]i, le[dʤ]e, le[dʤ]amo, le[dʤ]ete, le[ɡɡ]ono

b.

indu[k]o, indu[ʧ ]i, indu[ʧ ]e, indu[ʧ ]amo indu[ʧ ]ete, indu[k]ono

c.

cono[sk]o, cono[ ʃʃ ]i, cono[ ʃʃ ]e, cono[ ʃʃ ]amo, cono[ ʃʃ ]ete, cono[sk]ono

d.

pa[ɡ]o, pa[ɡ]i, pa[ɡ]iamo, pa[ɡ]ate, pa[ɡ]ano

In the subparadigms of the present indicative in (5a–c), palatalization occurs before the front vowels -i- and -e- – with one exception, i.e. 1PL . Here, the palatalized consonant is followed by /a/ (le[dʤ]amo, cono[ ʃʃ ]amo). At first glance, a phonological account would not be implausible since the 1PL suffix can underlyingly be analysed as the glide-initial phonological string /-jamo/, triggering palatalization. However, a pure phonological account of palatalization is not appropriate due to the following reasons: Firstly, and already mentioned, palatalization does not occur in the 1st conjugation although a front vowel follows the velar /ɡ/ in 2SG and 1PL (pa[ɡ]i, pa[ɡ]iamo, not *pa[ʤ]i, *pa[ʤ]amo).18 Secondly, palatalization also occurs in past participles in -uto (e.g. cono[ ʃʃ ]uto ‘know-PTCP ’), even though /u/ is not a front vowel. This means, that the levelling of palatalized stems must have occurred within the verb system (cf. Dressler 1985, 177). Therefore, palatalization in verb paradigms is often described in purely morphological terms (cf. Pirrelli 2000; Carstairs-McCarthy 2010, 149; Maiden 2011).19 Cf. the discussion in section 5 with respect to the autonomy of morphology. Nonetheless, the interaction between morphology and phonology has been at issue including palatalization, especially in studies settled in the framework of Optimality Theory and regarding Italian (e.g. Burzio 2004; Krämer 2009; ↗4 Phonology and morphology in Optimality Theory).20 Burzio (2004) analyses stem alternations (cf. below for velar insertion) in Italian verb inflection within an extended version of OT combining morphological with phonological constraints. He assumes that the participle cre[ ʃʃ ]uto is chosen as the optimal candidate against cre[sk]uto by ranking the phonological constraint ‘PALATALIZE ’ over ‘O(utput)O(utput)-FAITH (FULNESS ) (Infinite)’ dominating ‘I(nput)O(utput)-FAITH (FULNESS )’. OO-FAITH militates in favor of paradigm uniformity and, therefore, stem identity with the Infinitive, i.e. cre[ ʃʃ-], whereas IO-FAITH requires the identity of the input /kresk-/ and the output, but it is

18 Further evidence against an analysis in pure phonological terms is provided by Italian nominal inflection: Palatalization occurs only before plural -i (and not before -e), e.g. ami[k]o/ami[ʧ]i ‘friend (s)’ vs. ami[k]e ‘female friends’ (*ami[ʧ]e). Furthermore, the phenomenon is lexically restricted, e.g. bel[ɡ]a/bel[ʤ]i ‘Belgian-SG / M . PL ’ but colle[ɡ]a/colle[ɡ]i ‘colleague-SG / M . PL ’ (*colle[ʤ]i). 19 From an autonomous morphological perspective, the non-palatalized verb forms of Italian adhere to the U -pattern (cf. 5a–c). In Portuguese, however, the palatalized verb forms constitute the morphomic (L -)pattern (as mentioned in note 16 above, palatalization affects also consonants beside [ɡ]/[k]; here it is [n] for the verb vir ‘come’): venho, vens, vem, vimos, vindes, vêm (PRS . IND ) vs. venha, venhas, venha, venhamos, venhais, venham (PRS . SBJV ). 20 An OT-account on palatalization and velar insertion in Catalan has recently been provided by Querol i Cortiella (2009).

Inflectional verb morphology

157

outranked. Grounded in the standard version of OT, Krämer (2009, 73–84) proposes a similar solution. However, Krämer’s account is slightly more phonological in nature since it makes reference to phonological place features, i.e. [coronal] and [dorsal]. A minor group of Spanish verbs displays the presence of [ɡ k] within the root in the 1SG PRS . IND, and throughout PRS . SBJV, e.g. venir ‘come’ (6a), conocer ‘know’ (6b). For convenience and according to most of the literature in the field, we refer to it as velar insertion. (6)

Sp.

a.

vengo, vienes, viene, venimos, venís, vienen (PRS . IND ) venga, vengas, venga, vengamos, vengáis, vengan (PRS . SBJV )

b.

conoz[k]o, conoces, conoce, conocemos, conocéis, conocen (PRS . IND ) conoz[k]a, conoz[k]as, conoz[k]a, conoz[k]amos, conoz[k]áis, conoz[k]an (PRS . SBJV )

For the corresponding Italian verbs venire and conoscere, velar insertion occurs in the 1SG and 3SG PRS . IND, as well as in 1SG /2SG /3SG /3PL PRS . SBJV: (7)

It.

a.

vengo, vieni, viene, veniamo, venite, vengono (PRS . IND ) venga, venga, venga, veniamo, veniate, vengano (PRS . SBJV )

b.

conos[k]o, cono[ ʃʃ ]i, cono[ ʃʃ ]e, cono[ ʃʃ ]amo, cono[ ʃʃ ]ete, conos[k]ono (PRS . IND ) conos[k]a, conos[k]a, conos[k]a, cono[ ʃʃ ]amo, cono[ ʃʃ ]ate, conos[k]ano (PRS . SBJV )

Both types of velar insertion ([ɡ k]) exhibit the same linear as well as paradigmatic distribution. Their paradigmatic distribution is the reason why both may be treated as morphologically related to each other in terms of the L-/U-pattern (cf. Maiden e.g. 2011; cf. the discussion in section 5). Nonetheless, they are two different phenomena from a diachronic point of view: Verbs presenting -sc- within the root as (6b, 7b), may be attributed to spreading from Latin inchoative verbs which exhibited the augment -sc- (cf. section 4.3; cf. Schwarze 1999, among others). The insertion of /ɡ/, however, is at least partly related to phonology.21 In old Spanish, velar insertion is possibly built on the model of the verb decir ‘say’, exhibiting the velar in the 1SG PRS . IND and the subjunctive forms, i.e. digo, diga, etc. (according to Maiden 2011, 236). Insertion spread throughout Spanish afterwards, generally from verbs 21 For a purely phonological account, cf. Vennemann (1988, 52–53) who treats velar insertion as a syllable contact phenomenon where the glide [j] is strengthened, becoming the voiced velar plosive.

158

Sascha Gaglia and Marc-Olivier Hinzelin

with stem final /n/ (e.g. venir) into paradigms with stem final sonorants, e.g. doler ‘hurt’, salir ‘go out’, etc.). Moreover, the velar /ɡ/ was introduced in paradigms with a root final yod (< *GJ, *DJ ) in the 1SG , e.g. *AUDJO > oyo > oigo ‘hear-1SG ’, and throughout the present subjunctive forms (cf. Maiden 2011, 237). Similarly in Italian, the velar replaced regular alternants already created by yod, e.g. UENIO > old Tuscan vegno > Italian vengo (cf. Maiden 2011, 238; for a detailed diachronic analysis of velar insertion in Italian cf. also Spina 2007, 112–139). From a synchronic point of view, velar insertion with /ɡ/ requires as environment a following back vowel, i.e. /a/ or /o/ and a preceding coronal sonorant, i.e. /n l/ except /r/. But since it is not found before participles with the back vowel /u/, e.g. Italian venuto ‘come-PTCP ’ (*venguto; the same is true regarding insertion of /k/, e.g. It. cre[ ʃʃ ]uto ‘grow-PTCP ’, not *cre[sk]uto; cf. above), a pure phonological treatment (i.e. velar insertion before back vowels [u o a]) is not appropriate.22 Moreover, velar insertion is only encountered in a small group of verbs, as mentioned above, and never in the 1st conjugation, e.g. Italian cenare ‘have dinner’ completely lacks velar insertion in the given cells, i.e. ceno (*cengo 1SG ), cenano (*cengano 3PL).23 An account within Lexical Phonology would be the aforementioned study by Vogel (1993, 225–226). She claims that the derivation of non-velar stems occurs via deletion of the underlying velar in all environments except before ‑a and ‑o (for this reason, the velar is not inserted but is lexically present in Vogel’s account). We agree with Pirrelli (2000, 85) who plausibly objects that Vogel’s proposal is problematic because it does not only require deletion of the underlying velar before front vowels but an additional palatalization rule is required for verb stems of the type [tolɡ-] (tolgo ‘remove-1SG ’), [kolɡ-] (colgo ‘collect-1SG ’) deriving /l/ to [ʎʎ] before /i/ (togli ‘remove-2SG ’, cogli ‘collect-2SG ’).24 In his extended OT-approach already presented here with respect to palatalization, Burzio (2004) (cf. above) includes interaction with the phonological component by the implementation of a markedness constraint, expressed by a rule of entailment, which enables velar insertion to occur before back vowels, i.e. ‘-ɡ- ⇒ / _ [+back]’. But since the phenomenon cannot be analysed in pure phonological terms, as mentioned above, the phonological constraint is ranked in between 22 This is also confirmed by Bybee/Pardo (1981, 957) who conducted nonce-verb tests with speakers of Spanish (cf. also section 4.2). A statistical tendency of realizing velars before back vowels could not be found by the authors. 23 The subjunctive forms of cenare are neither affected by velar insertion because the suffixes of 1st conjugation class verbs only exhibit the initial front vowel in all cells, i.e. ‑i (ceni, ceni, ceni, ceniamo, ceniate, cenino). However, if velar insertion is purely phonological, then one would expect the theme vowel -a (a back vowel) to trigger velar insertion in 2PL present indicative as well – but this is not the case (cenate, *cengate). 24 A further derivational approach is Fanciullo (1998) who treats the insertion of /ɡ/ as velar ‘infix’ before back vowels (according to Malkiel 1974, 307) causing depalatalization of underlying /ʎ/ towards [l]. We refer to Pirrelli (2000, 181) and Maiden (2011, 241) for some important objections against Fanciullo’s account.

Inflectional verb morphology

159

two faithfulness constraints, i.e. ‘O(utput)O(utput)-FAITH (FULNESS ) (1,6 Indic.)’ and ‘I(nput)O(utput)-FAITH ( FULNESS ) ’. The former militates in favour of an output identity of the stems in 1SG and 3PL while the latter favors the identity of input and output.

2.2 Stress The following ways to account for main stress in Romance can usually be found: (i) phonological rules, (ii) (morpho-)lexical specification, and (iii) mapping morphological onto prosodic/phonological structures. According to this typology, the present section deals with (ii) and (iii) due to their involvement of interfaces. Main stress in classical Latin was prosodically predictable, i.e. it fell regularly on the heavy penult. If the penult was light, the antepenult had to be stressed. The modern Romance languages are, to a high degree (with some exceptions), still faithful to this so-called three syllable window but with the collapse of the quantitative vowel system in the evolution from classical Latin to proto-Romance, stress was no longer predictable from a phonological perspective (cf. Roca 1999 for a detailed overview regarding Romance languages as well as Vincent 1988b and Saltarelli 2000 for the development from Latin).25 This becomes very clear if we consider the following verb forms from standard Italian where stress placement varies from ultima even to preantepenultima: (8)

It.

a.

[mastiˈkɔ]

‘chew-3SG PRET. IND ’

(ult)

b.

[mastiˈkaːte]

‘chew-2PL PRS . IND ’

(penult)

c.

[ˈmastiko]

‘chew-1SG PRS . IND ’

(antepenult)

d.

[ˈmastikano]

‘chew-3PL PRS . IND ’

(preantepenult)

Hooper (1976, 23–31) suggests an analysis within Natural Generative Phonology which treats stress in Spanish verbs as morphologized (by referring to Harris 1969 who gives a purely phonological account). The rules governing stress assignment in Spanish verbs are directly associated with tense and with their relative positions to the stem in each subparadigm. Sluyters (1990) proposes a cyclic rule to account for Italian verb stress (cf. section 2.1 for Sluyter’s approaches to metaphony and diphthongization). E.g. in arrhizotonic verbs, stress is assigned to the last vowel of a derived verb stem, i.e. the theme vowel as for the imperfect forms parlavo (‘I spoke’), vendevo (‘I sold’),

25 Modern French is an exception since stress falls on the rightmost syllable of a rhythmic group and is, therefore, predictable. Individually, modern French may exhibit word-final schwa. However, schwa cannot be stressed.

160

Sascha Gaglia and Marc-Olivier Hinzelin

sentivo (‘I heard’), and in the corresponding preterite forms, e.g. parlai, vendei, sentii.26 Roca (1999) proposes the following, extraordinarily fine-grained account for Romance verb stress also making reference to many language-specific exceptions and depending on metrical parameters (extrametricality, syllable weight, binary foot construction, line conflation), as well as on morphological categories: (a) theme vowels are stressed for non-present tenses (‘Romance Verb Stress Rule’); (b) for the present tense, the singular forms and 3PL-form must be derived by the same stress algorithm as nouns, (c) in specific tenses, 1PL and 2PL must be stressed by a rule which assigns stress to the right of the theme vowel, e.g. imperfect indicative cantavamo, cantavate (‘1PL /2PL Accent Rule’), (d) for the future, the first vowel of the TMA-morpheme, specified with the lexical morphosemantic feature [+FUTURE ], is stressed (‘Future Accent Rule’ which also covers the conditional). Within the framework of Distributed Morphology (DM), Oltra-Massuet/Arregi (2005) as well as Pomino (2008a; 2008b) derive stress in Spanish from functional syntactic heads which determine word structure and supply a metrical grid.27 In these accounts, stress is placed on the rightmost foot preceding the functional head ‘T’ (Tense). Where stress is further to the left with respect to ‘T’, a stress deletion rule is assumed to be responsible for retraction: e.g., partes ‘leave-2SG ’, since stress would otherwise be placed on the theme vowel by the stress algorithm, i.e. *partés (cf. Oltra-Massuet/Arregi 2005, 61). Central to the analyses mentioned here is the claim that each functional head (besides Agr) exhibits a position for a theme vowel. Evidence is derived by the authors from verbs with multiple theme vowels, e.g. Sp. cant-á-b-a-mos ‘sing-1PL IMPF. IND ’.28 From our perspective, this assumption leads to an undesirable consequence: e.g. cf. the conditional, where three theme vowels must by assumed (adapted according to the verb structure proposed by Oltra-Massuet/Arregi 2005, 54). Moreover, the functional head F (Future) is stipulated in addition to T:29 (9)

Sp.

√ cant

v

Th a

F r

Th í

T Ø

Th a

Agr mos

26 Sluyters (1990, 79) includes also the V-final stressed future forms to be captured by the rule. However, this is doubtful because one is forced to assume an underlying stem final empty vowel which has to be deleted in order to assign stress to the inflectional suffix (e.g. /venderV-/ + /ɔ/ becoming [vendeˈrɔ] venderò ‘I will sell’). 27 With respect to stress in Catalan cf. especially Oltra-Massuet (1999). 28 Moreover, Oltra-Massuet/Arregi (2005, 55) assume Fusion of T/Agr if “T is [–Pst]”. For clarification it must be said that stems are not assumed in DM, only roots. Stem alternations are, thus, always a consequence of morphophonological readjustment rules. 29 The functional head ‘v’ is especially relevant for (in)transitivity and case.

Inflectional verb morphology

161

Meinschaefer (2011) combines a constraint-based (OT-)analysis incorporating McCarthy’s notion of “optimal paradigms” (cf. McCarthy 2005) in mapping the morphological onto a prosodic structure. She claims that stress in Spanish present and imperfect tenses is purely phonological. For Italian, where stress is phonologically irregular, she integrates constraints which make use of morphological notions and attributes a special role to the theme vowel. E.g. for arrhizotonic stress in 1PL /2PL PRS . IND she proposes a constraint which aligns “the left edge of every theme vowel [. . .] with the left edge of the head of the foot” (cf. Meinschaefer 2011, 59). This constraint dominates the purely prosodic constraints in the grammar. According to McCarthy (2005), she introduces a constraint for the 3PL , being identically stressed as the singular forms, militating in favor of paradigm uniformity, i.e. an optimal paradigm constraint (‘OPTIMAL-PARADIGM LEFT-ANCHOR-FOOT ’; cf. Meinschaefer 2011, 61). This constraint is also proposed by Meinschaefer with respect to the Spanish preterite, which cannot be purely analysed in prosodic terms.30

2.3 Phoneme loss and competition – morphological repercussions The origin of many syncretisms lies in sound change, e.g. the loss of consonants realizing person/number suffixes. This change is well-documented in the development from Latin to Ibero-Romance as the source of the 1SG =3SG syncretism observed in many tenses, e.g. (10) in Spanish. But once the sound change is established, phonology no longer plays a role in the functioning of syncretism, its interaction at other interfaces is described below (cf. sections 3.2 and 4.4). (10) Lat. CANTĀBAM /-ĀBAT > Sp. cantaba = cantaba ‘sing-1SG =3SG IMPF. IND ’ Suppletion may arise due to the lack of sufficient phonological substance after sound change as has been claimed for Latin īre (Darmesteter 1887, 162; Markun 1932, 89; FEW s.v. īre; Aski 1995, 408–409, 413): Already in Latin, it exhibits an extremely short one-phoneme stem (consisting of the root i- fused with the theme vowel -i-) with the monosyllabic forms īs and it – disyllabic eō and eunt would have become monosyllabic as well in Romance (Maiden 2004, 250 note 12). Most Romance languages have resorted to forms from other verbs in parts of the paradigm, most

30 Cf. also Krämer (2009, 188) for a constraint-based approach to Italian verb stress. The author claims default stress in the penultimate syllable, e.g. [kaˈde:ɾe] cadere ʻto fallʼ which is overridden by lexical stress on a given morpheme if the stress is not in the penultimate, e.g. [ˈɾideɾe] ridere ʻto laughʼ. The fact that the 3 PL present indicative is never stressed by default but always on the first root vowel is treated as a “paradigm effect” by Krämer (2009, 189).

162

Sascha Gaglia and Marc-Olivier Hinzelin

notably in the singular and 3PL present indicative, but sometimes (nearly) ousting the i-stem31 completely (cf. section 4.1). An uncertainty of the form to be produced by morphophonological rules may lead to defectiveness where competing morphological processes involving morphophonological rules exist, but also to overabundance by generating multiple forms for one paradigm cell (cf. section 3.3.3). Albright (2003, 13) discusses data from a production experiment on defectiveness in Spanish verbs and concludes: The overall picture that emerges is that the gaps that are listed in grammars lie at just one extreme of a gradient range of uncertainty that speakers feel when deciding whether or not to apply morphophonological alternations. This uncertainty is strongest when two factors collide: first, the word must be relatively infrequent or unfamiliar, so that the speaker is forced to synthesize a form. In addition, the lexicon must contain conflicting evidence about whether or not the alternation should apply.

Maiden/O’Neill (2010, 121), however, argue against the requirement of a morpho(phono)logical competition: Our interpretation of the Spanish and Portuguese data is that the primary motivation for defectiveness in them is likely to be preservation of the unique stem present in the etymological input of what are overwhelmingly loan words, for the most part restricted to learnèd vocabulary, to educated speakers and to elevated registers.

3 Interfaces with syntax This section describes interfaces with the syntax module, discusses a number of selected phenomena from the Romance languages, and highlights a sample of current theoretical analyses of these. In contrast to the preceding section on phonology interfaces, the interface with syntax has received considerably less attention which may partly be due to the practical problem of delimiting the two modules. As a sharp distinction between the two modules is notoriously difficult to draw, the question of the significance and the locus of the interface between morphology and syntax is one of the most challenging to address. Morpheme- vs. process- vs. wordbased models of morphology have been proposed (cf. Hockett 1954). Moreover, in some linguistic theories, morphology has even been denied an independent status from syntax and is subsumed under the more general and collective label ‘morphosyntax’ or it may be actually regarded as syntax below the word level, notably in many generative approaches (cf. the discussion in Hinzelin/Gaglia 2012, 2–3). From 31 In the French verb aller, the i-stem survives only in the future and conditional ir-stem, and in Catalan anar only as part of a hybrid anir-stem (cf. examples in section 4.1). Modern Italian andare shows no trace of the i-stem, old Tuscan gire has the typical present-indicative distribution of the two stems (Maiden 2004, 232).

Inflectional verb morphology

163

this perspective, linguistic processes commonly attributed to the morphology module are conceived as syntactic in nature and explained by genuinely syntactic operations like movement (e.g. affix hopping in verb inflection, cf. Chomsky 1957, 38–42, 113, here still named “Auxiliary Transformation”) or integrated into general phrase structure in the case of word formation (e.g. analysing the process of affixation as phrase structural, cf. Williams 1981, 246, or analysing compounds as syntactic phrases, cf., among others, Selkirk 1982; Lieber 1983; 1992). The Distributed Morphology approach is deeply rooted in this perspective on morphological phenomena.32 In these theories, there cannot exist a special interface with syntax as there is no morphology module separate from syntax to begin with (cf. Carstairs-McCarthy 2010 and Hinzelin/Gaglia 2012 for a critical discussion of this basic issue). Instead of this syntactic and morpheme-based conception, Anderson (1982), Aronoff (1994) and, more recently, Stewart/Stump (2007, 384) argue for a word-based conception where words are syntactic atoms and their internal structure is unavailable to syntax. We tend to follow the latter view for the conception of the interface.

3.1 Alignment of morphosyntactic properties Morphosyntactic properties are aligned with inflectional affixes: Person/number affixes have a direct correlate in syntax as they are morphological entities expressing a syntactic relation, i.e. usually the agreement between subject and verb as in (11). The subject-verb agreement in the Spanish examples is expressed by the absence of person/number marking (or a null morpheme) in the 3SG or the presence of the morpheme -n in the 3PL . (11)

Sp.

a.

El hombre canta.

‘the man sing-3SG ’

b.

Los hombres cantan.

‘the men sing-3PL ’

This correlation is one-to-one and direct, but need not always be the case as syncretism (cf. section 3.2) or fusional morphemes suggest as in (12). In cantó, the person/number morpheme is fused with the tense-mood morpheme, i.e. the portmanteau morpheme -ó expresses at the same time both the properties 3SG and preterite indicative. (12)

Sp.

a.

El hombre cantó.

‘the man sing-3SG PRET. IND ’

b.

Los hombres cantaron.

‘the men sing-3PL PRET. IND ’

32 Halle/Marantz (1993, 111–112) state: “We have called our approach Distributed Morphology (hereafter DM) to highlight the fact that the machinery of what traditionally has been called morphology is not concentrated in a single component of the grammar, but rather is distributed among several different components.”

164

Sascha Gaglia and Marc-Olivier Hinzelin

Morphosyntactic properties are aligned with stem allomorphs: Some morphosyntactic features seem to be aligned with a special stem allomorph in many Romance languages (for a discussion of a possible semantic explanation, cf. section 4.2).33 In Spanish and Portuguese irregular verbs, there is a special stem allomorph for the preterite (the Spanish indefinido) in all cells of this partial paradigm (among others; for the distribution, cf. Maiden 2001a; 2005). (13)

Sp.

hacer ‘make’ (preterite stem: hic-/hiz-) PRET. IND :

hice, hiciste, hizo, hicimos, hicisteis, hicieron

Nevertheless, this phenomenon (and many others) are better understood by taking the paradigm as a whole into account as the respective stem allomorphs show up in other cells of the paradigm as well (cf. section 2 for morphophonological processes behind stem allomorphy). In the case of the preterite allomorph, the past subjunctive, the pluperfect (in Portuguese only) and the future subjunctive (which is marginal in modern Spanish) are involved, displaying a typical distribution referred to as PYTA (perfecto y tiempos afines; cf. Maiden 2001a; 2005). As there is no commonly shared syntactic or semantic feature involved, the stem distribution may be analysed as being morphomic (in the sense of Aronoff 1994), i.e. autonomously morphological in nature (cf. also section 5; for a detailed discussion, cf. Maiden 2005).

3.2 Spread of syncretism and syntactic repercussions Syncretism is a prototypical example of a morphological phenomenon at the interface with syntax as suggested by the title of Baerman/Brown/Corbett’s (2005) seminal book The syntax-morphology interface: A study of syncretism. While rare in Latin verb paradigms, syncretism is commonplace in most Romance languages. Moderate syncretism is found in Ibero-Romance where 1SG and 3SG are usually syncretic in many tense/mood combinations throughout the paradigm, thus reducing the distinctions in form to five (its origin is discussed in section 2.3, its semantic/pragmatic implications in section 4.4): (14)

Sp.

hablaba

‘speak-1SG =3SG IMPF. IND ’

In some Gallo-Romance varieties (all Oïl dialects and Francoprovençal as well as some northern Occitan and some northern Italian varieties), syncretism is systemstructuring (Hinzelin 2012), often reducing the form distinctions to three, e.g. in the present indicative of French first-conjugation verbs: 33 Bybee (1985, 93) claims that inflectional splits may generally occur due to semantic reasons, i.e. along splits pertaining to aspect and, to a lesser extent, tense and mood (cf. the discussion in 4.2).

Inflectional verb morphology

(15)

Fr.

[paʁl]

‘speak-1SG =2SG =3SG =3PL ’;

[paʁˈlõ]

‘speak-1PL ’

[paʁˈle]

‘speak-2PL ’

165

Extensive syncretism, as found in Gallo-Romance verb paradigms, generally correlates with a frequent or obligatory use of subject pronouns. The use of clitic pronouns is obligatory in standard French as well as in Oïl, many Francoprovençal, and most northern Italian dialects (sometimes limited to a subset of grammatical persons, a phenomenon known as partial pro-drop, cf. Hinzelin/Kaiser 2012 and Heap 2000; 2002 – here ambiguously termed “split subject pronoun paradigms”). A frequent use of subject pronouns is also encountered in Caribbean Spanish and Brazilian Portuguese where sound change34 and/or the loss of dedicated forms of the paradigm35 have led to a reduced paradigm with respect to distinctive morphological markers. The loss of inflectional morphology frequently entails syntactic change to compensate for this lack of distinction by a syntactic element, a subject pronoun in this case.36 Person/number distinctions are thus expressed primarily by syntactic means.37 In the course of further language change, this pronoun is often reduced to a clitic and may be analysed even as an inflectional prefix in colloquial French (cf. e.g. Roberge 1990, 168–169; Auger 1993; Kaiser 2008; Culbertson 2010) and some northern Italian dialects (Brandi/Cordin 1989; Poletto 2000; Manzini/Savoia 2005), thereby completing the grammaticalization scale (cf. Lehmann 1985, 47; 1995, 39–42, 55; Meisenburg 2000). For the prominent position of clitics at the interface and the affix-vs.-clitic debate, ↗12 Object clitics. The loss of subject inversion and strict SVO-word order, which are also found in Brazilian Portuguese and Caribbean Spanish, are further syntactic changes which may be attributed, at least to some extent, to syncretism (Morales 1989; Hinzelin 2012).

34 In Andalusian and Caribbean Spanish, -s in coda position has been lost in most varieties leading to a syncretism of 2SG -forms (with the -s as 2SG - morpheme) with the 3SG . 35 E.g. the loss of dedicated verb forms for the 2SG , 1PL , and 2PL in Brazilian Portuguese, and for the 2PL in American Spanish in general (use of ustedes ‘you-PL (formal)’ with 3PL -agreement instead of vosotros ‘you-PL ’). In colloquial French, the expression of 1PL by the combination of the pronoun nous ‘we’ with a 1 PL -verb form, e.g. nous parlons, is supplanted by a construction employing the indefinite pronoun on ‘(some)one’ with 3SG -agreement, e.g. on parle, thus further reducing form distinctions to two. 36 A parallel development in nominal morphology can be found in the French determiner system: The (clitic) article has become obligatory in modern French as gender and number marking of consonant-initial nouns is reliably expressed by the article only. 37 This, as an anonymous reviewer points out, “can also be seen in the Italian subjunctive, where in a pronounless phrase (e.g. voglio che venga) the verb is interpreted by default as a 3SG , and the pronoun is obligatory in order to obtain another (e.g. 2SG ) interpretation (voglio che tu venga)”.

166

Sascha Gaglia and Marc-Olivier Hinzelin

3.3 Periphrases Grammatical functions may be expressed in two ways in languages: as a bound morpheme directly on the verb (i.e. synthetically) or through a separate word like an auxiliary verb (i.e. analytically/periphrastically). These analytic forms (or periphrastic constructions) are at the morphology-syntax interface as they express morphological properties by syntactic means (cf. Brown et al. 2012; Chumakina 2013; Sadler/ Spencer 2001). There is a long debate whether periphrastic forms are actually part of the morphological paradigm of a verb (Sadler/Spencer 2001; Ackerman/Stump 2004) or rather independent syntactic entities, interpreted compositionally, and with only a loose connection to the morphological domain inside morphosyntax, if any (this is usually the generative approach, cf. e.g. Giorgi/Pianesi 1997, 37–40). Again, this decision is theory-dependent and clear definitions of key terms like ‘morphological paradigm’ are needed. If the paradigm is “the set of all elements filling the cells defined by the inflectional categories that can be expressed for the lexeme” (Haspelmath 2000, 663), then “periphrastic forms are admitted as members of the paradigm, but at the price that the paradigm is no longer a purely morphological notion” (Haspelmath 2000, 663). Haspelmath (2000) distinguishes two kinds of periphrasis, suppletive and categorial. Whereas a suppletive periphrasis supplies forms that are deviating from the expected ones (hence its name and its connection to suppletion; cf. sections 2.3 and 4.1) and filling a gap38 (compare, e.g., the few analytic passive forms in Latin discussed below), a categorial periphrasis does not have this function but it simply “expresses a grammatical meaning in a multi-word construction” (Haspelmath 2000, 660). An example for this category are the analytic future forms discussed in section 3.3.1. Thus, the decision on the interpretation of periphrases depends to a large extent on the periphrasis under scrutiny: There is a gradual scale, on the one hand, with the combination of two largely independent verbs both contributing to the lexical meaning of the expression as a whole, and, on the other, highly grammaticalized periphrases where the participating inflected verb forms have lost all or most of their original meaning. To reach a decision, Cruschina (2013, 282–283) proposes that: [. . .] the degree of grammaticalization [. . .] must be considered as the main criterion for defining a morphological periphrasis, irrespective of whether synthesis competes with periphrasis as a mode of inflectional exponence. The more an analytic construction is grammaticalized, the more it becomes likely that it will be treated morphologically and will be subject to morphological rules and patterns of irregularity.

Olbertz (1998) divides Spanish constructions generally identified as periphrases into lexical, semi-auxiliary, and ‘true’ periphrastic constructions. The periphrases consist 38 Lexeme-specific gaps where no generally established periphrasis exists are labelled ‘defectiveness’ (cf. section 3.3.3).

Inflectional verb morphology

167

of the combination of two or more words, usually an auxiliary or modal verb, a preposition (optional), and a non-finite form of the main verb (infinitive, participle or gerund). Latin is a highly synthetic language (e.g. in perfect and future tense as well as in the passive; three different kinds of infinitives and participles are available), only very few forms are analytic. E.g. complex synthetic passive forms exist as in (16). But for the passives of the indicative and subjunctive of perfect (17), pluperfect, and future perfect, no synthetic forms are available (cf. Sadler/Spencer 2001; Vincent 2011). Romance languages use analytic structures for the passive in general (cf. section 3.3.2) as well as in many other tense/mood combinations, sometimes in competition with synthetic structures as in French and Spanish future tense (cf. section 3.3.1). (16)

Lat.

amābātur

‘love-3SG . PASS PST. IPFV. IND ’

(17)

Lat.

amātus est

‘love-PTCP + AUX-3SG PRS . IND ’ → 3SG PST. PFV PASS

3.3.1 Tense/mood/aspect: New Romance tenses The expression of tense, mood, and aspect (TMA) is situated at the interface of three modules, i.e. morphology with syntax and semantics. These concepts may be expressed morphologically by affixes or syntactically by the combination of words. Languages tend to make use of both devices for the expression of TMA-distinctions but often with a clear inclination for the one or the other. Latin TMA-verb morphology is predominantly synthetic in nature:39 (18)

Lat.

a.

amāverat ‘love-3SG PLPF. IND ’ (pluperfect)

b.

amāvisset ‘love-3SG PLPF. SBJV ’ (pluperfect subjunctive)

c.

amāverit ‘love-3SG FUT. PFV. IND ’ (future perfect)

As illustrated in the following sections, the Romance languages have replaced many Latin synthetic forms by periphrastic ones (Lausberg 21972, 13). Language change along the grammaticalization cline may lead to new synthetic forms, thus completing the cycle of periphrastic and synthetic forms (cf. below for future tense). Furthermore, there are numerous periphrases expressing different aspects as inceptive, progressive, etc. or tenses like recent past etc. (cf. Squartini 1998 for Romance in general; Olbertz 1998 for Spanish; Lehmann 1995, 133 for a discussion of the grammaticalization of periphrases in Portuguese). The lexical material used in these more syntactic periphrases differs considerably from language to language. 39 The diachronic origin of the formation of amāverat and amāverit is still quite transparent: the perfect stem takes inflected forms of the auxiliary ‘be’ as desinence. This development is similar to the evolution of the Romance synthetic future.

168

Sascha Gaglia and Marc-Olivier Hinzelin

The early Romance analytic future becomes synthetic. Latin uses a synthetic future: (19)

amābit ‘love-3SG FUT. IND ’

Lat.

This form has been lost in all Romance languages. In most Romance languages, a new synthetic future has evolved out of a vulgar Latin (and early Romance) periphrastic construction (Lausberg 2 1972, 228–231; Maiden 2011, 264–266):40 (20)

+ HABET ‘love-INF + have-3SG PRS . IND ’ > VLat. *amar-at > Sp./Pt. amará, Fr. aimera, Cat. amarà, It. amerà ‘love-3SG FUT. IND ’

Lat.

AMĀRE

This development has been used as a textbook example for grammaticalization in general (Hopper/Traugott 2003, 52–55; cf. also Roberts 1992). Modern analytic future formations are another way of expressing future tense, found in most Romance languages. This periphrastic future combines the motion verb ‘go’ with the infinitive in French and Portuguese or the preposition ‘to’ and the infinitive in Spanish: (21)

a.

Fr.

(il) va aimer ‘(he) go-3SG PRS . IND + love-INF ’

b.

Pt.

vou amar ‘go-3SG PRS . IND + love-INF ’

c.

Sp.

va a amar ‘go-3SG PRS . IND + P ‘to’ + love-INF ’

In Romanian, a further construction is used: a voi ‘want’ + INF (Lausberg 21972, 228; Iliescu/Popovici 2013, 236): (22)

Rom.

va cânta

or

o cânta

‘want-3SG PRS . IND + sing-INF ’

In other Romance languages, different constructions are encountered: dévere ‘must’ + a ‘to’ (P) + INF alongside àere ‘have’ + INF in Sardinian (Lausberg 21972, 228–229; Remberger 2006, 252–259), ‘come’ + P ‘to’ + INF in Sursilvan (Lausberg 21972, 228), and the present indicative with the adverb pouë ‘then’ in Valdôtain and Savoyard Francoprovençal (Hinzelin 2011b, 40–41). Another periphrastic tense is the perfect, formed with the auxiliary ‘have’ and the past participle; mainly with intransitive and reflexive verbs, French and Italian use ‘be’ + PTCP ( Vincent 1982). The use of the past participle in these constructions brings about a voice switch – the Latin past participle was passive (cf. above) – but the tense/aspect value is retained (Vincent 2011, 430).

40 In Romanian the periphrastic construction ‘want’ + INF is used (cf. below). Some southern Italian dialects have no dedicated future forms and use the present indicative to express future tense as well (Lausberg 21972, 227).

Inflectional verb morphology

(23)

Fr.

(il) a aimé

169

‘(he) have-3SG PRS . IND + love-PTCP ’

Whereas in most Romance languages a construction with the verb ‘go’ and the infinitive is used to refer to an event in the future (cf. above), in Catalan a periphrastic preterite has been coined using the same construction (cf. Squartini 1998, 189–192; Taylor 2011):41 (24)

va amar ‘go-3SG PRS . IND + love-INF ’

Cat.

In Sicilian, a similar construction expressing emphatic past but with both verbs inflected for present tense exists (Cruschina 2013, 279). The pluperfect is synthetic only in Portuguese (Spanish retains the forms but they now express the past subjunctive; old French used the forms for the perfect; cf. Lausberg 21972, 219–220). The other Romance languages use a periphrasis with the auxiliary ‘have’ in the imperfect and the past participle.

3.3.2 Voice The Latin synthetic passive (cf. section 3.3) has not survived in any Romance language, passive voice is expressed by periphrastic constructions instead: The construction with the reflexive clitic pronoun se/si can be used in the 3rd person with inanimate subjects (cf. Lausberg 21972, 239; Remberger 2006, 171–172; compare (25a)) or the auxiliary ‘be’ is combined with the past participle (cf. Lausberg 21972, 239–242; Remberger 2006, 159–167; Vincent 1982; compare (25b, 26a, b)). This last construction already existed in Latin (cf. section 3.3), but had a perfective interpretation. Romance passives are now ambiguous except in Spanish and Portuguese which express the difference by the choice of the auxiliary: ser denotes the dynamic passive (process), a state is expressed with estar (Lausberg 21972, 242; compare (27a, b)). Italian may distinguish the dynamic passive by the choice of the auxiliary venire ‘come’ (Lausberg 21972, 242; Remberger 2006, 185–187). Romansh uses only the combination of ‘come’ and the PTCP to express the passive (Lausberg 21972, 242). (25)

(26)

It.

a.

Questa macchina si costruisce in Italia. ‘this car REFL =construct-3SG PRS . IND in Italy’

b.

Questa macchina è costruita in Italia. ‘this car be-3SG PRS . IND construct-PTCP in Italy’

a.

Sp.

es amado ‘be-3SG PRS . IND + love-PTCP ’

b.

Fr.

(il) est aimé ‘(he) be-3SG PRS . IND + love-PTCP ’

41 This construction has replaced the synthetic one (e.g. amà) in spoken standard Catalan.

170

(27)

Sascha Gaglia and Marc-Olivier Hinzelin

Sp.

a.

la puerta es cerrada ‘the door be-3SG PRS . IND + close-PTCP ’

b.

la puerta está cerrada ‘the door be-3SG PRS . IND + close-PTCP ’

3.3.3 Defectiveness and overabundance in a syntactic perspective In defective paradigms, there are gaps for some paradigmatic cells. The missing forms may be substituted by a form of a modal verb and a non-finite form of the original verb, by a semantically related verb or a periphrasis. The periphrases used to express the meaning of the missing forms are occasional formations and cannot be grouped with other regular periphrastic forms of the verb discussed in section 3.3. Here, syntax may be used to fill an idiosyncratic gap in the paradigm of a verb which can often be attributed to morphophonological uncertainties (cf. section 2.3). (28)

Sp.

abolir ‘abolish’: 1SG —— (defective) → quiero abolir ‘want-1SG abolish-INF ’

Overabundance caused by the syntax exists if a language leaves a choice between a morphological (synthetic) and a syntactic (periphrastic/analytic) formation, as it is the case for the future tense in French, Spanish, and Portuguese (cf. section 3.3.1).

4 Interfaces with semantics and pragmatics In this section, we shall first focus on the morphophonological shape as well as on the distribution of stems and roots which can be associated to meaning (mainly with respect to person/number and to a lesser degree to tense; cf. sections 4.1–4.4). We reconsider certain observations already discussed in section 2. For a perspective on the morphology-semantics interface that considers tense/mood/aspect from the angle of formal semantics and logic, we refer to Becker (2010; 2013) among others (cf. more references therein) as we shall not address these issues in what follows. Regarding the morphology-pragmatics interface, we shall approach a phenomenon also related to person and number (cf. section 4.5).

4.1 Merger of different verbs as source of suppletion Verbs sharing some part of their meaning may merge into a single verb as evidenced by the origin of forms found in the verbs ‘go’ and ‘be’ in a number of Romance languages (29) (cf. Aski 1995; Hinzelin 2011a; Maiden 2004; Rudes 1980). This merger

Inflectional verb morphology

171

automatically leads to suppletion: the morphological paradigm displays stems coming from etymologically different verbs.42 In a next step of this fusion, even hybrid or intermediate forms may develop, blending the two stems to form a new one (30) (cf. Hinzelin 2011a). As a result, two distinct entities of the lexicon with partly overlapping semantics are united into a single new lexeme with completely unrelated stem alternants. The distribution of forms interacts here with lexical meaning, semantic and pragmatic principles like synonymy avoidance (cf. Maiden 2004, 228) or, more specifically, Clark’s Principle of Contrast (cf. Clark 1987; 1990).43 In many Romance languages, the paradigm of the verb ‘go’ integrates stems of two Latin verbs, ĪRE ‘go’ and VĀDERE ‘go (hastily)’, in Gallo-Romance and Catalan, additionally, AMBULĀRE ‘walk’ or *AMBITĀRE are found (cf. Hinzelin 2011a). Standard Italian has not retained any trace of Latin īre in andare. In Ibero-Romance, PYTA-cells (cf. section 3.1) typically exhibit the fu-stem which is taken from the (also suppletive) paradigm of Latin ESSE ‘be’. (29)

a.

Sp.

ir-INF, ido-PTCP, voy-1SG PRS . IND, vamos-1PL PRS . IND, vaya-1SG PRS . iba-1SG IMPF. IND, iré-1SG FUT. IND, fui-1SG PRET. IND

SBJV,

b.

Pt.

c.

Fr.

ir-INF, ido-PTCP, vou-1SG PRS . IND, vamos/imos-1PL PRS . IND, vá-1SG PRS . SBJV, ia-1sg IMPF. IND, irei-1SG FUT. IND, fui-1SG PRET. IND aller-INF, allé-PTCP, vais-1SG PRS . IND, allons-1PL PRS . IND, aille-1SG allais-1SG IMPF. IND, irai-1SG FUT. IND, allai-1SG PRET. IND

PRS . SBJV,

(30)

d.

Cat.

anar-INF, anat-PTCP, vaig-1SG PRS . IND, anem-1PL PRS . IND, vagi-1SG PRS . SBJV, anava-1SG IMPF. IND, aniré-1SG FUT. IND, aní-1SG PRET. IND

e.

It.

andare-INF, andato-PTCP, vado/vo-1SG PRS . IND, andiamo-1PL PRS . IND, vada-1SG PRS . SBJV, andavo-1SG IMPF. IND, andrò-1SG FUT. IND, andai-1SG PRET. IND

Cat.

future stem: anir- < an- (anar < *AMBITĀRE ) x ir- (< ĪRE )

4.2 Semantic alignment of stem distribution The relation between form and meaning may be explained in terms of iconicity. Iconicity in verbs is typically linked to suffixation: e.g., the present indicative does 42 Rudes (1980, 669) claims that already Latin īre “required the lexicalization of three separate stems”. These were replaced separately in later Romance “waves of suppletion”, the new forms are then “lexicalized as members of the conjugation of the irregular verb” (Rudes 1980, 670, 673). 43 Children acquiring their first language follow this pragmatic principle: “Every two forms contrast in meaning” (Clark 1987, 2).

172

Sascha Gaglia and Marc-Olivier Hinzelin

not exhibit tense/mood morphemes in Romance, while marked tenses and moods do. Zero suffixation of the singular verb forms in French present tense is a further example for iconicity with respect to person (cf. Kilani-Schoch/Dressler 2005, 146– 150).44 But also stem distribution may be aligned semantically: e.g., the Spanish irregular verb saber (‘know’) displays the stem sup- throughout the preterite (supe, supiste, etc.), while the present and imperfect exhibit the default stem sab-. The preterite stem differs from the present/imperfect stem with respect to markedness. The preterite stem represents iconicity since the semantically more marked tense is mirrored by a morphophonemically marked stem. This allomorphic difference also diagrams the (morpho)semantic relatedness between the given tenses according to Bybee (1985; 1988): the more similar two forms semantically are, the more morphophonemic similarity they will exhibit. Hence, two forms of the same tense tend to be more similar than two forms representing different tenses as is shown by the Spanish examples above. Bybee (1985, 93) claims that inflectional splits occur most likely with respect to aspect (most common), tense, and mood (less common).45 In Bybee/Pardo (1981) psycholinguistic evidence for the semantic alignment of stem distribution is provided by means of nonce-verb tests conducted with Spanish speakers. Semantic relations regarding person and number are also shown in Bybee/ Brewer (1980, 230). E.g., with respect to some dialects in Spain (the following examples are taken from the Asturian dialect of Lena), the authors suggest a split of the preterite paradigm along the categories ‘participant’ and ‘non-participant’, i.e. the category ‘participant’ exhibits the stressed vowel [e] coherently in the 1SG , 2SG , 1PL , 2PL (canté, cantéste, cantémos, cantésteis) whereas no coherence is observed with ‘non-participant’, i.e. 3SG and 3PL (cantó, cantáron). Iconicity is also a central notion within Natural Morphology (NM).46 E.g., KilaniSchoch/Dressler (2005, 146–150) claim that the relation between short and long stems in French verb conjugation is iconic (e.g. /fini/-/finis/ ‘finish’, /di/-/diz/ ‘say’, 44 Cf. Greenberg (1966/2005, 44), among others, who refers to zero marking of the 3SG in several languages. However, e.g. Dutch, German, English, and Latin are counterexamples since they use a person/number suffix in this cell of the paradigm. 45 For a critical discussion of Bybee’s (1985) assumptions and an interpretation of apparently inflectional splits in the light of the Autonomy of Morphology (cf. section 5), cf. Maiden (2004) and Hinzelin (2011a). 46 NM combines concepts of universal markedness and of natural phonology (cf. Wurzel 1984; Dressler et al. 1987, among others; cf. Dressler 2006a; 2006b for brief introductions). Its parameters on universal markedness refer to the morphology-phonology interface (segmentally and prosodically; e.g. the optimal shape of a word equals a metrical foot, the optimal shape of an affix one syllable; cf. Dressler 2006b, 540) as well as to the morphology-semantics interface (e.g. in terms of compositional meaning of an inflected and a derived form) and semiotics. With respect to Romance verb inflection, cf. especially Kilani-Schoch/Dressler (2005) who present a meticulous analysis of French. Accounts of Romance verb systems are also Dressler/Thornton (1991) and Spina (2007), among others, as well as Dressler et al. (2006) and other contributions in the same issue of Folia Linguistica dedicated to NM.

Inflectional verb morphology

173

/ba/-/bat/ ‘beat’, etc.), the latter being identical with the 3PL stem in Kilani-Schoch/ Dresslerʼs macroclass II. The alternation of short and long stems is generally observable in present, imperfect, and subjunctive forms (and to a lesser degree in the past participle and the preterite). The authors treat this alternation in terms of iconicity since the long stems can be interpreteted as marking the plural forms of the present indicative as well as all person forms of the subjunctive and the imperfect (cf. KilaniSchoch/Dressler 2005, 147 – also for exceptions). According to the authors, the iconicity of the French imperfect is also mirrored by the root allomorphs which are identical with the morphosemantically marked 1PL /2PL PRS . IND, e.g. buv- (vs. boi- in the paradigm of boire ‘drink’).

4.3 Loss of semantic distinctions In Latin, the verb augment -sc- was originally a derivational element marking inchoative (or inceptive) meaning, e.g. florēre (‘flower’) vs. florēscere (‘coming into bloom’). Later on, the augment becomes an inflectional formative (cf. especially Blaylock 1975 for a diatopic overview; cf. Maiden 2011 with respect to conjugation classes affected by the augment in the modern Romance languages. The latter also provides the most important references on the issue). For modern Romance languages, it is generally assumed that the augment is semantically empty (cf., among others, Harris 1988, 223; Maiden 2003b; 2005, 156–157; 2011, 249 and 710 note 39) and that the distribution of the augment was morphophonologically conditioned (cf. Zamboni 1982, 96, 118, note 27 for references cited from Blaylock 1975 regarding stress-based analyses). E.g. in Italian, the distribution of the augment provides a pattern where the word accent is realized on the theme vowel. Hence, in these paradigms, there is no longer a distinction between forms with rhizotonic and arrhizotonic stress. Zamboni (1982) prefers a functional-semantic analysis with respect to the Aktionsart and in terms of ‘transformativity’ (rather than of ‘inchoativity’) over a purely stress-based solution, arguing that the augment -sc- being incompatible with durativity (for an analysis of the augment *-idio, cf. Zamboni 1980). On the contrary, Maiden rejects a semantic approach (cf. Maiden 2003b), especially referring to Zamboniʼs analyses (cf. also Maiden 2011, 251, 710 note 49), as well as a treatment based on stress (cf. Maiden 2011, 249). Maiden (2003b, 13) argues with respect to -sc- that “[. . .] even if there are individual cases in which a meaning can be assigned to the augment, neither in modern Romance nor in those earlier stages [. . .] can any general meaning be allocated to it” (he argues similarly also with respect to -edj-).47 Instead, he makes an autonomously morphological claim, which we support, stating

47 With this respect, as already mentioned in section 2.1, Lat. -sc- spread from the augment also to the root of verbs which cannot be associated with inchoative meaning as in It. conoscere and Sp. conocer (cf., among others, Schwarze 1999).

174

Sascha Gaglia and Marc-Olivier Hinzelin

that the “attractive force” of the (morphomic) N-pattern was responsible for the paradigmatic distribution of the augments (cf. section 5).

4.4 Syncretism, defectiveness, and overabundance in a semantic/pragmatic perspective In a canonical perspective (Corbett 2007), semantic and pragmatic information should be mapped one-to-one to morphology: An inflectional affix is expected to flesh out the morphosyntactic properties obtained. In Stump’s Paradigm Function Morphology (PFM; cf. Stump 2001; 2002; 2006; Stewart/Stump 2007; a short introduction may be found in Hinzelin/Gaglia 2012), the Content-paradigm function is linked to the Form-paradigm function by rules. These rules of paradigm linkage are situated at the interface; mismatches are thus interface phenomena. Syncretism is such a mismatch: Different morphosyntactic values (semantic/ pragmatic information) like person/number are not expressed by different affixes, instead a single form has two different interpretations (cf. also section 3.2; Hinzelin 2012 provides a more in-depth discussion). The Form-paradigm fails to make a distinction present in the Content-paradigm: (31)

a.

Sp.

hablaba

‘speak-1SG =3SG IMPF. IND ’

b.

It.

parli

‘speak-1SG =2SG =3SG PRS . SBJV ’

c.

Fr.

[paʁl]

‘speak-1SG =2SG =3SG =3PL PRS . IND = PRS . SBJV ’

Another mismatch is defectiveness (cf. also sections 2.3 and 3.3.3), where no form is realized in the paradigm cell of the Form-paradigm. Verbs may be defective for semantic and pragmatic reasons (Baerman/Corbett 2010, 2–3, 8), e.g. impersonal verbs and weather verbs are only found in the 3SG as any other form is perceived as semantically and pragmatically odd as in (32). In this example, there is no morphological reason for the non-existence of this particular form. (32)

Sp.

*lluevo

‘rain-1SG PRS . IND ’

Overabundance (cf. also section 3.3.3) is the opposite phenomenon: Instead of the expected single form, there are “two or more forms realizing the same cell in an inflectional paradigm” (Thornton 2011, 360). A traditional and more general term for variation in form not restricted to morphology is ‘polymorphism’ (cf. Hinzelin 2011b, 41–42). The doublets may be sensitive to semantic features like [+human] and collocations, as they prove to be in the last two examples in (33) (Thornton 2011, 368).

Inflectional verb morphology

(33)

It.

a.

vado/vo

‘go-1SG PRS . IND ’

b.

faccio/fo

‘do-1SG PRS . IND ’

c.

devo/debbo

‘must-1SG PRS . IND ’

d.

sepolto/sepellito

‘bury-PTCP ’

e.

perso/perduto

‘lose-PTCP ’

175

4.5 Verb morphology and pragmatics Phenomena involving the interface between pragmatics and (inflectional) morphology are those that affect verb morphology with respect to the expression of discourse participants and politeness. We shall concentrate on the latter in this section (cf. section 4.2 for the participant vs. non-participant distinction in verb paradigms). A well-known case with a complex system of honorifics is Japanese. Regarding Romance languages, verb morphology is rather marginally involved within politeness systems (cf. Dressler/Merlini Barbaresi 1994, 58–72). They either use a 2PL pronominal form with the corresponding verb to refer to a singular addressee and to signal politeness/social distance (e.g. Fr. vous allez, archaic/regional It. voi andate ‘you go’) or 3SG (e.g. It. Lei va). In European Spanish, the pronoun of 2SG is tú and the politeness pronouns usted (SG ) and ustedes (PL) respectively, are combined with the verb forms of third person. Remember that Spanish, like Italian, is a pro-drop language and that politeness may be expressed also only by the verb desinence. In American Spanish, the picture is as follows: A variety exhibits either (i) tuteo, (ii) voseo or (iii) a mix of both (cf. Fontanella de Weinberg 1999 for an overview regarding usage and the diatopic distribution of the (sub)systems as well as, recently, de Jonge/Nieuwenhuijsen 2014 and the references therein; contributions to Spanish address systems from different linguistic angles, also including sociolinguistics, are provided in Hummel/Kluge/Vázquez Laslop 2010). In varieties exhibiting tuteo, the pronoun for singular addressee tú is combined with the corresponding verb form as it is the case for European Spanish. Regarding voseo, i.e. the etymological 2PL pronoun vos (< Lat. vōs) has ousted tú in 2SG and the verb form is distinct from 2SG in this cell of the paradigm; e.g. especially in Argentina, which is the most prominent case, showing the most homogeneous geographic distribution of voseo, and where the stress falls on the desinence (for an overview of the morphophonological differences between voseo-systems cf. Fontanella de Weinberg 1976); e.g. (vos) llegás ‘you arrive’.48 48 Politeness towards a singular addressee is always expressed by the use of usted and the corresponding verb form as in European Spanish. However, usted can also be used in intimate situations without any polite meaning in American Spanish, i.e. usted de cariño (affectionate usted, cf. Kany 21951, 93–95, also for a general view on the issue with respect to Latin America). The plural ustedes has been generalized to 2PL -reference and is no longer a politeness pronoun in American Spanish.

176

Sascha Gaglia and Marc-Olivier Hinzelin

In Chilean Spanish, the situation is far more complex and it belongs, thus, to the third category, exhibiting the co-occurrence of voseo and tuteo (cf. Torrejón 2010, among others) while voseo corresponds to less social distance and the use of usted in the singular usually to politeness, tuteo has an intermediate position. Syntactically the following sequence types consisting of an address pronoun and a verb form are possible (the corresponding infinitive is llegar; the meaning is ‘you arrive’): (tú) llegas (tuteo), (tú) llegái (tuteo + voseo), (vos) llegái (voseo), (vos) llegas (voseo + tuteo; according to Torrejón 2010 the existence of this type is doubtful in current speech). In Chilean Spanish, the choice of tuteo and voseo is pragmatically as well as sociolinguistically conditioned. However, the literature on Chilean voseo lacks studies that refer to the interface between pragmatics and morphology/ morphosyntax. In a first attempt, Gaglia (to appear) proposes an analysis politeness/ social distance by the means of a specific agreement feature. Work on this issue is still a desideratum in any case. A recent study on L1–acquisition of voseo and tuteo in Río de la Plata Spanish, referring especially to its verb morphology, is Moyna (2009).

5 The autonomy of morphology As already mentioned (cf. section 1), the interpretation of interface phenomena (or even their existence) very much depends on the researcher’s theoretical stance. The role of morphology as a separate module is a precondition for the existence of an interface. The conception of morphology as separate and, furthermore, possessing an autonomous morphomic level (Aronoff 1994) leads to a new analysis of many interface phenomena as purely morphological phenomena. These may have been phonological, semantic or syntactic in origin, but analyses of Romance verb morphology proposed by Maiden (e.g. 2005) suggest that this is very often no longer discernible: When the phonological condition that produced changes in the verb stem of e.g. L-/U-pattern verbs is no longer active, these changes can no longer be explained by phonological rules in a synchronic description. The alternation is now a purely morphological, i.e. morphomic, one. Descriptions that rely on phonological rules deriving allomorphy like Harris (1978) and Burzio (2004) would still see phonology (at the interface with morphology) at work. But proponents of a morphomic interpretation have found many counterexamples challenging the idea of active phonological rules still operating (Pirrelli 2000, 77–80, 178–184; Maiden 2001b; 2011, 241, 709 note 31). With respect to suppletion, Maiden (2004, 249) concludes: “We have seen that even where such alignments [of morphosyntactic properties] do occur they may be accidental, a function of purely morphological ‘fractures’ within paradigms, rather than of any morphosyntactic properties associated with them.”

Inflectional verb morphology

177

Hence, from the point of view of the autonomy of morphology, many interface phenomena described above are considered genuine and pure morphology involving Aronoffian ‘morphomes’ (cf. Hinzelin/Gaglia 2012 for a more detailed discussion of this issue and a number of examples as well as the notion of stem spaces in Bonami/Boyé 2002 and Montermini/Bonami 2013).

6 Conclusion and outlook We have highlighted a number of interface phenomena present in Romance verb inflection with an emphasis on stem allomorphy, syncretism, and periphrastic constructions. Throughout this chapter, it has been a striking observation how ill-defined the notion of “interface” and how theory-dependent the interpretation as an “interface phenomenon” is. Clearly, we are in dire need of a more precise definition of these terms. A promising topic for future research is the functioning of linguistic phenomena like syncretism which are situated at multiple interfaces. Is syncretism quintessentially an (autonomous) morphological phenomenon or is it primarily housed at the interfaces? An answer to these questions promises to provide new insights into the actual nature of ‘interfaces’ in general.

7 References Ackerman, Farrell/Stump, Gregory (2004), “Paradigms and periphrastic expression. A study in realization-based lexicalism”, in: Louisa Sadler/Andrew Spencer (edd.), Projecting Morphology, Stanford, CSLI Publications, 111–157. Albright, Adam (2003), “A quantitative study of Spanish paradigm gaps”, in: Gina Garding/Mimu Tsujimura (edd.), WCCFL 22. Proceedings of the 22nd West Coast Conference on Formal Linguistics, Somerville, MA, Cascadilla Press, 1–14. Anderson, Stephen R. (1982), “Where’s Morphology?”, Linguistic Inquiry 13, 571–612. Anderson, Stephen R. (1992), A-morphous morphology, Cambridge, Cambridge University Press. Aronoff, Mark (1976), Word formation in generative grammar, Cambridge, MA, MIT Press. Aronoff, Mark (1994), Morphology by itself. Stems and inflectional classes, Cambridge, MA, MIT Press. Aski, Janice M. (1995), “Verbal suppletion. An analysis of Italian, French, and Spanish ‘to go’”, Linguistics 33, 403–432. Auger, Julie (1993), “More evidence for verbal agreement-marking in colloquial French”, in: William J. Ashby/Marianne Mithun/Giorgio Perissinotto (edd.), Linguistic perspectives on the Romance languages. Selected papers from the 21st Linguistic Symposium on Romance Languages (LSRL XXI). Santa Barbara, California, 21–24 February 1991, Amsterdam/Philadelphia, Benjamins, 177–198. Baerman, Matthew/Brown, Dunstan/Corbett, Greville G. (2005), The syntax-morphology interface. A study of syncretism, Cambridge, Cambridge University Press.

178

Sascha Gaglia and Marc-Olivier Hinzelin

Baerman, Matthew/Corbett, Greville G. (2010), “Introduction. Defectiveness. Typology and diachrony”, in: Matthew Baerman/Greville G. Corbett/Dunstan Brown (edd.), Defective paradigms. Missing forms and what they tell us, Oxford, The British Academy/Oxford University Press, 1–18. Baerman, Matthew/Corbett, Greville G./Brown, Dunstan (edd.) (2010), Defective paradigms. Missing forms and what they tell us, Oxford, The British Academy/Oxford University Press. Becker, Martin (2010), “Die Ingredienzen des romanischen Imperfekts”, Linguistische Berichte 221, 779–108. Becker, Martin (2013), “La concordance des temps ‒ un cas d’absence ?”, in: Ludwig Fesenmeier/ Anke Grutschus/Carolin Patzelt (edd.), L’absence au niveau syntagmatique. Fallstudien zum Französischen, Frankfurt, Klostermann, 235–258. Blaylock, Curtis (1975), “The Romance development of the Latin verbal augment ‑SK-”, Romance Philology 28, 434–444. Bonami, Olivier/Boyé, Gilles (2002), “Suppletion and dependency in inflectional morphology”, in: Frank Van Eynde/Lars Hellan/Dorothee Beermann (edd.), Proceedings of the 8th International Conference on Head-driven Phrase Structure Grammar. Norwegian University of Science and Technology. Trondheim, Norway, 3–5 August 2001, Stanford, CSLI Publications, 51–70. Brandi, Luciana/Cordin, Patrizia (1989), “Two Italian dialects and the null subject parameter”, in: Osvaldo Jaeggli/Kenneth J. Safir (edd.), The Null Subject Parameter, Dordrecht, Kluwer, 111– 142. Brown, Dunstan/Chumakina, Marina/Corbett, Greville/Popova, Gergana/Spencer, Andrew (2012), “Defining ‘periphrasis’. Key notions”, Morphology 22, 233–275. Burzio, Luigi (2004), “Paradigmatic and syntagmatic relations in Italian verbal inflection”, in: Julie Auger/J. Clancy Clements/Barbara Vance (edd.), Contemporary approaches to Romance linguistics. Selected papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003, Amsterdam/Philadelphia, Benjamins, 17–44. Bybee, Joan L. (1985), Morphology. A study of the relation between meaning and form, Amsterdam/ Philadelphia, Benjamins. Bybee, Joan L. (1988), “Morphology as lexical organization”, in: Michael T. Hammond/Michael Noonan (edd.), Theoretical morphology, San Diego, Academic Press, 119–141. Bybee, Joan L./Brewer, Mary Alexandra (1980), “Explanation in morphophonemics. Changes in Provençal and Spanish preterite forms”, Lingua 52, 201–242. Bybee, Joan L./Pardo, Elly (1981), “On lexical and morphological conditioning of alternations. A nonce-probe experiment with Spanish verbs”, Linguistics 19, 937–968. Calabrese, Andrea (1985), “A constraint-based theory of phonological markedness and simplification procedures”, Linguistic Inquiry 26, 373–463. Calabrese, Andrea (2011), “Metaphony in Romance”, in: Marc van Oostendorp/Colin J. Ewen/Elizabeth Hume/Keren Rice (edd.), The Blackwell companion to phonology, vol. V: Phonology across languages, Chichester, Wiley-Blackwell, 2631–2661. Carstairs-McCarthy, Andrew (2010), The evolution of morphology, Oxford, Oxford University Press. Chomsky, Noam (1957), Syntactic structures, The Hague, Mouton. Chomsky, Noam (1980), Rules and representations, Oxford, Blackwell. Chumakina, Marina (2013), “Introduction”, in: Marina Chumakina/Greville G. Corbett (edd.), Periphrasis. The role of syntax and morphology in paradigms, Oxford, The British Academy/Oxford University Press, 1–23. Chumakina, Marina/Corbett, Greville G. (edd.) (2013), Periphrasis. The role of syntax and morphology in paradigms, Oxford, The British Academy/Oxford University Press. Clark, Eve V. (1987), “The principle of contrast. A constraint on language acquisition”, in: Brian MacWhinney (ed.), Mechanisms of language acquisition, Hillsdale, NJ, Erlbaum, 1–33. Clark, Eve V. (1990), “On the pragmatics of contrast”, Journal of Child Language 17, 417–431.

Inflectional verb morphology

179

Corbett, Greville G. (2007), “Canonical typology, suppletion, and possible words”, Language 83, 8– 42. Cruschina, Silvio (2013), “Beyond the stem and inflectional morphology. An irregular pattern at the level of periphrasis”, in: Silvio Cruschina/Martin Maiden/John Charles Smith (edd.), The boundaries of pure morphology. Diachronic and synchronic perspectives, Oxford, Oxford University Press, 262–283. Culbertson, Jennifer (2010), “Convergent evidence for categorial change in French. From subject clitic to agreement marker”, Language 86, 85–132. Darmesteter, Arsène (1887), La vie des mots étudiée dans leurs significations, Paris, Delagrave. Dressler, Wolfgang U. (1985), Morphonology. The dynamics of derivation, ed. Kenneth C. Hill, Ann Arbor, Karoma. Dressler, Wolfgang U. (2006a), “Introduction. Natural morphology”, Folia Linguistica 40, 1–6. Dressler, Wolfgang U. (2006b), “Natural morphology”, in: Keith Brown (ed.), Encyclopedia of Language & Linguistics, Oxford, Elsevier, 539–540. Dressler, Wolfgang U./Kilani-Schoch, Marianne/Gagarina, Natalia/Pestal, Lina/Pöchtrager, Markus (2006), “On the typology of inflection class systems”, Folia Linguistica 40, 51–74. Dressler, Wolfgang U./Mayerthaler, Willi/Panagl, Oswald/Wurzel, Wolfgang U. (1987), Leitmotifs in natural morphology, Amsterdam/Philadelphia, Benjamins. Dressler, Wolfgang U./Merlini Barbaresi, Lavinia (1994), Morphopragmatics. Diminutives and intensifiers in Italian, German, and other languages, Berlin/New York, De Gruyter. Dressler, Wolfgang U./Thornton, Anna M. (1991), “Doppie basi e binarismo nella morfologia italiana”, Rivista di Linguistica 3, 3–22. Fanciullo, Franco (1988), “Lukanien/Lucania”, in: Günter Holtus/Michael Metzeltin/Christian Schmitt (edd.), Lexikon der Romanistischen Linguistik, vol. 4: Italienisch, Korsisch, Sardisch, Tübingen, Niemeyer, 669–688. Fanciullo, Franco (1994), “Morfo-metafonia”, in: Palmira Cipriano/Paolo Di Giovine/Marco Mancini (edd.), Miscellanea di studi linguistici in onore di Walter Belardi, Roma, Il Calamo, 571–592. Fanciullo, Franco (1998), “Per una interpretazione dei verbi italiani a ‘inserto’ velare”, Archivio Glottologico Italiano 83, 188–239. FEW = Wartburg, Walther v[on] (1952), Französisches Etymologisches Wörterbuch, vol. 4: G H I, Basel, Helbing & Lichtenhahn. Fontanella de Weinberg, María Beatriz (1976), “Analogía y confluencia paradigmática en formas verbales de voseo”, Thesaurus 31, 249–272. Fontanella de Weinberg, María Beatriz (1999), “Sistemas pronominales de tratamiento usados en el mundo hispánico”, in: Ignacio Bosque/Violeta Demonte (edd.), Gramática descriptiva de la lengua Española, vol. I: Entre la oración y el discurso, Madrid, Espasa, 1401–1425. Frigeni, Chiara (2002), “Metaphony in Campidanian Sardinian. A domain-based analysis”, Toronto Working Papers in Linguistics 20, 63–91. Gaglia, Sascha (2009), Metaphonie im kampanischen Dialekt von Piedimonte Matese. Eine Analyse an der Schnittstelle zwischen Phonologie, Morphologie und Lexikon, PhD dissertation, Universität Konstanz, http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-86653 (06.05.2016). Gaglia, Sascha (2012), “Metaphonie im kampanischen Dialekt von Piedimonte Matese. Morphophonologischer Prozess oder morpho-lexikalische Selektion?”, Zeitschrift für romanische Philologie 128, 626–644. Gaglia, Sascha (to appear), Die diachrone und synchrone Spezifikation von phi-Merkmalen im Kontext pronominaler Anrede im Spanischen, Französischen und Italienischen, habilitation thesis, Georg-August-Universität Göttingen. Gaglia, Sascha/Hinzelin, Marc-Olivier (edd.) (2012), Inflection and word formation in Romance languages, Amsterdam/Philadelphia, Benjamins.

180

Sascha Gaglia and Marc-Olivier Hinzelin

Galani, Alexandra/Hicks, Glyn/Tsoulas, George (edd.) (2011), Morphology and its interfaces, Amsterdam/Philadelphia, Benjamins. Giorgi, Alessandra/Pianesi, Fabio (1997), Tense and aspect. From semantics to morphosyntax, Oxford, Oxford University Press. Greenberg, Joseph H. (1966/2005), Language universals. With special reference to feature hierarchies, The Hague, Mouton, 1966. Reprinted edition with a preface by Martin Haspelmath, Berlin/New York, De Gruyter, 2005. Halle, Morris/Marantz, Alec (1993), “Distributed morphology and the pieces of inflection”, in: Kenneth Hale/Samuel Jay Keyser (edd.), The view from building 20. Essays in linguistics in honor of Sylvain Bromberger, Cambridge, MA, MIT Press, 111–176. Harris, James W. (1969), Spanish phonology, Cambridge, MA, MIT Press. Harris, James W. (1978), “Two theories of non-automatic morphophonological alternations. Evidence from Spanish”, Language 51, 41–60. Harris, Martin (1988), “French”, in: Martin Harris/Nigel Vincent (edd.), The Romance languages, London, Croom Helm, 209–245. Haspelmath, Martin (2000), “Periphrasis”, in: Geert Booij/Christian Lehmann/Joachim Mugdan/ Wolfgang Kesselheim/Stavros Skopeteas (edd.), Morphologie/Morphology. Ein internationales Handbuch zur Flexion und Wortbildung/An International handbook on inflection and wordformation, vol. 1, Berlin/New York, De Gruyter, 654–664. Heap, David (2000), La variation grammaticale en géolinguistique. Les pronoms sujet en roman central, München, Lincom. Heap, David (2002), “Split subject pronoun paradigms. Feature geometry and underspecification”, in: Teresa Satterfield/Christina Tortora/Diana Cresti (edd.), Current issues in Romance languages. Selected papers from the 29th Linguistic Symposium on Romance Languages (LSRL), Ann-Arbor, 8–11 April 1999, Amsterdam/Philadelphia, Benjamins, 129–144. Hinzelin, Marc-Olivier (2011a), “Syncretism and suppletion in Gallo-Romance verb paradigms”, in: Martin Maiden/John Charles Smith/Maria Goldbach/Marc-Olivier Hinzelin (edd.), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press, 287–310. Hinzelin, Marc-Olivier (2011b), “L’analyse de la morphologie verbale du francoprovençal. Intégration de données dialectologiques hétérogènes – fatalité ou défi ?”, in: Christine Blauth-Henke/ Matthias Heinz (edd.), Où en sont les études des langues régionales en domaine roman? Données – méthodes – modèles de description, Tübingen, Stauffenburg, 25–47. Hinzelin, Marc-Olivier (2012), “Verb morphology gone astray. Syncretism patterns in GalloRomance”, in: Sascha Gaglia/Marc-Olivier Hinzelin (edd.), Inflection and word formation in Romance languages, Amsterdam/Philadelphia, Benjamins, 55–81. Hinzelin, Marc-Olivier/Gaglia, Sascha (2012), “Morphological theories, the autonomy of morphology, and Romance data”, in: Sascha Gaglia/Marc-Olivier Hinzelin (edd.), Inflection and word formation in Romance languages, Amsterdam/Philadelphia, Benjamins, 1–26. Hinzelin, Marc-Olivier/Kaiser, Georg A. (2012), “Le paramètre du sujet nul dans les variétés dialectales de l’occitan et du francoprovençal”, in: Mario Barra-Jover/Guylaine Brun-Trigaud/JeanPhilippe Dalbera/Patrick Sauzet/Tobias Scheer (edd.), Études de linguistique gallo-romane, Saint-Denis, Presses Universitaires de Vincennes, 247–260. Hockett, Charles F. (1954), “Two models of grammatical description”, Word 10, 210–231. Hooper, Joan B. (1976), An introduction to Natural Generative Phonology, New York, Academic Press. Hopper, Paul J./Traugott, Elizabeth Closs (2003), Grammaticalization, second edition, Cambridge, Cambridge University Press. Hummel, Martin/Kluge, Bettina/Vázquez Laslop, María Eugenia (edd.) (2010), Formas y fórmulas de tratamiento en el mundo hispánico, México, D.F., Colegio de México/Graz, Karl-FranzensUniversität.

Inflectional verb morphology

181

Iliescu, Maria/Popovici, Victoria (2013), Rumänische Grammatik, Hamburg, Buske. de Jonge, Bob/Nieuwenhuijsen, Dorien (2014), “Forms of address”, in: José Ignacio Hualde/Antxon Ollarea/Erin O’Rourke (edd.), The handbook of Hispanic linguistics, Malden, Wiley-Blackwell, 247–262. Kaiser, Georg A. (2008), “Zur Grammatikalisierung der französischen Personalpronomina”, in: Elisabeth Stark/Roland Schmidt-Riese/Eva Stoll (edd.), Romanische Syntax im Wandel, Tübingen, Narr, 305–325. Kany, Charles E. (21951), American-Spanish syntax, Chicago, University of Chicago Press. Kibort, Anna (2011), The feature of tense at the interface of morphology and semantics, in: Alexandra Galani/Glyn Hicks/George Tsoulas (edd.), Morphology and its interfaces, Amsterdam/Philadelphia, Benjamins, 171–193. Kilani-Schoch, Marianne/Dressler, Wolfgang U. (2005), Morphologie naturelle et flexion du verbe français, Tübingen, Narr. Kiparsky, Paul (1982), “From Cyclic Phonology to Lexical Phonology”, in: Harry van der Hulst/Norval Smith (edd.), The structure of phonological representations, Dordrecht, Foris, 131–175. Kiparsky, Paul (1985), “Some consequences of Lexical Phonology”, Phonology Yearbook 2, 85–138. Krämer, Martin (2009), The phonology of Italian, Oxford, Oxford University Press. Lausberg, Heinrich (21972 [1962]), Romanische Sprachwissenschaft. III: Formenlehre, Berlin/New York, De Gruyter. Lehmann, Christian (1985), “The role of grammaticalization in linguistic typology”, in: Hansjakob Seiler/Gunter Brettschneider (edd.), Language invariants and mental operations. International interdisciplinary conference held at Gummersbach/Cologne, Germany, September 18–23, 1983, Tübingen, Narr, 41–52. Lehmann, Christian (1995), Thoughts on grammaticalization, München, Lincom. Lieber, Rochelle (1983), “Argument linking and compounds in English”, Linguistic Inquiry 14, 251– 285. Lieber, Rochelle (1992), Deconstructing morphology. Word formation in syntactic theory, Chicago, The University of Chicago Press. Loporcaro, Michele (2011), “Phonological processes”, in: Martin Maiden/John Charles Smith/Adam Ledgeway (edd.), The Cambridge history of the Romance languages, vol. I: Structures, Cambridge, Cambridge University Press, 109–154, 689–698. Maiden, Martin (1991), Interactive morphonology. Metaphony in Italy, London, Routledge. Maiden, Martin (1996), “On the Romance inflectional endings ‑i and ‑e”, Romance Philology 50, 147–181. Maiden, Martin (2001a), “A strange affinity. ‘Perfecto y tiempos afines’”, Bulletin of Hispanic Studies 78, 441–464. Maiden, Martin (2001b), “Di nuovo sulle alternanze ‘velari’ nel verbo italiano e spagnolo”, Cuadernos de filología italiana 8, 39–61. Maiden, Martin (2003a), “Il verbo italoromanzo. Verso una storia autenticamente morfologica”, in: Mathée Giacomo-Marcellesi/Alvaro Rocchetti (edd.), Il verbo italiano. Studi diacronici, sincronici, contrastivi, didattici. Atti del XXXV congresso internazionale di studi [della Società di Linguistica Italiana], Parigi, 20–22 settembre 2001, Roma, Bulzoni, 3–21. Maiden, Martin (2003b), “Verb augments and meaninglessness in Romance morphology”, Studi di Grammatica Italiana 22, 1–61. Maiden, Martin (2004), “When lexemes become allomorphs. On the genesis of suppletion”, Folia Linguistica 38, 227–256. Maiden, Martin (2005), “Morphological autonomy and diachrony”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of Morphology 2004, Dordrecht, Springer, 137–175.

182

Sascha Gaglia and Marc-Olivier Hinzelin

Maiden, Martin (2011), “Morphophonological innovation”, in: Martin Maiden/John Charles Smith/ Adam Ledgeway (edd.), The Cambridge history of the Romance languages, vol. 1: Structures, Cambridge, Cambridge University Press, 216–267, 706–713. Maiden, Martin/O’Neill, Paul (2010), “On morphomic defectiveness. Evidence from the Romance languages of the Iberian Peninsula”, in: Matthew Baerman/Greville G. Corbett/Dunstan Brown (edd.), Defective paradigms. Missing forms and what they tell us, Oxford, The British Academy/ Oxford University Press, 103–124. Maiden, Martin/Smith, John Charles/Goldbach, Maria/Hinzelin, Marc-Olivier (edd.) (2011), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press. Maiden, Martin/Smith, John Charles/Ledgeway, Adam (edd.) (2011), The Cambridge history of the Romance languages, vol. 1: Structures, Cambridge, Cambridge University Press. Malkiel, Yakov (1974), “New problems in Romance interfixation (I). The velar insert in the present tense (with an excursus on -zer/-zir verbs)”, Romance Philology 27, 304–355. Manzini, Maria Rita/Savoia, Leonardo Maria (2005), I dialetti italiani e romanci. Morfosintassi generativa, vol. 1: Introduzione – Il soggetto – La struttura del complementatore. Frasi interrogative, relative e aspetti della subordinazione, Alessandria, Edizioni dell’Orso. Markun, Hans (1932), Ital. ‘ire’ und ‘andare’, Aarau, Sauerländer. McCarthy, John J. (2005), “Optimal paradigms”, in: Laura J. Downing/T. Alan Hall/Renate Raffelsiefen (edd.), Paradigms in phonological theory, Oxford, Oxford University Press, 170–210. Meinschaefer, Judith (2011), “Accentual patterns in Romance verb forms”, in: Martin Maiden/John Charles Smith/Maria Goldbach/Marc-Olivier Hinzelin (edd.), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press, 51–69. Meisenburg, Trudel (2000), “Vom Wort zum Flexiv? Zu den französischen Pronominalklitika”, Zeitschrift für französische Sprache und Literatur 110, 223–237. Montermini, Fabio/Bonami, Olivier (2013), “Stem spaces and predictability in verbal inflection”, Lingue e Linguaggio 12, 171–190. Morales, Amparo (1989), “Hacia un universal sintáctico del español del Caribe. El orden SVO”, Anuario de Lingüística Hispánica 5/6, 139–152. Moyna, Maria Irene (2009), “Child acquisition and language change. Voseo evolution in Río de la Plata Spanish”, in: Joseph Collentine/Maryellen García/Barbara Lafford/Francisco Marcos Marín (edd.), Selected proceedings of the 11th Hispanic Linguistics Symposium, Somerville, Cascadilla, 131–142. Olbertz, Hella (1998), Verbal periphrases in a functional grammar of Spanish, Berlin/New York, De Gruyter. Oltra-Massuet, Maria Isabel (1999), “On the constituent structure of Catalan verbs”, in: Karlos Arregi/Benjamin Bruening/Cornelia Krause/Vivian Lin (edd.), Papers on morphology and syntax, cycle one (MIT Working Papers in Linguistics 33), Cambridge, MA, MIT, 279–322. Oltra-Massuet, Maria Isabel/Arregi, Karlos (2005), “Stress-by-structure in Spanish”, Linguistic Inquiry 36, 43–84. Penny, Ralph J. (1994), “Continuity and innovation in Romance. Metaphony and mass-noun reference in Spain and Italy”, The Modern Language Review 89, 273–281. Pirrelli, Vito (2000), Paradigmi in morfologia. Un approccio interdisciplinare alla flessione verbale dell’italiano (Linguistica Computazionale, Supplemento ai vol. 17–20), Pisa/Roma, Istituti editoriali e poligrafici internazionali. Poletto, Cecilia (2000), The higher functional field. Evidence from Northern Italian dialects, Oxford, Oxford University Press. Pomino, Natascha (2008a), Spanische Verbalflexion. Eine minimalistische Analyse im Rahmen der Distributed Morphology, Tübingen, Niemeyer.

Inflectional verb morphology

183

Pomino, Natascha (2008b), “Aspekte der Relation zwischen Syntax, Morphologie und Phonologie: Spanische Verbalflexion im Rahmen neuer generativer Annahmen”, in: Eva-Maria Remberger/ Guido Mensching (edd.), Romanistische Syntax – minimalistisch, Tübingen, Narr, 229–252. Querol i Cortiella, Laia (2009), Aspectes morfonològics en la morfologia verbal del català nordoccidental, PhD dissertation, Universitat de Barcelona. Quicoli, Antonio Carlos (1990), “Harmony, lowering and nasalization in Brazilian Portuguese”, Lingua 80, 295–331. Ramchand, Gillian/Reiss, Charles (edd.) (2007), The Oxford handbook of linguistic interfaces, Oxford, Oxford University Press. Remberger, Eva-Maria (2006), Hilfsverben. Eine minimalistische Analyse am Beispiel des Italienischen und Sardischen, Tübingen, Niemeyer. Roberge, Yves (1990), The syntactic recoverability of null arguments, Kingston/Montreal, McGillQueen’s University Press. Roberts, Ian (1992), “A formal account of grammaticalisation in the history of Romance futures”, Folia Linguistica Historica 13, 219–258. Roca, Iggy (1999), “Stress in the Romance languages”, in: Harry van der Hulst (ed.), Word prosodic systems in the languages of Europe, Berlin/New York, De Gruyter, 659–811. Rudes, Blair A. (1980), “On the nature of verbal suppletion”, Linguistics 18, 655–676. Sadler, Louisa/Spencer, Andrew (2001), “Syntax as an exponent of morphological features”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of Morphology 2000, Dordrecht, Kluwer, 71–96. Saltarelli, Mario (2000), “From Latin metre to Romance rhythm”, in: John Charles Smith/Delia Bentley (edd.), Historical linguistics 1995. Selected papers from the 12th International Conference on Historical Linguistics, Manchester, August 1995, vol. 1, Amsterdam/Philadelphia, Benjamins, 345–360. Sánchez Miret, Fernando (1998), La diptongación en las lenguas románicas, München, Lincom. Savoia, Leonardo/Maiden, Martin (1997), “Metaphony”, in: Martin Maiden/Mair Parry (edd.), The dialects of Italy, London, Routledge, 15–25. Schwarze, Christoph (1999), “Inflectional classes in Lexical Functional Morphology. Latin -sk- and its evolution”, in: Miriam Butt/Tracy Holloway King (edd.), Proceedings of the LFG ‘99 conference, Stanford, CSLI, http://web.stanford.edu/group/cslipublications/cslipublications/LFG/ 4/lfg99schwarze.pdf (06.05.2016). Selkirk, Elisabeth O. (1982), The syntax of words, Cambridge, MA, MIT Press. Singh, Rajendra/Desrochers, Richard (edd.) (1996), Trubetzkoy’s orphan. Proceedings of the Montréal roundtable on “Morphonology. Contemporary responses” (Montréal, 30 September–2 October 1994), Amsterdam/Philadelphia, Benjamins. Sluyters, Willebrord (1988), “Vowel harmony, rule formats and underspecification. The dialect of Francavilla-Fontana”, in: Harry van der Hulst/Norval Smith (edd.), Features, segmental structure and harmony processes, Dordrecht, Foris, 161–184. Sluyters, Willebrord (1990), “Length and stress revisited. A metrical account of diphthongization, vowel lengthening, consonant gemination and word final vowel epenthesis in modern Italian”, Probus 2, 65–102. Spina, Rossella (2007), L’evoluzione della coniugazione italoromanza, Catania, Ed.it. Squartini, Mario (1998), Verbal periphrases in Romance. Aspect, actionality, and grammaticalization, Berlin/New York, De Gruyter. Stewart, Thomas/Stump, Gregory (2007), “Paradigm function morphology and the morphologysyntax interface”, in: Gillian Ramchand/Charles Reiss (edd.), The Oxford handbook of linguistic interfaces, Oxford, Oxford University Press, 383–421. Stump, Gregory T. (2001), Inflectional morphology. A theory of paradigm structure, Cambridge, Cambridge University Press.

184

Sascha Gaglia and Marc-Olivier Hinzelin

Stump, Gregory T. (2002), “Morphological and syntactic paradigms. Arguments for a theory of paradigm linkage”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of morphology 2001, Dordrecht, Kluwer, 147–180. Stump, Gregory T. (2006), “Heteroclisis and paradigm linkage”, Language 82, 279–322. Taylor, Catherine (2011), Periphrasis in Romance, in: Martin Maiden/John Charles Smith/Maria Goldbach/Marc-Olivier Hinzelin (edd.), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press, 401–416. Thornton, Anna M. (2011), “Overabundance (multiple forms realizing the same cell). A non-canonical phenomenon in Italian verb morphology”, in: Martin Maiden/John Charles Smith/Maria Goldbach/ Marc-Olivier Hinzelin (edd.), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press, 358–381. Torrejón, Alfredo (2010), El voseo en Chile. Una aproximación diacrónica, in: Martin Hummel/Bettina Kluge/María Eugenia Vázquez Laslop (edd.), Formas y fórmulas de tratamiento en el mundo hispánico, México, D.F., Colegio de México/Graz, Karl-Franzens-Universität, 413–428. Trubetzkoy, Nikolai S. (1929), “Sur la ‘morphonologie’”, Travaux du Cercle Linguistique de Prague 1, 85–88. Tuttle, Edward (1985), “Morphologization as redundancy in central Italian dialects”, Romance Philology 39, 35–43. Väänänen, Veikko (1963), Introduction au latin vulgaire, Paris, Klincksieck. van der Veer, Bart (2006), The Italian ‘mobile diphthongs’. A test case for experimental phonetics and phonological theory, PhD dissertation, Universiteit Leiden. Vennemann, Theo (1988), Preference laws for syllable structure and the explanation of sound change. With special reference to German, Germanic, Italian, and Latin, Berlin/New York, De Gruyter. Vincent, Nigel (1982), “The development of the auxiliaries ‘habere’ and ‘esse’ in Romance”, in: Nigel Vincent/Martin Harris (edd.), Studies in the Romance verb. Essays offered to Joe Cremona on the occasion of his 60th birthday, London, Croom Helm, 71–96. Vincent, Nigel (1988a), “Latin”, in: Martin Harris/Nigel Vincent (edd.), The Romance languages, London, Croom Helm, 26–78. Vincent, Nigel (1988b), “Non-linear phonology in diachronic perspective. Stress and word-structure in Latin and Italian”, in: Pier Marco Bertinetto/Michele Loporcaro (edd.), Certamen phonologicum. Papers from the 1987 Cortona phonology meeting, Torino, Rosenberg & Sellier, 421–432. Vincent, Nigel (2011), “Non-finite forms, periphrases, and autonomous morphology in Latin and Romance”, in: Martin Maiden/John Charles Smith/Maria Goldbach/Marc-Olivier Hinzelin (edd.), Morphological autonomy. Perspectives from Romance inflectional morphology, Oxford, Oxford University Press, 417–435. Vogel, Irene (1993), “Verbs in Italian morphology”, in: Geert Booij/Jaap van Marle (edd.), Yearbook of morphology 1993, Dordrecht, Kluwer, 219–254. Williams, Edwin (1981), “On the notions ‘lexically related’ and ‘head of a word’”, Linguistic Inquiry 12, 245–274. Wurzel, Wolfgang (1984), Flexionsmorphologie und Natürlichkeit, Berlin, Akademie-Verlag. Zamboni, Alberto (1980), “Un problema di morfologia romanza. L’ampliamento verbale in ‘‑idio’, ‘‑izo’”, Quaderni Patavini di Linguistica 2, 171–188. Zamboni, Alberto (1982), “La morfologia verbale latina in +sc+ e la sua evoluzione romanza. Appunti per una nuova esplicativa”, Quaderni Patavini di Linguistica 3, 87–138. Zwicky, Arnold M. (1983), “An expanded view of morphology in the syntax-phonology interface”, in: Shirô Hattori/Kazuko Inoue/Tadao Shimomiya/Yoshio Nagashima (edd.), Proceedings of the XIIIth international congress of linguists, 29 August–4 September 1982, Tokyo, Tokyo, Gakushuin University, 198–208.

II Structure and meaning

M. Teresa Espinal and Susagna Tubau

6 Meaning of words and meaning of sentences Abstract: This article addresses the general question of what the meaning of words, namely n-words, is and how this meaning may contribute to the meaning of sentences. We hypothesize that this contribution is not conceptual, but is fundamentally underspecified. An underspecified value can be attributed both to an inherent semantic feature and to a syntactic feature. In the specific case of n-words, we postulate a semantic feature that guarantees their behaviour as polarity items, and a syntactic feature that guarantees their occurrence in Negative Concord structures. The Principle of Compositionality is guaranteed by checking both semantic features and syntactically uninterpretable ones. Keywords: underspecified meaning, compositionality, n-words, Romance languages

1 Introduction The question we will address in this chapter is what the meaning of words, more precisely the meaning of lexical roots, is and how their meaning may contribute to the meaning of sentences, mainly within a minimalist semantics approach. A widespread assumption within classic lexical semantics (Cruse 1986; Levin/ Pinker 1991; Geeraerts 2010, inter alia) is that the lexical meaning is conceptual, which suggests that, for each word, its lexically encoded (context-independent) meaning is associated with either a concept or a set of concepts organized in specific ways and relationships. An alternative view, within the philosophical realism tradition, is that words denote things in the world (Frege 1892; Tarski 1944; Davidson 1967). These two views may also be conceived as complementary: Words can be said to correspond to mental representations that represent entities in the world (Fodor 1983). The idea of a conceptualist approach, as pointed out by Carston (2012, 607–608) is that: “We use sentences to express/communicate thoughts (truthconditional contents) and we use words to express/communicate concepts, which are constituents of thoughts (hence contribute to truth-conditional contents).” However, as pointed out by this author, it is now quite widely accepted that the meaning Acknowledgement: This research has been funded by two research grants awarded by the Spanish Ministerio de Ciencia e Innovación (FFI2014-52015-P), and by grants awarded by the Generalitat de Catalunya to the Centre de Lingüística Teòrica (2014SGR-1013). The first author also acknowledges an ICREA Academia award.

188

M. Teresa Espinal and Susagna Tubau

(or semantic content) that a word is used to express or communicate on an occasion of utterance is often distinct from the meaning it has as an expression type in a language system (that is, its standing or encoded meaning). This view is claimed to be shared by ‘contextualist’ philosophers of language, by some linguists, and by pragmaticists working within a cognitive framework such as Relevance Theory. Furthermore, as discussed by Carston (2012, 608), in spite of the fact that “word type meanings might be concepts, hence contentful entities that can be constituents of thought [. . .] there is an equally widely held view that word meanings are ‘underspecified’; that is, that they cannot contribute directly, without modification or transformation of some sort, to the thoughts/propositions that utterances in which they occur are used to express”. In this chapter we follow the latter approach, and we take seriously the idea that words, not only closed-class words (indexicals, determiners, quantifiers, prepositions, and other function words, such as connectives), but also open-class words (nouns, verbs, adjectives, and adverbs), may not encode concepts or map directly to contentful entities, but rather come with meaning-relevant components that are underspecified with respect to content. It is our aim to investigate how this underspecification can be understood, how it can be formalized, and how it can interact with other meaning components at the time of composing the meaning of sentences. It should be pointed out that this view not only makes specific predictions with regard to the meaning of words, but it also has specific consequences for the way we understand how the various types of meanings of words combine and compose to build the meaning of sentences. In this respect, we will have to evaluate to what extent the Principle of Compositionality (PoC; Frege 1892; Werning/Hinzen/Machery 2012; Pelletier 2011) might be accommodated when the meaning of words is highly underspecified. The PoC, which has been largely assumed by philosophers and semanticists to be a universal mode of composition of the meaning of a complex expression (e.g. a sentence), postulates that this meaning is determined by or is a function of the meaning of its constituent words and the rules used to combine them. This principle has been presented in the literature in different forms, depending on the degree of determinism that is postulated from the lexical meaning of words and the syntactic structure in which these words appear. More precisely, strong conceptual compositionality has been postulated in those circumstances where a full conceptual meaning for words directly combines with either the conceptual or functional meaning of other words to build the conceptual representation of a sentence (Jackendoff 1972). Strong compositionality of meaning has also been postulated in those circumstances where a syntactic algebra is interpreted through a meaning-assignment m, a function from E(xpression) to M(odel), and the set of available meanings for the expressions of E in order to build a logical representation for an expression E. This is the view developed in formal semantics (Montague 1974). Within the Principles and Parameters model of Generative Grammar a syntactic level of representation has been postulated, namely the Logical Form (LF), which corresponds to the meaning of a sentence at

Meaning of words and meaning of sentences

189

the syntax-semantics interface, as determined by lexical items occurring in specific positions within a syntactic structure (May 1985; Higginbotham 1985; Hornstein 1995). The PoC has been supplemented in the theory of language by various operations and processes that aim at accounting for the meaning of expressions that do not directly compose following the schema of function application. In this chapter we will explore the relevance of the PoC under the view that postulates that the meaning of words is not genuinely conceptual and that, on the contrary, it is (syntactically or semantically) underspecified. If we assume that lexical items (i.e. roots), no matter whether they correspond to open-class or closed-class words, contribute indirectly to the meaning of sentences by providing syntactic or semantic features whose contents will be valued in context by some process of enrichment that can either take place at the syntax-semantics interface (by means of a grammatically driven operation) or beyond grammar (by some process of pragmatic adjustment), we will have to seriously investigate what sort of features are involved in an underspecified characterization, and decide whether they are purely semantic, syntactic, or both.1 Our working hypothesis is that an underspecified meaning can be attributed both to an inherent semantic feature and to a syntactic feature with which a lexical root merges during a derivation. This suggests that an underspecified meaning, caused either by an underspecified semantic feature or by an underspecified syntactic feature is fully interpreted only when this feature is checked by an abstract operator, usually in a local domain. As we will discuss, a way to solve the compositionality of meaning at the syntax-semantics interface is to rely on the grammatical operation of Agree and the Principle of C-command (cf. below). The idea we will explore is that lexical items with underspecified meanings have underspecified features that must be submitted to highly constrained grammatical operations that guarantee their syntactic well-formedness and interpretation before these items can be claimed to contribute to a more complex meaning (that is, to the meaning of sentences) and, therefore, satisfy the PoC.2 In the rest of this introduction (section 1.1) we will sketch the relevance of this approach at the time of building the contribution of various types of pronouns and complementizers to the meaning of complex sentences. In section 2 we will focus on the central topic of this chapter: the compositionality of the meaning of n(egative)words (Laka 1990) to the meaning of sentences. In order to achieve this goal, we will focus on their features, their distribution, and the formal requirements for building 1 To our knowledge the antecedents of a theory of underspecification in linguistics have their origin in studies in phonology (Jakobson 1984a; 1984b), but have been extended to synchronic studies of morphology and syntax (Lumsden 1992; Farkas 1990; Rooryck 1994), as well as to diachronic studies on lexical change (Martins 1998). 2 Beyond grammar, the enrichment of underdetermined meanings up to the comprehension of a thought is not regulated by the PoC, which applies within the grammatical domain, but is pragmatically regulated by nonlinguistic principles, such as the Principle of Relevance (Sperber/Wilson 1986/ 1995), which attempts to account for the interaction between linguistically encoded meaning and nonlinguistically accessible information.

190

M. Teresa Espinal and Susagna Tubau

either a negative reading, distributed over multiple negative items (both in Strict and Non-Strict Negative Concord –NC– Romance languages), or a nonnegative reading (in interrogative, comparative, conditionals, and expletive contexts).3 In this section we will limit the data discussed to Catalan, French, Romanian, and Spanish.4

1.1 Pronouns and complementizers A classic example to illustrate the notion of underspecified meaning might be to consider the characterization of pronouns in relation to their distribution within a clausal structure. Traditional grammars already put forward the distinction between reflexive and third person clitics, which any native speaker of a language is supposed to be well aware of. For example, if we consider the following data from Spanish, a native speaker of this language realizes that, in spite of the fact that the occurrence of reflexive pronouns and the occurrence of third person clitics within sentences do show strong parallels, the meaning conveyed by the two types of pronouns is different, and also that this depends on the distribution and licensing conditions of formal features that characterize these linguistic objects. reprimen (a sí mismosi). (1) Sp. a. Los Hermanos Musulmanesi sei REFL repress to themselves the brothers Muslim ‘The Muslim Brothers repress themselves.’ reprimen. b. Los Hermanos Musulmanesi losj the brothers Muslim them repress ‘The Muslim Brothers repress them.’ c. Los Hermanos Musulmanesi afirman the brothers Muslim claim

que los jefes del ejército that the leaders of.the Army

reprimen (a sí mismosj). repress to Themselves ‘The Muslim Brothers claim that the army leaders repress themselves.’

sej

REFL

3 It should be remarked that in Romance a Double Negation (DN) reading (Horn 1989), by which a positive interpretation can be inferred from the co-occurrence of two or more negative expressions, is compositionally driven only under very restricted syntactic contexts: when two negative markers are distributed in a main and a subordinate clause and when a negative marker co-occurs with a negative prefix (cf. Bosque 1980; Sánchez 1999 for Spanish; and Solà 1973; Espinal 2002 for Catalan). In addition, recent studies show the possibility of DN interpretation at the syntax-prosody and prosody-gesture interfaces (cf. Espinal/Prieto 2011; Prieto et al. 2013; and Espinal et al. (2015) for studies on Catalan and Spanish), and with simple transitive sentences (cf. Déprez/Cheylus/Larrivée 2013 for French; Déprez et al. 2015 for Catalan). 4 Italian and European Portuguese are expected to work like Spanish (Zanuttini 1997; Matos 1999). Brazilian Portuguese is expected to work like Catalan (Teixeira de Sousa 2012). Cf. also de Swart (2010).

Meaning of words and meaning of sentences

d. Los Hermanos Musulmanesi afirman the brothers Muslim claim

191

que los jefes del ejército that the leaders of.the army

losi reprimen. them repress ‘The Muslim Brothers claim that the army leaders repress them.’ Identity of indices represents identity of reference, while disjoint indices represent disjoint reference. Thus, whereas in (1a) the reflexive is interpreted as sharing reference with the only possible nominal antecedent that occurs in subject position in the sentence, in (1c) the reflexive is interpreted as constraining identity of reference with the structurally most immediate nominal antecedent, that is, the subject of the subordinate clause. Similarly, whereas in (1b) the third person plural accusative clitic is interpreted as having disjoint reference with the only possible nominal antecedent that occurs in subject position in the sentence, in (1d) the same pronoun is interpreted as constraining disjoint reference with the structurally most immediate nominal antecedent that occurs in subject position of the subordinate clause (Chomsky 1981; Reinhart 1983). This structural constraint makes it possible for the clitic pronoun to be interpreted as coreferent with the subject of the main clause or to a third set of individuals (neither los Hermanos Musulmanes ‘Muslim brothers’ nor los jefes del ejército ‘the leaders of the army’). The question is: What is the meaning of a clitic pronoun that makes this possible? In accordance with a minimalist approach to semantics the most reasonable answer is that its meaning is grammatically underspecified, which makes it susceptible to some process of enrichment, either grammatically (when it is possible to postulate a formal checking relationship between the antecedent and the pronoun) or pragmatically (when this is not possible, but the pronoun still constrains the proposition expressed by providing instructions on the procedures required to find an appropriate referent for the pronoun; Wilson/Sperber 1993; Espinal 1996). Consider the lexical representations in (2).5 (2)

a.

se:

[REFLEX , uREF ]

b.

los:

[-IP, -IIP, +PL , -FEM , uREF ]

In these schemas [uREF ] stands for an uninterpretable reference formal feature, which means that the linguistic object that has this feature has no reference by itself; [uREF ] must be checked by an [iREF ] formal feature, which most characteristically characterizes proper names and definite descriptions of the sort exemplified in (1). For reflexive pronouns the formal condition is that the pronoun is dependent on 5 [REFLEX ] stands for reflexive, [uREF ] for uninterpretable reference, [-IP, -IIP ] for third person, [+PL] for plural, and [-FEM ] for masculine.

192

M. Teresa Espinal and Susagna Tubau

an antecedent [iREF ] that is a constituent of the same clause and, furthermore, the antecedent must c-command the anaphor. For third person clitic pronouns the requirement is that the pronoun does not have its antecedent in the same clause; in other words, [uREF ] is not dependent on an antecedent that occurs in the same structural domain, but in a higher structure; again the antecedent must c-command the pronoun.6 Within the Principles and Parameters model of Generative Grammar (Chomsky 1981 and subsequent work) these constraints on anaphor and pronoun binding are subsumed within the well-known Binding Principles, the general idea being that a linguistic theory of pronoun interpretation needs to bring together the ability to distribute and restrict the meaning of these linguistic objects within clausal domains. A second phenomenon suitable to illustrate how a minimalist semantics approach can explain the contribution of word meaning to the meaning of sentences has to do with modality and syntax, and more specifically with the distribution and meaning of various moods and complementizers. Let us consider the data in (3). (3)

Cat.

a.

Desitjo que sigui hope.1SG that be.3SG . SUBJ ‘I hope s/he is punctual.’

puntual. punctual

b.

Em pregunto si serà puntual. if be.3SG . FUT Punctual me ask.1SG ‘I wonder whether s/he will be punctual.’

These examples illustrate a dependency relationship between the type of modal domain of the subordinate clause (subjunctive (SUBJ) vs. indicative (IND)), the type of complementizer introducing the subordinate clause (que ‘that’ vs. si ‘whether’), and the type of verb of the main clause of which the subordinate clause is an argument (modal vs. interrogative).7 These examples show that the modal flavour of a specific mood in the subordinate clause is indirectly dependent on the modal anchor provided by the complementizer that heads the object argument of the verb of the main clause (Rigau 1984; Quer 1998). This observation supports the line of research developed by Kratzer (1981; 2012; 2013) and Hacquard (2006), according to which a hypothesis is developed on the projection, restriction, and syntactic representation of modal domains: “different types of modals select different types of anchors” and “modal anchors should be provided by the arguments of their modals” (Kratzer

6 [uREF ] must be checked within grammar. When the antecedent is in a position not c-commanding the pronominal or in a previous discourse other principles beyond coindexing and c-command apply to guarantee the interpretation of the anaphora. Cf. also Reinhart (1983) for disjoint reference and noncoreference. 7 Furthermore, these examples show a temporal correlation between the Tense and Aspect of the subordinate clause and the ones specified in the main clause. We leave this topic aside.

Meaning of words and meaning of sentences

193

2013, 191). The dependency we are postulating for (3) can also be formalized in terms of underspecified features, as illustrated in (4). (4)

a.

Vmodal[iM O O D ]

[ que[uM O O D ]. . . V[uM O O D ]. . . ]

b.

Vinterr[iM O O D ]

[ si[uM O O D ] . . . V[uM O O D ] . . . ]

What these rough representations make explicit is that the mood of the subordinate clause is underspecified and is dependent on the type of complementizer head, which, in turn, is dependent on the mood specification triggered by the verb of the main clause. This suggests that the underspecified mood feature of the verb in the subordinate clause will be checked and interpreted under a grammatical relation of Agree, similar to the one required for the checking of uninterpretable formal features in the case of pronouns. To sum up, in these two examples we have briefly sketched that the meaning of pronouns, as well as the meaning of complementizers and moods, is neither conceptual nor straightforwardly referential, but is underspecified for some features. The way we understand this underspecification is that some features formally characterizing lexical items have a value that is context-dependent. In the two cases considered here these features are checked locally and they get an interpretable content via a formal relationship of Agree defined on the basis of the Principle of C-command. In the next section we will present in more detail how this sort of analysis allows us to account for the meaning of n-words and its contribution to the meaning of negative sentences in various NC languages. We will focus on how an account of the meaning of words in nonconceptual terms can be extended to explain the meaning contribution of n-words and even the contribution of negative markers. The challenge of the characterization we aim to achieve is how to account for the distribution and meaning of n-words, and the compositionality of their meaning in negative (concord) contexts as well as in nonnegative ones. The ingredients of this minimalist semantics approach are the following: (i) lexical items (or roots) may be defined by underspecified semantic or syntactic features, (ii) some syntactic operations guarantee the formal interpretability and instantiation of formal features at morphophonology, and (iii) some semantic operations guarantee the compositionality of meaning at LF.

2 The distribution and meaning of n-words 2.1 Definition and distribution of n-words N-words are indefinite expressions that may encode a negative meaning, as shown in (5a, b) for Spanish. Notice that the preverbal n-word in (5a) or the isolated

194

M. Teresa Espinal and Susagna Tubau

n-word used as an answer in (5b) contribute a negative meaning either to the sentence (in (5a)) or to the fragment answer (in (5b)). (5)

Sp.

a.

Nadie ha llamado. nobody has called ‘Nobody called.’

b.

Q: ¿Quién ha who has ‘Who called?’ A:

llamado? called

Nadie. nobody ‘Nobody.’

N-words also seem to behave like polarity items (PIs), as shown in (6a–c) for Catalan.8 (6)

Cat.

a. *(No) he vist res. not have.1SG seen anything ‘I haven’t seen anything.’ b.

Has vist res? have.2SG seen anything ‘Have you seen anything?’

c.

Si veus res, avisa’m. if see.2SG anything warn.me ‘If you see anything, let me know.’

In this set of sentences the n-word needs to be licensed by a suitable operator (a sentential negation in (6a), an interrogative operator in (6b), and a conditional operator in (6c)). The term n-word, originally coined by Laka (1990), captures the fact that in Romance languages such as Spanish, Italian, and Portuguese, most −but not all− lexical items that participate in NC structures are spelled with an initial n- (e.g. Spanish nadie ‘nobody/anybody,’ nada ‘nothing/anything’; Italian nessuno ‘nobody/

8 In Catalan n-words can also be used as fragment answers to questions, as illustrated in (5b) for Spanish. That is, res (lit. thing) is also legitimate as a fragment answer with a negative meaning ‘nothing’. It should be noted that in Spanish n-words are less likely to be found in polarity contexts (Bosque 1980; Sánchez 1999).

Meaning of words and meaning of sentences

195

anybody,’ niente ‘nothing/anything’; Portuguese ninguém ‘nobody/anybody,’ nada ‘nothing/anything’).9 A formal definition of n-word is given in (7). The property in (7a) is illustrated by (6a), while the property in (7b) is exemplified in (5b). (7)

An expression α is an n-word iff: (Giannakidou 2006, 328) a. α can be used in structures containing sentential negation or another α-expression yielding a reading equivalent to one logical negation; b.

α can provide a negative fragment answer.

NC is commonly defined as the possibility that n-words have to combine with several manifestations of negation, although negation is logically computed just once. That is, multiple occurrences of negative constituents express a single negation (Labov 1972; Muller 1991; van der Wouden 1994; Acquaviva 1996; 1997; Déprez 1997; Giannakidou 2000; de Swart/Sag 2002; Corblin/Tovena 2003; Floricic 2005; Corblin et al. 2004; Tubau 2008, among others). In NC languages, n-words –some of which are not spelled with an initial n- (e.g. French personne ‘nobody/anybody,’ rien ‘nothing/anything’; Catalan res ‘nothing/ anything,’ cap ‘no/any’)– need to be licensed, in negative contexts, by the sentential negative marker (SNM) (8a) or by another n-word in preverbal position (8b). Without an appropriate licenser the sequence is ungrammatical (8c). (8)

Sp.

a.

No ha visto nada. not has seen anything ‘S/he didn’t see anything.’

b.

Nadie ha visto nada. nobody has seen anything ‘Nobody saw anything.’

c. *Ha has

visto seen

nada. anything

While all NC languages require postverbal n-words to be licensed by a negative expression in negative contexts –either a SNM or a preverbal n-word– a crucial difference exists among them when preverbal n-words are taken into account: While

9 The initial n- is not an indication of negative morphology, though. Etymologically, the Spanish n-words nadie and nada can be traced back to a Latin adjective: (homines) natī ‘(men) born’ and (res) nata ‘(thing) born’.

196

M. Teresa Espinal and Susagna Tubau

languages like Spanish and Italian are defined as having a system of NonStrict NC, languages like Romanian or Modern Greek have a system of Strict NC (Giannakidou 1998; 2000). In Non-Strict NC languages, the SNM is not compatible with n-words occurring preverbally if a single negation meaning is to be expressed. Conversely, in Strict NC languages, preverbal n-words and the SNM always co-occur. This is shown in (9) for Spanish (a typical Non-Strict NC language) and in (10) for Romanian (a typical Strict NC language). (9)

Sp.

(10) Rom.

Nadie (*no) ha visto nobody not has seen ‘Nobody saw anything.’ Nimeni *(nu) nobody not ‘Nobody calls.’

sună. calls

nada. anything

(Non-Strict NC)

(Strict NC)

Catalan is somewhat special with regard to the Strict vs. Non-Strict NC distinction, as for some speakers (including the authors of this chapter), the SNM can optionally co-occur with preverbal n-words with no difference in meaning. This is why the SNM no is in parentheses in (11). (11)

Cat.

Ningú (no) ha vist nobody not has seen ‘Nobody saw anything.’

res. anything

In the case of Standard French, n-words always co-occur with the negative scope marker ne, thus reproducing the pattern already illustrated in (10) for Romanian. This is shown in (12a). Example (12b) illustrates a postverbal n-word, a pattern similar to the one in (6a) for Catalan and (8a) for Spanish. (12)

Fr.

a.

Personne n’a nobody not.has ‘Nobody ate.’

mangé. eaten

b.

Jean n’a rien vu. Jean not.has anything seen ‘Jean didn’t see anything.’

However, if the SNM that combines with French n-words is pas, Double Negation (DN) arises, as shown in (13).

Meaning of words and meaning of sentences

(13)

Fr.

a.

Personne n’a pas mangé. nobody not.has not eaten ‘Nobody didn’t eat.’ (= Everybody ate)

b.

Jean n’a pas rien vu. Jean not.has not anything seen ‘Jean didn’t see nothing.’ (= Jean saw something)

197

In short, what the examples above show is that n-words may obtain a final negative reading, as in (5a, b), participate in NC structures, as in (6a), (8)–(12), or convey a nonnegative meaning, as in (6b, c). In NC structures, the SNM is sometimes overt with preverbal n-words (in Strict NC languages, as in (10)), but sometimes it is not (in Non-Strict NC languages, as in (9)). Other times it is truly optional (as in (11)). In view of this variation it is our aim to disentangle (i) what exactly the contribution of n-words to the meaning of sentences is, (ii) what the feature specification of n-words is, and (iii) how the licensing of the features we will postulate in section 2.2 will guarantee the interpretation of n-words both in NC (Strict and Non-Strict) contexts and in nonnegative ones. In the next section, we argue that n-words can be formally characterized with reference to two main kinds of features: the semantic feature [+σ] (Chierchia 2006) and the syntactic feature [uNeg] (Zeijlstra 2004).10 This characterization will be required in order to discuss in section 3 how the meaning of n-words contributes to the meaning of sentences. Special emphasis will also be put on explaining how the semantic composition of negation crucially diverges in Strict and Non-Strict NC languages, and is dependent on the characterization of negative markers.

2.2 Underspecified meaning of PIs and n-words We claimed in relation to (6) that when n-words are postverbal they behave like PIs, which can be licensed both in negative and nonnegative contexts. PIs, as defined by Giannakidou (2000), are linguistic expressions that bear a semantic requirement with respect to the kinds of contexts in which they can appear. As stated in the 10 It was already argued in Tubau/Espinal (2012) that Catalan n-words, namely res ‘anything,’ being a PI, carry the semantic feature [+σ]. However, Labelle/Espinal (2014) is, to our knowledge, the first study that postulates a combination of the semantic feature [+σ] and the morphosyntactic feature [uNeg] to account for the diachronic changes that affected French negative expressions: The semantic feature is postulated to be responsible for the interpretation of an expression as a PI, and a morphosyntactic feature is postulated to be responsible for the n-word behaviour of an expression. See this study for the proposal that, independently of having an initial n- or not, words may change from less negative to more negative or vice versa, a process that is explained in terms of feature changes that affect lexical items one by one. Cf. also Déprez’s (2011) microparametric approach.

198

M. Teresa Espinal and Susagna Tubau

definition in (14), PIs need to occur in contexts with a nonveridical operator, defined as in (15). (14)

A linguistic expression α is a polarity item iff: (Giannakidou 2000, 464) (i) the distribution of α is limited by sensitivity to some semantic property β of the context of appearance; and (ii) β is (non)veridicality.

(15)

(Non)veridicality (Zwarts 1995, 287) Let O be a monadic sentential operator. O is said to be veridical just in case Op ⇒ p is logically valid. If O is not veridical, then O is nonveridical. A nonveridical operator O is called a[nti]veridical iff Op ⇒ ~p is logically valid.11

If we consider again the data in (6), it should be noted that n-words in Romance may qualify as PIs, as they can be licensed by negation (an antiveridical operator, a subset of the nonveridical class of operators), as well as interrogative and conditional operators (both nonveridical).12 According to Chierchia (2006), the semantics of PIs makes them felicitous only in downward-entailing contexts (Ladusaw 1980), because they are scalar items that activate alternatives within smaller domains. That is, PIs must be interpreted within the largest relevant pragmatic context while also activating alternatives within smaller domains. Take, for instance, the PI any in a sentence like (16). (16) The students didn’t read any books. In (16), the relevant pragmatic context is not restricted to the books included in a course bibliography, for example, but it extends to a much larger set of books, namely all kinds one can think of. With respect to the activation of alternatives, the indefinite item any, in any books, implicates that the students did not read math books, poetry books, history books, and so on. Following work by Kadmon/Landman (1993), Krifka (1994), and Lahiri (1998), Chierchia (2006, 559) postulates a [+σ] feature to account for the fact that PIs induce the process of domain widening described above. Given that, as shown in (6a–c), Romance n-words behave like PIs, they can be assumed to be scalar terms that bear the semantic feature [+σ]. Thus, similar to any in (16), the n-word res ‘anything’ in a sentence like (6a) activates alternatives (e.g. a specific individual object, some-

11 We follow Giannakidou (1998) in substituting antiveridicality for Zwarts’s (1995) averidicality, since the meaning intended is ‘opposite to veridicality,’ not ‘without veridicality properties.’ 12 Not all Romance languages, however, show the same sort of distribution. For example, n-words in French and Romanian can only occur in negative contexts and cannot be used in other polar contexts such as conditional and interrogative sentences, for which other lexical items are required. Cf. the summary table below.

Meaning of words and meaning of sentences

199

thing, many things etc.) and introduces the implicature that res ‘anything’ is the pragmatically strongest alternative in this context (i.e. if I haven’t seen anything, then I haven’t seen something, I haven’t seen many things etc.). The [+σ] feature, associated with the scalar item, is uninterpretable and has to be checked in the syntactic representation of meaning by an interpretable abstract σ operator that can attach to negation as well as to other kinds of nonveridical operators, the idea being that the feature [+σ] linguistically encodes the need for an enriched interpretation (Chierchia 2006, 553–554). If the abstract σ operator attaches to a negative operator, the PI bearing a [+σ] feature will be interpreted as a Negative PI, while if the abstract σ operator attaches to a nonnegative operator the PI bearing a [+σ] feature will be interpreted as a Positive PI. Notice that a [+σ] feature is also found in PIs that are not n-words (i.e. that cannot be used as negative fragment answers), such as Catalan gaire ‘much’/ ‘many,’ French qui/quoi que ce soit (lit. who/what it may be) ‘anybody’/ ‘anything,’ Romanian cine știe ce (lit. who knows what) ‘much,’ and the Spanish postnominal indefinite (e.g. persona alguna, lit. person some, ‘anybody’), as shown in (17)–(20). (17) Cat.

a. *(No) he menjat gaire. not have.1SG eaten anything ‘I haven’t eaten much.’ b. Has menjat gaires pomes? have.2SG eaten many apples ‘Have you eaten many apples?’

(18) Fr.

a. Daniel *(n’)a *(pas) rencontré qui que ce soit. meet anybody Daniel NEG . has not ‘Daniel did not meet anybody.’ b. Si quoi que ce soit vous dérange faites-le nous savoir. if anything you bothers let.it us know ‘If anything bothers you, let us know.’ (adapted from Tovena/Déprez/Jayez 2004, 398)

(19) Rom. a. *(Nu) am mâncat cine știe ce la who knows what at not have.1SG eaten ‘I didn’t eat much at lunch.’

prânz. lunch

b. Dacă spui cine știe ce, le vei cauza probleme. if say.2SG who knows what them FUT make trouble ‘If you say anything, you will get them into trouble.’ (Elena Ciutescu, p.c.)

200

(20)

M. Teresa Espinal and Susagna Tubau

Sp.

a. *(No) he visto persona not have.1SG seen person ‘I haven’t seen anybody.’ b. ¿Quién ha dicho cosa who has said thing ‘Who said anything?’

alguna. some

alguna? some

These data show that PIs are dependent on a negative, interrogative or conditional operator, and we can formally represent this dependency by saying that PIs are encoded by a [+σ] semantic feature that must be c-commanded by an abstract σ operator adjoined to another operator in the C(omplementizer) domain, in NegP, or even in triggers of expletive negation (cf. below). For the set of examples given in (17) to (20) we postulate that the nonnegative reading is compositionally driven after an operation of checking between a PI semantically characterized with an inherent [+σ] feature that encodes enrichment and an Opσ, under local conditions that can be formally represented as in (21). In (21) Opσ c-commands the linguistic item specified [+σ], where Op is negative ((17a), (18a), (19a), (20a)), interrogative ((17b), (20b)), or conditional ((18b), (19b)), and the operator freezes the enrichment constraints encoded by [+σ]. (21)

Opσ. . .PI[+σ]

In this chapter we put forward the hypothesis that the [+σ] feature is also inherent in the French negative marker ne, which has been claimed to be a scope marker when it occurs as part of a negative dependency (Godard 2004).13 We entertain the idea that French ne indicates that the SNM pas has sentential scope, which means that ne itself is not a SNM but a mere marker of the limits of the scope of sentential negation. Since the real SNM pas acts as the negative operator in Standard French, a sentence such as (22a) is assigned the structure in (22b), which represents the freezing operation of a [+σ] feature by means of a σ operator attached to pas. In the possible situation in which pas is not explicit, a covert abstract σ operator must be attached to an expletive negation trigger, as in (23).14 Both (22b) and (23b) contain the ingredients to license the negative marker ne.

13 Ne is argued to be a PI in Zeijlstra (2009). 14 For references on the topic of expletive negation, cf. Jespersen (1917); Vendryès (1950); Bosque (1980); Martin (1984); Muller (1991); Espinal (1991; 1992; 2007); Horn (2010), among others.

Meaning of words and meaning of sentences

(22)

(23)

Fr.

Fr.

a.

Jean ne mange Jean NEG eats ‘Jean doesn’t eat.’

b.

[TP Jeani [NegP pasσ [Neg◦ ne[+σ]] [vP ti mange]]]

a.

Je crains qu’ il ne I fear that he NEG ‘I’m afraid he will come.’

b.

[TP je [vP crainsσ [CP que [TP ili [NegP ne[+σ] [vP ti vienne]]]]]]

201

pas. not

vienne. comeS B J V

We also postulate that Romanian, being a Strict NC language, also requires that nu is not a SNM, but a mere marker of the limits of the scope of sentential negation. In accordance with this hypothesis, our analysis for a sentence such as (24a) points at the presence of a covert abstract negative operator that takes sentential scope, to which the σ operator that binds the PI nu is adjoined. The relevant structure for this sentence is given in (24b). (24)

Rom.

a.

Elena nu sună. Elena NEG calls ‘Elena doesn’t call.’

b.

[TP Elenai [NegP Op¬σ [Neg◦ nu[+σ]] [vP ti sună]]]

So far it has been shown that PIs and negative heads like French ne and Romanian nu can be polar and, thus, bear a [+σ] feature. We would now like to extend this hypothesis to so-called expletive negation markers. In particular, we would like to postulate the [+σ] feature as part of the semantic make-up of expletive no ‘not’ in those languages that allow the presence of an overt negative marker that does not modify the truth value of the proposition in which it appears.15 In cases of expletive negation, as shown in (25) and (26) for Catalan and Spanish, respectively, no does not logically negate the proposition, but induces one to consider domains of states of affairs broader than what one would otherwise have considered within the scope of negation (Espinal 2007). Expletive negation is associated with linguistic expressions that constrain the nonveridicality of the context: nonaffirmative verbs (doubt),

15 Expletive, pleonastic or paratactic negation is more residual in some languages than in others. This is the case of Spanish in comparison to Catalan.

202

M. Teresa Espinal and Susagna Tubau

adversative predicates (be surprised), negative prepositions (without), temporal propositions (before), comparatives and superlatives etc. (Horn 2013). The consideration of a broader domain leads to a stronger and more informative proposition. (25) Cat. Gasta més ell en tres mesos que (no (pas)) tu en spends more he in three months than not you in tot l’any. all the.year ‘He spends more money in three months than you in a year.’ (Espinal 1991, 42) (26)

Sp.

Preferiría prefer+COND.1SG

salir go-out

con with

vosotros you

que than

(no) not

estarme be.me

en casa todo el fin de semana. at home whole the weekend ‘I would rather go out with you than stay at home the whole weekend.’ (Espinal 1997, 76) In these examples, similarly to what we have just proposed for the French example in (23a), the expletive no behaves like a PI dependent on the comparative degree adverb més ‘more, better’ or the comparative verb preferir ‘to prefer,’ to which an abstract σ operator is adjoined. It should be noted that expletive no cannot be sensitive to an interrogative or conditional operator because of its strong negative PI status (Zwarts 1981; Hoeksema 2012). In contrast to PIs in general, n-words behave most characteristically like Negative PIs (i.e. they are licensed in the context of a c-commanding negative operator, and are therefore dependent on an antiveridical operator; (6a), (8a)). However, they are also allowed as isolated fragment answers and in preverbal position. Thus, the Romance n-words that we introduced in section 2.1 do not seem to behave like PIs in all contexts. Furthermore, “the fact that n-words may occur as fragment answers with a negative reading makes them look like genuine negative quantifiers, but this is not compatible with the fact that they do not yield a double negation reading when they are used in combination with the sentential negation marker or another n-word” (Labelle/Espinal 2014, 199). N-words have a double-sided behaviour: They are PIs on some occasions, but are interpreted negatively in isolation and in preverbal position. In the model outlined here, this means that these lexical items (roots) bear an inherent semantic feature ([+σ]) that needs to be licensed at the level of abstract meaning representation (e.g. LF), which makes them PIs. However, during the course of the derivation these lexical items can also acquire a negative syntactic feature that will need to be checked in the syntax for a NC reading to be legitimated. The negative feature that

Meaning of words and meaning of sentences

203

n-words have been claimed to be associated with in negative contexts is defined as uninterpretable (henceforth [uNeg]) (Zeijlstra 2004; Biberauer/Zeijlstra 2012), as n-words are assumed to be semantically nonnegative, but syntactically active to participate in NC structures. We assume that n-words start as roots defined [+σ], and that in the course of the derivation these roots can merge with a [uNeg] feature to build a complex item, as in (27).16 Such an operation is optional and arguably takes place to ensure that a NC relationship with an [iNeg] constituent can be established. (27)

Being uninterpretable, a [uNeg] formal feature needs to be checked by an interpretable matching feature, [iNeg]. The relation between these two is one of Agree, namely Reverse / Inverse Agree, as defined in (28), the main characteristic of which is that “the goal may have an uninterpretable feature checked against a higher probe” (Zeijlstra 2012, 491).17 (28)

[Reverse / Inverse] Agree (Zeijlstra 2012, 514) α can Agree with β iff: a. α carries at least one uninterpretable feature and β carries a matching interpretable feature b.

β c-commands α

c.

β is the closest goal to α

16 We thus assume that the [uNeg] feature is not inherent to lexical items in all languages (as it is in Zeijlstra 2004; Labelle/Espinal 2014), but is part of syntax in some of them. Since pure PIs cannot stand on their own as negative fragment answers, we assume that the merge operation represented in (27) is not a possibility with these kinds of elements (e.g. Catalan gaire, Spanish persona alguna), which are considered to be weak PIs. In addition, (27) does not take place in French and Romanian, where n-words are inherently endowed with a [uNeg] feature and no [+σ]. This explains why n-words cannot occur in nonnegative contexts in these languages. 17 The classic definition of Agree (Chomsky 2000; 2001) is the one in (i). Notice that the crucial difference is the c-command relation between α and β (i.e. the probe and the goal). (i) Agree (Zeijlstra 2012, 493; after Chomsky 2000; 2001) α can agree with β iff: α carries at least one unvalued and uninterpretable feature and β carries a matching interpretable and valued feature α c-commands β β is the closest goal to α β bears an unvalued uninterpretable feature

204

M. Teresa Espinal and Susagna Tubau

3 The contribution of n-words in NC contexts Let us see how the definition in (28) allows us to explain the NC data we presented in the previous section. In Non-Strict NC languages like Spanish, the SNM no is assumed to carry an interpretable negative feature, [iNeg], and can, hence, check the [uNeg] feature of n-words by c-commanding them. The relevant configuration for an example like No ha visto nada (lit. ‘not has seen anything’ (8a)) is given in (29a). In turn, No ha visto nada en ningún cajón (lit. ‘not has seen anything in any drawer’), represented in (29b), illustrates that it is possible for more than one n-word to establish an Agree relationship with a licenser equipped with an [iNeg] feature. Syntactically, this is a case of multiple Agree (Hiraiwa 2011; Zeijlstra 2004), where one probe (the negative operator) agrees with more than one goal (the two n-words). Semantically, multiple Agree results in the combination of several expressions of negation into just one, hence facilitating the NC reading of the SNM and the postverbal n-words. (29)

Sp.

a.

[NegP [Neg◦ noσ] [vP ha visto nada[+σ]] [iNeg]

b.

[uNeg]

[NegP [Neg◦ noσ] [vP ha visto nada[+σ] en ningún[+σ] cajón]] [iNeg]

[uNeg]

[uNeg]

The overt SNM is not the only lexical item that can bear an [iNeg] feature, though. Recall that, as illustrated in (9) above, preverbal n-words in Non-Strict NC languages are predicted not to be able to co-occur with the SNM (Nadie (*no) ha visto nada, lit. nobody not has seen anything). How is the [uNeg] feature of the n-word in preverbal position checked? And, more importantly, how is the negative reading of this sentence compositionally driven, then? In other words, what is the licenser of the preverbal n-word? According to Zeijlstra (2004), the presence of uninterpretable negative features triggers the presence of a Last Resort abstract negative operator, Op¬[iNeg], which performs the checking operation, and ultimately guarantees a compositional interpretation. This is shown in (30), the structure corresponding to sentence (8b), which shows that both preverbal and postverbal n-words are underspecified for a negative feature, and therefore for their participation in a NC relationship. (30)

Sp.

[NegP Op¬σ nadie[+σ]i [vP ti ha visto nada[+σ]]] [iNeg]

[uNeg]

[uNeg]

As already mentioned, Spanish –like Italian and European Portuguese– is a NonStrict NC language, and therefore, the structure in (30) –for a sentence with preverbal and postverbal n-words– does not contain an overt SNM.

Meaning of words and meaning of sentences

205

The structures in (29a, b) and (30) should be contrasted with the ones postulated for French, where ne cannot merge with a [uNeg] feature.18 In (31a) –the syntactic structure for the French sentence Jean n’a rien vu in (12b above)– ne is licensed by the semantic σ operator adjoined to a covert Op¬ triggered by the n-word rien, and rien is checked by the syntactic feature [iNeg] that defines Op¬. On the other hand, in (31b) above –the syntactic structure for the French sentence Jean n’a pas rien vu in (13b) above– combining the SNM pas with the French n-word rien yields DN. Within the framework we are developing, we follow Zeijlstra (2009) in assuming that pas is a logical negative operator Op¬ that cannot bear the syntactic feature [iNeg]. Hence, it crucially contributes to the semantics of the sentence by introducing an instance of logical negation, but cannot participate in an Agree relationship with lexical items characterized with the feature [uNeg] (cf. footnote 16). The prediction is that pas does not participate in NC structures.19 The presence of the [uNeg] feature in the n-word triggers the insertion of a Last Resort Op¬[iNeg] that guarantees a successful checking of the uninterpretable feature, but also has an important consequence with regard to how the meaning of the sentence is composed: Pas, which is logically negative (i.e. Op¬), and the rescue operator (i.e. Op¬[iNeg]) cancel each other out, hence yielding a DN reading. (31)

Fr.

a.

[TP Jean [NegP Op¬σ [Neg◦ ne[+σ]] [vP a rien vu]]] [iNeg]

b.

[uNeg]

[TP Jean [NegP Op¬ [NegP pasσ [Neg◦ ne[+σ]] [vP a rien vu]]]] [iNeg]

[uNeg]

In Strict NC languages like Romanian, the presence of an n-word with a [uNeg] feature also triggers the insertion of a Last Resort Op¬[iNeg] that Agrees with the (various) [uNeg] feature(s) present in the sentence. We suggested in (24b) that the negative marker nu fulfils a similar function to French ne, namely signalling that it is a scope marker of sentential negation. Unlike French ne, however, nu can merge with a syntactic [uNeg] feature, which makes it possible for an Agree checking relationship to be established. This surfaces as Strict NC, as shown in (32a), which corresponds to (10) above, but it may also be postulated for postverbal n-words, as represented in (32b), corresponding to the sentence Nu sună nimeni (lit. not calls anybody, ‘Nobody calls’). 18 An argument in support of this claim is offered by *Jean ne mange. This sequence is ungrammatical because ne, which is specified as [+σ] but not as [uNeg], cannot trigger the insertion of a licensing Op¬[iNeg]. Only n-words, specified as [uNeg], can trigger the Last Resort operator. In short, it seems that in French the features [+σ] and [uNeg] do not merge in syntax. 19 As one of the reviewers pointed out, the situation is different in Québécois French, where pas can co-occur with n-words yielding a NC reading (Déprez 1997). In our analysis for this dialect of French pas is specified as [uNeg] and hence it is licensed by means of a Last Resort Op¬[iNeg].

206

(32)

M. Teresa Espinal and Susagna Tubau

Rom.

a.

[NegP Op¬σ nimenii [Neg◦ nu[+σ]] [vP ti sună]] [iNeg]

b.

[uNeg]

[uNeg]

[NegP Op¬σ [Neg◦ nu[+σ]] [vP sună nimeni]] [iNeg]

[uNeg]

[uNeg]

Finally, let us consider the contribution of n-words and no in NC contexts in Catalan, a language that is described as allowing an optional no after preverbal n-words. Catalan differs from Romanian in not having a Strict NC variety: There are no speakers for whom after a preverbal n-word no is obligatory, but still the combination n-word followed by no is a possibility. We explain the data in (6a) and (11) above in the following way: (33)

Cat.

a.

[NegP [Neg◦ noσ] [vP he vist res[+σ]]] [iNeg]

b.

[uNeg]

[NegP Op¬σ ningú[+σ]i [Neg◦ no[+σ]] [vP ti ha vist res[+σ]]] [iNeg] [uNeg]

[uNeg]

[uNeg]

Whereas in (33a) the obligatory SNM is specified with an inherent [iNeg] syntactic feature, which allows checking of the [uNeg] feature on res, in (33b) the optional no is merely a PI, an expletive form similar to the one exemplified in the comparative contexts illustrated in (25) and (26). In (33b) the items that participate in a NC relation, ningú, no, and res, all acquire [uNeg] during the course of the derivation, a feature that matches with a c-commanding [iNeg] formal feature that characterizes a covert Op¬.20 An overall picture of what has been discussed in this chapter regarding the contribution of PIs, n-words, and the negative marker to the meaning of sentences in Spanish, Catalan, French, and Romanian is summarized in the following table. To summarize, we suggest that [iNeg] and [+σ] are formal features that are inherent to some lexical items. While [iNeg] is syntactic and defines a word as inherently negative, [+σ] is a semantic feature associated with PIs and expletive negation. Unlike what has been proposed in the literature, we take the feature [uNeg] to be an inherent formal feature in French and Romanian, but a feature to which a root can merge in syntax to build an n-word in Spanish and Catalan. [uNeg] can be merged to lexical items carrying a [+σ] feature, resulting in the requirement that an Agree syntactic relationship is compulsory. In the case of n-words, the [+σ] feature

20 We would like to hypothesize that the optionality of Catalan no in (33b) vs. the obligatoriness of Romanian nu in (32a) has nothing to do with expletiveness, since they are both characterized as [+σ], but with the fact that in Catalan, but not in Romanian, expletive no has a homophonous negative no. Cf. the table below.

Meaning of words and meaning of sentences

Spanish

Catalan

French

Romanian

Polarity items

N-words

Negative marker

[+σ] (e.g. persona alguna)

[+σ], [uNeg] (e.g. nadie, nada)

Two homophones: 1. no[iNeg] (used in single negation and Non-Strict NC structures) 2. no[+σ] (expletive negation). Residual

[+σ] (e.g. gaire)

[+σ] (e.g. qui/quoi que ce soit)

[+σ] (e.g. cine știe ce)

[+σ], [uNeg] (e.g. ningú, res)

Two homophones: 1. no[iNeg] (used in single negation and Non-Strict NC) 2. no[+σ] (used in expletive negation)

[uNeg] (e.g. personne, rien)

1. pas. Semantic negation, Op¬ (used in single negation) 2. ne[+σ] (used in expletive negation and as a scope marker of sentential negation)

[uNeg] (e.g. nimeni, nimic)

nu[+σ]

207

Can [uNeg] associate with a negative marker with [+σ]?

No.

Yes. It results in no[+σ],[uNeg] (used in what look like Strict NC structures).

No.

Yes. It results in nu[+σ],[uNeg] (used in single negation and Strict NC structures).

accounts for the possibility for them to occur in nonnegative contexts, while the combination of [+σ] and [uNeg] accounts for their ability to participate in NC structures. [+σ] requires an operator σ that freezes the PI, and [uNeg] requires a checking [iNeg] feature that guarantees a NC interpretation. In both cases the relation is one of c-command.

4 Conclusions We started this chapter asking what the meaning of words is and how they contribute to the meaning of sentences, and we introduced the hypothesis that the contribution of words to the meaning of sentences is not conceptual, but underspecified. An underspecified meaning can be attributed both to an inherent semantic

208

M. Teresa Espinal and Susagna Tubau

feature and to a syntactic feature with which a lexical item might merge during a derivation. The Principle of Compositionality is guaranteed by checking both semantic features and uninterpretable syntactic features. This is how underspecification can be understood and formalized within a minimal approach to the compositionality of meaning. PIs are defined as being [+σ]. A prediction that is borne out from this characterization is that PIs are predicted not to occur in preverbal position in declarative clauses, because there is no operator to which the σ operator can attach. We have considered two classes of n-words: those that can only appear in negative contexts (e.g. the situation in French and Romanian), which are endowed with a [uNeg] feature, and those that may appear in nonnegative as well as in negative contexts (e.g. the situation in Catalan and Spanish), which are items inherently defined with a [+σ] feature that can merge with a [uNeg] feature in the course of the derivation. We predict that languages that have n-words of the former group have an independent set of lexical items that encode a [+σ] feature. This prediction is also borne out when we consider French and Romanian. With regard to negative markers the situation we have described is the following. In the case of French pas is a SNM that encodes logical negation (i.e. Op¬); this characterization predicts that this item cannot participate in NC structures because for this to be possible a syntactic correspondence between items characterized [uNeg] and an item characterized [iNeg] is required. By contrast, French ne is a scope marker defined [+σ]; this characterization predicts that it can be used in nonnegative (e.g. expletive) contexts. However, it cannot merge with a [uNeg] feature and trigger the insertion of a Last Resort Op¬[iNeg] that negates the clause. Rather, ne is licensed by expletive negation triggers or, in negative contexts, by pas. Otherwise, the Op¬[iNeg] is triggered by an n-word specified as [uNeg]. Unlike French ne, Romanian nu is specified as [+σ] and can merge with a [uNeg] feature to guarantee a single negation or NC reading. Spanish has two homophonous lexical items: (i) no1, defined [iNeg], which is a SNM that is required to bind postverbal n-words, as expected in a Non-Strict NC language, and (ii) no2, defined [+σ], which is residual in expletive negation contexts. Catalan also has two homophonous lexical items: (i) no1, defined [iNeg], which is a SNM that is required to bind postverbal n-words, as expected in a Non-Strict NC language, and (ii) no2, defined [+σ]. The difference between Spanish and Catalan is that in Catalan the possibility of expletive negation is much more productive than in Spanish. Moreover, no2 can merge with [uNeg] in what look like Strict NC contexts, similar to Romanian, although in Catalan but not in Romanian no2 is optional. This different use of no2 in the two languages predicts that those Catalan speakers with a scarce use of expletive negation are those that generally prefer Non-Strict NC structures, whereas those Catalan speakers with a broad use of expletive negation are those that generally prefer what look like Strict NC structures. We leave it for future research to investigate more extensively whether it is the case that those languages

Meaning of words and meaning of sentences

209

that show a higher use of expletive negation readings correlate with those languages that allow negative readings in what look like Strict NC contexts, and whether the fact that Catalan no2 is optional, but Romanian nu is not, in spite of both being characterized as [+σ], is due to the coexistence of a SNM no1 only in the former language.

5 References Acquaviva, Paolo (1996), “The logical form of negative concord”, University of Venice Working Papers in Linguistics 6, 1–29. Acquaviva, Paolo (1997), The logical form of negation. A study of operator-variable Structures in Syntax, New York, Garland. Biberauer, Theresa/Zeijlstra, Hedde (2012), “Negative concord in Afrikaans. Filling a typological gap”, Journal of Semantics 29, 345–371. Bosque, Ignacio (1980), Sobre la negación, Madrid, Ediciones Cátedra. Carston, Robyn (2012), “Word meaning and concept expressed”, The Linguistic Review 29, 607–623. Chierchia, Gennaro (2006), “Broaden your views. Implicatures of domain widening and the ‘logicality’ of language”, Linguistic Inquiry 37, 535–590. Chomsky, Noam (1981), Lectures on Government and Binding, Dordrecht, Foris. Chomsky, Noam (2000), “Minimalist inquiries. The framework”, in: Roger Martin/David Michaels/ Juan Uriagereka (edd.), Step by step. Essays on Minimalist syntax in honor of Howard Lasnik, Cambridge, MA, MIT Press, 89–155. Chomsky, Noam (2001), “Derivation by phase”, in: Michael Kenstowicz (ed.), Ken Hale. A life in language, Cambridge, MA, MIT Press, 1–52. Corblin, Francis/Déprez, Viviane/de Swart, Henriëtte/Tovena, Lucia (2004), “Negative concord”, in: Francis Corblin/Henriëtte de Swart (edd.), Handbook of French semantics, Stanford, CSLI, 417– 452. Corblin, Francis/Tovena, Lucia (2003), “L’expression de la négation dans les langues romanes”, in: Danièle Godard (ed.), Les langues romanes. Problèmes de la phrase simple, Paris, CNRS, 281– 343. Cruse, David (1986), Lexical semantics, Cambridge, Cambridge University Press. Davidson, Donald (1967), “Truth and meaning”, Synthese 17, 304–323. Déprez, Viviane (1997), “Two types of negative concord”, Probus 9, 103–143. Déprez, Viviane (2011), “Atoms of negation. An outside-in micro-parametric approach to negative concord”, in: Pierre Larrivée/Richard P. Ingham (edd.), The evolution of negation. Beyond the Jespersen Cycle, Amsterdam/Philadelphia, Benjamins, 221–272. Déprez, Viviane/Cheylus, Anne/Larrivée, Pierre (2013), “When and how is NC preferred. An experimental approach”, Paper presented at the 19 e Congrès International des Linguistes, Genève. Déprez, Viviane/Tubau, Susagna/Cheylus, Anne/Espinal, M. Teresa (2015), “Double negation in a negative concord language. An experimental investigation”, Lingua 163, 75–107. Espinal, M. Teresa (1991), “On expletive negation. Some remarks with regard to Catalan”, Linguisticae Investigationes 15, 41–65. Espinal, M. Teresa (1992), “Expletive negation and logical absorption”, The Linguistic Review 9, 338–358. Espinal, M. Teresa (1996), “On the semantic content of lexical items within linguistic theory”, Linguistics 34, 109–131. Espinal, M. Teresa (1997), “Non-negative negation and wh-exclamatives”, in: Danielle Forget/Paul Hirschbühler/France Martineau/María Luisa Rivero (edd.), Negation and polarity, Amsterdam/ Philadelphia, Benjamins, 75–93.

210

M. Teresa Espinal and Susagna Tubau

Espinal, M. Teresa (2002), “La negació”, in: Joan Solà/Maria Rosa Lloret/Joan Mascaró/Manuel Pérez-Saldanya (edd.), Gramàtica del Català Contemporani, vol. 3, Barcelona, Empúries, 2727– 2797. Espinal, M. Teresa (2007), “Licensing expletive negation and negative concord in Catalan and Spanish”, in: Franck Floricic (ed.), La négation dans les langues romanes, Linguistica e Investigationes Supplementa 26, Amsterdam/Philadelphia, Benjamins, 49–74. Espinal, M. Teresa/Prieto, Pilar (2011), “Intonational encoding of double negation in Catalan”, Journal of Pragmatics 43, 2392–2410. Espinal, M. Teresa/Tubau, Susagna/Borràs-Comes, Joan/Prieto, Pilar (2015), “Double negation in Catalan and Spanish. Interaction between syntax and prosody”, in: Pierre Larrivée/Chungmin Lee (edd.), Negation and polarity. Cognitive and experimental perspectives, Berlin, Springer, 145–176. Farkas, Donka (1990), “Two cases of underspecification in morphology”, Linguistic Inquiry 21, 539– 550. Floricic, Franck (2005), “La négation dans les langues romanes”, Lalies 25, 163–194, Actes des Sessions de Littérature et Linguistique (Aussois, 23–28 août 2004), Paris, Presses de l’École Normale Supérieure. Fodor, Jerry A. (1983), Modularity of mind. An essay on faculty psychology, Cambridge, MA, MIT Press. Frege, Gottlob (1892), “Über Sinn und Bedeutung”, Zeitschrift für Philosophie und philosophische Kritik 100, 25–50. Geeraerts, Dirk (2010), Theories of lexical semantics, Oxford, Oxford University Press. Giannakidou, Anastasia (1998), Polarity sensitivity as (non)veridical dependency, Amsterdam/Philadelphia, Benjamins. Giannakidou, Anastasia (2000), “Negative. . .concord?”, Natural Language and Linguistic Theory 18, 457–523. Giannakidou, Anastasia (2006), “N-words and negative concord”, in: Martin Everaert/Henk van Riemsdijk (edd.), The Blackwell Companion to Syntax, vol. 3, Malden, MA, Blackwell, 327–391. Godard, Danièle (2004), “French negative dependency”, in: Francis Corblin/Henriëtte de Swart (edd.), Handbook of French Semantics, Stanford, CSLI, 351–390. Hacquard, Valentine (2006), Aspects and modality, PhD dissertation, Massachusetts Institute of Technology. Higginbotham, James (1985), “On semantics”, Linguistic Inquiry 16, 547–594. Hiraiwa, Ken (2011), “Multiple Agree and the defective intervention constraint in Japanese”, MIT Working Papers in Linguistics 40, 67–80. Hoeksema, Jack (2012), “On the natural history of negative polarity items”, Linguistic Analysis 38, 3–33. Horn, Laurence R. (1989), A natural history of negation, Chicago, University of Chicago Press. Horn, Laurence R. (2010), “Multiple negation in English and other languages”, in: Laurence R. Horn (ed.), The expression of negation, Berlin/New York, Mouton de Gruyter, 111–148. Horn, Laurence R. (2013), “Revisiting the licensing question. Some negative (and positive) results”, Paper held at the 19e Congrès International des Linguistes, Genève. Hornstein, Norbert (1995), Logical Form. From GB to Minimalism, Oxford, Blackwell. Jackendoff, Ray S. (1972), Semantic interpretation in Generative Grammar, Cambridge, MA, MIT Press. Jakobson, Roman (1984a), “The structure of Russian case forms”, in: Linda Waugh/Morris Halle (edd.), Russian and Slavic grammar. Studies 1931–1981, Berlin/New York, de Gruyter, 105–133. Jakobson, Roman (1984b), “The structure of the Russian verb”, in: Linda Waugh/Morris Halle (edd.), Russian and Slavic grammar. Studies 1931–1981, Berlin/New York, de Gruyter, 1–14. Jespersen, Otto (1917), Negation in English and other languages, Copenhagen, Høst.

Meaning of words and meaning of sentences

211

Kadmon, Nirit/Landman, Fred (1993), “Any”, Linguistics and Philosophy 16, 353–422. Kratzer, Angelika (1981), “The notional category of modality”, in: Hans-Jürgen Eikmeyer/Hannes Rieser (edd.), Words, worlds, and contexts, Berlin/New York, de Gruyter, 38-74. Kratzer, Angelika (2012), Modals and conditionals, Oxford, Oxford University Press. Kratzer, Angelika (2013), “Modality for the 21st century”, in: Stephen R. Anderson/Jacques Moeschler/ Fabienne Reboul (edd.), L’interface language-cognition. The language-cognition interface, Genève/Paris, Droz, 179–199. Krifka, Manfred (1994), “The semantics and pragmatics of weak and strong polarity items in assertions”, in: Mandy Harvey/Lynn Santelmann (edd.), Proceedings from Semantics and Linguistic Theory IV, Ithaca, Cornell University, 195–219. Labelle, Marie/Espinal, M. Teresa (2014), “Diachronic changes in negative expressions. The case of French”, Lingua 145, 194–225. Labov, William (1972), “Negative attraction and negative concord in English grammar”, Language 48, 773–818. Ladusaw, William (1980), Polarity sensitivity as inherent scope relations, New York, Garland. Lahiri, Utpal (1998), “Focus and negative polarity in Hindi”, Natural Language Semantics 6, 57–123. Laka, Itziar (1990), Negation in syntax. On the nature of functional categories and projections, PhD dissertation, Massachusetts Institute of Technology. Levin, Beth/Pinker, Stephen (edd.) (1991), Lexical and conceptual semantics, Cambridge, MA, Blackwell. Lumsden, John S. (1992), “Underspecification in grammar and natural gender”, Linguistic Inquiry 23, 469–486. Martin, Robert (1984), “Pour une approche sémantico-logique du ne dit ‘explétif’”, Revue de Linguistique Romane 48, 99–120. Martins, Ana Maria (1998), “On the need of underspecified features in syntax. Polarity as a case study”, GLOW Newsletter 40, 46–47. Matos, Gabriela (1999), “Negative concord and the scope of negation”, Catalan Working Papers in Linguistics 7, 175–190. May, Robert (1985), Logical Form. Its structure and derivation, Cambridge, MA, MIT Press. Montague, Richard (1974), Formal philosophy. Selected papers of Richard Montague, ed. Richmond H. Thomason, New Haven, Yale University Press. Muller, Claude (1991), La négation en français. Syntaxe, sémantique et éléments de comparaison avec les autres langues romanes, Genève, Droz. Pelletier, Francis Jeffrey (2011), “Compositionality”, Oxford Bibliographies in Linguistics, doi: 10.1093/obo/9780199772810-0044 (Annotated Bibliography on “Compositionality”, Oxford Bibliographies Online). Prieto, Pilar/Borràs-Comes, Joan/Tubau, Susagna/Espinal, M. Teresa (2013), “Prosody and gesture constrain the interpretation of double negation”, Lingua 131, 136–150. Quer, Josep (1998), Mood at the interface, PhD dissertation, Universiteit Utrecht. Reinhart, Tanya (1983), Anaphora and semantic interpretation, London, Croom Helm. Rigau, Gemma (1984), “De com si no és conjunció i d’altres elements interrogatius”, in: Estudis Gramaticals. Working Papers in Linguistics, Bellaterra, Universitat Autònoma de Barcelona, 249–278. Rooryck, Johan (1994), “On two types of underspecification. Towards a feature theory shared by syntax and phonology”, Probus 6, 207–233. Sánchez, Cristina (1999), “La negación”, in: Ignacio Bosque/Violeta Demonte (edd.), Nueva Gramática Descriptiva de la Lengua Española, vol. 2, Madrid, Espasa-Calpe, 2561–2634. Solà, Joan (1973), “La negació”, in: Estudis de sintaxi catalana, vol. 2, Barcelona, Edicions 62, 87– 118.

212

M. Teresa Espinal and Susagna Tubau

Sperber, Dan/Wilson, Deirdre (1986/1995), Relevance. Communication and cognition, Oxford, Blackwell. de Swart, Henriëtte (2010), Expression and interpretation of negation. An OT typology, Dordrecht, Springer. de Swart, Henriëtte/Sag, Ivan (2002), “Negation and negative concord in Romance”, Linguistics and Philosophy 25, 373–417. Tarski, Alfred (1944), “The semantic conception of truth and the foundations of semantics”, Philosophy and Phenomenological Research 4, 341–375. Teixeira de Sousa, Lílian (2012), Sintaxe e interpretação de negativas sentenciais no Português Brasileiro, PhD dissertation, Universidade Estadual de Campinas, Campinas. Tovena, Lucia/Déprez, Viviane/Jayez, Jacques (2004), “Polarity sensitive items”, in: Francis Corblin/ Henriëtte de Swart (edd.), Handbook of French Semantics, Stanford, CSLI, 391–416. Tubau, Susagna (2008), Negative concord in English and Romance: syntax-morphology interface conditions on the expression of negation, PhD dissertation, Universitat Autònoma de Barcelona and University of Amsterdam. Tubau, Susagna/Espinal, M. Teresa (2012), “Doble negació dins l’oració simple en català”, Estudis Romànics 34, 145–164. Vendryès, Joseph (1950), “Sur la négation abusive”, Bulletin de la Société de Linguistique de Paris 46, 1–18. Werning, Markus/Hinzen, Wolfram/Machery, Édouard (edd.) (2012), The Oxford handbook of compositionality, Oxford, Oxford University Press. Wilson, Deirdre/Sperber, Dan (1993), “Linguistic form and relevance”, Lingua 90, 1–25. van der Wouden, Ton (1994), Negative contexts, PhD dissertation, University of Groningen. Zanuttini, Raffaella (1997), Negation and clausal structure. A comparative study of Romance languages, New York, Oxford University Press. Zeijlstra, Hedde (2004), Sentential negation and negative concord, PhD dissertation, University of Amsterdam. Zeijlstra, Hedde (2009), On French negation, Ms. University of Amsterdam, http://ling.auf.net/ lingbuzz/000885 (23.07.2015). Zeijlstra, Hedde (2012), “There is only one way to agree”, The Linguistic Review 29, 491–539. Zwarts, Frans (1981), “Negatief polaire uitdrukkingen I”, GLOT 4, 35–133. Zwarts, Frans (1995), “Nonveridical contexts”, Linguistic Analysis 25, 286–312.

Eva-Maria Remberger

7 Morphology and semantics: Aspect and modality Abstract: This chapter investigates the morphosyntactic and interpretive interactions between modality, mood, aspect and tense in Romance. I present an overview on common definitions of aspect, tense, mood and modality, applying these notions to the tense-aspect system and the exponents of mood and modality in Romance. To illustrate the interaction of grammatical levels (lexicon – morphosyntax – semantics – discourse) at the interfaces, a series of case studies from Romance are discussed: First, aspectual interactions regarding the stage-level/individual-level distinction; second, mood and modal meaning at the interface; and, third, the interactions of tense, aspect, and modality, and the shifted interpretations they produce. Keywords: aspect, mood, modality, tense, deixis, stage-level, individual-level, time relations, sequence of tense, interpretative shift

1 Introduction The main focus of this chapter is the interaction between structure and meaning in the field of two primarily semantic notions, aspect and modality. Aspect is a grammatical category related to temporal contour. It cannot be studied, at least not as a phenomenon relevant to interfaces, without taking into consideration the grammatical category of tense and its interpretive correlates. This is true to an even greater extent for Romance, where aspect scarcely exists as a stand-alone morphological category of verbs. Modality is a grammatical category through which a speaker can refer to possible worlds. It may have an exponent in Romance verbal morphology, namely mood, but it can also be encoded by other linguistic means. This chapter investigates the morphosyntactic and interpretive interactions between modality, mood, aspect, and tense in Romance.1 In section 2, I present an overview of common definitions of aspect (cf. section 2.1), tense (cf. section 2.2), and mood and modality (cf. section 2.3), applying these notions to the tense-aspect

1 Several recent publications have dealt with the subject at issue here (cf. Hogeweg/de Hoop/ Malchukov 2009) as have papers in edited volumes, e.g. Filip (2011), Hacquard (2011), Maienborn (2011), Smith (2012), Binnick (2012), Hamm/Bott (2014), and particularly Depraetere (2012), and Zagona (2012a; 2012b). Zagona (2012a; 2012b) are both based on a generative framework and Zagona (2012a) in particular can be read as a complement to the present article.

214

Eva-Maria Remberger

system and the exponents of mood and modality in Romance. Section 3 presents some case studies of interface phenomena between aspect, tense, and mood and modality which provide a good illustration of the interplay of morphosyntactic means, semantic encoding, and pragmatic interpretation.

2 Definitions 2.1 Aspect2 In Comrie (1976, 3, 5), aspect is described as “different ways of viewing the internal temporal constituency of a situation”.3 Aspect is a “temporal sub-system” as is tense, i.e. tense and aspect are “complementary domains” (cf. Smith 2012, 2581, 2605): While tense locates events in time with respect to the speech event, aspect defines their temporal quality. Aspect describes the internal temporal contour of the event situation expressed by the predication of a clause at the time that the speaker takes as the point of reference (cf. also the notion of viewpoint aspect in Smith 1997). Thus, in some models of the linguistic expression of time, aspect has been described as a second temporal relation (e.g. Giorgi/Pianesi 1997; Demirdache/Uribe-Etxebarria 2000, based on Reichenbach 1947 and Vikner 1985). Essential for the description of aspectual meaning is the notion of an event argument, as introduced by Davidson (1967).4 Events are referential objects in the world and have spatiotemporal properties (Partee 1973).5 A relevant distinction in the theory of aspect is that between lexical aspect (or Aktionsart) and grammatical aspect, expressed by verbal morphology (as in the well-known aspectual system of the Slavonic languages) or by verbal periphrases (more common in Romance):

2 A useful overview of different theories of aspect is given in the lecture notes by Bhatt/Pancheva (2005). 3 This definition is based on Holt (1943). 4 Davidson’s (1967) ‘event argument’ has also been expressed as ‘situation’ (Comrie 1985) or ‘eventuality’ (Bach 1986). Both terms are used in order to avoid the notion of ‘event’, which did not initially include states (Davidson 1967 only considered activities). The notion of eventuality or situation is instead intended to include (temporal) states. For the concept of event in its Neo-Davidsonian interpretation, cf. Maienborn (2011) for discussion; for event structure cf. also Rothstein (2004). 5 Other approaches interpret the encoding of time by using operators that apply to interpretation indices (starting from Prior 1967; Dowty 1979; Montague 1973; also Enç 1987; Guéron/Hoekstra 1995; Lohnstein 22011). For several reasons, the referential approach to tense and temporal interpretation has proven to be more appropriate, in particular in the generative framework (for some of these reasons, cf. the overview in Zagona 2012a, 748–753).

Morphology and semantics: Aspect and modality

(1)

215

Lexical and grammatical aspect a. Lexical aspect6 refers to the internal temporal contour of the event as provided by the lexical properties and the argument structure inherent to the predicate. b.

Grammatical aspect7 externally modifies or extends the temporal properties of the event in the situation the speaker refers to (usually) by grammatical means.

With reference to lexical aspect, Vendler (1967) distinguishes four classes of verbs, characterized by combinations of the aspectual features [±dynamic], [±telic], and [±durative]:8 (2)

a.

states [-dynamic, -telic, +durative], e.g. Fr. posséder ‘to own’, are not dynamic, have no end point and are characterized by duration. If the state P holds for an event situation, it also holds for every subinterval of this situation.

b.

activities [+dynamic, -telic, +durative], e.g. Fr. danser ‘to dance’, are dynamic (their subject is agentive), they last and do not have an inherent end point. If the activity P holds for an event, P is true also for every subinterval of this event.

c.

accomplishments [+dynamic, +telic, +durative], e.g. Fr. faire un gâteau ‘to bake a cake’, are dynamic (agentive) and telic, and before the end point is reached there is a phase of duration. Since the event of an accomplishment P is inhomogeneous, it cannot be subdivided into subintervals for each of which P holds.

d.

achievements [+dynamic, +telic, -durative], e.g. Fr. arriver ‘to arrive’, are dynamic and have a punctual end point which is not preceded by duration. Thus for an achievement P the subinterval property does not hold.

The classical tests to distinguish verb classes from each other proposed by Vendler (1967) and illustrated by Dowty (1979, 51–56) are as follows:9

6 Also called Aktionsart, situation aspect (Smith 1997; 2012), inner aspect, or aspectual class (Dowty 1979). 7 Also called viewpoint aspect (Smith 1997; 2012), or outer aspect. 8 Another feature would be [±homogeneous], which is dependent on the subdivisibility property and usually has the opposite value to [±telic]. Instead of [±durative] we also find its counterpart [±punctual]. 9 This is a selection; for details, cf. also Walková (2012), who gives a critical survey of Dowty’s (1979) tests.

216

Eva-Maria Remberger

(3)

States do not allow progressive aspect: Sp. *Juana está poseyendo un coche azul. ‘*J. is owning a blue car.’

(4)

Activities allow durative adverbials: It. Gianni ballava per un ora. ‘G. danced for one hour.’

(5)

Accomplishments and achievements allow telic/completive adverbials: a. Fr. Pierre a fait un gâteau en une heure. ‘P. baked a cake in one hour.’ b.

(6)

Por.

A Maria chegou em duas horas. ‘M. arrived in two hours.’

Achievements allow punctual adverbials: Rom. Ana a ajuns chiar în acea clipă. ‘A. arrived exactly at that moment.’

That lexical aspect is tightly connected to argument structure is immediately clear from the illustration of accomplishments, which, at least in Romance, usually come with a complement (cf. the direct object un gâteau in (2c) and (5a)). However, it is not only the presence of a complement that is relevant to a verbʼs class distinction; its semantic properties are too. Some verbs, the so-called incremental theme predicates, can be interpreted as an activity (i.e. [+telic]) or an accomplishment (i.e. [-telic]), depending on the referential properties of the complement DP. If the DP is cumulative, we have an activity (7a); if it is quantized, the aspect class changes into an accomplishment (7b).10 The role of the reference type of nominals and quantification in the temporal constitution of events has been extensively studied in the work of Krifka (1989; 1992; 1998), where he has shown that there is a similarity between nominal and verbal expressions insofar as they can be subdivided into parts. Nominal expressions are cumulative when they can be divided into equal subparts (as is the case for mass nouns or bare plurals like des gâteaux, cf. 7a); otherwise they are quantized (like le gâteau d’anniversaire, cf. 7b, or also deux gâteaux ‘two cakes’). This is referred to as the quantizing property. Verbal expressions encode events which either can be subdivided into subintervals (the subinterval property, as described above in 2a, b for states and activities), or cannot be subdivided in this way (cf. 2c, d for accomplishments and achievements). Smith also includes a fifth verb class, different from achievements, namely semelfactives like Fr. cesser ‘to stop’, i.e. “single-stage atelic events” (Smith 1997, 220). Furthermore, nonargumental elements, e.g. adverbs or adverbials, can also influence 10 As for French, cf. in particular Smith (1997, ch. 9, 193–222).

Morphology and semantics: Aspect and modality

217

class membership as in (7c), where the adverbial de repente ‘suddenly’ changes the state verb saber ‘to know’ into an achievement (cf. Zagona 2012b, 356; Smith 1997). It. starnutire ‘to sneeze’, an achievement (or semelfactive), can be turned into an activity of sneezing by an adverbial encoding duration like per tre ore ‘for three hours’ (7d): (7)

a.

Fr.

faire des gâteaux

‘to bake cakes’

b.

Fr.

faire le gâteau d’anniversaire

‘to bake the birthday cake’

c.

Sp.

saber algo de repente

‘suddenly (=get to) know something’

d.

It.

Gianni ha starnutito per tre ore.

‘G. sneezed for three hours.’

The distinction between verb class attribution and grammatical aspect is characterized by different types of coercion, as (7d) shows, where the VP is interpreted as an iteration of events (a plurality of punctual events which can be described as a single activity-event).11 Coercion phenomena also depend on the vagueness of linguistic expressions (such as the aspectual neutrality of the present tense in French, Italian, and Romanian), and have sometimes been analyzed as type shifting phenomena.12 In any case, aspectual interpretation is compositional and involves the lexicon,13 the event and argument structure, the referential properties, as well as morphosyntactic marking (grammatical aspect). States must be further distinguished into two types (cf. Maienborn 2003 and the references therein), which roughly correspond to the distinction between stage-level and individual-level predicates (analyzed by Kratzer 1995 but with earlier origins in Milsark 1974 and Carlson 1977): (8)

Stage-level (SL) and individual-level (IL) predicates a. SL-predicates hold at a specific reference time and thus appear as temporally bound (e.g. Fr. se situer ‘to be situated’). b.

IL-predicates are inherent to the element to which they refer and are not temporally bound to a specific reference time, thus holding generically (e.g. Fr. posséder ‘to own’, savoir ‘to know’).

11 Cf. Maienborn (2011, 824); cf. also Bertinetto (1986) for a study of Italian in this respect. 12 That is, aspect shift or hidden coercion operators (Zucchi 1998; de Swart 2011). 13 At this point, approaches to lexical decomposition must be mentioned, cf. Jackendoff (1990), where a lexical entry like Fr. tuer/It. uccidere/Sp. matar is decomposed into semantic-aspectual subcomponents, which represent aspectual primitives: [Event ACT([Entity x], CAUSE [BECOME [NOT [ALIVE([Entity y])]]])]; cf. also Pustejovsky’s (1991; 1995) lexical event structure and Parsons’ (1990) approach, which includes the aspectually relevant links between thematic roles and their arguments.

218

Eva-Maria Remberger

Kratzer (1995) assumes that IL-predicates have no event variable, whereas others follow the neo-Davidsonian approach that all predicates have event arguments.14 However, the distinction between SL and IL is not only lexical but also depends on the properties of the reference time for which the situation holds: If the reference time is specific (existentially quantified), the SL-reading obtains; if it is generic (generically quantified), the IL-reading arises (cf. section 3.1 for further discussion in a case study).15 Verb classes can then be further manipulated not only by the presence and by the properties of complements, but also by grammatical aspect (or aspect proper).16 Grammatical aspect interacts with tense, since both involve the reference time, i.e. the time the speaker refers to (cf. the discussion in section 2.2). Typical semantic interpretations of grammatical aspect in Romance concern perfectivity vs. imperfectivity, with the latter being able to encode further specifications such as the progressive (9a), continuative (9b), habitual etc. (cf. Comrie 1976); other aspectual notions are prospectivity or imminentiality (9c): (9)

a.

progressive, e.g. Sp. estoy comiendo ‘I’m eating’

b.

continuative, e.g. It. i prezzi andavano aumentando ‘the prices were increasing (continuously/step by step)’

c.

imminential, e.g. Fr. aller faire quelque chose, Sp. ir a hacer algo ‘to be going to do something’

Generative approaches to aspect sometimes reformulate the traditional aspectual values through a specification of time-relational notions (cf. Demirdache/UribeEtxebarria 2000; Giorgi/Pianesi 1997; Iatridou 2000; Zagona 2012a; 2012b). Others, like Smith (1997; 2012), strictly separate tense and aspect as two independent systems.

2.2 Tense Unlike aspect, tense is deictic, in the sense that in order to be interpreted it needs a direct reference to the language external world. In terms of semantic approaches to tense and its interpretation in the sense of time, two main proposals have been put 14 Cf. e.g. Chierchia (1995), who distinguishes location-dependent and location-independent events (cf. Maienborn 2011, 817). 15 The literature on this distinction is vast and growing; cf. e.g. the references in Maienborn (2003). For copula selection in Romance, which is dependent on this distinction, cf. also González-Vilbazo/ Remberger (2005; 2007), among many others. 16 When a state verb like Sp. saber appears in the indefinido, the synthetic past characterized by perfective grammatical aspect, a shift from state to (inchoative) achievement similar to (7c) occurs (cf. also Bhatt/Pancheva 2005, 27; de Swart 2011, 591).

Morphology and semantics: Aspect and modality

219

forward: In the tense operator approach, logic tense operators were introduced as part of the formal language describing the evaluation of the truth conditions of a proposition (Prior 1967; Montague 1973). However, times do not always show operator-like (scopal) behaviour: Times can be referred to and the occurrence of tenses in natural language is restricted to a finite inventory; furthermore, they are context-dependent, which would contradict their relevance at the level of truth conditions alone (Zagona 2012a, 749–751). For the purposes of this paper, only the time-relational approach will be examined, since it has proved more suitable for the modelling of the correlations of tense with aspect and modality. This approach analyzes times as argument-like, referential entities. Generative theory relies basically on the time-relational approach proposed by Reichenbach (1947): (10) Times a. E = event time, the time of the event or situation b. R = reference time, the time the speaker refers to, which can be made explicit by temporal adverbials c. S = speech time, the time of the utterance These times can be temporally ordered either in a relation of precedence (marked by the underscore) or in a relation of simultaneity (marked by the comma): (11)

Time relations (cf. Reichenbach 1947) S,E,R (present), E,R_S (imperfect), S_R,E (future), E_R,S (perfect),

E_R_S (pluperfect),

E_S_R, S,E_R, S_E_R (anterior future)

Vikner (1985) splits Reichenbach’s monolithic time relations into two separate relations, one between the event time and the reference time (E/R) and one between the reference time and the speech time (R/S), thus yielding two independent and only indirectly related time relations (which also gives a more reasonable representation of the various interpretations of an anterior future): (12) T1 / T2 a. [S,R]/[R,E] (present):

Sp. Juan come/está comiendo manzanas.

b. [R_S]/[R,E] (past):

Sp. Juan comía/comió manzanas.

c. [S_R]/[R,E] (future):

Sp. Juan comerá manzanas.

d. [S,R]/[E_R] (perfect):

Sp. Juan ha comido manzanas.

e. [R_S]/[E_R] (pluperfect):

Sp. Juan había comido manzanas.

f. [S_R]/[E_R] (anterior future): Sp. Juan habrá comido manzanas. ‘J. eats/is eating, ate, will eat, has eaten, had eaten, will have eaten apples.’

220

Eva-Maria Remberger

The perfect here is understood as a past event situation which is still relevant in the present, as is the case for Spanish; what is called perfect tense in Romance is not always used with the meaning of a perfect in this sense: The Italian compound perfect, the passato prossimo, in (13a), for example, clearly has the meaning of past, as in the representation in (12b) (the reference time is five years in the past relative to S). Spanish would choose the indefinido in this case (13b): (13)

a.

It.

Ho visto questo ritratto cinque anni fa.

b.

Sp.

Vi este retrato hace cinco años. ‘I saw this portrait five years ago’

However, all Romance languages have a distinction between a perfective and an imperfective past (cf. Sp. comía vs. comió in 12b). Spanish also has a grammaticalized generic reference time for the present (cf. the encoding of states or habitual events), whereas for a specific reference time the progressive periphrasis must be used (cf. also Bertinetto/Delfitto 1996; Squartini 1998, as well as the discussion in section 3.1). The list in (12) is not exhaustive; cf. e.g. the combination of [S,R]/[R_E], which would be a representation of a prospective future (or aspect) like (14a), or periphrases like the surcomposé (cf. Grevisse/Goosse 152011, 1090–1092; De Saussure/Sthioul 2012) in French (14b): (14)

Fr.

a.

Je vais manger quelque chose. ‘I’m going to eat something.’

b.

J’ai été parti. [I have.1SG be.PTCP leave.PTCP ]

Subsequent work based on the two time relations, like Klein (1994), Giorgi/Pianesi (1997), and Demirdache/Uribe-Etxebarria (2000; 2007; 2008), has sometimes distinguished between a syntactically higher T-relation (=T1) and a lower one (=T2), which has more aspectual than tense features.17 However, as can be seen in particular in (12a) and (12b), where Spanish has more than one verbal form encoding not only temporal but also different aspectual values, the temporal system according to Reichenbach (1947) and Vikner (1985) alone is not sufficient to unambiguously distinguish these aspectual values. Independently of whether we assume that the referential time is an argument proper or just a featural encoding of tense in the corresponding syntactic categories, the time-relational approach has been successfully adopted for the modelling of the interaction of the morphosyntax of tense and the interpretation of time. The role of a temporal anchor (the anchoring principle; cf. Enç 1987, 642), located in the

17 The relevant terminology varies here: Klein (1994) distinguishes between Time of Utterance, Topic Time, and Time of the Situation; Demirdache/Uribe-Etxebarria (2000; 2007; 2008) call them Utterance Time, Assertion Time, and Event Time.

Morphology and semantics: Aspect and modality

221

CP-domain, is also essential in this respect. This temporal anchor either provides the deictic reference of the speaker’s temporal coordinates or allows a temporally nonindependent clause to be temporally and referentially anchored (in the sense of Enç 1987) to a root clause. Direct and indirect tense relations also play a role in complex (biclausal) sentences, where sequence of tense (henceforth SOT) phenomena can be observed (cf. section 3.2 for a case study). As we have already seen, tense and aspect in Romance are not distinct morphological categories, as they are in Slavonic, but are marked syncretically in verbal inflection. The main aspectual distinction is that between imperfective and perfective. There are simple and compound tenses and a series of innovative verbal periphrases with aspectual values (GO-, COME-, STAND-, HOLD-, BE-, HAVE-periphrases). With regard to the simple tenses, the present and the imperfect express imperfective aspect. The simple future, meanwhile, which is sometimes interpreted not as a tense but as a modal category (or, at least, a category of uncertainty; cf. also Smith 2012, 259118), is neutral with respect to (im)perfectivity (at least in French, Italian, and Romanian). Compound tenses, i.e. the perfect, the pluperfect,19 and the anterior future, built with an auxiliary verb and the participle in most Romance languages, express perfective aspect. Most Romance languages also have a synthetic perfective past, like the Spanish indefinido, the Italian passato remoto, the French passé simple, the Portuguese pretérito perfeito simple, and the Romanian perfect simplu, all of which express perfective aspect.20 These tenses are derived from the old Latin perfect, which was used for both the values which now in some Romance languages are expressed by two different tenses: the time relation [R_S]/[R,E] (+perfective), for which in standard Spanish the synthetic perfect is used, and [S,R]/[E_R] (+perfective), which is now encoded by the compound perfect. This means that the Spanish indefinido encodes perfectivity in the past, whereas the compound perfect, the perfecto compuesto, encodes perfectivity that still holds in the present. In Romanian, French, and Italian (but not Southern Italian), however, the compound perfect can be used for both time relations, with the old synthetic perfect restricted to a purely literary register. Turning to the present tense, it must be noted that Spanish is quite strict in distinguishing a present tense which must obligatorily come with an unbound or generic reference time (15a) from a periphrastic aspectual construction with a specific reference time (15b) (cf. also the case study in section 3.1). This is less strict in Italian 18 Cf. also Lyons (1977, 677) quoted by Smith (2012): “Futurity is never a purely temporal concept; it necessarily includes an element of prediction or some related notion.” 19 In all Romance languages except Romanian the pluperfect is a compound tense; in Portuguese, there is a compound as well as a synthetic pluperfect, but the latter is only used in literary texts (cf. Gärtner 1998, 27). 20 In Catalan there is a synthetic and a more commonly used analytic preterite with auxiliary GO + infinitive; cf. Perea (2002, 630–631, 640–641).

222

Eva-Maria Remberger

(among other languages), which also has this periphrasis (15c; cf. also section 3.1), but is more flexible in the use of the present tense: (15)

a.

Sp.

Juan come mucho.

‘(In general) J. eats a lot.’

b.

Sp.

Juan está comiendo.

‘J. is eating (right now).’

c.

It.

Gianni sta mangiando.

‘G. is eating (right now).’

The gerund in Romance usually encodes imperfectivity, whereas the participle may encode perfectivity (but see the present passive). The infinitive is only indirectly tense related, since it has an underspecified first T-relation (cf. also Remberger 2006, 126–128). Thus a present infinitive can get a simultaneous, but also a posterior (future-oriented) reading (this is relevant for the discussion of modal expressions that embed an infinitive; cf. the case studies in section 3.3). A compound infinitive with the participle encodes anteriority and perfectivity. The substitute for the infinitive in Romanian, the subjunctive, is also temporally dependent and has a nonanterior and an anterior form.

2.3 Modality and mood Mood and modal systems are described in Palmer (²1990; ²2001) as “two ways in which languages grammatically deal with an overall category of modality” (Palmer ²2001, 4), but not all languages offer both possible ways. Mood is a morphosyntactic category that is not present in all languages but does clearly feature in Romance and has several values: the indicative (the default), the subjunctive, the conditional (cf. fn. 25), the (second person sg. and pl.) imperative, and some rarer values such as the presumptive in Romanian (cf. Squartini 2005; Academia Română 2008, 373–378; Pană Dindelegan 2013, 53–55). In Portner’s (2009, 1) words, modality is “the linguistic phenomenon whereby grammar allows one to say things about, or on the basis of, situations which need not be real”. In formal semantics, this is illustrated by the concept of sets of (possible) worlds. Modality semantically and pragmatically modifies a proposition by introducing sets of worlds which can, but need not, include the actual world. The subset of worlds where the proposition is held to be true is represented as a set of world variables bound by a quantifier. Modality in Romance can be encoded by morphological mood (↗5 Inflectional verb morphology), but also by functional and lexical means like modal verbs, modal adverbs and particles, modal nouns and adjectives, and other modal expressions. Based on Kratzer’s seminal work on modality and its description in possible worlds semantics (cf. Kratzer 1977; 1981; 1991), the following basic elements have been identified as the semantic “ingredients of modality” (cf. Remberger 2010;

Morphology and semantics: Aspect and modality

223

2011): First, modality consists of a modal relation or modal force (cf. Hacquard 2011), which can be a relation of possibility or necessity.21 Possibility can be described as existential quantification over a set of possible worlds, i.e. for at least one possible world w the proposition is held to be true. Necessity universally quantifies over a set of possible worlds, i.e. in all of them p is true. Another ingredient of modality is the conversational background against which the modal relation is evaluated. Conversational backgrounds, also called modal bases, are context-dependent sets of worlds. Modal bases of the type “as far as what is known” lead to an epistemic modal interpretation, while typical conversational backgrounds like “as far as what is needed, as far as what is feasible” lead to deontic (=root) interpretations of modality. Conversational backgrounds and modal bases are a means to formally describe the high context sensitivity of modally modified propositions. (16) It.

Piero deve lavorare molto. ‘P. has to/must work a lot.’

In (16) the modal relation is necessity and the conversational background can either include worlds containing rules, responsibilities, or prescriptive norms controlling Piero’s work, or it can represent the knowledge world that tells us, for instance, that somebody looking tired like Piero probably works a lot (for further discussion, cf. also the case studies in section 3.3). Depending on the conversational background the interpretation is either deontic (root) or epistemic. There is often an implicit source22 for the modal relation, and sometimes an explicit one, possibly linked to the modal base. An explicit source is most obvious for intensional verbs like Fr. vouloir/It. volere/Sp. querer, with the subject being the source of modality (cf. the case studies in section 3.3, where this observation becomes particularly relevant). Implicit sources of modality are more frequent, and might include a generic arbitrary reference or some referent retrievable only in the context, possibly the speaker. Finally, there is a target over which modality takes scope, usually a proposition. Another grammatical category, often subsumed under modality (e.g. by Palmer ²1990; ²2001) is evidentiality. Evidentiality is usually defined as the indication of the source of the information that the speaker gives in his assertion (Aikhenvald 2004). Evidentiality can be direct (speaker’s own experience), indirect (hearsay, reported), or inferential (deduced knowledge) (cf. Willett 1988; Palmer ²2001). The latter subtype overlaps with epistemic modality insofar as the epistemic state of the speaker is at stake. So with both epistemic modality and inferential evidentiality, the evaluation of the proposition is based on the conversational background of knowledge

21 For more fine-grained distinctions, cf. Palmer (21990; 22001) and also Depraetere (2012). 22 The notion of a ‘modal source’ was introduced by Calbert (1975).

224

Eva-Maria Remberger

worlds (of the speaker), but only modality is concerned with quantification over these worlds (for a case study concerning the shift of an imperfect tense to indirect evidentiality, cf. section 3.2). With regard to mood, the indicative is selected in assertions. The indicative establishes referentiality (cf. Zafiu 2013, 43) and, as discussed above, indicates that the proposition must be evaluated against a world representing the reality according to the speaker. The subjunctive, in contrast, seems to be an indicator of the fact that (i) the proposition must be interpreted independently of the speaker’s world model (the conversational background anchored to the speaker),23 and (ii) it must be interpreted dependently on other world models/modal bases (irrealis, emotive, potential etc.) that are introduced by the context, usually by a subordinating verb. In main clauses, the nonindicative moods usually introduce models of wish worlds (the optative), of (past or present) counterfactual worlds (the conditional), or of future worlds requiring the action of the hearer or the discourse participants (the imperative, the exhortative; cf. also Bosque 2012, 377). However, in contrast to the subordinate context, these models are always anchored to the speaker. Some groups of lexical items and particular sets of grammatical contexts select clauses in the subjunctive depending on modal operators in a broad sense. Which functional or lexical category introduces “states of affairs conceived through the angle of some evaluation, possibility, necessity, emotion, intention, causation, and other nonfactual or nonvericonditional [semantics]” is determined languagespecifically in the lexicon (cf. Bosque 2012, 379) and thus varies among Romance languages. Intensional verbs are quite stable crosslinguistically in that they obligatorily select the subjunctive.24 For other lexical classes this is not so clear: In Italian, for example, an epistemic verb such as credere ‘to believe’ selects the subjunctive, but in modern colloquial use it appears with the indicative (cf. Wandruszka 1991, 434), while in the other Romance languages examples such as Spanish creer ‘to believe’ usually take the indicative unless negated or in interrogative contexts (Bosque 2012, 379). Moreover, in French, Spanish, and Italian, epistemic verbs of uncertainty like ‘to doubt’ (Fr. douter, It. dubitare, Sp. dudar) usually take the subjunctive, but Romanian a se îndoi (cf. Zafiu 2013, 45) does not. The conditional as a form of the verbal paradigm can have the temporal futurein-the-past reading (exemplified later in section 3.2, 25) and thus is often treated

23 The notion of ‘world model’ was introduced by Farkas (1992) and Giannakidou (1998) and further formalized by Quer (1998). Models are part of the context (thus comparable to the conversational background) of an utterance and are anchored to individuals (Giannakidou 1998). 24 When there is subject identity, the infinitive is selected in all Romance languages but Romanian. The phenomenon whereby the pronominal subject of an embedded clause must be different in reference from the subject of the main clause in sentences like It. vuole che parta ‘He wants him to leave’ (cf. Fr. *je veux que je parte, Ruwet 1984) has been called the Obviation Effect.

Morphology and semantics: Aspect and modality

225

basically as a tense,25 although it manifests itself in a present and a past form. Furthermore, it has modal meaning in conditional constructions where the verbal form of the conditional introduces possible worlds (in the present or in the past) but some propositions belonging to these possible worlds (which can be expressed by an if-clause, e.g. It. canterei se potessi ‘I would sing if I could’26) logically restrict the truth of the proposition. Syntactically, morphological mood has either been located in the TP-domain, since it is connected to finiteness, or somewhere between the TP and the CP (cf. Zagona 2012b). In some constructions with verbal forms marked by the subjunctive, by the conditional, and in particular by the imperative, verb movement takes place. With respect to finiteness, there seems to be a hierarchy or gradually decreasing scale between finiteness and nonfiniteness, depending on mood morphology (cf. also Remberger 2006, 119). In finite subjunctive subordinate clauses, the Romanian subjunctive introduced by să is a finite form but – unusually in Romance – it also appears in obviation contexts, i.e. where the other Romance languages would insert an infinitive (which is rare in Romanian27). Nonfinite but inflected subordinate clauses are found in Portuguese and Sardinian, which have an inflected infinitive (derived from the Latin imperfective subjunctive).

25 The conditional mood represents a Romance innovation, as there was no conditional in Latin. It is synthetic in the present (except in Romanian), but is historically derived from a periphrasis which designated future-in-the-past. Bosque (2012) treats the conditional as a tense; Zagona also calls it a tense, but a tense encoding a modal feature, cf. Zagona (2012b, 365); in French, it is treated as a – past or present – tense form of the indicative, cf. Grevisse/Goosse (152011, 1027); in Italian, the conditional is classified as a mood, cf. Renzi/Salvi/Cardinaletti (1995); it is considered a mood in Portuguese by Gärtner (1998), but not by Mateus et al. (1983, 149, fn. 152); cf. Zafiu (2013, 51), where she refers to Thieroff (2010), who typologically distinguishes between two types of conditionals, the ‘Eastern conditional’, which is a mood proper, and the ‘Western conditional’, which he and others consider a tense. In Romanian, the conditional is periphrastic (e.g. aș vedea/aș fi văzut ‘I would see/I would have seen’). Only the Romanian conditional is a morphological mood proper, since it does not have the temporal future-in-the-past reading of the other Romance languages – in fact, it is also called condiţional-optativ. The optative meaning is derived from the conditional meaning, e.g. Rom. aș veni și eu ‘I would come, too’ > ‘I wished I could come, too (if some outer conditions obtain)’. 26 Note that in Standard Italian the protasis is in the past subjunctive and only the main clause is in the verbal form of the conditional. 27 Thus, in contexts where there is subject identity between the main and embedded clause, the Romanian subjunctive bears the function that other Romance languages encode with the infinitive (e.g. Rom. vreau să merg acasă ‘I want that I go home’ for ‘I want to go home’); cf. fn. 24. The subjunctive is in general (with the exception of some highly frequent verbs) inflectionally visible only in the third person. However, the Romanian subjunctive is always introduced by the subjunctive particle să (i.e. the subjunctive is always marked; the particle must be adjacent to the verbal form, separated from it only by clitics or negation).

226

Eva-Maria Remberger

Modality is located either in the TP- or in the CP-domain, depending on the language-specific encoding of modality, modal bases, and syntactic movement operations (cf. Cinque’s 1999 hierarchy of modal heads). As several studies have shown (e.g. Picallo 1990; Laca 2005; 2006, and many others), epistemic modality is located higher than root modality; cf. the following examples: (17) Sp. a. Este material debe poder resistir temperaturas muy altas, ¿no? ‘This material is probably able to resist very high temperatures, isn’t it?’ b. ‘It’s inferable from the extant information that this material is able to resist very high temperatures.’/ ‘In all worlds which are compatible with the speaker’s beliefs this material is able to resist very high temperatures.’ c. *root modality > epistemic modality

(following Laca 2006, 116)

(18) Cat. a. En Pere deu poder tocar el piano. (Picallo 1990, 294) ‘P. is probably able to play the piano.’ b. *En Pere pot deure tocar el piano. The first modal in (17a), debe, can only be interpreted related to an epistemic conversational background (17b). The modal verb deure in Catalan in (18) does not allow a deontic interpretation (for deontic necessity the verb haver de would have to be used; cf. Picallo 1990), thus in (18a) it comes first and is interpreted epistemically, whereas in (18b), in second position, it is uninterpretable. In combinations with two modals, the higher model must always be interpreted epistemically.

3 Interface phenomena: Romance case studies A first case study of the interaction of different grammatical levels is the IL- vs. SLdistinction (cf. section 3.1). The second set of case studies relate to mood at the interface (cf. Quer’s 1998 title) and its interplay with modal interpretation, two issues that are particularly relevant in the Romance languages, since they have complex mood systems (cf. section 3.2). Finally, I will present studies concerning the interaction of aspect, tense, and modality and the different types of interpretative shifts that arise depending on grammatical as well as contextual factors. These are particularly revealing for Romance, where modal verbs (in contrast to English) still fully inflect (cf. section 3.3).

Morphology and semantics: Aspect and modality

227

3.1 Aspect at the interface: the stage-level vs. individual-level distinction It has been stated that the distinction between SL- and IL-predicates is related to the speaker’s view of the reference time and is therefore an aspectual distinction. However, there are predicates in the lexicon that are semantically more appropriate for IL-readings and others that are more appropriate for SL-readings.28 Nevertheless, the lexical aspectual property can be further manipulated by other modules of the grammar. In Spanish, the SL/IL-distinction is grammaticalized in the choice between two copula verbs:29 Copulative constructions with the Spanish copula ser are ILconstructions, while with the copula estar they represent SL-constructions. ILconstructions hold for an unbound/generic reference time, whereas SL-constructions refer to a specific reference time. Example (19) shows that typically IL-predicates can be coerced to SL-predicates and vice versa: (19) Sp. a. ¡Qué inteligente que estás hoy! [inteligente is typically IL] ‘In what an intelligent manner you are behaving today!’ [SL-reading] b. Soy triste de tanto estarlo. ‘I’m a sad person for always feeling sad.’

[triste is typically SL] [IL-reading & SL-reading]

Inteligente in (19a) is typically an inherent property of individuals whereas triste normally designates a temporary state. Nevertheless, grammatical aspect (i.e. the aspectual copula selection) can change the prototypical lexical aspect.30 The interesting point here is that grammar can intervene at different levels to make a construction SL or IL. First, as mentioned previously, lexical aspect can intervene: Some predicates tend to be SL (adjectives like Sp. borracho ‘drunk’, cansado ‘tired’ etc.), while others are inherently IL (adjectives like Sp. inteligente ‘intelligent’, rubio ‘blond’ etc.). The copula estar has the property of introducing a specificity feature for the reference time of the clause, whereas with ser the reference time remains unspecified (or generic). Apart from the lexical level, grammatical aspect for SL-interpretations can be produced in many Romance languages through verbal periphrases such as gerund periphrases: (20) It.

Michele sta mangiando. / Sard. Michele est manikende.

Sp. Miguel está comiendo.

/ Cat.

En Miquel està menjant. ‘M. is eating.’

28 Nouns are never SL-interpretable, as opposed to adjectives, which can be both. 29 As it is, with a slight variation in the selectional criteria, in Portuguese and Catalan; cf. GonzálezVilbazo/Remberger (2007). 30 Cf. the coercion phenomena illustrated in de Swart (2011), especially the use of the so-called “active be” in English, which can be seen as parallel to the use of estar in (19a).

228

Eva-Maria Remberger

Whether these constructions encode an additional aspectual value or not depends on the language at issue. In Spanish, for example, this periphrasis has an additional interpretative value of duration, which can be bound, and it can thus appear in perfective tenses, whereas in (Modern) Italian the gerund periphrasis does not have a durative value and can appear only in imperfective tenses (cf. Bertinetto/Delfitto 1996; Squartini 1998; also Remberger 2006, 216–226 and 272–280 for a formalization of Italian and Sardinian data): (21)

Sp. It.

Miguel ha estado comiendo. ‘M. has been eating (for some time).’ *Michele è stato mangiando.

Furthermore, we also find examples of the interaction between grammatical and lexical aspect, such as the perfective-imperfective distinction and the IL/SL-property in this example: (22)

Sp.

a.

El portero del equipo era/*fue chileno. ‘The goal keeper was Chilean.’

[chileno is typically IL] [IL-interpretation]

b.

El portero del equipo fue chileno hasta que renunció a su nacionalidad [. . .] [SL-interpretation] ‘The goal keeper was Chilean until he renounced his citizenship. . .’ (cf. Zagona 2012b, 352, from De Miguel 1999)

Chileno is a typical IL-predicate and thus combines well with an imperfective ILcopula (ser) in the past. For a perfective form of the IL-copula a special context is required to make the clause acceptable. If the pragmatic context allows an ILpredicate to be combined with a perfective past reading so that the reference time from the speaker’s viewpoint can no longer be generic (since it is temporally bound), a SL-reading arises: Here, the inherent property of being Chilean becomes a temporal property which can be subject to change (i.e. the event is coerced from a state to a (terminative) achievement).

3.2 Mood and tense at the interface: subjunctive alternation, SOT, and free indirect discourse In complex sentences involving subordination, interface phenomena connected to the interplay between mood and tense (categories that express different degrees of finiteness) and other domains of grammar become relevant. The exchange of grammatical information between embedded and root clause and the interpretation of the embedded component with respect to the speaker coordinates will be discussed in reference to three phenomena: the interpretational effects of mood alternation in embedded relative clauses, SOT phenomena in complement clauses, and the

Morphology and semantics: Aspect and modality

229

use of imperfective verb forms in order to mark free indirect discourse, which is apparently no longer a morphosyntactically embedded context, but is somehow semantically selected. For these interface phenomena, the notion of world models as introduced by Farkas (1992) and Giannakidou (1998) and discussed by Quer (1998) (cf. also fn. 23) will be essential. In his work on mood at the interface, Quer (1998) conceives mood morphology as a grammatical means to manipulate world models. The main focus of his study is the function of the subjunctive in Romance (his data are principally from Catalan and sometimes from Spanish). In Romance, with some intralinguistic variation, the selection of subjunctive forms can be lexically conditioned (e.g. for the complements of certain intensional verbs), semantically conditioned (expressions of two different interpretations), or conditioned by operator scope (as in the case of negation or in interrogatives). There are several configurations where either the subjunctive or the indicative can be chosen: (23) Cat. a. El degà no creu [que els estudiants es mereixinsubjunctive un premi]. ‘The dean doesn’t believe that the students merit a prize.’ b. El degà no creu [que els estudiants es mereixenindicative un premi]. ‘The dean doesn’t believe (the fact) that the students merit a prize.’ Following Quer (1998), in (23a) the subjunctive facilitates an interpretation where the content of the embedded clause is independent from an evaluation against the world model of the common ground (the realistic world model according to the epistemic knowledge of the speaker). It is particularly the presence of the indicative, the referential mood, which in (23b) causes an additional model back-shift from a model of believe-worlds related to the subject of the embedding verb to the model of epistemic knowledge anchored to the speaker. The indicative, even when embedded, is able to refer to the speech context of the main clause. This causes the embedded proposition to be interpreted as true according to the epistemic model anchored to the speaker in (23b), whereas it is false according to the epistemic model anchored to the subject of the embedding verb; in (23a), with an embedded subjunctive, there is no such direct link to the speaker’s world model. Thus, even if morphological mood does not have a functional meaning on its own, it has interpretative effects in combination with other modules of the grammar, namely the insertion of semantic operators of modality as well as negation (as is the case in (23)) and interrogation. Another well-known example showing the alternation of subjunctive and indicative with an interpretative effect is the following: (24)

Sp.

a.

Juan busca (a) una secretaria que sepasubjunctive inglés. [P px] ‘J. is looking for a secretary who should know English.’

b.

Juan busca *(a) una secretaria que sabeindicative inglés. [px P] ‘J. is looking for a (certain) secretary who knows English.’

230

Eva-Maria Remberger

Here, in (24a) the subjunctive in the restrictive relative clause gives us the so-called de dicto reading, i.e. the indefinite DP which is the object of the intensional verb buscar ‘to look for’ is not specific and arbitrary (therefore Differential Object Marking by the preposition a is not obligatory; cf. (↗8 (In)definiteness, specificity, and differential object marking). The referent of secretaria ‘secretary’ only exists in those worlds in which the subject’s (i.e. Juan’s) desire of finding one is fulfilled (existential quantification px within the scope of the intensional operator of the predicate P buscar ‘to look for’). In example (24b), on the other hand, the indicative directly links the indefinite DP (which is obligatorily marked by the preposition) to the world model of the common ground of the root clause (the so-called de re reading), i.e. the interpretation of the DP is existentially quantified and it has a specific reference, since it is outside the scope of the intensional verb buscar ‘to look for’. The alternation between subjunctive and indicative has a specific interpretative effect, namely that, with the indicative, the argument is outside the scope of the operator and thus is directly connected to the speaker coordinates, in order to establish referentiality. If the intensional operator is within the scope of the embedding verb, no specific reading of the indefinite DP is instantiated and the finite form of the embedded verb is morphophonologically realized as a subjunctive form. As shown above (cf. section 2.2), tenses and the interpretation of times and the relations between them need to be anchored: Main clauses are anchored in the speech context, while embedded clauses usually take the event or the reference time of the main clause as their anchor (cf. Enç 1987). The SOT phenomena are a natural expression of this dependency: They clearly show that tenses can have relative (anterior, posterior, simultaneous) as well as absolute (past, present, future) values; they also show that the instantiation of temporal reference is usually established via the root clause. (25) Sp. a. Juan dice que José cantaráfuture a las tres. (Zagona 2012b, 358) ‘J. says that J. will sing at 3 o’clock.’ b. Juan dijo que José cantaríafuture-in-the-past a las tres. ‘J. said that J. would sing at 3 o’clock.’ In (25), the temporal connection (the time relation) between the root and embedded clauses is one of posteriority, i.e. the time interpretation of the embedded clause is posterior to that of the root clause. In (25a), the posteriority relation is clearly represented by the deictic future tense (cantará). In (25b), the same relation must be expressed by the conditional, because the main verb encodes a past perfective tense (dijo, in the indefinido). The tense of the embedded clause thus must follow SOT and to express the posteriority relation it needs to shift the future (the deictic expression for posteriority) to a relative future in the past, i.e. the conditional.31 31 The conditional in this case is a morphophonological tense proper, since it encodes [S_R]/[R,E] but independently of a main clause [R_S]/[R,E] (cf. section 2.2).

Morphology and semantics: Aspect and modality

231

If the temporal relation between root and embedded clause is simultaneous, this means that, morphophonologically, the same deictic tense has to appear in both clauses. Thus (26) is grammatical and (29) below is ungrammatical. Even if there is an explicit deictic reference time relative to S in the embedded proposition, like anterior ieri, simultaneous oggi, or posterior domani, the tense in the embedded clause must follow the SOT rules: (26) It. Gianni sperava che Maria partissepast.subjunctive ierianterior/oggisimultaneous/ domaniposterior.32 ‘G. hoped that Mary would leave yesterday/today/tomorrow.’ (Giorgi 2010, 34) However, an embedded tense is able to establish temporal reference directly, depending on the type of clause (adjunct clauses, relative clauses, complement clauses) and the language at issue. In the so-called double access reading (cf. Abusch 1997) an embedded tense (the first time relation), again in the indicative, can be directly anchored in the discourse: (27) a. It.

Gianni ha detto che Maria èpresent incinta. ‘G. said that M. is pregnant.’

b. Sp. María dijo que Pedro estápresent enfermo. ‘M. said that P. is ill.’

(Giorgi 2010, 13)

(Zagona 2012b, 367)

In (27), from Italian and Spanish, the main clause is in the past and the embedded clause is in the present tense: The time-relational interpretation of the present tense in the subordinate clause, if not explicitly specified otherwise by a temporal adverbial, is directly (deictically) anchored (D-anchored following Demirdache/UribeEtxebarria 2007; 2008) to the time of the speech act S (28a), and it is indirectly (anaphorically) anchored (A-anchoring following Demirdache/Uribe-Etxebarria 2007) via the event time of the root clause Eroot (i.e. the time of the speaking event of Gianni/ María) (28b). That is, for (27b) both (28a) and (28b) hold: (28) a. Rroot_S/Rroot,Eroot – S,Rsubordinate/Rsubordinate,Esubordinate

AND

b. Rroot_S/Rroot,Eroot – Eroot,Rsubordinate/Rsubordinate,Esubordinate The difference in meaning is that the ill person was ill at the time the subject of the main clause was speaking (28b), and that the person is also ill at the speech time (28a). In examples like the following from Giorgi (2010, 34), the Italian verb sperare

32 The verbal form is in the subjunctive since It. sperare is a verb that lexically induces mood.

232

Eva-Maria Remberger

‘to hope’ introduces a new evaluation model (of hopes, anchored to the subject of the main verb). In this case, there is no double access reading available and it is impossible to embed a present tense subjunctive (like parta) under a past tense subjunctive (the correct version is (26) with partisse): (29) It.

*Gianni sperava che Maria partapresent.subjunctive .

In Romanian, the double access reading with an embedded indicative (complement) clause is not obligatory with the present, i.e. in a sentence like (30) the embedded present tense can be deictic or anaphoric to the matrix tense: It can mean just (28b) but not necessarily (28a) (Giorgi 2010, 13; Zafiu 2013, 63). (30) Rom. Ion a zis că Maria epresent însărcinată. ‘I. said that M. is pregnant.’ Therefore, in Romanian the embedded present tense does not need to be used deictically, whereas in Italian and Spanish it does. In order to exclude (28b) for Italian or Spanish, the embedded tense must follow the SOT rules and be realized in the imperfective: (31) a. It.

Gianni ha detto che Maria eraimperfect incinta. ‘G. said that M. was pregnant.’

b. Sp. María dijo que Pedro estabaimperfect enfermo. ‘M. said that P. was ill.’ This is, in principle, also possible in Romanian, but represents the marked option (cf. Zafiu 2013, 64: the imperfect is necessarily deictic33): (32) Rom. Ion a zis că Maria eraimperfect însărcinată. ‘I. said that M. was pregnant.’ Thus, languages vary in this respect and not all languages obligatorily follow SOT. Another phenomenon that is prominent in many Romance languages is the use of the imperfect to mark reported discourse, which is introduced without a subordinating verb of saying: (33) Fr. Brigitte ouvrit la porte du petit salon et nous appela: ne voulionsimperfectnous pas un peu de thé? Cela nous réchaufferait aprés cette course. ‘B. opened the door of the small parlour and called to us: wouldn’t we like some tea? It would warm us up after this run.’ (Grevisse/Goosse 152011, 543) 33 In Spanish and Italian, an embedded clause in the imperfect can also have a double access reading (Zagona 2012b, 368).

Morphology and semantics: Aspect and modality

233

This phenomenon, called free indirect discourse, shows a shift from tense to evidentiality. The morphological tense of the imperfective (i.e. the verbal form voulionsnous) is not interpreted as the realization of an aspectual time relation ([R_S]/[R,E], (+imperfective)), but it marks a world model of reported speech that has not been explicitly introduced by a lexical verb of saying. The anchor to which the model is connected must be retrieved in the context (in this case it is Brigitte). Note, again, that the verbal form réchaufferait is in the conditional, since it is connected to the model of reported speech in a posteriority relation, but the reference time, which must also be retrieved in the context, is before the actual speech time.34 The tense phenomena observable in free indirect discourse obey the same SOT rules as in indirect speech, where an explicit and tensed verb selects a complement clause.35 A related phenomenon is the intentional (conative) reading of the imperfect (Zagona 2012b, 363): (34) Sp. Hasta ayer, íbamosimperfect a la playa de vacaciones pero hoy Pepa dijo que no hay dinero para eso. ‘Until yesterday we were going to the beach on holiday [= we planned to go] but today P. says that there is no money for that.’ Here too, the use of the imperfect signals the shift to a model of evaluation other than the one connected to the speaker’s coordinates: The verbal form íbamos signals an eventuality in the past, but the imperfective aspect in combination with a verbal phrase which is not inherently a state or a process (ir a la playa ‘to go to the beach’ is [+telic]; it is an accomplishment) changes the eventuality expressed into an imperfective eventuality, and the conative interpretation that the accomplishment was not completed therefore arises. The effect is due to the presence of a (past) world model of intentions anchored to the speaker and his/her group.36

3.3 The interaction of aspect, tense, and modality: conditional shift, modal shift, actuality shift, future shift This section examines the following phenomena: the conditional shift for an otherwise uninterpretable reading, the deontic/root vs. epistemic interpretation of modals, 34 Again, here, the conditional is the expression of a tense proper, a future in the past. 35 However, there is an essential difference between indirect speech and free indirect discourse: In the latter discourse markers (referring not to the speaker but to the anchor retrievable in the discourse) and other deictic expressions are permitted. 36 Another non-tense-reading of the imperfect is its use as a form of politeness with modal verbs of volition, e.g. Rom. voiam să vă întreb ceva ‘I wanted to ask you something’ (Zafiu 2013, 61). Moreover, in many Romance dialects and varieties, not only the subjunctive and the conditional, but also the imperfect can express counterfactuality, i.e. the introduction of a set of irrealis worlds.

234

Eva-Maria Remberger

the so-called implicative reading or actuality shift of modal interpretation (cf. Quer 1998; Hacquard 2006; Laca 2006), and the use of the future tense as a device to express epistemic modality. Root modality is future oriented, or, at least, nonanterior (cf. Quer 1998; also Remberger 2010, for the interplay of tense and volitionality). As a result, a modal verb embedding an anterior time relation might result in ungrammaticality: (35) a. It. *Gianni vuole aver lavorato. ‘G. wants to have worked.’ b. Sp. *María quiere haber asistido a la reunión. (Laca 2006, 120) ‘M. wants to have assisted at the meeting.’ Example (35) is ungrammatical because the modal verbs It. volere/Sp. querer ‘to want’ are restricted by a nonanteriority condition, i.e. the model of wish worlds anchored to the subject excludes past worlds, since past worlds are epistemically stable (cf. Laca 2006), i.e. they are truth-conditionally evaluable.37 However, the situation changes if the verb introducing modality is marked by mood morphology indicating the introduction of a further alternative world model, namely a model of conditional or irrealis worlds (excluding the actual world, the realistic model): (36) a. It.

Gianni vorrebbeconditional aver lavorato. ‘G. would like to have worked.’

b. Sp. Quisiera/querríaconditional haber llegado a tiempo. (Laca 2006, 119) ‘S/he would have liked/would like to have arrived in time.’ This also neutralizes the epistemic stability of the past encoded in the embedded sentence, since it is now in the scope of a modal operator; the sentence is therefore grammatical as a result of the conditional shift. However, with modal verbs where the source of modality is not explicit (other than volitional WANT), the modal relation can shift from a deontic/root to an epistemic modal base. This modal shift is a well-known effect present in more than just the Romance languages. In (37a), the modal reading is the deontic (root) reading, but in (37b), with a complex embedded infinitive, which encodes anteriority (the second time relation [E_R], or, aspectually speaking, perfectivity), an epistemic and posterior reading appears: (37) Cat. a. En Jordi no ha pogut sortir. [root reading] ‘J. was not able to go out.’

(Picallo 1990, 287)

b. En Jordi pot no haver sortit. [epistemic reading] ‘J. could not have gone out.’ 37 Note that the parallel construction in German is not ungrammatical since in German volitional modality can shift to an evidential reading; cf. Remberger (2010; 2011).

Morphology and semantics: Aspect and modality

235

However, the model shift from deonticity to epistemicity not only arises with embedded perfective aspect. The same happens if the embedded clause is marked as SL:38 (38) Sp. Juan puede estar comiendo. ‘J. might be eating.’

[epistemic reading]

Furthermore, there are still many examples where the reading remains ambiguous (especially with IL-verbs) and can only be established by the temporal reference retrievable in the context (cf. the ambiguous ex. (16) in section 2.3).39 As in the conditional shift, the tense and the aspectual features of the modal can also give rise to interpretative effects:40 (39) Cat. Va voler que el seu fill estudiés Belles Artes. (Quer 1998, 50, fn. 48) ‘S/he wanted her/his son to study fine arts (and he did).’ (40) It. Lizzie ha voluto parlare a Darcy, #ma non gli ha parlato. (Hacquard 2006) ‘L. wanted to (and did) talk to D., #but she didn’t talk to him.’ In (39) and (40) the modal verb is in the compound/periphrastic perfect and thus encodes a perfective reference time in the past. The modal base, the model of wishes according to the volition of the subject of the main verb, contains nonanterior worlds. Nevertheless the aspectual marking of the main volitional verb makes the embedded proposition true not only in the worlds that are included in the volitional model connected to the subject, but also according to the realistic world model anchored to the epistemic knowledge of the speaker. This might be explained by the fact that the volitional model also contains nonposterior, simultaneous worlds, and simultaneity with an event marked as perfective at the reference time gives rise to an interpretation in which the embedded proposition belongs to the actual world (the actuality shift; cf. Hacquard 200641). Thus, it would be strange to continue a sentence like (40) with an afterthought that explicitly negates the realization of the modalized proposition in the actual world.42 38 Cf. Depraetere (2012, 1011–1012) for English. 39 To be exact, the examples with embedded perfective infinitives could, in principle, also have a deontic/root reading, especially with explicit temporal adverbials referring to a posterior reference time and thus coercing a future-oriented but perfective modal meaning. However, the epistemic reading is the most natural when no other context is given. 40 This is a difference between Romance and English, since in the latter modals are grammaticalized auxiliaries which do not inflect (only some modals have past forms, often with a specialized interpretation). 41 The imperfect instead allows, but not necessarily forces, a “failed attempt” reading. 42 In Romanian a vrut (3SG compound perfect of WANT) has no actuality implication, but a putut (3SG compound perfect of CAN) and a trebuit (impersonal compound perfect of MUST) have.

236

Eva-Maria Remberger

The last shift to be discussed here is the epistemic interpretation of the future. It has been noted that there is an “asymmetry between a fixed past and an open future” (cf. Kaufmann/Condoravdi/Harizanov 2005, 99) – the future is always metaphysically uncertain (cf. also Copley 2002 and fn. 18). The future as well as the future in the past (the anterior future) can have an epistemic interpretation. A model of future “realistic” worlds (where the future is epistemically unstable) is easily substituted by a model referring to present epistemically anchored, but existentially quantified worlds: (41) a. It.

Saranno le tre.

‘It may be three o’clock.’

b. Sp. Serán las tres. (42) a. It.

Sarà stato il postino.

‘It may have been the postman.’

b. Sp. Habrá sido el cartero. In the simple future tense there is usually no modal relation. However, an epistemic meaning arises when the time-relational interpretation shifts from [S,R]/[R_E] to [S,R]/[R,E] and the future event is no longer interpreted as being located after the reference time and speech time, but is understood as simultaneous with the speech situation. This goes hand in hand with a shift from a temporal interpretation to an interpretation that evaluates the proposition against the model of epistemic knowledge of the speaker and quantifies over it: In at least one of the worlds compatible with the epistemic knowledge of the speaker it is true that p.43 In Romanian, where the future is not synthetic, various kinds of (more or less colloquial) future periphrases have developed, and these also often have an epistemic reading. A particular case is the so-called presumptive in (43) (cf. also Squartini 2005), which is a mood originating in a future periphrasis. (43) Rom. a. Acum o fi dormind/va fi dormind, că nu văd lumină. ‘(S)he may/might be sleeping now, as I can’t see any light.’ b. O fi rezolvat toate problemele, că pare mulțumit. ‘He may/might have solved all his problems, as he looks satisfied.’ (cf. Zafiu 2013, 54) The Romanian presumptive is an analytic form consisting of a verb form derived from ‘want’ (o, va) and either an infinitive or fi plus gerund (cf. Mihoc 2014, 65; Fălăuș 2014, 109). Its meaning has become epistemic or evidential (cf. Squartini 2005; Pană Dindelegan 2013). 43 Of course, another well-known change in interpretation, not only in Romance, is the opposite development, i.e. the grammaticalization of modal verbs into future auxiliaries or markers (cf. Fleischmann 1982; Bybee/Perkins/Pagliuca 1994, ch. 7), where an element expressing futureoriented modal meaning becomes a future tense marker.

Morphology and semantics: Aspect and modality

237

4 References Abusch, Dorit (1997), “Sequence of tense and temporal de re”, Linguistics and Philosophy 20, 1–50. Academia Română (2008), Gramatica limbii române, 2 vol., Bucureşti, Editura Academiei Române. Aikhenvald, Alexandra (2004), Evidentiality, Oxford, Oxford University Press. Bach, Emmon (1986), “The algebra of events”, Linguistics and Philosophy 9, 5–16. Bertinetto, Pier Marco (1986), Tempo, aspetto e azione nel verbo italiano, Firenze, Accademia della Crusca. Bertinetto, Pier Marco/Delfitto, Denis (1996), “Lʼespressione della ‘progressività / continuità’. Un confronto tripolare (italiano, inglese, spagnolo)”, in: Paola Benincà/Guglielmo Cinque/Tullio De Mauro/Nigel Vincent (edd.), Italiano e dialetti nel tempo. Saggi di grammatica per Giulio C. Lepschy, Roma, Bulzoni, 45–66. Bhatt, Rajesh/Pancheva, Roumyana (2005), “LSA 130. The syntax and semantics of aspect”, Lecture notes, MIT, http://web.mit.edu/rbhatt/www/lsa130/l1.pdf (04.01.2015). Binnick, Robert I. (ed.) (2012), The Oxford handbook of tense and aspect, Oxford, Oxford University Press. Bosque, Ignacio (2012), “Mood. Indicative vs. subjunctive”, in: José Ignacio Hualde/Antxon Olarrea/ Erin O’Rourke (edd.), The handbook of Hispanic linguistics, Hoboken/New York, Wiley, 373– 394. Bybee, Joan/Perkins, Revere/Pagliuca, William (1994), The evolution of grammar. Tense, aspect, and modality in the languages of the world, Chicago, University of Chicago Press. Calbert, Joseph (1975), “Toward the semantics of modality”, in: Heinz Vater/Joseph Calbert (edd.), Aspekte der Modalität, Tübingen, Narr, 1–70. Carlson, Gregory N. (1977), Reference to kinds in English, PhD dissertation, University of California at Irvine. Chierchia, Gennaro (1995), “Individual predicates as inherent generics”, in: Gregory N. Carlson/ Francis Jeffry Pelletier (edd.), The generic book, Chicago/London, University of Chicago Press, 176–223. Cinque, Guglielmo (1999), Adverbs and functional heads. A cross-linguistic perspective, New York/ Oxford, Oxford University Press. Comrie, Bernd (1976), Aspect, Cambridge, Cambridge University Press. Comrie, Bernd (1985), Tense, Cambridge, Cambridge University Press. Copley, Bridget Lynn (2002), The semantics of the future, PhD dissertation, Cambridge, MA, MIT. Davidson, Donald (1967), “The logical form of action sentences”, in: Nicholas Rescher (ed.), The logic of decision and action, Pittsburgh, University of Pittsburgh Press, 81–120. De Miguel, Elena (1999), “El aspecto léxico”, in: Ignacio Bosque/Violeta Demonte (edd.), Gramática descriptiva de la lengua española, vol. 2: Las construcciones sintácticas fundamentales. Relaciones temporales, aspectuales y modales, Madrid, Espasa-Calpe, 2977–3060. Demirdache, Hamida/Uribe-Etxebarria, Myriam (2000), “The primitives of temporal relations”, in: Roger Martin/David Michaels/Juan Uriagereka (edd.), Step by step, Cambridge, MA, MIT Press, 157–186. Demirdache, Hamida/Uribe-Etxebarria, Myriam (2007), “The syntax of time arguments”, Lingua 117, 330–366. Demirdache, Hamida/Uribe-Etxebarria, Myriam (2008), “Scope and anaphora with time arguments. The case of ‘perfect modals’”, Lingua 118, 1790–1815. Depraetere, Ilse (2012), “Time in sentences with modal verbs”, in: Robert I. Binnick (ed.), The Oxford handbook of tense and aspect, Oxford, Oxford University Press, 989–1019.

238

Eva-Maria Remberger

De Saussure, Louis/Sthioul, Bertrand (2012), “The surcomposé past tense”, in: Robert I. Binnick (ed.), The Oxford handbook of tense and aspect, Oxford, Oxford University Press, 586–610. Dowty, David (1979), Word meaning and Montague grammar, Dordrecht, Reidel. Enç, Mürvet (1987), “Anchoring conditions for tense”, Linguistic Inquiry 18, 633–657. Fălăuş, Anamaria (2014), “Presumptive mood, factivity and epistemic indefinites in Romanian”, Borealis. An International Journal of Hispanic Linguistics 3, 105–124. Farkas, Donka (1992), “On the semantics of subjunctive complements”, in: Paul Hirschbühler/ Konrad Koerner (edd.), Romance languages and modern linguistic theory, Amsterdam/ Philadelphia, Benjamins, 69–105. Filip, Hana (2011), “Aspectual class and Aktionsart”, in: Claudia Maienborn/Klaus von Heusinger/ Paul Portner (edd.), Semantics. An international handbook of natural language meaning, Berlin/Boston, Mouton De Gruyter, 1186–1217. Fleischmann, Suzanne (1982), The future in thought and language. Diachronic evidence from Romance, Cambridge, Cambridge University Press. Gärtner, Eberhard (1998), Grammatik der portugiesischen Sprache, Tübingen, Niemeyer. Giannakidou, Anastasia (1998), Polarity sensitivity as (non)veridical dependency, Amsterdam/ Philadelphia, Benjamins. Giorgi, Alessandra (2010), About the speaker. Towards a syntax of indexicality, Oxford, Oxford University Press. Giorgi, Alessandra/Pianesi, Fabio (1997), Tense and aspect. From semantics to morphosyntax, New York/Oxford, Oxford University Press. González-Vilbazo, Kay-Eduardo/Remberger, Eva-Maria (2005), “Ser and estar. The syntax of stage level and individual level predicates in Spanish”, in: Claudia Maienborn/Angelika Wöllstein (edd.), Event arguments. Foundations and applications, Tübingen, Niemeyer, 89–112. González-Vilbazo, Kay-Eduardo/Remberger, Eva-Maria (2007), “Die Kopula im Romanischen”, in: Ljudmila Geist/Björn Rothstein (edd.), Kopulaverben und Kopulasätze. Intersprachliche und intrasprachliche Aspekte, Tübingen, Niemeyer, 201–226. Grevisse, Maurice/Goosse, André (152011), Le bon usage, Bruxelles, De Boeck/Duculot. Guéron, Jacqueline/ Hoekstra, Teun (1995), “The temporal interpretation of predication”, in: Anna Cardinaletti/Teresa Guasti (edd.), Small clauses, San Diego, Academic Press, 77–107. Hacquard, Valentine (2006), Aspects of modality, PhD dissertation, Cambridge, MA, MIT Press. Hacquard, Valentine (2011), “Modality”, in: Claudia Maienborn/Klaus von Heusinger/Paul Portner (edd.), Semantics. An international handbook of natural language meaning, vol. 2, Berlin/New York, Mouton de Gruyter, 1484–1515. Hamm, Friedrich/Bott, Oliver (2014), “Tense and aspect”, in: Edward N. Zalta (ed.), The Stanford encyclopedia of philosophy (Spring 2014 Edition), http://plato.stanford.edu/archives/spr2014/ entries/tense-aspect/ (17.11.2015). Hogeweg, Lotte/de Hoop, Helen/Malchukov, Andrej (2009), Cross-linguistic semantics of tense, aspect, and modality, Amsterdam/Philadelphia, Benjamins. Holt, Jens (1943), Études d’aspect [Acta Jutlandica: Aarsskrift for Aarhus Universitet 15.2.], Copenhagen, Universitetsforlaget. Iatridou, Sabine (2000), “The grammatical ingredients of counterfactuality”, Linguistic Inquiry 31, 231–270. Jackendoff, Ray (1990), Semantic structures, Cambridge, MA, MIT Press. Kaufmann, Stefan/Condoravdi, Cleo/Harizanov, Valentina (2005), “Formal approaches to modality”, in: William Frawley/Erin Eschenroeder/Sarah Mills/Thao Nguyen (edd.), vol. 1: The expression of modality, Berlin/New York, De Gruyter, 71–106. Klein, Wolfgang (1994), Time in language, London/New York, Routledge.

Morphology and semantics: Aspect and modality

239

Kratzer, Angelika (1977), “What ‘must’ and ‘can’ must and can mean”, Linguistics and Philosophy 1, 335–355. Kratzer, Angelika (1981), “The notional category of modality”, in: Hans-Jürgen Eikmeyer/Hannes Rieser (edd.), Words, worlds, and contexts. New approaches in word semantics, Berlin/New York, De Gruyter, 38–74. Kratzer, Angelika (1991), “Modality”, in: Arnim von Stechow/Dieter Wunderlich (edd.), Semantik. Ein internationales Handbuch zeitgenössischer Forschung, Berlin/New York, De Gruyter, 639–650. Kratzer, Angelika (1995), “Stage-level and individual-level predicates”, in: Gregory N. Carlson/ Francis Jeffry Pelletier (edd.), The generic book, Chicago, University of Chicago Press, 125–175. Krifka, Manfred (1989), “Nominal reference, temporal constitution and quantification in event semantics”, in: Renate Bartsch/Johan von Benthem/Peter van Emde Boas (edd.), Semantics and contextual expression, Dordrecht, Foris, 75–115. Krifka, Manfred (1992), “Thematic relations as links between nominal reference and temporal constitution”, in: Ivan Sag/Anna Szabolcsi (edd.), Lexical matters, Stanford, CSLI, 29–53. Krifka, Manfred (1998), “The origins of telicity”, in: Susan Rothstein (ed.), Events and grammar, Dordrecht, Kluwer, 197–235. Laca, Brenda (2005), “Tiempo, aspecto y la interpretación de los verbos modales en español”, Lingüística 17 (ALFAL), 9–44. Laca, Brenda (2006), “Temporalidad y modalidad”, in: Miguel Casas Gómez/Ana Isabel RodríguezPiñero Alcalá (edd.), X Jornadas de Lingüística, Cádiz, Universidad de Cádiz, 109–136. Lohnstein, Horst (²2011), Formale Semantik und natürliche Sprache, Berlin/New York, De Gruyter. Lyons, John (1977), Semantics, Cambridge, Cambridge University Press. Maienborn, Claudia (2003), Die logische Form von Kopula-Sätzen, Berlin, Akademie-Verlag. Maienborn, Claudia (2011), “Event semantics”, in: Claudia Maienborn/Klaus von Heusinger/Paul Portner (edd.), Semantics. An international handbook of natural language meaning, vol. 1, Berlin/New York, De Gruyter, 802–829. Mateus, Maria Helena Mira/Brito, Ana Maria/Duarte, Inês Silva/Hub Faria, Isabel (1983), Gramática da língua portuguesa. Elementos para a descrição da estructura, funcionamento e uso do português actual, Coimbra, Livraria Almedina. Mihoc, Teodora (2014), “The Romanian future-and-presumptive auxiliary”, McGill Working Papers in Linguistics 24, 64–80. Milsark, Gary L. (1974), Existential sentences in English, PhD dissertation, Cambridge, MA, MIT. Montague, Richard (1973), “The proper treatment of quantification in ordinary English”, in: Jaakko Hintikka/Julius Moravcsik/Patrick Suppes (edd.), Approaches to natural language, Dordrecht, Reidel, 221–242. Palmer, Frank R. (²1990), Modality and the English modals, Essex, Longman. Palmer, Frank R. (²2001), Mood and modality, Cambridge, Cambridge University Press. Pană Dindelegan, Gabriela (ed.) (2013), The grammar of Romanian, Oxford, Oxford University Press. Parsons, Terence (1990), Events in the structure of English. A study of subatomic semantics, Cambridge, MA, MIT Press. Partee, Barbara (1973), “Some structural analogies between tenses and pronouns in English”, Journal of Philosophy 70, 601–609. Perea, Maria-Pilar (2002), “Flexió verbal regular”, in: Joan Solà/Maria-Rosa Lloret/Joan Mascaró/ Manuel Pérez Saldanya (edd.), Gramàtica del català contemporani, 3 vol., Barcelona, Empúries, 583–646. Picallo, Carme M. (1990), “Modal verbs in Catalan”, Natural Language and Linguistic Theory 8, 285– 312. Portner, Paul (2009), Modality, Oxford, Oxford University Press. Prior, Arthur N. (1967), Past, present, and future, Oxford, Oxford University Press.

240

Eva-Maria Remberger

Pustejovsky, James (1991), “The syntax of event structure”, Cognition 41, 47–81. Pustejovsky, James (1995), The generative lexicon, Cambridge, MA, MIT Press. Quer, Josep (1998), Mood at the interface, The Hague, Holland Academic Graphics. Reichenbach, Hans (1947), Elements of symbolic logic, New York, MacMillan. Remberger, Eva-Maria (2006), Hilfsverben. Eine minimalistische Analyse am Beispiel des Italienischen und Sardischen, Tübingen, Niemeyer. Remberger, Eva-Maria (2010), “The evidential shift of WANT”, in: Tyler Peterson/Uli Sauerland (edd.), evidence from evidentials, Vancouver, Canada, UBC Working Papers in Linguistics 25, 161–182, www.semanticsarchive.net/Archive/WI5NjEzY/ and http://www.linguistics.ubc.ca/UBCWPL/ (18.11.2015). Remberger, Eva-Maria (2011), “Tense and volitionality”, in: Renate Musan/Monika Rathert (edd.), Tense across languages, Berlin/New York, De Gruyter, 9–35. Renzi, Lorenzo/Salvi, Giampaolo/Cardinaletti, Anna (edd.) (1995), Grande grammatica italiana di consultazione, vol. 3: Tipi di frase, deissi, formazione delle parole, Bologna, il Mulino. Rothstein, Susan Deborah (2004), Structuring events: a study in the semantics of lexical aspect, Oxford, Blackwell. Ruwet, Nicolas (1984), “Je veux partir / *Je veux que je parte. On the Distribution of finite complements and infinitival complements in French”, Cahiers de grammaire 7, 75–138. Smith, Carlota (1997), The parameter of aspect, Dordrecht, Kluwer. Smith, Carlota (2012), “Tense and aspect. Time across languages”, in: Claudia Maienborn/Klaus von Heusinger/Paul Portner (edd.), Semantics. An international handbook of natural language meaning, Berlin/New York, De Gruyter, 2581–2608. Squartini, Mario (1998), Verbal periphrases in Romance. Aspect, actionality and grammaticalization, Berlin/New York, De Gruyter. Squartini, Mario (2005), “L’evidenzialità nel rumeno e nelle altre lingue romanze”, Zeitschrift für romanische Philologie 121, 246–268. de Swart, Henriëtte (2011), “Mismatches and coercion”, in: Claudia Maienborn/Klaus von Heusinger/ Paul Portner (edd.), Semantics. An international handbook of natural language meaning, Berlin/New York, De Gruyter, 547–597. Thieroff, Rolf (2010), “Moods, moods, moods”, in: Björn Rothstein/Rolf Thieroff (edd.), Mood in the languages of Europe, Amsterdam/Philadelphia, Benjamins, 1–29. Vendler, Zeno (1967), “Verbs and Times”, in: Linguistics in Philosophy, Ithaca/New York, Cornell University Press, 97–121. Vikner, Sten (1985), “Reichenbach revisited. One, two, or three temporal relations?”, Acta Linguistica Hafniensia 19, 81–98. Walková, Milada (2012), “Dowty’s aspectual tests. Standing the test of time but failing the test of aspect”, Poznań Studies in Contemporary Linguistics 48, 495–518. Wandruszka, Ulrich (1991), “Frasi subordinate al congiuntivo”, in: Lorenzo Renzi/Giampaolo Salvi/ Anna Cardinaletti (edd.) (1991), Grande grammatica italiana di consultazione, vol. 2: I sintagmi verbale, aggettivale, avverbiale. La subordinazione, Bologna, il Mulino, 416–481. Willett, Thomas (1988), “A cross-linguistic survey of the grammaticization of evidentiality”, Studies in Language 12, 51–97. Zafiu, Rodica (2013), “Mood, tense and aspect”, in: Gabriela Pană Dindelegan (ed.), The grammar of Romanian, Oxford, Oxford University Press, 24–65. Zagona, Karen (2012a), “Tense, aspect, and modality”, in: Marcel Den Dikken (ed.), The Cambridge handbook of generative syntax, Cambridge, Cambridge University Press, 746–792. Zagona, Karen (2012b), “Tense and aspect”, in: José Ignacio Hualde/Antxon Olarrea/Erin OʼRourke (edd.), The handbook of Hispanic linguistics, Malden, MA, Wiley-Blackwell, 355–372. Zucchi, Alessandro (1998), “Aspect shift”, in: Susan Rothstein (ed.), Events and grammar, Dordrecht, Kluwer, 349–370.

Luis López

8 (In)definiteness, specificity, and differential object marking Abstract: A number of Romance languages exhibit Differential Object Marking, a phenomenon in which a subset of direct objects is distinguished by means of a morphological mark or syntactic displacement. It has long been known that this mark correlates with semantic or pragmatic features such as specificity and topicality, which makes it an ideal playground to investigate the interfaces between syntax and interpretation. This chapter argues that DOM, in fact, provides interesting evidence for a particular theory of the role that syntactic configuration plays in semantic interpretation. Keywords: definiteness, indefiniteness, DOM, specificity, scrambling

1 Introduction Spanish, Romanian, and many other languages share the property that their objects may appear in two morphological shapes. In Spanish, some direct objects are preceded by a morpheme that surfaces as a: (1) Sp. a. Busco a un estudiante. seek.1SG DOM a student ‘I’m looking for a student.’ b. Busco un estudiante. seek.1SG a student ‘I’m looking for a student.’ Likewise, Romanian direct objects can be preceded by pe (and when that happens, the pe object is often doubled by a clitic): (2) Rom. a. Îl

caut pe un student. seek.1SG DOM a student ‘I’m looking for a student.’ CL . ACC

b. Caut un student. Seek.1SG a student ‘I’m looking for a student.’

(Mardale 2004, 64)

242

Luis López

The term Differential Object Marking (DOM) refers to this property. Traditionally studied in a language-by-language basis, after the work of Bossong (1985) it dawned on linguists that DOM is a phenomenon that can be found in hundreds of languages – and later typological research has expanded the data-base even more, cf. Iemmolo to appear for an overview). DOM can be expressed, as in the examples above, by means of a morpheme that derives diachronically from a preposition, but also by morphological case (Persian, Hindi, Inuit languages, Turkish), suppletive determiners (Maori) or agreement or clitics on the verb (Bantu family). If we adopt an abstract view of grammar that merges the traditional areas of morphology and syntax, the phenomenon of Scrambling (German, Dutch) is also a form of DOM, to the extent that a subset of objects behaves grammatically in a distinct manner. Even more interestingly, syntactic and semantic features associated with DOM appear in clusters in unrelated languages, which suggests that DOM can provide a door of entry toward an understanding of the features that make up the human faculty of language (López 2012). The Romance language family provides a healthy range of DOM phenomena. The best known cases are those of Spanish and Romanian, languages in which the usage of DOM extends to several types of DPs and on which there is an abundant literature (cf. Fábregas 2013 for a recent overview with focus on Spanish). DOM also appears in Catalan, Portuguese, Sardinian, Sicilian, Corsican and a number of Italian dialects. It even shows up sporadically in regional, non-standard forms of French and Italian, languages supposedly lacking DOM altogether (Iemmolo to appear). In these language varieties the distribution of DOM is considerably more restricted than in Spanish and Romanian. For instance, DOM in the variety of standard Catalan that appears in normative grammars only affects pronouns (cf. Escandell-Vidal 2009 for an overview of DOM in Catalan) and in Abruzzese only first and second pronouns (D’Alessandro 2012). In other varieties, DOM extends to definite DPs in dislocated position and in Sicilian it reaches any definite DP (but no indefinites). An extra requirement of animacy is present in almost all varieties. The goal of most DOM literature is to provide a simple, one-line statement that describes the distribution of DOM in a language. This goal is feasible in a language with limited DOM, but it is considerably harder to achieve in Spanish or Romanian. For instance, one can suggest the following generalization: (i) “marked objects are animate in Spanish, unmarked objects are inanimate” and take this to be the main rule of DOM in Spanish. Animacy is indeed a factor in Spanish DOM, as shown by the contrast between (3a) and (3b). But it can’t be the end of the story. (3c) shows that an animate object is not necessarily marked and (3d) shows that an inanimate one may: (3) Sp. a. Vio a un chico. saw.3SG DOM a boy ‘S/he saw a boy.’ b. *Vio a una mesa. table saw.3SG DOM a ‘S/he saw a table.’

(In)definiteness, specificity, and differential object marking

243

c. Vio un chico. saw.3SG a boy ‘S/he saw a boy.’ d. El camino sigue al río. The road follows DOM .the river ‘The road follows the river.’ Linguists have been aware of the difficulties in defining the distribution of DOM in a language like Spanish, and so they have taken two paths. One path is to make the statement vague enough that almost anything goes and find ad hoc accounts for contradictory evidence. Continuing with our example (3), the persistent linguist could say that in (3c) the object is de-animated while in (3d) it is en-animated (analyses that involve fluctuating definitions of animacy have, in fact, been proposed). It should be obvious that this strategy eventually leads to a cul-de sac. Along the same lines, the recent DOM literature also shows attempts at broader, cross-linguistic analyses, such as Aissen (2003), Dalrymple/Nikolaeva (2011), De Swart (2007), Iemmolo (to appear), Naess (2007), among others. Individually and as a group they have helped broaden the database considerably and they have provided insights into DOM as well as other topics. Just like the one-language specialists, these works try to account for the DOM phenomena within one generalization: Aissen and De Swart discuss specificity/definiteness and animacy within an OT framework, Naess discusses affectedness while Dalrymple/Nikolaeva and Iemmolo argue that Topicality is at the root of DOM. Given the difficulty of accounting for DOM in even one language, it is not surprising to find that these cross-linguistic investigations exert some considerable idealization of data (a perfectly acceptable strategy in science, in any case, as long as we are aware of it). A second, possibly more realistic strategy to describe DOM is to divide the problem in smaller pieces, accepting that each piece of the problem has its own analysis and that the analyses are related, at best, by a family resemblance. For instance, one could posit something like the following statement: (ii) “inanimate objects are marked if the external argument is not an agent/causer” as a statement added to (i). This additional clause takes care of (3d). This looks like a more promising path, even if less glamorous. It seems that our faculty of language – our Universal Grammar, if you will – includes DOM as a possible ingredient of a natural language. Once the grammar of a language adopts DOM, the usage may expand or contract along one of a number of significant linguistic features, most notably: animacy, theta structure, (in)definiteness, affectedness and information structure. Moreover, the features that trigger DOM may act in isolation or might be combined. And, as is often the case in morpho-syntax, there are patterns that seem circumscribed to very small sub-sets of the lexicon. An independent line of research on DOM explores the properties of this construction to the extent that it can shed light on specific theoretical debates. In particular,

244

Luis López

the grammatical availability of DOM in (1) and (2) turns out to be crucial data for our understanding of the connection between morphology, syntax and semantics. Within this line of research, there is already a wealth of work focusing on Spanish and Romanian: von Heusinger (2002), von Heusinger/Onea (2008), Leonetti (2004), López (2012), Rodríguez-Mondoñedo (2007), Torrego (1998). Although this approach is very recent, I think it is fair to say that it has made important contributions to our understanding of some theoretical problems and, in particular, to the syntaxsemantics interface. The rest of this chapter is organized as follows. Section 2 discusses the factors that affect the distribution of DOM in the Romance languages. Sections 3 and 4 present a summary of López (2012), submitted as an example of how the data of Romance DOM can provide insight to our understanding of the syntax-semantics interface. Section 5 presents the conclusions.

2 DOM factors in Romance We can divide the factors that affect DOM in three classes: (i) features of the object, (ii) features of the verb and (iii) features of the configuration.

2.1 Features of the object The factors involving the object in DOM phenomena have been studied quite extensively in many different languages. They can be summarized in three types of features: animacy, topicality and specificity. Moreover, as mentioned above, there are some varieties that accept DOM only on pronouns or a subset of pronouns. Let’s consider these categories in turn. We start with animacy: most forms of DOM in Romance affect only animate objects. This can be seen in the following contrast: (4) Sp. a. *Juan vio a todas las mesas. the tables Juan saw.3SG DOM all b. Juan vio todas las mesas. ‘Juan saw all the tables.’ There are only two types of exceptions. In Spanish, DOM on inanimate objects is possible when a constellation of conditions get together: the external argument is also inanimate and the verb assigns theme theta-roles (or, in any case, neither agentive nor experiencer) to both arguments (cf. García 2007 for a recent analysis):

(In)definiteness, specificity, and differential object marking

245

(5) Sp. El camino sigue al río. (=3b) The road follows DOM .the river ‘The road follows the river.’ The other type of exception to the animacy generalization comes from Balearic Catalan (Escandell-Vidal 2009). In this language, dislocated objects are marked even if they are inanimate. This can be seen in the following example (Escandell-Vidal 2009, 846): (6) Cat. Colliu-les a les peres, que ja són madures pick.IMP-Cl.ACC . PL DOM the pears that already are ripe ‘Pick the pears, for they are ripe.’ This segues quite naturally into the second factor: Topicality (cf. Laca 1995; Leonetti 2004; Iemmolo 2010; to appear). Following these authors, we take Topicality to be a term that embraces an aboutness dimension as well as a discourse-anaphoric dimension: we can take the Topic of a sentence to be a constituent that refers back to an antecedent in the previous discourse and fleshes out what the sentence is about. In the Romance tradition, it is often claimed that dislocated constituents are topics (cf. Vallduví 1992). It has been pointed out that dislocated objects accept DOM even in language varieties that do not normally accept it (Northern Italian, Standard Italian). The following pair of Italian examples is taken from Iemmolo (to appear): (7) It. a. A me, non mi convince questo. DOM me NEG me convince.3SG this ‘This does not convince me.’ b. *Non mi convince a me questo. Likewise, language varieties in which DOM is optional for a type of DP might regard it as obligatory when the object is dislocated. Dislocation is taken by Laca and Iemmolo as a sure indication of topicality and therefore a connection between topicality and DOM can be established via dislocation. Moreover, the diachronic studies of Iemmolo seem to indicate that for a number of Romance languages DOM began as a morpheme associated with dislocations which later spread to other uses. Finally, specificity is the DOM feature that has been explored in most depth, no doubt because it plays such an important role in contemporary linguistic theory: specificity effects in existential sentences, specific readings of scrambled objects, specificity in connection with telicity etc. Every discussion of specificity begins with the same complaint: there is no agreement on what it means. It would be presumptuous of me to try and provide a solution to the problem so I will content myself

246

Luis López

with an intuitive presentation (but cf. von Heusinger 2011 for a more detailed discussion). Consider the following example: (8) Mary is looking for a hammer. Mary may be the owner of a hammer and she may be looking for the hammer she lost. This hammer would be specific. Or she might be in a hardware store and looking to buy a hammer: that would be an unspecific hammer. Using specificity as a tool, we can separate two types of determiners with respect to their relationship with DOM (and always keeping animacy constant) in Spanish and Romanian. Determiners whose meaning entails specificity require DOM. In this category we have: each, every, most, some/many/few of the – Milsark’s (1974) “strong determiners”. Proper names and definite pronouns, which are specific by nature, also require DOM. Here are some examples of obligatory DOM in Spanish:1 (9) Sp. a. Juan me eligió a/*Ø mí. Juan Cl.1 chose.3SG DOM me ‘Juan chose me.’ b. Elegimos a/*Ø Miguel. Chose.1PL DOM Miguel ‘We chose Miguel.’ c. María ama a/*Ø todos los gatos. the cats Maria loves DOM all ‘Maria loves every cat.’ Determiners whose meaning does not entail specificity allow DOM. In this category fall Milsark’s “weak determiners”: a, some, many and numerals, provided that they are not constituents of a partitive construction. The presence of DOM with indefinite determiners allows for (but does not force) a specific reading for the DP. The absence of DOM makes this specific reading impossible. The following examples are Romanian (cf. 2): (10) Rom. a. Caut un student. seek.1SG a student ‘I’m looking for a student.’ b. Îl caut pe un student. Cl.ACC seek.1SG DOM a student ‘I’m looking for a student.’ (Mardale 2004, 64) 1 The “a/*Ø” formula means: obligatory a. The “a/Ø” formula means: optional a.

(In)definiteness, specificity, and differential object marking

247

The presence or absence of DOM on the object has additional effects on scope, which are discussed in section 3. Finally, as mentioned above, some Romance varieties accept DOM only on pronouns. Catalan (at least, the standard quasi prescriptive variety that appears in traditional descriptive grammars) accepts DOM only on pronouns: (11) Cat. a. Jo el vaig veure a ell. DOM him I CL .3 PST.1SG see ‘I saw him.’ b. Jo vaig veure la Maria. the Maria I PST.1SG see ‘I saw Mary.’ c. *Jo vaig veure a la Maria DOM the Maria I PST.1SG see In Abruzzese, only first and second person pronouns accept DOM (D’Alessandro 2012): (12) Abruzzese a. So vistə a tte. Be.1SG seen DOM you ‘I have seen you.’ b. *So vistə a jissə / a Marije. Be.1SG seen DOM them DOM Marije ‘I have seen them / Marije.’ Aissen (2003) regards the Catalan case as exemplifying the relevance of the specificity scale for DOM. Her idea is that noun phrases can be regarded to stand on a specificity scale such as the following: pronouns > proper names > definite noun phrases > indefinite noun phrases. According to Aissen, DOM distributes along this scale from left to right. D’Alessandro, on the other hand, ties the Abruzzese DOM with a broader investigation into the role of morphosyntactic features in microparametric variation.

2.2 Features of the verb The role of the verb in DOM has only been discussed on the wake of Torrego (1998) and remains, to this date, less understood than the role of the object. Moreover, all the Romance literature that I am aware of discusses only Spanish and so my discussion will also be limited to this language (outside of Romance, Kiparsky’s 1998 discussion of Finnish stands out, as well as Naess 2007).

248

Luis López

Torrego’s findings can be summarized as follows: several types of verbs select for obligatory DOM associated to the type of complement that they select or the type of event they denote. Consider first the affectedness variable: if a verb affects its complement (i.e.: exerts a permanent change on it), DOM becomes obligatory. Affectedness has also been investigated in a diachronic study of DOM in Spanish by von Heusinger/Kaiser (2011). In this respect, one can compare Sp. golpear ‘to hit’ with Sp. ver ‘to see’:2 (13) Sp. a. Tú golpearías a/*Ø un niño. you hit.COND.2SG DOM a child ‘You would hit a child.’ b. Tú verías a/Ø you see.COND.2SG DOM ‘You would see a child.’

un niño. a child

A second variable is humanness. Some verbs select for an obligatory human complement: to greet, to convince, to murder.3 These verbs also require DOM on the indefinite: (14) Sp. Tú asesinarías a/*Ø un miembro del congreso. you murder.COND.2SG DOM a member of.the congress ‘You would murder a member of congress.’ Relatedly, psych verbs also seem to prefer a marked object: (15) Sp. Tú admirarías a/*Ø un miembro del congreso. you admire.COND.2SG DOM a member of.the congress ‘You would admire a member of congress. Finally, Torrego (1998) argues that the Aktionsart of the verb is crucial to understand the DOM phenomena. Roughly put, telic verbs require DOM while atelic verbs are only compatible with it. For those verbs that could go one way or another, a

2 Notice that in these and the following examples I use a conditional form. The use of the conditional is meant to favor a non-specific reading for the complement and make sure that the specificity variable does not cloud the conclusion. Likewise, since the third person in Spanish ends in /a/, I switch to the second person, which ends in /s/, to make sure that the DOM morpheme is fully audible. 3 Although saludar ‘greet’ is in its normal use a verb that takes a human as an object, one can also see other metaphorical uses of the verb in which one greets the sun or the spring season. In these cases, DOM is also obligatory, which raises some questions that I will not be able to properly address in this paper. Thanks to Susann Fischer for pointing out the problem with this class of verb.

(In)definiteness, specificity, and differential object marking

249

marked object makes the predicate telic, an unmarked object makes it atelic. This can be seen in the following example: (16) Sp. Tú esconderías a/Ø un prisionero durante dos años. for two years you hide.COND.2SG DOM a prisoner ‘You would hide a prisoner for two years.’ According to Torrego, the version without DOM is an activity without an end point. Only the version with a marked object indicates an end point to the action of hiding – in effect, changing the Aktionsart from activity to achievement.

2.3 Configuration Finally, there are some syntactic configurations that require the presence of DOM in Romanian and Spanish (as well as other languages outside the Romance family). They are discussed in more detail in section 4.2, but here let’s just note the following example: (17) Sp. Tú considerarías a/*Ø un estudiante inteligente. intelligent you consider.COND.2SG DOM a student ‘You would consider a student intelligent.’ In small clauses like (17), the constituent that seems to bear the grammatical relation direct object (as shown by routine tests such as passivization and pronominalization) is not a complement of the verb – it is in fact an argument of the small clause predicate. This property is what makes DOM obligatory here, as I show below.

3 Scope and DOM: configuration and the modes of semantic composition The presence or absence of a morphological marking on an indefinite object dramatically affects its scope possibilities in Spanish and Romanian (as well as other unrelated languages such as Turkish, Persian, Hindi and Kiswahili, that I know of). It has been known since the 1980s that the scope of indefinites presents a challenge for the standard views of the syntax-semantics interface, which revolve around the notion that quantifier scope can be predicted using syntactic c-command and Quantifier Raising, as I explain below. López (2012) argues that the DOM facts, together with some recent developments in semantic theory, shed light on the empirical problems and lead to a new understanding of the interface. Sections 3 and 4 summarize some of the results of López (2012).

250

Luis López

Section 3.1 presents the canonical view of quantifier scope and the challenges presented by indefinite objects, section 2 discusses the scope of indefinite objects in DOM languages, section 3.3 shows that marked objects scramble and section 3.4 connects all three strands: scope, DOM and scrambling.

3.1 The syntax-semantics interface The classical formal semantics that grew out of the works of Richard Montague and Barbara Partee assumes one mode of semantic composition, called Function Application (FA), which has the following structure: (18) λx [P(x)](a) → P(a) Generative grammar incorporated formal semantics and proposed a simple hypothesis to approach the translation of syntactic structures onto semantic structures: the c-command configuration gets translated as scope. If x c-commands y, then x has scope over y. However, there are apparent mismatches, such as the one shown in (19): (19) a. Some man loves every cat. b. ∃ > ∀ ∀>∃ As shown in (19b), the universal quantifier may take scope over the existential one, giving rise to a reading in which for every cat there is some man or other that loves it. But ‘every cat’ does not, on the face of it, c-command ‘some man’. This empirical problem is solved by means of an inaudible syntactic operation (although very much audible in Hungarian and many other languages) called Quantifier Raising (QR). QR adjoins the quantifier to TP (May 1985, i.m.a.): (20) [TP every cat [TP some man [VP t loves t ]]] Thus, with a helping hand from QR, classic generative grammar was able to maintain the idea that there is a direct mapping of syntax to semantics. However, indefinite NPs remain an empirical problem for the framework sketched above. The problems come from their unusual scope properties and the phenomenon of specificity. I focus now on scope. Indefinite NPs can have a very wide scope beyond the limits of what QR can allow. Consider the following sentence:

(In)definiteness, specificity, and differential object marking

251

(21) a. If Bert invites a philosopher, Lud will be upset. b. → > ∃ ∃>→ This sentence allows a wide scope for ‘a philosopher’, with a reading that can be paraphrased as follows: “there is a philosopher such that, if Bert invites him, Lud will be upset”. But in order to obtain this reading via QR, the indefinite would have to move outside of a conditional clause, which is a strong island for movement. Indefinite NPs can also have an obligatory narrow scope. This is exemplified in the following Persian example (from Karimi 2003, 111): (22) Persian a. Kimea ye ketâb na-xarid. Kimea a book NEG .compró ‘Kimea didn’t buy a book.’ b. ¬ > ∃ *∃ > ¬ As indicated, this sentence means that Kimea did not buy any books. It does not mean that there is a book that Kimea did not buy, although the second reading could easily be obtained by QR. Notice that ye ketâb is not a negative polarity item. Thus, the question that this sort of example poses is the following: If QR is available in the grammar, why is it that some indefinite NPs cannot use it and obtain wide scopes? And, even more intriguingly, why is it that the indefinites that take obligatory narrow scope are always object indefinites? Let’s start with the first problem: indefinites with wide scope. In an influential article, Reinhart (1997) argues that indefinites can be interpreted by means of choice functions. The starting point is the assumption that indefinite NPs are predicates of type , denoting sets of entities. A choice function applies on an indefinite NP and chooses an element from the set, with the result that the type of the indefinite is raised to , ready to combine with a predicate by Function Application. The function must be closed by an existential quantifier. This existential quantifier can be located in any syntactic position without respecting syntactic islands, which gives rise to the possibility of wide scopes. Thus, sentence (23a) can have the logical form (23b): (23) a. If Bert invites a philosopher, Lud will be upset. b. ∃f CH(f) ∧ Bert invites f(philosopher) → Lud will be upset The logical form in (23b) can be paraphrased as follows: ‘there is (a choice function that picks) a philosopher such that if Bert invites the philosopher picked by the function Lud will be upset.’

252

Luis López

Notice that in Reinhart’s approach to wide scopes, the scope of an indefinite NP is independent of its syntactic position. The connection between syntax and semantics becomes less direct than in the classic approach. Let’s consider now the second problem: some indefinite objects have obligatory narrow scope. The literature on this topic has grown substantially in recent years and I can’t present an overview in these pages. Instead, I adopt without argument Chung/Ladusaw’s (2004) approach. Their point of departure is again that indefinite NPs are of type . Lexical verbs are also of type and consequently cannot combine with indefinites by means of Function Application. Chung/Ladusaw propose that indefinite objects and lexical verbs may combine by Restrict, a mode of semantic composition in which the indefinite NP does not saturate the predicate – rather, it behaves like a modifier. The resulting predicate is then immediately closed by an existential quantifier. Consider (24) as an example. We are focusing on the semantic composition of the VP ‘buys a book’. The output of applying Restrict to ‘buy’ and ‘a book’ is the conjunct shown in (24b). (24c) shows existential closure of the two variables. The resulting LF could be paraphrased as ‘there is an x such that x has the properties of being bought and being a book:’ (24) a. Mary [VP buys a book] b. RESTRICT (λx [ buy’(x)], book’) → λx [buy’(x) ∧ book’(x)] c. ∃x [buy’(x) ∧ book’(x)] One semantic effect of this immediate closure is that an indefinite object composed by Restrict can only have narrow scope. Notice that Chung/Ladusaw (2004) takes another step in divorcing syntax from semantics. This is because application of Restrict depends on the type of noun phrase that we have, not on the syntactic position where it is found. In theory at least, an indefinite noun phrase could be in any syntactic position, be composed by Restrict and therefore take a narrow scope over anything else in the clause.

3.2 The scope of the marked and the unmarked object In Spanish and Romanian (and other languages) the indefinite marked object has all the properties of an indefinite NP subjected to a choice function while an unmarked object looks like it has been composed by Restrict. The marked object may take wide scope beyond any islands, as can be seen in the following examples: (25) Sp. a. Todo hombre amó a una mujer. ‘Every man loved DOM a woman.’ =for every man x there was a woman y such that x loved y =there was a woman x such that every man y loved x

∀> ∃ ∃>∀

(In)definiteness, specificity, and differential object marking

253

b. La mayoría de los hombres amó a una mujer. ‘Most men loved DOM a woman.’ ∃ > Most / Most > ∃ c. Juan no amó a una mujer. ‘Juan didn’t love DOM a woman.’ ≠there is no x, x a woman, such that Juan loved x. *¬ > ∃ =there was a woman x such that Juan didn’t love x. ∃ > ¬ d. Juan no amó a ninguna mujer. ‘Juan didn’t love DOM any woman.’ ‘Juan didn’t love any woman.’ *∃ > ¬ / ¬ > ∃ e. Si Lud invita a un filósofo, Bert se ofenderá. ‘If Lud invites DOM a philosopher, Bert will be upset.’ =Bert doesn’t want any philosophers at the party →>∃ =there is a philosopher that Bert doesn’t want invited ∃ > → The examples (25a) and (25b) show that Spanish DOM can have wider or narrower scope with respect to other quantifiers. (25c) shows that Spanish DOM has wide scope over negation if the determiner is un/a. But DOM is compatible with a narrow scope with respect to negation too, as shown in (23d). In (23d), the negative word ningun- must take narrow scope and must also be prefixed by DOM. Finally, (23e) shows that a marked object in Spanish may take scope outside a conditional island. The ease with which marked objects in Spanish can take a variety of scopes suggests that their interpretation is mediated by choice functions. Romanian pe indefinite objects have the same properties as Spanish a indefinite objects. For instance, they can take scope over a conditional island: (26) Rom. a. Dacă Bert invită pe un filozof, Lud se va supăra. If Bert invites DOM a philosopher Lud CL AUX . FUT annoy ‘If Bert invites a philosopher, Lud will be annoyed.’ b. → > ∃ ∃>→ The examples in (25) contrast with those in (27), with an unmarked object. Wide scope is impossible in these examples: (27) Sp. a. Todo hombre amó una mujer. ‘Every man loved a woman.’ =for every man x there is a woman y such that x loves y ∀ > ∃ ≠there was a woman x such that every man y loved x *∃ > ∀

254

Luis López

b. La mayoría de los hombres amó una mujer. *∃ > Most / Most > ∃ ‘Most men loved a woman.’ c. Juan no amó una mujer. ‘Juan didn’t love a woman.’ =there is no x, x a woman such that Juan loved x ≠there was a woman x such that Juan didn’t love x

¬>∃ *∃ > ¬

d. Juan no amó ninguna mujer. ‘Juan didn’t love a woman.’ e. Si Lud invita un filósofo, Bert se ofenderá. ‘If Lud invites a philosopher, Bert will be ofended.’ =Bert doesn’t want philosopers at the party →>∃ ≠there is a philosopher that Bert doesn’t want invited. *∃ > → Likewise, the Romanian indefinite object without pe cannot take wide scope. (28) contrasts with (26): (28) Rom. a. Dacă Bert invită un filozof, Lud se va supăra. if Bert invites a philosopher Lud CL AUX . FUT annoy ‘If Bert invites a philosopher, Lud will be annoyed.’ b. → > ∃ *∃ > → The obligatory narrow scope of unmarked objects tells us that they must be composed by Restrict. The last conclusion allows us to address an old empirical problem regarding the distribution of DOM. It has often been noticed that existential verbs do not tolerate DOM: (29) Sp. a. Hay

un señor en el jardín. a gentleman in the garden ‘There is a gentleman in the garden.’ HAVE

b. *Hay a un señor en el jardín. This has often been taken to derive from the supposed specificity property of DOM: if marked objects are specific and existential predicates do not admit specific objects, it follows that existential predicates cannot have marked objects. However, the first premise of the argument is false: marked objects are not necessarily specific. This can be seen in the following examples:

(In)definiteness, specificity, and differential object marking

255

(30) Sp. a. Estoy buscando a un señor en el jardín. a gentleman in the garden be.1SG searching DOM ‘I am looking for a gentleman in the garden.’ estoy buscando a nadie. be.1SG searching DOM anyone ‘I am not looking for anyone.’

b. No

NEG

In (30a) it is not necessarily the case that the gentleman that I am looking for is specific. In (30b), nadie is unambiguously non-specific. DOM makes specificity possible but does not force it. But then, why is (29b) ungrammatical? I argue that the ungrammaticality of (29b) depends on the mode of semantic composition associated with existentials. Let’s take it seriously that an existential predicate in fact involves the conjunction of two predicates, as is standard in predicate logic and formal semantics: (31) ∃x gentleman(x) ∧ in the garden(x) If the noun phrase of an existential must be a predicate, it follows that it can’t be subjected to choice functions because choice functions raise the type of the indefinite to that of an entity . Since marked objects are of type , it follows that marked objects can’t be components of an existential predicate.

3.3 Syntax of DOM Classical c-command tests reveal that the marked object is in a hierarchically higher position than the unmarked object, which I take to indicate that marked objects undergo a Scrambling operation (but cf. Leonetti 2004 for a different take on similar data). The Scrambling we are referring to here is very short. It can only be perceived in relation to the indirect object. Consider the following example: (32)

Context: What did the enemies do? They delivered X to Y, but . . . Sp. Los enemigos no entregaron a sui hijo a/Ø ningúni prisionero. prisoner The enemies NEG entregaron DAT his son DOM any ‘The enemies delivered no prisoner to his son.’

The NP ningún prisionero ‘no prisoner’ can be marked. Interestingly, the presence or absence of DOM alters the range of possible interpretations for this sentence. When the object is marked, it is possible to obtain a quantifier-variable interpretation for ‘sons’ and ‘prisoners’. This interpretation could be paraphrased as ‘the enemies

256

Luis López

didn’t deliver John to John’s son, the enemies didn’t deliver Mary to Mary’s son, the enemies didn’t deliver Pat to Pat’s son, etc.’ Under the common assumption that quantifier-variable interpretations require c-command configurations, it follows that marked objects c-command indirect objects. The unmarked object does not allow for a quantifier-variable interpretation: the only interpretation available is something like ‘the enemies didn’t deliver his son any prisoners’. This tells us that the unmarked object does not c-command the indirect object. Notice that the structural difference is not reflected in the word order: the left-to-right linearization of complements in Spanish seems to be fairly independent from structure. Anaphora binding confirms the higher position of the marked object. Consider the following examples: (33) Sp. a. María le entregó a Juan un hombre. ‘Maria delivered a man to Juan.’ b. *María le entregó a sí mismo un hombre. a man Maria CL . DAT delivered DAT REFL c. María le entregó a sí mismo a un hombre. DOM a man Maria CL . DAT delivered DAT REFL ‘María delivered a man to himself.’ Take (33a) as the baseline example, a regular ditransitive (here the morpheme a is a dative marker, not the DOM marker). (33b) shows that an unmarked object cannot bind an indirect object reflexive. (33c) shows that a marked object can. The marked direct object does not scramble very high (unless some other operation, such as dislocation, affects it). In fact, the marked object does not c-command an external argument, even if the latter remains in situ: (34) Sp. Ayer no castigó su padre a ningún niño. child yesterday NEG punished.3SG his father DOM no ‘Yesterday his father punished no child.’ It is very difficult to interpret example (34) as involving a pairing of children and fathers. Thus, the scrambling that DOM gives rise to is very short, since it stays below the initial merge position of the external argument. Once again, the properties of marked and unmarked objects in Romanian parallel point by point those of Spanish: the marked object c-commands the indirect object, the unmarked object does not: (35) Rom. Duşmanii nu l-au cedat pe niciun prizonier fiului lui. enemies.DEF NEG CL . ACC -AUX yielded DOM NEG a prisoner son.DEF. DAT his ‘The enemies yielded no prisoner to his son.’

(In)definiteness, specificity, and differential object marking

257

Moreover, Romanian provides an extra piece of evidence for scrambling because in this language the left-to-right order of complements reflects structure. Unmarked objects show up to the right of indirect objects, while marked objects are located preferably to the left of indirect objects. cedat fiului lui pe niciun prizonier (36) Rom. ? Duşmanii nu l-au enemies.DEF NEG CLACC -AUX yielded son.DEF. DAT his DOM NEG a prisoner (37) Rom. Duşmanii nu i-au cedat fiului lui un prizonier. enemies.DEF NEG CLDAT-AUX yielded son.DEF. DAT his a prisoner ‘The enemies yielded no prisoner to his son.’ The configurational difference between the marked and unmarked object in Spanish and Romanian can be shown graphically: (38) [vP EA v [αP OB(DOM) IO α [VP V OB]]] The structure represented in (38) includes a number of structural hypotheses, some commonly assumed in generative grammar, others less so. The only crucial hypothesis for this work is the relative c-command relations of the arguments in the clause. The external argument c-commands the other arguments (unless the latter are dislocated, not discussed here). The marked object c-commands the indirect object while the latter c-commands the unmarked object. Moreover, (38) posits the presence in the structure of a functional head that here noncommittally I refer to as α. α introduces the indirect object and offers a spec for the marked object. It is related to the applicative heads that have been discussed in the literature since Marantz (1993) as well as the Inner Aspect of Travis (1992; 2010), although a detailed consideration goes beyond the limits of this chapter.

3.4 Scrambling and semantic composition From the previous two sections we can extract two conclusions: (39) a. The marked object (i) undergoes a Scrambling operation and (ii) is composed semantically by Function Application (after type-shifting.) b. The unmarked object (i) stays in Compl,V and (ii) is composed semantically by Restrict.

258

Luis López

It is tempting to link the two contrastive properties of direct objects. Thus, I submit the following hypothesis: (40) [vP EA v [αP OB(DOM) IO α [VP V OB]]] ↑ FA

↑ Restrict

The question now is: why is it that DPs found in the VP are composed by Restrict? And why is it that DPs outside of the VP are interpreted by Function Application? In the following, I borrow and freely adapt some ideas sketched in Carlson (2003). Let’s assume the model includes a set E of eventualities and every lexical verb denotes a member of this set (i.e.: a type of event). Let’s further assume that this denotation must be kept for every projection of the lexical verb and the resulting VP also denotes a member of E. We additionally maintain the assumption that indefinite NPs are predicates: thus, we have a set P of predicates and indefinite NPs denote members of this set. Within these conditions, Restrict is allowed as a mode of semantic composition within the VP. That is because Restrict limits itself to increasing the specificity of the eventuality that the verb refers to. But Function Application is not permitted within the VP because it alters the type of the constituent it composes with. Since within the VP one can only combine and element of E with an element of P, it follows that individuals and generalized quantifiers do not have a denotation within the VP and therefore must raise to a VP-external position. Among the phrases that must move we find the indefinite NPs that have been type-shifted from type to type by means of a choice function.

3.5 Conclusion Recall that both Reinhart (1997) and Chung/Ladusaw (2004) have the effect of making the syntax-semantics mapping opaque because they connect the scope of indefinites to semantic operations or modes of semantic composition regardless of syntactic configuration. The proposal sketched in these pages recovers a predictable syntax-semantics mapping because there is a connection between syntactic position and the mode of semantic composition and, consequently, the range of indefinite scopes.

4 Extensions and predictions In this section I develop further the ideas introduced in section 3 and explore some empirical consequences.

(In)definiteness, specificity, and differential object marking

259

4.1 The structure of marked and unmarked objects Linguists working on a variety of languages have noticed that a subset of indefinite direct objects bear a combination of two properties: (i) they can only take narrow scope and (ii) they are found either adjacent to the verb or fully integrated in the morphological structure of the main clause predicate. Cf. for instance Massam (2001) on Palauan, Dayal (2011) on Hindi, Karimi (2003; 2005) on Persian, van Geenhoven (1998) on West Greenlandic. These authors have utilized the concept of (pseudo)incorporation to account for this phenomenon. Let’s assume that unmarked objects in Spanish are pseudo-incorporated into the lexical verb. There are several ways to implement this idea technically, here I propose one that seems attractive. Unmarked nominals are relatively small, maybe consisting only of a Number head that takes an NP as an argument and provides a label for the resulting structure. Marked objects include a determiner that selects a Number Phrase as well as a K head that selects the determiner. The K head is the one that, under appropriate conditions, may spell out as DOM: (41) a. Marked object:

K[DP[NumP[NP]]]

b. Unmarked object: Num[NP] Further, let’s take K to be a phase head in the sense of Chomsky (2000), while the smaller Number Phrase is not. An immediate consequence of this assumption is that the unmarked object transfers to the interpretative mechanisms together with the verb while the marked object is transferred independently. I suggest that the syntactic and semantic effects of incorporation are a direct consequence of the simultaneous transfer of verb and object. I already suggested that K may spell out as DOM. It seems quite natural to take K to be the syntactic category that is associated with the choice function in the semantics: (42)

K also takes generalized quantifiers and definite DPs as complement. In these cases, K is simply a formal feature without a semantic function. Let’s briefly turn to specificity. We have seen that specificity is a property of marked objects. Bearing the function f is also a property of marked objects. It seems

260

Luis López

very reasonable to regard specificity also as a (optional) property of f (as proposed by von Heusinger 2002). Let me now show a prediction that derives from these assumptions. Brugè/ Brugger (1996) show that bare plurals are incompatible with DOM in Spanish (in fact, I have found out that the same restriction holds of the other languages I have investigated): (43) Sp. a. Juan vio futbolistas en el Camp Nou. ‘Juan saw footballers in the Camp Nou.’ b. *Juan vio a futbolistas en el Camp Nou. If, as Martí (2008) argues, bare plurals consist of a Number Phrase, it follows that bare plurals cannot be marked. More consequences ensue: since bare plurals are unmarked indefinites, they have to be predicates and must be composed by Restrict. If they are composed by Restrict, they can only take narrow scope. This prediction holds. Example (44) only has the reading in which Juan saw no footballers; it does not have the reading according to which there are some footballers he did not see: (44) Sp. a. Juan no vio futbolistas en el Camp Nou. ‘Juan didn’t see footballers in the Camp Nou.’ b. ¬ > ∃ *∃ > ¬ Within the set of hypotheses that constitute generative grammar, the motivation for incorporation and scrambling is Case Theory. As Baker (1988) argues, incorporation satisfies the Case Filter. We can implement this idea by assuming that the object can incorporate into V and V incorporates into v where the [assign accusative] feature is lodged. Since the incorporated object is contained within v, the need of Case is satisfied without further ado. KP does not incorporate. Why does it scramble to Spec,ɑ? Let’s assume that Case assignment is carried out in a strictly local configuration, what used to be called government. If so, KP needs to raise to Spec,ɑ to be governed by v.

4.2 Configurational DOM As mentioned in section 2.3, certain configurations require DOM even on their indefinite objects, regardless of any other properties of these objects (once again, keeping animacy constant). These configurations can be understood as having the following property in common: the grammatical object of the matrix predicate

(In)definiteness, specificity, and differential object marking

261

is not a complement of the predicate. Some of these grammatical objects are constituents of the clause that is the complement of the main predicate. In broad terms, they can be described with this schema: (45) V [XP OB(DOM) . . .] The following are some Spanish examples: (46) Small clauses Sp. El profesor considera a/*Ø un estudiante inteligente. the profesor considers DOM a student intelligent ‘The professor considers a student intelligent’ (47) Causatives Sp. María hizo llegar tarde a/*Ø un niño. Maria made.3SG arrive late DOM a child ‘Maria made a child arrive late.’ (48) Permissives Sp. María dejó llegar tarde a/*Ø un niño. Maria let.3SG arrive late DOM a child ‘Maria let a child arrive late.’ (49) Predicates of perception Sp. María vio llegar tarde a/*Ø un niño. Maria saw.3SG arrive late DOM a child ‘Maria saw a child arrive late.’ Here are some Romanian examples:4 (50) Small clauses Rom. Ion (îl) consideră pe/*Ø student tîmpit student stupid John CL . ACC considers ‘John considers a/the student stupid.’ (51) Small clauses Rom. Obama (l)-a numit pe/*Ø senator ministr-ul sănătăţi-i senator secretary.DEF health.GEN Obama CLACC . AUX named ‘Obama designated a senator health secretary.’ 4 In Romanian, there is no clause union proper. Instead, we have subjunctive clauses, which raise some complications that we will have to ignore for the purposes of this short paper. The same goes for object control, discussed below.

262

Luis López

The fourth type of object that requires DOM is an object control. The following is an example: (52) Sp. El maestro forzó a/*Ø un niño a hacer los deberes. the homework the teacher forced DOM a child to do ‘The teacher forced a child to do his homework.’ As in the previous cases, in this sort of structure the object is not the complement of the main predicate. Instead, it has to be located in a specifier position so that it can control into the infinitival complement: (53) [VP Obji [V’ V [CP PROi . . .]]] Thus, the common property shared by these instances of obligatory DOM is that the object is not merged initially as a complement of the lexical verb. And this follows directly from the proposal outlined above. These objects cannot incorporate into the lexical verb because only objects in complement position can do so (Baker 1988, i.m.a.). Since they can’t incorporate, their unvalued Case feature leads them to scramble. And unscrambled animate objects are marked.

5 Conclusions I started the chapter by presenting some of the variety of DOM phenomena that are encountered in the Romance languages. In particular, I showed how features of the object, the verb and the configuration itself impinge on the possibility of DOM. Then I moved on to a discussion of how DOM on indefinite objects affects scope in Spanish and Romanian. To summarize, marked objects can take wide scope beyond islands while unmarked objects can only take narrow scope. Using traditional binding tests, I also showed that marked objects are located in a higher syntactic position than unmarked objects. This triple correlation between morphology, syntax and semantics is what makes DOM an exemplary grammatical interface phenomenon. Next, I used the scope properties of marked and unmarked objects as a springboard from which to revisit the classical theory of the syntax-semantics mapping. As developed in the May’s seminal work (May 1985), this classical theory is based on two features: Function Application as the mode of semantic composition for noun phrases and a direct correspondence between c-command and scope. I proposed an analysis in which the mode of semantic composition of an indefinite noun phrase depends on its position in the syntactic tree: Restrict (as defined in Chung/Ladusaw 2004) for objects inside the VP, Function Application (after type-shifting) for objects

(In)definiteness, specificity, and differential object marking

263

outside the VP. The different modes of composition yield different semantic representations and different scope possibilities. A happy result of this approach is that we reestablish a predictive approach to the syntax-semantics interface.

6 References Aissen, Judith (2003), “Differential object marking. Iconicity vs. economy”, Natural Language and Linguistic Theory 21, 435–483. Baker, Mark (1988), Incorporation, Chicago, University of Chicago Press. Bleam, Tonia (2005), “The role of semantic type on differential object marking”, Belgian Journal of Linguistics 19, 3–27. Bossong, Georg (1985), Empirische Universalienforschung. Differentielle Objektmarkierung in den neuiranischen Sprachen, Tübingen, Narr. Brugè, Laura (2000), Categorie funzionali del nome nelle lingue romanze, Milan, Cisalpino. Brugè, Laura/Brugger, Gerhard (1996), “On the accusative a in Spanish”, Probus 8, 1–51. Carlson, Greg (2003), “Weak indefinites”, in: Martine Coene/Yves D’Hulst (edd.), From NP to DP. On the syntax and pragma-semantics of noun phrases 1, Amsterdam/Philadelphia, Benjamins, 195–210. Chiriacescu, Sofiana (2009), “Indefinite NPs and PE-marking in Romanian”, Ms., University of Stuttgart. Chomsky, Noam (2000), “Minimalist inquiries. The framework”, in: Roger Martin/David Michaels/ Juan Uriagereka (edd.), Step by step. Essays on Minimalist syntax in honor of Howard Lasnik, Cambridge, MA, MIT Press, 89–156. Chung, Sandra/Ladusaw, William (2004), Restriction and saturation, Cambridge, MA, MIT Press. D’Alessandro, Roberta (2012), Merging probes. A typology of person-driven, differential object marking, Ms., University of Leiden. Dalrymple, Mary/Nikolaeva, Irina (2011), Objects and information structure, Cambridge, Cambridge University Press. Dayal, Veneeta (2011), “Hindi pseudo incorporation”, Natural Language and Linguistic Theory 29, 123–167. De Hoop, Helen (1996), Case configuration and noun phrase interpretation, New York, Garland. De Swart, Peter (2007), Cross-linguistic variation in object marking, Utrecht, LOT. Dobrovie-Sorin, Carmen (1994), The syntax of Romanian, Berlin/New York, Mouton de Gruyter. Enç, Mürvet (1991), “The semantics of specificity”, Linguistic Inquiry 22, 1–25. Escandell-Vidal, Victoria (2009), “Differential object marking and topicality. The case of Balearic Catalan”, Studies in Language 33, 832–884. Fábregas, Antonio (2013), “Differential object marking in Spanish. State of the art”, Borealis. An International Journal of Hispanic Linguistics 2, 1–80. García, Marco (2007), “Differential object marking with inanimate objects”, in: Georg A. Kaiser/ Manuel Leonetti (edd.), Proceedings of the workshop “Definiteness, specificity and animacy in Ibero- Romance languages”, Arbeitspapier 122, Fachbereich Sprachwissenschaft, Universität Konstanz, 63–84. van Geenhoven, Veerle (1998), Semantic incorporation and indefinite descriptions, Stanford, CA, CSLI. Ghomeshi, Jila (2008), “Markedness and bare nouns in Persian”, in: Simin Karimi/Vida Samiian/ Donald Stilo (edd.), Aspects of Iranian linguistics, New Castle upon Tyne, UK, Cambridge Scholars Publishing, 85–111.

264

Luis López

von Heusinger, Klaus (2002), “Specificity and definiteness in sentence and discourse structure”, Journal of Semantics 19, 245–274. von Heusinger, Klaus (2011), “Specificity”, in: Klaus von Heusinger/Claudia Maienborn/Paul Portner (edd.), Semantics. An international handbook of natural language meaning, vol. 2, Berlin/New York, De Gruyter, 1025–1058. von Heusinger, Klaus/Chiriacescu, Sofiana (2009), “Definite ‘Bare’ Nouns and pe-Marking in Romanian”, in: Maria T. Espinal/Manuel Leonetti/Louise McNally (edd.), Proceedings of the IV Nereus International Workshop “Definiteness and DP structure in Romance languages”, Arbeitspapier 124, Fachbereich Sprachwissenschaft, Universität Konstanz, 63–82. von Heusinger, Klaus/Kaiser, Georg (2011), “Affectedness and differential object marking in Spanish”, Morphology 21, 593–617. von Heusinger, Klaus/Onea, Edgar (2008), “Triggering and blocking effects in the diachronic development of DOM in Romanian”, Probus 20, 71–116. Iemmolo, Giorgio (2010), “Topicality and differential object marking”, Studies in Language 34, 239– 272. Iemmolo, Giorgio (to appear), Differential object marking, Oxford, Oxford University Press. Karimi, Simin (2003), “Object positions, specificity and scrambling”, in: Simin Karimi (ed.), Word order and scrambling, Oxford, Blackwell, 91–125. Karimi, Simin (2005), A minimalist approach to scrambling, Berlin/New York, Mouton de Gruyter. Kiparsky, Paul (1998), “Partitive case and aspect”, in: Miariam Butt/Wilhelm Geuder (edd.), The projection of arguments. Lexical and syntactic constraints, Stanford, CA, CSLI, 265–308. Klein, Udo (2007), “Clitic doubling and differential object marking in Romanian”, Ms., University of Stuttgart. Laca, Brenda (1995), “Sobre el uso del acusativo preposicional en español”, in: Carmen Pensado (ed.), El complemento directo preposicional, Madrid, Visor Libros, 61–91. Leonetti, Manuel (2004), “Specificity and differential object marking in Spanish”, Catalan Journal of Linguistics 3, 75–114. López, Luis (2012), Indefinite objects. Scrambling, choice functions and differential marking, Cambridge, MA, MIT Press. Marantz, Alec (1993), “Implications of asymmetries in double object constructions”, in: Sam A. Mchombo (ed.), Theoretical aspects of Bantu grammar, Stanford, CA, CSLI, 113–151. Mardale, Alexandru (2004), “Sur l’object direct prépositionnel en roumain”, in: Actes des VIIèmes RJC ED268 langage et langues, Paris, Paris III, 62–67. Martí, Luisa (2008), “The semantics of plural indefinite noun phrases in Spanish and Portuguese”, Natural Language Semantics 16, 1–37. Massam, Diane (2001), “Pseudo noun incorporation in Niuean”, Natural Language and Linguistic Theory 19, 153–197. May, Robert (1985), Logical form, Cambridge, MA, MIT Press. McNally, Louise (2004), “Bare plurals in Spanish are interpreted as properties”, Catalan Journal of Linguistics 3, 115–133. Milsark, Gary (1974), Existential sentences in English, PhD dissertation, Cambridge, MA, MIT. Naess, Ashild (2007), Prototypical transitivity, Amsterdam/Philadelphia, Benjamins. Pensado, Carmen (1995), “El complemento directo preposicional. Estado de la cuestión y bibliografía comentada”, in: Carmen Pensado (ed.), El complemento directo preposicional, Madrid, Visor, 11–59. Reinhart, Tanya (1997), “Quantifier scope. How labor is divided between QR and choice functions”, Linguistics and Philosophy 20, 335–397.

(In)definiteness, specificity, and differential object marking

265

Ritter, Elizabeth (1991), “Two functional categories in noun phrases. Evidence from Modern Hebrew”, in: Susan Rothstein (ed.), Syntax and semantics 26, perspectives on phrase structure: heads and licensing, New York, Academic Press, 37–62. Ritter, Elizabeth/Rosen, Sarah T. (2001), “The interpretive value of object splits”, Language Sciences 23, 425–451. Rodríguez-Mondoñedo, Miguel (2007), The syntax of objects. Agree and differential object marking, PhD dissertation, University of Connecticut. Torrego, Ester (1998), The dependencies of objects, Cambridge, MA, MIT Press. Torrego, Ester (1999), “El complemento directo preposicional”, in: Ignacio Bosque/Violeta Demonte (edd.), Gramática descriptiva de la lengua española, vol. 2, Madrid, Espasa Calpe/Real Academia Española de la Lengua, 1779–1806. Travis, Lisa (1992), “Inner aspect and the structure of VP”, Cahiers Linguistique de l’UQAM 1, 130– 146. Travis, Lisa (2010), Inner aspect, Dordrecht, Springer. Vallduví, Enric (1992), The informational component, New York, Garland.

Roberta D’Alessandro and Diego Pescarini

9 Agreement restrictions and agreement oddities Abstract: Romance languages mostly exhibit uniform agreement patterns: The finite verb shows full agreement with the subject, the modifiers agree with the head of the noun phrase, and past participles agree with promoted and clitic objects in those Romance languages that display participial agreement. There are however some exceptional agreement patterns in Romance, which will be examined here. This chapter is divided into three main parts. The first concerns agreement restrictions (or oddities) on Romance pronouns and pronominal clusters, the second examines some agreement facts in verb/argument structures, and the third one targets agreement restrictions/oddities within the DP. Keywords: agreement, agreement restrictions, PCC, impersonal si, anti-agreement effects

1 Introduction – Agreement in Romance Romance languages mostly exhibit uniform agreement patterns: The finite verb shows full agreement with the subject, the modifiers agree with the head of their noun phrase, and past participles agree with promoted objects in those Romance languages that display participial agreement. There are however some exceptional cases, i.e. constructions featuring unexpected patterns that are not found elsewhere in Romance. This chapter deals with such irregularities, which have been referred to as ‘constraints’ on agreement (for instance by Bonet 1991), agreement ‘restrictions’ (D’Alessandro/Fischer/Hrafnbjargarson 2008), ‘eccentric’ agreement (Hale 2002; Bobaljik/Branigan 2006), ‘anti-agreement’ (Ouhalla 1993), ‘exceptional’ agreement (Zwicky 1986), agreement ‘displacement’ (Bright 1957; Harris 1981), and agreement ‘mismatch’ (Corbett 1990). In the rest of the chapter, we will use the neutral terms ‘restrictions’ for those cases in which the occurrence of one element restricts or limits the occurrence of another, and ‘oddity’ for an unexpected agreement pattern that is not caused by any element involved but holds for the construction as a whole. This chapter is divided into three main parts. The first concerns agreement restrictions on pronouns and pronominal clusters, the second looks at some agreement facts in verb/argument structures, and the third one examines agreement restrictions/oddities within the DP.

268

Roberta D’Alessandro and Diego Pescarini

The first and the second part inevitably overlap: A pronoun can be an argument, thus restrictions on verb/argument agreement will also concern pronouns when we are dealing with a pronominal argument. We have nevertheless decided to keep the two sections separate, with the intention of giving more emphasis to one aspect or the other: If it seems that the pronominal nature is central to the restriction, then the phenomenon will be listed under ‘pronominal restrictions’. If it is instead the construction itself that appears to be responsible for bringing in the restriction, and this restriction/oddity also targets full DPs, the phenomenon will be filed under ‘argumental restrictions’. We have tried to provide an overview of the better-known agreement phenomena in Romance. Some have been studied for several years, in which case we have reproduced the main theoretical insights into the constructions and the most widely received analyses. Some other phenomena, on the other hand, are understudied, or almost unknown. We have decided to include these phenomena too, and to report what is known, in an attempt to draw a picture of agreement restrictions and oddities in Romance that is as accurate and inclusive as possible.

2 Agreement restrictions with pronouns This section deals with agreement restrictions involving pronouns. Before moving on to the various constructions under investigation, some general remarks are in order. First, pronouns in Romance usually encode number and person information. Gender information is restricted to 3rd person pronouns. Case was lost on full DPs in most Romance languages, with the exception of Romanian, which retains a direct/oblique distinction, and of some southern Italian varieties and again Romanian, which have a dedicated marker for vocative. However, all Romance languages have retained case distinctions on pronouns. Case in fact proves to be a crucial factor in determining some restrictions on pronominal clusters. In this section, we consider restrictions involving pronouns that do not affect the corresponding full DPs. We start by examining one of the most widely studied restrictions, the so-called PCC (Person Case Constraint), first analysed by Bonet (1991). We then turn to agreement restrictions on courtesy pronouns, which to our knowledge have not been addressed by any study so far. We continue with some PCC-like restrictions in causative constructions, which are much less well known and less widely studied. Lastly, we examine a different sort of restriction, which is not caused by the pronominal nature of the element involved in it, but does involve a pronominal element: impersonal si/se in Romance. This pronoun, when used impersonally or as an impersonal passive construction, only allows a 3rd person internal argument (Cinque 1988; D’Alessandro 2004; 2007).

Agreement restrictions and agreement oddities

269

2.1 The PCC in Italo-Romance varieties, Italian, Romanian, and French dialects Almost none of the Romance languages allow combinations of a 3rd person dative clitic and a 1st/2nd person accusative clitic. (1)

It. *Giorgio gli ti ha presentato. Giorgio to.him= you= has introduced ‘Giorgio introduced you to him.’

The restriction holds even if the dative clitic stands for a nonargumental dative1 (e.g. benefactive/malefactive adjuncts or datives of inalienable possession), or when it is the complement of a preposition as in the following example. (2)

It. *Non mi gli posso not me= to.him= can-1SG ‘I cannot sit next to him.’

sedere sit

accanto. near

The PCC holds even if the 3rd person dative is reflexive, as in (3). Notice that the same clitic combination is fine if the reflexive clitic stands for the direct object as in (3): (3)

It.

a. *Giorgio ti si è comprato Giorgio you= for.himself= is bought ‘Giorgio bought you as his slave.’ b.

come as

Giorgio ti si è presentato come Giorgio to.you= himself= is introduced as ‘Giorgio introduced himself to you as a doctor.’

schiavo. slave

dottore. doctor

Bonet (1991, 192) notices that true 1st/2nd person reflexives are better tolerated than inherent reflexives, i.e. reflexive clitics marking a particular set of unaccusative verbs deriving from transitive ones; cf. Reinhart/Reuland (1993), among others. vaig recomanar (jo mateix) ahir. (4) Cat. ??A en Pere, me li yesterday to the Pere, me= to.him= go-1.SG recommend (I self) ‘I recommended myself to him (Pere) yesterday.’ 1 An anonymous reviewer points out that bene-/malefactives are not always adjuncts and that consequently a different label should be used such as ‘applicative datives’ or ‘free adjunct datives’ in order to distinguish them from true ‘nonargumental datives’ – e.g. ethical datives – which cannot even be expressed as full DPs/PPs. We are using the term ‘nonargumental’ in a very descriptive way to refer to nonobligatory complements.

270

Roberta D’Alessandro and Diego Pescarini

Ethical datives tend to escape the restriction (Perlmutter 1971; Rouveret/Vergnaud 1980, 169–171; Bonet 1991, 197). (5)

Cat.

No me li diguis not to.me= to.him/her= tell.SUBJ ‘Don’t tell him/her lies (on me).’

mentides. lies

The acceptability of combinations of 1st/2nd person clitics is subject to crosslinguistic variation. In some languages, like Spanish or French (Bonet 1991), these combinations are reported to be completely ungrammatical (although there is no full consensus; cf. Nicol 2005), while in other languages, like Italian, some clusters are in fact very marginal, but still interpretable, at least when both elements are singular (we will see that combinations of plural clitics are generally more degraded than those formed by singular pronouns). ‘me to you / you to me’

(6) It. a.

%Mario

mi

ti

ha

presentato.

b.

??Mario

mi

vi

ha

presentato/i. ‘me to you-PL / you-PL to me’

c.

??Mario

ti

ci

ha

presentato/i. ‘you to us / us to you’

d.

*?Mario

vi

ci

ha

presentati.

‘you-PL to us’

Mario cl= cl= has introduced Romanian exhibits a different pattern, as some of the above combinations are not subject to the PCC (Săvescu 2007). In proclisis, Romanian allows combinations including a second person singular accusative clitic, as in (7), and, to a lesser extent, a first person singular accusative clitic, as in (8). (7)

(8)

Rom.

Rom.

a.

Mi tea prezentat Ion la to.me= you= has introduced John at ‘John introduced you to me at the party.’

b.

I teau recomandat ieri. to.him/her= you= has recommended yesterday ‘They recommended you to him yesterday.’

a. *Ţi ma prezentat Ion la to.you= me= has introduced John at ‘John introduced me to you at the party.’ b.

%I

petrecere. party.

petrecere. party

m- au recomandat ieri. to.him/her me has recommended yesterday. ‘They recommended me to him yesterday.’

Agreement restrictions and agreement oddities

271

Proclitic combinations are ungrammatical when the 3rd person dative clitic is reflexive, as in (9a), or when 1st/2nd person clitics are plural, as in (9b): (9)

Rom.

a. *Maria si m/te a luat drept sclav. Mary herself= me/you= has taken as slave ‘Mary has taken me/you to be her slave (for herself).’ b.

*/?? Ni

v a recomandat to.us= you.PL = has recommended ‘Mary has introduced you.PL to us.’

Maria. Mary

Combinations of singular enclitics, conversely, are always permitted. This is consistent with the above observation that clusters of plural clitics are more degraded than the others, although Nevins/Săvescu (2010) argue for an alternative explanation elaborating on the hypothesis that singular clitics in Romanian are not subject to the PCC – but only in enclisis – because they are not case-syncretic. Lastly, there are varieties in which the PCC does not hold. This is the case in several southern Italian dialects, such as that spoken in Arielli, where all the above clitic combinations are in fact allowed: (10)

Ariellese

a.

Giorgə ji t’ a prisindatə. Giorgio to.him= you= has introduced ‘Giorgio introduced you to him.’

b.

Ni mmi ji pozzə not me= to.him= can-1SG ‘I cannot sit near him.’

c.

Giorgə ti z’ a Giorgio you= for.himself= has ‘Giorgio bought you as his slave.’

assəttà sit

m’baccə. near

‘ccattatə bought

pi for

sservə. slave

The above data mean that, descriptively speaking, the PCC is a constellation of restrictions, some of which are subject to linguistic variation, rather than a single constraint. This said, there is no consensus on the nature of the restriction. Functionalist accounts observe that PCC combinations correspond to infrequent argument configurations (Haspelmath 2004), but it is not clear to us how to demonstrate that the constraint results from frequency effects and not the other way around. Moreover, it is somewhat unclear why the PCC targets combinations of clitic pronouns, while strong pronouns – which are expected to occur with the same low frequency – are unconstrained. Formal accounts differ as to whether the constraint is morphological or syntactic in nature. Morphological accounts argue that the constraint does not follow from syntactic principles (ultimately, from Agree-like procedures, Chomsky 2001), but

272

Roberta D’Alessandro and Diego Pescarini

from an extra-syntactic (say, morphological) filter preventing certain clitic pronouns or agreement affixes from co-occurring (Perlmutter 1971). In this view (cf. Bonet 1991; 1995, among others), the constraint filters certain feature bundles at the syntax-PF interface; this hypothesis explains why certain combinations of clitic pronouns or agreement markers are subject to the constraint although the corresponding featural configuration is legible at the syntax-LF interface. Alternatively, it is argued that the constraint follows from an agreement restriction (lato sensu) which occurs as a consequence of a multiple Agree configuration (Anagnostopoulou 2003; Adger/Harbour 2007; Nevins 2007, among others) or is due to a minimality restriction (Bianchi 2006; Săvescu 2007). Anagnostopoulou (2005) argues that the PCC arises as two goals compete to check the same features against a single probe. In a nutshell, let us suppose that both objects have to check against a head endowed with an uninterpretable feature F: If the indirect object checks F, the direct object cannot enter an Agree relation with the same probe and consequently the derivation ends up crashing. Conversely, if the indirect object does not check the feature F, the sentence is grammatical as the direct object is allowed to enter the Agree relation. According to this kind of explanation, the PCC ultimately resides on the featural specification of each element: F clitics trigger the PCC, while non-F clitics (hence, 3rd person accusative clitics) can occur in any clitic combination. The fact that the constraint is subject to crosslinguistic variation (cf. above) may be problematic for accounts that suggest that the restriction follows directly from a basic mechanism of Narrow Syntax. To overcome the objection, we can either argue that crosslinguistic variation depends on the featural specifications of each item (i.e. on whether or not the clitic bears a valued/interpretable feature F) or, following Nevins (2007), one may argue that the agree relation is parameterized: F stands for a constellation of binary features and, given a specific feature (e.g. [participant]), the probe can search for a single value (positive, negative, or contrastive) of that specific feature. Bianchi (2006) departs from a multiple Agree analysis and argues instead for an explanation based on Rizzi’s Relativized Minimality (Rizzi 1990). She proposes that each clitic pronoun is in a dependency relation with a Person head in the CP layer. Since such Person projections are rigidly ordered in a cartographic-like fashion, the dependency relations in a ditransitive construction may either cross each other as in (11a) or one may be nested into the other as in (11b). In the latter configuration, Relativized Minimality is violated as the lower clitic enters a dependency with the higher PersonP rather than with the nearest one: (11)

a.

Person1P

Person2P

...

clitic

clitic

b. *Person1P

Person2P

...

clitic

clitic

Agreement restrictions and agreement oddities

273

The Romance languages differ further with respect to strategies of PCC avoidance. In Ibero-Romance, for instance, the PCC is avoided by replacing the dative clitic with a strong pronoun, which is not mandatorily focused in these cases (Bonet 1991, 204). For instance, in (12a), a PCC environment, the dative clitic can be replaced by a nonfocused strong pronoun, while in (12b), where the PCC does not hold, the strong dative pronoun is mandatorily focused (focus is represented conventionally with capital letters):2 (12)

Sp.

a.

Me (*le) recomendaron a (*to.him) recommended.they to ‘They recommended me to him.’

él/ÉL. him

b.

Lo recomendaron a *él/ÉL. it/him recommended.they to him ‘They recommended it/him to him.’

Another strategy for avoiding PCC violations is the substitution of the 3rd person dative clitic with a locative exponent. This pattern is allowed in Barceloní Catalan (Bonet 1991, 209; 2008), French (Rezac 2010), and, marginally, in Italian (Pescarini 2010). In the following Catalan example, for instance, the substitution of the 3rd person dative clitic li with the locative item hi seems to overcome the PCC: (13)

Cat.

A en Pere m’ hi/*li va to the Pere me there/*to.him go.3SG ‘Josep recommended me to him (Pere).’

recomanar recommend

en the

Josep. Josep

Another controversial aspect is the type of feature(s) triggering the constraint. As the name suggests, the PCC is often regarded as a restriction on Person and Case: Abstracting away from crosslinguistic variation, the core restriction is that which prevents clitic combinations in which the accusative pronoun is [+participant]. Further research, however, has revealed that the PCC might be a constraint on animacy-related features and that person features are involved insofar as they are related to animacy: In particular, 1st/2nd person pronouns are intrinsically animate and, in many (but not all) Romance languages, the 3rd person dative clitic has become a [+human] pronoun. In French, Catalan, and Italian, for instance, only human referents can be pronominalized by the dative clitic, as in (14), while the locative clitic is used to reference nonhuman datives, as in (14) (Rigau 1982; Bonet 2008). 2 In other languages, however, the restriction in (12b) is not attested as 3rd person strong pronouns can co-occur with another focused element even in combination with a 3rd person accusative clitic: (i) MATTIA l’ ha raccomandato a lui. Mattia him= has recommended to him ‘Mattia recommended it/him to him.’

274

(14)

Roberta D’Alessandro and Diego Pescarini

It.

a.

Gli dedico molto tempo to.him= dedicate-1SG much time ‘I dedicate much time to him (Carlo).’

b. *Gli/ci dedico molto tempo to.it= dedicate-1SG much time ‘I dedicate much time to it (soccer).’

a Carlo. (to Carlo)

al calcio. (to soccer)

The fact that the dative clitic is restricted to human referents is, historically speaking, a puzzle. In fact, third person dative clitics – like third person accusative clitics – derive from the Latin demonstrative ILLE ‘that’, which in origin did not exhibit any animacy-related restriction. This means that the above restriction emerged as soon as the dative determiner ILLI (S ) became a clitic pronoun, but to the best of our knowledge no proposal has been put forward in the recent literature that could explain a development of this type (cf. Pescarini 2015 for a tentative analysis based on the parallelism between Romance cliticization and English double object constructions). The hypothesis that the PCC is an animacy restriction may shed light on the contrast in (14): Since the locative clitic is not endowed with an animacy-related feature, it is not subject to the PCC even if it exceptionally stands for a 3rd person animate. Not in all Romance languages, however, are 3rd person dative clitics restricted to human referents. In Ibero-Romance, for instance, the 3rd person clitic le(s) may stand for an inanimate noun (notably, this is allowed in a language that displays no locative clitics like It. ci, Fr. y, or Cat. hi, which are normally used to pronominalize inanimate datives). Even if they reference an animate element, however, 3rd person datives are subject to the PCC, cf. (15) (Ormazabal/Romero 2007), although Bonet (2008) observes that the ungrammaticality of (15) persists even if the dative clitic is omitted, as in (16). The ungrammaticality of (15) must therefore follow from some orthogonal constraint. (15)

Sp. *Te le pongo a ti (de pata) you= to.it= I.put a you (as leg) ‘I assemble you as a leg of the table.’

(16)

Sp. *Te pongo a ti (de pata) a you= I.put a you (as leg) to ‘I assemble you as a leg of the table.’

la the

a to

la the

mesa. table

mesa. table

Further evidence for an animacy-based analysis of the PCC comes from Leista Spanish, namely those Ibero-Romance dialects in which the dative clitic le (pl. les) may stand for human direct objects. In these varieties, the clitic le is subject to the PCC even if it stands for the direct object: As shown in (17), the le meaning ‘him’ cannot combine with a 1st/2nd person dative clitic. In this environment, Leista speakers must retreat to the exponent lo, as in (17) (Ormazabal/Romero 2007).

Agreement restrictions and agreement oddities

(17)

Sp.

275

a. *Te le di. to.you= him= give-PST.1SG ‘I give him to you.’ b.

Te lo di. to.you= it= give-PST.1SG ‘I give it to you.’

Things become even more complicated in dialects of French and Italian where the 3rd person dative clitic is always expressed by the locative clitic (e.g. Fr. y instead of lui/leur ‘to him/her/them’) as a consequence of historical changes that made the etymological 3rd person dative clitic form fall out of use (Calabrese 1994). The dative/locative syncretism affects the PCC in two opposite ways: It may prevent the PCC, i.e. the 3rd person dative clitic is free to occur with a 1st/2nd person clitic (Rezac 2011), or the PCC may be extended to the locative clitic, i.e. the dative/ locative clitic cannot occur with 1st/2nd person clitics even when it has a locative interpretation. For instance, in certain northern Italian dialects like Vicentino the locative clitic ghe, which is syncretic with the 3rd person dative clitic, is free to combine with 1st/2nd person singular pronouns, while the combinations with plural clitics are – again – more degraded. (18)

Ventino

??ne

us=

ghe there=

porta brings

Carlo. Carlo

b. ??ve you.PL =

ghe there=

porta brings

Carlo. Carlo

a.

2.2 PCC-like effects in causative constructions In various Romance languages, the causee can occur as either a dative complement or as a PP headed by the preposition da (It.), par (Fr.) etc. (Kayne 1975). For the sake of consistency, many of the following data are from Italian, but the same holds for other Romance languages. (19)

It.

a.

Micol fa pettinare Giulia a Giulia to Micol make-3SG comb ‘Micol makes Carlo comb Giulia’s hair.’

Carlo. Carlo

b.

Micol fa pettinare Giulia da Giulia by Micol make-3SG comb ‘Micol makes Carlo comb Giulia’s hair.’

Carlo. Carlo

Unlike the a-causee, the da-phrase cannot be resumed by a dative clitic:

276

(20)

Roberta D’Alessandro and Diego Pescarini

It.

a.

A Carlo, Micol gli fa Micol to.himi= make-3SG to Carloi ‘Micol makes Carlo comb Giulia’s hair.’

pettinare comb

Giulia. Giulia

b. *Da Carlo, Micol gli fa by Carloi Micol to.himi= make-3SG ‘Micol makes Carlo comb Giulia’s hair.’

pettinare comb

Giulia. Giulia

What is of interest here is that dative causees, regardless of their clitic or phrasal status, trigger a sort of PCC: In fact, 1st/2nd person clitic pronouns cannot co-occur with the a-causee, while they can occur when the causee is introduced by the preposition da; cf. (21a) vs. (21b). As in the canonical PCC pattern, 3rd person clitics, by contrast, are always unconstrained; cf. (22a) vs. (22b). (21)

(22)

It.

It.

a. *Micol Micol

mi me=

pettinare comb

a to

Carlo. Carlo

b.

Micol mi fa pettinare Micol me= make-3SG comb ‘Micol makes Carlo comb my hair.’

da by

Carlo. Carlo

a.

Micol Micol

pettinare comb

a to

Carlo. Carlo

b.

Micol la fa pettinare Micol her= make-3SG comb ‘Micol makes Carlo comb her hair.’

da by

Carlo. Carlo

la her=

fa make-3SG

fa make-3SG

The restriction in (21) is sometimes referred to as the Fancy Constraint (Postal 1989) and has received far less attention than the canonical PCC. In fact, the link between the PCC and the Fancy Constraint is far from straightforward. Besides the fact that the former, unlike the latter, targets only clitic combinations, they differ with respect to the behaviour of reflexives. In fact, the Fancy Constraint targets every type of reflexive clitic, including 3rd person, while 3rd person direct objects are not subject to the PCC, regardless of whether or not they are reflexives: (23)

It.

Si fa visitare da(/*a) to/*from him/her-self= make-3SG visit ‘He/she makes Linda visit him/her.’

Linda. Linda

Moreover, causative environments allow us to observe how two dative arguments (namely, the causee and the indirect object) do co-occur. Speakers allow sentences in which the causee is clitic and the indirect object is phrasal, as in (24a), while the opposite configuration is rejected, namely, phrasal causee and clitic indirect object, cf. (24b), unless the causee is expressed by a PP as in (24b’):

Agreement restrictions and agreement oddities

(24)

It.

Le faccio telefonare to.her= make-1SG phone ‘I make her phone Carlo.’

a to

Carlo. Carlo

b. *Le faccio telefonare to.her= make-1SG phone ‘I make Carlo phone her.’

a to

Carlo. Carlo

a.

b’.

Le faccio telefonare to.her= make-1SG phone ‘I make Carlo phone her.’

da from

277

Carlo. Carlo

When both arguments are 3rd person clitics, the sentence is ungrammatical if both clitics climb; cf. (25a) vs. (25b) (Kayne 1975, 297): (25)

Fr.

a. *Elle she b.

me me=

Elle me she me=

lui to.him

présentera, introduce-FUT.3SG

présentera introduce-FUT.3SG

à to

lui. him

Double dative constructions are allowed if and only if the causee is 1st/2nd person and the indirect object is 3rd person. The acceptability of such combinations is subject to crosslinguistic variation: They are not allowed in Italian and Spanish, while some Italian dialects and French are more liberal (on French, cf. Strozer 1976, 171; Rezac 2010). Notice that the acceptability does not depend on the linear order of clitics. The following set of examples from French and Italian dialects shows that the restriction holds regardless of the linear order of pronouns. (26) Fr.

Vicentino

a. Je vais te le lui faire donner. I= go to.you.CAUSEE = it.DO = to.him.IO = make give b. Te ghe lo fasso portare. to.you.CAUSEE = to.him.IO = it.DO = I.make bring

S. Val., Abr.3 c. jə tə lu faccə purtà. to.him.IO = to.you.CAUSEE = it.DO = I.make bring ‘I make you bring it to him.’

2.3 Courtesy forms Courtesy forms are normally used to avoid direct reference to the hearer. In Italian, either 3rd person feminine pronouns or 2nd person plural pronouns may be used 3 Southern Italian dialect of San Valentino in Abruzzo Citeriore.

278

Roberta D’Alessandro and Diego Pescarini

as courtesy forms. With respect to the PCC, courtesy forms behave like 2nd person clitics, regardless of their apparent 3rd person morphology: (27)

It. *Giorgio glie l’ ha Giorgio to.him= her(‘you’)= has ‘Giorgio introduced you to him.’

presentata. introduced

Courtesy forms exhibit a rather puzzling pattern of agreement. They normally exhibit grammatical agreement with the inflected verb, but semantic agreement with other constituents such as adjectives or past participles. Hence, if the hearer is masculine, adjectives and participles will display the masculine singular ending regardless of the morphology of the courtesy pronoun: (28)

It.

Lei è simpatico. she(‘you’) is nice-M . SG ‘You are nice.’

(29)

It.

Voi siete simpatico. you.PL are nice-M . SG ‘You.SG are nice.’

If the courtesy form is an object clitic, however, the honorific systems of Italian diverge as the 3rd person feminine singular courtesy form always exhibits grammatical agreement: (30)

It.

L’ ho vist -a/*-o her(=you) I.have seen-F. SG /*M . SG ‘I have often seen you here.’

(31)

It.

Vi ho vist-o/-a you.PL = I.have seen-M . SG /F. SG ‘I have often seen you.SG here.’

spesso often

spesso often

qui. here

qui. here

To sum up, with respect to the PCC, courtesy forms behave like 2nd person singular pronouns: They cannot co-occur with a dative clitic even if they are morphologically 3rd person. Furthermore, courtesy forms display a puzzling mismatch between grammatical and semantic agreement when the courtesy form agrees with a nominal element (a predicative adjective or the past participle): Normally agreement is controlled by the referent save for the clitic la, which triggers grammatical feminine agreement (on further grammatical/semantic gender mismatches, cf. section 4.2).

Agreement restrictions and agreement oddities

279

2.4 Impersonal si/se constructions in Italian, Spanish, and Romanian The se/si pronoun in Romance has several uses: It can be used as a reflexive (in which case it displays a fully-fledged paradigm), as an inchoative, as an aspectual marker, and as an impersonal ‘subject’. It is the agreement patterns of the last of these that interests us here, in sentences like (32): (32)

It.

Si vedono molte automobili in questo in this si= see.3PL many.F. PL car.F. PL ‘One sees many cars in this neighbourhood.’

quartiere. neighbourhood

The exact status of si in (32) is much debated, but it is not strictly relevant here. What matters is that in these constructions there is quirky agreement between the finite verb and the internal argument (in the case of the example, both automobili and vedono are plural), which bears nominative. This construction can only have a 3rd person internal argument (Burzio 1986; Cinque 1988). A 1st or 2nd person pronoun is banned. (33)

It. *Vi si vedono in you.PL = si= see.3PL on ‘One can see you on TV.’

televisione. TV

Note that there is a parallel construction in which the internal argument does not agree with the finite verb and carries accusative case. This construction is illustrated in (34) and does not present the person restriction. (34)

It.

Lo/vi si vede. it/you= si= sees.3SG ‘One sees him/you.PL .’

In (34) the internal argument carries accusative, and the verb is inflected as 3rd person singular. This has led many linguists (most notably, Cinque 1988) to assume that si has a different status, i.e. argumental or nonargumental, in agreeing vs. nonagreeing constructions, respectively. If si is argumental, it withdraws the external theta-role and blocks accusative, thus making it impossible for the object to receive Accusative case. For the Case filter, the object will then need to agree with the inflectional head T. Cinque does not discuss the agreement restriction in detail. If si is not argumental, accusative can be assigned to the object. A further complication arises when observing the difference between impersonal si constructions with unaccusative verbs, which show plural agreement on the

280

Roberta D’Alessandro and Diego Pescarini

predicative adjective or on the past participle, and impersonal si constructions with unergative verbs, which show default masculine singular agreement instead. (35)

It.

Si è arrivat-i. si= is.3.SG arrived-M . PL ‘One has arrived/we have arrived.’

(36)

It.

Si è lavorat-o. si= is.3.SG worked-M . SG ‘One has worked/we have worked.’

An easy way to account for the singular/plural alternation on the participle would be to consider si as argumental in each of these constructions. In (35), as the verb is unaccusative, si can be the internal argument. In (36), with an unergative verb, si is an external argument. The obvious question then is why we do not see plural agreement on the auxiliary in (35). In sentences with a bona fide pronominal argument, like for instance loro, we see agreement on both the auxiliary and the participle, as illustrated in (37). (37)

It.

Loro sono they are.3.PL ‘They arrived.’

arrivat-i. arrived-M . PL

This leaves us with a dilemma: If impersonal si is argumental, why do agreement paradigms arise like those seen above? If it is not, then what exactly is it? And, for the purposes of the present chapter: Why does si trigger agreement restrictions on Nominative objects and only partial agreement on the auxiliary? Different answers have been provided to these questions over the years. Cinque (1988), as summarized above, has proposed a different argumental status for the two sis in the two constructions. In one case si is a quasi-argument which cannot absorb accusative; hence, accusative is assigned normally to the internal argument. In the object-agreeing constructions, instead, si is an argument that creates a semi-passive construction by absorbing the external theta role and blocking accusative assignment, in compliance with Burzio’s generalisation. Given that accusative cannot be assigned to the internal argument, this must take nominative. The agreement restriction is due to the arbitrary nature of si. According to D’Alessandro (2004; 2007), si does not have two different statuses, nor does it absorb or block Case in any way. Si is a 3rd person pronoun, bearing a 3rd person feature as well as an unvalued number. This pronoun incorporates on the T head, hence valuing the verb as 3rd person. The constructions in which agreement takes place between T and the internal argument are similar to Icelandic quirky dative constructions: Si is in any case an external argument, but T is not a

Agreement restrictions and agreement oddities

281

fully transitive head. Si is 3rd person, and it is incorporated on T; T agrees with the internal argument to get its features valued. At this point, a condition on multiple agreement applies (the condition proposed by Anagnostopoulou 2003 and discussed in 2.1. for the PCC), licensing only the internal argument that hosts the same person feature as si (namely, 3rd). A different feature specification would cause a feature mismatch on the T head, with consequent derivation crash. The inclusive reading that emerges in sentences like (35) or (36) is due to a semantic/pragmatic feature, parasitic on person, which is directly linked to the Speech Act projection. The inclusive reading (i.e. the ‘we’ reading) is shown to be determined by event boundedness, and to be available with all verb classes (contra Cinque 1988). The same multiple agreement restriction holds, as mentioned, in Icelandic quirky dative constructions as well as in Spanish olvidarse constructions, which cannot have an inner argument other than 3rd person. Thus, the agreement restriction arises according to D’Alessandro because of the syntactic structure in which si occurs, not because of its different status. The two constructions, with and without object agreement, are structurally different in that one denotes a bounded event and one does not. Where there is a bounded event, an inner aspectual head is present in the v field. The mismatch between the singular auxiliary and the plural participle illustrated in (35) is again due to the fact that si is a 3rd person pronoun that incorporates on T, valuing it as 3rd person singular. Number remains unvalued, and it is marked as default at lexical insertion as a Match of two unvalued features. In unaccusatives, the past participle probes si, which is merged as an internal argument, because of the fact that v is not a phase head in unaccusatives. The plural value is assigned to the participle by [arb] feature, which once again gets valued by the referents of the Speech Act. The construction always has an inclusive reading, as also noted by Cinque (1988).

3 Agreement restrictions with arguments 3.1 Anti-agreement effects with postverbal and/or dislocated subjects in Tuscan and Ligurian A number of Romance dialects, most notably those spoken in central Italy, have been reported to display a curious agreement effect. This effect, which we will call anti-agreement, as it is reminiscent of a similar phenomenon found in Arabic and Berber, consists in a lack of agreement with subjects in postverbal position (Corbett 1979; Brandi/Cordin 1989; Fassi Fehri 1993; Saccon 1993).

282

Roberta D’Alessandro and Diego Pescarini

This agreement pattern is discussed Saccon (1993) and Cardinaletti (1997) from a formal syntactic point of view. Cardinaletti reports, for the variety of Anconetano, sentences like (38), where we see full agreement between the subject and the finite verb only if the subject is preverbal. (38) Anconetano a. Questo, lo fa sempre i bambini. this.ACC it does always the children.NOM b. *Questo, i bambini lo fa this.ACC the children.NOM it-ACC does

sempre. always

c. Questo, i bambini lo fanno sempre. this.ACC the children.NOM it-ACC do always In (38), we see that full agreement takes place when the subject is in its canonical preverbal position. If the subject i bambini is in postverbal position, the finite verb will show a default 3rd singular ending. Saccon and Cardinaletti both analyse this construction along the same lines, namely by arguing that full agreement only takes place when the subject is VP-internal or in canonical Spec,TP position. When the subject is extraposed, the verb agrees with a pro which forms a chain with the overt subject for case assignment. Recently these data have been brought to the centre of syntactic debate by Noam Chomsky, who, in a series of talks about labelling, as well as in a paper (Chomsky 2013), has argued that anti-agreement can be attributed to a labelling issue. Chomsky mentions Rizzi’s observation regarding the fact that in an XP YP configuration (i.e. when the subject is in Spec,TP in traditional terms) agreement must always be full. For further speculations on this, cf. D’Alessandro (2013).

3.2 Inflected infinitives in Portuguese and Sardinian Romance languages display head movement of the finite verb to T, the head hosting tense and agreement.4 Infinitives are not inflected for number and person in Romance. This means that they are generally assumed not to move to T. There are however some well-known exceptions: European Portuguese (Por.), and to some extent Brazilian Portuguese, Galician, and Sardinian, all have inflected infinitives of the form illustrated in (39) for Portuguese:5 4 In previous stages of generative syntax, this head was split into an I/T head and an AgrS head, the former encoding tense/aspectual information, the second proper φ-agreement information. Before then, I or Infl was considered to host an agreement feature (Agr) and a tense feature (T), the latter assigning Nominative to the subject under government (later, simply in a specifier-head configuration). 5 It has been shown that, although Romance infinitives do not move to T, they move to some intermediate position between V and T. For a detailed overview of the position occupied by infinitive as well as finite verbs in Romance, cf. Ledgeway/Lombardi (2005) and Ledgeway (2012).

Agreement restrictions and agreement oddities

(39)

Por.

283

a.

(para) eu falar

(for) I to-speak-1.SG

b.

(para) tu falares

(for) you to-speak-2.SG

c.

(para) ela falar

(for) she to-speak-3.SG

d.

(para) nós falarmos

(for) we to-speak-1.PL

e.

(para) vocês falarem

(for) you to-speak-2.PL

f.

(para) elas falarem

(for) they to-speak-3.PL (from Madeira 1994, 180)

Inflected infinitives can occur with overt as well as null referential subjects, and are mainly licensed in embedded clauses (Raposo 1987; Madeira 1994; Ambar 1994; Sitaridou 2002; Mensching 2000), in infinitival subject clauses, or in adjuncts headed by a preposition. The following example, taken from Raposo’s (1987) influential work, exemplifies some of the contexts in which inflected infinitives appear in EP. (40) Sp. a. Será difícil [eles aprovarem a proposta]. be.FUT.3SG difficult they.3PL approve.INF.3PL the proposal ‘It will be difficult for them to approve the proposal.’ b. *Será

difícil

[eles

aprovar___ a proposta]. (adapted from Raposo 1987, 86)

Raposo’s analysis accounts for the presence of inflection on an infinitive (which does not have Tense) by adopting a model whereby tense and agreement are separate features on Infl, and by proposing that they can be specified independently. A head specified as [+T] is usually able to assign Nominative. A head specified as [+Agr] can assign case, when T is not finite, only when Agr is itself marked for case. Take for example (40). Its structure is as in (41): (41)

284

Roberta D’Alessandro and Diego Pescarini

In (41), the subject infinitive [eles aprovar a proposta] is extraposed and coindexed with pro in Spec,Infl2. Since Infl2 is specified as [+T], it can assign Nominative case to pro. The case of pro is then “passed” to the InflP with which it forms a CHAIN . From InflP, Nominative percolates down to its head Infl, and hence to +Agr. At this point, this Infl head with no +T can assign Nominative to eles in its specifier. The basic idea of having Nominative assigned by noninflected T in some languages, or in some specific constructions, is also adopted by Mensching (2000) and Ledgeway (2000), for Sardinian and some southern Italian dialects, respectively. Whether a [–T] head can or cannot assign Nominative is considered to be a parameter. Inflected infinitives, as we have seen, are mainly restricted to embedded clauses. Not all verb types can license an inflected infinitive, however. Madeira (1994) provides a list of possible contexts for inflected infinitives in Portuguese, which are licensed as complements to declarative/epistemic predicates, complements to factive predicates, complements to perception verbs, and complements to causative predicates. As we have seen in the case of (40), they can also appear in infinitival subject clauses, and in adjunct clauses introduced by a preposition. Finally, observe that overt subjects can be licensed in some contexts by infinitives in Romance. We will not address this issue here as it is not directly relevant to agreement facts, given that we do not see inflection. The reader is referred to Ledgeway (2000) and Mensching (2000) for an overview of these constructions.

3.3 Agreement mismatch marking and omnivorous agreement in Abruzzese Finite verbs in Romance do not show gender agreement. The variety spoken in Ripatransone (Ascoli Piceno, Ripano henceforth), however, does. This variety has a fully-fledged paradigm for masculine and feminine finite verbs, as exemplified in (42). (42) Ripano a. I’ ridu (‘I laugh’-M . SG ) tu ridu (‘you laugh’-M . SG ) issu ridu (‘he laughs’-M . SG ) noja ridemi voja rideti issi ridi c. i’so risu (‘I have laughed-M . SG ) tu sci risu issu e risu noja semi risi voja seti risi

b. ìa ride (‘I laugh’-F. SG ) tu ride (‘you laugh’-F. SG ) esse ride noja ridema voja rideta essa ride d. ìa so rise (‘I have laughed’-F. SG ) tu si rise esse e rise noja sema risa voja seta risa (Rossi 2008, 3)

Agreement restrictions and agreement oddities

285

Interestingly, Ripano also displays agreement mismatch marking in transitive constructions: If the arguments of the verb have different gender or number specifications, the finite verb (and often the auxiliary) will exhibit a special agreement mismatch marker, namely -ə: (43)

Ripano

a.

Babbu dicə dad-M . SG says-3RDSG . N ‘Dad tells the truth.’

le the-F. SG

vərità. truth-F. SG (Mancini 1988, 107)

b.

So magnatə lu pani. the-M . SG bread-roll-M . SG am eaten-N ‘I(F ) have eaten the bread roll.’

These patterns have not been widely studied. D’Alessandro (2016) proposes that these mismatches are due to the fact that v in this variety is a complex probe, which is made up of two probes SHARING their features (Ouali 2008). Each of the two vs targets one argument. Given a requirement on uniformity of agreement, of the sort proposed by Anagnostopoulou (2003; 2005) and already discussed for the PCC, the two vs will have to show the same agreement ending. If this uniformity is not granted, because the two arguments have different featural specifications, the agreement ending of the verb will be a mismatch marker. The same complex v structure is found, according to D’Alessandro, in neighbouring dialects. In particular, the dialect spoken in Arielli shows omnivorous number (D’Alessandro/Roberts 2010; D’Alessandro/Ledgeway 2010; D’Alessandro 2016): The finite verb (and the auxiliary) will agree with whichever argument is plural. Plural agreement is thus selected whenever plural appears on any argument of a transitive (or even a ditransitive) verb, as exemplified in (44). (44) Ariellese a. Giuwannə a pittatə nu murə. have-3 painted-PP. SG a.SG wall-M John-SG ‘John has painted a wall.’ [sg SUBJ – sg OBJ] b. Giuwannə a pittitə ddu murə. have-3 painted-PP. PL two walls-M John-SG ‘John has painted two walls.’ [sg SUBJ – pl OBJ] c. Giuwannə e Mmarijə a pittitə nu murə. John and Mary-PL have-3 painted-PP. PL a.SG wall-M ‘John and Mary have painted a wall.’ [pl SUBJ – sg OBJ] d. Giuwannə e Mmarijə a pittitə ddu murə. John and Mary-PL have-3 painted-PP. PL two walls-M ‘John and Mary have painted two walls.’ [pl SUBJ – pl OBJ] (D’Alessandro/Roberts 2010, 45)

286

Roberta D’Alessandro and Diego Pescarini

Here, as in Ripano, the complex v targets both arguments. If one of the two arguments is specified as plural, the agreement ending inserted in this case will be plural.

4 Agreement restrictions within the DP 4.1 Agreement asymmetries In various Romance languages, DP-internal agreement is asymmetric: Agreement is mandatory with postnominal modifiers, while prenominal modifiers lack agreement. Rhaeto-Romance varieties, for instance, exhibit a pattern of partial agreement which Haiman/Benincà (1992, 219–222) term the ‘Ladin lazy agreement rule’. In some Central Ladin varieties only the element in DP-final position exhibits feminine plural morphology, expressed by the suffix -es. The leftmost elements of the DP, by contrast, never display feminine plural endings. This is exemplified in the following examples from a dialect spoken in the Fassa Valley (Rasom 2008): (45)

Fassa Valley

l-a cès-es the-F. SG house-F. PL ‘the houses’

With adjectives, the possible patterns are as follows: With a prenominal adjective, the plural ending occurs only on the noun; with a postnominal adjective, the plural ending occurs either on the adjective or on both the noun and the adjective. (46)

Fassa Valley

a.

la pìcola cès-es the-F. SG small-F. SG house-F. PL ‘the houses, which are all small’

b.

la cèsa-F. SG pìcol-es small-F. PL the-F. SG house ‘those houses that are small’

c.

la cès-es pìcol-es the-F. SG house-F. PL small-F. PL ‘the houses, which are all small’

Observe that (46c), where the adjective and the noun agree in number, has the same restrictive interpretation as (46a) with the prenominal adjective. This led Rasom (2008) to argue that the differences between (46a) and (46b) in terms of agreement morphology and interpretation follow from two different syntactic sources of adjectives (Cinque 2010): Attributive adjectives are generated as reduced relative clauses

Agreement restrictions and agreement oddities

287

above appositive adjectives. Furthermore, she argues that Number is encoded in a dedicated projection close to the noun and that number features spread within the DP: (47) [DP . . . [Reduced Relative

Clause

A

. . . [AP A . . . [Num {pl}

[NP N

The hypothesis is that, in Ladin, number spreads downwards, i.e. an adjective will exhibit number inflection if the noun moves above it. The problematic example is therefore (46b) in which the noun preceding the adjective does not display plural -es. According to Rasom, the exceptionality of this pattern depends on the clausal nature of attributive adjectives. With attributive adjectives, the noun acts as the antecedent of the (reduced) relative clause, while the adjective occupies a predicative position inside the clause. As such, the NP is not required to move to Spec,NumP (and, consequently, to exhibit number morphology), while the adjective is free to agree in number under clausal agreement, which is not subject to the same restriction of DP-internal agreement/concord. Another pattern of lazy agreement is shown in Ibero-Romance with feminine nouns (e.g. agua ‘water’) which, when singular, select for a masculine article (arguably, the phenomenon originated from a dissimilation rule as normally happens before words beginning with a). (48)

standard Sp.

el/*la agua

In some dialects, however, the lack of agreement has been extended to other prenominal modifiers: (49)

dialects of Sp.

a.

el the.M

nuevo new.M

arma weapon.F

secreta secret.F

b.

el the.M

mismo same.M

agua water.F

parecerá will.seem

fría cold.F

Cardinaletti/Giusti (2011; 2015) deal with another asymmetric pattern displayed by some Italo-Romance dialects. It concerns three nominal modifiers: the partitive article del ‘of the’, the distal demonstrative quel (‘that’), and the adjective bel ‘nice’. In Italian, as well as in several Italo-Romance varieties, their endings coincide with the form of the definite article. Like the definite article, the endings of del, bel, and quel are subject to context-determined allomorphy (e.g. M . SG dello, bello, quello occur before words beginning with sC; del, bel, quel before other Cs; del’, bell’, quell’ before Vs).

288

Roberta D’Alessandro and Diego Pescarini

M . SG

F. SG

M . PL

F. PL

a.

Partitive article

de-l/lo

de-lla

de-i/gli

de-lle

b.

Adjective

be-l/llo

be-lla

be-i/gli

be-lle

c.

Demonstrative

que-l/llo

que-lla

que-i/gli

que-lle

(50) It.

Furthermore, in Anconetano (a central Italian dialect), the plural ending -i can be dropped giving rise to the forms de’, be’, and que’. When dei/quei and bei co-occur, the possible patterns of i-dropping are the following: a) both is occur, b) the higher is dropped, or c) both are dropped. Otherwise, if only the intermediate i is dropped, the sequence becomes ungrammatical, cf. (51d/52d). Descriptively, -i can only spread bottom-up. (51)

Anconetano

a.

dei

bei

fioli

b.

de’

bei

fioli

c.

de’

be’

fioli

d. *dei be’ fioli some nice boys ‘some nice boys’ (52)

Anconetano

a.

quei

bei

fioli

b.

que’

bei

fioli

c.

que’

be’

fioli

d. *quei be’ fioli those nice boys ‘those nice boys’ To account for bottom-up effects, Cardinaletti/Giusti argue that DP-internal agreement follows from a peculiar feature sharing mechanism (Giusti 2008) due to the combination of two basic operations: projection (bottom-up feature sharing across the functional spine of the DP) and concord (i.e. feature sharing between a head and its specifier). Bottom-up agreement is therefore due to the combination of projection along the structure of the DP and concord between a functional head F° and the modifier hosted in its specifier. (53)

Agreement restrictions and agreement oddities

289

Moreover, Cardinaletti/Giusti argue that de-, be-, and que- are inflectionless elements: Their “endings” are therefore the spell-out of the head F°, whose φ features are projected from below. Not all the asymmetries attested in Romance, however, are consistent with bottom-up effects. Asturian, for instance, exhibits a mixed pattern in which gender spreads to prenominal elements while mass/count agreement spreads postnominally: Prenominal adjectives always display masculine/feminine agreement (e.g. -u/a), while postnominal adjectives exhibit mass/count agreement (e.g. -o/u): (54)

(55)

Asturian

Asturian

fierru

a.

duru

ferruñosu

M, COUNT

b.

duru fierro ferruñoso hard iron rusty ‘hard rusty iron’

a.

guapa

manzana

madura

F, COUNT

b.

guapa manzana good apple ‘good ripe apple’

maduro ripe

F, MASS

M, MASS

This shows that gender features spread bottom-up, while mass/count features seem to spread top-down. Another case of top-down agreement is exhibited by a number of southern Italian varieties, where prenominal modifiers (determiners and some adjectives) show overt agreement endings, while the endings of the noun and its postnominal modifiers are subject to centralization (namely, -a/e/i/o/u > -ə): (56)

Southern It.

a.

‘o the

bellu nice

ciorə flower

b.

‘o the

ciorə flower

bellə nice

The fact that both bottom-up and top-down asymmetries are found in Romance (sometimes in the same language, as in the case of Asturian) has led to the postulation of different types of feature sharing operations such as agree vs. projection/ concord. Furthermore, on the basis of these data, different ‘layers’ of agreement may be postulated (cf. also Ackema/Neeleman 2012), i.e. a syntactic mechanism responsible for postnominal concord, usually via specifier-head agreement (Guasti/ Rizzi 2002), and a postsyntactic one wherein agreement is obtained by means of output constraints (cf. Samek-Lodovici 2002 on clausal agreement; Bonet/Mascaró 2011; Bonet 2013; Bonet/Lloret/Mascaró 2015).

290

Roberta D’Alessandro and Diego Pescarini

4.2 Gender agreement restrictions in conjoint DPs in French Sleeman/Ihsane (2013) deal with gender agreement in French with nouns displaying a conflict between grammatical and semantic gender (namely, sex) such as enfant ‘child’, professeur ‘teacher, professor’, sentinelle ‘sentinel’ etc. These nouns may reference either masculine or feminine individuals. In the latter case, grammatical gender always controls agreement inside the strict DP, while semantic gender may control agreement outside the strict DP. Take for instance a noun like professeur ‘teacher, professor’, which triggers masculine agreement even if it refers to a female as in (57). (57) Fr. le bon professeur the.M good.M teacher.M However, DP-external agreement, as in the case of a predicative adjective, must be feminine, i.e. it must agree with the semantic sex rather than with the grammatical gender: (58) Fr. Mon ancien professeur de français était toujours content-*(e) my.M former.M professor of French was always satisfied-F de mon travail. of my work ‘My former French teacher was always satisfied with my work.’ The authors argue that grammatical gender is a grammatical uninterpretable feature (Zamparelli 2008, among others) which is encoded separately from semantic gender. On the separation between semantic and grammatical gender there are several possible views. Kramer (2009) argues for a morphological analysis in which nouns are formed through the combination of a nominalizing head n with a categoryneutral root √ (Marantz 1997; 2001). Semantic gender is encoded by n and, if n lacks gender, the agreeing gender is the grammatical one, encoded on the root. Sleeman/ Ihsane, conversely, adopt a syntactic view in which the extended DP contains a Gender projection GenP encoding semantic gender, while grammatical gender is encoded by the NP.

5 Summary In this chapter we have dealt with a series of prima facie irregularities regarding the realization of agreement endings and the distribution of pronominal elements.

Agreement restrictions and agreement oddities

291

Two sections are devoted to clausal agreement: one focusing on mutual exclusion patterns between pronouns and the other addressing patterns of agreement between a verbal form and one or more nominal forms. The last section has taken DP agreement into consideration. In the first section we summarized data and analyses concerning the distribution of (clitic) pronouns. In fact, Romance clitics cannot occur freely as their combinations are subject to systematic gaps. Besides the canonical PCC (which is subject to a certain degree of crosslinguistic variation), we observed the behaviour of clitic combinations in causative constructions, the syntax of honorific systems, and the agreement possibilities of impersonal si/se. The second section deals with verbal agreement: We focused on languages lacking agreement on finite forms (such as central Italian dialects) and, conversely, languages showing person agreement on nonfinite forms (e.g. European Portuguese and Sardinian). Lastly, we mentioned cases of varieties like Ripano, in which verbal morphology exhibits gender agreement. The third section is about DP agreement/concord. We observed that the Romance languages show both top-down and bottom-up effects, i.e. either prenominal or postnominal modifiers may fail to agree with the noun on a language-specific basis. We submitted the hypothesis that this might be due to the existence of various kinds of feature-sharing operations within the DP, possibly applying in different stages of the derivation. Lastly, we observed patterns of DP-external agreement in cases of a mismatch between semantic and grammatical gender.

6 References Ackema, Peter/Neeleman, Ad (2012), “Agreement weakening at PF. A reply to Benmamoun and Lorimor”, Linguistic Inquiry 43, 75–96. Adger, David/Harbour, Daniel (2007), “Syntax and syncretisms of the person case constraint”, Syntax 10, 2–37. Ambar, Manuela (1994), “Aux-to-Comp and lexical restrictions on verb movement”, in: Guglielmo Cinque/Jan Koster/Jean-Yves Pollock/Luigi Rizzi/Raffaella Zanuttini (edd.), Paths towards universal grammar, Washington, D.C., Georgetown University Press, 1–24. Anagnostopoulou, Elena (2003), The syntax of ditransitives. Evidence from clitics, Berlin/New York, De Gruyter. Anagnostopoulou, Elena (2005), “Strong and weak person restrictions. A feature checking analysis”, in: Laurie Heggie/Francisco Ordóñez (edd.), Clitic and affix combinations, Amsterdam/Philadelphia, Benjamins, 199–235. Bianchi, Valentina (2006), “On the syntax of personal arguments”, Lingua 116, 2023–2067. Bobaljik, Jonathan David/Branigan, Phil (2006), “Eccentric agreement and multiple case checking”, in: Alana Johns/Diana Massam/Juvenal Ndayiragije (edd.), Ergativity. Emerging issues, Dordrecht, Springer, 47–77. Bonet, Eulàlia (1991), Morphology after syntax. Pronominal clitics in Romance, PhD dissertation, Massachusetts Institute of Technology.

292

Roberta D’Alessandro and Diego Pescarini

Bonet, Eulàlia (1995), “Feature structure of Romance clitics”, Natural Language and Linguistic Theory 13, 607–647. Bonet, Eulàlia (2008), “The person-case constraint and repair strategies”, in: Roberta D’Alessandro/ Susann Fischer/Gunnar Hrafn Hrafnbjargarson (edd.), Agreement restrictions, Berlin/New York, Mouton de Gruyter, 103–128. Bonet, Eulàlia (2013), “Agreement in two steps (at least)”, in: Ora Matushansky (ed.), Distributed morphology today: morphemes for Morris Halle, Cambridge, MA, MIT Press, 167–184. Bonet, Eulàlia/Lloret, Maria-Rosa/Mascaró, Joan (2015), “The prenominal allomorphy syndrome”, in: Eulalia Bonet/Maria-Rosa Lloret/Joan Mascaró (edd.), Understanding allomorphy. Perspectives from Optimality Theory, London, Equinox, 4–55. Bonet, Eulàlia/Mascaró, Joan (2011), Asimetrías de concordancia en el SD. El rasgo de masa en asturiano, Ms. Universitat Autònoma de Barcelona. Brandi, Luciana/Cordin, Patrizia (1989), “Two Italian dialects and the null subject parameter”, in: Osvaldo Jaeggli/Ken Safir (edd.), The null subject parameter, Dordrecht, Kluwer, 111–142. Bright, William (1957), The Karok language, Berkeley, University of California Press. Burzio, Luigi (1986), Italian syntax. A Government and Binding approach, Dordrecht, Reidel. Calabrese, Andrea (1994), “Syncretism phenomena in the clitic systems of Italian and Sardinian dialects and the notion or morphological change”, in: Jill N. Beckman, (ed.), Proceedings of NELS 25, vol. 2, Amherst (Mass.), GLSA, 151–174. Cardinaletti, Anna (1997), “Subjects and clause structure”, in: Liliane Haegeman (ed.), The new comparative syntax, London, Longman, 33–63. Cardinaletti, Anna/Giusti, Giuliana (2011), “L’opzionalità alle interfacce sintassi – morfologia – fonologia”, in: Daniela Veronesi/Giovanna Massariello Merzagora/Serena Dal Maso (edd.), I luoghi della traduzione. Le interfacce, Società di Linguistica italiana, Roma, Bulzoni, vol. SLI 54, 865–879. Cardinaletti, Anna/Giusti, Giuliana (2015), “Cartography and optional feature realization in the nominal expression”, in: Urs Shlonsky (ed.), Beyond functional sequence. The cartography of syntactic structures, vol. 10, Oxford/New York, Oxford University Press, 151–172. Chomsky, Noam (2001), “Derivation by phase”, in: Michael Kenstowicz (ed.), Ken Hale: A life in language, Cambridge, MA, MIT Press, 1–50. Chomsky, Noam (2013), “Problems of projection”, Lingua 130, 33–49. Cinque, Guglielmo (1988), “On si constructions and the theory of arb”, Linguistic Inquiry 19, 521– 582. Cinque, Guglielmo (2010), The syntax of adjectives. A comparative study, Cambridge, MA, MIT Press. Corbett, Greville (1979), “The agreement hierarchy”, Journal of Linguistics 15, 203–224. Corbett, Greville (1990), Agreement, Cambridge, Cambridge University Press. D’Alessandro, Roberta (2004), Impersonal ‘si’ constructions, PhD dissertation, University of Stuttgart. D’Alessandro, Roberta (2007), Impersonal ‘si’ constructions. Agreement and interpretation, Berlin/ New York, Mouton de Gruyter. D’Alessandro, Roberta (2013), Merging Probes. Ms, Leiden University, http://ling.auf.net/lingbuzz/ 001771 (17.03.2016). D’Alessandro, Roberta (2016), “When you have too many features. Auxiliaries, agreement and clitics in Italian varieties”, Ms., Leiden University, submitted. D’Alessandro, Roberta/Fischer, Susann/Hrafnbjargarson, Gunnar H. (edd.) (2008), Agreement restrictions, Berlin/New York, Mouton de Gruyter. D’Alessandro, Roberta/Ledgeway, Adam (2010), “The Abruzzese T-v system. Feature spreading and the double auxiliary construction”, in: Roberta D’Alessandro/Adam Ledgeway/Ian Roberts, Syntactic variation. The dialects of Italy, Cambridge, Cambridge University Press, 201–209. D’Alessandro, Roberta/Roberts, Ian (2010), “Past participle agreement in Abruzzese. Split auxiliary selection and the null-subject parameter”, Natural Language and Linguistic Theory 28, 41–72.

Agreement restrictions and agreement oddities

293

Fassi Fehri, Abdelkader (1993), Issues in the structure of Arabic clauses and words, Dordrecht, Kluwer. Giusti, Giuliana (2008), “Agreement and concord in nominal expressions”, in: Cécile De Cat/ Katherine Demuth (edd.), The Bantu-Romance connection, Amsterdam/Philadelphia, Benjamins, 201–238. Guasti, Teresa/Rizzi, Luigi (2002), “Agreement and tense as distinct syntactic positions. Evidence from acquisition”, in: Guglielmo Cinque (ed.), The structure of DP and IP. The cartography of syntactic structures, vol. 1, Oxford, Oxford University Press, 167–194. Haiman, John/Benincà, Paola (1992), The Rhaeto-Romance languages, London, Routledge. Hale, Ken (2002), “Eccentric agreement”, in: Beatriz Fernández/Pablo Albizu (edd.), Kasu eta Komunztaduraren gainean [On case and agreement], Vitoria-Gasteiz, Euskal Herriko Unibetsitatea, 15–48. Harris, Roy (1981), The language myth, London, Duckworth. Haspelmath, Martin (2004), “Explaining the ditransitive person-role constraint. A usage-based account”, Constructions 2/2004, 1–71. Kayne, Richard (1975), French syntax: the transformational cycle. Cambridge, MA, MIT Press. Kramer, Ruth (2009), Definite markers, phi features and agreement. A morphosyntactic investigation of the Amharic DP, PhD dissertation, University of California. Ledgeway, Adam (2000), A comparative syntax of the dialects of southern Italy. A Minimalist approach, Oxford, Blackwell. Ledgeway, Adam (2012), From Latin to Romance. Morphosyntactic typology and change, Oxford, Oxford University Press. Ledgeway, Adam/Lombardi, Alessandra (2005), “Verb movement, adverbs and clitic positions in Romance”, Probus 17, 79–113. Madeira, Anam (1994), “On the Portuguese inflected infinitive”, UCL Working Papers in Linguistics 6, 179–203. Mancini, Anna Maria (1988), “Le caratteristiche morfosintattiche del dialetto di Ripatransone (AP), alla luce di nuove ricerche”, Quaderni di proposte e ricerche 6, 3–28. Marantz, Alec (1997), “No escape from syntax. Don’t try morphological analysis in the privacy of your own lexicon”, University of Pennsylvania Working Papers in Linguistics, vol. 4.2, 201–225. Marantz, Alex (2001), “Words”, 20th West Coast Conference on Formal Linguistics Paper held at, USC, February 2001. Mensching, Guido (2000), Infinitive constructions with specified subjects. A syntactic analysis of the Romance languages, Oxford, Oxford University Press. Nevins, Andrew (2007), “The representation of third person and its consequences for person-case effects”, Natural Language and Linguistic Theory 25, 273–313. Nevins, Andrew/Săvescu, Oana (2010), “An apparent number case constraint in Romanian. The role of syncretism”, in: Karlos Arregi/Zsuzsanna Fagyal/Silvina Montrul/Annie Tremblay (edd.), Romance Linguistics 2008, Amsterdam/Philadelphia, Benjamins, 185–203. Nicol, Fabrice (2005), “Romance clitic clusters. On diachronic changes and cross-linguistic contrasts”, in: Laurie Heggie/Francisco Ordóñez (edd.), Clitic and affix combinations, Amsterdam/ Philadelphia, Benjamins, 141–197. Ouali, Hamid (2008), “On C-to-T phi-feature transfer”, in: Roberta D’Alessandro/Susann Fischer/ Gunnar H. Hrafnbjargarson (edd.), Agreement restrictions, Berlin/New York, De Gruyter, 159– 180. Ormazabal, Javier/Romero, Juan (2007), “The object agreement constraint”, Natural Language and Linguistic Theory 25, 315–347. Ouhalla, Jamal (1993), “Subject extraction, negation and the anti-agreement effect”, Natural Language and Linguistic Theory 11, 477–518.

294

Roberta D’Alessandro and Diego Pescarini

Perlmutter, David (1971), Deep and surface structure constraints in syntax, New York, Holt, Rinehart, and Winston. Pescarini, Diego (2010), “Elsewhere in Romance. Evidence from clitic clusters”, Linguistic Inquiry 41, 427–444. Pescarini, Diego (2015), “A note on Italian datives”, in: Maria Grazia Busà/Sara Gesuato (edd.), Studi in onore di Alberto Mioni, Padova, CLEUP, 491–501. Postal, Paul (1989), Masked inversion in French, Chicago, IL, The University of Chicago Press. Raposo, Eduardo (1987), “Case Theory and Infl-to-Comp. The inflected infinitive in European Portuguese”, Linguistic Inquiry 18, 85–109. Rasom, Sabrina (2008), Lazy concord in the central Ladin feminine DP. A case Study on the interaction between morphosyntax and semantics, PhD dissertation, Università degli Studi di Padova. Reinhart, Tanya/Reuland, Eric (1993), “Reflexivity”, Linguistic Inquiry 24, 657–720. Rezac, Milan (2010), “On the unifiability of repairs of the person case constraint. French, Basque, Georgian, and Chinook”, Anuario del Seminario de Filología Vasca “Julio de Urquijo”, Special issue of XLIII, 1–2 (Ricardo Etxepare/Ricardo Gómez/Joseba A. Lakarra (edd.), Beñat Oihartzabali Gorazarre. Festschrift for Beñat Oyharçabal), 769–790. Rezac, Milan (2011), Phi-features and the modular architecture of language, Dordrecht, Springer. Rigau, Gemma (1982), “Inanimate indirect objects in Catalan”, Linguistic Inquiry 13, 146–150. Rizzi, Luigi (1990), Relativized minimality, Cambridge, MA, MIT Press. Rossi, Alfredo (2008), Dizionario del dialetto ripano, Monteprandone, Linea Grafica. Rouveret, Alain/Vergnaud, Jean-Roger (1980), “Specifying reference to the subject. French causatives and conditions on representations”, Linguistic Inquiry 11, 97–202. Saccon, Gabriella (1993), Postverbal subjects, PhD dissertation, Harvard University. Samek-Lodovici, Vieri (2002), “Agreement impoverishment under subject inversion. A crosslinguistic analysis”, Linguistische Berichte 11 (Gisbert Fanselow/Caroline Féry (edd.), Resolving conflicts in grammar), 49–82. Săvescu, Oana (2007), “Challenging the person case constraint. Evidence from Romanian”, in: José Camacho/Nydia Flores-Ferrán/Liliana Sánchez/Viviane Déprez/María José Cabrera (edd.), Romance Linguistics 2006. Selected Papers from the 36th Linguistic Symposium on Romance Languages (LSRL), Amsterdam/Philadelphia, Benjamins, 261–275. Sitaridou, Ioanna (2002), The synchrony and diachrony of Romance infinitives with nominative subjects, PhD dissertation, University of Manchester. Sleeman, Petra/Ihsane, Tabea (2013), “Gender mismatches, locality and feature checking”, Ms. University of Amsterdam and University of Geneva. Strozer, Judith (1976), Clitics in Spanish, PhD dissertation, UCLA. Zamparelli, Roberto (2008), “Bare predicate nominals in Romance languages”, in: Henrik Høeg Müller/Alex Klinge (edd.), Essays on nominal determination, Amsterdam/Philadelphia, Benjamins, 101–130. Zwicky, Arnold M. (1986), “German adjective agreement in GPSG”, Linguistics 24, 957–990.

Jaume Mateu Fontanals

10 Auxiliary selection Abstract: Auxiliary selection is a complex interface phenomenon that has been claimed to be sensitive to factors ranging from (in)transitivity, argument structure (cf. unaccusativity vs. unergativity), lexical semantics, and Aktionsart to tense, modality, clausal aspect, and subject person. Such a variety of factors has led different authors to approach this topic from a syntactic perspective or from a semantic one. There are also a few authors who have argued for a unified analysis of the morphosyntactic and the semantic factors involved in auxiliary selection. Indeed, there appears to be an impressive range of variation attested in Romance languages in terms of how the selection works and what it is sensitive to. In this chapter, I give a survey of the different auxiliary splits attested in Romance languages (section 1) and then provide an overview of some relevant syntactic and semantic approaches (section 2). The conclusion is that the two theoretical perspectives are not to be regarded as incompatible. Although I concentrate on the synchronic aspects of auxiliary selection in Romance, some diachronic issues are also taken into account (section 3). Finally, this chapter contains some concluding remarks (section 4). Keywords: auxiliary selection, auxiliary splits, unaccusativity, argument structure, lexicon-syntax interface

1 The empirical evidence As is well known, in Romance languages like Italian and French (or in Germanic languages like German or Dutch), there is an important division or split in the formation of compound forms of verbs: Transitive verbs and some intransitive verbs select auxiliary HAVE (e.g. It. avere, Fr. avoir), while other intransitive verbs select auxiliary BE (e.g. It. essere, Fr. être). Consider some relevant examples from Italian: transitive verbs like vedere ‘see’ or salutare ‘greet’ in (1), intransitive verbs like lavorare ‘work’, ballare ‘dance’ or rintoccare ‘ring’ in (2), and intransitive verbs like arrivare ‘arrive’, arrossire ‘blush’ or esistere ‘exist’ in (3). (1) It. a. Gianni ha visto Maria. Gianni has seen Maria ‘Gianni saw Maria.’ b. Gianni l’ha vista. Gianni her.has seen.F. SG ‘Gianni saw her.’ Acknowledgement: This work has been supported by grants from the Spanish Ministry of Economy and Competitiveness (FFI2014-56968-C4-1-P) and from the Generalitat de Catalunya (2014 SGR-1013).

296

Jaume Mateu Fontanals

c. Gianni li ha salutati. Gianni them.M . PL has greeted.M . PL ‘Gianni greeted them.’ (2) It. a. Gianni ha lavorato molto. Gianni has worked a-lot ‘Gianni worked a lot.’ b. Gianni ha ballato per ore. Gianni has danced for hours ‘Gianni danced for hours.’ c. La campana ha rintoccato the bell has tolled ‘The bell rang.’ (3) It. a. Maria è arrivata. Maria is arrived.F.SG ‘Maria arrived.’ b. Maria è arrossita. Maria is blushed.F.SG ‘Maria blushed.’ c. L’unicorno è esistito. the unicorn is existed.M .SG ‘The unicorn existed.’ Transitive verbs consistently select HAVE in Italian, whereby there is no variation in this respect.1 In striking contrast to this, there appears to be great variation involved with the intransitive verbs: For example, as pointed out by traditional grammarians like Battaglia/Pernicone (1987, 186), “per i tempi composti nella forma attiva dei verbi intransitivi, non si possono dare norme sicure: alcuni prendono avere, altri essere, e l’insegnamento può venire solo dall’uso, dalla lettura dei buoni scrittori, dalla consultazione del vocabolario”2. In the following sections other factors and splits are presented that have been claimed to be relevant to auxiliary selection in Romance: lexicosemantic and lexicoaspectual factors (section 1.1), reflexives (section 1.2), subject person (section 1.3), and tense, clausal aspect, and modality (section 1.4). 1 As we will see below, in some Italian dialects where the auxiliary split is sensitive to other factors (e.g. person and number of the subject), BE (It. essere) can also be found with transitive verbs. 2 “For the compound tenses in the active form of intransitive verbs, one cannot provide absolute rules: some take HAVE , others BE , and only usage, the reading of good authors, and the consultation of the dictionary can teach one how to use them.”

Auxiliary selection

297

1.1 Lexicosemantic and lexicoaspectual factors In spite of the abovementioned quote, certain intuitive patterns have been claimed to emerge when dealing with intransitive verbs: Basically, It. essere ‘be’ is selected with the predicates where the subject is a theme that undergoes a change of location (3a) or a change of state (3b) or where the subject is a static theme (3c), whereas It. avere ‘have’ is selected with the predicates that do not express a directed change but rather an action or internal process (be it agentive, cf. 2a and 2b, or not, cf. 2c). HAVE selection with intransitive verbs is often associated with agentivity in the literature, but examples like (2c) or (4) make it clear that this association is not accurate enough. This notwithstanding, an interaction between agentivity and auxiliary selection has been claimed to be involved in contrasts like the one exemplified in (5), drawn from Sorace (2000). (4) It. Maria {è caduta / *ha caduto} apposta. on-purpose Maria is fallen.F.SG / has fallen ‘Maria fell on purpose.’ (5) It. a. Il pilota {ha / ?è} atterrato sulla pista di emergenza. the pilot has / is landed on-the runway of emergency ‘The pilot landed on the emergency runway.’ b. Lʼaereo {?ha / è} atterrato sulla pista di emergenza. the plane has / is landed on-the runway of emergency ‘The plane landed on the emergency runway.’ However, as we will see below, it is the internal creation process (Hale/Keyser 2002; Mateu 2002), rather than agentivity, that determines the structural meaning of the intransitive verbs that select avere: That is, what is semantically construed as relevant in (5a) is not the change of location/state of the pilot but rather the action carried out by him (i.e. making a landing), whereas what is semantically construed as relevant in (5b) is the result of the process. Similarly, the following examples in (6) show that agentivity per se does not determine auxiliary selection since that notion could in principle be applied not only to (6a) but also to (6b). (6) It. a. Gianni ha corso (per ore). Gianni has run for hours ‘Gianni ran (for hours).’ b. Gianni è corso a casa (in tre minuti). Gianni is run.M .SG to home in three minutes ‘Gianni ran home (in three minutes).’

298

Jaume Mateu Fontanals

A paradigmatic contrast like the one exemplified in (6) has often been taken to show the relevance of Aktionsart (i.e. lexical aspect) in BE -selection: The addition of a telic PP in (6b) determines essere selection. As is well known, telicity has been argued to be an aspectual notion that turns out to be crucially relevant when determining BE -selection with intransitive verbs (e.g. cf. Zaenen 1993; Sorace 2000, or van Hout 2004). Similarly, the addition of a reflexive pronoun (It. si) to a verb like It. bruciare ‘burn’ changes its Aktionsart: The atelic reading in (7a) is associated with avere, whereas the telic one in (7b), which is claimed to be determined by the addition of the reflexive, is associated with essere (cf. Sorace 2000; 2004; Jezek 2003). Similar facts are also discussed by Labelle (1990; 1992) for French (e.g. cf. (8)). As pointed out by Labelle (1990, 309), “the intransitive construction states what the subject does. So, one can say it focusses on the process itself. In contrast, the reflexive construction states what is happening to the object. Thus, it focusses on the result of the process”. (7) It. a. La carne ha bruciato (per alcuni minuti). the meat has burned for some minutes ‘Meat has burned for some minutes.’ b. La carne si è bruciata (#per alcuni minuti). some minutes the meat SI is burned.F.SG for (8) Fr. a. Le poulet a cuit pendant 3 heures. the chicken has cooked for 3 hours b. Le poulet s’es cuit en exactement 3 heures. the chicken REFL .is cooked.M .SG in exactly 3 hours Nonetheless, the relevance of telicity at the lexicon-syntax interface has been called into question. For example, Levin/Rappaport Hovav (1995) point out that there are atelic intransitive verbs that select BE , contrary to what is predicted by telicity-based approaches to BE -selection: For example, indefinite change of state verbs select essere in Italian (and zijn ‘be’ in Dutch); cf. (9a) and (9b).3 In spite of this, the aspectual approach can be claimed to account for the fact that these verbs in French do not select être, but avoir; cf. (9c).

3 Indefinite change of state verbs are also referred to in the literature on aspect as “degree achievements” (cf. Dowty 1979 and Bertinetto/Squartini 1995).

Auxiliary selection

(9) It.

299

a. Maria è cresciuta / ??ha cresciuto molto quest’anno. a lot this year Maria is grown.F.SG / has grown

Dutch b. Haar baby is / *heeft deze maand enorm gegroeid. her baby is / has this month enormous grown Fr.

c. Marie a grandi / *est grandie depuis l’an dernier. Marie has grown / is grown.F.SG since the year last

Furthermore, telicity-based approaches to BE -selection run into nontrivial problems when dealing with stative verbs. Although these approaches could be said to account for the fact that stative verbs in French select avoir, once again their predictions are not borne out in Italian; cf. (10a) and (10b). (10) Fr. a. Les dinosaurs ont existé. the dinosaurs have existed It.

b. I dinosauri sono esistiti. the dinosaurs are existed.M .PL

Moreover, dative psychological verbs of the piacere class (cf. e.g. Belletti/Rizzi 1988) also offer potential counterexamples to aspectual approaches to auxiliary selection; e.g. cf. (11). (11) It.

a. Questo film mi è piaciuto molto. this film me.DAT is pleased.M .SG a-lot

Dutch b. Dat is mij bevallen / tegengevallen. that is me.DAT pleased / disappointed Finally, the existence of (admittedly exceptional) telic intransitive verbs that select in Romance languages is also problematic for aspectual approaches to auxiliary selection. For example, consider the English examples in (12), which involve verbs of birthing (Hale/Keyser 1993). The existence of these verbs is especially relevant when dealing with auxiliary selection in French, where être has been argued to be mainly related to telicity (cf. e.g. Sorace 2000): Atelic intransitive verbs typically select avoir in French (cf. 9c and 10a), which is consistent with the claim that telicity is relevant in this Romance language. However, despite their telicity, intransitive verbs like Fr. pouliner ‘to foal’ do not select être but avoir (13a). Avere is also selected for this class of verbs in Italian (p.c. Paolo Lorusso); cf. (13b).

HAVE

300

Jaume Mateu Fontanals

(12) a. The mare foaled {in / ??for} two hours. b. The cow calved {in / ??for} two hours. (13) Fr. a. La the

jument {a pouliné / *est poulinée} en / ??pendant mare has foaled / is foaled.F.SG in / for

deux heures. two hours ‘The mare foaled in/??for two hours.’ It. b. La giumenta {ha figliato / ??è figliata} in / ??per the mare has foaled / is foaled.F.SG in / for due ore. two hours ‘The mare foaled in/??for two hours.’ All in all, the challenge to be drawn from the present discussion on the relevance of thematic properties (e.g. agentivity) or aspectual properties (e.g. telicity) in auxiliary selection in Romance is how one should cope with (i) the apparently contradictory facts briefly reviewed above (e.g. telicity is (ir)relevant to auxiliary selection with intransitive verbs) and (ii) the nontrivial crosslinguistic differences involved. We will come back to this issue in section 2.

1.2 Reflexives As is well known, the presence of the reflexive element (It. si) correlates with the selection of essere quite systematically, as shown in some relevant examples in (14) through (17). For example, consider the following readings associated with si: reflexive in (14), anticausative in (15), impersonal in (16), or passive in (17). (14) It. a. Gianni si è pettinato. Gianni SI is combed.M .SG ‘Gianni combed himself.’ b. Gianni e Carlo si sono guardati. Gianni and Carlo SI are looked.M .PL ‘Gianni and Carlo looked at each other.’ c. Gianni si è comprato una macchina. Gianni SI is bought.M .SG a car ‘Gianni bought himself a car.’

Auxiliary selection

301

(15) It. a. Il bicchiere si è rotto. (cf. Gianni ha rotto il bicchiere) the glass SI is broken.M .SG ‘The glass broke.’ b. Maria si è svegliata. (cf. Gianni ha svegliato Maria) Maria SI is woken.F.SG ‘Maria woke up.’ c. Maria si è arrabbiata. (cf. *Gianni ha arrabbiato Maria) Maria SI is angered.F.SG ‘Maria got angry.’ (16) It. a. Li si è salutati. them.ACC SI is greeted.M .PL ‘One has greeted them.’ b. Si è andati al mare. SI is gone.M .PL to.the sea ‘One went to the seaside.’ c. Si è camminato molto ieri. SI is walked a-lot yesterday ‘One walked a lot yesterday.’ (17) It. Le camicie si sono lavate a secco. the shirts SI are washed.F.PL to dry ‘The shirts have been dry-cleaned.’ Some comments are in order. First, it is important to point out that the presence of reflexive si is not incompatible with an accusative object, i.e. with a transitive construction (e.g. cf. the reflexive construction in (14c) and the impersonal one in (16a)). So it appears that the notions of transitivity and voice (Manzini/Savoia 2011, 197, gloss si as a “middle-passive” marker) cannot be collapsed; for example, (16a) is not an intransitive construction, even though it has been said to be “middle” (e.g. cf. Manzini/Savoia 2011, 198). Second, it is also interesting to note that many authors have related BE -selection in the reflexive reading (e.g. (14a)) to the fact that the pronoun si is a clitic. The fact that the auxiliary selected in the Germanic counterpart of (14a) is not BE but HAVE is often attributed to the fact that Ger. sich or Dutch zich have nonclitic status. For example, according to Reinhart/Siloni (2005), the different auxiliary selection with It. si and Dutch zich is due to their different clitic vs. nonclitic status, respectively (cf. also Haider/Rindler-Schjerve 1987). However, Manzini/Savoia (2011, 198–200) show that this proposal is not correct. For example, these Italian linguists point

302

Jaume Mateu Fontanals

out that in Romance varieties such as Soazza, the pronoun found in reflexive, anticausative, or impersonal constructions has a clitic status (s), but this clitic does not influence auxiliary selection.4 For example, as can be seen in the following data drawn from Manzini/Savoia (2011, 199), in Soazza auxiliary selection in impersonal constructions (e.g. cf. (18a–b) with (16b–c)) is determined by verb class (cf. (19)). (18) Soazza a. S a sempro dormit ben. SI has always slept well ‘One has always slept well.’ b. S e sempro rivo tart. SI is always arrived late ‘One has always arrived late.’ (19) a. O dormit. I.have slept ‘I have slept.’ b. Som {rivo / rivada}. I.am arrived.M .SG / arrived.F.SG ‘I have arrived.’ Manzini/Savoia (2011, 200) also point out that “Soazza has the same perfect participle agreement system as standard Italian, even if the auxiliary is have, for instance in

4 Cf. also Kayne (1993) and Cennamo/Sorace (2007) for discussion of some Italo-Romance varieties (e.g. Paduan) which allow HAVE with reflexive clitics. Interestingly, Cennamo/Sorace (2007) show that for those speakers of Paduan who accept both auxiliaries in anticausative constructions, the alternation may reflect different semantic conditions. According to them, “the auxiliary HAVE in (ia) would be appropriate if the event resulted from an external Causer, whereas BE would be used if no external Causer is implied (ib) and the door opened by itself, or by virtue of an inanimate (and unexpressed) causer (e.g. the wind). This is clearly shown in (ii)” (p. 90). (i) Paduan a. La porta se ga verto de colpo. the door SI has opened suddenly ‘The door has suddenly opened.’ b. La porta se ze verta de colpo. the door SI is opened. F. SG suddenly (ii) Paduan a. Go provà metare la ciave ma la porta no se ga verto. I.have tried put.INF the key but the door not SI has opened ‘I tried to put the key in but the door didn’t open’ (i.e. I did not manage to open it). b. Go provà metare la ciave ma la porta no *se ze verta. I.have tried put.INF the key but the door not SI is opened.F. SG

Auxiliary selection

303

the reflexive in (20). In other words, perfect participle agreement is entirely independent of the selection of have and be” [emphasis mine]. Cf. also Loporcaro (1998) for further discussion. (20) Soazza El/la s a lavo / lavada. he/she SI has washed.M .SG / washed.F.SG ‘He/she has washed himself/herself.’ In contrast, in Standard Italian avere is only allowed in reflexive constructions when a reflexive clause is constructed with the nonclitic version se stesso in place of the clitic si; cf. (21) with (14a). (21) It. Gianni ha pettinato se stesso. Gianni has combed himself ‘Gianni combed himself.’

1.3 Person-based splits Another pattern of the have vs. be split found in some (nonstandard) Italian dialects is the one which opposes the 1st and 2nd person with be to the 3rd person with have (cf. Kayne 1993; Cocchi 1994; 1995; D’Alessandro/Roberts 2010; Manzini/Savoia 2011). This classic split, which is the most common one, is oblivious to both verbal class and reflexive marker si. For example, cf. the following data in (22), drawn from a dialect of Neapolitan (Ledgeway 2000). (22) Neapolitan a. So’visto a Ciro. / So arrevato. ACC . Ciro / am arrived am.seen ‘I have seen Ciro. / I have arrived.’ b. Ha visto a Ciro. / Ha arrevato. has seen ACC . Ciro / has arrived ‘He has seen Ciro. / He has arrived.’ It is important to point out that there is no intrinsic association of be with 1st and 2nd person and of have with 3rd person. In fact, as pointed out by Manzini/Savoia (2011, 202), some varieties present a different pattern, which is less robustly attested, i.e. the 1st and 2nd person are associated with have, while the 3rd is associated with be, as in Morcone (Campania).

1.4 Tense, clausal aspect, and modality splits Auxiliary selection according to tense is also well attested in the Italian dialects. For example, as pointed out by Ledgeway (2000, 202f.), in the dialect of San Leucio del

304

Jaume Mateu Fontanals

Sannio, HAVE has generalized in the present perfect to all verb classes (the passive excluded); e.g. cf. (23a) and (23b). In contrast, in the pluperfect, BE is the universal auxiliary with all verb classes; e.g. cf. (24a) and (24b). (23) San Leucio a. Eggio (*so’) fatto tutto quello which have.PRS .1SG be.PRS .1SG done all ch’eggio pututo. that have.PRS .1SG be.able.PTCP ‘I have done all that I could.’ b. Iti (*siti) venuto priesto? early have.PRS .2PL be.PRS .2PL come ‘Did you arrive early?’ (24) San Leucio a. Illu era (*aveva) venuto priesto. early he be.PST.3SG have.PST.3SG come ‘He had come early.’ b. Erem’ (*avevamo) auta dice quello be.PST.1PL have.PST.1PL must.PTCP say that che dicevono loro. which say.PST.3PL they ‘We had had to say what they said.’ Tense splits found in some Italo-Romance dialects often interact with sensitivity to person and number. Manzini/Savoia (2005) make the important observation that the majority of dialects which show person-driven auxiliary selection in the present perfect do not show it in the pluperfect or in counterfactual tenses, either HAVE or BE being consistently found here (cf. e.g. Tuttle 1986; Ledgeway 2000 and D’Alessandro/ Roberts 2010 for further discussion). Splits according to clausal aspect are also well attested and can be exemplified with the contrast in (25) from Romanian, drawn from Dragomirescu/Nicolae (2009). These linguists show that this contrast can be explained from an aspectual point of view: The regular HAVE variant expresses a completed action, an action which took place in the past. In contrast, the BE variant expresses an action which started in the past and may continue at the moment of the utterance. In particular, these authors point out that the former means ‘John left at a certain moment and maybe he came back’, whereas the latter means ‘John left at a certain moment and he definitely has not returned yet.’ (25) Rom. Ion {a / e} plecat. Ion has / is left ‘Ion left.’

Auxiliary selection

305

It seems then that the auxiliary selection in (25) depends on which kind of perfect is being expressed, i.e. the Experiential Perfect or the Perfect of Result (cf. e.g. Iatridou/Anagnostopoulou/Pancheva 2003; McFadden 2007; McFadden/Alexiadou 2010). According to McFadden (2007), “the Experiential Perfect describes an eventuality which occurred previous to some reference time (which is equivalent to the speech time in a present perfect), often an experience that the subject has had. The Perfect of Result, on the other hand, describes a state holding at the reference time, which is the result of the underlying eventuality described by the VP”. For example, I have been sick before expresses an Experiential Perfect, whereas I have lost my cellphone (Could you help me find it?) expresses a Perfect of Result. A similar pattern is also found in Quebec French, as shown in (26) (from Manente 2009): (26) Quebec Fr. a. Jean a arrivé / parti / entré à huit heures. Jean has arrived / left / entered at eight hours b. Jean est arrivé (= là) / parti (= absent) / entré (= dedans). Jean is arrived (= there) / left (= absent) / entered (= inside) The use of BE in Romanian (25) and in Quebec French in (26b) can be related to the stative resultative interpretation that McFadden/Alexiadou (2010) attribute to Earlier English I am come. For example, the contrast in (27) given by Manente (2009, 43) can be accounted for on the basis that BE is related to a perfect-of-result reading where the target state holds of the subject, as in Earlier English. (27) Quebec Fr. Maintenant qu’il {*a / est} arrivé chez lui, now that.he has / is arrived at.his.house, il ne voudra plus jamais ressortir. he not want.FUT any longer exit Finally, modality has also been shown to play a role in auxiliary selection in Romance languages like Romanian and Old Neapolitan. For example, Avram/Hall (2007) show that HAVE is the regular auxiliary for perfects in Romanian, except in irrealis clauses, where BE is used instead (cf. 28a). In contrast, as shown by Ledgeway (2003), the opposite pattern is found in Old Neapolitan, where HAVE was favoured over BE in irrealis clauses (cf. 28b). But cf. Cennamo (2002) for some relevant qualifications. (28) Rom.

a. Ar fi plecat. would.3SG be left ‘S/he would have left.’

ONeapolitan b. Se illo avesse arrivato in Grecia, . . . if he had.3SG arrived in Greece ‘If he had arrived in Greece, . . .’

306

Jaume Mateu Fontanals

To conclude, we have seen that auxiliary selection in Romance languages is a complex interface phenomenon that is sensitive to factors ranging from (in)transitivity, lexical semantics, and Aktionsart to subject person, tense, clausal aspect, and modality (cf. also McFadden 2007 for a recent review). For reasons of space, in this section I have only been able to provide an outline of the existing variation and comment on the most important factors on the basis of some relevant examples.

2 Theoretical approaches In this section, we will see that the different factors reviewed above have led different authors to approach auxiliary selection from syntactic perspectives (e.g. cf. Burzio 1986; Perlmutter 1989; Kayne 1993; Hoekstra 1994, among many others) or from semantic ones (e.g. cf. Shannon 1990; Centineo 1996; Sorace 2000; Aranovich 2003; Bentley/Eythórsson 2003, among many others). Furthermore, there are a few descriptive attempts where a unified account of the morphosyntactic and semantic factors involved in auxiliary selection has been pursued (e.g. cf. Legendre/Sorace’s 2003 and Legendre’s 2007 O(ptimality)T(heory)-based accounts). In this section, I provide an overview of some relevant theoretical approaches –of course, a review of all syntactic and semantic approaches is impossible– and I will concentrate on showing where both perspectives, the syntactic and the semantic one, can be reconciled. Since both syntactic and semantic factors have been shown to be involved in the analysis of the complex phenomenon of auxiliary selection in Romance, it seems to be natural to try to find an interface strategy that allows one to incorporate the insights from both perspectives. Crucially, when dealing with the semantic factors, we will see that purely conceptual ones are not involved in auxiliary selection but only those that encode “structural meaning”, a domain that is not oblivious to syntax (cf. e.g. Hoekstra 1999; Mateu 2002 for some relevant discussion).

2.1 Syntactic approaches The starting point of many syntactic analyses of auxiliary selection is the so-called Unaccusative Hypothesis, initially formulated by Perlmutter (1978) in the Relational Grammar framework and later developed by Burzio (1981; 1986) in the Government and Binding framework. According to this hypothesis, intransitive verbs (or clauses; cf. Perlmutter 1978) divide up into two classes on the basis of which status is assigned to their argument: The argument of unergatives is just like the subject of transitives, whereas the argument of unaccusatives is more like an object in important respects, though it may look like a subject on the surface. Burzio (1981; 1986) attributes to them two different D(eep)-structure configurations. Unergative verbs

Auxiliary selection

307

like It. lavorare ‘to work’, telefonare ‘to phone’, dormire ‘to sleep’, giocare ‘to play’ etc. occur in the syntactic frame in (29a), while unaccusative verbs like It. venire ‘to come’, uscire ‘to go out’, salire ‘to go up’, morire ‘to die’ etc. enter into the configuration in (29b), where [NP e] expresses an empty NP subject. There is an important single split represented in (29): Unergatives have an external argument (i.e. the NP is external to VP), while unaccusatives have their argument internal to VP. (29) a. [NP [VP V]] b. [[NP e] [VP V NP]] Perlmutter (1978) and Burzio (1981; 1986) argue that the grammatical behaviour of unaccusative verbs (clauses), which can be defined through characteristics such as BE -selection in Dutch or in Italian, ne-cliticization in Italian or their lack of impersonal passivization in Dutch, can be explained in a uniform way by postulating an underlying structure in which their surface subject originates in an internal argument position (cf. 29b). Burzio points out that the pattern of auxiliary selection in Italian is parallel to that of the distribution of ne-cliticization, and that it reflects the different D(eep)structure configurations of unaccusative verbs (cf. 29b) vs. unergative ones (cf. 29a). Unaccusative verbs select the auxiliary essere, while nonunaccusatives (i.e. both unergatives and transitives) select the auxiliary avere. Burzio (1986, 30) points out that “ne-cliticization is possible with respect to all and only direct objects” and defines unaccusative verbs as those verbs whose subject can be substituted for the direct object clitic ne. Two related facts are then accounted for: First, the contrast between unergatives (e.g. telefonare ‘to phone’) in (30a) and unaccusatives (e.g. arrivare ‘to arrive’) in (30b) (NB: Burzio renamed Perlmutter’s unaccusatives as “ergatives”) and, in addition, the parallelism between the unaccusative subject in (30b) and the transitive object in (30c). As expected, ne-cliticization fails with unaccusative verbs if the pronominalized NP is in subject position (cf. 30d). Finally, as pointed out by Perlmutter, it seems more convenient to use “unaccusative” as applied to structures (or clauses) rather than to verbs; for example, the passive construction can also be analysed as unaccusative (cf. (30f) and (30b)). (30) It. a. *Ne

hanno telefonato tre. (cf. Hanno telefonato tre ragazze) have.3PL phoned three (cf. have.3PL phoned three girls) ‘Three of them have phoned.’

PART. CL

b. Ne

sono arrivate tre. (cf. Sono arrivate tre ragazze) are.3PL arrived.F. PL three (cf. are.3PL arrived.PL three girls) ‘Three of them have arrived.’ PART. CL

c. Ne

hanno comprato tre. (cf. Hanno comprato tre macchine) have.3PL bought three (cf. have.3PL bought three cars) ‘They have bought three of them.’ PART. CL

308

Jaume Mateu Fontanals

d. *Tre ne sono arrivate. three PART.CL are.3PL arrived ‘Three of them have arrived.’ e. *Tre ne hanno {comprato due macchine / telefonato}. two cars / phoned three PART.CL have.3PL bought ‘Three of them have {bought two cars / telephoned}.’ f. Ne

saranno comprate molte (di macchine). be.FUT.3PL bought.F. PL many (of cars) ‘Many of them will be bought.’ PART. CL

Burzio (1986, 55–56) formulates the rule in (31) for essere assignment as follows: (31) The auxiliary will be realized as essere when a binding relation exists between the subject and a nominal contiguous to the verb (where ‘a nominal contiguous to the verb’ is a nominal which is either part of the verb morphology, i.e. a clitic, or a direct object). Although impressive, Burzio’s syntactic account based on distributional arguments and on binding principles has been said to present some shortcomings (cf. e.g. Centineo 1996 for a critical review). For reasons of space, next I will concentrate on Burzio’s important correlation between essere selection and ne-cliticization, which has been called into question in the literature. Consider the relevant examples in (32b) and (32d), taken from Lonzi (1985) and also revisited by Levin/Rappaport Hovav (1995, 276–277, ex. (106)–(107)). These data have been claimed to be counterexamples to Burzio’s (1986) claim that ergative (i.e. unaccusative) verbs are the only monadic verbs that admit ne-cliticization of their argument. Following Lonzi (1985), Levin/Rappaport Hovav (1995, 275) point out that “a variety of verbs that take the auxiliary avere ‘have’ do permit ne-cliticization, but only when they are found in a simple tense; ne-cliticization is not possible when these verbs are found in a complex tense in which the auxiliary is expressed”. (32) It. a. *Di ragazze, ne hanno lavorato molte nelle fabbriche di Shanghai. of girls, PART. CL . have.3PL worked many in.the factories of Shanghai b. Di ragazze, ne lavorano molte nelle fabbriche di Shanghai. PART. CL . work.3PL . many in.the factories of Shanghai of girls, ‘There are many girls working in the factories of Shanghai.’ c. *Di ragazzi, ne hanno russato molti nel corridoio del treno. PART. CL have.3PL snored many in.the corridor of.the train of boys d. Di ragazzi, ne russavano molti nel corridoio del treno. PART. CL snored many in.the corridor of.the train of boys, ‘There were many boys snoring in the corridor of the train.’

Auxiliary selection

309

Levin/Rappaport Hovav (1995, 277) conclude that “phenomena said to involve ‘surface unaccusativity’ (. . .) are not unaccusative diagnostics strictly speaking, but rather to a large extent receive their explanation from discourse considerations” (cf. also Lonzi 1985 and Maling/Calabrese/Sprouse 1994 for similar remarks). In particular, they point out that “unergative verbs are found in this construction under circumstances similar to those that sanction the appearance of English unergative verbs in locative inversion –that is, in contexts where the verb describes a characteristic activity or process of the entity it is predicated of” (p. 276).5 This said, it is important to realize that Levin/Rappaport Hovav’s (1995) discourse-based observation does not account for the fact that avere-selection is not allowed in the examples in (32a) and (32c). Such a restriction is not found in the Romance languages that have lost auxiliary selection; for example, HAVE is possible in their Catalan counterparts (e.g. cf. 33).6 (33) Cat. a. De dones, a les fàbriques, n’hi {treballen / of women, in the factories, PART. LOC . work.3PL / han treballat} moltes. have.3PL worked many ‘There are many women working/who have worked in the factories.’ b. N’hi

{treballen / han treballat} moltes. PART. LOC . CL work.3PL / have worked many ‘Many work/have worked there.’

One could claim that Levin/Rappaport Hovav’s (1995) remark on Italian only holds for imperfective tenses, since these can be regarded as the idoneous ones for expressing habitual activities. However, the following triplet from Centineo (1996, 230–231, fn. 6) shows that this is not the case, since in the so-called passato remoto (lit. ‘remote past’) these alleged unergative verbs are also compatible with necliticization; cf. (34c). (34) It. a. Ce ne nuota tanta di gente, in quella piscina. pool there PART.CL swims much of people in that ‘Lots of people swim in that swimming pool.’ 5 But cf. Culicover/Levine (2001) for a critical review of Levin/Rappaport Hovav’s (1995) discoursebased analysis of locative inversion. According to the former, the traditional unaccusative diagnostic provided by locative inversion must again be accepted once this unaccusative construction is separated from heavy NP inversion constructions with unergative verbs. 6 The locative marker hi ‘there’ is obligatory in this existential construction (cf. Rigau 1997; Mateu/ Rigau 2002). (i) Cat. *En PART. CL

{treballen / han treballat} moltes. work.3PL / have worked many

310

Jaume Mateu Fontanals

b. ??Ce ne ha nuotato molta di gente in quella piscina. much of people in that pool there PART.CL has swum c. Ce ne nuotò molta di gente in quella piscina. pool there PART.CL swam much of people in that ‘Lots of people swam in that swimming pool.’ The ungrammaticality of (32a,c) and (34b) is actually predicted by Burzio’s (1986) correlation between ne-cliticization and unaccusativity. Indeed, there is some evidence that points to the fact that the constructions in (32) and (34) are unaccusative. Avere would not then be the expected auxiliary in (34b) if the Italian existential construction in (34) turns out to be unaccusative. In this sense, notice Centineo’s (1996, 231, fn. 6) remark: “[It] must also be added that some of the native speakers consulted about these data attempted to use essere as the auxiliary for (iv)” (= 34b). Indeed, assuming that the construction in (34b) is unaccusative, one wonders why essere-selection was only “attempted”. This requires further research. Furthermore, as pointed out by Rigau (1997) and Mateu/Rigau (2002), a Romance language like Catalan also offers an interesting piece of evidence for the unaccusative status of existential constructions. A well-known crosslinguistic generalization is that bare NP plurals cannot be postverbal subjects of unergative verbs in free inversion contexts (e.g. cf. 35b) but are only possible as postverbal subjects of unaccusatives (e.g. cf. Cat. Vénen joves, lit. ‘Come boys’, i.e. ‘There come boys’) or as direct objects of transitive verbs (e.g. Cat. Les drogues maten joves ‘Drugs kill boys’). Given this, the existential construction in (35c) should be unaccusative. As expected, the postverbal bare subject in (35c) is pronominalized by partitive ne; cf. (35d). (35) Cat. a. Els joves canten. the boys sing.3PL b. *Canten joves. sing.3PL boys c. (En aquesta coral) hi canten joves. in this choir LOC .CL sing.3PL . boys ‘There are boys singing (in this choir).’ d. (En aquesta coral), (de joves) n’hi canten molts. in this choir of boys PART.CL . LOC .CL sing.3PL many ‘There are many boys singing (in this choir).’ Such a syntactic flexibility or “elasticity” is not expected under a lexicalist account à la Levin/Rappaport Hovav (1995), who, following Lonzi (1985), argue that some Italian examples similar to (35c) and (35d) are not unaccusative. However, following

Auxiliary selection

311

Torrego (1989), Rigau (1997) argues that the unaccusativity of the existential construction in (35c) or (35d) is possible because of the presence of the obligatory locative marker hi ‘there’. Similarly, as is well known (cf. e.g. Hoekstra 1999; Sorace 2000, and Mateu/Rigau 2002), the presence of a directional PP like into the hall has been shown to be crucial in the explanation of why the constructions in (36b) and (37b) are unaccusative (cf. e.g. BE -selection in (37b)). The German data in (37) are adapted from Sorace (2000, 876, ex. (40)). (36) a. They danced (for hours). b. They danced into the hall. (37) Ger. a. Hans und Rita haben getanzt. Hans and Rita have danced ‘Hans and Rita danced.’ b. Hans und Rita sind *(in den Saal) getanzt. Hans and Rita are into the hall danced ‘Hans and Rita danced into the hall.’ As pointed out by Rigau (1997) and Mateu/Rigau (2002), an even clearer piece of evidence for the unaccusativity of the existential constructions in (35c) and (35d) can be found in the Northwestern variety of Catalan, where there is no agreement between the indefinite argument joves ‘boys’ and the verb; cf. (38b). Indeed, the lack of agreement in (38b) would be unexpected if the bare plural NP were the subject/external argument of an unergative verb/construction (cf. the unaccusativity of the example in (39b)). (38) NW Cat. a. Els joves canten. the boys sing.3PL b. (En aquesta coral) hi canta joves. in this choir LOC .CL sings boys c. (En aquesta coral), (de joves) n’hi canta molts. PART. CL . LOC . CL sings many in this choir of boys (39) Central Cat. a. Vénen (els) joves. come.3PL (the) boys NW Cat.

b. Ve joves. comes boys

Drawing on Hale/Keyser’s (1993) configurational theory of argument structure, Mateu/ Rigau (2002) claim that the syntactic analysis of the agentive unergative structure in

312

Jaume Mateu Fontanals

(35a) or (38a) is the one depicted in (40a), whereas that of the existential unaccusative structure in (35c) or (38b) is the one shown in (40b). (40) a. [vP Els joves [v DO √CANT-]] b. [vP [v √CANT- HAVE (=BE + PrepC C R )] [P P hi [PrepC C R joves]]]

Following Hale/Keyser (1993; 2002), the formation of unergatives can be argued to involve conflation of a nominal (or a simple root, e.g. √CANT- ‘s[o]ng’),7 which occupies the complement position in (40a), with a null light verb (e.g. an agentive DO ); cf. sing – DO song. The formation of the unaccusative argument structure in (40b) is quite different: A null possessive light verb HAVE , which is conceived of as the result of conflating a null P with the more basic light verb BE (cf. Kayne 1993 and Rigau 1997, among others, for the proposal that HAVE = BE + Prep), subcategorizes for a Small-Clauselike PP as complement: The Preposition that expresses a Central Coincidence Relation (PC C R ) in (40b), which is crucial when dealing with possessive relations (cf. Hale 1986 and Hale/Keyser 2002, chap. 7, among others), is conceived of as a birelational element that relates a possessor (hi ‘there’)8 with a possessee ( joves ‘boys’).9

7 Conflation is understood as a local operation whereby the phonological matrix of a head is transmitted to its phonologically defective sister head (cf. Hale/Keyser 2002 for more discussion). Thus, for instance, in (40a) the phonological matrix of the root √cant conflates into its phonologically defective (null) sister node v, giving rise to the verb cantar. 8 Cf. Mateu/Rigau (2002) for the claim that the locative clitic hi ‘there’ acts as an impersonalizer in (40b). As predicted, Nominative case is then impossible in unaccusative existential constructions, e.g. *Hi ha(n) ells ‘there are theynom’ (cf. Rigau 1997 for more relevant discussion). 9 In (i) are some examples of the haver-hi construction; cf. Fr. Il y a; Sp. Hay; Cat. Hi ha. For some prominent syntactic analyses of this existential construction, cf. Hoekstra/Mulder (1990); Moro (1997); and Rigau (1997), among others. The main difference between the examples in (i) and the ones in (ii) is that the latter involve an additional conflation of the root √CANT- with the null light verb HAVE (cf. (40b)). (i) Central Cat. a. (En aquesta coral) hi han joves {que canten cantant}. in this choir LOC . CL have.3PL boys who sing.3PL / singing b. (En aquesta coral) hi ha joves {que canten / cantant}. in this choir LOC . CL has boys who sing.3PL / singing ‘There are boys singing in this choir.’ (ii) Central Cat. a. (En aquesta coral) hi canten joves. in this choir LOC . CL sing.3PL boys NW Cat.

b. (En aquesta coral) hi canta joves. boys in this choir LOC . CL sings ‘There are boys singing in this choir.’

Auxiliary selection

313

As shown by Mateu/Rigau (2002), the conflation of √CANT- with HAVE depicted in (40b) is similar to that of √DANCE - with GO involved in the unaccusative structure of The boys danced into the kitchen, analysed in (41) (cf. e.g. Hoekstra 1999 for the claim that the unaccusative construction in (41) involves a Small Clause Result (SCR) as complement of the verb). Both (40b) and (41) are unaccusative argument structures, i.e. there is no argument occupying the specifier position of v, the one that corresponds to the external argument of unergatives or transitives.10 (41) [vP [v √DANCE GO ] [S C R /P P the boys into the kitchen]] With this background in mind, it seems natural to consider the Italian examples in (42), which are counterparts of the Catalan existential constructions analysed above, as unaccusative structures. Following the present syntactic analysis of similar data from Catalan (see above), one can then assume that the clauses/structures (as in Perlmutter 1978, rather than the verbs, as in Burzio) in (42) are unaccusative; cf. the syntactic analysis in (43a), which involves an obligatory Small-Clause-like PP and a conflation process similar to the one shown in (41). That is to say, in (43a) a root designating an event (√LAVOR ‘work’) is conflated with a null light unaccusative verb BE , which subcategorizes for a Small Clause whose inner predicate has a locative nature.11 10 It is often said that a Romance language like Italian also shows a similar polysemy (e.g. cf. (ia)– (ib)), but at the same time it is clear that it lacks the regular Germanic polysemy, as shown in (ic); cf. e.g. Mateu (2002) and Mateu/Rigau (2010) for further discussion. (i) It. a. Gianni ha {corso / ballato} per molte ore. Gianni has run / danced for many hours b. Gianni è corso via / Gianni è corso alla cucina. Gianni is run away / Gianni is run to.the kitchen ‘Gianni ran away. / Gianni ran to the kitchen.’ c. *Gianni è ballato via. / *Gianni è ballato alla cucina. Gianni is danced away / Gianni is danced to.the kitchen ‘Gianni danced away. / Gianni danced to the kitchen.’ 11 In fact, things turn out to be more complex. For example, the data in (i), taken from Maling/ Calabrese/Sprouse (1994), do not involve any surface locative PP. However Maling/Calabrese/ Sprouse (1994, 5) point out that (ib) is possible only on a very specific reading –namely, many people are calling in one specific place relevant to the speaker. A similar comment could be argued to be appropriate for (ia). Alternatively, temporal phrases like domani ‘tomorrow’ in (ia) or la domenica ‘on Sunday’ in (ib) can be claimed to play an important role as well. The relevant conclusion seems to be that a spatiotemporal element is compulsory in the syntactic structure in order to license these existential constructions. (i) It. a. Domani ne parleranno molti. tomorrow NE will.speak.3PL many b. Ne telefonano molti, di tifosi, la domenica! NE phone.3PL many of fans on Sunday

314

Jaume Mateu Fontanals

(42) It. a. Di ragazze, ne lavorano molte nelle fabbriche di Shanghai. PART. CL work.3PL many in.the factories of Shanghai of girls, b. Di ragazzi, ne russavano molti nel corridoio del treno. of boys, PART. CL snored many in.the corridor of.the train (43) a. [vP [v √LAVOR- BE ] [P P/S C molte ragazze [P ’ nelle fabbriche di Shanghai]]] b. There are many girls in the factories of Shanghai who are working. Given the unaccusative structure in (43a), which lacks an external argument occupying the specifier position of the v(erbal) head, avere-selection is expected to be blocked. As noted, this result is compatible with (and actually predicted by) Burzio’s (1986) classic analysis. However, as noted above, since the existential construction in (43a) is unaccusative, one wonders why essere-selection is not possible. This requires further research. To conclude, some important theoretical assumptions have been made in this section: Unaccusativity (and argument structure, in general) is not a property of verbs, as argued by proponents of projectionism, but rather of constructions/structures, as argued by proponents of neoconstructionism/constructivism (cf. Marantz 2013 and Mateu 2014 for recent relevant discussion of these two theories of the lexicon-syntax interface). Furthermore, the simpler syntactic structures shown in (44) have developed into the more complex ones represented in (45) thanks to the important work by Hale/Keyser (1993; 2002) and Hoekstra (1988; 1999), among others. (44) a. [NP b. [NP

[VP V NP]] [VP V]]

c. [[NP e] [VP V NP]] (45) a. [vP NP [v’ V C AUS E / DO [S C /P P NP Pred]]] (transitive structure) b. [vP NP [v’ V D O NP]] c. [vP

(unergative structure)

[v’ V B E ( C O M E ) [S C /P P NP Pred]]] (unaccusative structure)

Given the syntactic argument structures in (45), the relevant question with respect to auxiliary selection is why HAVE is associated with (45a) and (45b), whereas BE is associated with (45c). Hoekstra (1999, 82) points out that the starting point of this endeavour is the observation that the verb BE is similar to unaccusatives in a way in which the verb HAVE is similar to transitives and unergatives. Notice in passing that Hale/Keyser (1993; 2002) do provide an explicit relation of unergativity with transitivity by claiming that unergatives are underlying transitives as well (cf. Hale/ Keyser’s unergative structure in 45b with the classic one in 44b). According to Hoekstra (1994; 1999), the crucial difference between HAVE and BE is that the former brings in a transitivity feature. There is a debate in the syntactic

Auxiliary selection

315

approaches to auxiliary selection as to what the exact nature of this feature is. According to Hoekstra (1999, 82), the representation of a HAVE -perfect tense construction would be something like [BE [Participial Phrase . . . TF. . .]], with TF (short for “transitivity feature”; Kayne (1993) suggests it is a determiner/preposition; den Dikken (1994) and Hoekstra (1994) assume it is a preposition, Mahajan (1994) a case-marker) raising to BE , which overtly appears as HAVE . In this sense, there is in fact no selection, but rather the difference between HAVE and BE results from a different make-up of the internal structure of the participial complement of BE . Cf. also Ledgeway (2000) and Manzini/Savoia (2011), after Kayne (1993), for some important syntactic proposals that try to integrate the different morphosyntactic factors involved in the complex phenomenon of auxiliary selection in Italian dialects (cf. section 1).

2.2 Semantic approaches The proponents of semantic approaches to auxiliary selection often concentrate on distinctions based on thematic and aspectual properties (cf. section 1.1); e.g. cf. Shannon (1990; 1995), Van Valin (1990), Zaenen (1993), Centineo (1996), Lieber/ Baayen (1997), Sorace (2000; 2004), Aranovich (2003), Bentley/Eythórsson (2003), and Stolova (2007), among others. The most influential semantic approach to auxiliary selection in Romance is the one pioneered by Antonella Sorace and her colleagues, who take the systematic linguistic variation to suggest that unaccusativity is determined by a semantic notion whose components are organized along a (proto)typicality scale ranging from core to periphery. Sorace (2000; 2004) argues that a more nuanced descriptive approach than a simple two-way split is needed in order to account for the attested variation. In particular, Sorace (2000; 2004) shows that in Italian some intransitive verbs (e.g. the ones in 46a–b and 46k) select an auxiliary more categorically than other verbs do (e.g. the ones in 46c through 46j). The former are called ‘core verbs’, while the latter are ‘noncore verbs’.12 (46) It. a. Gianni è / *ha arrivato. Gianni is / has arrived b. Gianni è / *ha morto. Gianni is / has died c. La pianta è fiorita / ha fiorito due volte quest’anno the plant is blossomed.F. SG / has blossomed twice this year 12 It is interesting to point out that similar ideas can already be found in the literature on auxiliary selection in Germanic languages. For example, consider Shannon’s (1990, 476) proposal: “[in German and Dutch] verbs closely approximating the transitive prototype take HAVE , whereas clear mutatives take BE . However, the farther away from the prototypical extremes we get, the more room for variation we find.”

316

Jaume Mateu Fontanals

d. I miei nonni sono sopravvissuti / ?hanno the my grandparents are survived.M . PL / have sopravvissuto al terremoto. survived to-the earthquake e. La guerra è durata / ?ha durato a lungo. the war is lasted.F. SG / has lasted for long f. I dinosauri sono esistiti / ??hanno the dinosaurs are existed.M . PL / have esistito 65 milioni di anni fa. existed 65 million of years ago g. Il nuovo ballo brasiliano è / ha attecchito anche in Italia. the new dance Brazilian is / has taken-root also in Italy h. La campana ha rintoccato / ?è rintoccata. the bell has tolled / is tolled.F. SG i. Maria ha corso / è corsa velocemente. Maria has run / is run.F. SG fast j. È corsa / ?ha corso voce che Maria si sposa. is run.F. SG / has run rumor that Maria REFL marries k. Gianni ha lavorato / *è lavorato. Gianni has worked / is worked

As shown by Sorace, both native and nonnative speakers of Italian can have more doubts when establishing auxiliary selection grammaticality judgements of nonprototypical intransitive verbs (e.g. verbs of appearance and existence) than when establishing those of prototypical verbs (e.g. verbs of telic change of location/state). In (47) the relevant Auxiliary Selection Hierarchy (ASH) is depicted, as argued for by Sorace (2000, 863; 2004) and Keller/Sorace (2003). It basically embodies two main factors: telicity and agentivity.13 (47) Auxiliary Selection Hierarchy (ASH) CHANGE OF LOCATION

selects BE

– least variation

CHANGE OF STATE CONTINUATION OF A PRE - EXISTING STATE EXISTENCE OF STATE UNCONTROLLED PROCESS CONTROLLED PROCESS (MOTIONAL) CONTROLLED PROCESS (NONMOTIONAL)

selects HAVE – least variation

13 Zaenen (1993) and van Hout (2004), among others, also argue that telicity is the main semantic notion that is characteristic of unaccusative verbs in Dutch.

Auxiliary selection

317

In particular, Keller/Sorace (2003, 60f.) explain the ASH as follows: verbs at the BE end of the ASH are core unaccusatives and denote telic change; verbs at the HAVE end are core unergatives and denote agentive activity in which the subject is unaffected. Intermediate verbs between the two extremes incorporate telicity and agentivity to lesser degrees, and tend to have a less specified (basically stative) event structure (. . .). Core verbs are those on which native grammaticality judgments are maximally consistent, and are acquired early by both first and second language learners. In contrast, intermediate verbs are subject to crosslinguistic differences and exhibit gradient auxiliary selection preferences.

Sorace claims that the crosslinguistic variation depends on the location of the relevant cut-off point along the ‘Auxiliary Selection Hierarchy’ (ASH) in (47). For example, the main cut-off point in Italian can be empirically claimed to be drawn between the lexical semantic class expressing ‘existence of state’ and the one expressing ‘uncontrolled process’, whereas the main cut-off point in French can be drawn between telic changes and the rest of the lexical semantic classes, as shown in (48). (48) CHANGE OF LOCATION

selects BE – least variation

TELIC CHANGE OF STATE

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - “cut-off point” (French) ATELIC CHANGE OF STATE CONTINUATION OF A PRE - EXISTING STATE EXISTENCE OF STATE

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - “cut-off point” (Italian) UNCONTROLLED PROCESS CONTROLLED PROCESS (MOTIONAL) CONTROLLED PROCESS (NONMOTIONAL)

selects HAVE – least variation

Sorace’s (2000) work on the ASH has influenced other accounts of auxiliary selection like the semantic one put forward by Bentley/Eythórsson (2003) and the O(ptimality) T(heory) proposal by Legendre/Sorace (2003). This OT account is especially valuable since these authors try to argue for a unified analysis of the morphosyntactic ingredients (e.g. the reflexive clitic) and the semantic factors (e.g. telicity, stativity or control) involved in auxiliary selection in French. In order to deal with the systematic être selection in reflexive constructions the authors posit a constraint against linking morphosyntactic reflexives as unergatives. This crucially outranks all the semantic constraints, ensuring that reflexives will always select être, no matter what their semantics. Despite their descriptive insights, the formal limits on the semantic ingredients involved in auxiliary selection are not provided by Sorace (2000; 2004) nor by Legendre/Sorace (2003). In the next section this issue will be examined by using a syntactic theory of the structural meaning that is systematically associated with the

318

Jaume Mateu Fontanals

argument structure configurations (cf. e.g. Hale/Keyser 1993; Mateu 2002; Hoekstra 1999 or Marantz 2013 for the relevance of syntax when dealing with the ‘structural meaning’ and its irrelevance when dealing with the ‘idiosyncratic/conceptual/ encyclopaedic meaning’).

2.3 A syntactic approach to the structural meaning of unergativity and unaccusativity Despite Sorace’s (2000; 2004) descriptive merits, it is not clear from her approach what the formal constraints are that led her to posit seven or eight (but not eleven or twenty) lexical semantic classes of verbs when dealing with the auxiliary selection problem.14 Mateu (2002) argues that his syntactic approach to thematic structure, which is based on the ones put forward by Hale/Keyser (1993) and Hoekstra (1988; 1999), can provide one with some formal constraints concerning the ingredients of structural meaning involved in Sorace’s Auxiliary Selection Hierarchy: Meaning components like process, on the one hand, and change or existence, on the other, are the relevant ingredients at the syntax-semantics interface precisely because these notions can be argued to be filtered into the abstract relational semantics associated with the unergative and unaccusative syntactic argument structures in (49a) and (49b), respectively. (49) a. [X1 Z1 [X1 X1[±R] Y1]] b.

(Unergative argument structure)

[X1 X1[±T] [X2 Z2 [X2 X2[±r] Y2]]] (Unaccusative argument structure)

In (49a) the [+R] feature encodes an agentive/volitional immediate cause function (cf. Levin/Rappaport Hovav 1995), while the [-R] feature subsumes the nonagentive one. The nonrelational elements Z1 and Y1 are interpreted as Originator/Causer and Incremental Theme, respectively. Y1 is the created object that can typically be conflated into the unergative verbal head X1 (cf. Hale/Keyser 1993; 2002; Mateu 2002). In (49b), in contrast, an eventive head X1 subcategorizes for a birelational noneventive head X2 , which relates two nonrelational elements, Z2 and Y2 . The T(ransition) features, [+T] and [-T], which are associated with the unaccusative verbal head X1 in (49b), encode the CHANGE and BE semantic functions, respectively. Moreover, the [+r] and [-r] features correlate with Hale/Keyserʼs (1993; 2002) terminal coincidence

14 Sorace (2000, 861) is aware of this nontrivial problem: “(. . .) there are some important questions that I do not attempt to address. First, the reader will not find an explanation of why particular semantic components are more crucial to the selection of particular auxiliaries than others.”

Auxiliary selection

319

relation and central coincidence relation, respectively:15 The birelational element X2 relates two nonrelational elements Z2 and Y2 , Figure and Ground, respectively (Talmy 2000). Importantly, the relational features [T] and [R] are configurational (or derivative) in the sense that they can be read off from the mere syntactic argument structure, i.e. X1 is the very same eventive head in both (49a) and (49b). It is just the case that this head is realized as [R] (‘source/immediate cause’) if there is an external argument (Z1 in (49a)); otherwise, it is realized as [T] (‘transition’), as in (49b). In contrast to the lack of formal constraints involved in Sorace’s (2000; 2004) lexical semantic classes (their number is not formally limited), Mateu (2002; 2009) argues that the possible combinations of relational semantic features that can be drawn from the syntactic argument structures in (49) turn out to be formally limited or reduced to the ones in (50):16 (50) a. [[+T] [+r]] (cf. telic change of {location/state}) b. [[+T] [-r]] (cf. atelic change of {location/state}) c. [[-T] [-r]]

(cf. {continuation of a pre-existing state / existence of state})

d. [-R]

(cf. nonvolitional internal cause)

e. [+R]

(cf. volitional internal cause)

The relational semantic features in (50) can be associated with the syntactic argument structures depicted in (51), where the most relevant ‘cut-off points’ in Romance languages like French or Italian have been represented (Mateu 2002). (51)

a. b. c.

[X1 X1[+T] [X2 Z2 [X2 X2[+r] Y2]]] ----------------------------[X1 X1[+T] [X2 Z2 [X2 X2[-r] Y2]]]

selects BE cut-off point (French)

cut-off point (Italian)

d.

[X1 X1[-T] [X2 Z2 [X2 X2[-r] Y2]]] ----------------------------[X1 Z1 [X1 X1[-R] Y1]]

e.

[X1 Z1 [X1 X1[+R] Y1]]

selects HAVE

15 Cf. Hale (1986) for relevant discussion on the semantic notions associated with {terminal/central} coincidence relations. Basically, a terminal coincidence relation involves a coincidence between one edge or terminus of the theme’s path and the place, whereas a central relation involves a coincidence between the centre of the theme and the centre of the place. 16 Mateu (2002) claims that the [[-T] [+r]] combination can be excluded in virtue of the semantic fact that the unaccusatives involving [+r](esultative) are always associated with a positive Transition (i.e. [+T]). Alternatively, this combination could be assigned to stative constructions that take a result phrase as complement: e.g., Lat. abesse ‘be away’. In contrast, [[+T] [-r]] accounts for Sorace’s (2000) indefinite change of state verbs (i.e. Dowtyʼs 1979 ‘degree achievements’); e.g. cf. It. Gianni è cresciuto molto quest’anno ‘Gianni has grown a lot this year’.

320

Jaume Mateu Fontanals

3 Some diachronic issues In this section, I will not deal, for reasons of space, with the issue of how perfect systems with auxiliary splits were created in Romance languages (cf. e.g. Vincent 1982; Cennamo 2008; Ledgeway 2012). Rather I will make some brief remarks on how the auxiliary selection has subsequently been lost in some Romance languages (e.g. in Catalan and Spanish).17 For example, in Mateu (2009) the following descriptive generalization was drawn: The intransitive verbs that exhibit gradient auxiliary selection preference in Contemporary Italian (Sorace 2000; 2004) typically coincide with the ones that earlier lost the BE auxiliary in both Old Catalan (Batlle 2002) and Old Spanish (Aranovich 2003).18 Drawing on data from these two sources,19 Mateu (2009) points out that it cannot be a mere coincidence that in both Old Catalan and Old Spanish verbs of existence and appearance20 were among the first ones to admit the HAVE auxiliary, the rest of the unaccusative verbs being more reluctant to accept it. As shown above, Sorace (2000; 2004) shows that intransitive verbs of existence and appearance show gradience in Italian. In contrast, core or prototypical unaccusatives, the ones expressing change of location, do not show any gradient variation at all. However, by looking at more data from Old Catalan, Mateu/Massanell i Messalles (2014) conclude that, as far the diachronic process of auxiliary selection in this language is concerned, there are not “core verbs” but rather “core constructions”. Core constructions are the last ones that were affected by the relevant grammatical change. These authors show that it seems plausible to take subject-of-result structures whose DP subject is definite and typically preverbal (e.g. OCat. El cavaller és 17 Concerning Catalan, it is interesting to point out that ésser (BE ) still remains as a perfect auxiliary in some dialectal varieties: mainly Balearic, Rossellonian, and Alguerese (cf. Batlle’s 2002 appendix). 18 For example, cf. Aranovich (2003, 5–6): “A quick glance at the verbs (. . .) reveals that the degree of affectedness of the subject is a factor in the displacement of ser by haber as the perfect auxiliary. At one end of the continuum are the subjects of stative verbs of existence and appearance like quedar ‘remain’. The subjects of these verbs do not suffer any changes in state or location, hence they are not affected in any way by the event. This is the first class to lose its ability to select ser. At the opposite end are subjects of verbs of directed motion and verbs of change of state. These subjects are affected since they are in a new location or state as a consequence of the event. These classes are the last ones for which haber displaces ser as the perfect auxiliary of choice. In between these two extremes are verbs of manner of motion like correr ‘run’, and dynamic verbs of existence and appearance like desaparecer ‘disappear’. (. . .) The chronology of split auxiliary selection in Spanish, then, falls under the generalization that the less affected the subject, the earlier a verb lost its ability to select auxiliary ser.” 19 Unfortunately, Sorace’s (2000) important work on gradience in auxiliary selection with intransitive verbs was mentioned neither by Batlle (2002) nor by Aranovich (2003). 20 Levin/Rappaport Hovav (1995, 282) include remain, endure, come, and exist within this class. Come is also classified by Levin/Rappaport Hovav (1995, 281) within the class of “verbs of inherently directed motion”.

Auxiliary selection

321

arribat, lit. ‘The knight is arrived’) as the “core” ones, whereas existential structures whose NP is indefinite and typically postverbal (e.g. OCat. Ha arribat correu, lit. ‘Has arrived mail’, i.e. ‘There arrived mail’)21 can be regarded as “noncore” constructions and are then predicted to be the first ones affected by the relevant grammatical change. Mateu/Massanell i Messalles (2014) show that variation is not found in core constructions (i.e. subject-of-result structures, e.g. OCat. El cavaller {és/*ha} arribat ‘lit. The knight {is/*has} arrived’) until the very latest stages of the diachronic process of auxiliary substitution, whereas noncore constructions (e.g. existential structures like OCat. {Ha/és} arribat correu, lit. ‘Has/is arrived mail’, i.e. ‘There arrived mail’) do present variation long before this process is reaching its final stages. It seems then that the following descriptive generalization holds for Old Catalan: Subject-ofresult constructions systematically involve BE -selection, whereas existential constructions often involve HAVE -selection (though BE -selection is also attested). Indeed, it is important to point out that this fact is compatible with Keller/Sorace’s (2003, 60) conception of “intermediacy” if (and only if ) we replace “verbs” by “constructions”, i.e. “[i]ntermediate [constructions] between the two extremes incorporate telicity and agentivity to lesser degrees, and tend to have a less specified (basically stative) event structure”. As predicted, existential constructions in Old Catalan like the ones where HAVE is typically selected are atelic and more stative than subject-of-result constructions. Importantly, Mateu/Massanell i Messalles’s (2014) claim that the fact that in Old Catalan even Sorace’s (2000) core unaccusative verbs like venir ‘come’, entrar ‘enter’, anar ‘go’, arribar ‘arrive’ or eixir ‘exit’ can enter into an existential argument structure where HAVE is typically selected can be accommodated in (neo)constructionist approaches to the lexicon-syntax interface (e.g. Marantz 2013) in a more appropriate way than in projectionist approaches (e.g. Levin/Rappaport Hovav 1995).22

4 Concluding remarks Auxiliary selection in Romance languages is a complex interface phenomenon that has been shown to be sensitive to factors ranging from (in)transitivity, argument structure (cf. unaccusativity vs. unergativity), lexical semantics, and Aktionsart 21 Cf. also Rosemeyer (2014) for the relevance of (in)definiteness in auxiliary selection in Old Spanish. 22 Recall, however, that, according to Sorace (2000; 2004), core unaccusative verbs like It. venire ‘come’ or arrivare ‘arrive’ are predicted to select only BE , no matter what the quantificational nature of their subject is; e.g. cf. It. *Ha arrivato posta vs. È arrivata posta ‘There arrived mail’. As pointed out by Sorace, proponents of projectionism can account for this situation (and the lack of elasticity of core verbs; e.g. cf. It. *Gianni ha venuto/morto/. . . ‘Gianni has come/died/. . .’) in a more appropriate way.

322

Jaume Mateu Fontanals

to subject person, tense, and modality. Such a variety of involved factors has led different authors to approach auxiliary selection from a syntactic perspective (cf. e.g. Burzio 1986; Perlmutter 1989; Kayne 1993; Hoekstra 1994; Ledgeway 2000) or from a semantic one (cf. e.g. Centineo 1996; Sorace 2000; 2004; Bentley/Eythórsson 2003; Aranovich 2003). Syntactosemantic approaches where unaccusativity is semantically determined but syntactically represented can also be found in the literature (cf. e.g. Perlmutter 1978; Cennamo 2002). For reasons of space, in this chapter I have concentrated on offering an overview of some relevant syntactic and semantic approaches to auxiliary selection in Romance and have concluded that the two perspectives are not incompatible. Future research is needed to see if the resulting unifying perspective can accommodate the morphosyntactic factors reviewed in section 1 (cf. e.g. Legendre/Sorace’s 2003 and Legendre’s 2007 OT-based accounts for a unified analysis of the morphosyntactic and semantic factors involved in auxiliary selection). Assuming an important theoretical distinction between syntactically nontransparent conceptual content and syntactically transparent semantic construal (cf. e.g. Mateu 2002; 2009; Marantz 2013), Sorace’s (2000; 2004) Auxiliary Selection Hierarchy can be shown to be accounted for in terms of syntactic representations that encode structural meaning (vs. the idiosyncratic meaning, which is encoded by the conceptual root). One relevant theoretical consequence is that there are no unaccusative verbs but rather unaccusative structures/constructions that encode relational meaning. Such a view does not exclude the fact that some roots (which encode conceptual content) tend to be lexically associated to some specific syntactic constructions (which encode structural meaning). (Neo)constructionist approaches to auxiliary selection can also be claimed to be more compatible with those accounts that posit the intervention of “nonlexical” level factors (cf. e.g. Ledgeway 2000; 2003; Manzini/Savoia 2011) and with those accounts that allow both the integration of so-called “core unergative verbs” (e.g. It. lavorare ‘to work’) into unaccusative structures (e.g. It. Di ragazze, ne lavorano molte nelle fabbriche di Shanghai ‘There are many girls in the factories of Shanghai who are working’) and the integration of so-called “core unaccusative verbs” (e.g. OCat. arribar ‘to arrive’) into existential constructions where HAVE is typically selected (e.g. OCat. Ha arribat correu, lit. ‘Has arrived mail’). Indeed, insertion of roots into syntactic argument structures is not as free as some radical proponents of (neo)constructionism would predict (e.g. Old Catalan venir ‘come’ cannot be associated to an unergative argument structure; cf. also footnote 22) but is much freer than proponents of projectionist approaches to the lexicon-syntax interface predict. Quite probably, the proper account(s) of auxiliary selection in Romance languages, i.e. the one(s) that turn(s) out be successful in connecting both the thematic/Aktionsart factors with the “nonlexical” ones (e.g. person, tense, clausal aspect, and modality), will have to find the right balance between these two approaches to the lexicon-syntax interface.

Auxiliary selection

323

5 References Aranovich, Raúl (2003), “The semantics of auxiliary selection in Old Spanish”, Studies in Language 27, 1–37. Avram, Larisa/Hall, Virginia (2007), “An irrealis BE auxiliary in Romanian”, in: Raúl Aranovich (ed.), Split auxiliary systems. A cross-linguistic perspective, Amsterdam/Philadelphia, Benjamins, 47–64. Batlle, Mar (2002), L’expressió dels temps compostos en la veu mitjana i la passiva pronominal. El procés de substitució de l’auxiliar ésser per haver, Barcelona, Institut d’Estudis Catalans/ Publicacions de l’Abadia de Montserrat. Battaglia, Salvatore/Pernicone, Vincenzo (1987), Grammatica italiana, Torino, Loescher. Belletti, Adriana/Rizzi, Luigi (1988), “Psych verbs and theta theory”, Natural Language and Linguistic Theory 6, 291–352. Bentley, D elia/Eythórsson, Thórhallur (2003), “Auxiliary selection and the semantics of unaccusativity”, Lingua 114, 447–471. Bertinetto, Pier M./Squartini, Mario (1995), “An attempt at defining the class of ‘gradual completion verbs’”, in: Pier M. Bertinetto (ed.), Temporal reference, aspect and actionality, Torino, Rosenberg and Sellier, 11–26. Burzio, Luigi (1981), Intransitive verbs and Italian auxiliaries, PhD dissertation, Massachusetts Institute of Technology. Burzio, Luigi (1986), Italian syntax, Dordrecht, Reidel. Cennamo, Michela (2002), “La selezione degli ausiliari perfettivi in napoletano antico. Fenomeno sintattico o sintattico-semantico?”, Archivio Glottologico Italiano 87, 175–222. Cennamo, Michela (2008), “The rise and development of analytic perfects in Italo-Romance”, in: Thórhallur Eythórsson (ed.), Grammatical change and linguistic theory. The Rosendal Papers, Amsterdam/Philadelphia, Benjamins, 115–142. Cennamo, Michela/Sorace, Antonella (2007), “Auxiliary selection and split intransitivity in Paduan”, in: Raúl Aranovich (ed.), Split auxiliary systems. A cross-linguistic perspective, Amsterdam/ Philadelphia, Benjamins, 65–99. Centineo, Giulia (1996), “A lexical theory of auxiliary selection in Italian”, Probus 8, 223–271. Cocchi, Gloria (1994), “An explanation of the split in the choice of perfect auxiliaries”, Probus 6, 87– 102. Cocchi, Gloria (1995), La selezione dell’ausiliare, Padua, Unipress. Culicover, Peter W./Levine, Robert D. (2001), “Stylistic inversion in English. A reconsideration”, Natural Language and Linguistic Theory 19, 283–310. D’Alessandro, Roberta/Roberts, Ian (2010), “Past participle agreement in Abruzzese. Split auxiliary selection and the null-subject parameter”, Natural Language and Linguistic Theory 28, 41–72. Dikken, Marcel den (1994), “Auxiliaries and participles”, Proceedings of the Northeastern Linguistics Society (NELS) 24, 65–79. Dowty, David (1979), Word meaning and Montague Grammar, Dordrecht, Reidel. Dragomirescu, Adina/Nicolae, Alexandru (2009), Relics of auxiliary selection in Romanian, Ms. Iorgu Iordan-Al.Rosetti Institute for Linguistics/University of Bucharest. Haider, Hubert/Rindler-Schjerve, Rosita (1987), “The parameter of auxiliary selection. Italian-German contrasts”, Linguistics 25, 1029–1055. Hale, Kenneth L. (1986), “Notes on world view and semantic categories. Some Warlpiri examples”, in: Peter Muysken/Henk Van Riemsdijk (edd.), Features and projections, Dordrecht, Foris, 233– 254. Hale, Kenneth L./Keyser, Samuel J. (1993), “On argument structure and the lexical expression of syntactic relations”, in: Kenneth L. Hale/Samuel J. Keyser (edd.), A view from Building 20. Essays in linguistics in honor of Sylvain Bromberger, Cambridge, MA, MIT Press, 53–109.

324

Jaume Mateu Fontanals

Hale, Kenneth L./Keyser, Samuel J. (2002), Prolegomenon to a theory of argument structure, Cambridge, MA, MIT Press. Hoekstra, Teun (1988), “Small clause results”, Lingua 74, 101–139. Hoekstra, Teun (1994), “HAVE as BE plus or minus”, in: Guglielmo Cinque (ed.), Festschrift for Richard Kayne, Washington, Georgetown University Press, 199–215. Hoekstra, Teun (1999), “Auxiliary selection in Dutch”, Natural Language and Linguistic Theory 17, 67–84. Hoekstra, Teun/Mulder, René H. (1990), “Unergatives as copular verbs. Locational and existential predication”, The Linguistic Review 7, 1–79. van Hout, Angeliek (2004), “Unaccusativity as telicity checking”, in: Artemis Alexiadou/Elena Anagnostopoulou/Martin Everaert (edd.), The unaccusativity puzzle. Explorations of the syntaxlexicon interface, Oxford, Oxford University Press, 60–83. Iatridou, Sabine/Anagnostopoulou, Elena/Pancheva, Roumyana (2003), “Observations about the form and meaning of the perfect”, in: Artemis Alexiadou/Monika Rathert/Arnim von Stechow (edd.), Perfect explorations, Berlin/New York, De Gruyter, 153–204. Jezek, Elisabetta (2003), Classi di verbi tra semantica e sintassi, Pisa, ETS. Kayne, Richard (1993), “Toward a modular theory of auxiliary selection”, Studia Linguistica 47, 3–31. Keller, Frank/Sorace, Antonella (2003), “Gradient auxiliary selection and impersonal passivization in German. An experimental investigation”, Journal of Linguistics 39, 57–108. Labelle, Marie (1990), “Unaccusatives and pseudo-unaccusatives in French”, Proceedings of the Northeastern Linguistics Society (NELS) 20, 303–317. Labelle, Marie (1992), “Change of state and valency”, Journal of Linguistics 28, 375–414. Ledgeway, Adam (2000), A comparative syntax of the dialects of southern Italy. A minimalist approach, London, Blackwell. Ledgeway, Adam (2003), “The distribution of the perfective auxiliary avere in Early Neapolitan. Split intransitivity conditioned by modal factors”, Archivo Glottologico Italiano 88, 29–71. Ledgeway, Adam (2012), From Latin to Romance. Morphosyntactic typology and change, Oxford, Oxford University Press. Legendre, Géraldine (2007), “Optimizing auxiliary selection in Romance”, in: Raúl Aranovich (ed.), Split auxiliary systems. A cross-linguistic perspective, Amsterdam/Philadelphia, Benjamins, 145–180. Legendre, Géraldine/Sorace, Antonella (2003), “Split intransitivity in French. An optimality-theoretic perspective”, in: Danièle Godard (ed.), Les langues romanes. Problèmes de la phrase simple, Paris, CNRS Éditions, 185–234. Levin, Beth/Rappaport Hovav, Malka (1995), Unaccusativity. At the syntax-lexical semantics interface, Cambridge, MA, MIT Press. Lieber, Rochelle/Baayen, Harald (1997), “A semantic principle of auxiliary selection in Dutch”, Natural Language and Linguistic Theory 15, 789–845. Lonzi, Lidia (1985), “Pertinenza della struttura tema-rema per l’analisi sintattica”, in: Harro Stammerjohann (ed.), Theme-rheme in Italian, Tübingen, Narr, 99–120. Loporcaro, Michele (1998), Sintassi comparata dell’accordo participiale romanzo, Torino, Rosenberg and Sellier. Mahajan, Anoop (1994), “The ergativity parameter. Have and Be alternation, word order and split ergativity”, Proceedings of the Northeastern Linguistics Society (NELS) 24, 317–331. Maling, Joan/Calabrese, Andrea/Sprouse, Rex A. (1994), “‘Domani ne parleranno molti’. The Independence of ne-cliticization and essere-selection in Italian”, Ms., Brandeis University, Harvard University, Indiana University. Manente, Mara (2009), L’aspect, les auxiliaires “être” et “avoir” et l’hypothèse inaccusative dans une perspective comparative français/italien, PhD dissertation, Università Ca’ Foscari di Venezia, http://arcaold.unive.it/handle/10278/965 (20.11.2015).

Auxiliary selection

325

Manzini, Maria Rita/Savoia, Leonardo Maria (2005), I dialetti italiani e romanci. Morfosintassi generativa, Alessandria, Edizioni dell’Orso. Manzini, Maria Rita/Savoia, Leonardo Maria (2011), Grammatical categories. Variation in Romance languages, Cambridge, Cambridge University Press. Marantz, Alec (2013), “Verbal argument structure. Events and participants”, Lingua 130, 152–168. Mateu, Jaume (2002), Argument structure. Relational construal at the syntax-semantics interface, PhD dissertation, Universitat Autònoma de Barcelona, http://hdl.handle.net/10803/4828 (20.11.2015). Mateu, Jaume (2009), “Gradience and auxiliary selection in Old Catalan and Old Spanish”, in: Paola Crisma/Giuseppe Longobardi (edd.), Historical syntax and linguistic theory, Oxford, Oxford University Press, 176–193. Mateu, Jaume (2014), “Argument structure”, in: Andrew Carnie/Dan Siddiqi/Yosuke Sato (edd.), The Routledge Handbook of Syntax, New York, Routledge, 24–41. Mateu, Jaume/Massanell i Messalles, Mar (2014), “A constructional approach to auxiliary selection. Evidence from existential constructions”, in: Rolf Kailuweit/Malte Rosemeyer (edd.), Auxiliary selection revisited. Gradience and gradualness, Berlin/Boston, De Gruyter, 183–211. Mateu, Jaume/Rigau, Gemma (2002), “A minimalist account of conflation processes. Parametric variation at the lexicon-syntax interface”, in: Artemis Alexiadou (ed.), Theoretical approaches to universals, Amsterdam/Philadelphia, Benjamins, 211–236. Mateu, Jaume/Rigau, Gemma (2010), “Verb-particle constructions in Romance. A lexical-syntactic account”, Probus 22, 241–269. McFadden, Thomas (2007), “Auxiliary selection”, Language and Linguistics Compass 1, 674–708. McFadden, Thomas/Alexiadou, Artemis (2010), “Perfects, resultatives, and auxiliaries in Earlier English”, Linguistic Inquiry 41, 389–425. Moro, Andrea (1997), The raising of predicates. Predicative noun phrases and the theory of clause structure, Cambridge, Cambridge University Press. Perlmutter, David (1978), “Impersonal passives and the unaccusative hypothesis”, in: Jeri J. Jaeger/ Anthony C. Woodbury/Farrell Ackerman/Christine Chiarello/Orin D. Gensler/John Kingston/Eve E. Sweetser/Henry Thompson/Kenneth W. Whistler (edd.), Proceedings of the fourth annual meeting of the Berkeley Linguistic Society, Berkeley, CA, Berkeley Linguistics Society, 157–189. Perlmutter, David (1989), “Multi-attachment and the unaccusative hypothesis. The perfect auxiliary in Italian”, Probus 1, 63–119. Reinhart, Tanya/Siloni, Tal (2005), “The lexicon-syntax parameter. Reflexivization and other arity operations”, Linguistic Inquiry 36, 389–436. Rigau, Gemma (1997), “Locative sentences and related constructions in Catalan. Ésser / haver alternation”, in: Amaya Mendikoetxea/Myriam Uribe-Etxebarria (edd.), Theoretical issues at the morphology-syntax interface, Bilbao, Donosti-San Sebastián, Euskal Herriko Unibertsitatea, Gipuzkoa Foru Aldundia, 395–421. Rosemeyer, Malte (2014), Auxiliary selection. Gradience, gradualness, and conservation, Amsterdam/ Philadelphia, Benjamins. Shannon, Thomas (1990), “The unaccusative hypothesis and the history of the perfect auxiliary in Germanic and Romance”, in: Henning Andersen/Konrad Koerner (edd.), Historical Linguistics 1987. Papers from the 8th International Conference on Historical Linguistics, Amsterdam/ Philadelphia, Benjamins, 461–488. Shannon, Thomas (1995), “Toward a cognitive explanation of perfect auxiliary variation. Some modal and aspectual effects in the history of Germanic”, American Journal of Germanic Linguistics and Literatures 7, 129–163. Sorace, Antonella (2000), “Gradients in auxiliary selection with intransitive verbs”, Language 76, 859–890.

326

Jaume Mateu Fontanals

Sorace, Antonella (2004), “Gradience at the lexicon-syntax interface. Evidence from auxiliary selection and implications for unaccusativity”, in: Artemis Alexiadou/Elena Anagnostopoulou/Martin Everaert (edd.), The unaccusativity puzzle. Explorations of the syntax-lexicon interface, Oxford, Oxford University Press, 243–268. Stolova, Natalya I. (2007), “Italian split intransitivity and image schemas”, Annual Review of Cognitive Linguistics 5, 77–106. Talmy, Leonard (2000), Toward a cognitive semantics. Typology and process in concept structuring, Cambridge, MA, MIT Press. Torrego, Esther (1989), “Unergative-unaccusative alternations in Spanish”, MIT Working Papers in Linguistics 10, 253–272. Tuttle, Edward F. (1986), “The spread of esse as a universal auxiliary in Central Italo-Romance”, Medioevo Romanzo 11, 229–287. Van Valin Jr, Robert D. (1990), “Semantic parameters of split intransitivity”, Language 66, 221–260. Vincent, Nigel (1982), “The development of the auxiliaries habere and esse in Romance”, in: Nigel Vincent/Martin Harris (edd.), Studies in the Romance verb, London, Croom Helm, 71–96. Zaenen, Annie (1993), “Unaccusativity in Dutch. Integrating syntax and lexical semantics”, in: James Pustejovsky (ed.), Semantics and the lexicon, Dordrecht, Kluwer, 129–161.

III Sound, structure, and meaning

Michelle Sheehan

11 Subjects, null subjects, and expletives Abstract: This chapter provides an overview of the status of overt and null subjects in Spanish, Catalan, European Portuguese and Italian. The two main Minimalist approaches to null subjects imply that either: (a) the verbal morphology in null subject languages (NSLs) is pronominal (Barbosa 1995; Alexiadou/Anagnostopoulou 1998); or (b) a pronominal, which in some circumstances can be null, functions as the subject in these languages (Sheehan 2006; Roberts 2010a, amongst others). Crucially, these two approaches make very different empirical predictions regarding the status of overt subjects in pre- and postverbal position as well as the (non-)existence of null expletives, the exploration of which forms the basis of this chapter. Interestingly, it seems that while an (a)-type approach is more apt for some Romance NSLs, others require a (b)-type analysis. The role of the interfaces with morphology and information structure in licensing null subjects is also discussed. Keywords: pro-drop, A-bar, morphology, information structure, inversion, null subjects, dislocation, expletives

1 Introduction Many of the “national” Romance languages have been characterized as pro-drop or null subject languages (NSLs) because they allow pronominal subjects to remain implicit/null in the correct information-structure context. Consider, for example, the following examples from Italian, Spanish, European Portuguese (Por.) and Catalan, compared with French and English, which are standardly held to be non-NSLs:1 (1)

a.

It.

Canti sing.2SG

bene. well

b.

Sp.

Cantas sing.2SG

bien. well

c.

Por.

Cantas sing.2SG

bem. well

Acknowledgement: The writing of this paper was funded by the European Research Council Advanced Grant No. 269752 “Rethinking Comparative Syntax”. 1 The licensing of 1st/2nd-person null subjects behaves differently from the licensing of 3rd-person null subjects. For discussion of 3rd-person null subjects and their need to be linked to a topic antecedent cf. Frascarelli (2007).

330

Michelle Sheehan

d.

Cat.

Cantes bé. sing.2SG well ‘You sing well.’

e.

Fr. *(Tu) you

chantes sing.2SG

f.

Fr. *(You) sing well.

bien. well

While this basic property fairly obviously holds at a descriptive level, the correct analysis of null subjects has remained elusive. One of the difficulties involves the apparent role of the interfaces with information structure and morphology in the licensing of Romance null subjects, raising some challenges for modular theories of syntax. Moreover, other surface effects which have been connected to the availability of null subjects such as ‘free inversion’ and the violation of the that-trace filter also seem to be sensitive to prosodic factors, further complicating the modular view (cf. Zubizarreta 1998 and Kandybowicz 2006 respectively). In this chapter we review the main analyses of NSLs within Government and Binding (GB) and Minimalism in relation to a number of Romance languages and examine the empirical predictions of these approaches in terms of the distribution and status of null/overt subjects/ expletives in these languages. The structure of the chapter is as follows. Section 2 considers the main analyses of NSLs in GB and Minimalism, making clear how their empirical predictions differ and how they tackle the syntax-morphology interface. Section 3 considers the distribution and status of overt subjects in a number of Romance NSLs and the implications for the Minimalist analyses expounded in section 2. Section 4 briefly considers the status of inversion and the (non-)existence of null expletives in Romance null subject languages. Finally section 5 concludes and raises some issues for future research.

2 Analyses of Romance null subjects 2.1 Null subjects in Government and Binding Theory In GB Theory, Rizzi’s influential analysis of NSLs involves pro, an empty pronominal which needs to be both: (2)

a.

licensed by Xy0

b.

identified by binding from features on the local head Xy0. (Rizzi 1982; 1986)

331

Subjects, null subjects, and expletives

If a given head Xy0 licenses pro then null (expletive) subjects will be available in a position local to Xy0, but the interpretation of pro (as referential or quasi-argumental) will depend on the extent to which pro is also identified (by rich agreement morphology on Xy0). This provides a fine-grained typology of NSLs, based on the properties of finite I in a given language (Rizzi 1986): (3)

a.

Full NSL: both referential and non-referential pro (proref and proexpl) are identified – e.g. Italian, Spanish, Greek

b.

Semi NSL Type I: only null non-referential pro, i.e. quasi-argumental and non-argumental expletives are identified – e.g. Icelandic, Yiddish

c.

Semi NSL Type II: only null non-argumental pro, but not referential or quasi-argumental pro is identified – e.g. Dutch, German

d.

Non-NSL: pro is not licensed at all – e.g. English, French

The typology in (3) concerns the properties of finite I across languages. The prediction, though, is that even within a given language, heads might potentially differ with respect to their ability to license/identify pro. Thus as Rizzi (1986) shows, Italian has different heads which behave like each of the finite ‘I’s in (3a–d) in their licensing/ identification possibilities. In finite clauses in Italian, pro is licensed and identified as referential by rich agreement on I, as per (3a): (4)

It.

a.

Ritengo [che pro sia be.SBJV.3SG believe.1SG that ‘I believe that (he) is nice.’

simpatico]. nice

b.

Ritengo [che pro sia be.SBJV.3SG believe.1SG that ‘I believe that (it) is too late for S.’

troppo too

c.

Ritengo [che pro sia be.SBJV.3SG believe.1SG that ‘I believe that (it) is likely that S.’

probabile che S]. probable that S (adapted from Rizzi 1986, 541)

tardi late

per for

S]. S

In Italian non-finite Aux-to-Comp complements (discussed by Rizzi 1982, 127–129), pro is licensed in Italian, but only quasi-argumental pro (the subject of weather predicates) can be identified:2 2 The subject of weather predicates is taken to be quasi-argumental because, amongst other things, it can control PRO: (i) It rained after PRO having snowed.

332

(5)

Michelle Sheehan

It.

a. *Ritengo [essere pro believe.1SG be.INF ‘I believe to be (he) nice.’

simpatico]. nice

b.

Ritengo [essere pro troppo too believe.1SG be.INF ‘I believe to be (it) too late for S.’

tardi late

c.

Ritengo [essere pro probabile probable believe.1SG be.INF ‘I believe to be (it) likely that S.’

per S]. for S

che S]. that S (adapted from Rizzi 1986, 542)

In Italian small clauses, pro is licensed, but neither referential nor quasi-argumental pro can be identified. All that is possible is expletive pro, which does not require licensing: (6)

It.

a. *Ritengo believe.1SG

[pro

simpatico]. nice

b. *Ritengo believe.1SG

[pro

troppo too

c.

tarde late

Ritengo [pro probabile believe.1SG probable ‘I believe it likely that S.’

per for

che that

S]. S

S]. S (adapted from Rizzi 1986, 542)

Finally, Rizzi shows that in Italian Control clauses, introduced by di ‘of’, pro is not even licensed (because of the lack of Case) and only PRO (a null subject with very different properties) is possible:3 (7)

It.

a.

proi

ritengo [di PROi believe.1SG of ‘I believe that I am nice.’

b. *proi

ritengo [di proj believe.1SG of ‘I believe him to be nice.’

essere be.INF

essere be.INF

c. *Ritengo [di pro essere troppo be.INF too believe.1SG of ‘I believe it to be too late for S.’

simpatico]. nice

simpatico]. nice

tardi late

per for

S]. S

3 Note, however, that in French/English pro fails to be licensed even in contents where Case is otherwise available, so, while pro requires Case (like all DPs), it also requires licensing of a more specific nature.

Subjects, null subjects, and expletives

333

d. *Ritengo [di pro essere probabile che S]. be.INF probable that S believe.1SG of ‘I believe it to be likely that S.’ (adapted from Rizzi 1986, 541s) These facts can be taken as evidence that (i) the availability of null subjects must be relativized to specific heads in a given language and (ii) there are different kinds of null subjects (referential, quasi-argumental and expletive), the availability of which in a given language is also relativized to specific contexts. Note that these facts rule out a purely lexical approach to NSLs whereby pro is simply made available or not in the lexicon of a given language, as well as macro-parametric accounts where an entire language permits null subjects across the board. Rather, on Rizzi’s view, pro is universally available in natural languages but its distribution is constrained by (2). One of the key issues with this approach is the question of morphological richness and its connection to identification. In finite clauses, Italian (like to a slightly lesser extent Spanish varieties and Portuguese) has a rich morphological paradigm in that all person/number combinations are differentiated: (8)

Rich agreement: Italian present tense paradigm, regular ‑are verb singular plural 1st canto cantiamo 2nd canti cantate 3rd canta cantano

Formally, Rizzi (1986, 543) claims that pro can be referential only if it is licensed for both person and number, as is clearly the case with the subjects of finite I, given (8). In (5), on the other hand, non-finite I in Aux-to-Comp contexts is not specified for person, but is abstractly specified for number, and so only a quasi-argumental reading is possible. The difference between (5) and (6) is that, in (6), pro is not theta-marked by the licensing head and so no identification at all is possible. In the absence of any identification for person or number only an expletive interpretation is possible. The following table summarises: Table 1: feature specification of pro. Interpretation of pro

person

number

Referential pro Quasi-argumental pro Expletive pro

+ – –

+ + –

A potential weakness with this proposal is that syntactic specification for a given feature and morphological realisation of that feature do not always go hand in hand. Thus Agr in Aux-to-Comp contexts is only abstractly specified for number, as

334

Michelle Sheehan

no number inflection is present morphologically. This fact can be considered an advantage or a flaw depending on how modular a view of syntax one wants to maintain. On the plus side, positing syntactic features which only partially overlap with morphological realisation maintains an autonomous syntactic component. On the downside, though, the empirical predictions of such a model are weaker. What independent evidence is there that Agr in Aux-to-Comp contexts bears a number specification except for the licensing of quasi-argumental pro? Another question is why the presence of +person always implies the presence of +number. Why is it not possible for a head to be +person only? Would this be sufficient to license referential pro? Following work by Perlmutter (1971) and Taraldsen (1980), Rizzi (1982) further proposes that several additional surface effects follow from the licensing and identification of referential pro: (9)

Obligatorily null expletive/quasi-argumental subjects It. (*egli / ciò) sta piovendo. it / this is.3SG raining ‘It’s raining.’

(10)

Apparent violations of the that-trace filter It. Chi credi che partirà? who think.2SG that come.FUT.3SG ‘Who do you think (*that) will come.’

(11)

Free inversion in simple clauses It. Ha telefonato Gianni. has.3SG telephoned Gianni ‘Gianni rang.’

The Rizzian approach to NSLs provides an elegant and explanatory account of (9–11). The obligatory availability of null expletives/quasi-argumental pro where referential pro is identified follows (a) from the licensing/identification distinction (for pro to be identified it must also be licensed) and (b) because (+person, +number) morphology rich enough to identify pro as referential will automatically be rich enough to identify quasi-argumental pro (as +number). The possibility of free inversion is due, in turn, to the availability of a null expletive which satisfies the subject requirement (the EPP), precluding the need for subject movement to Spec,IP. The that-trace filter is then avoided by extracting the subject from its (governed) post-verbal position (12), something which is also possible in Englishtype languages with overt expletives (13): (12)

It.

che pro partirà Chii credi come.FUT.3SG who think.2SG that ‘Who do you think (*that) will come.’

ti?

Subjects, null subjects, and expletives

335

(13) Whati did you say that there was ti in the box? Crucially, Rizzi (1982) further argues that Italian covert movement is subject to the that-trace filter, following work on French by Kayne (1981). As such, it is not that the that-trace filter is itself parameterised, but rather that it is avoided in Italian and other NSLs because of the availability of pro and a postverbal A-position. All in all, then, Rizzi’s account of NSLs is highly elegant and in the context of GB theory provides an explanatory account of the properties of Romance NSLs as well as the means for a fine-grained typology of null arguments across and within languages. Questions remain, however, concerning the nature of the syntaxmorphology interface, notably the nature of the specifications for +person/number and how/why exactly these serve to license the various kinds of pro.

2.2 Null subjects in minimalism In the content of the Minimalist Program MP, Rizzi’s analysis is, however, problematic on conceptual grounds because (i) the availability of an entity such as pro which, when it enters the derivation lacks content at both the LF and PF interface is highly suspect; (ii) licensing relies crucially on government, a stipulative language-specific relation which is rejected in Minimalist approaches; and (iii) the mechanism of identification cannot easily be restated in terms of the Minimalist operation Agree (cf. Holmberg 2005). Agree involves the valuation of uninterpretable features [uF] by interpretable features [iF] in a Probe-Goal configuration (Chomsky 1993): (14) Probe[uF] . . .Goal[iF] Holmberg notes that there are two possible ways to restate Rizzi’s (1982; 1986) theory of identification in terms of Agree: A. The agreement morphology on I is interpretable (cf. Barbosa 1995; Alexiadou/ Anagnostopoulou 1998). B. The agreement morphology on I is uninterpretable and valued by interpretable Ф-features originating on the null pronominal (cf. Holmberg 2005; Sheehan 2006; Roberts 2010a). In simplified terms, either: (a) pro does not exist, or (b) pro is an “ordinary” pronoun which simply lacks a PF realization. The third option, most similar to Rizzi’s proposal, whereby pro bears only uninterpretable Ф-features, which are valued by interpretable features on I, is not theoretically viable. This is because pronouns are arguably just bundles of Ф-features with a PF representation (Ritter 1995). A null pronoun without interpretable Ф-features would therefore be uninterpretable at both the PF and LF interfaces.

336

Michelle Sheehan

Crucially, A and B (above) assume different things about the syntax-morphology interface and make subtly different empirical predictions. It is the aim of this chapter to examine these in some detail. We first consider these competing analyses before testing them in sections 3 and 4.

2.2.1 Interpretable agreement morphology Following Jelinek (1984) and Barbosa (1995), Alexiadou/Anagnostopoulou (1998) propose that the verbal agreement morphology (Agr) in a rich agreement language “includes a nominal element ([+D, +interpretable phi-features, potentially +Case])” (Alexiadou/Anagnostopoulou 1998, 516). Thus Agr in NSLs like Greek and Spanish has “exactly the same status as pronouns in the English paradigm” (ibid.) and Agr is stored as a lexical item in the lexicon, of category D. There are two different interpretations of this proposal, however, depending on whether Agr can also absorb theta-roles (ibid. section 6.3): A.1 (15), and A.2 (16):4 A.1 A.2

Agr can absorb theta-roles – we don’t need pro at all. Agr cannot absorb theta-roles – we still need referential pro in thematic positions

(15)

A.1.

(16)

A.2.

In both (15) and (16), v+V+AgrD movement to T serves to satisfy the EPP (subject requirement), precluding the need for an expletive in Spec,TP because of (17). (17)

Parameterised mode of EPP-checking: Move/merge XP vs. move/merge X0. (Alexiadou/Anagnostopoulou 1998)

5

4 For various reasons, the inflectional head of the sentence which was referred to as I in the GB era is now usually referred to as T. I adopt the Minimalist terminology here, though it is of no real consequence. 5 Alexiadou/Anagnostopoulou (1998) actually argue that (17) is a parameter associated with AgrP, rather than TP, with a separate parameter determining the availability of Spec,TP as a subject position. We simplify matters somewhat here for ease of exposition.

Subjects, null subjects, and expletives

337

In A.1 (15), roughly equivalent to Jelinek’s (1984) account of Walpiri, we also eliminate the need for a referential pro in Spec,vP as Agr absorbs the relevant thetarole, whereas in A.2 (16), the more conservative position taken by Alexiadou/ Anagnostopoulou (1998) and Barbosa (1995; 2009), referential pro remains in Spec,vP (cf. Jelinek 1984 for related discussion). These two approaches therefore make different predictions with respect to the status of overt pre- and postverbal subjects. According to A.1, any overt subject will be dislocated in an A-bar position as there is no A-position for overt subjects (with the exception of the pronominal suffix Agr). According to A.2, however, overt subjects can be base-generated or moved to a preverbal A-bar position or remain in a postverbal A-position in spec vP. In the latter case, obligatory verb movement to T will result in the only A-position for overt subjects being postverbal. Both A.1 and A.2 thus have the apparent advantage of explaining the general absence of overt expletives in NSLs because verb movement to T serves to satisfy the EPP in both cases (but cf. section 4 below). It is not clear, however, that they can account for the more fine-grained distinction between the availability of quasiargumental and referential pro in the range of contexts discussed by Rizzi. Nonetheless, their proposal has the advantage of offering an elegant account of the NSP which attributes it to the status of Agr in a given language: whereas, in an NSL, Agr is an independent lexical entry, in non-NSLs, verbs are stored along with their inflections in the lexicon. Under A.2, which is the approach actually adopted by Alexiadou/Anagnostopoulou (1998), the Rizzian connection between free inversion and (expletive) null subjects is also maintained: both arise as surface effects where movement of a verbal complex into the inflectional domain serves to satisfy the EPP. In terms of the syntax-morphology interface, Alexiadou/Anagnostopoulou (1998) explicitly state that their analysis is syntactic: what counts for the setting of the NSP are the syntactic effects they discuss, with morphological richness being an additional (optional) morphological surface manifestation of this syntactic parameter.

2.2.2 Uninterpretable agreement morphology A version of B, whereby the agreement morphology in NSLs bears uninterpretable phi-features is explored by Roberts (2010a), Sheehan (2006; 2010) and Saab (to appear): the idea being that the uninterpretable Ф-features on T are valued by the interpretable Ф-features of the pronominal/DP subject, as in non-NSLs like English and French. Where the subject is pronominal, though, the idea is that it can then be deleted at PF if certain conditions hold (going back, broadly speaking, to the analysis of Perlmutter 1971). Here I give a hybrid version of this account, drawing on the above proposals. Following Roberts (2010a), assume that T bears an uninterpretable D feature [uD] in NSLs as the narrow syntactic correlate of rich agreement. Assume, further, that pronouns are just D heads bearing phi- and Case features and

338

Michelle Sheehan

that nominative Case is an uninterpretable Tense feature (following Pesetsky/Torrego 2001). If these things hold then pronouns, unlike other arguments, will constitute ‘defective goals’ in Roberts’ (2010a) terms: once T has probed and agreed with a pronoun, the features of said pronoun will constitute a proper subset of the features of T. Consider the following Spanish example by way of illustration: (18)

Once T has agreed with the 1sg subject its phi-features and uD feature are valued, as is the subject’s [uT] (Case) feature. As such, the valued features of T form a superset of the valued features of the subject. Although LF cares which of these features are interpretable and which are not, PF arguably does not as morphology commonly spells out uninterpretable valued features. There are then two options regarding the EPP in such a system: B.1 T lacks an EPP feature – the subject is incorporated into T B.2 T bears an EPP feature – the subject raises to Spec,TP If T lacks an EPP feature (B.1) then the valued features [1s T, D] on T form a chain at PF with [1s T, D] on the pronoun and the latter is deleted by virtue of being the lowest link in said chain (in the same way that chain reduction happens more generally – cf. Nunes 1999). This is formally equivalent to cliticisation in Roberts’ (2010a) terms, with the subject agreement functioning as a syntactic clitic. Such a system is basically equivalent to A.2 discussed above (i.e. the system proposed by Alexiadou/Anagnostopoulou 1998) with the added advantage that it permits the elimination of referential pro. In both cases (A.2 and B.1) the postverbal position is the only A-position. Alternatively, under B.2, T bears an EPP feature which attracts the subject to its specifier. (19)

Subjects, null subjects, and expletives

339

This gives rise to a three-link chain at PF containing the two copies of the subject plus the intermediate verbal inflection. The question here is why it is that this three-link chain leads to deletion of the highest and lowest copies of the subject and leaves only those features on T to be spelled out. A potential answer to this comes from the fact that the features [1s T, D] form part of a larger bundle of features on T which in the partly fusional morphology of Romance languages must be spelled out by a morpheme which realises also mood and tense features. It is not possible, then, to delete the intermediate copy of [1s T, D] as this would result in a combination of features with no morphological exponent in the language. As such, unusually, the intermediate copy in the chain is privileged and both the highest and lowest copies deleted (for economy reasons) (cf. also Nunes 1999 and Saab to appear for discussion). An additional prediction of this approach is that where a pronominal subject bears additional focus/topic features it cannot be deleted. This accounts for the special discourse interpretation associated with overt pronouns in null subject languages. This B.2 approach makes very different predictions regarding the distribution and status of overt subjects in Romance NSLs, and indeed has much in common with Rizzi’s original analysis of null subjects. Where the goal is a full DP, additional features will mean that it fails to be a defective goal, and so will not be deleted. As T still bears an EPP feature (*), by hypothesis, there will be two potential A-positions for such subjects in Romance NSLs: Spec,TP and Spec,vP. The default position of overt subjects will be Spec,TP, but Spec,vP is expected to be available wherever the EPP is satisfied in some other way. As with analysis A.2/B.1, verb movement to T will make Spec,vP a postverbal position. Crucially, though, while preverbal subjects can occupy an A-position (20), it will also be possible for them to occupy an A-bar position, either as a base-generated CLLD topic (21) or via A-bar movement from a preverbal position (cf. also Camacho 2013). (20)

(21)

Analysis B also retains a slightly loose account of the link between rich agreement and referential null subjects: the presence of [uD] is taken to be connected to the presence of rich agreement because of acquisition pressures, but, as with Alexiadou/Anagnostopoulou (1998) account, it would presumably be possible to

340

Michelle Sheehan

detect this feature via the other syntactic effects of the NSP.6 Several further questions remain. For one, how can this kind of approach deal with the difference between licensing and identification argued to be necessary by Rizzi? Moreover, how can this approach also explain the distribution of null expletive and quasiargumental subjects? One possible way to account for the differing distribution of expletive and quasiargumental null subjects both across and within languages, is to posit differing structures for referential, quasi-argumental and expletive pronouns along the following lines (cf. also Déchaine/Wiltschko 2002): (22)

a.

Referential pronouns = [DP D [nP n [ΦP Φ ]]]

b.

Quasi-argumental pronouns = [nP n [ΦP Φ ]]

c.

Expletive pronouns = [ΦP Φ ] or even some subset of phi-features

If referential pronouns are DPs, they will only be deleted where T bears a [uD] feature. If, in such contexts, T also bears a [un] feature then Rizzi’s typology and implications can be maintained: quasi-argumental pronouns will only be able to delete where a probe bears a [un] feature, expletives, however, as pure ΦPs will be able to delete even where the probe bears no additional nominal features. The various contexts discussed above for Italian where only expletive null subjects are licensed or where expletive and quasi-argumental but not referential subjects are licensed will equate to the presence of [uD] or [un] features on the probe (with the same questions about independent falsifiability). The discussion thus far has focused on the mechanism equivalent to identification of null subjects: in a sense the [uD] feature “identifies” a referential null subject in permitting its deletion, [un] does the same for a quasi-argumental and [uΦ] is sufficient to do this for expletive subjects. We still need some mechanism equivalent to licensing in order to explain why, in some languages with agreement morphology, even expletive subjects cannot be deleted (e.g. English, French). Following Roberts/ Roussou (2001), Holmberg (2005) and Landau (2007) the licensing of null subjects (i.e. the availability of deletion independently of the features of the probe) can be attributed to an additional parameter determining whether a given EPP feature must be satisfied at PF or not: (23) The EPP associated with a head H holds/does not hold at PF. 6 An anonymous reviewer asks me to specify what this implies for languages which lack agreement morphology but which nonetheless have null subjects (e.g. Chinese, Japanese, Mauritian Creole). Crucially, the null subjects exhibited by these kinds of languages have different properties to the null subjects of null subject languages with rich agreement, suggesting that they require an entirely different analysis (cf. Huang 1989; Adone 1994; Tomioka 2003; Neeleman/Szendrői 2007; Takahashi 2008; Şener/Takahashi 2010; Barbosa 2013, amongst others, but also Duguine 2014 for the opposite view).

Subjects, null subjects, and expletives

341

Clearly only languages in which finite T has a negative setting for (23) will permit deletion of subject pronouns, regardless of the features of T. In a language like English, therefore, where (23) has a positive setting, even expletives must remain overt even where they agree with T.

3 The status of overt subjects Recall the different predictions of the Minimalist analyses regarding overt lexical subjects: Table 2: Predictions of the three analyses regarding overt subjects in NSLs.

Preverbal subject Postverbal subject

A.1

A.2/B.1

B.2

A-bar position A-bar position

A-bar position A-position

A-bar/A-position A-position

A number of different kinds of tests can be used to tease apart these predictions. I assume that DPs which are base-generated in a left-peripheral A-bar position are clitic left dislocated (CLLD) and thus predicted to share the properties of other CLLD arguments (occurring in virtually any subordinate clause, displaying obligatory connectivity effects, always taking wide scope over TP-internal elements, failing to license parasitic gaps, being impossible with non-referential DPs and being island-sensitive – cf. Cinque 1990).7 Of course, the possibility remains, in analyses A.2/B.1 and B.2, that preverbal A-bar subjects may be derived rather than basegenerated, a point to which we return below. The status of overt subjects in Romance NSLs has been the subject of much heated debate and, given the brevity of this chapter, it will not be possible to provide a comprehensive overview of the literature. Instead, we focus on some of the most cited and/or compelling arguments in favour of the three distinct positions.

3.1 Test 1: Adverb placement Alexiadou/Anagnostopoulou give potential evidence for A.1/A.2 from the placement of adverbs in NSLs. They claim that, in NSLs, adverbs can intervene between the

7 Cinque (1990) claims that CLLD is sensitive only to strong and not weak islands, but as López (2009, 6) shows, once indefinites are extracted, a sensitivity to weak islands also emerges.

342

Michelle Sheehan

verb and the subject, whereas this is not the case in non-NSLs. Compare Spanish with French (a non-NSL):8 (24)

Sp.

Juan ya quiere irse. Juan already want.3SG go=SELF.CL ‘Juan already wants to leave.’

(25)

Fr.

Jean {*déjà} veux {déjà} Jean already want.3SG already ‘Jean already wants to leave.’

s’en aller. SELF.CL =PART =go

A potential explanation for this contrast is that, in Spanish, all preverbal subjects occupy an A-bar position above Spec,TP, whereas in French all preverbal subjects occupy Spec,TP. If it is assumed that adverbs cannot adjoin to the X-bar level, then the contrast in (24)–(25) follows naturally: there is simply no position for the adjunct in French (25). It is not so clear, however, that this contrast can be attributed to the NSP per se. While Portuguese patterns with Spanish, Italian, an NSL, patterns with French in this respect:9 (26)

It.

Maria {*già} vuole {già} Maria already want.3SG already ‘Maria already wants to leave.’

andarsene. go=SELF.CL =PART

An alternative explanation is that the French/Italian vs. Spanish/Portuguese contrast is due to differences in verb movement in the two groups of languages (Emonds 1978; Pollock 1989; Belletti 1990, 39–42; Cinque 1999; Schifano to appear). If the verb raises higher in modern Italian/French than in Spanish/Portuguese then this explains the contrast in adverb placement possibilities. As such, these facts cannot be taken to clearly support any of A.1, A.2/B.1 or B.2.

3.2 Test 2: Wide/narrow scope of preverbal subjects Alexiadou/Anagnostopoulou also give evidence which seems to support A.1 from the scope of quantificational preverbal subjects, which, they claim, always take wide scope with respect to object quantifiers. Consider the following evidence from Greek: 8 Note, however, that Alexiadou/Anagnostopoulou do not actually cite Spanish examples like this but rather cases where a fronted adverbial competes with the subject for the preverbal position (cf. Zubizarreta 1998). They do, however, discuss Greek examples equivalent to (23) to illustrate the same point. 9 Although, as an anonymous reviewer notes, in operatic registers of Italian, già often does occur before the finite verb, as in Spanish.

343

Subjects, null subjects, and expletives

(27)

kapios fititis stihiothetise kathe arthro. some student filed every article ‘Some (particular) student filed every single article.’ (wide scope only) (Alexiadou/Anagnostopoulou 1998, 505)

Greek

Alexiadou/Anagnostopoulou (1998) note that, in this respect, preverbal subjects pattern with CLLD objects in Greek: (28)

Greek

kapjo pedi to eksetase kathe kathigitis. some child CL . ACC examined every professor ‘Some child is such that every professor examined that child.’ (wide scope only) (Alexiadou/Anagnostopoulou 1998, 505)

This contrasts, interestingly, with the behaviour of postverbal subjects, which, according to Alexiadou/Anagnostopoulou (1998), have ambiguous wide/narrow scope in Greek: (29)

Greek

stihiothetise filed

kapios some

fititis student

kathe every

arthro. article

a.

‘Every article was filed by some student (or other).’

b.

‘Some (particular) student filed every single article.’ (Alexiadou/Anagnostopoulou 1998, 505)

These facts are actually problematic for both A.1 and A.2. Under A.1, both pre- and postverbal subjects occupy A-bar positions. A.2 also allows for the possibility that preverbal subjects are A-bar moved from a postverbal A-position (cf. Barbosa 1995). Given standard assumptions about A-bar movement, this means that at least some preverbal subjects would be able to reconstruct into a postverbal A-position, receiving ambiguous scope as in (29). The contrasts in question are also obviously problematic for analysis B, which would allow the possibility of a preverbal A-position for subjects, which would also permit optional reconstruction (Fox 1999). The facts themselves are less than clear in Romance NSLs, where the quantifiers used seem to affect which scope is preferred. Consider the following Spanish examples from Suñer (2002): (30)

Sp.

Algún some

estudiante student

sacó took.3SG

prestados lent.PL

todos all

a.

?? ‘All the books were borrowed by some student.’ (narrow scope subject)

b.

‘Some (particular) student borrowed all the books.’ (wide scope subject)

los the

libros. books

344

(31)

Michelle Sheehan

Sp.

Algún some

estudiante student

sacó took.3SG

prestado lent

cada each

libro. book

a.

‘Each book was borrowed by some student (or other).’ (narrow scope subject)

b.

‘Some (particular) student borrowed each book.’ (wide scope subject)

Similarly ambiguous examples can be created in Italian (as noted by Alexiadou/ Anagnostopoulou 1998, 511, fn 22, attributed to Jean-Yves Pollock), and Portuguese (cf. Sheehan 2006, chapter 2). More clear cut is the behaviour of the indefinite subjects of intransitive verbs headed by the indefinite article in Spanish, Portuguese, Italian, and Catalan, as noted by Barbosa (1995, 36). Whereas in English and French, such preverbal subjects are ambiguous in their scope, in Romance NSLs they seem to get an obligatorily ‘strong’ (wide scope) reading: (32) A letter of recommendation is required. (33)

Fr.

Une lettre de recommendation est a letter of recommendation is ‘A letter of recommendation is required.’

(34)

Cat. #Una a

(35)

Cat.

carta letter

de of

recomanació reccomendation

és is

requise. required

necessari. required

És necessari una carta de recomanació. is required a letter of reccomendation ‘A letter of recommendation is required.’ (examples from Barbosa 1995, 36s., also in Spanish, Portuguese, and Italian)

Taken in isolation, these contrasts again seem to suggest that all preverbal subjects are CLLD, favouring A.1 over A.2/B.1 or B.2. There are several potential complicating factors, however, which make these data slightly suspect. Firstly, it is well known that indefinites behave differently from other quantifiers. Secondly, while the French example involves a passive, the Catalan examples do not. Thirdly, this effect seems to be restricted to (some) intransitive verbs. In transitive contexts (and indeed with some intransitive verbs), both weak and strong readings are available in preverbal position (cf. Pinto 1994; 1997, 202 on Italian):

Subjects, null subjects, and expletives

(36)

Sp.

Un plato les a dish CL . DAT ‘A (single) dish was ‘All the guests were

345

fue servido a todos los huéspedes. was served to all the guests served to all the guests.’ served with a (possibly different) dish.’

The interpretation of quantificational subjects in Romance NSLs is therefore puzzling but does not seem to provide conclusive evidence for or against any of A.1, A.2/B.1 or B.2. Further careful comparative research is clearly required in this domain.

3.3 Test 3: Binding of postverbal subjects Data from binding is often cited in favour of the view that only the postverbal subject position in Romance NSLs is an A-position. If solid, then this would be strong evidence in favour of A.2/B.1 and against A.1 and B.2. Montalbetti (1986) notes that preverbal overt pronouns in NSLs cannot be bound (cf. also Rigau 1988 on Catalan), and this is consistent with the fact that they have a special discourse status. Solà (1992), citing Rosselló (1986), further notes, however, that postverbal subjects in Catalan behave differently and can be bound:10 (37) Cat.

Tots els estudiants1 es pensen que ells1 aprovaran. all the students self= think.3PL that they passed a. All the students believe that they as a group will pass. (mutually encouraging) b. *For all the students it is true that x thinks x will pass. (egotistical)

(38) Cat.

que guanyaran ells1. Tots els jugadors1 estan convençuts all the players are.3PL persuaded.PL that win.FUT.3PL they a. All the players believe that they as a team will win. (football) b. For all the players it is true that x thinks x will win. (tennis) (Solà 1992, 290)

A similar effect also holds in Portuguese (Barbosa 1995), Spanish, and Italian (Sheehan 2006), illustrated here for Italian: (39)

It. *Nessuno1 ha detto che lui1 l’avrebbe nobody has said that he it=would.have.3SG ‘Nobody said that he would have done it.’

fatto. done

10 An anonymous reviewer notes that (37) actually sounds ungrammatical to him/her.

346

(40)

Michelle Sheehan

It.

Nessuno1 ha detto che l’avrebbe nobody has said that it=would.have.3SG ‘Nobody said that he would have done it.’

fatto done

lui1. he

Further data from preverbal foci suggest, however, that this is actually a difference between topics vs. foci, whereby topics resist binding (cf. Barbosa 1995; Sheehan 2006): (41)

Sp.

(42) It.

Nadie dijo que tan solo él quisiera nobody said that only he want.SBJV.3SG ‘Nobody said that only he would like an apple.’

una an

manzana. apple

Nessuno1 ha detto che soltanto lui1 l’avrebbe fatto. nobody has said that only he it=would.have.3SG done ‘Nobody said that only he would have done it.’

As such, it appears that the possibility of binding overt pronouns correlates with their information structure as foci, presumably because overt topics signal a change of topic (disjoint reference). As such, these facts may not bear on the issue of an A-position for pronominal subjects, and provide no conclusive evidence either for or against any of A.1, A.2 or B.

3.4 Test 4: Non-referential subjects If all preverbal subjects are CLLD, then the naïve prediction is that non-referential preverbal subjects will not be possible, because CLLD objects, as noted above, cannot be non-referential. In this connection, Solà (1992) notes that bare indefinite/ negative quantifiers often cannot be preverbal subjects in Catalan, Spanish or Italian: (43)

It. *Studenti sono arrivati. students are.3PL arrived ‘Students have arrived.’

(44)

It. *Niente è successo. nothing is happened ‘Nothing has happened.’

(Solà 1992, 272)

This effect is limited to certain intransitive verbs, however, and negative quantifiers, at least, can surface preverbally with transitive predicates in Spanish, Italian, and Portuguese:

347

Subjects, null subjects, and expletives

(45)

Sp.

Nadie quiere ser político. nobody want.3SG be.INF politician ‘Nobody wants to be a politician.’

Examples such as (45) seem to pose a problem for analysis A.1.11 For analysis A.2, however, the possibility remains that such examples involve A-bar movement of the QP from a postverbal A-position, as Barbosa (1995) proposes, rather than base generation. In fact, Barbosa (1995; 1996; 2009) gives evidence from clitic placement in Portuguese to this effect. A preverbal quantifier, whether a subject or object, triggers proclisis in Portuguese, whereas a preverbal non-quantificational subject (or topic) triggers enclisis in matrix clauses. Matters are more complex in embedded clauses, however, where clitics are generally proclitic. As such, these data are potentially consistent with analysis A.2/B.1. They are also potentially consistent with analysis B.2, according to which the negative subject in (45) might occupy either a preverbal A-position or derived A-bar position, though this would leave the Portuguese clitic facts unexplained.

3.5 Test 5: Floating quantifiers Data from Cardinaletti (1997) and Sheehan (2006), (citing Rizzi 1982 and Burzio 1986) concerning the distribution of floating quantifiers in Romance NSLs suggests that not all preverbal subjects can be CLLD. Consider the following data from Spanish, which suggests that preverbal subjects allow quantifiers to be stranded in Spec,vP: (46)

Sp.

a.

b.

[Todos all [Los the

c. *Se SE

d.

los the

chicos] boys han have.3PL

chicos] boys se SE

se SE

han have.3PL

han have.3PL

comprado bought

SE

un a

coche. car

todos all

un a

coche. car

comprado bought

todos all

un a

han comprado todos un have.3PL bought all a ‘All the boys bought themselves a car.’ Se

comprado bought

coche car

[los the

chicos]. boys

coche. car

In order for the quantifier to be stranded in Spec,vP in examples like (46b), preverbal subjects would have to be (at least optionally) derived via A- or A-bar movement from a post-verbal A-position. This is consistent with A.2/B.1 or B.2 but again not A.1. Interestingly, post-verbal subjects do not share this property (46c), while null 11 Though cf. Alexiadou/Anagnostopoulou (1998), Camacho (2013) for a critique of this argument.

348

Michelle Sheehan

subjects do (46d). This is taken as evidence by Rizzi (1982) that pro occupies a preverbal position, which is consistent only with B.2, if correct.

3.6 Test 6: Subjects vs. topics Goodall (2001) notes that, in Spanish, clauses with fronted topics are islands for extraction, whereas clauses with preverbal subjects are not (at least for many speakers):12 (47) Sp. a. *A quién crees [que el premio se lo dieron]? CL . DAT CL . ACC gave.3PL to whom think.2.SG that the prize (Lit. ‘Who do you think that the prize they gave it to?’) b. A quién crees [que Juan le dio el premio]? to whom think.2.SG that Juan CL . DAT gave.3SG the prize ‘Who do you think that Juan gave the prize to?’ (Goodall 2001, 201) This seems to provide further evidence that not all preverbal subjects can occupy a CLLD position. In Catalan and Portuguese, the same contrast holds:13 (48) Cat. a. *A qui creus que el premi el van_donar? to whom think.2SG that the prize CL . ACC = gave.3PL b. A qui creus que en Joan va_donar el premi? to whom think.2SG that the Joan gave.3SG the prize ‘Who do you think that Juan gave the prize to?’ (49)

Por.

a. *A to

quem whom

no in.the

ano year

achas think.2SG

que that

o the

prémio prize

o CL . ACC

deram gave.3PL

passado]? last

b. A quem achas que o Rei deu o prémio to whom think.2SG that the king gave.3SG the prize no ano passado? in.the last year ‘Who do you think that the King gave the prize to last year?’ 12 Some speakers report this effect to be slightly weaker than that reported by Goodall but nonetheless find a contrast of the right kind. 13 A reviewer points out that the Spanish example in (47) has clitic doubling of the dative argument, whereas the Catalan example in (48) does not. For the speaker I consulted, however, clitic doubling of datives is strongly dispreferred, whereas it is often preferred in Spanish and is sometimes obligatory in Spanish with true datives (cf. Curvo 2003).

Subjects, null subjects, and expletives

349

Italian, too, displays the same contrast, though the topic island effect is much less robust: (50) It. a. ?A chi credi [che il premio lo abbiano dato]? CL . ACC have.SBJV.3PL given to whom think.2SG that the prize b. A chi credi [que Gianni abbia dato il premio]? to whom think.2SG that Gianni has.SBJV.3SG given the prize ‘Who do you think that Juan gave the prize to?’ These examples provide strong support that not all preverbal subjects can be CLLD, posing a serious problem for A.1. Indeed, assuming that (i) all derived preverbal subjects are foci and (ii) there can be only one focus per clause, such examples suggest that at least some preverbal subjects can occupy an A-position in all the Romance NSLs under discussion, favouring analysis B.2.

3.7 Test 7: Hortative contexts Another piece of evidence in favour of the view that not all preverbal subjects can be CLLD, hence against A.1, comes from the position of subjects in hortative constructions in Spanish (cf. Villa-García 2012). As Demonte/Fernández-Soriano (2009) show, CLLD elements in Spanish hortative constructions always precede que: (51)

Sp.

a.

El tenedor, ¡que lo cojan! the fork, that it take.SBJV.3PL ‘The fork, let them take it.’

b. *¡Que that

el the

tenedor fork

lo it

cojan! take.SBJV.3PL

(Villa-García 2012, 152)

The pattern is different with subjects, which can either precede or follow que, suggesting that preverbal subjects can either be CLLD or not: (52)

Sp.

a.

b.

Antonio, ¡que no lo vea! Antonio that NEG it see.SBJV.3SG ‘Antonio, may he not see it.’ ¡Que Antonio no lo that Antonio NEG it ‘May Antonio not see it.’

(Villa-García 2012, 155)

vea! see.SBJV.3SG (Demonte/Fernández-Soriano 2009, 39)

The same effect holds in Italian (Villa-García, citing Ledgeway 2005) and Catalan. As Villa-García notes, these facts seem to show that not all preverbal subjects can be CLLD in these languages. Depending on the structure of the left periphery, they might also be taken as evidence that some preverbal subjects occupy Spec,TP.

350

Michelle Sheehan

Whether Spec,TP is an A-position in Romance NSLs, however, is somewhat controversial (cf. Zubizarreta 1998; Gallego 2007 for the claim that Spec,TP is an A-bar position in Spanish).

3.8 Test 8: Basic word order Thus far, the various diagnostics discussed strongly militate against an analysis along the lines of A.1 for any of the main Romance NSLs. It is more difficult, however, to decide between A.2/B.1 and B.2, with the possibility remaining that both analyses are correct for a subset of languages. One potential piece of evidence that some preverbal subjects occupy an A-position, hence in favour of B.2, comes from basic word order facts. The three analyses make different predictions concerning the basic word order in NSLs: A.1 No ‘basic’ subject position A.2/B.1 VS(O) basic word order (subject base generated in Spec,vP) B.2 SV(O) basic word order All else being equal, as long as the object does not move and V raises past the subject to the T-domain, A.2/B.1 predicts a basic VS(O) word order whereas B.2 predicts SV(O), because S is attracted past the verb to Spec,TP. Although it is difficult to determine what the basic order of a language is, the consensus following Chomsky (1971) and Jackendoff (1972) is that basic word order corresponds to ‘wide focus’ sentences which are felicitous answers to questions like ‘what happened?’.14 In many transitive/ditransitive contexts in Spanish, Portuguese, and Italian, SVO is the word order generally required: (53)

It.

a.

Cos’ è successo? what is happened ‘What happened?’

b.

Gianni ha dato un libro Gianni has given a book ‘Gianni gave a book to Maria.’

a to

Maria. Maria

c. *Ha dato Gianni un libro has given Gianni a book ‘Gianni gave a book to Maria.’

a to

Maria. Maria

d. #Ha dato un libro a Maria has given a book to Maria ‘Gianni gave a book to Maria.’

Gianni. Gianni

14 But cf. Solà (1992) and Camacho (2013) for a critique of this view.

Subjects, null subjects, and expletives

351

Indeed, Hulk/Pollock (2001, 3) claim that “[t]here is a consensus among both traditional and generative grammarians that the canonical surface word order of the Romance languages is subject-verb-object”. If true, this would fall out naturally from a type B.2 analysis, but be mysterious under type A.1 and A.2/B.1 analyses. In fact, as A.1 allows for no subject A-position, it makes unclear predictions about basic word order. Things are not actually so clear cut, however. In fact, for all of the languages under discussion, opposing views have been presented regarding the status of SVO order: cf. Costa/Duarte (2002); Costa (2004) vs. Barbosa (1995; 2009) on Portuguese; Suñer (2002) vs. Leonetti (2008; 2014) on Spanish; Rizzi (1997); Cardinaletti (1997) vs. Moro (1997); Manzini/Savoia (2002) on Italian; and Forcadell (2013) vs. Vallduví (1993) on Catalan. As Alexiadou/Anagnostopoulou (1998) note, Spanish like Greek allows VSO orders (unlike Italian and Catalan) in certain out of the blue contexts (cf. Zubizarreta 1998, chapter 3; Gallego 2013): (54)

Sp.

a. ¿Qué what b.

Sp.

a. ¿Qué what b.

pasado? happened

Juan ha ganado la Juan has won the ‘Juan won the lottery.’

b’. ??Ha has (55)

ha has

ganado won

Juan Juan

pasó happened.3SG

la the

lotería. lottery

lotería. lottery

ayer? yesterday

Ayer ganó Juan la Yesterday won.3SG Juan the ‘Yesterday Juan won the lottery.’

lotería. lottery

For some speakers, though, it seems that VSO orders are only fully felicitous in Spanish where some other XP surfaces preverbally (Zubizarreta 1998; Sheehan 2010).15 If this is the case then such orders do not provide very strong evidence against a type B.2 and in favour of a type A.2/B.1 analysis. For those speakers who allow VSO orders without any preverbal XP, however, such orders might be taken as evidence for an analysis along the lines of A.2/B.1. There appears to be variation across varieties in this respect. It is a much-discussed fact that Italian and Catalan, 15 Interestingly, the addition of que ‘that’ to (54b’) renders it much more acceptable. Cf. Etxepare (2010); Demonte/Fernández-Soriano (2013) on the status of quotative que.

352

Michelle Sheehan

unlike Spanish, do not appear to permit VSO orders with DP objects (cf. Belletti/ Shlonsky 1995; Gallego 2013). Portuguese does allow VSO orders but not, it seems, in out of the blue contexts (Costa 2004). Intransitive predicates paint a very different picture. In all of Catalan, Spanish, Italian, and Portuguese at least some intransitive verbs permit VS as well as SV order in out of the blue contexts (cf. Pinto 1994; 1997; Adger 1996; Zubizarreta 1998; Sheehan 2006; Corr 2012, amongst others). Consider the following from Portuguese: (56)

Por.

a.

O que é que the what is that ‘What happened?’

b.

Chegou arrived.3SG

c.

A avó chegou. the grandmother arrived.3SG ‘Grandmother arrived’

a the

foi? was

avó. grandmother

(Corr 2012, 16)

There is microparametric variation across these languages with respect to which predicates permit inversion of this kind, in all cases, though, it seems that inversion of the subject gives rise to a special ‘deictic’ interpretation (cf. Pinto 1994; 1997). Thus while (56c) simply states that grandmother has arrived somewhere, (56b) implies that she has arrived at the place where the speaker is. Crucially, this distinction does not track the unaccusative/unergative distinction. There are unaccusative verbs which typically disallow inversion in Spanish (change of state verbs like ‘to blush’) and unergatives which typically allow it (‘to call’, ‘to contribute’). According to Pinto (1994; 1997), in such contexts, it is a covert PP which satisfies the EPP in such contexts, permitting subject inversion. As predicted, where an equivalent overt PP remains postverbal, the subject must generally raise in Italian and Portuguese, and to a lesser extent Spanish (cf. Sheehan 2006; Corr 2012): (57)

Por.

Entrou entered.PST.3SG

o the

Nuno Nuno

(*no in.the

cinema). cinema

As such, even “free inversion” cannot be taken as strong evidence for a type A.2/B.1 analysis, as the distribution of postverbal subjects is actually fairly constrained. It is also not immediately accounted for by a type B.2 analysis, however, though the possibility remains that it instantiates a kind of locative inversion, whereby movement of a deictic PP mitigates the need for subject movement to Spec,TP (cf. Pinto 1994; 1997). We return to this issue in section 4.

353

Subjects, null subjects, and expletives

3.9 Test 9: Disambiguation In Romance NSLs, it is generally accepted that overt pronouns are usually used only for emphasis (Rigau 1988). The majority of overt pronouns, then, serve a special discourse function as A-bar foci or topics (including switch reference topics). Nevertheless it has been claimed that some overt pronouns do function as true A-subjects in Romance NSLs (cf. also section 3.1.3). Cardinaletti (1997) has shown that where verb forms are ambiguous in Italian, pro-drop is limited to certain persons. In the present subjunctive of Italian, 1st, 2nd and 3rd person singular verb endings are all syncretic. Whereas 1st and 3rd person permit pro-drop, 2nd person requires an overt pronoun: (58)

It.

Che possa riuscirci non è chiaro. that can.SBJV manage=there not is clear ‘It isn’t clear that I/*you/he can manage it.’

(59)

It.

Che tu possa riuscirci that you can.SBJV manage=there ‘It isn’t clear that you can manage it.’

non not

è is

chiaro. clear

Interestingly, this effect seems to hold not only in out of the blue contexts but even where there is a contextually salient 2SG antecedent (Luigi Rizzi, p.c.): (60)

It.

So know.1SG

che that

hai have.2SG

provato tried

ma but

non not

è is

facile easy

che that

*(tu) possa riuscirci. you can manage=there ‘I know that you’ve tried but it’s not going to be easy for you to manage it.’ The overt preverbal subject tu in (60) does not function as a topic or a focus, but rather serves to add essential morphological information to the underspecified verb form. Cole (2000) argues that a similar effect holds more generally of Romance NSLs, wherever morphological ambiguity of this kind arises. Consider, for example, the contrast between (61) and (62): (61) Sp. María y yo llegamos a casa. Encontré las llaves. . . Maria and I arrived.1PL at home found.1SG the keys. . . (62) Sp. María y yo llegamos a casa. *(Yo/ella) tenía las llaves. . . Maria and I arrived.1PL at home had.1SG /3SG the keys (Cole 2000)

354

Michelle Sheehan

In both (61) and (62), the 1SG pronoun yo is equally contextually salient. However, in (62), as opposed to (61), the following verb is morphologically ambiguous, requiring an overt subject to be used. Cole claims where agreement identification fails, NSLs have recourse to discourse pro-drop strategies (i.e. they look for a single salient discourse topic), and where these fail, require an overt subject. In such contexts, then, the overt preverbal pronoun seems to serve purely a disambiguating function rather than behaving like an A-bar topic or focus. These facts raise some issues for the Minimalist analyses discussed in section 2. According to A.1 and A.2/B.1, it is not possible for an overt preverbal subject to occupy an A-position, making examples like (60) potentially problematic. Analysis B.2, however, leaves open the possibility that deletion might fail in certain contexts leaving an overt pronoun in Spec,TP. The mechanism whereby deletion fails, of course, needs to be specified. Does the probe lack person features in such contexts or is the [uD] feature missing only in part of the paradigm?

3.10 Test 10: Parasitic gaps Thus far, then, the diagnostics clearly disfavour A.1 and some arguably show a preference for B.2 over A.2/B.1 across the Romance NSLs under discussion. The data from parasitic gaps, however, suggests that while B.2 is the correct account for some NSLs (Italian), some (minimally revised) version of A.2/B.1 may be the correct account for others (Spanish, Portuguese, and perhaps Catalan). Consider the predictions of the three approaches vis-à-vis the licensing of parasitic gaps. According to standard assumptions, parasitic gaps are only licensed where an A-bar moved XP c-commands a parasitic gap but its trace does not (Kayne 1983). Putting to one side the correct analysis of parasitic gaps (but cf. Nunes 1999), based on this description, the three approaches make the following predictions: A.1 A.2/B.1 B.2

preverbal subjects will not license parasitic gaps as they are always basegenerated preverbal subjects may license parasitic gaps as they can be A-bar moved from a low postverbal A-position preverbal subjects will not license parasitic gaps as A-bar subjects are moved through a preverbal A-position which c-commands the adverbial.

To our knowledge, these predictions have not previously been tested in Romance NSLs. Now consider the following facts from Spanish. Example (63) shows that an in-situ object does not license a parasitic gap: (63)

Sp. *Archivaste filed.2SG

el the

documento document

sin without

abrir. open.INF

Subjects, null subjects, and expletives

355

Example (64) shows that a wh-moved object does license a parasitic gap: (64)

Sp. ¿Qué documento archivaste sin abrir? what document filed.2SG without open.INF ‘What document did you file without opening?’

Now consider the preverbal subject of a passive. The fact that (65)–(66) are acceptable suggests that preverbal subjects can be derived via A-bar movement from a post-verbal position without having to move through Spec,TP, as predicted by A.2/B.1: (65)

Sp. ?El documento fue archivado sin the document was filed without ‘The document was filed without opening.’

abrir. open.INF

(66)

Sp. ?Ningún documento fue archivado sin no document was filed without ‘No document was filed without opening.’

abrir. open.INF

Interestingly, while Portuguese and perhaps Catalan seem to pattern like Spanish in this respect, Italian seems to behave differently, with preverbal subjects failing to license parasitic gaps: (67) It. *Hai archiviato il documento senza aprire. have.2SG filed the document without open.INF (68) It. ?Quale documento hai archiviato senza aprire? without open.INF which document have.2SG filed ‘Which document did you file without opening?’ (69) It. *Il documento è stato archiviato senza (prima) aprire. the document is been filed without first open.INF (70) It. *Nessun documento è stato archiviato senza (prima) aprire. no document is been filed without first open.INF The implication of these facts seems to be that in Spanish, Catalan, and Portuguese, preverbal subjects can be moved directly from a postverbal A-position without transiting through Spec,TP, whereas in Italian they cannot. This is consistent with the claim that in Italian preverbal subjects always need to raise to/through a preverbal A-position (as in analysis B.2), whereas in the other languages, they do not (as in analysis A.2/B.1).

356

Michelle Sheehan

A problem remains for the facts discussed in sections 3.1.6–3.1.8, however, which seem to suggest that preverbal subjects can occupy a preverbal A-position also in Spanish, Catalan, and Portuguese. One possible way of resolving this problem is to revise analysis A.2 slightly so that subjects can, but need not raise to Spec,TP to satisfy the EPP. If Spanish, Portuguese, and Catalan permit the EPP to be satisfied by either a head or phrase, a DP subject in Spec,vP, whether overt or null is actually equidistant from T with its head complex V+v+AgrD. It is plausible then, that either movement of V+v+AgrD to T or movement of DP to Spec,TP can satisfy the EPP in Spanish, Catalan, and Portuguese. Where the DP raises, the verb also raises to T for purely morphological reasons and the result is an SVO order with a preverbal A-subject. Where the subject fails to raise, however, V+v+AgrD movement serves to satisfy the EPP and the result is also grammatical. The equidistance of V+v+AgrD and Spec,vP from T makes both options available. In these terms, Italian would differ from these languages in requiring the EPP to be satisfied by an XP (cf. Alexiadou 2006 on further differences between Italian and Spanish). Although this analysis of Spanish, Portuguese, and Catalan may seem unparsimonious, something along these lines seems to be necessitated by the diagnostics discussed above. Of course such an account raises many questions which cannot be addressed here for reasons of space, notably regarding the status of postverbal narrowly focused subjects. We return briefly to issues of word order in section 4.

3.11 Summary of results Although there are gaps in the data and certain inconsistencies arise across the ten tests discussed above, the data seem to show overwhelmingly that an analysis along the lines of A.1 cannot be correct: preverbal subjects are not all CLLD in Romance NSLs. Analyses A.2 and B fare better. In Italian the data, particularly from parasitic gaps, seem to suggest that B is correct, as derived preverbal subjects appear to raise obligatorily to/through Spec,TP. In Spanish, Portuguese, and Catalan matters are more complex. It seems that subjects can occupy a preverbal A-position, but can also raise directly from a post-verbal A-position to a preverbal A-bar position, licensing parasitic gaps, at least for some speakers. This suggests that a revised version of A.2 is necessary for these languages whereby the EPP can be satisfied by either XP or head-movement.

4 On inversion and expletive pro When discussing subject-verb inversion in Romance NSLs, it is important to control for information structure. As discussed in section 3.1.8, the basic word order in many Romance languages seems to be SVO with transitives (with some apparent exceptions in Spanish). With intransitives, VS is often possible in out-of-the-blue contexts,

Subjects, null subjects, and expletives

357

but there is semantic evidence that such orders involve a kind of locative inversion with a covert PP. Where the subject is narrowly focused (in answer to “who did Y?” questions), however, inverted V(O)S orders become more generally possible in Romance NSLs, and there is a question of how this relates to the analyses under discussion. Unfortunately, a discussion of this marked inversion is beyond the scope of this chapter (cf. Sheehan 2010). An important remaining question concerns the (non-)existence of covert expletives in Romance NSLs. Expletive pro formed a crucial part of Rizzi’s (1982; 1986) classic analysis but has become suspect in the Minimalist context. As Alexiadou/ Anagnostopoulou (1998) note, expletive pro has no PF interpretation and the apparent lack of definiteness effects in VS(O) orders in NSLs suggests that it also lacks an LF interpretation. A.1 and A.2/B.1 are thus apparently attractive in avoiding the need for null expletives.16 Interestingly, though, while the standard wisdom is certainly that Romance null subject languages lack definiteness effects of the English type, in certain controlled contexts, a kind of definiteness effect occurs. First consider existential constructions, as discussed by Leonetti (2008) and Fischer (2013). Spanish displays a definiteness effect in such contexts and Leonetti (2008) argues convincingly that where a locative PP is included inside VP, the effect is also observed in Italian and to some extent in Catalan: (71)

It.

??C’è LOC =is

la the

statua di Michelangelo in Piazza della Signoria. statue of Michelangelo in Piazza della Signoria (Fischer 2013, 13)

A similar effect can be observed with unaccusative verbs in some Romance languages. While it is true that the inverted subjects of unaccusatives fail to display a straightforward definiteness effect in Romance null subject languages (Fischer 2013), this is as expected if this inversion is triggered by locative inversion, which is not subject to definiteness effects (cf. Pinto 1994 and 1997 on Romance and Freeze 1992 on the connection between locatives and existentials). In out-of-the-blue utterances, however, with unaccusative verbs and overt postverbal PPs, a definiteness effect is attested in Portuguese and Italian, though not, apparently, in Spanish. In such contexts only indefinite subjects seem to be possible in a postverbal position (Sheehan 2006; 2010; Corr 2012):

16 What have been called “overt expletives” are attested in some Romance varieties (cf. Toribio 1996; Hinzelin/Kaiser 2007 on Caribbean Spanish; Carrilho 2005 on colloquial EP; Uriagereka 1995 on Galician; Maiden/Parry 1997 on Corsican, Sicilian, Neapolitan and Campanian; and Kaiser 2006, for an overview). In most cases, however, restrictions on the distribution of overt expletives in Romance NSLs make them look more like discourse particles than true expletives, but cf. Hinzelin/ Kaiser (2012) on Occitan varieties on the Francoprovençal border.

358

(72)

Michelle Sheehan

Por.

a.

O que é que the what is that ‘What happened?’

aconteceu? happened.3SG

b.

Chegou alguém ao arrived.3SG someone to-the ‘Someone arrived at school.’

b’. *Chegou o João ao arrived.3SG the João to-the ‘João arrived at school.’

colégio. school

colégio. school

This follows if, where an overt PP remains postverbal, the subject must raise to Spec,TP all else being equal. The contrast between (72b) and (72b’) suggests that a null expletive can satisfy the EPP in such contexts, permitting indefinite subjects to remain low. The fact that Spanish seems to lack this effect (Corr 2012) whereas Italian appears to have something similar (Belletti 1988) suggests that further parameterization is required here, particularly given the differences between existential and unaccusative contexts.

5 Conclusions This chapter has examined the behaviour of some Romance NSLs in the context of GB and the Minimalist Program, arguing that they fall into at least two groups. Though the data are complex and many questions remain, it would appear that Spanish behaves like a variant of a type A.2/B.1 language in which either XP or X-movement can satisfy the EPP. Italian, on the other hand, appears to be a type B.2 language in which some XP must always satisfy the EPP and null subjects are simply deleted. Interestingly, many of the analyses of NSLs face the same issue concerning the syntax-morphology interface. While it is recognized that rich agreement morphology seems to be involved in the licensing of null subjects in these languages (cf. especially 3.1.9), there is only ever a loose connection between the syntactic feature which is responsible for null subjects and its surface morphological manifestation. It is hoped that future research will serve to (i) resolve the issues surrounding some of the diagnostics discussed above and (ii) discover new diagnostics which illustrate more clearly how the various Romance NSLs should be syntactically distinguished.

Subjects, null subjects, and expletives

359

6 References Adger, David (1996), “Economy and optionality. Interpretations of subjects in Italian”, Probus 8, 117–135. Adone, Dany (1994), The acquisition of Mauritian creole, Amsterdam/Philadelphia, Benjamins. Alexiadou, Artemis (2006), “Uniform and non-uniform aspects of pro-drop languages”, in: Peter Ackema/Patrick Brandt/Maaike Schoorlemmer/Fred Weermann (edd.), Arguments and agreement, Oxford, Oxford University Press, 127–158. Alexiadou, Artemis/Anagnostopoulou, Elena (1998), “Parametrizing AGR. Word order, V-movement and EPP-checking”, Natural Language and Linguistic Theory 16, 491–539. Barbosa, Pilar (1995), Null subjects, PhD dissertation, Massachusetts Institute of Technology. Barbosa, Pilar (1996), “Clitic placement in European Portuguese and the position of subjects”, in: Aaron Halpern/Arnold Zwicky (edd.), Approaching second. Second position clitics and related phenomena, Stanford, CSLI Publications, 1–40. Barbosa, Pilar (2009), “Two kinds of pro”, Studia Linguistica 68:1, 2–58. Barbosa, Pilar (2013), “Pro as a minimal NP. Towards a unified theory of pro-drop”, Ms., Universidade do Minho, http://ling.auf.net/lingbuzz/001949 (11.06.2015). Belletti, Adriana (1988), “The case of unaccusatives”, Linguistic Inquiry 19, 1–34. Belletti, Adriana (1990), Generalized verb movement. Aspects of verb syntax, Torino, Rosenberg & Sellier. Belletti, Adriana/Shlonsky, Ur (1995), “The order of verbal complements in Italian and in Hebrew”, Natural Language and Linguistic Theory 13, 489–526. Bentley, Diane (2013), “Subject canonicality and definiteness effects in Romance there-sentences”, Language 89, 675–712. Burzio, Luigi (1986), Italian syntax. A government-binding approach, Dordrecht/Boston Hingham, MA, Reidel. Camacho, José (2013), Null subjects, Cambridge, Cambridge University Press. Cardinaletti, Anna (1997), “Subjects and clause structure”, in: Liliane Haegeman (ed.), The new comparative syntax, London/New York, Longman, 33–63. Carrilho, Ernestina (2005), Expletive “ele” in European Portuguese dialects, PhD dissertation, Universidade de Lisboa. Chomsky, Noam (1971), “Deep structure, surface structure, and semantic interpretation”, in: Danny D. Steinberg/Leon A. Jakobovits (edd.), Semantics, Cambridge, Cambridge University Press, 183–216. Chomsky, Noam (1993), “A Minimalist Program for linguistic theory”, in: Ken Hale/Samuel Jay Keyser (edd.), A view from building 20, Cambridge, MA, MIT Press, 1–52. Cinque, Guglielmo (1990), Types of A’-dependencies, Cambridge, MA, MIT Press. Cinque, Guglielmo (1999), Adverbs and functional heads. A cross-linguistic perspective, Oxford, Oxford University Press. Cole, Melvyn (2000), The syntax, morphology, and semantics of null subjects, PhD dissertation, University of Manchester. Corr, Alice (2012), Subject inversion in Ibero-Romance, MPhil Dissertation, University of Cambridge. Costa, João (2004), Subject positions and interfaces. The case of European Portuguese, Berlin/New York, De Gruyter. Costa, João/Duarte, Inês (2002), “Preverbal subjects in null subject languages are not necessarily dislocated”, Journal of Portuguese Linguistics 2, 159–176. Cuervo, María Cristina (2003), Datives at large, PhD dissertation, Massachusetts Institute of Technology.

360

Michelle Sheehan

Déchaine, Rose-Marie/Wiltschko, Martina (2002), “Decomposing pronouns”, Linguistic Inquiry 33, 409–442. Demonte, Violeta/Fernández-Soriano, Olga (2009), “Force and finiteness in the Spanish complementizer system”, Probus 21, 23–49. Demonte, Violeta/Fernández-Soriano, Olga (2013), “El que citativo, otros que de la periferia izquierda y la recomplementación”, in: Daniel Jacob/Katja Ploog (edd.), Autour de que – El entorno de que, Frankfurt am Main, Lang, 47–70. Duguine, Maia (2014), “Argument ellipsis. A unitary approach to pro-drop”, The Linguistic Review 31, 515–549. Emonds, Joe (1978), “The verbal complex V’-V in French”, Linguistic Inquiry 9, 151–175. Etxepare, Ricardo (2010), “From hearsay evidentiality to samesaying relations”, Lingua 120, 604– 627. Fischer, Susann (2013), “Unaccusatives vs. existentials. Explorations of the syntax-semantic interface”, in: Sofiana Chiriacescu (ed.), Theoretical implications at the syntax-semantics interface in Romance. Proceedings of the VI Nereus International Workshop, Konstanz, Universität Konstanz, Konstanzer Arbeitspapiere des Fachbereichs Sprachwissenschaft 127, 1–21. Forcadell, Montserrat (2013), “Subject informational status and word order. Catalan as an SVO language”, Journal of Pragmatics 53, 39–63. Fox, Danny (1999), “Reconstruction, binding theory, and the interpretation of chains”, Linguistic Inquiry 30, 157–196. Frascarelli, Mara (2007), “Subjects, topics, and the interpretation of referential pro”, Natural Language and Linguistic Theory 25, 691–734. Freeze, Ray (1992), “Existentials and other locatives”, Language 68, 553–595. Gallego, Ángel (2007), Phase theory and parametric variation, PhD Dissertation, Universitat Autònoma de Barcelona. Gallego, Ángel (2013), “Object shift in Romance”, Natural Language and Linguistic Theory 31, 409– 451. Georg A. Kaiser (2006), “Pronombres sujeto en construcciones impersonales de lenguas iberorrománicas”, in: B. Beatrix Fernández/Itziar Laka (edd), Andolin gogoan. Essays in Honour of Professor Eguzkitza, Vitoria-Gasteiz, Universidad del País Vasco/Euskal Herriko Unibertsitatea, 513–530. Goodall, Grant (2001), “The EPP in Spanish”, in: William D. Davies/Stanley Dubinsky (edd.), Objects and other subjects, Dordrecht, Kluwer, 193–223. Hinzelin, Marc-Olivier/Kaiser, Georg (2007), “El pronombre ‘ello’ en el léxico del español dominicano”, in: Wiltrud Mihatsch/Monika Sokol (edd.), Lenguas en contacto y cambio lexico-gramatical en el Caribe. Análisis de casos concretos – aspectos tipológicos – implicaciones teóricas. Language contact and lexico-grammatical change in the Caribbean. Analysis of particular cases – Typological aspects – Theoretical implications, Frankfurt am Main, Lang, 171–188. Hinzelin, Marc-Olivier/Kaiser, Georg (2012), “Le paramètre du sujet nul dans les variétés dialectales de l’occitan et du francoprovençal”, in: Mario Barra-Jover/Guylaiine Brun-Trigaud/Jean-Philippe Dalbera/Patric Sauzet/Tobias Scheer (edd.), Études de linguistique gallo-romane, Saint-Denis, Presses Universitaires de Vincennes, 247–260. Holmberg, Anders (2005), “Is there a little pro? Evidence from Finnish”, Linguistic Inquiry 36, 533– 564. Huang, C. T. James (1989), “Pro-drop in Chinese. A generalized control theory”, in: Osvaldo Jaeggli/ Ken Safir (edd.), The null subject parameter, Dordrecht, Kluwer, 185–214. Hulk, Aafke C./Pollock, Jean-Yves (2001), “Subject positions in Romance and the theory of universal grammar”, in: Aafke C. Hulk/Jean-Yves Pollock (edd.), Subject inversion in Romance and the theory of universal grammar, Oxford, Oxford University Press, 3–19.

Subjects, null subjects, and expletives

361

Jackendoff, Ray (1972), Semantic interpretation in generative grammar, Cambridge, MA, MIT Press. Jelinek, Eloise (1984), “Empty categories, case and configurationality”, Natural Language and Linguistic Theory 2, 39–76. Kaiser, Georg A. (2006), “Pronombres sujeto en construcciones impersonales de lenguas iberorrománicas”, in: Beatrix Fernández/Itziar Laka (edd.), Andolin gogoan. Essays in Honour of Professor Eguzkitza, Vitoria-Gasteiz, Universidad del País Vasco/Euskal Herriko Unibertsitatea, 513–530. Kandybowicz, Jason (2006), “Comp-trace effects explained away”, in: Donald Baumer/David Montero/Michael Scanlon (edd.), WCCFL 25. Proceedings of the 25th West Coast Conference on Formal Linguistics, Somerville, MA, Cascadilla Proceedings Project, 220–228. Kayne, Richard S. (1981), “EPP extensions”, Linguistic Inquiry 12, 93–133. Kayne, Richard S. (1983), “Connectedness”, Linguistic Inquiry 14, 223–249. Landau, Idan (2007), “EPP Extensions”, Linguistic Inquiry 38, 485–523. Ledgeway, Adam (2005), “Moving through the left periphery. The dual complementizer system in the dialects of Southern Italy”, Proceedings of the Philological Society 103, 229–396. Leonetti, Manuel (2008), “Definiteness effects and the role of the coda in existential constructions”, in: Alex Klinge/Henrik Høek Muller (edd.), Essays on nominal determination. From morphology to discourse management, Amsterdam/Philadelphia, Benjamins, 131–162. Leonetti, Manuel (2014), “Spanish VSX”, in: Karen Lahousse/Stefania Marzo (edd.), Romance languages and linguistic theory 2012, Amsterdam/Philadelphia, Benjamins, 37–62. López, Luis (2009), A derivational syntax for information structure, Oxford, Oxford University Press. Maiden, Martin/Parry, Mair (1997), The dialects of Italy, London/New York, Routledge. Manzini, Rita/Savoia, Leonardo (2002), “Parameters of subject inflection in Italian dialects”, in: Peter Svenonius (ed.), Subjects, expletives and the EPP, Oxford, Oxford University Press, 157– 200. Montalbetti, Mario (1986), “How pro is it?”, in: Osvaldo Jaeggli/Carmen Silva-Corvalán (edd.), Studies in Romance Linguistics, Dordrecht/Riverton, Foris, 137–152. Moro, Andrea (1997), The raising of predicates, Cambridge, Cambridge University Press. Neeleman, Ad/Szendrői, Kriszta (2007), “Radical pro-drop and the morphology of pronouns”, Linguistic Inquiry 38, 671–714. Nunes, Jairo (1999), “Linearization of chains and phonetic realization of chain links”, in: Samuel David Epstein/Norbert Hornstein (edd.), Working Minimalism, Cambridge, MA/London, MIT Press, 217–250. Nunes, Jairo (2004), Linearization of chains and sideward movement, Cambridge, MA, MIT Press. Perlmutter, David (1971), Deep and surface constraints in syntax, New York, Holt, Rinehart and Winston. Pesetsky, David/Torrego, Esther (2001), “T-to-C movement: causes and consequences”, in: Michael Kenstowicz (ed.), Ken Hale. A life in language, Cambridge, MA, MIT Press, 355–426. Pinto, Manuela (1994), “Subjects in Italian. Distribution and interpretation”, in: Reineke BokBennema/Crit Cremers (edd.), Linguistics in the Netherlands, Amsterdam/Philadelphia, Benjamins, 175–187. Pinto, Manuela (1997), Licensing and interpretation of inverted subjects in Italian, PhD dissertation, Utrecht University. Pollock, Jean-Yves (1989), “Verb movement, universal grammar and the structure of IP”, Linguistic Inquiry 20, 365–424. Rigau, Gemma (1988), “Strong pronouns”, Linguistic Inquiry 19, 503–511. Ritter, Elizabeth (1995), “On the syntactic category of pronouns and agreement”, Natural Language and Linguistic Theory 13, 405–443. Rizzi, Luigi (1982), Issues in Italian syntax, Dordrecht, Foris. Rizzi, Luigi (1986), “Null objects in Italian and the theory of pro”, Linguistic Inquiry 17, 501–557.

362

Michelle Sheehan

Rizzi, Luigi (1997), “The fine structure of the left periphery”, in: Liliane Haegeman (ed.), Elements of grammar, Dordrecht, Kluwer, 281–337. Roberts, Ian (2010a), “A deletion analysis of null subjects”, in: Theresa Biberauer/Anders Holmberg/ Ian Roberts/Michelle Sheehan, Parametric variation. Null subjects in Minimalist theory, Cambridge, Cambridge University Press, 58–87. Roberts, Ian (2010b), Agreement and head movement. Clitics, incorporation, and defective goals, Cambridge, MA/London, MIT Press. Roberts, Ian/Holmberg, Anders (2010), “Introduction. Parameters in Minimalist theory”, in: Theresa Biberauer/Anders Holmberg/Ian Roberts/Michelle Sheehan (edd.), Parametric variation. Null subjects in Minimalist theory, Cambridge, Cambridge University Press, 1–57. Roberts, Ian/Roussou, Anna (2001), “The EPP as a condition on the tense dependency”, in: Peter Svenonius (ed.), Subjects, expletives and the EPP, Oxford, Oxford University Press, 125–157. Rosselló, Joana (1986), Gramàtica, configuracions i referència. Per una teoria alternativa el PROdrop romànic, PhD dissertation, Universitat de Barcelona. Saab, Andrés (to appear), “On the notion of partial (non-) pro-drop in Romance”, in: Mary A. Kato/ Francisco Ordóñez (edd.), The morphosyntax of Portuguese and Spanish in Latin America, Oxford, Oxford University Press. Schifano, Norma (to appear), “Le lingue romanze. Verso una cartografia del movimento del verbo”, in: Éva Buchi/Jean-Paul Chauveau/Jean-Marie Pierrel (edd.), Actes du XXVIIe Congrès international de linguistique et de philologie romanes, Strasbourg, Société de linguistique romane/ ÉliPhi. Şener, Serkan/Takahashi, Daiko (2010), “Ellipsis of arguments in Japanese and Turkish”, Nanzan Linguistics 6, 79–99. Sheehan, Michelle (2006), The EPP and null subjects in Romance, PhD dissertation, Newcastle University. Sheehan, Michelle (2010), “‘Free’ inversion in Romance and the null subject parameter”, in: Theresa Biberauer/Anders Holmberg/Ian Roberts/Michelle Sheehan (edd.), Parametric variation. Null subjects in Minimalist theory, Cambridge, Cambridge University Press, 231–262. Solà, Jaume (1992), Agreement and subjects, PhD dissertation, Universitat Autònoma de Barcelona. Suñer, Margarita (2002), “The lexical preverbal subject in a Romance null subject language. Where art thou?”, in: Rafael Núñez-Cedeño/Luis López/Richard Cameron (edd.), A Romance perspective on language knowledge and use, Amsterdam/Philadelphia, Benjamins, 341–357. Takahashi, Daiko (2008), “Noun phrase ellipsis”, in: Shigeru Miyagawa/Mamoro Saito (edd.), The Oxford Handbook of Japanese Linguistics, New York, Oxford University Press, 394–422. Taraldsen, Knut (1980), “On the NIC, vacuous application, and the that-trace filter”, Paper presented at Indiana Linguistics Club. Tomioka, Satoshi (2003), “The semantics of Japanese null pronouns and its cross-linguistics implications”, in: Kerstin Schwabe/Susanne Winkler (edd.), The interfaces. Deriving and interpreting omitted structures, Amsterdam/Philadelphia, Benjamins, 321–339. Toribio, Almeida Jacqueline (1996), “Dialectal variation in the licensing of null referential and expletive subjects”, Paper presented at Aspects of Romance linguistics. Selected papers from the Linguistic Symposium on Romance Languages XXIV, Washington. Uriagereka, Juan (1995), “An F position in Western Romance”, in: Katalin É. Kiss (ed.), Discourse configurational languages, Oxford, Oxford University Press, 153–175. Vallduví, Enric (1993), “Catalan as VOS. Evidence from information packaging”, in: William Ashby/ Marianne Mithun (edd.), Linguistic perspectives on the Romance languages, Amsterdam/ Philadelphia, Benjamins, 335–350. Villa-García, Julio (2012), “Spanish subjects can be subjects. Acquisitional and empirical evidence”, Iberia 4:1, 124–169. Zubizarreta, Maria Luisa (1998), Prosody, focus, and word order, Cambridge, MA, MIT Press.

Susann Fischer and Maria Goldbach

12 Object clitics Abstract: This chapter provides an overview of the phonological, morphological, semantic, and syntactic properties of Catalan, French, Italian, Portuguese, and Spanish object clitic pronouns, before discussing some of the most prominent generative approaches that have been suggested in order to explain the distribution and behaviour of object clitics in Romance. Since this overview can of course not be exhaustive, the chapter seeks to present the main questions that have been asked and the main difficulties the approaches have been confronted with. The discussion focuses on the fact that various components of grammar – syntax, morphology, phonology, semantics – and the interfaces between these, are involved in describing and analysing Romance clitics. Keywords: base-generation, movement hypothesis, distributed morphology, PersonCase-Constraint, clitic climbing, clitic doubling, mesocliticization, interpolation

1 Introduction Clitics have always posed a great challenge for linguistic research. Even before Jacob Wackernagel published his extensive examination of the syntactic behaviour of certain word-like elements in ancient Indo-European languages in 1892, the phenomenon had been recognized (cf. Tobler 1875; Mussafia 1888). For more than a hundred years now clitics have been a source of controversy in linguistic theory. Especially within generative grammar, clitics have been a very prominent topic of research. The tremendous interest in this topic might be due to the fact that the status of clitics, as elements which exhibit properties of fully fledged words, but which lack the independence usually associated with words, is particularly challenging for modular theories of grammar, in that clitics clearly operate at multiple levels of representation. The main point of disagreement among the different analyses has been the extent to which the various components of grammar – syntax, morphology, phonology, semantics, and pragmatics – are involved, and the type of interaction that is required in determining the position, linear order, and interpretation of clitics. What seems to be clear after these many years of interest is the fact that clitics seem to be more easily defined by what they are not than by what they are, as a comprehensive theory of clitics is still not available. Given the vast amount of Acknowledgement: The first author would like to acknowledge the DFG grant FI 875/2-1 “Clitic doubling across Romance” for financial support of her research.

364

Susann Fischer and Maria Goldbach

literature on clitics (cf. e.g. the articles and all the references in van Riemsdijk 1999, as well as the bibliography for the period 1892–1992 in Nevis et al. 1994), what follows cannot be an exhaustive overview of clitics and cliticization; instead, we will restrict ourselves to object clitic pronouns1 in Romance. After a brief general overview of the classic criteria used to define clitics and the specific properties shared by Romance object clitics, we will discuss some synchronic and diachronic differences. This will be followed by a short overview of the most prominent (classic) theories and a few recent approaches that have been proposed in order to account for the specific behaviour of Romance clitics.

2 General definition The origin of the word clitic (derived from Greek κλίνειν ‘lean, incline’) reflects, first of all, a prosodic conception: In essence, the term describes forms that do not define a phonological word domain by themselves, but only in conjunction with the appropriate lexical material, i.e. a clitic must attach to another prosodic word in order to be pronounced. Typically clitics are function words, such as modal particles (e.g. interrogative particles), conjunctions, complementizers, determiners,2 or auxiliary verbs. The traditional observation that clitics always supply grammatical rather than lexical information has led to the assumption that any functional category, but no lexical category, can in principle be a clitic. In fact, it has been shown that there are no good examples of clitic main verbs, clitic nouns, or clitic adjectives (cf. Sadock 1991, 112; Spencer/Luís 2012). Historically, clitics generally develop from fully fledged words and frequently develop into inflectional affixes (cf. Givón 1979; Zwicky 1985; Wandruszka 1992). In Zwicky (1977) a first overview and an attempt to give a definition of different forms of these elements is carried out, and a classification into simple clitics, special clitics, and bound words is proposed. Zwicky (1985) gives up the notion of bound words3 and defines the difference between simple and special clitics in that simple clitics are mere reduced phonological forms whereas special clitics have a different form and a different syntax compared to the full form (cf. also Anderson 2005). 1 We will restrict ourselves to object clitics and not treat subject clitics in this chapter because of space limitations. Please refer to Rizzi (1986), Kaiser (1992), Sportiche (1999a), Poletto (1995; 1999), Goria (2004), Manzini (2015), among many others, for extensive discussions concerning subject clitics in Romance languages. 2 In the generative literature pronominals are treated as determiners (DPs or Ds). 3 In Zwicky (1977) bound words are words which do not correspond to a full form, but which nevertheless need a host and are in some cases restricted to a particular sentence position, such as Latin ‑que ‘and’. In Zwicky (1985) the bound words are more or less subsumed under the notion special clitics. Nevertheless, it seems that the notion of boundness is less clear-cut. Take for example the English article the: it cannot occur on its own, nor can it be considered a syntactic constituent without a following noun. Nevertheless this determiner is commonly not considered to be a clitic.

Object clitics

365

Under the broad notion of cliticization as mere reduction in phonological form most languages – very possibly, all except those of the most isolating type – have clitics. As for English, the contracted equivalents ’m, n’t, ’s etc. of the full forms him, not, has etc. can be seen as clitics. (1)

Eng.

a.

I can’t stand’m.4

c.

*I can’t m’stand

b.

I can’t stand him.

d.

*I can’t him stand

The form of the simple clitics (to use Zwicky’s term) seems to be dictated by phonology and can thus be affected by speech rate, level of formality, and the like (cf. Prinz 1991). The clitic in (1a) can be replaced by the full form as in (1b) and does not seem to have a special syntax compared to the full form (cf. 1c, d), i.e. the clitics preserve the linear order of their noncliticized counterparts. The bulk of the work on clitics however has targeted special clitics (cf. Nevis et al. 1994). In Zwicky’s (1977; 1985) view, special clitics are those that are not derived from full form equivalents by phonological reduction processes, and therefore, do not depend on factors such as speech rate. Special clitics act as variants of stressed full forms and show a special syntax: In the following Catalan (2), French (3), and Spanish (4) examples, the pronominal clitic cannot be replaced by a full form (d), and the full form cannot be replaced by a clitic (c), i.e. both the full form and the clitic occupy a different syntactic position (this holds for all Romance varieties except for Brazilian Portuguese). (2) Cat. a. La Mercè va veure en Joan. PST see the Joan the M. b. La Mercè el va veure. the M. him PST seen c. *La Mercè va veure-lo.5 d. *La Mercè en Joan va veure. (3) Fr. a. Zoé a vu Pierre. Z. has seen P. b. Zoé l’a vu. Z. him’has seen c. *Zoé a vu le. d. *Zoé Pierre a vu. 4 Throughout this chapter clitic elements are set in bold. 5 Since the 19th century Catalan clitics consistently change their form depending on whether they are in a pre- or postverbal position (Fischer 2002; 2006).

366

Susann Fischer and Maria Goldbach

(4) Sp. a. María ha visto a Juan. M. has seen to J. b. María lo ha visto. M. him has seen c. *María ha visto lo. d. *María a Juan ha visto. Looking at simple clitics, the interface between morphology and phonology seems to be affected, whereas with special clitics the interfaces between syntax, morphology, and phonology are addressed. Different authors have proposed different criteria to clarify the status of clitics. Kayne (1975), working on French pronominal clitics, is mainly concerned with establishing a coherent syntactic movement theory of clitics in which general constraints would do most of the work in explaining the major properties of the French clitics. Since the French (like the other Romance) clitics do not preserve the linear order of their full counterparts, he proposes criteria to distinguish clitics from full forms and full phrases, whereas Zwicky (1977) and Zwicky/Pullum (1983), working within the field of morphology and often working on polysynthetic languages, are mainly interested in distinguishing clitics from affixes.6 In contrast to the approaches which stress the differences between clitics and words and clitics and affixes, Klavans (1982), building on Zwicky (1977), considers their similarities and proposes that clitics could universally be accounted for by assuming that they are phrasal affixes, i.e. clitics operate on the phrase level, whereas affixes operate on the word level. In Klavans’ (1982; 1985) view, three different parameters are responsible for the distribution of clitics: (i) initial/final with respect to their domain, (ii) before/after the peripheral constituent of their domain, and (iii) left/right regarding attachments to their hosts. When all combinations of these three parameters are considered, the result is that the clitic can appear anywhere in a sentence except for initially or finally, i.e. eight different positions are possible according to which eight different clitic types are assigned. Empirical evidence for this theory has not been found in total, since some clitics seem to avoid some of the proposed positions whereas others appear in places where they are, according to Klavans (1982), unexpected. Looking at Romance clitics, Klavans’ proposal in its strict form needs to be dismissed: first, because Modern Romance clitics sometimes precede and sometimes follow their host verb, depending on whether the latter is finite or nonfinite, and second, regarding all Old Romance languages and Modern European Portuguese, clitics could/can precede and follow the finite verb. In Klavans’ 6 There are many more authors who have proposed criteria to define the notion clitic, but nowadays Kayne’s (1975) and Zwicky’s (1977) work is still considered the most influential.

Object clitics

367

taxonomy the enclitics and proclitics, although both pronominal in character, would be analysed as belonging to different clitic types, which is a serious drawback. Klavans (1985, 103) herself admits that the Romance type of verbal clitics pose a problem for her theory in its strict form. Although some authors have nourished the idea that Romance clitics are better analysed as agreement affixes (cf. section 4), Romance clitics nevertheless display the typical properties and behaviour that have been assigned to clitic elements in general.

3 Properties of Romance object clitics Romance clitics display peculiar behaviour regarding all grammatical modules: they display a deficient phonology, an anomalous syntax and morphology (Anderson 2005, 33), and a specific semantics.7 A survey of the different properties in Romance will be given in the following.

3.1 Phonological properties The phonological properties that are usually discussed with respect to clitics in general, i.e. the lack of stress and their prosodic status (cf. Selkirk 1996; Halpern 1995), are also important for Romance clitics. Clitic pronouns lack stress, while strong pronouns bear stress (cf. Kuchenbrandt 2009 for an extensive discussion on the prosodic properties of French and Spanish clitics). The following examples from French, Italian and Spanish form minimal pairs and clearly show that also the Romance clitics lack word stress. In Catalan and Portuguese the vowels in unstressed environment even need to be reduced. (5) Cat. a. te [tə] ‘you’

vs. te [ˈtɛ] ‘tea’

Fr.

b. la [la] ‘her’

vs. là [ˈla] ‘there’

It.

c. la [la] ‘her’

vs. là [ˈla] ‘there’

Por. d. se [sɯ] REFL vs. sé [ˈse] ‘cathedral’ Sp.

e. te [te] ‘you’

vs. té [ˈte] ‘tea’

Cardinaletti (2015, 598) goes one step further in arguing that clitic pronouns do not form a single phonological word with their host since the process of s-sonorization which is found word-internally in intervocalic contexts (as in It. re[z]istere (‘resist’), and It. ca[z]a ‘house’ does not take place between a proclitic and the verb (e.g. It. lo [so] / lo *[z]o) ‘I know it’ and neither with enclitics (e.g. It. mettendo[s]i / 7 Cf. also Kuchenbrandt (2009) for an overview of the phonological, morphological, syntactic, and semantic characteristics of French and Spanish clitics.

368

Susann Fischer and Maria Goldbach

*mettendo[z]i ‘put.GER . REFL ’). The process of s-sonorization in Castilian Spanish does apply in the context of a following voiced consonant as in mi[z]mo ‘same’ (cf. Hualde 2005) and in contact varieties of Spanish also in intervocalic contexts (e.g. Ecuadorian Spanish lo[z] otros ‘the others’) (Lipski 1989; Chappell 2011; Hualde/ Prieto 2014). As attested by the examples above s-sonorization applies across wordboundaries in Spanish which also holds true for Catalan (e.g. do[z] ami[ɡz] íntims ‘two close friends’) and French (e.g. le[z] hommes ‘the men’). Thus this test seems to work for Italian clitics but not for Spanish, Catalan or French clitics.

3.2 Morphological properties The morphological properties of clitics that are most often discussed are their reduced morphological form compared to the full forms, the opaque forms, the person-caserestriction when appearing in a cluster, and their phi-8 and case features. Clitics are monosyllabic words, while strong pronouns can be bi- or even trisyllabic. Since, for instance, in French clitic pronouns are homophonous with strong pronouns (nous ‘us’, vous ‘you.PL ’), the generalization has been formulated that clitic pronouns are equal to or smaller than their strong counterparts: clitic ≤ strong (cf. Cardinaletti/Starke 1999, 174). This holds true for Romance object clitics. (6) Cat. a. el/lo ‘him’ vs. ell ‘he’ / a ell ‘to him’ Fr.

b. le ‘him’

vs. il ‘he’ / à lui ‘to him’

It.

c. lo ‘him’

vs. egli, lui ‘he’ / di lui ‘of him’

Por. d. o ‘him’ Sp.

e. lo ‘him’

vs. ele ‘he’ / a ele ‘to him’ vs. él ‘he’ / de él ‘of him’

In Romance and other languages combinations of clitic pronouns, so-called clitic clusters, often trigger opaque forms which have been shown to always coincide with clitics that independently coexist in the language (Perlmutter 1970; 1971; Bonet 1995; Gerlach 2002; Cardinaletti 2008). See the following examples (7), in which in a combination of a 3sg accusative and 3sg dative clitic, the dative is substituted by the locative hi (a), the pronoun ci (b), and the reflexive se (c): (7) Cat. a. lo ACC .3SG

It.

Sp.

+ li DAT.3SG

b. si 3SG

+ si

c. le

+ lo

DAT.3SG

REFL

ACC .3SG

→ l’hi ACC .3SG ’LOC

→ ci si LOC . REFL

→ se lo REFL . ACC . 3SG

8 Phi-features in generative syntax are the grammatical features of person, number, and gender.

Object clitics

369

Furthermore, both types of the person-case-constraint (PPC) can be attested (8), which sometimes leads to a mismatch between overt morphological marking and the syntactic-semantic interpretation. (8) PCC a. Strong Version In a combination of a weak direct object and an indirect object [of clitics, agreement markers, or weak pronouns] the direct object has to be third person. b.

Weak Version In a combination of a weak direct object and an indirect object, if there is a third person it has to be the direct object. (Bonet 2008, 104)

French, for example, shows the strong version (9) and Spanish (10), Italian (11), and Catalan (12) show the weak version of the PCC. (9) Fr. a. *Roger te nous /nous t’avait recommandés. DAT.2SG ACC .1PL /ACC .1PL DAT.2SG ’had recommended R. ‘Roger had recommended us to you.’ b. Roger nous avait recommandés à toi. ACC .1PL had recommended to DAT.2SG R. ‘Roger had recommended us to you.’ Example (9a) shows that combinations of first and second person clitics are judged ungrammatical. The only way to utter this sentence is to use a strong pronoun (9b) instead of the clitic (cf. Perlmutter 1970, 222). Spanish, Italian, and Catalan allow combinations of first and second persons, but whenever a third person indirect clitic is included the sentence is ungrammatical. In order to overcome this constraint Spanish (10) and Italian (11) use a strong pronoun, whereas Catalan uses the locative clitic hi (12) (for more examples and a discussion of other types of agreement restrictions, cf. Fischer 2011; and ↗9 Agreement restrictions and agreement oddities). (10) Sp. a. Te

me

presentaste. introduced ‘You introduced yourself to me.’

ACC .2SG

b. *Me

DAT.1SG

presentaron. ACC .1SG DAT.3SG introduced

c. Me

le

presentaron a él. introduced to him ‘They introduced me to him.’

ACC .1SG

370

Susann Fischer and Maria Goldbach

(11) It. a.

Mi

ti

ACC .1SG

DAT.2SG

presantarono. introduced ‘They introduced me to you.’ gli

b. *Mi ACC .1SG

DAT.3SG

presentarono. introduced

c. Mi

presentarono a lui. ACC .1SG introduced to him ‘They introduced me to him.’

(12) Cat. a. Te

m’ha recomanat en Miquel. ACC .2SG DAT.1SG has recommended the M. ‘Miquel has recommended me to you.’

b. *Me

li

ACC .1SG

DAT.3SG

ha recomanat en Miquel. has recommended the M.

c. M’hi ha recomanat en Miquel. ACC .1SG ’LOC 9 has recommended the M. ‘Miquel has recommended me to him.’ It is a well-known fact that full noun phrases in Romance do not display morphological case; clitic pronouns however exhibit case distinction, namely accusative (e.g. Italian/Spanish lo), dative (e.g. Italian/Spanish gli/le), partitive (e.g. Italian/ Catalan ne/en), and with a locative interpretation (e.g. Italian/Catalan/French ci/hi/y). Concerning grammatical features, it has been shown that Romance clitics encode different phi-features (cf. Bonet 1995; Fischer 2002; Cardinaletti 2008). Consider the following feature matrix for Spanish (Table 1). Table 1: The Spanish feature matrix (cf. Halle/Marantz 1994, 280) 3p

ACC

DAT

2p

M

F

SG

lo

la

te

me

PL

los

las

os

nos

SG

le

PL

les

M

1p F

M

F

same as ACC REFL

SG

se PL

9 Rigau (1982) has suggested that the clitic hi can be used as a locative clitic, but also as an inanimate dative. For a similar approach, cf. Hualde (1992), who gives a brief description of hi and other clitics which can have different uses.

Object clitics

371

What can be seen in Table 1 is that 3p is represented by , ACC . M by , ACC . F by , dative by , and plural by , deriving the forms lo, la, los, las for the different accusative clitics and le, les for the dative clitics.

3.3 Syntactic properties Syntactically, clitics have often been argued to be deficient in that they cannot be modified, coordinated, contrastively stressed, or occur in isolation, and in that they cannot appear in the same position as their strong equivalents. Romance object clitics show all of these properties: they obligatorily appear in a position next to the verb, a position which is not available to full DPs or full pronouns (recall example 2). They are not allowed to appear in isolation (13), which includes that they cannot appear at a peripheral position in a sentence where they are separated by a prosodic break (14). (13) Cat. a. Qui who

coneix knows

la the

Núria? N.

Ella/*La. she/ her

Fr.

b. Qui connait Zoé ?/Zoé, elle connait qui ? Lui/*La.

It.

c. Chi conosce, Fiammetta?

Lui/*Lo.

Por. d. Quem é que a Joana conhece?

Ele/*O.

Sp.

A él/*A lo.

e. ¿A quién conoce María?

(14) Cat. a. A ell/*El, la Núria el coneix. Fr.

b. Lui/*La, Zoé la connait.

It.

c. Lui/*Lo, Fiammetta lo conosce.

Por. d. É ele/*O que a Joana conhece. Sp.

e. A él/*A lo, María lo conoce.

Furthermore, Romance clitics can neither be modified (15), nor conjoined (16), nor contrastively focused (17): (15) Cat. a. La Núria la coneix solament a ella/*la. Fr.

b. Zoé ne connait qu’elle/*que la.

It.

c. Fiammetta conosce solo lei/*la.

Por. d. A Joana conhece só ele /*o. Sp.

e. María la conoce solamente a ella/*la.

372

Susann Fischer and Maria Goldbach

(16) Cat. a. La Núria els coneix a ella i a ell/*La Núria [la i lo] coneix. Fr.

b. Zoé connait [il et elle]/*Zoé [le et la] connait.

It.

c. Fiammetta conosce [lui e lei]/*Fiammetta [lo e la] conosce.

Por. d. A Joana conhece ele e ela /*A Joana conhece [o e a]. Sp.

e. María los conoce a [ella y a él]/*María [lo y la] conoce.

(17) Cat. a. La Núria el coneix a ELL, no a ella/*La Núria EL coneix, no a ella. Fr.

b. Zoé connait LUI, y pas elle/*Zoé LE connait, et pas elle.

It.

c. Fiammetta conosce LUI, non lei/*F. LO conosce, non lei.

Por. d. A Joana conhece ELE, mas não ela /*A Joana O conhece, não ela. Sp.

e. María lo conoce a ÉL, no a ella/*María LO conoce, no a ella.

3.4 Semantic properties The semantics of clitics have often been called upon in order to motivate the high position in the phrase-structure, i.e. the position in the front of the sentence. As for the motivation for this position, an idea has been proposed that links clitics to specificity/referentiality (Martins 1994; Uriagereka 1995; Fischer 2002; Roberts 2010, among many others). As referential/specific elements, they refer to something already mentioned in the discourse (18). It has been proposed that TP (the place where the finite verb is located in Romance) is the border for specificity, which can be deduced if it is accepted that all predicates have a Davidsonian event argument, and that this is also true of nouns, even noneventive ones (Higginbotham 1987). Specificity is just an element taking wide scope with respect to the Davidsonian event operator in the sentence; in other terms, a DP is specific if and only if its event variable is not bound by the event operator, being instead bound by the discourse (Herburger 1994). Thus clitics must move out of the vP/VP (cf. section 4). (18) Cat. a. Speaker A: Jo conec en Joan. I know the J. Speaker B: Sí, jo el conec també. yes I him know too Fr.

b. Speaker A: Moi, je connais Pierre. Speaker B: Oui, je le connais aussi.

It.

c. Speaker A: Io conosco Fiammetta. Speaker B: Anch’io la conosco.

Por. d. Speaker A: Eu conheço a Joana. Speaker B: Eu também a conheço. Sp.

e. Speaker A: Yo (lo) conozco a Jorge. Speaker B: Sí, yo lo conozco también.

Object clitics

373

Another fact worth mentioning is that clitics can refer to both human and nonhuman entities, whereas strong pronouns can only refer to human entities (19). Ethical datives can only be represented by clitics and never by full pronouns (20). (19) Fr.

a. [Le nouveau livre de Éric Laurent]i il me plait beaucoup de lei lire. the new book of E. L. it me pleases a lot it read ‘It pleases me a lot to read the new book of Éric Laurent.’ b. *[Le livre de Éric Laurent] i . . . je ne veux pas parler de luii the book of E. L. I not want not speak of it

Sp. a. [El nuevo libro de Carlos Ruiz Zafón] i . . . me gustaría leerloi b. *[El libro de Carlos Ruiz Zafón] . . . no quiero hablar de él. (20) Cat. a. Aquell nen no em menja res. that child not me eat nothing Lit. ‘This child doesn’t eat anything for me.’ It.

b. Mi è nato un bambino. me is born a child Bible English. ‘Unto me was born a child’

What has become evident in this discussion of the different properties is the fact that the individual properties apply on more than one level, for example the PCC seems to be a morphological and a syntactic effect, and the same holds for contrastive focus, which is plainly phonological but also syntactic. This clearly shows that the interfaces at the different modules of grammar are important for the analysis of clitics.

3.5 Clitic climbing, clitic doubling, mesocliticization, and interpolation In addition to the properties that are shared by the clitics of the different modern Romance languages, we also find language-specific characteristics: for instance, clitic climbing is attested in Spanish, Italian, European Portuguese, Romanian, and Catalan, but not in French. Clitic doubling is attested for Portuguese, Spanish, Romanian, and Catalan, but not for Standard Italian and Standard French.10 Interpolation, postverbal clitics with finite verbs, and mesocliticization are found in European Portuguese but not in any of the other modern Romance languages. 10 French and Italian allow doubling of quantifiers and full pronouns. However, indirect objects cannot be doubled in the standard written varieties (Fischer/Rinke 2013).

374

Susann Fischer and Maria Goldbach

However, starting out, the Romance languages were not as diverse as they are now (cf. Wanner 1987; Kuchenbrandt 2009; Hinzelin 2007); instead the O(ld) Romance languages shared many of the abovementioned characteristics and they all displayed postverbal clitics with finite verbs. Cf. the examples in (21), all taken from Fischer (2002, 35–36). (21) OCat. a. e donà-la per muller a l’emperador de Castela and gave-her as wife to the emperor of Castille OFr.

b. et demande li and ask him

OIt.

c. offerse-gliene due marchi di guadagno offered-him-of.it two marks of interest

OPor. d. perguntou-lhe e disse-lhe asked-him and told-him OSp.

e. e fizo-lo traer preso and made-him bring prisoners

The fact that Old Romance clitics appear following the finite verb has been argued to be an effect of the Tobler-Mussafia law, which states that unstressed object pronouns cannot stand in absolute initial position, thus due to a phonological constraint. However, Fischer (2002) and Hinzelin (2007) have found object clitics in embedded sentences, which shows that there must be an additional factor allowing/forcing clitics in a postverbal position (22), especially since sentences can be attested which show exactly the same words but in a different linear order (23). Cf. the following examples from Old Catalan. (22) OCat. . . . qui per justícia seguex-se la fi per què . . . who for justice follows-ref the purpose for that ‘. . . who – out of justice – follows the purpose for which. . .’ (Fischer 2002, 177) (23) OCat. a. I. jorn se sdevench que lo ermità splugave one day ref happened that the hermit scrutinized son cilici, his penitential robe, ‘And it came to pass that the hermit scrutinized his penitential robe,’

Object clitics

b. I. one

375

jorn sdevench-se que I. juheu vench a aquell day happened-ref that a Jew came to that

sant hom, holy man, ‘And it came to pass that a Jew came to that holy man,’ (Fischer 2002, 178) In modern Romance, postverbal clitics with finite verbs are still allowed in European Portuguese (24), however never in embedded sentences, in contrast to interpolation (25a) (where an item, in (25a) the negation não, can split the clitic-verb unit) and mesocliticization (25b).11 (24) Por. a. Ele conhece-me. he knows-me ‘He knows me.’ (25) Por. a. O João pediu que o não acordassem. the J. asked that him not should wake up ‘João asked them not to wake him up.’ (Mateus et al. 2003, 866) b. Se me fizesse essa pergunta, recusar-me-ia a responder. if me made this question, refuse-me-would to answer ‘If s/he asked this question, I would refuse to answer.’ (Mateus et al. 2003, 865) In (25b) me is the ethical dative object clitic occurring before the conditional inflectional ending , thus the clitic is located between the verb and the verbal inflection, which is called mesocliticization. Interpolation and mesocliticization are also attested in Old Romance. Cf. the following examples of interpolation, the separation of the clitic-verb sequence, in Old Portuguese (26), Old French (27), and Old Spanish (28) (cf. Fischer 2002, 40–42). (26) OPor. E sse and if

pela uentujra uos alguen enbargar by chance you someone blocks

a dita vya . . . the said vineyard ‘And if by chance someone blocks the vineyard from you. . .’ (Antt, Chelas, 1294) 11 All examples from the Old Romance languages have been assembled by Fischer (2002) and are cited here accordingly; however some of these examples have been taken from different authors (Kaiser 1992; Martins 1994; Wanner 1987 etc.).

376

Susann Fischer and Maria Goldbach

(27) OFr. Ke il te plus face aorer . . . that he you more make ? ‘That he lets you ? . . .’ (La vie de Saint Eustache, Paris, 1928) (28) OSp. pero que lo non fallamos en toda la estoria . . . but that it not find in all the story ‘but we do not find it in the whole history . . .’ (Alfonso X, Estoria de España II/11) Interpolation has not been attested in Old Italian or in Old Catalan (Fischer 2002), however Old Catalan (30) shows mesocliticization exactly like Old Portuguese (29) and Old Spanish (31) (Fischer 2002, 44-48). (29) OPor. e vós sabè-lo-edes . . . and you know-it-will ‘and you will know it . . .’

(Cantigas d’Escarnho, 66,34)

(30) OCat. E seguir-vos he ab .LXXX. cavalers, and follow-you have with LXXX cavaliers, ‘I will follow you with eighty cavaliers,’ (31) OSp. dezir lo hedes al rey? tell it will to.the king ‘will you tell it to the king?’

(Desclot 77,13)

(Libro del caballero Zifar 124)

A further peculiarity of the Romance languages concerns clitic climbing. Most scholars working on clitic climbing see it as a sort of optional movement which may, but need not, apply (cf. Rizzi 1982; Roberts 1997; and Rooryck 2000 for an overview). As for the modern Romance languages which display this phenomenon, the optionality of the clitic to climb to the verb or stay with the infinitival verb is considered to have already been available in the Old Romance languages: “patterns such as Yo quiero comerlo and Yo lo quiero comer ‘I want to eat it’ are attested from the earliest documents” (Rivero 1991, 241). The modern Romance languages displaying clitic climbing with causative as well as with modal verbs are Spanish, Italian, Portuguese, and Catalan. In contrast to other Romance languages, modern French only admits clitic climbing with causative verbs; with modal verbs the clitic has to remain with the infinitive. In the Old Romance languages clitic climbing is also attested for French. However, clitic climbing was not optional in the medieval languages; on the contrary it was obligatory (cf. Pearce 1990; Fischer 2002). Cf. the Old French example with a modal verb (32) and the Old Catalan examples with a causative (33a) and a modal verb (33b). In all instances of a causative or modal verb in Old French and Old Catalan, the clitic had to climb to the finite verb (cf. Fischer 2002, 42–43).

Object clitics

(32) OFr. Mes ele ne la pot veoir . . . but she not her can see ‘But she cannot see her . . .’

377

(La Chastelaine de Vergi, 729)

(33) OCat. a. . . . e sia plaer de Déu que.ls vos fa comensar, and is pleasure of God that.they you make begin ‘. . . and it is God’s pleasure that makes you begin,’ (Llull, 79,10) b. Fort ho volria saber strong it want know ‘Urgently I would like to know it’

(Metge, 73,2)

The last characteristic we need to look at before discussing the proposed analyses is clitic doubling. Clitic doubling is understood as a construction in which a clitic co-occurs with a full DP in argument position and shares with it one syntactic and one semantic function. It is different from left and right dislocation, since dislocation necessarily involves a prosodic break whereas this is not true for doubling structures (Jaeggli 1986; Gabriel/Rinke 2010). This is illustrated in (34), an example from Spanish, where le and a Juan denote the same referent without a prosodic break being involved. (34) Sp. Le dimos el libro a Juan. him.DAT gave.1PL the book to Juan ‘We gave the book to Juan.’ Clitic doubling is obligatory with full pronouns and optional with indirect objects in Catalan, Portuguese, Romanian (35), and Spanish (34) (Fernández Soriano 1999; Iliescu/Popovici 2013; Fischer/Rinke 2013). In the Old Romance languages clitic doubling was not yet a coherent characteristic (Gabriel/Rinke 2010). It did appear with full pronouns (36), but even with full pronouns it was still optional (37) (cf. Fischer 2002, 44–45). (35) Cat.

Por.

a. Li dono el llibre a en Joan. him give the book to the J. b. Dei-lhe o livro à Maria. gave-her the book to M.

Rom. c. I-am dat o carte Mariei. her-have given a book to M. (36) OCat. Prec-vos que m’ojats tots a mi un poc. ask-you that me’listen all to me a bit ‘I ask that you all listen to me a while.’

(Metge, 18,5)

378

Susann Fischer and Maria Goldbach

(37) OCat. a. . . .e tan amarg és a mi que . . . and so bitter is to me that ‘. . . and it is so bitter for me that . . .’ OSp.

(Martorell, 31,26)

salir. b. al logar onde dios mando ami to.the place where god ordered to.me exit ‘to the place where God had ordered me to get out.’ (General Estoria Part I.65r)

The characteristics presented in this section have often been used as arguments in favour of or against various theoretical approaches to cliticization in Romance. Thus in the next section a brief overview of some prominent generative approaches is given.

4 Cliticization in generative grammar The questions that have not only received different answers in the various versions of the generative model (Government-Binding (GB), cf. Chomsky 1981; Minimalist Program, cf. Chomsky 1995; 2001 etc.),12 but also within the same version of the model, address the status of clitics as heads or phrases, the relationship between clitics and the corresponding (canonical) nonclitic position, and what functional category hosts the clitic element.

4.1 The movement hypothesis The recognition of two morphologically and syntactically distinct series of Romance pronouns, i.e. clitic and strong pronouns, has influenced all subsequent research on pronouns. Kayne’s (1975) fundamental contribution to the theory of clitics in generative grammar consists in the proposal that cliticization is a movement rule. One simple motivation for his analysis is the apparent complementary distribution between clitics and their associated full DPs. Kayne (1975) proposes that Romance clitics are base-generated in the object position of the verb, where full DPs are also generated, and are subsequently moved to a position left-adjoined to the verb (38).

12 It should be mentioned that contributions to the theory of clitics are found in most of the frameworks of generative-inspired formal grammar, including Lexical Functional Grammar (LFG) (for example Grimshaw 1982), Head-driven Phrase Structure Grammar (HPSG) (for example Monachesi 1999), and of course Optimality Theory (OT) (cf. Gerlach 2002; Grimshaw 2001). However, in this chapter we will concentrate on the transformational theorizing, where most of the leading questions have been asked.

Object clitics

(38)

Cat. La Mercè [v eli [ the Mercè him

v

379

veu ] ] [NP ti] sees

One type of evidence Kayne (1975) uses in order to corroborate his conclusions is the blocking effects of intervening subjects on (some type of) clitic placement. A sentence like (39) is ungrammatical in that it implies movement of the dative clitic leur from its base-generated complement position across the (inverted) embedded subject Lucille, thus suggesting a Specified Subject Condition effect. (39) Fr. Le directeur *leur a fait répondre Lucille. the director them.DAT has made answer.INF L. ‘The director made Lucille answer *them.’ (Kayne 1975, 301) More direct evidence in favour of a movement analysis comes from the locality effects displayed by clitic placement, which are typical of movement operations: for example extraction from a PP or out of an adverbial prepositional phrase, or extraction out of a noun phrase whose highest specifier is filled with a demonstrative. Such data strongly suggest that movement is involved. The same holds for stranding under clitic placement (cf. Cardinaletti/Starke 1999 for an extensive discussion on this matter) (40–41): (40) Cat. a. Has vingut amb la Maria? have.2SG come with the Maria ‘Did you come with Maria?’ b. *Qui has vingut amb? who have.2SG come with (41) Cat. a. En Joan ha vingut amb ella. the Joan have.3SG come with her ‘Joan has come with her.’ b. *En Joan li ha vingut amb. the Joan her have.3SG come with A further argument in favour of such an analysis has always been seen in the fact that some Romance languages trigger past-participle agreement; for instance, Catalan, Italian, and French participles may (or must, depending on the variety) agree with their accusative direct object when they precede the participle (42). (42) Cat. a. La Mercè ha vist les nenes. the Mercè has.3SG seen the girls ‘Mercè has seen the girls.’

380

Susann Fischer and Maria Goldbach

b. La Mercè les ha vistes. the Mercè them.F. PL has.3SG seen.F. PL ‘La Mercè has seen them.’ Under a movement analysis these data are easily accounted for if, as Kayne (1989) suggests, there is an intermediate specifier (of the participial morphology) through which the moved object may or must move. Furthermore, clitic climbing is another set of data that clearly shows that whatever hypothesis one prefers in order to explain the Romance object clitic, at least some constraints on their placement are to be interpreted in terms of conditions on movement. The beginning of the movement hypothesis can be seen as an attempt to deny that clitics have any special properties other than their obvious phonological ones and to account for their distribution by independently necessary syntactic principles on movement. Within GB theory, clitics were regarded as categories. Depending on the analysis adopted they were sometimes considered heads (X°), e.g. Kayne (1975; 1991), or phrases (XPs), e.g. Platzack (1995), which were displaced. Consequently, a trace was no longer a copy but a category with properties which are potentially distinct from those of the moved element. Chomsky (1993) proposes that movement is not displacement but copying of a category into another position, with subsequent deletion of one of the copies in PF. Chomsky (1995) outlines a theory where movement applies to features instead of whole categories. On moving to the checking domain of F, F’ may have to pied-pipe minimally a word, sometimes a phrase. Feature movement theory can be combined with the copy theory (at least this seems to be what Chomsky 1995 has in mind), i.e. chain reduction deletes copies creating the appearance of movement. Movement is triggered then by a formal feature F of a functional category which needs checking, and therefore attracts a feature F’ to the checking domain of F. F’ together with the features that have been pied-piped are copied into the checking domain of F. The overt vs. covert movement (LF-movement) distinction is a matter of whether the features are spelled out upstairs or downstairs. Strong features trigger overt movement whereas weak features only trigger movement after spell-out, i.e. covert at LF. With respect to clitic movement, this means that in order for the clitic to move, it has to be attracted by a formal feature on one of the functional categories which needs to be checked. In later approaches in the Minimalist Program, a difference between interpretable/uninterpretable and valued/unvalued features was proposed (Chomsky 2000 and subsequent work). In Roberts (2010) clitics are viewed as minimal/maximal categories that move by head-movement and as such have a semantic effect (cf. Roberts 2010; Lechner 2005). Clitics enter an Agree relation to value their probe’s features. This Agree relation however is seen as incorporation by which the clitic’s features are copied onto the probe. The most important aspect of the copying of the features of the clitic is that valuing the features of the probe exhausts the content of

Object clitics

381

the goal. Therefore the operation is not distinguishable from the copying involved in movement. According to Roberts (2010, 60) in the case of incorporation Agree and Move/Internal Merge are formally indistinguishable. Since clitics have interpretable phi-features, they can value the uninterpretable phi-features of v*. The idea that v*min is a phase then explains why once all its features have been valued (and deleted, where necessary), it is sent to the interfaces as a unit (v*min+Dmin/max). At every stage of the generative framework a great deal of work has been devoted to determining the concrete landing site of the clitics. Most of the work has converged on the idea that this is a functional category at the left periphery of the clause between C and T. Although the various proposals offer different explanations of what the exact functional category is, there is a general consensus that this category serves as an interface between syntax and discourse (cf. Raposo/Uriagereka 2008). Thus, whatever triggers clitic movement seems to be a cluster of syntactic and semantic constraints.

4.2 The base-generation hypothesis The attempt to capture the generalization that clitics and DPs display complementary distribution is directly contradicted by the phenomenon of clitic doubling (recall examples (34) and (35)). In early GB studies clitic doubling constructions were seen as the major argument in favour of a base-generation analysis of clitics. The phenomenon of clitic doubling was first discussed by Strozer (1976) and Rivas (1977), who point out that in some constructions in Spanish in contrast to French and Italian clitic objects co-occur with nonclitic objects; for example full pronouns are always doubled and indirect objects may be doubled by a clitic pronoun in Spanish as well as in the Catalan sentences above. They judge the data to be direct proof against a movement analysis and propose instead that the clitics in question are base-generated in their surface position. According to such analyses, the pronominal clitic in the VP is generated to the left of the verb, while the coindexed DP is generated in the object position, as required by the selecting verb. This idea has been further developed in Jaeggli (1982; 1986) and Borer (1984), who give a more detailed proposal in that the clitic is generated in a special position, neither A nor A’ (next to the verb), and that the A-position related to the clitic is occupied either by the empty pronominal pro or by a lexical noun. (43)

382

Susann Fischer and Maria Goldbach

The presence of the complex Cl-V under V reflects the affix-like status of the clitic in this kind of analysis. Borer (1984) considers clitics spell-out features of the V projecting the VP; for Jaeggli (1986) they are syntactic affixes forming a V with the V they attach to. The theoretically interesting question is how it can be that the clitic and the full NP share one theta-role and one Case between them. One possible explanation is given by Jaeggli (1982) and Borer (1984), who assume that the complement DP position is an argument position, which is theta-marked by the verb. In this analysis the presence of a full DP depends on whether the clitic absorbs the Case which the verb assigns. Jaeggli (1982) suggests parameterizing the ability of the clitic to absorb Case and proposes that in French and Italian the clitic obligatorily absorbs the accusative or dative Case of the verb and an DP cannot appear, whereas in Spanish the clitic absorbs Case only optionally, and leaves open the possibility of having a lexical DP that would receive Case from the verb, in exactly the same way as if the clitic were not present. Jaeggli (1986) suggests an analysis in which the preposition, which is obligatory in Spanish clitic doubling constructions, doesn’t itself assign Case, but rather transmits the Case assigned by the verb, thus combining Kayne’s generalization that in clitic doubling constructions a Case marker is present on the doubled DP with Borer’s proposal (1984) that the clitic absorbs the spell-out or morphological realization of the verb’s Case feature. Jaeggli (1986) assumes a process of Case matching rather than Case assignment, under which the Case assigner, the verb, is matched with either a recipient or a transmitter of Case (here the preposition a). However, Uriagereka (1995), Suñer (1988; 2006), and Fischer/Rinke (2013) show that in quite a lot of doubling constructions no preposition is present. Thus, what seems to be the problem if we adopt the movement hypothesis for the Romance languages is the fact that in, for example, Spanish and Catalan, indirect object DPs can be doubled by a clitic, and that in Romanian and some varieties of South American Spanish indirect as well as direct object clitics are doubled. There seems to be no easy solution: if we adopt the base-generation approach, we cannot explain why the relation between the clitic and the corresponding full phrase position is subject to locality constraints in much the same way as other movement processes are; on the other hand, if we adopt the movement hypothesis, we are able to explain the locality constraints, but get into trouble because of the clitic doubling phenomenon, which is a phenomenon that applies at the interface of syntax and semantics (cf. Fischer/Navarro/Vega 2016).

4.3 Reconciling the approaches while accounting for the diverse behaviour Developments in syntactic theory, such as the proliferation of functional projections (Pollock 1989; Poletto 1999; Cardinaletti 2015), the typology of A- and A’-positions, or the approach of movement as triggered by morphosyntactic features, have led to

Object clitics

383

a reconciliation of the base-generation and movement approaches to cliticization. Several scholars have proposed a combined base-generation and movement approach to clitics (cf. Sportiche 1999b; Uriagereka 1995). For instance, Sportiche (1999b) argues that clitics are heads in the extended projection of V while the doubled DP is generated in a VP-internal argument position and undergoes movement to or through the clitic position at some stage in the derivation. Unlike Sportiche (1999b) however, Uriagereka (1995) claims that clitics are determiners, i.e. heads in the extended projection of N, while doubled DPs are in the specifier position of a complex DP headed by the doubling clitic, which undergoes movement to its surface site. This implies that the nature of the determiners in a language determines whether this language can or cannot have clitic doubling, i.e. clitic doubling is present in a language if the determiners can be the head of a complex DP. Camacho Taboada (2006) suggests that in languages which use clitic doubling, clitics are base-generated in a functional projection whereas in languages without doubling clitics are D-heads which are moved to a functional position. In all the approaches mentioned here, clitics have the same syntax across languages. A different approach rejects a homogeneous account of clitics in languages. Gabriel/Müller (2005) suggest that 1st and 2nd person and nonreflexive forms of 3rd person clitics are generated in and subsequently moved from the complement position, whereas reflexive se is base-generated with the verb. Bleam (1999) and Anagnostopoulou (2003) regard the clitic’s syntax as being defragmented. Bleam (1999) argues that accusative clitics are determiners and dative clitics are inflections, which is taken up by Déchaine/Wiltschko (2002), who propose that doubling clitics come in two guises: D-clitics and phi-clitics. In this approach D-clitics trigger semantic effects while phi-clitics are purely syntactic markers. Anagnostopoulou (2003, 14) then relates the phi-features to double object constructions, where they are seen as matching/doubling the features of the objects. In this approach clitic doubling is then accounted for via overt feature movement with a PF reflex (Anagnostopoulou 2003, 214). In current research, the operation long distance AGREE is discussed regarding cliticization and/or clitic doubling (Preminger 2008). In these discussions cliticization is regarded as a form of agreement that mirrors subject-verb agreement, implying that clitics are a purely syntactic phenomenon and that the semantic factors that make, for instance, doubling possible are of minor importance and basically depend on the doubled DP and not on the clitic.

4.4 The distributed morphology approach The ideas of distributed morphology (DM) arose in the early nineties of the last century, mainly inspired by Bonet (1991) and more concisely published in Natural Language and Linguistic Theory as Bonet (1995). The starting point of DM-considerations

384

Susann Fischer and Maria Goldbach

was Romance object clitics and especially object clitic sequences (opaque clitic clusters such as in section 3.2, example (7)). But soon afterwards these concepts were adapted to inflectional morphology (cf. Halle/Marantz 1994; Noyer 1997; Bobaljik 2000, and more recently Halle/Marantz 2008; ↗5 Inflectional verb morphology). According to, for example, Embick/Noyer (2001), syntactic structure is derived without phonological material; rather it is built up exclusively on the basis of abstract morphosyntactic features, and the morphological processes are distributed over the syntactic structure. In other words, syntax computes the terminal nodes by Merge and Move and only after this is finished are phonological features added. Within DM, terminal nodes are thus complexes of semantic and syntactic features that systematically lack all phonological features. The phonological features are supplied on a postsyntactic level, namely morphological structure, by the insertion of vocabulary items into the terminal nodes. The procedure of vocabulary insertion operates according to underspecification and the elsewhere-condition (cf. Kiparsky 1982). This means that in the component of the lexicon, the morphosyntactic features of the vocabulary items are stored not fully specified with respect to their insertion conditions. Let us illustrate this concept on the basis of the Spanish clitic combination se lo, resulting from dative singular le preceding accusative singular masculine lo. According to Bonet (1995), the vocabulary insertion conditions for these clitic object pronouns are the following (44).13 (44) le ↔ {clitic {argument, third person {oblique}}} lo ↔ {clitic {argument, third person}} se ↔ {clitic {argument}} It is evident that se is the least specified of the three object clitics in (44) while le is the most specified. The double-headed arrow ‘↔’ in the lexical insertion conditions means: The item can only be inserted into a terminal node in the context of the syntactic feature structure specification on the right side of the arrow. Thus, the clitic le can only be inserted when the terminal node of the syntactic structure is specified for {clitic {argument, third person {oblique}}}. According to the elsewhere-condition no other clitic can be inserted, since the most specific item must always be inserted before a less specific item comes into play. But then, the question remains as to what happens to the syntactic structure in examples such as (45) (compare Bonet 1995)? (45) Sp. A Pedro, el premio, se lo (*le lo, *lo le) dieron ayer. to P. the prize him it gave.3PL yesterday

13 Bonet (1991; 1995) uses a tree structure, but the bracket illustration in (44) is equivalent to Bonet’s tree structure.

Object clitics

385

More plausibly, the terminal node in the syntactic structure must be specified as in (46), since we don’t get any reflexive reading, which is the function of se outside of clitic clusters. (46) {clitic {argument, third person {oblique}}} {clitic {argument, third person}} The idea of Bonet (1995) is that surface filters operate on the syntactic structures in the morphological component. Due to these filters, morphological operations may apply in the morphological component. One such operation is for example impoverishment, i.e. impoverishing the syntactic feature structure of the terminal node. (47)

Thus, in a syntactic structure such as in (47), a filter cuts the features {third person {oblique}} from the terminal node. This means that the lexical insertion conditions of le are no longer met and as a consequence the less specific se must be inserted. These impoverishment operations are morphological adaptions of proposals in the phonology of feature geometry concerning sound deletion operations (cf. Clements 1985; Halle 1995). One advantage of Bonet’s DM approach to opaque clitic sequences is that the syntactic component can be kept constant and the surface structure of the clitic sequences results from a language-specific morphological filter. Her proposal rather convincingly explains opaque clitic sequences in different languages. Goldbach (2007) adopts Bonet’s DM-model for all sequences in French preverbal clitic clusters, and Fischer (2002; 2006) explains the allomorphy of Old Catalan clitics and the change concerning the form of the clitics regarding the pre- vs. postverbal position within a DM-model. Thus, clitics are moved to a surface position, and there vocabulary insertion derives the clusters. Nevertheless, so far none of the proposals has been able to explain the object clitics of all Romance languages and neither all the divergent behaviour.

5 Conclusion As far as we can see, the linguistic community has not reached a consensus on the theoretical treatment of object clitics, not even in the Romance languages. The current state of research on the topic is far from accounting for broadly accepted generalities, not even in the relatively homogeneous group of Romance object

386

Susann Fischer and Maria Goldbach

clitics. The debate concerning movement vs. base-generation, adjunction, or their affix character is still open to further contribution. We think that one has to accept that while object clitics do have quite generalizable properties in several respects, it is also the case that grammars tolerate morphosyntactic islands of diachronically inherited detritus, such as the interpolation examples illustrated in (24) and (25) in section 3.5 (you can think of (21), (22) and (23b), postverbal clitics with finite verbs, in these terms as well). As concerns cliticization in general, it has become clear from the above discussion on Romance clitics that even if we adopt a DM-approach and treat morphology as divided between syntax and phonology, we still have to admit that clitic placement and distribution depend on syntactic and phonological constraints. Furthermore, some phenomena related to cliticization can clearly not be explained by syntactic operations only, but seem to be better analysed in terms of the syntax/ morphology as well as syntax/semantic interfaces. One such phenomenon is clearly clitic doubling, where doubling depends on the interaction of syntax (Acc/Dat, full pronoun etc.), word order in general and additionally on the semantics (specificity, definiteness, and animacy) of the doubled DP involved.

6 References Anagnostopoulou, Elena (2003), The syntax of ditransitives. Evidence from Clitics, Berlin/New York, Mouton de Gruyter. Anderson, Stephen R. (2005), Aspects of a theory of clitics, Oxford, Oxford University Press. Bleam, Tonia (1999), Leísta Spanish and the syntax of clitic doubling, PhD dissertation, University of Delaware. Bobaljik, Jonathan D. (2000), “The ins and outs of contextual allomorphy”, in: Kleanthes Grohmann/ Caro Struijke (edd.), University of Maryland Working Papers in Linguistics 10, 35–71. Bonet, Eulàlia (1991), Morphology after syntax. Pronominal clitics in Romance, PhD dissertation, Massachusetts Institute of Technology. Bonet, Eulàlia (1995), “Feature structure of Romance clitics”, Natural Language and Linguistic Theory 13, 607–647. Bonet, Eulàlia (2008), “The person-case constraint. A morphological approach”, in: Roberta D’Alessandro/Susann Fischer/Gunnar Hrafn Hrafnbjargarson (edd.), Agreement restrictions, Berlin/New York, Mouton de Gruyter, 103–128. Borer, Hagit (1984), Parametric syntax. Case studies in Semitic and Romance languages, Dordrecht, Foris. Camacho Taboada, María Victoria (2006), La arquitectura de la gramática. Los clíticos pronominales románicos y eslavos, Sevilla, Universidad de Sevilla. Cardinaletti, Anna (2008), “On different types of clitic clusters”, in: Cécile de Cat/Katherine Demuth (edd.), The Bantu-Romance connection, Amsterdam/Philadelphia, Benjamins, 41–82. Cardinaletti, Anna (2015), “Syntactic effects of cliticization”, in: Tibor Kiss/Artemis Alexiadou (edd.), Syntax – theory and analysis. An international handbook, Berlin/New York, Mouton de Gruyter, 595–653.

Object clitics

387

Cardinaletti, Anna/Starke, Michael (1999), “The typology of structural deficiency. A case study of the three classes of pronouns”, in: Henk van Riemsdijk (ed.), Clitics in the languages of Europe, Berlin/New York, Mouton de Gruyter, 145–234. Chappell, Whitney (2011), “Intervocalic voicing of /s/ in Ecuadorian Spanish”, in: Jim Michnowicz/ Robin Dodsworth (edd.), Selected proceedings of the 5th workshop on Spanish sociolinguistics, Somerville, MA, Cascadilla Proceedings Project, 57–64. Chomsky, Noam (1981), Lectures on government and binding, Dordrecht, Foris. Chomsky, Noam (1993), “A Minimalist Program for linguistic theory”, in: Kenneth Hale/Samuel J. Keyser (edd.), The view from building 20, Cambridge, MA, MIT Press, 1–52. Chomsky, Noam (1995), The Minimalist Program, Cambridge, MA, MIT Press. Chomsky, Noam (2000), “Minimalist inquiries. The framework”, in: Roger Martin/David Michaels/ Juan Uriagereka (edd.), Step by step, Cambridge, MA, MIT Press, 89–155. Chomsky, Noam (2001), “Derivation by phase”, in: Michael J. Kenstowicz (ed.), Ken Hale: A life in Language, Cambridge, MA, MIT Press, 1–52. Clements, George N. (1985), “The geometry of phonological features”, Phonology Yearbook 2, 225– 252. Déchaine, Rose-Marie/Wiltschko, Martina (2002), “Decomposing pronouns”, Linguistic Inquiry 33, 409–442. Embick, David/Noyer, Rolf (2001), “Movement operations after syntax”, Linguistic Inquiry 32:4, 555– 596. Fernández Soriano, Olga (1999), “El pronombre personal. Formas y distribuciones. Pronombres átonos y tónicos”, in: Ignacio Bosque/Violeta Demonte (edd.), Gramática descriptiva de la lengua española, Madrid, Espasa, 1209–1274. Fischer, Susann (2002), The Catalan clitic system. A diachronic perspective on its syntax and phonology, Berlin/New York, Mouton de Gruyter. Fischer, Susann (2006), “Degrammaticalization or the historical distribution of the epenthetic vowel”, in: Claus D. Pusch (ed.), The grammar of Catalan pronouns. Variation – evolution – function, Beihefte zur Zeitschrift für Katalanistik 5, 1–27. Fischer, Susann (2011), “Some notes on the rise of mismatches”, in: Natascha Pomino/Elisabeth Stark (edd.), Mismatches in Romance, Arbeitspapier 125, Konstanz, Fachbereich Sprachwissenschaft der Universität Konstanz, 113–127. Fischer, Susann/Navarro, Mario/Vega, Jorge (2016), “The clitic doubling cycle. A diachronic reconstruction”, Paper held at the 26th Colloquium on Generative Grammar, Cáceres 13.–14. April 2016. Fischer, Susann/Rinke, Esther (2013), “Explaining the variability in clitic doubling across Romance. A diachronic account”, Linguistische Berichte 236, 455–472. Gabriel, Christoph/Müller, Natascha (2005), “Zu den romanischen Pronominalklitika. Kategorialer Status und syntaktische Derivation”, in: Georg Kaiser (ed.), Deutsche Romanistik – generativ, Tübingen, Narr, 161–180. Gabriel, Christoph/Rinke, Esther (2010), “Information packaging and the rise of clitic-doubling in the history of Spanish”, in: Gisella Ferraresi/Rosemarie Lühr (edd.), Diachronic studies on information structure. Language acquisition and change, Berlin/New York, De Gruyter, 63–86. Gerlach, Birgit (2002), Clitics between syntax and lexicon, Amsterdam/Philadelphia, Benjamins. Givón, Talmy (1979), “From discourse to syntax. Grammar as a processing strategy”, in: Syntax and semantics, vol. 1: Discourse and syntax, ed. Talmy Givón, New York, Academic Press, 81–112. Goldbach, Maria (2007), “The distributed morphology of object clitics in modern French”, Zeitschrift für Sprachwissenschaft 26:1, 41–81. Goria, Cecilia (2004), Subject clitics in the Northern Italian dialects, Dordrecht, Kluwer.

388

Susann Fischer and Maria Goldbach

Grimshaw, Jane, (1982), “On the lexical representation of Romance reflexive clitics”, in: Joan Bresnan (ed.), The mental representation of grammatical relations, Cambridge, MA, MIT Press, 87–148. Grimshaw, Jane (2001), “Optimal clitic positions and the lexicon in Romance clitic systems”, in: Geraldine Légendre/Jane Grimshaw/Sten Vikner (edd.), Optimality Theoretic Syntax, Cambridge, MA, MIT Press, 205–240. Halle, Morris (1995), “Feature geometry and feature spreading”, Linguistic Inquiry 26:1, 1–46. Halle, Morris/Marantz, Alec (1994), “Some key features of distributed morphology”, MIT Working Papers in Linguistics 21, 275–288. Halle, Morris/Marantz, Alec (2008), “Clarifying ‘blur’. Paradigms, defaults, and inflectional classes”, in: Asaf Bachrach/Andrew Nevins (edd.), Inflectional identity, Oxford, Oxford University Press, 55–72. Halpern, Aaron (1995), On the placement and morphology of clitics, Stanford, CA, CSLI Publications. Herburger, Elena (1994), “Focus and the LF of NP Quantification”, in: Proceedings of Salt III, Ithaca/ New York, Cornell University Press, 77–96. Higginbotham, James (1987), “Indefiniteness and Predication”, in: Eric Reuland/Alice G. B. ter Meulen (edd.), The representations of (in)definiteness, Cambridge, MA, MIT Press, 43–70. Hinzelin, Marc (2007), Die Stellung der klitischen Objektpronomina in den romanischen Sprachen. Diachronie, Perspektive und Korpusstudie zum Okzitanischen sowie zum Katalanischen und Französischen, Tübingen, Narr. Hualde, José Ignacio (1992), Catalan, London, Routledge. Hualde, José Ignacio (2005), The sounds of Spanish, Cambridge, Cambridge University Press. Hualde, José Ignacio/Prieto, Pilar (2014), “Lenition of intervocalic alveolar fricatives in Catalan and Spanish”, Phonetica 71:2, 109–127. Iliescu, Maria/Popovici, Victoria (2013), Rumänische Grammatik, Hamburg, Buske. Jaeggli, Osvaldo A. (1982), Topics in Romance syntax, Dordrecht, Foris. Jaeggli, Osvaldo A. (1986), “Three issues in the theory of clitics. Case, doubled NPs, and extraction”, in: Hagit Borer (ed.), The syntax of pronominal clitics, Orlando, Academic Press, 15–42. Kaiser, Georg (1992), Die klitischen Personalpronomen im Französischen und Portugiesischen. Eine synchronische und diachronische Analyse, Frankfurt, Vervuert. Kayne, S. Richard (1975), French syntax. The transformational cycle, Cambridge, Cambridge University Press. Kayne, S. Richard (1989), “Facets of Romance past participle agreement”, in: Paola Benincà (ed.), Dialect variations and the theory of grammar, Dordrecht, Foris, 85–104. Kayne, S. Richard, (1991), “Romance Clitics, Verb Movement, and PRO”, Linguistic Inquiry 22, 647– 686. Kiparsky, Paul (1982), “Lexical phonology and morphology”, in: In-Seok Yang (ed.), Linguistics in the morning calm, Seoul, Hanshin, 3–91. Klavans, Judith (1982), Some problems in a theory of clitics, Bloomington, IN, Indiana University Linguistic Club. Klavans, Judith (1985), “The independence of syntax and phonology in cliticization”, Language 61, 95–120. Kuchenbrandt, Imme (2009), Prosodische Aspekte in der Entwicklung der spanischen und französischen Klitika, PhD dissertation, Universität Hamburg. Lechner, Winfried (2005), “Interpretative effects of head-movement”, Ms., University of Tübingen, http://ling.auf.net/lingbuzz/000178 (02.02.2016). Lipski, John (1989), “/s/-voicing in Ecuadorian Spanish. Patterns and principles of consonantal modification”, Lingua 79, 49–71.

Object clitics

389

Manzini, Rita (2015), “On the substantive primitives of morphosyntax and their parametric notion. Northern Italian subject clitics”, in: Marc van Oostendorp/Henk van Riemsdijk (edd.), Representing structure in phonology and syntax. Berlin/Boston, Mouton de Gruyter, 167–194. Martins, Ana Maria (1994), Clíticos na história do português, PhD dissertation, Universidade de Lisboa. Mateus, Maria H. M./Brito, Ana Maria/Duarte, Inês/Hub Faria, Isabel (2003), Gramática da língua portuguesa, Lisboa, Caminho. Monachesi, Paola (1999), A lexical approach to Italian cliticization, Cambridge, Cambridge University Press. Mussafia, Adolf (1888), “Enclisi o proclisi del pronome personale atono quale oggetto”, Romania 27, 145–146. Nevis, Joel A./Joseph, Brian D./Wanner, Dieter/Zwicky, Arnold. M. (edd.) (1994), Clitics. A comprehensive bibliography 1892–1991, Amsterdam/Philadelphia, Benjamins. Noyer, Rolf (1997), Features, positions, and affixes in autonomous morphological structure, New York/London, Garland. Pearce, Elizabeth, (1990), Parameters in Old French syntax. Infinitival complements, Dordrecht, Kluwer. Perlmutter, David M. (1970), “Surface structure constraints in syntax”, Linguistic Inquiry 1:2, 187– 255. Perlmutter, David M. (1971), Deep and surface structure constraints in syntax, New York, Rinehart and Winston. Platzack, Christer (1995), “The loss of verb second in English and French”, in: Adrian Battye/Ian Roberts (edd.), Clause structure and language change, New York, Oxford University Press, 200–226. Poletto, Cecilia (1995), “The diachronic development of subject clitics in North Eastern dialects”, in: Adrian Battye/Ian Roberts (edd.), Clause structure and language change, New York, Oxford University Press, 295–324. Poletto, Cecilia (1999), “The internal structure of AgrS and subject clitics”, in: Henk van Riemsdijk (ed.), Clitics in the languages of Europe, Berlin/New York, Mouton de Gruyter, 581–620. Pollock, Jean-Yves (1989), “Verb movement, universal grammar, and the structure of IP”, Linguistic Inquiry 20, 365–424. Preminger, Omer (2008), Breaking agreements. Distinguishing agreement and clitic doubling by their failures, Cambridge, MA, MIT Press. Prinz, Michael (1991), Klitisierung im Deutschen und Neugriechischen. Eine lexikalisch-phonologische Studie, Tübingen, Niemeyer. Raposo, Eduardo P./Uriagereka, Juan (2008), “Clitic placement in Western Iberian”, in: Guglielmo Cinque/Richard S. Kayne (edd.), The Oxford handbook of comparative syntax, Oxford, Oxford University Press, 639–697. Riemsdijk, Henk van (ed.) (1999), Clitics in the languages of europe, Berlin/New York, De Gruyter. Rigau, Gemma (1982), “Inanimate indirect objects in Catalan”, Linguistic Inquiry 13, 146–150. Rivas, Alberto (1977), A theory of clitics, PhD dissertation, MIT. Rivero, María Luisa (1991), “Clitic and NP climbing in Old Spanish”, in: Héctor Campos/Fernando Martínez-Gil (edd.), Current studies in Spanish linguistics, Washington, D.C., Georgetown University Press, 241–282. Rizzi, Luigi (1982), Issues in Italian syntax, Dordrecht, Foris. Rizzi, Luigi (1986), “On the status of subject clitics in Romance”, in: Osvaldo Jaeggli/Carmen SilvaCorvalán (edd.), Studies in Romance linguistics, Dordrecht, Foris, 391–419. Roberge, Yves (1990), The syntactic recoverability of null arguments, Kingston/Montreal, McGillQueen’s University Press.

390

Susann Fischer and Maria Goldbach

Roberts, Ian (1997), “Restructuring, head movement and locality”, Linguistic Inquiry 28, 423–460. Roberts, Ian (2010), Agreement and head movement, Cambridge, MA, MIT Press. Rooryck, Johan (2000), Configurations of sentential complementation. Perspectives from Romance languages, London, Routledge. Sadock, Jerrold, M. (1991), Autolexical syntax, Chicago, University of Chicago Press. Selkirk, Elisabeth O. (1996), “The prosodic structure of function words”, in: James L. Morgan/Katherine Demuth (edd.), Signal to syntax. Bootstrapping from speech to grammar in early acquisition, New York, Erlbaum, 187–213. Spencer, Andrew/Luís, Ana R. (2012), Clitics. An introduction, Cambridge, Cambridge University Press. Sportiche, Dominique (1999a), “Subject clitics in French and Romance inversion and clitic doubling”, in: Kyle Johnson/Ian Roberts (edd.), Beyond principles and parameters. Essays in memory of Osvaldo Jaeggli, Dordrecht, Kluwer, 189–221. Sportiche, Dominique (1999b), “Pronominal clitic dependencies”, in: Henk van Riemsdijk (ed.), Clitics in the languages of Europe, Berlin/New York, Mouton, 679–708. Strozer, Judith (1976), Clitics in Spanish, PhD dissertation, UCLA. Suñer, Margarita (1988), “The role of agreement in clitic-doubled constructions”, Natural Language and Linguistic Theory 6, 391–434. Suñer, Margarita (2006), “Left dislocations with and without epithets”, Probus 18, 127–158. Tobler, Adolf (1912 [1875]), “Besprechung von J. Le Coultre, De l’ordre des mots dans Chrétien de Troyes”, in: Vermischte Beiträge zur französischen Grammatik 5, Leipzig, Hirzel, 395–414. Uriagereka, Juan (1995), “Aspects of the syntax of clitic placement in Western Romance”, Linguistic Inquiry 26:1, 79–123. Wackernagel, Jakob, (1892), “Über ein Gesetz der indogermanischen Wortstellung”, Indogermanische Forschungen 1, 333–436. Wandruszka, Ulrich (1992), “Zur Suffixpräferenz. Prolegomena zu einer Theorie der morphologischen Abgeschlossenheit”, Papiere zur Linguistik 46, 3–27. Wanner, Dieter (1987), The development of Romance clitic pronouns. From Latin to Old Romance, Berlin/New York, De Gruyter. Zwicky, Arnold (1977), “On clitics”, in: Wolfgang U. Dressler/Oskar E. Pfeiffer (edd.), Phonologica, Akten der dritten internationalen Phonologie-Tagung, Wien, 1.–4. Sept. 1976, Innsbruck, Institut für Sprachwissenschaft der Universität Innsbruck, 29–39. Zwicky, Arnold (1985), “Clitics and particles”, Language 61, 283–305. Zwicky, Arnold/Pullum, Geoffrey (1983), “Cliticization vs. inflection. English n’t”, Language 59, 503– 513.

Judith Meinschaefer

13 Nominalizations Abstract: Nominalizations are complex words, which, under a classic view of the architecture of the language faculty, belong to the domain of word structure, or morphology. In the simplest cases, nouns can be derived from one-place predicates, like adjectives and intransitive verbs. In more complex cases, they derive from complex verbs and retain much of the syntactic and semantic complexity of their bases. Two fundamental questions have been at the core of the study of nominalization: First, which aspects of the event-structural and of the argument-structural complexity seen in the verbal domain are visible in the nominal domain, and which aspects are not? Second, how does the morphological complexity of nominalizations relate to their event- and argument-structural complexity? With a focus on these two questions, the present article aims at presenting a state-of-the-art perspective on nominalization and its implications for the interfaces between morphology, syntax, and semantics, with specific reference to Romance languages. Keywords: nominalization, deverbal noun, event structure, argument structure, morphology

1 Introduction Nominalization, and in particular deverbal nominalization, has served as a test case for the theoretical modelling of the interfaces between modules of the language faculty from early generative linguistics onwards (e.g. Lees 1966; Chomsky 1970). Nominalizations are complex words, which, under a classic view of the architecture of the language faculty, fall into the domain of word structure, or morphology. At the same time, being derived from verbs, which may be considered the core elements of sentences and which determine many aspects of sentential syntax and interpretation, deverbal nominalizations exhibit much of the same syntactic and semantic complexity, an explanation of which lies in the scope of syntax and semantics. The study of nominalization thus presupposes – and fosters – an understanding of the interaction of morphology, syntax, and semantics. Two basic questions have been at the core of the study of nominalization: First, which aspects of the event-structural and of the argument-structural complexity seen in the verbal domain are visible in the nominal domain? Second, how does the morphological complexity of nominalizations relate to their event- and argumentstructural complexity? The present article aims to present a state-of-the-art perspective on these two questions, with specific reference to Romance languages. In doing so,

392

Judith Meinschaefer

we focus on deverbal nominalizations, (1a), leaving aside deadjectival nominalizations, (1b). (1)

a.

Smoke obstructs the view – the obstruction of the view by smoke

b.

The teacher is ill – the illness of the teacher, the teacher’s illness

Among deverbal nominalizations, two types may be distinguished: nouns referring to events, (2a), and nouns referring to participants of events, (2b). The latter may or may not have an event-related semantics (cf. McIntyre 2014; Roy/Soare 2014; Alexiadou/Schäfer 2010; Bowers 2011; Baker/Vinokurova 2009 for recent studies). (2)

teach English to foreigners a.

the teaching of English to foreigners

b.

a teacher of English to foreigners

In the following, we focus on deverbal nominalizations with event readings. Two major topics to be dealt with are the event structure and the argument structure of deverbal nouns. Another important issue relates to the varying degree of verbality or nominality that may be found in different types of nominalizations, which in previous studies has often been taken as evidence for the syntactic as opposed to the morphological, or lexical, derivation of one or the other type. Before addressing these core topics, we start with an overview of the various event-related readings presented by nominalizations and of the distributional criteria by means of which they can be distinguished. Although studying the implications of nominalization for a theory of linguistic interfaces unavoidably entails certain theoretical commitments as to how the linguistic modules interact, we aim, so far as possible, at a theory-neutral discussion. Still, an important issue is whether morphology works piece-based, arranging morphemes into complex words, or process-based, deriving complex lexemes from base lexemes. Given that much of the research on nominalization of the last two decades has been framed within Distributed Morphology, a piece-based approach in which nominalizations appear as complex syntactic structures containing a verbal base, we shall adopt the terminology of that approach, without endorsing any deeper theoretical commitments. The article is structured as follows. Section 2 illustrates the array of event-related readings presented by nominalizations. In section 3, a typology of syntactically varying nominalization structures is introduced, some of which have verbal properties, while others have a purely nominal nature. Section 4 addresses the question of how verbal event structure and aspectual structure appear in nominalizations; their argument-structural complexity will be dealt with in section 5.

Nominalizations

393

2 Readings of nominalizations 2.1 Different types of event nouns Of crucial importance to the discussion of nominalization and its challenges for the interface between morphology, syntax, and semantics is a distinction introduced by Grimshaw (1990), who shows that deverbal nouns present systematic differences related to event and argument structure properties. She introduces a distinction between complex event nouns (CEN ), simple event nouns (SEN ), and result nouns (RN ). Only the first share with their base verbs the property of taking arguments. The three classes can be distinguished on the basis of various properties. CEN and SEN pattern together in referring to events, and are thus compatible with predicates like take place, (3a, b), while RN , referring to concrete or abstract objects, are not, (3c). SEN and RN , on the other hand, pattern together in not having an argument structure, (3b, c), while CEN inherit the argument structure of their base nouns, cf. (3a). (3)

Fr.

a.

Complex event noun (CEN ) L’examination des dossiers par le conseil a eu lieu hier. ‘The examination of the files by the board took place yesterday.’

b.

Simple event noun (SEN ) Plusieurs examens ont eu lieu hier. ‘Various exams took place yesterday.’

c.

Object/result noun (RN ) *Tous ces examens sur la table ont eu lieu hier. ‘All these exams on the table took place yesterday.’

As to the distinction between CEN and SEN , an item’s having or not having a complex event structure and argument structure has often been claimed to be associated with various distributional differences, such as the compatibility with different types of determiners and the availability of pluralization, (4a, b), as well as with the compatibility of the singular noun with modifiers referring to event structure, (5a, b). (4)

Fr. a.

b.

CEN : Only definite article, no pluralization possible ?*{Plusieurs + ces + des} utilisations du service par des mineurs ont eu lieu. ‘{Various + these + some} uses of the service by minors took place.’ SEN : Different types of determiners and pluralization {Plusieurs + ces + des} examens ont eu lieu hier. ‘{Various + these + some} exams took place yesterday.’

394

(5)

Judith Meinschaefer

Fr. a.

b.

CEN : Modification referring to event structure is possible L’examination continuelle des dossiers par le conseil est importante. ‘The continuous examination of the files by the board is important.’ SEN : Modification referring to event structure is impossible *L’accident fréquent/continuel sur cette route effraie tout le monde. ‘The frequent/continuous accident on this road frightens everyone.’

The distinction between CEN and SEN is not always straightforward, and a given deverbal noun often can have both readings. More recent research has shown that, for example, whether a CEN allows pluralization or not depends, among other things, on its aspectual properties (Mourelatos 1978; Roodenburg 2010; Iordăchioaia/Soare 2009; 2008; Alexiadou/Iordăchioaia/Soare 2008). Other classifications have been proposed, as well. Sleeman/Brito (2010a; 2010b) draw a distinction between five types of deverbal nominalizations; Borer (2003; 2013), in contrast, abstracts away from the difference between complex and simple event nouns, focusing on the distinction between argument-taking event nouns and result nouns that lack argument structure (cf. also Alexiadou 2010a; Roy/Soare 2011a; 2011b).

2.2 More fine-grained semantic distinctions In addition to event-related readings, deverbal nouns can have a variety of other interpretations (e.g. Melloni 2006; 2012; Ježek 2007): Among other things, they can denote the object resulting from the event referred to by the base verb, (6a), the means by which this event is brought about, (6b), or a state resulting from the event, (6c). One of the central questions addressed in research on this topic has been whether it is possible to predict the availability of different readings of the deverbal noun from the meaning of the base verb (Ježek/Melloni 2009; Melloni 2010; Bisetto/ Melloni 2005). (6)

Readings of Italian result nominals

(examples from Melloni 2006)

a.

Result object

La costruzione è crollata inaspettatamente. ‘The construction collapsed unexpectedly.’

b.

Means

L’argentatura di questo anello è molto sottile. ‘The silver-plating of this ring is very thin.’

c.

State

Occorre rimuovere l’ostruzione di questa valvola. ‘It is necessary to remove the obstruction of this valve.’

Nominalizations

395

Another strand of research, in which Romance languages have figured less prominently, has been concerned with the contribution of the context of the utterance to the interpretation of deverbal nouns, with the aim of identifying distributional diagnostics that allow unambiguous differentiation between the different readings (cf. in particular Ehrich/Rapp 2000; cf. also Roßdeutscher/Kamp 2010; Hamm/ Kamp/Van Lambalgen 2006; Brandtner/von Heusinger 2010; Dölling 2013).

3 Lexical and syntactic nominalization Since the publication of Chomsky’s (1970) “Remarks on Nominalization”, much research has focused on the distinction between syntactically derived nominalizations, (7a), and nominalizations that are derived lexically from a verbal base, (7b). On this view, nominalizations sharing most of their syntactic and semantic properties with their verbal bases, such as taking a direct object, are derived in the syntax, while, roughly speaking, those which differ semantically and syntactically from their base verbs (presenting differences that go beyond the mere contrast in having the outward distribution of a verb vs. of a noun) are taken to be derived in the lexicon. (7)

English syntactic and lexical nominalizations; examples from Chomsky (1970) a.

John’s refusing the offer

b.

John’s refusal of the offer

Chomsky points out three differences between lexical and syntactic nominalization, as exemplified by the English examples in (7): (i) Syntactic nominalization is fully productive, while lexical nominalization may be blocked in certain cases; (ii) the semantic relation between base and syntactic nominalization is regular, while that between a lexical nominalization and its base may be idiosyncratic; and (iii) the internal structure of a syntactic nominalization may be clause-like, while that of a lexical nominalization is nominal. Most approaches that distinguish between lexical and syntactic derivation are, however, not entirely clear with respect to how and where exactly the dividing line between syntactic and lexical nominalizations is to be drawn (cf. Kornfilt/Whitman 2011a for a discussion). Under the assumption that all morphological derivation is carried out in the syntax (as in Distributed Morphology), the distinction between lexical and syntactic nominalization reemerges as one relating to functional structure, with syntactic nominalizations presenting more complex functional structure than lexical nominalizations (Harley/Noyer 1997; 1998; Alexiadou 2001). Taking seriously Chomsky’s (1970, 185) view that “the proper balance between various components of the grammar is entirely an empirical issue”, so that “we have no a priori insight into the trading relation between the various parts”, in what follows we retain his distinction between syntactic and lexical

396

Judith Meinschaefer

nominalization, discussing ‘syntactically’ derived nominalization in this section, and focusing on ‘lexically’ derived nominalizations in following sections. Romance nominalized infinitives (with the exception of French and Romanian, cf. below), as opposed to lexically derived deverbal nouns, present a case of syntactic nominalization. As shown by Alexiadou/Iordăchioaia/Schäfer (2011), drawing on Romanian and Spanish data, Romance nominalized infinitives come in different types, which are distinguished by a number of syntactic differences that may be related to the locus at which nominalization of a verbal projection takes place in the syntactic configuration.

3.1 Iberoromance and Italian The Spanish infinitive, when preceded by a determiner, allows nominal uses, with (8b) or without (8a) a complement, as well as fully verbal uses, followed by a direct object DP (8c) or even by a (pronominal) subject DP (8d) (Plann 1981; Hernanz Carbó 1999; Demonte/Varela Ortega 1997; 1998; Yoon/Bonet-Farran 1991; Miguel 1996; Hare 2001; Fábregas/Varela Ortega 2006). (8)

Nominalized infinitives Sp.

a.

[El vestir y calzar (de las mujeres)] era lo más importante. ‘Clothing and putting on shoes was the most important thing.’

b.

Se escucha [el lento abrir de una cerradura]. ‘One hears the slow opening of a lock.’

c.

Fue fácil y rápido [el aprobar nuestra participación en este proyecto]. ‘Approving our participation in this project was easy and quick.’

d.

Nos favorece [el pertenecer nosotros al municipio]. ‘Our belonging to the municipality favours us.’

Verbal and nominal uses may be distinguished with regard to a variety of properties, not all of which are exemplified here: optionality vs. obligatoriness of complements, cf. (8a) as opposed to (8b), compatibility with case-marked complements vs. prepositional complements, (8c, d) as opposed to (8b), modification by adverbials vs. adjectives, compatibility with clitic pronouns and auxiliaries, and the acceptability of plurals. Let us briefly mention that Spanish even allows nominalization of finite clauses, (9). (9)

Sp.

El que tú vengas no me importa. the that you come does not matter to me ‘That you come does not matter to me.’

(Plann 1981, 204)

Nominalizations

397

Italian basically allows the same constructions as the Spanish nominalized infinitive, (10) (Salvi 1985; Zucchi 1993), though certain differences have been claimed to exist between the two languages (Szilágyi 2009; Pérez Vázquez 2002). (10)

Nominalized infinitives It.

(examples from Salvi 1985, 247)

a.

l’avviarsi lento del treno ‘the slow starting of the train’

b.

l’aver-gli Giorgio sempre Giorgio always the have.INF-him.DAT ‘Giorgio’s always having conceded to him’

dato given

ragione right

3.2 French French allows purely verbal infinitives, as in (11a), and purely nominal infinitives, as in (11b), but no event-denoting nominal infinitives that have argument structure, as in (11c) (Kerleroux 1990; George 1976; Umbreit 2014; Sleeman 2010; Marzo/Umbreit 2013). (11)

Fr.

a.

Je l’ai entendu parler très bien l’italien. ‘I have heard him speak Italian very well.’

b.

Le déjeuner a eu lieu à midi. ‘Lunch took place at noon.’

c. *le rire des faiblesses d’autrui ‘laughing about other persons’ weaknesses’ Old French, however, had nominal infinitival constructions similar to those attested in Spanish (Sleeman 2010; Buridant 2005; 2008; Kerleroux 1990; Schaefer 1911), where an argument of the infinitive could be realized as a prepositional complement, (12a, b), or as a direct object, (12c). (12)

Examples from Schaefer (1911) OFr.

a.

li trembler del leün ‘the trembling of the lion’

b.

a l’eschevir del seiremant ‘at the taking of the oath’

c.

au conter le duel qu’ele fist ‘at the telling of the sorrow she had’

398

Judith Meinschaefer

Infinitives preceded by determiners are attested in documents of the 12th century, with their frequency decreasing from the 13th century onwards (Schaefer 1911). From the 16th century onwards, infinitives preceded by determiners do not have argument structure; only infinitives of the type le boire ‘the drink’, le dîner ‘the dinner’, le manger ‘the food’ are attested (Ewert 21949, 183; Kukenheim 1967, 78; Schaefer 1911). Unlike in Spanish, neither in Old nor in Middle French are any examples attested for nominal infinitives with an overt subject (Mensching 2000, 20). To date, the reasons for the ungrammaticality of event-denoting nominalized infinitives in French, and more generally, for the differences presented by Romance languages with regard to the availability of more or fewer verbal nominalization structures remain unclear.

3.3 Romanian Romanian is like French in not allowing nominalized infinitives with a verbal internal structure. The infinitive can be verbal, or nominal, as in (13), but in the latter case it cannot assign nominative or accusative case and it allows only adjectival, but no adverbial modification (Soare/Mardale 2007; Cornilescu et al. 2013; Cornilescu 1999; 2001; 2004). (13)

Infinitive Rom. Cumpărarea casei a fost inutilă. house.the.GEN was useless buy.INF.the ‘The buying of the house was useless.’

(Cornilescu 1999, 213)

However, Romanian differs from French in that the nominalized infinitive can denote events and does have argument structure, realizing its arguments in the same way as lexical nominalizations. Note that Old Romanian, like Old French, allowed nominalized infinitives to occur with direct objects, (14). (14)

ORom.

tăierea capul cut.INF.the head.the.ACC ‘the cutting his head off’

lui he.GEN (Soare 2007, 177)

Differently from other Romance languages, Romanian presents syntactic nominalizations of a second verbal category, the supine, morphologically a past participle (Soare 2007), which is like the Spanish nominalized infinitive in being either verbal, (15a), or nominal, (15b). The nominal supine allows adverbial modification, (15c).

Nominalizations

(15)

Supine Rom. a.

Am de cules căpşuni. have.1SG to collect.SUP strawberries ‘I have to collect strawberries’

399

(Soare 2007)

b.

Cumpăratul casei a fost inutilă. useless buy.SUP.the house.the.GEN was ‘The buying of the house was useless.’ (Cornilescu 1999, 213)

c.

cititul constant al ziarelor read.SUP.the constantly of journals.GEN ‘the constant reading of journals’ (Alexiadou/Iordăchioaia/Schäfer 2011, 27)

3.4 Summary Drawing on current versions of minimalist syntax, the parametric variation found in Romance nominalization structures has been related to differences in functional structure, in particular to the presence vs. absence of verbal functional nodes like tense and aspect and of nominal functional nodes like number and classifier (encoding the mass/count distinction) (Alexiadou/Iordăchioaia/Schäfer 2011). A challenge for future research is to explain why certain languages have more restricted or less restricted possibilities to form syntactic nominalizations than others, that is, why some languages allow more complex combinations of verbal and nominal functional structures than others. Similarly, the surprising fact that the combination of nominal and verbal functional structure within a single projection is possible, while being subject to rather severe restrictions, calls for an explanation (Kornfilt/Whitman 2011a). Finally, a point that has not been addressed here is the fact that nominalized infinitives and deverbal nouns derived by affixation typically present referential and distributional differences like those exemplified in (16), as first pointed out by Vendler (1967). A possible explanation might draw on the intuition that more strongly verbal nominalizations tend to denote facts, while more nominal nominalizations tend to denote events (van Lambalgen/Hamm 2005; Asher 1993; Bücking 2012). (16)

Sp.

a.

El {establecer una + establecimiento de una} relación es improbable. ‘Establishing a relation + the establishment of a relation is improbable.’

b.

El {?*establecer una + establecimiento de una} relación era gradual. ‘Establishing a relation + the establishment of a relation was gradual.’

400

Judith Meinschaefer

4 Event structure and aspectual structure The present study focuses on event-denoting nominalizations, that is, nominalization structures that refer to the unfolding of an event in time. In section 2 above, it has already been noted that reference to events in the nominal domain occurs in two variants: Nouns can have a complex event structure (as is the case for CEN ), or they can be simple event nouns (SEN ). In what follows, we take a closer look at the eventrelated properties of deverbal nouns. Two dimensions of the temporal structure of events are relevant in this regard: First, the temporal complexity of the event itself, for example whether it comprises more than one subevent or not, and whether it implies a result state or not; second, the focusing or defocusing of other time-related aspects of the event, for instance whether it is seen as unfolding in time, as completed, as incipient, or as occurring repeatedly or habitually.

4.1 Complex event structures That there is a correlation between the presence or absence of a nominalizing affix and the temporal complexity of the event denoted by a deverbal noun has often been noted. In particular, it has been claimed that only if there is an overt nominalizer can the derived noun have complex event structure (Smith 1972; Grimshaw 1990; Alexiadou/Grimshaw 2008; but cf. Fábregas 2014; Newmeyer 2009; Harley 2009 for counterevidence). Interpretational differences between nouns with, (17a), and without, (17b, c), a nominalizer are illustrated below. (17)

Fr. a. Il a accroché les véhicules à la queue du convoi. ‘He bumped into the cars at the tail of the convoy.’ On a assisté à l’{accrochage + *accroc} progressif des véhicules à la queue du convoi. ‘We witnessed the progressive bumping into the cars at the tail of the convoy.’ b.

Un clou a accroché son pantalon. ‘A nail snagged his trousers.’ *L’accroc à/de son pantalon s’est produit hier. ‘The snag of his trousers happened yesterday.’ Il y a un {accroc + *accrochage} à son pantalon. ‘There is a snag in his trousers.’

Nominalizations

c.

401

Ce film accroche les spectateurs dès le début. ‘This film hooks the viewers right from the beginning.’ *L’accroche du spectateur doit se produire immédiatement. ‘The hook of the viewers must happen immediately.’ L’accroche du film paraît conventionnelle. ‘The trailer of the film seems conventional.’

In recent research, a fine-grained view of the syntactic dimension of event structure has gained ground, following research on the syntax-semantics interface by Ramchand (2008) and others. A crucial development has been the integration of a decompositional representation of event structure into syntactic tree configurations, so that interactions between the morphosyntactic structure of a nominalization and its event-structural interpretation can be more directly represented. One hypothesis that has been pursued in this framework assumes that nominalizing affixes relate to different subconfigurations of such syntactic-semantic tree structures (Fábregas 2010; Sleeman/Brito 2010a). Thus, deverbal nouns can lexicalize the full array of the three subcomponents of vP, that is, the initiation (initP), process (procP), and result (resP) components, cf. (18), or they can lexicalize only individual subcomponents, cf. (19) and (20); examples adapted from Sleeman/Brito (2010a). (18)

Lexicalization of initP, procP, and resP Por. A construção do campo de jogos para entreter as crianças trouxe benefícios para a comunidade. ‘The building of the playground to entertain the children benefitted the community.’ (Sleeman/Brito 2010a, 212)

(19)

Lexicalization of procP Por. A construção do campo está parada há um ano. ‘The building of the playground stopped a year ago.’

(20)

Lexicalization of resP Por. A obstrução é persistente. ‘The obstruction persists.’

4.2 Nominal and verbal aspect As to the second dimension, i.e. the focusing or defocusing of other time-related aspects of the event, it has long been noted that there is a nominal correlate to verbal aspect (Mourelatos 1978), that is, countability. Questions relating to the possibility of pluralizing event-denoting nouns are to be seen in this context, as well

402

Judith Meinschaefer

(Mourelatos 1978; Roodenburg 2010; Iordăchioaia/Soare 2008; 2009; Alexiadou/ Iordăchioaia/Soare 2008). It has been proposed by Jackendoff (1991) and others that both verbal and nominal aspect can be subsumed under the broader term ‘boundedness’. Before continuing the discussion of aspect, let us briefly note that what is meant by verbal aspect is sometimes event-structural complexity as discussed in the previous paragraphs, but this term may also refer to grammatical aspect as found, for example, in Slavic languages. Here we follow Smith (1997), who distinguishes between lexical aspect or Aktionsart on the one hand, that is, whether a verb phrase denotes, for instance, a state or an accomplishment, and viewpoint aspect on the other, that is, whether the event denoted by a verb phrase is seen as, for example, being in progress (imperfective aspect), as being completed (perfective aspect), or as occurring repeatedly (iterative aspect). Having said this, it should also be pointed out that there are complex interactions between both levels, in particular in the nominal domain. Note also that in languages with a rich verbal aspectual system, such as Russian and Polish, aspectual markers can occur inside nominalizations (Rozwadowska 2000). The same is true for Romance ‘syntactic’ nominalizations, such as the Spanish or Italian infinitive, which can occur in the form have + past participle; cf. (10) above. A frequently made claim is that deverbal nouns inherit the aspectual features of their base verbs (Mourelatos 1978; Brinton 1995; 2008; Fábregas/Marín 2011; Fábregas/ Marín/McNally 2012; Haas/Huyghe/Marín 2008; Haas/Huyghe 2010; Meinschaefer 2005a). Hence, deverbal nouns derived from unbounded, i.e. atelic, verbal bases are unbounded themselves, i.e. mass nouns, cf. (21), while nouns derived from bounded, i.e. telic, verbs are bounded, i.e. count nouns, (22). Some of the distributional criteria for detecting boundedness in the nominal domain used in the following are adapted from Haas/Huyghe/Marín (2008). (21)

(22)

Fr.

Fr.

a.

Max a patiné {pendant deux heures + *en deux heures}. ‘Max skated {for two hours + in two hours}.’

b.

Max a fait {du patinage + *plusieurs patinages}. ‘Max did {some skating + various skatings}.’

a.

Le service secret a mis deux mois à assassiner le président. ‘It took the secret service two months to assassinate the president.’

b.

On a vu beaucoup {*d’assassinat + d’assassinats}. ‘We saw {much murder + many murders}.’

At the same time, it has often been pointed out that the ‘Aspect Preservation Hypothesis’ does not always hold, so that, for instance, nouns derived from atelic, and thus unbounded, activity verbs, (23a), may refer to individualized instances of an event and may be countable, (23c).