Formal Approaches to DPs in Old Romanian [1 ed.] 9789004292550, 9789004287716

Citation preview

Formal Approaches to DPs in Old Romanian

Brill’s Studies in Historical Linguistics Series Editor Jóhanna Barðdal (Ghent University) Consulting Editor Spike Gildea (University of Oregon) Editorial Board Joan Bybee (University of New Mexico) – Lyle Campbell (University of Hawai’i Manoa) – Nicholas Evans (The Australian National University) Bjarke Frellesvig (University of Oxford) – Mirjam Fried (Czech Academy of Sciences) – Russel Gray (University of Auckland) – Tom Güldemann (Humboldt-Universität zu Berlin) – Alice Harris (University of Massachusetts) Brian D. Joseph (The Ohio State University) – Ritsuko Kikusawa (National Museum of Ethnology) – Silvia Luraghi (Università di Pavia) Joseph Salmons (University of Wisconsin) – Søren Wichmann (MPI/EVA)

VOLUME 5

The titles published in this series are listed at brill.com/bshl

Formal Approaches to DPs in Old Romanian Edited by

Virginia Hill

LEIDEN | BOSTON

Library of Congress Cataloging-in-Publication Data Formal approaches to DPs in Old Romanian / Edited by Virginia Hill.   pages cm. — (Brill’s Studies in Historical Linguistics; Volume 5.)  Includes index.  ISBN 978-90-04-28771-6 (hardback : alk. paper) — ISBN 978-90-04-29255-0 (e-book) 1. Romanian language—Syntax. 2. Romanian language—To 1500. 3. Romanian language—History. 4. Discourse analysis—History. 5. Pragmatics—History. 6. Historical linguistics. 7. Romania—History—To 1711. I. Hill, Virginia, editor.  PC713.F67 2015  459’.5—dc23

2015010249

This publication has been typeset in the multilingual “Brill” typeface. With over 5,100 characters covering Latin, IPA, Greek, and Cyrillic, this typeface is especially suitable for use in the humanities. For more information, please see www.brill.com/brill-typeface. issn 2211-4904 isbn 978-90-04-28771-6 (hardback) isbn 978-90-04-29255-0 (e-book) Copyright 2015 by Koninklijke Brill NV, Leiden, The Netherlands. Koninklijke Brill NV incorporates the imprints Brill, Brill Hes & De Graaf, Brill Nijhoff, Brill Rodopi and Hotei Publishing. All rights reserved. No part of this publication may be reproduced, translated, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior written permission from the publisher. Authorization to photocopy items for internal or personal use is granted by Koninklijke Brill NV provided that the appropriate fees are paid directly to The Copyright Clearance Center, 222 Rosewood Drive, Suite 910, Danvers, MA 01923, USA. Fees are subject to change. This book is printed on acid-free paper.

Contents Abbreviations vii List of Contributors ix Introduction 1 Virginia Hill

PART 1 The Internal Structure of the DP The Parameter of Definiteness in Romanian: Diachronic and Synchronic Evidence 17 Alexandru Nicolae Polydefinite DPs in Old Romanian 62 Daniela Isac Classified Proper Names in Old Romanian: Person and Definiteness 100 Alexandra Cornilescu and Alexandru Nicolae Agreeing and Non-agreeing Genitives in Old Romanian and the History of Romanian Genitive Constructions 154 Ion Giurgea Differential Object Marking in the First Original Romanian Texts 200 Alexandru Mardale Two DP Configurations for Supine-Based Nouns 246 Adina Dragomirescu

Part 2 The DP and the Clause Object Pronouns in the Evolution of Romanian: A Biolinguistic Perspective 269 Anna Maria Di Sciullo and Stanca Somesfalean

vi

Contents

DPs in Αdjectival Small Clauses in Romanian: A Diachronic Perspective 290 Monica Alexandrina Irimia Restrictive and Appositive Relatives 329 Anca Sevcenco Index 365

Abbreviations ABL ablative ACC accusative ADV adverb AUX auxiliary verb COND conditional DAT dative DEF definite article DEM demonstrative DO direct object DOM differential object marking ERG ergative ESS essive F feminine FUT future GEN genitive GER gerund IMP imperative IMPER imperfective INDEF indefinite INDIC indicative INF infinitive INF infinitive INST instrumental IRR irrealis IO indirect object M masculine N neuter NEG negation NOM nominative NPI Negative Polarity Item PART partitive PASS passive PFV perfective PL plural PLUPERF pluperfect PPART past participle PRES present

viii

abbreviations

PROG progressive PS simple past REFL reflexive SE ARB arbitrary se SG singular SUBJ subjonctif Note: ‘=’ attached to an item indicates its clitic/affixal status; ‘/’ between two abbreviations indicates syncretism (e.g., Case syncretism: NOM/ACC

List of Contributors Alexandra Cornilescu [email protected] University of Bucharest Anna Maria Di Sciullo [email protected] Université du Québec à Montréal Adina Dragomirescu [email protected] The I.Iordan-Al.Rosetti Institute of Linguistics Ion Giurgea [email protected] The I.Iordan-Al.Rosetti Institute of Linguistics Virgina Hill [email protected] University of New Brunswick - Saint John Monica Alexandrina Irimia [email protected] University of York & University of Toronto Daniela Isac [email protected] Concordia University Alexandru Mardale [email protected] INALCO de Paris Alexandru Nicolae [email protected] The I.Iordan-Al.Rosetti Institute of Linguistics

x Anca Sevcenco [email protected] University of Bucharest Stanca Somesfalean [email protected] Université du Québec à Montréal

list of contributors

Introduction Virginia Hill This volume provides a number of studies that bring a formal perspective to the discussion of diachronic changes the noun phrases underwent from Old to Modern Romanian. It is the first time that a collection of such studies is published in English, outside Romania, so the respective data and tests become accessible to linguists all over the world. Why would the Old Romanian DP be interesting for cross-linguistic studies?1 In order to answer this question, I will briefly summarize the main points of contention in the current DP theory, and then point out how the studies included in this volume contribute to a better understanding of the relevant issues. In a nutshell, the data from Old Romanian increase the empirical basis needed to revise or refine current analyses of DPs with respect to their internal structure and to the constraints such phrases impose on the structure and interpretation of clauses. The empirical observations lead to precise theoretical proposals that benefit from the diachronic perspective adopted across the board in this volume. 1

Data and Methodology

Old Romanian texts have been preserved starting with mid16th century. Thus, what we call old for Romanian grammar corresponds to the early modern stages in other languages (see also Hill & Alboiu 2015). Thus, Old Romanian is the language of the texts created, translated or printed from the 16th century up to the end of the 18th century (Gheție et al. 1997). The Institute of Linguistics in Bucharest compiled a digital corpus comprising most of these texts. All the contributors to this book used texts included in this corpus, in their digital or original versions, with the occasional addition of other documents. The search is manual and, for the digital version, it relies on the word search key in the software. The formal analyses proposed on the basis of these data are couched in the framework of generative grammar, in particular, the minimalist theory (Chomsky 1995 et seq).

1  The term Determiner Phrase (DP) refers to the complex structure generated by a noun; that is, the argumental structure of the noun plus its functional domain.

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_002

2

hill

2 Typology It is well known that the Romanian morphosyntax displays a typological mix of Romance and Balkan parametric settings: while the morphological items are essentially inherited from Latin (Iordan & Manoliu 1965), the structure in which they merge is derived according to Balkan patterns (Mladenova 2009). For nouns, this typological mix is best illustrated by the definite article, which has a Latin etymon, but it is enclitic on the noun stem in Romanian—following the Balkan pattern—whereas in other Romance languages it is prenominal (e.g., Rom. fiul ‘son=DEF’ versus It. il figlio ‘DEF son’). This parametric setting remains constant in Old and Modern Romanian, and is responsible for a wide range of cross-linguistic variation concerning the word order within the DP. For example, among Romance languages, only Romanian may display definite articles in true vocatives (e.g., fiule ‘son.DEF.VOC’; Hill 2014);2 and may allow adjectives to carry the article (e.g. bunul fiu ‘good=DEF son’; Giusti 1993). Diachronic changes affect only the micro-parameters, and they make the topic of the articles included in this volume. That is, further ramifications of the parametric setting for the definite articles are discussed (among other properties of the DP), and changes are pointed out in the area of: multiple agreement, where both adjectives and nouns carry the definite article; definiteness marking with proper nouns; the development of prenominal genitive markers; and the syntax of wh-phrases that had enclitic articles in Old Romanian. 3

Formalization: The Internal Structure of DPs

Since the formalization of nominal phrases as DP structures (Brame 1982; Abney 1987), the linguistic theory strived to identify the finer articulations of these configurations, in a way that would capture the morphosyntactic and interpretive variations observable intra- and cross-linguistically. For example, variations arise regarding the codification of phi-features (i.e., number and gender), the expression of possession, the definiteness marking, the word order within the DP and so on. Typically, the theoretical approach to these issues relies on a finer articulation of the functional field (e.g., Cinque 1990; Cornilescu 1992; Giusti 1993 a.o.) and/or a reassessment of the DP internal

2  French direct addresses of the type les amis! ‘friends!’ are not true vocatives (i.e., where the D field must be absent), in terms of Espinal (2013), but contain the D field and qualify as DPs.

3

introduction

displacements (Kayne 1994; Ștefănescu 1997; Corver 2003 a.o.), in terms of A or A’-movement (see an overview in Coene & D’Hulst 2003). In both cases, most analyses rely on the mapping of semantic features as functional features which, in turn, act as triggers for movement. 3.1 Functional Features As far as the articulation of the functional domain is concerned, noun phrases have been argued to map number (NumP; Ritter 1988; Carstens 1991 a.o.), agreement (AgrP; Kayne 1994), possession (PossP; Valois 1991), Case (KP or FP; Valois 1991; Giusti 1993), gender/classification (ClassP; Ndayiragije & Nikema 2011), and definiteness (Danon 2010). Accordingly, the internal structure of the DP may consist of one of the hierarchies in (1), where the features listed above are not necessarily mapped to a functional head distinct from D (i.e., there is no one-to-one mapping requirement). (1)

a. b. c. d. e.

DP > NumP > CaseP > NP FP/KP > DP > NP DP > NumP > NP DP > PossP > NP* > NP DP > NumP > ClassP > NP

(Valois 1991) (Giusti 1993) (Ritter 1988) (Valois 1991) (Picallo 2008)

Crucially, there is wide disagreement as to the exact representation of the functional domain of a noun phrase, as the variations in (1) indicate. Practically, the representations differ according to the empirical data that motivate them. However, the inventory of functional features is generally agreed upon, variation arising with respect to their ordering and their substantiation. Modern Romanian noun phrases have been shown to display all the projections in (1), the DP exhibiting morphological Case (e.g., Cornilescu 1992), agreement endings (e.g., Ștefănescu 1997), evidence for the number feature (Dobrovie-Sorin 2012), special morphology for expressing possession (Cornilescu 1992; Ștefănescu 1997; Dobrovie-Sorin & Giurgea 2011 a.o.), as well as morphosyntactic sensitivity to the contrast between count and mass nouns (e.g., Iordăchioaia & Soare 2008), and expletive articles (Nicolae 2013). Old Romanian noun phrases are shown in this volume to follow the same pattern, but with non-trivial variations. In particular, Old Romanian displays multiple agreement within DPs, which is not the case in Modern Romanian, as shown in (2). That is, the syncretic ending for the definite article (–lu-) and the Genitive Case (-i) appears on three items in (2a) but only on one item in (2b).

4

hill

(2) a. păntru sufletul răposatului jupânului Predei OR for soul.DEF late.DEF.GEN master.DEF.GEN Preda.GEN ‘for the soul of the late master Preda’ (DÎ.1600: LVI) b. pentru sufletul răposatului jupân Preda for soul.DEF late.DEF.GEN master Preda ‘for the soul of the late master Preda’

MR

Alexandru Nicolae argues that the distribution of the definite article in structures such as (2) justifies an analysis whereby the definiteness feature is mapped as a functional feature in both Old and Modern Romanian. This has been proposed for other languages (e.g., Danon 2010; Kibort 2010) and finds confirmation in the Romanian data. In Old Romanian, this parameter combines with the Long Distance Agree operation, which has been lost across the board towards Modern Romanian, where only Local Agree is possible. Hence, multiple agreement is possible (but not obligatory) in Old Romanian, as in (2a), but impossible in Modern Romanian (2b). For a subset of the same type of constructions, Daniela Isac argues that the mapping of definiteness is split in Old Romanian, over a definiteness [def] feature and a reference [Ref] feature, both unvalued and associated with D. The definite article is spelled out disjointly both on items that probe D to check/ value the [def] feature, and on items that check/value the [Ref] feature on D. Modern Romanian lost this distinction (i.e., it lost [Ref]), which explains the contrast between (2a) and (2b). This analysis keeps the diachronic change within one parameter and dispenses with the parametrization of the Agree operation. To complicate the picture, proper nouns may display the definite article in Old Romanian, as in (3a), but not in Modern Romanian, as shown in (3b), for equivalent versions of the same example. (3) a. făcu nuntă nepotului său, Vladul vodă OR made.3 wedding nephew.DEF.DAT his Vlad.DEF king ‘he threw a wedding for his nephew, King Vlad’ b. făcu nuntă pentru nepotul său, Vlad vodă MR made.3 wedding for nephew.DEF his Vlad king ‘he threw a wedding for his nephew, King Vlad’ Aiming to account for the structure of proper nouns as in (3), Alexandra Cornilescu & Alexandru Nicolae adopt Longobardi’s (2008) proposal that such

introduction

5

DPs map a [Person] feature (which is not present on the D of common nouns). Their contribution is to show the effects arising from the mapping of [Person] to a D that is positive for the definiteness parameter. It is argued that this featural make-up is distributed over Picallo’s (2008) hierarchy shown in (1e). The diachronic change in the use of the definite article shown in (3) would then follow from the same change in the parameter for the application of Agree (i.e., Long Distance in Old Romanian but Local in Modern Romanian). Numerous studies on the DP domain focus on the structure of possessive phrases. In fact, the mapping of a [possessive] feature has been a major argument for the proposal of D as a functional head in Abney (1987). Valois (1991) proposes a PossP within a DP structure that replicates Larson’s (1988) VP shell. That is, in the hierarchy DP > PossP > NP* > NP, PossP is assigned the external th-role of N. Kayne (1994) also treats possessive phrases as having an argument structure from which the items move to the functional domain to satisfy agreement requirements. Romanian provides important empirical support in this respect, especially due to the fact that possessiveness is marked both on the prenominal definite article and on the noun, as [definite] and genitive, as in (4). (4) un frate al femeii a.MSG brother GEN.MSG woman.DEF.FSG.GEN ‘a brother of the woman’ Within the group of Romance languages, the genitive possessive article al has no counterpart, as these languages developed a partitive pattern, based on de, which did not spread to Romanian (Iordan & Manoliu 1965). Understanding the emergence and the properties of al is crucial for confirming or infirming theoretical generalizations regarding the mapping of possessiveness within DP. Ion Giurgea treats al-phrases as agreeing genitives and argues that they emerged in a system that allowed for the existence of two definite articles, a suffix and an independent (DP-initial) one: the DP-initial article became a genitive K in attested Romanian, but kept its agreement properties. This K bears, besides the adnominal structural Case feature (treated as uPoss), uPhi-features that are valued by the case licensor (the item that checks uPoss). The challenging aspect of this configuration is that it indicates upward probing (since the probe is K and the goal is the case licensor, external to the KP), instead of the standard downward probing assumed in the grammatical theory (Chomsky 1995, 2000). Upward probing has been noticed for other configurations, crosslinguistically (see Christodoulou & Wiltschko 2012 and references within), so this paper contributes as well to that theoretical debate.

6

hill

3.2 D = C? A different line of inquiry regarding the internal structure of DPs focuses on the uniformity of functional hierarchies: how different are the verbal and nominal functional domains? Since Szabolcsi (1983) and Abney (1987), the debate surrounds the functional identity between D and I(nflection) or between D and C(omplementizer) domains. More recently, arguments have been made that the DP domain allows for the encoding of discourse features, in addition to grammatical features (Aboh 2004; Haegeman 2004 a.o.), in the same way we see such encoding at the left periphery of clauses (i.e., in the CP field). Thus, Topic and Focus phrases arise at the left periphery of the DP, and they are responsible for a number of A’-movements inside the DP, which leads to variation in word order (Villalba 2006). Romanian has usually been involved in the comparative paradigms that provided empirical evidence for a discourse domain within DP (e.g., Giusti 2002; Villalba 2006), mainly because of the variation in the [adjective > noun] versus [noun > adjective] word order. Recently, a different type of evidence has been brought to light, attesting to the Topic mapping at the left periphery: the merging of the differential object marker pe has been shown to arise from the presence of a [topic] feature on D, rather than as a side effect of clitic doubling or Case assignment requirements (Hill & Tasmowski 2008; Hill 2013). This marker is obligatory with [human] and some [animate] nouns in Modern Romanian, as in (5c), but is optional in Old Romanian, as in (5a, b); also, in Old Romanian, the marker may occur with non-animate nouns. (5) a. au bătut pre unguri OR has= beaten DOM Hungarians ‘he beat he Hungarians’ (Ureche 100) b. au lăsat ungurii și au fugit OR has= left Hungarians.DEF and has= run ‘he left the Hungarians and ran’ (Ureche 118) c. (i)-a lăsat *(pe) unguri și a fugit MR them=has= left DOM Hungarians and has= run ‘he left the Hungarians and ran’ Alexandru Mardale proposes a diachronic investigation of differential object marking in Romanian. He establishes the axis of grammaticalization for pe, which undergoes semantic bleaching from a concrete to an abstract lexical preposition (equivalent to the topicalizing expression ‘as far as’), and is then

introduction

7

reanalyzed as a D with the [topic] feature; eventually, topic pe-D is reanalyzed as a grammatical marking strategy for the direct object position (complete semantic bleaching). Old Romanian reaches the D/Topic stage, whereas Modern Romanian progressed to the grammatical DOM stage. This approach allows the author to demonstrate that DOM is excluded with direct objects having a property reading (type ); that is, with direct objects that are (semantically) non-referential and (morpho-syntactically) non-argumental, possibly resulting in structures with pseudo-incorporation. 3.3 Variations in the Functional Field The inventory of the functional features that led to the representations in (1) is likely to be universal for DP fields. However, the empirical evidence often indicates that not all the features are substantiated in a systematic way. This kind of variation has been observed not only cross-linguistically, but intra-linguistically as well, from one class of nouns to another. For example, if highly inflected nouns project to K, where the [Case] feature is mapped, non-inflected nouns do not display the K level. Should we consider that, for the latter, the internal structure stops at DP while the KP layer is not projected? Boucher (2003) proposes such an approach for the diachronic changes in the French nominal system, postulating that Old French K is reanalyzed as D, which further changes the positive setting for the definiteness parameter (K being inherently definite) to a system where the definiteness on D needs to be valued syntactically. Every feature listed for (1) may occur in some derivations but be absent from other derivation, even within the same language. Is there a systematic and predictable mechanism that can account for this variation? So far, there are as many proposals in this respect as there are case studies. Old Romanian is an underexploited gold mine, providing, in this volume, an example of such variation within the class of supine nouns: some of these nouns have all the expected D features, whereas others lack phi-features in this domain. Regular supines are shown in (6a), displaying the number feature. Deficient supines shown in (6b) ban the number feature. How can these two types of derivations concur in the same grammar? (6) a. începutul / începuturile begin.SUP.SG.DEF begin.SUP.PL.DEF ‘the beginning/ the beginnings’ b. distrusul / *distrusurile destruct.SUP.SG.DEF destruct.SUP.PL.DEF ‘the destruction/ the destructions’

8

hill

Adina Dragomirescu treats the regular supine DPs as having Picallo’s (2008) hierarchy, and argues that the deficient supines lack NumP and ClassP. She relates this variation to a switch in the feature value for inner Aspect, which is a property of the supine root. That is, the author follows Wiltschko (2014) for assuming that a verbal or nominal root has an inner Aspect feature that is substantiated as [+/– bounded] in Indo-European languages. Thus, regular supine DPs have this feature valued freely for one of these settings, whereas the deficient supine DPs have a fixed [– bounded] value, which blocks the substantiation of phi-features and leads to the inactive state of NumP and ClassP (or their deletion). The contrasting behaviour of the two classes of supine nouns can be reduced to the parametric variation in the setting for the [bounded] feature. The diachronic change consists in the preference for the deficient class, the regular class being unproductive in Modern Romanian. 4

Formalization: The DP and the Clause

It has long been noticed that the distribution of DPs in the clause is related to the variation in their internal properties (Crisma 1990; Longobardi 1994 a.o.). Pronouns stand out in this respect since their distribution is constrained by binding principles according to their featural make-up (see Déchaine & Wiltschko 2002 and references within). In Romanian, the featural make-up of pronouns must also include discourse features, such as [topic], mapped through the differential object marker pe, which became part-and-parcel of strong direct object pronouns since Old Romanian (see Mardale, this volume). Furthermore, topicalization involves inherent specificity, so the distribution of pe-DPs (nouns or pronouns) will further reflect the interaction between the available syntactic positions (e.g., for arguments) and the semantic requirements. Several options are available in the Romanian grammar for balancing this interaction. One is the change in the inventory of features (i.e., grammaticalization, by which a full pronoun becomes a clitic); another one is the change in the merge position, by which the full pronoun is moved to a functional projection that ensures the necessary checking and valuation. Both operations are discussed in this volume: Anna Maria Di Sciullo & Stanca Somesfalean point out that 16th century texts attest to a situation where full pronouns where exclusively post-verbal in argumental position (7a), whereas clitic pronouns could be either pre- (7c) or post-verbal (7b).3 3  Strong pronouns can be preverbal if they are fronted to non-argumental positions (Topic or Focus) in the CP field.

introduction

9

(7) a. cu slavă priimiși mine with glory received.2SG me.ACC ‘With glory you received me.’ (Coresi, 137r) b. fără dereptate mânară-mă without justice led.3PL= me.ACC ‘they chased me unjustly’ (Coresi, 238v) c. nu mă rușinez când caut not me.ACC/ REFL embarrass.1SG when search.1SG ‘I’m not ashamed when I search’ (Coresi, 231v) In Modern Romanian, the option in (7a) has been lost (unless the pronoun is under DOM), and only the preverbal position for the clitic, as in (7c) is productive. The analysis argues that examples as in (7) indicate a fluctuation in which the distribution of clitics followed a symmetric pattern of distribution (pre- and post-verbal), which was unsustainable, on theoretical grounds: the tendency in natural languages is to resolve such fluctuations through a symmetry breaking and promote asymmetry (Di Sciullo 2005). This is achieved in Modern Romanian, where the clitics are pre-verbal by default. Monica-Alexandrina Irimia focuses on the subjects of secondary predicates, and notices that, in Old Romanian, strong pronouns can be either DOM-ed or not, as in (8a, b); the nouns, however, are obligatorily under DOM unless they carry the definite article. In Modern Romanian, both the strong pronouns and the [human] nouns are obligatorily under DOM with clitic doubling (8c). (8) a. Dumnezeu ispiti pre ei, și află ei God tested.3 DOM them and found.3 them destoinici luiș. loyal to.him ‘God put them to test and found them loyal to him.’ (Coresi EV 260) b. Ispitind pre el diavolul . . . află pre el . . . testing DOM him devil.the found.3 DOM him ‘When the Devil put him to test, he found him invincible.’ (Coresi EV 520) c. Ispitindu-l pe el diavolul . . . îl găsi testing=him DOM him devil.the him= found.3 *(pe) el . . . de nebiruit. DOM him as invincible ‘When the Devil put him to test, he found him invincible.’

10

hill

The author analyzes the change from optional DOM in Old Romanian, as in (8a, b), to obligatory DOM in Modern Romanian (8c) by postulating a development of the definiteness scale for DOM in Romanian, complementary to the animacy scale. It is proposed that a complex predicate analysis (rather than a small clause analysis) of constructions as in (8) is able to derive the strong/ specific readings of the shared arguments under discussion by employing a Multiple Agreement Mechanism and exploiting Raising to Object options. Any overview of the DP syntax includes a discussion of relative clauses. In this respect, there are significant changes from Old to Modern Romanian in restrictive and appositive relatives. More precisely, the inventory of relative pronouns and complementizers has changed, and so did the distribution of the clause, since Old Romanian allows for distance relatives, whereas in Modern Romanian the relative is obligatorily adjacent to its antecedent. Anca Sevcenco points out that, structurally, the Old Romanian restrictive relatives were of two types (see also Grosu 1994): (i) wh-relatives introduced by an inflected or an invariable relative pronoun that have a gap at the relativization site and involve displacement of the relative phrase to the left periphery of the clause; and (ii) restrictive relatives introduced by relative complementizers, which also have a gap at the first merge position and are further derived by the internal merge of the nominal relative head with the Focus Phrase in the left periphery. Diachronic changes affect the list of relative complementizers (i.e., de and ce become unproductive), and the status of relative pronouns; namely, relative pronouns lose their inflected definite article, and the emerging invariable form is eventually reanalyzed as non-phrasal, merged directly in the C head, instead of being a phrase that moves to the CP. Changes in the derivational operations apply accordingly, to accommodate this upward reanalysis. 5 Conclusions The papers assembled in this volume provide original data on the syntax of DPs in Old Romanian. These data are introduced descriptively, at the beginning of each paper, so the material is accessible to linguists working in any theoretical framework. Then, the analyses are couched in the framework of generative grammar and contribute to the theoretical debate regarding the internal structure of DPs and their distribution in specific constructions (e.g., in relation to the verb, in secondary predicates and in relative clauses). This volume does not intend to provide a reference grammar of the DP in Old Romanian, but only to offer a sample of theoretical issues that may benefit from the inclusion of Old Romanian data in the comparative paradigms. We

introduction

11

sincerely hope that this sample will stir the specialists’ interest and that the authors will be contacted in the future for further data and discussions. References Aboh, Enoch. 2004. Topic and focus within D*. Linguistics in the Netherlands 21: 1–12. Abney, Steven P. 1987. The English Noun Phrase in its sentential aspect. Ph.D. dissertation, MIT. Boucher, Paul. 2003. Determiner Phrases in Old and Modern French. In M. Coene & Y. D’Hulst (eds), From NP to DP: The syntax and semantics of noun phrases. 47–70. Amsterdam: John Benjamins. Brame, Michael. 1982. The head-selector theory of lexical specifications and the nonexistence of coarse categories. Linguistic Analysis 10(4): 139–183. Cinque, Guglielmo. 1990. Agreement and the head-to-head movement in the Romance Noun Phrase. Invited talk. LSRL 20, University of Ottawa. Coene, Martine & Yves D’Hulst. 2003. Introduction: The syntax and semantics of noun phrases. Theoretical background. In M. Coene & Y. D’Hulst (eds), From NP to DP: The syntax and semantics of noun phrases. 1–46. Amsterdam: John Benjamins. Cornilescu, Alexandra. 1992. Remarks on the determiner system of Rumanian: the demonstratives al and cel. Probus 4(3): 189–260. Corver, Norbert. 2003. On three types of movement within the Dutch nominal domain. In M. Coene & Y. D’Hulst (eds), From NP to DP: The syntax and semantics of noun phrases. 297–328. Amsterdam: John Benjamins. Crisma, Paola. 1990. Functional categories inside the NP: A study on the distribution of nominal modifiers. Ph.D. thesis, University of Venice. Christodoulou, Christiana & Wiltschko, Martina. 2012. Function without content: Evidence from Greek Subjunctive na. In A.-M. Di Sciullio (ed.), Towards a Biolinguistic Understanding of Grammar: Essays on Interfaces. 117–140. Amsterdam: John Benjamins. Danon, Gabi. 2010. The definiteness feature at the syntax-semantics interface. In A. Kibort & Gr G. Corbett (eds), Features. Perspectives on a Key Notion in Linguistics. 144–165. Oxford: Oxford University Press. Déchaine, Rose-Marie & Martina Wiltschko. 2002. Decomposing Pronouns. Linguistic Inquiry 33: 409–422. Di Sciullo, Anna Maria. 2005. Asymmetry in Morphology. Cambridge, MA: The MIT Press. Dobrovie-Sorin, Carmen. 2012. Number as a Feature. In A. Cardinaletti; N. Munaro; G. Giusti; C. Poletto (eds), Functional Heads. 304–324. New York: Oxford University Press.

12

hill

Dobrovie-Sorin, Carmen & Ion Giurgea. 2011. Pronominal possessors and Feature Uniqueness. Language 87(1): 126–157. Espinal, Maria-Teresa. 2013. On the structure of vocatives. In B. Sonnenhauser & P. Noel Aziz Hanna (eds.), Vocatives! Addressing between system and performance. 109–132. Berlin: Mouton de Gruyter. Giusti, Giuliana. 1993. La sintassi dei determinanti. Padova: Unipress. Gheție, Gheorghe et al. (eds). 1997. Istoria limbii române literare. Epoca veche. București: Editura Academiei Române. Grosu, Alexandru. 1994. Three studies in locality and Case. London: Routledge. Haegeman, Liliane. 2004. DP-Periphery and Clausal Periphery: Possessor Doubling in West Flemish. In D. Adger, C. de Cat & G. Tsoulas (eds.), Peripheries. Syntactic Edges and their Effects. 211–240. Dordrecht: Kluwer. Hill, Virginia. 2014. Vocatives: How syntax meets with pragmatics. Leiden: Brill. ———. 2013. The Direct Object Marker in Romanian: A Historical Perspective. Australian Journal of Linguistics 33: 140–151. Hill, Virginia & Gabriela Alboiu. 2015. Verb movement and clause structure in Old Romanian. Oxford: Oxford University Press. (forthcoming). Hill, Virginia & Liliane Tasmowski. 2008. Romanian Clitic Doubling: A view from pragmatics-semantics and diachrony. In D. Kalluli & L. Tasmowski (eds.). Clitic Doubling in the Balkan Languages. 135–164. Amsterdam: J. Benjamins. Iordan, Iorgu & Maria Manoliu. 1965. Introducere în lingvistica romanică. București: Editura didactică și pedagogică. Iordăchioaia, Gianina & Elena Soare. 2008. Two kinds of Event Plurals: Evidence from Romanian Nominalizations. In O. Bonami & P. Cabredo Hofherr (eds), Empirical issues in Syntax and Semantics7, 193–216. Paris: CNRS. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass: MIT Press. Kibort, Anna. 2010. Towards a typology of grammatical features. In A. Kibort & G.G. Corbett (eds), Features. Perspectives on a Key Notion in Linguistics. 64–106. Oxford: Oxford University Press. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N movement in syntax and Logical Form. Linguistic Inquiry 25(4): 609–665. ———. 2008. Reference to Individuals, Person, and the Variety of Mapping Parameters. In H.H. Müller & A. Klinge (eds), Essays on Nominal Determination. 189–213. Amsterdam: John Benjamins. Mladenova, Olga. 2009. Definiteness in Bulgarian: Modelling the Processes of Language Change. Berlin: Mouton de Gruyter. Ndayiragije, Juvenal & Emmanuel Nikema. 2011. Why classifiers are not determiners: from Chinese to Bantu. Paper presented at the Fifth International Conference on Formal Linguistics, Guangzhou, Dec. 10–12, 2011.

introduction

13

Nicolae, Alexandru. 2013. Notă de sintaxă comparată: parametrul [+definit] în sintaxa grupului nominal românesc. Limba română, LXII, 2: 186–208. Picallo, M. Carme. 2008. Gender and number in romance. Lingue e Linguaggio 1: 47–66. Ștefănescu, Ioana. 1997. The syntax of Agreement in Romanian. MIT-WPL 14 monograph. Villalba, Xavier. 2006. The topic-focus articulation and (non)spurious articles in Germanic and Romance. Presented at “DP-internal information structure”, Utrecht. Handout at: http://www.let.uu.nl/~Marjo.vanKoppen/personal/ISDP/abstract%20 website/Xavier1.doc. Wiltschko, Martina. 2014. The universal structure of categories. Cambridge: Cambridge University Press.

Part 1 The Internal Structure of the DP

∵

The Parameter of Definiteness in Romanian: Diachronic and Synchronic Evidence Alexandru Nicolae This paper brings diachronic and synchronic evidence for the claim that the Parameter of Definiteness coined by Danon (2010) is active in the syntax of the Romanian nominal phrase. We analyze the status of the definiteness feature in the typology of grammatical features put forth by Kibort (2010), and show that definiteness is realized as a grammatical suffix and behaves like a morphosyntactic feature in Romanian as it is involved in syntactic agreement and, marginally, in licensing. The morphosyntactic nature of definiteness opens the way towards a proper parameterization. We then thoroughly discuss the patterns of definiteness realization which indicate that the definite article starts low, as a suffix on the noun (as chiefly shown by the “low definite article” of Old Romanian identified by Cornilescu and Nicolae 2011a), and its (sometimes multiple) realization is driven by Agree.

1 Introduction The goal of this paper is to support, with diachronic and synchronic evidence coming from Romanian, the claim put forth by Danon (2010) (see also Ledgeway 2013) in the analysis of Hebrew, namely that there exists a Parameter of Definiteness whose specification accounts for some of the most radical disanalogies between the DP-syntax of languages like Hebrew and Arabic or, as claimed here, Romanian (see also Nicolae 2013d), specified as positive for this parameter, and languages like English or French, in which definiteness is a morphosemantic feature with a very reduced degree of narrow syntax significance.

*  This paper starts from the intuitions expressed in Cornilescu & Nicolae (2011a) and brings cross-linguistic evidence for the analysis advocated there, which is exclusively based on (Old) Romanian data. In the present paper, we provide additional diachronic and synchronic evidence for the claims advanced in Cornilescu & Nicolae (2011a), through a formal analysis of multiple definiteness agreement.

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_003

18

Nicolae

We will be mostly concerned with the analysis of Romanian definite DPs, focusing on the patterns of definiteness in Old Romanian (henceforth, OR) and Modern Romanian (henceforth, MR) and with the (highly diversified and syntactically distinct) patterns of multiple realization of the definite suffix within the same DP from the same double, synchronic and diachronic, perspective. Other phenomena, for example the emergence of the genitival marker al, a, ai, ale (specific to Romanian in Romance, see Dragomirescu & Nicolae 2015a) or the suffixation of the definite article onto indefinite quantifiers (e.g. unul one. DEF, altul other.DEF), also bring support for our analysis. However, a proper account of these phenomena involves a thorough discussion of the syntax of Romanian genitives (for Rom. al, ai, a, ale) (see Cornilescu 1992, 2003) and of the syntax of nominal ellipsis (for Rom. unul and altul) (see Cornilescu & Nicolae 2012, Nicolae 2013a: ch. 3), two chores which are beyond our space limitations here, but are addressed in Giurgea (this volume). Following previous literature, we adopt the DP-hypothesis (Abney 1989, Longobardi 1994), and assume that the DP consists of (at least) the functional projections in (1) (following Longobardi 2001; Giusti 2005; Borer 2005; Julien 2005; Stan 2009 and Tănase-Dogaru 2009 for Romanian). (1) DP > QP > NumP > NP Adjectives are assumed to merge as specifiers of functional projections in the space above NP (Cinque 2010); however, the prenominal space in which the adjectives merge is not uniform (Cornilescu 2006, 2009a), and adjectives from different classes merge in distinct areas of the prenominal space (Cornilescu & Nicolae 2011b). The outline of the paper is as follows: section 2 is devoted to the introduction of the theoretical ingredients on which the analysis is based; section 3 presents the empirical data which indicate that the Romanian definite article starts out as a suffix on the noun; section 4 analyzes the patterns of (multiple) definiteness realization of Old and Modern Romanian; finally, section 5 draws the conclusions. We start by introducing the theoretical ingredients necessary for the analysis, the most important of which are the following: (i) Pesetsky & Torrego’s (2007) distinction between valuation and interpretability of features and their feature-sharing model of Agree; and (ii) a finer-grained typology of features, modelled on recent work by Corbett and his collaborators (see especially Corbett 2011, 2012; Kibort 2010; Kibort & Corbett 2010). We thus bridge the gap between two traditions of research, linguistic typology and generative grammar,

the parameter of definiteness in romanian

19

endeavouring to formally implement the typological results, a fact which opens the way to place them in a minimalist approach to parameterization (cf. Roberts 2012; Kayne 2013).1 2

Theoretical Framework

2.1 Interpretability and Valuation Minimalist research as well as linguistic typology regard features as being ordered pairs of type (Adger & Svenonius 2011: 38) or (Kibort 2010: 66). This perspective on feature is convergent with the Pesetsky & Torrego (2007) theory of Agree, which manipulates two featural dimensions: valuation and interpretability. The framework assumed here, the second variant of Minimalism (Chomsky 2000 and ssq. work), disposes of the Spec-Head agreement operation and refines the conception on feature mechanics: features are no longer checked, but valued. The operation that drives feature valuation is Agree. Pesetsky & Torrego (2007) distinguish between interpretable and uninterpretable features, on the one hand, and valued and unvalued features, on the other hand. The syntactic derivation is driven by the need to delete uninterpretable features (Chomsky 2000, 2001). In Pesetsky & Torrego’s (2007) framework, Agree is conceived as a form of feature sharing (see also Frampton & Gutmann 2006 for a similar proposal).

1  In the current minimalist framework, there are two mainstream approaches to parameterization, not necessarily antagonistic (cf. Gallego 2011): the macroparametric approach (Baker 2008) and the microparametric approach (Roberts 2012; Kayne 2013). While not dismissing microparameters whose source are the lexical properties of functional items (The “Borer– Chomsky” Conjecture, formulated by Baker 2008: 353), the macroparametric approach advocates the existence of macroparameters whose source of variation is the grammatical component of the faculty of language, delineable by a major comparison of languages from different families (Baker 2008). By contrast, in the microparametric framework, macroparameters are not dismissed; rather, they come about as clustering effects (aggregates) of microparameters, and thus the source of parametric variation is still lexical (Roberts 2012). For example, assuming that the canonical UG ordering is head-complement (Kayne 1994), rigidly head-final languages (e.g. Japanese, Korean, Dravidian) are characterized by the presence of a head-final feature (i.e. a diacritic which triggers comp-to-spec movement) on all heads (Roberts 2012: 320–321).

20

Nicolae

(2) Agree (Feature sharing version) (Pesetsky & Torrego 2007) (i) An unvalued feature F (a Probe) on a head H at syntactic location α (Fα) scans its c-command domain for another instance of F (a Goal) at location β (Fβ) with which to agree. (ii) Replace Fα with Fβ, so that the same feature is present in both locations. Agree is initiated by some head at a location α (the Probe), provided with an unvalued, uninterpretable or interpretable feature Fα; the c-command domain is scanned for another instance of F, in some (Goal) phrase β, Fβ, with which it agrees. Conceived of as feature sharing, Agree consists in replacing Fα with Fβ, so that the same feature is present in both locations. A link accessible throughout the derivation is thus established. By combining valuation and interpretability, Pesetsky & Torrego (2007) arrive at a fourfold typology of features, given below (feature participating in Agree will bear the same numerical index; an empty pair of brackets signals that a feature has not participated in Agree). (3) [uF] [1]: uninterpretable, valued [iF] [1]: interpretable, valued [uF] [ ]: uninterpretable, unvalued [iF] [ ]: interpretable, unvalued Adger & Svenonious’ (2011) propose a slightly different minimalist perspective on features, according to which a valued feature is a feature whose attribute is satisfied by a value chosen from a set of values, as in (4). (4)

Valued feature: a. A valued feature is an ordered pair where b. Att is drawn from the set of attributes, {A, B, C, D, E, . . . } c. and Val is drawn from the set of values, {a, b, c, d . . . }

If we correlate (3) with (4), there are two logical possibilities to supply a value for an attribute: Agree or Merge. Note that the second option, Merge, is not directly included in the system established by Pesetsky & Torrego (2007) (Ian Roberts, p.c.), but has been however advocated by several authors, most prominently by Rouveret (2012). Assume that, in UG, the D head is minimally endowed with an interpretable unvalued definiteness feature [idef _ ] and with interpretable unvalued φ-features [iφ _ ].

21

the parameter of definiteness in romanian

(5)

DP

D [idef _] [] [iφ _] []

NP

In structures with freestanding articles, the definiteness requirement on D is satisfied by the direct merger in D of the respective articles; e.g. the freestanding definite and indefinite articles of English (6a, b), or the freestanding indefinite article of Romanian (6c). (6) a. the b. a c. un a

man (English) man om (Romanian) man

These articles are lexically specified as uninterpretable valued [+ / – definite], as in (7), while the φ-feature requirements are satisfied via Agree with the head-noun. (7) DP D NP [i+def] [1] [iφ] [SG, MASC] [2] Δ

DP D NP [i-def] [1] [iφ] [SG, MASC] [2] Δ

the man a man (Engl.) [u+def] [1] [uφ] [SG, MASC] [2] un om (Ro.) [u-def] [1] [uφ] [SG, MASC][2] Affixes may also mark definiteness,2 either as suffixes, as in (8) for Romanian, Scandinavian varieties like Norwegian, Swedish, and Faroese (Julien 2005: ch. 2), or as a prefix, as in (9), for Hebrew (Danon 2010) or Arabic (Hoyt 2008). In these structures, the Agree relation established between the definite noun and the D head suffices to satisfy the requirements of D (Cornilescu & Nicolae 2  The realization of definiteness by means of affixation is richly attested cross-linguistically; in the WALS map devoted to definite articles, Dryer (2005a) documents 84 languages with definiteness affixation.

22

Nicolae

2011a), as shown in (10). Agree may or may not involve movement of the affixed noun to Spec, DP; this is subject to parametric variation (see the discussion in Ledgeway 2013: section 3.3). (8) a. carte-a book-DEF.F.SG ‘the book’

Romanian

b. skjort-a shirt-DEF.F.SG ‘the shirt’

Norwegian

c. hus-et house-DEF.N.SG ‘the house’

Swedish

d. kettlingur-in Faroese kitten-DEF.M.SG.NOM ‘the kitten’ (9) a. ha-sepr DEF-book ‘the book’

Hebrew

b. ʾal-kitāb-u DEF-book-NOM ‘the book’

Arabic

(10) DP D NP [i+def] [1] Δ [iφ] [val][2] carte-a / skjort-a / hus-et / kettlingur-in ha-sepr / ʾal-kitāb-u [u+def][1] [uφ] [val][2]

Agree

the parameter of definiteness in romanian

23

Cross-linguistically,3 indefiniteness may also be marked by grammatical affixes (11) (see Dryer 2005b); [idef _ ] valuation as [i-def] is assumed to proceed in a similar manner. (11) abül-fekha man-INDEF(SPECIFIC) ‘a (certain) man’

Korowai

However, not all features may be syntactically manipulated in the same manner. In the next section, we present Kibort’s (2010) typology of grammatical features and discuss the variable status of the definiteness feature. 2.2 Features. The Status of Definiteness Kibort (2010) puts forth a threefold typology of features and distinguishes between morphosemantic / morphological / morphosyntactic features, taking as criteria the availability of the respective feature to participate in agreement and government / assignment in a given language, and its semantic import. Morphosemantic features do not participate in agreement phenomena, but have semantic effects; semantic tense features likes [Past] / [Present] fall into this class. Morphological features do not participate in agreement phenomena and they do not have semantic implications; inflectional class features (e.g. [Conjugation I] of Latin verbs) are purely morphological features: they do not have semantic correlates, and they do not morphosyntactically infringe upon their dependents (i.e. there is no “agreement in conjugation” or the like). Morphosyntactic features are defined as features whose values are involved in either syntactic agreement or assignment / government. A few comments are in order with respect to the notion assignment / government. Kibort (2010) employs the concept government, also glossed as assignment, in the traditional GB sense: a governor is a head which assigns a certain grammatical value to a governee. This process is illustrated with Case assignment. As shown by Pesetsky & Torrego (2004, 2011) and Pesetsky (2013), Case assignment is actually resolved also by means of the Agree relation established by a nominal endowed with [uT] features against a [iT]-bearing head (a verb, a preposition, etc.). Thus, if both syntactic agreement and assignment / government are driven by Agree, then Kibort’s definition of morphosyntactic features can be simplified as follows: (12) Morphosyntactic features are features whose values are involved in Agree. 3  The map in Dryer (2005b: 160–161) documents 31 languages in which the indefinite article is affixal, all of them outside Europe.

24

Nicolae

This perspective on features neatly corresponds to Corbett’s (2011: 448) definition of canonical morphosyntactic features: (13) A canonical morphosyntactic feature is one that has robust formal marking and is manipulated by the rules of syntax. The affixal status of the definite article (in Romanian, Norwegian, Swedish, Faroese, Hebrew and Arabic) (see section 3.1 below for evidence that the Romanian definite article is a suffix and not a (second position) clitic) proves important in deciding whether the definiteness feature is morphosyntactic, as the inflectional realization of a certain feature is indicative of its status, as noted by Corbett (2011: 458): “another perspective on canonical [morphosyntactic] features [ . . . ] is that they be realized by canonical inflectional morphology”. Affixation represents a canonical mechanism of encoding morphosyntactic information. From this featural perspective, we can now turn to definiteness. It has been observed that “definiteness is a particularly difficult feature since the need for it, or not, in various languages is the source of some disagreement” (Kibort & Corbett 2010: 2). In Kibort’s (2010: 83) survey of the status of grammatical features across the world’s languages, definiteness is qualified as rarely participating in agreement; participation in assignment / government is not attested. Danon (2010) advocates the existence of a parameter of definiteness distinguishing languages like Hebrew and Arabic (in Danon’s formulation; we may add to this list at least Romanian, Norwegian, Swedish, and Faroese4), in which definiteness is a morphosyntactic feature, from languages like English and French, in which there is definiteness marking, but the definiteness feature is morphosemantic. Danon characterizes morphosyntactic definiteness as being a privative / monovalent feature: thus, the alternation is between having a [+definite] feature and lacking it, not between [+definite] and [-definite]. Lacking morphosyntactic definiteness is not equivalent to being semantically indefinite; semantic definiteness is supplied by freestanding articles or other definite determiners. The discussion above indicates that there is an important correlation between the realization of a grammatical feature and its status in the typology of grammatical features, which opens the possibility of a proper parameteriza4  This Romanian-Scandinavian similarity was first noticed by B.P. Hașdeu (1879 [1984]).

the parameter of definiteness in romanian

25

tion. With respect to definiteness, the affixal status of the definite marker alongside its morphosyntactic manifestation (participation in syntactic agreement) qualifies definiteness as a morphosyntactic feature in a given language. This correlation is supported empirically: a language like Bulgarian, in which the definite article is a second position clitic (see Dobrovie-Sorin & Giurgea 2006 and references therein), displays neither syntactic definiteness agreement of the sort found in Old Romanian, see (14) (Croitor 2008) and Modern Romanian, see (15) (Nicolae 2013a: ch. 3; cf. also Iordan 1956), Hebrew, see (16) (Wintner 2000) and Arabic, see (17) (Hoyt 2008); nor double definite constructions in which definiteness is marked by different elements (an affixal article and a freestanding article / a demonstrative determiner), found in Old Romanian, as in (18) (Stan 2013) and in Modern Romanian, as in (19) (Nicolae 2013b), or in Scandinavian, as in (20) (Julien 2005).

• Definiteness agreement (14) a. păntru sufletul răposatului jupânului Predei for soul.DEF late.DEF.GEN master.DEF.GEN Preda.GEN ‘for the soul of the late master Preda’ (DÎ.1600: LVI) b. zidul cetății marei și frumoasei wall.DEF city.DEF.GEN big.DEF.GEN and beautiful.DEF.GEN ‘the wall of the big and beautiful city, (Cantacuzino, apud Croitor 2008) (15) a. bietul bărbatul meu poor.DEF man.DEF my ‘my poor husband’ b. muncitorul ală vrednicul worker.DEF that hardworking.DEF ‘that hardworking worker’ (16) ha-sepr ha-gado (Hebrew) DEF-book DEF-big ‘the big book’ (17) ʾal-kitāb-u l-kabīr-u (Arabic) DEF-book.M.SG-NOM DEF-big.M.SG-NOM ‘the big book’

26

Nicolae

• Double definites (18) popa cela greșitul5 priest.DEF that make-mistake(PPLE).DEF ‘that priest who made a mistake’ (CPrav.1560–1562: 9r) (19) cartea cea / aceea interesantă book.DEF the that interesting ‘the / that interesting book’ (20) a. den gul-e skjort-a DEF yellow-W6 shirt-DEF ‘the yellow shirt’

(Norwegian)

b. det gul-a hus-et DEF yellow-W house-DEF ‘the yellow house’

(Swedish)

c. tann svart-i kettlingur-in (Faroese) DEF black-W kitten-DEF ‘the black kitten’ Recent theorizing indicates that affixation (in our case, the suffixation of the definite article) is a pre-syntactic phenomenon; in other words, lexical items may bear the definite affix at Merge. This idea is well-captured by Faarlund (2009), as in (21), and is further supported by Kibort (2010), who clearly indicates that inflection (and, implicitly, affixation) is a lexical property (22). (21) “inflectional forms of the word are created by morphosyntactic feature specification, rather than just added as separate morphemes” (Faarlund 2009: 623, building on Stump 2001 and Corbett 2006) (22) “the ‘rule’ that determines which elements have to realize particular inflections is found in the lexicon in the form of a generalization over the relevant part of speech or a subclass within a part of speech” (Kibort 2010: 69) 5  Old Romanian also has the option of active past participles with subject externalization; this option is lost in Modern Romanian, being replaced by a relative clause strategy (see Dragomirescu & Nicolae 2015b). 6  “ W” stands for weak inflection in the Scandinavian examples.

the parameter of definiteness in romanian

27

To sum up, the hypothesis that we entertain, which will be supported throughout the paper, is that the parameter of definiteness proposed by Danon (2010) translates as the ability of lexical items to be specified as [+definite] at Merge.7 The highly diversified patterns of definiteness agreement in Romanian will be shown to follow from this assumption in conjunction with the regular syntax of Romanian DP-internal constituents (adjectives, demonstratives and possessives). 3

On the Status and Merger of the Romanian Definite Article

3.1 The Suffixal Status of the Definite Article A series of distributional, inflectional as well as phonological characteristics indicate that the Romanian definite article is a suffix (Lombard 1974: 2; Halpern 1992; Ortmann & Popescu 2000; Dobrovie-Sorin & Giurgea 2006; Nicolae 2012a), not a second position clitic (Renzi 1993). 3.1.1 Distribution (i) In Modern Romanian, with the exception of the definiteness agreement constructions in (15), the definite article occurs in a constant position, i.e. on the first noun (23a) or adjective (23b) in the group. Its distribution is thus limited, being hosted only by constituents of a certain type, i.e. [+N] constituents. (23) a. fata girl.DEF

frumoasă c. *fată frumoasa beautiful girl beautiful.DEF

b. frumoasa fată d. *frumoasă fata beautiful.DEF girl beautiful girl.DEF ‘the beautiful girl’ A formal implementation of the suffixation of the definite article onto prenominal adjectives will be discussed in section 4.1 below. (ii) In coordination, the article attaches to both conjuncts, an unexpected repetition for a clitic (cf. Zwicky & Pullum 1983; Dobrovie-Sorin & Giurgea 2006).8 7  In a HPSG framework, Barbu (2004: 84, fnt. 48) also suggests that definite affixation is a Lexicon property. With reference to Swedish, Hankamer & Mikkelsen (2005) also adopt the idea that definite affixation is a pre-syntactic process. 8  Consider again the contrast with the Bulgarian definite article, which is a second position clitic (Dobrovie-Sorin & Giurgea 2006): the Bulgarian article occurs only once, on the highest adjective.

28

Nicolae

(24) a. frumosul și marele oraș beautiful.DEF and big.DEF city ‘the big and beautiful city’ b. *frumosul și mare oraș beautiful.DEF and big.DEF city By contrast, with stacked prenominal adjectives, the definite article surfaces on the highest one: (24) a. frumosul mare oraș beautiful.DEF big city ‘the big beautiful city’ b. *frumosul marele oraș beautiful.DEF big.DEF city (iii) The article is not always string-second in DP, since degree words may intervene between the D position and the adjectives to which the article attaches: (25) atât de lungile drumuri such of long.DEF roads ‘the very long roads’ (iv) In contrast to clitics, which exhibit a low degree of selection with respect to their hosts, affixes exhibit a high degree of selection with respect to their stems. The Romanian definite article attaches to nouns and φ-complete adjectives. Suffixation is blocked with cardinal numerals, which are φ-deficient

(i) a. novata

new.DEF b. *novata new.DEF (ii) a. *noua new.DEF b. noua new.DEF

i interesna and interesting i interesnata and interesting.DEF și interesantă and interesting și interesanta and interesting.DEF

kniga (Bulgarian) book kniga book carte (Romanian) book carte book

the parameter of definiteness in romanian

29

(26a), and definiteness valuation is satisfied by the merger of the freestanding definite article cel (‘the’) in D. (26) a. *cincile fete five.DEF girls b. cele cinci fete DEF five girls ‘the five girls’ With “quantifying adjectives” (quantifiers with adjectival morphology) (Pană Dindelegan 2003; Cornilescu 2009b), both patterns of definiteness valuation are available: (27) a. foarte puținii elevi de aici very few.DEF schoolchildren of here b. cei foarte puțini DEF very few ‘the very few schoolchildren here’

elevi de aici schoolchildren of here

The property which appears to be relevant is φ-completeness. Except for certain simple numerals which may encode gender distinctions (see Stan 2010 for extensive discussion), Romanian numerals are morphologically defective: they can neither bear the definite suffix (26), nor be inflected for Case; in the Genitive and Dative, quantified phrases headed by numerals are introduced by the prepositions a and, respectively, la (28). (28) a / la trei copii GEN DAT three children ‘of / to the three children’ In languages in which numerals are not morphologically impoverished (e.g. Modern Standard Arabic), they can bear Case inflection (29a) and be affixed by the definite article (29b) (Bardeas 2009: 37–38). (29) a. fataataani iṭnataani (Arabic) girls.F.NOM two.F.NOM ‘two girls’

30

Nicolae

b. an-nisaaɁu aṭ-ṭalaaṭatu DEF-women.F.NOM DEF-three.NOM ‘the three women’ 3.1.2 Inflection (i) Allomorphy. The definite article changes its form depending on the gender of the noun’s stem and on the last phoneme. For example, masculine nouns whose final segment is the singular inflectional ending -e (30a) take the allomorph -le (30b), while singular masculines with a final consonant (31a) reactivate the ancient singular inflectional ending -u and take the allomorph –l (31b). (30) a. frat-e brother-SG b. frat-e-le brother-SG-DEF (31) a. băiat boy b. băiat-u-l boy-SG-DEF Singular feminine nouns present an even more complex situation: the singular inflectional ending-ă (32a) is replaced by the definite allomorph -a (32b); the definite allomorph -a directly attaches (32d) to nouns whose singular inflectional ending is the full vowel -e (32c) (but not to those ending in a yod+[e] sequence like odaie [odaje] ‘small room’); by contrast, in the case of feminine nouns whose singular is marked by a silent allomorph (33e), the definite article allomorph is –ua (33f). (32) a. fat-ă girl-SG b. fat-a girl-DEF

the parameter of definiteness in romanian

31

c. cart-e book-SG d. cart-e-a book-SG-DEF e. pijama-Ø pyjama-SG f. pijama-Ø-ua pyjama-SG-DEF The suffixation of article onto prenominal adjectives strengthens the claim that the article is an allomorph of the stem on which it surfaces, not a clitic which raises to the topmost position of the DP. Singular masculines ending in a consonant take the allomorph -l (33a); when preceded by adjectives whose singular inflectional ending is -e (corresponding to the nouns in (30)), the definite allomorph is -le (34b). (33) a. băiat-u-l boy-SG-DEF

tenac-e tenacious-SG

b. tenac-e-le băiat tenacious-SG-DEF boy ‘the tenacious boy’ (ii) Genitive-Dative inflection. The definite article may shares the GenitiveDative inflectional endings with demonstrative, indefinite and relative pronouns. Compare the paradigms in (34) and (35). (34) a. băiat-u-lui (35) a. băieț-i-lor boy-SG-DEF.GEN/DAT boy-PL-DEF.GEN/DAT.PL b. acest-ui b. acest-or this-GEN/DAT.SG these-GEN/DAT.PL un-ui un-or one-GEN/DAT.SG one-GEN/DAT.PL căr-ui căr-or which-GEN/DAT.SG which-GEN/DAT.PL

32

Nicolae

3.1.3 Phonological Evidence The suffixation of the definite article has the effect of turning semi-vocalic endings into full vowels, thus altering the syllabic structure of nouns (36). (36) a. leu [leŭ] → leul[le-ul] lion.SG lion.DEF b. lei [leĭ] → lei[le-i] lion.PL lion.PL.DEF Definite affixation also has the reverse effect, that is, it may turn full vowels into semivowels (37). (37) floare [floare]→ floarea [floarěa] flower.SG flower.SG.DEF One other phonological effect of definite affixation is stress shift (38). (38) radio → radioul radio.SG radio.SG.DEF 3.1.4 Summary The tests presented in this section indicate that the Romanian enclitic definite article is a suffix, not a second position clitic. According to Corbett’s (2011) definitions, this item qualifies as the marker of the morphosyntactic [definiteness] feature. The lexical base to which the enclitic article attaches consists of adjectives and nouns (most inflectional classes of adjectives overlap with the inflectional classes of nouns; Brăescu 2013). 3.2 The Article Starts Out as a Suffix on the Noun There are numerous diachronic and synchronic empirical facts indicating that the Romanian definite article starts out as a suffix on the noun, and its (sometimes multiple) realization on adjectives represents a syntactic agreement phenomenon driven by Agree. Such an approach, developed in this section, allows one not only to properly account for definiteness agreement, but also to account for the situations in which definiteness agreement is not available, ultimately providing a correct characterization of the distribution of definiteness in Romanian.

the parameter of definiteness in romanian

33

3.2.1 The Low Definite Article of Old Romanian The first piece of evidence indicating that the definite article starts out as a suffix on the noun is the “low definite article” of Old Romanian, first discussed by Cornilescu & Nicolae 2011a). This construction is attested since the 16th century texts, and is present both in original documents and in translations. In this construction, the noun suffixed by the definite article may be preceded by constituents that in Modern Romanian count as interveners in definiteness valuation. These constituents are of several types: prenominal non-definite adjectives (39), prenominal genitives (40), and prenominal quantifiers (41); there are instances in which more than one constituent precedes the definite noun (42). (39) a. cu [DPcinstită cartea mării tale] with honoured letter.DEF highness.DEF.GEN your ‘with your highness’ honoured letter’ (DÎ.1596: CVI) b. tu tinde cumilă cătră noi [DPsvântă mâna ta] you extend with mercy towards us holy hand.DEF your ‘extend your holy hand towards us with mercy’ (FT.1570–1575: 3v) c. că văzuiu [DPluminată fața ta] that see.PS.1SG bright face.DEF your ‘that I saw your bright face’ (A.1620: 58r) d. s-au oploșit în [DPvicleană făgăduința lui] REFL=has sheltered in sly promise.DEF his ‘he took shelter in his sly promise’ (CLM.1700–1750: 186v) e. iară [DPascunsŭ giudéțul lui Dumnedzeu] and hidden judegement.DEF GEN God toate gândurile omenești le strămută all thoughts.DEF human CL.ACC.3PL moves ‘and God’s hidden judgement troubles all human thoughts’ (CLM. 1700–1750: 242r) f. să potoale Vasilie vodă [sunate zarvele] SUBJ mitigate Vasilie prince noisy quarrels.DEF ‘in order for Prince Vasilie to mitigate the noisy quarrels’ (CLM. 1700–50: 234v)

34

Nicolae

(40) a. au aflat cap și începătura moșilor[ . . . ] have.3= found head and beginning ancestors.DEF.GEN ca să nu se înéce [DP a toate țărâle so SUBJ not REFL drown GEN all countries.DEF anii trecuți] years.DEF passed ‘They found the origin and the beginning of their ancestors so that the history of all countries may not be drowned into oblivion’ (ULM.~1725: A-1 f.2) b. Umblăm după [DPa lumii înșelătoare fața] go.1PL after GEN world.DEF.GEN deceitful face.DEF ‘We are after the world’s deceitful face’ (CVL.1672: 34v) (41) a. deade Dumnezeu [DP zeace cuvintele sale] gave God ten words.DEF his ‘God gave his ten commandments’ (CCat.1560: 4r) b. arătarea [a dooa venireei lui] showing.DEF second coming.DEF.GEN his ‘the showing of his second coming’ (CC2.1581: 536/16) c. însă [câte trei morții] nu sunt închipuiri but each three deceased.DEF not are illusions ‘but the three deceased each are not illusions’ (AD.1722–1725: 131r) (42) închinra-se-vor înraintea-i [DPtoată a moșilor limba] bow=REFL=will before=DAT.1SG all GEN ancestors people.DEF ‘all the ancient people will bow before him’ (PH.1500–1510: 18r) The low definite article occurs overwhelmingly in configurations in which the definite noun is followed by a genitival phrase, a demonstrative adjective or another DP-internal modifier. Occurrence without postnominal dependents is rare, but not unattested (see (39f), (40b), (41c), (42)). Prenominal adjectives may take over the definite article in Old Romanian (43), similarly to Modern Romanian, and the multiple realization of the definite article is richly attested in Old Romanian as well (44) (Croitor 2008). (43) păn nu s-au săvârșit [DPsfânta slujbă] until not REFL=has ended sacred.DEF service ‘until the sacred service ended’ (CLM.1700–1750: 161v)

the parameter of definiteness in romanian

35

(44) naintea sfântului papei before holy.DEF.GEN Pope.DEF.GEN ‘before the holy Pope’ (DÎ.1593: XCII) As such, with non-definite prenominal adjectives, the realization of the suffixal definite article onto the post-adjectival noun is syntactically required for economy reasons: in Old as well as in Modern Romanian, post-nominal inflectional genitival phrases need not be introduced by the genitival marker al if they are strictly adjacent to the definite article (affixed onto a noun or an adjective); similarly, post-nominal demonstratives are also licensed in the presence of the definite article to their left. Thus, in the phrases in (40), it might be that prenominal adjectives undergo definiteness agreement with the noun and satisfy [idef _ ] on D via Local Agree, but the definite article is realized onto the lower noun for syntactic reasons. However, a larger picture, which takes into consideration quantifiers and prenominal genitives points to the fact that in Old Romanian it is indeed possible to value definiteness across an intervening phrasal constituent which may itself value definiteness (the prenominal genitive, the cardinal and ordinal numeral) or take over the definite suffix (the prenominal adjective). Consider the DP in (41a) and its derivation given in (45): definiteness valuation takes place via Long Distance Agree, across the intervening quantifier in D. Definiteness agreement of the quantifier and the lower realization of the definite suffix is not an option, as Romanian quantifiers are incompatible with the definite suffix or with other inflectional markers (see example (26) and the discussion surrounding it). (41) a. zeace cuvintele sale ten words.DEF his (45) DP D QP [i+def] [1] CardP Q’ Q NP [u+def] [2] Δ zece cuvintele

Agree

36

Nicolae

Given such facts, all the examples in (39)–(42) above, including those with a non-definite prenominal adjective, may be interpreted as displaying Long Distance Agree: a phrase situated between the probe in D (the [idef _ ] feature of D) and the goal lower down in the structure (the [u+def] feature of the noun) does not give rise to defective intervention effects. Cross-linguistically, there are other cases of definiteness agreement across an intervening phrase; consider (46) from Icelandic, which features the same Long Distance Agree option. (46) gamli maður-inn old man.DEF ‘the old man’ (from Stroh-Wollin 2009: 4) The conclusion that Agree could operate long distance in Old Romanian is further supported by the existence of Long Distance Movement in Old Romanian (Cornilescu & Nicolae 2011a). Long Distance Movement is shown by the existence of constructions in which specifiers can be crossed over by phrasal constituents. The examples in (47) below feature long phrasal demonstratives preceded by complex phrasal constituents; (48) illustrates prenominal determiner genitives (see Cornilescu & Nicolae 2011c on the determiner genitive of Old Romanian) crossed-over by complex adjectival phrases; in (49), an adjective suffixed by the definite article by-passes a cardinal numeral and reaches Spec, DP. Note also that the demonstrative adjective in (47a) and the prenominal genitives in (48) are the sole bearers of definiteness in the respective DPs; they too value definiteness by Long Distance Agree. (47) a. arătându-i [DPmare treabă aceasta] showing=him.DAT big affair this ‘showing him this big affair’ (CLM.1700–1750: 166v) b. pă [DP ticălosul pământu acesta] să vină DOM wretched.DEF earth this Să come.SUBJ ‘that he should come on this wretched earth’ (Greceanu.1711: 99) c. până la domniia lui aceasta until reign.DEF his this ‘until this reign of his’ (CLM.17s00–1750: 203v)

the parameter of definiteness in romanian

37

(48) a. [DPfrumos mirositoare a dragostei flori] a răsări sweetly smelling GEN love.DEF.GEN flower FUT spring ‘the sweet-smelling flower of love will spring’ (Cantemir.1705 II: 4) b. pre [DPmai mare a vicleșugului căptușală]i PE more big GEN guile.DEF.GEN hiding oi făcea it.ACC made.IMPERF.3SG ‘he resorted to a deeper hiding of his guile’ (= ‘he hid his guile deeper’) (Cantemir.1705 I: 307) (49) cântecele [DPsfinților trei feciori] songs.DEF sacred.DEF.GEN three sons ‘the song of the three sacred sons’ (CP1.1577: 307v) If internal merge is preceded by Agree (Chomsky 2000), then the examples (47)–(49) featuring movement across a phrasal specifier further testify to the availability of Long Distance Agree in the Old Romanian DP. In conclusion, the low definite article of Romanian indicates that the definite article starts out as a suffix on the noun. The Long Distance Agree option of Old Romanian allows the definite article to value the [idef _ ] feature of D across a potential intervener, and to remain on the noun when it is syntactically required to license post-nominal genitives, demonstratives or other modifiers (more rarely, it may remain on the noun even without being syntactically required to do so). The overall diachronic change from Old to Modern Romanian is the loss of the Long Distance Agree option in DPs. Thus, the low definite article of Old Romanian (examples (39)–(42)) has been totally eliminated; relics may be found in frozen phrases characteristic of the ecclesiastical style. The construction in (39), which features a non-definite adjective preceding a definite noun, has been replaced by structures in which the definite article surfaces on the prenominal adjective of the type illustrated in (43). Low definite article structures featuring a prenominal quantifier have been replaced by structures in which definiteness valuation is fulfilled by the freestanding definite article cel (Nicolae 2013a: ch. 3); the Modern Romanian examples in (50a) and (50b) correspond to the Old Romanian DPs in (41a) and (41b). (50) a. cele zece cuvinte ale sale the ten words GEN his ‘his ten words’

38

Nicolae

b. cea de-a doua venire a lui the second coming GEN his ‘his second coming’ The loss of this option also affected the constructions based on Long Distance Movement. Phrasal movement across demonstratives of the type exemplified in (47) is no longer permitted (see (51b), (51b’)), except for the restricted case of affective adjectives which authorize multiple definiteness realization in Modern Romanian (discussed in section 4.3 below); the standard option of Modern Romanian is head-movement of definite nouns across the phrasal demonstrative (51a) (Cornilescu 2005). (51) a. treaba aceasta mare affair.DEF this big ‘this big affair’ b. *[treaba mare] aceasta b’. *[marea treabă] asta affair.DEF big this big.DEF affair this Structures of the type (48) and (49) have been completely lost in Modern Romanian. 3.2.2 Morphologically Defective Adjectives and Definiteness Realization Let us now turn to the Modern Romanian data indicating that the definite article merges low, as a suffix on the noun. Romanian possesses a restricted class of morphologically defective (i.e. φ-defective) adjectives with superlative meaning and evaluative semantics which cannot take over the definite suffix: ditamai, ditai, cogeamite, coșcogeamite ‘huge, very big’. Despite being obligatorily prenominal, these adjectives cannot be inflected for definiteness (52b) like regular, non-defective adjectives (52a) and trigger definiteness realization on the post-adjectival head noun (52c). However, they do not block definiteness valuation, and thus contrast with cardinal numerals, which count as interveners in definiteness valuation by a definite noun (53a) and call for the insertion of the freestanding definite article cel ‘the’ (53b). (52) a. frumoasele (φ: F.PL +def) beautiful.DEF ‘the beautiful houses’

case (F.PL) houses

the parameter of definiteness in romanian

b. *cogeamitele very-big.DEF

39

(φ: −, *def) case (F.PL) houses

c. cogeamite (φ: −) very-big ‘the very big houses’

casele (F.PL, +def) houses.DEF

(53) a. *două casele two houses b. cele două case the two houses ‘the two houses’ Definiteness realization on the lower noun with defective adjectives also constitutes evidence that the definite article merges as an affix on the noun. 3.3 Summary In this section, we brought evidence for two distinct but related claims: (i) the Romanian definite article is part of the inflection of Romanian nominals; in other words, Romanian nominals may be lexically specified for definiteness; (ii) of the two possible candidates that may be suffixed by the definite article at Merge, i.e. nouns and adjectives, it appears that nouns enter the derivation in this manner, and the realization of the definite article on adjectives represents a syntactic agreement phenomenon. 4

Definiteness: Spell Out and Agreement

The results of the previous section allow us to investigate, in this section, how definiteness is substantiated in Old and Modern Romanian DPs. In particular, the conclusion that definiteness is involved in syntactic agreement phenomena driven by the Agree is taken as an indication that definiteness is a morphosyntactic feature in Romanian—which is the line of analysis we pursue. 4.1 Spelling Out Definiteness on Prenominal Adjectives As already mentioned, the definite article may surface on prenominal adjectives, in both Old (54a) and Modern Romanian (54b).

40

Nicolae

(54) a. păn nu s-au săvârșit [DPsfânta slujbă] until not REFL=has ended sacred.DEF service ‘until the sacred service ended’ (CLM.1700–1750: 161v) b. frumosul copil beautiful.DEF child ‘the beautiful child’ In what follows, we illustrate definiteness marking on prenominal adjectives with Modern Romanian examples; the distributional features discussed below are characteristic of Old Romanian as well. Definiteness marking on prenominal adjectives represents a phenomenon of agreement in definiteness. The position of the adjectival head in the DP is highly relevant: not all configurations with adjectival modification allow definiteness agreement. Actually, only attributive (qualifying or intensional) prenominal adjectives can undergo definiteness agreement (55). In the Modern Romanian DP, the definite article generally surfaces only once, on the highest [+N] constituent, see (55b–c); DP-internal adverbials (i.e. [-N] categories) can precede the constituent with the enclitic definite article and do not trigger (defective) intervention effects, see (55a). Note that intensional adjectives (like former, poor, etc.) are exclusively prenominal, see (55d, e).9 (55) a. ( foarte) frumosul trandafir very beautiful.DEF rose ‘the very beautiful rose’ b. *frumos trandafirul beautiful rose.DEF c. *frumosul trandafirul beautiful.DEF rose.DEF

9  Certain prenominal adjectives like sărac may also be post-nominal, but the reading is different; e.g., post-nominal sărac ‘impecunious’ versus prenominal evaluative sărac ‘poor, pitiable’; evaluatives are exclusively prenominal, e.g. fost (‘former’), viitor (‘future’) or biet (‘pitiable).

the parameter of definiteness in romanian

41

d. fostul șef former.DEF boss ‘the former boss’ e. *șeful fost boss.DEF former Post-nominal adjectives, be they qualifying (56a) or relative (56c), are excluded from definiteness agreement, see (56b, d).10 (56) a. trandafirul frumos rose.DEF beautiful ‘the beautiful rose’ b. *trandafirul frumosul rose.DEF beautiful.DEF c. comedie americană comedy.DEF American ‘the American comedy’ d. *comedia americana comedy.DEF American.DEF Predicative adjectives do not undergo definiteness agreement either (57): (57) a. Trandafirul este frumos. rose.DEF is beautiful ‘The rose is beautiful’ b. *Trandafirul este frumosul. rose.DEF is beautiful.DEF c. *Trandafir este frumosul. rose is beautiful.DEF

10  The class of relative adjectives includes thematic (e.g. British victory) and classifying (e.g. romantic poem) adjective (see Cornilescu & Nicolae 2011b for details).

42

Nicolae

Focusing on prenominal adjectives, we point out that there is a relevant difference between coordination and stacking (see also Scott 2002): with stacked prenominal adjectives, the article surfaces only once, on the highest adjective, as in (58a) versus (58b) and (58c), whereas, with coordinated adjectives, the article surfaces on both conjuncts (59). (58) a. marele bătrân continent, Europa big.DEF old continent Europe ‘the big old continent, Europe’ b. *marele bătrânul continent, Europa big.DEF old.DEF continent Europe c. *mare bătrânul continent, Europa big old.DEF continent Europe (59) a. lunga și călduroasa vară long.DEF and warm.DEF summer.DEF ‘the long and warm summer’ b. *lunga și călduroasă vară long.DEF and warm summer.DEF c. *lungă și călduroasa vară long and warm.DEF summer.DEF The relevance of the disparity between stacking and coordination proves relevant, as stacking involves hierarchical structure (Scott 2002), whereas coordination is essentially a phenomenon of symmetry (Munn 1993). In order to derive definiteness agreement, we assume that definiteness is among the agreement features of the adjective. This affirmation finds support first and foremost in the inflectional behaviour of adjectives (see Brăescu 2013: 410–414). Modern Romanian possesses four inflectional classes of adjectives: four-form adjectives, three-form adjectives, two-form adjectives, invariable adjectives. All these adjectival inflectional classes dissolve into the inflectional classes of the noun, as also remarked by Brăescu (2013: 410): “[a]djectives share a series of inflectional affixes with nouns, and the variation of the stem generally displays morpho-phonological alternations common to both classes”. The inflectional relation between adjectives and nouns is actually a set-superset

the parameter of definiteness in romanian

43

relation: there is no adjectival inflexion which is not present in the noun set of inflexions, but not vice versa.11 On morphological grounds, the assumption that definiteness is among the agreement features of the adjective is thus well supported. The definiteness feature on adjectival heads will be unvalued uninterpretable [definite] and the feature set contains the unvalued uninterpretable [φ] and [definiteness]. According to Cinque (2010), prenominal adjectives merge as specifiers of functional projections. Consider example (55a), repeated here, whose internal structure is given in (60)(only the relevant features are present in the tree). The noun enters the derivation with the suffixed definite article. (55) a. frumosul trandafir beautiful.DEF rose ‘the beautiful rose’ (60) DP D FP [idef] [] [uφ] [] AP F’ [uφ] [] [udef][] F NP [iφ] [] [u+def] [] Δ frumos trandafirul beautiful rose.def In this configuration, the adjective c-commands the noun and takes over the values of the matching features on the noun. In its turn, the adjectival head equipped with matching features is probed by the features on the D-head: Agree between the higher c-commanding D-head and the lower c-commanded adjective ensures that the features of D get valued, and the uninterpretable features on the adjective and the noun get deleted, so the derivation converges.

11  As also remarked by Brăescu (2013: 410): “there also exist dissimilarities between the inflection of the noun and that of the adjective: some inflectional endings [of nouns] (-uri for plural, -o for vocative) never occur with adjectives”.

44

Nicolae

(61) DP D FP [i+def] [2] [uφ] [1] AP F’ [uφ] [1] [u+def] [2] F NP [iφ] [1] [u+def] [2] Δ frumosul trandafir Agree Agree The derivation of stacked adjectives featuring the definite article on the DP-initial adjective (62a) is similar (62b) (only the final step of the derivation is shown): successive Agree relations ensure that the [+definite] specification is transmitted upwards from the definite noun to the highest prenominal adjective; Agree between D and the highest prenominal adjective ensures well-formedness. (62) a. marele bătrân continent, Europa big.DEF old continent Europe ‘the big old continent, Europe’ b. DP D FP [i+def] [2] [uφ] [1] AP F’ [uφ] [1] [u+def] [2] F FP AP [uφ] [1] [u+def] [2] F marele bătrân Agree

Agree

F’ NP [iφ] [1] [u+def] [2]

Δ

continent Agree

the parameter of definiteness in romanian

45

This analysis makes important predictions that are supported by our data: (i) Predicative postcopular adjectives do not display definiteness agreement. If we agree that copular sentences involve a small clause (den Dikken 2006 and references therein) in which the specifier is occupied by a DP subject and the complement position is occupied by the predicative adjective, then at no point in the derivation, prior to the valuation of D features, does the predicative adjective c-command the noun. Therefore, Agree between the adjective and the noun fails, and definiteness agreement does not take place. This holds for both Modern (Dragomirescu 2013) and Old Romanian (Guruianu 2005: 31–35; Carabulea 2007; Frâncu 2009: 162–165; Dragomirescu mss. for the 16th– 18th centuries; Zafiu 2012: 303–305 for the 19th century). (ii) It is important to distinguish between the presence of the definite article and the valuation of definiteness. While definiteness is phonologically realized only once, on the highest [+N] constituent of the phrase, it is present on all [+N] heads in the extended projection of the noun. Evidence for the presence of the definiteness feature on all [+N] heads is given by the existence of multiple realization of the definite article, both in Old Romanian (section 4.2) and in Modern Romanian (section 4.3). 4.2 Multiple Definiteness Marking in Old Romanian So far, the analysis accounts for the relatively high frequency of multiple definite constructions in Old Romanian, and for the distribution of multiple definites (Croitor 2008; Stan 2013): the multiple marking of definiteness takes place in structures with prenominal adjectives, i.e. in A + N structures (63a,b) or in stacked adjectival A + A + N structures (63c); the structures of the form N + A in which both the N and the A bear the article are very rare (attested only with DPs in the Genitive Case) (64). (63) a. și naintea [sfântului papei] and before saint.DEF.GEN pope.DEF.GEN ‘and before the saint pope’ (DÎ.1600: XCII) b. Pănă la [cumplita domniia lui Aaron vodă] until at terrible.DEF reign.DEF GEN Aaron prince ‘until Prince’s Aaron terrible reign’ (CLM.1700–1750: 159r) c. moaștiile a [sfintei prepodobnei Paraschevei] relics.DEF GEN saint.DEF beautifully-adorned.DEF Parascheva.GEN ‘the relics of the holy beautifully-adorned Parascheva’ (CLM.1700–1750: 235r)

46

Nicolae

(64) zidul [cetății marei și frumoasei] wall.DEF city.DEF.GEN big.DEF.GEN and beautiful.DEF.GEN ‘the wall of the big and beautiful city’ (Cantacuzino, apud Croitor 2008) The fact that the multiple realization of the definite article occurs almost exclusively with prenominal adjectives verifies the mechanism proposed above to derive the realization of the definite article in Modern Romanian: the definiteness feature percolates upwards via Agree from the lower definite noun to the higher c-commanding adjectives, being thus present as an agreement feature on all prenominal [+N] heads. Rare examples like in (64), where multiple definiteness marking occurs with post-nominal adjectives, are not problematic: if DP-internal adjectives merge as specifiers of prenominal FPs (Cinque 2010), then the structures featuring post-nominal adjectives are derived via movement of the definite noun across the adjective. Thus, before the completion of the DP-phase and the valuation of definiteness, the adjective and the noun are in a c-command configuration, ensuring definiteness agreement. The contrast between Old and Modern Romanian lies in the spell out of the definite article. In Old Romanian, the article may be phonologically realized on all these heads, while in Modern Romanian spell out is restricted to the DP-initial adjectives. This is presumably related to the diachronic shift form analyticity to syntheticity present throughout the history of Romance and Romanian, characterized in the nominal domain by a tendency towards the single marking of the grammatical categories on the first element of the DP (Repina 1971; Stan 2008).12 The multiple spell out of the definite article briefly discussed here was present throughout the entire period of Old Romanian (Croitor 2008: 213). Certain relics survive at the beginning of the 19th century (Nicolae 2012b: 117). 4.3 Multiple Definiteness Spell Out in Modern Romanian In this section, we show that the multiple spell out of the definite article is still available in some contexts in Modern Romanian. This is highly significant, as it validates the idea that definiteness agreement targets all the prenominal [+N] heads in the extended projection of the DP.

12  In the verbal domain, the change from syntheticity to analyticity is most prominently visible in the replacement of the Latin synthetic verb forms by the modern Romance analytic verb forms; in the adjectival domain, the change from synthetic to analytic has affected the marking of intensity, with the synthetic markers being replaced by analytic marker (see Ledgeway 2012: ch. 2 for details).

the parameter of definiteness in romanian

47

The multiple spell out of the definite article in Modern Romanian is triggered by post-nominal demonstratives and by post-nominal genitives or possessive adjectives. The examples below illustrate two different patterns: the first pattern, in (65), features a definite A + a definite N + a post-nominal demonstrative / possessive adjective or genitive DP; the second pattern, in (66), illustrates a definite N + a post-nominal demonstrative + a definite A. (65) a. săracul băiatul ăla pitiable.DEF boy.DEF that ‘that pitiable boy’ b. bietul bărbatul meu / ei / Mariei poor.DEF man.DEF my her(GEN) Mariei.GEN ‘my poor husband’ (66) a. muncitorul ăla vrednicul worker.DEF that hardworking.DEF ‘that hardworking worker’ b. fata asta mica girl.DEF this little.DEF ‘this little girl’ In the remainder of this section, we focus on the post-nominal demonstrative constructions, and keep in mind that the same analysis can be extended to the possessive adjective / genitive DP pattern, which is derivationally similar (modulo the different projection hosting the genitive / possessive adjective). The relevant interpretative characteristic of both patterns in (65a), (66) is that they are pragmatically marked: in both cases, the nominal phrase expresses the speaker’s evaluation. Intensive research on the Romanian demonstratives has shown that, from a functional perspective (Tasmowski 1990; Manoliu 2000; Cornilescu 2005; Vasilescu 2009), the post-nominal demonstrative is an emphatic element (possibly endowed with a specificity feature), and thus behaves like a focus. The presence of evaluative adjectives in these structures is thus not surprising. As shown by previous research (Cornilescu 1992; Dimitrova-Vulchanova & Giusti 1998; Cornilescu 2005), the derivation of the simple post-nominal demonstrative construction (67) standardly relies on head-movement of the definite noun across the demonstrative in the case of Romanian.

48

Nicolae

(67) a. muncitorul acesta worker.DEF this ‘this worker’ b. DP D’ D FP [idef] [uφ] DemP F’ [+EPP] | Dem F NumP [u+def] [uφ] Num NP | N [u+def] [iφ] muncitorul acesta Without going into details regarding the derivation of the two definiteness agreement patterns presented above, it follows from structure (67b) that both are derived by movement across the demonstrative. The essential point is that only adjectives that are or may be prenominal (intensional and qualifying adjectives) are allowed to occur in these constructions. The first pattern (definite A + definite N + post-nominal demonstrative) in (65a) allows only intensional, evaluative adjectives which are restricted to the prenominal position (biet ‘piteous’, pretins ‘alleged’), see (68), or which display intensional, evaluative meanings only in prenominal position (sărac ‘piteous’ versus ‘poor, penniless, impecunious’), as in (69). (68) a. bietul băiat piteous.DEF boy ‘the pitiable boy’ b. *băiatul biet boy.DEF piteous

the parameter of definiteness in romanian

49

(69) a. săracul băiat piteous.DEF boy ‘the piteous boy’ b. băiatul sărac boy.DEF penniless ‘thepenniless / impecunious boy’ The derivation of the first multiple definiteness pattern in (65a) can be accounted for as follows: the prenominal adjective c-commands the noun and agrees with it, and thus takes over the morphosyntactic specification for definiteness. Subsequently, there is phrasal movement of the [definite A + definite N] sequence across the demonstrative, yielding the word order A > N > Dem. The obligatory presence of the definite article on both constituents is due to economy considerations (similar to those responsible for the realization of the low definite article in Old Romanian; see section 3.2. above): on the one hand, post-nominal demonstratives are licensed only in the presence of the definite article to their left (thus, the definite article has to surface on the noun in order to license the post-nominal demonstrative); on the other hand, the locality condition for Agree forces the article to surface on the prenominal adjective as well, in order to satisfy featural requirements on D (Local Agree). The second pattern, in (66), with the sequence definite N + post-nominal demonstrative + definite A) allows for typical qualifying adjectives which may be placed both prenominally (70b) and post-nominally (70a), with no differences in meaning; however, prenominal adjectives encode speaker evaluation. (70) a. muncitorul vrednic worker.DEF hardworking b. vrednicul muncitor hardworking.DEF worker ‘the hardworking worker’ In prenominal position, as in (70b), these adjectives take over the definite article, c-commanding the noun. The prenominal configuration (71b) feeds the multiple definiteness configuration in (66). As in the simple [definite noun + demonstrative] construction in (67), there is head movement of the definite noun across the definite adjective and across the demonstrative, yielding the order [definite N > Dem > definite A]. The mechanics of this derivation is more

50

Nicolae

complex: in order for the definite noun to crossover the adjective placed in Spec, FP, an equidistant derivation in the sense of Lasnik (2009) has to apply. The noun undergoes movement to the F0 head (in whose Specifier we find the definite adjective). At this point, the adjective and the noun are in the same minimal domain and equidistant from the target of movement,13 since head movement extends the domain of a head (Chomsky 1995;14 Roberts 2011). A natural question at this point is why the reverse order, [definite A > Dem > definite N], is not possible (as in (71)). In other words, what makes the definite noun a better candidate for movement than the definite adjective, since they are both equidistant to the target? (71) *vrednicul ală muncitorul hardworking.DEF that worker.DEF The solution adopted by Cornilescu (2005) in the derivation of the post-nominal demonstrative construction can be successfully extended to the construction analyzed here as well. For example, in (72), the two candidates for movement have a different structural status: the noun is a head, while the demonstrative is phrasal (for evidence that post-nominal demonstratives are phrasal, while prenominal demonstratives are heads, see Cornilescu 2005; Nicolae 2013c). The preference for head movement (N0) over phrasal movement (AP movement in this case) follows from an economy principle requiring the piping of only as much material as is needed for convergence (Pied Pipe Less Weight, Stateva 2002; or Attract / Move Smallest, Akiyama 2004). (72) muncitorul ăla vrednic worker.DEF that hardworking.DEF ‘that hardworking worker’ Mention should also be made of the fact that the post-nominal adjective structure in (70a) feeds a construction without definiteness agreement (this is because post-nominal adjectives do not c-command the noun, so there is no agreement). More interestingly, the two minimally different structures correlate with the expected interpretative contrast: while the structure with multiple

13  As in Lasnik (2009): “If α, β are in the same minimal domain, they are equidistant from γ”. 14  Chomsky (1995) formulates this idea with respect to verbal head movement.

the parameter of definiteness in romanian

51

definiteness marking in (66a) expresses focus and speaker evaluation, the structure in which the definite suffix is expressed once, as in (72), expresses only focus. This interpretative contrast validates our assumptions with respect to which structure lies at the basis of the post-nominal demonstrative patterns: structure (70a), with a post-nominal adjective which does not express speaker evaluation, constitutes the basis for (72) (in which the definite suffix is realized only once), while the basis for (66a) is (70b), a structure in which the adjective expresses speaker evaluation. To conclude, the multiple spell out of the definite article is triggered by the presence of the post-nominal demonstrative in the structure, which are licit only in the presence of the definite article to their left. Compare the examples in (73). (73) a. muncitorul acesta worker.DEF this ‘this worker’ b. *muncitor acesta worker.DEF this The same reasoning applies to constructions of type in (65b), where the definite article surfaces both on the prenominal adjective and on the post-adjectival noun when the A + N sequence is followed by a possessive adjective. It is well known that, in Romanian, possessive adjectives or genitival DPs may be introduced directly only in the presence of the definite article to their left (74a); in the case of non-adjacency (resulting either from the fact that the selecting head is indefinite (74b) or from the presence of an intervening constituent (74c)), the insertion of the genitival marker al is required. (74) a. caietul fetei / meu notebook.DEF girl.DEF.GEN my ‘the girl’s / my notebook’ b. un caiet al fetei / al meu a notebook GEN girl.DEF.GEN GEN my ‘a notebook of the girl / of mine’ b’. *un caiet fetei / meu a notebook girl.DEF.GEN my

52

Nicolae

c. caietul de matematică al fetei / al meu notebook.DEF of mathematics GEN girl.DEF.GEN GEN my ‘the girl’s / my mathematics notebook’ c’. *caietul de matematică fetei / meu notebook.DEF of mathematics girl.DEF.GEN my All these facts amount to the conclusion that the multiple spell out of the definite article in the post-nominal demonstrative construction and in the possessive adjective / genitival DP construction is due to economy considerations: in the former construction, the lower definite article licenses the post-nominal demonstrative, while in the latter, it permits the direct insertion of the possessive adjective / genitive DP, without the mediation of the genitival marker AL. Constraints of the similar type also license the low definite article of Old Romanian (section 3.2.1 above): recall that the spell out of the definite article on a lower noun occurs overwhelmingly in the presence of post-nominal genitive / possessive or of a post-nominal demonstrative (see the statistics in Cornilescu & Nicolae 2011a: 208). At the same time, the strengthening of the locality conditions on the application of Agree requires that the definite article be spelled out on the first [+N] constituent of the DP whenever possible.15 The multiple marking of the definite article is, of course, possible due to its presence on all the [+N] constituents of the DP which c-command the noun. 4.4 Summary In this section, we have discussed the following phenomena: definiteness marking on DP-internal prenominal adjectives and multiple definiteness marking in Old and Modern Romanian. We have shown that these data can be easily accommodated by the analysis put forth in the previous section: the definite article starts low, as a suffix on the noun, and its propensity towards the left edge of the DP is syntactically driven by Agree.

15  Violations of this condition occur in two situations, discussed in the previous sections: (i) with prenominal quantifiers, which block definiteness valuation and call for the insertion of the freestanding article cel (see (26) in section 3.2.1); (ii) with φ-defective adjectives, which cannot inflectionally realize the definite article, but do not block definiteness valuation (see (52) in section 3.2.2).

the parameter of definiteness in romanian

53

5 Conclusions In this paper, we argued for the following analysis: (i) The definite article is a suffix in Romanian; nouns are lexically equipped with the definite suffix at Merge. From this perspective, the traditional idea that there exists a “definite declension” in Romanian (see Rizescu 1966: 85–88) is justified. (ii) The realization of the definite article on prenominal adjectives, a pattern characteristic for both Old and Modern Romanian, represents a syntactic agreement phenomenon, driven by Agree. The fact that definiteness is manipulated by the basic rules of syntax indicates that it is a morphosyntactic feature in the sense of Kibort (2010), Danon (2010), Corbett (2011, 2012). (iii) The multiple spell out of the definite article signals that, despite the typical presence of the article on the first [+N] constituent of the DP, the article is computed on all the [+N] heads in the extended projection of the noun. A second, lower spell out of the definite article is required syntactically, to introduce post-nominal demonstratives and genitive DPs / possessive adjectives without the mediation of the genitival marker AL. From this perspective, the definite article is involved in “licensing”: Romanian thus helps complete Kibort’s (2010) characterization of definiteness as a morphosyntactic feature, as it illustrates an option (participation in government / licensing) which, according to Kibort (2010: 83), is not attested.16 The morphosyntactic status of definiteness allows one to properly parameterize this feature. The view on parameterization adopted here is the microparametric one (see footnote 2); the source of parametric variation is the Borer-Chomsky Conjecture: (75) The Borer–Chomsky Conjecture (Baker 2008: 353) All parameters of variation are attributable to differences in the features of particular items (e.g. the functional heads) in the lexicon. Thus, definiteness is a lexical property of nominals, present as a private morphosyntactic feature in the lexical make up of Romanian nominals (and also, Hebrew, Arabic, Norwegian, Swedish, etc.). All the languages specified for this

16  Recall that the table in Kibort (2010: 83) qualifies definiteness as rarely participating in agreement; the participation of definiteness in government/licensing is described as not attested.

54

Nicolae

parameter present a cluster of properties: definiteness agreement, double definite constructions, multiple definiteness spell out. (iv) The main diachronic change from Old to Modern Romanian in the nominal domain is the restriction of Agree, from Long Distance Agree to Local Agree. This change has led to the disappearance of the low definite article of Old Romanian (see section 3.2.1 above) and to the general condition that determiners occupy the DP-initial position. Another consequence of this change is the emergence of the freestanding article cel in contexts in which φ-defective quantifiers act as defective interveners and block definiteness valuation (see the discussion surrounding example (50)). Yet another consequence is the disappearance of the structures based on Long Distance Movement (illustrated in (47)–(49)): if Move is preceded by Agree (Chomsky 2000), then the disappearance of Long Distance Movement is predicted by the disappearance of Long Distance Agree. Acknowledgements The first version of this paper has been presented at the Surrey Morphology Group in March 2012; I am grateful to Greville G. Corbett and Antonio Fortin for the invitation and for their constructive comments. I would also like to express my gratitude to the following colleagues whose comments have been extremely insightful: Alexandra Cornilescu, Adina Dragomirescu, Virginia Hill, Adam Ledgeway, Gabriela Pană Dindelegan, Ian Roberts, Camelia Stan and Rodica Zafiu. This paper is supported by the Sectorial Operational Programme Human Resources Development (SOP HRD), financed from the European Social Fund and by the Romanian Government under the contract number SOP HRD/159/1.5/S/136077. Corpus A.1620

Zgraon, F. (ed.). 2005. Alexandria. Bucharest, Fundația Națională pentru Ştiință și Artă. AD.1722–1725 Ştrempel, G. (ed.). 1997. Antim Ivireanul, Didahii. In Antim Ivireanul, Opere. Bucharest: Minerva, p. 3–210. Cantacuzino Constantin Cantacuzino, Istoria Ţării Românești, Litera Chișinău, 1997 (apud Croitor 2008).

the parameter of definiteness in romanian

55

Cantemir.1705

Verdeș, I. & P.P. Panaitescu (eds.). 1983. Dimitrie Cantemir, Istoria Ieroglifică. Bucharest: Minerva. CC2.1581 Pușcariu, S. & A. Procopovici (eds.). 1914. Diaconul Coresi, Carte cu învățătură. Bucharest: Atelierele Grafice Socec & Co. CCat. 1560. Roman-Moraru, A. (ed.). 1982. Catehismul lui Coresi. In I. Gheție (coord.), Texte românești din secolul al XVI-lea. Bucharest: Editura Academiei, p. 21–97. CLM.1700–1750 Panaitescu, P.P. (ed). 1958. Miron Costin, Letopisețul Ţării Moldovei. In: Opere. Bucharest: ESPLA, p. 41–201. CP1.1577 Toma, S. (ed.). 1976. Coresi, Psaltirea slavo-română (1577) în comparație cu psaltirile coresiene din 1570 și din 1589. Bucharest: Editura Academiei. CPrav. 1560–1562 Chivu, Gh. (ed.). 1982. Coresi, Pravila. In I. Gheție (coord.), Texte românești din secolul al XVI-lea. Bucharest: Editura Academiei, p. 218–230. CVL.1672 Panaitescu, P.P. (ed). 1958. Miron Costin, Viața lumii. In Opere. Bucharest: ESPLA, p. 318–326. DÎ Chivu, G.; M. Georgescu; M. Ioniță; A. Mareș; A. RomanMoraru (eds.). 1979. Documente și însemnări românești din secolul al XVI-lea. Bucharest: Editura Academiei. FT.1570–1575 Gheție, I. (ed.). 1982. Fragmentul Todorescu (Carte de cântece). In I. Gheție (coord.), Texte românești din secolul al XVIlea. Bucharest: Editura Academiei, p. 332–43. Greceanu.1711 Piru, A. (ed.). 1964. R. Greceanu, Cronica. In Al. Piru (ed.), Cronicarii Munteni. Bucharest: Editura Tineretului. PH.1500–1510 Gheție, I. & M. Teodorescu (eds.). 2005. Psaltirea Hurmuzaki. Bucharest: Editura Academiei Române. PO.1582 Pamfil, V. (ed.). 1968. Palia de la Orăștie. Bucharest: Editura Academiei Române. ULM.~1725 Panaitescu, P.P. (ed). 1955. Grigore Ureche, Letopisețul Ţării Moldovei. Bucharest: ESPLA, p. 57–209. References Abney, Steven Paul. 1987. The English Noun Phrase in its Sentential Aspect. Doctoral dissertation, MIT.

56

Nicolae

Adger, David, Peter Svenonius. 2011. Features in minimalist syntax. In Cedric Boeckx (ed.), The Oxford Handbook of Linguistic Minimalism. 27–51. Oxford: Oxford University Press. Akiyama, Masahiro. 2004. Multiple nominative constructions in Japanese and economy. Linguistic Inquiry 34(4): 671–683. Baker, Mark C. 2008. The macroparameter in a microparametric world. In Theresa Biberauer (ed.), The Limits of Syntactic Variation. 351–373. Amsterdam, John Benjamins. Barbu, Ana-Maria. 2004. Sintaxa determinatorilor. Analiză lingvistică și computațională. Bucharest: All. Bardeas, Suzanne Mahmoud. 2009. The Syntax of the Arabic DP. Doctoral dissertation, University of York. Borer, Hagit. 2005. Structuring Sense: vol. I. In Name Only. Oxford: Oxford University Press. Brăescu, Raluca. 2013. Adjectives and adjectival phrases. In Gabriela Pană Dindelegan (ed.), The Grammar of Romanian. 410–431. Oxford: Oxford University Press. Carabulea, Elena. 2007. Predicatul nominal. In Mioara Avram (ed.). Sintaxa limbii române în secolele al XVI-lea–al XVIII. 46–53. Bucharest: Editura Academiei Române. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Massachusetts: MIT Press. ———. 2000. Minimalist inquiries: the framework. In: Roger Martin, David Michaels & Juan Uriagereka (eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik. 89–156. Cambridge, Massachusetts: MIT Press. ———. 2001. Derivation by phase. In Michael J. Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge, Massachusetts: MIT Press. Cinque, Guglielmo. 2010. The Syntax of Adjectives. Cambridge, Massachusetts: MIT Press. Corbett, Greville G. 2006. Agreement. Cambridge: Cambridge University Press. ———. 2011. The penumbra of morphosyntactic feature systems. Morphology 21: 445–480. ———. 2012. Features. Cambridge: Cambridge University Press. Cornilescu, Alexandra. 1992. Remarks on the determiner system of Rumanian: the demonstratives al and cel. Probus 4(3): 189–260. ———. 2003. Romanian genitive constructions revisited. Bucharest Working Papers in Linguistics 5(1): 45–70. ———. 2005. Romanian demonstratives and minimality. Bucharest Working Papers in Linguistics 7(1): 102–116. ———. 2006. Modes of semantic combination: NP/DP adjectives. In Jenny Doetjes, Paz González (eds.), Romance Languages and Linguistic Theory 2004. 43–71. Amsterdam: John Benjamins.

the parameter of definiteness in romanian

57

———. 2009a. On the linearization of adjectives in Romanian”. In Danièle Torck, W. Leo Wetzels (eds.), Romance Languages and Linguistic Theory 2006. 33–52. Amsterdam: John Benjamins. ———. 2009b. Measure phrases and the syntax of Romanian nouns and adjectives. Bucharest Working Papers in Linguistics 11(1): 35–66. Cornilescu, Alexandra, Alexandru Nicolae. 2011a. On the syntax of Romanian definite phrases: changes in the patterns of definiteness checking. In Petra Sleeman, Harry Perridon (eds.), The Noun Phrase in Romance and Germanic. Structure, Variation, and Change. 193–221. Amsterdam: John Benjamins. ———. 2011b. Nominal peripheries and phase structure in the Romanian DP. Revue roumaine de linguistique 56(1): 35–68. ———. 2011c. On the history of Romanian genitives: the prenominal genitive. Bucharest Working Papers in Linguistics 13(2): 37–55. ———. 2012. Nominal ellipsis as definiteness and anaphoricity. Lingua 122: 1070–1111. Croitor, Blanca. 2008. Aspecte privind acordul în determinare în limba română veche. In Gabriela Pană Dindelegan (ed.), Limba română. Dinamica limbii, dinamica interpretării. 213–218. Bucharest: Editura Universității din București. Danon, Gabi. 2010. The definiteness feature at the syntax-semantics interface. In Anna Kibort, Greville G. Corbett (eds.), Features. Perspectives on a Key Notion in Linguistics. 144–165. Oxford, Oxford University Press. Den Dikken, Marcel. 2006. Relators and Linkers. The Syntax of Predication, Predicate Inversion, and Copulas. Cambridge, Massachusetts: MIT Press. Dimitrova-Vulchanova, Mila, Giuliana Giusti. 1998. Fragments of Balkan nominal structure. University of Venice Working Papers in Linguistics 8(1): 141–172. Dobrovie-Sorin, Carmen, Ion Giurgea. 2006. The suffixation of definite articles in Balkan languages”. Revue roumaine de linguistique 61(1): 73–103. Dragomirescu, Adina. 2013. The subjective predicative complement. In Gabriela Pană Dindelegan (ed.), The Grammar of Romanian. 160–166. Oxford: Oxford University Press. ———. mss. The subjective predicative complement. In Gabriela Pană Dindelegan (ed.), The Syntax of Old Romanian. Oxford: Oxford University Press. (in preparation) Dragomirescu, Adina, Alexandru Nicolae. 2015a. 55 Case. In Adam Ledgeway, Martin Maiden (eds.), The Oxford Guide to the Romance Languages. Oxford: Oxford University Press. (to appear) ———. 2015b. L’ellipse nominale avec article défini de l’ancien roumain au roumain moderne: Le cas du participe passé. In Éva Buchi, Jean-Paul Chauveau & Jean-Marie Pierrel (eds), Actes du XXVIIe Congrès international de linguistique et de philologie romanes (Nancy, 15–20 juillet 2013), 3 volumes. Strasbourg: Société de linguistique romane / ÉliPhi. (to appear)

58

Nicolae

Dryer, Matthew S. 2005a. 37 Definite articles. In: Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures (WALS). 154–157. Oxford: Oxford University Press. ———. 2005b. 38 Indefinite articles. In: Martin Haspelmath, Matthew S. Dryer, David Gil & Bernard Comrie (eds.), The World Atlas of Language Structures (WALS). 158– 161. Oxford: Oxford University Press. Faarlund, Jan Terje. 2009. On the history of definiteness marking in Scandinavian. Journal of Linguistics 45(3): 617–639. Frampton, John, Sam Gutmann. 2006. How sentences grow in the mind: Agreement and selection in an efficient minimalist syntax. In Cedric Boeckx (ed.), Agreement Systems. 121–157. Amsterdam: John Benjamins. Frâncu, Constantin. 2009. Gramatica limbii române vechi (1521–1780). Jassy: Casa Editorială Demiurg. Gallego, Ángel J. 2011. Parameters. In Cedric Boeckx (ed.), The Oxford Handbook of Linguistic Minimalism. 523–550. Oxford: Oxford University Press. Giusti, Giuliana. 2005. At the left periphery of the Romanian noun phrase. In Martine Coene, Liliane Tasmowski (eds.), On space and time in language. 23–49. ClujNapoca: Clusium. Guruianu, V. 2005. Sintaxa textelor românești originale din secolul al XVI-lea. Sintaxa propoziției. Bucharest: Editura Universității din București. Halpern, Aaron. 1992. “The Balkan definite article and pseudo-second position”. In L. Buszard-Welcher, L. Wee & W. Weigel (eds), Proceedings of the Eighteenth Annual Meeting of the Berkeley Linguistics Society. 338–349. Berkeley: Berkeley Linguistics Society. Hankamer, Jorge, Line Mikkelsen. 2005. When movement must be blocked: a reply to Embick and Noyer. Linguistic Inquiry 36(1): 85–125. Hasdeu, Bogdan Petriceicu. 1879 [1984]. O pagină din sintaxa româno-albaneză. Reduplicarea și triplicarea articolului hotărât. In Cuvente din bătrâni, vol. II, 1879; ed. by Gh. Mihăilă. Bucharest: Editura Didactică și Pedagogică, 432–481. Hoyt, Frederick. 2008. The Arabic noun phrase. In: The Encyclopedia of Arabic Language and Linguistics. Leiden: Brill. Iordan, Iorgu. 1956. Limba română contemporană. Bucharest: Editura Ministerului Învățământului. Julien, Marit. 2005. Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins. Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge, Massachusetts: MIT Press. ———. 2013. Comparative syntax. Lingua 130: 132–151.

the parameter of definiteness in romanian

59

Kibort, Anna. 2010. Towards a typology of grammatical features. In Anna Kibort, Greville G. Corbett (eds.), Features. Perspectives on a Key Notion in Linguistics. 64–106. Oxford: Oxford University Press. Kibort, Anna, Greville G. Corbett. 2010. Introduction. In Anna Kibort, Greville G. Corbett (eds.), Features. Perspectives on a Key Notion in Linguistics. 1–13. Oxford: Oxford University Press. Lasnik, H. 2009. Shortest Move and Equidistance. http://ling.umd.edu/~lasnik/ LING611%202009’/Equidistance.pdf Ledgeway, Adam. 2012. From Latin to Romance. Morphosyntactic Typology and Change. Oxford: Oxford University Press. Ledgeway, Adam. 2013. Greek disguised as Romance? The case of Southern Italy. In Mark Janse, Brian D. Joseph, Angela Ralli & Metin Bagriacik (eds.), Proceedings of the 5th International Conference on Greek Dialects and Linguistic Theory. 184–228. UPatras: Laboratory of Modern Greek Dialects, University of Patras. Lombard, A. 1974. La langue roumaine. Une presentation. Paris: Klincksieck. Longobardi, Giuseppe. 1994. Reference and proper names: a theory of N-movement in syntax and logical form. Linguistic Inquiry 25(4): 609–665. ———. 2001. The structure of DPs: some principles, parameters and problems. In Mark Baltin, Chris Collins (eds.), The Handbook of Contemporary Syntactic Theory. 562–603. Oxford: Blackwell. Manoliu, Maria. 2000. Demonstratives, story-world and talk-interaction. In Martine Coene, Walter De Mulder, Patrick Dendale, and Yves D’Hulst (eds.), Traiani Augusti Vestigia Pressa Sequamur. Studia Linguistica in Honorem Lilianae Tasmowski. 583– 600. Padua: Unipress. Munn, Alan. 1993. Topics in the Syntax and Semantics of Coordinate Structures. Doctoral dissertation, University of Maryland. Nicolae, Alexandru. 2012a. Gramaticalizarea articolului hotărât românesc. Noi rezultate. Limbă și literatură 53 (1–2): 7–19. ———. 2012b. Secolul al XIX-lea. Articolul. In Gh. Chivu, Gabriela Pană Dindelegan, Adina Dragomirescu, Isabela Nedelcu & Irina Nicula (eds.), Studii de istorie a limbii române. Morfosintaxa limbii literare în secolele al XIX-lea și al XX-lea. 111–120. Bucharest: Editura Academiei Române. ———. 2013a. Types of Ellipsis in Romanian. The interpretation of structures containing ellipsis sites and the syntactic licensing of ellipsis. Doctoral dissertation, University of Bucharest & University of Cambridge. ———. 2013b. Polydefinite structures. In Gabriela Pană Dindelegan (ed.), The Grammar of Romanian. 318–319. Oxford: Oxford University Press. ———. 2013c. Demonstratives. In Gabriela Pană Dindelegan (ed.), The Grammar of Romanian. 294–300. Oxford: Oxford University Press.

60

Nicolae

———. 2013d. Notă de sintaxă comparată: parametrul [+definit] în sintaxa grupului nominal românesc. Limba română 62(2): 186–209. Ortmann, Albert, Alexandra Popescu. 2000. Romanian definite articles are not clitics. In: Birgit Gerlach, Janet Grijzenhout (eds.), Clitics in Phonology, Morphology and Syntax. 295–324. Amsterdam: John Benjamins. Pană Dindelegan, Gabriela. 2003. Elemente de gramatică. Dificultăți, controverse, noi interpretări, Bucharest: Humanitas. Pesetsky, David. 2013. Russian Case Morphology and the Syntactic Categories. Cambridge, Massachusetts: MIT Press. Pesetsky, David, Esther Torrego. 2004. Tense, Case, and the nature of syntactic categories. In Jacqueline Guéron & Jacqueline Lecarme (eds.). The Syntax of Time. 355– 426. Cambridge Massachusetts, London, England: The MIT Press. ———. 2007. The syntax of valuation and the interpretability of features. In: Simin Karimi, Vida Samiian & Wendy K. Wilkins (eds.), Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation. 262–294. Amsterdam: John Benjamins. ———. 2011. Case. In Cedric Boeckx (ed.), The Oxford Handbook of Linguistic Minimalism. 55–72. Oxford: Oxford University Press. Renzi, Lorenzo. 1993. L’articolo posposto rumeno in diacronia e in sincronia. Revue roumaine de linguistique 38(4): 307–323. Repina, Tamara. 1971. Caracterul sintagmatic al declinării românești. Studii și cercetări lingvistice 22(5): 459–469. Rizescu, I. 1966. Substantivul. In Al. Graur, Mioara Avram, Laura Vasiliu (eds.), Gramatica limbii române, vol. I. 55–94. Bucharest: Editura Academiei. Roberts, Ian. 2011. Head movement and the minimalist program. In Cedric Boeckx (ed.), The Oxford Handbook of Linguistic Minimalism. 195–219. Oxford: Oxford University Press. ———. 2012. Macroparameters and minimalism. A programme for comparative research. In Charlote Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo & Juantio Avelar (eds.), Parameter Theory and Linguistic Change. 320–335. Oxford: Oxford University Press. Rouveret, Alain. 2012. VP ellipsis, phases and the syntax of morphology. Natural Languages and Linguistic Theory 30(3): 897–963. Scott, Gary-John. 2002. Stacked adjectival modification and the structure of nominal phrases. In Guglielmo Cinque (ed.), Functional Structure in DP and IP: the Cartography of Syntactic Structures, vol. 1. 91–120. Oxford: Oxford University Press. Stan, Camelia. 2008. Grupul nominal românesc (aspecte diacronice). In Gabriela Pană Dindelegan (ed.), Limba română. Dinamica limbii, dinamica interpretării. 239–243. Bucharest: Editura Universității din București.

the parameter of definiteness in romanian

61

———. 2009. Ierarhia constituenților în grupul nominal. In Rodica Zafiu, Blanca Croitor, Ana-Maria Mihail (eds.), Studii de gramatică. Omagiu Doamnei Profesoare Valeria Guțu Romalo. 231–236. Bucharest: Editura Universității din București. ———. 2010. On the grammaticality status of numerals in Romanian. Revue roumaine de linguistique 55(3): 237–246. ———. 2013. Sulla sintassi dei sintagmi nominali con più determinant nel rumeno. In Emili Casanova Herrero, Cesáreo Calvo Rigual (eds.), Actas del XXVI Congreso Internacional de Lingüística y de Filología Románica, volumen II. 397–407. Berlin: De Gruyter. Stateva, Penka. 2002. Possessive clitics and the structure of nominal expressions. Lingua 112: 647–690. Stroh-Wollin, Ulla. 2009. On the development of definite markers in Scandinavian. Working Papers in Scandinavian Syntax 83: 1–25. Stump, Gregory T. 2001. Inflectional Morphology: A Theory of Paradigm Structure. Cambridge: Cambridge University Press. Tasmowski, Liliane. 1990. Les démonstratifs français et roumains dans la phrase et dans le texte. Langages 97: 82–99. Tănase-Dogaru, Mihaela. 2009. The Category of Number. Its Relevance for the Syntax and the Semantic Typology of the Nominal Phrase. Bucharest: Editura Universității din București. Vasilescu, Andra. 2009. Strategii pragmatice de reluare, gramaticalizate ca relații apozitive de tip GN1−GN2. Limba română 58(2): 275–284. Wintner, Shuly. 2000. Definiteness in the Hebrew Noun Phrase. Journal of Linguistics 36 (2): 319–363. Zafiu, Rodica. 2012. Secolul al XIX-lea. Sintaxa. In Gh. Chivu, Gabriela Pană Dindelegan, Adina Dragomirescu, Isabela Nedelcu & Irina Nicula (eds.), Studii de istorie a limbii române. Morfosintaxa limbii literare în secolele al XIX-lea și al XX-lea. 283–378. Bucharest: Editura Academiei Române. Zwicky, Arnold M., Geoffrey K. Pullum. 1983. Cliticization vs. inflection: English n’t. Language 59(3): 502–513.

Polydefinite DPs in Old Romanian Daniela Isac This paper focuses on Old Romanian DPs in which the definite article surfaces more than once: on the noun and on another item, typically an adjective. While the literature analyzes these DPs as involving multiple agree with the definiteness feature on the D head, I propose that Old Romanian made use of a feature that is no longer syntactically active in Modern Romanian, namely referentiality. Thus the analysis I propose computes over two morphosyntactic features: definiteness and referentiality. Both of these are unvalued features of D and the proposal is that the definite article is spelled out disjointly both on items that probe D to check/value the [def] feature, and on items that check/value the [Ref] feature on D. One advantage of this approach over a multiple agree approach is that long distance agree, which was posited for Old Romanian in previous work on definite DPs, can be dispensed with.

1 Introduction This paper focuses on Old Romanian DPs in which definiteness is phonologically expressed more than once: on the noun and on another item within the same DP, typically an adjective, as in (1). (1)

să te lipsești de dorita fața lui SUBJ you.REFL= deprive of desired.DEF face.DEF of Dumnezeu God ‘to deprive yourself of the cherished face of God’ (Antim {353})

In this paper I will refer to these DPs as polydefinite DPs, following Alexiadou & Wilder (1998), Lekakou & Szendrői (2007), Cinque (2010), a. o. 1.1 Background The existing literature on Romanian definite DPs (e.g., Cornilescu & Nicolae 2009, 2011; Cornilescu 1995, 2004; Croitor 2008; Dobrovie-Sorin 1987; DobrovieSorin & Giurgea 2006; Grosu 1988; Pană-Dindelegan 2008) proposes that definiteness is a morphosyntactic feature that is shared by determiners, nouns and adjectives, although the nature of this feature differs across these categories. In Cornilescu & Nicolae’s work (see also this volume), definiteness is assumed © koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_004

Polydefinite DPs in Old Romanian

63

to be an uninterpretable but valued feature on the N(oun) ([udef:+]),1 an interpretable (though unvalued) feature on the D(eterminer) ([idef: ]), and an uninterpretable, unvalued feature on A(djectives) ([udef: ]). N agrees with A and values the latter’s [udef:] feature as ‘+’. The [idef:] feature on D gets valued either by entering Agree with A, as in (2a), represented in (2c), or by Agree with N, as in (2b), represented in (2d). (2) a. frumosul trandafir beautiful.DEF rose ‘the beautiful rose’ b. trandafirul frumos rose.DEF beautiful ‘the beautiful rose’ c. DP D FP [idef:+] AP

F’

A F NP [udef:+] N frumosul [udef:+] trandafir Agree

1  Throughout the paper I will follow Pesetsky & Torrego (2007) and assume that the interpretable/uninterpretable distinction is independent from valuation. Thus, interpretable features can be valued or unvalued and similarly, uninterpretable features can be valued or unvalued. The notation I will use indicating the value of a feature is [feature:val]. If no value is indicated, that means the respective feature is unvalued, as in [feature: ]. Moreover, I will specify for each feature whether it is uninterpretable or interpretable by using the letters ‘u’ and ‘i’ respectively in front of the name of that feature, as in [ufeature:val] or [ifeature:val].

64

Isac

d. DP  D NP [idef:+] NP AP N [udef:+]

A

trandafirul frumos Agree In Modern Romanian the probe that values the definiteness feature on D has to be local (a [+N] constituent immediately below D). The two representations in (2c, d) are thus relevant for Modern Romanian: in (2c) the [idef:] feature on D is valued by the AP, which is the [+N] constituent closest to D, while in (2d) the [idef:] feature on D is checked by agree with N, which in this case is the [+N] constituent closest to D. However, in Old Romanian the definiteness feature on D could be valued by a distant goal according to Cornilescu & Nicolae (2009, 2011). In other words, in a representation like (2c), it could be either the [udef:+] feature on the AP or the [udef:+] feature on the NP that values the [idef:] feature on D, irrespective of the word order. When the goal is the adjective, the definite determiner is spelled out as a suffix on the adjective, and when the goal is the noun, the definite determiner is suffixed to the noun. Examples of both situations are given in (3). In (3a, c) D is adjacent to the goal, while in (3b, d) it is not. (3) a. au purces fără numai din vechea has= acted without only from old.DEF și rânceda pizmăluire and rancid. DEF envy ‘He acted out solely from the old and rancid envy’ (Cantemir I 67) b. că mare scrâșnetul roatelor that great grinding.DEF wheels.DEF.GEN ‘that the strong grinding of the wheels’ (Cantemir I 67)

65

Polydefinite DPs in Old Romanian

c. La cinstita mănă a Mării Sale to honest.DEF hand of highness.DEF.GEN his.GEN lui Ghiorgu the.GEN Ghiorgu ‘of the honest hand of his highness Ghiorgu’ (PH 1339)

a of

d. La cinstită măna Măriei Sale to honest hand.DEF highness.DEF.GEN his.GEN birăului de Bistriță mayor.DEF.GEN of Bistrita ‘of the honest hand of his highness the mayor of Bistrița’ (PH.1340) Moreover, Old Romanian also allowed for DPs in which the definite determiner is spelled out on both the local goal (i.e. the closest [+N] constituent that can value the [idef:] feature on D) and the more distant one, as in (4). (4) înfricoșata zioa aceaia a răsplătirii scary.DEF day.DEF that of doom.DEF.GEN ‘that scary doomsday’ (Antim {392}) Croitor (2008) and Nicolae (2013) distinguish between two types of polydefinites in Old Romanian: (a) DPs in which the definiteness feature on D is the only probe and there are several potential goals that can value this feature within the DP: in this case the definiteness feature is phonologically realized on multiple sub-constituents of the DP as a result of multiple agreement with one and the same D head (either by local Agree or by distant Agree); and (b) DPs which contain multiple definiteness probes (multiple Ds), with different syntactic and interpretive properties, each of which has its own goal. The definiteness feature is phonologically realized on multiple DP sub-constituents as a result of independent Agree relations with independent (but multiple) Ds. 1.2 Aim of the Paper In this paper I argue that at least some of the instances of polydefinites that have been analyzed as involving multiple agree in Old Romanian are actually instances that involve two different features of D: definiteness [def] and referentiality [Ref]. I will thus argue that determiners carry a referentiality feature on top of the definiteness feature and that these two features can be valued by one and the same item in the DP or else by independent items. When the latter occurs, the result is a polydefinite DP in which the definite D is spelled

66

Isac

out twice, once on the item that values the [idef:] feature on D and once on the item that values the [iRef:] feature on D. If this is on the right track, then long distance agree can be dispensed with and the differences explaining the existence of polydefinites in Old Romanian, but not in Modern Romanian, could be accounted for by the hypothesis that [Ref] used to be a syntactically active feature in Old Romanian, but ceased to be so in Modern Romanian. This account would be more in line with the Minimalist assumptions that linguistic variation is restricted to the lexicon, more precisely to the featural properties of lexical items. 2 Data Old Romanian polydefinite DPs are made up of two terms in very limited cases, as in (5), but otherwise these DPs always involve a third term, as in (6). (5) a. întru lauda a Preacuratei Ficioarii Mariei in praise.DEF of pure.DEF.GEN virgin.DEF.GEN Mary.GEN ‘In praise of the pure Virgin Mary’ (Ureche. 6978 [1470]) b. în zioa sfintei blagoiavleniei in day.DEF.GEN saint.DEF.GEN baptism.of.the.Lord.DEF.GEN ‘the day of the saint baptism of the Lord’ (LC {224}) c. Iar diavolul diavolul nu putu and enemy.DEF devil.DEF not could ‘and the evil devil could not bear’ (LC {195})

răbda bear

What is striking about two-term polydefinite DPs is that nouns that occur in these contexts designate unique individuals: they are either proper nouns, as in (5a), or common nouns that refer to single individuals or entities in the world, as in (5b,c).2 Notice that in addition to the noun, the prenominal adjective also has the definite article suffixed onto it, which is what makes these DPs polydefinite. In polydefinite DPs that contain three terms, the definite article is also overtly realized on both the noun and a preverbal adjective, just as in twoterm polydefinite DPs. However, in three-term polydefinite DPs the noun is 2  Even though the DP in (5a) contains 3 words, Fecioarii Mariei counts as one constituent, i.e. as a complex Proper Name. Thus, the DP in bolded letters in (5a) is a two-term DP.

Polydefinite DPs in Old Romanian

67

accompanied by a genitive DP, as in (6); a post-nominal demonstrative phrase, as in (7); a modifier, as in (8); or a complement to the noun, as in (9). (6) a. nemăsurata mila lui Dumnezeu boundless.DEF mercy.DEF of God ‘the boundless mercy of God’ (Antim {160}) b. mărg la svânta icoana lui Hristos de să-nchină go.3PL at saint.DEF icon.DEF of Christ to REFL=-pray ‘they go to the saint icon of Christ to pray’ (Dosoftei {16}) c. pentru buna stâmpărarea vânturilor for good.DEF calming.DEF winds.DEF.GEN ‘for the good calming of the storm’ (Dosoftei {59}) (7) pentru svânta lăcuința aceasta for saint.DEF home.DEF this ‘for this saint home’ (Dosoftei {38}) (8) a. la luminatul craiul leșescu at enlightened.DEF prince.DEF Polish ‘at the enlightened Polish prince’ (DÎR, XVIII, Scrisoare, 5 sept. [1599]) b. cetății marei și frumoasei în Spania fortress.DEF.GEN big.DEF.GEN and beautiful.DEF.GEN in Spain ‘of the big and beautiful fortress in Spain’ (CC {37}) (9) la sfânta purtătoarea de viață a at holy.DEF bearer.DEF DE life of mântuitoriului nostru Iisus Hristos savior.DEF.GEN our Jesus Christ ‘to the holy mother of Jesus Christ’ (Antim {105}) Sometimes the noun could be accompanied by more than one of the elements listed above, as in (4), both a demonstrative and a genitive phrase follow the noun. The DPs in (6)–(9) can be split into two groups depending on whether the overt realization of the definite article on the noun is obligatory or not in those contexts. In particular, the suffixation of the definite article on the noun is optional when the noun is accompanied by a modifier or a complement. Compare in this sense (8), in which the definite article is suffixed both onto the

68

Isac

prenominal adjective and onto the noun, and (10), in which the definite article is overtly realized on the adjective only. (10) pentru marile faceri de bine for big.DEF doings DE good ‘for the big good deeds’ (Antim {87}) The same optionality can be observed with mono-definite DPs: the definite article can be overtly realized on the noun, as in (11a,b) or not, as in (11c). (11) a. tăietorul de pietri cutter.DEF DE stones ‘stone-cutter’ (Cheia {4r}) b. trupul omenesc body.DEF human ‘the human body’ (Cheia {18r}) c. adevărări luminate truths enlightened ‘enlightened truths’ (Cheia {2r}) On the other hand, in all the examples of polydefinite DPs, including genitive phrases and demonstratives that I found, the definite article is overtly realized on the noun, which seems to indicate that the definite article must be overt in these cases. The same observation is supported by examples of mono-definite DPs: whenever a genitive phrase or a post-nominal demonstrative is present, the D must be suffixed on the noun in an overt way.3 Interestingly, the pattern described above can be observed with objects of prepositions and with examples involving non-local agree in Old Romanian. When occurring alone, objects of prepositions can be bare nouns in Old Romanian.4 (12) a. poarta pre carea scotea din cetate gunoaele gate through which took.3SG out.of city garbage.DEF ‘the gate through which he would take the garbage out of the fort’ (Antim {302}) 3  As will become apparent below, only certain genitive phrases support this generalization, namely the so-called ‘anchoring’ ones. 4  This option became the rule in Modern Romanian.

69

Polydefinite DPs in Old Romanian

b. să propoveduiască la oameni venirea SUBJ preach.SUBJ.3 to men arrival.the ‘to preach to the men about his arrival’ (Antim {4})

lui his

The alternative is that the definite article can be overtly realized on the noun, as in (13) and (14); this happens in the same environments in which the article is suffixed on the noun in polydefinite DPs, as described above. In particular, the object of propositions can be the host of an overt definite article if (i) the noun has unique reference, as in (13); (ii) the noun is accompanied by a genitive phrase, a demonstrative, a modifier or a complement, as illustrated in (14). (13) a. s-au închinat la soltanul REFL=have bowed to sultan.DEF ‘they bowed to the Sultan’ (Amiras {302r}) b. scris-au la împăratul cum nu va fi written-has to empreror.DEF that not will.3SG be craiul să margă prince.DEF SUBJ go.SUBJ.3 ‘they wrote to the Emperor that the Prince was not going to come’ (Amiras {265r}) (14) a. pentru suspinurile săracilor Genitive DP for sighs.DEF poor.DEF.GEN ‘for the suffering of the poor’ (Neculce 107) b. în cuptoriul acela in oven.DEF that ‘in that oven’ (Antim {295})

Demonstrative DP

c. în mănăstirea armenească AP Modifier in monastery.DEF Armenian ‘in the Armenian monastery’ (Neculce 107) Moreover, the same optionality observed in the case of polydefinite DPs including modifiers and complements can be noticed with objects of pre­ positions. The definite article can be overtly realized on the noun when a modifier or complement is present, as illustrated in (14c, d) or not, as illustrated in (15).

70

Isac

(15) a. cu milă îndestulată with mercy abundent ‘with abundant mercy’ (Antim {346}) b. pentru slujbe ce-au făcut for liturgies that-have.3.PL= done ‘for the liturgies they have held’ (Neculce 153) c. în vreame de nevoe in time DE need ‘in times of need’ (Antim {307}) Similarly, non-local Agree takes place in the same type of situations. The definite article can be suffixed onto a noun across an intervening item if (i) the noun has unique reference, as in (16); (ii) the noun is accompanied by a genitive phrase, a demonstrative, a modifier or a complement, as in (17). (16) cu fericită sănătatea să să with happy health.DEF SUBJ REFL= ‘may he have happiness and health’ (PH 1193)

dea give.SUBJ.3

dumisalea him

(17) a. ca să nu se înece a toate that SUBJ not REFL= drown of all țările anii trecuți countries.DEF years.DEF past ‘so that the past years of all countries may not be drowned (into oblivion)’ (Ureche 57) b. înpreaună cu fericită casa Mării together with happy house.DEF highness.DEF.GEN Tale your.DEF.GEN ‘together with the happy house of your highness’ (PH 1399) 3 Proposal The proposal is that all types of polydefinite DPs contain nouns that are referential. Two-term polydefinites contain a noun that is intrinsically referential,

Polydefinite DPs in Old Romanian

71

whereas three-term polydefinite DPs contain a noun whose reference is anchored by a genitive phrase, a demonstrative, a modifier or a complement. The syntax of these polydefinite DPs computes not only over definiteness [def] features, but also over referentiality [Ref] features. The basic intuition is that genitive phrases, demonstratives, modifiers and complements are in some sense deictic and they help anchor the reference of the noun. More specifically, I propose that the D head has an interpretable but unvalued [iRef: ] feature, and the noun has a matching [Ref] feature as well.5 Names and nouns with a unique designator have an uninterpretable but valued [Ref] feature, while the rest of the nouns have and uninterpretable and unvalued [Ref] feature. In the latter cases, the value of the [uRef:] feature of the noun is supplied by the genitive phrase, the demonstrative, the modifier or the complement. Genitive phrases and post-nominal demonstratives have an intrinsic referential feature which is uninterpretable but valued as ‘+’ (i.e. [uRef:+]), whereas modifiers and complements carry this value optionally (i.e. they can be either [uR:+], just as genitive phrases and demonstratives, or can have an unvalued but interpretable [iRef: ] feature). Given this split between genitive phrases and demonstratives, on the one hand, which obligatorily carry a valued [Ref] feature, and modifiers and complements, on the other hand, which only optionally carry such a feature, nouns which are accompanied by genitive phrases or demonstratives will always have their [uRef:] feature valued, while nouns which are accompanied by modifiers or complements will have their [uRef:] feature valued only in some cases. The [uRef:] feature on the noun, which gets valued by the [Ref] feature of the genitive phrase, demonstrative, modifier or complement, is still uninterpretable and needs to find a match to check off its uninterpretability. This match is provided by D, which carries an interpretable but unvalued [iRef: ] feature. Checking will result in erasing the uninterpretability of the [Ref] feature on N. I will adopt Pesetsky & Torrego’s (2007) proposal that interpretability and valuation are distinct from each other (see footnote 1). While Pesetsky & Torego (2007) argue that interpretability and valuation should be dissociated from each other in the lexicon, I propose that the two are also distinguished in the computation and that the two are processed independently by the grammar. In particular, feature valuation can take place independently from checking 5  The idea that D carries referentiality features is not new. Many researchers have proposed that it is the D head that contributes to turning the NP into an argument; see Stowell (1989), Longobardi (1994), Giusti (1996, 2005), Roehrs (2006), among others.

72

Isac

(and erasing) an uninterpretable feature. In Pesetsky & Torrego’s (2007) view it is the need for feature valuation that initiates the search for a matching feature and that drives Agree. Moreover, the directionality of this search is always top-down in their view. Thus, Agree is initiated by some head at a location α (the probe), that has an unvalued, (uninterpretable or interpretable), feature Fα; and the head searches in its phase (the DP in our case) for another instance of F, a goal at location β, Fβ, with which it agrees. However, if search and Agree are exclusively for valuation purposes, the probe itself could in principle have either interpretable or uninterpretable features (as long as the respective feature is unvalued), and the goal that is found could also in principle have interpretable or uninterpretable matching features (as long as they are valued). Given the particular feature composition of the relevant heads that we described above, the D head is the probe searching for a match, since it is the D head that has unvalued features. Notice though that D has interpretable (even if unvalued) features. The goal that D finds must thus have a matching feature that is valued, but in principle that matching feature could be interpretable or uninterpretable. The claim of this article is that the grammar distinguishes between these two situations, and thus that the grammar is sensitive to (un) interpretability independently of feature valuation. The more specific rule that I claim captures this distinction is stated in (18). (18) Spell out of the definite article in Old Romanian The definite article is spelled out as a suffix on an item X within the DP iff (i) D has an unvalued feature F that probes for a matching valued feature F, and finds feature F on X, and (ii) X has an uninterpretable feature F and D provides an interpretable match for the feature F on X. What (18) says is that the definite article is spelled out on a head X in the DP domain only if D and X act as each other’s matches (a feature of X is the match/ goal of an unvalued feature on D and D’s feature simultaneously acts as the match for the same uninterpretable feature on X).6 6  Notice that this proposal is different from Reverse Agree. In other words, I am not claiming that Agree can be initiated by a feature in a domain that is higher than the respective feature. I am instead adopting the view that Agree is unidirectional, namely top-down, and thus that an unvalued feature can search for a valued matching feature only downwards, in its c-command domain. Crucially, I am assuming that only unvalued features initiate search and Agree and that uninterpretable features, although they must be eliminated, don’t initiate search and Agree. However, uninterpretable features do require matching (interpretable)

Polydefinite DPs in Old Romanian

73

Since in our case D probes for N (given that D’s [Ref] feature is unvalued and N has a matching [Ref] feature that is valued), and at the same time D provides a match for N (given that N has an uninterpretable [Ref] feature and D has a matching interpretable [Ref] feature), D and N are each other’s matches and D will get spelled out on the noun. The definite article is spelled out twice in these structures because D has two unvalued features that probe for agreeing valued features: apart from the [iRef:] feature, which probes for the [uRef:+] feature on the noun, D also has an [idef:] feature, which agrees with the uninterpretable [udef] feature on the prenominal item (usually an adjective). The article is spelled out on the adjective because the adjective has an uninterpretable [def] feature and the [def] feature on D acts as its match. Thus, the D and the A are each other’s matches and hence the definite article is spelled out on the adjective. The correlation between the occurrence of the definite article as a suffix on the noun and the presence of a genitive phrase, a demonstrative, a modifier or a complement can thus be accounted for by the fact that these constituents are able to value the [uRef:] feature on the noun, which in turn is a necessary step in order for the noun to act as the goal for the [iRef:] feature on D. A sequence like *frumoasa fata/ ‘beautiful.the girl.the’ is predicted to be ungrammatical because in this analysis one of the conditions for the definite article to be spelled out as a suffix on the noun is if the latter acts as the goal for D and provides a value for a feature on D. The [idef:] feature on D is valued by the adjective and so the definite article gets spelled out on the adjective, but its [iRef:] feature cannot be valued in this sequence. In the absence of any value for the [Ref] feature on N and hence on D, no spell out of the definite article can occur on the N. In the next section I will discuss each case in turn.

features in order to get erased and matches for uninterpretable features can be positioned either down, in the c-command domain of the uninterpretable feature, or up, in positions that c-command the uninterpretable feature. In both cases, the uninterpretable feature enters an Agree relation with an interpretable match as a piggy back on the Agree relation initiated by an unvalued feature. If the uninterpretable feature is valued, it can only be a goal for a probe that is unvalued and that is placed higher in the structure. In this case, the match for the uninterpretable (and valued) feature will be higher in the structure. If on the other hand the uninterpretable feature is unvalued, it will act as a goal itself and it will initiate search for a valued match downwards, in its c-command domain. In this case, the match for the unintepretable feature could be either downwards, in its c-command domain or higher in the structure (in case the goal is valued but uninterpretable.

74

Isac

4 Analysis 4.1 Two-term Polydefinites Including Nouns with Unique Reference Examples of polydefinite DPs that include nouns that designate a unique referent were given in (5), repeated below for convenience. (19) a. întru lauda a Preacuratei Ficioarii Mariei in praise.DEF of pure.DEF.GEN virgin.DEF.GEN Mary.GEN ‘in praise of the pure Virgin Mary’ (Ureche. 6978 [1470]) b. în zioa sfintei blagoiavleniei in day.DEF.GEN saint.DEF.GEN baptism.of.the.Lord.DEF.GEN ‘the day of the saint baptism of the Lord’ (LC {224}) c. Iar diavolul diavolul nu putu and enemy.DEF devil.DEF not could ‘and the evil devil could not bear’ (LC {195})

răbda bear

In the analysis proposed here, the crucial property that accounts for the overt realization of the definite article on the noun in these cases is the fact that these nouns have a valued (although uninterpretable) [Ref] feature. (20) DP D FP [idef: ] [iRef: ] AP

F’

A F NP [udef:] N [udef:+] [uRef:+] vrăjmașul diavolul evil.DEF devil.DEF The valued but uninterpretable [uRef:+] feature on the noun acts as a goal for the [iRef:] feature on D and provides a value for the latter. At the same time, the noun’s uninterpretable [uRef] feature gets erased by the matching interpretable [Ref] feature on D. Hence, D and N provide matching features for

Polydefinite DPs in Old Romanian

75

each other and the definite article gets spelled out on the noun in accordance with (18). On the other hand, the article also gets overtly suffixed onto the prenominal adjective, because the adjective acts as a goal for the unvalued [idef:] feature on D, and at the same time D provides a matching feature for the uninterpretable [udef] feature on A (since D has a matching [def] feature which is interpretable). Given that D and A are each other’s matches, the definite article gets spelled out on the A as well. Although in this analysis the adjective bears an unvalued [udef:] feature, the latter gets a value by agreement with the noun, which has a valued [udef:+] feature. Once its [def] feature is valued, the adjective can act as a goal for D and value its [idef:] feature. 4.2 Three-term Polydefinites Including a Genitive Phrase Examples of this type of polydefinite DPs are given in (6) above, repeated below for convenience. (21) a. nemăsurata mila lui Dumnezeu boundless.DEF mercy.DEF of God ‘the boundless mercy of God’ (Antim {160}) b. mărg la svânta icoana lui Hristos de să-nchină go.3PL at saint.DEF icon.DEF of Christ to REFL=-pray ‘they go to the saint icon of Christ to pray’ (Dosoftei {16}) c. pentru buna stâmpărarea vânturilor for good.DEF calming.DEF winds.DEF.GEN ‘for the good calming of the storm’ (Dosoftei {59}) The analysis of these polydefinite DPs is similar to the one proposed for twoterm polydefinites, except that in these cases the noun is not necessarily intrinsically referential. If it is, then the analysis is identical to the one proposed above. If, on the other hand, the noun is not a name or name-like, I propose that its [Ref] feature is unvalued. Valuation is obtained from the genitive phrase, which bears a valued [Ref] feature. Thus, (21c) has the configuration in (22).

76

Isac

(22) DP D FP [idef: ] [iRef: ] AP

F’

A F NP [udef:] N GenP [udef:+] [uRef:+] [uRef:] buna stâmpărarea vânturilor good.DEF calming.DEF winds.DEF.GEN Once the [Ref] feature on the noun is valued from the genitive phrase, the derivation proceeds exactly as the one described above, for two-term polydefinite DPs. In particular, the definite article is spelled out as a suffix on the noun because the noun acts as a goal for the [iRef:] feature on D and provides the value for it and at the same time D provides an interpretable match for the uninterpretable [uRef] feature on N. The article is also spelled out as a suffix on the adjective because the adjective is the goal for D’s other feature, namely [idef:], and provides the value for the latter. At the same time D provides the interpretable match for the uninterpretable [udef] feature on A. Positing a [Ref] feature on the genitive phrase can be justified by the fact that genitives have a deictic, anchoring interpretation (Koptjevskaya-Tamm 2002, 2005; Cornilescu & Nicolae 2009, 2011; Pană Dindelegan 2008). However, not all genitives share this property. The literature distinguishes between several types of genitives, depending on their relative position with respect to the noun, depending on their referential properties, and depending on their morphology. Thus, genitives could be post-nominal or prenominal from the point of view of their position, and anchoring or non-anchoring from the point of view of their referential potential. Moreover, from the point of view of their morphological make-up, Old Romanian distinguishes between three ways of expressing a genitive (see Giurgea, this volume): (i) an inflected article a/al followed by a noun inflected for Genitive Case (al genitives); (ii) only a noun inflected for Genitive Case (bare inflectional genitives); (iii) a PP with P de followed by a noun (de genitives). I will first discuss post-nominal genitives and turn to prenominal ones later on.

77

Polydefinite DPs in Old Romanian

4.2.1 Post-nominal Genitives Within the class of post-nominal genitives, inflectional genitives introduced by a/al and bare inflectional genitives are consistently anchoring from the point of view of their reference, while the genitive introduced by de could be both an anchoring and a property genitive (Pană Dindelegan 2008; Cornilescu & Nicolae 2009, 2011). Given this distinction, we expect only al genitives and bare inflectional genitives to always bear a [Ref] feature, but not de genitives. Thus, we expect to find examples in which de genitives co-occur with a noun that bears the definite article, as well as examples in which the noun does not bear the article. The expectation is borne out in (23). (23) a. preasfintei născatoarei de Dumnezeu anchoring de genitives holy.DEF.GEN bearer.DEF.GEN DE God ‘of the holy bearer of God’ (Antim {16}) b. pentru marile faceri de bine for big.DEF doings DE good ‘for the big good deeds’ (Antim {78v})

property de genitives

In (23a) the de genitive follows a noun that hosts the definite article. In our analysis this indicates that the de genitive has a valued [Ref] feature that enters Agree with the [Ref] feature on the noun, and values the latter. This now valued [Ref] feature on the noun can in turn act as a goal for the unvalued [Ref] feature on D and thus the definite article is spelled out on N. On the other hand, in (23b) the de genitive follows a bare noun. In our analysis this indicates that the de genitive has no [Ref] feature in this case. In spite of the fact that both al genitives and bare inflectional genitives are anchoring and referential, their properties are not identical. In particular, al genitives can co-occur with bare nouns or with nouns suffixed with the definite article, whereas bare inflectional Genitives seem to only co-occur with nouns that are suffixed with the definite article. Al genitives are illustrated in (24a,b) and bare ones in (24c). (24) a. la sfânta mănăstire a Homorului at saint.DEF monastery AL Homor.DEF.GEN ‘at the saint monastery of Homor’ (PH 1174) b. urdzirea a lumiei making.DEF AL world. DEF.GEN ‘the making of the world’ (PH {12r})

78

Isac

c. pentru buna stâmpărarea vânturilor for good.DEF calming.DEF winds.DEF.GEN ‘for the good calming of the storm’ (Dosoftei {59}) In order to account for this asymmetry, I propose that both al and bare inflectional genitives bear a valued [Ref] feature, but with al genitives this feature is interpretable, while with bare inflectional genitives, it is uninterpretable. The features and the structure of polydefinites including a post-nominal al genitive is given below. (25) DP D FP [idef: ] [iRef: ] AP

F’

A F NP [udef:] N AL GenP [udef:+] [iRef:+] [uRef:] sfânta mănăstire a Homorului saint.DEF monastery AL Homor.DEF.GEN The effect of positing an interpretable [Ref] feature on al genitives is that the uninterpretable [Ref] feature on the noun can be checked against the matching feature of the al genitive, and thus D is cut out of the loop. More precisely, even though N acts as a goal for D (since the valued [Ref] feature on N values the [iRef:] feature on D), D does not provide a match for the uninterpretable [uRef] feature on N (since N checks its uninterpretable [Ref] feature against the al genitive). The definite article will thus not be spelled out on N, since D and N do not act as each other’s matches. We still need to account for the fact that al genitives can also co-occur with nouns bearing the definite article in Old Romanian, as in (24b). The crucial observation is that (24b) is not a polydefinite DP but a monodefinite one. I have not found any example of polydefinite DPs that show an overt definite article suffixed onto the noun when an al genitive is present. I propose that this gap is not accidental. In monodefinites such as (24b), the overt realization of the definite article on the noun is not the result of a mutual match of the [Ref] feature on N and D, because the nouns’s [uRef:] feature has already been checked against the interpretable and valued [iRef:+] feature on the al

Polydefinite DPs in Old Romanian

79

genitive. Instead, the overt definite article suffixed on the noun is the result of N and D providing matches for each other’s [def] feature. This contrast is supported by the fact that in monodefinite structures al genitives are compatible with both definite nouns and indefinite ones, as can be seen in (26a,b). (26) a. urdzirea a lumiei making.DEF AL world. DEF.GEN ‘the making of the world’ (PH {12r}) b. bun priiaten a dumilor voastre good friend AL highness.DEF.GEN your ‘a good friend of your highness’ (PH {888}) If the definite article suffixed on the noun in (24a) were the effect of the checking of the [uRef] feature on N, we would expect the definite article to always be spelled out on the noun, contrary to fact. If instead the definite article suffixed on the noun in (26a) is the PF realization of the noun checking its [udef] feature against the [idef] feature on D, then it is to be expected that the [udef] feature on N could be valued either as [+] or as [-]. (27) a. DP D NP [idef: ] [iRef: ] N AL GenP [udef:+] [iRef:+] [uRef:] urdzirea a lumiei making.DEF AL world. DEF.GEN b. DP D NP [idef: ] [iRef: ] N AL GenP [udef:-] [iRef:+] [uRef:] priiaten a dumilor voastre friend AL highness.DEF.GEN your

80

Isac

On the other hand, in all the examples that I found of monodefinites that include a bare inflectional genitive, the noun is suffixed with the definite article, even when an adjective precedes the noun, as in (28a). The same observation is made in Cornilescu & Nicolae (2009, 2011), who propose to account for this restriction by invoking a principle of economy. In their view the use of a bare inflectional genitive is less costly than an al genitive and, in order to make this more economical option, the definite article must be suffixed onto the noun, rather than onto the adjective. The fact that the bare inflectional genitive cannot occur in the absence of a definite article suffixed onto the noun is simply stipulated as a distributional restriction on the bare inflectional genitive. In the analysis proposed here this constraint can be accounted for. The definite article must be spelled out on the noun because the unvalued [iRef:] feature on D probes for the valued [Ref] feature on N and the uninterpretable [uRef] feature on the noun simultaneously finds a match in the [iRef] feature of D. N and D act as each other’s matches and the definite article is thus spelled out on N, as in (27)/(28). Whether the definite article is also spelled out on the adjective or not is an independent issue (given that adjectives do not bear [Ref] features and that the agreement between the adjective and D involves a different feature-[def], rather than [Ref]). (27) a. înaltă socotiala mării sale high business.DEF highness.DEF.GEN his ‘the important business of his highness’ (Amiras {316r}) b. voia vizirului will.DEF minister.DEF.GEN ‘the will of the minister’ (Neculce 59) (28) DP D FP [idef: ] [iRef: ] AP

F’

A F NP [udef:] N GenP [udef:+] [uRef:+] [uRef:] înaltă socotiala mării sale high business.DEF highness.DEF.GEN his

Polydefinite DPs in Old Romanian

81

The same [Ref] feature can also account for why bare inflectional genitives are licensed only by definite nouns. Given that the bare inflectional Genitive has an uninterpretable [Ref] feature valued as ‘+’, this type of genitives is compatible only with N heads that are definite. This is because only nouns that have a [+def] feature offer a way for the genitive to check its uninterpretable [Ref] feature. Nouns whose [def] feature is valued as [-] do not provide a proper match for the bare inflectional genitives, since the [-] value of the nouns will clash with the [+] value of the genitive, and thus there can be no Agree between the [Ref] feature on the genitive and the [Ref] feature on N. As a result, the uninterpretable [Ref] feature on the genitive remains unchecked and the derivation crashes. The relevant representation is given in (28a). If, on the other hand the noun has a [udef:+] feature, it will also have a [uRef:] feature which will be valued as [+] by agreement with the genitive and then will be checked against the interpretable matching [iRef:] feature of the D head. This will secure the spell out of the definite article on the noun, since it is the noun that is the goal for the unvalued [iRef:] feature on D, and D provides a match for the uninterpretable [uRef:+] feature on N. However, in these cases, the noun does double duty: apart from checking its [uRef:] feature against the matching [Ref] feature on D, it also checks its [udef:] feature against the [def] feature of the same D head. Thus both uninterpretable features of N are matched by the features of D and both features of D are valued by the matching features on N. The relevant structure is given in (29b). (29) a. DP D NP [idef: ] [iRef: ] N GenP [udef:-] [uRef:+] [uRef:-] *știre vizirului knowledge minister.DEF.GEN

82

Isac

DP b. D NP [idef: ] [iRef: ] N GenP [udef:+] [uRef:+] [uRef:] știrea vizirului knowledge.DEF AL highness.DEF.GEN your 4.2.2 Prenominal Genitives In addition to their post-nominal position, genitive phrases could also occur prenominally in Old Romanian. According to Cornilescu & Nicolae (2011) there are two types of prenominal genitives: the so-called determiner genitive, which is referential (30); and the attributive/property genitive, which is nonreferential (31). (30) a. după a țării poftă according.to AL country.DEF.GEN desire ‘in compliance with the country’s desire’ (RG {16}) b. a toată lume cap să piară de o muiere AL all people head SUBJ perish.SUBJ.3 by a woman ‘that everyone’s head should perish because of a woman’ (Alexandria 197) c. frumos mirositoare a dragostei flori sweetly smelling AL love.DEF.GEN flower ‘the sweet-smelling flower of love’ (Cantemir II, 4) (31) a. a ceriului împăratul AL sky.DEF.GEN emperor.DEF ‘the emperor of the sky’ (PH 143) b. a lor feciorii AL their sons.DEF ‘their sons’ (PH 207)

83

Polydefinite DPs in Old Romanian

c. să înece a toate țările anii trecuți SUBJ drown AL all countries.DEF years.DEF passed ‘(. . .) would drown the by-gone years of all the countries’ (Ureche 57) d. deci cu reaua a lui so with bad.DEF AL his ‘so with his bad service’ (LC 40)

slujbă service

e. aceste ale Ciacalului (. . .) cuvinte these AL Jackal.DEF.GEN words ‘these words of the Jackal’ (Cantemir I, 72) f. altă a trupului parte other AL body.DEF.GEN part ‘some other part of their body’ (Cantemir I, 30) Determiner genitives (illustrated in (30)) occur with bare nouns, but their presence is enough to render the whole DP definite. Cornilescu & Nicolae (2011) analyze determiner genitives as always involving movement of the genitive to Spec,DP, where the determiner genitive checks the [def] feature on D. I will adopt their analysis here, with the addition of the [Ref] feature introduced above. In other words, I will assume that in addition to the [def] feature, al genitives also have an interpretable [iRef] feature valued as ‘+’ (see the discussed above, in relation to the structure in (25)). The analysis in Cornilescu & Nicolae (2011) also proposes that the remnant NP raises to the Specifier of NumP, but in (32) I will simplify and just represent movement of the genitive to SpecDP. (32) DP ALGen D’ [idef:+] [uRef:+] D NP [idef: ] [iRef: ] N ALGenP [udef:+] [idef:+] [uRef:] [iRef:+] a țării poftă a țării AL desire AL country.DEF.GEN country.DEF.GEN

84

Isac

When merged as a sister to N, the al genitive checks and values the [uRef:] feature on N, as well as the latter’s [udef] feature. Once it raises to Spec,DP, it can also act as a goal for the [idef:], as well as the [iRef:] feature on D. Given that N does not probe for D (since all of its uninterpretable features are checked and valued by agree with the features of the genitive), the definite determiner will not be overtly realized as a suffix on N, but the whole structure is still interpreted as definite, since ultimately the [idef:] feature on D is valued as ‘+’. Property genitives, on the other hand, can co-occur with the definite article (which can be suffixed onto a prenominal adjective or onto the noun, as in (31a–d)), with demonstrative articles, as in (31e), as well as with indefinite articles, as in (31f). Their co-occurrence with both definite and indefinite articles suggests that property genitives do not have a definiteness feature. I propose that these genitives also lack a [Ref] feature. Given that property genitives lack both a [def] and a [Ref] feature, they are basically irrelevant for the computation of definiteness and referentiality in the DPs in which they occur. These features are checked and valued independently from the genitive. I will leave aside examples like (31f) in which the DP is headed by an indefinite determiner, since my interest here is in polydefinite DPs. I will therefore focus on the definite DPs with prenominal al genitives in (31a–e). If the DP is headed by a demonstrative, as in (31e), the demonstrative determiner, which is specified as [+def] and as [+Ref] will check and value the corresponding features on D. As a result, the definite article will not be suffixed onto the noun, since it is not the noun that probes for D, but the demonstrative. Given that the demonstrative checks both the [idef:] and the [iRef:] feature on D, there is no possibility for disjoint checking of these features (i.e. by two different items) and therefore no polydefinites are expected to be found in these cases. When the property genitive co-occurs with an adjective or a noun that hosts the definite article, as in (31a–d), it is either the prenominal adjective or the noun that will act as a goal for the [idef:] feature on D and at the same time the prenominal adjective or the noun will check their [udef] feature against D. Thus, the definite article will be suffixed either on the adjective or on the noun. When the adjective finds a match for its [udef:] feature in D, as in (31d), the possibility is open, at least in principle, for the [iRef:] feature on D to enter Agree with a matching feature on a different item, i.e. on the noun. This structure could thus support a polydefinite DP, as in (33). However, even though the noun could indeed find a match in D for its [uRef: ] feature, there is no item in this structure that could supply a value for the [Ref] feature. Since N cannot act as a goal for D, the definite article will not be spelled out on N.

85

Polydefinite DPs in Old Romanian

(33) DP D FP1 [idef: ] [iRef: ] AP

F’

A F1 FP2 [udef:] PropertyGenP F2’

F2 NP

N PropertyGenP [udef:+] [uRef:] reaua a lui bad.DEF AL his

slujbă service

a lui AL his

The question now is: how is the [Ref] feature supposed to be interpreted in the absence of a value for this feature? I propose that in these cases the [Ref] feature acts like a variable and that a default existential operator will be inserted to bind this variable. The operator is most likely anaphoric, relating the DP to the preceding discourse, but I will leave the precise characterization of this operator for further research. What is important for our analysis is to notice that when the value of the [iRef:] feature is supplied by Operator binding, the definite article is not spelled out on the noun, even when the noun probes for the [iRef:] feature on D (i.e. even when the noun checks off the uninterpretability of its [uRef:] feature be agree with D). This is because, as stated in (18), the spell out of the definite article on an item inside the DP depends on whether the respective item and D can be each other’s matches. Finally, when the definite determiner is spelled out on the noun, as in (31a–c), this is because the noun values some feature on D and D provides a match for the same feature on N. In (31a), both the [idef:] feature on D and the latter’s [iRef:] feature are valued by matching features on N, while D provides at the same time interpretable matches for the uninterpretable [udef] and [uRef] features on the noun. The N and D act as each other’s matches, so the definite article is overtly suffixed on the noun. Notice that in (31a) the noun has its [Ref] feature valued intrinsically, and not due to the presence of a genitive,

86

Isac

demonstrative, modifier, or complement. The nouns in these constructions are Proper Name-like and have their referential feature valued as [+], as in (34). DP (34) D [idef: ] [iRef: ]

FP PropertyGenP



F’

F NP

N PropertyGenP [udef:+] [uRef:+] a ceriului împăratul a ceriului AL sky.DEF.GEN emperor.DEF AL sky.DEF.GEN However, in (31b,c) the noun does not bear an intrinsically valued [Ref] feature and yet the noun carries the overt definite article. This is because in these cases, even though the [Ref] feature on N is not valued, its [def] feature still is. Thus, N and D match each other’s [def] feature, and this is enough for the definite article to be realized overtly on N. The [Ref] feature on N and D receives a value by virtue of the fact that it is bound by an operator. 4.3 Three-term Polydefinites Including a Demonstrative Phrase Examples of this type of polydefinites are given in (7) and repeated below. (35) pentru svânta lăcuința aceasta for saint.DEF home.DEF this ‘for this saint home’ (Dosoftei {38}) Demonstratives that occur in this type of polydefinites are always postnominal. The morpho-phonological shape of the paradigm of post-nominal demonstratives differs from that of prenominal demonstratives. This indicates that the two occupy different positions in the tree. The two paradigms for masculine forms are given in (36). (36) prenominal acest acești

Post-nominal acesta aceștia

87

Polydefinite DPs in Old Romanian

Post-nominal demonstratives have been analyzed as occupying the Specifier position of a functional projection higher than NP. The post-nominal position of these demonstratives is assumed to be the result of movement of the noun head past the demonstrative, to a higher head (Cornilescu 1992, 1995, 2004; Cornilescu & Nicolae 2009, 2011). I will follow Tasmowsky (1990), Manoliu (2000), Bernstein (2001), Brugè (2002), Cornilescu (2005), Vasilescu (2009), and propose that post-nominal demonstratives are indexical and I will thus posit that they bear a [+Ref] feature. Possibly, post-nominal demonstratives might also have a [def] feature (just like prenominal ones), but the data can be accounted for without committing to the existence of this feature. I will therefore leave it aside. The tree in (37) uses the labels and hierarchy of projections proposed in Cornilescu & Nicolae (2011) for demonstrative structures. (37)

D

D EmphP [idef: ] [iRef: ] NP/FP

Emph’

Emph ArtP

DemP Art’ [uRef:+] Art FP AP F’ [udef:] F NP [udef:+] [uRef:] a. dunga line.DEF

aceasta this

vearde green

dunga line

b. [sfinta lăcuința] saint.DEF home.DEF

aceasta this

sfinta saint

lăcuința home

c. nenorocitele unfortunate.DEF

acestea these

nenorocitele unfortunate

vremi times

88

Isac

This derivation is very similar to the one discussed above for polydefinites involving a bare inflectional genitive. The noun gets its [uRef:] feature valued by the demonstrative, which has a valued referential feature. As with all demonstrative structures the noun moves past the demonstrative to a higher position. The literature assumes that what moves is actually a phrasal constituent including the noun: either the NP by itself, as in (37a), or the FP including both the NP and the AP, as in (37b) or else the AP by itself, as in (37c). If the noun is part of the XP, that raises past the demonstrative, once it raises, it checks its [Ref] feature against the matching feature on D. Given that D also probes for the [Ref] feature on N in order to get its own [iRef:] feature valued, the definite article is spelled out on the noun. On the other hand, the definite article can also be spelled out on the adjective, if the adjective is taken along by XP movement, as in in (37b), or if the adjective phrases raises by itself, as in (37c). If however, the adjective is left behind and only the NP moves to Spec of EmphP, then the noun will probe both the [idef:] feature on D and its [iRef:] feature, and the noun will be the only one bearing the definite suffix. Old Romanian also provides examples in which the adjective carries the definite article suffix even if it stays low and does not raise in front of the demonstrative. Such constructions, as in (38), were first noticed by Nicolae (2013). (38) a. locul cela strimtul place.DEF that narrow.DEF ‘that narrow place’ (DÎR, I, Scrisoarea lui Neacșu, c. 1521) b. pizma ceaia reaoa să o strife.DEF that bad.DEF SUBJ it= ‘let us resolve that discord’ (Coresi 54)

spargemǔ break.1.PL

These constructions can easily be accounted for under the assumption that Old Romanian uses distant Agree as well as local Agree, as proposed in Cornilescu & Nicolae (2009, 2011). However, under the assumption that Agree is strictly local, these DPs are problematic, because the more local noun could probe and value both the [idef:] feature and the [iRef:] feature on D and the adjective is expected to occur without a definite article, as in (37a), and yet it seems that at least one of these features is checked by the more distant adjective, since the adjective is suffixed with the definite article. Notice that thus far, we have managed to account for the data by assuming local Agree only. From a theoretical point of view, sticking to local agree only and shifting the locus of variation in the lexicon is more consistent with the assumptions of the Minimalist program.

Polydefinite DPs in Old Romanian

89

I will therefore tentatively propose that these constructions involve two DP layers, each of them making its own semantic contribution, on a par with CEL constructions (Cornilescu 1992). It is probably significant in this sense that CEL forms are derived historically from demonstratives, as proposed by Iordan & Manoliu (1965), Dimitrescu (1975), Marchis & Alexiadou (2009), Giurgea (2012), Pană Dindelegan (2013), a.o. Another observation which points in the same direction is that Modern Romanian has a limited number of contexts in which the definite article is suffixed on both the noun and the adjective and all those contexts involve a demonstrative (Iordan 1956), as in (39). (39) a. fata aia girl.DEF that ‘that little girl’

mica little.DEF

b. muncitorul ăla vrednicul worker.DEF that diligent.DEF ‘that diligent worker’ Given that the generally agreed assumption is that Modern Romanian does not have long distance Agree, an alternative analysis needs to be proposed for these structures. Since a double DP layer structure has already been shown to work for CEL constructions, it is not unlikely that the same type of analysis could be extended to the DPs in (39) as well, and implicitly to the ones in (38). 4.4 Three-term Polydefinites Including a Modifier or a Complement Examples of polydefinite DPs including modifiers were given above in (8) and (9) and are repeated below. (40) a. la luminatul craiul leșescu at enlightened.DEF prince.DEF Polish ‘at the enlightened Polish prince’ (DÎR, XVIII, Scrisoare, 5 sept. [1599]) b. cetății marei și frumoasei în Spania fortress.DEF.GEN big.DEF.GEN and beautiful.DEF.GEN in Spain ‘of the big and beautiful fortress in Spain’ (CC {37})

90

Isac

(41) la sfânta purtătoarea de viață a at holy.DEF bearer.DEF DE life of mântuitoriului nostru Iisus Hristos savior.DEF.GEN our Jesus Christ ‘to the holy mother of Jesus Christ’ (Antim {105}) As illustrated in section 2, the overt realization of the definite article on the noun is optional when the noun is accompanied by a modifier or a complement. In (10), repeated below as (42), the definite article is overtly realized on the adjective only, in contrast to (40) and (41), in which the definite article is suffixed both onto the noun and onto the adjective. (42) pentru marile faceri de bine for big.DEF doings of good ‘for the big good deeds’ (Antim {78v}) It is important in this respect to distinguish between adjectival modifiers, that are attributive and prenominal, and adjectives that are post-nominal and predicative. According to Cornilescu & Nicolae (2009, 2011), only prenominal adjectives carry a [udef:] feature. What this means in our analysis is that only prenominal adjectives can carry an overt definite article as a suffix. In order to deal with the optionality of the suffix on the noun I will assume that modifiers and complements optionally carry a valued [Ref] feature. This captures their potential of anchoring the reference of the noun. When these items carry this feature, it values the [Ref] feature on the noun, which will in turn be able to probe for the [iRef:] feature on D and value it. In these cases the definite article will be suffixed on the noun. Thus in (42), the prenominal adjective is suffixed with the definite article as a consequence of the fact that D probes for the adjective in order to value its own [idef:] feature and at the same time A has an uninterpretable [udef] feature that gets matched by the [idef:] feature on D. On the other hand, the definite article is also suffixed on the noun because D probes for the valued [Ref] feature on N and N’s uninterpretable [uRef] feature finds a match in the interpretable [iRef] feature on D. The [uRef:] feature on N is first valued by the predicative adjective.

Polydefinite DPs in Old Romanian

91

(43) DP D FP [idef: ] [iRef: ] AP

F’

A F NP [udef:] NP AP N A [udef:+] [uRef:+] [uRef:] luminatul craiul leșescu enlightened.DEF prince.DEF Polish However, if the modifier/complement does not carry a valued [Ref] feature, the [uRef:] feature on the noun cannot be valued. As a consequence, N cannot act as a goal for D, since its [Ref] feature is not valued. Thus, the definite article cannot be spelled out on the noun in these cases. This would account for examples such as (42). As discussed above, in these cases, in which an uninterpretable [Ref] feature finds an interpretable match but no valuation can occur, we propose that the value of the [Ref] feature is supplied by inserting a default operator that binds the referential variable. Old Romanian also provides examples like (40b), in which both the noun and the adjective carry the definite article suffix, but the adjective follows, rather than precedes the noun. This is puzzling under the assumption that only prenominal, attributive adjectives can carry the definite suffix. In order to account for this, I will assume that in (40b) the adjective is indeed attributive and prenominal, but that the derivation of (40b), as opposed to (40a) involves an additional step, namely noun movement to a position that is higher than the adjective.

92

Isac

(44) DP D FP1 [idef: ] [iRef: ] NP

F’

N F1 FP2 [udef:+] [uRef:] AP F’ A F2 NP [udef:] NP PP N [udef:+] [uRef:+] [uRef:] cetății marei și frumoasei cetății în Spania fortress.DEF.GEN big.DEF.GEN and fortress.DEF.GEN in Spain beautiful.DEF.GEN I do not have much to say about what motivates this movement of the noun to a higher position except that it is most likely related to a focused/topicalized interpretation of the noun. This could be related to the ‘weight’ of the prenominal adjective (notice that the AP is complex here, being made up of a conjunction of two adjectives) and to some discourse related operation whose effect is to place heavy constituents towards the end of the phrase. The same order can be observed in (45). Here again, the AP is rather heavy and thus is shifted to the end of the phrase. (45) iar diavolul neiubitorul de neamul omenesc and devil.DEF not.fond.DEF DE kind. DEF human ‘and the devil, not fond of the human kind’ (LC 39) 4.5 Monodefinite DPs This proposal can account not only for polydefinites but also for instances in which the article gets spelled out only once.

93

Polydefinite DPs in Old Romanian

(46) a. Și au trimis cărțile cu călărașul and have.3PL sent books.DEF with soldier.DEF ‘and they sent the books with the soldier’ (Amiras {249v}) b. Ș-au rămas țara în bună and-have stayed country.DEF in good ‘and the country stayed calm’ (Amiras {312v})

odihnă rest

c. astupa voroava cărăușilor covered conversation.DEF wagon.drivers.DEF.GEN ‘covered the conversation of the wagon drivers’ (Cantemir I 67) d. au purces fără numai din vechea has= acted without only from old.DEF și rânceda pizmăluire and rancid. DEF envy ‘He acted out solely from the old and rancid envy’ (Cantemir I 67) e. că mare scrâșnetul roatelor that great grinding.DEF wheels.DEF.GEN ‘that the strong grinding of the wheels’ (Cantemir I 67) In simple DPs like (46a) the noun needs to check its uninterpretable [uRef] feature and the [idef:] feature on D provides a match, and on the other hand D needs to valued its [idef:] feature and probes for the valued [udef:] feature on N. This is enough for the definite article to be spelled out on N. However, even though the uninterpretable [Ref] feature on N can find an interpretable match in D, D can’t find a match in the [Ref] feature of N, as the latter is not valued. Therefore a default existential operator will bind the referential variable and the interpretation will most likely be anaphoric (i.e. the operator will relate the noun to the previous discourse). If the simple noun is Proper Name-like, as in (46b), then the noun and the D will probe both for each other’s [def] feature and each other’s [Ref] feature. The result, again, is overt suffixation of the definite article on N. If the noun is accompanied by a genitive phrase, as in (46c), the genitive, which carries a valued [Ref] feature in our account, will value the [Ref] feature on N and this will allow N to value both the [def] and the [Ref] feature on D. The definite article is therefore suffixed onto the noun.

94

Isac

If an attributive adjective precedes the noun, as in (46d), D probes for the [def] feature on A since the latter is valued and D also provides a match for the uninterpretable [udef] feature on A. As a result, the definite article will be suffixed onto the adjective. On the other hand, the unvalued [Ref] feature on D will seek for a valued matching goal, but N is not a suitable match, since the [Ref] feature on N is not valued. Thus, the definite article on N does not encode the [Ref] feature, but only the [def] feature. The [Ref] feature will be treated like a variable and will be bound by a default operator. Finally, in (46e) the definite article is suffixed onto the noun. This indicates that the noun probes and values some feature of D and at the same time that D provides a match for the same feature in N. Notice that in (46e) the noun is accompanied by a predicative adjective—omenesc/‘human’-which values its [Ref] feature. Therefore the noun is able to value both the [def] feature on D and its [Ref] feature. On the other hand, D is able to offer a match for both the uninterpretable [def] feature on N and the latter’s uninterpretable [Ref] feature. The puzzling thing about (46e) is not to explain why the definite article is spelled out as a suffix on N, but to explain why the prenominal adjective doesn’t value the [def] feature on D and therefore why the prenominal adjective does not carry a definite suffix. The answer provided by Cornilescu & Nicolae (2009, 2011) is that Old Romanian made use of long distance agree in addition to local agree. In this view, (46e) is an instance of distant agree in which the [def] feature on D is valued not by the more local adjective, but by the more distant N. Notice that our account explained away all the other instances of distant agree. In all the situations discussed in this paper, the features of D were valued by the closest item bearing matching features. Thus, instances like (46e) would be the only ones in which distant agree would apply. It is thus very tempting to propose a different analysis of (46e), one that does not rely on distant agree. Although I will leave the matter open, an alternative analysis for (46e) would have to assume that the prenominal adjective occurring in this example is actually a predicative one and therefore it starts off as adjoined to NP. The adjective then moves to a prenominal position for discourse related purposes, but given its predicative nature, it does not bear a [def] feature in our analysis. Therefore even if it is closer to D then the noun, the adjective does not have the right features to be able to value the [def] feature on D. The definite article will not be spelled out on A in this case and the noun will value both the [def] feature on D and its [Ref] feature.

Polydefinite DPs in Old Romanian

5

95

Diachronic Remarks

Modern Romanian still has contexts in which the definite suffix is attached to more that one constituent within a DP. According to Iordan (1956) these contexts all involve a demonstrative, as in (47). (47) a. fata aia girl.DEF that ‘that little girl’

mica little.DEF

b. muncitorul ăla vrednicul worker.DEF that diligent.DEF ‘that diligent worker’ c. săraca fata poor.DEF girl.DEF ‘that poor girl’

aia that

In section 4.3, in which we discussed polydefinite DPs including demonstratives, we suggested that these examples should be analyzed as involving two DP layers, on a par with CEL constructions (Cornilescu (1992)). Apart from these contexts, polydefinite DPs are no longer possible in Modern Romanian. I suggest that this is because in Modern Romanian the [Ref] feature is no longer a morpho-syntactic feature that participates in the syntactic derivation of DPs. As a consequence, the only feature that can be checked reciprocally by D and another item in the DP is the [def] feature, and the definite article can be spelled out on only one item in the DP. 6 Conclusions The proposal made in this paper is that polydefinite DPs in Old Romanian compute syntactically over two features: the [def] feature and the [Ref] feature. The D head carries both of these features and they are both unvalued on D. These features can be valued by items within the DP carrying matching features. If the matching features are carried by one and the same item in the DP, then the definite article is overtly realized only once. But if the matching features are carried disjointly by independent items, then the definite article is overtly expressed twice within the DP.

96

Isac

I also proposed that there is a correlation between the occurrence of the definite article as a suffix on the noun and the presence of a genitive phrase, a demonstrative, a modifier or a complement. This correlation can be accounted for by the fact that these constituents are able to value the [uRef:] feature on the noun, which in turn is a necessary step in order for the noun to value the [iRef] feature on D. A sequence like *frumoasa fata/ ‘beautiful.the girl.the’ is predicted to be ungrammatical because in our analysis the definite article is spelled out as a suffix on the noun only if D and the noun act as each other’s matches. In this case the N head cannot act as a goal for the D head that needs to value its [iRef:] feature, because the [Ref] feature on N is unvalued. The paper also argued that (un)interpretability and valuation are treated differently by the grammar and thus validated Pesetsky & Torrego’s (2007) proposal that interpretability and valuation are not biconditionally related in the lexicon. Finally, the paper also proposed an account for the contrast between al genitives and bare genitives (bare inflectional genitives must co-occur with a definite article whereas al genitives don’t have to) by positing that al genitives can check the [uRef] feature on the noun by virtue of the fact that they carry an interpretable [iRef] feature. Thus, when an al genitive is present, D does not provide a match for the uninterpretable [uRef] feature on N and the definite article is not spelled out on the noun. In contrast, bare genitives have an uninterpretable and valued [Ref] feature. Even though bare genitives can value the [Ref] feature on N, they cannot provide a match for the uninterpretable [uRef] feature on N. As a consequence, the uninterpretable [uRef] feature on N will be matched by n interpretable [Ref] feature on D and the definite article gets spelled out on N. If this paper is on the right track, it might be possible to discard the hypothesis that Old Romanian had distant Agree (as opposed to Modern Romanian, which only has local agree) and implicitly the hypothesis that Old Romanian had multiple agree. What looks like multiple agree has been accounted for in this paper as two separate instances of agree, with two different features. Crucially, each of these Agree instances is local. The only instances in which multiple agree might still be be at work are instances in which multiple adjectives agree with the same noun as in (48). (48) preaînălțatului marelui domn highness.DEF.GEN great.DEF.GEN sire ‘of his great highness the Sire’ (Antim {241} apud Croitor 2008) I leave it for further research to determine what the differences are between constructions like (48) and the constructions discussed in this paper.

Polydefinite DPs in Old Romanian

97

Corpus Alexandria 1620

Zgraon, Florentina (ed.). 2006. Cele mai vechi carți populare în iteratura română, XI, Alexandria, Cea mai veche versiune păstrată. București: Fundația Națională pentru Stiință și Artă. Amiras 1661–1729 Simionescu, Dan (ed.). 1975. Cronica anonima a Moldovei, 1661–1729: Pseudo-Amiras. Bucharest: Editura Academiei Române. Antim 1692–1714 Gabriel Ștrempel (ed.). 1972. Antim Ivireanu, Opere. București: Editura Minerva. CC 1678–1688 Gregorian, Mihai. 1984. Constantin Cantacuzino, Istoria Ţării Românești. București: Editura Minerva. Cantemir Verdeș, Ion & P.P. Panaitescu (eds.). 1983. Dimitrie Can­ temir, Istoria ieroglifică. Bucharest: Editura Minerva. Coresi 1581 Pușcariu, Sextil & Alexie Procopovici(eds.). 1914. Dia­ conul Coresi, Cartea cu învățătură [1581]. București: Atelierele Grafice Socec. DÎR 1521–1600 Chivu, Gheorghe; Magdalena Georgescu; Magdalena Ioniță; Alexandru Mareș; Alexandra Roman-Moraru (eds.). 1979. Documente si însemnări românești din secolul al XVI-lea. București: Editura Academiei. Cheia 1678 Varlaam. 1678. Cheia ințelesului. București: Mitropolia. Dosoftei 1679 Ursu, N.A. (ed.). 1980. Dosoftei, Dumnezeiasca Liturghie. Iasi: Mitropolia. PH Gheție, Ion & Mirela Teodorescu (eds.). Psaltirea Hurmuzaki (sec. al XVI-lea). București, Editura Academiei. LC 1650–1716 Gregorian, M. (ed.). 1964. Letopisețul Cantacuzinesc [Istoria țării Rumânesti de când au descălescat pravoslanicii creștini], in Cronicarii Munteni. București: Editura Tineretului. Neculce 1672–1745 Iordan, Iorgu (ed.). 1959. Ion Neculce, Letopisețul țării Moldovei si O samă de cuvinte. Bucuresti: Editura de Stat pentru Literatură si Artă. RG 1688–1698 Gregorian, M. (ed.). 1964. R. Greceanu, Cronica. In Cronicarii Munteni. Bucharest: Editura Tineretului. Ureche 1590–1647 Panaitescu, PP. (ed). 1987. Grigore Ureche, Letopisețul Ţării Moldovei. 1987. București: Editura Minerva.

98

Isac

References Alexiadou, Artemis & Chris Wilder. 1998. Adjectival modification and multiple determiners. In A. Alexiadou & C. Wilder (ed.), Possessors, Predicates and Movement in the DP. 303–332. John Benjamins. Brugè, Laura. 2002. The positions of Demonstratives in the Extended Nominal Projection. In Cinque, G. (ed.), Functional Structure in DP and IP. The Cartography of Syntactic Structures, vol. 1. 15–53. Oxford University Press. Cinque, Gugliemo. 2010. The Syntax of Adjectives. Cambridge, MA: The MIT Press. Cornilescu, Alexandra. 1992. Remarks on the Determiner System of Rumanian: the Demonstratives AL and CEL. Probus 4. 189–260. ———. 1995. Romanian Genitive Constructions. In Guglielmo Cinque & Giuliana Giusti (ed), Advances in Rumanian Linguistics. 1–52. Oxford: Oxford University Press. ———. 2004. Romanian Genitives Revisited. Bucharest Working Papers in Linguistics V(1): 45–70. Cornilescu, Alexandra & Alexandru Nicolae. 2009. Evoluția articolului hotărît și genitivul în româna veche. In R. Zafiu, G. Stoica & M.N. Constantinescu (eds.), Limba română: teme actuale. 647–669. Bucuresti: Editura Universității din Bucuresti. ———. 2011. On the syntax of Romanian definite phrases: Changes in the patterns of definiteness checking. In Petra Sleeman & Harry Perridon (eds.), The noun phrase in Romance and Germanic. 193–222. Amsterdam: John Benjamins. Croitor, Blanca. 2008. Aspecte privind acordul în determinare în limba română Veche. In Pană Dindelegan, G. (ed), Limba română. Dinamica limbii, dinamica interpretării. 213–218. Bucuresti, Editura Universității din Bucuresti. Dimitrescu, Florica. 1975. Introducere in morfosintaxa istorică a limbii române. Bucharest: Universitatea din Bucuresti. Dobrovie-Sorin, Carmen. 1987. A propos de la structure du groupe nominale en roumain. Rivista di gramatica generativa 12, 123–152. Dobrovie-Sorin, Carmen & Ion Giurgea. 2006. The Suffixation of Definite Articles in Balkan Languages. Revue roumaine de linguistique LI, 1, 73–104. Giurgea, Ion. 2012. The Origin of the Romanian “Possessive Genitival Article” AL and the Development of the Demonstrative System. Revue Roumaine de Linguistique, LVII, 1. 35–65, București. Giusti, Giuliana. 1996. Is There a Focus and a Topic Phrase in the Noun Phrase? ms., University of Venice ———. 2005. At the Left Periphery of the Romanian Noun Phrase. In M. Coene & L. Tasmowski (ed), On Space and Time in Language. 23–49. Cluj Napoca: Clusium. Grosu, Alexandru. 1988. On the Distribution of Genitive Phrases in Romanian. Linguistics 26: 931–949.

Polydefinite DPs in Old Romanian

99

Iordan, Iorgu & Maria Manoliu. 1965. Introducere in linguistica romanică. București: Editura didactică și pedagogică. Koptjevskaja Tamm, M. 2002. Adnominal possession in the European languages: form and function. Sprachtypologie und Universalienforschung (STUF) 55: 141–172. ———. 2005. Maria’s ring of gold: Adnominal Possession and Non-anchoring Relations in European Languages. In K. Ji-Young et al. (eds), Possessives and Beyond: Semantics and Syntax. Amherst, MA: GLSA Publications, 155–181. Lekakou, Marika & Kriszta Szendrői. 2007. Eliding the noun in close apposition, or Greek polydefinites revisited. UCL Working Papers in Linguistics 19: 129−154. Longobardi, Giuseppe. 1994. Reference and Proper Names: a Theory of N-movement in Syntax and Logical Form. Linguistic Inquiry 25(4): 609–665. Marchis, Alexandra & Artemis Alexiadou. 2009. On the distribution of adjectives in Romanian: the cel construction. In Aboh, Enoch O., Elisabeth van der Linden, Josep Quer & Petra Sleeman (eds.), Romance Languages and Linguistic Theory: Selected papers from ‘Going Romance’ 2007. 161–178. Amsterdam: John Benjamins. Nicolae, Alexandru. 2013. Notă de sintaxă comparată: parametrul [+definit] în sintaxa grupului nominal românesc. Limba română LXII (2): 186–208. Pană Dindelegan, Gabriela. 2008. Tipuri de gramaticalizare. Pe marginea utilizărilor gramaticalizate ale prepozițiilor de și la. In Gabriela Pană Dindelegan (ed), Limba română. Dinamica limbii, dinamica interpretării. 227–239. Bucharest, EUB. ———. (ed), 2013. The Grammar of Romanian. Oxford: Oxford University Press. Roehrs, Dorian. 2006. The Morpho-Syntax Of The Germanic Noun Phrase: Determiners Move Into The Determiner Phrase. PhD dissertation, Indiana University. Stowell, Tim. 1989. Subjects, Specifiers, and X-Bar Theory. In M. Baltin & A. Kroch (eds) Alternative Conceptions of Phrase Structure. The University of Chicago Press.

Classified Proper Names in Old Romanian: Person and Definiteness Alexandra Cornilescu and Alexandru Nicolae In this paper, we analyze the structure of Romanian complex/descriptive proper names (PNs) from a double perspective, synchronic and diachronic. Synchronically, for Modern Romanian, two main configurations are distinguished, a [classifier + PN] structure and a [PN + classifier] structure. Diachronically, these two structures arise from changes in the position and the definite/non-definite marking of adjectives and other PN-dependents, the marking of the PNs themselves for definiteness and Case, and the formation of new PNs out of nouns and adjectives which originally functioned as classifiers. In addition to classified structures, we also bring into discussion the other patterns of complex PN formation, the most relevant of which are the genitival and locative constructions.

1

Preliminary Remarks

Philosophers have always considered proper names (= PNs) as paradigmatic examples of referential expressions, and are currently still debating on the two hypotheses regarding the functioning of PNs, namely: (i) the theory of PNs as definite descriptions (in the wake of Frege 1892, an approach recently illustrated by Geurts 1997, Matushansky 2006), and (ii) the theory of direct reference/rigid designation (a theory initiated by Mill 1843 and Russell 1905, and made famous by Kaplan 1964 and Kripke 1980). On the other hand, linguists have concentrated on other problems, one of them being the extent to which PNs represent a class distinct from common nouns (= CNs). The distinction was usually set in semantic terms: CNs have descriptive sense, while PNs are devoid of descriptive content. The idea that PNs do not have meaning is apparently contradicted by the possibility of using PNs as predicates: e.g. Rom. El este un Shakespeare (‘He is a Shakespeare’). In the present paper we disregard predicative uses of PNs, and restrict the discussion to argumental uses of PNs, which, in terms of a formal grammar, should be distinguishable from CNs in their distinct formal features. It is thus incumbent on us to identify these features. The aim of this paper is twofold. On the one hand we develop a hypothesis on the structure of DPs headed by PNs, focusing mostly on Romanian © koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_005

Classified Proper Names in Old Romanian

101

(­sections 1–4). This will allow us to account for constructions typical of PNs in Modern Romanian. In the second part (sections 5–6) we discuss the syntax of PNs in Old Romanian, focusing on the transition from CNs to PNs as attested in Romanian, and describing the more complex structure of PNs in the older grammar. 2

Syntactic Features Specific to PNs

There is general agreement that PNs are DPs (Longobardi 1994; Borer 2005), and thus phases, hence expressions with an interface potential which might account for their being referential (cf. Hinzen 2007). If PNs count as DPs even in the absence of an overt determiner (D), then they must be endowed with a feature that is valued in D, and may thus activate the D position. There are two influential proposals regarding the identity of this feature. One proposal (Longobardi 1994; Borer 2005) is that PNs are inherently definite, coming from the lexicon with an valued uninterpretable1 definite feature (i.e. [u+def][]). The assumed inherent definiteness of PNs relies on a variety of semantic and morphosyntactic arguments. From a very general semantic perspective, the inherent definiteness of PNs is in harmony with their interpretation as referential definite descriptions (rather than attributive, in the sense of Donnellan 1966), since they presuppose the existence of a referent which is unique in a particular context. Since the proper identification of the intended unique referent of a PN depends on shared contextual knowledge, Borer (2005) interprets PNs as situational anaphors. From a syntactic perspective, PNs are often in complementary distribution with the definite article occurring in positions which are inaccessible to CNs, as in Longobardi’s (1994) classical examples in (1) (1) a. il mio Gianni (Italian) the my Gianni b. Gianni mio Gianni my ‘my Gianni’

1  We adopt the distinction between interpretability and valuation of features, as in Pesetsky & Torrego (2007).

102

Cornilescu and Nicolae

The initial position of the PN in (1b), replacing rather than preceding the definite article, made Longobardi (1994) propose that an essential aspect of PN syntax is (overt or covert) movement of the PN to the D position (N-to-D), so that at LF the PN occupies the D-position in UG, a widely accepted proposal also adopted here, though in a modified form (more on this below). Beyond these very general hypotheses, the syntax of PN in a particular language depends on other general morphosyntactic properties of determiners and nouns. Thus, there are languages where article insertion is a semanticpragmatic process (cf. Roehrs 2006, Leu 2008). English is a case in point; in such languages, PNs do not take articles and it is the presence/ absence of the article which distinguishes between PNs and the CNs, as in (2). (2) a. Cat came in running. b. The cat came in running. In other languages, article insertion is more of a morphosyntactic process, having to do with agreement and locality problems. In languages of this type, PNs are always (e.g. Greek, Portuguese) or at least sometimes (e.g. Romanian) morphologically provided with an expletive definite article, which is erased upon raising to D (e.g. many Romance languages, such as Italian, illustrated above). In a subset of the languages which also use the article for syntactic purposes, nouns or nominal stems are specified for definiteness in the lexicon, as a further parametric property, characterizing Semitic languages (Hebrew, Arabic) and Romanian, in the Romance family (see Borer 1999 and Danon 2010 for Semitic languages, and Nicolae 2013, present volume for Romanian). In Romanian, nouns may enter the derivation with a valued uninterpretable definite feature [u+def][], realized as asuffixal definite article. As marked on nouns, definiteness is clearly a syntactic feature, since semantic definiteness is an attribute of definite determiners. Danon (2010) argues that the existence of a formal definiteness feature in the grammar of a given language must be demonstrated using specific diagnostics, rather than merely assumed. Among such diagnostics, Danon includes definiteness agreement of the attributive adjectives with the noun, a property of Semitic languages shared by Romanian prenominal adjectives. Further evidence for the presence of a syntactic definiteness feature in Romanian is supplied by the existence of a class of polydefinite constructions (see Isac, this volume; Nicolae 2013, this volume). The realization of this feature as a suffix on the noun imposes a particular definiteness valuation mechanism based on Local Agree, a mechanism at work for both common nouns and PNs (see Cornilescu & Nicolae 2011 for details). Summing up, the first hypothesis regarding the difference between CNs and PNs is that only the latter are inherently definite.

Classified Proper Names in Old Romanian

103

Longobardi (2008) revisits his analysis of PNs starting from the interpretative similarity between PNs and personal pronouns: both display direct reference and lack descriptive content. In his interpretation, the similarity between PNs and personal pronouns springs from the fact that both possess a [person] feature, which is also valued in D. Actually, Longobardi claims that [person] is the minimal content of the category D, and it is Person, not inherent definiteness, that distinguishes PNs from CNs. The presence of the definite article on PNs in some languages would then be a by-product of [person], since, as acknowledged by many researchers, [person] entails [definiteness] (Harley & Ritter 2002), but not the other way round. Naturally, since [person] is also valued in D, (some form of) N-to-D continues to exist. In sum, the second influential hypothesis is that the feature which distinguishes PNs form CNs is Person. The grammar of Romanian PNs apparently confirms Longobardi’s second hypothesis, in the sense that, whenever there are major distributional differences between PNs and CNs, they follow from the presence of Person, not definiteness. The definite article on PNs, i.e. the proprial definite article, is an expletive element with a morphosyntactic role, as has often been remarked in work on Romanian or other languages (Tomescu 1998; Cornilescu&Nicolae 2015 for Romanian; cf. also van Langendonck 2007). Intuitively, the special property of the proprial definite article is precisely to signal [person] on nouns. At first sight, it seems counterintuitive to accept that the proprial definite article is an expletive, since a PN—whether endowed with a proprial article or not—expresses the very idea of existence and uniqueness of the referent, which is precisely the semantic content of the definite article (Russell 1905). Moreover, the syntactic position D is usually defined in UG as the locus of definiteness (Lyons 1999), and if PNs are DPs, the proprial definite article would be expected to have its usual semantic role. This is precisely the point made by the descriptive theory of PNs. A possible solution to this paradox is to redefine the minimal (semantic) content of the category D, as done by Gillon (2009) with reference to Squamish,2 so as to better understand under what circumstances an article counts as an expletive. The determiners of Squamish are not sensitive to the definiteness contrast, in the sense that the same determiner (e.g. ka or kwi) may be used both to introduce a new referent (the function of the indefinite article in European languages) and to make reference to a known object, unique and already mentioned in the discourse/context (the function of the definite article in European languages). Squamish cancels out the novelty/ familiarity distinction (Heim 1982), which is the content of the (in)definiteness feature. Thus, with respect to their content, the determiners of English 2  A Salish language spoken in British Columbia, Canada.

104

Cornilescu and Nicolae

and Squamish do not behave like a homogeneous class, and are not universally employed to express semantic (in)definiteness (cf. Lyons 1999). There is, however, a semantic property common to all determiners, which unifies this class, namely the fact that determiners introduce domain restrictions; more exactly, determiners combine with nouns which have a certain extension in context. According to Gillon (2009), a lexical item is a determiner if and only if it introduces a domain restriction (NP). The morphosyntactic reflex of the relation between the operator and the restriction is inheritance by the determiners of the noun’s φ-features. The definiteness feature in European languages includes both the domain restriction and uniqueness (i.e. the opposition novelty/ familiarity). One may wonder what the content of Squamish determiners is, except for the introduction of a domain restriction. According to Gillon (2009), Squamish determiners are sensitive to a deictic feature, and indicate proximity or distance with respect to the speaker, instead of familiarity or novelty. The minimal content of the category D in UG is thus the introduction of a nominal restriction. In Gillon’s (2009) terms, the definite article in familiar languages is defined by two semantic features: the nominal restriction and (Russellian) maximality / uniqueness. An expletive article may lose one or both features. In our opinion, the proprial expletive article suspends both semantic features, merely retaining a morphosyntactic role: it retains syntactic definiteness, with the incumbent properties, and the ability to be marked for Case (see Stan 2015 on the relation between definiteness and Case marking in Romanian). The question arises as to why the article on PNs is expletive, if present. Intuitively, this follows from the absence of the descriptive content of the PN, which picks out no restriction (set) in the context. The result of the absence of descriptive content is that PNs cannot have interpretable φ-features and cannot in and of themselves transmit their φ-features to the article. The φ-features on the PN itself are uninterpretable or, rather, unspecified. One might say that PNs are morphologically opaque: by themselves, they cannot indicate grammatical gender and, as a result, cannot vary according to number, given that Number always selects Gender, at least in Romance (see Picallo 2008 for Romance; Tănase-Dogaru 2009 for Romanian). With Romanian PNs, absence of both grammatical and semantic gender is apparent in at least two situations: (i) Their morphophonemic structure does not determine grammatical gender, as CNs normally do. The PN Toma in (3a) ends in the vowel -a, which is normally interpreted as the singular feminine form of the definite article (or at least as a feminine inflectional marker) (compare with (3b)). Nevertheless, the PN Toma is usually assigned to male referents, and this is visible in the masculine form of the predicative adjective in copular sentences (3a). The

Classified Proper Names in Old Romanian

105

uninterpretable feminine gender feature in the article however determines the feminine Genitive-Dative Case form, as in (3c) (also compare with (3d)). (3) a. Toma este viteaz / *vitează Thomas is courageous.M courageous.F b. Maria este vitează / *viteaz. Mary is courageous.F courageous.M c. Duminica Tomii / Tomei Sunday.DEF Thomas.GEN d. rochia Mariei dress.DEF Mary.GEN In the same vein, the PN Carmen in (4) ends in a consonant, a typically masculine morphophonemic form, but it is mostly assigned to feminine referents, as again shown by adjectival agreement. (4) Frumoasa Carmen este lingvistă. beautiful.DEF.F Carmen is linguist.F (ii) Contemporary Romanian possesses quite a few PNs conventionally compatible with referents of both sexes (e.g. Irinel, Alex, Leonida, Pusi), in spite of their morphophonemic form. In other words, the grammatical gender of these PNs varies. (5) a. Irinel e doctor / doctoriță. Irinel is doctor(M) doctor(F) b. Alex e frumos / frumoasă. Alex is beautiful.M beautiful.F We claim that precisely because of their unspecified φ-features, PNs require the presence of a classifier, as un/o ‘a’ in (6). The classifier introduces the intended interpretable grammatical gender/number features of the expression, as will be discussed below. (6) a. (Doctorul) Irinel este un bun specialist. Doctor(M).DEF Irinel is a good.M specialist(M)

106

Cornilescu and Nicolae

b. (Doctorița) Irinel este o bună specialistă. Doctor(F).DEF Irinel is a good.F specialist(F) To conclude, PNs have a specific feature matrix which includes: an interpretable [person] feature, as well as an entailed uninterpretable definiteness feature, (often) realized as an expletive article. The functional structure of nominal phrases headed by PNs includes (at least) a Classifier Phrase (ClassP) that supplies interpretable φ-feature, and a DP projection, as in (7). (7) DP > ClassP [iφ] > PN [iperson, uφ, (u+def)] 3

Expletive Article vs. Referential Article

In agreement with previous research on Romanian, we preserve the hypothesis that the definite article is affixal and the definiteness feature is valued in a strictly local configuration (Cornilescu & Nicolae 2011). In this section, we review the syntactic differences between the expletive and the referential definite article. While in languages like English it is the absence of the definite article which announces a PN, in Romanian the PN is hailed precisely by the very different distribution of the definite article with PNs as compared with CNs. Simplifying somewhat, we might say that the referential and the expletive articles are morphologically identical, but differ considerably from a syntactic perspective. As with the referential article, the morphology of the expletive article expresses gender (9), number (10), Case (11), and even definiteness (12) differences (if the name also has an indefinite form); the singular paradigm of the definite article is given in (8). An important aspect in the morphology of Romanian PNs (detailed in the second part of the paper) is the difference in the realization of the Case paradigm between masculine and feminine proper names.3 Briefly, feminine PNs have the same morphology as feminine definite common nouns, with limited exceptions exhibiting the regular enclitic article throughout the declension (12). In contrast, masculine PNs have lost the article in the Nom-Acc Cases, and show a proclitic definite article in the Gen-Dat cases (13). The masculine PNs with a feminine form (14a) exhibit a dual GenDat Case marking pattern, being compatible both with the feminine strategy (14b) and with the masculine strategy (14c). 3  In the Gen-Dat., the Case inflection is enclitic for feminine PNs, and proclitic (i.e. a proclitic article) for masculine PNs. Old Romanian also had the option of an enclitic definite article on masculine PNs (see section 5 below).

Classified Proper Names in Old Romanian

107

(8) the paradigm of the singular definite article as realized on PNs: masculine: -(u)l feminine: -a (9) Ionel / Ionel-a John Johana-DEF (10) Ionescu / Ioneșt-i-i Ionescu Ionescu-PL-DEF (11) Mari-a / Mari-ei Mary-DEF Mary-DEF.GEN (12) Aceeași Mărie cu altă pălărie. / Maria the-same Mary with different hat Mary-DEF ‘The same Mary with a different hat’ ‘Mary’ (13) Ionel // caietul lui Ionel John notebook.DEF DEF.GEN John ‘John’ ‘John’s notebook’ compare with elevul // caietul elevului student.DEF notebook.DEF student.DEF.GEN ‘student’ ‘the student’s notebook’ (14) a. Toma, Preda, Șendrea b. Duminica Tomii / Tomei Sunday.DEF Thomas.GEN ‘Thomas’ Sunday’ c. caietul lui Toma notebook.DEF DEF.GEN Thomas ‘Thomas’ notebook’ Despite morphological similarity, the syntactic differences between the two types of articles are considerable. Notice first the co-occurrence of a prenominal definite adjective with an article-suffixed PN (15a), against the constraint of spelling out the definiteness feature only once on the highest [+N] term of the DP, see (15b) vs. (15c)) active with DPs headed by CNs.

108

Cornilescu and Nicolae

(15) a. frumoasa Maria vs. b. *frumoasa fata beautiful.DEF girl.DEF beautiful.DEF Mary.DEF ‘beautiful Mary’ c. frumoasa fată beautiful.DEF girl.DEF ‘the beautiful girl’ Secondly, PNs may also be followed by definite suffixed adjectives (16a), unlike CNs (16b); occurrence in the adjectival article construction headed by cel (‘the’) is available for both PNs (17a) and CNs (17b). (16) a. Mihai Viteazul Michael Brave.DEF ‘Michael the Brave’ b. *conducătorul viteazul // conducătorul viteaz leader.DEF brave.DEF leader.DEF brave ‘the brave leader’ (17) a. Ștefan cel mare Stephen CEL great ‘Stephen the Great’ b. conducătorul cel mare leader.DEF CEL great ‘the great leader’ Furthermore, PNs may appear not only in double definite, but also in polydefinite structures. A polydefinite structure has two definite adjectives preceding the PN, as in (18a) vs. (18b). The adjectives are stacked, not coordinated. In contrast, in the CN structure only the highest adjective is definite (19b) vs. (19a), unless there is coordination (19c). (18) a. ticăloasa Sângeroasa Maria mean.DEF bloody.DEF Mary.DEF ‘mean Bloody Mary’ b. *ticăloasa Sângeroasă Maria mean.DEF bloody Mary.DEF

Classified Proper Names in Old Romanian

109

(19) a. *ticăloasa sângeroasa femeie mean.DEF bloody.DEF woman b. ticăloasa sângeroasă femeie mean.DEF bloody woman ‘the mean bloody woman’ c. ticăloasa, sângeroasa femeie mean.DEF bloody.DEF woman ‘the mean, bloody woman’ In a second polydefinite structure, unlike CNs, PNs may be preceded and followed by nouns or adjectives suffixed by the definite article, as in (20). (20) a. mult discutatul Călin Nebunul much discussed.DEF Călin Madman.DEF ‘the intensely discussed Călin the Madman’ b. regretatul Ion Zamfirescu criticul literar late.DEF John Zamfirescu critic.DEF literary ‘the late literary critic John Zamfirescu’ Another difference occurs in PPs, where the preposition may be followed by a definite PN (21b), though it may not by followed by a definite (unmodified) CN (21a). (21) a. Am văzut-o pe fată / *fata. AUX.PERF.1SG seen=ACC.3SG.F DOM girl girl.DEF



b. Am văzut-o pe Maria. AUX.PERF.1SG seen=ACC.3SG.F DOM Mary.DEF ‘I have seen the girl/Maria.’

One more piece of evidence that the article on PNs is not referential comes from its possible co-occurrence with an indefinite article, as in (22). (22) O a doua Maria Sângeroasa nu există în istoria a second Mary.DEF Bloody.DEF not exists in history.DEF Angliei. England.DEF.GEN ‘A second Bloody Mary does not exist in the history of England.’

110

Cornilescu and Nicolae

To conclude, the syntax of the two articles is completely different. In our view, these differences follow from the presence of the [person] feature, typical of PNs. 4

The Internal Structure of PNs: Classifiers and Close Appositions

Given the numerous syntactic differences between PNs and CNs, an analysis of PNs as definite descriptions, which assimilates PNs to definite CNs in an attempt at unification and simplicity (Matushansky 2006), encounters substantial empirical problems. We therefore opt for a syntax that acknowledges the difference between PNs and definite descriptions. Moreover, in line with Longobardi (1994, 2008) and Hinzen (2007), we believe that the much discussed rigidity of the PNs is an effect of their syntax, which somehow guarantees their atomicity, hence their lack of variation in different contexts or possible worlds. Of the several internalist accounts of PNs that we are aware of (Longobardi 1994, 2008; Uriagereka 1998; Hinzen 2007) we have opted for a variant of Longobardi’s account, as already stated. Essentially, Longobardi’s view is that definite DPs are compositional, minimally formed out of an operator (D) + restriction (NP). The operator takes scope at LF (adjoining to the DP; cf. Dobrovie Sorin 1994) and finally to the sentence (Heim 1982). At LF, the D position is filled by a variable and the DP exhibits a quantifier + variable structure. It is the operator + variable + restriction syntactic configuration which secures variability of reference for definite descriptions, different members of the restriction set possibly being picked in different possible worlds (23). (23) DP the DP D0 NP x king In sharp contrast to definite descriptions, PNs do not have a compositional interpretation; the PN itself fills the D position through (overt or covert) movement of the PN to the D position. In the resulting configuration, there is no variable and no nominal restriction. The PN is found in an unanalyzable atomic structure. In Longobardi’s view, it is precisely the absence of an operator-variable-restriction structure which produces the rigid designation of PNs (24).

Classified Proper Names in Old Romanian

111

(24) DP D0 NP [person] Petru N Thus the essence of PN syntax is that the PN itself becomes an exponent of the D category via raising to D, thus cancelling out the quantificational structure. Cross-linguistic differences of PN syntax follow from several parameters, such as: (i) the interplay of Person and Definiteness; (ii) the level where N-to-D occurs, syntax or LF. According to Longobardi (2008), in languages where PNs do not raise in syntax, a CHAIN is formed between the D position and the lower NP position. The same chain values the uninterpretable φ-features in D. The D position will be filled by the PN at LF. This is the case of English-type languages. There are certain problems that must be solved before extending Longobardi’s analysis to Romanian. At this point it is enough to show that Romanian PNs do move inside the DP, as shown by their initial (25b) or final position with respect to classifiers (25a). (25) a. actorul George Vraca (PN in a lower position) actor.DEF George Vraca ‘actor George Vraca’ b. George Vraca actorul George Vraca actor.DEF ‘George Vraca the actor’

(movement of the PN to the D-domain)

In the remainder of the section, we detail these two PN patterns in Romanian, neither of which is possible for definite CNs; compare (26) to (25). (26) a. *actorul eroul actor.DEF hero.DEF b. *eroul actorul hero.DEF actor.DEF

112

Cornilescu and Nicolae

4.1 On Classifiers and PNs As in Cornilescu (2007), we assume that in the functional structure of a PN there is a qualitative classifier (Class). This hypothesis reflects the intuition that an adequate understanding of a PN requires understanding the type of entity denoted by the PN, i.e. there is an implicit categorization process, at the end of which the PN is minimally supplied with interpretable φ-features. It follows that Class is a nominalizer, a form of little n, which supplies valued interpretable φ-features for the PN, in agreement with the entity denoted by the PN in some particular context of use. The presence of Class accounts for the many instances of gender variation in the use of the same PN. Thus in (27a), Venus denotes a name, having what Kaplan (1964) called syntactic reference. Accordingly, its grammatical gender is neuter. In (27b), Venus refers to the goddess, its grammatical gender is feminine, and there is differential object marking. (27) a. Venus are / este format din cinci litere. [neuter] Venus has is formed.N of five letters ‘Venus has / is composed of five letters.’ b. Toți o iubeau pe Venus. [feminine] all her.ACC.3SG.F= loved.3PL DOM Venus ‘Everybody loved Venus.’ The PN enters the derivation with an unspecifed gender feature, the gender of the DP being given by Class. Gender specification makes possible number specification, given that, in Romance, the NumP selects for a Gender Phrase (Picallo 2008; cf. also Bernstein 1993). Class has an EPP-feature triggering movement of PN to Spec, ClassP, as in (28), or merger of a lexical classifier as will be shown below. Technically, Class selects an N [person], an analysis which may be extended to personal pronouns as well (Cornilescu & Nicolae 2014). (28) a. DP D0 ClassP [uPers] [uφ] Class0 NP [iφ] [ipers] ([GENDER: NEUT] [φ___] ([GENDER: __) FEM] Venus [EPP]

Classified Proper Names in Old Romanian

113

b. DP D0 ClassP [uPers] [uφ] NP Class’ [ipers] [φ___] ([GENDER: __) Class0 tNP [iφ]

([GENDER: NEUT] FEM] [EPP] Venus The suggestion of using classifiers to specify the nature of the entity denoted by the PN is directly supported by languages that employ classifiers to denote types of entities, such as Mandjaku in Kihm (2005: 474).4 A root like Ölik (‘to draw water’) may be classified in different manners, by choosing different nominalizing affixes such as pë- / i- / m- / ka-, thus deriving words like pë-lik / i-lik (‘well(s)’), m-lik (‘water’) și ka-lik (‘fruit juice’). These examples show that classifying elements have a descriptive content, denoting a certain kind of object. Our proposal is that classifiers also operate on PNs, specifying the type of entity which bears the respective name. In addition to this semantic role, classifiers play an essential syntactic role, specifying interpretable φ-features for (particular uses of) the PN, interpretable gender in the first place. Class contains semantic interpretable features such as PERSON (anthroponyms), PLACE (toponyms), OBJECT (e.g. the word Venus), and so on. Agree between Class and the PN endows the latter with interpretable φ-features, necessary for DP-internal and DP-external concord processes. As will be immediately shown, the existence of classifiers is fully confirmed by the category of “descriptive PNs” / “complex PNs” in the sense of Soames (2002). 4.2 Complex Proper Names 4.2.1 Overt Classifiers A central set of data, whose significance has not always been appreciated, is that of “partially descriptive names”, first included in the semantic category of rigid designators by Soames (2002), who offers a semantic account of sentences with descriptive PNs. A partially descriptive proper name n is made up of a CN + a bare PN, as in (29), and is semantically associated with both 4  A Bak language, spoken in Guinea-Bissau and Senegal.

114

Cornilescu and Nicolae

a descriptive property PD, contributed by the CN, and some referent Object, contributed by the PN. The proposition expressed by a sentence n is F, with n a partially descriptive PN is the same as that expressed by the proposition: [the x: PD x & x=y] Fx. This proposition is true at w iff the Object has the property PD and Fx is true at w. In other words the descriptive content (PD ) is a means of fixing the referent (in the sense of Kripke 1980) but is not part of the proposition Fx evaluated as true or false in a context, a proposition which has an atomic subject (PN). From now on, we will refer to partially descriptive PNs as “complex PNs” (=CPNs) or “descriptive names” and will be paying special attention to their syntax. The category of CPNs brings valuable information regarding the syntactic structure of PNs. As mentioned, CPNs are made up of a common name plus a proper name. The CN designates a social role (kinship, profession, institutional role, etc.), a type of location (city, street, village, river, etc.), other entities (institutions, names, planets, etc.), as in (29). (29) a. Profesorul Ionescu; Regina Elisabeta; Mătușa Tamara professor.DEF Ionescu queen.DEF Elisabeth aunt.DEF Tamara b. Orașul București; Strada Paris; Râul Dâmbovița city.DEF Bucharest street.DEF Paris river.DEF Dâmbovița c. Teatrul I.L. Caragiale; Planeta Venus theatre.DEF I.L. Caragiale planet.DEF Venus It has been suggested that these are appositive constructions rather than CPNs (cf. GLR 1966). English however shows their status as PNs through the obvious absence of the definite article: (30) Prince Charles not*The Prince Charles The hypothesis that there is a classifier in the extended projection of PNs offers a natural analysis of CPNs, since the classifier is overt in their structure. Intuitively, the CN in the structure of CPNs has a classifying role, denoting the kind of entity which is the referent of the PN as in Professor Ionescu. If for simple PNs it is the PN itself that values the classifying feature of Class0 via movement, in the case of descriptive PNs, the CN merges in Spec, ClassP, since the semantic feature of Class0 (semantic PERSON for anthroponyms, semantic PLACE for toponyms, etc.) is among the semantic features of the CN (prince, king; street, park; etc.). In other words, with CPNs, the classifier is overt.

Classified Proper Names in Old Romanian

115

In CPNs, the PN is not an appositive structure because appositional modification is a form of DP-modification, in contrast with NP-modification (Potts 2005, cf. Larson & Marušič 2004): both the modifier and the modifee are DPs, often interchangeable, see (31). (31) a. Îl admirau pe autorul piesei, him.ACC.3SG.M= admired.3PL DOM author.DEF play.DEF.GEN cel mai bun dramaturg englez. the best playwright English ‘They admired the author of the play, the best English playwright.’ b. Îl admirau pe cel mai bun dramaturg him.ACC.3SG.M= admired.3PL DOM the best playwright englez, autorul piesei. English author.DEF play.DEF.GEN ‘They admired the best English playwright, the author of the play.’ Notice that the PN in the descriptive PN construction is an NP, not a DP, as can be seen from the impossibility of replacing the PN with a definite noun or a personal pronoun (i.e. a complete DP) in this construction, as in (32). (32) a. profesorul Popescu professor.DEF Popescu b. *profesorul el / muncitorul professor.DEF he worker.DEF In genuine appositive structure, PNs are interchangeable with personal pronouns and represent DPs; the two DPs are separated by an intonational break, as in (33). (33) a. Brâncuși, cel mai mare sculptor român// Brâncuși the greatest sculptor Romanian cel mai mare sculptor român, Brâncuși the greatest sculptor Romanian Brâncuși ‘Brâncuși, the greatest Romanian sculptor’ b. el, cel mai mare sculptor român// he the greatest sculptor Romanian

116

Cornilescu and Nicolae

cel mai mare sculptor român, el the greatest sculptor Romanian he ‘he, the greatest Romanian sculptor’ Classifiers may be considered semi-lexical categories, in the sense of Löbel (2001): semi-lexical nouns become relational, selecting a non-theta marked complement. They are not referential, since in a phrase like Mister Popescu, as shown in Soames’ (2002) analysis given above, there is only one referent, namely that of the PN. The descriptive term may sometimes be abbreviated, which could be interpreted as a linguistic marker of a functional element (dna. Ionescu, Dr. Jones, Mr. Smith). Finally, with certain (masculine) classifiers, the definite article may be dropped, which shows that the classifiers has become a part of the PN, see (34). (34) Sunt doctor(ul) Ionescu. I.am doctor(.DEF) Ionescu ‘I am doctor Ionescu.’ One last remark: descriptive nouns may head an extended ClassP. In conventionalized CPNs, the classifier is simple, but in the classifier + PN structure the classifier may attract any modifier, prenominal (35a) or postnominal (35b), and the classifier structure is recursive (36). (35) a. profesorul de matematică Marin Miroiu professor.DEF of mathematics Marin Miroiu omul de afaceri Dinu Patriciu man.DEF of business Dinu Patriciu ambasadorul UNESCO al Romaniei la Paris ambassador.DEF UNESCO GEN Romania.GEN at Paris Valentin Lipatti Valentin Lipatti b. celebrul actor (de drama și comedie) Toma Caragiu famous.DEF actor of drama and comedy Toma Caragiu ilustrul ambassador Mihai Zamfir illustrious.DEF ambassador Michael Zamfir (36) a. Domnul Ionescu mister.DEF Ionescu

Classified Proper Names in Old Romanian

117

b. Domnul doctor Ionescu mister.DEF doctor Ionescu c. Domnul doctor rezident Ionescu mister.DEF doctor resident Ionescu The presence of the nominal classifier above the PN is also shown by the possible presence of intensional modifiers in the class fost(ul) ‘(the) former, fals(ul) ‘(the) false’, pretinsul ‘(the) alleged’, adevăratul ‘(the) real/genuine’ etc. These adjectives are reference modifiers, not referent modifiers, in the sense that they are functions whose argument is a property, not an object. (37) falsul (țar) Dimitri false.DEF tsar Dmitriy pretinsul Boris alleged.DEF Boris fostul (rege) Mihai de Romania former.DEF king Michael of Romania adevăratul Sfânt(ul) Niculai genuine.DEF saint Nicholas It is plausible that in examples like (37) the adjective is a modifier of the classifier, overt or null. 4.2.2 Valuing Person and Definiteness. A Few Important Assumptions The time has come to have a closer look at the feature valuation mechanisms involved in generating complex PNs. We will adopt the following hypotheses: (i) Anthroponymical PNs always value a Person feature in D. (ii) Since [person] entails [definiteness], a syntactic feature in Romanian, definiteness valuation is always involved in deriving PNs. The mechanism of definiteness valuation is the same as for DPs headed by CNs, always involving Local Agree (see Cornilescu & Nicolae 2011, and Nicolae, present volume). The definiteness feature in D is valued by a definite [+N] constituent in the specifier of the projection immediately below D. The definite feature should be realized only once, on the highest [+N] constituent. Apparent exceptions in polydefinite structures will be interpreted as PF effects of the coexistence of [person] and [definiteness]. More specifically, the definiteness-realizing morpheme is not erased if definiteness is dominated by [person] in the feature hierarchy. This amounts to saying that the

118

Cornilescu and Nicolae

d­ efiniteness morpheme is realized or at least realizable (other things being equal) on nouns (or adjectives) which are (part of) proper names or are treated as such. However, it remains true that outside of the PN itself, only one definite constituent, the one in the highest specifier projection below D realizes definiteness. [Person] and [definiteness] may have distinct valuation properties. The fact that [person] and [definiteness] are different features, even if they are both valued in D, is immediately apparent in English. Definiteness of a common noun is a strong feature and requires the overt presence of an article throughout narrow syntax. [Person] may be valued through Long Distance Agree, but Long Distance Agree is followed by movement to D at LF. This contrast is apparent in pairs like the following: (38a) with Long Distance Agree is a PN checking [person], (38b) is a PN reanalyzed as a CN, hence the obligatory merger of the article. (38) a. Tall Kim b. the tall Kim (iii) In line with other researchers (Julien 2005; Roehrs 2006; Leu 2008), we will also assume that articles raise to the D position from a lower Article Phrase, on the edge of the lexical nP domain. The assumed functional structure of a PN is then the following: (39) DP > ArtP > ClassP > (NumP>) NP 4.2.3 Deriving CPNs in the Structure [Classifier + PN] Recall the two complex proper name constructions mentioned above, which differ in the position of the PN with respect to the classifier. To account for this contrast, we propose that, while PNs always check Person, the [person] feature in D is strong only in examples of type (40b), which feature the raising of the PN across the classifier. (40) a. actorul George Vraca actor.DEF George Vraca b. George Vraca actorul George Vraca actor.DEF c. *actorul el / el actorul actor.DEF he he actor.DEF

(the PN is in a lower position) (the PN raises to the D domain)

Classified Proper Names in Old Romanian

119

Notice that only in configuration (40b) can the PN be substituted by a personal pronoun (40c) and it has been shown that Romanian personal pronouns also raise to D (Cornilescu & Nicolae 2014), if not further. Thus in (40b) the Person-bearing element is in initial position. In (40a) the definite classifier (not endowed with [person]) occupies the highest position. Before giving a (roughly) Longobardian analysis of Romanian PNs, we should somewhat reconsider Longobardi’s analysis, from two points of view: First, late Minimalism (Chomsky 2000 and ssq.) dispenses with covert movement in favour of Long Distance Agree, involved in valuing weak features of heads. Hence, N(P)-to-D should be used only if there is evidence for it in narrow syntax. In the particular case of Romanian it has long been known that both nouns and adjectives move as phrases (Grosu 1988; Cornilescu 1992; Cinque 2004 i.a.) hence, an N-to-D analysis for PNs is undesirable. Second, the LF representation of a DP headed by a PN should be such that the D position has no content, so that no quantificational interpretation arises. N-to-D achieves this end by having a meaningless (i.e. featureless) constant (the PN) fill the D position, an operation which not only fills D, but also vacates the NP restriction. We propose an alternative implementation of the same idea. At LF, D0 contains no interpretable features, being radically empty, since valued uninterpretable features are deleted. At the same time, since PNs are DPs, if the D head is not itself visible at LF because it is empty, it is the Spec, DP position that must be filled by phrasal movement. Simply put, we propose the replacement of the PN LF representation in (41a), by the one in (41b). Essentially, we capitalize on the fact that valued uninterpretable features are cyclically erased, and stipulate that the [person] feature on PNs is interpretable, while the Person feature in D is uninterpretable and strong (EPP). The representation in (41b) has the same interpretative property as the one in (41a), D is featureless and may not count as an operator, and the restriction position (lexical NP) has been vacated. (41) a. DP N0+D0 NP [person] John tn

120

Cornilescu and Nicolae

b.

DP NP D’ [iperson] John D0 [uperson] [EPP]

tnp

We adopt (41b) for Romanian, because of the activated definiteness parameter (see section 2 above), and in this parametric setting, [person] is accompanied by syntactic definiteness. In our interpretation, it is the feature structure of D and of the PN which ultimately determines the internal syntax of PN-headed DPs, more exactly the position of the PN with respect to the classifier. As far as [person] is concerned, we will assume that, as in English, [person] is always interpretable on the PN, but may be valued or unvalued. Moreover, when [person] on PN is unvalued, it will be valued in the same local configuration as definiteness. Again as in English, [person] in D is uninterpretable and unvalued. As to definiteness, definite Romanian nouns are uniformly specified as [u+def] and valuation takes place in a strictly local configuration. The definiteness feature in D is valued by the specifier of the projection under D and it may as well be valued by the head of this first lower projection. Given that the proprial article is expletive, the definiteness feature in D is also uninterpretable. Since in D, both [person] and [definiteness] are uninterpretable, at least one of them (possibly) both must have the EPP property to secure movement to Spec, DP, and thus visibility of the DP projection. Consider now the following examples which illustrate the [classifier + PN] configuration. Example (42a) is a bare NP configuration, (42b) involves a nominal classifier, and (42c) involves an evaluative adjective, representing (part of) the ClassP. (42) a. Maria Mary.DEF b. secretara Maria secretary.DEF Mary.DEF c. marele Ștefan great.DEF Stephen (42b) and (42c) have a more perspicuous structure, exhibiting the [classifier +PN] configuration. Since the PN does not occupy the first position, it is safe to assume that the uninterpretable [person] feature in D does not have the

Classified Proper Names in Old Romanian

121

EPP property; to secure visibility of the DP projection, the (uninterpretable) [definiteness] feature in D must have the EPP property. In other words, the feature structure of D is as follows: D[[udef:__, EPP], uPers:__]]. The PN has avalued interpretable[person] feature, which entails the presence of the usual uninterpretable unvalued [definiteness] feature of Romanian nominal stems, i.e. Maria [iperson:val, u+def]. We can now detail the derivational steps involved in generating (43b), where the classifier is overtly expressed and undergoes definiteness agreement, the feature [u+def] becoming part of the classifier’s matrix. Recall that PERSON given in small caps represents semantic person, an attribute of Class0 which secures the selection of an anthroponymic name (PLACE ensures the selection of a toponym, etc.). The first step of the derivation is as described above: a ClassP supplying interpretable φ-features takes as its complement the PN equipped with Person and Definiteness. (43) ClassP NP Class’ PERSON Gen:F Class0 NP Nr:SG PERSON [iperson: val] [u+DEF] Gen: F [u+DEF] Nr: SG secretara Maria At the next step, the Art head merges so that both of the unvalued uninterpretable [definite] features under ClassP are valued (44a). Since definiteness in ArtP is strong, after definiteness valuation by Agree, ClassP raises to Spec, ArtP (44b). (44) a. ArtP Art0 ClassP [idef:__] [EPP] NP Class’ PERSON Gen:F Class0 NP Nr:SG PERSON [iperson: val] [u+def] Gen: F [u+def] Nr: SG secretara Maria

122

Cornilescu and Nicolae

b. ArtP ClassP

Art’

NP Class’ Art0 tClassP PERSON [u+DEF] Gen:F Class0 NP [EPP] Nr:SG PERSON [iperson] [u+DEF] Gen: F [u+DEF] Nr: SG secretara Maria In the final step, D[[udef:___, EPP], [uperson:__]] merges and should get its uninterpretable features valued. It should be obvious that both features may be valued by Agree with constituents within ArtP. The Art head values definiteness in a strictly local configuration, since Art [i+def] is the head of the projection immediately below D. Next, ArtP is attracted to Spec, DP, since uninterpretable definiteness is strong. The [uperson:__ ] feature in D may be valued by the interpretable matching feature on the PN, which is in the c-­command domain of D (Long Distance Agree). (45) a. DP D0 ArtP [uperson: val] [EPP] ClassP

Art’

NP Class’ Art0 tClassP PERSON [i+DEF] Gen:F Class0 NP [EPP] Nr:SG PERSON [iperson] [u+DEF] Gen: F [u+DEF] Nr: SG secretara Maria b. DP ArtP secretara Maria

D’ D0 [uperson:val] [EPP]

tArtP

Classified Proper Names in Old Romanian

123

At the end of the derivation ArtP is in Spec, DP, there are no interpretable features in D and the NP position, i.e. the restriction, is empty, this being a legitimate LF representation for a DP which has rigid designation. In the [classifier + PN] structure, as seen in (42), the [person] feature of the PN plays an important syntactic and morphophonological role, even if it is interpretable and valued. Syntactically, [person] is responsible for the generation of the ClassP, as well as for introducing syntactic definiteness. Morphophonologically, [person] determines the conditions of realization of the definiteness feature, to which we now turn. In a DP headed by a CN, only the higher definiteness feature on the DP-initial constituent would have been realized. However, definiteness features which are dominated by [iperson] may or must be realized, irrespective of their position with respect to D. In our example (42b), definiteness is visible not only on the highest definite NP, the classifier secretara, but also on the feminine PN, Maria, where definiteness is below [iperson]. Structure (42b) is thus polydefinite. In contrast, in Modern Romanian, a valued definite feature ([u+def:val]) under [iperson] is not realized on a masculine PN in the Nom-Acc Cases (42c). When a classifier no longer realizes definiteness, behaving as a masculine PN, it becomes part of the PN, a complex PN which may be reclassified (46). (46) a. [ClassP rezidentul [PN Alin Popescu]] resident.DEF Alin Popescu b. [ClassP doctorul [PN rezident Alin Popescu]] doctor.DEF resident Alin Popescu Classifiers proper, which function as such remaining outside of the PN, are always definite because of definiteness agreement. This double treatment of categorizing phrases leads to instances of systematic variation in the realization of definiteness, as shown in (47). To maintain the generalization that with the exception of masculine “traditional” PNs, definiteness is realized only once, on the highest [+N] constituent, one should analyze doctorul in (47b) as part of the PN. In other words (47a) and (47b) have different constituent structure. (47) a. tânărul doctor Popescu young.DEF doctor Popescu [ClassP [tânărul doctor] Class0 [NP [N Popescu]]

124

Cornilescu and Nicolae

b. tânărul doctorul Popescu young.DEF doctor.DEF Popescu [ClassP [tânărul] Class0 [NP doctorul Popescu]] Conclusions so far: (i) Romanian PNs come from the lexicon with a (valued) interpretable [person] feature and an uninterpretable definite feature. (ii) [Person] is responsible for projecting ClassP, whose role is to supply interpretable φ-feature for the PN, which, by virtue of being meaningless, is also devoid of interpretable φ-features. ClassP is a functional projection typical of PNs. (iii) In Romanian-type languages (i.e., with the definiteness parameter), [person] entails [definiteness], realized as the suffixal definite article. Having no contentful features, the article is expletive so that the article-bearing PN is semantically atomic. The syntactic definiteness feature is active, and it induces definiteness agreement between the PN and the classifier, if the latter is overt. Definiteness valuation involves strictly Local Agree as with CNs. (iv) The main syntactic property of the expletive article is its occurrence in polydefinite DPs. This is a property which distinguishes DPs headed by CNs, which spell out definiteness only once on the highest [+N] term, and DPs headed by PNs. (v) Following Longobardi (1994, 2008) we have proposed an account in which PNs are rigid since they are atomic and do not exhibit a [quantifier + variable + restriction] structure. In our analysis, the D position is uninterpretable since it contains no contentful features. The PN (construction) moves as a phrase to Spec, DP, so that there is no restriction. 4.2.4 Integrated Appositions: The [PN + Classifier] Structure Romanian PNs also appear in a classifier construction with a second integrated apposition structure (in the sense of Potts 2005), which is available for both personal pronouns (48b) and PNs (48a), but not for CNs (48c). (48) a. Eminescu gazetarul este la fel de viu ca Eminescu Eminescu journalist.DEF is as alive as Eminescu poetul. poet.DEF ‘Eminescu the journalist is as alive as Eminescu the poet’

Classified Proper Names in Old Romanian

125

b. El lingvistul știe cum să scrie. he linguist.DEF knows how SUBJ write ‘He the linguist knows how to write’ c. *M-am întâlnit cu eroul poetul. REFL.1SG=AUX.PERF.1SG met with hero.DEF poet.DEF Just as with the [classifier +PN] construction, there is no intonation break between the PN and the appositive classifier, a hint that this is a close or integrated apposition. The striking difference with the preceding CPN structure lies in word order, in this second case the order being not [classifier + PN] (e.g. poetul Eminescu ‘poet.DEF Eminescu’), but [PN + classifier] (e.g. Eminescu poetul ‘Eminescu poet.DEF’). In Modern Romanian, this construction has the following properties: (i) It is available to both PNs (48a) and personal pronouns (48b), but not to common nouns (48c), as shown above. (ii) In the [PN+ classifier] construction, the classifier is prosodically marked, representing an information focus. Since this second structure allows both PNs and personal pronouns, but excludes definite CNs, it is likely to be based on the [person] feature, since this is the feature that distinguishes pronouns and PNs from definite descriptions. Moreover, since the PN or the personal pronoun occupies the first position in the DP, the [uperson] feature in D must be strong (endowed with the EPP diacritic). Accordingly, D will attract to its specifier (only) the constituent bearing the [iperson] feature, i.e. the PN. The feature structure of D is as before, except that both [uperson] and [udefinite] are EPP. We will assume that [uperson] is strong in D because, in this construction [iperson] on the PN is unvalued and valuation of [iperson] is strictly local. Accordingly, an integrated appositive construction as in (49) has the representation in (50). For (50), assume that ClassP has merged and there has been definiteness agreement. (49) Vraca actorul Vraca actor.DEF

126

Cornilescu and Nicolae

(50) ClassP NP Class’ Gen:m Nr:SG Class0 NP [u+DEF] Gen: m Nr: SG N0 [ipers: __] [u+DEF] actorul Vraca As in the preceding construction, the two [u+def] features of the Class and of the PN are deleted upon Agree with the interpretable counterpart hosted by the Art head. (51) ArtP Art0 ClassP [i+def] [EPP] NP Class’ [iφ:val] [u+DEF] Class0 NP [iφ:val] N0 Agree [iPers: __] [u+DEF] actorul Vraca At this point, the PN still has an unvalued feature, [ipers:__ ], which is too low to be accessible to D when it merges. To get its [person] feature valued, the PN raises to Spec, ArtP by means of an equidistance derivation (see Lasnik 2009). The resulting configuration is given in (52a). Note that, in this configuration, the interpretable feature on the PN is sufficiently close to D0 to secure the deletion of the unintepretable person feature hosted by D0 via Local Agree. Finally the PN legitimately moves to Spec, DP to satisfy the EPP property of the features in D (52b).

Classified Proper Names in Old Romanian

127

(52) a. DP D0 ArtP [upers] NP Art’ N0 Art0 ClassP [ipers:__] [i+DEF] [uDEF: val] [EPP] NP Class’ [iφ:val] Agree [u+DEF] Class0 tNP [iφ:val] Vraca actorul b. DP NP D’ N0 D0 ArtP [ipers:val] [upers:val] [u+DEF] [EPP] tNP Art’ Art0 ClassP [i+DEF] [EPP] NP Class’ [iφ:val] [udef:val] Class0 tNP Vraca actorul The noun phrase in Spec, DP continues to be interpreted as a proper name, even if the post-nominal classifier is stressed and highlights the nature of the referent. In terms of Soames (2002), the descriptive content of the classifier is a means of fixing a rigid referent. This is in line with the representation of the close appositive construction in (52), where there are no interpretable features in D so that no operator-variable structure is created. This derivation is not permitted to common nouns which do not have a [person] feature. Predictably, this structure, based on a feature common to PN in UG is presumably available in most languages that have a definite article, unlike the preceding structure,

128

Cornilescu and Nicolae

which is based on a syntactic definiteness feature, available only in certain ­languages (syntactic definiteness). This prediction is confirmed, as shown by the sameness of the examples in (53) and the contrast in (54). (53) a. Michael the Brave / Richard the Lion-Hearted b. Mihai Viteazul Michael brave.DEF (54) a. King John

/ *The King John

b. *Rege Ioan / Regele Ioan king John king.DEF John

(English) (Romanian) (English) (Romanian)

4.3 Other Types of Complex Proper Names There are also other types of complex PNs, a particularly frequent class being that based on possessive constructions such as in (55). From the syntactic point of view, these complex PNs are standard definite descriptions, containing two DPs, unlike the previously discussed structures, which contain only one DP. In the following examples the classifier is a nominal head which assigns genitive to the PN. (55) a. Cetatea Branului fortress.DEF Bran.DEF.GEN ‘Bran Fortress’ b. Apa Sâmbetei water.DEF Saturday.DEF.GEN ‘Sâmbăta river’ At the same time, since the PN itself is a nominal definite head, it may license the genitive of any other name, proper or common, this also being a productive source of complex PN formation. This is illustrated in (56). (56) a. Ion al Glanetașului John GEN Glanetaș.DEF.GEN b. Nică al lui Ștefan a Petrei / Apetrei Nică GEN GEN Stephen GEN Peter.GEN GEN-Peter.GEN

Classified Proper Names in Old Romanian

129

c. Ștefan al Moldovei Stephen GEN Moldavia.DEF.GEN d. Elisabeta a doua a Angliei Elizabeth the second GEN England Finally, as well-known, Romanian has a large number of so-called double definite constructions. PNs are expectedly possible in all of them, occupying the highest position in the double definite construction (Spec, DP). Since, as their name shows, these structures involve a double valuation of definiteness and proper names are syntactically definite, it is expected both that they are possible in these constructions, and that in these constructions they are interchangeable with common definite NPs. In (57) we list the double definite constructions and show the alternation of definite CNs and PNs in these structures. (57) a. the post-nominal demonstrative structure CN: eroul acesta hero.DEF this PN: Ștefan ăsta Stephen this b. the adjectival article construction CN: eroul cel viteaz hero.DEF the brave PN: Ștefan cel Mare Stephen the great c. the post-nominal ordinal numeral construction CN: competitorul al doilea competitor.DEF the-second PN: Elizabeta a doua Elizabeth the-second We conclude that double definite structures are not specific to PNs and do not deserve any special considerations in presenting the syntax of PNs. 4.4 Conclusions In this first part of the paper, we have reviewed the major syntactic constructions available to PNs in Modern Romanian. Peculiar to Romanian PNs is the interplay of [person] and [definiteness]. Romanian PNs are syntactically

130

Cornilescu and Nicolae

d­ efinite. As such they appear in all the so-called double definite constructions. In addition, they also appear in constructions where [person] plays the crucial role. These are classifier constructions: the [classifier + PN] construction and the [PN + classifier] construction. On the basis of this analysis in the next two sections we investigate Romanian PNs from a diachronic perspective. 5

Complex PNs in the History of Romanian

In this section we turn to the description of PN structures in Old(er) Romanian. Before we start, a methodological comment would be appropriate. For the diachronic analysis, we have analyzed a corpus of Romanian texts, ranging from the earliest attested original writings of the 16th century (the documents comprised in DÎ) to the late Wallachian chronicles of the 18th century / early 19th century. Generally speaking, there is continuity in the syntax of PNs, in the sense that the two constructions we consider typical of PNs have both been available ever since the oldest 16th century texts. However, at first sight one is struck by dissimilarities rather than similarities in the structure of anthroponimical PNs. The following important differences in the structure of CPNs between Old Romanian and Modern Romanian will be documented below: (i) In Old Romanian, CPNs could exhibit both the structure [classifier + PN] and the structure [PN + classifier]. In other words the more or less marked Modern Romanian “integrated apposition structure”, i.e. [PN + classifier], functioned as a genuine CPN in Old Romanian and held a central role in the economy of Old Romanian PN constructions. Modern Romanian has retained only the [classifier + PN] structure for PNs, and it is incumbent on us to understand why. Through the existence of CPNs exhibiting the [PN + classifier] structure, person names differed considerable from place names, the latter exhibiting the [classifier + PN] structure in Old Romanian as well. Exceptions exhibiting the [PN + classifier] structure for place names were insignificant; here is one in (58a), which was probably used under the influence of some foreign template. The same name appears in the regular [classifier + genitive PN] two pages later (58b). (58) a. la Cladova cetate at Cladova fortress b. Cetatea Cladovei fortress.DEF Cladova.DEF.GEN

(CAB 155) (CAB 156)

131

Classified Proper Names in Old Romanian

(ii) Secondly, in Old Romanian there is a sharp contrast between masculine and feminine PNs. There is first a considerable quantitative difference, feminine CPNs (like feminine bare PNs) being much less frequent than masculine ones. Interestingly, feminine CPNs strongly if not exclusively prefer the [classifier + PN] structure, where the classifier is often derived from a corresponding masculine form (e.g. jupân ‘M.boyar(d)’ → jupân-easă ‘F. titled-lady’; ban ‘M. ban’ → F. băn-easă), as in the following pairs of examples; compare the feminines in (59a,c) with their masculine counterparts in (59b,d). (59) a. băneasa Ilinca, sora jupânesii ban-F.DEF Ilinca sister titled-lady.DEF.GEN Stancăi Stanca.DEF.GEN b. Gheorghe banul Băleanul George ban.DEF Bălean.DEF Jupan Bărcan vistiiar boyar Bărcan treasurer c. călugărița Mariia nun.DEF Mary.DEF

(DÎ VIII)

d. Rafail călugărul Raphael monk.DEF

(DÎ VII)

(CAB 163)

(LB 79) (DÎ XXXVII)

Noticeably, in the pair in (59c–d) the masculine classifier follows the PN, while the feminine one precedes the PN. As a result, feminine CPNs have the same structure as in Modern Romanian, while masculine PNs of the type [PN+ classifier] exhibited in (59b) are no longer available. Thus, morphosyntactic changes primarily affect masculine CPNs, the loss of the [PN + classifier] structure being accompanied by the generation of new PNs, through the gradual change of final classifiers into PNs. We hypothesize that, in the [PN + classifier] structure, the PN status is directly signalled by the initial position of the PN with respect to the classifier. Hence, the definite article on masculine PNs and then on the (former) masculine classifier was no longer necessary and was gradually lost in Nom-Acc positions. Concomitantly, a proclitic definite article (lui) emerged with the Gen-Dat nouns for inflectional reasons specific to Romanian; namely, in this language, genitive morphology is realized only on articles in the masculine declension. In our interpretation, the morphophonological changes in the Case paradigm of masculine PNs and the

132

Cornilescu and Nicolae

loss of the definite article should be viewed as consequences of the existence of the [PN + classifier] structure. (iii) A third significant difference between Old and Modern Romanian (which will be discussed below) is the existence in Old Romanian of a large number of PNs based on genitive and locative constructions. The genitives have often been turned into PNs, as expected, given that the PN + genitive structure could be analyzed as a [PN + classifier] construction, while locatives have been retained for historical names (Ioan de Hunedoara ‘John of Hunedoara’), but have often replaced by place-derived adjectives (Țarigrădeanul ‘from Țarigrad’ DP), the root being verbal in nature; however, our data indicate that verbal roots do not need (and should not) be involved in the analysis of the deficient supines. Indeed, all the effects discussed in Iordăcheoaia & Soare can be differently derived, by identifying the parametric change that triggered the gradual restrictions we observed in the derivation of these DPs.

Two Dp Configurations For Supine-based Nouns

259

4 Analysis The point of revision we propose for Iordăcheoaia & Soare (2008) is the location of the Aspectual feature involved in the constraints noticed for the deficient supines: instead of attributing this feature to a syntactic AspP (i.e., outer Aspect), typical for the functional extended domain of verbs, we attribute it to a semantic specification on the aspectual value of the root (i.e., inner Aspect). We borrow the application of this distinction from Wiltschko (2014), where the inner Aspect maps the boundedness feature to syntax: regular DPs (e.g., supine nouns of sub-division (i)) are either + or – bounded; whereas the deficient supines have an exclusive [– bounded] value, which naturally constrains the phi-feature values, as well as the combinatorial possibilities with items valuing the outer Aspect in the clause (see also Hill & Dragomirescu 2014 and Dragomirescu & Hill 2014 for an application of this contrast to verbal supines). The remainder of this paper details the implementation of this proposal. The Universal Spine Hypothesis in Wiltschko (2014), illustrated in (22), allows for the same root to be mapped either on the nominal or the verbal spine. Accordingly, mixed verbal and nominal projections, as in (5), are excluded. (22) CP ----- Linking ----- KP







IP -------Anchoring ------DP

AspP ------- Viewpoint -------- PhiP



iAspP -------- Classification ------- iAspP

The basis of this Spine contains the inner Aspect, which may be substantiated differently in different languages. For example, inner Aspect shows the parametric option [+/− bounded] in English, but [+/− animate] in Blackfoot. Romanian has the same setting as English. Importantly, the parametric setting for inner Aspect constrains the type of features that are mapped upwards on the Spine.

260

Dragomirescu

How is this analysis correlated with our data? The diachronic data presented above show that the OR regular supine nouns project an inflectional field that maps a complete set of features, on a par with infinitive-based nouns. Hence, we can say that the inner Aspect of these supine roots are either + or – bounded, which entails free mapping of the phi-features, on the nominal spine, as PhiP, which corresponds to NumP in Iordăcheoaia & Soare. In the data, this is shown in examples where the regular supines behave as count nouns, and display singular or plural endings, as well as some variation in gender endings. These examples do not show any restrictions with respect to the classes of adverbs they may co-occur with in the clause, which is expected since PhiP is the counterpart of AspP on the verbal spine (i.e., unrestricted PhiP means unrestricted AspP when they co-occur). On the other hand, the restrictions illustrated for the deficient supine nouns indicate a deficiency at the level of the inner Aspect. More precisely, as shown in Iordăcheoaia & Soare, such nouns are systematically in the masculine singular, cannot be count nouns, and display non-trivial constraints on the type of adverbials they can co-occur with in the clause. These restrictions follow naturally if the inner Aspect value is set exclusively for [– bounded], since this parametric setting blocks the substantiation of phi-features upwards on the spine. Since PhiP is restricted (to the default masculine singular), its AspP correlate on the verbal spine is also restricted. Thus, the theory allows us to derive the deficient supine nouns from the regular supine nouns by a switch in the parametric setting for a more restricted value. Consequently, the formal representations in (4) and (5) must be revised as in (23) and (24), respectively. Crucially, this revision discards AspP from the extended domain of N, as the effects attributed to AspP can be captured through the setting for boundedness in the stem. (23) DP D NumP -l/-a Num ClassP [sg/pl] Class NP -u/-ă[m/n/f] -ø- răsărit- [+/– bounded]

Two Dp Configurations For Supine-based Nouns

261

(24) DP D -(u)l -ø-

NP citit- [– bounded]

In (23), -l and -a are the definite articles, the gender specification located in Class is u (răsăritul) or ă (răsărită). For phonological reasons, the inflectional ending -ă is not visible when the definite article -a is present (*răsărit-ă-a), hence the definite article replaces the inflectional ending and comes to cover its functions as well. In (24), u may represent the presence of Class and Num, but its valuation is fixed (i.e., the default M/N and obligatorily singular) and predictable from the [– bounded] value of the root. The comparison between (23) and (24) provides an important diachronic clue: the structure in (23) is the structure of the Latin nominal supine, which was flexible for morphological Case and aspectual interpretations, as shown in (25). This is in line with previous proposals in Dragomirescu (2013a,b) that the Romanian supine is inherited from Latin as a noun; hence the participial suffix -t/-s does not merge with the root in the syntax; rather, it is lexically assigned to the supine root. (25) a. Spectātum veniunt seen.SUP.ACC came ‘They come to see the show’ [single event interpretation] (Ov., A.A., I, 99, in Gildersleeve 1974 [1867]: 284) b. facile factu easily done.SUP.ABL ‘easy to do’ [habitual interpretation] (Class.Lat., in Woodcock 1959: 112) The ensuing question concerns the triggers responsible for the change in the value of the bounded feature of the supine stem. Although we do not have evidence for the changes (and their triggers) that occurred at the time when the configuration in (24) emerged, we can however comment on the factors that promoted this configuration. Namely, the class of verb based nouns increasingly favored the long infinitive stems over the supines. For example, the 16th century texts display pairs of infinitive and supine stems for the same noun, as shown in (26) to (28), and the infinitive version is favored.

262 (26)

Dragomirescu

Și înaintea născutului lu Hristos, and before birth.SUP.DEF.GEN.M/N.SG GEN Christ și după nașterea lui and after birth.INF.DEF.GEN.F.SG his.GEN ‘before and after Christ’s birth’ (CC1.1567–1568: 180v)

(27) a. Și unde iaste ertatulu and where is forgive.SUP.DEF.M/N.SG ‘And where is the absolution of sins?’ (CC2.1581: 118)

păcatelor sins.GEN

b. de ertarea păcatelorŭ of forgive.INF.DEF.F.SG sins.GEN ‘of the absolution of sins’ (CC2.1581: 118) (28) a. Dentâi, despre îngrupatul first about bury.SUP.DEF.M/N.SG ‘First, about Christ’s burial’ (CC1.1567–1568: 14r)

lu GEN

Hristos Christ

b. Acest obiceai și nărav al îngroparei this custom and habit GEN bury.INF.DEF.GEN.F.SG era al jidovilor was GEN Jews.GEN ‘This custom of the burial was Jewish’ (CC1.1567–1568: 16r) As early as the 17th century, the supine version is eliminated from the nominal paradigm while the long infinitive version is preserved. This may be a side effect of a more general trend in the grammar, whereby long infinitives lose their verbal categorization, being replaced by short infinitives (which are exclusively [V] categories), and later by subjunctives (Pană Dindelegan 2013). Hence, the long infinitives are massively re-categorized as nouns, and became the set option for the lexical renewal of the deverbal noun class. Most supine nouns have been replaced by their infinitive counterparts, although some have been preserved until Modern Romanian. The point is that the supine in (23) became unproductive because the competing infinitive provided the same configuration. On the other hand, the deficient supine nouns, with the configuration in (24), had no competition and continued to allow for lexical renewal along the same pattern. Furthermore, the set value for the inner Aspect in the root led, in certain contexts, to categorial ambiguity, in the sense that the root started to

Two Dp Configurations For Supine-based Nouns

263

generate derivations on both nominal and verbal spines in (22). Hence, supine clauses as in (3) emerged, and became productive in Modern Romanian. 5 Conclusions This paper revisited the analysis of supine nouns proposed in Iordăcheoaia & Soare (2008, 2011) from a diachronic perspective. First, this paper signaled the need to distinguish between two classes of supine nouns in Old Romanian, one with regular nominal properties, the other deficient. The former became unproductive, the latter is still productive in Modern Romanian. The investigation of the two classes of supines followed the formalizations in Iordăcheoaia & Soare (2008, 2011) insofar as the difference between regular and deficient supines has been captured in terms of presence or absence of NumP and ClassP in the structure of the DP. However, this paper argued against the use of AspP in the inflectional domain of the noun and argued for a different justification of the structural contrast between regular and deficient nouns. In our view, the lack of phi-features and the restriction to mass noun interpretation for the deficient supines follow from a change in the parametric setting for the value of the inner Aspect on the supine root: whereas regular supine nouns have an optional [+/– bounded] feature for the inner Aspect, the deficient supines have it set to [– bounded]. The type of aspectual adverbs with which the deficient supine may co-occur in the clause is naturally restricted on functional grounds (i.e., the correlation between AspP and PhiP as nodes mapping derivatives of the same primitive ‘viewpoint’ feature), and on semantic grounds (i.e., it is subject to the Principle of Compositionality, which causes incompatibility between non-boundedness and telicity). The diachronic changes witnessed for the Romanian nominal supine (simultaneous with the preservation of the infinitive nominals) offer an interesting picture of the way in which the functional structure of deverbal nouns can evolve. Following a universal path of evolution, the supine evolved in two opposite, but converging ways, from a bona fide noun to a defective, verbal noun, on the one hand, and to a defective verb form, on the other hand. Corpus AB.1706–1717 Bert.1774

Gregorian, M. (ed.). 1961. Anonimul Brâncovenesc, in Cronicari munteni II. Bucharest: Editura pentru Literatură: 273–352. Verebceanu G. (ed.). 2002. Viața lui Bertoldo. Un vechi manuscris românesc. Chișinău: Museum.

264 CC1.1567–1568

Dragomirescu

Drimba, V. (ed.). 1998. Coresi. Tâlcul Evangheliilor, in Tâlcul evangheliilor și molitvenic românesc. Bucharest: Editura Academiei Române: 31–187. CC2.1581 Pușcariu, S. & A. Procopovici (eds). 1914. Diaconul Coresi, Carte cu învățătură. Bucharest: Atelierele Grafice Socec & Co. CDicț.1691–1697 Gherman, A.-M. (ed.). 2001. T. Corbea, Dictiones Latinæ cum Valachica interpretatione. Cluj-Napoca: Clusium. CLM.1700–1750 Onu, L. (ed.). 1967. Miron Costin, Letopisețul tărâi Moldovei, in Opere alese. Letopisețul țărîi Moldovei. De neamul moldovenilor. Viiața lumii. Bucharest: Editura Științifică. CM.1567–1568 Drimba, V. (ed.). 1998. Coresi. Molitvenic românesc, in Tâlcul evangheliilor și molitvenic românesc. Bucharest: Editura Academiei Române: 189–211. PV.~1705 Gherman, A.-M. (ed.). 2010. Theodor Corbea, Psaltirea în versuri. Bucharest: Editura Academiei Române: 143–449. CV.1563–1583 Costinescu, M. (ed.). 1981. Codicele Voronețean. Bucharest: Editura Academiei. DÎ Chivu, G., M. Georgescu, M. Ioniță, A. Mareș, A. RomanMoraru (eds.). 1979. Documente și însemnări românești din secolul al XVI-lea. Bucharest: Editura Academiei. DPar.1683 Ungureanu, M. (ed.). 2012. Dosoftei, Parimiile preste an. Iași: Editura Universității “Alexandru Ioan Cuza”. NL.~1750–1766 Iordan, I. (ed.). 1955. Ion Neculce, Letopisețul Țării Moldovei, in Letopisețul Țării Moldovei și O samă de cuvinte. Bucharest: Editura de Stat pentru Literatură și Artă. PH.1500–1510 Gheție, I. & M. Teodorescu (eds). 2005. Psaltirea Hurmuzaki. Bucharest: Editura Academiei Române. PO.1582 Pamfil, V. (ed.). 1968. Palia de la Orăștie. Bucharest: Editura Academiei Române. RG.1711 Gregorian, M. (ed.). 1961. Radu Greceanu, Începătura ­istoriii, in Cronicari munteni II. Bucharest: Editura pentru Literatură: 5–272. SA.1683 Goția, A. (ed.). 2008. Monumenta linguae. Sicriul de aur. Cluj: Editura Napoca Star. SVI.~1670 Stanciu Istrate, M. (ed.). 2013. Varlaam și Ioasaf, in Reflexe ale medievalității europene în cultura română veche: Varlaam și Ioasaf în cea mai veche versiune a traducerii lui Udriște Năsturel. Bucharest: Editura Muzeului Național al Literaturii Române. ULM.~1725 Onu, L. (ed.). 1967. Grigore Ureche, Letopisețul Țării Moldovei. Bucharest: Editura Științifică.

Two Dp Configurations For Supine-based Nouns

265

References Alexiadou, Artemis. 1999. Remarks on the Syntax of Process Nominals: An Ergative Pattern in Nominative-Accusative Languages. Proceedings of NELS 29: 1–15. Borer, Hagit. 2005. Structuring Sense. Oxford: Oxford University Press. Brâncuș, Grigore. 2007 [1967]. O concordanță gramaticală româno-albaneză: modul supin. In Studii de istorie a limbii române. 167–173. Bucharest: Editura Academiei Române. Cornilescu, Alexandra. 2001. Romanian nominalizations: case and aspectual structure. Journal of Linguistics 37: 467–501. Diaconescu, Ion. 1971. Supinul în limba română din secolele al XVI-lea–al XVIII-lea. Analele Universității din București. Limba și literatura română 20: 151–163. Dragomirescu, Adina. 2013a. Particularități sintactice ale limbii române în context romanic. Supinul. Bucharest: Editura Muzeului Național al Literaturii Române. ―――. 2013b. Du latin au roumain. Une nouvelle hypothèse sur l’origine du supin en roumain. Revue de linguistique romane 77 (305–306): 51–85. Dragomirescu, Adina & Virginia Hill. 2014. A diachronic perspective on de-supine complements. Paper presented at DiGS 16, Budapest, July 3–5. Dragomirescu, Adina & Alexandru Nicolae. 2015. L’ellipse nominale avec article défini de l’ancien roumain au roumain moderne: Le cas du participe passé. In Éva Buchi, Jean-Paul Chauveau, Jean-Marie Pierrel (eds). Actes du XXVIIe Congrès international de linguistique et de philologie romanes (Nancy, 15–20 juillet 2013), 3 volumes. Strasbourg: Société de linguistique romane/ÉliPhi (in press). Gherman, Cristina. 2007. Schimbarea valorii gramaticale. In Magdalena PopescuMarin (ed.), Formarea cuvintelor în limba română din secolele al XVI-lea–al XVIII-lea. 271–296. Bucharest: Editura Academiei Române. Gildersleeve. 1974 [1867]. Latin Grammar. Third edition, revised and enlarged by B.L. Gildersleeve and Gonzales Lodge. London and Basingstoke: Macmillan. Hill, Virginia. 2013. The emergence of the Romanian supine. Journal of Historical Linguistics 3.2: 230–271. Hill, Virginia & Adina Dragomirescu. 2014. De-supine complements in Old Romanian. Paper presented at LSRL 44, The University of Western Ontario, Canada, May 2–4. Iordăchioaia, Gianina & Elena Soare. 2008. Two kinds of Event Plurals: Evidence from Romanian Nominalizations. In: Olivier Bonami, Patricia Cabredo Hofherr (eds), Empirical issues in Syntax and Semantics 7, 193–216. Paris: CNRS. ―――. 2011. A further insight into the syntax-semantics of pluractionality. Proceedings of SALT 21: 95–114. ―――. 2013. Deverbal Nominalization with the Down Operator in Romanian. Talk given at Going Romance, Amsterdam, November 28–30.

266

Dragomirescu

Nedelcu, Isabela. 2013. Particularități sintactice ale limbii române în context romanic. Infinitivul. București: Editura Muzeului Național al Literaturii Române. Pană Dindelegan, Gabriela. 2013. The supine. In Gabriela Pană Dindelegan (ed.), The Grammar of Romanian. 233–245. Oxford: Oxford University Press. Picallo, Carme M. 2006. Some notes on grammatical gender and l-pronouns. In Klaus von Heusinger, Georg A. Kaiser and Elisabeth Stark (eds), Proceedings of the workshop ‘Specificity and evolutional/emergence of nominal determination systems in Romance’. 107–121. Konstanz: Fachbeireich Sprachwissenschaft der Universität Konstanz. Stan, Camelia. 2003. Gramatica numelor de acțiune din limba română. București: Editura Universității din București. ―――. 2013. La nominalizzazione dell’infinito in rumeno—osservazioni diacronicotipologiche. Revue roumaine de linguistique 51.1: 31–40. Wiltschko, Martina. 2014. The universal structure of categories. Cambridge: Cambridge University Press. Woodcock, E.C. 1959. A new Latin syntax. London: Methuen and Co. Ltd.

Part 2 The DP and the Clause

∵

Object Pronouns in the Evolution of Romanian: A Biolinguistic Perspective Anna Maria Di Sciullo and Stanca Somesfalean We examine pronominal objects in Old Romanian and show that the fluctuation in their position (pre-/post- verbal) and in their form (clitic/strong pronoun) is the result of the Directional Asymmetry Principle (DAP), a complexity-reducing principle proposed in Di Sciullo (2011), according to which language evolution is symmetry breaking. We show that DAP is sensitive to both derivational and representational complexity. Under its effects, on grounds of derivational complexity reduction, Romanian lost the discourse-driven verb movement that yielded enclisis. On grounds of representational (sensori-motor) complexity reduction, Romanian lost the use of strong pronouns in contexts that now only allow clitics. Thus, a fluctuating phase in the evolution of pronominal objects is followed by a phase where a preponderant use is attested (i.e. proclitics in Modern Romanian). We confirm previous findings on the diachronic development of the Romanian DP under the effects of DAP, showing the role of complexity reduction in language change.

1

The Issue

In Old Romanian (OR), object personal pronouns in argumental position1 can be post-verbal, (1a), (2a), (3a), and pre-verbal (4a), (5a), with what looks like2 a strong preference for a post-verbal positioning for both clitics and strong pronouns. In contrast, Modern Romanian (MR) manifests an exclusive proclitic use in the same contexts (1b), (2b), (3b), (4b), (5b).

* This work is supported in part by funding from the Social Sciences and Humanities Research Council of Canada to the Major Collaborative Research on Interface Asymmetries, grant number 214-2003-1003, and by funding from the FQRSC on Dynamic Interfaces, grant number 137253. 1  Strong pronouns can be pre-verbal when they move to discourse related A’-positions, in which case they occur higher than negation and T related clitics. Those instances are not relevant to this discussion. 2  In the corpus we examined.

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_009

270

Di Sciullo and Somesfalean

(1) a. cu slavă priimiși mine [. . .] OR with glory received.2SG me.ACC ‘With glory you have received me.’ (Coresi, 137r) b. cu slavă mă primiși MR with glory me.ACC= received.2SG  ‘With glory you have received me.’ (2) a. Doamne, cântec nou cânt ție, [. . .] Lord, song new sing.1SG you.DAT ‘Lord, a new song I sing to you.’ (Coresi, 274r) b. Doamne, cântec nou îți cânt [. . .] Lord, song new you.DAT= sing ‘Lord, a new song I sing to you.’

OR

MR

(3) a. fără dereptate mânară-mă, [. . .] OR without reason led=me.ACC ‘without reason they led me’ (Coresi, 238v) b. fără dereptate mă mânară [. . .] MR without reason me.ACC= led ‘without reason they led me’ (4) a. nu mă rușinez când caut [. . .] not me.REFL= embarrass when search ‘I am not ashamed when I search’ (Coresi, 231v)

OR

b. nu mă rușinez când caut [. . .] not me.REFL= embarrass when search ‘I am not ashamed when I search’

MR

(5) a. Și acestu sfat îi da [. . .] OR and this advice him.DAT= give ‘And she/he gave him this piece of advice’ (Amiras, 253v) b. Și acest sfat îi dădea and this advice him.DAT= ‘And she/he gave him this piece of advice’

[. . .] MR give

271

Object Pronouns in the Evolution of Romanian

We assume that linguistic variation, including diachronic change, is dependent on the effects of feature valuation. The set of unvalued features varies cross-linguistically as well as in the historical development of languages. The fluctuation in word order between pre- and post- positions for a given category in a projection chain is a function of the availability of both a valued and an unvalued feature F, giving rise to movement if F is unvalued, and not otherwise. We interpret this choice as a point of symmetry and we explore the effect of factors that are external to the Language Faculty on the reduction of complexity brought about by this choice, as well as its gradual elimination over time. Specifically, we examine the fluctuation in the position of the object pronoun in Old Romanian illustrated above, and propose that it is an instance of the Directional Asymmetry Principle (DAP), a complexity-reducing principle proposed in Di Sciullo (2011) and linked to the symmetry-breaking laws active in the natural world (Lewontin 1970, 1974; Graham, Freeman & Emlen 1993; Palmer, 2004, a.o.). A biolinguistic explanation for a diachronic phenomenon has the advantage of further bridging the explanatory gap between language development and biology. 2

An Instance of the Directional Asymmetry Principle

Symmetry breaking is part of the natural laws affecting the evolution of the shape of biological organisms. Seen as a dynamic force external to the Language Faculty, it provides a biolinguistic explanation for language variation and evolution with respect to the position of a head and its dependent. A head-dependent structure includes a prominent element, the head (H). H can be to the right or to the left of its dependent (XP), as in (6), which depicts the position of H with respect to its sister for simplicity; however the position of H with respect to XP is set in an extended projection chain of H. It has been observed that both (6a) and (6b) are attested at some stage of the historical development of languages, while in a subsequent stage only one of the two options is available. In order to account for this phenomenon, the Directional Asymmetry Principle is formulated in Di Sciullo (2011) as in (7), where symmetry breaking applies to the availability of structures such as (6a) or (6b) in the course of language diachronic development. (6) a. b. H H H XP XP

H

272

Di Sciullo and Somesfalean

(7) Directional Asymmetry Principle (DAP) Language evolution is symmetry breaking: fluctuating asymmetry is followed by directional asymmetry. Fluctuating asymmetry (random left or right positioning of a head) > Directional asymmetry (exclusive left or right positioning of a head). Di Sciullo (2011) provides evidence that DAP makes correct predictions for language historical evolution on the basis of the development of possessive pronouns from Genitives in the evolution of Classical Greek to Modern Greek and Greek dialects, as well as in the evolution of Latin to Italian and Italian dialects. Namely, the pre- and post-nominal positions are possible for the Genitive theme in Classical Greek, while only the post-nominal position is possible in Modern Greek. This is also the case for the possessive clitics in Modern Greek and Grico, a Greek dialect spoken in the Italian regions of Calabria and Puglia, as well as in Modern Italian and in dialects spoken in the regions of Abruzzo, including Pescasseroli, Fallo and Arielli. The predictions of DAP have also been shown to cover the development of prepositions in Indo-European languages (in Di Sciullo & Nicolis 2013), and of the definite determiner in Romanian (in Di Sciullo & Somesfalean 2013). According to DAP, language evolution is symmetry breaking. Symmetry introduces choice-points, thus instability in a system that seeks to eliminate it in order to reinstate an asymmetrical stable state. The effects of symmetry breaking in language historical development are legible at the sensory-motor interface. As predicted, the fluctuating stage of pronoun position in Old Romanian is followed by a phase where a preponderant location is attested: in MR the predominant use of the object personal pronoun clitic is preverbal, i.e. proclitic.3 The Old Romanian use of enclitics, the development of Differential Object Marking (DOM), and the rise of Clitic Doubling (CD) constructions are all phenomena that have been addressed and discussed in recent works (Zafiu 2014; Hill 2013; Chiriacescu & Von Heusinger 2009; Alboiu & Hill 2012; Von Heusinger & Onea Gaspar 2008, a.o.). We consider some of these facts in a broader perspective, as instances of language evolution processes. The notion of language evolution goes beyond the classical notion of language change and grammaticalization (Roberts & Roussou 2003) by incorporating recent results from evolutionary developmental biology. This incorporation has both descriptive

3  While post-verbal strong pronouns exist in Modern Romanian, they are part of DOM/CD constructions, i.e. the presence of the clitic is obligatory.

Object Pronouns in the Evolution of Romanian

273

and explanatory advantages over classical notions of language change and grammaticalization. The descriptive advantage is that fluctuating stages are predicted to occur and can be described systematically. The explanatory advantage is that questions such as why languages change and why grammaticalization exists can be addressed on the basis of the existence of general laws governing the development and evolution of biological form. 3 Analysis 3.1 Enclisis in Old Romanian Enclisis is a generalised characteristic of Old Romanian, evidenced not only with pronominal objects, as seen in examples (1a), (2a), (3a), but also with verbal clitics, i.e. auxiliaries, as in (8a) below, and adverbial clitics. While we mention the other constructions marginally in our analysis, this paper is concerned with the behaviour of pronominal objects only. (8) a. ales-au 12 oameni de țară OR chosen=have.2PL 12 men of country ‘They have chosen 12 countrymen’ (Amiras, 249r) b. au ales 12 oameni de țară MR have.2PL chosen 12 men of country ‘They have chosen 12 countrymen’ The massive use of enclitics in Old Romanian may seem like instances of Wackernagel’s law, very strong in Slavic languages and assumed to have greatly influenced the written Romanian language (Frâncu 2009). However, Alboiu & Hill (2012) argue that Wackernagel’s law is not active in Old Romanian, given the fluctuation in the placement of clitics in Old Romanian, i.e. clitics are not consistently in second position (cf. ex (9), from Alboiu & Hill 2012); clitics may also be preverbal (cf. ex. (10) from Alboiu & Hill 2012), and finally, the rise of proclitics is independent of Wackernagel’s law. They conclude that the enclisis that characterises Old Romanian is discourse-driven. (9) cu pizmă huluiia-l with hate cursed=him.ACC ‘cursed him with hate’ (Frâncu 2009: 277)

274

Di Sciullo and Somesfalean

(10) să vedea că după acest război fără noroc [. . .] SE arb= saw that after this war without luck ‘one could see that after this war without luck [. . .]’ (Ureche 115) Our focus in this paper is a particular case of the enclisis that characterises Old Romanian, namely of pronominal objects. Moreover, Old Romanian allows for both preverbal and postverbal pronominal objects, a fluctuation thus coexists at a given moment in the evolution of Romanian, illustrated in (11) below. Interestingly, while the pronominal object enclisis is wide-spread in Old Romanian—for both clitics and strong pronouns, the tendency in Modern Romanian is towards the use of clitic objects, exclusively preverbal. Object strong pronouns, when used, can only be doubling constituents (i.e. the presence of the preverbal clitic is necessary for CD). This is illustrated in (12). (11) a. pădzească tine cel mare domnul și OR protect.SUBJ.3 you.ACC the great lord=DEF and te alduiască [. . .] you.ACC= bless. SUBJ.3 ‘May the great Lord protect you and bless you.’ (Frag.Tod. 4r) b. să te păzească pe tine domnul MR SUBJ you.ACC= protect.SUBJ.3 DOM you.ACC lord=DEF cel mare și să te binecuvânteze [. . .] the great and SUBJ you.ACC= bless.SUBJ.3 ‘May the great Lord protect you and bless you.’ (12)

Old Romanian

Modern Romanian

primiiși

mine

mă

primiși

received.2SG

me.ACC

me.ACC=

received.2SG

cânt

ție

îți

cânt

sing.1SG

you.DAT

you.DAT=

sing.1SG

mânară-mă

mă

mânară

led=me.ACC

me.ACC=

led.3PL

English

(pe mine) ‘you received me’ DOM me.ACC (ție) ‘I sing to you’ you.DAT (pe mine) ‘they led me’ DOM me.ACC

Object Pronouns in the Evolution of Romanian

275

We provide an account for the observed fluctuation of the position of pronominal objects in Old Romanian and of the Modern Romanian tendency towards proclisis. This reinforces previous findings about the same trend in the evolution of the Romanian DP, as shown in Di Sciullo & Somesfalean (2013). Specifically, we have shown that Modern Romanian is in a phase of directional asymmetry with respect to the behaviour of the Gen/Dat forms of the definite determiner. Thus, while Nom/Acc forms of the definite determiners have reached a phase of directional asymmetry (i.e. enclisis), the Gen/Dat forms are still allowing fluctuation, with a strong tendency towards proclisis. These two patterns of evolution are summarised in (13), from Di Sciullo & Somesfalean (2013). (13) Evolution of the Nominative/Accusative form of the definite determiner: Phase 1 – fluctuating asymmetry: Danubian Latin prenominal and Post-nominal demonstratives (from Graur 1929) Prenominal: (i) homo ille man.NOM that.NOM ‘That good man.’

bonus good.NOM

Post-nominal: (ii) ille homo bonus that.NOM man.NOM good.NOM ‘That good man.’ Phase 2 – directional asymmetry: Stabilization of Post-nominal definite article in Old and Modern Romanian (i) a. omul bun (MR) man.DEF.NOM/ACC good ‘the good man’ b. cartea din librărie book.DEF.NOM/ACC from bookstore ‘the book from the bookstore’

276

Di Sciullo and Somesfalean

(ii) a. acesta iaste ce aud cuvântul [. . .] this is what hear.1SG word.DEF.ACC ‘this is the word I hear’ (Tetra. 43v) b. să nu poftești vecinului tău [. . .], SUBJ not covet .2SG neighbor.DEF.DAT your nece feciorul, nece boul, nece asinul, [. . .] neither boy.DEF.ACC nor ox.DEF.ACC nor ass.DEF.ACC ‘To your neighbour, you shall not covet the son, the ox or the ass.’ (Î.C.5r) Prenominal Nom/Acc definite articles are not attested in Old or Modern Romanian. (iii) *ul DEF.NOM/ACC ‘the good man’

om man

bun good

(14) Evolution of the Oblique Genitive/Dative form of the definite determiner: Phase 1 – fluctuating asymmetry: Prenominal and Post-nominal definite determiners in Old Romanian Prenominal: (i) lui Hotchevici DEF.DAT/GEN.M.SG Hotchevici (ii) ei Maria DEF.DAT/GEN.F.SG Maria Post-nominal: (iii) Radului Radu.DEF.DAT/GEN.M.SG. (iv) Mariei Maria.DEF.DAT/GEN.F.SG. (v) băiatului boy.DEF.DAT/GEN.M.SG. (vi) copilei girl.DEF.DAT/GEN.F.SG.

Object Pronouns in the Evolution of Romanian

277

Phase 2 – (strong tendency of) directional asymmetry: Prenominal definite determiner /prepositional marker in Modern Romanian,4, 5 (i) lui / lu’ Ioan DEF.DAT/GEN.M.SG. Ioan (ii) lui / lu’ DEF.DAT/GEN.M.SG.

copil /copilul6 boy/the boy

(iii) lui / lu’ DEF.DAT/GEN.M.SG.

fată / fata girl/the girl

Romanian constructions with definite determiners allowed us to confirm the prediction of DAP, i.e. that a directionality is observed in the evolution of the language from an older stage to a modern stage: while the fluctuating asymmetry is brought about by a choice in the valued or unvalued properties of Case features in Old Romanian, the symmetry brought about by these choice points is gradually eliminated in the modern stage of the language, where eventually only one of the two options remains available. Based on these conclusions, we expect DAP to be active also in the evolution of Romanian pronominal object constructions, with the effect of reducing complexity. In what follows we will appeal to two notions of complexity proposed in Di Sciullo (2012): (i)

Internal complexity (I-complexity) is derived by the operations of the Language Faculty and is measured in terms of length of derivations. Thus, a derivation of a linguistic expression that involves

4  lu’ is a phonetically shortened form of lui. 5  la is also a possible substitution for the Dative Case marked DP that can be employed with both masculine and feminine forms. However, while for most speakers lu(i) is best followed by a definite noun, la is mostly followed by an indefinite. 6  When the prepositional marker is used, the definite forms seem to be more easily acceptable than the indefinite forms. Still, in non-standard registers, for some speakers of Romanian the following constructions are acceptable: (i) I-am spus lu’ copil să nu întârzie. him.DAT= have told DEF.DAT child SUBJ not be late ‘I have told the child not to be late.’ (ii) I-am spus și la fată. him.DAT= have told also to girl ‘I have also told the girl.’

278

Di Sciullo and Somesfalean

fewer operations will be preferred over a more ‘costly’ derivation on grounds of computational efficiency. (ii) External complexity (E-complexity) is legible at the sensori-motor (SM) interface and is calculated in terms of density of representations, which is not limited to string linear measure, but includes supra-segmental material such as tone, as discussed in Di Sciullo (2005), and stress. Thus, a representation that contains less SM material will be less ‘costly’ on grounds of representational efficiency. We propose that the change in the pattern of pronominal objects from Old to Modern Romanian is the result of a bi-fold complexity reduction mechanism, namely the reduction of both I-complexity, which is basically derivational, and the reduction of E-complexity, which is basically representational. We now turn to the structure of the constructions under investigation and discuss their complexity. We will first discuss I-complexity (henceforth, derivational complexity) and then E-complexity (henceforth, representational complexity). 3.2 Derivational Complexity We assume that pronouns are determiners, as in Postal (1969), and that clitics and strong pronouns differ in their level of complexity, as in Kayne (1991, 1994), Uriagereka (1995), Cardinaletti & Starke (1999), Sportiche (1999), Di Sciullo (1990), Di Sciullo & Aguero (2008), a.o. More specifically, we assume, with Uriagereka (1995), that pronominal clitics, as anchors on new information, are in the head of a functional projection at the periphery of IP, illustrated below in (15). As mentioned above, we assume that movement is driven by feature checking/valuation (cf. Chomsky 1995, 2000, 2001; Pesetsky & Torrego 2006; a.o.). We will not discuss the feature checking/valuation in the derivations, as the identification of the features checked/valued is orthogonal to our discussion. (15)

FP cl

IP I

. . . . . . vP V

DP strong pron.

Object Pronouns in the Evolution of Romanian

279

Thus, in a construction such as (16) below, the clitic is in (Head, FP) and the proclitic positioning of the object pronoun follows, as illustrated in (17). (16) Costantin Ciobanul [. . .], de altă parte, Constantin Shepherd.DEF on other side, îl sfătuia să vie [. . .] him.ACC= advised.3 SUBJ come.SUBJ.3 ‘Constantin The Shepherd, on the other hand, advised him to come back’ (Amiras, 253v) (17)

FP F îl

IP I sfătuia

. . . . . . vP CP

V sfătuia

. . . . . .

We illustrate the strong pronoun object example (18) with the structure in (19) below. (18) cu slavă priimiși mine with glory received.2SG me.ACC ‘With glory you received me.’ (Coresi, 137r) (19)

IP I priimiși

vP V priimiși

DP mine

280

Di Sciullo and Somesfalean

However, as mentioned above, Old Romanian also presents evidence of postposed clitic constructions in declaratives, as in (20). (20) a. Domnul fereaște-te de tot răul. OR Lord.DEF protects=you.ACC from all harm ‘The Lord protects you from all harm.’ (Coresi, 248v) b. Domnul te ferește de tot răul. MR Lord.DEF you.ACC= protects from all harm ‘The Lord protects you from all harm.’ We analyse these as instances of verb movement, in the sense of Emonds (1978), Polock’s (1989), a.o., a movement that is still active in Modern Romanian, but only in imperative and gerund constructions, such as (21), illustrated in (22) with a structure based on Isac (1998).7 (21) Ferește-te protect.IMP.2SG=you.ACC.REFL ‘Protect yourself from the sun!’ (22)

de from

soare! MR sun

MoodOp M ferește

FP te

MP ferește

vP V ferește

PP . . . . . .

7  In Isac (1998) the (Modern) Romanian verb moves to Mood projection, while the subjunctive particles, imperatives and gerunds move to a higher MoodOp projection in order to check strong irrealis features.

281

Object Pronouns in the Evolution of Romanian

The question that immediately arises is what drives generalized V-to-C in Old Romanian. We assume that V-to-C in Old Romanian is discourse-related, as in Alboiu, Hill & Sitaridou (2014). Adopting a cartographic approach (Rizzi 1997), the authors analyse V>clitic order (i.e. all instances of enclisis in Old Romanian: verb>auxiliary clitics, V>adverbial clitics, V>pronominal clitics,) as the result of verb movement to a focus head, triggered for discourse purposes, an optional movement that depends on whether certain discourse features (i.e. a focus operator feature optionally associated with the CP field) are present in the derivation. In their analysis, four types of focus operators (following Hohle 1992; Krifka 2007; Richter & Mehlhorn 2006) are present in Old Romanian and operator feature checking is satisfied either by constituent movement to (Spec, FocP)—for instance, a wh-phrase—or by head-to-head movement of the verb to Foc—in which case the operator is null. Thus, V>clitic order obtains when the verb moves higher than the tense-bearing projection T, to Foc, for checking emphatic focus (EF) or verum focus (VF) features. This analysis predicts that the V>clitic order would be optional in declaratives, as in (23), obligatory in yes/no questions, as in (24), and excluded in wh-questions, as in (25). (23) a. afla-să această tară happened= SE ARB this country ‘This country happened to . . .’ (Ureche 67)



[. . .]

b. să vedea că după acest război [. . .] SE ARB= saw that after this war ‘You could see that after this war . . .’ (Ureche 115)

(24) Cunoști-mă pre mine, au ba? know=me.ACC DOM me.ACC or not ‘Do you recognize me or not?’ (Costin 33) (25) Cum ar hi împăratu să hie drag tuturora? how AUX be king.DEF SUBJ be dear all.DAT ‘How should the king be to be loved by all?’ (Costin 33) Thus the partial structure of an Old Romanian declarative such as (20a), for instance, would look like (26) below, where the verb moves to Foc.8 8  The constituents preceding FocP are in a topicalised projection TopP. The proposed cartography of the Old Romanian clause in Alboiu & Hill (2012); Alboiu, Hill & Sitaridou (2014) is

282

Di Sciullo and Somesfalean

(26) . . . . . .  FocP Foc fereaște

FP te

IP fereaște

vP V fereaște

PP . . . . . . 

The fact that Modern Romanian still allows stylistic, highly focused expressions where indicatives or conditionals precedes the clitic, as in (27)–(28) below, is an argument in favour of such an analysis, where the preverbal position of the verb is associated with an emphatic reading. (27) Pare-se că vrea să seems= SE ARB that wants SUBJ ‘It seems that she/he wants to leave.’

plece. leave

(28) Mira-m-aș că vine și wonder=REFL=AUX that comes also ‘I would be surprised if she came along.’

MR

ea! her.

MR

We adopt this movement of the verb to a Focus projection in order to derive the V>clitic constructions. These constructions are attested started from mid17th century. In the constructions involving strong pronominal arguments, on the other hand, the verb may also move, but not as high as FocP. Zafiu (2010) and Alboiu & Hill (2012) show that 16th century Old Romanian does have (nonemphatic) verb movement in translated texts, but it targets a lower projection than FocP, namely Fin (the equivalent of I in our structures). TopP>FocP>FinP/IP>TP. We differ from their clause structure assumptions only with respect to clitic placement: while they place clitics adjoined to T, we have them in F, as in Uriagereka (1999), above IP. We also assume Long Head Movement (Rivero 1993, a.o.), ensuring that the verb is able to move to higher projections such as FocP.

283

Object Pronouns in the Evolution of Romanian

Based on our assumptions on derivational complexity mentioned above and the structures in (17), (19) and (26), the derivation of cl>V or V>strong pronoun constructions involves fewer operations than the derivation of V>cl constructions, where the verb has moved to FocP. It follows that the derivation of (17), as well as that of (1), (2), (3b) is less costly from a computational point of view than the derivation of (26), as well as that of (3a), hence it is preferable for efficiency reasons. Thus, in a fluctuation period such as the one observed in Old Romanian, our analysis predicts that given DAP and the fact that the derivation of post-verbal clitic objects is more derivationally costly, proclisis will be preferred. Our prediction is confirmed by Modern Romanian data. 3.3 Representational Complexity We have seen that verb movement yields variation in the position of pronominal objects in Old Romanian. But we have not said anything about the variation in the choice between a strong and a clitic form of the object pronoun in the evolution of Romanian, cf. the contrast between (1a), (2a) and (1b), (2b). What drives the choice of a clitic over a strong pronoun in the evolution of Romanian? Why are patterns (29a) and (29d)—i.e. with strong object pronouns— attested in Old Romanian, rather than simply (29b) and (29e)—i.e. with clitic objects, which incidentally is also what survived in highly stylistic Modern Romanian (cf. (29c) and (29f))? (29) Summary of diachronic changes Old Romanian (strong pronoun use)

a) primiiși

received. 2SG d) cânt

mine

me.ACC

ție

sing.1SG you.DAT

Modern Romanian (clitic use)

b) mă

primiși

me.ACC= received.2SG c) stylistic: e) îți

primiși-mă cânt

you.DAT= f) stylistic:

sing.1SG cântu-ți

English

(pe mine)

DOM me.ACC (ție) you.DAT

‘you received me’

‘I sing to you’

284

Di Sciullo and Somesfalean

We believe the answer to this question lies in the degree of complexity at the SM interface that differentiates strong pronouns and clitics. In other words, the choice of a strong pronoun (in Old Romanian) implies the choice of a more complex SM form, an option which is gradually eliminated in favour of a less complex form, in our case, a clitic (in Modern Romanian). This implies that DAP is equally sensitive to another measure of complexity, i.e. E-complexity. Numerous studies have investigated the difference in structure between strong pronouns and clitics. Cardinaletti & Starke (1999), Déchaine & Wiltscko (2002), Di Sciullo (2005), a.o., have proposed systems where the visible lack of structure in the clitic form is the reflection of a deficient internal structure. The fact that deficient pronouns have an impoverished structure is shown by their morpho-phonological form (l, i, etc.), while strong pronouns are more articulated, often including determiner forms (lui, ei, etc.). Since features have a morphological reflex in the theory we are adopting, we should expect that clitics lack encoding some of the features that are present in the strong forms. In other words, the features of the clitic should constitute a subset of the features of the strong pronoun. The exact feature argued to be deficient in clitics as opposed to strong pronouns may differ according to the approach. It has been proposed that Romance clitics lack encoding a [person] feature, cf. Uriagereka (1995), a.o. In Cardinaletti & Starke (1999), the difference between strong pronouns and deficient ones is the presence/absence of the Case projection in their structure. In Déchaine & Wiltschko (2002) strong pronouns and clitics have a different categorical status. We adopt Di Sciullo (2005), where the internal structure of functional elements is an operator shell. The upper layer of the shell is the locus of the operator feature, such as a wh- or a th- operator, and the lower layer is the locus of the restrictor of the variable bound by the operator. The difference between strong and weak pronouns is the occurrence of a Focus feature in the upper part of the Op-Shell in the case of strong pronouns, and its absence in the case of weak pronouns. Thus it is the presence of a Focus feature in the structure of the strong pronouns that makes them differ from the clitic counterpart, as the Op-Shells in (30) illustrate. (30)

a. [Opd Foc b. [Opd

F

[α

Re

β]]

strong pronoun

F

[α

Re

phi-features β]] phi-features

clitic

Object Pronouns in the Evolution of Romanian

285

Considering the upper layer of the OP-shells in (30), the pronominal operator (Opd) is associated with a focus feature (Foc) in the case of a strong pronoun (30a), but not in the case of a clitic, (30b). The F head is the locus of the variable bound by the operator. The variable is linked to its restrictor (Re) in the lower layer of the Op-Shell. For example, with pronouns, Re can be +Human, e.g. he/him or –Human, e.g. it. The dependent of the restrictor is associated with phi-features, including person, number, gender and Case. In this framework, “the Op-Shell covers the morphological properties of [±Q] elements, including question words and complementizers, and [±D] elements, including definite, indefinite, and expletive determiners, demonstratives and pronouns, which have the same asymmetric form.” (Di Sciullo 2005: 121). The morphological form of the functional elements differs however with respect to their feature structures. The difference in SM representational complexity may not always arise from string-linear properties. In our case, while a string-linear difference is visible in certain forms of the strong pronoun/clitic pair, such as mine PRON. ACC.1SG. vs. mă ACC.1SG=, this difference may not be obvious in other forms, such as ție PRON.DAT.2SG. vs. îți DAT.2SG=. On the other hand, the notion of density of SM representations encompasses string-linear properties and suprasegmental features, in our case stress. Thus, in both mine / mă and ție / îți pairs, for instance, only the strong pronoun can bear stress. This difference in SM representational complexity leads to the preference of the clitic (the less complex form) over the strong pronoun (the more complex form). Our analysis predicts that, whenever a choice is possible, a clitic will be preferred over a strong pronoun, which concurs with what is found in L1 acquisition studies (Granfeldt & Schlyter 2004). It also concurs with principles such as Avoid Pronoun (Chomsky 1981) or Minimise Structure (Cardinaletti & Starke 1999, essentially stating that whenever a weak pronoun is available it must be chosen over a stronger pronoun). In the constructions under examination here, after a period of fluctuation in Old Romanian where both strong and clitic object pronouns coexist in a given configuration, only the clitic form survives in later stages of the language. As mentioned above, this is not to say Modern Romanian doesn’t have strong pronouns. As shown in (29b) and (29e), Modern Romanian allows strong pronouns, but only in CD constructions, i.e. the presence of the clitic is required.9 In other words, the default pronominal argument seems to be the clitic, and when the strong pronoun is present, DOM is too. Irimia (this volume) shows that strong pronouns have indeed 9  One could wonder why these doubling cases are not subject to DAP. It is generally assumed that CD structures are highly emphatic and thus motivated on grammar external grounds.

286

Di Sciullo and Somesfalean

different properties in Old Romanian as opposed to Modern Romanian. Namely, while DOM and CD are not obligatory in Old Romanian, they are in Modern Romanian. She links this difference to the levels of prominence active in the language at a given point in time. While in Modern Romanian the DOM constructions are subject to both the animacy and the definiteness scale, presumably in Old Romanian the definiteness scale is not fully implemented, yielding a variable behaviour of strong pronouns. Thus, it may be the case that, as the later stages of Romanian impose more constraints on the presence of the strong pronouns (i.e. DOM becoming increasingly obligatory), the choice of pronominal arguments gradually reduces to clitic forms, a ‘simpler’ choice on both E-complexity and I-complexity grounds. 3.4 Summary Our study of the change in form of pronominal objects constructions from Old to Modern Romanan reveals that they are the result of two phenomena: on one hand, there is an a optional, discourse-related movement of the verb to a position higher than the tense-bearing node; on the other hand, there is a choice between strong and clitic forms of the object pronoun. We propose that the form of the pronominal object constructions in the evolution of Romanian is the result of the reduction of two types of complexity: i)

I-complexity or derivational CI complexity (as a measure of the number of syntactic operations that apply in the derivation of a linguistic expression)

Under its effect, Modern Romanian gradually lost the verb movement motivated by discursive features.10 The only verb movement to a projection higher than the tense-bearing node that is still attested in Modern Romanian is not discourse-related (i.e. not optional), but motivated by syntactic-feature checking/valuation in imperative and gerund constructions. ii) E-complexity or representational SM complexity (as a measure of the SM density of a representation) Under its effect, MR gradually lost the choice of strong pronouns in favour of clitics in argument positions (i.e. examples such as (29a), (29d) are disallowed, in favour of (29b) and (29e)), independently of verb movement. 10  Reminiscent of these are cases of stylistic emphasis in MR such as in (27) and (28) above, crucially involving clitics.

Object Pronouns in the Evolution of Romanian

287

The combined effect of the two complexity-reduction mechanisms is the complete loss of postverbal strong pronouns (without DOM) in Modern Romanian. 4 Conclusion We have proposed that the change in the pattern of pronominal objects from Old to Modern Romanian is the result of DAP, a complexity-reduction mechanism sensitive to both derivational and representational complexity. We further confirmed its effects in the evolution of Romanian. We have shown that in addition to derivational complexity reduction, representational complexity reduction is also a factor of language change. Corpus Amiras

Simonescu, D. (ed.). 1975. Cronicile medievale ale României. Vol. 9. Bucharest: Editura Academiei. Coresi Toma, S. (ed.). 1976. Coresi, Psaltirea slavo-română (1577) în comparație cu psaltirile coresiene din 1570 și din 1589. Bucharest: Editura Academiei. Costin Panaitescu, P.P. (ed.). 1979. Miron Costin, Letopisețul Țării Moldovei. Bucharest: Editura Minerva. Frag.Tod. Fragmentul Todorescu. In Buză, E., G. Chivu, M. Georgescu, I. Ghetie, A. Roman-Moraru & F. Zgraon (eds.). 1982. Texte românești din secolul al XVI-lea. Bucharest: Editura Academiei. Î.C. Întrebare creștinească. In Buză, E., M. Georgescu, A. Mareș & F. Zgraon (eds.). 1994. Crestomatia limbii române vechi. vol I (1521–1639). Institutul de lingvistica Bucuresti. Neculce Iordan, I. (ed.). 1955. Ion Neculce, Letopisețul Țării Moldovei. Bucharest: Editura de Stat. Tetra. Tetraevanghel. In Buză, E., M. Georgescu, A. Mareș & F. Zgraon (eds.). 1994. Crestomatia limbii române vechi. Vol. I (1521–1639). Institutul de lingvistica Bucuresti. Ureche Panaitescu, P.P. (ed.). 1958. Grigore Ureche, Letopisețul Țării Moldovei. Bucharest: Editura de Stat.

288

Di Sciullo and Somesfalean

References Alboiu, Gabriela & Virginia Hill. 2012. Early Modern Romanian and Wackernagel’s law. Journal of the Linguistic Association of Finland 25: 7–28. Alboiu, Gabriela; Virginia Hill & Ioanna Sitaridou. 2014. Discourse driven V-to-Focus in Early Modern Romanian. Natural Language and Linguistic Theory. DOI 10.1007/ s11049-014-9270-8. Cardinaletti, Anna & Michael Starke. 1999. The typology of structural deficiency: A case study of the three classes of pronouns. In H. van Riemsdijk (ed.), Clitics in the Languages of Europe. 145–235. Berlin/New York: Mouton de Gruyter. Chiriacescu, Sofiana & Klaus Von Heusinger. 2009. Pe-marking and Referential Persistence in Romanian. In A. Riester & E. Onea (eds.), Focus at the SyntaxSemantics Interface. Working Papers of the SFB 732, Vol. 3. University of Stuttgart. Chomsky, Noam. 2001. Derivation by phase. In M. Kenstowicz (ed.), Ken Hale: A life in language. 1–52. Cambridge, MA: MIT Press. ———. 2000. Minimalist inquiries. In R. Martin, D. Michaels, & J. Uriagereka (eds.), Step by Step. 89–155. Cambridge, MA: MIT Press. ———. 1995. The Minimalist Program. Cambridge, MA: The MIT Press. ———. 1981. Lectures in Government and Binding. (Studies in generative grammar 9.) Dordrecht: Foris. Déchaine, Rose-Marie & Martina Wiltschko. 2002. Decomposing Pronouns. Linguistic Inquiry 33: 409–422. Di Sciullo, Anna Maria. 2012. Perspectives on Morphological Complexity. In F. Kiefer, M. Ladanyi & P. Siptar (eds.), Morphology. (Ir)regularity, Frequency, Typology. 105– 135. Amsterdam: John Benjamins. ———. 2011. A Biolinguistic Approach to Variation. In A.M. Di Sciullo & C. Boeckx (eds.), The Biolinguisitic Entreprise. New Perspectives on the Evolution and Nature of the Human Language Faculty. Oxford: Oxford University Press. ———. 2005. Asymmetry in Morphology. Cambridge, MA: The MIT Press. ———. 1990. On the Properties of Clitics. In A.M. Di Sciullo & A. Rochette (eds.), Binding in Romance. Canadian Journal of Linguistics. 209–223. Di Sciullo, Anna Maria & Marco Nicolis. 2013. Third factor in the development of P. NELS 42. University of Toronto. Di Sciullo, Anna Maria & Stanca Somesfalean. 2013. Variation in the Position of the Definite Determiner in Romanian: A Biolinguistic Perspective. Australian Journal of Linguistics 33(2): 121–139. Di Sciullo, Anna Maria & Calixto Aguero Bautista. 2008. The delay of Condition B Effect and its Absence in Certain Languages. Language and Speech 51: 77–100. Emonds, Joseph. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9: 49–77.

Object Pronouns in the Evolution of Romanian

289

Frâncu, Constantin. 2009. Gramatica limbii române vechi (1521–1780). Iasi: Demiurg. Granfeldt, Jonas & Suzanne Schlyter. 2004. Cliticisation in the acquisition of French as L1 and L2. In J. Paradis & P. Prévost (eds.), The Acquisition of French in Different Contexts: Focus on Functional Categories. 333–370. Amsterdam/Philadelphia: John Bejamins. Hill, Virginia. 2013. The direct object marker in Romanian: a historical perspective. Australian Journal of Linguistics 33(2): 140–151. Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, MA: The MIT Press. ———. 1991. Romance Clitics, Verb Movement, and PRO. Linguistic Inquiry 22: 647–686. Krifka, Manfred. 2007. Basic Notions of Information Structure. In C. Fery, G. Fanselow & M. Krifka (eds.), Interdisciplinary Studies of Information Structure 6. Potsdam. Pesetsky, David & Ester Torrego. 2006. Probes, Goals and Syntactic Categories. In Y. Otsu (ed.), Proceedings of the Seventh Tokyo Conference on Psycholinguistics. 25–60. Pollock, Jean-Yves. 1989. Verb Movement, Universal Grammar, and the Structure of IP. Linguistic Inquiry 20: 365–424. Postal, Paul M. 1969. On so-called ‘Pronouns’ in English. In D.A. Reibel & S.A. Schane (eds.), Modern Studies in English: Readings in Transformational Grammar. 201–224. Englewood Cliffs, N.J.: Prentice Hall. Richter, Nicole & Grit Mehlhorn. 2006. Focus on contrast and emphasis: Evidence from prosody. In V. Molnar & S. Winkler (eds.), The architecture of focus. 347–373. Berlin: Mouton de Gruyter. Rivero, Maria-Luisa. 1991. Long Head Movement and Negation: Serbo-Croatian vs Slovak and Czech. The Linguistic Review 8: 319–351. Roberts, Ian & Anna Roussou. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Sportiche, Dominique. 1999. Pronominal clitic dependencies. In H. van Riemsdijk (ed.), Clitics in the Languages of Europe. 679–708. Berlin/New York: Mouton de Gruyter. Uriagereka, Juan. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79–123. Von Heusinger, Klaus & Edgar Onea Gaspar. 2008. Triggering and blocking effects in the diachronic development of DOM in Romanian. Probus 20: 67–110. Zafiu, Rodica. 2014. Auxiliary encliticization in the 16th century Romanian: restrictions and regularities. Linguistica Atlantica 33(2): 71–86.

DPs in Adjectival Small Clauses in Romanian: A Diachronic Perspective Monica Alexandrina Irimia This paper focuses on Romanian diachronic data with adjectival predicates under intensional verbs (consider-Adj. types). Due to their salient diachronic stability w.r.t important structural properties, these constructions are relevant to the research on small clauses and to the investigations on Romanian DPs and DOM strategies. It is proposed that a complex predicate analysis employing a Multiple Agreement Mechanism is able to derive the strong/specific readings of the shared arguments under discussion; the variation in the DOM marking of pronouns is correlated to a plausibly more recent development of the definiteness scale for differential marking in Romanian, complementary to the animacy scale.

1 Introduction Constructions involving adjectival predicates embedded under intensional predicates like consider, see, want, and so on, although less studied as compared to other non-verbal counterparts, pose numerous conceptual challenges. One the one hand, their syntactic structure is still an open issue, theoretical accounts alternating between small clause structures (Stowell 1981, 1991, a.o.) or complex predicate configurations (Chomsky 1975, a.o.). On the other hand, these constructions are highly relevant for understanding the nature of DPs functioning as shared arguments, as well as for nominal syntax in general. These two aspects are interconnected as the special morphosyntactic and interpretive properties of these DPs in turn provide crucial hints into the structure of such non-finite embedded projections. This paper discusses adjectival predicate data from 16th century Romanian, as compared to modern Romanian, illustrating both interpretative stability as well as variation in the morphological marking of the shared argument. The diachronic picture confirms the configurational distinctions, as well as the interpretive restrictions of the shared DPs, allowing us to better refine this explanatory domain.

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_010

DPs in Adjectival Small Clauses in Romanian

291

1.1 Specific Shared DPs As initially noticed in Williams’ (1983) seminal paper, and subsequently confirmed by various cross-linguistic data (see Irimia 2011 for an overview), shared arguments with adjectival embedded predicates can normally only be interpreted as specific. Hence the morphological indefinite in (1) is felicitously interpreted as a specific indefinite (de Houp 1996; Heim 1982). ENGLISH (1) I considered a student sick. = I considered a specific student sick. (student is salient in the context, and known to the speaker, if not the wider audience) ≠ I considered some student or other sick. Note that this restriction holds not only with individual-level adjectives (as argued for in Basilico 2003), but also with stage-level ones; e.g., in (1) sickness is not understood as an immutable characteristic of the student. As expected, this specificity requirement also applies in Modern Romanian. Moreover, as in many other languages, object (human) morphological indefinites in these contexts are generally overtly marked as differential objects introduced by the marker pe, and clitic-doubled (register-dependent), as shown in (2). (2) (L)- am considerat *(pe) un student bolnav. him= have.1.SG= considered DOM a.M.SG student sick.M.SG ‘I considered a student sick.’ The diachronic picture is nevertheless more complex in Romanian. Although in Old Romanian such arguments normally carry strong/specificity readings, as predictable, their morphological marking is not as strict as in Modern Romanian, in that the DOM material can be omitted in some instances in which it would be obligatory in Modern Romanian. The most important question posed by this paper is how semantic stability (i.e., immutable restriction to specificity) can be reconciled with a non-systematic morphological marking of some classes of shared DPs. The answer has non trivial consequences regarding the syntactic configurations adjectival predicates project, as explained below in 1.2. 1.2 Specific DPs and the Small Clause/Complex Predicate Debate As is well known from the extensive literature on scope, canonical specific (indefinite) DPs are generally analyzed as taking wide-scope with respect to

292

Irimia

the (intensional) predicate (Milsark 1974; May 1985; Heycock 1994; Diesing 1992; de Houp 1996; Enç 1991; various papers in Reuland & ter Meulen 1987, a.o.). Hence specificity in (1) and (2) can only be explained if the shared DP is interpreted above the matrix predicate. Raising for Case (to a position above the matrix predicate) can straightforwardly derive wide-scope; however, it leaves unexplained why narrow-scope reading under reconstruction (Chomsky 1995; Lasnik 1999; May 1985; Boeckx 2001, a.o.) is not possible. A comparison with other non-finite embedded predicates (e.g., infinitives) unambiguously indicates that in these latter contexts both wide-scope and narrow-scope readings are possible. Consider the contrast between (3) and (4). (3) A student seems to be sick. Wide-scope: a student >> seems Narrow-scope: seems >> a student (4) A student seems sick. Wide-scope: a student >> seems No narrow-scope reading Williams (1983) attributed the lack of narrow-scope readings of shared DPs in (1, 2, 4) to their complex predicate structure, strongly denying a small clause analysis. More precisely, the shared argument is base-generated above the complex formed by the two predicates, and hence cannot reconstruct to a position lower than the matrix predicate. Therefore, specificity is the only possible interpretation of these types of strong/wide-scope DPs. When looking at non-pronominal forms, the data collected from Old Romanian behave as expected: only specific indefinites (introduced by pre) or definites have been found in these contexts. Complications are found with pronouns: as seen in (5) as opposed to (6), the DOM marker can sometimes be absent with pronouns. (5) Dumnezeu ispiti pre ei și află ei destoinici luiș. God tested DOM they and found them loyal.M.PL he.DAT ‘That God put them to test and found them loyal to him.’ (Coresi EV 260) (6) Ispitind pre el diavolul  . . ., află pre el nebiruit Test.GER DOM he devil.the found DOM he invincible.SG.M ‘When the Devil put him to test, he found him invincible.’ (Coresi EV 520)

DPs in Adjectival Small Clauses in Romanian

293

Notably, in Modern Romanian DOM is obligatory with (shared) object pronominal arguments. Hence, two other questions follow: (i) Why do pronouns exhibit this fluctuating behavior in Old but not Modern Romanian?1 (ii) Are there changes with respect to what DOM (i.e., the preferred strategy for marking wide-scope and specificity) actually encompasses morphologically (Silverstein 1976; Aissen 2003)? Again, what is interesting in these examples is that the DOM marking can be either present (6) or absent (5) with shared object pronouns. However, in the Old Romanian structures examined here non pronominal DPs can only carry DOM or be definite (typical instantiations of strong/specific DPs). The data under analysis mainly come from some of the earliest attested Romanian texts, especially Coresi’s Evanghelia cu învățătură (CC 1581) which exhibits a variety of non-finite adjectives embedded under intensional predicates. This permits a non-trivial testing of both stability as well as diachronic differences in the structure of Romanian DPs and small clauses. Given what is known about the systematicity of adjectival embedded predicates, the working hypothesis is that the stage of Romanian described here illustrates the introduction and early development of a definiteness scale, alongside the more prominent animacy scale in the differential marking of objects (see also Mardale, this volume). It is also argued that other systematic properties of the construction, such as the strong readings of shared nouns, support a complex predicate analysis. However, as adjectival predicates display phi-feature agreement with the shared argument, the mechanics of a derivationally simultaneous multiple checking of features by a unitary projection (Hiraiwa 2005) is implemented. On the basis of this working hypothesis and after a detailed diachronic examination of Romanian adjectival predicates under intensional predicates, the main proposal is that an enriched complex predicate structure is better equipped to account for the data. The discussion is contained in four main parts, starting with section 2 which further introduces the relevant examples and their diachronic picture. Section 3 emphasizes the commonalities of the Romanian examples against a cross-linguistic background, and their stability with respect to the presence of strong readings of shared arguments. In 1  Di Sciullo & Somesfalean (this volume) ask the same question from a biolinguistic perspective. While their answer grasps changes in the computational pattern (i.e., the breaking of symmetry), my purpose is to see how the stability solution for this construction can be compared to similar changes in other contexts where specificity is mapped through DP morphology.

294

Irimia

section 4 further remarks about the structure of non-finite embedded adjectives and their interaction with differential object marking are made. Section 5 contains an evaluation of previous theoretical approaches to embedded adjectives which paves the path for the analysis in section 6. Embedded adjectives under unitary multiple agreement are demonstrated to construct a complex predicate structure (Chomsky 1975): a functional projection (vCMPL.) is responsible for checking the relevant features of multiple predicates simultaneously derivationally; section 7 addresses the problem of the unstable morphological marking of pronouns and proposes that this is a result of a shift in the scope of differential marking. The last part (section 8) contains the conclusions. 2

Romanian Small Clauses

As mentioned in the introduction, Romanian texts from the 16th century contain a variety of adjectival small clauses under intensional predicates (consider, want, think, etc.); such configurations are also common in Modern Romanian. As the structure of such constructions is still problematic for many modern syntactic analyses, any empirical data that can contribute further relevant details deserve careful investigation. This paper will focus on the contribution embedded small clauses bring to the analysis of shared argument DPs, more clearly their interaction with specificity and DOM. Moreover, although constructions of this type have received a great amount of attention crosslinguistically from a synchronic perspective, not much has been said about their diachronic behavior, which however appears to provide crucial clues into their nature. In order to illustrate the facts more precisely, it is very useful to present the data from Modern Romanian and then compare them to the facts from older stages of the language. As seen below, this strategy reveals both stability at the interpretive level and some important differences in the morphological marking of shared DPs. Building on the example in (2), repeated here in (7), in Modern Romanian shared DPs in small clauses under intensional predicates can only have the following properties: a) indefinites generally carry DOM, (optional) clitic doubling, and specificity related intepretations (7) (see Dobrovie Sorin 1994; von Heusinger & Onea 2009; Chiriacescu & von Heusinger 2009; Hill 2013, etc.); b) if DOM is not possible (as in Modern Romanian DOM is generally restricted to DPs with the feature [+human]), the DPs must be definite (8a), interpreted cardinally in the singular (8b), or obligatorily specific in the plural (8c) (von Heusinger & Onea 2009; Chiriacescu & von Heusinger 2009); c) bare DPs

DPs in Adjectival Small Clauses in Romanian

295

(singular or plural) are ungrammatical without differential marking (Chiriacescu & von Heusinger 2009); d) if pronouns function as the shared argument, they must have DOM, as well as clitic doubling, as in (11) (Chiriacescu & von Heusinger 2009). (7) (L)- am considerat *(pe) un student bolnav. him= have.1.SG= considered DOM a.M.SG student sick.M.SG ‘I considered a (specific) student sick.’ (8) a. Am considerat cărțile prea scumpe. have.1.SG= considered book.PL.F.the.PL.F too expensive.F.PL ‘I considered the books too expensive.’ b. Am considerat o carte scumpă. have.1.SG= considered a.F.SG book.F.SG expensive.F.SG ‘I considered a book expensive.’ c. Am considerat niște / unele2 have.1.SG= considered some.F.PL. some.DEF.F.PL cărți scumpe. book. F.PL expensive. F.PL ‘I considered some (specific) books expensive.’ (9) *Am considerat cărți prea scumpe3. have.1.SG considered book.PL.F too expensive.F.PL Intended: #‘I considered books too expensive.’ (10) *Am vrut pisică sănătoasă3. have.1 wanted cat.F.SG healthy.F.SG Intended: #‘I wanted a cat healthy’. 2  Many native speakers will only accept the indefinite plural form unele in this context. As opposed to the indefinite niște which can alternate between a weak and strong reading in the plural, the determiner constructed from the base un ‘one’, and which obligatorily takes the definite marker (une + le), can only have specific readings. 3  As the unmarked position of the attributive adjectives is post-nominal in Modern Romanian the N+Adj string in (9) and (16) is grammatical with the adjective interpreted attributively (i.e., I considered expensive books, I wanted a healthy cat), a reading which is irrelevant for the discussion.

296

Irimia

(11) *(Ne)- au considerat *(pe) noi inteligenți/inteligente. us= have.3.PL= considered DOM us smart.M.PL/smart.F.PL ‘They considered us intelligent.’ All these examples also illustrate obligatory phi-feature (gender, number) agreement between the shared DP and the embedded adjective. If agreement is not overt, the embedded predicate can in some instances be interpreted as a manner adverb of the matrix predicate, as in (12); hence such examples are also left aside and omitted from the discussion, since the DP might not have to comply with the same (interpretive and morphological requirements) as the shared DPs in embedded small clauses. (12) *(Ne)- au

us= have.3.PL=



considerat *(pe) noi inteligent.

considered

DOM

us intelligent.ADV

‘They considered us in an intelligent manner.’

Table 1 sums up the properties of the shared DP in Romanian embedded small clauses. table 1

Shared DPs in small clauses in Modern Romanian Bare DPs

Pronouns

Definite DPs

Indefinite DPs

√

* (both singular and – only possible if DOM – require DOM if and clitic doubling plural), unless [+human] are present DOM-ed – possible in the singular if interpreted cardinally – possible in the plural with a specific interpretation

As already mentioned, the data from older stages of Romanian is both convergent and divergent. The similarity resides in the fact that only definite DPs, DOM indefinites (17) and specific indefinites (ceva in 13, non-DOM marked DPs as in 15, etc.) appear to be found in the corpora. Further examples are given below:

DPs in Adjectival Small Clauses in Romanian

297

(13) Unde să veți afla ceva neisprăvit where if FUT.2.PL find something unaccomplished.M.SG bine sau greșit . . .  well or wrong.M/N.SG ‘Where, if you find something which is not well accomplished or wrong . . .’ (Coresi EV, 63) (14) Pentru aceaia fericat iaste omul cela ce for that blessed.M.SG is man.the.M.SG that.M.SG who se află nevinovat. REFL.3 finds innocent.M.SG. ‘For that reason, happy is the man who is innocent.’ (Coresi EV 48) (15) Cine va vedea mortul înviat și să who FUT.3.SG. see.INF dead.the.M.SG ressurected.M.SG and SUBJ nu se veselească? not REFL.3 rejoice.SUBJ.3.SG ‘Who is that person who will see the dead resurrected and not rejoice for it?’ (Coresi EV 23) (16) Când ar fi văzut îngerul din cer cu when COND.3.SG be.INF seen angel.the.M.SG from sky with veste bună și de izbândă pogorit. tidings.SG good.F.SG and with victory descended. M.SG ‘When he would have seen the angel from the sky with good news and descended with victory . . .’ (Coresi EV 85) (17) Pre cazaci să -i lase DOM Cossack.M.PL SUBJ =them.M.PL leave.SUBJ.3 neatinși. untouched.M.PL ‘They should leave the Cossacks untouched.’ (Ureche 204) At this stage of the language, pronouns can also function as shared arguments with embedded non-finite adjectives. However, differently from Modern Romanian the differential object marking is not obligatory. The examples in (5) and (6) are repeated here as (18), and (19). Note that in (18) the DOM on the shared argument is missing in the small clause (such a structure would therefore be ungrammatical in Modern Romanian), while it is found with the

298

Irimia

pronoun in the first matrix clause (ispiti pre ei). In (19) on the other hand, DOM is seen on the shared argument (pre el). (18) Dumnezeu ispiti pre ei și află ei destoinici luiș. God tested DOM they and found them loyal.M.PL to.him ‘That God put them to test and found them loyal to him.’ (Coresi EV 260) (19) Ispitind pre el diavolul  . . ., află pre el nebiruit test.GER DOM he devil.the found DOM he invincible.SG.M ‘When the Devil put him to test, he found him invincible.’ (Coresi EV 520) More examples illustrating this alternation are provided below. Note that in (20), (21) and (22) the pronoun has DOM. (20) Să nu ne afle pre noi mirele SUBJ nu us= find.SUBJ.3.SG DOM us groom. DEF.M.SG adurmiți și leninindu -ne asleep.M.PL and slack off.GER =REFL.1PL ‘So that the groom doesn’t find us asleep and slacking off.’ (Coresi EV 90) (21) Deci de-aciia vrea vedea pre el răstignit. Hence from-here will.3.SG = see DOM he crucified.M.SG ‘So he will see him crucified.’ (Coresi EV 115) (22) Nici pre noi nu lăsă săraci. nor DOM us not left.3SG poor.M.PL ‘Nor did he leave us poor’. (Coresi EV 88) Once again, what differentiates Modern Romanian from older stages is the obligatoriness of DOM with pronouns in the former variant. The similarities, as well as this relevant difference are summarized in Table 2. These patterns require a unified explanation. There are two aspects an analysis needs to capture with respect to the structure of embedded small clauses and the nature of their shared DPs. First of all, what type of configuration predicts the stable character of non-pronominal DPs? More specifically, why are weak/non-specific DPs not possible in these contexts? And secondly, given the specificity/strong readings imposed on the shared arguments in these constructions, why do pronouns have a variable behavior? An investigation into the second question requires a clear answer to the first question. More simply put,

299

DPs in Adjectival Small Clauses in Romanian table 2

Shared DPs in small clauses in Old versus Modern Romanian

Definite DP s

Indefinite DP s

Bare DP s Modern Romanian

Modern Romanian

16th c Romanian

Modern Romanian

16th c Romanian

√

√

– require DOM if [+human] – possible in the singular if interpreted cardinally – possible in the plural with a specific interpretation

Pronouns 16th c Romanian

* (both singular and plural); unless DOM-ed (if respecting the conditions for DOM, i.e., animacy, etc.)

Modern Romanian

16th c Romanian

only possible if DOM and clitic doubling are present

DOM and clitic doubling are not obligatory

it is necessary to understand what the structure of embedded small clauses is and its interactions with specificity on shared DPs. Section 3 presents a more comprehensive, cross-linguistic picture of the behavior of shared arguments in embedded small clauses, demonstrating without doubt that the restrictions on specificity are a defining property of this configuration, and not a quirk of Romanian. This solid conclusion in turn allows us to tackle the problem of the structure of embedded small clauses in section 4. 3

Small Clauses and Their Shared DPs: Beyond Romanian

The observation that shared arguments with embedded adjectival predicates carry interpretations generally identified under the class of specificity is crosslinguistically robust. In fact, it can be seen in language after language that various specificity strategies are employed in order to construct strong DPs (Milsark 1974) in these instances, which I illustrate in this section for consider + AdjP constructions. Moreover, traditional grammars systematically signal the special morphological behavior of objects in such contexts, connecting it to ‘broad specificity’ interpretation: an entity is made salient/made specific/ identified/individualized within someone’s mind, but the audience might not precisely identify which entity the mind has individualized (see de Hoop 1996

300

Irimia

for Finnish; Kachru 2006 for Hindi-Urdu, Mandarin Chinese; general discussion in Postal 1974, a.o.). Some of the morphological strategies for shared DPs are listed below: – Bare indefinite DPs must carry DOM and be interpreted specific; see (23) for Hindi and (27) for Turkish. – If an indefinite DOM is impossible, only definites are allowed; see mass nouns in (25) for Romanian and (31) for Italian. – If weak Case is used, as in languages that lack morphological (in)definiteness (e.g., Finish), there is a mismatch between morphology and semantics; e.g., no weak readings are allowed in (26), although the essive is not the Case ‘strong DPs’ normally carry. – Strong Case must be further ‘strengthened’ by overt specificity marking; e.g. Arabic in (28) – In classifier languages, demonstratives are obligatory in these instances; see Chinese in (29). – Split behavior of mass nouns: in some languages mass nouns are only possible if definite; e.g., Romanian (25) and (31) in other languages bare/mass DPs are possible with generic interpretation only; e.g., English in (30)

• •

HINDI-URDU (INDO-IRANIAN) (23) admi kitab-ko/*Ø acha səməjhta man.M.SG book.DOM good.M.SG think.PRES.PRT.M.SG. hɛ. be.PRES.3.SG. ‘The man considers the/a book good.’ (a book >> consider; #consider>> a book) SPANISH (ROMANCE) (24) El professor consideró a/*Ø un estudiante intelligente. The professor considered (a= DOM) a student intelligent. ‘The professor considered a specific student intelligent.’ (a student >> consider) (25)

ROMANIAN (ROMANCE) Vânzătorul consideră mierea/*miere ieftină. Store clerk.the considers honey.the.F.SG./honey cheap.F.SG ‘The store clerk considers (the) honey cheap.’

DPs in Adjectival Small Clauses in Romanian

301

FINNISH (FINO UGRIC) (26) Miehet pitävät oppilaita ilois-i-na. Man.PL.NOM consider.PRES.3.PL student.PART.PL happy.PL.ESS ‘The men consider the/ specific students happy.’ (27)

TURKISH (ALTAIC) Ali bir öḡrenc-i-yi/*Ø zeki bulu-yor. Ali a student-EP.V.DOM/*Ø intelligent find-PRES.PROG.3.SG ‘Ali finds/considers a (specific) student intelligent.’ (a student >> find; *find>> a student)

ARABIC (SEMITIC) (28) ʔəʕtəbiru Ta:lib-ən *(bi-ʕəyni-hi) kəsu:l-ən 1.SG.consider student-ACC. in-same-him lazy-ACC ‘I consider a specific student lazy.’ (29)

MANDARIN CHINESE (SINO-TIBETAN) Jiaoshou renwei nage/*Ø shuesheng hen congming. Teacher consider DEM. student very smart. ‘The teacher considers that/the student smart.’

ENGLISH (GERMANIC) (30) These people consider meat healthy. ITALIAN (ROMANCE) (31) Considero il miele/*Ø costoso. Consider.1.SG. the.MS.SG honey expensive.M.SG. ‘I consider (the) honey expensive.’ Both the overt morphological markings, as well as the interpretations these types of shared DPs receive correspond to the so called strong readings, in Milsark’s (1974) typology. Based on syntactic and semantic properties, Milsark (1974) pioneered a classification of nominal phrases into two important subgroups: a) weak/narrow scope/non-specific DPs (which are characteristic to existential contexts, and can be used bare); b) strong/wide scope/specific DPs (which are normally more complex morphologically, take wider scope with respect to other operators and cannot be used bare). Milsark (1974), as well as subsequent work, have also identified interpretational properties of each of the two classes. Given that shared DPs in embedded small clauses are uniformly marked as strong DPs which exhibit a relevant host of properties, an

302

Irimia

analysis must be formulated that can derive the wide scope as well as the differential marking. As will be shown in section 4, base generating a small clause containing the shared argument and the embedded adjectival predicate does not predict the right output. Hence strong DPs in these configurations indicate something non-trivial about the structure of embedded adjectival predicates and the nature of wide scope readings. Before addressing the structure in more detail, one more aspect needs to be made precise about the ‘consider-AdjP’ contexts. As seen from the various examples above, the embedded adjective requires phi-feature agreement (excluding person, as adjectives do not normally carry person morphology) with the shared DP. Romanian, Italian, and Hindi-Urdu illustrate this pattern. The same facts hold for a language like French; see (32). Languages like Arabic also require agreement in Case, as shown in (28), although cross-linguistically a dedicated Case strategy is also a prevalent option, as seen in Finnish (33), with essive Case on all embedded adjectives (depictives and complement adjectival predicates): or in Russian—see (34) with instrumental Case. FRENCH (32) Jean considère les femmes intelligentes/*intelligent. Jean.M.SG considers the.PL women.F intelligent.F.PL/*intelligent.M.SG ‘Jean considers the women intelligent.’ FINNISH (33) a. Sö-i-n tomaati-n raaka-na. (Depictive) eat-PST-1.SG tomato-ACC raw-ESS ‘I ate a tomato raw.’ (Pylkkänen 2008, ex. 34) b. Miehet pitävät oppilaita ilois-i-na. man.PL.NOM consider-PRES.3.PL. student.PART.PL happy.PL.ESS ‘The men consider the/specific students happy.’ RUSSIAN (34) Ja sčitaju Ivana umnym/*umnogo. I.NOM consider Ivan.ACC.M.SG intelligent.INSTR./intelligent.ACC ‘I consider Ivan intelligent.’ Exhaustive agreement between the shared argument and the embedded predicate is a strong indication of a small clause configuration, which however is problematic for deriving the wide scope readings. Hence the challenge for an analysis that could reconcile the two aspects becomes even more difficult.

DPs in Adjectival Small Clauses in Romanian

303

Section 4 discusses and evaluates possible accounts, and proposes that a complex predicate implementation is equipped best to derive the right results. 4

Embedded Adjectival Predicates and Their Shared DPs

When examining the properties of adjectival secondary predicates, the linguist is puzzled to see that they are crosslinguistically uniform, in the sense that they respect other sets of common characteristics besides the property in a) below: a) b) c) d)

the requirement that the shared argument receive a strong interpretation (as opposed to other non-finite instances which might permit weak readings); binding facts, which demonstrate that the adjectival secondary predicate does not project an independent binding domain, separate from the matrix predicate; lack of subject control readings; occurrence with cross-linguistically stable classes of matrix predicates.

Each of these diagnostics is presented below, comparing Romanian with English in the paradigms in (35) to (38). 4.1 One Universal Fact: Strong Readings of Shared DPs As already shown the shared argument with adjectival small clauses must be interpreted as specific; this entails a wide scope reading with intensional predicates, as in the English (35). ENGLISH (35) John considers a student intelligent. a student » consider = John considers a (specific) student intelligent. #consider » a student # John considers some student or other intelligent. Embedded non-finite adjectives contrast in this respect with embedded infinitivals under main predicates like consider, in that they might allow weak readings of the shared argument: (36)

ENGLISH John considers a student to be intelligent. a student » consider = John considers a (specific) student to be intelligent consider » a student = John considers some student or other to be intelligent.

304

Irimia

As already mentioned, this property also holds in both Old and modern Romanian; moreover, the two stages also provide morphological overt evidence that signals the specificity/wide-scope status of the shared argument. As seen throughout the paper and in Tables 1 and 2, the object with adjectival secondary predicates under intensional verbs is normally either definite, specific indefinite or differentially marked with pre. The Old Romanian example (17) containing a differentially marked indefinite is repeated here as (37), for convenience. The absence of narrow scope readings is also strengthened in example (38) from Modern Romanian where the shared argument cannot receive an existential reading: (37)

Pre cazaci să -i lase DOM Cossack.M.PL SUBJ =them.ACC.M.PL leave.SUBJ.3 neatinși. untouched.M.PL ‘They should leave the Cossacks untouched.’ (Ureche 204)

(38)

(L)- am considerat *(pe) un student bolnav. him= have.1.SG considered DOM a.M.SG student sick.M.SG ‘I considered a student sick.’ a) ‘I considered a (specific) student sick.’ b) ≠ ‘I considered some student or other sick.’

At this stage, it is opportune to examine in more detail the nature of pe/premarked arguments, as there are at least two logical possibilities regarding the source of wide-scope readings. On the one hand, it is often assumed that these types of interpretations arise as a result of a special syntactic configuration, generally derived by movement of the DP to a position above the intensional predicate. On the other hand, it has also been demonstrated that there are situations in which a position of the DP below the matrix predicate does not necessarily prohibit the argument from taking wide scope. The correct answer for the configurations under discussion here will have syntactic consequences on the structure of secondary predicates. More specifically, it helps disambiguate between a small clause account (39 a) and a complex predicate analysis (39 b). The major difference between the two structures in (39) is that in (39 a) the shared argument is base-generated below the matrix predicate, and further raising to a position above the matrix predicate entails the wide-scope reading. In (39 b), on the contrary, the shared argument is merged high to begin with, predicting the correct high scope facts.

305

DPs in Adjectival Small Clauses in Romanian

(39) a. . . . . .

b. ... . . consider

SC

pe/pre un student

pe/pre un student

inteligent

Small clause structure

consider

inteligent

Complex predicate structure

(39) b. . . . . .

pe/pre un student consider intelligent Complex predicate structure

There are various diagnostics whose application can detect the presence of a copy below the intensional predicate; for example, the position should be available for reconstruction processes which could potentially restore the narrow scope reading. However, as observed crosslinguistically, narrow scope interpretations are generally blocked in these environments. However, in order to proceed with a discussion of reconstruction, and to unambiguously decide which of the structures in (39) is to be assumed, let’s examine first the nature of the pre/pe marker. 4.2 Differential Object Marking in Romanian Traditional Romanian grammars, as well as more recent formalizations (Leonetti 2007; Klein 2007; von Heusinger & Onea 2009; Hill 2013; Cornilescu & Dobrovie Sorin 2008; Farkas & von Heusinger 2003, etc.) have long been puzzled by the nature of this prepositional-like marker, being caught in a debate on whether it encodes a Case marker (the accusative Case, as in Cornilescu & Dobrovie Sorin 2008) or something else (discourse features in Mardale, this volume; Hill 2013). The Academy Grammar notes that in Romanian the marker pe/pre is normally restricted to indefinites or bare nouns, excluding definites. The following contexts of use are further mentioned: a) when the indefinite noun (phrase) is animate, especially human; or b) if inanimate, the noun (phrase) is specific; or c) the noun phrase is used with secondary predicates.

306

Irimia

From a more recent perspective, the collapse of specificity with sensitivity to animacy suggests that pe/pre marks a mixed differential object (DO) strategy. If this assumption turns out to be correct, Romanian mirrors the morphosemantic picture these types of objects have cross-linguistically. For example, languages which encode such objects overtly require the differential encoding of the object in embedded small clause overriding features [+human] (as seen in the examples in 23–31, etc.). Also, differential objects are normally interpreted as specific, and must take wide scope (Bossong 1991; Bittner 1994; Torrego 1998; Aissen 2003; Öztürk 2005; Rodríguez-Mondoñedo 2007; de Swart 2007, etc.). And, lastly, in the majority of languages their special status is normally indicated by adpositional material. The question is now: why are differential objects the unmarked option with secondary predicates? The two structures in (39) provide a straightforward answer in orthogonal ways. If the secondary predicates project a small clause (39 a), the shared argument has to move to a position above the matrix predicate in order to have its Case features checked (assuming a pre-Derivation By Phase framework); adjectival small clauses are non-finite domains in which the relevant structural Case projection is not available. The differential marking signals the derived high position of the argument. The complex predicate analysis on the other hand assumes that the two predicates are merged directly, and the shared argument is independently introduced to a position above the complex formed. Given its high position, it will carry differential marking. Distinguishing between these two hypotheses turns out to be a complex task; however, as shown is section 5 the small clause analysis ends up in circularity when pushed to its limits. A complex predicate analysis makes better predictions. 5

Theoretical Debates on the Structure of Adjectival Embedded Predicates

The observation that shared DPs with adjectival Secondary Predicates must take wide scope was discussed at length by Williams (1983), who examined sentences like (40) contrasting an embedded adjectival predicate (40 a), and an infinitival (40 b). ENGLISH (40) a. A student seems sick. b. A student seems to be sick.

(Williams 1983, 293, ex.40a)

DPs in Adjectival Small Clauses in Romanian

307

In order to make the distinction more transparent, assume a context like the following for the two sentences above: (41) I walk into the classroom and I see some pill cases on one of the desks. Only sentence (40 b) is possible as a continuation to this small fragment; it reports that the evidence indicates that a student is sick, but not necessarily a particular one. This is the non-specific indefinite reading, as the indefinite DP can refer to a non-specific entity. Sentence (40 a) on the other hand would be infelicitous for describing the context in (41), as the shared DP with a secondary predicate does not permit the non-specific indefinite reading. In order for (40 a) to be acceptable it must be the case that there is a specific, previously introduced, discourse or contextually salient student who is sick. The canonical generative grammar approach to interpretative contrasts along the specificity line follows the pioneering analysis of scopal relations formulated in May (1985), who scrutinized sentences like (42). (42) Some politician is likely to address John’s constituency. May’s (1985) crucial observation is that the sentence above is ambiguous. Its two readings can be further paraphrased as follows: (i) There is a politician, e.g. Rockefeller, who is likely to address John’s constituency, or (ii) It is likely that there is some politician (or other) who will address John’s constituency. Crucially, May (1985) further connected the two readings to structural differences. The interpretation in (i), which introduces a specific referent, was assumed to correspond to a structure in which the DP is (interpreted) above the matrix predicate (is) likely, as shown simplistically in (43b)/(44b). Specificity is thus associated with the wide scope of the shared DP. The non-specific reading, May (1985) argued, must be read off a configuration in which the DP is below the predicate (is) likely, most probably inside the infinitival clause, as in (43a)/ (44a). A process of covert quantifier lowering inside/adjoined to the non-finite clause allowed the reconstruction of the (existential) quantifier in the embedded domain.

308

Irimia

(43) a. . . . . . .

is likely

IPinfinitival



a politician to address . . . narrow scope reading

A politician is likely to address. . . . . [Some politician or other is likely to . . .] is likely >> a politician (43) b. . . . . . . . . a politician is likely to address wide scope reading A politician is likely to address. . . . . [A (specific) politician is likely to. . . . .] a politician >> is likely (44) a. . . . . . .

seem IPinfinitival

a student to be sick narrow scope reading A student seems to be sick. [Some student or other seems to be sick] Seems >> a student

309

DPs in Adjectival Small Clauses in Romanian

(44) b. . . . . . .

a student

seem sick wide scope reading A student seems sick. [A (specific) student seems sick] a student >> seems Following May (1985), Williams (1983) took the impossibility of the narrow scope reading with secondary predicates to indicate that the existential quantifier is not found in the embedded subject position at any stage in the derivation. If the subject were indeed base-generated lower than the matrix predicate (as the subject of a small clause à la Stowell 1981, or Chomsky 1981) then the possibility of the existential (narrow scope) interpretation would be expected to arise (by quantifier lowering), on a par with the wide scope one (obtained after the subject has raised). Simplifying the structure of small clauses, the configuration in (45) would be expected if the quantifier were allowed to lower inside secondary predicate domains. Tp

(45) a student a student >> seem Wide scope of quantifier

.. . . . . . . seem S(mall) C(lause)

sick

Seem >> a student Narrow Scope of quantifier (Weak indefinite reading) Impossible

310

Irimia

The question is what the absence of narrow scope readings tells us about the structure of secondary predicates. A radical conclusion, argued for by Williams (1983), is that since the small clause subject position is not available for reconstruction, it is probably not present in the syntax. More straightforwardly, it must be the case small clause structures don’t exist for non-finite embedded adjectives, and therefore the grammar should not contain a specific theoryinternal subcomponent dedicated to them. More importantly, by eliminating small clauses, the conceptual force of the theory is not in any way weakened; there are other ways in which the puzzle of secondary predicates can be adequately explained. Williams’ (1983) account is not the only possible answer to the interpretational puzzle of shared arguments with secondary predicates. Other researchers, among which Lasnik (1999), Chomsky (1993, 1995), Sportiche (2005), Matushansky (2002), and Basilico (2003), argue in favour of a small clause account for secondary predicates, attributing the lack of narrow scope readings to other, independent reasons (among which the impossibility of A-movement to feed reconstruction). Some of the small-clause accounts will be briefly discussed below, with their shortcomings. What these analyses have in common is the idea that reconstruction into an original subject position inside the small clause is not possible due to independent reasons. Three hypotheses are salient: a) A-movement does not feed reconstruction (Chomsky 1995; Lasnik 1999); b) the subject of adjectival secondary predicates has a topic status small-clause internally, and hence is interpreted as taking wide scope even if embedded under the intensional predicate (Basilico 2003); c) small clauses are not quantification domains, or crucial structural layers pertaining to quantification are absent in them (Sportiche 2005). Pushing to its limits the assumption that A-movement does not feed reconstruction leaves the infinitival cases (40b, 43b) unexplained. On the one hand, there are situations in which reconstruction under A-movement could be possible (Boeckx 2001). In order to propose a unified analysis for both infinitives and adjectival secondary predicates, one must make the assumption that embedded adjectival predicates do not instantiate small-clause configurations. A possible solution according to which the various readings with shared DPs in these instances are due to the vagueness of indefinites (Lasnik 1999) is not totally satisfactory either, as it cannot predict how ‘vagueness’ works in these instances (i.e., where strong vs. weak readings are obtained). Turning to the topic status of the shared DPs, Basilico’s (2003) analysis implies that narrow scope readings must be systematically absent from these configurations. But this is not empirically correct. Crucially, narrow scope interpretations are possible with modal adjectives only, as seen in (46).

DPs in Adjectival Small Clauses in Romanian

311

ENGLISH (46) The man considers a book necessary. considers » a book (i.e. the man considers some book or other necessary) a book » consider (i.e., the man considers a specific book necessary) Note that the observations extend to the Romanian data analyzed in this paper; that is, instances containing modal adjectives functioning as secondary predicates allow specific and existential readings. Observe that, as in example (47), in these instances the specificity differential object marker –pre can be dropped more easily in modern Romanian; see (47). (47) Profesorul consideră un student necesar pentru proiect. teacher.DEF considers a.M.SG student necessary.M.SG for project a. The professor considers a (one) student necessary for the project. b. The professor considers some student or other necessary for the project. As topics have specificity as one of their semantic correlates, examples like (46) and (47) are not only unexpected, but also impossible to derive under Basilico’s account. A more plausible explanation is that the low scope interpretation is obtained in these instances because of the inherent semantics of modal adjectives, which have been argued to normally take wide scope with respect to other quantificational elements (see Moulton 2013). Under a complex predicate account the data can be derived straightforwardly; the modal adjectives take scope over the high merged shared DP inside the complex. Lastly, the reasoning that narrow scope readings are not possible with subjects of adjectival predicates because small clauses are not full domains of quantification has its problems, too. As Williams (1983) already remarked, if it were true that small clauses are scopally defective in important respects, blocking certain types of quantification, we would expect weak interpretations of indefinites to be systematically absent inside small clauses. But again, this doesn’t seem to be the case. In sentences (48 a and 48 b) from English, the indefinite in the complement to the adjectival secondary predicate can be interpreted both weak and strong. ENGLISH (48) a. John seemed upset with a friend. = John seemed upset with a specific friend. = John seemed upset with some friend or other.

312

Irimia

b. They considered the student happy with a book. = They considered the students happy with a specific book. = They considered the student happy with some book or other. There does not seem to be any non-stipulative way to explain the absence of the narrow scope readings in a putative subject position inside the small clause except for the assumption that that position is not generated to begin with. Hence a complex predicate structure makes the right predictions without further unnecessary machinery. This proves to be on the right track because it also predicts the binding facts (discussed in 5.1). 5.1 Binding Facts In many languages negation on the adjectival secondary predicate takes main clause scope (i.e., it behaves as if interpreted in the domain of the matrix verb). To better understand this point, let’s examine a language where such characteristics are more prominent. For example, let’s look at dialectal Hindi-Urdu adjectival embedded predicates illustrated in (49), and (50), where the positioning of negation is variable. HINDI-URDU (INDO-IRANIAN) (49) ek-bhii laṛke-ne machli nahĩ: kacch-ii kaa-i. one-NPI. boy-ERG. fish.F.SG NEG raw-F.SG eat-PFV.F.SG ‘One boy ate the fish not raw.’ (i.e., some fish-eating might have taken place, but the fish wasn’t raw)// ‘Not even a single boy ate the fish raw.’ (dialectal variant) (49) is not not truth-conditionally equivalent to a sentence in which negation appears on the main predicate, as in (50). The reading of the latter is that no fish eating has taken place, no matter whether the fish would have been raw or cooked. HINDI-URDU (INDO-IRANIAN) (50) ek-bhii laṛke-ne machli kacch-ii nahĩ: kaa-i. one-NPI boy-ERG. fish.F.SG raw-F.SG NEG eat-PFV.F.SG ‘Not even a single boy ate the fish raw.’ (i.e., the boy did not eat fish at all) These two sentences contain a negative polarity item on the subject of the matrix predicate. Crucially, in (49) negation is not placed adjacently to the matrix predicate (as in 50), but rather in the domain of the adjectival secondary predicate. The fact that this sentence is not interpreted as a negation on the main predi-

DPs in Adjectival Small Clauses in Romanian

313

cate indicates that the negation is not found syntactically in a position above the main predicate only. Sentence (49) is rather felicitous in a context in which some fish eating might have taken place, but crucially the fish consumed wasn’t raw.4 What is relevant is that even though negation is in the domain of the embedded adjective, it can take scope over and bind the negative polarity item (ek-bhii) in the matrix predicate domain. A complex predicate analysis derives the facts straighforwardly. The negation diagnostic strengthens the complex predicate analysis. In Section 5.2 another diagnostic that supports these two conclusions is also introduced. 5.2 No Subject Control Verbs A classic observation regarding the constructions under scrutiny here is that verbs selecting adjectival predicates are not normally subject–control classes (Chomsky 1981; Postal 1974; Hornstein 1999, a.o.): ENGLISH (51) *John considers intelligent. (i.e., John considers himself intelligent). ENGLISH (52) *John wants happy.

(i.e. John wants to be happy).

There has been extensive discussion as to why this restriction holds (see especially the detailed presentation in Chomsky 1981). A complex predicate structure predicts the facts. Under a small clause analysis, intensional predicates like consider would require a control-type configuration in which co-reference is mediated by PRO in the subject position of the small clause headed by the adjectival predicate (assuming here a very simple syntax of control). But if there is no such position, PRO can not be part of the structure.5 As expected, the no-subject-control restriction is valid in Romanian:

4  For speakers of the standard variety of Hindi-Urdu, the sentence in (49) appears to be degraded (but not completely out; rather it should be marked as ??). I thank Rajesh Bhatt for clarification. 5  Seem-type contexts, which are not discussed in this paper, might appear to be an exception, because of examples like He seems sick. However, this is only apparent. In language after language, these structures exhibit properties which are clearly distinct from canonical control.

314

Irimia

ROMANIAN (ROMANCE) (53) *Profesorul consideră bun teacher.DEF.M.SG considers good.M.SG Intended: ‘The professor considers himself good.’ ROMANIAN (ROMANCE) (54) *Regele vrea viu. king.DEF.M.SG wants alive.M.SG Intended: ‘The king wants to be alive.’ Lastly, to finish the comparison between Romanian and their correspondent structures elsewhere, it can also be mentioned that the inventory of matrix predicates selecting adjectival secondary predicates is the typical one: vrea ‘want’; considera ‘consider’; declara ‘declare’; spune ‘tell’; găsi ‘think’, ‘find’, ‘consider’, etc. 6

Multiple Agreement and Complex Predicates

The discussion of the last two sub-sections concluded that a complex predicate analysis can straightforwardly derive the lack of narrow scope readings of DPs with non-modal adjectival embedded predicates. In this section, the precise mechanics of shared agreement is investigated: 6.1 starts by eliminating some logically plausible explanations for the propagation of exhaustive parasitic agreement to adjectival secondary predicates. Sub-section 6.2 introduces an enriched complex predicate analysis under which shared agreement arises as a result of a Multiple Agree operation applying simultaneously with two predicates. 6.1 Previous Analyses Configurations of shared, multiple agreement have been under intense scrutiny in the minimalist program. A recent structural evaluation is found in the Derivation by Phase-based (DBP)—style syntactic implementations. Following the explicit remarks in Chomsky (1993) adjectival secondary predicate configurations are collapsed with participial parasitic agreement instances. More specifically, a minimalist theory of agreement has to explain the mechanics of agreement on participial forms like caught (55). Note that in inflectionally richer languages, φ and Case agreement on the participial is overt in these instances, as shown in example (57 a) from Icelandic, where it tracks its semantic argument. In Icelandic, adjectival secondary predicates also show exhaustive

DPs in Adjectival Small Clauses in Romanian

315

agreement with the semantic argument in consider contexts, as illustrated in (57 b). (55) There seem to have been caught many fish. (Chomsky 2001, ex. 18) . . . . . . . . expect to have been caught many fish (56) a. C[β T seem b. [β v expect

[Expl to have been [α caught several fish]]

ICELANDIC (57) a. Participial overt exhaustive agreement Það virðist sem margir fiskar hafi verið veiddir There seems that many fish.PL.N have been caught.PL.N ‘Many fish seem to have been caught.’ b. Secondary predicate overt exhaustive agreement Monica tel barnið gafað. Monica considers child.the.N.SG.ACC smart.N.SG.ACC ‘Monica considers the child smart.’ The DBP model assumes double Agree operations generate sentences like (57). The probes (T or v) agree with Expl (this step is not discussed here for reasons of space) and fish. T deletes the uninterpretable feature of Expl (inducing raising) and values Nominative on fish (56a). v deletes the uninterpretable feature of Expl (without raising to [Spec,v], and values Accusative Case on fish (56 b). The second Agree operation involves the participial/adjectival secondary predicate. More specifically, the participle is assumed to establish a phiAgree relation with the argument fish. But since in languages like Icelandic Case agreement is also visible, its presence on the participial must also be explained. The Case facts that have to be captured by the theory are that, normally, the participial is Nominative with probe T, and Accusative with probe v. Crucially, in order for the Agree relation between the participial/adjectival secondary predicate and the shared argument to be straightforwardly established, a small-clause configuration needs to be taken for granted. Assuming, for the sake of the argument, that the object is licensed as an argument of a small clause projected by the participial head, let’s focus our attention on that stage of the derivation in which the shared argument and the participial are in a local configuration (58). In order to simplify the discussion, that stage is labeled here ‘cycle α’.

316

Irimia

(58) [α PPART [catch [DO several fish]]] u# # uγ γ uC uC The precise mechanics of cycle α, which is also the main interest for the embedded adjectives, is as follows. As the PPART head has an adjectival nature, “its φ-set may consist of (unvalued) number, gender, and Case, but not person” (Chomsky 2001). Its unvalued features initiate the operation Agree, targeting the shared argument (closest, in fact only DP in its domain). The φ-sets of PPART and DO match (DO is φ-complete), inducing Agree. Number and gender features of the PPART receive values and delete. At the next stage of the cycle (stage β, see 56), Case is assigned to the DO: nominative with probe T and accusative with probe v. But this leaves the PPART with its Case feature unvalued, leading to a crash in the derivation. Chomsky’s idea for a repair strategy builds on the notion of phase. The φ-features of PPART are still visible at stage β of the cycle, though deleted, because T and PPART are not strong phases. Valued features will only disappear at the strong-phase level CP or vP, as the phase is transmitted to the phonological component. As a result of this feature visibility extension, PPART can also enter into a Case agreement relation. Hence, PPART and DO agree directly for number/gender, and indirectly for structural Case (since each agrees with the probe). Another important remark concerns the basic nature of predicative adjective features. The general assumption in DPB, as well as MPLT (Chomsky 1993) is that predicative adjectives establish, in cycle α, automatic agreement with the closest argument (the specifier/complement in a small clause configuration). The two frameworks a-priori block a mechanics in which the unvalued features of the predicative adjective could skip the closest DP, preferring later valuation. The challenge presented now is how to reconcile these agreement patterns, which suggest a small clause configuration, with the absence of narrow scope readings. 6.2 New Analysis To recapitulate, the various diagnostics introduced support one important conclusion: a complex predicate analysis is more adequate than a small clause analysis to explain the scope facts. Remember that a skeletal configuration that predicts default scope readings on the shared DP in canonical adjectival secondary predicate contexts like (2a), repeated here as (59), is as in (60). The two predicates merge first, and then the shared argument is introduced. Crucially,

DPs in Adjectival Small Clauses in Romanian

317

the shared argument and the secondary predicate do not form a small clause constituent at any stage in the derivation: (59) (L)- am considerat *(pe) un student bolnav. him= have.1.SG= considered DOM a.M.SG student sick.M.SG ‘I considered a student sick.’ (60)  . . . . . . .

pe un student



considerat bolnav

A complex predicate syntax along the lines in (60) is by no means new. It has been proposed in Chomsky (1975), to cite just a classic reference. However, several aspects of it have to be further worked out in order to provide an adequate account of the Romanian data. Two questions are particularly important: (i) What is the status of the shared argument? (ii) What is the specific mechanics of the operation of complex predicate formation? My proposal is that the structure in (60) has to be enriched as in (61). (61) . . . . . . . . vcmpl book

Shared Arg. Introducer



good consider

Based on cross-linguistic consistent morphological markings, it seems safe to assume that the shared argument is introduced by a dedicated functional projection. But nothing hinges on this. Note that the facts would also be explained by assuming that the argument is compositionally introduced by the predicate complex; the only reservation is that in this case more sophisticated mechanics would be needed. What is important for the current analysis is that the shared argument be merged high. The more precise nature of the projection introducing it would require a discussion that goes beyond the space limits of this paper, and is thus left aside.

318

Irimia

The main assumption regarding the structure of these embedded predicates is that they undergo a process of complex predicate formation in syntax, and don’t project small clauses. This analysis further builds on Rothstein’s (1985) idea: there are predicates which cannot saturate their features directly. More specifically, these are predicates that cannot take their subject directly— they have to be predicated of an argument introduced by/in the domain of another predicate. The computational system of FL contains a procedure by which the [+pred] feature of more than one head can be checked simultaneously. The ingredients of this simultaneous checking operation are the process of Multiple Agree (Hiraiwa 2004) and its implementation to the domain of predicates. Hiraiwa (2004) formalized a process of multiple agreement which operates in those instances in which multiple simultaneous identical Case markings spread to more than one syntactic object. He focused mainly on sentences like (62) from Japanese: JAPANESE (62) Taro-ga/ni Hanako-ga me-ga waru-ku kanji-rare-ta. Taro-NOM/DAT Hanako-NOM eye-NOM bad.INF think-PST ‘Taro thought that Hanako had a bad eyesight.’ (Hiraiwa 2004, ex. 2.4) These types of sentences are puzzling because the Nominative Case ‘propagates’ across a non-finite complementation domain. The subject of the matrix clause is expected to bear Nominative Case, but the morphological markings of the subject and the object of the embedded non-finite clause are surprising. In Japanese, just like in English, the verb think can function as an ECM-inducing predicate, checking Accusative Case. But the Case on the embedded subject, the DP mega ‘eye’, is Nominative. Even more unexpected is the presence of the Nominative on the embedded clause possessor, Hanako. In this context, the Genitive Case would normally be predicted in Japanese. In order to account for this apparently strange state of affairs, Hiraiwa (2004) assumes that a single probe (matrix T) can check the uninterpretable Case features of several goals which happen to be in the required space at some moment in the derivation. Implementing Ura’s (1996) observations about multiple feature checking, Hiraiwa (2004) formalizes the mechanism of Multiple Agree as in (63), (64), and (65): (63) MULTIPLE AGREE (multiple feature checking) with a single probe is a single simultaneous syntactic operation; AGREE applies to all the matched goals at the same derivational point derivationally simultaneously. (Hiraiwa 2004, page 38)

319

DPs in Adjectival Small Clauses in Romanian

(64) MULTIPLE AGREE (P, ∀G) Agree is a derivationally simultaneous operation AGREE (P, ∀G)

P > G1 > . . . . . . . > Gn

(65) THE PRINCIPLE OF SIMULTANEITY Apply operations simultaneously at a probe level.

(Hiraiwa 2004, 2.9)

Assuming a correlation between multiple Case spreading with arguments, and shared agreement with 2> predicates, this paper extends and specifies the nature of Multiple Agree to the domain of predicative complexes. When applied to complex predication formation, Multiple Agree is an operation that values the Pred (and other uninterpretable) features of (two) predicates. The process is realized simultaneously and initiated by a functional projection endowed with the capacity of valuing and transmitting the relevant features of more than one predicate. The principle of complex predicate formation is given in (66). (66) PRINCIPLE OF COMPLEX PREDICATE FORMATION [uPredicate/uφ] features of more than one predicate in the same phase are checked derivationally simultaneously by a probe which can establish an AGREE relation with a goal containing the relevant interpretable [φ]features. Multiple agreement initiated by the functional projection labeled vCMPLX in order to emphasize its contribution to the formation of the complex. Checking is initiated at v, instead of T (see also Béjar & Rezac 2009), in order to explain the common cross-linguistic object agreement patterns with such constructions, as well as their complex predicate nature (as was seen from anti-reconstruction patterns in the interpretation of shared arguments, binding effects, etc.). More specifically, vCMPLX establishes first an agreement relation (1) with the closest DP and obtains the relevant [φ]features features which are then (2) transmitted simultaneously to the multiple predicates in the configuration. A sample derivation is provided in (67).

320

Irimia

(67) (1) vcmplx > G1 > . . . . > Pred1 > . . . . . . . . Predn (2) A sample derivation: (68) The man considers the women intelligent. Num/Lexicon = {the, man, Sit0, vCMPLX , considers, DEP, a, √good} [Sit0 = to introduce the shared argument]

i) Assemble Dep:

Dep

Dep Individual (= IL/SL) Individual/Stage

a a √good

The label DEP is used here for the embedded adjectival predicate in order to disambiguate it from a resultative constituent. INDIVIDUAL/STAGE stands for the possible individual-level or stage-level specification of a predicate like good. As further details of the internal structure of the secondary predicate are irrelevant here, they will be left aside. ii) Select consider (more detailed structure not shown here) Vconsider iii) Dep & Vconsider: Merge Vconsider Vconsider Dep Dep Individual (= IL) Individual

a a √good

321

DPs in Adjectival Small Clauses in Romanian

iv) Merge Sit0, and shared argument . . . . .

the women

Sit0 Vconsider Vconsider Dep Dep Individual (= SL) Individual

a

a √good

v) Merge vCMPLX MULTIPLE AGREE OPERATION, responsible for THE INTEGRATION OF INDEPENDENT PREDICATES INTO A COMPLEX .......... vCMPLX [u CMPLX ] the women Value u CMPLX 1.Agree with DP

2. Initiate Multiple Agree on multiple predicates

Sit0

Vconsider

Vconsider Set 2 . . . . . Set 1 uγ uγ u# u# ....

dep DEP

Individual (= IL)

Individual

a √good a [uPred] [uγ] [u#] [uCase] . . . . .

322

Irimia

Note that the derivation above represents the situation in which the shared argument and the secondary predicate also show Case match. As already mentioned, in many languages (Finnish, Russian, etc., see examples 8 and 9 above) secondary predicates might carry dedicated Cases. These patterns require a somehow distinct implementation of the Agree operation, but as their analysis goes beyond the topic of this paper, the precise mechanics will be omitted here. Multiple Agree predicts uniform agreement with the shared object, as applied to the process of complex predicate formation. The mechanics introduced above derives both the wide scope readings and differential morphology on the shared DP in both Old and Modern Romanian. If these constructions are instances of complex predicates, their diachronic stability is straightforwardly explained. However, the complex predicate analysis remains agnostic to the diverging marking of pronominals. This aspect is briefly touched upon in section 7. 7

Pronouns and Differential Object Marking

Remember that a striking difference between Old Romanian and Modern Romanian is seen in the marking of the pronouns functioning as shared arguments in embedded adjectival contexts. Namely, if in Modern Romanian all personal pronouns must carry differential marking in these instances, as in example (11), repeated here as (69), in Old Romanian there appears to be optionality in that pronouns can either have the differential marking or not. Diachronic examples (5) and (6) are also repeated here under (70) and (71). In (70) the shared pronoun does not carry differential marking. (69) (Ne)- au considerat *(pe) noi inteligenți/inteligente. us= have.3.PL= considered DOM us smart.M.PL/smart.F.PL ‘They considered us intelligent.’ (70) Dumnezeu ispiti pre ei și află ei destoinici luiș. God tested DOM they and found them loyal.M. PL him ‘That God put them to test and found them loyal to him.’ (Coresi EV 260) (71) Ispitind pre el diavolul  . . .  află pre el nebiruit test.GER DOM he devil.the found DOM him invincible.SG.M ‘When the Devil put him to test, he found him invincible.’ (Coresi EV 520)

DPs in Adjectival Small Clauses in Romanian

323

Given this split, the question is the following: do pronominal morphological alternations indicate structural diachronic differences in the construction of small clauses? Or are the differences to be found somewhere else? Given the fact that non-pronominal DPs exhibit the same morphological marking restrictions diachronically (i.e, prohibition on bare indefinites, restriction to wide scope and specificity readings), structural differences at the clausal level are not plausible. It rather appears to be more probable that a shift in the setting of the differential marking took place, extending its coverage to pronouns. In order to better motivate this preliminary observation, a few more words are necessary regarding differential object marking. The typological literature following the pioneering work by Bossong (1991) has established that differential object marking is a strategy encoding the prominence of an object when compared to the subject. This generalization, as formulated is Aissen (2003) is spelled out in (72). (72) The higher in prominence a direct object, the more likely it is to be overtly case marked. Another very important observation typological studied have made is that the differential object marking strategy is not unitary. Human languages employ this overt morphological marking based on a variety of factors broadly related to prominence, among which are animacy and definiteness. Aissen (2003) further identified the following rankings that hold cross-linguistically with respect to the differential status of objects: (73) Animacy scale: Human > Animate > Inanimate (74) Definiteness scale: Personal pronoun > Proper name > Definite NP > Indefinite specific NP > Indefinite NP What these scales are supposed to capture is not only semantic/structural prominence but also its correspondence to morphology. Hence, as [+human] is situated higher on the scale of animacy, human DPs are statistically more probable to receive differential marking. The reasoning goes on a similar way for the definiteness scale; as personal pronouns are the highest, they are more likely to receive differential marking. Romanian, as opposed to many languages in which only one of the scales is used for differential object marking, exhibits a complex mix. On the animacy scale, DPs with the features human and animate can only receive differential

324

Irimia

marking. On the definiteness scale, all DPs with the puzzling exception of definites (and non-specific indefinite NPs) can take the marker pe. This is schematized in (75). (75) Differential marking in Romanian: a. Animacy scale: Human > Animate b. Definiteness scale: Personal pronoun > Proper name > Indefinite specific NP However, out of the two scales, the definiteness one is less stable. Examples from the same period indicate that some definites accepted differential marking, especially if they encoded unique entities (e.g., God): (76) Cine va căuta aceasta milostiv afla-va pre Domnul who will= seek this merciful find=will DOM God.DEF ‘And who will seek this, he will find God merciful.’ (Coresi EV 278) However, such examples are not statistically prominent, and appear to be marginal in Modern Romanian. This fluctuating status could probably indicate that the texts from Old Romanian witness the development of the second definiteness scale, complementary to the animacy one. This could explain why the pronouns can receive differential marking or not; in the latter case, they illustrate a stage in the history of Romanian when the definiteness scale had not been fully implemented, leaving the pronouns unaffected. 8

Conclusions and Further Issues

This paper investigated some previously unaddressed diachronic data with adjectival secondary predicates in Romanian, focusing on the most economical and empirically motivated strategy to derive restrictions to wide scope on shared DPs. The main conclusion is that typical intensional predicates with adjectival secondary predicates form a predicate complex syntactically, resulting in obligatory multiple agreement inside the complex. Preliminary investigation into the nature of differential marking in Romanian also supports a hypothesis according to which the alternating differential marking of pronouns (as opposed to non-pronominal DPs) is due to a shift in differential marking strategy, namely the introduction of a definiteness scale, complementary to the more canonical animacy scale.

DPs in Adjectival Small Clauses in Romanian

325

Corpus CC2 1581 Pușcariu, S. & A. Procopovici (eds). 1914. Diaconul Coresi, Cartea cu învățătură. Bucharest: Atelierele Grafice Socec. CPr 1566–1567 Bianu, I. (ed.). 1930. Coresi, Apostol. Texte de limbă din secolul XVI, IV, Lucrul apostolesc tipărit de diaconul Coresi la 1563. Ureche Onu, L. (ed.). 1967. Grigore Ureche, Letopisețul Țării Moldovei. Bucharest: Editura Științifică. References Aissen, Judith. 2003. Differential object marking: iconicity vs. economy. Natural Language and Linguistic Theory 21: 435–448. Andrews, Avery D. and Christopher Manning. 1999. Complex predicates and information spreading in LFG. CSLI Publications. Stanford California. Basilico, David. 2003. The topic of small clauses. Linguistic Inquiry 34(1): 1–35. Béjar, Susana & Milan Rezac. 2009. Cyclic agree. Linguistic Inquiry 40.1: 35–73. Bittner, Maria. 1994. Case, scope and binding. Boston: Kluwer. Boeckx, Cedric. 2001. Scope reconstruction and A-movement. Natural Language and Linguistic Theory 19: 503–548. ———. 2004. Long-distance agreement in Hindi: some theoretical implications. Studia Linguistica 58 (1): 23–36. Bossong, Georg. 1991. Differential object marking in Romance and beyond. In D. Wanner and D. Kibbee (eds.), New Analyses in Romance Linguistics, Selected Papers from the XVIII Linguistic Symposium on Romance Languages 1988,143–170. Amsterdam: John Benjamins. Chiriacescu, Sofiana & Klaus von Heusinger. 2009. Discourse prominence and pemarking in Romanian. International Journal of Pragmatics 2: 298–332. Chomsky, Noam. 1975. The logical structure of linguistic theory. New York: Plenum Press. ———. 1981. Lectures on government and binding. Dordrecht: Foris. ———. 1986. Knowledge of language: its nature, origin, and use. Santa Barbara, CA: Praeger. ———. 1993. A minimalist program for linguistic theory. In Hale, Kenneth & Samuel Jay Keyser (eds.), The view from building 20: essays in linguistics in honor of Sylvain Bromberger, 1–52. Cambridge, MA: MIT Press. ———. 1995. The minimalist program. Cambridge, MA: MIT Press. ———. 2001. Derivation by phase. In Kenstowicz, Michael (ed.), Ken Hale. A life in language, 1–53. Cambridge, MA: MIT Press.

326

Irimia

———. 2007. Approaching UG from below. In Sauerland, Uli & M. Gaertner (eds.), Interfaces + Recursion = Language? Chomsky’s minimalism and the view from syntaxsemantics, 1–30. Berlin/New York: Mouton de Gruyter. ———. 2008. On Phases. In Freidin, Robert et al. (eds.), Foundational Issues in Linguistic Theory, 133–166. Cambridge, MA: MIT Press. Cornilescu Alexandra & Carmen Dobrovie-Sorin. 2008. Clitic doubling, complex heads and interarboreal operations. In D. Kalluli & L. Tasmowski (eds.), Clitic Doubling in the Balkan Languages, 289–320. Amsterdam: John Benjamins. Dalmi, Grete. 2005. The role of agreement in non-finite predication. Amsterdam: John Benjamins. De Swart, Peter. 2007. Cross-linguistic variation in object marking. Doctoral dissertation. University of Njimegen. Diesing, Molly. 1992. Indefinites. Cambridge, MA: MIT Press. Dobrovie-Sorin, Carmen. 1990. Clitic doubling, wh-movement, and quantification in Romanian. Linguistic Inquiry 21: 351–398. ———. 1994. The syntax of Romanian. Berlin. Mouton de Gruyter. Enç, Mürvet. 1991. The semantics of specificity. Linguistic Inquiry 22: 1–25. Farkas, Donka & Klaus von Heusinger. 2003. Stability of reference and object marking in Romanian. Unpublished manuscript, UC Santa Cruz / Universität Stuttgart. Frampton, John &Sam Gutman. 2006. How sentences grow in the mind. Agreement and selection in efficient minimalist syntax. In Cedric Boeckx (ed.), Agreement systems, 123–157. Amsterdam: John Benjamins. Heim, Irene. 1982. The semantics of definite and indefinite noun phrases. Doctoral dissertation. University of Massachusetts. von Heusinger, Klaus. 2002. Specificity and definiteness in sentence and discourse structure. Journal of Semantics 19: 245–274. von Heusinger, Klaus & Edgar Onea. 2009. Triggering and blocking effects in the diachronic development of DOM in Romanian. Probus 20: 67–110. Heycock, Caroline. 1994. Layers of predication: the non-lexical syntax of clauses. New York: Garland. Hill, Virginia. 2013. The direct object marker in Romanian: a historical perspective. Australian Journal of Linguistics 33: 140–151. de Houp, Helen. 1996. Case configuration and noun interpretation. Garland. New York. Hiraiwa, Ken. 2004. Dimensions of agreement. Doctoral dissertation. MIT. Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30: 69–96. Irimia, Monica Alexandrina. 2011. Secondary predicates. Doctoral dissertation. University of Toronto. Kachru, Yamuna. 2006. Hindi. Amsterdam/Philadelphia. John Benjamins.

DPs in Adjectival Small Clauses in Romanian

327

Klein, Udo. 2007. Clitic doubling and differential object marking in Romanian. Unpublished manuscript, Universität Stuttgart. Koul, Omar N. 2008. Modern Hindi grammar. Springfield, VA: Dunwoordy Press. Lasnik, Howard. 1999. Chains of arguments. In Epstein, Samuel & Horstein, Norbert (eds.), Working Minimalism, 189–215. Cambridge, MA: MIT Press. Leonetti, Manuel. 2007. Clitics do not encode specificity. In Georg A. Kaiser & Manuel Leonetti (eds.), Proceedings of the Workshop on Definiteness, Specificity and Animacy in Ibero-Romance Languages, 111–139. Universität Konstanz: Fachbereich Sprachwissenschaft (Arbeitspapier 122). López, Luis. 2012. Indefinite objects. Cambridge, MA: MIT Press. Matushansky, Ora. 2002. Tipping the scales: The syntax of scalarity in the complement of seem. Syntax 5: 219–276. Milsark, Gary. 1974. Existential sentences in English. Doctoral dissertation. MIT. May, Robert. 1985. Logical form. Its structure and derivation. Cambridge, MA: MIT Press. Moulton, Keir. 2013. Raising from the dead. Linguistic Inquiry 44 (1): 157–167. Öztürk, Balkiz. 2005. Case, referentiality and phrase structure. Amsterdam: John Benjamins. Postal, Paul. 1974. On raising: an inquiry into one rule of English and its theoretical implications. Cambridge, Massachusetts. MIT Press. Pylkkänen, Liina. 2008. Introducing arguments. Cambridge, Massachusetts. MIT Press. Reuland, Eric & Alice ter Meulen (eds.). 1987. The representation of (in)definiteness. Cambridge, MA: MIT Press. Rodríguez-Mondoñedo, Miguel. 2007. The syntax of objects: Agree and differential object marking. Doctoral dissertation. University of Connecticut. Rothstein, Susan. 1985. The syntactic forms of predication. Bloomington: Indiana UniversityPress. Sigurðsson, Halldór A. 2006. Agree in syntax, agreement in signs. In Boeckx, Cedric (ed.), Agreement systems, 201–239. Amsterdam: John Benjamins. Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Dixon, R.M.W. (ed.), Grammatical categories in Australian languages, 112–171. Canberra: Australian National University Press. Sportiche, Dominique. 2005. Division of labour between Merge and Move: strict locality of selection and apparent reconstruction paradoxes. LingBuzz/000163. Stowell, Timothy. 1981. Origins of phrase structure. Doctoral dissertation. MIT. ———. 1991. Small clause restructuring. In Freidin, Robert (ed.), Principles and parameters in comparative grammar, 182–218. Cambridge, MA: MIT Press. ———. 1995. Remarks on clause structure. In Cardinaletti, Ana & Maria Teresa Guasti. (eds.). 1995. Syntax and semantics 28. Small clauses. London: Academic Press. Torrego, Esther. 1998. The dependencies of objects. Cambridge, MA: MIT Press.

328

Irimia

Ura, Hiroyuki. 1996. Multiple feature checking: a theory of grammatical function splitting. Doctoral dissertation. MIT. Webelhuth, Gert. 1995. X-bar theory and Case theory. In Webelbuth, Gert (ed.). Government and binding theory and the minimalist program: principles and parameters in syntactic theory. Oxford, England/Cambridge, Massachusetts. Blackwell. Williams, Edwin. 1983. Against small clauses. Linguistic Inquiry 14 (2): 287–308.

Restrictive and Appositive Relatives Anca Sevcenco This paper focuses on restrictive and appositive relatives in Old Romanian (mid17th to mid18th century), which are compared to equivalent structures in Modern Romanian. The paper identifies and formalizes the patterns that underly restrictive and appositive relativization in Romanian, arguing that (i) a uniform account based on HEAD raising of the nominal relative head is best suited to account for restrictive relatives; and (ii) appositive relatives are better analyzed as coordinated conjuncts with a specifying function. These patterns provide an adequate account for diachronic changes, mainly, the loss of inflection on the relative pronoun care ‘which’, arising from a reanalysis of this item as an invariable relative complementizer with head status (versus the phrasal status, in its inflected state).

Introduction This paper starts with the syntactic investigation of Old Romanian restrictive and appositive relative clauses, and follows with a diachronic discussion of the results. The theoretical tools are provided by the Minimalist Program (Chomsky 1995, 2008) and the cartographic approach (Rizzi (1997). The data are strictly limited to clauses with finite verbs, without involving non-finite relatives. The data that I draw upon come from three historical chronicles: (i) The Chronicle of Moldavia by Grigore Ureche, from mid17th century; (ii) The Chronicle of Moldavia. About the Moldovans’ Genealogy by Miron Costin, from the end of the 17th century and (iii) The Chronicle of Moldavia and Some Chosen Words by Ion Neculce, from mid18th century. I argue that restrictive relativization in Old Romanian relies on two main patterns (and some variation within each of them) that I will briefly list now, leaving the precise details regarding their derivation for later discussion in this paper. The first pattern comprises wh-relatives, introduced by an inflected or an invariable relative pronoun, which have a gap at the relativization site and involve displacement of the relative phrase to the left periphery of the clause. It is possible for direct and indirect object wh-relatives with a gap to also include a clitic determiner that corefers with the relative head and remains stranded after displacement of the relative phrase has taken place. The second pattern

© koninklijke brill nv, leiden, ���5 | doi ��.��63/9789004292550_011

330

sevcenco

features restrictive relatives introduced by relative complementizers, which also have a gap at the first merge position and are further derived by internal merge of the nominal relative head with the Focus Phrase in the left periphery. As a slight variation within the second pattern, there are also direct and indirect object restrictive relatives introduced by a complementizer, with internal merge of the nominal head with Focus phrase and stranding of a coreferring clitic determiner at the relativization site. This concise picture hints at the fact that restrictive relativization in Old Romanian can be uniformly explained in terms of derivations that result from displacement. Appositive relativization in Old Romanian also involves two patterns. The first pattern features wh-relatives introduced either by an inflected or an invariable relative pronoun, whereas the second pattern brings into the picture appositive relatives introduced by a complementizer. I will show that both options lend themselves to an account based on the proposal that appositive relatives are appositions to their antecedent with which they are linked up by means of coordination (i.e. the asyndetic coordination approach put forth in de Vries 2002, 2006a a.o.). The paper is organized as follows. Section 2 presents the properties of restrictive relative clauses in Old Romanian and outlines the proposal regarding the syntactic derivation of the two patterns that I have briefly presented above. The proposal is based on the HEAD raising analysis, recently put forth in Donati & Cecchetto (2011). Section 3 focuses on the properties of appositive relatives in Old Romanian and spells out the details of an analysis in terms of coordination with the antecedent (following de Vries 2006a, a.o.). The last section draws the conclusions and provides comments about syntactic changes that occurred in the transition from earlier stages of Romanian to Modern Romanian, with a view to highlighting aspects related to the evolution of inflected relative pronouns into (uninflected relative) complementizers. 1

Old Romanian Restrictive Relatives

This section gives an overview of the properties that characterize restrictive relatives in Old Romanian, which are all of the externally headed kind. I identify the available types (i.e. wh-relatives and relatives introduced by complementizers) and discuss their features. I will touch upon the following issues: properties of the relative connectors, the presence of a gap at the relativization site and the occurrence of resumptive Accusative clitics in direct object restrictive relatives. I do not discuss indirect object relatives with Dative coreferrring clitics.

331

Restrictive And Appositive Relatives

1.1 Properties of Restrictive Relatives in Old Romanian 1.1.1 Care Relatives Old Romanian wh-relatives are introduced by the relative pronoun care ‘who, which’. This pronoun occurs as an inflected form that carries number and gender features; it has distinct Case endings; and it displays the enclitic definite article –l/a (Frîncu 1997; Stan 2013). Table 1 sums up the morphology of Old Romanian relative pronouns. Examples (1) illustrate the number and gender agreement that holds between the inflected pronoun and the relative nominal head. table 1  The morphology of the inflected relative pronoun care. Case

Nom/Acc Gen / Dat

Singular Masc

Fem

Plural Masc

carele căruia

carea căriia

carii cărora

Fem

carele

(1) a. plumbul cu careli au fost lead.DEF.SG.M with which.SG.M have.AUX.3PL been acoperită mănăstirea Putna covered monastery Putna ‘the lead with which Putna monastery was covered’ (IN: 115) b. oamenii carii altă hrană nu știe, fără oștenia people.PL.M who.PL.M other food not know except military.service ‘people who know no other reward than the military service’ (MC: 103) c. osteneala cu carea să diprinsése tiredness.DEF.SG.F with which.SG.F REFL = accustomed ‘the tiredness with which he had got accustomed’ (GU: 97) d. cetațile carile le luase fortresses.DED.PL.F which.PL.F them = took ‘the fortresses which the heathens took’ (GU: 92)

păgânii heathens.DEF

Stan (2013) notes that the definite article is part of the morphological make-up of pronominal forms and convincingly argues that the function of this article is

332

sevcenco

to express ‘gender and number distinctions’ (Stan 2013: 3). In other words, this definite article plays the part of an inflectional marker. The relative pronoun also has the uninflected, invariable form care, which occurrs with a much more limited frequency than its variable counterpart, and is employed mostly in appositive relatives. Below is an illustration of uninflected care in a restrictive relative. (2) a. oștile care lăsase la Leva armies.DEF.PL.F which left at Leva ‘the armies which he left at Leva’ (IN: 124) b. boierii care era zălog atunce boyars.DEF.PL.M who were hostage then ‘the boyars who were hostages then’ (IN: 157) All care relatives introduced in (1) and (2) behave as wh-relatives insofar as they leave an empty site (a gap) at the relativization site, as illustrated in (3). (3) a. istoricii leșești carii_ historians.DEF.PL.M Polish.PL.M who.DEF.PL.M au scris have.AUX.3PL written ‘the Polish historians who wrote’ (MC: 7) b. oștile care lăsase_ la Leva armies.DEF.PL which left at Leva ‘the armies which (he) left at Leva’ (IN: 125) In very few cases, direct object relatives have an Accusative clitic that corefers with the head of the structure, agreeing with it in number and gender featues, as shown in (4). This clitic is generally preverbal, hence, it is not at the relativization gap, but higher. I take the clitic to represent an instantiation of resumptive pronoun, and mention that this option was not readily available with Old Romanian restrictive relatives. (4) cetățile carile le luase păgânii fortresses.DEF.PL.F which.PL.F them.ACC.3PL.F = took heathens.DEF ‘the fortresses which the heathens took’ (GU: 92)

Restrictive And Appositive Relatives

333

In prepositional object wh-relatives, the relative pronoun always pied-pipes the preposition: (5) drumul pe care venisă way.DEF on which came ‘the way on which they came’ (MC: 93) 1.1.2 Relative Complementizers Relatives introduced by the complementizers ce and de (which I gloss as ‘that’) are extensively used in Old Romanian restrictive relatives, to a much larger extent than their relative pronoun counterparts. The complementizer status of these items can be determined by two properties: they do not have inflected forms and they never show up with pied-piped prepositions. Ce is compatible with relativization from all syntactic positions: subject (6), direct object (7), indirect object (8) and prepositional object (9). Most of the relatives introduced by it have a gap and no resumptive pronoun, but, rarely, it is also possible for the gap to be accompanied by a resumptive clitic (see examples (8) and (10)). (6) o samă de boieri ce era de casa Ducăi-vodă a group of boyars that were of house.DEF Ducăi-king.GEN ‘a group of boyars who belonged to king Duca’s house’ (IN: 129) (7) supărarea ce făcè boierilor trouble.DEF that made boyars.DAT ‘the trouble that (he) caused to the boyars’ (IN: 122) (8) apa ce-i dzice Moldoa water.DEF that = it.DAT.3SG say Moldoa ‘the water that they call Moldova’ (GU: 66) (9) văzându-să în boala grea ce zăcuse seeing.GER = REFL in disease serious that lay ‘seeing himself bedridden by a serious disease’ (GU: 193) (10) tălhari ce-i adusă într-ajutoriu thugs.PL.M that=them.ACC.3PL.M brought in help ‘thugs that they brought as help’ (GU: 22)

334

sevcenco

De also introduces a wide range of relatives: subject (11), direct object (12) and indirect object (13). It is also possible for restrictive relatives introduced by de to have a gap and a resumptive clitic (as shown in (13)). (11) așa scrie și cronicariul lor de zice că thus writes also historian.DEF their.POSS.3PL that says that ‘their historian also writes thus, saying that . . .’ (GU: 86) (12) acest lucru de nu this thing that not l-au scris it.ACC.3SG.M= have.AUX.3PL= written ‘this thing that they did not write’ (IN: 109) (13) acel loc de-i dzic Vadul Jorăi that place that=it.DAT.3SG say Vadul Jorăi ‘that place to which they refer as Vadul Jorăi’ (GU: 102) It is important to mention that neither of these two complementizers are used exclusively as relative complementizers. De introduces purpose and result clauses, as well as complements, in addition to being a conditional complementizer and a clause coordinator. Also, ce has more than one use. It is an interrogative pronoun, it functions as an adversative coordinating conjunction and, last but not least, it introduces free relatives. In this latter case, ce takes on the role of relative determiner (14). (14)

Cu ce măsură măsuri, with what measure measure.2PRES măsura-ți-să-va measure to.you = DAT se = REFL will.AUX.FUT ‘whatever measure you use will also be used for you’ (MC: 20)

The last property I bring into the discussion concerns exclusively direct object restrictive relatives, where the diachronic change is outstanding: In Modern (but not in Old) Romanian, these relatives feature an obligatory Accusative clitic pronominal that corefers with the head of the relative and agrees with it in number and gender, as shown in (15).

335

Restrictive And Appositive Relatives

(15) băiatul pe care boy.DEF.SG.M pe who.ACC ‘the boy whom I draw’

îl desenez him.ACC.3SG.M=draw

In the Old Romanian equivalents of (15), a coreferring Accusative clitic is used only optionally. In fact, these clitics are absent in most cases (even if not altogether so). Let us now take stock of the relativization options that were available in Old Romanian. I have identified two basic patterns: (i) wh-relatives with a gap at the relativization site and (ii) complementizer relatives with a gap. Both patterns show slight variations because it is also possible for a resumptive pronominal clitic to occur in direct object relatives and to remain stranded at the relativization site. Examples with resumptive forms are few and far between, though. These patterns differ from Modern Romanian in two important respects. First, the inflected pronoun that introduces wh-relatives in Old Romanian has lost its number and gender inflection in Modern Romanian, and has become invariable when it occurs in the Nominative / Accusative Case. However, the number and gender inflections have survived when the pronoun takes the Genitive / Dative forms. Second, direct object relatives with gaps and no resumptive clitic are no longer grammatical (irrespective of whether they are of the wh or the complementizer kind). In other words, the resumptive Accusative clitic has become mandatory. 1.2 The Syntax of Restrictive Relatives in Old Romanian This section focuses on the structure of restrictive relatives. It will be argued that two patterns are at play in these derivations, according to whether direct merge or movement is involved in the checking operation. 1.2.1 Wh-relatives: Theoretical Framework The first theoretical question is whether the wh-relative pronoun is base generated in or moved to the clause initial position. I mentioned in the previous subsection that wh-restrictive relatives have a gap at the relativization site. It is generally acknowledged that the presence of a gap is an indication that displacement has taken place. Therefore I will use the presence of a gap as a criterion for identifying a movement derivation for wh-relatives, and support it with additional tests.

336

sevcenco

In Modern Romanian, when relatives are introduced by the (invariable) relative pronoun care ‘who, which’, the head of the relative cannot be separated from its argumental position, at the relativization site, by a strong DP island (16). (16) *Băiatul pe care ți-am arătat o fată boy.DEF pe which to.you = have.AUX.1SG shown a girl [care [îl simpatizează]]. who him = ACC.3SG.M likes *‘The boy whom I showed you a girl who likes him’. (Grosu 1994: 234) Such (ungrammatical) examples are not attested in the corpora investigated and, given their absence, I assume they were impossible in Old Romanian as well. The situation becomes a bit more complex in those cases in which a resumptive pronominal clitic occurs, because there is a line of thinking that associates resumptives with a base generation derivation (Shlonsky 1992; Adger & Ramchand 2005, a.o.). Note, however, that relatives with resumptive clitics are also compatible with a movement derivation (Boeckx 2003; Donati & Cecchetto 2011). I assume this second view and propose that wh-relatives are the result of a syntactic derivation that involves raising of the relative phrase to a position in the left periphery of the subordinate clause. The second theoretical question concerns the derivational mechanism: how do we formalize the relation between the head and the relative clause? The raising analysis for the relative clause has been advocated for in a huge body of work (Vergnaud 1974; Kayne 1994; Bianchi 1999, 2000; Bhatt 2002; Donati & Cecchetto 2011 a.o.). I have chosen the version presented in Donati & Cecchetto (2011) as a means of formalizing the syntax of Old Romanian whrelatives because it provides an approach that deals with the issues of both theoretical and empirical nature brought up by the analyses in Kayne (1994) and Bianchi (1999). In what follows, I will briefly present the main tenets of the raising analysis and outline the theoretical details of the Donati & Cecchetto implementation. The raising analysis is associated with a complementation structure for relatives (versus adjunct in Demirdache 1991). More precisely, the proposal is that relative clauses are the complement of an external determiner (D). Relative pronouns spell out the D head of the nominal projection that includes the head of the relative. This constituent is usually referred to as the relative DP.

Restrictive And Appositive Relatives

337

The framework put forth in Donati & Cecchetto (2011) rests on a couple of tenets about phrase structure which I will briefly go over in what follows. Upon merger of two syntactic objects, one of them gives the label of the new projection. This label “is bound to be a subset of the features of the items that are merged” (Donati & Cecchetto 2011: 521). For instance, when a transitive verb merges with its direct object, the V head provides the label of the new projection, which is dubbed VP. Cecchetto & Donati (2010) propose a Probing Algorithm that accounts for the way the labeling of newly formed projections is determined: (17) Probing Algorithm The label of a syntactic object {α, β} is the feature(s) that act(s) as a probe of the merging operation creating {α, β}. (Cecchetto & Donati 2010: 245) It is further assumed (following Chomsky 2008) that lexical items have edge features, which are, actually, the categorial features of the respective words. Upon merger in the syntactic structure, a lexical item becomes a probe because it has an edge feature. The categorial feature on the item will give the label of the new projection. Let us see how the derivation of a wh-relative proceeds in Donati & Cecchetto (2011), starting from an example whose bracketed representation they provide (18b is their example 21b): (18) a. the book which John saw b. [DP the [NP [[NP book] [CP [DP which book] John saw which book]]]] The relative DP, formed by the relative D and its nominal complement, externally merges in argumental position inside the relative clause. It undergoes subsequent movement to a position in the left periphery, Spec,CP. The head noun, book, is a lexical item and has an edge feature, i.e. its categorial N feature. Upon internal merger with the C projection, it turns into a probe and it relabels the projection, which becomes thus an NP projection. The external D further merges externally with this NP. Donati & Cecchetto leave for further research the exact position in which the raised head ends up; they simply conjecture that it must be located somewhere in the fine-grained left periphery of the CP field. I will go back to this detail later on. For now, I give the tree diagram of a wh-relative and henceforth place between angled brackets the constituent that is displaced (Donati & Cecchetto’s example (22b)):

338

sevcenco

(19) wh-relative wh-relative DP D

NP

the N

CP

book which John saw The proposal for the derivation in (19) has been dubbed by its authors the HEAD-raising analysis. This is so because the displaced element is always a lexical item, in other words the nominal head of the relative. When a relative clause has a phrasal head, Donati & Cecchetto (2011) have to assume that whatever material modifies the nominal head enters the derivation by late merger. 1.2.2 Applying the Theory to the Old Romanian Data Two issues need clarification before we proceed: the internal structure of the Old Romanian relative phrase and the landing site of the relative DP. As mentioned in section 1.1, Old Romanian relative pronouns come in two flavors, with inflected and uninflected forms. I propose that inflected care is a noun that undergoes head-to-head movement to D, which contains the enclitic definite article; this is shown in (20) for the masculine plural form carii. On the other hand, uninflected care (with the phonological variant cari), has been reanalyzed in D, and thus replaces the definite article; it takes as complement the NP that contains the relative clause, as in (21). (20) [DP [D cari-i [NP cari]]] (21) [DP [D cari [NP (e)]]] The difference between (20) and (21) indicates a trend of grammaticalization, by which the lexical head in (21) is reanalyzed as the functional head in (22), in the same structure. Irrespective of whether care has the internal structure in (20) or (21), we have to determine its landing site within the clause. For that purpose, I consider the position of the pronoun with respect to: the inflected auxiliary in T, clitics that use the inflected auxiliary as a host, and the position of the pronoun with respect to focalized and topicalized constituents. In order to achieve this,

Restrictive And Appositive Relatives

339

I resort to the fine-grained structure of the left periphery in Rizzi (1997) presented in (22). (22) [ForceP [TopP [FocP [TopP [FinP [TP ]]]]]] Within a hierarchy as in (22), the Old Romanian data indicate that the inflected relative pronoun precedes the inflected auxiliary in T and the clitics that adjoin to this verbal form (23 and 24) (see Alboiu, Hill & Sitaridou 2014 for tests identifying the T field as the host of clitic auxiliaries and clitic pronouns in Old Romanian); it also precedes focalized (25) and topicalized constituents (26). (23) istoricii leșești carii historians.DEF.PL.M Polish.PL.M who.PL.M au scris have.AUX.3PL= written ‘the Polish historians who wrote’ (MC: 7) (24) cetațile carile le luase fortresses.DEF.PL.F which.PL.F them = took ‘the fortresses which he took’ (GU: 92) (25) mulți carii nici iniceri nu era many who.PL.M neither soldiers not were ‘many who were not even soldiers’ (MC: 62) (26) boieri . . .  carii mai sus boyars.PL.M whom.PL.M more above s-au scris REFL = have.AUX.3PL= written ‘boyars about whom it was written above’ (IN: 218) Uninflected care may also come before the inflected lexical verb that raises to T and clitics that take T as a host (27 and 28), focalized phrases (29) and topicalized constituents (30). (27) unu diiac care era . . .  a scribe.SG.M who was . . . ‘a scribe who was . . .’ (MC: 33)

340

sevcenco

(28) treaba care o luase task.DEF.SG.F which it=took ‘the task which he took’ (MC: 115) (29) toate care la locul său all.PL.F which at place its.POSS.REFL le-am tocmit them = have.AUX.1SG done ‘all which we did one by one’ (GU: 104) (30) celui târgșorŭ la care puțintei haiduci era that town at which few outlaws were de paza cetății of guard city.DEF.GEN ‘that town in which only a few outlaws guarded the fort’ (MC: 83) Therefore, irrespective of whether the internal structure of the wh-relative is (20) or (21), the location in the clause is the same, namely, at the highest level of the articulated CP field (above TopP), which means in ForceP. Now everything is in place for formalizing the syntactic derivation of Old Romanian wh-relatives. Let us start with a subject relative introduced by the inflected relative pronoun—(31)—and give its tree diagram in (32). (31) împărații carii au urmatu emperors.DEF.PL.M who.PL.M have.AUX.3PL followed (de aicea înainte) from then on ‘the emperors that followed from then on’ (GU: 130) Given that the inflected relative pronoun projects a DP, I suggest that this DP externally merges in argumental subject position and then undergoes displacement to Spec,ForceP (the highest projection in (22)). The nominal antecedent of the relative externally merges with ForceP (see Aoun & Li 2003 for the proposal that the relative head merges directly in the left periphery and only the relative pronoun undergoes displacement). Due to its edge categorial feature N that turns it into a probe, the head noun relabels the newly formed projection as N.

341

Restrictive And Appositive Relatives

(32) subject wh-relative with inflected relative pronoun DP

D

NP

împăraţii N

ForceP

DP

...

carii ...

TP

au urmatu Remember that direct object wh-relatives occasionally feature an Accusative clitic that agrees in φ features with the inflected pronoun and the nominal head, example (1d) above repeated here as (33), for the reader’s convenience. (33) cetațile carile le fortresses.DEF.PL.F which.PL.F them.ACC.PL.F= ‘the fortresses which (he) took’ (GU: 92)

luase took

I suggest that the inflected pronoun fills in the specifier position of a big DP whose head is the clitic determiner and whose complement position is occupied by a null pronominal (in line with the big DP structure proposed in Torrego 1986; Uriagereka 1995, a.o.). Displacement of the inflected pronoun strands the clitic, which, in its turn, subsequently cliticizes onto the finite T. Here I depart from Donati & Cecchetto (2011) who propose that only relatives introduced by complementizers strand a resumptive form at the relativization site. Their observation is based on the idea that substandard Romance varieties (which have complementizer relatives) have direct object relatives with resumptive clitics. On the basis of the data presented in (9)/(10) and (12)/(13), I propose that Old Romanian is one language in which the stranded determiner may become lexically realized in both wh and complementizer direct object relatives. The distribution of the Accusative resumptive clitic in Old Romanian relatives is not subject to the same conditions as that of the Accusative clitic in modern Romance (substandard) relatives. The representation for (33) is given in (34) below. It closely follows (31) in that the phrasal r­ elative pronoun raises

342

sevcenco

to Spec,ForceP and the head noun enters the derivation by external merge with Force. Relabeling as above applies. (34) wh-relative with a resumptive Accusative clitic DP D cetăţile

NP N

ForceP DP

...

carile . . .

TP

T CL

VP T

le lăsase

V

DP

DP

V' D

pro

I still have to spell out the details regarding the syntactic derivation of relatives introduced by the uninflected pronoun such as (2a), repeated for convenience as (35): (35) oștile care lăsase la Leva armies.DEF.PL.F which left at Leva ‘the armies which he left at Leva’ (IN: 124) Example (35) differs from (33) in that the relative DP [DP care oști] ‘which armies’ merges in argumental position. Subsequently, the entire DP raises to Spec,ForceP and the head noun merges with the Force projection and relabels the newly formed projection in the already known fashion. The derivation is illustrated in (36).

Restrictive And Appositive Relatives

343

(36) wh-relative with uninflected relative pronoun DP D

NP

oștile

N

ForceP

DP

... ...

TP

lăsase la Leva A summary of what I have proposed so far: Wh-relatives with phrasal care involve raising of phrasal care to the left periphery of the clause, followed by the external merge of the head noun to the Force node and by the relabeling of the newly formed projection as an NP. Wh-relatives with determiner care are different. The constituent formed by the relative determiner and the head noun raises to the left periphery, after which the head noun internally merges with Force and relabeling takes place. The difference thus resides in the fact that the noun head merges externally (in wh-relatives with phrasal care) or internally (in wh-relatives with determiner care). Ultimately, I do not think that this is a relevant matter since whenever a lexical items merges (no matter by what sort of merge), it becomes a probe because it has an edge feature and thus can label the new projection. As Donati & Cecchetto (2011) remark, the Probing Algorithm applies to both internal and external merge. 1.2.3 Complementizer Relatives This section looks into the syntactic derivation of the kind of restrictive relatives most frequently encoutered in Old Romanian, i.e. the relatives introduced by the complementizers ce and de. I will use the HEAD raising analysis from Donati & Cecchetto as a starting point for my account. I give in (37) the derivation of English that-relatives as an illustration of this pattern.

344

sevcenco

(37) That-relative (Donati and Cecchetto example 28b) example 28b) That-relative (Donati2011: and530, Cecchetto 2011:530, DP

D

P

the N

CP

book that you saw [D ] The relative phrase in this derivation is a DP headed by a null non-wh-­ determiner, which remains stranded at the relativization site. The nominal head merges with the CP and relabels the newly formed projection as a NP. The external D selects the relative clause as its complement. In what follows, I will briefly mention two arguments in favor of raising inside the relative in Old Romanian. Just as with wh-relatives, I take the presence of the gap at the relativization site as a clue about the application of displacement. Displacement is also detectable in terms of the presence of reconstruction effects. In Modern Romanian relatives, displacement from the relativization site can be determined by means of the standard battery of tests that involve reconstruction for idiom interpretation, binding and scope (Sevcenco in prep.). The Old Romanian corpora I have studied have instances of reconstruction for possessive reflexive anaphor binding. Let us consider (38). (38) datoriia sa ce avè spre împărăție debt his.POSS.REFL.3SG that had towards government ‘the debt that he had to the government’ (MC: 18) The possessive reflexive sa ‘his’ is subject-oriented. It needs a subject binder and the null subject inside the relative clause binds it. For this binding relation to obtain, the head of the relative, with the embedded reflexive, must be within the c-command domain of the subject and this is possible only if reconstruction to the relativization site occurs. Let us now consider the position of the relative complementizer ce in the left periphery of the clause. The data indicates that the complementizer may precede: inflected auxiliaries (39), finite verbs that raise to T in Old Romanian (Alboiu, Hill & Sitaridou 2014), clitics that adjoin to these auxiliaries (40) and the negative operator nu ‘not’ that heads NegP (41). (39) niște sași ce au fost olari some Saxons that have.AUX.3PL= been pot.makers ‘some Saxons that were pot makers’ (GU: 28)

345

Restrictive And Appositive Relatives

(40) cel ce m-au that.one that me = have.AUX.3PL= ‘the one who educated me’ (IN: 122)

învățat carte taught book

(41) celora ce nu ascultă to.those that not listen ‘to those that do not listen’ (GU: 72) Given these word order details, I submit that the relative complementizer occupies a position in the CP field. In principle, it could be either in the Foc or in the Fin head. Since ce does not co-occur with contrastive focus, I propose the complementizer occupies the Foc head. The Fin option is excluded because there are no contexts in which constituents that bear contrastive focus precede the complementizer. The relative phrase is a DP, with a null D head that takes as complement the nominal head of the relative (to illustrate, conside the relative phrase in (42) below: [DP [D [NP vitejia]]]). Upon merger of the noun head with Force, the determiner remains stranded at the relativization site. Example (42) has the derivation in (43). (42) (pentru) vitejia ce avea (for) fearlessness.DEF.SG.F that had ‘for the fearlessness that he had’ (GU: 137) (43) complementizer relativerelative with gap complementizer with gap DP D

NP

vitejia N

ForceP

. . . FocP ...

Foc ce . . . T

TP VP

ave [D ]

346

sevcenco

Remember from section 1.1 that complementizer relatives occasionally feature a resumptive clitic that corefers with the nominal relative head, as in (44). (44) robilor ce îi adusese slaves.DEF.GEN.PL.M that them.ACC.3PL.M= brought din Țara Leșească from country Polish ‘to the slaves that he had brought from the Polish country’ (GU: 139) I propose that the clitic in (44) represents the overt realization of the determiner that remains stranded in (43). I take the clitic to head a big DP projection whose specifier position is occupied by the nominal head of the relative, as detailed in the representation in (45). (45) complementizer relative with gap and resumptive clitic DP D

NP

robilor N

...

. . . FocP Foc

...

ce

. . . TP VP

T CL

T

îi lăsase

Spec

V V

Spec

DP D

D pro

The complementizer de has the same properties as ce. It may precede inflected auxiliaries, finite verbs raised to T, the clitics adjoined to T and the nega-

Restrictive And Appositive Relatives

347

tive head nu, as summed up in (46). I suggest it also occupies the Foc head.1 Relatives introduced by de have the same syntactic derivation as those with complementizer ce. (45) au acoperit acest lucru have.AUX.3PL covered this thing.SG.M de nu l-au scris that not it=have.AUX.3PL= written ‘he hid this thing that he did’t not write (about)’ (IN: 109) To sum up, in this section I discussed the derivation of direct object relatives with and without an Accusative coreferring clitic. In those relatives that don’t have the clitic, the relative DP is made up by a null determiner and the head noun. The head noun merges internally in the left periphery and strands the empty D, which I cliticizes onto finite T. In the direct object relatives with the clitic, the head noun occupies the specifier position of a big DP, whose head is the clitic. After internal merge of the head noun in the left periphery, the stranded D, realized as a clitic this time, cliticizes onto finite T. 1.2.4 Theoretical Spill Off The analysis of Old Romanian relatives in terms of Donati & Checchetto’s (2011) framework comes with a nice bonus. In earlier versions of the raising analysis, the head noun and the relative clause formed a constituent to the exclusion of the external D (Kayne 1994). The HEAD raising proposal fixes this unwelcome constituency problem because the external D selects the relative nominal projection, so the relative CP remains a constituent of its own. This detail is quite useful. Even if splits between the relative DP and the relative clause are outside the scope of this paper, I must mention that extraposition (of both restrictives and appositives), as shown in (47), and head noun topicalization, see (48), occured quite frequently in Old Romanian. The constituency proposed in Donati & Cecchetto (2011) theoretically allows for the existence of these phenomena, although they do not exploit it.

1  The placement of relative de in Foc instead of Fin also accounts for the fact that the indicative clause remains finite. That is, Hill & Alboiu (2015) show that whenever de merges in Fin in Old Romanian, the clause becomes non-finite and has the same properties as an infinitival, even if the embedded verb comes in indicative inflection.

348

sevcenco

(46) a. Chiematu-o-au unii și Flachia, called =it= have.AUX.3PL some also Wallachia ce scriu letopisețile latinești that write.3 chronicles Latin ‘Some who write Latin chronicles also called it Wallachia’ (GU: 66) b. Le-au ieșit înainte turcilor din them = have.AUX.3PL = come before Turks.DEF from sus de Vaslui, de la Podul Înalt, above of Vaslui from Podul Înalt, pre carii i-au biruit Ștefan vodă, DOM who.PL.M them = have.AUX.3PL defeated Ștefan king, nu așa cu vitejia, cum not thus with bravery, but cu meșterșugul. with cunning. ‘They came from up above Vaslui, from Podul Înalt, before the Turks, whom King Ștefan defeated, not with bravery, but with cunning.’ (GU: 100) (47) Amândoao cetățile cu multă moarte și peire both fortresses.PL.F with much death and loss de ai săi le dobândi, of his.kind them=got [carile întărindu-le cu bucate și slujitori] which stregnthening.GER = them with food and servants ‘With much death and loss of his men, he got both fortresses, which he strengthened with food and servants’ (GU: 92) 1.3 Section Conclusions The present section looked into the syntax of externally headed restrictive relatives in Old Romanian and ended up with the following results: (i) Descriptively, Old Romanian has wh-relatives and relatives introduced by the complementizers ce and de, both to be translated as ‘that’. The complementizers are widely spread in Old Romanian relatives (but have been lost in Modern Romanian) and occur in other clause types. (ii) Both types of relatives display a gap at the relativization site and the occasional occurrence of a resumptive pronominal form. This indicates derivation through displacement, further confirmed by recon-

Restrictive And Appositive Relatives

349

struction effects for binding and the absence of examples with island violations. (iii) Following Donati & Cecchetto (2011), I proposed that both types of relatives are derived by displacement that takes place inside the wh-relative phrase and continues, across the DP phase, into the relative CP, due to the selectional properties of the nominal head. (iv) Regarding resumptive clitic relatives, I have taken them as instantiating structures with a D (the clitic) that has been stranded at the relativization site and undergoes subsequent cliticization onto the finite T. I depart from Donati & Cecchetto (2011) by suggesting that the stranded D can be lexically realized not only in the complementizer relatives, but also in the relatives of the wh-type. 2

Old Romanian Appositive Relatives

Appositive relatives are of two kinds: wh-relatives, introduced by the inflected relative pronoun (48) and by the uninflected one (49); and appositives introduced by the complementizers ce and de ‘that’ (50). The most frequent in terms of occurrence are wh-appositives. (48) a. Constantin vodă, carele ținea pre doamna Mariia Constantin king.M who.SG.M kept DOM lady Maria ‘King Constantin, who was married to lady Maria’ (MC: 31) b. domnii tăi, carii stăpânescŭ în tine lords.PL.M yours.PL.M who.PL.M reign in you ‘your lords, who reign over you’ (MC: 39) c. Marea Albă, carea să întinde pănă în Țarigrad Sea White.F Sea which.SG.F REFL=stretch up to Țarigrad ‘The White Sea, which stretches up to Țarigrad’ (MC: 194) (49) moartea, care multe lucruri taie death which many things cuts ‘death, which ends many things’ (MC: 13) (50) a. de răul lui, ce era fruntea boierilor of evil his that was front.DEF boyars.GEN ‘because of his wickedness, who was the leader of the boyars’ (MC: 13)

350

sevcenco

b. Ieremie vodă, ce era om întreg la toate Ieremie king who was man whole at all ‘King Ieremie, who was a just man’ (MC: 37) c. au adus și capul Sfântului have.AUX.PL brought also head.DEF Saint.GEN Grigorie Bogoslav, de stă pănă astădzi Grigorie Bogoslav that stays until today ‘they also brought the head of Saint Grigorie Bogoslav, which remains until today’ (IN: 111) 2.1 Properties of Appositive Relatives in Old Romanian First, when the antecedent is modified by more than one relative structure, some of which are restrictive whereas the others are appositives, it is always the case that restrictive relatives come first: (51)

Acéste povești ce sântu scrise mai sus, carile spun these stories.PL.F that are written above which.PL.F say de Iliașu Vodă of Iliașu King ‘these stories that are written above, which tell of King Iliașu’ (GU: 171)

Second, appositive relatives have the multiplicity property, as in (52), which shows stacked relatives of different types—subject and object. (52) fost-au și Dragoș, carile au been have.AUX.3PL also Dragoș.M who.SG.M have.AUX.3PL= venitu de la Maramoroș, carile să vediia come from Maramoroș, who.SG.M REFL= saw și mai de folos decâtu toți, even more of and than all pre toții l-au pus mai mare și purtătoriul lor above all him=have.3PL= put more great and spokesman theirs ‘Dragoș was there as well, who had come from Maramoroș, who saw himself as more honourable and more useful than all of them, whom they all named their leader and spokesman’ (GU: 72) Third, just like restrictive relatives, Old Romanian appositives can be extraposed, as shown in (53).

Restrictive And Appositive Relatives

351

(53) într-acesta an venit-au Hroiot cu in this year come=has Hroiot.M with oaste de la unguri asupra lui Ștefan vodă, army from Hungarians against Ștefan king, căruia i-au ieșitu Ștefan vodă înainte cu oaste whom.SG.M to.him = has= come Ștefan king ahead with army ‘this year, Hroiot came with an army from the Hungarians against king Ștefan—Hroiot whom king Ștefan also met with an army’ (GU: 108) Fourth, occasionally, appositive relatives may have an internal head as well, as shown in (54). (54) să ia și să dezbată Cetatea Albă și SUBJ take and SUBJ free Cetarea Albă and Chiliia, care cetăți luase de la Ștefan vodă Baiazitu Chiliia which fortresses took from Ștefan king Baiazitu ‘to take and free Cetatea Albă and Chiliia, which fortresses Baiazid had taken from King Ștefan’ (GU: 109) Quite interestingly, internally headed appositives like (54) are possible only if the relative connector is the uninflected relative pronoun.2 Note that this characteristic finds a straightforward account if we consider the status of the pronoun relative connectors. I claimed that uninflected care is a D head; thus, its noun complement position can be filled in by the internal head. Inflected care, on the other hand, is a DP projection whose NP complement position is already occupied by the pronoun that will subsequently raise to the expletive enclitic article head. The last property to mention regards direct object appositives. Many of them have an Accusative clitic that corefers with the relative head and agrees with it in number and gender features:

2  Only Neculce’s chronicle has an internally headed relative introduced by inflected care (see 1). I have come across no such relatives with a complementizer. (i) Constantin Catacozino, carele acela au cădzut Constantin Catacozino.M who.SG.M that.one has= fallen ‘Constantin Catacozino, who fell’ (IN: 236)

352

sevcenco

(55) a. feciorul lui, carele îl avea numai acela son.DEF.SG.M his.SG.M who.SG.M him = had only that ‘his son, who was the only one he had’ (MC: 113) b. au venit și sol, carele l-au has= come also messenger.SG.M who.SG.M him=has= trimis Brâncovanul vodă sent Brîncovanul king ‘there came a messenger also, whom king Brîncovanul had sent’ (IN: 293) 2.2 The Syntax of Appositive Relatives in Old Romanian 2.2.1 Theoretical Framework In this section, I introduce the framework for the analysis of appositive relatives and I show that the Old Romanian data are compatible with this approach. Basically, I formalize the structure of Old Romanian appositive relatives as being similar to clausal coordination. A sizeable body of research has been dedicated to the investigation of the syntax of appositive relatives, and it is outside the scope of the present paper to give an overview of this. However, I want to hightlight a couple of facts that will be relevant to the theoretical framework I adopt. It has been claimed that appositives are main clauses, which do not form a constituent with their antecedent (Emonds 1979 and subsequent work along this line). It has also been submitted that appositives are subordinate clauses (Jackendoff 1977). This second proposal does not question the fact that the appositive and its antecedent do not form a constituent and advances various hypotheses regarding the relationship between the two: complementation (Platzack 2000), right adjunction (Smits 1988) or coordination (de Vries 2002, 2006a, a.o.). Finally, mixed approaches claim that appositives form a constituent with their antecedent at deep and surface structure, but are detached at LF and adjoin to the main clause (see Demirdache 1991; Bianchi 1999). I will use the complementation and specifying coordination approach in de Vries (2000, 2006b) to give an account for the Old Romanian data, but add a slight modification. The gist of de Vries’s approach is that appositive relatives are part of a coordination structure, in which the first conjunct is the overt antecedent whereas the relative instantiates the second conjunct and has a specifying function, as exemplified in (56), example (14) from de Vries (2006a).

Restrictive And Appositive Relatives

353

(56) [DP1 Annie]i &: [DP2 Øk [CP whok is our manager]]i The indices in (56) mark the fact that both the antecedent and the relative refer to the same entity. As for the status of the appositive, de Vries (2006a) suggests that it is a false free relative, because it has an empty pronominal antecedent, as indicated by the represention in (9). Standard free relatives, on the other hand, have an antecedent that is implied in the relative pronoun. Going back to the derivation of the false free relative, de Vries submits the following proposal: An external determiner selects the relative CP. The relative DP moves to the left periphery to check the [wh] feature on the C head. The empty NP inside the relative DP moves to Spec,D and checks abstract agreement features with the D head. Subsequently, the empty noun moves to the external D to check agreement and abstract case. de Vries further suggests that “the N + D complex corresponds to an independent personal or demonstrative pronoun” (de Vries 2006a: 17). The tree representation in (57) details the derivation of the appositive relative in (56): (57) derivation of appositive relatives de Vries (2006a) (example (34)) CoP DPi

Co'*

Annie &:

DPj

D N

CP D Øk

DP rel NPi tn

C' D'

Drel

(C) ti

IP t dpr

whok The analysis presented above relies on (i) the presence of a empty pronominal antecedent, which may be phonologically realized as he or the one and (ii) raising of the empty noun complement to the specifier of the relative D.

354

sevcenco

My analysis of Old Romanian constructions starts with Grosu’s observation on Modern Romanian appositives: “the appositive expression is not even clausal; rather, it appears to be a DP that properly includes an RC and is headed by a demonstrative pronoun” (Grosu 2013: 526). Accordingly, I propose that the Old Romanian wh-appositive is a complement to a demonstrative pronoun that could be either empty or lexically realized. There is a matching relationship between the demonstrative pronominal head and the relative item in the left periphery of the subordinate clause. This matching relationship consists in φ feature agreement between the two elements. The corpus I have studied has examples of appositives with a lexically overt demonstrative pronoun, as in (58), which brings support to the line of thought I am pursuing. (58) Trimisă solii săi la craiul sent messengers his to prince leșescu, de-și ceriia robii, ceia Polish that=REFL ask slaves.DEF.PL.M those PL.M ce-i luase Tarnovschii hatmanul that = them took Tarnovschii commander.DEF ‘He sent his messengers to the Polish prince to ask for his slaves, those that commander Tarnovschii had taken’ (MC: 163) What motivates the proposed matching relationship? It has been argued that relative pronouns in appositive clauses display anaphoric behavior (Jackendoff 1977; Demirdache 1991; del Gobbo 2003; Grosu 2013 a.o). The anaphoric status of the pronoun makes it enter the φ feature agreement with the pronominal head. As for appositives introduced by a relative complementizer, I will argue in section 2.2.3 that they are derived by the displacement of a null operator (instantiating a bundle of abstract nominal features) in the left periphery of the relative. This null operator needs thus licensing and the φ feature agreement with the pronominal head provides this necessary licensing. 2.2.2 Wh-appositives Let us begin by considering the derivation of wh-appositives introduced by the inflected relative pronoun. As is also the case with wh-restrictive relatives, the pronoun precedes finite verbs in T and the clitics adjoined to them (59), topicalized constituents and focalized ones. Example (60) shows the relative pronoun preceding the focalized constituent multă urgiie ‘much suffering’ and the topicalized pre-verbal subject părinții noștri ‘our parents’ (assuming that the basic word order in Old Romanian is VSO, as in Alboiu, Hill & Sitaridou 2014).

Restrictive And Appositive Relatives

355

(59) feciorul lui, carele îl avea numai acela son.DEF.SG.M his.SG.M who.SG.M him = had only that ‘his son, the only one he had’ (MC: 113) (60) Vasilie vodă, de la carele multă urgiie părinții noștri Vasilie king.M from whom.SG.M much suffering parents ours au petrecut have.AUX.3PL endured ‘King Vasilie, from whom our parents endured much suffering’ (MC: 137) Given these word order clues, I submit that the inflected relative pronoun occupies the specifier of the highest ForceP projection in the left periphery of the subordinate clause. The derivation of a subject wh-appositive as in (61) proceeds as in (62). The relative phrase merges in argumental position and undergoes subsequent merge to Spec,ForceP. There, it enters φ feature agreement with the demonstrative pronominal head. (61) Buda, carele era scaun Crăiei Ungurești Buda, who.SG.M was seat kingdom Hungarian ‘Buda, that was the seat of the Hungarian kingdom’ (GU: 134) (62) wh-appositive with a gap CoP DP

Co'*

Buda

& D

pro dem

DP Force P DP carele

... ...

TP

era scaunul Craiei Direct object wh-appositives have a different structure because they also feature (in most cases) an Accusative resumptive clitic pronominal that corefers with the null demonstrative head and the relative head. I suggest that the ­relative pronoun is part of a big DP projection used to represent clitic

356

sevcenco

doubling structures. More precisely, this DP is headed by the resumptive clitic and the relative DP occupies its specifier position. I give in (64) below a proposal for the syntactic derivation of the wh-appositive with a coreferring clitic in (59), repeated here for convenience as (63). I suggest that the demonstrative acela ‘that’ lexically realizes the null nominal complement of the clitic determiner: (63) feciorul lui, carele îl avea numai acela son.DEF.SG.M his.SG.M. who.SG.M him = had only that ‘his son, who was the only one he had’ (MC: 113) (64) wh-appositive with resumptive clitic CoP DP

Co'*

feciorul lui & D

DP Force P ...

pro dem DP carele . . .

TP

T CL

VP T

V

îl avea DP

DP D

D pro (acela)

Finally, let us look at wh-appositives introduced by the uninflected relative pronoun and focus on an example with an overt internal head (65). In the representation in (66), the relative DP, made up by the relative determiner and its head merges first in argumental position and then raises to Spec,Force. φ features agreement with the emtpy demonstrative head ensues.

Restrictive And Appositive Relatives

357

(65) dincolo de mare, care mare să chiamă . . .  beyond of sea, which sea REFL = calls . . . ‘beyond the sea, which one calls sea . . .’ (MC: 22) (66) wh-appositive with internal head CoP DP

Co'*

mare &

DP

D

Force P

pro dem DP

...

care mare . . . TP se chiamă 2.2.3 Complementizer Appositives The complementizer ce in appositive relatives may precede clitics and finite verbs in T (67), topicalized constituents (68), and the negative operator nu ‘not’ (69). (67) cele 7 sate de țară, ce să dzicŭ hănești those 7 villages of country that se ARB = call hănești ‘those seven villages, which are called hănești’ (MC: 15) (68) aceste povești, ce într-înse spune cursul anilor these stories that in them tell course.DEF years.GEN ‘these stories, which tell about the course of the years’ (GU: 64) (69) Jolcovschii, ce apoi nu l-au mai Jolcovschii that then not him = has = more = slobodzitŭ Schinderŭ-pașea released Schinder-pasha ‘Jolcovschii, whom Schinder pasha never released again’ (MC: 48)

358

sevcenco

Starting from word order data, I suggest that the complementizer merges in the head of the Focus projection. Let us consider an exemplification of an appositive with complementizer ce and a resumptive clitic (see (70). The null operator merges in the specifier position of a DP that accomodates clitic doubling structures. The clitic heads this projection. The operator undergoes displacement to the specifier of Focus and enters φ feature agreement with the empty pronominal head, as shown in (71). (70) Pocuțiia ce o luase Ștefan vodă Pocuțiia that her.3SG.F= took Ștefan king ‘Pocuțiia, which king Ștefan had conquered’ (GU: 120) (71) complementizer appositive with resumptive clitic CoP DP

Co'*

Pocuţiia &

DP

D

Focus P

pro dem DP Op

Focus ...

Focus ...

ce

TP

T CL o

VP T

V

DP

luase DP

D

D

pro

The derivation of subject appositives with complementizer requires the presence of the null operator first in argumental position and then in the specifier of Focus. The complementizer de behaves similary to ce; it precedes clitics and finite verbs in T (72).

Restrictive And Appositive Relatives

359

(72) Radului-vodă . . .  de –i dzicu Radul-vodă cel Mare Radu king.DAT . . .  that = to.him say Radu king the Great ‘to King Radu, whom they call King Radu the Great’ (MC: 36) I therefore assume that appositives headed by de involve a similar derivation with their ce counterparts, with de merging in the head of the focus projection. The analysis of appositive relatives as structures that rely on specifying coordination manages to give a satisfying account of the properties that these appositives have. de Vries 2006a discusses them in detail so I will not dwell on this issue because the explanations regarding ordering effects, stacking, extraposition and the availability of an internal head carry over to the Old Romanian data too. I just want to remark that the specifying coordination analysis could also accomodate a type of appositive that has the antecedent in a previously mentioned sentence (i.e. the relatif de liaison). The relatif de liaison structure is quite frequent in Old Romanian and it was introduced by both the inflected and the invariable relative pronouns, as illustrated in (73). (73) a. și Nicolai vodă au făcut pre Constantin-vodă, and Nicolai king has=sired DOM Constantin king.M careli au fostu domnu aice la noi who.SG.M has= been king here at us în Moldova în anii de la zidirea lumii 7242. in Moldova in years from at creation world.GEN 7242. Careli se trage de pre strămoașă-sa neam who.SG.M REFL=originates from of ancestor-his kin din domnii cei vechi moldovenești. from lords DEF old Moldavians ‘and King Nicolai sired King Constantin. Who reigned here, in Moldova in the years from the creation of the world 7242. Who comes by his ancestors from the old Moldavian kings’ (IN: 112) b. Și au trăit Cârnul până la o a and has= lived Snub.Nose until at a of doa domnie a lui Mihai-vodă Racoviță, și atunce second reign of DEF Mihai-king Racoviță and then a murit. Care mare cinste i-au făcut has= died. which great honour to.him = has= made

360

sevcenco

împăratul la moartea lui. emperor.DEF at death.DEF his ‘And Cârnu lived until the second reign of king Mihai Racoviță and after that he died. For whom the emperor did him great honours at his death’ (IN: 122) The structures in (73) find a natural explanation if we take them to be instances of sentence coordination. The first conjunct contains the overt head Constantin-vodă and the relative DP whereas the second has an elided head, non-distinct from the overt one in the first conjunct, and the second appositive. 2.3 Section Conclusions This section has proposed a uniform syntactic analysis for Old Romanian appositives, both of the wh type and the type introduced by relative complementizers. The analysis formalizes the semantic intution that appositive relatives represent, in fact, appositions coordinated with their antecedents. I have used a slightly modified version of the asyndetic coordination explanation exhaustively outlined in de Vries 2002, 2006a to account for the properties of Old Romanian appositives. The modification concerns the relationship between the appositive and the empty (demontrative) pronominal head whose complement the appositive is. While de Vries goes for displacement of the relative phrase followed by raising of an empty noun to the external D, I suggested that a matching relationship, in terms of φ feature agreement, holds between the empty demonstrative pronoun (which can sometimes be overtly realized) and the relative determiner / operator in the left periphery of the subordinate clause. 3

A Few Comparative Considerations

In this section, I will make a few remarks about the overall picture that emerges from the present paper, especially from a diachronic view. Regarding restrictive relatives in Old Romanian, there are two important facts that need higlighting. The relative pronoun that introduces these relatives has two realizations: the inflected form is integrated in a DP projection and is, therefore, phrasal; its uninflected counterpart instantiates the D head within the DP. Restrictives introduced by complementizers (ce / de) are a lot more frequent than their wh counterparts. These two complementizers cumulated other functions, besides being relative clause subordinators. Ce is an ­interrogative pronoun, a conjunction and a relative determiner (when it introduces free relative). De, on the other hand, goes with other types of subor-

Restrictive And Appositive Relatives

361

dinates, such as purpose and result clauses; it may also be a conditional complementizer and a clause coordinator. This means that there was no dedicated relative complementizer in Old Romanian. Modern Romanian presents a different picture. Complementizer restrictives with ce and de are quite infrequent and limited to relativization from subject, direct and indirect object position. Invariable care is also used as a relative complementizer (Grosu 1994, 2013). The inflected relative pronoun has distinct number and gender forms only when in the Genitive or Dative Case (there is no number and gender inflection when the pronoun is spelled out in the Nominative/Accusative form). As for appositive relatives in Old Romanian, a look at the connectors that introduce them gives us the mirror image of the situation that holds for restrictives. More precisely, appositives with a relative pronoun (inflected or invariable) are much more frequent than their complementizer counterparts. Another important difference concerns the presence of the Accusative clitic that corefers with the relative head. The clitic occurs more frequently with direct object appositives than with direct object restrictives. At this time, I cannot capture the significance of this contrast. The diachronic changes I can discuss in light of the analysis in this paper concern (i) the grammaticalization of the relative pronoun care and (ii) the loss of relatives introduced by ce and de. I conjecture that, from the stage discussed in this paper, when inflected care and invariable care occured in free variation, the loss of inflectional features (number and gender) on the inflected form led to the stablization of the use of invariable care, which could, in principle, be used either as a D head or a C head. Modern Romanian shows variation between D care (present in indirect and prepositional object relatives) and C care (that introduces direct, indirect and prepositional object relatives). It seems that a gradual preference for the use of the C option is on the rise (at least in what is regarded informal speech style). This preference can be translated in syntactic terms as development from XP displacement in relatives (the relative D moves along with its nominal complement, the head of the relative) to direct X merge in C (or Spec,ForceP). This development might be captured by two approaches to grammaticalization. It is compatible with the proposal in Roberts & Roussou (2003) according to which a derivation in terms of Merge (External Merge/EM in more recent minimalist terminology) is less economical than one in terms of Move (Internal Merge/IM). However, this has come to be a controversial claim. Chomsky (2013) notes: “If anything, internal merge is simpler, since it requires vastly less search than external merge (which must access the workspace of already generated objects and the lexicon” (Chomsky 2013: 41). An alternative is the approach in van Gelderen (2004, 2009 a, b, 2011). She argues

362

sevcenco

that economy principles (understood as principles of efficient computation, a third factor component in language growth (Chomsky 2005:6)) guide linguistic change. She suggests that the Head Preference Principle: Be a head rather than a phrase (van Gelderen 2004: 11) explains the development of relative pronouns into complementizers. However, choosing between these two alternatives is outside the scope of this paper, as it requires further inquiry. As for the loss of ce and de as complementizers in standard Modern Romanian, one factor has to be taken into account: both of them cumulate other functions than that of relative connector. Ce is a wh interrogative pronoun, de still introduces other subordinates (adjunct clauses). Care, on the other hand, is a dedicated relative complementizer, it does not go with other types of subordinates. Hence, I suspect that care has come to take over the relativizer role from the other two complementizers because this is the only use it has in Modern Romanian. 4 Conclusions This paper made a set of proposals about the syntactic structure and derivation of restrictive and appositive relatives in Old Romanian and argued that both of them can be given a uniform account. Restrictive relatives are derived by means of HEAD raising, i.e. merging of the the nominal head in the left periphery followed by relabeling of the CP projection (Donati & Cecchetto 2011). Appositive relatives are argued to be appositions to their overt antecedents, with which they are coordinated (de Vries 2006a, a.o.). They were analyzed as a structure involving an empty demonstrative pronominal head (which can occasionally become lexically realized) that takes as complement the relative CP. The empty pronominal head enters a matching relation with the relative pronoun / a relative operator in the left periphery of the CP. The matching relation is expressed as agreement in terms of φ features. Corpus IN Iordan, I. 1955. Ion Neculce, Letopisețul Țării Moldovei. Bucharest: Editura de Stat. GU  Panaitescu. P.P. 1958. Grigore Ureche, Letopisețul Țării Moldovei. Bucharest: Editura de Stat. MC  Panaitescu. P.P. 1979. Miron Costin, Letopisețul Țării Moldovei. Bucharest: Editura Minerva.

Restrictive And Appositive Relatives

363

References Adger, David & Gillian Ramchand. 2005. Merge and Move: Wh-Dependencies Revisited. Linguistic Inquiry 36(2): 161–193. Alboiu, Gabriela; Virginia Hill & Ioanna Sitaridou. 2014. Discourse Driven V-to-C in Early Modern Romanian. Natural Language and Linguistic Theory. DOI 10.1007/ s11049-014-9270-8. Aoun, Joseph & Ye-hui Audrey Li. 2003. Essays on the Representational and Derivational Nature of Grammar. The Diversity of Wh-Constructions. Cambridge, MA: MIT Press. Boeckx, Cedric. 2003. Islands and Chains. Amsterdam: John Benjamins. Bhatt, Rajesh. 2002. The Raising Analysis of Relative Clauses: Evidence from Adjectival Modification. Natural Language Semantics 10: 43–90. Bianchi, Valentina. 1999. Consequences of Antisymmetry: Headed Relative Clauses. Berlin: Mouton de Gruyter. ———. 2000. The Raising Analysis of Relative Clauses: A Reply to Borsley. Linguistic Inquiry 31: 123–140. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press. ———. 2005. Three Factors in Language Design. Linguistic Inquiry 36(1): 1–22. ———. 2008. On Phases. In Robert, Freidin, Carlos P. Otero, Maria Luisa Zubizaretta (eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. 133–166. Cambridge, MA: MIT Press. ———. 2013. Problems of Projection. Lingua 130: 33–49. Demirdache, Hamida. 1991. Resumptive Chains in Restrictive Relatives, Appositives and Dislocation Structures. Doctoral dissertation, MIT. Cecchetto, Carlo & Ecaterina Donati. 2010. On Labeling: Principle C and Head Movement. Syntax 13: 241–278. Donati, Caterina & Carlo Cecchetto. 2011. Relabeling Heads: A Unified Account for Relativization Structures. Linguistic Inquiry 42(4): 519–560. Emonds, Joseph. 1979. Appositive Relatives Have No Properties. Linguistic Inquiry 10: 211–243. Frîncu, Constantin. 1997. Structura limbii române literare. Morfologia. In Ion Gheție (ed.), Istoria limbii române literare. Epoca veche. 113–147. București: Editura Academiei Române. del Gobbo, Francesca. 2003. Appositives at the Interface. Doctoral dissertation, University of California, Irvine. Gelderen, Elly van. 2004. Grammaticalization as Economy. Amsterdam: John Benjamins. ———. 2009a. Renewal in the Left Periphery: Economy and the Complementizer Layer. Transactions of the Philological Society 10: 131–195. ———. 2009b. Feature Economy in the Linguistic Cycle. In Paola Crisma & Pino Longobardi (eds.), Historical Syntax and Linguistic Theory. 93–109, Oxford: Oxford University Press.

364

sevcenco

———. 2011. The Linguistic Cycle. Language Change and the Language Faculty. Oxford: Oxford University Press. Grosu, Alexander. 1994. Three Studies in Locality and Case. London: Routledge. ———. 2013. Relative Clause Constructions and Unbounded Dependencies. In Carmen Dobrovie-Sorin & Ion Giurgea (eds.), A Reference Grammar of Romanian: Volume 1: the Noun Phrase. 597–662. Amsterdam: John Bejamins. Hill, Virginia & Gabriela Alboiu. 2015. Verb movement and clause structure in Old Romanian. Oxford: Oxford University Press. (forthcoming). Jackendof, Ray. 1977. X-bar Syntax. A Study of Phrase Structure. Cambridge, MA: MIT Press. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Platzack, Christer. 2000. A Complement-of-N0 Account of Restrictive and Nonrestrictive Relatives. In Artemis Alexiadou, Paul Law, André Meinunger, Chris Wilder (eds.), The Syntax of Relative Clauses. 265–308. Amsterdam: John Benjamins. Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery. In Liliane Haegeman (ed.), Elements of Grammar. 281–337. Dordrecht: Kluwer. Roberts, Ian & Roussou, Anna. 2003. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Shlonksy, Ur. 1992. Resumptive Pronouns as a Last Resort. Linguistic Inquiry 23: 438–468. Smits, Rik. 1988. The Relative and Cleft Constructions of the Germanic and Romance Languages. Dordrecht: Foris. Sevcenco, Anca. in prep. Romance Restrictive Relatives. A Typological and Theoretical Study. Ms. University of Bucharest. Stan, Camelia. 2013. Some Functions of the Definite Article in Old Romanian. In Gabriela Pană Dindelegan, Rodica Zafiu, Adina Dragomirescu, Irina Nicula, Alexandru Nicolae (eds.), Diachronic Variation in Romanian. Newcastle: Cambridge Scholars Publishing (to appear). Torrego, Esther. 1986. Determiners and Pronouns: a DP Analysis of Noun Phrases in Spanish. Ms., University of Massachussetts, Boston. Uriagereka, Juan. 1995. Aspects of the Syntax of Clitic Placement in Western Romance. Linguistic Inquiry 26: 79–123. Vergnaud, Jean-Roger. 1974. French Relative Clauses. Doctoral dissertation, MIT. Vries, Mark de. 2002. The Syntax of Relativization. Utrecht: LOT. ———. 2006a. The Syntax of Appositive Relativization: On Specifiying Coordination, False Free Relatives and Promotion. Linguistic Inquiry 37(2): 229–270.

Index adjectives embedded 294, 296, 302, 313, 316 non-finite 293, 297 adverb 251–252, 256, 260, 296 agent 187 n. 36, 209, 218, 222 agentivity 218 Agree Local 4, 35, 49, 54, 65, 88, 94, 96, 102, 117, 124, 126 Long Distance 4, 35–37, 54, 62, 66, 89, 94, 118–119, 122 multiple 62, 65, 96, 314, 318–319, 321–322 agreement agreeing 5, 73, 154–155, 160–161, 165 n. 19, 167–169, 172–174, 178, 184–186, 188–190, 192 multiple 2–4, 10, 65, 290, 294, 314, 318–319, 324 non-agreeing 160–161, 168, 185, 188 al genitive 76–80, 83–84, 96, 155 n. 4, 159, 162 n. 15, 163, 165, 168, 173, 185 ambiguity structural 209, 211 animate 6, 213, 218, 222–223, 231, 233, 235–236, 240, 259, 305, 323–324 animacy scale 10, 290, 293, 323–324 article enclitic 2, 32, 106, 252, 351 definite 2–5, 9–10, 17–18, 21, 23 n. 3, 24–40, 43–47, 49, 51–54, 62, 66–70, 73–81, 84–86, 88–91, 93–96, 101–104, 106–107, 109, 114, 116, 124, 127, 131–132, 134, 140, 149, 154–159, 178, 184, 192, 246, 261, 275–276, 331–332, 338 genitival 98, 196 suffixal 154, 156–158, 161, 178, 192 aspect / AspP inner 8, 246, 259–260, 262–263 outer 246, 259 attrition 213, 258 Beneficiary 218 binding 8, 85, 303, 312, 319, 344, 349

bounded 8, 246, 259–261, 263 boundedness 259–260, 263 Case Accusative 200, 210–211, 305, 315, 318, 335 Dative 218–219, 277, 361 Genitive 3, 45, 76, 188, 246–247, 253, 318 marker 2, 156, 158–159, 167, 184–185, 253 morphological 3, 154, 261 oblique 154–155 structural 5, 154, 187–189, 192, 218–219, 306, 316 classifier 100, 105–106, 110–114, 116–121, 123–125, 127–128, 130–132, 134–140, 142–150, 300 ClassP 3, 8, 106, 112–114, 116, 118, 120–127, 134, 143–146, 246, 248–249, 252, 260, 263 clitic doubling 209 object 274, 283, 285 complement 45, 67, 69–71, 73, 86, 89–91, 96, 116, 121, 154, 175, 180–181, 187, 207, 253, 302, 311, 316, 334, 336–338, 341, 344–345, 351, 353–354, 356, 360–362 complex predicate 10, 290–294, 303–304, 306, 311–314, 316–319, 322 concord 113, 187 count noun 260 de genitive 76–77 deficient noun 247–248, 257, 263 deficiency 260 definite DP 110, 164, 176, 229, 234, 296, 299 feature 24, 101–102, 117, 121, 123–124 definiteness double 158, 162–163, 165, 175 scale 10, 286, 290, 293, 323–324 demonstrative phrase 67, 86 denotation filter 214, 235 desemantization 213, 220 determiner genitive 36, 82–83 Differential Object Marking 6, 112, 200, 240, 272, 294, 297, 305, 322–323

366 Directional Asymmetry Principle 269, 271–272, 275, 277 dislocation 159, 180, 214, 219, 239 discourse 85, 92–94, 103, 213, 222, 239, 269, 273, 281, 286, 307 DP non-specific 298, 301 specific 233, 291, 293, 301 episodic event 251 External complexity 278 E-complexity 278, 284, 286 evolutionary developmental biology 272 feature definite 24, 101–102, 117, 121, 123–124 discourse 6, 8, 239, 281, 305 referential 71, 86, 88 fluctuating asymmetry 272, 275–277 gender personal 209, 239 generic reading 252 genitive phrase 67–71, 73, 75–76, 82, 93, 96 habitual reading 257 indefinite specific 223, 233, 323–324 strong 295 n. 2, 311 weak 295 n. 2, 311 infinitive stem 246, 249, 261 long 261–262 short 262 Internal complexity 277 I-complexity 277–278, 286 Kayne’s generalization 200, 209–210, 212 Language Faculty 271, 277 level individual 291, 320 stage 291, 320 measure of complexity 284 mixed projection 259 modifier 67, 69–71, 73, 86, 89–91, 96, 115–117, 141, 147, 258

Index nominal ellipsis 18, 256 n. 5 non-episodic event 255 noun bare 68, 77, 83, 155, 168–169, 174–175, 178–179, 191, 222, 305 mass 3, 263, 300 relational 154, 171–172, 174, 185, 223, 228, 233–234, 240 NumP 3, 8, 18, 48, 83, 112, 118, 189–190, 246, 248–250, 252, 260, 263 Op-Shell 284–285 operator binding 85 pluractional 248 Parameter of Definiteness 17, 24, 27 participial nominal 256 n. 5 polydefinite DP 62, 65–66, 68–71, 74–76, 78, 84, 89, 95, 124 predicative nouns 175 NP 180 bare nouns 175, 178–179 bare NP 175, 178–181, 186 preposition lexical 6, 200–202, 219, 239 property denotation 201, 240 reading 214 possessive 5, 47, 51–52, 128, 155, 157–159, 163, 165, 167, 169, 175–179, 181, 183–188, 191, 223, 233–234, 272, 344 possessor pronominal 164 pseudo-incorporation 7, 201, 235 quantifier generalized 214 reading existential 304, 311 generic 252 individual 214 referential 214 specific 10, 290, 295 n. 2 strong 290, 293, 295, 298 weak 300, 303, 310 re-categorization 246–247

367

Index Recipient 218, 238 reconstruction 154, 292, 305, 307, 310, 319, 325, 344 reference unique 69–70 salience 22 secondary predicate 9–10, 303–307, 309–312, 314–317, 320, 322, 324 scope narrow 292, 301, 304–305, 308–312, 314, 316 wide 291–293, 301–304, 306–311, 322–324 selection double 216 shared argument 10, 290–295, 297–299, 302–304, 306, 310, 315–317, 319–322 specificity strategies 299

strong pronoun 8–9, 269, 272, 274, 278–279, 283–287 subjunctive 262, 280 n. 7 supine based noun 246–247, 253, 258 nominal 246–249, 261, 263 verbal 246–248, 253, 259 stem 246–247, 249, 253, 261 Topic contrastive 201, 214, 234 familiar 234 marker 200–201, 213, 219–220 Universal Spine Hypothesis 259 valency frame 213, 216 variation free 218, 361