Minority Languages, Microvariation, Minimalism and Meaning : Proceedings of the Irish Network in Formal Linguistics [1 ed.] 9781443853309, 9781443850360

Citation preview

Minority Languages, Microvariation, Minimalism and Meaning

Minority Languages, Microvariation, Minimalism and Meaning: Proceedings of the Irish Network in Formal Linguistics

Edited by

Catrin S. Rhys, Pavel Iosad and Alison Henry

Minority Languages, Microvariation, Minimalism and Meaning: Proceedings of the Irish Network in Formal Linguistics Edited by Catrin S. Rhys, Pavel Iosad and Alison Henry This book first published 2013 Cambridge Scholars Publishing 12 Back Chapman Street, Newcastle upon Tyne, NE6 2XX, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2013 by Catrin S. Rhys, Pavel Iosad, Alison Henry and contributors All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-4438-5036-5, ISBN (13): 978-1-4438-5036-0

TABLE OF CONTENTS Preface ..................................................................................................... viii Acknowledgements ................................................................................... ix Introduction ................................................................................................ x Catrin S. Rhys, Pavel Iosad and Alison Henry Part I: Microvariation Chapter One ................................................................................................ 2 Multiple Subjects in Flemish: The External Possessor Liliane Haegeman and Lieven Danckaert Chapter Two ............................................................................................. 24 Object Shift in Belfast English Alison Henry Chapter Three ........................................................................................... 36 Diachronic and Synchronic Variation in the Auxiliary Selection of Eastern Abruzzese Francesca La Morgia Chapter Four ............................................................................................. 53 The Syntax of the After Perfect in Hiberno-English Mariachiara Berizzi and Silvia Rossi Part II: Celtic Languages Chapter Five ............................................................................................. 70 Tense and Features in Irish Verbal Morphology Paolo Acquaviva Chapter Six ............................................................................................... 91 Word-Level Phonological Conditions on Linearization in Scottish Gaelic Andrew Carnie

vi

Table of Contents

Chapter Seven......................................................................................... 101 Local Agreement in Irish and Welsh: Theoretical and Typological Implications Alessio S. Frenda Chapter Eight .......................................................................................... 131 Demonstratives in Irish Frances Kane Part III: Formal Approaches to Syntax, Semantics and Discourse Chapter Nine........................................................................................... 150 A Phase-Theoretic Analysis of Root Compounds (from a Cross-Linguistic Perspective) Leah S. Bauke Chapter Ten ............................................................................................ 171 Recoverability in Slovene Relative Clauses Marko Hladnik Chapter Eleven ....................................................................................... 186 Generic Subjects and Voice in Polish Impersonal Constructions 0DáJRU]DWD.U]HN Chapter Twelve ...................................................................................... 206 On Feature Inheritance and Cyclicity of Derivation Masakazu Kuno Chapter Thirteen ..................................................................................... 222 Inference and Grammar: Intersectivity, Subsectivity, and Phases Ulrich Reichard Chapter Fourteen .................................................................................... 245 What Went Wrong in “Parky” vs. Ryan? Formal Ambiguity in the Emergence of Disaffiliation Catrin S. Rhys Chapter Fifteen ....................................................................................... 266 “Is that the way I look like”: Hegemonic Femininity as an Interactional Accomplishment Catrin S. Rhys and Grace Burke

Minority Languages, Microvariation, Minimalism and Meaning

vii

Contributors ............................................................................................ 290 Index ....................................................................................................... 292

PREFACE THE IRISH NETWORK IN FORMAL LINGUISTICS We are delighted to welcome the publication of papers presented at the Irish Network in Formal Linguistics conferences. The Irish Network in Formal Linguistics was founded in 2009 with the support of a research networking grant from the Arts and Humanities Research Council (UK) to Prof. Alison Henry of the University of Ulster, and to Prof. Eithne Guilfoyle of Dublin City University from the Irish Research Council for Humanities and Social Sciences. They were joined on the organizing committee by Dr Aidan Doyle (University College Cork) and Prof. Jamal Ouhalla (University College Dublin). The aim of the network is to support researchers working on theoretical linguistics (broadly construed) on the island of Ireland; to offer opportunities for them to meet and share their research; and to bring other researchers in theoretical linguistics to Ireland. Prior to the network being established, there was much good research being undertaken, but often by researchers who were the only theoretical linguists in their departments. The network has so far organized three workshops and a major international conference whose products appear in this volume. We hope this will be the first of a series of such conferences in the future. EITHNE GUILFOYLE ALISON HENRY

ACKNOWLEDGEMENTS This volume would not exist were it not for the inspired work of Prof. Alison Henry of the University of Ulster and Prof. Eithne Guilfoyle of Dublin City University in establishing the Irish Network in Formal Linguistics. We would like to thank the Arts and Humanities Research Council and the Irish Research Council for Humanities and Social Sciences for their support for the Irish Network in Formal Linguistics. Finally, many thanks are due to Carol Koulikourdi and colleagues from Cambridge Scholars Publishing, for accepting our proposal to publish this volume and for the help and guidance they have provided all along the process.

INTRODUCTION CATRIN S. RHYS, PAVEL IOSAD AND ALISON HENRY Part 1: Microvariation Microvariation approaches the analysis of dialect variation with a specific focus on how it can contribute to the understanding of linguistic theory. The synchronic and diachronic variation examined in this volume includes Irish English (Henry, chapter 2 and Rossi and Berizzi, chapter 4), dialects of Italian with a focus on Eastern Abruzzese (La Morgia, chapter 3) and dialects of Flemish (Haegeman and Danckaert, chapter 1). Haegeman and Danckaert (chapter 1) discuss a previously undocumented structure in West Flemish where a possessor can, under certain conditions, be realized as a subject. The external possessor pattern described in this chapter sheds new light on the structure of the (high) middle field of Flemish embedded clauses. Henry’s study of object shift in Belfast English (chapter 2) shows that the phenomenon of object shift, generally considered characteristic of the Scandinavian languages and the subject of much debate in the generative literature, occurs in imperatives in Belfast English, and shows how this can contribute to the understanding of object shift in general. Henry’s chapter also highlights the importance of individual-level variation in work on non-standard language varieties and the theoretical insights that such variation can provide. La Morgia (chapter 3) considers diachronic and synchronic patterns of variation in auxiliary selection in Eastern Abruzzese, which displays a person-driven type of auxiliary selection, with the third person singular and plural selecting avé (have) and the other persons esse (be). The analysis considers language-internal as well as language-external influences on patterns of lexical variability in resistance to contact-driven change. Rossi and Berizzi (chapter 4) provide a cartographic account of the “after perfect” in Hiberno-English, a well-known feature of the dialect which has been little studied from a theoretical perspective. The chapter

Minority Languages, Microvariation, Minimalism and Meaning

xi

applies the fine structure proposed for spatial PPs to the P appearing in this aspectual construction but differs from other analyses of aspectual construction involving P in that it analyses the preposition not as the head of an aspectual projection but as a PP hosted in the Spec position.

Part 2: Celtic Languages The linguistic study of Celtic languages provides a natural focus for INFL. This volume includes four papers on the topic; all of them are broadly set in the tradition of generative syntax, but they also exemplify a variety of approaches to important architectural questions, such as the interaction between morphosyntax and semantics (Acquaviva, chapter 5 and Kane, chapter 8), phonology (Carnie, chapter 6) and language use (Frenda, chapter 7). The papers collected in this section also challenge some notions with a long history in traditional descriptions of the Celtic languages, for instance by providing critical examination of variation often said to be “free” (Carnie) and of the monolithic notion of “gender agreement” (Frenda). The chapters by Acquaviva and Kane both reveal important insights into the morphological structure of Irish verbs and demonstratives respectively and into the interaction between syntax and semantics. Acquaviva examines the morphological structure of Irish verbs, showing how the structure of verbal paradigms illuminates the finely grained structure of the left periphery of the clause. He shows that his approach can be used to understand both the morphological make-up of the “irregular” paradigms and the semantics of the Irish tense and aspect distinctions, thus underscoring the importance of the connection between syntax and semantics. Carnie’s chapter sheds new light on the well-known and recently much-researched pattern of pronoun postposing in the Goidelic languages. He argues that, unlike Irish, the rightward dislocation of phonologically “light” object pronouns in Scottish Gaelic cannot be motivated either syntactically or with reference to information structure: instead, prosodic factors, more specifically foot structure, play a decisive role. These data are an important challenge to the well-established principle of PhonologyFree Syntax, and more generally to any strictly modular approach to grammatical architecture. The topic of the chapter by Frenda is the nature of agreement in Irish and Welsh and of its interaction with discourse and with morphophonology. Frenda argues that the patterns of consonant mutation and of gender agreement within the DP can only be understood in a highly

xii

Introduction

formalized, representationally rich theory of grammar which is able to take into account a variety of morphosyntactic and discourse-related factors. Along the way, he argues for a more nuanced understanding of the phenomena covered by the traditional blanket notion of “agreement”, incorporating more than one grammatical mechanism. Finally, Frenda’s study underscores how corpus-based methods and the analysis of spontaneous speech provide insights into the grammar of the Celtic languages that are unavailable in more traditional descriptive frameworks. Kane argues for a novel analysis of the internal structure of the Irish demonstrative DP. She suggests that, in certain contexts, the Irish element an, traditionally described as a determiner, demonstrates the syntactic and semantic properties of a demonstrative (from which it derives diachronically). In particular, she argues that the Irish demonstrative is phrasal, involving two separate components; an adjectival/deictic element and a second component encoding definiteness, the definite article. Kane’s chapter, along with Acquaviva’s, therefore highlights the importance of detailed examination of finely grained syntactic and semantic structure.

Part 3: Formal approaches to syntax, semantics and discourse This general section brings together papers on topics of current theoretical interest in the formal analysis of syntax, semantics and discourse. Several of the papers adopt a phase theoretic approach to address different topics including nominal root compounding (Bauke, chapter 9), Feature Inheritance and derivation (Kuno, chapter 13) and interface conditions on the entailment properties of Adjective Noun modification (Reichard, chapter 13). Hladnik give a minimalist account of Case marking issues in the analysis of relative clauses in Slovene (chapter 10). Bauke, Krzek and Reichard (chapters 9, 11 and 13) all address phenomena involving syntactic conditions on semantic interpretation. The final two papers of this section focus at the level of discourse and demonstrate the advantages of a formal micro-level sequential analysis of talk-in-interaction. Bauke provides an analysis of compounding that is in line with recent minimalist approaches to Phases. In particular, she argues against the claim that nominal root compounding is parameterized across languages and proposes an account of compounding in narrow syntax that relies on a distinction between uncategorized roots, specified only for edge features, and categorizing x-heads that are Phase heads and the standard assumptions for the operation of Merge and Spell-Out. Bauke thus ties

Minority Languages, Microvariation, Minimalism and Meaning

xiii

cross-linguistic differences in the patterns of compounding to a microparameter that is sensitive to the properties of number-checking. Hladnik also relies on the operation of Spell-Out in his analysis of resumption in Slovene restrictive relative clauses in chapter twelve. His investigation of Slovene shows that a relative clause is introduced either by the pronoun kateri, or by ki, a complementizer. Moreover, when the relative clause is introduced by the complementizer, the head noun is obligatorily resumed by a clitic pronoun. This pattern of resumption is argued to operate as a “case-recoverability mechanism” driven by spellout restrictions in the CP domain. In chapter thirteen, Krzek examines the morphosyntactic properties of two Polish impersonal constructions and provides syntactic evidence for the projection of a pro with a generic interpretation in Polish. Krzek also claims that the –NO/–TO impersonal suffix is a base generated head of VoiceP and hence blocks passivization whereas the SIE impersonal can be passivized. These Polish impersonals thus present a significant challenge to both Holmberg’s (2005) generalization that generic null subjects are not found in consistent null subject languages and Fassi-Fehri’s (2009) supplementary generalization that generic interpretation of third person singular null subjects is available in consistent null subject language, but only with a specific Voice, namely Passive. Kuno, in chapter 13, focuses on Chomsky’s (2008) account of Feature ,QKHULWDQFHLQZKLFK7LQKHULWVXQLQWHUSUHWDEOHDJUHHLQJij-features and the EPP property from C. The paper contrasts the morphosyntactic properties of cases where Feature Inheritance is motivated and cases where Feature Inheritance is blocked and argues that where Feature Inheritance is motivated, the derivation inherently proceeds counter-cyclically. Kuno also provides independent evidence for counter-cyclicity and so argues that given that counter-cyclicity is a logical consequence of Feature Inheritance and is also independently motivated, it is preferable to Chomsky’s (2008) proposals for derivational simultaneity. Reichard examines the entailment properties of different categories of adjective and provides an alternative to the formal semantic lexicalist approach, in which he argues that the difference between intersective and subsective interpretations of adjectives follows from interface conditions. Given a phasal architecture of grammar and recent claims that intersective modifiers are clausal whereas non-intersective modifiers are not, Reichard’s overall proposal is that, due to the asymmetric structure of grammar, modifiers are interpreted relative to their hosts. Subsectivity is thus the ‘normal case’. However, if the modifier is clausal, it is also a phase. Since phases are taken to be units of semantic interpretation, the

xiv

Introduction

interpretation of the modifier is fixed at the time of modification and the modifier can therefore not be relativized to its modifiee. An intersective interpretation is therefore the only remaining possibility. The final two papers in this section adopt a formal perspective not to aspects of linguistic structure but to actual stretches of language use in context. Both papers examine conversational data within the conversation analytic approach in which the sequential, formal properties of the talk are examined to draw conclusions at a more macro-level about ambiguities in the emergence of disaffiliation (Rhys, chapter 14) and about the hegemonic character of gender categories in interaction (Rhys and Burke, chapter 15). Chapter 14 draws on formal micro analysis to examine the interaction between discursive practices that are oriented to alignment to/disalignment from the actional content of the talk and the emergence of disaffiliation between the participants. Chapter 15 combines the formal analytical practices of CA with Membership Categorization Analysis in examining mundane conversations of women who smoke and shows the formal interactional practices through which participants resist incumbency of a negatively appraised category of woman smoker

PART I MICROVARIATION

CHAPTER ONE MULTIPLE SUBJECTS IN FLEMISH: THE EXTERNAL POSSESSOR LILIANE HAEGEMAN AND LIEVEN DANCKAERT 1.1 Introduction In this paper, we give an overview of the main properties of some novel data from Flemish which involve a possessor DP that appears separated from the possessee with which it is associated. We refer to this phenomenon as the “external possessor”: we show that both the high possessor and the lower possessee have a number of properties that are usually associated with subjects: this will lead us to the claim that both are hosted in separate subject positions. The external possessor pattern described here sheds new light on the structure of the (high) middle field of Flemish embedded clauses.

1.2 Presentation of the data 1.2.1 The External Possessor The basic data we will be concerned with is a marked word order pattern available in a number of Flemish dialects, which involves a possessor DP which is separated from the possessum by an intervening temporal adjunct. The basic data are exemplified in (1–2): 1 a

dat [Jehan [zenen kleenen]] toen juste in de klinieke was. that Jehan his little then just in the hospital was “... that just then John’s little one was in hospital.”.’

Multiple Subjects in Flemish

3

b % dat [Jehan] toen juste [zenen kleenen] in de klinieke was. that Jehan then just his little in the hospital was “… that just then John had his little one in hospital.” 2 a

b

dat [men moeder [euren pols]] toen juste in de ploaster zat. that my mother her wrist then just in the plastercast sat “... that just then my mother’s wrist was in a plaster cast.” % dat [men moeder] toen juste [euren pols] in de ploaster zat. that my mother then just her wrist in the plaster sat “… that my mother just then had her wrist in a plastercast.”

In the a-sentences in (1–2), the subject is a complex DP (bracketed) consisting of a possessor (leftmost) and a possessee, which is introduced by a possessive pronoun (in boldface). The pattern is referred to as possessor doubling (Haegeman 2004a, 2013). The entire DP containing possessor and possessum sits to the left of the temporal adjunct toen juste ‘just then’ (underscored), right adjacent to the complementizer, which is the canonical subject position in Flemish. As will be shown in section 1.3, the complex possessor–possessee forms a constituent: for this reason, we could call the possessor in (1a–2a) an “internal” possessor (cf. Vergnaud and Zubizaretta 1992). The b-sentences on the other hand show a pattern in which the possessor and the possessee are separated by the adjunct. The possessee appears, somewhat surprisingly, in a position below the adjunct. We will call this pattern the external possessor (cf. Vergnaud and Zubizaretta 1992). Schematically, the structure of the internal and external possessor pattern could be represented as in (3): 3 a b

Internal possessor: C[TP [DP possessor possessum] adjunct … External possessor: %C[TP [DP possessor] adjunct [DP possessum] …

As suggested by the bracketing while the internal possessor forms a constituent with the possessum (see also section 1.3.1.1.), the external possessor and the possessum do not form a constituent. For instance, while the the possessor and the possessum in (1a) can be right dislocated (1c), this is not possible in (1b):

Chapter One

4

1 c

d

dat jei toen juste in de klinieke was, [Jehan zenen kleenen]i. that-he then just in the hospital was, Jehan his little one “... that he just then was in hospital, John’s little one. *dat je in de klinieke was, Jehan toen juste zenen kleenen. that-he in the hospital was, Jehan then just his little one

Crosslinguistically, the phenomenon of external possession (in a very broad sense) is very common (see for instance Szabolcsi 1983, 1994 (on Hungarian), Chinese (Xu 2004; Hsu 2009) and the contributions in Payne and Barshi 1999), but the Flemish data have thus far received very little attention1. Our contribution is organized as follows. In the remainder of section 1.2, we will highlight some important features that characterize the Flemishexternal possessor. Section 1.3 provides some background about the syntax of possession in Flemish DPs. Section 1.4 is devoted to the syntactic analysis of the split possessor pattern. We show that both the external possessor and the lower possessum DP have subject properties. We will propose that both the higher and the lower DP are hosted in a TPinternal dedicated subject position.

1.2.2 Some salient properties of the external possessor First of all, some remarks about the regional distribution and acceptability of the EP are in order. The pattern is accepted both in some regional dialects and in the so called tussentaal, the informal regiolect that is used across Flanders. Asked whether utterances like the b-examples in (1–2) would be acceptable in their dialect or in an informal regiolect, 14 out of 24 Flemish informants accepted the pattern in sharp contrast to Dutch speakers from the Netherlands who uniformly reject the pattern. However, there is much variation in acceptance among Flemish speakers and at this stage this variation is not clearly linked to a specific regional dialect. The external possessor pattern was accepted by 6 out of 9 West Flemish informants, by 6 out of 10 East Flemish informants, by 1 out of 4 Brabant informant, and by one bilingual (Flemish-French) speaker with West Flemish background; an informant from Antwerp rejected the pattern. Some speakers who reject the pattern do admit that it sounds like something that is possible in other dialects, and associate it with West

1

For a first discussion, see Haegeman and Van Koppen (2012).

Multiple Subjects in Flemish

5

Flemish. Speakers who do not accept the pattern do share (some of) the judgements discussed below. Second, unlike comparable phenomena in French (Vergnaud and Zubizaretta 1992; Guéron 2006) and German (Burridge 1990; Hole 2006; Lee-Schoenfeld 2006), the Flemish external possessor requires the presence of a possessive pronoun in the DP that expresses the possessum. This is illustrated in the examples (4), where the determiner of the possessum-DP is an indefinite article rather than a possessive pronoun2. 4 a b

*dat [Jehan] toen juste [nen zeune] in de klinieke was. that Jehan then just a son in the hospital was *dat [men moeder] toen juste [nen pols] in de ploaster zat. that my mother then just a wrist in the plaster sat

Because the external possessor pattern seems to depend on the possessor doubling pattern, we will elaborate briefly on the syntactic encoding of possession in Flemish DPs in section 1.3 below. Third, the external possessor exhibits a number of properties which are usually attributed to subjects: the most conspicuous of these properties are (i) the ability of the external possessor to establish an agreement relation with a complementizer and (ii) the fact that (weak) indefinite possessor DPs can trigger insertion of the expletive er. Fourth, there are significant restrictions of an interpretive nature on the external possessor, which do not (or in any case less strongly) apply to the internal possessor pattern. In general, external possessors are subject to a strong animacy requirement and they are always in some sense “affected” by the event expressed by the TP. The subject-like behaviour and the interpretive nuances of the external possessors will be the subject of section 1.4. Finally, the external possessor is only available in subordinate clauses, finite and non-finite alike: in root clauses, the pattern is strongly degraded. For reasons of space, this point will not be addressed in this paper: the reader is referred to Haegeman and Van Koppen (2012) for an analysis.

2

At this point, it remains unclear to us why the external possessor phenomenon is only available when the possessum DP contains a doubling pronoun. We hope to return to this question in future research.

Chapter One

6

We now turn to some background on the structure of Flemish DPs, concentrating on two different ways in which possessor–possessee relations can be expressed.

1.3 Background: possessor relations in the Flemish DP In the previous section, we saw that the split possessor pattern is only acceptable if the a possessive determiner is present in the possessum -DP (cf. the * examples in (4)). However, as shown in the data in (5), not any possessive element can license the external possessor: 5 a

b

dat [DP men moeder sen pols] toen juste in de ploaster zat. that my mother sen wrist then just in the plastercast sat “that just then my mother’s wrist was in a plaster cast.” *dat [DP men moeder] toen juste [DP sen pols] in de ploaster zat. that my mother then just sen wrist in the plaster sat

In the example in (5a), the subject is a complex DP containing the prenominal “genitive” marker se(n), which resembles the Saxonian genitive in English. (5b) shows that the se(n)-element does not allow allow the split possessor pattern.

1.3.1 Prenominal possessors in (West) Flemish: a brief survey The two basic ways in which a relation between a possessor and a possessee can be expressed in (West) Flemish are exemplified in (6): 6 a

b

DP + sen + NP = “prenominal genitive” Valère sen hoed Valère sen hat (DP) + poss.pronoun + NP = “prenominal doubling” (Valère) zenen hoed (Valère) his hat

Multiple Subjects in Flemish

7

1.3.1.1 Some similarities Before we look at the differences between (6a) and (6b), it should be pointed out that the genitival and the doubling pattern have two important characteristics in common. First of all, the complex “possessor–possessee” form a single constituent. For instance, both can fill the preverbal slot in a verb second clause: 7 a b

[DP Lieven sen computer] is weeral kapot. Lieven sen computer is again broken. [DP Lieven zijnen computer] is weeral kapot. Lieven his computer is again broken. “Lieven”s computer is broken again.”

Second, both the se(n)-genitive and the doubling pattern are subject to a generalized animacy restriction: the prenominal possessor DP cannot refer to an inanimate object3: 8 a b

*[die deure] eur klinke that door her latch

c

*[die deure] se klinke that door sen latch

*[dienen rok] zenen zuom that skirt his hemline

d

*[dienen rok] sen zuom that skirt sen hemline

1.3.1.2 Differences between the two patterns However, the two patterns also differ along a number of dimensions. A first difference is that the se(n)-element does not vary with respect to the gender and number of the possessum DP it precedes (9b)4. On the other

3

In Flemish, relations of possession in which the possessor is inanimate can be expressed by means of a PP headed by van “of”: (i) de klink the latch

4

[van of

[de deur]] the door

This is not to say that the element is strictly speaking “invariant”. However, the alternation between the forms sen and se is phonologically determined: sen is used before a word starting with a vowel, and se before a word starting with a consonant (Taeldeman 1995; Haegeman 2013: 224-226).

Chapter One

8

hand, the doubling possessive determiner displays agreement with the DP it modifies (9a): 9 a b

MASC SG

a Marie euren-hoed Marie her.M.SG hat b Marie sen hoed Marie sen hat

FEM SG

Marie eur veste Marie her jacket Marie sen veste Marie sen jacket

NEUT SG

Marie eur kleed Marie her dress Marie se kleed Marie sen dress

Second, the se(n)-genitive cannot appear with ellipsis of the head noun (10b–c). NP-ellipsis is possible after a doubling possessive determiner, albeit only if a definite article is inserted between the possessor and its doubling pronominal (11b). 10

a Marie se boeken Marie se book b *(de) Marie sen c *Marie de sen ‡

11

a Marie eur boeken Marie her books b Marie d’eure ‡ Marie de her+e c *Marie eur ‡ d *de Marie eure e * Marie eure

Third, the se(n)-genitive has a more restricted syntactic distribution: it cannot be used if the possessor is a plural. Such a restriction does not hold for the doubling pattern (12b): 12 a

*djungers sen hus b djungers under hus the children sen house the children their house “the children’s house”

Fourth, the possessor and se(n) need to be linearly adjacent, whereas a possessor can be separated from the possessum by a universal quantifier as in (14b). The fact the examples in (13–14) involve topicalized direct objects occupying the first slot of a declarative verb second clause shows again that the possessor and the possessee are contained in a single constituent. 13 a

[Al [Marie sen boeken]] een-k gezien. Q Poss sen NP all Marie sen books have-I seen

Multiple Subjects in Flemish

b

14 a

b

*[Marie al sen boeken] een-k *Poss Q sen NP “All Marie’s books I have seen.” [Al [Marie Q Poss all Marie

eur eur her

gezien.

boeken]] een-k NP books have-I

[Marie al eur boeken] Poss Q eur NP “All Marie’s books I have seen.”

9

een-k

gezien seen gezien

Other differences between the two patterns include the possibility of the possessor to be modified by deictic markers and appositive relative clauses: modification of this kind is only possible with possessors that come with a doubling pronoun. With the se(n) pattern the possessor can be reciprocal, this is not an option with the doubling pattern (cf. Haegeman 2004b). For reasons of space, these phenomena are not illustrated here.

1.3.2 Summary The representation in (15), based on Haegeman (2004a), tries to capture the syntactic differences between the two possessor constructions. In line with much work on the syntax of possessors, a specialized functional projection “PossP” is postulated in the extended projection of the NP which heads the containing DP: According to Haegeman (2004a), the main syntactic difference between the two possible ways of expressing possession relations in Flemish DPs is the position of the possessor DP in the functional structure of the complex DP. In the se(n)-pattern, the possessor sits in Spec,PossP, fairly deeply embedded in the entire structure. On the other hand, the possessor is in a higher position in the doubling pattern, say in Spec,DP5. 5

However, it is probably not the case that the doubling possessor sits in a DPinternal TopP or FocP (on TopP and FocP in DP, see Giusti 1996 and Aboh 2004). The possessor is in no sense emphatic or contrastive, which makes FocP an unlikely option. Moreover, bare quantifiers, which are known to be “unlikely” topics (see Rizzi 1997), can occur as possessor in the doubling pattern: (i) Dat zijn [niemand zen that are no one his “Those are no one's affairs.”

zaken]. businesses

Chapter One

10

15

DP Spec

D’ PossP6

D

Spec

Poss’ Poss

NumP Spec

Num’ Num

NP

a

DP/pro

den

euren

Ø

b

DP/pro

euren

euren

boek

c

DP

sen

boek

1.4 The syntax of the Flemish external possessor We now return to the external possessor and we start our discussion by showing that both the possessum and the external possessor behave in various ways as subjects.

1.4.1 Subject properties of the possessum First of all, and most prominently, the possessum DP invariably show agreement with the finite verb, which by assumption is encoded in the Tnode in the clausal spine: 16 a

dat [Valère] toen juste [zen ouders] niet in Gent woaren/*was. that.SG Valère then just his parents not in Gent be.PL/*SG. “that just then Valère’s parents weren’t in Ghent.”

The nature of the position occupied by the doubling possessor remains to be clarified. Haegeman (2004a) suggests it is similar to the initial position in a subject initial V2 clause. 6 Possibly the possessor in Spec, PossP originates in a lower position, for instance as a predicate of a small clause (cf. den Dikken 1998). PossP can perhaps be compared to the clausal SubjP (Rizzi 2004, Rizzi and Shlonsky 2005, Tortora and den Dikken 2010).

Multiple Subjects in Flemish

11

b dat [Valère] tegenwoordig [zen twee GSMs] atent that.SG Valère these days his two mobiles always an stoan/*stoat. on stand.PL/*.SG. “that nowadays Valère’s two mobiles are always switched on.” Furthermore, while a direct object containing a possessor doubling pattern can be A’-extracted (17a) across a subject, the possessum cannot be extracted across the external possessor (17b–c). Though we cannot go into the details of extraction here, it is tempting to interpret the contrast between (17a) and (17c) in terms of an object/subject asymmetry. 17 a

b

t’Is [Valère zenen sloapkoamer]i dan ze toen juste ti gingen It is Valère his bedroom that they then just went schilderen redecorate “It’s Valère’s bedroom that they were going to redecorate.” da Valère toen juste [zenen sloapkoamer] geschilderd was that Valère then just his bedroom redecorated was “... that just then Valère’s bedroom was being redecorated”

c *t’Is [zenen sloapkoamer]i da Valère toen juste ti geschilderd was. It is his bedroom that Valère then just redecorated was

1.4.2 Subject properties of the possessor As hinted at above, the external possessor also behaves in some ways like a subject. First, some speakers allow a nominative pronoun as the external possessors: 18 a

%? dat [zie ier] toen juste [eur scheerapparaat] kapot was. that she.NOM here then just her razor broken was

b

%??da-n [zunder doar] toen juste [underen computer] kapot was. that.PL they.NOM there then just their computer broken was

Although pronominal external possessors bearing nominative case are always marginal at best (they are most acceptable if modified by a deictic marker like ier 'here' and doar 'there'), they are definitely better than

Chapter One

12

pronouns bearing dative case, which are completely unacceptable both as an internal (not illustrated) and as external (cf. (19)) possessor: 19 a b

*dat [eur ier] toen juste [eur scheerapparaat] kapot was. that.SG her.DAT here then just her razor broken was *dat/da-n [under ier] toen juste that.SG/that.PL them.DAT here then just kapot was. broken was

[underen computer] their computer

This is not because realization of dative case is no longer available and/or problematic in West Flemish7 (cf. (20b)): 20 a

da [dienen GSM] Marie nie anstond that that mobile Marie not on.stood “that that mobile did not please Marie”

b

da that

[dienen GSM] that mobile

eur niet anstond her.DAT not on.stood

c

*??da [dienen GSM] ze niet anstond that that mobile her.ACC not on.stood “that that mobile did not please her”

Finally, in West Flemish as well as in many other Flemish varieties, indefinite external possessors give rise to insertion of the expletive element (d)er (which is roughly equivalent to the English there-expletive) (21a). (D)er-insertions is typically triggered by indefinite subjects (cf. Haegeman 2004a for more discussion). The indefinite subject in (21b) triggers (d)er-insertion, the indefinite object in (21c) does not. A subject containing an indefinite possessor also triggers (d)er-insertion (cf. Haegeman 2004a) (21d). Observe that it is not possible that (d)er-insertion in (21a) is triggered by the lower DP underen GSM 'their mobile phone', since DPs containing a possessive determiner without an indefinite doubling possessor qualify as definite and thus not able to give rise to (d)er-insertion (cf. (21e)).

7

This is contrary to much of the literature in which it is proposed that Flemish dialects no longer have the dative case.

Multiple Subjects in Flemish

21 a

dan *(der) that.PL there af stoat off stands

13

[veel studenten] atent [underen GSM] many studenten always their.MSG mobile

b

dan *(der) veel studenten that *(there) many students atent afzetten always off.switch

underen GSM their.M.SG mobile

c

dat (*er) Lieven that (*er) Lieven

d

dat *(er) that.SG there af stoat off.stands

[veel studenten many students

e

dat (*der) that (*there)

underen GSM atent af stoat their mobile always off.stands

veel studenten many students

eet has

underen GSM] their.M.SG mobile

atent always

Third, further evidence for the subject-like quality of the external possessor comes from dialects which display the phenomenon of Complementizer Agreement (CA), like West Flemish (see also Haegeman and Van Koppen 2012). In the canonical case, West Flemish complementizer agreement qua number targets the subject, i.e. the external argument that also controls agreement on the finite verb: 22 a

b

'k peinzen da-n/*da die venten I think that.PL/that-SG those men “I think that those men know Marie.” ’k peinzen da/ *da-n dienen vent I think that.SG/that-PL that man “I think that that man knows Marie.”

Marie kenn-en. Marie know.PL Marie kenn-t. Marie know.SG

Note that CA is not simply dependent on an adjacency requirement (contra Ackema and Neeleman 2004; Miyagawa 2009). To the extent that a (scrambled/focused) object DP can (very marginally, cf. ?? (23b)) intervene between C and the subject DP, CA targets the non-adjacent subject DP (23b) and, crucially, it cannot target the intervening object DP (23c).

Chapter One

14

23 a

’k peinzen da zelfs Valère zukken boeken niet leest. not reads I think that.SG even Valère such books

b

?? ’k peinzen da zukken boeken I think that.SG such books

c

* ’k peinzen da-n zukken boeken I think that.PL such books

zelfs Valère niet leest. even Valère not reads zelfs Valère even Valère

niet leest. not reads

Interestingly, CA is controlled by external possessors. Such is the case in (24a), in which the complementizer agrees with the plural external possessor rather than with the singular possessum which itself induces number agreement on the finite verb: 24 a

b

omda-n/*omdat [André en Valère] toen juste then just because.PL/because.SG ANDRÉ and Valère gebeld oan/*oat phoned had.PL/*SG “…because André and Valère called just then.” omda-n/*omdat [André en Valère] toen juste [underen because.PL/*because.SG André and Valère then just their computer] kapot was/*woaren. computer broken was.SG/were.PL “…because André and Valère’s computer broke down just then.”

However, note that the external possessor is not dependent on (= licensed by) CA (pace Haegeman and Van Koppen 2012). This can be seen in non-finite (infinitival) clauses, in which CA can never occur but which are nevertheless compatible with the external possessor pattern: 25 a b

Mee Valère toen juste With Valère then just

nen nieuwen oto gekocht te een… a new car bought to have

Mee [Valère] toen juste With Valère then just

[zenen computer] kapot te zyn… his computer broken to be

Multiple Subjects in Flemish

15

1.4.3 Interpretive features of the external possessor The external possessor is always in some sense affected by the event (or more precisely the state resulting from the event) expressed in the proposition. Moreover, there is a very strong requirement for the split possessor to be alive at the time the event or state expressed in the clause takes/took place. Consider for instance the examples in (26), with an internal possessor in (26a) and an external one in (26b): 26 a

omdat [men grootvader zijnen fiets] dan juist kapot was because my grandfather his.M.SG bicycle then just broken was “... because my grandfather's bike was broken just then.”

b

omdat [men grootvader] dan juist [zijnen fiets] kapot was because my grandfather then just [his bicycle] broken was “... because my grandfather had just then his bike broken.”

For all the speakers that we consulted, (26b) can only be uttered felicitously if the grandfather was alive at the moment his bike was broken. In addition, the sentence also implies that the broken bike is not just any bike owned by grandfather, it is his “personal” bike, the one he uses daily8. (26a), on the other hand, could, at least for some speakers, also refer to a bike inherited by one of the grandfather's grandchildren after the grandfather himself has died. Observe that the ban on the dead external possessor is much stronger than the animacy constraint on possessor doubling and possessive pronouns discussed above (section 1.3.1, cf. the examples in (8)). There is no general ban on a dead possessor in the internal possessor pattern with pronominal doubling, as witnessed by the (attested) example in (27b): 27 a

8

[Zijn auto] “His car

werd gevonden was found

aan de kaai. on the quay.”

Cf. the extended use of the term “inalienable possession” in Vergnaud and Zubizarreta (1992).

Chapter One

16

b

Voorbijgangers hebben woensdagmiddag in Hemiksem het lichaam van L.B. aangetroffen in de Schelde. De man verdween in de nacht van 1 op 2 april. De doodsoorzaak ligt nog niet vast. [L.B. zijn auto] was eerder al gevonden aan de Scheldekaaien in Antwerpen. (http://www.nieuwsblad.be/article/detail.aspx?articleid=DMF201 10413_153) “Wednesday afternoon, passers-by found the dead body of L.B. in the river Scheldt. The man had disappeared in the night of 1 and 2 April. The cause of his death is yet to be determined. [L.B. his car] had been found earlier at the Scheldekaaien in Antwerp.”

c

omdat [L.B. zijnen auto] gisteren al gevonden is because L.B. his.M.SG car yesterday already found is

However, the external possessor would be completely unacceptable in this context: d

*omdat [L.B.] because L.B.

gisteren yesterday

[zijnen auto] his.M.SG car

gevonden is found is

We now turn to the closing section of the paper, in which we will argue that the external possessor is located in a high TP-internal argument position with subject properties. Our conclusion supports a line of research that distinguishes more than one subject position in the clausal spine (see Henry 1995 on Hiberno English; É. Kiss 1996; Cardinaletti 1997, 2004; McCloskey 1997 and Tortora and den Dikken 2010).

1.4.4 A high subject position In a first analysis of the Flemish external possessor data, Haegeman and Van Koppen (2012) proposed the following structure, in which the KLJK SRVVHVVRU LV ORFDWHG LQ D SURMHFWLRQ ODEHOOHG Į3 LQVSLUHG E\ Miyagawa 2009).

Multiple Subjects in Flemish

17

CP

28

Į3

&Xij SRVVHVVRULij

Į

Į

FP adjunct

F’ F

TP T’ 7Xij

VP possessum Lij

One central question that remains is to identify the exact nature of ĮP? Recall that we would like to capture both the syntactic (subject-like) properties of the external possessor and its interpretive characteristics. 1.4.4.1 ĮP is not part of the clausal left periphery First of all, it is important to observe that the external possessor pattern is available in a wide range of embedded clauses (29), including those that are not normally compatible with Root Transformations (Emonds 1976) or Main Clause Phenomena (Hooper and Thompson 1973), such as conditional clauses (29b), subject clauses (29c), relative clauses (29d,e) or infinitival mee-clauses (29b). Indeed, as shown by Haegeman and Van Koppen (2012) it is not available in root clauses. 29 a

Het moest lukke dat [Valère] toen juist [zijnen GSM] It must happen that Valère then just his.M.SG mobile afstond. off.stood “It so happened that just then Valère had his mobile switched off.”

Chapter One

18

b Als [Valère] dan juist [zijnen GSM] afstaat kunnen we if Valère then just his.M.SG mobile of.stands can we hem niet bereiken. him not reach. “If at that time Valère has his mobile switched of, we won’t be able to get in touch.” c

Da [Valère] toen juist [zijnen GSM] afstond that Valère then just his.M.SG mobile of.stood was toeval. was coincidence “It was a coincidence that just then Valère had his mobile switched off.”

d

Dat was in de tijd dat [mijn broer] toen juist that was in the time that my brother then just [zijnen computer] kapot was. broken was. his.M.SG computer “That was in the time that my brother’s computer was broken.”

e

Dat is die kliniek [waar da [mijn zus] verleden jaar that is that hospital where that my sister last year [haren kleinen] opgenomen was]. her.M.SG little.M.SG up.taken was. “that’s the hospital where my sister had her little one last year.”

f

*Valère stond toen juist zijnen GSM af. Valère stood then just his mobile off.

Since argument fronting typically qualifies as a Main Clause Phenomenon (Haegeman 2006), the data in (29) strongly suggest that the external possessor is TP-internal rather than in a left peripheral/dislocated position. The next section reviews the hypothesis that ĮP is an A-position. 1.4.4.2 ĮP as an A-position Evidence for A-VWDWXVRI6SHFĮ3FRPHVILUVWRIDOOIURPWKHIDFWWKDW the external possessor shows complementizer agreement, under the assumption that agreement relations are established in A-positions. Furthermore, there is a locality restriction on the relation between the external possessor and the possessum in that an external possessor DP

Multiple Subjects in Flemish

19

cannot be associated with (extracted from?) a direct object across a subject. In (30a) the object DP contains a doubling possessor, but this possessor cannot be externalized and located above the subject DP 30 a

b

omdat Valère toen juste [Marie euren computer] because Valère then just Marie her computer gerepareerd oat repaired had “because Valère just then had repaired Marie’s computer’ *omdat [Marie]iValère toen juste [eureni computer] gerepareerd because Marie Valère then just her computer repaired had gerepareerd oat repaired had

Whether the pattern in (30b) is derived through movement or not, one could reasonably think that the anaphoric dependency between the leftward possessor and the possessive determiner euren 'her' is blocked by the intervening subject DP Valère. 31 a C POSS

subject object predicate

b *C POSS subject object predicate Assuming that the external possessor is an A-position then the ungrammaticality of (31b) is ascribed to A- intervention. 32 * A

A

A

We could hypothesize that the split possessor DP is located in a dedicated SubjP, which is a canonical subject position in the high middle field, specialized to host “categorical” subjects, typically full DPs (cf. Cardinaletti (2004) and Rizzi and Shlonsky (2005, 2006); see also Alexopoulou, Doron and Heycock (2004) on “broad subjects”). However, this analysis has the drawback that it does not explain the “affectedness” reading that is characteristic of the external possessor.

20

Chapter One

1.4.4.3 Accounting for the affectedness effect: a high ApplP? An alternative approach would be to assume that the affectedness reading associated with the external possessor is the result of it being located in an applicative phrase: this would not only account for the idiosyncratic interpretive properties of the external possessor (we could consider those to be a lexical property of the applicative head), it also would explain why an additional argument can be present in the structure, without being part of the selectional frame of the predicate of the clause. A problem that immediately arises for such an analysis is the fact that the Flemish external possessor sits much higher than the ApplPs proposed in recent work on the syntax of applicatives (see esp. Pylkkänen 2008): a low ApplP is usually taken to be located inside the verb phrase, whereas a high ApplP immediately dominates it. However, with Rivero (2009) and Rivero, Arregui and )UąFNRZLDN (2010) we could assume an ApplP which takes the entire TP as its complement, yielding a reading in which the applied argument is effected, which is exactly what we find in the case of the Flemish external possessor. The structure would be as in (33): 33

[CP [ApplP DPPOSS.OR [FP Adj [TP [vP DPPOSS.EE ]]]]]

1.4.5 Remaining issues There remain a number of issues not dealt with in this paper. First, it remains to be determined whether the external possessor is merged directly in the high position (SubjP or ApplP) or is extracted from the lower position containing the possessum DP. Moreover, it is not clear why the external possessor pattern should only be available in embedded clauses (section 1.3.2, see Haegeman and Van Koppen 2011 for an analysis). For reasons of space we cannot develop these points here.

1.5 Conclusion In this paper, we discussed the syntax of the Flemish external possessor construction. We showed that such an external possessor always occurs with a clausemate possessum DP, and that both of these constituents have subject properties. We concluded that the external possessor occupies a high argument position in the middle field, possibly an ApplP that takes the entire TP as its complement. Our analysis lends support to the idea that the notion of “subject” is not a syntactic primitive,

Multiple Subjects in Flemish

21

but that multiple positions for subjects are present in the clause, which under certain circumstances can be filled simultaneously.

References Aboh, E. 2004. Topic and focus within D. Linguistics in the Netherlands 2004: 1–12. Ackema, P. and A. Neelema. 2004. Beyond morphology: interface conditions on word formations. Oxford: Oxford University Press. Alexopoulou, T., E. Doron and C. Heycock. 2004. Broad subjects and Clitic Left Dislocation. In D. Adger, C. De Cat and G. Tsoulas (eds.), Peripheries: syntactic edges and their effects, 329–358. Dordrecht: Kluwer. Burridge, K. 1990. Sentence datives and grammaticalization of the dative possessive. La Trobe University Working Papers in Linguistics 3: 29– 47. Cardinaletti, A. 1997. Subjects and clause structure. In Haegeman, L. (ed.), The new comparative syntax, 33–63. London: Longman. —. 2004. Towards a cartography of subject positions. In L. Rizzi (ed.), The cartography of CP and IP, 115–165. Oxford: Oxford University Press. Den Dikken, M.1998. Predicate inversion in DP. In A Alexiadou and C. Wilder (eds.), Possessors, predicates and movement in the determiner phrase, 177–214. Amsterdam: Benjamins. Emonds, J. 1976. A transformational approach to English syntax. New York: Academic Press. É.Kiss, K. 1996. Two subject positions in English. Linguistic Review 13: 119–142. Giusti, G. 1996. Is there a FocusP and a TopicP in the Noun Phrase structure? University of Venice Working Papers in Linguistics 6:105– 128. Guéron, J. 2006. Inalienable possession. In M. Everaert and H. van Riemsdijk (eds.), The Blackwell companion to syntax, vol. 2, 589–638. Oxford: Blackwell. Haegeman, L. 2004a. DP periphery and clausal periphery: possessor doubling in West Flemish. In D. Adger, C. de Cat and G. Tsoulas (eds.), Peripheries: syntactic edges and their effects, 211–240. Dordrecht: Kluwer. —. 2004b. A DP-internal anaphor agreement effect. Linguistic Inquiry 35: 704–712.

22

Chapter One

—. 2006. Argument fronting in English, Romance CLLD and the left periphery. In R. Zanuttini, H. Campos, E. Herburger and P. Portner (eds.), Cross-linguistic research in syntax and semantics: negation, tense and clausal architecture, 27–52. Washington (DC): Georgetown University Press. —. 2013. Two prenominal possessors in West Flemish. In K. Börjars, D. Denison and A. Scott (eds.), Morphosyntactic categories and the expression of possession, 219–252. Amsterdam: Benjamins. Haegeman, L. and M. Van Koppen. 2012. Complementizer agreement and the relation between C° and T°. Linguistic Inquiry 43: 441–454. Henry, A. 1995 Belfast English and Standard English. Oxford: Oxford University Press. Hole, D. 2006. Extra argumentality – affectees, landmarks and voice. Linguistics 44: 383– 424. Hooper, J. and S. Thompson. 1973. On the applicability of root transformations. Linguistic Inquiry 4: 465–497. Hsu, Y.-Y. 2009. Possessor extraction in Mandarin Chinese. Penn Working Papers in Linguistics 15. 95–104. Lee-Schoenfeld, V. 2006. German possessor datives: raised and affected, Journal of Comparative Germanic Linguistics 9: 101–142. McCloskey, J. 1997. Subjecthood and subject positions. In L. Haegeman (ed.), Elements of grammar, 197–235. Dordrecht: Kluwer. Miyagawa, S. 2009. Why Agree? Why move? Cambridge, MA: MIT Press Payne, D. and I. Barshi (eds.), 1999. External possession. Amsterdam: Benjamins. Pylkkänen, L. 2008. Introducing arguments. Cambridge, MA: MIT Press. Rivero, M.-L. 2009. Intensionality, high applicatives and aspect: involuntary state constructions in Bulgarian and Slovenian. Natural Language and Linguistic Theory 27: 151–196. Rivero, M.-L., A. $UUHJXL DQG ( )UąFNRZLDN  Variation in circumstantial modality: Polish versus St’át’imcets. Linguistic Inquiry 41:704–714. Rizzi, L. 1997. The fine structure of the left periphery. In Haegeman, L. (ed.), Elements of grammar: handbook in generative syntax, 281–337. Dordrecht: Kluwer. —. 2004. Locality and left periphery. In A. Belletti (ed.), Structures and beyond, 223–251. Oxford: Oxford University Press Rizzi, L. and U. Shlonsky. 2005. Strategies of subject extraction. In H.-M. Gärtner and U. Sauerland (eds.), Interfaces + Recursion = Language? 115– 160. Berlin: Mouton de Gruyter.

Multiple Subjects in Flemish

23

—. 2006. Satisfying the Subject Criterion by a non-subject: English locative inversion and heavy NP shift. In M. Frascarelli (ed.), Phases of interpretation, 341–361. Berlin: Mouton de Gruyter. Szabolcsi, A. 1983. The possessor that ran away from home. Linguistic Review 3: 89–102. —. 1994. The noun phrase. In F. Kiefer and K. É. Kiss (eds.), The syntax of Hungarian, 179–274. New York: Academic Press. 7DHOGHPDQ--DQ=‫ތ‬1QRPHQRYHUHHQEH]LWVFRQVWUXFWLHLQGH Vlaamse dialecten. Taal en Tongval 47: 220–228. Tortora, C. and M. den Dikken. 2010. Subject agreement variation: support for the configurational approach. Lingua 120: 1089–1108. Vergnaud, J.-R. and M.-L. Zubizarreta. 1992. The definite determiner and the inalienable constructions in French and English. Linguistic Inquiry 23: 595–652. Xu, J. 2004. Possessor raising in Chinese and Korean. Languages in Contrast 5: 245–290.

CHAPTER TWO OBJECT SHIFT IN BELFAST ENGLISH ALISON HENRY 2.1 Introduction Object shift in the mainland Scandinavian languages and Icelandic has been the focus of much research in generative linguistics (see for example Holmberg 1986, 1999; Sells 2001; Fox & Pesetsky 2005; Hosuno 2010). Particular focuses of this research have been on how to capture “Holmberg’s Generalization”, that is, the generalization that object shift only occurs when the verb raises out of the VP, and on how to explain the distinction between object shift which applies to weak pronouns only (as in the Mainland Scandinavian languages), and which is obligatory, and object shift of full DPs, as occurs in Icelandic, which is generally optional. Most of this work has consisted in the reanalysis of well-established data, and little new data, particularly from different languages, has been available to help in assessing the validity of the various theories which have been proposed, which range from the phonological (eg ErteschikShir 2005) to the semantic-pragmatic (Holmberg 1999) to the determination of word order by derivational syntactic principles (Fox & Pesetsky 2005). This paper considers the position of objects in Belfast English imperatives, which at first sight might seem to bear little relation to Object Shift in the Nordic languages, but which, it will be argued here, is essentially the same type of phenomenon. It will consider, in particular, the relevance of Belfast English object shift for the understanding of the difference between varieties with full DP object shift and those where only weak pronouns can shift. Object shift does not generally occur in modern English. However, it is not clear whether this is because English simply does not have this process, or whether it is because English lacks the circumstances necessary for it to apply. Object shift is well known to apply only if the verb moves out of the VP. But it is well known since at least Pollock

Object Shift in Belfast English

25

(1989) that English verbs do not leave the VP, and thus cannot precede negation or adverbs that appear in pre-VP position, and cannot move to C in questions 1 2 3 4

*John likes not coffee. John does not like coffee. *The girl sang not a song. The girl did not sing a song.

5 6 7 8

*John read carefully the book. John read the book carefully. *Bill brings always the books with him. Bill always brings the books with him.

9 10 11 12

*Likes John coffee? Does John like coffee? *Sang the girl a song? Did the girl sing a song?

Although the data above apply to Belfast English also, that dialect does have a structure in which lexical verbs can leave the VP, that is in overt subject imperatives (see Henry 1995: ch. 3). 13 Go you on ahead. 14 Take you a rest. 15 Read you that book. Assuming that the subject is first merged in vP, and the verb first merged in VP, then the verb must have raised out of vP in these structures; to put it simply, since it precedes the position of the subject, and if we are right about where the subject and verb are first merged, and movement, perhaps construed as internal merge, is uniformly leftward, then the verb must, quite simply, have moved out of VP to give this order. It is not important for current purposes whether it has moved to C (as proposed in Henry 1995) or to a lower projection such as T or another head above V (as proposed for example in Rupp 2007), but rather that it raises out of the VP, thus potentially creating the environment for object shift. In Belfast English inverted imperatives, the object can be found to precede adverbs. 16 Make her always you a cup of tea after work. 17 *Make always her you a cup of tea after work.

Chapter Two

26

Unfortunately, one crucial test for the position of the verb and the object in the Nordic languages, i.e. their position relative to negation, cannot be tested in Belfast imperatives. It is generally assumed that the negative element is first merged above vP, either as the head of a NegP above vP, or adjoined to VP. Thus, elements that precede the negative element can be assumed to have left the VP. This is illustrated in the following examples from Holmberg (1999): 18 Jag kysste henne inte. I kissed her not “I did not kiss her.” 19 *Jag kysste I kissed

inte not

henne. her

20 *Jag har henne inte I have her not “I have not kissed her.”

kysst. kissed

21 Jag har I have

henne. her.

inte not

kysst kissed

Unfortunately, this test is not available in Belfast imperatives, because, just as in standard English, the usual pre-VP verbal negative not is not used in overt subject imperatives. Rather, imperatives have an invariant sentence-initial negative element don’t . 22 *Do not you forget your homework. 23 *Do you not forget your homework. 24 Don’t you forget your homework. Note that in Belfast imperatives, one of the main indicators of object shift is that the object precedes the subject. Note that the subject is low, below adverbs such as always. 25 Make always you a good effort. It seems thus that in these imperatives, the subject remains in SpecVP and does not raise to a higher position. We assume that here the subject does not have to raise to SpecTP because imperatives do not project T, so

Object Shift in Belfast English

27

there is no SpecTP position for imperatives to move into. This is rather plausible because although imperatives marginally allow aspect: 26 Be doing your homework when I get back! 27 Have finished your homework before I get back! they never show tense: rather, they always refer, in a sense, to the future.

2.2 Data collection While data collection methods are not often discussed in papers on theoretical syntax, it is important when considering data from nonstandard dialects to give some consideration to how data has been collected. Data collection from speakers of nonstandard dialects is not always straightforward. Fortunately, it is easier in relation to structures which are not highly stigmatized. This is the case with inverted imperatives in general and also with the object shift which appears in these. Speakers do not regard these structures as non-standard or stigmatized, and, for example, teachers do not “correct” their pupils when they use these, unlike for other features of the dialect such as negative concord. The data was collected by informal questioning of speakers as argued for in Henry (2005), asking “Could you say...”, and if a negative response was obtained, by asking “What would you say?” in order to be sure the sentence was not being rejected because of some feature other than the object shift being investigated. Speakers were asked about what they themselves could say, rather than, for example, what could be said in their dialect or what they had heard said, since unreliable intuitions can be gathered from “passive” dialect users, who have heard the structure but do not use it themselves. As we will see below, careful data collection methods are important in order to establish the variety of grammars possible in a speech community.

2.3 Imperative inversion in Belfast English We begin here by briefly introducing Belfast English inverted imperatives in a little more detail. There are two subdialects of Belfast English, termed Dialect A and Dialect B in Henry (1995: ch. 3). We shall be concerned here only with Dialect B. Dialect A only allows verb-subject order in imperatives where the verb is unaccusative, whereas in Dialect B, verbs which take overt objects can show verb-subject order. It is only of course in the latter dialect that object shift is possible, since in Dialect A

28

Chapter Two

there is no verb-subject order with objects. Henry (1995) proposes that in Dialect B, the verb moves to C in imperatives. The important point for our purposes here is that the verb clearly raises out of the VP, if we accept the almost universally accepted verb phrase internal subject analysis of the merge position of subjects.

2.4 Is object shift in Belfast English imperatives similar to the well-known object shift in Scandinavian languages? It will be argued here that the key characteristics of object shift in Belfast English imperatives are similar to that in Scandinavian languages. First, object shift is entirely dependent on verb movement, as Holmberg’s Generalization states for the Scandinavian languages. This is the key characteristic of Scandinavian object shift. The object can only precede the subject or adverbs if the verb has moved; it cannot do so if there is an auxiliary, where presumably the auxiliary is in T or C, and the lexical verb remains in the VP. 28 *Be you it reading when I get back. 29 *Be it you reading when I get back. 30 Be you reading it when I get back. For some speakers, in imperatives with an auxiliary, the lexical verb can move to adjoin to the auxiliary; since the verb has moved, though not in this case because it has alone head-moved to a higher head, the object also shifts. 31 Be giving her you the prize when the press arrives. 32 Have given her you the prize before the press arrives. Secondly, verb movement differs across varieties in whether it is available only to weak object pronouns or to full DPs. In one subdialect, only weak objects can undergo shift, while in another, full DPs can shift. 33 Give her you that book. (grammatical for all speakers) 34 Give your mother you that book (only grammatical for a subset of speakers) Thirdly, in those varieties that have only pronominal object shift, the items targeted are the same as in the Mainland Scandinavian languages

Object Shift in Belfast English

29

where full DPs do not shift, that is, only weak pronouns can shift, and not strong, stressed or conjoined pronouns. 35 36 37 38

Show them you to me. *Show themuns you to me. *Show them you to me. *Show her and them you to me.

Fourthly, even in the subdialect where shift of full DPs is available, the DP must be definite. As in Icelandic, which allows full DPs to shift, indefinite DPs may not shift. 39 40 41 42

Tell the teacher you that. *Tell a teacher you that. Read the book you before next week. *Read a book you before next week.

We assume that here the subject does not raise to SpecTP, arguably because imperatives do not project T, so there is no SpecTP position for imperatives to move into. Note that this is consistent with the fact noted above that imperatives marginally allow aspect but not tense (see examples (26–27)).

2.5 What distinguishes language varieties with full DP object shift from those where only weak pronominals shift? Full DP object shift has been argued to be part of a cluster of properties which differentiate Icelandic from Mainland Scandinavian languages, including V-to-T movement in main and embedded clauses, stylistic fronting and transitive expletives (see for example Hiraiwa 2001). Belfast English lacks at least some of these properties, so it is not generally an “Icelandic-type” language. Thus, lexical verbs do not move out of VP; they do will not raise past negation, do not appear before adverbs and do not move to C in questions; in this respect Belfast English behaves just like standard English 43 44 45 46

*John likes not coffee. John does not like coffee. *The girl sang not a song. The girl did not sing a song.

Chapter Two

30

47 48 49 50

*Likes John coffee? Does John like coffee? *Sang the girl a song? Did the girl sing a song?

Belfast English, like standard English, shows no evidence of stylistic fronting. Stylistic fronting is a process in Icelandic in which an element other than the subject fronts to clause-initial position in a subordinate clause. 51 *They said that yesterday went he to school. 52 *He thought that quickly ran they along the road. A property which might seem to unite Icelandic and Belfast English is the availability of transitive expletives. Henry & Cottell (2007) show that Belfast English, unlike standard English, does have transitive expletives. It is not altogether straightforward to establish whether full DP object shift and transitive expletives coincide in the same speaker grammars. This is because the existence of object shift is dependent on verb movement out of VP in imperatives, which generally occurs in older speakers; without this verb movement, the key known characteristic of object shift—its dependence on the verb moving out of VP, known as Holmberg’s Generalization, does not apply. On the contrary transitive expletives appear to be an innovation and do not occur in the oldest speakers, so it is difficult to establish to what extent these coincide. However, it does seem that for those speakers who have transitive expletives and verb movement in imperatives, full DP object shift is available, indicating that there may be a link between the two. Bernstein & Zannutini (2010) argue that Appalachian English, a variety showing many similarities to Belfast English, has transitive expletives and singular concord because it has an additional projection above T, which they call FP and which hosts person agreement. However, Henry and Cottell (2007) show that in Belfast English, the associate of a transitive expletive is not restricted to a single position below T but rather can appear in a range of positions between T and V. There is no indication in the literature that Appalachian English has either imperative inversion or object shift. Thus, it seems reasonable to conclude that while full DP object shift is available in a subdialect of Belfast English, it does not seem to be part of a parameter which also licenses transitive expletives. Rather, these seem to be unrelated features which do not necessarily co-occur in a language variety.

Object Shift in Belfast English

31

2.6 Why two subdialects? We noted above that for some speakers only weak pronoun objects shift while for others full definite DPs can shift. This is a difference which does not correspond to dialect areas, or to other characteristics such as age or social class of speakers. Moreover and more importantly, it does not seem to necessarily correspond to any other feature in speakers’ grammars (of course speakers’ grammars must have verb movement in imperatives in order to create the environment for object shift in the first place). This suggests that trying to find characteristics that differentiate languages which have full DP object shift, from those where object shift is restricted to weak pronouns are essentially misguided. Rather, UG allows grammars with either type of object shift, and thus we can find, within dialects, either type of grammar. (It should be noted that when we refer to subdialects, we are really referring to differences in one single area of grammar. It is not the case that the “subdialects” are to be construed as entities that differ from one another in a range of areas of grammar). This approach suggests that we need to make a further move away from the idea that differences between languages can be accounted for by large-scale parameters (such as the null-subject parameter for example) and towards a view where UG, though highly constrained, nevertheless leaves open a limited range of possibilities, any of which we may find in a given grammar. Object shift is only available where verb movement takes place, and it is only available to definite DPs; moreover, it is to be noted, at least as far as we have observed, that it cannot apply to only full DPs and not pronouns; no such language or dialect, where full definite DPs shift but not pronouns, has been found. Rather, it can either apply only to weak object pronouns, or to weak object pronouns plus all other definite DPs. We have identified other aspects of variation of a similar type in Belfast English, and have no reason to doubt that such examples could be found elsewhere if difference, rather than uniformity, between speakers’ grammars were sought out. Thus, a similar phenomenon emerges when one considers negative concord in Belfast English. it is well known that languages with negative concord fall into two distinct types (see for example Giannakidou 2000). In one type, exemplified by Spanish, postverbal n-words must be accompanied by a verbal negator, but preverbal n-words must not be. 53 No veo a nadie. Not I-see particle nobody “I do not see anybody.”

Chapter Two

32

54 *Nadie no Nobody not “Nobody helps me.”

me me

55 Nadie me Nobody me “Nobody helps me.”

ayuda. helps

ayuda. helps

In another type, exemplified by the formal register of French, preverbal and postverbal n-words are both accompanied by a verbal negator. 56 Je ne vois personne. I NEG see nobody “I don’t see anybody.” 57 Personne ne Nobody NEG “Nobody helps me.”

m’ me

aide. helps

Now there are two subdialects of Belfast English: one behaves like Spanish, disallowing a verbal negator after an n-word. 58 I can’t see nobody. 59 *Nobody wouldn’t tell us. The other behaves like French, and in this both of the sentences above are grammatical. It is important to note that such differences may not show up if standard methodologies are used; often, grammaticality judgements are sought from groups (often class groups of linguistics students) rather than individuals. Apart from leading to sheep-like behaviour among the more reticent, who may follow the majority view rather than giving their own judgements, if the investigator is not inclined to look for difference, but instead expects to find a dialect with a single grammar, they may overlook or dismiss different judgements being expressed. Seeking to account for the grammars of individuals, however, which is of course the avowed goal of work in generative grammar but which in practice in general is pursued as some kind of generalization across many individual grammars, may lead to the finding that grammars show more evidence of UG-constrained variation than one would expect. This is a way of investigating the limits of variation. It appears to show that the limited range of options made

Object Shift in Belfast English

33

available by UG are more widely distributed than generally envisaged, and certainly that the earlier ideas of large and wide-ranging parameters – such as those envisaged to differentiate languages which have full DP object shift, stylistic fronting, V-to-T movement, and transitive expletives, like Icelandic, from those that have weak pronoun object shift, but lack stylistic fronting, V-to-T movement and transitive expletives like Danish, Swedish or Norwegian—do not really exist but may appear to do so through historical accident. Another important point to note is the subset relationship between grammars with full DP object shift, and those with only pronominal shift. The latter grammar is a subset of the former, since it allows more elements to shift. It is very noticeable, even though a full quantitative study is beyond the scope of the present paper, that many fewer speakers have the grammar allowing full DP shift, which is rather rare, appearing in the grammars of only 8 of the 85 speakers we studied. Similarly, French-type negative concord, where the order n-word followed by verbal negator is grammatical as well as the order verbal negator followed by n-word, is rare. Here again we have a subset relationship, with the Spanish-type negative concord allowing negative concord in a subset of cases where the French type allows it. This suggests that the subset principle may indeed be operative in grammars and in acquisition. This principle, which has received little attention recently, was first proposed by Wexler and Manzini (1987) in relation to short-distance versus long-distance anaphora. They proposed that settings of a parameter were necessarily in a subset relationship, and that learners would adopt the “smallest” grammar in the first instance, moving to a superset grammar only in the presence of positive evidence; this, they argued, offered a solution to the problem of how parameters are set by learners. Now if some learners in a community hear sufficient input indicative of a superset grammar (one with full DP object shift or French-type negative concord) to make them reset their grammars to a superset setting, while many speakers do not have sufficient input to make them reset their grammars, this would offer an explanation for why superset grammars are less widely distributed in the community than subset grammars, but nevertheless exist; some speakers, presumably at a critical time in acquisition, have enough input to make them adopt a superset grammar. This might also lead us to predict that generally, languages with the superset grammar (for example Icelandic in the case of object shift, or formal French in the case of negative concord) would be rarer and more vulnerable to change than those, like the mainland Scandinavian languages or Romance languages other than French, This indeed seems to be the

34

Chapter Two

case, with the verbal negator disappearing in informal French. We might also expect some varieties of a language which generally has a subset grammar, such as the Mainland Scandinavian languages in relation to object shift, to have a small number of speakers who allow full DP object shift among a majority who only have the subset grammar, and this indeed appears to be the case for some few varieties of Norwegian. Note that we only discovered the existence of full DP object shift in Belfast English by seeking out differences in speakers’ intuitions, and thus grammars, rather than trying to account only for the majority grammar of the dialect. It seems important, then, that we should seek to understand the type of options left available by UG by actively seeking out the differences between very closely related dialects and indeed idiolects.

References Bernstein, J. and R. Zanuttini. 2010. What is verbal -s in English? Ms., William Patterson University Erteschik-Shir, N. 2005. Sound patterns of syntax: object shift. Theoretical Linguistics 31: 470–493. Fox, D. and D. Pesetsky. 2005. Cyclic linearization of syntactic structure. Ms., MIT Giannakidou, A. 2000, Negative... concord? Natural Language and Linguistic Theory 18: 457–523 Henry, A. 1995. Belfast English and Standard English. Oxford: OUP. —. 2005. Non-standard dialects and linguistic data. Lingua 115(11). 1599–1617. Henry, A. and S. Cottell. 2007. A new approach to transitive expletives: evidence from Belfast English. English Language and Linguistics 11(2). 279–299. Hiraiwa, K. 2001. EPP and object shift in Scandinavian. In Proceedings of the 20th West Coast Conference in Formal Linguistics, 290-303. Somerville, MA: Cascadilla Press. Holmberg, A. 1986. Word order and syntactic features in the Scandinavian languages and English. PhD diss., University of Stockholm. —. 1999. Remarks on Holmberg’s Generalization. Studia Linguistica 53: 1–39. Hosuno, M. 2010. Scandinavian Object Shift as the cause of downstep. Working Papers in Scandinavian Syntax 85: 1–36. Pollock, J.-Y. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20: 365-424.

Object Shift in Belfast English

35

Rupp, L. 2007. “Inverted” imperatives. In W.van der Wurff (ed.), Imperative clauses in generative syntax, 297–322. Amsterdam: John Benjamins Sells, P. 2001. Structure, alignment and optimality in Swedish. Stanford: CSLI Publications. Wexler. K. and M. R. Manzini. 1987. Parameters and learnability in binding Theory. In T. Roeper and E. Williams (eds.), Parameter setting, 41–76. Dordrecht: Reidel.

CHAPTER THREE DIACHRONIC AND SYNCHRONIC VARIATION IN THE AUXILIARY SELECTION OF EASTERN ABRUZZESE FRANCESCA LA MORGIA 3.1 Introduction1 This paper aims to provide some insight into the evolution of auxiliary selection in Eastern Abruzzese from the end of the 19th century until the present day, by presenting results from the analysis of a written corpus and of data elicited from speakers of this dialect. Eastern Abruzzese displays a person driven type of auxiliary selection, with the third person singular and plural selecting avé (have) and the other persons esse (be). This study examines the variation and change in the third person singular auxiliaries used to form the present perfect and evaluates factors such as the contact with neighbouring dialects, the influence of Italian and the existence of different types of linguistic registers which may account for variability in the selection of auxiliaries. The hypothesis proposed is that some verbs are more resistant to contact-driven change than others. As discussed in this paper, the reasons may be sought in language-internal as well as languageexternal causes. The paper is organized as follows: the main features of auxiliary selection in Eastern Abruzzese are reviewed; then the results from the analysis of the written corpus and the elicitation tasks are presented and discussed.

1

I wish to thank Emiliano Giancristofaro for the insightful discussions and for kindly granting me access to his library and Jacqueline Laws for her helpful comments.

Auxiliary Selection of Eastern Abruzzese

37

3.2 Auxiliary selection in Eastern Abruzzese The dialects of Italy exhibit various patterns in the selection of auxiliaries in the present perfect. The dialects spoken in Abruzzo display a pattern of auxiliary selection which differs from that of Italian. In Italian the selection of the auxiliary in the formation of the present perfect and other compound tenses is determined by the type of verb: transitive and unergative verbs select avere, while unaccusative verbs select essere. This split intransitivity is governed by the argument realization patterns of the verbs and has been attributed to a gradient semantic interpretation of the verbs (Sorace 2000, 2004). In Eastern Abruzzese and in other dialects spoken in Abruzzo the selection of auxiliaries in the present perfect depends on the person, as shown in 3.1. Other compound tenses may display different patterns. For example, in Eastern Abruzzese the past perfect only selects avè irrespectively of person, number or argument realization of the verb, as shown in examples 1-32. 1 Li citl’ avé’ rijite The children have.PAST return.PART “The children had returned home”

a la cas’. to the house

2 Sorem’ s’avé’ ‘ngullat’ ‘nu guaj’. My sister pro.have.PAST put on.PART a problem “My sister had caused herself a problem” 3 Nu’ avam’ magnat’ naprese. We have.PAST eat.PART a lot “We had eaten a lot” The existence of a person driven type of auxiliary selection has been explained in terms of sentence structure (Manzini and Savoia 2005) and hierarchy of verb persons (Tuttle 1986). The distinction between esse in the first and second person and avé in the third would reflect the fact that the first two persons convey a meaning of inclusion and empathy (Cordin 2009), which contrast with the role of the third person, which has a status of “non-person” (Benveniste 1946).

2 Unless stated otherwise, the spelling adopted in this article is based on the Vocabolario dell’uso abruzzese (1893).

Chapter Three

38

The three Abruzzese varieties classified by Giammarco (1979), exhibit three different patterns of auxiliary selection, which converge only in the second and third person singular (Table 3.13). Table 3.1. Auxiliary selection in Aquilano, Eastern and Western Abruzzese Aquilano Aquilano st

Esse

Abruzzese Adriatico Eastern Abruzzese esse

nd

Esse

esse

esse

rd

avé

avé

avé

st

esse

esse

avé

nd

esse

esse

avé

rd

esse

avé

avé

1 sing. 2 sing. 3 sing. 1 plur. 2 plur. 3 plur.

Abruzzese Occidentale Western Abruzzese avé

As Giammarco notes (1973: 163), the patterns shown in Table 3.1 are representative of the general trend in auxiliary selection in each dialect group, but they do not account for the variation existing within the dialect subgroups. As the examples from Lancianese (an Eastern Abruzzese dialect) show, some persons may select both auxiliaries (Table 3.2). This optionality was described by Giammarco as part of “the dynamic phases of conflict”, phases in which each dialect variety established an individual pattern of auxiliary selection. Lancianese selected esse in the first and second person singular, following other Eastern dialects, with the second and third person plural adapting to the first two. Another typical feature of Lancianese is the use of avé in the first person singular of unaccusative and reflexive verbs, probably as a result of the influence of western varieties (Giammarco 1973).

3

Tables 3.1 and 3.2 are adapted from Giammarco (1973, 1979).

Auxiliary Selection of Eastern Abruzzese

39

Table 3.2. Auxiliary selection in Lancianese st

magná” ‘to eat’

jiì” ‘to go’

1 sing.

sò magnat’ (esse)

ajjԥ jit’ (avé)

2 sing.

si,sci magnat’ (esse)

si, sci jit’(esse)

3 sing.

a jit’ (avé) avamԥ jit’ (avé)

2 plur.

a magnat’ (avé) avamԥ, sèmԥ magnat’ (avé, esse) avate, sète magnat’ (avé, esse)

3 plur.

a magnat’ (avé)

a jit’ (avé)

nd rd st

1 plur. nd rd

séte jit’ (esse)

The examples from Table 3.2 demonstrate that the second and third person singular are resistant to the acceptance of both auxiliaries4. However, the picture presented by Giammarco is not complete, as this study will demonstrate. The analysis of the examples from Finamore’s dictionary and other texts written in Lancianese shows that the second person has kept a consistent selection of esse, while the third person singular has started to move towards an alternation of the two auxiliaries. This alternation is already evident in the tales transcribed by Finamore5 (1882) and in the examples from his dictionary of use (4–7), which is based on the dialect spoken in Lanciano (Finamore 1893). 4 Ajje sendut’ have.PRES.1SG hear.PART “I have heard” 5 So vist’ be.PRES.1SG see.PART “I have seen” 6 Ha remast” have.PRES.3SG remain.PART “he/she has remained”

4

It also needs to be noted that the 3rd person plural resembles the third singular. The Novelle Popolari Abruzzesi were collected and transcribed by Finamore with the aim to document popular traditions as well as the different dialects spoken throughout the region.

5

40

Chapter Three

7 È menut’ be.PRES.3SG come.PART “he/she has come” Person driven auxiliary selection has been documented in other Italian dialects, such as Maceratese (Marche), the dialect of Altamura (Puglia) and some varieties spoken in Trentino (Paciaroni 2009, Loporcaro 1988, Cordin 2009). Like in Lancianese, also in Maceratese pronominal verbs select either essere or avere in the third person. 8 Marì s a laato le ma Mary herself HAVE.PRES.3SG. wash.PART the hands 9 Marì s ũ llaata le ma Mary herself BE.PRES.3SG. wash.PART the hands “Mary has washed her hands” (Paciaroni 2009: 37) In the dialect of Altamura (Puglia) transitive and unergative verbs select either have or be in the 3rd person singular, and have in the 3rd plural, while unaccusative, reflexive and pronominal verbs select be in the 3rd person singular and either be or have in the 3rd person plural. Even though this pattern is the most frequently attested, it seems that many informants tend to use also have in the third person plural with unaccusative verbs (Loporcaro 1988: 309). The examples from the dialect of Altamura are a further demonstration that in some dialects speakers find both auxiliaries acceptable with some verbs. As it has been shown, person driven auxiliary selection is common to several dialects. The distinction in auxiliary selection in the three different Abruzzese dialect groups does not fully represent the use across several dialects spoken within each group. It is reasonable to assume that in the phase of transition from late Latin the contact between different dialects resulted in the interchangeable use of auxiliaries, with the exception of the second and third person singular, which in all Abruzzese dialects respectively select esse and avé. This complex picture has been further enriched by the influence of the Italian language in the late 19th century. This study aims to discover whether some verbs were subject to change more than others and whether the analysis of a written corpus and of elicited data from speakers of an Eastern Abruzzese variety can reveal insights in the process of change during a transitory phase in which both auxiliaries are becoming acceptable.

Auxiliary Selection of Eastern Abruzzese

41

3.3 Diachronic analysis of auxiliary selection in EA This study combines two methodologies with the aim to document the selection of auxiliaries in the present perfect over time. The first is a comparative analysis of different genres of written documents. The second is a real time study, based on a translation task and a grammaticality judgement task.

3.3.1 The written corpus The diachronic investigation of the use of auxiliaries in EA is based on texts which vary in genre and date of publication, with the aim to have a rich and various corpus. The corpus includes poems, plays, short stories, songs and a vocabulary of use (appendix A). All the authors are speakers of Eastern Abruzzese varieties from the province of Chieti (Casalbordino, Lanciano, Guardiagrele, San Vito). In each text, all instances of present perfect were isolated and classified according to person, number and combination of auxiliary and main verb type. A total of 500 present perfect tokens were found. As expected, all transitive and unergative verbs consistently select avere (have), while more variability in auxiliary selection is exhibited by unaccusative, pronominal and reflexive verbs in the third person singular. The analysis of the written corpus is therefore narrowed down to those verbs, for a total of 140 tokens. The first finding is that the more the text conforms to literary standards, the less it resembles spoken language, and this is also reflected in the choice of auxiliary. For example, in De Titta’s play (1920) we find a consistent selection of esse with unaccusative verbs, while we find avé in texts such as children’s rymes, songs, and tales, which more closely reflect the spoken language (Lupinetti 1983, Finamore 1882). The examples which more closely resemble the spoken language come from the transcriptions of rhymes (Lupinetti 19836). The rhymes7 in examples 10– 13 show a consistent use of avé in the third person with unaccusative and pronominal verbs. 10 s’á ‘ncagnate la ciuvètte Pi na cose che j’aje dètte J’aje dètte si vvô lu marite La ciuvètte s’á mèsse a ride 6

The malicious woman got irritated for something I have said I asked her if she wants to get married the malicious woman laughed

The data collected by Lupinetti over more than 50 years was published in different journals and in many volumes of the Rivista Abruzzese. 7 Transcription system used by Lupinetti.

42

Chapter Three

11 chi s’á magnate lu casce? - La hátte da duv’ á ‘ndráte? - da la buscèlle dov’ á-riscite? - da la finistrèlle […]

Who has eaten the cheese? - The cat where has it entered from? -from the hole where has it come out from? -from the window

12 si n’á minute l’ore, o cara spose: è ll’ore di partì da la tua case.

If the time has not come, o dear bride: it”s time to leave your house

13 l’amore mìje si n’á jìte a mmète […] My love has gone to harvest. While the choice of style and register may affect the selection of auxiliaries, it is also important to consider the difference between urban and rural dialects. Lupinetti (1983) distinguishes between a forma volgare, which is used by farmers and among family members and friends, and a forma civile, a formal register used generally by the middle class. The use of volgare or civile therefore depends on the situation, the social class and the level of instruction of the speakers. The presence of separate registers is reflected in the use of auxiliaries and this emerges clearly in the different texts which constitute the corpus analysed in this study. The authors who use a formal register or conform to the literary standards of their time are subject to the influence of the national language. The texts from the 19th century and early 20th century which represent transcriptions of novels, songs, and rymes which have been transmitted orally reflect the way the language was spoken at the time, and these show that avé is the preferred auxiliary in combination with any verb type. This clear distinction which reflects the use of different registers has been lost in contemporary EA. In more recent writings we find esse and avé used interchangeably. For example, in Rosato’s (2009) poetry, we find examples such as è ite and á ite ‘has gone’. Interestingly, if we compare the use of auxiliaries in diachrony we can observe a progressive increase in the use of avé. In Finamore’s novels (1882) avé is used with 6% of unaccusative, reflexive and pronominal verbs. The percentage increases to 47% in the poetry of Della Porta (1933) and to 63% in that of Rosato (2009). We would expect writers to conform more and more to standard Italian, however we see a shift in the opposite direction. This shift can be attributed to the increase in awareness of the distinction between the local dialect and the national language, and probably in a renewed interest in

Auxiliary Selection of Eastern Abruzzese

43

reproducing the dialect as it is spoken rather than creating an idealized written standard. The results which emerge from the analysis of the written data demonstrate that the selection of auxiliaries is dictated by the choice of register and by the distinction between a rural and a more refined variety. The texts which reflect more closely the spoken dialect consistently show the selection of avé in the third person singular, irrespectively of the verb type.

3.3.2 The experimental tasks In order to gather data on the current use of auxiliaries and to control for variables such as verb type and word frequency, 58 speakers of Lancianese residing in the urban area in the town of Lanciano were asked to participate in two tasks. Of the 58 participants, 17 were males and 41 females, with ages ranging between 8 and 81. Each age group (ie.10-19; 20-29) included at least 5 individuals and there was an equal distribution of participants under and over the age of 50. Participants were presented first with a translation task (henceforth TT) and subsequently with a grammaticality judgement (henceforth GJ). The TT required participants to translate 10 sentences from Italian to Lancianese. The sentences included 8 unaccusative verbs (jì’ ‘to go’) and 2 with dative pronominal verbs (farse mal’ ‘to hurt oneself’). The responses from the TT showed that the general distribution of auxiliaries follows the original pattern, with avé being used with 82.9% of verbs. The verbs which were more commonly combined with esse were stá, and jì’. It is possible to hypothesize a gradience in the selection of auxiliaries, which may follow the Italian pattern. However, as shown in Table 3.3, there is no correspondence between verb category and the trend in auxiliary selection. If we examine the individual responses, we find that only 8 participants selected avé in every sentence and nobody used only esse. Gender and age proved not to be significant variables overall. However, it is interesting to note that the age of the 8 participants who only use avé ranges between 19 and 66, and that those participants who are aged 70 and over use esse in 70-90% of sentences. Therefore, it emerges from this analysis that only a small percentage of speakers uses only avé, and none use only esse, and there is no significant correspondence between argument structure or verb category and auxiliary selection.

Chapter Three

44

Fig. 3.1. Auxiliary selection in the translation task – percentage use of avé and esse

100 90 80 70 60 50 40 30 20 10 0

avé esse

Table 3.3. Verb categories, in order of percentage of use of avé Verb CALAT’ ‘descend’ FATT’ MAL’ ‘hurt’ MENUT’ ‘come’ FIRMAT’ ‘stop’

Category Change of location Uncontrolled process Change of location Controlled processes

NAT’

‘born’

Change of state

‘NDES’ MORT’ REMAST’ STAT’ JIT’

‘feel’ ‘die’ ‘remain’ ‘be’ ‘go’

Uncontrolled processes Change of state Continuation of condition Existence of state Change of location

Another task was presented to the same participants in order to determine which auxiliaries were considered acceptable with different verbs, some of which are the same as used in the TT. This type of study is complementary to the TT, because it allows us to draw more firm conclusions over the issue of the current usage and acceptability of the two auxiliaries. Participants were not aware of the purpose of the task, and were asked to determine whether the sentence was correct or incorrect, and to rephrase

Auxiliary Selection of Eastern Abruzzese

45

the incorrect sentences. The 16 sentences were constructed using terms that are likely to be known by all speakers, and were read by a native speaker from Lanciano8. As shown in appendix B, each sentence includes at least one verb in the present perfect for a total of 19 verbs in the third person singular. Nine of the sentences include verbs that were used in the translation task. The verb jì’ ‘go’ appears 4 times, 3 with esse and 1 with avé. The verb menì ‘come’ appears 3 times, two with avé and one with esse. The verb calá ‘descend’ appears twice, once with esse and once with avé. There is a mixture of transitive, unaccusative and pronominal verbs. The choice of verbs was based on frequency and on the occurrence of ambiguity in examples from the written corpus. Sentences 1 and 16 were accepted by all participants. Sentence 16 includes a transitive verb followed by a direct object, therefore the auxiliary avé is accepted by all, as expected. However, sentence 1 includes the verb calá ‘descend’, which in this sentence is intransitive9. Both the judgements on sentences 1 and 4 show a significant trend which matches the results from the TT. The more consistent responses from the GJ came from the older population. Participants over 70, who used mostly esse, also considered the sentences with avé incorrect, and rephrased them using esse. However, the responses from the younger participants were not as consistent and they often contradicted the responses given in the first task. For example, almost 50% of participants selected esse with rimanè ‘remain’ in the translation task, but 77% accepted avé; moreover, 57% of participants who used esse in the translation task accepted avé in the grammaticality judgement. Another interesting finding emerges from the analysis the verb jì’ ‘go’. Out of the 60% of participants who used esse in the TT, 50% selected avé in the GJ. 20% of participants used esse in the translation and corrected sentence 15 using avé. This inconsistency also emerges in the results from the analysis of the verbs nasce ‘to be born’ and murì’/murirse ‘to die’. As shown in Table 3.4, avé is preferred over esse with both verbs. However, not all the participants who use esse in the translation task, correct á nate with è nate. It is interesting to note that almost 70% of participants accept á nate and s’è morte and almost 30% of participants use a different auxiliary with the same verb in the two tasks. 8 The native speaker is an adult who acquired Lancianese as an L1. Participants can hear the sentence, but they cannot read it. The rationale behind this choice lies in the intention to replicate a “listening environment”, which is likely to bring out the participants’ perception of what “sounds right/wrong”. 9 Calá can be used as a transitive, meaning ‘to bring down’.

46

Chapter Three

Table 3.4. Comparison of results for verbs nasce and murì NASCE Translation Task Esse: 34.6% Avé: 65.4% Grammaticality Judgement (sentence 10) Á nat’ 27% correct á nat’ with è nat’

MURì’ - MURIRSE Translation Task Esse: 46.1% Avé: 53.9% Grammaticality Judgement (sentence 11) S’è mort’ 30.3% correct s’è mort’ with s’á mort’

An example which further confirms the existence of inconsistencies in the choice of auxiliary is that of the verb minì ‘come’. In the TT 20% of participants select esse. The GJ presents two sentences, one with esse (sentence 9) and one with avere (sentence 14) and 40% of participants accept both.

3.4 Discussion This study aimed to observe the patterns of auxiliary selection in Eastern Abruzzese, examining a written corpus of literary texts and administering two controlled tasks. The first finding is that both avé and esse are currently acceptable auxiliaries in combination with unaccusative, reflexive and pronominal verbs in the third person singular. Another important finding is based on the written data. The texts written in the 19th and first half of 20th century show a consistent use of esse with these verbs, similar to standard Italian. It is possible to assume that Italian (Tuscan) set the standards of a literary model of language, and that those who were writing in their native Abruzzese dialect were using a formal register, which was dissimilar from the spoken language. It needs to be noted that many of those authors were teachers or writers who had good command of Italian. Therefore, we can assume that the influence of the literary model has not affected the lexicon, which is strictly conforming to the original, but has started to emerge in the syntax and morphology. The texts which reflect more closely the spoken language are songs, rhymes and tales. These texts, which have been collected by interviewing speakers from different areas in the East of Abruzzo, show a consistent use of avé. In studying the evolution in the use of auxiliaries, we need to account for the strong influence of the national language, which has a different typology of

Auxiliary Selection of Eastern Abruzzese

47

auxiliary selection, based on the argument structure of the verb, as well as the influence of the neighbouring dialects. As Giammarco had initially claimed (1973), Lancianese presents optionality in auxiliary selection in the 1st person singular and plural, but not in the other three persons. However, the analysis of the written corpus shows that while the 2nd singular and the 3rd plural are more conservative, the 3rd singular is following the optionality system of the other 3 persons. The analysis of the data elicited from speakers of Lancianese shows that the auxiliary avé is still used more than esse in the third person even with verbs which denote change of state, such as jì’ and menì, which would select essere in Italian. The fact that many speakers in the age range of 20-50 prefer avé while speakers over the age of 70 almost solely choose esse can be an indication of the influence of schooling, which brought about the perception that speaking and writing in the local dialect was “wrong”. Moreover, as noted by Lupinetti, there are further distinctions in the use of different varieties of the same dialect, which reflect the social status. It is possible that the forma civile of the dialect, spoken by urban middle class people and used for public occasions may have started to accept more freely the introduction of the Italian type of auxiliary selection. The choice of auxiliary could, among other linguistic features, characterize a more “polished” dialect, and become a marker which distinguishes urban and rural dialect. At the same time, the behaviour of the younger speakers, who tend to select avé more than older ones, can be a result of the progressive loss of the knowledge of the dialect. We can assume that the existence of a distinction between volgare/civile or urban/rural may have been lost since Italian (or its regional variety) has become the first language of the majority especially in urban areas, and that each speaker may have been exposed to a different variety or to a mixture of more than one variety. The hypothesis of the contact with a non-urban type of dialect which has become confounded with the urban dialect and the hypothesis of the influence of schooling (not to mention of the prestige of national language), are external variables which certainly play an important role. However, we still need to explain why some verbs such as calá select avé significantly more than verbs such as rimanè. Since it appears that the type of verb does not have an influence, the answer to this question may lay in a phonological phenomenon, namely “raddoppiamento fonosintattico” (phonosyntactic doubling), which results in the doubling of the initial consonant of the verb which follows the auxiliary è.

Chapter Three

48

As shown in Table 3.5, the third person auxiliary è triggers this phenomenon with all verbs, except those starting with vowels, consonant clusters or the alveolar trill r. Table 3.5. Combination of past participle with avé and esse avé

esse

avé

esse

á calat’

è ccalat’

s’á mort’

s’è mmort

s’ á fatt’ mal’

s’è ffatt’ mal’

s’ á ‘ndes’

s’ á firmat’

s’è ffermat’

á it’

s’è ‘ndes’ (s’è sendit’10) é it’

á minut ‘

é mminut’

á stat’

è stat’

á nat’

è nnat’

á rimast’

è rimast’

This phonological difference may affect the perception of speakers and this may be reflected in auxiliary selection. If we compare the examples below, we see that in sentence 14a the raddoppiamento of the voiced velar introduces an additional phoneme, making the difference between 14a and 14b more salient than that between 15a and 15b. 14 a 14 b

è calat’ á calat’

/’ũ/ /kkş’lştԥ/ /’ş/ /kş’lştԥ/

15 a 15 b

è jt á jt

/’ũ/ /’it/ /’ş/ /’it/

It is possible that the raddoppiamento, by making the difference in pronunciation more salient, also aids the selection of auxiliary. This explanation would account for the consistent use of á calat’. This verb is the one that obtained the widest agreement both in the translation task and in the GJ. All speakers perceived é calat’ as incorrect. Firstly, this can be related to frequency of the two words in combination with each other. It is likely that the frequency of use of the combination á calat makes the doubling è calat sound not familiar and therefore “sound wrong” to the speakers. 10

In the TT the Italian participle sentito was translated as sendut, sendit and ‘ndes’.

Auxiliary Selection of Eastern Abruzzese

49

It is possible to assume that speakers accept both auxiliaries with the same verb more easily if there is no raddoppiamento. Therefore, the change is likely to be more rapid for these verbs. This hypothesis, which was not envisaged at the initial stages of this study, would need to be further tested to be validated.

3.5 Conclusion This study aimed to contribute to the understanding of the process of change in a dialect which exhibits a person driven type of auxiliary selection. The main hypothesis is that there are language-internal as well as language external forces which drove the change. The language external force is the influence of the neighbouring dialects and more importantly of the national language. While these are significant influences, the change cannot be fully explained without accounting for language internal factors. In the case examined in this study, the third person auxiliary è triggers raddoppiamento fonosintattico (phonosyntactic doubling), a phenomenon which causes the germination of the initial consonant of the past participle verb. This phenomenon occurs with all verbs, except those starting with vowels, consonant clusters or the alveolar trill r. The steps in the change can be thus summarized as follows: originally, speakers of Eastern Abruzzese dialects would have selected avè in the third person irrespectively of verb type. The influence of Italian drove speakers to start accepting esse (as well as avè) with unaccusative, reflexive and pronominal verbs. However, some verbs were more resistant to this influence than others. This resistance is not rooted in the argument realization or the meaning of the verb, but in a phenomenon which causes the gemination of the initial consonant of the past participle verb when preceded by esse. Raddoppiamento makes the distinction between auxiliaries more salient, and therefore when this phenomenon is not triggered, speakers have more difficulty choosing the “right” auxiliary. This ambiguity could have been the initial driving force in the acceptability of both auxiliaries. This theory needs to be further validated, and future research will need to investigate the factors which drove the change in all persons across the three dialect groups. Moreover, the analysis of spoken corpora is needed to complement the findings from this study.

50

Chapter Three

Appendix A Texts included in the written corpus De Titta, C. 1919. Canzoni Abruzzesi. Carabba editore. —. 1920. La scuncòrdie. Carabba editore. Della Porta, M. 1933. Ta-Pu. Lu trombone d’accumpagnamente. Carabba editore. Fagiani, A. 1933. Lu done. Mancini. Finamore, G. 1885. Novelle popolari abruzzesi. Carabba editore. —. 1893. Vocabolario dell’uso abruzzese. Tipografia dello stabilimento S. Lapi. Iannucci, A. 2006. Lu decotte e Il ragazzo, il poeta e le sue donne. Esperienze teatrali. Punto d’incontro editore. Lupinetti, D. 1983. Tradizioni socio-letterarie-melodiche dell’Abruzzo. Rivista Abruzzese, 36 (1). Rosato, G. 2009. Lu scure che s’attonne. Versi in dialetto abruzzese, 1990–2007. Raffaelli Editore. Sigismondi, G. 1966. Da cente e cente vocche... Di Paolo Editore. Sigismondi, V. (ed.) 1991. Canzune nustre. Canti popolari abruzzesi di Giulio Sigismondi. Lanciano: Mancini..

Appendix B Sentences employed for the grammaticality judgements with percentages of rejection Sentence 1. Á calat' da la mundagn. Have.PRES.3SG descend.PART from the mountain “He-she has descended from the mountain” 2. Fijme á jit' a Langian a ccattá 'na cose. My son have.PRES.3SG go.PART to Lanciano to buy.INF one thing “My son has gone to Lanciano to buy something”

Auxiliary % rejection Avé 0 Avé 36.5

3. Lu citile s' è ' mbuss' nghe l' acque. Esse The child refl.PRO be.PRES.3SG wet.PART with the (pronominal) water 30.7 “The child got wet with water”

Auxiliary Selection of Eastern Abruzzese

4. Li so chiamat' ma n” è calat'. Him be.PRES.1SG call.PART but not be.PRES.3SG descend.PART “I called him, but he didn’t come down” 5. Á stat' cundende cá la fijje s” á spusat'. have.PRES.3SG be.PART happy that the daughter PRO have.PRES.3SG marry.PART “He-she was happy that his-her daughter got married” 6. Se n' è jit' nghe la machine. Himself PRO be.PRES.3SG go.PART with the car “He/she has left in the car” 7. Marije á remast'a lu mare. Mary have.PRES.3SG remain.PART at the sea “Mary remained at the beach” 8. L' so' chiamat' ma ngi á vulut menì. Him be.PRES.1SG call.PART but not+PRO have.PRES.3SG want.PART come.INF “I called him but he didn’t want to come” 9. Giuwanne è menut' a la casa me. John be.PRES.3SG come.PART at the house my “John has come to my house” 10. Chillè á nat' a Castellenove. She have.PRES.3SG born.PART in Castellenove “She was born in Castellenove (Castelnuovo)” 11. Marije s' è mort'. Mary PRO+be.PRES.3SG die.PART “Mary died” 12.Ci ha stat’ ‘nu probleme. Expl. have.PRES.3SG be.PART a problem “There has been a problem” 13. Á finite a magná e s’ha ‘ndes’ mal. Have.PRES.3SG finish.PART to eat.INF and PRO+have.PRES.3SG feel.PART bad “He/she finished to eat and he felt sick”

51

Esse 90.3 Avé, Avé 30.7 - 17.3

Esse (pronominal) 21.15 Avé 26.9 Avé 1.9 Esse 36.5 Avé 26.9 Esse (pronominal) 34.6 Avé 51.9 Avé, Avé (pronominal) 0 - 43.3

52

Chapter Three

14. Chillè ha menut’ e se n’ é jit’. She have.PRES.3SG come.PART and PRO+PRO be.PRES.3SG go.PART “She has come and gone” 15. È jit’ a ‘ccattá lu furmagg’. be.PRES.3SG go.PART to buy.INF the cheese “He/she has gone to buy cheese” 16. Giuwanne ha savt’ le scale. John have.PRES.3SG climb.PART the stairs “John has climbed the stairs”

Avé, esse 13.4 - 19.2 Esse 21.1 Avé 0

References Benveniste, E. 1946. Structure des relations de personne dans le verbe. Bulletin de la Société de Linguistique de Paris 42: 1–12. Cordin, P. 2009. Gli ausiliari essere e avere nell’italiano regionale trentino. In A. Cardinaletti and N. Munaro, Italiano, italiani regionali e dialetti. Milano: Franco Angeli. Giammarco, E. 1973. Selezione del verbo ausiliare nei paradigmi dei tempi composti Abruzzo 11: 152–178. —. 1979. Abruzzo. Pisa: Pacini Editore. Loporcaro, M. 1988. Grammatica storica del dialetto di Altamura. Pisa: Giardini. Lupinetti, D. 1983. Tradizioni socio-letterarie-melodiche dell’Abruzzo. Rivista Abruzzese 36 (1). Manzini, R. and L. Savoi. 2005. I dialetti italiani e romanci. Morfosintassi generativa. Alessandria: Dell’Orso. Paciaroni, T. 2009. Sull’ausiliazione perfettiva in maceratese. In B. Moretti, E.M. Pandolfi and M. Casoni, Linguisti in contatto. Ricerche di linguistica italiana in Svizzera. Atti del Convegno dell’Osservatorio linguistico della Svizzera italiana (OLSI), Bellinzona, 16-17 November 2007: 37–58. Sorace, A. 2000. Gradients in auxiliary selection with intransitive verbs. Language 76: 859–890. —. 2004. Gradience at the lexicon-syntax interface: evidence from auxiliary selection and implications for unaccusativity. In A. Alexiadou, E. Anagnostopoulou and M. Everaert, The Unaccusativity puzzle. Oxford: Oxford University Press. Tuttle, E. F. 1986. The spread of ESSE as Universal Auxiliary in Central Italo-Romance. Medioevo Romanzo11: 229-287.

CHAPTER FOUR THE SYNTAX OF THE AFTER PERFECT IN HIBERNO-ENGLISH MARIACHIARA BERIZZI AND SILVIA ROSSI 4.1 Introduction1 In this paper we take into consideration a very well-known aspectual construction of Hiberno English (henceforth HE), the so-called AfterPerfect (AFP; Filppula 1999). This construction, which is not attested in other varieties of English, involves a tensed form of the verb to be; the preposition after and a lexical verb in the -ing form plus its arguments: 1 She's after breaking the window (DHE, s. v. After, p. 3) The AFP conveys a specific perfective aspect variously defined as “hot news perfect” (Harris 1984), “retrospective aspect” (Cinque 1999) or “perfect of recent past” (Cottell 2003). The linguistic literature on HE considers the AFP an equivalent of Standard English (SE) have just + past 1

We would like to thank our patient informants: Paddy Comber, Patrick Connolly and Jennifer Sullivan. Thanks also to the audiences of New Perspectives on Irish English (University College Dublin, 11-13 March 2010), ScanDiaSyn 2010 (Sommarøya, 6-9 June 2010) and the Irish Network of Formal Linguistics Conference (University of Ulster at Jordanstown, 25-27 May 2011), at which preliminary versions of this work have been presented. This paper has also much profited from the very helpful comments and suggestions of Andrew Bailey, Paola Benincà, Davide Bertocci, Eithne Guilfoyle, Liliane Haegeman, Alison Henry, Mark Henry, Mair Parry, Lukas Pietsch and Christina Tortora. The present contribution is the result of the constant collaboration of the two authors; nonetheless, for the purposes of the Italian Academy, Mariachiara Berizzi is responsible for sections 1, 2, 3.1 and 3.2 and Silvia Rossi for the sections 3.3, 4, 5 and the appendix.

54

Chapter Four

participle, (cf. 2), even though, (2) is only an approximate translation that, for many HE speakers, does not capture completely the meaning of (1): 2 She has just broken the window It is traditionally assumed (Harris 1984, Kallen 1991, Filppula 1999, a.o.) that the AFP is to be considered a direct calque on the Irish construction here exemplified in (3): 3 Tá Mairéad tar éis amhrán a chasadh is Margaret after song PRT singing-VN “Margaret has just sung a song” The correspondence between the AFP and the construction in (3) is evident. They both express retrospective aspect and they both present the same type of elements: a form of the copula bi 'to be there', usually tá; the compound prepositions tar éis or i ndiaidh, both meaning after and a verbal noun (VN). We will not discuss here the complex relation between the AFP and the Irish construction. The main concern of the present contribution is to focus on the syntactic representation of the AFP within the grammar of HE, providing an alternative point of view to the more acknowledged treatment of AFP as a calque on Irish. In what follows we give a description of the distributional properties of the AFP (Section 4.2) which provides useful hints in order to put forward a formal analysis of the construction. The analysis is laid out in detail in Section 4.3 where we integrate the most recent Cartographic developments on the internal structure of PPs (Koopman 2000, Tortora 2006, 2008, Cinque 2010 and Svenonius 2010 a.o.) with the results on the functional architecture of the sentence as given in Cinque (1999). In Section 4.4 we tentatively extend our analysis to other languages that present aspectual periphrases involving prepositions. Section 4.5 concludes. After the conclusions a brief appendix presents some observations on the British (and American) English interpretation of AFP. The data for HE has been culled from various material, mainly from the existing literature on the phenomenon and from dictionaries like the Oxford English Dictionary (OED), the English Dialect Dictionary (EDD) and Dolan's (2006) Dictionary of Hiberno-English (DHE). We integrate this data with other examples collected in fieldwork by submitting questionnaires and recording spontaneous speech of two HE varieties, the variety spoken in Dublin and that spoken in Ennis, Co. Clare (western

The Syntax of the After Perfect

55

Ireland). Lastly, we also draw on some examples from the online editions of the major national newspapers, since the phenomenon is so pervasive of the language that it also appears in the standard written language.

4.2 The After-Perfect in Hiberno-English: a brief description In the present section we describe the distributional properties of the AFP. In particular, we will focus on the type of subjects allowed, the type of lexical verbs it combines with, its possible restrictions with the different tenses, its compatibility with modality and its behaviour with respect to negation. Like any other aspectual constructions, the AFP cannot occur with stative verbs (cf. Kallen 1991): 4 a b

*I'm after needing it *I'm after disliking it

As regards the type of subjects it allows, the AFP occurs with all persons and also with the expletives it (5a) and there (5b): 5 a

b

It's after raining not so long ago. (www.homepage.eircom.net/~ceindreadh/ ceindreadh/roadtrip4.htm) There's after being a sea change. (André Lyder, Pushers Out. The Inside Story of Dublin's AntiDrugs Movement, 2006)

The AFP can occur in the past tense (6a) and in the future with the modal will (6b), even though this latter possibility is not accepted by all speakers of HE: 6 a

I was in the market, and I was after buyin' a load of strawberries (Filppula 1999:99)

b

I will be after finishing my work tomorrow by the time you arrive (Co. Clare)

Chapter Four

56

As for its use with modals other than will, the AFP seems to be restricted to epistemic modality of the irrealis type (usually with might): 7 a b

He might be after falling asleep already when I get back (Figini 1996) The simple fact is that you can come off the field after scoring 3 or 4 points, but, your man might be after doing a lot of harm. (Irish Examiner online: http://archives.tcm.ie/irishexaminer/1999/06/19/shead.htm)

Testing the distribution of the APF with negation, it appears clear from the judgments of our informants that its occurrence with not is restricted to highly pragmatically marked contexts. Consider the following: 8 a b

You're after breaking the window! No, I'm NOT after breaking the window!

9 a b

I'm not after breaking the window, am I? Tell me I'm not after breaking the window!

As the examples (8) and (9) show, it is not the aspectual (perfective) value of the AFP that is negated: in (8)b, not scopes over the truth value of the whole sentence, while in (9) it negates the speaker's presuppositions. Intuitively, this is because aspect cannot be negated, only the lexical content of the verb can. This is true also with other aspectual constructions, as (10) shows: 10

I'm not reading a book = I am doing something which is not reading a book

As a final characteristic, we notice that the AFP can occur with temporal adverb(ial)s that in SE usually occur with the present perfect, as for instance, just or already. However, what seems more interesting, it can also occur with adverb(ial)s that in SE can occur only with the present or with the simple past, like now or ago: 11 a b

They are just after playing a county final. (The Kingdom, 24/11/2006, http://www.thekingdom.ie/news/cwcweyidgb/) I'm nearly after spilling the coffee all over the table (Dublin)

The Syntax of the After Perfect

12 a

b

57

Jaysus, have you heard what Ryanair are after doing now? (Irish Independent online: http://www.independent.ie/lifestyle/state-policy----the-herodway-1818029.html) I'm only after sleeping three hours (Dublin)

4.3 A syntactic analysis In the previous section we have given a description of the main distributional properties of the AFP that will serve as starting point for our analysis of its syntactic representation. The proposal we put forward combines the results of Cinque (1999) as regards the mapping of the functional portion of the sentence with the most recent developments of the linguistic research on the internal structure of spatial PPs (Koopman 2000; Cinque 2010; Svenonius 2006, 2010).

4.3.1 Some notes on the Cartography of PPs The presence of functional projections above the lexical projection of P(reposition) has long been argued for in the generative literature (at least since van Riemsdijk 1990). Koopman (2000), den Dikken (2006) and Tortora (2006, 2008), for instance, have identified various dedicated projections and their relative order by analysing the behaviour of locative PPs in Dutch and Italian. More recently, based upon cross-linguistic evidence, Cinque (2010) and Svenonius (2010) have further refined the internal structure of spatial PPs. We report in (13) Cinque's (2010) proposal:2 13 [PPdir [PPstat AT [DPPlace [DegP right [ModeDirP diagonally [AbsViewP north/south [RelViewP up/down [RelViewP in/out [DeicticP here/there [AxPartP [PP P [NPPlace Ground [PLACE]]]]]]]]]]]]]

2

Svenonius' (2010) proposal differs only slightly from that of Cinque (2010):

(i) p — Deg(ree) — Deix — Place — AxPart — K(ase) — DP(Ground) On a par with what has been proposed for the extended projection of V (vP-VP, cf. Larson 1988), Svenonius also proposes a “small p” p, the head of a functional projection whose Spec introduces the Figure (the object that has to be located with respect to the Ground).

58

Chapter Four

In the above structure we can identify a higher functional portion and a lower lexical part. The higher portion is constituted by two projections PPdir and PPstat, the projections in which directionality and stativity (Path and Place in Jackendoff's 1983 terms) are encoded respectively. PPdir is further articulated in PPsource ('from'), PPgoal ('to') and PPpath ('across'). PPstat encodes stativity by means of the preposition AT, which is assumed to be always present (either lexicalize d or silent) in both stative and directional PPs. The lower lexical part contains an NPPlace which hosts the “Ground”, the reference point. This NPPlace is embedded in a bigger DPPlace, under a series of projections hosting different kinds of modifiers of the Ground, as for instance AxPartP which hosts “axial part” prepositions (Jackendoff 1996. Svenonius 2006), i.e., prepositions that define a portion of space with respect to the Ground.

4.3.2 Applying the split-PP hypothesis to the AFP As we have already mentioned, the structure in (13) has been proposed for spatial PPs. Why should we apply such a structure to a chiefly temporal P like after? And most importantly, is it correct to apply an analysis developed for PPs to a P in an aspectual construction? Roy and Svenonius (2008: 4-5) observe that “a notable difference between the spatial dimension and the temporal dimension is the switch from three dimensions to one”. They show that temporal (and causal) complex Ps have the same structure of spatial complex Ps, or at least they share the same three “pieces”: Place (corresponding to Cinque's 2010 PPstat), AxPartP and KP (for case). Moreover, many, even though not all temporal Ps, are also spatial (cf. at, on etc.), or are etymologically so. After is a case in point: it derives from OE æfter, a compound of of/af ('off') designating origin (cf. Latin ab, Greek apo-) and the IE comparative element *-ter. Thus, after originally means 'more away than, further off than', (cf. also OED s.v. after). As a purely temporal preposition, after defines a portion of the timeline in relation to a specific moment or period that provides the Ground of the P. Consider (14), for instance, where Christmas, the party and midnight provide the Ground: 14 We met after Christmas/the party/midnight After is then an “axial part” P in the sense of Svenonius (2006) and is hosted in AxPartP. As such, the PP in (14) can be analysed as follows:

The Syntax of the After Perfect

59

15 We met [PPdir [PPstat AT [DPplace ... [AxPartP after AxPart° [PP P° [NPplace Christmas [PLACE/TIME] Evidence in favour of such analysis comes from the fact that the null preposition AT is overtly lexicalized in the preposition (adverb and subordinating conjunction) at-after, 'after', attested by both the EDD and the OED for some northern British varieties: 16 a b c d

It's my turn at-after thee (w. Yorkshire) Ay, it is a bonny nee, for sure, at-after this storm (w. Yorkshire, Lancashire) Shall you come nae (=now) or at-after? (ne. Lancashire, e. Lanchashire, Cheshire) That happened at efther Jack had geean heeam (e. Yorkshire) (EDD, s. v. at-after, 1. adv., 3. prep., 5. conj.)

To answer the second question, whether it is correct to apply the structure in (13) to Ps appearing in aspectual constructions, we can capitalize on the idea—more or less explicitly present in the abovementioned works—that PPs, adverbs and particles are different realization s of the same underlying structure. Thus, prepositional aspectual markers should also be realizations of (13) to some extent, or, at least, their syntactic behaviour should be partly related to this detailed structure. Similarly for what we have shown in (14)-(15), we claim that in the AFP after defines a portion of the time-line w.r.t. the Ground, that, in this case, is provided by the event described by the –ing form of the lexical verb (plus its arguments). Then, the example in (1)—here repeated in (17a)—can be rephrased as (17b) and be analysed as in (18): 17 a b

She's after breaking the window She's [AT after THE PLACE/TIME OF breaking the window]

18

[PPdir [PPstat AT [DPplace ... [AxPartP after [PP P° [NPplace breaking the window [PLACE/TIME]

As partial evidence in favour of this, we notice the parallelism in meaning between the AFP and the HE to be after something meaning to be

60

Chapter Four

over with, to have finished.3 The example (19a) is rather close in meaning to (17a), as its rephrased version in (19b) shows. Yet, in this latter case, the “axial” meaning of after appears somewhat more evident. We propose then that (19a) could be given the analysis in (20): 19 a b

I'm after my breakfast = I have (already) had my breakfast I'm [AT after THE PLACE/TIME OF my breakfast]

20

[PPdir [PPstat AT [DPplace ... [AxPartP after [PP P° [NPplace my breakfast [PLACE/TIME]

4.3.3 Going a step further... Following Cinque (1999), the different aspectual values are encoded in distinct and ordered functional projections according to the universal hierarchy in (21): 21 [frankly MoodSpeechAct[fortunately MoodEvaluative [allegedly ModEvidential [probably ModEpistemic [once T(past) [then T(future) [perhaps MoodIrrealis [necessarily ModNecessity [possibly ModPossibility [usually AspHabitual [again AspRepetitive(I) [often AspFrequentative(I) [intentionally ModVolitional [quickly AspCelerative(I) [already T(Anterior) [no longer AspTerminative [still AspContinuative [always AspPerfect [just AspRetrospective [soon AspProximative [briefly AspDurative [AspGeneric/Progressive [almost AspProspective [completely AspSgCompletive(I) [tutto AspPlCompletive [well Voice [fast/early AspCelerative(II) [again AspRepetitive [often AspRepetitive(II) [completely AspSgCompletive(II)

3

Please notice that we are simplifying matters for the sake of argument. In this respect it is interesting to note that there are HE speakers whose grammar admits to be after something but does not admit the AFP (Mark Henry p.c.). Clearly, the AFP and the HE verb combination to be after something are two different types of constructions, which are ultimately to be given different structural representations at the clause level. It is possible that for these speakers to be after is available as any other verb particle combination of the sort of to look after, to name after, while the AFP is not available presumably because their grammar does not allow AspRepetitive to be filled by a PP (cf. the following section).

The Syntax of the After Perfect

61

“Retrospective” aspect—the aspectual value we are directly concerned with here—is conveyed by various strategies in the languages of the world: by affixes on the verb, like those of “immediate or momentaneous aspect” of the Indo-Pacific languages Yimas and Una (cf. Cinque 1999:97); by adverbs, like the English just and the Italian appena 'just'; and by verbal periphrasis, like the French venir de, the Spanish and Portuguese acabar de and the HE AFP itself. Cinque (1999) proposes that this aspectual value is encoded in AspRetrospective, highlighted in bold in the above hierarchy, a projection whose head hosts affixes while its Spec hosts adverbs. In the AFP the retrospective value is conveyed neither by an affix nor by an adverb. Rather, it seems to be conveyed solely by the P after. One might then ask whether this P is to be hosted in the head or in the Spec of AspRetrospective. Poletto (2008) analyses the progressive construction of a North-Eastern Italian dialect (Venetan) that presents the preposition drio, 'behind, after' and an infinitive, a nominal form of the verb, cf. (22): 22 El ze drio magnar (Poletto 2008, ex. 28) he is behind/after to-eat “He is eating” Following Kayne's (2004) proposal that Ps are probes, Poletto (2008) proposes that the P drio is base-generated is the head of the projection for continuous aspect (AspProgressive in [21]) and being a probe, it attracts the infinitive magnar in its Spec.4 This accounts for the fact that the P and the infinitive behave as a unit, as a constituent. Poletto's (2008) analysis could be nicely applied to the AFP as well, with the only difference that after is base-generated in the head of AspRetrospective, as illustrated by (23): 23 [XP X° after [AspRetrospectiveP breaking the window Asp° after [ …. [ breaking the window]]]] However, we put forward a slightly different analysis. We propose that after and the –ing form of the lexical verb form a PP as in (17)-(18), which

4

The superficial order drio plus infinitive is derived by a successive movement of the P out of its aspectual projection to a higher functional projection: (i) [ XP

X°

P [AspP DP Asp° P [ …. [ DP]]]]

Chapter Four

62

behaves like an adverbial (on a par with just) and is thus hosted in the Spec of AspRetrospective: 24 [TP She T° is [AspRetrospective after breaking the window Asp° … 5 We believe that the analysis in (24) accounts rather straightforwardly for the following apparent problematic facts: (i) Assuming that the PP containing after is hosted in the Spec of AspRetrospective, we expect that just—the prototypical adverb in Spec AspRetrospective—should not co-occur with the AFP. However, as we have seen in section 4.2, the AFP and just do co-occur, cf. (25a) and (11a) here repeated in (25b):6 25 a

b

Mystery Jets are just after finishing a short run of gigs in the States (Galway Independent, 25/02/2009, http://www.galwayindependent.com/entertainment /buggin-outwithmystery-jets-/) They are just after playing a county final. (The Kingdom, 24/11/2006, http://www.thekingdom.ie/news/cwcweyidgb/)

In fact, judgements of native speakers of HE confirm that there is a difference between just + AFP and the AFP alone: just draws the event of AFP closer to the moment of utterance. Thus, the temporal scale in (26) can be established, in which have just + past participle—an option always 5

Here a number of questions arise. Is the after-PP directly generated in AspRetrospective or is it moved there? Is the event expressed by the –ing form basegenerated in the Ground of the PP? It could be that the –ing form of the verb is derived by a series of syntactic operations targeting a lower projection (in or immediately outside the vP area) and that after is the only element to be basegenerated in AspRetrospective, cf. (i): (i) [AspRetrospective after Asp° [ … [ZP breaking the window Z° [vP …]]]] We will explore this promising line of research in future works. For the time being, the point we would like to make here is that after should be considered an XP rather than a head. 6

Indeed, for some speakers the occurrence of just with the AFP seems to be obligatory (Alison Henry, p. c.).

The Syntax of the After Perfect

63

present in the grammar of HE speakers7—is the furthest in time with respect to the moment of utterance: 26 have just + past participle > AFP > just AFP > moment of utterance Just then modifies the AFP in exactly the same way just modifies other temporal PPs as in (27): 27 a b

I saw him just after the party It happened just after midnight

We notice that in the fine structure of PPs as reported in (13), there is a projection Deg(ree)P in which adverbial modifiers like right and measure phrases like two inches are hosted (Koopman 2000)8 and that this projection is above AxPartP. We propose then that just appearing in the AFP is not the aspectual adverb of English but a modifier within the DPPlace containing the AxPart after and its Ground breaking the window: 28 [PPdir [PPstat AT [DPplace [DegP just ... [AxPartP after [PP P° [NPplace breaking the window [PLACE] Thus, we propose that (25b), here repeated in (29), should be analysed as (30): 29 They are just after playing a county final. 30 [TP They T° are [AspRetrospective [PP just after playing a county final]Asp° .... (ii) As Cecilia Poletto (p. c.) notes, the subject of the AFP is also the subject of the lexical verb in the –ing form, i.e., she in (1) is the agentive subject of break the window. To capture this fact we assume the presence of a PRO co-indexed with the subject in TP which is part of the thematic grid of the nominalization in the –ing form: 31 [TP Shei T° is [AspRetrospective after PROi breaking the window Asp° .... 7

The scale in (26) may account for the observation made in section 4.1 that HE speakers do not consider the AFP equivalent to SE have just + past participle. 8 Cinque (2010, fn. 24) hints at the possibility that adverbs like right might in fact be hosted in a projection dominating PPdir (as in right from there, in which Cinque proposes that right modifies a null away, right (away) from there). This view does not substantially change the point that concerns us most here, namely, that there is a projection inside the PP dedicated to adverbial modifiers like just and right and that this projection is above AxPartP.

Chapter Four

64

4.4 Aspectual constructions with prepositions in other European languages It is a characteristic of many languages that prepositions can express various aspectual and actional (in the sense of Aktionsart) values. Prepositions conveying aspectual values typically occur as: particles in verb particle constructions, or phrasal verbs, a very well-known feature of English but also attested in the Venetan variaties of Italy (cf. spacar fora, lit. 'to smash out', 'to break completely'); as verbal prefixes as in Latin, Dutch or German (cf. Lat. ebibere or Ger. austrinken, both lit. out-drink 'drink up'); as aspectual markers in periphrastic constructions. We have already mentioned the progressive construction of the Venetan dialects (Northern Italy), here repeated in (32): 32 El ze drio magnar he is behind/after to-eat “He is eating” Prepositional constructions conveying progressive aspect are also attested in a number of German varieties: 33 a

Ich bin beim Lesen der Zeitung (Southern German varieties) I am by-the.dat reading of-the.gen newspaper “I'm reading the newspaper”

b

Ich bin das Buch am Lesen (Rhenish; north-western German) I am the.acc book on-the.dat reading “I'm reading the book”

Moreover, it is well known that the Celtic languages convey the different aspectual values by means of prepositional constructions: 34 a

b

9

(Irish9)

Tá Mairtín ag casadh is Martin at sing.VN “Martin is singing song”

amhráin song.GEN

Tá Mairéad tar éis amhrán is Margaret after song “Margaret has just sung a song”

a chasadh PRT sing.VN

Examples from Stenson (1981: 138, her examples [18]).

The Syntax of the After Perfect

c

Tá Rónán le amhrán a is Ronan with song PRT “Ronan is about to sing a song”

65

chasadh sing.VN

All the constructions in (32-34) involve (i) a form of the verb to be, (ii) a preposition and (iii) a nominal form of a lexical verb, i.e., the same grammatical elements that appear in the AFP. Even though they express different types of aspect (progressive, retrospective, proximative), the analysis we have proposed for the AFP can be applied also to these cases: 35 [PPdir [PPstat AT [DPPlace [AxPartP drio/bei [PP P [NPPlace magnar/Lesen [PLACE]]]]]]]] The above examples present peculiarities we will not treat in detail here, as for instance the position of the direct object with respect to the verb, the morphological case of the direct object (genitive or accusative) and the nature of the preposition that each construction involve as some are axial and some are not. It is clear that all these facts should be accounted for in any comprehensive analysis of these constructions, as they would probably lead to some modifications of the structure in (25). We will come back to this most interesting matter to future research.

4.5 Conclusions In the present contribution we have put forward a structural representation of a very well-known construction of the Hiberno English, the After-Perfect. We have applied the fine structure, which has been proposed for spatial PPs, to P appearing in an aspectual constructions. Under such view the preposition after of the AFP is considered a modifier of the Ground, the event expressed by the nominal form of the lexical verb (the –ing form and its internal arguments). Assuming Cinque's (1999) hierarchy, the retrospective aspect is encoded in a specific projection of the functional domain, AspRetrospective. Under our analysis the AFP, that is to say the whole PP encoding the aspectual value, is located in the Spec of this projection. Our proposal differs from other analysis put forward for aspectual constructions involving Ps, as for instance Poletto (2008), in that it analyses the preposition not as the head of the dedicated aspectual projection, but as a PP, thus an XP, hosted in the Spec. Such analysis accounts rather straightforwardly for the cases of occurrences of the AFP

66

Chapter Four

with just, as the fine structure of PP envisages a dedicated projection for this type of modifier, DegP.

Appendix The British (and American) English interpretation of the AFP10 It is a well-known fact that BrE speakers tend to interpret the AFP as a future of intention with the meaning of 'to be going to', (cf. Cottell 2003: fn. 2): 36 She is after breaking the window. = She is going to break the window Clearly, this interpretation is closely connected to the verb particle combination to be after something, which has the meaning of 'to be in pursuit of, trying to reach or get in the company of, trying to get or do', as illustrated by the example in (37): 37 He's after her money Intuitively, the meaning of after in this verb particle combination can be associated to directionality (cf. 'go after', 'follow'). We suggest here that the difference between the HE and the BrE interpretations is due to the different positions occupied by after in the PP. In the light of this, we propose that in the BrE interpretation, after is hosted in PPdir: 38 [PPdir after [PPstat AT [DPPlace ... [AxPartP [PP P [NPPlace NP [PLACE]]]]]]] Moreover, it could be the case that, when interpreting the AFP, nonHiberno English speakers locate the PP containing after in another functional projection in Cinque's (1999) hierarchy, probably ModVolitional (cf. 21).

10

We have tested the AFP with speakers of both British and American English. What we report in this section are only acceptability judgments: to our BrE informants, for instance, the construction sounds rather obsolete and would not be produced spontaneously.

The Syntax of the After Perfect

67

References Cinque, G. 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford-New York: Oxford University Press. —. 2010. Mapping Spatial PPs: An introduction. In G. Cinque and L. Rizzi (eds.), Mapping Spatial PPs: The Cartography of Syntactic Structures. Vol. 6. Oxford-New York: Oxford University Press. Cottell, S. M. 2003. The Syntax of Perfective Aspect in Hiberno-English. Belfast Working Papers in Language and Linguistics 16: 1–21. den Dikken, M. (2006). On the functional structure of locative and directional PPs. Ms. CUNY. Dolan, T. P. 2006 (1998). A Dictionary of Hiberno-English. Dublin: Gill and MacMillan. Figini, V. E. 1996. Due caratteristiche sintattiche dell'inglese d'Irlanda a confronto: il “present perfect” e il “presente abituale”. Quaderni Patavini di Linguistica 16: 95–158. Filppula, M. 1999. The Grammar of Irish English. Language in Hibernian Style. London: Routledge. Harris, J. 1984. Syntactic variation and dialect divergence. Journal of Linguistics 20: 303–327. Jackendoff, R. 1983. Semantics and Cognition. Cambridge MA: MIT Press. —. 1996. The architecture of the linguistic-spatial interface. In P. Bloom, M. A. Peterson, L. Nadel and M. F. Garrett (eds.), Language and Space, 1–30. Cambridge, MA: MIT Press. Kallen, J. L. 1991. Sociolinguistic variation and methodology. After as a Dublin variable. In J. Cheshire (ed.), English around the worlds, 61– 74. Cambridge: Cambridge University Press. Kayne, R. 2004. Prepositions as Probes. In A. Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures Vol. 3, 192–212. Oxford-New York: Oxford University Press. Koopman, H. 2000. Prepositions, Postpositions, Circumpositions and Particles: The Structure of Dutch PPs. In H. Koopman, The Syntax of Specifiers and Heads. Collected Essays of Hilda J. Koopman. London: Routledge. Larson, R. K. 1988. On the Double Object Construction. Linguistic Inquiry 19(3). 335–391. OED, The Oxford English Dictionary. 1989. Simpson, J. A and Weiner (eds.), E. S. C., 2nd edition, Oxford: Clarendon, Poletto, C. 2008. Syntactic encoding of Aspect in some Northern Italian dialects. In Á. Corrasco Gutiérrez (ed.). Tiempos Compuestos y

68

Chapter Four

Formas Verbales Complejas, Ch. 11, 499–518. Frankfurt-Madrid: Vervuert-Iberoamerica. Riemsdijk, H. van. 1990. Functional Prepositions. In H. Pinkster and I. Genee (eds.), Unity in Diversity, 229–241. Dordrecht: Foris. Roy, I. and P. Svenonius. 2008. Complex Prepositions. Available at http://ling.auf.net/lingbuzz/@EMhqsuvusZJothBZF/PbWkvatp?8. Stenson, N. 1981. Studies in Irish Syntax. Tübingen: Gunther Narr. Svenonius, P. 2006. The Emergence of Axial P. In P. Svenonius and M. Pantcheva (eds.), Adpositions. Special issue of Nordlyd: Tromsø Working Papers in Linguistics, 33(1).49–77. Available at (http://www.ub.uit.no/munin/nordlyd/). —. 2010. Spatial P in English. In G. Cinque and L. Rizzi (eds.), Mapping Spatial PPs. The Cartography of Syntactic Structures. Vol. 6. OxfordNew York: Oxford University Press. Tortora, C. 2006. On the aspect of space: The case of PLACE in Italian and Spanish. In N. Penello and D. Pescarini (eds.), Atti dell'undicesima giornata di dialettologia (Quaderni di lavoro ASIS, v. 5). pp. 50–69. CNR: Padova. Available at: http://asis-cnr.unipd.it/ql-5.it.html. —$VSHFWLQVLGH3/$&(33V,Q$$VEXU\-'RWODþLO%*HKUNH and R. Nouwen (eds.), Syntax and Semantics of Spatial P, 273–301. Amsterdam: John Benjamins. Wright, J. 1898–1905. The English Dialect Dictionary. London: Henry Frowde.

PART II CELTIC LANGUAGES

CHAPTER FIVE TENSE AND FEATURES IN IRISH VERBAL MORPHOLOGY PAOLO ACQUAVIVA 5.1 Introduction: Forms and categories The verbal system of a language relates an array of semantic values, defined in terms of tense, aspect, and mood, to a system of forms defining the verbal inflection of the language. This paper attempts an analysis of the latter. Following on the pioneering study of Wigger (1972), it analyzes the structure of Irish verbal paradigms in terms of combinations of grammatical primitives, expressed as formal features, and proposes an answer to the questions of what features are associated to what morphemes to express what semantic values. The point of departure is the catalogue of inflected forms of a regular verb, as in (1), organized by traditional tense and mood labels, and abstracting away from the pronominal endings that stand in complementary distribution with pronominal subjects in some person/number combinations. The list includes the marginal subjunctive (in earlier historical stages opposed to a past subjunctive, now moribund), and a separate category of present habitual indicative which only applies to the existential verb. The verbal forms are flanked by a simple analysis into their visible constituent parts: root, ending, a word-initial element here indicated as DO but appearing as lenition of the initial consonant on this verb, and a root-extending suffix schematically indicated as F: 1 Modern Irish tenses and moods (bris ‘to break’) present indicative briseann ROOT-ending (present habitual indicative) past indicative bhris DO-ROOT past habitual indicative bhriseadh DO-ROOT-ending future indicative brisfidh ROOT-F-ending (subjunctive) brise ROOT-ending

Tense and Features in Irish Verbal Morphology

conditional imperative

bhrisfeadh bris

71

DO-ROOT-F-ending ROOT

DO appears as lenition on consonant-initial roots (bris > bhris ‘broke’), as d- prefixation of vowel-initial roots (ól > d’ól ‘drank’), and as both with roots beginning by /f/ > Ø (fan > d’fhan ‘waited’, freagair > d’fhreagair ‘answered’). F stands for the ‘future’ suffix spelled -f- (whose realizations range from /h/ to zero) or -ó- (realized as /oඌ /). The verbal paradigm is complemented by periphrastic forms like those exemplified in (2): 2 a tá Eoin ag is Eoin AT “Eoin is reading.”

léamh reading.VN

b tá an leabhar léite ag is the book read.VA AT “Eoin has read the book.”

progressive

Eoin Eoin

perfective

c tá leabhar le léamh agam prospective is book WITH reading.VN AT.1SG ‘I have a book to read’ These forms are anything but marginal in an account of the Irish tense/aspect system; cf. especially Ó Sé (2001), and Bloch-Trojnar (2006, 2008) for the morphology of the verbal noun and Ó Donnchadha (2010) for its syntax. Our main concern here, however, is the morphological system defined by the inflected forms in (1), and not the semantic categories and oppositions of the Irish verbal system as a whole. I will pursue a constructional approach, where the semantic range of a tense arises from the content of its morphemes and from the way they are syntactically combined. An important background assumption of this approach is that the temporal interpretation defined by a tense is a property of the whole clause. To offer a featural analysis capable of revealing a system behind the list in (1), we must then view inflected verbs as parts of the structure of a clause. The “meaning” of tenses will then emerge from the value of markers in their syntactic context. The analysis will be empirically successful to the extent that it derives not only paradigmatic, but also syntagmatic regularities characterizing Irish inflected verbs from the syntactic locus of features.

Chapter Five

72

5.2 Syntax of the Irish finite clause The structure in (3) summarizes the structure assumed for the Irish clause, along with the position of the key features before syntactic and morphological operations: 3.

CP C

Polarity/FinitenessP Pol/Fin [±modal]

TP T [±Past]

AspP Asp

vP v [±Future] ([±Perfective])

ROOT

This representation reflects the arguments for distinct inflectional heads below the complementizer C (Guilfoyle 1990, 2000, Cottell 1995, Doherty 1996, McCloskey 1996a, b, Doyle 2002) and the evidence for at least two grammatical heads between the inflected verb and the head of VP (McCloskey 1996a, b, 2001, 2009, Ó Donnchadha 2010). The element marked as v is a verbalizing head which takes as complement a categoryfree root. This follows a line of analysis arguing for the syntactic decomposition of lexical categories (Harley and Noyer 1999, Embick and Marantz 2008), so that a lexical verb is a lexical root in a verbal environment. I assume a separate head Asp(ect), host of aspectually used particles like ag (McCloskey 1996a, 2007: 847-851, Doyle 2002 106-111; this corresponds to the head Voice in Ó Donnchadha 2010). While required by syntax, this head plays no role in the make-up of inflected verbs as word forms—it is natural, in this light, that aspectual oppositions tend to be expressed periphrastically. T, for Tense, is the locus of interpretation of the key feature in the semantics of Tense, namely [±Past], interpreted as “±Exclusion” between two temporally extended arguments (see Demirdache and Uribe-Etxebarria 2000, Stowell 2007, Mezhevich  7KHODEHO3RO)LQVLJQDOVWKHLGHQWLILFDWLRQRI3RODULW\1HJȈ DQG Finiteness (Rizzi 1997) as the same structural position in Irish. Following McCloskey (2010), this is the host of raised finite verb, and the locus of

Tense and Features in Irish Verbal Morphology

73

interpretation for negative features (Acquaviva 1996). The crucial evidence that a full finite clause is not a TP but a larger projection PolP is provided by coordinations like (4), from McCloskey (2010): 4 mur dtéighinn agus iad cailleadh, if.NEG go.CONDIT.1SG and 3.PL lose.VN mhuirbhfeadh iad mé kill.CONDIT 3PL 1SG “If I were not to go and they were to lose, they would kill me.” Only the first conjunct has negative sense, with a schematic reading if [[not p] and [q]] (otherwise the protasis would mean ‘unless I were to go and they were to lose’, contrary to fact). Negation is however realized higher than both conjuncts, on C, as an instance of left conjunct agreement: 5

CP C mur

X [NEG-clause1] and

[clause2]

More generally, C is the single structural location for preverbal particles (McCloskey 2001, contra Duffield 1995).

5.3 Post-syntactic morphological realization The syntactic approach followed here involves a realizational view of morphology, as a series of operations on the output of syntax. Their point of departure is depicted in (6), after the root has adjoined to v, the resulting complex head has adjoined to T, and T has adjoined to Fin/Pol (henceforth just Fin):

Chapter Five

74

6

FinP Fin

TP

T

Fin

T

v v

ROOT

The word-form corresponding to an inflected (tensed) verb arises from the morphological interpretation of this complex Fin head. This input is not immediately translated into a phonological string, but into a morphological structure whose elements have well-formedness requirements which drive structural adjustments: Fusion, Fission, or Merger of local nodes. (Halle and Marantz 1993, Harley and Noyer 1999, Embick and Marantz 2008). (7) shows the proposed map between the syntactic input and the word structure anticipated in 1: 7

Fin T T

Fin v v

ROOT

DO

STEM

(ENDINGS)

5.4 The syntactic construction of Tense Another set of assumptions concerns the interpretation of Tense. I will follow Iatridou (2000), in turn based on Smith (1991) and Klein (1995), in taking the central Tense opposition to be [±Past], and in assigning to this feature the basic value of a two-place relation paraphraseable as “Exclusion”. A past reading arises when the feature valued [+Past] establishes an exclusion relation between “now” and the time at which the verb-denoted event is asserted to take place. An event asserted to obtain but not now can only have taken place in the past, under the common and plausible assumption that future statements are inherently modal, and cannot strictly be true or false in the actual world but only in a

Tense and Features in Irish Verbal Morphology

75

continuation of it. Further, Iatridou accounts for the crosslinguistic tendency for the same morphology to express past and counterfactual readings by hypothesizing that both arise from the same “exclusion” relation, which relates two times in one case and two possible worlds in the other: 8 T(x) excludes C(x)

(Iatridou 2000: 246)

T = ‘the x that we are talking about’ (Topic x) C = ‘the x that for all we know is the x of the speaker’ (Utterance x) For x = time: ‘Topic time excludes Utterance time’ > past For x =world: ‘Topic world excludes Utterance world’ > counterfactual The interpretation of [Past] as “Exclusion” not only unifies the semantics of past Tense and counterfactual mood; it also provides a unified structural account for the semantics of Tense and Aspect. Demirdache and Uribe-Etxebarria (2000), Stowell (2007), and Mezhevich (2008) have articulated a constructional approach that views Tense and Aspect as syntactic heads interpreted as two-place relations between temporally extended arguments expressed by syntactic projections. The difference between the two categories is that Tense relates the Utterance time to the Topic time (the relevant interval for interpretation), while Aspect relates the Topic time to the Event time (the interval at which the event occurs); for example, an imperfective reading corresponds to a Topic time entirely contained within the Event time, so that the boundaries of the latter are interpretively irrelevant. To apply this general approach to the analysis of Irish verbal Tense, I identify three features, referenced by syntax and morphology, and having the following semantic values: [Past], which licenses the reading of T as [Exclusion] as outlined above; [Future], which licenses the reading of Fin as [Modal] (providing access to alternative worlds for the semantic evaluation of the sentence); and [Perfective], which signals coincidence between Assertion time and Event time. The next three sections outline an analysis of Irish verbal inflection based on these features.

76

Chapter Five

5.5 Features and morphology: [Future] DO and F in (1) above are a shorthand for an array of exponents, some in a relation of phonologically conditioned allomorphy (like the exponents of DO on regular verbs, whose choice depends on the root onset), some in paradigmatic contrast across lexical roots (like the exponents of Future). As a first step, let us identify the morphemes referred to by the abstract label DO with the exponence of [Past], and the morphemes referred to by F with the exponence of [Future]. These features, as is well known, group together Irish tenses in the following way: DO = [Past] characterizes simple past, past habitual, and conditional; F = [Future] characterizes future and conditional. This leaves out the copula is, which has no future or habitual, but just a past form with a past or counterfactual reading. Correspondingly, the forms of the copula have neither F- nor DOmorphemes. This is not an accidental defectiveness in its paradigm, but the sign that the paradigm of the copula is organized differently from that of lexical verbs, in keeping with the standard view that the Irish copula is not a lexical verb but a spellout of a tensed inflectional head. In particular, the copula lacks the future category altogether, either as a morphological feature (it has neither a future nor a conditional tense) or as a notional category, no matter how expressed. This finds a natural explanation if the notional content of future is mediated by the morphosyntactic feature [Future], and this is a feature of the head v, as anticipated in (3). Since v combines with lexical roots, the property of having a [Future]v amounts to a particular way to be a word; specifically, a particular way to be a lexical verb. To say that the copula is not lexical, under this analysis, mean that it lacks a lexical root and the v head which categorizes it as a verb (I will assume that it spells out T). In this way, positing v as the structural locus of [Future] explains why both the morphological marker and the notional category are relevant for lexical verbs but not for the copula, whose make-up does not involve v. In addition, since v is immediately adjacent to the root, the hypothesis explains why the exponence of [Future] can depend on the choice of the root, as a case of local root-governed allomorphy (the traditional conjugations, illustrated by bris-f-idh ‘will break’ contrasting with ceangl-ó-idh ‘will bind’). Thirdly, the locality with a root also explains why certain verbs use special suppletive roots for the future, like rachaidh ‘will go’. As I will argue below, such special stems arise when a root is fused with the head hosting a Tense feature; an independent locality requirement on Fusion explains why not every tense can have a special stem.

Tense and Features in Irish Verbal Morphology

77

To make the analysis more precise, (9) structurally represents the adjunction of a root to v (a general process) when the latter is marked [Future]: 9

vP v ROOT

t

v[Future]

The simple hypothesis that v has the exponents symbolized as F when [Future], but no exponent otherwise, accounts for the general pattern corresponding to the first conjugation: 10 First conjugation: ROOT -F -ENDINGS future ROOT -ENDINGS all other tenses

(bris-f-idh) (bris-eann)

The second conjugation requires a more complex analysis. Apart from the fact that its future suffix -ó- co-occurs with -f- in impersonal forms (ceangl-ó-f-ar ‘one will bind’), it is subject to a prosodic output constraint to the effect that all its forms are at least bisyllabic, with a palatal suffix -íin forms like the present ceangl-í-onn ‘binds’ but not in the bare stem ceangail ‘bind!’ (imperative), nor in the future (cf. Ó Sé 1991). 11 Second conjugation: ROOT -O -ENDINGS ROOT -I -ENDINGS

future all other tenses

(ceangl-ó-idh) (ceangl-aí-onn)

This suggests that second-conjugation roots are in fact complex roots comprising a lexical root and a particle (which I will notate IGH), neither of which is intrinsically a verb. This is the analysis Basilico (2008) proposed for English phrasal verbs like build up: 12

¥build up ¥EXLOG

up

Chapter Five

78

Unlike English particles, IGH is a bound affix, and so must attach to a host as a morphological requirement. Without [Future], this results in the structure ROOT-IGH-. But when v is [Future], what we observe is not a tripartite ROOT-IGH-F-, but a bipartite ROOT-O-. We can interpret this as another facet of the morphological requirement of -IGH-: it attaches to a lexical root, or fuses with a [Future] affix which attaches to the root. To model this analysis in more precise terms, -IGH- triggers Lowering of a structurally adjacent v[Future] head (Embick and Noyer 2001; cf. McCloskey 1996a: 55), followed by Fusion with it: 13

v ROOT

v

v[Future]

ROOT

ĺ

ROOT

IGH

ROOT

IGH 14

Lowering

IGH v[Future]

IGH ĺ

[IGH, v[Future] ]

Fusion

IGH v[Future] The fused terminal [IGH, v[Future]] is then realized as -ó-, while -IGH- alone is spelled out as -í-; without -IGH-, that is in the first conjugation, v[Future] is realized as -f- or zero, that is, as one of the exponents designated by F: 15 a b c

[IGH, v[Future]] ļ [IGH] ļ v[Future] ļ

o i f

Realization rules (Vocabulary Items)

5.6 Features and morphology: [Past] As we have seen, the exponents represented by DO characterize the simple past, habitual past, and conditional of lexical verbs. The copula displays instead a simple part-nonpast opposition (is / ba). In addition, an analysis of past morphology in Irish must take into account the distinctive “past” form of preverbal particles, characterized by a suffixed -r (ar, níor). Crucially, however, this does not apply to all particles, and appears only when the tense is simple past, not whenever its independent form (in a context without a particle) has DO.

Tense and Features in Irish Verbal Morphology

79

All of this points to [Past] being interpreted on T, as is natural for its value of “Exclusion”, and appearing as an uninterpretable feature on Fin. Locating this feature higher than v accords with the facts that it applies to the copula (which spells out T), and its realization is sensitive to C. More precisely, [Past] values the feature [Tense: __ ] on T, where it is interpreted as indicated. This featural content of T gets spelled out (on Fin) as DO for lexical verbs, and as the past/conditional forms notated BA for the copula. The difference is that lexical verbs are roots categorized by v: the different realizations of T can then be made sensitive to the presence of an uninterpretable [v] feature on T, which attracts lexical verbs (=[v [root]]): 16 7ĺDO (lexical verbs) [Tense: Past] [v] 17 7ĺBA (elsewhere = copula) [Tense: Past] As noted, DO represents an abstract morphological element; it is the context which determines its concrete exponence on [Past] verbs. The first distinction is whether the verb is or is not preceded by a particle; secondly, preverbal particles may trigger lenition, nasalization, or no mutation at all. This is schematized and exemplified in (18) and (19): 18 For [Past] verbs not preceded by a preverbal particle: d-*[Lenition]*STEM d’fhan sí liom DO-LEN.wait 3SG.FEM with.1SG “She waited for me.” 19 for [Past] verbs preceded by a preverbal particle X: a If X is a (direct relative particle), má ‘if’, or ó ‘because’: same as (18) an bhean a d’fhan liom DEF woman PRT DO-LEN.wait with.1SG “The woman that waited for me”

Chapter Five

80

b

If X is a leniting particle other than a (direct) or má: X*[Lenition]*STEM níor fhan sí liom NEG.PAST LEN.wait 3SG.FEM with.1SG “She did not wait for me.”

The distinctive behavior of a (direct), má, and ó prompted Duffield (1995) to propose structurally different positions for them with respect to all other particles. However, there is no syntactic evidence for such a split (McCloskey 2001). What seems relevant, instead, is that these three particles are the only ones without past or negative alternant. It is an independently established observation that Irish morphology has a fused expression of Tense and Negation on C; more precisely, negation of a finite verb is a form of Tense (which for Acquaviva 1996 explains the lack of negative quantifiers). Let us suppose, then, that the expressions of C with past-nonpast and positive-negative alternants are formally marked as [Tense]. In this analysis, then, the morphological centrality of this feature corresponds to an articulated syntax: it originates and is interpreted on T, is an uninterpretable feature on Fin, where it triggers raising of the tensed T-v complex, and also marks C, when present, with an uninterpretable feature; a formally tensed C is spelled out by a particle, not by the raised verb. Some realizations of C instead lack this uninterpretable [Tense] feature, and they are identified (in a non-arbitrary way) by the lack of a paradigmatic tense alternation. Since finite clause negation in Irish morphologically coincides with the expression of Tense on C, the same particles also lack a paradigmatic contrast between affirmative and negative forms. So, what distinguishes the particles a (direct), má, and ó is that they fill a C which lacks [Tense]. The realization rules for DO, then, will reference [Tense] and not specifically the three exceptional particles: 20 a

DO

ļ

[Lenition] / C [Tense]_____ for C = gur, nár, níor...

b

DO

ļ

d- + [Lenition]

elsewhere

In sum, [Tense] is on C when a particle lexicalizing this position has a paradigmatic set of oppositions; all and only C marked [Tense] have past and negative alternants; the three particles that lack both are not marked [Tense]. But the realization rule for DO references [Tense]; the three special particles do not define a natural class, but fall within the elsewhere

Tense and Features in Irish Verbal Morphology

81

case of C without [Tense]. It is no accident, then, that DO in their context is realized as if there were no particles.

5.7 Features and morphology: [Perfective] and “habitual” As noted, the “past” alternants of particles (suffixed by -r) in fact occur only with the simple past, not generally with all tenses that involve DO: 21 a

aƍ

22 a

d’fhan sí PAST.wait SG.FEM “She waited”

simple past, affirmative

ní-or fhan NEG-PAST wait “She did not wait”

simple past, negative

sí 3SG.FEM

d’fhan-adh sí PAST.wait-HAB 3SG.FEM “She used to wait”

past habitual, affirmative

aƍ

ní fhan-adh sí NEG wait-HAB 3SG.FEM “She used not to wait”

past habitual, negative

23 a

d’fhan-f-adh sí PAST.wait-F-HAB 3SG.FEM “She would wait”

conditional, affirmative

ní fhan-f-adh sí NEG wait-F-HAB 3SG.FEM “She would not wait”

conditional, negative

aƍ

This is one of several ways in which the simple past stands out morphologically from other tenses that involve the “past” affixes here labelled DO. In addition to the alternation between DO and past (C-r) particles, the simple past defines a tense-specific set of endings, which does not appear on other tenses; it is also the only tense that suspends the realization of DO with the impersonal form (mhol sé ‘he praised’, but moladh ‘one praised’); also in the impersonal form, it is the only tense that does not involve a “pre-ending” -t- or -f- (Wigger 1972, Ó Siadhail 1989); and finally, it is unique among DO-tenses in admitting root suppletion in

82

Chapter Five

irregular verbs (cf. below). There is thus ample evidence that Irish verbal inflection can reference the simple past on its own. We can keep calling [Past] the feature realized as DO, even though not all tenses having it are notionally past (again, this will be discussed shortly), but some other formal marking is needed to single out the simple past among DO-marked tenses and also among notionally past tenses (that is, simple and habitual past, to the exclusion of the conditional). Semantically, what singles out the simple past is that it is episodic, in contrast with the generic and/or frequentative function of the habitual past. An example by Ó Sé (2001: 142) brings this out clearly (emphasis added): 24 nuair a bhí cónaí orm i Maigh Nuad when PRT be.PAST home on.1SG in Maynooth shiúlaíos go dtí mo chuid oibre uair amháin walk.PAST.1SG to my COUNT work time one “When I worked in Maynooth I walked to my work once.” The simple past shiúlaíos (Kerry dialect) must refer to one completed event, not to a habit as would be expressed by the past habitual shiúlaínn ‘I used to walk’ (Ó Sé notes that the English walked admits both readings, unlike shiúlaíos). Crucially, however, this does not mean that a simple past form should always be translatable by a perfect in another language. Consider bhí ‘was’ in the temporal clause introducing (24): the living event referred to is a prolonged state, not a bounded episode, and as such would be translated by an imperfect in a Romance language, for example. The Irish simple past, then, is not aspectually perfect (certainly not in the sense of expressing a viewpoint later than the event’s completion), but rather ‘perfective’, in the sense of ‘completive’ (Binnick 1991: 210, 296– 297): it expresses the total overlap between Assertion time (the interval relevant for interpretation) and the Event time (the interval at which the verbal predicate is true). This is compatible with the meaning of bhí in (24), which denotes a long event whose length coincides with the interval on which the assertion is made; but it is not compatible with a macro-event structured into smaller completed events, as in the frequentative reading. The simple past, then, is marked [Perfective] in addition to [Past], which makes it aspectually completive. I take [Perfective] to originate on v, like [Future]. This explains why the copula, lacking v, also lacks a past form corresponding to the simple past, and distinct from the one marked [Past] used in the conditional. The copula also lacks the completive-frequentative opposition altogether, but this is due to its value of individual-level predicator, which is incompatible

Tense and Features in Irish Verbal Morphology

83

with episodic readings. Locating [Perfective] on v also correctly predicts the possibility of inherently simple past roots, analyzed as fusion between a lexical root and a [Perfective] v, parallel to inherently [Future] roots (see below). Finally, the fact that “past” alternants of C particles appear only with the simple past is accounted for if their appearance is conditional not on [Past] alone, but on [Past] and [Perfective]. Normally [Perfective] implies [Past], but the existential verb bí represents a single lexical exception in which the habitual-episodic opposition is morphologically expressed in the present too. An important result of this analysis of the simple past as [Perfective] is that there is no need to posit a “habitual” feature, to mark off either a habitual interpretation, or a morphological marker: what is “habitual” is simply not episodic, with a contrast that is morphologized in the past through the presence or absence of [Perfective]. Semantically, a habitual value is a possible reading of the present tense, without special morphology (except for the verb bí, which opposes a regular “habitual” bíonn to a suppletive non-habitual tá); and it is a necessary reading of the habitual past, but not of other tenses. Morphologically, habitual marking identifies a set of forms shared by past habitual and conditional (Wigger 1972: 177, Ó Siadhail 1989: 178), with habitual meaning only in the former. There is thus a mismatch between habitual semantics and morphology. The puzzle disappears if the marked value is [Perfective] instead. Morphologically, the habitual endings emerge in our featural analysis as the endings of all and only the tenses that are [Past] but not [Perfective], namely habitual past and conditional. These endings, then, are realizations of the head [Fin] conditioned by [Past] (the realization of [Past] is DO, which spells out T and ends up in prefixal position; cf. (7)). There is no puzzle caused by a semantically inert habitual marking in the conditional, because there is no such marking. Semantically, “habitual” readings comprise in fact two distinct varieties, the frequentative proper and the generic, illustrated by the examples in (25): 25 a

b

bíonn an fhirinne searbh be.HAB the LEN.truth bitter “The truth is bitter.” san oirthear a éiríonn an ghrian (Ó Sé 2001: 124) in.the east PRT rise.HAB the sun “The sun rises in the east.”

84

Chapter Five

The timeless statement of a proverb like (25a) differs from (25b), which attributes a permanent property arising from cyclic re-occurrence of an event, and both differ markedly from the core frequentative value of the past habitual, which refer to an indeterminate but bounded plurality of events (as ‘I used to walk’). But these readings are all non-episodic: either for not denoting a particular event (generic), or for denoting a series of such events (frequentative). Semantics and morphology thus converge in suggesting a negative characterization of “habituals”. This approach to habituality offers a new perspective on the category of realis in Irish. Taking Fin to encapsulate the deictic centre of the clause, we can distinguish clauses with and without an abstract deictic “now” encoding the Utterance time. Non-episodic clauses are those where the verbal event is not related to such a “now”. This naturally distinguishes generic from time-specific readings in the indicative, but does not account for the conditional, which is irrealis but has nothing to do with genericity. However, Fin is also plausibly the locus of modalization (cf. (3); there is no evidence in Irish for a Mood head distinct from Fin), interpreted as access to alternative worlds. A notional category of realis, not directly morphologized, emerges from the conjunction of deictic “now” and lack of modalization, both encoded on Fin. For different reasons, then, the habitual past and the conditional end up being both irrealis, which sits naturally with their merger in several dialects (e.g. Hughes 1994: 645 for Ulster). More generally, this analysis also accords with Iatridou’s (2000) generalization that “When the temporal coordinates of an eventuality are not set with respect to the utterance time, morphology is always Imp[erfective]”: so-called habitual morphology is the lack of a [Perfective] marking with [Past], and it appears on the conditional, which is always irrealis because Fin is modal, and on the habitual past, which is irrealis only when it does not relate the verbal predication to the Utterance time, namely in generic clauses.

5.8 Suppletive roots as fused verbal stems [Root + v] As anticipated, the structural-constructional approach to Irish inflected verb forms explored here allows us to see root suppletion as more than just an unpredictable idiosyncrasy. The lexical items involved and the forms of the alternants remain (synchronically) unpredictable, but the distribution of suppletive alternants in the paradigm is not random, and supports a constructional analysis of tenses. The main claim is that a suppletive form like the simple past chonaic ‘saw’ arises by the Fusion of the lexical root with a Tense value on v (here, [Perfective]). Because they are (abstractly)

Tense and Features in Irish Verbal Morphology

85

composed with a verbalizing head, I will call these fused forms stems rather than roots, even though they may appear to be monomorphemic (modulo lenition, as in this case). Table 5.1 shows that there is no direct correlation between stems and tenses, but some patterns obtain. Table 5.1. Distribution of stem allomorphs by tense in the standard dialect future/ conditional dear bear be

ABAIR BEIR BÍ CLOIS DÉAN

gheobh

FAIGH FEIC ITH TABHAIR TAR TÉIGH

íos tabhar tioc rach

imperative

present

abr

past habitual deir

beir bí (present non-hab: tá) clois déan faigh feic ith tug tar (2sg) tag té

simple past dúr rug

chuala rinne fuair chonaic

tháinig chuaigh

The key generalizations on stem allomorphs (S) and tenses are laid out in (26): 26 a

If S expresses past, it is used for the simple past and for no other tenses

b

The S used for the conditional is always a future S plus the regular past marker (no past S + regular future, or a specifically conditional S) [PAST [ FUT+ROOT ]] *[FUT [ PAST+ROOT ]] *[FUT+PAST+ROOT ]

c

There is no past habitual S

These now follow from the locality of Fusion, under the structural analysis proposed. Recall that only simple past forms are marked [Perfective] on v, close enough to the root to allow for Fusion; the “past”

Chapter Five

86

S, then, are limited to the simple past because they are in fact [Perfective]. As for the conditional, it involves [Past] and [Future], but a root can only fuse with [Future], which is on v. Finally, the past habitual is excluded because it is marked only [Past] and not [Perfective]: 27

[ C [ Fin [T [ v [ root ]]]]] ±modal ±Past ±Future ±Perfective

local Fusion

These paradigmatic patterns are not the only empirical observations that our structural analysis can account for. A particular class of suppletive stems only occur in negative or syntactically subordinating contexts, traditionally termed “dependent”; accordingly, I will label these forms dependent stem allomorphs, or DS. Table 5.2 lists them, again limited to the standard dialect: Table 5.2. S with a special form for syntactically dependent contexts

‘to say’ ‘to be’ ‘to carry’ ‘to hear’ ‘to do’ ‘to get’ ‘to see’ ‘to eat’ ‘to give’ ‘to come’ ‘to go’

tense stem allomorphs (verbal noun in brackets) abair, deir, déar, dúr, dúirt (rá) bí, tá, beidh beir, béar, rug clois / cluin, chuala déan, rinne / dhein faigh, gheobh, fuair (fáil) feic, chonaic ith, íos tabhair, tug tar, tag, tioc, tháinig (teacht) té, rachaidh, chuaigh (dul)

dependent stem allomorphs fuil, raibh dearna faighid faca

deachaigh

The main regularities concerning DS are summarized in (28): 28 a b c

DS occur only in verbs that have a S DS are always sensitive to Tense DS occur after preverbal particles with a negative and a past alternant: not after a (direct), má, ó

From this perspective, DS are best seen as positionally conditioned allomorphs of S. This makes sense of the fact that they do not cross-

Tense and Features in Irish Verbal Morphology

87

classify with S, but systematically appear on verbs that allow root suppletion (a generalization that apparently holds across dialects). Besides, DS do not just occur in dependent contexts, but only in certain tenses; finally, the very same set of particles that are “invisible” for the dependent realization of DO also fail to count as triggers for DS. The context for allomorphy, therefore, is the same as that reconstructed in (20), namely a position following C when this is marked [Tense].

5.9 Conclusions The constructional analysis explored here has attempted to derive a number of systematic regularities from independently justifiable assumptions about the syntactic structure of Irish clauses and about the operations and mutual relation of syntax and morphology. The key assumption that syntax generates paradigm space makes it possible to unify the explanation for syntagmatic/distributional and paradigmatic restrictions, and to constrain what morphology can express. Within this approach, I have applied to Irish a line of inquiry into verbal tense and aspect based on the views that tense meanings are functions of feature content and of syntactic context. These background assumptions have prompted some novel proposals about Irish verbal inflection. The doctrine that views past as a topological relation defining exclusion between two arguments has made possible a featural analysis of Irish tenses based on [Past], [Future], and [Perfective]. Some innovative claims of this analysis may be highlighted: that secondconjugation verbs are complex roots, parallel to English phrasal verbs (but with an affixal radical); that “habitual” morphology and semantics represents the unmarked value, contrasting with [Perfective]; that morphemes on C are abstractly marked for [Tense], even in the absence of overt past markers -r—except for a set of particles that consistently prove “invisible”—and that positional allomorphy on C is in fact triggered by the [Tense] feature on it, without the need for a [Dependent] feature; finally, that suppletive roots involve the Fusion [root + v]. Throughout, the goal of this attempt has been not only to offer a revealing analysis of Irish verb morphology, but also, in so far as possible, an empirically falsifiable one. The rich domain offered by Irish dialects will provide a testing ground for several of the generalizations stated here, like the principled exclusion of a specific S for past habitual or for conditional (distinct from that of the future), or the claim that DS are positional allomorphs of existing S. In addition, the empirical domain should be broadened to include categories that have been left aside here,

88

Chapter Five

namely the subjunctive (not everywhere moribund), or modal uses of the future. A further direction for development, in a necessarily open list of desiderata, is represented by a closer and textually supported semantic analysis of the values here proposed for Irish tenses, in particular for [Perfective].

References Acquaviva, P. 1996. Irish negation and the representation of monotone decreasing quantifiers. In R. Borsley and I. Roberts (eds.), The syntax of the Celtic languages, 284–313. Cambridge: Cambridge University Press. Basilico, D. 2008. Particle verbs and benefactive double objects in English: High and low attachments. Natural Language and Linguistic Theory 26: 731–773. Binnick, David. 1991. Time and the verb. Cambridge: MIT Press. Bloch-Trojnar, M. 2006. Polyfunctionality in morphology: A study of verbal nouns in Modern Irish (Lublin Studies in Celtic Linguistics 3). Lublin: Wydawnictwo KUL. —. 2008. The morphophonology of verbal nouns in Modern Irish. Éigse 36: 63–81. Cottell, S. 1995. The representation of Tense in Modern Irish. GenGenP 3: 105–124. Demirdache, H. and M. Uribe-Etxebarria. 2000. The primitives of temporal relations. In R. Martin, D. Michaels, J. Uriagereka (eds.), Step by step: Essays on Minimalist syntax in honor of Howard Lasnik, 157–186. Cambridge: MIT Press. Doherty, C. 1996. Clausal structure and the Modern Irish copula. Natural Language and Linguistic Theory 14: 1–46. Doyle, A. 2002. Covert and overt pronominals in Irish (Lublin Studies in Celtic Linguistics 1). Lublin: Wydawnictwo Folium. Duffield, N. 1995. Particles and projections in Irish syntax. Dordrecht: Kluwer. Embick, D. and A. Marantz. 2008. Architecture and Blocking. Linguistic Inquiry 39: 1–54. Embick, D. and R. Noyer. 2001. Movement operations after syntax. Linguistic Inquiry 32: 555–595. Guilfoyle, E. 1990. Functional categories and phase-structure parameters. PhD diss., McGill University.

Tense and Features in Irish Verbal Morphology

89

—. 2000. Tense and N-features in Irish. In A. Carnie and E. Guilfoyle (eds.), The syntax of verb-initial languages, 61–73. Oxford: Oxford University Press. Halle, M. and A. Marantz. 1993. Distributed Morphology and the pieces of inflection, in K. Hale and S. J. Keyser (eds.), The View from Building 20, 111–176. Cambridge: MIT Press. Harley, H. and R. Noyer. 1999. State-of-the-article: Distributed Morphology. GLOT International 4: 3–9. Hughes, A. 1994. Gaeilge Uladh. In K. McCone et al. (eds.), Stair na Gaeilge, 611–660. Maigh Nuad: Roinn na Sean-Ghaeilge, Coláiste Phádraig. Iatridou, S. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry 31: 231–270. Klein, W. 1995. A time-relational analysis of Russian aspect. Language 71: 669–695. McCloskey, J. 1996a. On the scope of verb raising in Irish. Natural Language and Linguistic Theory 14: 47–104. —. 1996b. Subjects and subject positions in Irish. In R. Borsley and I. Roberts (eds.), The syntax of the Celtic languages, 241–283. Cambridge: Cambridge University Press. —. 2001. The morphosyntax of wh-extraction in Irish. Journal of Linguistics 37: 67–100. —. 2007. The grammar of autonomy in Irish. Natural Language and Linguistic Theory 24: 825–857. —. 2009. The shape of Irish clauses. Ms, University of California, Santa Cruz. —. 2010. Polarity, ellipsis, and the construction of finite verbs. Paper presented at the second INFL workshop, Dublin, Dublin City University, 17/9/2010. Mezhevich, Ilana. 2008. A feature-theoretic account of tense and aspect in Russian. Natural Language and Linguistic Theory 26: 359–401. Ó Donnchadha, G. 2010. Syntactic structure building and the verbal noun in Modern Irish: A minimalist approach. PhD diss., University College Dublin. Ó Sé, D. 1991. Verbal inflection in Modern Irish. Ériu 42: 61–81. —. 2001. Gnáthach agus leanúnach i mbriathar na Gaeilge. In B. Ó Catháin and R. Ó hUiginn (eds.), Béalra: aistí ar theangeolaíocht na Gaeilge, 123–145. Maigh Nuad: An Sagart. Ó Siadhail, Micheál. 1989. Modern Irish: grammatical structure and dialectal variation. Cambridge: Cambridge University Press.

90

Chapter Five

Rizzi, L. 1997. The fine structure of the left periphery. In L. Haegeman (ed.), Elements of grammar: Handbook of generative syntax, 281–337. Amsterdam: Kluwer. Smith, C. 1991. The parameter of aspect. Dordrecht: Kluwer. Stowell, T. 2007. The syntactic expression of Tense. Lingua 117: 437– 463. Wigger, A. 1972. Preliminaries to a generative morphology of the modern Irish verb. Ériu 23: 163–213.

CHAPTER SIX WORD-LEVEL PHONOLOGICAL CONDITIONS ON LINEARIZATION IN SCOTTISH GAELIC ANDREW CARNIE 6.1 Introduction It is an incontrovertible fact that phonological operations can make reference to syntactic structures. Phenomena such as nuclear stress, intonation and sandhi have all been shown to be sensitive to constituency, both surface and underlying (Selkirk 1984, 1986 and many others). What is more controversial are phenomena, such as heavy NP shift or focus movement (Szendröi 2003), where it appears as if the syntactic rule is referencing phonological information. Most putative instances have semantic or pragmatic correlates, so the directionality of the causation between syntax and phonology is unclear. Pullum and Zwicky (1988) proposed the principle of Phonology-Free Syntax, which drew the line in the sand against syntactic rules directly referencing the phonological properties of the clause. In this short paper, I present some evidence from Scottish Gaelic of a phenomenon that seems to directly reference word-level phonological properties of the words themselves: the phenomenon of weak pronoun postposing. David Adger (1997, 2007) has written extensively on this topic, and shown that the phenomenon operates in the PF component. He has shown that the positioning of weak pronouns is sensitive to where the nuclear stress of the phrase is. His evidence (presented below) clearly demonstrates the existence of a syntactic phenomenon sensitive to prosody. However, his approach is open to an alternative syntactocentric view, whereby the pronoun postposing is sensitive to precisely the syntactic configurations that give rise to nuclear stress, rather than to nuclear stress itself. In this paper, I add to the mix the fact that pronoun postposing in Gaelic seems to be sensitive to not only syntactically derived nuclear stress, but to the syntax-blind and word-internal stress patterns of

Chapter Six

92

individual lexemes as well. This in turn points towards a principle of linearization that is truly phonologically dependent.

6.2 Pronoun postposing in Scottish Gaelic All the members of the Goidelic group have a productive pattern, where light object pronouns of a particular grade (taking the forms e, i and ’ad in Scottish Gaelic, é, í and (i)ad in Irish, and eh, ee and ad in Manx) shift away from their base positions towards the end of the sentence. Full object NPs cannot do this (as seen in examples 1a-b), but it is the preferred order for pronominal objects (as in examples 1c-d). 1 a

Chuir Muriel am ball air put.PAST Muriel the ball on “Muriel put the ball on the table.”

b

*Chuir

Muriel air

c

?Chuir Muriel e air put.PAST Muriel 3SM on “Muriel put it on the table.”

a’ the

d

Chuir

e

Muriel air

a’ bhòrd V S O PP the table

a’ bhòrd am ball

a’ bhòrd

bhòrd table

*V S PP O ?V S e PP

V S PP e

There have been a variety of theoretical approaches to this phenomenon. Chung and McCloskey (1989) propose that it is a rightward movement rule similar to Heavy NP shift. Duffield (1995) adopts an approach where the pronoun moves to a Wackernagelian second position clitic position, and then the clausal remnant shifts around it into a topic position. Adger (1997) provides extensive critique of these syntactic accounts, so I won’t discuss them in detail here. But Duffield’s account, in particular, suffers from the fact that pronouns don’t postpose after embedded clauses, as would be expected with a remnant approach. Adger (2007) convincingly shows that the phenomenon—in Gaelic— at least is post-syntactic. It appears to have no semantic or pragmatic effects (however, cf. Mulkern 2011, who demonstrates real information structure effects in the cognate Irish construction). His argument works as follows: he claims that VP ellipsis is VP deletion in the PF component. Elements syntactically extracted from the VP survive ellipsis. It then follows that if pronoun postposing is syntactic, then the postposing

Linearization in Scottish Gaelic

93

pronoun should survive ellipsis. It does not (2), so postposing must also be postsyntactic. 2 a

Smaoinich mi gun do dh’fhàg [VP mi mo think.PAST 1S that PAST leave.PAST 1S 1S.POSS leabhar] agus dh’fhàg [VP mi e] aig an sgoil book and leave.PAST at the school “I thought I left my book and I had left it at school.” (Adger 2007) ... V [VP S O] PP

b

*Smaoinich mi gun do dh’fhàg mi mo leabhar think.PAST 1S that PAST leave.PAST 1S 1S.POSS book agus dh’fhàg [VP mi ti] aig an sgoil ei and leave.PAST at the school 3SM “I thought I left my book and I had left it at school.” (Adger 2007) *... V [VP S ti] PP ei

Adger (1997) bases his analysis of pronoun postposing on the Nuclear Stress Rule: 3 Nuclear stress in SG falls: a on the main stress last word of the object DP, unless it is a pronoun b if there is no object, then on the main stress of the final nonfunction word in the VP [Temporal adverbs like today, yesterday, Monday etc. behave as if they are VP-internal for these purposes] There is a technical implementation of this which references mergebased constituent structures, but that need not concern us here, except to note that nuclear stress is determined by referring to syntactic structure. The pronoun-postposing rule (again slightly paraphrased) that Adger proposes is given in (4) 4. Encliticize e-grade pronouns to the non-verb, non-subject word bearing i. focal stress or ii. nuclear stress

Chapter Six

94

Variation in placement of the pronoun comes down to whether the pronoun is attached to the nuclear stressed item (5) or some focal-stressed item (6). 5 Cliticized to nuclear stress Chunnaic Mòrag [ann an] Lunnainn [an +dè] i V S PP Adv i see.PAST Mòrag [in] London [yesterday] 3SF “Mòrag saw her in London yesterday.” (Adger 1997) 6 Cliticized to focal stress. Q. Càit’ am faic Mòrag do mhàthair? where wh-C see.PAST Morag 2S.POSS mother “Where did Mòrag see your mother?” (Adger 1997) A.

Chunnaic i [ann an] LUNNainn i [an dè] V S PP i Adv see.PAST 3SF [in] London 3SF [yesterday] “She saw her in London yesterday.” (Adger 1997)

Elfner (2011) has a related OT analysis of Irish, where an ordering of a weak pronoun at the beginning of a prosodic constituent results in a violation of a prosodic constraint known as STRONGSTART, which is ranked above the LCA (see also López’ 2009 violable LCA), so pronouns prefer to encliticize to the end of final stress bearing prosodic constituents. Both Elfner and Adger’s analyses are important in providing the central insight that stress bears a critical role in determining the position of these pronouns. However, note that the stress principles in question are all ones that refer to the syntactic structure itself. This lays their analyses open to the possibility that the pronoun shifting principles are actually simply sensitive to the same syntactic configurations that the phonological principles of nuclear stress are dependent upon.1 In the next section, I provide some evidence that Gaelic pronoun post-posing requires more phonological information than can be derived from syntactically determined prosody. In particular, I show that pronoun postposing in Gaelic is also sensitive to both the lexically determined syllable count and segmental structure of the words the pronouns attach to.

1

One might object that Adger’s argument from VP ellipsis does not suffer from this complaint, but note that it critically relies on the assumption of PF deletion rather than other common approaches to VP ellipsis, such as LF-copying of the VP.

Linearization in Scottish Gaelic

95

6.3 Lexical conditions on pronoun postposing Linguists often suffer from the conceit that they consciously understand the nature of linguistic structures far better than the native speakers they are working with. Even the most experienced fieldworkers, tempered by years of dealing with prescriptive nonsense in their classrooms, are tempted to dismiss analytic insights from their native speaker consultants as folk myths or half-formed naïve generalizations. This is, alas, a mistake, I think; linguists who do so, do it at their peril. In the case of pronoun postposing, while working with my native speaker, Muriel Fisher—a speaker from the village of Glendale on the tip of the Duirinish Peninsula on Skye—I was dismissive of her characterization of the placement of these pronouns. On one occasion she said, “It’s the rhythm of the balance of the sentence that determines how we’re going to say it.” On another, she said, “It’s affected by the particular words you’re using and the way that they sound.” Coming into this with my preconceived notions of the organization of the grammar, I was at least inwardly (and possibly outwardly) dismissive of these insights. But I have come to believe that they are correct. The placement of Gaelic pronouns is not only sensitive to principles of nuclear and focal stress as observed by Adger, but also to the internal phonological properties of the host words themselves. If this is true, then the phenomenon can’t be covertly syntactic, but must really be determined by the phonology. All else being equal, the postposing process appears to be optional. Speakers will accept (and generate) both of the sentences in (7). When asked to choose, they’ll express a preference for (7b), but most speakers characterize both as “fine”. 7 a

b

Thug Muriel e do give.PAST Muriel 3SM to “Muriel gave it to Paul.” Thug

Muriel do Phòl e

Phòl. Paul

V S e PP

V S PP e

However, the phenomenon appears to become obligatory if failure to move would create an iambic structure as in (8):

Chapter Six

96

8 a

b

× (× ×) *Thug Muriel e dha Dihaoine *V S e PP Adv give.PAST Muriel 3SM to.3SM Friday “Muriel gave it to him on Friday.” × (× ×) Thug Muriel dha e Dihaoine

V S PP e Adv

The obligatoriness of this shift is surprising under either syntactic accounts or Adger’s account, where (8a) would be predicted to be as acceptable as (7a). The picture becomes more complicated when we look at cases where pronoun postposing is blocked, even when the movement would attach the pronoun to a nuclear or focal stressed host. This happens when, for example, the focalized or nuclear stressed element is an underlying trochee or dactyl. In (9a), we have focal stress on the medial syllable of the word Dihaoine ‘Friday’. When the object is changed to a weak pronoun, the preferred position is in situ (9b) rather than in the predicted post-focal stress position in (9c). The only situation in which the postposed pronoun is preferred is when the final vowel of adverb is deleted. The resulting form ends in a trochee, which we will see appears to be the preferred pattern more generally. 9 a

b c d

× (× ×) Phòg Muriel Pòl DiHAOIne kiss.PAST Muriel Paul Friday “Muriel Kissed Paul on Friday” Phòg Muriel e Dihaoine *Phòg Muriel Dihaoine e ?Phòg Muriel Dihaoin’ e

V S e Adv *V S Adv e ?V S Adv e

In (10a–c) we see that an unshifted order is preferred to one where postposing would result in attaching the pronoun to an already existing polysyllable, like the adverb cùramach. By contrast, the shifted order is preferred when attaching to a stressed monosyllable (11a–c), like the adverb tric. In this circumstance again, we are creating a trochaic foot by postposing. Finally, given the choice between attaching to a polysyllable

Linearization in Scottish Gaelic

97

or a monosyllable, the monosyllable is always preferred, no matter what position it appears in, as seen in (12). In each case it is the word-internal phonological structure of the clitic host that determines whether the pronoun can attach or not. 10 a

b c 11 a

b c 12 a

Leugh Bob an leabhar read.PAST Bob the book “Bob read the book carefully.” Leugh Bob ?Leugh Bob

gu cùramach ADV careful

e gu cùramach gu cùramach e

Leugh Bob an leabhar read.PAST Bob the book “Bob read the book frequently.”

V S e Adv ?V S Adv e

gu tric ADV frequent

?Leugh Bob e gu tric Leugh Bob gu tric e

?V S e Adv V S Adv e

?Chic Muriel e gu cùramach gu Pòl kick.PAST Muriel 3SM ADV careful to Paul “Muriel kicked it to Paul carefully.” ?V S e Adv PP

b c

Chic Muriel gu cùramach gu Pòl e ?Chic Muriel gu cùramach e gu Pòl

d

?Chic Muriel e gu Pòl kick.PAST Muriel 3SM to Paul “Muriel kicked it to Paul carefully.”

V S Adv PP e (?)V S Adv e PP

gu cùramach ADV careful ?V S e PP Adv

e f

Chic Muriel gu Pòl e gu cùramach ?Chic Muriel gu Pòl gu cùramach e

V S PP e Adv ?V S PP Adv e

Stress and syllable count are not the only phonological constraints that limit the position of postposed pronouns. Scottish Gaelic generally disprefers sequences of identical vowels across morpheme boundaries. When underlying pronominal position creates such sequences, pronoun postposing seems to be obligatory (13). If the movement would create

Chapter Six

98

such a sequence, either that order is dispreferred (14), or the final vowel of the host is deleted (see 9d above). ?Bhris e e leis Break.PAST 3SM 3SM with.3SM “He broke it with it.”

?V S e PP2

b

Bhris e leis e

V S PP e

c

?Chuir Muriel e air a’ bhòrd put.PAST Muriel 3SM on the table “Muriel put the ball on the table

?V S e PP

d

Chuir Muriel air a’ bhòrd e

V S PP e

13 a

14 a

b c

Dh’fhàgar e ’na laighe PAST.give.IMPERS 3SM in.3SM.POSS lie.VN air an làr V S e AP PP on the ground “It was left lying on the ground.” *Dh’fhàgar ’na laighe e air an làr Dh’fhàgar ’na laighe air an làr e

*V S AP e PP V S AP PP e

Finally, the phonological properties of the pronoun itself determine whether it can be shifted or not. Orthographically there are three pronouns that shift in Scottish Gaelic (e, i, and iad). The last of these is trochaic DOUHDG\ LԥW :KHQLWLVLQWKLVIXOO\IRRWHGIRUPLWGRHVQRWVKLIW:KHQ it postposes, it must take the reduced monosyllabic form ’ad ԥW  2 Note, however, this similar paradigm where a VV sequence appears to be acceptable:

(i) a. Bhris e e leis break.PAST 3SM 3SM with “He broke the chair with the hammer.” b. Bhris e leis an ord e

an the

ord hammer

While (i-b) is preferred to (i-a), example (i-a) did not receive the negative response that (13a) elicited in my native speaker. This may have to do with the fact that the PP here is a full phrase rather than an inflected preposition. The correct analysis of this subtlety eludes me.

Linearization in Scottish Gaelic

15 a

b

Dh’inns i iad do PAST.say 3SF 3PL to “She said them to James.”

99

Sheumais James

Dh’inns i do Sheumais ’ad

6.4 Conclusions What is clear, then, is that not only are syntactically determined nuclear and focal stress patterns critical to understanding pronoun postposing, but that phonological properties of the lexical items themselves play a role. This means that the potential objection to Adger’s analysis such that the phenomenon is covertly syntactic falls apart. The phenomenon really has to be phonological in nature and consequently principles of linearization must apply reasonably late in the derivation, and be flexible enough to include phonological conditioning like those described above.

References Adger, D. 1997 VSO order and weak pronouns in Goidelic Celtic. Canadian Journal of Linguistics 42: 9–29 —. 2007. Pronouns Postpose at PF. Linguistic Inquiry 38: 343–349. Chung, S. and J. McCloskey. 1987. Government, barriers and small clauses in Modern Irish. Linguistic Inquiry 18: 173–237. Duffield, N. 1995. Particles and projections in Irish syntax. Dordrecht: Kluwer. Elfner, E. 2011. The interaction of linearization and prosody: evidence from pronoun post-posing in Irish. In A. Carnie (ed.), Formal Approaches to Celtic Linguistics, 17–40. Newcastle upon Tyne: Cambridge Scholars Press. López, L. 2009. Ranking the Linear Correspondence Axiom. Linguistic Inquiry 40: 239–276. Mulkern, A. E. 2011. Left right behind: Irish pronoun postposing and information structure. In A. Carnie (ed.), Formal Approaches to Celtic Linguistics, 179–202. Newcastle upon Tyne: Cambridge Scholars Press. Pullum, G. and A. Zwicky. 1988. The syntax-phonology interface. In F. Newmeyer (ed.), Linguistics: The Cambridge Survey, vol. 1, 255– 280. Cambridge: Cambridge University Press. Selkirk, E. 1984. Phonology and syntax. Cambridge, MA: MIT Press.

100

Chapter Six

—. 1986. On derived domains in sentence phonology. Phonology Yearbook 3: 371–405. Szendröi, K. 2003. A stress-based approach to the syntax of Hungarian Focus. The Linguistic Review 20: 37–78.

CHAPTER SEVEN LOCAL AGREEMENT IN IRISH AND WELSH: THEORETICAL AND TYPOLOGICAL IMPLICATIONS ALESSIO S. FRENDA 7.1 Introduction* Agreement is a relationship between two linguistic elements, whereby one displays changes of a formal nature in response to some property or properties of the other. Grammatical gender is a morphosyntactic property of the noun; the natural gender of the referent is a semantic property. Agreement itself can reflect either morphosyntactic or semantic properties: in the former case, we may speak, with Corbett (1991), of syntactic agreement; in the latter, of semantic agreement.1 Still following Corbett, we call the noun the agreement controller, since it is the element whose properties influence the form of agreeing elements; the latter are called agreement targets, or simply targets. Since gender distinctions are not always (consistently) marked on the noun, evidence that they exist can only come from agreement (Corbett 1994: 1348). In other words, very often we can only tell that two nouns belong to two different genders if, all other things being equal, agreeing elements consistently appear in different forms. In most cases, the difference between grammatical and semantic agreement is not apparent, either because there is no conflict between * This research was funded by a Government of Ireland Exchange Scholarship and by an IRCHSS Postgraduate Scholarship. 1 Other terms have been proposed: Audring (2009: 16) cites, among the others, “ad formam”, “formal”, “morphosyntactic”, “lexical” and “grammatical” for “syntactic”, and “ad sensum”, “notional”, “logical”, “pragmatic” and “referential” for “semantic”.

102

Chapter Seven

grammatical and semantic gender or because semantic gender is not a relevant property of the controller. However, there are cases in which grammatical and semantic gender appear to be at odds. Textbook examples from Irish and Scottish Gaelic include two grammatically masculine nouns with female referents, cailín ‘girl’ and boirionnach ‘woman’, respectively. As example (1) shows, when this happens we have syntactic local agreement and semantic anaphoric agreement. 1. Irish (Wigger 2000: 1-01-21)2 d’oscail a(n) cailín is a mhac open.PST ART.NOM.M.SG girl(M).SG and oh son nuair a chonaic sí an … when see.PST 3SG.F ART “The girl opened [the door] and boy, when she saw the…” This agreement pattern represents a typologically common solution to the problem of conflicting agreement criteria and is captured by Corbett’s (1991) Agreement Hierarchy. 2. The Agreement Hierarchy (Corbett 1991) attributive < predicate < relative pronoun < personal pronoun The hierarchy predicts that the likelihood of semantic agreement will increase monotonically from the left of the hierarchy to its right; in other words, semantic agreement with a given target in a given language will not be found unless it is also found with all the targets to its right. In this paper, I deal with an innovative agreement pattern which would seem to require the insertion of an additional split within the Agreement Hierarchy, and I argue that such a solution would be inadvisable on both theoretical and typological grounds and then propose an alternative analysis.

7.2 Functional approaches to the study of language Functional linguistic theories see language primarily as a means of communication between human beings in a given social and cognitive context, and therefore as a system shaped by specific cognitive and sociocultural factors (Dik 1986: 3; Butler 2009: 3). The main aim of a 2

See section 7.5.2 for more information on the source of this example and the citation format.

Local Agreement in Irish and Welsh

103

functional theory of language must be, as stated by Dik (1997a: 6), to describe and explain linguistic phenomena by taking into account the instrumental nature of language as a means of social interaction. The natural communicative use that is made of the language is therefore of paramount importance for a functionalist, which requires that samples of language use in natural communication be studied in order to supplement native speakers’ intuition and data collected in experimental settings (Butler 2009: 12–14). In addition, as the focus is on the communicative purpose, functionalists regard syntax as subservient to semantics, i.e. to the expression of some content, as motivated by the pragmatic context (Dik 1997a: 2, 8). This defining perspective on the study of language distinguishes functional approaches from another important linguistic paradigm, usually referred to as “formal” or “generative”,3 which traces its origin to the theories elaborated by Chomsky in the late 1950s: the latter tends to confine the study of variability in language to the background, to reject the use of performance data as a basis for description and theorizing, and to regard syntax as a self-contained system of rules independent from meaning and context. Human language, according to Chomsky (1980: 239), is a “system for free expression of thought, essentially independent of stimulus control, need-satisfaction or instrumental purpose”, and the object of linguistic study is the competence of the ideal language user “in his or her (fictitiously) homogeneous speech community” (Butler 2009: 6). Since competence is assumed to be only dimly reflected by performance (Chomsky 1965: 4), “data about usage and frequency—or even the issue of whether a given possibility is ever actually instantiated in any language” are of “marginal interest” (Anderson 1999: 11). A functional approach therefore appears particularly appropriate for a study of linguistic variation such as the present one, whereas a major difficulty in applying generative principles to such a study is that variation is at odds with the defining notion of an ideal language user in his or her homogeneous community:4 3

Although it has become customary to apply the labels “generative” and “formal” restrictively to this latter paradigm, theoretical models such as Simon Dik’s Functional Grammar (Dik 1978; 1997a;b) and its successors, e.g. Functional Discourse Grammar, are themselves generative theories which employ their own formalisms (cf. Anderson 1999: 2). 4 Indeed, if the sole object of linguistic investigation is the language user’s competence, and if variation is confined to performance, the study of variation is outside the scope of interest of the formalist programme even when it is invoked to explain language change (cf. Harris and Campbell 1995: 37; Matthews 2003).

104

Chapter Seven Variation exists because speech communities are not homogeneous, and speaker-listerners in speech communities must be able to negotiate a variety of styles, dialects, registers, and the like in their use of the language. (Kallen 1986: 81)

The study of variation data from a functional-typological perspective has already proved useful to provide linguistically relevant explanations to the distribution of grammatical markers. Cases in point are Siemund (2008) and Audring (2009), whose studies, conducted on various types of corpus evidence, were able to explain the distribution of pronominal gender exponents in a number of English varieties (Siemund) and Dutch (Audring) in terms of factors that are cognitively, semantically and in some cases syntactically relevant: degree of individuation and distinctions such as mass vs. count, concrete vs. abstract and animate vs. inanimate. At the same time, explicit formulation of the facts and structures under inverstigation, as well as of any relevant hypothesis, is clearly desirable when descriptive adequacy is being sought: this is why a solid formal infrastructure is necessary. Functional Discourse Grammar (FDG) (Hengeveld and Mackenzie 2008) is what Butler (2008: 2) calls a “structural-functional grammar”, a term by which Butler alludes to the generative character of the theory, i.e. to the fact that it provides “sets of rules and principles” to assign “structures and interpretations to linguistic expressions” (Butler 2003: 30; cf. Chomsky 1965: 4). Structural-functional grammars therefore combine the advantages of a functional approach with the type of rigorous formalism that is necessary to attain descriptive adequacy. A more detailed treatment of FDG would be impossible within the limited scope of this article; the reader is referred to the presentation of the theory made by Hengeveld and Mackenzie (2008).

7.3 The Insular Celtic language group Insular Celtic is a branch of Proto-Celtic, itself a reconstructed member of the Indo-European language family; it includes two subgroups, Goidelic (or Gaelic) and Brythonic (or Brittonic or British). To the former belong modern Irish, Scottish Gaelic and Manx, while Welsh, Breton and Cornish continue Brythonic (Fife 1993: 6; Watkins 1999).

Local Agreement in Irish and Welsh

105

All of these languages have or had grammatical gender systems in which nouns can be masculine or feminine;5 those that are still natively spoken are nowadays minority languages, spoken by bilingual speakers who have either learned them as another L1 (alongside English, French, or Spanish)6 or as an L2; some of them, like Irish and Welsh, enjoy official status and have been the object of efforts aimed at revitalization and revival, but may nevertheless be described as obsolescent according to Jones’s definition of language obsolescence as a “gradual reduction in use, due to domain-restriction, [which] may result in the emergence of historically inappropriate morphological and/or phonological forms” (Jones 1998: 5f.). Others, like Cornish and Manx, are no longer spoken natively but are officially recognized and supported at least to some extent (Akutagawa 1987; Dorian 1994; George and Broderick 1993; Watson 1989).

7.4 Gender agreement in Insular Celtic Gender agreement in Insular Celtic involves a number of targets within the controller’s noun phrase, such as the article and various attributive elements, as well as pronominal targets, and is limited to the singular. In Brythonic, the number of different agreement target types is greater than in Goidelic, in that it includes, at least in more conservative varieties, agreeing elements such as deictics and certain numerals which are not inflected elements in Goidelic; but apart from the range of agreeing elements, the agreement systems of Goidelic and Brythonic are largely comparable. Within the noun phrase, agreement is shown predominantly by initial mutations, a particularly salient feature of Insular Celtic (cf. Hamp 1951: 230). The initial mutation of a word involves either a change in its initial consonant, or the prefixing of a limited set of consonants to vowel-initial words. Historically, they represent the fossilized continuation of previously productive sandhi phenomena, like intervocalic lenition (Hamp 1951; Kortlandt 1982; Hickey 1995).7 5

There is evidence, both direct and indirect, that Proto-Celtic had inherited the Indo-European neuter gender as well: cf. Thurneysen (1980: §245), Lewis and Pedersen (1974: 159). 6 Most Insular Celtic languages are, or were prior to extinction, in contact with English; Breton is in contact with French, and Patagonian Welsh, in Argentina, with Spanish. 7 Although initial mutations may be used to mark agreement, they also appear in a number of unrelated contexts, including ones where they serve no apparent

Chapter Seven

106

In the case of article–noun agreement, as in (3) for Brythonic and (4) for Goidelic, it is the noun, i.e. the agreement controller, which is marked. 3 Welsh cath / y cat(F).SG ART “(a) cat / the cat”

gath cat(F).SG

4 Irish céim / an chéim degree(F).NOM.SG ART.NOM.F.SG degree “(a) degree / the degree” Examples (3) and (4) also show how initial consonant mutation works: for instance, initial /k/ (c-) corresponds to soft-mutated /g/ in Welsh and to lenited /x/ (ch-) in Irish. In these examples, all other conditions being equal, the presence or absence of mutation depends on the grammatical gender of the noun, but case (in Goidelic) and number are also relevant to the occurrence of this phenomenon: for instance, no gender-dependent mutation occurs in the plural, and while in the nominative singular mutation affects feminine nouns, in the genitive singular it affects masculines, as illustrated in (5).8 5 Irish a an chinn ART.GEN.M.SG head(M).SG “of the head” b na ART.GEN.F.SG “of the degree”

céim-e degree(F)-GEN.SG

purpose. In example (i-a), lenition has a distinctive function, being used to mark the gender of the possessor: its presence or absence causes different interpretations. In (i-b), on the other hand, it has no distinctive function: its absence would be unexpected, but not meaningful. i

8

Irish béal EƍH‫ޝ‬OµPRXWK¶ a a bhéal /ԥ YƍH‫ݱޝ‬/ ‘his mouth’ : a béal /ԥ EƍH‫ݱޝ‬/ ‘her mouth’ b mo bhéal /mԥ YƍH‫ݱޝ‬/ ‘my mouth’ : ?mo béal /mԥ EƍH‫ݱޝ‬/ ‘my mouth’

Welsh does not have nominal case inflection, which is why case is not specified for Welsh in the interlinear glosses.

Local Agreement in Irish and Welsh

107

In Goidelic, but not in Brythonic, agreement between the noun and the article may also be marked on the article itself, as also shown in example (5), where the masculine form of the article (genitive singular), an, contrasts with the feminine form, na. Lenition is also used to mark grammatical gender with adjectives, as in (6). 6 Irish bean mhaith woman(F).NOM.SG good.NOM.F.SG “a good woman” (cf. unmutated maith ‘good’) In Functional Grammar, agreement is modelled as feature copying from a source to one or more targets (Dik 1997a: 377; Hengeveld and Mackenzie 2008: 350, 394), i.e. from the constituent to which a feature or set of features is inherent to other constituents; this is a model which clearly emphasizes the asymmetric nature of agreement and the displacement of information that it involves (Corbett 2006: 114f.). Corbett (1991) uses the term “controller” for Dik’s “source”, a terminological choice also adopted here. More specifically, according to this model, agreement targets are first introduced in the derivation as underspecified abstract forms (“placeholders”) during morphosyntactic encoding; then, the relevant features are copied to them from the controller, at which stage the placeholder can be replaced by the appropriate phonological representation during phonological encoding (Hengeveld and Mackenzie 2008: 352). This is exemplified in (8) for the phrase in (7); phonological notations indicate forms that have already been finalized at a given point in the derivation, orthographic notations forms that need further encoding. 7 Cois Fhairrge Irish (Ó Siadhail 1989: 113) móin dhubh peat(F).NOM.SG black.NOM.F.SG “black peat” 8 Derivation of (7) IL: (–id R) (assuming the noun phrase is used referentially) RL: (mxi : [(fi: móinNf (fi)) (xi)]: [(fj : dubhA (fj)) (xi)] (xi)) ML: 1S>1ZPR‫݆ޝ‬ƍ1Z $S$ZGXEK$Z $S I!@1S PL: (PPi : [(PWi PR‫݆ޝ‬ƍ3:i)) (PWj : /ܵX‫ޝ‬3:j))] (PPi))

Chapter Seven

108

The representations in (8) read as follows: at the Interpersonal Level (the level of pragmatics), the element is represented as non identifiable (– id) and used referentially (R). At the Representational Level (i.e. the level of semantic encoding), an entity characterized as a mass (mxi) is represented, of which a Property (fi) lexically specified by móin is predicated. Another Property (fj) is predicated of xi; this second Property is lexically specified by an adjective (dubhA). The RL representation further indicates that móin is a Lexeme belonging to the class of feminine Nouns (Nf). At the Morphosyntactic Level, móin dhubh is represented in terms of syntactic constituency, i.e. as a Noun Phrase (Np) made up of a Noun Word (Nw: móin) and an Adjective Phrase (Ap), the latter containing an Adjective Word (Aw: dubh). The Adjective Phrase is represented as having the feature (feminine), which pertains to the Lexeme móin, copied on it. The feature must be copied onto the Adjective Phrase as a whole, since (in Irish as in Welsh) the initial mutation, where applicable, applies to the leftmost word in the phrase, which may or may not coincide with its head. Agreement between the article and the noun, as we have seen, may be marked on the controller. A possible way of modelling this is by positing distinct underlying forms of the article which are selected after feature copying and are responsible for causing the following noun to mutate, as in example (9); this is possible under various morphophonological/ autosegmental analyses such as those discussed below. 9

Welsh Irish

underlying representation /ԥSM ‫ޖ‬NDș /ԥ݆L ‫ޖ‬NƍH‫ޝ‬Pƍ

surface realization /ԥ‫ޖ‬JDș /ԥ݆‫[ޖ‬ƍH‫ޝ‬Pƍ

(y gath ‘the cat’) (an chéim “the degree”)

Under this view, there would be a placeholder for the article inserted at the ML, to which the relevant feature is copied; this results in a form of the article bearing the appropriate autosegment, as follows. 10 ML: an PL: /ԥ݆L/ However, a similar model presents a number of problems when considered from a synchronic point of view. To begin with, initial mutations do not always continue erstwhile sandhi lenition. For instance,

Local Agreement in Irish and Welsh

109

attributive adjectives do as a rule mutate after all feminine nouns, including those which historically terminated in a consonant: initial mutation was intepreted as a mark of agreement on adjectives and generalized to all appropriate contexts (McCone 1994: 120f.). Equally disjoined from their historical causes are examples where the inital mutation is distinctive and occurs independently of any preceding word, as in example (11). 11 Irish a EƍUƍLVƍ H‫ޝ‬

break-IMP.2SG 3SG.M “break it” b

/vƍUƍLVƍ break-PST “he broke”

VƍH‫ޝ‬

3SG.M

Nevertheless, attempts to provide a unified and purely phonological explanation to this phenomenon have been numerous, based either on abstract (non-surfacing) phonemes or on diacritics and readjustment rules. Kallen (1979: 19–27) and Stewart (2004: ch. 3) survey a number of them, pointing out their ad-hoc nature, the fact that they lead to a proliferation of underlying abstract forms, and the impossibility of subjecting any of these hypotheses to empirical testing. An attempt to overcome the limitations of a purely phonological analysis by overtly involving morphology is represented by the morphophonological approach, first proposed by Hamp (1951), who posits distinctive elements (morphophonemes) whose presence triggers the initial mutation of the following word. These elements are quasi-segmental in that they do not have autonomous phonological realization; rather, their presence is revealed only by the modification of the following sound. In this sense their operation pertains to phonology, but their distribution does not, since it may be either lexical or morphological.9 In a similar model of mutation, it is possible to represent the two synchronically distinct forms of the article as an (masculine singular) and anL (feminine singular), as in (12).

9

As both Kallen (1979: 29) and Stewart (2004: 39) further observe, morphophonemes are positioned a posteriori, based on the observed distribution of mutated segments, rather than a priori, based on some independent motivation for their presence.

Chapter Seven

110

12 Irish NP an béal an bean

underlying representation /ԥ݆ EƍH‫ݱޝ‬/ /ԥ݆L EƍD݆/

surface realization /ԥ݆ EƍH‫ݱޝ‬/ /ԥ݆L va݆/

Following Hamp (1951) and Oftedal (1962), superscript L represents the morphophoneme which triggers lenition. In a similar fashion, superscript SM represents the Soft Mutation trigger and so on. A similar analysis is proposed by Lieber (1983), who considers mutations to be triggered by a “floating autosegment” which attaches to, or is located at, the end of a word and instructs the phonological component to change the initial segment of the following word. Green (2006: 1948) points out two major difficulties of morphophonological approaches: (i) there is no single phonological feature or bundle of features in whose terms “the wide variety of alternations found within a single mutation” may be described; (ii) the phonological component does not treat all morphophonemes equally; sometimes it just ignores them, or skips them. Irish lenition provides good examples of both problems. For instance, lenition “turns oral stops and m (but not n) into fricatives, debuccalizes coronal obstruents, ‘laxes’ tense coronal sonorants, and deletes f” (Green, 2006: 1959): [+continuant] alone will trigger only the spirantization, not the debuccalization, sonorant laxing, or f-deletion. If a case can be made that what distinguishes җ җƍ Ѭ Ѭƍ from O Oƍ Q Qƍ is the feature [tense] […] then [–tense] could account for the sonorant laxing, but not the other cases. And it is difficult to conceive of any feature that could be added to f to induce deletion. (Green 2006: 1959)

In other words, “[u]nlike truly phonological processes […], the mutations do not target natural classes of sounds or have uniform effects” (ibid.). The second problem is exemplified in (13). 13 Irish (Green, 2006: 1966) a cuid ‘part’ b árN gcuid ‘our part’ c dháL chuid ‘two parts’ d. árN dháL gcuid ‘our two parts’

Local Agreement in Irish and Welsh

111

As this example shows, árN and dháL trigger nasal mutation and lenition, respectively, as expected; but when both are present, as in (13d), the noun is nasalized rather than lenited, despite following dháL. While there might be possible explanations for (13d), e.g. that the semantic expression of the possessor overrides adjacency, phonology per se is blind to such considerations.10 In other words, the observed mutation patterns need to be explained in syntactic and possibly semantic terms: phonology alone cannot account for all the data (Borsley et al. 2007: 192). In addition to these theoretical difficulties, the morphophonological model runs into another problem, posed by the emergence of differential agreement patterns within the noun phrase. Hints to this phenomenon are found in the literature, with reference to both surviving and extinct varieties. In Irish, word-final inflection of the adjective in the singular “has largely been abandoned”, mostly surviving in set phrases (Mac Eoin 1993: 115f.). In spoken Welsh, adjective agreement by vowel alternation, while still found in the literary language, is now quite rare, especially among younger speakers (Jones 1998: 68, 173f.; Thomas 1992b: 292; Thorne 1993: 194; Watkins 1993: 312); in addition, a tendency has been reported to leave attributive adjectives and especially attributive nouns unmutated, or to mutate them when not appropriate, as in (14) (Jones 1998; Thomas 2001; Thomas and Gathercole 2005).11 14 Welsh (Thomas, 2001: §2.11.3) a y gath du ‘the black cat’ (expected: y gath ddu) b y pêl goch ‘the red ball’ (expected: y bêl goch) The Agreement Hierarchy (2), as discussed above, predicts that if multiple agreement forms are allowed, semantic agreement will be more likely for targets on the right-hand side of the hierarchy, but it has no further position to the left of the attribute’s for determiners; Dikker (2004: 10

In spoken Irish, moreover, a certain amount of variation may be found: a small survey of Gaeltacht speakers revealed that árN ndá cuid (with no “trigger skipping” and no mutation after the numeral) and ár dháL chuid (with lexicalized mutation on the numeral and the regular mutation after the numeral) are also accepted forms (Denis Kelliher, personal communication). 11 In Manx (Thomson,1992) and Cornish (George 1993; Thomas 1992a), neither of which is spoken natively any longer, gender marking within the noun phrase had completely fallen in disarray, with both initial mutations eventually and article inflection (Manx) being no longer applied productively and surviving mostly in set phrases (Broderick 1984: 25; Thomas 1992a: 368; Thomson 1992: 119).

112

Chapter Seven

39) explicitly includes the latter “under the attributive agreement targets” in her FDG-based model of agreement. From a theoretical point of view, Dikker’s (ibid.) model inteprets the typological possibilities summarized in the Agreement Hierarchy in terms of syntactic structure: the more phrasal borders intervene between the agreement controller and a particular target, the more likely it is “for semantic information to take the lead over syntactic information” in the selection of the appropriate agreement form; to add a further position to the hierarchy just to the left of the attribute would entail to posit a further type of phrase which contained just the noun and the determiner, but not the attribute, as shown in (15). 15. [[Article Noun]?p Adjective]Np This hypothesis would posit a problematic syntactic constituent, here labelled “?p”, within what we assume to be a noun phrase. The problem is that since the constituent labelled Np is obviously not an adjective phrase, its head cannot be the adjective: this would leave to the absurd conclusion of a phrase without a head. A constituent like ?p in (15) is therefore of dubious theoretical validity, which means that the difference in agreement rate we have observed cannot be explained by positing one. The problem can therefore be summarized as follows: if syntactic agreement is “a purely syntactic operation, involving the copying of syntactic features of controllers onto targets” (Dikker, 2004: 37) and sensitive to syntactic boundaries, then the difference in agreement rates exhibited by the article and the attribute is unexpected, since the syntactic distance between the article and the noun is the same as that between the noun and the attribute. The purpose of this study is therefore to ascertain whether the difference in agreement rates between the article and the attributive adjective can be said to be statistically significant in Irish and Welsh. This study of usage variation is based on data from actual spoken usage; the data sets employed for this purpose are described in the following section.

Local Agreement in Irish and Welsh

113

7.5 Data sets and methodology The data on which my analysis was carried out pertain to the colloquial register and come from two types of source: 1. interviews broadcast in the context of Irish- and Welsh-language radio shows; 2. dialectological studies. Type-1 data were recorded in the last decade and represent contemporary usage; the material to include was selected on the basis of external criteria only, as recommended by best practice (cf. Sinclair 2005), namely context (radio broadcast) and type of interaction (informal conversation). The size of the contemporary component of each corpus (in the region of 20,000 words—see below) approaches that of the “Broadcast Interviews” component (20,000 words) in the design of the International Corpus of English.12 The use of broadcast material in linguistic inquiry is discussed by Labov (1972: 211) and applied to Welsh by Ball (1982). This type of data is representative of the language spoken in a semi-institutional and nationwide context, in which various types of speakers may be heard. Three types in particular are identified by Ó Giollagáin and Mac Donnacha (2008: 111f.): (a) “traditional” native speakers; (b) non-native (i.e. L2) speakers; (c) “non-traditional” native speakers, i.e., those whose parents are not traditional native speakers, but who still regard themselves as native speakers, having being raised through Irish (typically outside the Gaeltacht), and who have subsequently attended Irish-language schools. To include non-native and “non-traditional” native varieties was deemed crucial on account of the considerable exposure that these varieties get through the media. Type-2 data represent the variety of older generations and consist of two sets of recordings, involving speakers who were born around the beginning of the 20th century: a collection of recordings from Connemara, made in 1964 (Wigger 2000) and a collection of transcribed recordings 12

http://ice-corpora.net/ice/design.htm.

114

Chapter Seven

from a dialectological survey of Welsh (Thomas and Thomas 1989). Because of the constraints imposed on this work in terms of time and resources, the choice of the material to include as type-2 was restricted to conversational recordings for which transcriptions were already available. The data on traditional native speakers (older components) is meant to provide a term of comparison against which to quantify and qualify the degree of innovation exhibited by the gender system of the contemporary varieties, i.e. the varieties that are heard and accepted as contemporary Welsh and Irish, and not just the language spoken by traditional native speakers. For reasons of space only a summary description of these data sets is possible here; further details may be found in Frenda (2011).

7.5.1 The Welsh corpus The Welsh corpus includes two components. The contemporary Welsh component (CC-W) is a collection of interviews and informal conversations; it is entirely transcribed and freely available online as part of the Talkbank database, along with the original recordings.13 For the present study, an hour and a half of spoken interaction (contained in five files) was analyzed. The older component (OC-W) includes fourteen recorded conversations with different informants, along with their transcriptions, and runs for about one hour. 7.5.1.1 The older component (OC-W) OC-W contains the material published in the Enghreifftiau (‘Examples’) section of Thomas and Thomas (1989: 86–155). It includes 14 conversations, each with a different speaker and from a different location in Wales, totalling about 9,400 words. The speakers were all born between 1891 and 1922, except one speaker who was born in 1943, but whose data were included because they do not differ significantly, in pattern, from the other speakers’. His contribution is also quite limited. The date of the various recordings is not indicated.

13 http://talkbank.org/data/BilingBank/Bangor/Pilot.zip, last accessed 21 November 2010.

Local Agreement in Irish and Welsh

115

7.5.1.2 The contemporary component (CC-W) CC-W includes about 26,200 words and is comprised of five recordings from the BBC Welsh-language radio channel, Radio Cymru, broadcast in 2003 and 2004. There are 13 different speakers featured in this data set. Based on the estimates provided, they were born between 1945 and 1975. Excerpts from this component are identified by the name of the file in the online repository and line number (e.g. “beti.cha 38” refers to line 38 of the file beti.cha).

7.5.2 The Irish corpus The Irish corpus also includes two components. A corpus of presentday Irish usage comparable to the Welsh one was not available and was built as part of this study using the same type of sources. The sources were both Irish-language radio stations and Irish-language shows broadcast by English-language stations. This component contains excerpts from eleven radio shows, featuring interviews and conversations with different guests and conducted by different presenters, and runs for approximately one hour and 40 minutes. Because there are in Ireland statutorily defined Irish-language districts (the Gaeltachtaí), this component includes both data from Dublin-based radio stations and data from the Gaeltacht-based Raidió na Gaeltachta. The older generations’ variety is represented by a selection of the transcriptions published by Wigger (2000); the corresponding audio files are freely available from the website of the Dublin Institute for Advanced Studies.14 7.5.2.1 The older component (OC-I) OC-I is comprised of six texts, totalling 12,200 words approximately. All recordings were made as informal conversations in 1964 and all speakers were males from Ros Muc in the Connemara Gaeltacht. All were born between 1896 and 1937. Every excerpt from this component is identified by the bibliographical reference “(Wigger, 2000)” followed by the three-number code given to it by the corpus editor, e.g. 4-01-02. The first number (4 in this example) refers to the conversational session (referred to as caint by Wigger, 2000) in which the excerpt is found; the second (01 in the example) to a specific 14 http://www.celt.dias.ie/publications/online/caint_ros_muc/, last accessed 21 November 2010.

116

Chapter Seven

part in the caint (each part has a title which refers to the topic of the conversation); the third number (02 in the example) identifies the stretch of discourse in which the example is found (this subdivision is found in Wigger (2000) and refers to stretches generally not exceeding one minute in duration). 7.5.2.2 The contemporary component (CC-I) CC-I contains eleven texts and about 20,000 words. It includes seven texts from Raidió na Gaeltachta (RnaG, approximately 13,000 words, broadcast in 1997 and 2007) and four from non-Gaeltacht sources (approximately 7,000 words, broadcast in late 2006 and 2007). Excerpts from CC-I are identified by citing the source (radio station and programme name) and the date of the recording in the format dd/mm/yyyy (e.g. RnaG “Cumarsáid” 06/11/1997).

7.5.3 Data retrieval and statistical testing After each text was checked against the corresponding recording to make sure the relevant details had been transcribed faithfully, the data was scanned manually to identify and classify the relevant instances of agreement according to whether (i) the expected form was found, (ii) an unexpected form was found, or (iii) it was impossible to decide (e.g. if the form was unintelligible). Each observation was recorded in a database. All observations classified as uncertain or ambiguous were discarded. The agreement rate for each target type was calculated as the proportion of realizations of grammatical gender agreement out of the total number of observations. The statistical test chosen for the difference between agreement rates is Fisher’s exact test, since some comparisons are based on small sample sizes (cf. Baayen, 2008: 113). A difference was regarded as statistically significant for p < 0.05.

7.5.4 Semantic vs. non-semantic agreement A further distinction was made between semantic and non-semantic agreement. If information (lexical or otherwise) about the sex of the referent (when relevant) was available to the speaker; otherwise it was counted as representing non-semantic agreement.

Local Agreement in Irish and Welsh

117

7.6 Variation in gender agreement within the noun phrase Due to space limitations, only a cursory overview of the relevant findings will be given in what follows, limited to the two local agreement targets for which a larger number of observations was available: articles and attributive adjectives. A complete exposition of the findings relative to these and the remaining target types can be found in Frenda (2011). The traditional system of gender agreement within the noun phrase is well reflected in OC-I and OC-W. As regards CC-I and CC-W, the picture emerging from the data is at variance with the traditional system, especially as regards the treatment of prenominal agreement targets as opposed to postnominal ones. Grammatical gender agreement between the article and the noun is regularly observed with both masculine and feminine controllers in both OC-I and CC-I, although in the latter it is significantly less frequent with feminines than with masculines. In OC-I, agreement rate with the article is almost the same for both masculine (98%) and feminine controllers (97%), with no statistically significant difference, while in CC-I it appears to be significantly more frequent with masculine (97%) than with feminine ones (88%), as shown in Table 7.1. Table 7.1. Agreement with the article in Irish

Agreeing Total % agreeing

OC-I masc. fem. 227 107 232 110 98 97 p = 0.7154

CC-I masc. fem. 176 136 182 155 97 88 p = 0.0028

Agreement rate with feminine controllers is higher in OC-I than in CCI (97% vs. 88%, respectively), which is a statistically significant difference (p = 0.0059) not observed in the case of masculine controllers (p = 0.546). No evidence of any significant difference in agreement rates was found in the Welsh corpus in relation to the article: the expected agreement forms appear more often than 95% of the time with controllers of either gender, in both OC-W and CC-W (Table 7.2).

118

Chapter Seven

Table 7.2. Agreement with the article in Welsh

Agreeing Total % agreeing

OC-W masc. fem. 174 51 174 53 100 96 p = 0.0537

CC-W masc. fem. 176 136 182 155 97 88 p = 0.2108

Article agreement with nouns of both genders also appears to be stable across the two components. As regards the attributive adjective as a target of gender agreement (Table 7.3), no statistical evidence of any between-gender difference in agreement rates was found in OC-I: grammatical agreement was observed 94% of the time with masculines and 92% with feminines, a negligible difference (p = 0.6853). In CC-I, on the other hand, adjectives grammatically agree with masculines significantly more often than they do with feminines (p < 0.0001): the rate of grammatical agreement is 91% with masculines and only at 45%, i.e. less than chance level, with feminines. Table 7.3. Agreement with the adjective in Irish

Agreeing Total % agreeing

OC-I masc. fem. 49 33 52 36 94 92 p = 0.6853

CC-I masc. fem. 51 9 156 20 91 45 p < 0.0001

As already observed in relation to the article, the agreement rate of attributive adjectives with feminine controllers is significantly higher in OC-I than in CC-I (92% and 45%, respectively, p = 0.0002), which is not the case with masculine controllers (p = 0.7175). In Welsh, syntactic agreement in OC-W is slightly more frequent with masculine (100%) than with feminine controllers (96%), but the difference between these two agreement rates is not statistically significant (p = 0.4286). In CC-W, on the other hand, syntactic agreement with feminine controllers was found to be statistically less likely than with masculine ones (77% vs. 99%, p < 0.0001).

Local Agreement in Irish and Welsh

119

Table 7.4. Agreement with the adjective in Welsh

Agreeing Total % agreeing

OC-W masc. fem. 36 26 36 27 100 96 p = 0.4286

CC-W masc. fem. 82 47 83 61 99 77 p < 0.0001

As was already the case for the Irish data set, syntactic agreement with feminines turned out to be significantly more frequent in OC-W than in CC-W (p < 0.05), while no significant between-component difference was observed in the case of masculine controllers (p = 1).

7.7 An FDG-based model of gender agreement for Insular Celtic In the light of the findings presented in section 7.6, a model of gender agreement in Celtic should be able to explain the fact that agreement, in contemporary speakers, is not marked consistently on all types of target within the noun phrase. In FDG, agreement is described as a process whereby “information properly pertaining to a single element of the Clause (or of the Phrase) is copied to one or more other elements” (Hengeveld and Mackenzie 2008: 350, 394). This is referred to as the feature-copying model of agreement. The terminology employed by Hengeveld and Mackenzie is broad enough to cover both syntactic agreement (which depends on grammatical information) and semantic agreement (which depends on semantic information). The characterization of agreement as the copying of information or features captures the asymmetric nature of agreement, based on the displacement of some information, which appears to be a fairly consensual notion in current models of agreement (Corbett 2006: 114–116). In FDG, agreement rules are applied by the morphosyntactic encoder, after all the slots in the clausal template have been filled, either with lexical material or with placeholders: this ensures that all agreement controllers and all agreement targets are already in place and available (Hengeveld and Mackenzie 2008: 350f.). The final PL representation of the agreement target (initially an underspecified abstract form) depends on features that are inherent to the agreement controller; the agreement rules copy these features to the agreement target and the phonological encoder

120

Chapter Seven

then selects the appropriate form of the latter based on the copied features (ibid., p. 352). Agreement between the noun and the attributive adjective was modelled as in (8) above. It is now necessary to clarify what is involved in what we have referred to as article–noun agreement. As illustrated in section 7.4, agreement between the article and the noun is marked on both the dependent and the head in Goidelic, only on the head in Brythonic. Here we add that the form of the article may depend on the initial segment of the following word. In Welsh, for instance, and simplifying somewhat, ’r occurs after a vowel when the article cliticizes to the preceding word; otherwise, yr occurs before a vowel and y before a consonant. This is illustrated by examples (16a), (16b) and (16c), respectively. 16 Welsh a gyda’r ci with=ART dog “with the dog” b yr afon ART river “the river” c y ci ART dog “the dog” If y is the bare lexical form of the article, as is commonly assumed (Watkins 1993: 313), then its final phonological representation will be available at the Morphosyntactic Level in the case of (16c), but only at the Phonological Level in the case of both (16a) and (16b). If on the other hand as the bare lexical form is yr, then the article will be finalized at the Morphosyntactic Level in (16b) and only at the Phonological Level in the other two examples. In Irish, the situation is more nuanced in that stylistical factors and formality, as well as the dialect, may affect the realization of the article, specifically as regards the realization of the final /n/ (cf. Ó Curnáin, 2007: 1330–1332). What seems to be regular, however, is the deletion of /n/ when the article occurs between two consonants within a larger constituent (cf. Ó Curnáin, 2007: 1330 for a Connemara variety and Ó Sé 2000: 166f., 176–197 for a Munster variety): cf. an sagart (17a) vs. bean a(n) tí (17b).

Local Agreement in Irish and Welsh

121

17 Irish (Ó Sé, 2000: 184, 191) a an sagart ART priest “the priest” b bean a’ tí woman ART.GEN house “the housewife” Assuming that the bare lexical form of the article is an, this is identical to its final phonetic representation in (17a), but not in (17b); so the article would be finalized at the Morphosyntactic Level in the former case, but only at the Phonological Level in the latter. This state of affairs can be represented as in (18); as usual, words presented orthographically symbolize phonologically non-finalized representations. 18 Irish an fear cliste ‘the clever man’ IL: (+id R) (assuming the noun phrase is used referentially) RL: (1xi : [(fi: fearNm (fi)) (xi)]: [(fj: clisteA (fj)) (xi)] (xi)) ML: 1S >*Z DQQRPPVJ! *Z  1Z IƍDU 1Z  $S $Z cliste (Aw)) (Ap))] (Np)) PL: (PPi : [(PWi : /ԥ݆IƍDU3:i)) (PWj NƍOƍLVƍWƍԥ/ (PWj))] (PPi)) The copying of morphosyntactic features onto the article, indicated in (18) as a Grammatical Word (Gw), can only account for the selection of its own form (an vs. na in Irish), but not for the initial mutation of the noun, since we have rejected the autosegmental trigger hypothesis. On the other hand, since the noun only mutates when it is preceded by the article, its mutation is not triggered just by a specific bundle of morphosyntactic features (e.g. ) interacting with definiteness (represented as +id in the IL representation): in the absence of the article, the combination of all these features does not trigger mutation, as shown in examples (19a) and (19b), where both Cáit and bean are nominative feminine singular nouns heading definite noun phrases yet neither is lenited.15 19 Irish a Tá Cáit / *Cháit in-a múinteoir anois be.PRS Cáit(F).NOM.SG in-POSS:3SG.F teacher now “Kate is a teacher now.” 15

In both Irish and Welsh an embedded definite noun phrase causes the matrix noun phrase to be interpreted as definite, hence the head of the latter need not, and cannot, be preceded by an article: cf. in example (19b) [bean [an tí]] vs. *[an bhean [an tí]].

Chapter Seven

122

b

Bhí bean / *bhean be.PST woman(F).NOM.SG tí deas house(M).GEN.SG nice “The housewife was nice.”

an ART.GEN.M.SG

(The same considerations apply to Welsh.) In both Irish and Welsh the effect of initial mutation can be the deletion of an initial consonant (e.g. /f/ in Irish under lenition, /g/ in Welsh under soft mutation); this means that a noun beginning in a consonant may surface as beginning in a vowel, which will in turn affect the form of the article. In other words, the mutated form of the noun must be available before the form of the article is finalized. This is suggested by Hannahs and Tallerman (2006: 781), in whose model content words, like nouns, are inserted in the syntactic derivation before grammatical words, like the article. For instance, the derivation of Welsh yr ardd ‘the garden’ is modelled as in (20). 20 Welsh yr ardd ‘the garden’ (Hannahs and Tallerman, 2006: 808) syntactic structure: [Article Noun] i) insertion of content word ___ gardd ii) triggering of mutation ___ ardd iii) insertion of article yr ardd In (20), a gap (___) represents the placeholder for the article, whose phonological form is finalized only in the last step; in the FDG model, this corresponds to the Phonological Level. However, we also need to assume that both steps (ii) and (iii) in (20) (triggering of mutation and insertion of article) take place at the Phonological Level. The alternative would be to say that the mutated form of the noun is already available at the Morphosyntactic Level, a situation which could be represented as in (21). 21 Welsh yr ardd ‘the garden’ IL: (+id R) (assuming the noun phrase is used referentially) RL: (1xi: [(fi: garddNf (fi)) (xi)] (xi)) ML: (Np: [(Gw: y (Gw)) (Nw: /arð/ (Nw))] (Np)) PL: (PP: [(PW: /ԥU‫ޖ‬DUè3: @33 The problem with (21) lies in the implication that /arð/ is the basic form of the noun, which is untenable for morphosyntactic reasons (the

Local Agreement in Irish and Welsh

123

mutated form would not appear if the word was cited in isolation) as well as historical reasons (we know that the mutated form is secondary as it derives from the grammaticalization of a sandhi phenomenon). Nevertheless, all the morphosyntactic information needed to finalize the form of the noun (gender, number, case and whether there will be an article in the final utterance) is already available at the Morphosyntactic Level and can therefore be sent down to the Phonological Level, where the selection of the appropriate form of the article is possible because the form of the following noun can already be determined. Given an ML representation such as that in (22), the Phonological Encoder delays the finalization of the form of the article, “looks ahead” to ascertain the final representation of the noun, and then finally encodes the article. 22 ML: (Np: [(Gw: y (Gw)) (Nw: gardd (Nw)) (Ap: (Aw: mawr (Aw)) (Ap))] (Np)) PL: (PP: [(PWi : /ԥU‫ޖ‬DUè3:i)) (PWj : /va‫ݜ‬r/ (PWj))] (PP)) It is important to notice that article–noun agreement (insofar as it is head-marked) crucially differs from noun–attribute agreement (dependentmarked): while the latter depends on feature copying, the former does not; instead, it depends (a) on a bundle of morphosyntactic features (i.e. gender, number and case) and (b) on the syntactic configuration of the final utterance (i.e. whether the article will be present or not). Because agreement thus depends on two separate mechanisms, one for head marking and one for dependent marking, we can make sense of the differential agreement patterns revealed by the data and described above. If feature copying ceases to operate, it follows that attributive elements will no longer mutate in response to the gender of the noun; that is, they will cease to be agreement targets, in line with the observations already made in section 7.4. Nouns can still be mutated after the article because their mutation does not depend on feature copying, whereas the selection of the article morpheme in the genitive singular in Irish, which still does, should cease: in fact, as has already been observed, genitive marking is increasingly rare in spoken Irish (cf. Hughes 1994: 628–635; Ó hUiginn 1994: 563–567; Ua Súilleabháin 1994: 492–493) and Scottish Gaelic (MacAulay 1992: 162f.; also cf. Gillies 1993: 172); in our data, we find that the occurrence of gen. sg. na in CC-I is linked to high-frequency and/or formulaic terms, such as na hÉireann ‘of Ireland’ and na Gaeilge ‘of the Irish language’, mí na Nollag ‘December’, and the official names of Raidió na Gaeltachta and of

Chapter Seven

124

a research institute (Institiúid Taighde na Fiontraíochta Digití ‘The Digital Enterprise Research Institute’). The process whereby feature copying within the noun phrase ceases to operate is gradual. This means that certain speakers may retain feature copying in controlled speech, that is, as a feature of a higher register, and ignore it in lower registers; in this sense, adjective agreement may be, for those speakers, a stylistic choice. Other speakers may have ceased to apply feature copying altogether, but may retain some fixed mutations in particular phrases (idiomatic or otherwise high-frequency phrases). The following example (from CC-I) is a case in point. 23 Irish (RnaG “Barrscéalta” 23/10/1997) a tá dainséar ann … go gcaillfimís be.PRS danger(M).NOM.SG there … COMP lose.COND.1PL cuid mhaith den rud share(F).NOM.SG good.NOM.F.SG of.ART.SG thing(M).SG so DEM.PROX “We risk to lose a considerable amount of this thing” b

ba mhaith linn … go mbeadh COP good with.1PL … COMP be.COND tuiscint maith understanding(F).NOM.SG good.NOM.M.SG sa Ghaelainn acu in.ART.PREP.SG Irish(F).SG at.3PL “We would like for them to have a good comprehension of Irish.”

Both (23a) and (23b) were uttered by the same speaker; cuid mhaith (‘a good amount’, ‘a great deal’) occurs twice with this speaker and tuiscint maith (‘a good comprehension’) just once. Both cuid and tuiscint are feminine, but cuid mhaith is a frequent collocation and as such is also found in standard dictionaries (cf. Ó Dónaill 1977, s.v. cuid), whereas tuiscint m(h)aith is not. The presence of initial mutation on certain adjectives in speakers who would otherwise omit it might be explained as an application of the principle of lexical priority, formulated by Dik as follows: Whenever a rule is encountered of the general form: M[X] = Y, where Y is the form of X under the operation M, first check the lexicon to see whether the M form of X is stored there ready-made. If so, select this form; otherwise apply the rule (Dik 1997a: 345)

Local Agreement in Irish and Welsh

125

In other words, the form cuid mhaith with the meaning ‘a great deal’ would be stored in the lexicon as a whole and this would prevent the occurrence of the productively combinatorial phrase cuid maith.

7.8 Conclusions In this paper I presented the findings of a corpus-based investigation into gender agreement marking in the Irish and Welsh noun phrase and demonstrated that an incremental model of speech production such as FDG, capable of handling processes that are not strictly linear, can adequately describe the mechanisms of agreement in this context and account for the differential agreement patterns that are observed in contemporary spoken varieties. The problem this differential pattern creates in the traditional model of agreement as the copying of morphosyntactic features is solved by distinguishing dependent-marked from head-marked agreement, with only the former depending on feature copying. By doing so, we can explain the difference in agreement patterns without adding a further position for the determiner to the Agreement Hierarchy, an addition which is typologically unwarranted and theoretically problematic: the lower agreement rates that characterize attributive elements within the noun phrase are interpreted as indicative of a gradual loss of the attributive elements as agreement targets.

Abbreviations 3 Ap ART Aw Dp F, f. GEN Gw IL L M, m. ML

third person adjective phrase article adjectival word determiner phrase feminine genitive grammatical word interpersonal level lenition masculine morphosyntactic level

NOM, nom. Np Nw PL PP PW PRS PST RL SG, sg. SM

nominative noun phrase nominal word phonological level phonological phrase phonological word present past representational level singular soft mutation

126

Chapter Seven

References Akutagawa, M. 1987. A linguistic minority under the protection of its own ethnic state: a case study in an Irish Gaeltacht. In G. Mac Eoin, A. Ahlqvist and D. Ó hAodha (eds.), Third International Conference on Minority Languages: Celtic papers, 125–146. Clevedon: Multilingual Matters. Anderson, S. R. 1999. A formalist’s reading of some functionalist work in syntax. In M. Darnell, E. Moravcsik, F. Newmeyer, M. Noonan and K. Wheatley (eds.), Functionalism and Formalism in Linguistics, vol. 1, 111–135. Amsterdam: John Benjamins. Audring, J. 2009. Reinventing pronoun gender. PhD diss., University of Amsterdam. Baayen, R. H. 2008. Analyzing linguistic data: A practical introduction to statistics using R. Cambridge: Cambridge University Press. Ball, M. J. 1982. Stylistic variation in radio broadcast: an introductory study. Papurau Gwaith Ieithyddol Cymraeg Caerdydd/Cardiff Working Papers in Welsh Linguistics 2: 17–24. Borsley, R. D., M. Tallerman and D. Willis. 2007. The syntax of Welsh. Cambridge: Cambridge University Press. Broderick, G. 1984. A handbook of late spoken Manx. Volume 1: Grammar and texts. Tübingen: Max Niemeyer. Butler, C. S. 2003. Structure and function: A guide to three major structural-functional theories. Part 1: Approaches to the simplex clause (Studies in Language Companion Series 63). Amsterdam and Philadelphia: John Benjamins. —. 2008. Cognitive adequacy in structural-functional theories of language. Language Sciences 30: 1–30. —. 2009. Criteria of adequacy in functional linguistics. Folia Linguistica 43 (1). 1–66. Chomsky, N. 1965. Aspects of the theory of syntax. Cambridge, MA: The MIT Press. —. 1980. Rules and representations. New York: Columbia University Press. Corbett, G. G. 1991. Gender. Cambridge: Cambridge University Press. —. 1994. Gender and gender systems. In R. E. Asher and J. M. Y. Simpson (eds.), The encyclopaedia of language and linguistics, vol. 3, 1347–1353. Oxford: Pergamon Press. —. 2006. Agreement. Cambridge: Cambridge University Press. Dik, S. C. 1978. Functional grammar. Amsterdam: North-Holland Publishing Co.

Local Agreement in Irish and Welsh

127

—. 1986. On the notion “functional explanation”. Belgian Journal of Linguistics 1: 11–52. —. 1997a. The theory of functional grammar. Part 1. The structure of the clause. (2nd revised edn prepared by Kees Hengeveld.) Berlin: Mouton de Gruyter. —. 1997b. The theory of functional grammar. Part 2. Complex and derived constructions. (2nd revised edn prepared by Kees Hengeveld.) Berlin: Mouton de Gruyter. Dikker, S. 2004. On the whereabouts of gender and number assignment: location and accessibility. Working Papers in Functional Grammar 79: 35–51. Dorian, N. C. 1994. Purism vs. compromise in language revitalization and language revival. Language in Society 23 (4). 479–494. Fife, J. 1993. Introduction. In M. J. Ball and J. Fife (eds.), The Celtic languages, 3–25. London: Routledge. Frenda, A. S. 2011. Gender in Insular Celtic: A functionalist account of variation and change in Irish and Welsh. PhD diss., Trinity College Dublin. George, K. 1993. Cornish. In M. J. Ball and J. Fife (eds.), The Celtic languages, 410–468. London: Routledge. George, K. and G. Broderick. 1993. The revived languages: modern Cornish and modern Manx. In M. J. Ball and J. Fife (eds.), The Celtic languages, 644–663. London: Routledge. Gillies, W. 1993. Scottish Gaelic. In M. J. Ball and J. Fife (eds.), The Celtic languages, 145–227. London: Routledge. Green, A. D. 2006. The independence of phonology and morphology: the Celtic mutations. Lingua 116: 1946–1985. Hamp, E. P. 1951. Morphophonemes of the Keltic mutations. Language 27(3). 230–247. Hannahs, S. J. and M. Tallerman. 2006. At the interface: selection of the Welsh definite article. Linguistics 44(4). 781–816. Harris, A. C. and L. Campbell. 1995. Historical syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. Hengeveld, K. and J. L. Mackenzie. 2008. Functional Discourse Grammar: A typologically-based theory of language structure. Oxford: Oxford University Press. Hickey, R. 1995. Sound change and typological shift: initial mutation in Celtic. In J. Fisiak (ed.), Linguistic typology and reconstruction, 133– 182. Berlin: Mouton de Gruyter.

128

Chapter Seven

Hughes, A. 1994. Gaeilge Uladh. In K. McCone, D. McManus, C. Ó Háinle, N. Williams and L. Breatnach (eds.), Stair na Gaeilge, 611–660. Maigh Nuad: Roinn na Sean-Ghaeilge, Coláiste Phádraig. Jones, M. C. 1998. Language obsolescence and revitalization: Linguistic change in two sociolinguistically contrasting Welsh communities. Oxford: Clarendon Press. Kallen, J. L. 1979. Initial mutation in Modern Irish. MA diss., University of Washington. —. 1986. Linguistics fundamentals for Hiberno-English syntax. PhD diss., Trinity College Dublin. Kortlandt, F. 1982. Phonemicization and rephonemicization of the Old Irish mutations. Eriú 33: 73–83. Labov, W. 1972. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press. Lewis, H. and H. Pedersen 1974. A concise comparative Celtic grammar (3rd edn). Göttingen: Vandenhoeck & Ruprecht. Lieber, R. 1983. New developments in autosegmental morphology: consonant mutation. In M. Barlow, C. A. Ferguson, D. Flickinger and M. Wescoat (eds.), Proceedings of the West Coast Conference on Formal Linguistics, 165–175. Stanford, CA: Stanford University Press. Mac Eoin, G. 1993. Irish. In M. J. Ball and J. Fife (eds.), The Celtic languages, 101–144. London: Routledge. MacAulay, D. (1992). The Scottish Gaelic language. In D. MacAulay (ed.), The Celtic languages, 137–248. Cambridge: Cambridge University Press. Matthews, P. 2003. On change in “E-language”. In R. Hickey (ed.), Motives for language change, 7–17. Cambridge: Cambridge University Press. McCone, K. 1994. An tSean-Ghaeilge agus a réamhstair. In K. McCone, D. McManus, C. Ó Háinle, N. Williams and L. Breatnach (eds.), Stair na Gaeilge, 61–219. Maigh Nuad: Roinn na Sean-Ghaeilge, Coláiste Phádraig. Ó Curnáin, B. 2007. The Irish of Iorras Aithneach, County Galway, Volume II. Dublin: Dublin Institute for Advanced Studies. Ó Dónaill, N. 1977. Foclóir Gaeilge-Béarla. Baile Atha Cliath: Roinn Oideachais agus Eoilaíochta. Ó Giollagáin, C. and S. Mac Donnacha (2008). The Gaeltacht today. In C. Nic Pháidín and S. Ó Cearnaigh (eds.), A new view of the Irish language, 108–120. Dublin: Cois Life.

Local Agreement in Irish and Welsh

129

Ó hUiginn, R. 1994. Gaeilge Chonnacht. In K. McCone, D. McManus, C. Ó Háinle, N. Williams and L. Breatnach (eds.), Stair na Gaeilge, 539– 609. Maigh Nuad: Roinn na Sean-Ghaeilge, Coláiste Phádraig. Ó Sé, D. 2000. Gaeilge Chorca Dhuibhne. Baile Átha Cliath: Institiúid Teangeolaíochta Eireann. Ó Siadhail, M. 1989. Modern Irish: Grammatical structure and dialectal variation. Cambridge: Cambridge University Press. Oftedal, M. 1962. A morphemic evaluation of the Celtic initial mutations. Lochlann 2: 93–102. Siemund, P. 2008. Pronominal gender in English. New York: Routledge. Sinclair, J. 2005. Corpus and text – basic principles. In M. Wynne (ed.), Developing linguistic corpora: A guide to good practice, 1–16. Oxford: Oxbow Books. Stewart, Jr., T. W. 2004. Mutation as morphology: bases, stems, and shapes in Scottish Gaelic. PhD diss., Ohio State University. Thomas, A. R. 1992a. The Cornish language. In D. MacAulay (ed.), The Celtic languages, 346–369. Cambridge: Cambridge University Press. —. 1992b. The Welsh language. In D. MacAulay (ed.), The Celtic languages, 251–345. Cambridge: Cambridge University Press. Thomas, B. and P. W. Thomas (1989). Cymraeg, Cymrâg, Cymrêg… Cyflwyno’r tafodieithoedd. Caerdydd: Gwasg Taf. Thomas, E. M. 2001. Aspects of gender mutation in Welsh. PhD diss., University of Wales. Thomas, E. M. and V. C. M. Gathercole. 2005. Minority language survival: obsolescence or survival for Welsh in the face of English dominance? In J. Cohen, K. T. McAlister, K. Rolstad and J. MacSwan (eds.), Proceedings of the 4th International Symposium on Bilingualism, 2233–2257. Somerville, MA: Cascadilla Press. Thomson, R. L. 1992. The Manx language. In D. MacAulay (ed.), The Celtic languages, 100–136. Cambridge: Cambridge University Press. Thorne, D. A. 1993. A comprehensive Welsh grammar. Oxford: Blackwell. Thurneysen, R. 1980. A grammar of Old Irish. Revised and enlarged edition with supplement. Dublin: Dublin Institute for Advanced Studies. Ua Súilleabháin, S. 1994. Gaeilge na Mumhan. In K. McCone, D. McManus, C. Ó Háinle, N. Williams and L. Breatnach (eds.), Stair na Gaeilge, 479–538. Maigh Nua: Roinn na Sean-Ghaeilge, Coláiste Phádraig.

130

Chapter Seven

Watkins, C. 1999. Two Celtic notes. In P. Anreiter and E. Jerem (eds.), Studia Celtica et Indogermanica: Festschrift für Wolfgang Meid zum 70. Geburtstag, 539–544. Budapest: Archaeolingua. Watkins, T. A. 1993. Welsh. In M. J. Ball and J. Fife (eds.), The Celtic languages, 289–348. London: Routledge. Watson, S. 1989. Scottish and Irish Gaelic: the giant’s bed-fellows. In N. C. Dorian (ed.), Investigating obsolescence: Studies in language contraction and death, 41–59. Cambridge: Cambridge University Press. Wigger, A. (ed.). 2000. Caint Chonamara: Bailiúchán Hans Hartmann. Imleabhar IX. Ros Muc. Bonn: Universität Bonn.

CHAPTER EIGHT DEMONSTRATIVES IN IRISH FRANCES KANE 8.1 Introduction This paper is an investigation into the position and the distribution of demonstratives in Irish. The Irish demonstrative element appears strictly postnominally, and it has been documented in traditional grammars (Christian Brothers 2004) that this demonstrative element features a strict co-occurence constraint with the prenominal definite article: 1 *(an)duine seo the person this “this person” Following the constraint shown above in (1), this paper will argue that rather than corresponding directly to the English demonstrative element this as represented in the gloss, the postnominal element seo above is in fact the spelled-out locative part of a phrasal demonstrative that has remained stranded low down in the projection of the NP. A similar phrasal approach to demonstratives has been adopted for other languages (see Leu 2008 for a phrasal analysis of Scandinavian, Ihsane and Puskás 2001 for French and Hungarian). What has been previously analysed as the definite article—the prenominal element—is in fact another part of the demonstrative that has moved to initial position in the nominal projection. This paper will argue for a phrasal approach to the Irish demonstrative based on three lines of evidence: (i) Analyses of other languages where the demonstrative is considered to be phrasal, a combination of (at least two) elements. These analyses seem to transfer quite neatly to the Irish data. (ii) A consideration of the diachronic development of both the definite article and the demonstrative in Irish and in typologically

Chapter Eight

132

related languages. By looking at the historical development of the two elements that, as we propose, make up the phrasal demonstrative, we hope to be able to show that the two elements were at one stage adjacent to each other, or two constituents of one single element. (iii) An observation of the internal structure of the locative adverbial that exists in the modern form of the language. Analysis of the morphology of locative adverbs of Irish seems to show that they are a combination of the two elements that we propose make up the phrasal demonstrative. This will have implications for both the fine structure of the Irish noun phrase and for demonstratives in general.

8.2 Distribution of the demonstrative in Irish Irish is a head-initial language. This characteristic is observed in NPs, where the noun consistently appears in first position. There is no indefinite article (2), and in cases where the definite article accompanies the head noun, it appears strictly prenominally (3). Any modifying adjectives appear to the right of the head noun (4): 2 madadh mór dog big “a big dog” 3 an madadh the dog “the dog” 4 an madadh mór big the dog “the big dog” Irish demonstrative strings follow the same head-initial pattern. The definite article an precedes the noun, as in the simple NPs shown above, while the demonstrative appears postnominally as an independent element. Irish has three demonstrative particles that make distinctions between proximate (seo) that loosely corresponds to ‘here’, intermediate (sin) that

Demonstratives in Irish

133

corresponds to ‘there’ and distal (siúd1), loosely corresponding to a more distant ‘there’ than sin, often referred to in non-standard English as ‘yonder’ or ‘thonder’: 5 an madadh seo the dog DEM1 “this dog” 6 an madadh sin the dog DEM2 “that dog” 7 an madadh siúd the dog DEM3 “that dog yonder” Demonstratives appear strictly in final position. When the head noun is modified by both a postnominal demonstrative and an adjective (that also appears postnominally), the demonstrative must appear on the far right of the construction: 8 an madadh mór the dog big “that big black dog”

dubh black

sin DEM2

9 *an madadh sin the dog DEM2

mór big

dubh black

10 *an madadh mór big the dog

sin DEM2

dubh black

Due to a strict co-occurrence constraint in Irish, the prenominal article is obligatory when the head noun is followed by one of the demonstrative elements shown above. A bare noun cannot usually be modified by the postnominal demonstrative alone:

1

The forms úd and údaí also appear as dialectal alternatives (McCloskey 2004) but we will not discuss these at this stage, as they are interchangeable with the standard siúd.

134

Chapter Eight

11 *(an) madadh seo the dog DEM1 “this dog” 12 *(an) madadh sin the dog DEM2 “that dog” 13 *(an) madadh siúd the dog DEM3 “that dog yonder” The demonstrative can also appear with the prenominal possessive pronoun:2 14 a mac 3SG.FEM son ?”this woman’s son”

seo DEM1

The only case where the head noun can appear bare is when it is followed by a genitive possessor in a construct-state nominal (CSN) construction: 15 madadh seo an fhir dog DEM1 the man.GEN “this dog (of the man)” The formation of the type shown in (12) above, termed construct-state nominals, is an issue that is well documented in the literature both in the Celtic languages and elsewhere (principally the Semitic languages). In the remainder of this paper we will deal only with “simple” demonstrative DPs (as in examples (5–7)) The pattern found in Irish demonstratives can be observed (to different degrees) in other languages as well. In Spanish, for example, there is optionality with respect to demonstrative placement. The demonstrative can appear prenominally, in parallel with English:

2

Note the “?” alongside the translation of (14). There is debate in the literature as to whether the demonstrative in this type of construction modifies the possessee or the possessor. This will not be tackled at this stage.

Demonstratives in Irish

135

16 este libro DEM book “this book” Spanish also provides the option of a postnominal demonstrative, on a par with Irish. When the demonstrative element in Spanish appears postnominally, the appearance of a prenominal definite article is obligatory (17), in parallel with the Irish construction in (1) but unlike English (18–20): 17 *(el) libro the book “this book”

este DEM

18 *The this book 19 *This the book 20 *The book this Although the construction is ungrammatical in English (and in other languages), many languages permit appearance of the definite article with the demonstrative (examples from Brugè 1996: 3): 21 ika n this the “this child”

anak child

(Javanese)

22 ez a this the “this boy”

fiú boy

(Hungarian)

23 sa maðr-inn this man-the “this man”

(Old Icelandic)

24 þan wig the way “this way”

(Gothic)

jainan this

Chapter Eight

136

8.3 Demonstrative reinforcers In an analysis similar to our phrasal approach to the demonstrative, Bernstein (1997) provides an account of demonstratives in both Romance and Germanic languages, where she proposes that “adverbial-like” elements appear alongside the demonstrative and the article, and that these “adverbial” elements act as reinforcers. Consider the example below: 25 den här this here “this man”

mann-en man-the

(Swedish)

In this example, the adverb här is assumed to be a reinforcer, strengthening the locative element of the demonstrative den.3 In the case of Swedish, it appears as an independent lexical element. In French, however, a similar element appears as a phonologically enclitic element on the head noun. Bernstein (1997) assumes that this is another example of a reinforcer element: 26 cette femme-ci this woman-here “this woman” This enclitic reinforcer element appears along with an independent prenominal demonstrative. By examining the morphology of this element, we can show that what appears as a “reinforcer” in the sense of Bernstein’s (1997) analysis is in fact a reduced form of the full locative adverbial ici meaning ‘here’. 27 Jean habite ici John live.PRES here “John lives there” The same is found with the French distal demonstrative “reinforcer” enclitic element -là that appears to be a reduced form of the locative adverbial là-bas meaning ‘there’: 3

Note that in Swedish the element den is both the demonstrative and the definite article. When the element appears as the demonstrative it bears phonological stress. The fact that in this language both the article and the demonstrative are represented morphologically by the same element is an argument that the two elements are closely related (at least in this language).

Demonstratives in Irish

137

28 ce livre-là that book-there “that book” 29 Jean habite là-bas John live.PRES there “John lives there.” The contrast between the reinforcer elements shown above and their locative adverbial counterparts bears a strong resemblance to what is observed in a comparison between the Irish demonstratives outlined in (5– 7) above and their corresponding adverbs: Table 8.1. Irish demonstrative elements and locative adverbs demonstrative element seo DEM1 (proximal) ‘this’ sin DEM2 (intermediate) ‘that’ siúd DEM3 (distal) ‘that’ (distal)

locative adverb anseo ‘here’ ansin ‘there’ ansiúd ‘there’ (distal)

If we assume that the two corresponding elements are related, which is feasible given their locative semantics, we can make the preliminary conclusion that to form the locative adverb we add a prefix to the postnominal demonstrative element. Therefore we can propose the following template for locative adverbials in Irish: 30 AN + DEM = LOCATIVE ADVERB Note that the prefix added to the demonstrative element in the formation of the locative adverb as shown above is an-, which is morphologically identical to the Irish definite article as shown in example (3) repeated below:

Chapter Eight

138

31 an madadh The dog “the dog” By adopting Bernstein’s (1997) analysis for Irish, we can predict that the definite article in Irish is in fact a pronominal demonstrative, rather than an article. Therefore, we assume the postnominal element as a “reinforcer” rather than a demonstrative. This idea is supported by the ungrammaticality of these elements by themselves (without an): 32 *madadh seo dog DEM1 33 *madadh sin dog DEM2 34 *madadh siúd dog DEM3 Our proposal is that what was previously analysed as a prenominal definite article actually has demonstrative properties and may be analysed as a demonstrative (or part of a phrasal demonstrative consisting of more than one element), at least in cases where it appears obligatorily along with the postnominal demonstrative element. A similar proposal is put forward by Henry (2010) for Belfast English. This analysis will be discussed at a later stage in terms or a comparison between Belfast English and Irish: 35 an madadh seo demonstrative locative reinforcer The next section investigates if there is any motivation to analyse an, which has previously been assumed to be the definite article, as a demonstrative. What (if any) locative properties does this element possess, apart from the strict constraint on co-occurrence with the demonstrative?

Demonstratives in Irish

139

8.4 The relationship between the demonstrative and the definite article How can we say that an in Irish is a demonstrative, when in all other cases in the language it operates exactly as a definite article? First, there is crosslinguistic evidence showing that the demonstrative and the definite article are inherently related, given their distribution across languages. For example in English, their co-occurence is strictly disallowed: 36 *The this book 37 *The this book In Spanish, their co-occurrence is allowed but is restricted. When both the demonstrative and the article appear, the demonstrative element must be postnominal: 38 el libro the book “this book”

este DEM

39 *el este the DEM

libro book

Alternatively, the demonstrative can appear prenominally. However, crucially, in this case the demonstrative must appear without the article. This could mean that the demonstrative and the definite article either “compete” for the same position or are in fact the same element: 40 este libro DEM book “this book” The distribution of the demonstrative and the article is much freer in languages such as Greek. In Greek the appearance of the article is required with the demonstrative; however, these elements can occur in more than one position: 41 afti i DEM the “this cat”

ghata cat

Chapter Eight

140

42 i ghata the cat “this cat”

afti DEM

Finally, in some languages the co-occurence of both elements is obligatory and they must appear in fixed positions. The above Irish examples in (45–46), although similar to the Spanish and Greek ones, show that the elements must appear in fixed positions and cannot appear in positions other than those shown in (43): 43 *(an) madadh this dog DEM1 “this dog” 44 *an seo this DEM1 45 *madadh an dog the

seo

madadh dog seo DEM1

46 *madadh seo an dog DEM1 the The examples above show that the distribution of the two elements varies greatly across languages, but that at the same time there exist consistently strict rules on both their co-occurrence and their placement. Consider Latin, a language that does not have a definite article: 47 illa dies ueniet mea TXƝ OnjJXEULD SǀQDP DEM day come.FUT.3SG my when mourning put.1SG “The day will come when I will put aside my mourning.” (Ovid, Trista 4.2.73) In Latin the demonstrative element operates exactly as a definite article. This is evidence that the demonstrative can perform the function of an article, which was what we proposed for Irish in (30). Furthermore, in languages that have a definite article it seems that the article has developed from a demonstrative through a process of “grammaticalization” (Harris 1980, TekavþLü  (SVWHLQ  9LQFHQW   7KH IROORZLQJ WDEOH shows Romance articles that have developed from Latin demonstratives (adapted from Vincent 1997: 151):

Demonstratives in Irish

141

Table 8.2. Romance articles that have developed from Latin demonstratives Latin (Nom) (Acc) French Italian Portuguese

masculine ille/illi illum/illos le/les il/i o/os

feminine illa/illae illam/illas la/les la/le a/as

neuter illud/illa illud/illa

According to Harris (1978), phonologically weak demonstratives were eliminated from the language by the time of Vulgar Latin, the form of the language from which the Romance languages subsequently developed. In Vulgar Latin, the forms of the demonstrative element that remained in the language could appear in prenominal position, where their person feature was no longer specified (although the personal pronouns that emerged did retain their person specification). The demonstrative ille, however, retained its definiteness feature and was able to appear with a noun and function as the definite article. Following this, it seems plausible that the loss of the demonstrative feature from a demonstrative is responsible for the emergence of an article. Thus far, we have proposed that what has been previously analysed as the definite article in Irish, an, is in fact a demonstrative that has lost its deictic feature when appearing in a simple NP such as that in the example below, even though it appears to be a definite article akin to English the. 48 an madadh the dog “the dog” Can we show that the article in Irish has evolved from a demonstrative, as shown for the Romance articles above? If so, this may provide evidence in support of our hypothesis that the article (in some cases) has retained some deictic features and can operate as a demonstrative when it appears with postnominal elements, as shown in (5–7).

8.5 The diachrony of the Irish article In Old Irish (ca 600–900 AD), most forms of the article are based on a stem sindo- or VLQGƗ- (Thurneysen 1946: 293). As the article has always

142

Chapter Eight

been proclitic, phonological reduction as part of grammaticalization has led to the loss of the initial s-. Consider the example below: 49 in fer bec the man small “the small man”

(Old Irish, ca 600–900) McCone (1994:119)

As the example above shows, by the stage of Old Irish, the initial s- on the definite article is not observed. However, there are contexts where the initial s- reappears in this later form of Irish: when a preposition precedes the definite article, the two form a compound element. In this case the initial s- can be observed when the preposition le ‘with’ combines with the article in: 50 la-sin fer with-the man “with the man” What is observed here is evidence in support of the theory that the definite article in Old Irish in has emerged from a longer version with an initial s-. This intermediate element may have been similar to the distal demonstrative element sin that exists in the modern form of the language. Example (49) shows a change in the language that is occurring at that stage. Regarding demonstratives in Old Irish, the construction observed is similar to the modern demonstrative DP in Irish. Adverbs of place (so, sa, se, sin) may follow a noun that is modified by the article, as in (43). The fact that there exist two separate elements in a demonstrative construction as far back as Old Irish means that, in order to find out if the article does in fact originate as a demonstrative, we need to look back further into the history of the language (Common Celtic) and at Indo-European languages. Another piece of data from Old Irish shows that at some stage in the development of the language both the article and demonstrative could appear adjacent to one another, in the postnominal position: 51 iss ed laithe in sin COP 3SG.MASC day the DEM “That is the day.” Caoimhín Ó Dónaill (p.c., 2011) This word order is not permitted in Modern Irish, where the article must appear in the prenominal position:

Demonstratives in Irish

52 *is é lá COP 3SG.MASC day

an the

143

sin DEM2

Recall Table 8.1, however, where it was proposed that the demonstrative element and the preceding definite article could combine to form the locative adverb. In this example it is clear that at some stage the two elements were adjacent to one another. The positioning of these elements on the right periphery of the construction also provides evidence for our proposal that the demonstrative is phrasal and is generated low. To generate the correct word order for modern Irish, however, we need to motivate movement of one of the components past the head noun, and the stranding of another element so that it can appear on the right of the construction.

8.6 The phrasal demonstrative It has been suggested in the literature (Vangsnes 1999, Elbourne 2005, Leu 2008) that the demonstrative is phrasal, involving (at least) two separate components; an adjectival/deictic element and a second component encoding definiteness (often the definite article). According to Leu (2008:16), demonstratives are morphologically complex, spelling out different heads in an extended adjectival projection. The analysis of demonstratives as being complex, and possibly made up of more than one lexical element, may tell us more about the Irish demonstrative construction with its two obligatory components. Recall that when the head noun is modified by a pronominal demonstrative element, the appearance of a pronominal definite article is obligatory: 53 *(an) madadh the dog “this dog”

seo DEM1

It is possible, under a phrasal analysis of the demonstrative, that the definite article is part of a demonstrative complex, corresponding to the definite element outlined by Leu (2008) and others. Under such an analysis, the DEM particle that has been analysed so far as the demonstrative may in fact be an adjectival or deictic marker. Recall Bernstein’s (1997) analysis that we previously adopted for Irish in which the demonstrative is supported by a reinforcer element:

144

54 cette femme-ci this woman-here “this woman” 55 an madadh seo the dog DEM1 “this dog”

Chapter Eight

(French) (Bernstein 1997: 91) (Irish)

Here, we modify this analysis: rather than a demonstrative reinforcer, we analyse the postnominal demonstrative element as an overt locative element that makes up one part of the phrasal demonstrative. The pronominal element that was previously analysed as the definite article is then the second element in the phrasal demonstrative. This pronominal element is assumed to carry the definiteness element of the phrasal demonstrative in the sense of Leu (2008). This follows a similar analysis by Henry (2010) of Belfast English, a non-standard variety of English, which permits the following types of construction: 56 This here book is very hard to understand 57 That there coffee is very weak

(Henry 2010: 1)

These are distinguished from the standard English constructions where the postnominal adverb functions like a “true” adverb: 58 This book here is very hard to understand 59 That coffee there is very weak Rather than following Bernstein’s (1997) approach involving demonstrative reinforcers, Henry (2010) argues that these pronominal elements (shown in bold above in (54)) are simply spelled-out versions of the deictic component found in all demonstratives, diverging from Bernstein’s initial analysis. By adopting this approach for the Irish data, we can provide an analysis for the demonstrative DP repeated below: 60 an madadh seo the dog DEM1 “this dog”

Demonstratives in Irish

145

We can therefore analyse the Irish demonstrative DP as shown below: 61 an

madadh seo

[+DEF]

[+LOC]

What is still left to find out is how the pronominal element of the phrasal moves to end up on the left periphery and where exactly it moves from and to within the extended projection of the noun.

8.6.1 Further support for “composite” nature of demonstratives In all of the languages discussed so far, the idea of demonstratives has been applied only to languages where demonstrative DPs consist of more than one independent lexical element. English needs only one demonstrative element, and the co-occurrence of the demonstrative and the definite article is strictly ungrammatical: 62 *The this book 63 *This the book 64 *The book this Kayne and Pollock (2009) provide an analysis of the English demonstrative where both the definite and deictic components of the phrasal demonstrative are contained in one independent lexical element. Kayne and Pollock deconstruct the demonstrative as is shown below: 65 th- + -at 66 th- + -is In the examples above, the initial component th- is assumed to perform the function of the definite element of the phrasal demonstrative, expressed in simple NPs as the independent definite article the. Superficially, it seems that this definite article may be an instantiation of the definiteness marker that creates a standalone element through combination with a schwa, resulting in ‘the’. The final components -at and -is are overt deictic elements realized as bound morphemes, and represent distal and proximal meanings, respectively. The ungrammaticality of the definite article alongside the demonstrative elements this and that could then be reduced to an issue of economy. Placing the definite article

146

Chapter Eight

alongside the demonstrative results in the definite element being included in the construction twice, which goes against the idea that natural language uses the most economical derivation. The duplication of information is clearly not economical and results in ungrammatical derivations such as those outlined in (62–64).

8.7 Conclusion and suggestions for further research This paper has provided a body of evidence to show that the demonstrative in Irish is phrasal and consists of two elements: a pronominal element that was previously analysed as a definite article providing definiteness features for the head noun, and a postnominal locative element that provides deictic information for the head noun. By examining morphological evidence and data from Old Irish, as well as applying ideas from phrasal analyses of demonstratives in other languages, an alternative analysis for the Irish demonstrative DP has been proposed. Now that functions for each element of the phrasal demonstrative have been identified, the next step is to work out where to locate each of these elements within the extended projection of the noun, as well as to understand the motivation for movement that yields the article—noun— demonstrative word order that is observed in the language. One possibility is that the phrasal demonstrative is generated low, as proposed for other languages (Bernstein 1997 on Romance and Germanic, Vangsnes 1999 on Scandinavian, Ihsane and Puskás 2001 for French and Hungarian, and Roehrs 2006 for German), based on the fact that demonstratives appear consistently on the far right of the NP. If the locative element is generated low, is the definite element also located low down in NP? Recall example (51) repeated below: 67 iss ed laithe in sin COP 3SG.MASC day the DEM “That is the day.” Caoimhín Ó Dónaill (p.c., 2011) This raises questions regarding the motivation for the positioning of these two elements within the Irish demonstrative construction. Given the general ban on downward movement, the example above shows that (at least in Old Irish) the definite element of the demonstrative is generated low, to the right of the NP. If we assume that this is consistent with the formation of demonstrative DPs in modern Irish, we must motivate movement of the definite element past the head noun and any modifying APs to a higher position. It is clear that much more research is needed to

Demonstratives in Irish

147

find out where these elements are located both before and after any movement operations. If the demonstrative elements are in fact generated low, why then do they move? Where do they move? The demonstrative construction in Irish poses an important problem, which has implications for both the fine structure of the Irish demonstrative DP, as well as the Irish DP in general. The analysis of the distribution of demonstrative elements in Irish must also be considered in a general sense in relation to the syntax of demonstratives across languages.

References Bernstein, J. 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87-113 Brugè, L. 1996. Demonstrative movement in Spanish: A comparative approach. University of Venice Working Papers in Linguistics 6 (1). 1– 53. Christian Brothers. 2004. New Irish Grammar. Dublin: CJ Fallon. Elbourne, P. 2005. Situations and individuals. Cambridge, MA: MIT Press. Epstein, S. D. 1994. The derivation of syntactic relations. Ms., Harvard University Harris, M. B. 1980. The marking of definiteness in Romance. In J. Fisiak (ed.), Historical morphology, 141–156. Berlin: Mouton de Gruyter. —. 1978. The evolution of French syntax. London.: Longman Henry, A. 2010. Demonstratives in Belfast English: Microvariation and externalisation. Ms., University of Ulster. Ihsane, T., and G. Puskás. 2001. Specific is not Definite. Generative Grammar in Geneva 2: 39–54. Kayne, R and J. Pollock. 2009. Notes on French and English demonstratives. Ms., New York University and Université de Paris. Leu, T. 2008. The internal syntax of determiners. PhD diss., New York University. McCloskey, J. 2004. Irish Nominal Syntax 1: Demonstratives. Ms., University of California, Santa Cruz. Available at http://ohlone.ucsc. edu/~jim/PDF/demon.pdf McCone, K. 1994. An tSean-Ghaeilge agus a réamhstair. In K. McCone, D. McManus, C. Ó Háinle, N. Williams and L. Breatnach (eds.), Stair na Gaeilge, 61–219. Maigh Nuad: Roinn na Sean-Ghaeilge, Coláiste Phádraig. Roehrs, D. 2006. The morpho-syntax of the Germanic noun phrase. Bloomington: Indiana University Press

148

Chapter Eight

TekavþLü3Grammatica storica dell’italiano. Bologna: Il Mulino. Vangsnes, Ø. A. 1999. The identification of functional architecture. PhD diss., University of Bergen. Vincent, N. 1997. The emergence of the D-system in Romance. In A. van Kemenade and N. Vincent (eds.), Parameters of morphosyntactic change, 149–169. Cambridge: Cambridge University Press.

PART III FORMAL APPROACHES TO SYNTAX, SEMANTICS AND DISCOURSE

CHAPTER NINE A PHASE-THEORETIC ANALYSIS OF ROOT COMPOUNDS (FROM A CROSS-LINGUISTIC PERSPECTIVE) LEAH S. BAUKE 9.1 Introduction It is standardly assumed that nominal root compounding is parameterized across languages (cf. among many others Roeper, Snyder & Hiramatsu 2002; Roeper & Snyder 2005; Delfitto, Fábregas & Melloni 2008). Along these lines Romance languages are characterized as unproductive, non-compositional, non-recursive, whereas Germanic languages are standardly described as productive, compositional and recursive. So, the patterns in (1) and (2) emerge: 1 Romance: a hombre rana (Spanish) “frogman” b homme grenouille (French) “frogman” c uomo rana (Italian) “frogman”

2 Germanic: a coffee cup b c

Kinderbett (German) “children’s bed” baby doll napkin

Recursive forms like the one in (2c) hardly ever exist in Romance and in those rare cases where they occur they are even described as a deliberate act of coinage or even as coming close to the invention of a new morpheme.1 This, of course, does not mean that lexicalization in Germanic languages does not exist. The example of frogman is a case in point here. 1

Roeper & Snyder (2005) point out gateau fôret-noir as the only instance of a recursive compound that is recursive, albeit fully lexicalized in French.

A Phase-Theoretic Analysis of Root Compounds

151

But this lexicalization is hardly ever total. Hence, next to the interpretation of ‘combat diver’ this can also mean something like ‘man who looks like frogs’, ‘man who is an expert on frogs’ etc. (cf. Roeper, Snyder & Hiramatsu 2005). Equally standardly, the distinction between the two patterns of compounding is associated with the differentiation between lexical and syntactic word-formation (cf. e.g. Giegerich 2007). Here Romance, being the language with irregular forms, is frequently equated with lexical wordformation and Germanic, being the language that displays regular, and compositional forms, is ascribed to syntactic word-formation.

9.2 The conundrum If the pattern described above is on the right track, the question that emerges at this point is why Romance does have a productive pattern of phrasal compounding and why some Germanic forms are clearly drifted: 3 a

tasse à café (French) “coffee cup”

b

Kindbett (German) “childbed”

The French form in (3a) is often identified as a phrasal compound (cf. Delfitto, Fábregas & Melloni 2008), which is the standard productive pattern of compounding in this language. This type of compound is also recursive and compositional. The German form in (3b), on the other hand, is clearly non-compositional and has a very drifted and lexicalized meaning. Thus, both forms constitute a clear violation of the alleged parameter setting, and this violation of the parameter is operative in both directions, i.e. the prepositional compound in French operates against the non-compositional nature of the compounding process and the German form with its lexicalized and fixed reading operates against the parameter setting for the Germanic form. Another example that gives reason for some doubt comes from the Swedish forms in (4): 4 a

barn bok club (Swedish) “book club for children”

b

barn boks club (Swedish) “children’s book club”

The relevant structures for the forms in (4) are given in (5) (cf. also Roeper & Snyder 2005): 5 a

barn [bok club]

b

[barn boks] club

Chapter Nine

152

What this example shows is that Swedish seems to be parameterized along the distinction between left- and rightward branching compounds. The latter require a linking element in contrast to the former to be well-formed. Finally, Chinese is yet another language that potentially conflicts with the alleged parameter setting (cf. e.g. Zhang 2007 and references therein). Here so called breakable compounds, which are frequently analyzed as exocentric compounds, are a case in point: 6

dan xin (Chinese) carry heart “worry”

The combination of verb and noun in (6) is interpreted as a simple verb.2 7 a

Ta hen dan xin zhe jian shi (Chinese) he very carry heart this CL matter “he is very worried about this matter”

7 b

*Xin , wo yi-dian dou bu dan zhe jian shi (Chinese) heart, I one-bit all not carry this CL matter “I don’t worry about this matter”

The dan xin complex takes a complement in (7a), which illustrates that xin is not a complement of dan. Similarly, xin cannot be extracted as is shown in (7b). On the other hand, the same complex can function as a V + N complex and extraction of the nominal complement is possible, with virtually the same interpretation and the same intonation: 8 a

Ta dan xin (Chinese) he carry heart “He was worried”

8 b

Xin, wo yi-dian dou bu dan (Chinese) heart, I one-bit all not carry “I am not worried at all”

In light of the fact that these examples are problematic under the alleged parameter setting discussed above, let us now take a closer look at 2

And in principle all sorts of combinations are possible, such as e.g. noun plus adjective leading to a verb etc. (cf. Zhang 2007 and references therein for several further examples).

A Phase-Theoretic Analysis of Root Compounds

153

one language of the Germanic language family and see how the parameter setting fares here.

9.3 The case of German Like Chinese, German seems to have a pattern of compounding that generates non-compositional forms and one that is truly compositional: 9 a.

Kindbett (German) “childbed”

b

Kinderbett (German) “children’s bed”

The difference between the two forms is that the form in (9b) contains a compound internal plural marker, while the form in (9a), which consists of just two stems, does not: 10

Kinderbett (German) child.PL + bed

This compound-internal plural marker is somewhat surprising, because it is standardly assumed that inflectional material is confined to the edges of a compound (but cf. e.g. Selkirk 1982 for an early rebuttal). Still, this is a reoccurring pattern that goes along with the frequently attested flexibility in interpretation for the German forms: 11 a

Bettlaken (German) bed + sheet “bedsheet”

b

Bettenburg (German) bed.PL + castle “big ugly hotel”, “beds arranged into a castle”, etc.

12 a

Landkarte (German) country + card “map”

b

Länderspiel (German) country.PL + game “match (between two national teams)”, “children’s game that involves knowledge about certain countries”, etc.

Chapter Nine

154

13 a

Wortwitz (German) word + wit “pun”

b

Wörterbuch (German) word.PL + book “dictionary”, ”book with words on its cover” etc.

14 a

Glasdach (German) glass + roof

b

Gläsertuch (German) glass.PL + cloth “glass roof”, “dish towel”, “spectacle wiping cloth”, etc.

The a-forms in (11–14) do not contain any compound internal inflectional material and they all have just one interpretation. The b-forms, in contrast, all contain a compound internal inflectional marker and they all have one preferred interpretation, which is listed first here, but they also have a variety of other interpretations, of which one is given as second for each example. The preferred interpretation may of course be a lexicalized and rather drifted reading3 (cf. e.g. (11b)), but the important point is that other interpretations are recoverable as long as the inflectional marker is present. So, it is along this dividing line that the distinction between recursive and compositional vs. non-recursive and non-compositional compounds in German can be drawn (cf. also Bauke 2009). Thus what emerges here is an intra-language distinction rather than the inter-language distinction based on a macro-parameter with several dubious exceptions.

9.4 Towards an analysis The discussion in the preceding section has revealed that German displays a pattern of compounding in which compositional compounds show compound internal inflectional material. This strongly suggests that this type of compound is built in syntax.4 In essence, this has the effect of reiterating the distinction between regular and irregular processes of word formation along the line of the syntax vs. lexicon distinction (cf. Chomsky 1970), where all irregularities are attributed to the lexicon (cf. also Borer 1984, Chomsky 1995) and where regular processes of word formation are located in syntax. 3

This is actually not very surprising and is what is expected from any word formation process. 4 The other type of compound, i.e. the one without compound internal inflectional material that is not compositional, may well be derived in the lexicon as I originally argued in Bauke (2009), but may also be built in syntax (cf. below and Bauke 2012).

A Phase-Theoretic Analysis of Root Compounds

155

The remaining question is, how these syntactic processes of word formation can be described. One possibility is to follow the approach laid out in e.g. Roeper & Snyder (2005) and Roeper, Snyder & Hiramatsu (2002), where root compounding is framed in terms of a Root Compounding Parameter that is linked to the Abstract Clitic Hypothesis: 15 RCP (Root Compounding Parameter) Merger can combine (two) non-maximal categories 16 ACH (Abstract Clitic Hypothesis) All verbs in English have an invisible Clitic position that may be occupied by particles (or other designated markers) So, according to (15) Merge creates syntactic objects with the status of complex words, provided that the parameter is set in such a way that Merge is allowed to do this. Roeper & Snyder (2005) argue that for Romance the parameter is set negatively, which is why productive, recursive and compositional root compounding is not an option in this language family. For Germanic languages the parameter is set positively and thus allows for productive word formation. The positive setting of the parameter is instantiated by input, i.e. by recursive compounds.5 This RCP is then combined with (16) in order to account for the left- and rightward building pattern of nominal root compounds. This is achieved by extending the ACH, which originally applied to verbs only (cf. Keyser & Roeper 1992) to all lexical categories: 17 Original ACH: V V

ACP

write

up

So, instead of verbs, nouns can also have an ACP (Abstract Clitic Position), which can host e.g. other nouns. A rule of compounding then moves the element in the ACP to the left, thus rendering the correct surface order: 5

In fact, the RCP is not only operative in nominal root compounding but also in verb-particle structures, locative constructions, causative constructions, resultative constructions, double object constructions, middle constructions, perceptual constructions and others (cf. Roeper, Snyder & Hiramatsu 2002).

Chapter Nine

156

18a

N N

b

N

ACP

N

cup coffeeN

N

coffee N ACP cup coffeeN

Crucially, this ACP is a non-argument position, which means that the ACP allows for multiple insertion, because elements moved out of this position do not leave behind a copy (cf. Roeper & Snyder 2005). This has the effect that recursivity automatically follows. 19

N N restaurant

N N

N

coffee N cup

ACP coffeeN restaurantN

So after moving coffee out of the ACP by abiding to the rule of compounding, which is not specified any further in Roeper & Snyder (2005) nor in the original analysis of Keyser & Roeper (1992), the ACP is available for insertion of another element; restaurant in this case, which will also be moved to the left via the compounding rule. This analysis is immediately applicable to the German forms discussed above (cf. Bauke 2009, 2012 for exposition). However, some unresolved questions remain: 1 2 3 4

Why does the ACP lead to a compositional reading? What triggers leftward movement? How can the ACP assumption be reconciled with BPS? How does the distinction between lexical and syntactic word-formation follow from the ACP and how can interaction between the two be accounted for?

These questions will be tackled in the next section where the distinction between syntactic and lexical word formation is subjected to closer scrutiny.

A Phase-Theoretic Analysis of Root Compounds

157

9.5 Syntactic and lexical word-formation? The data discussed so far suggest that a parameter that accounts for inter-language variation is too broad. On the other hand, the intra-language variation is aligned with the distinction between where the compounds are derived, i.e. syntax vs. lexicon. This, however, is hard to encode in parametric terms, at least in the absence of any solid distinctive criteria. Also, the parameter argument is hard to reconcile with a minimalist (BPS) approach (cf. among others Boeckx 2010). To name only a few of the relevant questions that emerge here, we could ask why Romance does not show the same intra-language variation that is attested in Germanic. Or, in broader terms, it could be asked why exactly the deviance from the default parameter opens up the option for recursive and syntactic compounding in Germanic. And eventually postulating two different places for the derivation of the two types of compounds instantiates a violation of the strongest minimalist thesis (SMT), which states that syntax is an optimal solution to interface requirements (cf. Chomsky 2005 and many others). So, what I intend to do in this chapter is to offer an SMT analysis based on the hypothesis in (20): 20 Both types of compounds are derivable in narrow syntax. To see how this is brought about, let me start with two, independently motivated prerequisites. The first is the assumption that compounds, like all other syntactic objects (SOs) of a certain amount of structural complexity6, are formed by Merge. Secondly, for Merge to be operative, lexical items (LIs) must be equipped with edge features (EFs) (cf. e.g. Chomsky 2008, Narita 2011). From this latter assumption (21) falls out naturally: 21 EFs signal mergability What (21) entails is that LIs with EFs are potential candidates for Merge. Complex SOs, which are the product of Merge do not have EFs though (cf. Narita 2011 for detailed exposition).7 6

Determining what this amount of structural complexity might be this involves a fundamental discussion of what the building blocks of syntax are. These matters are certainly beyond the scope of this paper but cf. Bauke (2012). 7 Bauke (2012) has some very detailed analysis of the consequences of this assumption, however, it is somewhat irrelevant to the matter discussed here and will therefore be put aside as well.

Chapter Nine

158

With that much in place let us now look at some sample derivations. According to standard assumptions in DM, roots enter the derivation ZLWKRXW ij-features (or categorial or any other syntactic features for that matter). Rather, these are provided by categorizing x-heads (cf. Marantz 2007). Given (21) holds at all stages of the derivation, roots and categorizing x-heads must bear EFs. So, the following picture emerges: 22 nEF, ij) ¥ĮEF

n()ij) ¥ȕEF

5RRWV Į DQG ȕ ZKLFK DUHERWK HTXLSSHG ZLWK ()V HDFK PHUJH ZLWK D categorizing x-head, (n in our case) which are likewise bearers of EFs. The assumption that roots bear EFs is the null-assumption, because otherwise roots could never enter the derivation. Now, it could be asked whether an EF on the categorizing x-heads is crucial, because according to (22) it should be enough that one of the Merge-partners bear an EF. The answer to this question falls out naturally from (23): 23

vP v()ij) ¥393 ¥ĮEF

DP

0HUJHU RI URRW Į LQ   ZLWK WKH '3 ZKLFK LV D 62 ZLWKRXW DQ () LV JXDUDQWHHGE\WKH()RQĮ,IKRZHYHUWKHFDWHJRUL]LQJ x-head, v in this case, did not bear an EF, Merger of v ZLWK¥3ZKLFKLVDOVRD62ZLWKRXW an EF, would never be possible.8 If (21) is the only prerequisite for Merge, Merger of two roots should, in principle, be possible, as illustrated in (24): 24 ¥ĮEF

¥ȕEF

+HUHWKHWZRURRWVĮ DQGȕ are merged via their EFs. Now, the question that emerges at this point is, how the two derivational alternatives sketched in (22) and (24) go together with differences in interpretation, which are expected under the SMT. 8

It could, of course, be argued that only some and not all x-heads bear EFs and that transitive v is such a head, whereas n maybe is not, but this leads to an unmotivated complication in the characterization of EFs that better be avoided (cf. Bauke 2012 for further elaboration).

A Phase-Theoretic Analysis of Root Compounds

159

Provided that the analysis outlined here is on the right track and that URRWVHQWHUWKHGHULYDWLRQZLWKRXWij-features, roots are never independently interpreted. Categorizing x-heads, on the other hand, are Phase-heads (cf. Marantz 2007) and as such they trigger Spell-Out and interpretation. Under this view, two derivational paths ensue. The first path, which is an instance of Merger of two roots does not involve a Phase: 25 a

0HUJH^Į`DQG^ȕ`ĺQR3KDVH

25 b ¥ĮEF

¥ȕEF

In a second step, the complex root in (25b) is then merged with a categorizing x-head. This x-head is a Phase head that triggers Spell-Out: 26 a

Merger of categorizing little x-head (n LQWKLVFDVH ĺ6SHOO-Out RIFRPSOHPHQWRIWKH3KDVHKHDGĺQRLQGHSHQGHQWPHDQLQJ UHDOL]DWLRQRIURRWVĮDQGȕGULIWHGUHDGLQJ

26 b

nP nEF ¥ĮEF

¥ȕEF

Compounds derived this way are not expected to show any compound LQWHUQDO LQIOHFWLRQDO PDUNLQJ EHFDXVH WKHUH VLPSO\ LV QR SODFH ZKHUH ijfeatures could be realized. The second path is one under which a root is merged with a categorizing x-head, which involves a Phase and thus triggers Spell-Out and interpretation: 27 a

0HUJHURIURRWĮDQGȕUHVSHFWLYHO\ZLWKFDWHJRUL]LQJOLWWOHx-head (n LQWKLVFDVH ĺ6SHOO-Out of complement of the Phase head ĺLQGHSHQGHQWPHDQLQJUHDOL]DWLRQRIURRWĮDQGURRWȕ respectively: compositional reading

27

nP nEF

¥ĮEF

nP nEF

¥ȕEF

Chapter Nine

160

In a subsequent step, the two categorized roots in (27b) are merged, yielding the structure in (28): 28

nP nP nEF

nP

¥ĮEF

nEF

¥ȕEF

Here compound internal inflectional material receives a natural H[SODQDWLRQ,WVLPSO\LVDUHIOH[RIWKHUHDOL]DWLRQRIWKHij-features of the categorizing n-head inside the compound. Hence, the two derivational paths that ensue here comport well with the observed differences in the realization of inflectional markers inside compounds and corresponding interpretational variants.

9.6 PoS (Points of symmetry) Both of the derivational paths outlined in section 9.5 and repeated here in (29), lead to a PoS. So, the question is which of the two nodes projects.9 29 a

PoS ¥ĮEF

¥ȕEF

b

PoS nP nEF ¥ĮEF

nP nEF

¥ȕEF

According to standard assumptions from the Linear Correspondence Axiom (LCA) (cf. Kayne 1994), this PoS has the potential of stalling the derivation when it remains unresolved. I here suggest that the PoS is dissolved in both (29a) and (29b), yet differently. Let me start by looking at (29b) first. Here I follow Roberts’ (2010) purely syntactic account of head-movement. In his approach to clitic incorporation Roberts argues for a syntactic analysis of head-movement that cannot be relegated to PF alone. Under this analysis clitics are identified as a minimal category, the relevant definition for minimal category that I use here is the following (cf. also Narita 2009 for a similar analysis): 9

Alternatively, both types of compounds could, of course, be analyzed as exocentric compounds (cf. also the discussion of the Chinese examples in section 3), but this comes at the cost of blurring the distinction between the two types that has been established in the preceding section.

A Phase-Theoretic Analysis of Root Compounds

161

30 A minimal category is an LI that bears an EF (adapted from Narita: 2011) Further, Roberts (2010) argues that the constraint on incorporation provided in (31) must hold and that the clitic must be a defective goal defective goals being defined as in (32): 31 Incorporation constraint Incorporation can take place only where the label of the incorporee is non-distinct from that of the incorporation host. (adapted from Roberts 2010) 32 Defectiveness A Goal G is defective iff G’s formal features are a proper subset of those of G’s Probe P. (adapted from Roberts 2010) The incorporation constraint is somewhat reminiscent of the conflation mechanism in other approaches (cf. e.g. Hale & Keyser 1993; Harley 2004), a point that I will address in some more detail in due course. Before, however, let me note that defectiveness, as defined by Roberts (2010) is a relative notion, that is inherently bound to the subset relation. Thus, the configuration that Roberts (2010) assumes for clitic incorporation is the following: 33 Trigger for Agree v* [Pers: unvalued, Num: unvalued] After Agree v >3HUVĮ1XPȕ@

FOLWLF>3HUVĮ1XPȕ@ FOLWLF>3HUVĮ1XPȕ@

Since the clitic element GRHVQRWEHDUDQ\&DVHWKHij-features on the FOLWLF FRQVWLWXWH D SURSHU VXEVHW RI WKH ij-features on v* . In consequence, WKHFOLWLFLQFRUSRUDWHVLQWRWKHYHUEDQGYDOXHVLWVij-features. Now let us see, how this reasoning plays out in nominal root compounds. Just like the v* -head, the n-head is a Phase head (cf. Marantz 2007), in contrast to the verbal Phase head, however, the nominal head does not bear unvalued person and number features.10 This does not mean 10

I here follow Roberts (2010) in assuming that [Person] and [Number] are the relevant features. Contrary to assumptions made elsewhere in the literature (cf. e.g. Delfitto, Fábregas & Melloni 2008), gender is not relevant here. For some insight into why this is the case, the reader is referred to the appendix. For a more detailed discussion cf. Bauke (2012).

Chapter Nine

162

though, that this head cannot act as a Probe. Following Gallego (2010) let us assume that syntax does not operate on the specifics of the interpretable features on LIs. All that syntax really cares about is whether a feature is valued or unvalued (cf. Gallego 2010; Chomsky 2001). Furthermore, a default value for any of the features should, in principle, be sufficient for marking a feature as valued. Under this reasoning EFs have a special status, in so far as they never receive a value, but can pass Spell-Out without causing a crash (cf. Chomsky 2008; Narita 2011). Finally, Match does not operate on feature values, it only operates on feature attributes (cf. again Gallego 2010 for a similar assumption). With that much in place, the picture in (34) emerges: 34 Trigger for Agree n >3HUVĮ1XPȕ@>()XQYDO@ After Match n >3HUVĮ1XPȕ@>()XQYDO@

n >3HUVȖ1XPį@>()XQYDO@ n >3HUVȖ1XPį@>()XQYDO@

In (34) both n-KHDGVKDYHYDOXHGij-features, so here it is only the EFs that can act as Probes. Despite their probing, the EFs will not be valued, which is unproblematic, because they can pass Spell-Out regardless. Yet, after Probing Match, which, as mentioned above, does not operate on feature values, but on feature attributes only, will identify that the features on both n-heads match and as a result, incorporation ensues: 35

n nĮ

nȕ

The remaining question still is, which of the two n-heads is the projecting head. This is determined by the feature configuration on the two heads. Both n-heads lack a specification for Case.11 The person-features on the two n-heads are both default: [3]. So what remains is the Number feature. Here at least one head must have a non-default, i.e. marked [PL], VSHFLILFDWLRQ RWKHUZLVH WKH ij-feature configurations on the two n-heads would be so similar that instead of incorporating one head into the other, one of the two heads would be deleted from the derivation entirely (cf. 11

At least they are not specified for structural Case. Inherent or lexical Case may well be of relevance as is briefly discussed below (cf. also Bauke 2012) for further elaboration.

A Phase-Theoretic Analysis of Root Compounds

163

Roberts 2010 for the specifics). So, let us assume the following feature configurations for the two n-heads in (36): 36 nȕ [Pers: default, Num: plural/marked] [EF: unval.] 37 nĮ [Pers: default, Num: random but valued] [EF: unval.] [Case: unvalued] (36) is a proper subset of (37) and thus incorporation ensues. Crucially, KRZHYHUWKH1XPEHUIHDWXUHRQȕLVPDUNHG>3/@ZKLFKPDNHVWKHWZRnheads sufficiently distinct. Notice further that the plural marker on the incorporated n-head is embedded in the incorporation host. This quite naturally predicts that this plural is not interpreted as a genuine plural, but rather as opening up the various interpretations listed in section 9.2. Should both, nĮ and nȕbear a default value on [Number], genitive Case may function as a last resort. This is briefly outlined in (38): 38 a

Länderspiel (German) country.PL + game

b

Landeskriche (German) country.GEN + church

In (38b) Land does not bear plural morphology. It is marked for genitive Case instead. The effect is the same though, i.e. that a variety of interpretations emerges that is not available for the contrastive form Landkarte (cf. (12a)) where only one interpretation is available.12 Before, briefly looking at how the PoS in the other type of nominal root compound identified here is dissolved, let me now pause and compare the present analysis to the approach on primary compound incorporation in Harley (2004, 2009). Conceptually, both approaches are very similar. As already mentioned above Harley’s (2004) head conflation mechanism is somewhat similar to the incorporation constraint of Roberts (2010). And the trigger for incorporation in Harley (2009) is a Case-feature or a [± affix] feature - both of which are not specified any further. The only significant difference between the incorporation approach in Harley (2009) and the analysis outlined here then boils down to the incorporation site. In 12

i

Notice also that in some cases even two genitive forms are available:

Landsmann country.GEN + man All of this strongly suggests that neither the genitive marker nor the plural marker can be analyzed as a linking element, as this would immediately beg the question of why more than one linker is available.

Chapter Nine

164

Harley (2009) the modifying noun, i.e. the incorporated noun, is fully FDWHJRUL]HG SUHVXPDEO\ LQFOXGLQJ D FRPSOHWH ij-feature specification). This noun then incorporates into a root, which is categorized only in a second step: 39

nP n°

¥3

¥Qƒ ¥Į nP nq

¥Į

‡

Į

¥ȕQƒ Į ȕ

n°

¥ȕ

‡

ȕ

‡ (adapted from Harley 2009)

In contrast, in the analysis here incorporation takes place only after both roots have been categorized and the trigger for incorporation is not an DIIL[ RU &DVH IHDWXUH EXW UHVXOWV IURP WKH ij-feature configurations of the two n-heads instead. 40

nP n° nP n° ¥ȕQƒ

nP n°

¥ȕ

n°

n°

¥ȕ

¥Į ¥ȕQƒ

¥Į n°

This has the effect that recursive structures are derivable. This is not the case, when incorporation does not go into another noun but into a root instead - at least not when the compounds are strictly rightward branching: 41 [restaurant [coffee [cup]]] 42 [[[gourmet] [coffee]] [cup]]

A Phase-Theoretic Analysis of Root Compounds

165

So the compound in (42) is equally derivable under both analyses, whereas the compound in (41) can only be derived under the analysis outlined here. This is immediately obvious from the structure in (39). The result of incorporation is a categorized noun. This categorized noun can, of course, be incorporated into another root that is then categorized in a second step. This, however, is the derivation for (42). What is impossible is that the complex n serves as an incorporation host under Harley's (2009) analysis, because this would require that it had the status of a root, i.e. the already categorized n would need to be reanalyzed as a root. This, however, is a somewhat ad hoc assumption that seems to suggest that reanalysis is possible randomly and it can be avoided by the analysis suggested here. Finally, let us take a short look at how the PoS is dissolved for the other type of compound: 43

PoS ¥ĮEF

¥ȕEF

The two roots both contain EFs, which are, of course, unvalued and can thus act as Probes. However, again no valuation and no labeling will ensue for these features, which will pass Spell-Out unmolested though (cf. Chomsky 2008). So what we are looking at here is the structural configuration that leads to total conflation in Roberts (2010). Incidentally, this is what explains the lack of compositionality and recursivity observed for this type of compound. The PoS can then simply be dissolved by standard assumptions on antisymmetry (cf. Kayne 1994, Moro 2000 and several others), i.e. in terms of PF-movement. This again comports well with the fact that there are no alternative readings available for this type of compound. This is what is expected, when the two roots are unanalyzed in syntax and the PoS is dissolved only after Spell-Out and at the PFinterface. At the same time this means that the compound must be noncompositional, because when merging a third root in the same derivational cycle this would not lead to any embedding but would derive a flat structure that cannot be linearized at PF. Ultimately, this line of reasoning is also in line with parametric variation. As has been shown above, Merger of three roots is generally unavailable across languages. Merger of two n-heads is possible in Germanic languages, but unavailable in Romance, because here [Number] is not associated with n (cf. e.g. Alexiadou, Haegeman & Stavrou 2007). Thus, Number here emerges as the place for parametric variation (cf. also Roberts 2010 for a similar conclusion).

Chapter Nine

166

Appendix ,Q WKH DQDO\VLV SUHVHQWHG KHUH WKH ij-features relevant for the incorporation analysis are [Person] and [Number]. [Gender] has been explicitly excluded from the considerations here. However, in alternative analyses, such as the one provided by Delfitto, Fábregas & Melloni 2008) it is precisely the Gender feature that is identified as the feature that dissolves the PoS. Delfitto, Fábregas & Melloni (2008) define a PoS as a structural configuration that emerges from the Merger of two elements that share the same level of structural complexity. This PoS is then dissolved by a functional head F, which is merged on top of the PoS, acting as a Probe and that ultimately leads to PoS dissolution. So, in essence, the approach by Delfitto, Fábregas & Melloni (2008) is similar to that in Moro (2000) though it is not as general as his account is. While Moro (2000) relates all movement operations to PoS dissolution and ties this to the movement operation to structural configurations rather than to the properties of specific features, Delfito, Fábregas & Melloni (2008) identify a PoS on the basis of structural configurations as well, but relate the dissolution of the PoS to specific features probed by the F-head. The difference between Germanic and Romance languages then results from where the PoS arises - and is dissolved. The structure that they provide for Germanic is given below: 44 Germanic: FP nP n

FP

¥Į) PoS nP

nP

n ¥Į n ¥ȕ (adapted from Delfitto, Fábregas & Melloni 2008) So, here F acts as a Probe that targets an interpretable but unvalued feature on the n-head, thus instantiating a Compound Phase that is subject to an earliness principle. The feature targeted is a Gender feature that is derivable from the declension class of the noun. The F-head is realized as a linking element then and constitutes the head of the compound. However, F does not activate the lexical semantics on any of the two

A Phase-Theoretic Analysis of Root Compounds

167

nouns, which is what accounts for the flexible interpretation (cf. Delfitto, Fábregas & Melloni 2008). In Romance languages, where Gender and declension class are not UHODWHG)3WDUJHWVij-phrases instead of nPs and the feature that is targeted by F is a Quale-oriented feature instead: 45 Romance: FP ij3 ij nP n

FP F

PoS

¥Į ij

ij3

ij3 nP ij

nP

n ¥Į n ¥ȕ (adapted from Delfitto, Fábregas & Melloni 2008) So the difference between Romance and Germanic languages in this approach comes down to a distinction between F targeting the Gender feature in the latter language family, while targeting the Quale-feature in the former. This has the additional effect that the projecting head is the non-moved head in Germanic, while it is the moved element in Romance.13 However, a number of questions remain open under this analysis.14 First of all, it is not clear how it can be guaranteed that the derivation “knows” when to perform Parallel Merge (i.e. at the level of nP in *HUPDQLFDQGDWWKHOHYHORIij3LQ5RPDQFH *LYHQWKDWD3R6LVGHILQHG as arising from the same amount of structural complexity on the two constituents, it seems unavoidable to give way to at least some amount of look-ahead to calculate the Merger site of F in the two language families. 13

Actually, the situation is a bit more complex here and Delfitto, Fábregas & Melloni (2008) make a distinction between syntactic and semantic heads and between Phase and non-Phase heads. However, a thorough analysis of the distinctions they make would lead us too much astray here. I refer the interested reader to Delfitto, Fábregas & Melloni’s original analysis here and to Bauke (2012) for a critical reflection. 14 For reasons of space, I will simply list the unresolved issues here. For discussion and analysis the reader is again referred to Bauke (2012).

168

Chapter Nine

Closely related to that, is the question of how the derivation keeps track of the internal structural make-up of the two Merge partners. It has repeatedly been argued that counting mechanisms are generally unavailable in grammars (cf. e.g. Boeckx 2008, Hornstein 2009), however it is not clear how “the same amount” of structural complexity can be calculated otherwise. This, in turn, brings up the more general question of how it can be guaranteed that the PoS dissolving and Phase-instantiating head has the required language specific features. Or to phrase it differently it is not clear whether the F-head that probes Gender features in Germanic and Quale features in Romance is really the same head. Talking of Phase heads, it could also be asked why the feature on the Phase head is valued in Delfitto, Fábregas & Melloni’s (2008) account and why it targets an unvalued feature lower in the tree - rather than the reverse. Under the assumption that the F-head is realized as a linking element in Germanic, it is also unclear, how the correct linking element is chosen. According to Delfitto, Fábregas & Melloni (2008) this is regulated by the correlation between declension class and gender in these languages. But how then can the compounds that allow more than one linking element per stem be factored in and what is the distinction between a linking element that is not realized phonologically and one that is? This latter question has deep consequences for the analysis laid out in this paper, because it undermines the distinction between compositional and non-compositional compounds. In the same vein the question of what happens in those Germanic languages where [Gender] is never overtly realized needs to be addressed, just as much as the reverse question of how those Romance forms that display a clear correlation between Gender and declension class (cf. Alexiadou, Haegeman & Stavrou 2007) can be factored in. And ultimately the question of how recursivity can be accounted for remains unresolved. Admittedly, none of the questions directed towards the approach by Delfitto, Fábregas & Melloni (2008) is answered in full by the analysis presented in this paper. However, a more principled analysis seems feasible, when pursuing the path laid out here.

References Alexiadou, A., L. Haegeman and M. Stavrou. 2007. Nouns Phrase in the Generative Perspective. Berlin Mouton de Gruyter. Bauke, L. 2009. Nominal Root Compounds in German. Ms. UMass.

A Phase-Theoretic Analysis of Root Compounds

169

—. 2012. Lexical and non-lexical symmetry breaking: an analysis of nominal root compounds, nominal gerunds and small clauses. PhD diss., University of Frankfurt. Boeckx, C. 2008. Bare Syntax. Oxford: Oxford University Press. —. 2010. What Principles and Parameters got wrong. Ms. Available at http://ling.auf.net/lingbuzz/001118. Borer, H. 1984. Parametric Syntax. Dordrecht: Foris. Chomsky, Noam 1970. Remarks on Nominalization. In R. Jacobs and P. Rosenbaum (eds.), Readings in transformational grammar, 184–221. Washington: Georgetown UP. —. 1995. The Minimalist Program. Cambridge, MA: MIT Press. —. 2001. Derivation by Phase. In M. Kenstowicz (ed.), Ken Hale: a life in language, 1–52. Cambridge, MA: MIT Press. —. 2005. Three factors in language design. Linguistic Inquiry 36: 1-22. —. 2008. On Phases. In R. Freidin, C. Otero and M.-L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory, 133–166. Cambridge, MA: MIT Press. Delfitto, D., A. Fábregas and C. Melloni. 2008. Compounding at the interfaces. Paper presented at NELS. Gallego, A. 2010. Phase Theory. Amsterdam: John Benjamins. Giegerich, H. 2007. Compounding and lexicalism. In R. Lieber and P. Štekauer (eds.), The Oxford Handbook on Compounding. Oxford: Oxford University Press. Harley, H. 2004. Merge, conflation and head movement. In K. Moulton & M. Wolf (eds.), Proceedings from NELS 34, 239–254. Amherst MA: GLSA. —. 2009. Compounding in Distributed Morphology. In R. Lieber and P. Štekauer (eds.), The Oxford Handbook on Compounding, 129–144. Oxford: Oxford University Press. Hale, K. and Keyser, S. 1993. On the argument structure and the lexical expression of syntactic relations. In K. Hale and S. Keyser (eds.), The view from building 20: essays in linguistics in honor of Sylvain Bromberger, 53–109. Cambridge MA: MIT Press. Hornstein, N. 2008. Theory of syntax. Cambridge: Cambridge University Press. Kayne, R. 1994. The antisymmetry of syntax. Cambridge, MA: MIT Press. Keyser, S. J and T. Roeper 1992. Re: the abstract clitic hypothesis. Linguistic Inquiry 23: 89–125. Marantz, A. 2001. Words. Ms. MIT. —. 2007. Words and Phases. Ms. NYU. Moro, A. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press.

170

Chapter Nine

Narita, H. 2008. Full interpretation of optimal labeling. Biolinguistics 3(2-3). 213-254. —. 2011. Phasing in Full-interpretation. PhD diss., Harvard University. Roberts, I. 2010. Agreement and head-movement. Cambridge, MA: MIT Press. Roeper, T., Snyder. W and Hiramatsu. K. 2002. Learnability in a minimalist framework: Root compounds. merger and the syntaxmorphology interface. In I. Lasser (ed.), The process of language acquisition. Frankfurt: Peter Lang. Roeper, T and W. Snyder. 2005. Language learnability and the forms of recursion. In A. M. DiSciullo (ed.), UG and external systems, 155– 169. Amsterdam: John Benjamins. Selkirk, L. 1982. The syntax of words. Cambridge MA: MIT Press. Zhang, N. 2007. Root Merger in Chinese Compounds. Studia Linguistica 61(2). 170–184.

CHAPTER TEN RECOVERABILITY IN SLOVENE RELATIVE CLAUSES MARKO HLADNIK1 10.1 Introduction This paper discusses resumption patterns in Slovene relative clauses based on data from the standard variety of the language as well as from dialectal varieties, gathered in a larger survey by means of a questionnaire and additional interviews with speakers.2 The central claim of the paper is that resumption in relatives is a result of the interplay between spell-out restrictions in the CP domain and the requirement that the case features identifying the gap be recoverable. The paper describes the available relative clause constructions in Slovene, focusing on the elements introducing relative clauses and on resumption facts before providing an analysis of the resumption patterns identified.

10.2 Descriptive Facts 10.2.1 Two Relative Constructions In all varieties of Slovene, standard and dialectal, we find two ways of constructing a short distance relative clause. They are exemplified by the sentences in (1) and (2), with the differences highlighted. Note that I resort to writing examples in standard Slovene rather than trying to reflect the 1

The author would like to thank Sjef Barbiers, Norbert Corver, Marjo van Koppen, and audiences of the Syntax Circle in Utrecht, the TiN Dag 2011, and the INFL Conference 2011 for useful discussions and guidance. All errors are entirely my own. This research has been supported by the Netherlands Organisation for Scientific Research (NWO). 2 Over 70 responses have been collected from 55 different test locations.

Chapter Ten

172

dialectal pronunciation, since in the provided examples this does not obscure possible syntactic differences (but cf. footnote 6). 1 Poznam þORYHND katerega so iskali.3 know.1SG man.ACC KATERI.ACC AUX.3PL looked-for 2 Poznam þORYHND ki so ga iskali. know.1SG man.ACC KI AUX.3PL he.ACC.CL looked-for Both: ”I know the man they were looking for.” The first possibility is to introduce the relative clause by the element kateri in an appropriate morphological form (see subsection 10.2.2). The other one is to introduce it by ki, and use a resumptive clitic within the relative clause (see subsection 10.2.3). Schematically, the two constructions can be represented as in example (3). 3 a b

[matrix clause ... head noun [relative clause kateri ... gap]] [matrix clause ... head noun [relative clause ki ... resumption]]4

10.2.2 Relative Elements De Vries (2002, 155-62 and references therein) relates the key relative construction properties to the syntactic characteristics of the relative elements. Subordination is expressed by relative complementizers, which carry neither M-features nor case, while attribution to the head of the relative clause and the identification of the gap within the clause are associated with relative pronouns bearing both. The Slovene element kateri appears in any of the six case forms of Slovene, which relate the head noun to a position in the relative clause. At the same time it agrees with the head noun in gender and number. This leads us to conclude that kateri is a relative pronoun. Typologically, these often show demonstrative or wh-morphology (de Vries 2002, 173), which

3

Slovene orthography requires a comma even with restrictive relative clauses, in contrast with the English rule. Though this paper focuses on restrictive relative constructions, appositional relatives do not formally differ from restrictive ones in Slovene. 4 Similar patterns of relativization exist in other Slavic languages. For Serbo-&URDWLDQ VHH *RRGOXFN  6WRMDQRYLü    RU DQ HDUOLHU GHVFULSWLRQ E\ Browne (1986).

Recoverability in Slovene Relative Clauses

173

also holds true for kateri – it is identical in form to the interrogative ‘which’. In Table 10.1, the singular paradigm of kateri is shown. Table 10.15 masculine

feminine

neuter

nominative

kateri

katera

katero

genitive

katerega

katere

katerega

dative

kateremu

kateri

kateremu

accusative

katerega

katero

katero

locative

(pri) katerem

(pri) kateri

(pri) katerem

instrumental

(s) katerim

(s) katero

(s) katerim

The element ki, on the other hand, has the characteristics of a relative complementizer; it is invariable in form, and caries no M-features or case. Its role is specialized, i.e. it appears only in relative constructions.6 There are limitations on what can be spelled out in the CP domain of Slovene relative constructions. Firstly, a Doubly Filled COMP Filter type of restriction in the sense of Chomsky & Lasnik (1977) applies: not more than one element can be expressed (cf. example 4). Secondly, zero relativization is not allowed: one element must always be expressed (cf. example 5). The presence or absence of resumption plays no role when these requirements are violated. 4 *Poznam þORYHND katerega ki so (ga) iskali. know.1SG man.ACC which.ACC C-rel AUX.3PL (him.CL) looked-for 5 3R]QDPþORYHND so (ga) iskali. know.1SG man.ACC AUX.3PL (him.CL) looked-for Intended: “I know the man they were looking for.” 5

Locative and instrumental are cases whose assignment in Slovene is limited to certain prepositions. The prototypical ones included in the table are pri 'at' for locative and s 'with' for instrumental. 6 This is true for the standard variety and the majority of dialects. In a geographically contained group of dialects in eastern Slovenia, however, a complementizer with a form resembling ki is used both in general subordinate contexts and in relative ones.

Chapter Ten

174

This leaves us with the two options presented at the beginning (see the schematic representations in example 3): either only the pronoun or only the complementizer is present.

10.2.3 Resumption As described in section 10.2.1, resumption takes place when the relative clause is introduced by the complementizer ki. This applies to object relatives, whereas in subject relative clauses there is no resumption even with the complementizer. With object relatives, resumption is obligatory, as shown in (6a). Furthermore, the resumptive pronoun must be a clitic, tonic pronouns will not do, as demonstrated by (6b). 6 a

b

To MHþORYHN ki *(ga) LãþHMR this is man.NOM C-rel he.ACC.CL look-for.3PL “This is the man they are looking for.”

7RMHþORYHNNL QMHJDLVþHMR this is man.NOM C-rel he.ACC look-for.3PL Intended: “This is the man they are looking for.”

There exist no locative or instrumental clitics in Slovene. Locative, instrumental, and other prepositional object relative clauses require resumption in the form of a prepositional phrase with a tonic pronoun as its complement. The overview of resumptive elements in the singular paradigm is given in Table 10.2. I put forward an analysis of the resumption patterns in the following section, starting with the shaded area of the table, i.e. clitic resumption in object relatives. Table10. 27 nominative genitive dative accusative locative instrumental 7

masculine

feminine

neuter

-ga mu ga pri njem z njim

-je Ji jo pri njej z njo

-ga mu ga pri njem z njim

The forms of Slovene masculine and neuter personal pronouns, both tonic and clitic, are syncretic (also in the dual and the plural).

Recoverability in Slovene Relative Clauses

175

10.3 Proposed Analysis 10.3.1 Resumption in Object Relative Clauses Relative clause constructions are generally assumed to involve whmovement of a pronoun (or an operator) to the left edge of the clause. The positions of the relative pronoun and of the resumptive element at the extraction site are thus related, and as observed in Slovene object relatives, the presence of one precludes the presence of the other.8 My assumption is that the two available relative constructions share one and the same underlying derivation. Both alternatives, also the one with no overt relative pronoun, involve wh-movement as we show below. Examples (7a) and (7b) illustrate the impossibility of wh-extraction out of the relative clause constructions. Parallel to (7a), the example in (7b) contains material in the intermediate [Spec,CP] position, preventing the movement of kaj ‘what’ through it and thus rendering the example ungrammatical. The difference is only that in (7a) the element in question is phonetically null. 7 a

b

.DMYLGLPþORYHNDNDWHUHPXVHPGDOB" what see.1SG man.ACC which.DAT AUX.1SG gave *Kaj vidim þORYHNDNL sem mu dal _? what see.1SG man.ACC C-rel AUX.1SG he.DAT.CL gave Both: *”What do I see a man whom I gave _?”

The fact that the element in the intermediate [Spec,CP] has moved there, rather than being base-generated in that position, can be demonstrated using known islands as a diagnostic tool for movement. Slovene displays sensitivity to strong islands (8), but as shown in Golden (1997), embedded wh-interrogative clauses such as in (9) do not constitute a syntactic island [original example (18a)]. je Peter odpustil ti? 8 *Kogai je Janez jezen, ker who is J. angry because AUX.3SG P. fired *”Who is John angry because Peter fired?”

8

That resumption in relative constructions precludes the use of relative pronouns is a typological universal (de Vries 2002, 37 and 165), or at the very least a strong tendency. Slovene data complies with these facts.

176

Chapter Ten

9 Katere knjigei se je Peter spraševal, kdoj tj bo izdal ti? which books SELF AUX.3SG P. wondered who would publish “For which books did Peter wonder who would publish them?” When we form relative clauses introduced by the complementizer ki that involve an adjunct island, the result is ungrammatical. Relativization out of an embedded wh-clause, on the other hand, is licit. 10 þORYHNNL je Janez jezen, ker ga je Peter odpustil man C-rel is J. angry because him AUX.3SG P. fired *”the man that John is angry because Peter fired” 11 þORYHNNLMH Janez pozabil, kje ga je spoznal forgot where him AUX.3SG met man C-rel AUX.3SG J. “the man that John forgot where he met him” Serbo-Croatian D UHODWHG 6ODYLF ODQJXDJH DFFRUGLQJ WR %RãNRYLü (2009) shows both wh- and adjunct island effects. This is reflected in the ungrammatical status of relativization from within these islands as well [original examples (6) and (2b)]. 12 þRYMHNãWR si otišao zato što ga je Petar otpustio fired man C-rel AUX.2SG went-away because him AUX.3SG P. *”the man that you went away because Peter fired.” 13 þRYMHNãWR je zaboravio gdje ga je upoznao where him AUX.3SG met man C-rel AUX.3SG forgot “the man that he forgot where he met him” The parallelism in islandhood and relative clause acceptability suggests that movement is involved also in the formation of Slavic relatives introduced by complementizers, and not only those with an overt wh-pronoun. I argue that the observable differences between pairs of sentences such as in (14) and (15) arise from the choice of which element in the CP domain is spelled out. 14 To MHþORYHNkaterega LãþHMR this is man.NOM which.ACC look-for.3PL

Recoverability in Slovene Relative Clauses

177

15 To je þORYHN ki ga LãþHMR this is man.NOM C-rel he.ACC.CL look-for.3PL Both: “This is the man they are looking for.” Let us take a closer look at the two alternatives in turn. The two positions in the CP domain are boxed by a dashed line in (14’) and (15’). The relative complementizer one is labelled C-rel and the specifier position the relative pronoun moves to is labelled D-rel. Only one of the two can host an overt element at the same time (as discussed in 2.2 above). 14’

To je þORYHN katerega C-rel LãþHMR this is man which look-for.3PL Features: wh/relative wh/relative NOM ACC M M “This is the man they are looking for.”

katerega wh/relative ACC M .

In (14’), the relative pronoun kateri is spelled out in the [Spec,CP] position as the highest copy in its movement chain, thereby suppressing the pronunciation of the complementizer. The pronoun carries a wh-feature (or a special relative feature), M-features matching those on the head noun, and a case feature that identifies the gap position in the embedded clause. The key mechanism of the analysis comes into play when the second option is chosen.9 In (15’), it is the complementizer that is overt, and to comply with the limitations on the CP-domain spell-out, the relative pronoun cannot be pronounced. The unique case feature on the pronoun is thus not spelled out. In order to preserve the case and ensure that the gap information it encodes is recoverable, the minimal element that spells it out – a clitic – becomes overt in the lower copy position. Note that, as is generally true for all clitics in Slovene, the resumptive clitic is pronounced in the second position (or as part of a second position clitic cluster). 15’

7RMHþORYHN D-rel ki LãþHMR D-rel this is man C-rel look-for.3PL Features: wh/relative wh/relative wh/relative NOM ACC ACC M M . M “This is the man they are looking for.” = ga (he.ACC.CL)

9

I treat the two options as equal, although in any given context speakers prefer one over the other. Crucially, both constructions are available in their grammars.

178

Chapter Ten

Not all the features of the lower copy are selected for spell-out (otherwise the result would be kateri), only the key case feature and the associated M-features necessarily encompassed by the clitic. Together, the complementizer and the resumptive clitic, the overt elements, spell out the same set of features in sentence (15) as the overt relative pronoun does in the alternative construction in (14).10 Support for this type of case-recoverability mechanism in language can be found in the work of Alexiadou et al. (2010), dealing with a different domain. They analyze Greek and Romanian control as raising from the embedded clause, in the course of which the case on the subject may get rewritten in the matrix clause (when there is a case mismatch due to a quirky subject in the embedded clause). The case of the embedded subject then needs to be realized by an obligatory clitic spelling out the relevant case and M-features. In the Greek example in (16) this is genitive [original example (51a)], and in the Romanian one in (17) the case is dative [original example (52b)]; the clitics are highlighted in bold print in both examples. 16 I Mariai tolmise na tisi aresun i operes the MaryNOM dared3SG Csubj she.CL.GEN please.3PL the operas.NOM “Mary dared to like operas.”

[Greek]

17 Ioni a UHXúLWVă nu-ii [Romanian] John.NOM AUX.3SG managed C-subj not-he.CL.DAT VFDSHPXOWHJUHúHOL escape.3PL many mistakes “John managed so that not many mistakes escaped his attention.”

10.3.2 No Resumption in Subject Relative Clauses As discussed, subject relative constructions involve no resumption. There are no nominative clitics in Slovene, and using a tonic pronoun is not an alternative. Compare the examples in (18): 18 a

10

3R]QDPþORYHND ki LãþH VOXåER know.1SG man.ACC C-rel look-for.3SG job “I know a man who is looking for a job.”

The first to operate with the notion of obligatory case recoverability was Broihier (1995), who in his unpublished manuscript discussing Polish and Russian relatives in an Optimality Theory framework writes that we “[need to consider the] possibility that certain case morphology also qualifies as unrecoverable”.

Recoverability in Slovene Relative Clauses

b

179

3R]QDPþORYHND ki on LãþH VOXåER know.1SG man.ACC C-rel he.NOM look-for.3SG job Intended: “I know a man who is looking for a job.”

This absence of resumption in the subject position is a near-universal property of languages,11 and not specific to Slovene constructions. It has been proposed by McCloskey (1990 and 2005) that the following restriction holds: 19 Highest Subject Restriction (HSR) A resumptive pronoun cannot occupy a subject position immediately subjacent to its binder. Informally speaking, resumptive elements are obligatorily bound, and since they are by virtue of their form and their properties pronouns, they also behave like pronouns in that they cannot be bound too locally. HSR thus rules out subject resumption in the highest clause, while embedded subject resumption still takes place – in Irish and Hebrew, for example – and is fully grammatical (ibid.). In Slovene long relatives, however, subject resumption is ruled out even in long distance relative constructions (see section 10.3.4 below for a discussion of long relatives): 20 a b

Poznam þORYHND ki mislim, da LãþHVOXåER know.1SG man.ACC C-rel think.1SG that look-for.3SG job

3R]QDPþORYHNDNL mislim, da on LãþHVOXåER know.1SG man.ACC C-rel think.1SG that he.NOM look-for.3SG job “I know a man who I think is looking for a job.”

This suggests that HSR may in fact not apply in Slovene. The following non-nominative subject constructions point in the same direction. The accusative and the dative case of the nominals in the highest position in (21) and (22) are obligatorily preserved in the corresponding relativizations (21’) and (22’). 21 Janeza zebe. John.ACC be.cold “John is cold.”

11

Cf. Comrie & Kuteva (2005) for documented exceptions.

180

Chapter Ten

22 Janezu paše plesati salso. John.DAT pleases to.dance salsa “John likes to dance salsa.” 21’ Poznam þORYHND ki ga zebe. know.1SG man.ACC C-rel he.ACC.CL be.cold “I know a man who is cold.” 22’ Poznam þORYHNDNL mu paše plesati salso. know.1SG man.ACC C-rel he.DAT.CL pleases to.dance salsa “I know a man who likes to dance salsa.” If we subscribe to Neeleman & Weerman (1999), who argue that nominative is not a case, but rather the absence of it, resumption facts follow from the analysis as presented above. When there are no case features to be recovered, there is no (nominative) resumption either. Slovene, however, need not be a counter-example to McCloskey's restriction in (19), since not all resumptive pronouns appear to have been created equal. Engdahl (1985) in her study of Swedish establishes that resumptives in Swedish have trace properties, and concludes they are to be analysed as overt realization of traces, and not as bound pronouns in syntax. If Slovene resumptive pronouns likewise share properties with traces, that would explain why HSR is not applicable and would in fact be the desired state of affairs given that the analysis presented considers resumption in Slovene to be the (partial) spell-out of moved material. One of the properties of empty categories is that they license parasitic gaps. In Slovene, parasitic gaps are rather limited, but they are possible, same as in Russian, and in contrast to Polish and Serbo-Croatian (Franks 1995) when it comes to comparing Slavic languages. An environment where parasitic are most acceptable in Slovene is relative clauses (Golden 1997). If we adapt an example from Golden (1997) and turn a kateri-type relative into a ki-type one with resumption, the result is grammatical: 23 To je predavatelj, ki ga vsak, ki spozna _pg_, ceni. values this is lecturer.NOM C-rel he.ACC.CL everyone C-rel meets “This is the lecturer everyone who gets to know appreciates.” Despite the resumptive in the relative clause [ki ga vsak ceni ‘who everyone appreciates’], the parasitic gap in the embedded relative clause [ki spozna __ ‘who gets to know __’] is licensed. Thus, at least when we

Recoverability in Slovene Relative Clauses

181

consider the licensing of parasitic gaps, Slovene resumptives behave like traces or copies.

10.3.3 Prepositional Relative Constructions Preposition stranding is banned in Slovene. Compare the ungrammatical example (24a) to the grammatical (24b): 24 a b

*Kom govoriš o? who.LOC talk.2SG about O kom govoriš? about who.LOC talk.2SG “Who are you talking about?”

The exact same restriction of course applies to relative constructions as well. When the relative clause involves an overt relative pronoun, the pronoun cannot move on its own as in (25a), but must rather pied-pipe the entire prepositional phrase to the left edge of the relative clause as in (25b). 25 a

*Poznam þORYHND katerim govoriš s. know.1SG man.ACC which.INSTR talk.2SG with

b

Poznam þORYHND s katerim govoriš. know.1SG man.ACC with which.INSTR talk.2SG “I know the man you are talking with.”

Choosing the option of introducing the relative clause with an overt complementizer means that the highest PP copy in [Spec,CP] must be left unpronounced. The lowest chain link position then necessarily gets spelled out as a PP the nominal part of which is a [-wh/relative] pronoun, as we see in (26).12 Such spell-out preserves both the case in question as well as the preposition. 26 Poznam þORYHND ki govoriš z njim. know.1SG man.ACC C-rel talk.2SG with he.INSTR “I know the man you are talking with.” 12 Although this construction is considered archaic or stilted, it is nevertheless attested and available.

Chapter Ten

182

It is expected that such resumption obligatorily takes place in prepositional object relatives not introduced by a relative pronoun in all languages without preposition stranding.13

10.3.4 Long Distance Relatives and an Additional Type of Resumption Long distance relative constructions are those associated with the head noun does not appear subordinated clause, but rather in the clause that is deeper. This is schematically shown in (27) and a given in (28):

where the position in the immediately embedded one level Slovene example is

[matrix clause head noun [relative clause relative element … … [fin. embedded clause gap/resumption]]]

27

28 Poznam þORYHNDki mislim, da LãþH VOXåER know.1SG man.ACC C-rel think.1SG that look-for.3SG job “I know a man who I think is looking for a job.” In (28) man plays a role in the matrix clause (as the internal argument of know) and in the most deeply embedded clause (as the external argument of look for), and a possible paraphrase of the sentence is: I know a man and I think that this man is looking for a job. The construction can be analysed on a par with short distance relative clauses, with further steps of cyclic movement, regardless of what the precise syntactic analysis of relative clauses is. In Slovene, long relative constructions pattern with short relative clauses in the relevant aspects of obligatory and illicit resumption as well. Long relative constructions of the ki-type thus obligatorily involve resumption (29a) with object relatives, while the kateri-type relatives do not (29b), for example. 29 a b

13

3R]QDPþORYHND ki mislim, da *(ga) LãþHMR know.1SG man.ACC C-rel think.1SG that he.ACC.CL look-for.3PL 3R]QDPþORYHND katerega mislim, da LãþHMR think.1SG that look-for.3PL know.1SG man.ACC which Both: “I know the man I think they are looking for.”

Hebrew, for example, is a case in point (cf. Borer 1984).

Recoverability in Slovene Relative Clauses

183

We need to distinguish several types of resumption. A two-way distinction has been noted in the literature, already by e.g. Sells (1984). The first type is resumption as a last-resort operation, where resumptive elements occur in movement islands. They grant immunity from constraints on movement and save or improve (as in 30a) the otherwise ungrammatical utterance (30b). 30 a b

?I wonder who they think [that [if Mary marries him] then everybody will be happy]]. *I wonder whoi they think [that [if Mary marries ti] then everybody will be happy]].

The other type is the resumption we find it in relative clauses introduced by complementizers in languages like Slovene – the phenomenon discussed in this paper. There are no movement islands involved, yet resumption needs to take place and the sentences have no degraded status whatsoever. In addition to these two, however, yet another type has been identified in Slovene long relative clauses. Several speakers report the possibility of resumption in long relatives introduced by the pronoun kateri, as in (31).14 31 Poznam þORYHND katerega mislim, da (ga) LãþHMR know.1SG man.ACC which.ACC think.1SG that he.ACC.CL look-for.3PL “I know a man who I think they are looking for.” Such constructions are ungrammatical in Standard Slovene. They are not geographically limited, i.e. their availability is not dependent on the dialect, and the use of this type of resumption is optional. The construction seems to calls for a processing account, since it is restricted to the spoken varieties of language and because this third resumption type apparently facilitates the formation of the long-distance dependency.

10.4 Conclusion This paper has shown how the effects of a necessary choice as to what element gets spelled out in the complementizer domain of the clause can account for resumption patterns in relative clauses in Slovene. Resumption 14 Contrastively, in short relatives, using resumption in the construction with an overt relative pronoun is ungrammatical in all varieties of Slovene.

184

Chapter Ten

takes place to ensure the recoverability of case features identifying the gap the head noun is related to. More work is needed to examine resumption in other relative constructions, and especially for the analysis to encompass the facts of cross-linguistic variation within Slavic languages and beyond, as languages may and do differ as to the conditions on recoverability. Likewise, it needs to be explored how resumption as a reflex of recoverability relates to instances of resumptive pronoun use in other contexts. With the analysis presented in the paper, this endeavour appears to be off to a promising start.

References Alexiadou, A., E. Anagnostopoulou, G. Iordachioaia and M. Marchis. 2010. No objections to Backward Control?. In N. Hornstein and M. Polinsky (eds.), Movement Theory of Control, 89–118. (= Linguistik Aktuell/Linguistics Today. 154). Amsterdam and Philadelphia: John Benjamins. Borer, H. 1984. Restrictive relatives in modern Hebrew. Natural Language and Linguistic Theory 2(2). 219–260. %RãNRYLüäµ2QUHODWLYL]DWLRQVWUDWHJLHVDQGUHVXPSWLYHSURQRXQV¶ FDSL 7 Proceedings, 1–13. Broihier, K. 1995. Slavic Relatives. Resumptive Pronouns and Learnability. Ms. MIT. Browne, W. 1986. Relative Clauses in Serbo-Croatian in Comparison With English. Institute of Linguistics, University of Zagreb, Croatia. Chomsky, N and H. Lasnik. 1977. Filters and Control. Linguistic Inquiry 8(3). 425–504. Comrie, B and T. Kuteva. 2005. Relativization strategies. In M. Haspelmath. M.S. Dryer. D. Gill and B. Comrie (eds.), The World Atlas of Language Structures, 494–501. New York and Oxford: OUP. Engdahl, E. 1985. Parasitic Gaps. Resumptive Pronouns and Subject Extractions. Linguistics 23: 3–44. Franks, S. 1995. Parameters of Slavic Morphosyntax. Oxford: OUP. Golden, M. 1997. Parasitic Gaps in Slovene. Proceedings of the First European Conference on Formal Description of Slavic Languages (University of Leipzig. Germany), 387–398. Frankfurt am MaIn Vervuert Verlag. *RRGOXFN + DQG ' 6WRMDQRYLü 1996. The structure and acquisition of relative clauses in Serbo-Croatian. Language Acquisition 5: 285–315.

Recoverability in Slovene Relative Clauses

185

McCloskey, J. 1990. Resumptive pronouns. A’-binding and levels of representation in Irish. In R. Hendrick (ed.), Syntax of the Modern Celtic Languages. 199–248. New York and San Diego: Academic Press. —. 2005. Resumption. In M. Everaert and H. van Riemskdijk (eds.), The Blackwell Companion to Syntax, 94–117. Oxford: Blackwell. Neeleman, A and F. Weerman. 1999. Flexible Syntax. Dordrecht: Kluwer. Sells, P. 1984. Syntax and Semantics of Resumptive Pronouns. PhD diss., University of Massachusetts. de Vries, M. 2002. The Syntax of Relativization. PhD diss., (= LOT Dissertation Series 53.) University of Amsterdam.

CHAPTER ELEVEN GENERIC SUBJECTS AND VOICE IN POLISH IMPERSONAL CONSTRUCTIONS MAàGORZATA KRZEK 11.1 Introduction This paper provides a general overview of the morphosyntactic properties of Polish impersonal 6,ĉ and –NO/–TO constructions. It demonstrates that whereas the –NO/–TO suffix is best analysed as the head of VoiceP (following Fassi Fehri 2009), the particle 6,ĉ cannot be given a similar analysis (contra Fassi Fehri 2009). What is more, Fassi Fehri’s (2009) generalization, according to which, generic pros cannot be associated with the third person active inflection, but only with the third person passive, although true for Arabic, Irish, and Italian, cannot account for the phenomena found in Polish impersonal constructions. This paper is organized as follows: section 11.2 recapitulates general features of consistent null-subject languages of which Polish is generally said to be one, with special reference to impersonal constructions; section 11.3 presents two types of impersonal constructions found in Polish with a concise description of their morphosyntactic properties; section 11.4 provides evidence that both constructions can be interpreted generically; section 11.5 discusses some properties of canonical passive in Polish and passive in two impersonal constructions in question; section 11.6 reveals that the 6,ĉ particle and the –NO/–TO suffix cannot be analysed in a uniform way. Following Fassi-Fehri (2009), it is argued that the –NO/–TO suffix is base-generated as the head of a functional projection VoiceP. This projection is situated between TP and vP (Kratzer 1996; Collins 2005; and Sigurdsson 2011, to mention just a few), and is the locus of the impersonal active voice. Evidence of the cooccurrence of 6,ĉ and passivization is argued to indicate that a similar analysis is not available for the 6,ĉ particle, which can be analysed as either a functional head or an argument.

Polish Impersonal Constructions

187

11.2 Polish as a consistent null subject language Polish is considered a consistent null subject language just like Italian and Spanish (Sigurðsson & Egerland 2009) and as such displays an interesting array of properties. The ones that are crucial for this study are listed below: I. Missing subjects (Chomsky 1982). a

Jesz GXĪR malin. eat.2SG.PRES a lot of raspberries.GEN “[You] eat a lot of raspberries.”

b

ĪH poszli. went.3PL “[He] did not know that [they] went.”

Nie ZLHG]LDá

NEG knew.3SG.MASC that

Being easily recoverable from the morphological marking on the verb, subject pronouns in both main and embedded clauses are always null unless they receive contrastive focus or there is a topic shift1. II. Rich agreement morphology. To give an example, the past tense paradigm of the Polish verb F]\WDü ‘to read’ looks as follows:

Singular

1st 2nd 3rd

Feminine czytDáDP czytDáDĞ czytDáD

Masculine Neuter czytDáHP czytDáHĞ czytDá czytDáR

Plural

1st 2nd 3rd

czytDá\ĞP\ czytDOLĞFLH czytDá\ĞFLH czytDOLĞFLH czytDá\ czytali

1

Note, however, sentences such as (i) below where the subject of the embedded clause on ‘he’ is obligatorily overt. i

MDU\VLDPyZLĪHàXNDV] nie mówi po angielsku, ale on mówi. Marysia says that Lukasz NEG speaks for English but he speaks. “Marysia says that Lukasz does not speak English but he does.”

See Holmberg (2005) for a detailed discussion of this phenomenon, which is found in a number of other null-subject languages.

Chapter Eleven

188

Polish verbal morphology encodes not only person and number but also gender in the past tense and the past participle of the future tense. Gender distinction does not occur in present tense. III. Free inversion in simple clauses a

']ZRQLáD Marysia. called.3SG. FEM Marysia.NOM “Marysia called.”

Polish does not make great use of word order to indicate the theta roles of arguments and grammatical relations. Instead, it uses case endings on nouns, and prepositions in combination with case endings on nouns. Differences in word order are used to express not theta roles, but aspects of information structure, such as emphasis, something which in English is usually done by means of intonation (Swan 2002). IV. Lack of null generic 3rd person subject. There is another property that has been argued to be shared by null subject languages, but Polish does not have it, thus casting some doubt on its generality. The property concerns the possibility of having a null subject with generic interpretation. According to Holmberg (2005: 555-556), consistent null subject languages have a D-feature2 in I(inflection) which means that when a null ĭ3“phi-phrase”; deficient pronoun) enters into an Agree relation with I, it can be interpreted as definite, referring to an individual or a group. But it also means that a null subject cannot have a generic interpretation, that is, it cannot refer to people in general. Therefore, in order to express generic meaning, consistent null subject languages have to resort to a variety of “overt strategies”. Thus, they may express it with an overt pronoun of si/se-type. The pronoun si, which is lexicalized as 6,ĉ in Polish, is considered the subject (Holmberg 2005). More recently, Fassi Fehri (2009) observed, however, that that third person null generic pronouns are in fact present in consistent null subjects languages but only in Passive Voice.

2 For Holmberg (2005), D is a label for the feature that distinguishes arguments inherently capable of reference from referentially deficient arguments.

Polish Impersonal Constructions

189

11.3 Two types of impersonal constructions Polish has a large number of constructions that can be regarded as impersonal. This paper is, however, concerned with only two types of impersonal constructions, namely the reflexive (6,ĉ) construction and the –NO/–TO construction. Both of these have distinct morphological marking on the main verb. Despite the fact that the morphological marking is different for the two constructions, they nevertheless share a number of morphosyntactic properties.

11.3.1 The 6,ĉconstruction This structure is a cognate of the so-called si/se construction in Italian and Spanish. It consists of a verb and a 6,ĉ SDUWLFOH. The verb is in the default 3SG NEUTER form, as in (1): 1 a

7DĔF]\ VLĊ tutaj F]ĊVWR dance.3SG.NEUT 6,ĉ here often “[One] dance here often.”

b

Buduje VLĊ WXWDMV]NRáĊ build.3SG.NEUT 6,ĉ here school.ACC “[One] is/are building/builds/build/ a/the school here.”

c

-DGáRVLĊ GXĪR malin. ate.3SG.NEUT 6,ĉ a lot of raspberries.GEN “[One] ate a lot of raspberries.”

d

Da VLĊ wszystkim prezenty. give.FUT.3SG 6,ĉ everybody.DAT gifts.ACC “[One] will give everybody gifts.”

The sentences in (1) show that reflexive impersonals can be used with intransitive (1a), monotransitive (1b)/(1c), and ditransitive(1d) verbs with distinct forms for each tense (Kibort 2008). In (1a) and (1b) the impersonal verb refers to the present, in (1c) to the past, and in (1d) to the future. When an impersonal reflexive refers both to the present and the past, the verb form is marked 3SG.NEUT. It is also clear that accusative objects are retained, as in (1b), (1c), and (1d).

Chapter Eleven

190

As noted by Kibort (2008), reflexive impersonals can also be formed from both unergative (2) and unaccusative predicates (3). .

2 3UDFRZDáR VLĊ FLĊĪNR worked.3SG.NEUT 6,ĉ hard. “[One] worked hard.” 3 W tym domu umiera VLĊ spokojnie. In this house dies.3SG.NEUT 6,ĉ peacefully. “In this house [one] dies peacefully.”

11.3.2 The –NO/–TO construction The –NO/–TO construction uses otherwise an uninflected verb form with a –NO/–TO suffix as exemplified in (4) and (5): 4 .XSRZDQRGXĪR kwiatów. bought.IMPERS a lot of flowers.GEN “[One] bought a lot of flowers.” 5 7DĔF]RQRGRELDáHJRUDQD danced.IMPERS to white morning “[One] danced until dawn.” The construction can be formed from both transitive (4) and intransitive (5) verbs. Just like in the case of reflexive impersonals, transitive verbs in the –NO/–TO construction also retain the accusative object. However, unlike reflexive impersonals, the –NO/–TO construction can only refer to the past tense. Another property of this construction, as noted by Kibort (2008), is that it can be formed from both unaccusative (7) and unergative predicates (6) including the iterative form of the verb ‘be’ (8), as is illustrated by the following examples: 6 Pracowano FLĊĪNR worked.IMPERS hard. “[One] worked hard.” 7 Umierano W\VLąFDPL died.iterative.IMPERS thousands “[They] died in thousands.”

Polish Impersonal Constructions

191

8 Bywano F]ĊVWRZWHM restauracji. was.iterative.IMPERS often in this restaurant “[One/They] was/used to come to this restaurant often.” The occurrence of unaccusative verbs in –NO/–TO impersonals is, however, restricted to imperfective verbs with an iterative interpretation, such as such as umierano in (8) and similar verbs (e.g. upadano ‘fell.iterative.IMPERS’, dojrzewano ‘ripened.iterative.IMPERS’). A perfective verb form, denoting a singular action, e.g. *umarto ‘died.IMPERS’, *XSDGQLĊWR ‘fell.IMPERS’, *dojrzano ‘ripended.IMPERS’ is ungrammatical. A similar observation is made by Lavine (2005: 7), who claims that this constraint is most likely related to the availability of a generic interpretation for the subject of the –NO/–TO impersonal. There are, however, some noniterative unaccusative predicates that can are compatible with the –NO/– TO impersonal, such as QDWNQLĊWRVLĊ3 ‘came-across.IMPERS.’ (Kibort 2004: 256), ]DGUĪDQR ‘trembled.IMPERS.’ (Kibort 2004: 256), and osiwiano ‘grew grey.IMPERS’.

11.3.3 Subjects in impersonal constructions The following subsections provide evidence that in Polish impersonals active syntactic subjects are projected. In order to substantiate the presence of subjects, a number of tests will be employed (e.g. availability of agent oriented adverbials, availability of PRO in the embedded clause, availability of a reflexive or other anaphor, availability of the agentive phrase). 11.3.3.1 Availability of agent oriented adverbials The availability of agent oriented adverbials, such as celowo ‘on purpose’, dobrowolnie ‘voluntarily’, confirms the presence of an agent (or experiencer/undergoer4). Agent oriented adverbials are allowed in both impersonal constructions, as the following examples demonstrate:

3 The verb QDWNQLĊWRVLĊis inherently reflexive, and the reflexive particle VLĊ is part of its meaning. 4 As demonstrated in section 3.2.1, the 6,ĉand the –NO/–TO impersonals can be formed from unaccusative verbs. The subjects of these verbs, as noticed by Kibort (2004), are not agents but experiencers/undergoers, thus agent-oriented adverbials will not be felicitous in their environment. This fact, however, cannot be taken as an argument that the syntactic subject is not projected in these clauses.

Chapter Eleven

192

11 a

b

%XGRZDQRWXWDMDXWRVWUDGĊ celowo. built.IMPERS here motorway.ACC on purpose. “[One/They] built a motorway here on purpose.” -DGáR VLĊ celowo maliny. ate.3SG.NEUT 6,ĉ on purpose raspberries.ACC “[One/They/We] ate raspberries on purpose.” 11.3.3.2 Control and binding

Participation in control and binding relations constitutes another argument supporting the presence of underlying subjects. Bondaruk & &KDU]\ĔVND-Wójcik (2003) observe that the 6,ĉ and the –NO/–TO impersonals can share their subjects with embedded infinitive clauses (12), with present and past participle forms (13), and in subject-raising constructions (14). 12 a

Próbowano ]UR]XPLHü ten problem. tried.IMPERS understand.INF this.ACC problem.ACC “[One/They] tried to understand this problem.”

b

&KFLDáR VLĊ RGNU\ü SUDZGĊ wanted.3SG.NEUT 6,ĉ discover.INF truth.ACC “[One/They/We] wanted to discover the truth.”

13 a

b

14 a

b

3U]\JRWRZXMąF przedstawienie, zapomniano o kostiumach. prepare.PART.present performance.ACC forgot.IMPERS about costumes “In preparing the performance, [one/they]forgot about costumes.” :\FKRG]ąF z GRPX]DP\NDáR VLĊ drzwi na klucz. leave.PARTpresent from house closed.3SG.NEUT 6,ĉ door.ACC on key “On leaving the house, [one/they/we] locked the door.” =DF]ĊWR Z\JOąGDü EDUG]RSRZDĪQLH serious.ADV began.IMPERS look.INF very “[One/They] started to look very serious.” Zdawano VLĊ tego nie GRVWU]HJDü seemed.3SG.NEUT 6,ĉ this.GEN NEG notice.INF “[They/One/We] seemed not to notice this.” (adapted from Kibort 2004 :272)

As regards binding, Kibort (2004) observes that the covert subject of the 6,ĉ and the –NO/–TO impersonals is also capable of binding reflexive

Polish Impersonal Constructions

193

and reflexive-possessive pronouns that need to be bound by the subject, as illustrated by examples in (15) and (16) respectively. 15 a

b

16 a

b

2JOąGDQR VLHELHVLĊ5 w lustrze. looked.IMPERS self/ 6,ĉ in mirror “[One/They] looked at oneself/themselves in the mirror.” (adapted from Kibort 2004:273) .XSRZDáR VLĊ sobie prezenty. bought.3SG.NEUT 6,ĉ self.DAT gifts.ACC “[One/They/We] bought oneself/themseves/us gifts.” Czytano swoje NVLąĪNL read.IMPERS own.REFL.ACC books.ACC “[One/They] read one’s/their books.” 6áXFKDáR VLĊ swojej muzyki. listened.3SG.NEUT 6,ĉ own.REFL.GEN music.GEN “[One/They/We] listened to one’s/their/our music.” 11.3.3.3 Nominative subjects and oblique phrases

Additional evidence supporting the claim that Polish impersonals do have covert subjects comes from the observation that they do not accept surface expression of a nominative subject, as in (17). They also do not accept expression of an agent in an oblique phrase, as demonstrated (18): 17 a

b

18 a

5

*Maria czytano ciekawe NVLąĪNL Maria.NOM read.IMPERS interesting books.ACC (intended) “Maria read interesting books.” *Maria je VLĊ GXĪR malin. Maria.NOM eat.3SG 6,ĉ a lot of raspberries.GEN (intended) “Maria eats a lot of raspberries.” */?Kupowano samochód przez ojca. by father bought.IMPERS car.ACC (intended) “A car was bought by my/his/her father.”

In Polish the reflexive pronoun siebie ‘selfACC’ is interchangeable, in restricted contexts, with a multifunctional enclitic form VLĊ (Nagórko 1998, Kibort 2004).

Chapter Eleven

194

b

*/?Je VLĊGXĪR malin SU]H]0DU\VLĊ eat.3SG.NEUT 6,ĉ a lot of raspberries by Marysia (intended) “A lot of raspberries are eaten by Marysia.”

As the examples in (18) demonstrate, the status of impersonal sentences containing oblique phrases is somewhat unclear. The fact that these examples are to some extent acceptable is not because the agent oblique phrases are allowed. These are excluded from Polish impersonal sentences, as the agentive meaning they are intended to express is incompatible with impersonal predicates. The reason why the sentences in (18) are more or less acceptable must be that the NPs in the prepositional phrases (PPs), although superficially similar to agent oblique phrases, are not in fDFWWUXHDJHQWV%RQGDUXN &KDU]\ĔVND-Wójcik 2003). They are not instigators of the action denoted by the impersonal verbs; they only participate in them and may express either the cause of negative circumstances, or may be regarded as a kind of instrument. The instrumental and causative roles of the PPs in (18a) and (18b) respectively become clear when impersonal sentences are compared with their personal equivalents in (19). 19 a

b

.XSLáDPVDPRFKyGSU]H]RMFD bought.1SG car.ACC by father “I bought a car through my/his/her father.” Nie PRJĊWXWDMSDOLü przez Piotra. can.1SG here smoke.INF by Piotr “I can’t smoke here because of Peter.”

NEG

In (19), the role of agent is assigned to the covert subject pro, whose agreement features are visible as inflectional marking on the main verbs. Therefore, the PPs must have a different role, namely that of an instrument (19a), or causer (19b).

11.4 Interpretation of the impersonal 6,ĉ and –NO/–TO constructions This section presents a very general overview of interpretation of the two impersonal constructions in question. This is to illustrate that in both constructions generic reading is available. It is, however, dependent on a

Polish Impersonal Constructions

195

number of factors such as type of a predicate, aspect and temporal/spatial frames. With unspecific time reference and imperfective aspect both constructions are interpreted generically regardless of the type of predicate. This is illustrated by sentences in (20) and (21)6. 20 a

b

21 a

b

Je VLĊ WXWDMGXĪR owoców. eat.3SG.NEUT 6,ĉ here a lot of fruit “[One] eats here a lot of fruit.”

(Transitive)

1DWąFKRUREĊ umiera VLĊSRZROL(Unaccusative) from this disease.ACC die.3SG.NEUT 6,ĉ slowly “From this disease [one] dies slowly.” Jedzono GXĪR owoców. ate.IMPERS a lot of fruit. “[One] bought a lot of fruit.” Umierano powoli. died.iterative..IMPERS slowly “[One] was dying slowly.”

(Transitive)

(Unaccusative)

With the specific time reference and imperfective aspect, the –NO/–TO construction is still interpreted as generic (as illustrated by examples in 22), whereas the 6,ĉ construction clearly acquires new readings. It can be interpreted as referring to a group of people (arbitrary reading) that may or may not include the speaker (specific reading). The generic reading is not that clear, but may be, nevertheless, still available with transitive and unergative predicates. With an unaccusative predicates, the generic interpretation is highly improbable. It is exemplified by sentences in (23). 22 a

b

6

:;,;ZLHNXMHG]RQRGXĪRRZRFyZ in XIX century ate.IMPERS a lot of fruit. “[One] ate a lot of fruit in the XIX century.” W XIX wieku umierano powoli. in XIX century died.iterative.IMPERS slowly “[One] was dying slowly in the XIX century.”

(Transitive)

(Unaccusative)

Due to the space limitation, sentences with unergative predicates have been omitted. For a more detailed analysis of the interpretation of impersonal constructions in Polish, see Krzek (2010).

Chapter Eleven

196

23 a

-DGáR VLĊ wczoraj GXĪR owoców. (Transitive) ate.3SG.NEUT 6,ĉ yesterday a lot of fruit “[They/Some people including the speaker/?One] ate a lot of fruit yesterday.”

b

3UDFRZDáR VLĊ ZF]RUDMFLĊĪNR (Unergative) hard worked.3SG.NEUT 6,ĉ here “[They/Some people including the speaker/ ?One] worked hard yesterday.”

c

Wczoraj ]DGUĪDáR VLĊ ze strachu nie raz. (Unaccusative) yesterday trembled.3SG.NEUT 6,ĉ from fear not once “Yesterday [they/some people including the speaker/ ??one] trembled from fear more than once.”

Examples in (20)-(23) demonstrate that in certain contexts the subjects of both constructions can be interpreted as generic.

11.5 Voice Phrase and generic subjects In this section, I attempt to see whether the analysis of impersonal constructions as expounded by Fassi Fehri (2009) can be extended to account for both the interpretation of the two constructions in question and the function of the 6,ĉ particle and the –NO/–TO suffix, and as a consequence whether Holmberg’s generalization (2005) can be maintained. It will be argued that the 6,ĉ particle and the –NO/–TO suffix cannot be given a uniform analysis, as the two elements perform different functions within a clause. The –NO/–TO suffix, but not the 6,ĉ particle, is best analysed as the head of the functional projection (VoiceP) located between vP and TP. Crucially, it will be demonstrated that the –NO/–TO suffix heads the active impersonal Voice Phrase, not a passive impersonal Voice Phrase, as argued by Fassi Fehri (2009) for Arabic, Italian and Irish. What follows is that third person generic pro is also available in active sentences in consistent null subject languages (CNLS) of which Polish is an example, not only in passive ones as maintained by Fassi Fehri (2009). This observation, in turn, provides counterevidence to Holmberg’s generalization according to which, in consistent null subject languages, null subjects are interpreted exclusively as referential, whereas a third person pro is exclusively generic in partial null subject languages (e.g. Finnish).

Polish Impersonal Constructions

197

11.5.1 Theoretical assumptions Despite the fact that there is no agreement as to what constraints on Voice and Voice-like operations exist, where it fits in the overall organization of the grammar, or even precisely how voice is to be defined, there is an agreement that voice is a universal linguistic phenomenon (Babby 1998). In traditional definitions, voice is primarily concerned with the relation between the initial and derived subjects of the sentence that is active vs. passive and active vs. middle. For Babby (1998), Voice is a subset of a base verb’s derived diatheses, and it should be understood as an alternation in the lexical representation of an argument structure of a given verb. Within the generative framework, Passive Voice has been extensively studied (Jaeggli 1986, Baker at al. 1989, Collins 2005, etc.). Other types of Voice and the category of Voice have not been as widely investigated. There is, however, a growing consensus that Voice is a functional head located between v and T, either morphologically marked or unmarked (e.g. Kratzer, 1996, Sigurðsson 2011, Fassi Fehri 2009). According to Sigurðsson (2011), any predicate is embedded under some Voice head, and a Voice head commonly alters the argument structure of basic (i.e. lexical, non-derived) predicates. Voice is also a cover term, much as Aspect in Cinque (1999), that is there are number of mutually exclusive Voice-type heads, including passive and active Voice, VoicePASS and VoiceACT (Sigurðsson 2011). This is, in essence, the view that I adopt here, pointing out, however, that in the case of Polish impersonal –NO/– TO construction, Voice head does not affect the argument structure of a predicate, that is, arguments are not added, incorporated/suppressed, expletivized or demoted. This fact is based on the observation that in this construction verbs behave in the same way as in sentences in active personal Voice, that is they assign accusative case to their objects, oblique phrases are not allowed, and agent oriented adverbials are possible with verbs that assign agent theta roles to their external arguments. Most importantly, however, the presence of the –NO/–TO suffix excludes passivization. The reason being that if it is assumed that there can only be one Voice head per phrase, then the fact that passive auxiliaries cannot be merged suggests that the position reserved for this type of auxiliaries is filled in by some other element. It is argued that this element is the –NO/– TO suffix (see section 11.5.4). The same analysis, however, is not available for the particle 6,ĉ, as in this case passivization although restricted is nevertheless possible with the 6,ĉ construction (see section 11.5.3).

Chapter Eleven

198

11.5.2 Passive in Polish Passive in Polish is formed from an auxiliary E\ü ‘be’ or ]RVWDü ‘become’ plus a past participle, which can be perfective or imperfective. Combinations of different forms of passive auxiliaries and past participles yield either actional or stative passives, as follows: the auxiliary byü ‘to be’ with an imperfective past participle is used to express an ongoing or repetitive action, as illustrated by example in (24). 24 .VLąĪND jest czytana SU]H]0DU\VLĊ Marysia book.NOM is read.IMPERF by “This book is (being) read by Marysia.” E\üplus perfective past participle expresses stative passive7, as in (25). 7

The analysis of stative passives as passives is somewhat controversial because of the status of perfective past participle that exhibits a number of properties typical of adjectives. This is confirmed by the following observations (Anagnostopoulou 2003): ‡DGMHFWLYDOSDUWLFLSOHVPD\DSSHDULQDSUHQRPLQDODGMHFWLYDOSRVLWLRQ i

QDJURG]RQ\RVNDUĪRQ\V]F]ĊĞOLZ\DXWRU awarded/ accused/ happy author

‡ RQO\ DGMHFWLYDO SDUWLFLSOHV PD\ DSSHDU DV FRPSOHPHQWV RI  Z\JOąGDü QD (look), SR]RVWDZDü(remain), ]DFKRZ\ZDüVLĊ(act, behave), Z\GDZDüVLĊ(seem) ii

a

2QSR]RVWDZDáV]F]ĊĞOLZ\PRVNDUĪRQ\P]PDUWZLRQ\P he remained happy/ accused/ worried

b

2QZ\GDZDáVLĊV]F]ĊĞOLZ\P]PDUWZLRQ\PGRFHQLDQ\P he seemed REFL happy/ worried/ appreciated

‡ RQO\ DGMHFWLYDO SDUWLFLSOHV SHUPLW SUHIL[DWLRQ RI D QHJDWLYH SUHIL[ nie- (un-), as illustrated by (iii): iii a

b

2QE\áQLHGRFHQLDQ\SU]H]SXEOLF]QRĞü he was unappreciated by public “He was unappreciated by the public.” 2QE\áQLHOXELDQ\SU]H]SXEOLF]QRĞü he was not-liked by public “He was not liked by the public.”

Polish Impersonal Constructions

199

25 Marysia E\áD RVNDUĪRQDRNUDG]LHĪSU]H]3DZáD by Pawel Marysia.NOM AUX.3SG.FEM accused of theft “0DU\VLDZDVDFFXVHGRIWKHIWE\3DZHá” the auxiliary zostaü ‘to become’ and a perfective past participle yield a resultant state, as in (26). 26 Ta ĞFLDQD ]RVWDáD SRPDORZDQDSU]H]0DU\VLĊ Marysia This.NOM wall.FEM.NOM AUX.3SG.FEM painted.PERF by “This wall was painted by Marysia.” the auxiliary zostawaü ‘to become.imperf’ plus an imperfective past participle expresses a process, as in (27). 27 Ten projekt zostaje F]ĊĞFLRZRUHDOL]RZDQ\8. This.NOM project.NOM AUX.3SG partially realized.IMPERF “This project is getting partially realized.” As discussed in detail in sections 11.5.3 and 11.5.4, although the passivization in the 6,ĉ construction is fairly constrained, it is still c

‡

Ta wyspa jest niezamieszkana. this island is uninhabited “This island is uninhabited.”

Verbs and verbal participles do not permit nie-prefixation, as shown in (iv)

iv a

*Ludzie go niedoceniali. people him unappreciated “(intended) People did not appreciated him.”

b

*Ludzie nielubili go. people not-liked him “(intended) People did not like him.”

c

*Ludzie niezamieszkiwali tej wyspy people uninhabited this island “(intended) People did not inhabit this island.”

The observations above demonstrate that stative passives are subject to dual analysis. Namely, they can be analysed as examples of canonical passives or as examples of a copula plus a predicative adjectival complement. 8 This sentence is accepted only by some of my informants. It does not, however, has a bearing on the analysis presented.

Chapter Eleven

200

possible with E\ü‘to be’ and ]RVWDü‘to become’ auxiliaries and with both perfective and imperfective past participles. This is, however, not the case for the –NO/–TO construction.

11.5.3 Passivization in the 6,ĉ construction The 6,ĉ impersonals can be formed from a passivized predicates, as shown by examples in (28). 28 a

%\áR VLĊOXELDQ\PRVNDU]RQ\P AUX.3SG.NEUT 6,ĉ liked/accused “[One] was liked/accused.”

b

=RVWDáR VLĊSRELW\P AUX.3SG.NEUT 6,ĉ beaten.PERF “[One] was beaten.”

c

=RVWDZDáR VLĊ powolLZGUDĪDQ\Pw ELHĪąFH sprawy. AUX.iterative.3SG.NEUT 6,ĉ slowly introduced in current affairs “[One] was slowly introduced in the current state of affairs.”

Examples in (28) demonstrate that passivized 6,ĉ impersonals can be formed with both passive auxiliaries and with perfective and imperfective past participles. This process is not, however, entirely free and there are a number of constraints. These are illustrated by examples in (29): 29 a

%\áR VLĊ kupionym. AUX.3SG.NEUT 6,ĉ bought “?[One/They] was/were bought.”

b

%\áR VLĊF]HNDQ\P AUX.3SG.NEUT 6,ĉ waited “?[One/They] was/were waited for.”

c

%\áR VLĊ RVNDUĪRQ\Po NUDG]LHĪ of theft AUX.3SG.NEUT 6,ĉ accused “[One/They/We/I] was/were accused of theft.”

The analysis of examples in (29) reveals that 6,ĉ impersonals are only possible with those passivized predicates whose active transitive counterpart are not prepositional verbs (29b) and which take [+human]

Polish Impersonal Constructions

201

direct objects (compare (29a) with (29c)). The second observation follows naturally from the assumption that subjects of 6,ĉ impersonals must be human. One of the characteristic features of passives is that object is promoted to subject position, and at the same time subject is demoted and can be expressed by means of an oblique phrase. Examples in (30), however, illustrated that this is not the case here. The object of the 6,ĉ impersonal cannot be not promoted to the subject position, and the subject cannot be demoted. This seems to suggest that passivized impersonal 6,ĉ constructions might not be derived from its active counterpart. 30 a

1DJUDG]DáR GRFHQLDáRVLĊFLHELH awarded.3SG.NEUT/ appreciated.3SG.NEUT 6,ĉ you. “[One/They/We/I] awarded/appreciated you.”

b

*Ciebie/ty E\áR VLĊQDJUDG]DQ\PGRFHQLDQ\P you.GEN/you.NOM AUX.3SG.NEUT 6,ĉ awarded/appreciated “You were awarded.”

c

%\áR VLĊQDJUDG]DQ\PGRFHQLDQ\P]PDUWZLRQ\P AUX.3SG.NEUT 6,ĉ awarded/ appreciated/ worried “[One/They/We/I] was/were awarded/appreciated/worried.”

What is more, this kind of transformation, namely from the active impersonal VLĊ(30a) to the passive impersonal 6,ĉ (30c), does not satisfy the semantic transitivity which is a prerequisite for a construction to be considered passive. In other words, there is no transfer from an Agent/Actor in (30a) to a Patient/Undergoer in (30c); the Agent/Actor in (30a) and the Patient/Undergoer remain the same. All of the constraints on the formation of passive stem from the constraints imposed on the subject of impersonal 6,ĉ construction such that it cannot be specified for person and that it has to be [+human/animate], and that is why they cannot be treated as an argument that in 6,ĉ construction passivization is excluded.

11.5.4 Passivization in –NO/–TO construction Contrary to 6,ĉ impersonals, –NO/–TO impersonals cannot be formed from passivized predicates. This is illustrated by examples in (31). 31 a

(*ZoVWDáR  siano SV]HQLFĊQDZLRVQĊ in spring AUX.3SG.NEUT sowed.IMPERS wheat.ACC “Wheat will be sowed in spring.” (adapted from Dziwirek 1994)

Chapter Eleven

202

b

 %\áR ]QDOH]LRQRQLHPRZOĊ w koszu (*przez lekarzy). AUX.3SG.NEUT found.IMPERS baby.ACC in basket by doctors “A baby was found in a basket by doctors.” (Lavine 2005)

As sentences in (31) illustrate, the passivization of the –NO/–TO construction yields ungrammatical results, regardless of the type of an auxiliary (both are excluded) and a past participle (both perfective and imperfective are excluded) used. As shown in section 11.5.2 both passive auxiliaries and oblique phrases are available in Polish canonical passive so the fact that they are impossible in the –NO/–TO construction is not a fact about passive in Polish in general but rather specific to the –NO/–TO construction. The fact that passivization is excluded is additionally confirmed when Polish –NO/–TO construction is compared with the cognate construction in Ukrainian. As illustrated by the examples in (32), the Ukrainian –NO/– TO construction allows overt tense marking auxiliaries and by-phrases, both of which are excluded in Polish –NO/–TO construction. 32 a

Nemovlja bulo znajdeno u košyku likarjami. (Ukrainian) baby.ACC AUX.pass found.IMPERS in basket doctors “A baby was found in a basket by doctors.” (Lavine 2005)

b

Inozjemcja bude SRVDGåHQR do v’jaznyci (Ukrainian) Foreigner.ACC AUX.fut placed.IMPERS to prison hlavoju urjadu. head.INSTR government “A foreigner will be put in prison by the head of government.” (Lavine 2005)

These data confirm that it is the –NO/–TO suffix in Polish that is responsible for the unavailability of passive in the –NO/–TO construction9; this gives rise to the analysis of the –NO/–TO suffix as the head of VoiceP. 9

Kibort (2008) claims that they can be formed from the iterative form of the verb ‘to be’ as illustrated in (i). The following example contains an impersonal form of the verb ‘to be’ (bywano) with an imperfective passive participle (beaten): i

?*Bywano bitymi. was.iterative.IMPERS beaten.PL “[One] was beaten.” (adapted from Kibort 2008)

Polish Impersonal Constructions

203

11.5.5 Generic subjects and VoiceP We are now in a position to discuss the function of the –NO/–TO suffix and the 6,ĉ particle and to demonstrate how the analysis suggested for the aforementioned elements relates the hypothesis put forward by Fassi Fehri (2009), according to which in consistent null subject languages generic subjects are only available in Passive Voice. Having assumed that different types of Voice heads are mutually exclusive, and that there can only be one Voice head in a clause, the fact that the –NO/–TO suffix blocks passivization suggests that it is merged in the head of VoiceP. However, the –NO/–TO suffix is not a passive morpheme and does not introduce a passive voice head as confirmed by the observations that both oblique phrases and overt expressions of nominative subject are excluded. In other words, the argument structure of a predicate in the –NO/–TO construction remains intact. Thus, the construction is still active. It is assumed here what the –NO/–TO suffix does is to introduce an impersonal voice. This analysis coupled with the fact that generically interpreted subjects are readily available in this construction shows that generic subjects are available not only with third Kibort (2008) seems to overstate the ability of the –NO/–TO construction to passivize as all of my informants considered this sentence ungrammatical. However, the acceptability of sentences containing the impersonal iterative form of the verb ‘to be’ rises somewhat when they are coupled with perfective passive participles. This is illustrated by the sentence in (ii). ii

?Bywano RVNDUĪRQ\PL was.iterative.IMPERS accused.PL “[One] was accused.”

The examples containing the iterative form of the verb ‘to be’ with the -NO/-TO suffix become fully acceptable when followed by an adjectival/nominal/PP complement, as illustrated by the following example. iii %\ZDQRV]F]ĊĞOLZ\PLZW\PGRPXU]DGNR was.iterative.IMPERS happy.PL/ in this house rarely “[One] was happy/in this house rarely.” The reason behind this upsurge in acceptability seems to be that what the -NO/-TO suffix attaches in the sentences above is not a passive auxiliary but a copula. As such the copula bywano is incompatible with imperfective passive participle, as in (i). Because, as discussed in footnote 6, the constructions such as the one in (ii) are subject to dual interpretation as examples of either stative passive or a copula plus deverbal adjective, they are higher on the acceptability scale.

204

Chapter Eleven

person passive inflection as argued by Fassi Fehri (2009) but also when the inflection is non-agreeing. Therefore, it appears that Fassi Fehri’s (2009) generalization cannot be extended to account for the derivation and interpretation of the –NO/–TO construction in Polish. As for the 6,ĉ construction, it is argued that the 6,ĉ particle cannot reside in the head of VoiceP, as it does not block the merge of a passive auxiliary. In other words, the 6,ĉ construction, just like the –NO/–TO construction, is active as both oblique phrases and overt expressions of nominative subject are excluded, but contrary to the –NO/–TO suffix, the SIĉ particle does not block passivization. As the generic interpretation is readily available, it is evident, contra Fassi Fehri (2009), that generic subjects can be associated with third person active inflection, not only with the passive one. As regards the possible functions of the 6,ĉ particle, it may be that of either an argument or a functional projection. However, much more work needs to be done before any such conclusion is made.

11.6 Conclusion This paper provides a general overview of morphosyntactic properties of two Polish impersonal constructions: the –NO/–TO construction and the 6,ĉ construction. It has been demonstrated that the –NO/–TO suffix should be best analysed as the head of VoiceP. A similar analysis is not available for the 6,ĉ particle. It has also been shown that generic pros are found in active Voice in consistent null subject languages, not only in passive Voice, as argued by Fassi Fehri (2009). Consequently, it appears that Holmberg’s (2005) generalization, according to which generic subjects are not found in consistent null subject languages, will need to be revisited.

References Anagnostopoulou, E. 2003. Participles and voice. In Alexiadou, A., M. Rathert and A. von Stechow (eds.), Perfect Explorations, 1–36. Berlin/New York: Mouton de Gruyter. Babby, L. 1998. Voice and diathesis in Slavic. Ms. Retrieved from citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.73.6619 on 22/12/ 2010. Baker, M., K. Johnson and I. Roberts. Passive arguments raised. Linguistic Inquiry 20:219–251

Polish Impersonal Constructions

205

%RQGDUXN $ DQG 0 &KDU]\ĔVND-Wójcik. 2003. Expletive pro in Impersonals Passives in Irish. Polish and Old English. Linguistische Berichte 195: 325–362. Chomsky, N. 1982. Concepts and consequences of the theory of government and binding. Cambridge, MA: MIT Press. Cinque, G. 1988. On si constructions and the theory of Arb. Linguistic Inquiry 19(4). 521–581. —. 1999. Adverbs and functional heads. A cross-linguistic perspective. Oxford: Oxford University Press. Collins, C. 2005. A smuggling approach to the Passive in English. Syntax: 8 (2). 81– 120. Dziwirek, K. 1994. Polish Subjects. New York: Gardland. Fassi Fehri, A. 2009. Arabic silent pronouns. person and voice. Brill’s Annual of Afroasiatic Languages and Linguistics 1: 1–38. Holmberg, A. 2005. Is there a little pro? Evidence from Finish. Linguistic Inquiry 36: 533–564. Jaeggli, O. 1986. Arbitrary plural pronominals. Natural Language and Linguistic Theory 4: 43–76. Kibort, A. 2004. Passive and passive-like constructions in English and Polish. Ms. University of Cambridge. —. 2008. Impersonals in Polish: an LFG perspective. Transactions of the Philological Society 106 (2). 246–289. Kratzer, A. 1996. Severing the external argument from its verb. In Rooryck. J. and L. Zaring (eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer Academic. Krzek. M. 2010. Some aspects of subjects of impersonal constructions in Polish. Newcastle Working Papers in Linguistics 16: 66-87. Lavine, J. 2005. The morphosyntax of Polish and Ukrainian -no/-to. Journal of Slavic Linguistics 13(1). 75–117. Nagórko. A. 1998. Zarys Gramatyki Polskiej, Warszawa: PWN Sigurðsson H. Á. 2011. On the new passive, Syntax 14:148–178. . Sigurðsson, H. A. and V. Egerland. 2009. Impersonal null-subjects in Icelandic and elsewhere. Studia Linguistica 63(1). 158–185. Swan, E.O. 2002. A grammar of contemporary Polish. Indiana: Slavica Publishers. Wyner, A.Z. 2008. Towards flexible types with constraints for manner and factive adverbs. In L. Mcnally and Ch. Kennedy (eds.), Adjectives and Adverbs. Oxford: OUP.

CHAPTER TWELVE ON FEATURE INHERITANCE AND CYCLICITY OF DERIVATION MASAKAZU KUNO 12.1 Introduction It is proposed in Chomsky (2008) that in building CP phases, T gets uninterpretable agreeing features and the EPP property from C. This mechanism is called Feature Inheritance. 1 Feature Inheritance (Chomsky 2008): T LQKHULWVDJUHHLQJij-features and the EPP property from C. When Feature Inheritance takes place, the derivation proceeds as follows. 2

CP

C[uij(33@ T Feature Inheritance

3 TP

CP

C vP

DP[iij]

4 TP

T[uij, EPP] Agree

CP

C vP

DP[iij]

TP DP[iij]

T’ T[uij, EPP]

Move

vP

tDP

(2) is a stage of the derivation where Feature Inheritance takes place, by which T inherits the uninterpretable ij-features and the EPP property from C. Now that T has the uninterpretable ij-features and the EPP property, it can act as a probe. Thus, it agrees with the DP within vP, which bears interpretable ij-features, as shown in (3). Then, the DP moves to

On Feature Inheritance and Cyclicity of Derivation

207

Spec-TP in response to the EPP property of T, deleting the uninterpretable ij-features of T, as illustrated in (4). This is a sketch of the derivation where Feature Inheritance takes place. Note that when Feature Inheritance takes place, the derivation proceeds counter-cyclically in part, as stated in (5). 5 Feature Inheritance involves inherent counter-cyclicity in the sense that T starts probing only after C is merged. Note also that movement of the DP into Spec-TP in (4) violates the Extension Condition. This paper discusses the cases where Feature Inheritance takes place as well as those where it is blocked. We will see that to the extent that Feature Inheritance is motivated, counter-cyclic derivation is also motivated. To begin with, we will take up two cases where Feature Inheritance is blocked.

12.2 The Two Cases where Feature Inheritance is Blocked The first case where Feature Inheritance is blocked is the one stated in (6) and sketched in (7). 6 Feature Inheritance does not occur when C directly agrees with the subject DP. 7

CP TP

C[uij] T

vP

Agree DP[iij@ In this derivation, the uninterpretable ij-features of C directly agree with the DP within vP, skipping T. This makes Feature Inheritance unnecessary. Then the question is what happens in such a derivation. A natural speculation is that in such a derivation T will not agree with anything as it lacks any agreeing features of its own. This can be found in the distribution of subject clitic in Fiorentino dialect of Italian. This dialect

Chapter Twelve

208

has two kinds of subject clitics, agreeing and non-agreeing. The agreeing subject clitic is shown in (8). Subject Clitic in Fiorentino dialect of Italian 8 a Mario e parla. Mario SCL speaks b

E parla.

c

*Mario parla

d

*Parla

(Agreeing Subject Clitic)

(Brandi and Cordin 1989: 113) (8a) is a case where a subject clitic appears in the agreeing form. (8b) shows that the subject clitic appears with a null subject. (8c) and (8d) illustrate that the subject clitic cannot be omitted. We take the agreeing subject clitic as a morphological realization of T. Now consider (9), where the non-agreeing subject clitic appears. In these cases, the subject appears in a postverbal position, which we take as the vP-internal subject position. 9 a

Gli ha telefonato delle ragazze (Non-Agreeing Subject Clitic) SCL(3sg) has telephoned some girls “Some girls have telephoned.”

b

Gli è venuto delle ragazze (Non-Agreeing Subject Clitic) SCL(3sg) is come.3ms some girls “Some girls have come.” (Brandi and Cordin 1989: 121-122)

In this derivation, C agrees directly with the vP-internal subject without the mediation of T, as illustrated in (10). 10

[CP C[uI] [TP Gli T é [vP venuto delle ragazze[iI]]]] SCL is3.Sg come3.Sg.Masc some girls3.Pl.Fem Agree

On Feature Inheritance and Cyclicity of Derivation

209

This is the case stated in (6). Since in such a derivation, T does not inherit the uninterpretable ij-features, it does not show subject agreement. We assume the condition in (11). 11 When T does not Agree with any DP, it is realized as a non-agreeing subject clitic gli. The other case in which Feature Inheritance is blocked and thus the subject clitic shows up in the non-agreeing form is the one stated in (12). 12 The non-agreeing subject clitic is predicted to occur when the subject undergoes Wh-movement directly into Spec-CP. This derivation proceeds as follows. 13

CP

14

C[uij, uWH, EPP] TP T Agree

CP

DP[iij, iWH] vP

C’ C[uij, uWH, EPP] TP T gli

DP[iij, iWH] Move

vP tDP

In (13), C bears uninterpretable ij-features, an uninterpretable WHfeature and the EPP property and directly agrees with the subject DP within the vP, which, in turn, bears interpretable ij-features and an interpretable WH-feature. The subject DP then moves into Spec-CP, deleting the uninterpretable ij-features and the uninterpretable WH-feature of C. Given this derivation, it is predicted that when a subject DP is WHmoved into Spec-CP, only non-agreeing subject clitic can occur. This prediction is borne out, as shown by the contrast between (15) and (16). Matrix Subject Question 15 a Quante ragazze gli è venuto con te? how many girls SCL is come with you “How many girls have come with you?”

(Non-Agreeing Subject Clitic)

Chapter Twelve

210

b

Quante ragazze gli ha parlato con te? how many girls SCL has spoken with you “How many girls have spoken to you?”

(Non-Agreeing Subject Clitic)

16 a

*Quante ragazze le sono venute con te? (Agreeing Subject how many girls SCL are come with you Clitic) “How many girls have come with you?”

b

*Quante ragazze le hanno parlato con te? (Agreeing Subject how many girls SCL have spoken with you Clitic) “How many girls have spoken to you?” (Brandi and Cordin 1989: 125)

Note that the unavailability of the agreeing subject clitic in (16) shows that in such a derivation, T lacks agreeing features, which in turn shows that Feature Inheritance does not occur in this derivation. The same contrast can be seen when a WH-subject is extracted from an embedded clause, as shown in (17) and (18). Comp-trace environment 17 a Quante ragazze tu credi [che e’ sia venuto]? (Non-Agreeing how many girls you believe that SCL be come? Subject Clitic) “How many girls do you believe have come?” b

Quante ragazze tu credi [che gli abbia parlato]? (Non-Agreeing how many girls you believe that SCL have spoken. Subject Clitic) “How many girls do you believe have spoken?”

18 a

*Quante ragazze tu credi [che le siano venuto]? (Agreeing how many girls you believe that SCL are come? Subject Clitic) “How many girls do you believe have come?”

b

*Quante ragazze tu credi [che le abbiano parlato]? (Agreeing how many girls you believe that SCL have spoken? Subject “How many girls do you believe have spoken?” Clitic) (Brandi and Cordin 1989: 125)

Just like the contrast between (15) and (16), the subject clitic must appear in the non-agreeing form.

On Feature Inheritance and Cyclicity of Derivation

211

In this section, we have seen two cases where the non-agreeing subject clitic is forced to occur. The two cases can be unified by the present proposal that when C directly agrees with a subject DP within vP, Feature Inheritance is blocked, which can be seen in the failure of T to show ijfeature agreement with a subject DP.

12.3 The Cases Where Feature Inheritance is Motivated Let us now turn to the case where Feature Inheritance is motivated. This case can be seen in a simple WH-object question like the one below. 19 Whati did John buy ti The derivation of this sentence proceeds as follows. 20

CP C[uij uWH, EPP] TP T

vP whati[iij iWH]

v’

v’ John[iij] *Agree (intervention by what) v[Xij, EPP]

VP V buy

twhat

Consider the stage of the derivation where C is merged with TP. C bears uninterpretable ij-features, an uninterpretable WH-feature and the EPP property. Suppose that the uninterpretable ij-features act as a probe and try to agree with the subject DP within vP. This relation cannot be established because of the intervening WH-object raised to Spec-vP, which also bears interpretable ij-features. Therefore, the alternative operation to be executed at this stage is agreement between the uninterpretable WHfeature of C and the interpretable WH-feature of the WH-object, as illustrated in (21).

Chapter Twelve

212

21

CP C[uij uWH, EPP] TP T Agree

vP v’

whati[iij iWH]

v’

John[iij] v[Xij, EPP]

VP V buy

twhat

Then, the WH-object moves into Spec-CP in response to the EPP property of C and deletes the uninterpretable WH-feature thereof, as shown in (22). 22

CP whati[iij iWH]

C’

C[uij uWH, EPP]

TP

T Move

vP v’

ti John[iij]

v[Xij, EPP]

v’ VP V buy

twhat

Now notice that C still contains an uninterpretable ij-feature, which cannot be deleted by the moved WH-object since it has already established an agreement relation with v. Here is the point where Feature Inheritance becomes necessary. The uninterpretable ij-features and the EPP property,

On Feature Inheritance and Cyclicity of Derivation

213

which we assume remains even after it has attracted a phrase to the specifier of its bearer, are inherited onto T, as shown in (23). 23

CP C’

whati[iij iWH]

C[uij uWH, EPP]

TP vP

T[uij EPP] Feature Inheritance

v’

ti

v’

John[iij] v[Xij, EPP]

VP V buy

twhat

Then, the uninterpretable ij-features of T probe and find the DP subject John as a goal. 24

CP whati[iij iWH]

C’

C[uij uWH, EPP]

TP vP

T[uij, EPP] ti Agree

v’ John[iij]

v’ v[Xij, EPP] V buy

VP twhat

Chapter Twelve

214

Finally, the DP subject moves into Spec-TP in response to the EPP property of T, as illustrated in (25). 25

CP whati[iij iWH]

C’

C[uij uWH, EPP]

TP

Johnj[iij]

T’ T[uij, EPP]

vP v’

ti Move

v’

tj

v[Xij, EPP]

VP V buy

twhat

Obviously, Feature Inheritance is involved in run-of-the-mill sorts of sentences, which in the same breath means that derivation proceeds in counter-cyclically in part and violates the Extension Condition. Indeed, Feature Inheritance and the counter-cyclic derivation are both motivated on independent grounds. Consider (26). Raising to Subject across an Experiencer in Italian 26 a *[TP Giannii sembra a Maria [ti essere stanco]] Gianni seems to Mary to be ill “Gianni seems to Mary to be ill.” b

[CP A Mariaj, [TP Gianni sembra tj [ti essere stanco]]] to Maria Gianni seeems to be ill

c

[CP A chij [C’ sembra C [TP Giannii tj [ti essere stanco]]]] to whom seems Gianni to be ill “To whom does Gianni seem to be ill?” (Boeckx 2009)

On Feature Inheritance and Cyclicity of Derivation

215

(26a) shows that raising to subject across an experiencer is impossible in Italian. However, as shown in (26b) and (26c), when the experiencer is WH-moved or topicalized, raising becomes possible. This puzzle can be explained by assuming the following derivation. We focus on (26c) for illustration, but the same analysis can be carried over to (26b). Derivation of (26c) --- verb movement omitted 27 a Agree (C, a chi) [CP C[uijuWH, EPP] [TP T a chi[iWH] [Gianni[iI] essere stanco]]] b

Move (C, a chi) [CP a chij[iWH] C[uijuWH, EPP] [TP T tj [Gianni[iI] essere stanco]]]

c

Feature Inheritance [CP a chij[iWH] C[uWH, EPP] [TP T[uij, EPP] tj [Gianni[iij] essere stanco]]]

D

Agree(T, Gianni) [CP a chij[iWH] C[uWH, EPP] [TP T[uij, EPP] tj [Gianni[iij] essere stanco]]]

e

Move(T, Gianni) [CP a chij[iWH] C[uWH,EPP] [TPGiannii[iij] T[uij,EPP] tj [ti essere stanco]]]

First, the uninterpretable WH-feature of C agrees with the experiencer with an interpretable WH-feature, as in (27a), and then the WHexperiencer moves to Spec-CP, as in (27b). Then, the uninterpretable ijfeatures and the EPP property of C are inherited to T, as in (27c), which acts as a probe and agrees with the DP subject Gianni, as in (27d). Finally, Gianni moves into Spec-TP, as in (27e). At this point of the derivation, the intervening experiencer has already been moved to Spec-CP, hence no intervention effect occurs. Essentially the same account can be extended to raising to subject across experiencer in Icelandic, as shown in (28). Raising to Subject across an Experiencer in Icelandic 28 a Hafði Olafur virst [t vera gáfaður] has Olaf.NOM seemed to be intelligent “Did Olaf seem intelligent?” b

*Hafði Olafur  eim virst [t vera gáfaður] has Olaf.NOM them.DAT seemed to be intelligent “Did it seem to them that Olaf was intelligent?”

Chapter Twelve

216

c

*Hafði Olafur virst  eim [t vera gáfaður] has Olaf.NOM seemed them.DAT to be intelligent

d

Hafði  eim virst has them.DAT seemed

[Olafur vera gáfaður] Olaf.NOM to be intelligent

As shown in (28a), raising to subject is possible if there is no experiencer. However, if there is an experiencer, raising to subject becomes impossible, as shown in (28b) and (28c). The only possible derivation is (28d), where the subject stays within vP. However, there is a way in which raising to subject can be made possible, which is by moving the intervening experiencer to Spec-CP before raising to subject takes place, as in (29). 29 a

b

Hverjum hefur Ólafur virst tD [tO vera gáfaður] Who.DAT has Olaf.NOM seemed to be intelligent “Who has found Olaf intelligent?” strákarnir virst tD [tboys vera gáfaðir] HverjumWh hafa to be intelligent Who.DAT have.PL the boys.NOM seemed “Who has found Olaf intelligent?” (Sigurðsson 1996: 25-6)

This counter-cyclic derivation becomes possible only when the derivation proceeds in the fashion sketched in (27). English also exhibits a similar effect. But this time, the countercyclicity has to do with agreement, rather than movement. Consider (30). Agreement across an Experiencer 30 a John seems to Mary to be the best. b John and Bill/The men seem to Mary to be the best. c There seems to Mary to be a man in the room. d There ?*seem/seems to Mary to be men in the room. e To Mary, there seem/?*seems to be men in the room. f John wonders to whom there seem/?*seems to be men in the room. (Boeckx 2009:151-152) As shown in (30a) and (30b), English allows raising to subject across an experiencer. Thus, at this point, one might think that English does not

On Feature Inheritance and Cyclicity of Derivation

217

provide any significant contrast like those seen in Italian and Icelandic. However, when it comes to agreement, English shows a similar contrast. Compare (30c) and (30d). In (30c), the matrix verb shows 3rd person singular agreement, which seems to be agreement with the associate noun phrase a man. However, in (30d), even if the associate noun phrase is a plural noun like men, the matrix verb still shows 3rd person singular agreement. According to Boeckx (2009), the impossibility of plural agreement between the matrix verb and the associate noun phrase in (30d) is due to the intervening experiencer phrase. He supports this claim by showing that when the intervening experiencer is dislocated by topicalization or WH-movement, agreement with the plural associate noun phrase is back, as shown in (30e) and (30f). This counter-cyclic effect can be accounted for if we assume that sentences like (30e) and (30f) undergo the derivation sketched in (27), namely, the intervening experiencer is first moved to Spec-CP and then Feature Inheritance assigns the remaining uninterpretable ij-features of C to T, which then agrees with the associate noun phrase, with no intervening experiencer phrase in-between. In this section, we first saw that Feature Inheritance is needed in runof-the-mill cases of WH-object movement, which as a consequence makes counter-cyclic derivation possible. Then, we further saw that the proposed derivation has explanatory power to account for otherwise puzzling phenomena in Italian, Icelandic and English. Therefore, to the extent that the arguments given in this section are empirically supported, Feature Inheritance and the counter-cyclic derivation as well as violations of the Extension Condition are motivated.

12.4 Subject-Object Asymmetries Let us now revisit Fiorentino dialect of Italian. The relevant examples are repeated below. Fiorentino dialect of Italian: Revisited 31 a Quante ragazze gli è venuto con te? how many girls SCL is come with you “How many girls have come with you?” b

Quante ragazze gli ha parlato con te? how many girls SCL has spoken with you “How many girls have spoken to you?”

(Non-Agreeing Subject Clitic) (Non-Agreeing Subject Clitic)

Chapter Twelve

218

32 a

b

*Quante ragazze tu credi [che le siano venuto]? (Agreeing how many girls you believe that SCL are come? Subject Clitic) “How many girls do you believe have come?” *Quante ragazze tu credi [che le abbiano parlato]? how many girls you believe that SCL have spoken? “How many girls do you believe have spoken?”

(Agreeing Subject Clitic)

What we learn from this dialect is that when a subject is WH-moved, subject clitic appears in the non-agreeing form. Given the discussion of the previous section, it is predicted that when a WH-object is extracted, the subject clitic has to occur in the agreeing form, since such a derivation requires Feature Inheritance and T would get an agreeing ij-features. Unfortunately, we do not have relevant evidence from Fiorentino dialect of Italian, but a language which provides relevant evidence is Haitian Creole. Consider (33). Haitian Creole 33 a Kilès *(ki) e wè Mari who KI ANT see Mari “Who saw Mari?” b

*Kilès (*ki) Mari te wè who KI Mari ANT see “Who did Mari see?”

Haitian Creole exhibits a clitic-like element ki when a subject is WHmoved. We take this element as a morphological realization of T when it fails to inherit agreeing ij-features from C. By contrast, when an object is WH-moved, T is realized as zero, which we take as an agreeing form of T. The same contrast shows up in cases of extraction from the embedded clause, as shown in (34). 34 a

b

Kilès (*ki) Mari panse *(ki) renmen Jan? who KI Mari think KI like Jan “Who does Mari think likes Jan?” Kilès (*ki) Mari panse (*ki) Jan renmen? who KI Mari think KI Jan like “Who does Mary think that Jan likes?” (Takahashi and Graþanin-Yusek 2008)

On Feature Inheritance and Cyclicity of Derivation

219

To put it more concretely, when the WH-phrase is a subject and agrees directly with C, as in (35), T will not inherit any agreeing ij-features, in which case it is realized as ki, as shown in (36). 35

CP

36 TP

C[uij, uWH, EPP]

CP

DP[iij, iWH]

T

C’

C[uij, uWH, EPP] TP

vP

T

DP[iij, iWH]

Agree

vP

ki tDP

Move

However, when C agrees with a WH-object and moves it to its Spec while T inherits agreeing ij-features from C and agrees with a subject DP, forming Spec-TP, T is realized as zero, as shown in (37). 37

CP DP[iWH,iij]

   



C’

C[uWH,EPP]

TP

DP[iij]  

T’

T[iij,EPP]

vP

This way, Haitian Creole fills in the gap, which was left open if we look at Fiorentino dialect of Italian alone. That is, whether or not T obtains agreeing ij-features via Feature Inheritance depends on whether a moved WH-phrase is a subject or an object. When it is a subject, Feature Inheritance does not take place, with the result that T will not inherit any agreeing ij-features whereas when it is an object, Feature Inheritance takes place and T will obtain agreeing ij-features. Finally, let us examine an extraction asymmetry pointed out by Chomsky (2008).

220

Chapter Twelve

Extraction Asymmetry 38 a Of which car did they find [the driver ____ ] b *Of which car did [the driver ____ ] cause a scandal? c Of which car was [the driver ____ ]i awarded ti a prize? (Chomsky 2008: 147) As is well known, a WH-phrase can be extracted out of object, but not out of subject, as indicated by the contrast between (38a) and (38b). However, Chomsky points out that when a WH-phrase is extracted out of a subject of a passive sentence, it is relatively acceptable. In order to account for this fact, he proposes that WH-extraction takes place from the containing noun phrase when it is in the complement position of a verb and also that in order to avoid a violation of the Extension Condition, the WH-extraction and the NP movement into Spec-TP take place simultaneously. The derivational simultaneity is odd in its own right. Rather we propose that (38c) should be analyzed as involving Feature Inheritance and counter-cyclic derivation, as shown in (39). 39 a [CPC[uWH, uij, EPP] [TPT was awarded [the driver [of which car][iWH]] [iij] a prize]] Agree b [CP[of which car][iWH] C[uWH, uij, EPP] [TPTwas awarded [the driver ti ][iij] a prize]] Move c [CP[of which car][iWH]C[uWH,EPP][TPT[uij,EPP]was awarded [the driver ti][iij]a

prize]]

Feature Inheritance d [CP[of which car][iWH]C[uWH,EPP][TPT[uij,EPP]was awarded [the driver ti][iij]a

prize]]

Agree e [CP[of which car][iWH]C[uWH,EPP][TP[the driver ti]j[iij]T[uij,EPP]was awarded tj a

prize]]

Move In this derivation, first C agrees with the WH-phrase within the to-bemoved NP and attracts it to its Spec, as in (39a) and (39b). Then, Feature

On Feature Inheritance and Cyclicity of Derivation

221

Inheritance takes place, which gives the uninterpretable ij-features and the EPP property to T, as in (39c). Now that T obtains agreeing ij-features, it can search for a matching goal and agree with it. This happens in (39d). Finally, the goal is moved to Spec-TP, as in (39e).

12.5 Conclusion One may think that it is difficult to decide whether our proposal is superior to Chomsky’s (2008) because both proposals have a potential weakness: derivational simultaneity for Chomsky’s approach and countercyclicity for ours. However, notice that both approaches share Feature Inheritance and that counter-cyclicity is a logical consequence of Feature Inheritance. In other words, if one adopts Feature Inheritance, it means that one allows counter-cyclicity. Therefore, if the choice between derivational simultaneity and counter-cyclicity is at issue, one has to go for counter-cyclicity as long as one adopts Feature Inheritance. For this reasoning, our proposal is superior to Chomsky’s as long as Feature Inheritance is adopted.

References Boeckx, C. 2009. Aspects of the Syntax of Agreement. New York: Routledge. Brandi, L and P. Cordin. 1989. Two Italian Dialects and Null Subject Parameter. In O. Jaeggli and K. Safir (eds.), The Null Subject Parameter, 111–142. Dordrecht: Kluwer. Chomsky, N. 2008. On Phases. In R. Frieden. C. Otero and M. L. Zubizarreta (eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud, 133–165. Cambridge, MA: MIT Press. Richards, M. 2007. On Feature Inheritance: An Argument from the Phase Impenetrability Condition. Linguistic Inquiry 38: 563–572. Sigurðsson, H. 1996. Icelandic Finite Verb Agreement. Working Papers in Scandinavian Syntax 57: 1–46. Takahashi, S. and M. Graþanin-Yusek. 2008. Morphosyntax of Movement Dependencies in Haitian Creole. Syntax 11: 223–250.

CHAPTER THIRTEEN INFERENCE AND GRAMMAR: INTERSECTIVITY, SUBSECTIVITY, AND PHASES1 ULRICH REICHARD In the semantic tradition, the difference between intersective, subsective and non-subsective adjectives has often been taken to be a lexical phenomenon. The difference between them can be encoded in “meaning postulates” which are part of the lexical entry of the adjective (Kamp and Partee 1995). Alternatively, intersective and non-intersective adjectives may be treated as two different lexical categories (Siegel 1976). After a short exposition of the semantic perspective in section 13.1, I argue against this “lexicalist” treatment of adjectives by showing that, for most adjectives, both intersective and non-intersective readings are systematically available. This suggests that the phenomenon is grammatical rather than lexical (section 13.2). In section 13.4, I argue that the difference between intersective and subsective occurrences of adjectives follows from interface-conditions, given we assume a phasal architecture of grammar (Chomsky 2001) and adopt the recent thesis that intersective modifiers are clausal whereas non-intersective modifiers are not (Alexiadou, et al. 2007; Cinque 2010). Some of the independent grammatical evidence for this thesis is shortly summarized in section 13.3. To cut a long story short: my overall proposal is that, due to the asymmetric structure of grammar, it is expected that modifiers are interpreted relative to their hosts. Subsectivity is thus the “normal case”. However, if the modifier is clausal, it is also a phase. Phases are taken to be units of semantic interpretation. The interpretation of the modifier is 1

Many thanks to Michelle Sheehan, Alex Drummond, and Wolfram Hinzen for conversations, suggestions, and very useful comments on earlier drafts of this paper; to the participants of the INFL conference for an interesting discussion; and to Ustinov College for a travel grant which allowed me to present this paper.

Inference and Grammar

223

thus already fixed at the time of modification and the modifier can therefore not be relativized to its modifee. An intersective interpretation is the only remaining possibility.

13.1 Different kinds of adjectives: The perspective from formal semantics The question which inferences are valid and which ones are not has probably always been of interest in human scientific reflection, and reflection on that topic can be traced back at least to Plato. The validity of inferences can in many cases be tracked by formal features of natural language expressions. Thus, the aim of Aristotle’s Topic was to “develop a method which allows us to deduce what follows from any accepted premises in respect to any argument someone else raises, and to avoid contradictions when we argue ourselves” (Aristotle, Topic: 100a). Aristotle’s logic covers mainly the interrelation between negative and affirmative existential and universal statements.2 Yet, as has long been noted, not only do quantification and negation exert systematic influence upon possible inferences, also modification allows in many cases for a formal treatment of the validity of inferences.3 For example, if something is a blue house, it is both blue and a house. Adopting a standard settheoretic semantics, the simplest way of capturing this phenomenon is to treat both the adjective and the common noun as predicates. Both will then have sets as denotations. The inference pattern is captured by the 2

Two general problems of Aristotle’s logic were that, first, it did not even nearly cover all possible inferences and, second, the inference rules seemed somehow unmotivated. However, later Peripatetics and the medieval logicians improved both coverage and simplicity of traditional logic. As Ludlow (2002) calls it, the “Holy Grail” of medieval natural logic was to reduce the number of inference rules to two. 3 Traditionally, attributive adjectives have been treated as “normal” predicates. This, however, is problematic in some cases for reasons reviewed below. A note on terminology: I shall reserve the terms attributive and predicative to specify the position in which an adjective occurs. Thus, blue is used predicatively in The house is blue and attributively in The blue house. In contrast, I shall use the terms intersective and subsective to refer to the inferential behaviour of the relevant expressions. This contrast has also often been captured with the terms predicative and attributive (cf. Davidson 1967 and subsequent literature). Finally, the term predicative has been used to classify those adjectives which can appear in predicate position (cf. Alexiadou, et al. 2007: 290-92). According to this classification, kind is predicative but mere is not, because we can say that the student is kind but we cannot say that the fact is mere.

224

Chapter Thirteen

assumption that the denotation of the complex phrase [Adj, N] is, in such a case, the intersection of the two sets denoted by the two predicates as illustrated in (1).4 For example, the meaning of blue house is, according to this theory, the intersection of the denotation of blue and the denotation of house as illustrated in (2).5 1 Intersective adjectives: [[Adj N]] >>$GM@@ŀ>>1@@ 2 [[house]] ^[_KRXVH[ ` [[blue]] ^[_EOXH[ ` [[blue house]] ^[_KRXVH[ `ŀ^[_EOXH[ ` ^[_KRXVH[   EOXH[ ` However, there are adverbs and adjectives for which this rule does not hold when they are combined with a noun.6 For example, a skilful surgeon does not need to be a skilful violinist, even if she is a violinist. Thus, it does not follow from the fact that someone is a skilful N that she is skilful per se. However, a skilful surgeon is a surgeon, as shown in (4). In set theoretic terms, the denotations of the two predicates do not intersect, but the denotation of the whole phrase can be classified as a subset of the denotation of the noun, as in (3). Following the terminology of Kamp and Partee (1995), adjectives which behave in this way are called “subsective”. 3 Subsective Adjectives: [[Adj N]] ‫[[ ك‬N]] 4 [[skilful surgeon]] ‫_[^ ك‬VXUJHRQ[ ` [[skilful surgeon]] ‫_[^ م‬VNLOIXO[ `

4

Katz and Fodor’s (1963) theory of feature conjunction follows the same intuition and has similar formal effect. 5 It has to be noted though that even in these cases, there is an asymmetry between the modifier and the modifee: whether the colour of a certain object counts as red, for example, is not independent of whether the object is a car or an apple (cf. Kamp 1975). 6 This observation is often traced back to Davidson (1967). For the adjectives good and evil see already the discussion in Geach (1956). To my knowledge, the first to explicitly address subsectivity is Aristotle, for example in chapter 11 of De Interpretatione. This chapter also contains a short description of privative adjectives.

Inference and Grammar

225

Formal semantics has mainly followed Kamp (1975) in the suggestion that there is a difference between truly subsective adjectives on the one hand and intersective, but vague or context dependent, ones on the other.7 Thus, whether something is tall, fat or small depends on what you compare it with; and the comparison class does not always have to be provided by the predicate that is modified by the adjective in question.8 As Higginbotham (1985: 563) notes, a big butterfly is not a big animal, but in principle, you can also refer to a butterfly that is big compared to other butterflies as “the small butterfly”. However, relativized to the comparison class, which is generally taken to be provided by context, these adjectives, when combined with a noun, behave like other intersective adjectives.9 7

Prior to Kamp, these adjectives were often treated as subsective. Parsons (1970), for example, proposes treating vague adjectives intensionally which may be seen as the most economical treatment if one accepts Montague’s conjecture of always generalizing to the “worst case”. However, Kamp’s initial motivation for his alternative theory is to develop a proper analysis of comparatives and superlatives which the original Montagovian theory could not analyze. He claims that “what underlies the possibility of making comparative claims is that adjectives can apply to things in various degrees” (Kamp 1975: 128). However, the standard for the grading depends on context – whether Smith is cleverer than Johns may depend on whether you take the ability of solving mathematical problems or the ability to read quickly as central for cleverness. This standard is usually provided by context. And since we need the contextual parameter anyway for comparatives and superlatives, we can also use it for an analysis of the positive. Furthermore, whenever the contextual parameter is provided by the modified noun, the two theories will provide equivalent results. However, if the adjective is interpreted relative to a different standard, the context-free theory is in trouble. 8 However, in some cases there seem to be syntactic constraints on comparison classes. See Bresnan (1973), Ludlow (1989) and Sadler and Arnold (1994) for discussion. There are furthermore syntactic constraints on whether an absolute or relative reading is possible (cf. Cinque 2010: ch. 2.5). 9 Nonetheless, Kamp and Partee (1995: 143) stress that there are many borderline cases and that the distinction between vague and non-intersective adjectives is thus not clear cut: “One might argue that some or all of these other supposedly nonintersective adjectives like skillful might also be better analyzed as contextdependent intersective adjectives, differing from adjectives like tall only in the nature and extent of the contextual effects.” On the other hand, adjectives may be both vague and subsective as in (i), making this a genuine distinction. i

very good as a diagnostican for someone with so little experience (Partee 2007: 154)

A related question is whether subsectivity in this sense is best analyzed in terms of intensionality. McConnell-Ginet (1982) has argued that substitution failure is not

226

Chapter Thirteen

Finally, there are some adjectives for which not even (3) seems to hold. The former president is in most cases not a president anymore and an alleged murderer may not have committed any murder at all. These adjectives are called “non-subsective”. 5 Non-subsective adjectives [[Adj N]] ‫[[ م‬Adj]] & [[Adj N]] ‫[[ م‬N]] 6 [[former president]] ‫_[^ م‬SUHVLGHQW[ ` [[former president]] ‫_[^ م‬IRUPHU[ ` They also differ syntactically from most subsective and intersective ones in that they usually cannot be used predicatively. Thus, whereas the sentences in (7) are fine, those in (8) are not. This is consistent with the intuition that there is no single set denoted by all former things, making the second line of (6) essentially meaningless. 7 He is a former/alleged candidate. 8 *The candidate is former/alleged. In this article, I shall not say much about this last kind of adjective. The fact that most of them are modal or temporal may suggest that they require a special treatment.10 In order to capture the properties of subsective and non-subsective adjectives formally, they are standardly analyzed as properties, in the sense of “property” developed by Montague (1969: 152), rather than as sets. That is, they are taken to be functions which map the intensions of necessarily due to intensionality but can be due to a hidden relationality. Intuitively, the reason for why someone who is a skilful surgeon may not be a skilful violinist doesn’t seem to consist in properties the person has in other possible worlds, but rather in a feature of his in the actual world, namely that he is skilful only in some, but not in all, respects. Furthermore, we can say that a skilful surgeon is skilful (as a surgeon). The following argument depends upon the presupposition that we, at least heuristically, accept the traditional picture, which, as noted, is not unproblematic. See also n. 16 and 19. 10 Schlenker (2006) argues that quantification over, and reference to, objects, times and possible worlds form different but structurally parallel systems in natural language. Sigurðsson (2004) has described such systems for objects and times from a grammatical point of view. In case it is possible to identify a similar system for modality, it could be argued that modal and temporal adjectives interact with these systems, which could explain their non-subsectivity in respect to individuals.

Inference and Grammar

227

noun phrases into such intensions of noun phrases. Subsective and nonsubsective adjectives, therefore, modify the intension, rather than the extension, of a noun. In part criticizing her own earlier work, Partee (2007; 2009; 2010) argues that privative adjectives like fake and counterfeit, which have often been treated as a fourth class of adjectives, are not a special kind of adjective but “normal” subsective ones.11 Originally privative adjectives have been analyzed such that the combination of a privative adjective and a noun implies the negation of what the noun denotes. In set theoretic terms the intersection between the denotation of the noun and the denotation of the phrase is the empty set, as illustrated in (9). A fake gun, for example, is not a gun, as exemplified in (10). 9 Privative Adjectives >>1@@ŀ>>$GM1@@ ‫׎‬ 10 ^[_IDNHJXQ[ `ŀ^[_JXQ[ ` ‫׎‬ However, if these adjectives are indeed privative in this sense, the wellformedness of sentences like (11) is surprising. 11 Is this gun real or fake? (Partee 2007: 153) Therefore, Partee suggests that “in the absence of a modifier like fake or real, all guns are understood to be real guns” (Partee 2007: 156). But when such an adjective is there, the meaning of the noun is expanded such that the denotation of gun includes both fake and real ones.12 Further evidence for treating privative adjectives as subsective comes from syntax: Polish, for example, disallows NP splitting only for non-subsective adjectives. In this respect, privative adjectives form a natural class with intersective and subsective adjectives. It could be added that many privative adjectives allow a predicative use. Montague (1970), Parsons (1970), and Clark (1970) originally proposed a uniform treatment of lexical categories. To allow for this, 11

Some of the privative adjectives could even be treated as intersective: Whilst we may riddle about the question of whether a fake gun is a gun, a fake gun is clearly a fake. 12 An alternative explanation which Partee does not consider could be that the meaning of gun is not by itself definite enough to exclude fake guns. The fact that we usually mean real gun when we use gun in a sentence could then be due to a default which in turn could be pragmatic or grammatical. For the latter possibility compare fn. 10. See also Sheehan and Hinzen (2011).

228

Chapter Thirteen

every expression of a certain category has to be generalized to the “worst case”. In the case of adjectives, this means that all adjectives have to be treated as non-subsective. Predicative occurrences of adjectives are treated as elliptical attributive occurrences. In order to capture the different inferential properties of the different classes of adjectives discussed above, additional specifications may be part of the lexical entry. These may be formulated as “meaning postulates” (Kamp and Partee 1995; Partee 2007). An alternative way of capturing the data is to give up the aim of providing a unified treatment of lexical classes. In this case, intersective adjectives, for example, can be treated as normal sets (Siegel 1976). In both cases, the difference between the different classes of adjectives is treated as lexical. There is, however, a phenomenon that suggests that the difference between subsective and intersective adjectives is not a lexical but rather a grammatical phenomenon: many adjectives which are traditionally classified as subsective also have an intersective reading and most, if not all, adjectives that are traditionally classified as intersective also have a subsective reading – even if in many cases one of the readings is marginal for pragmatic reasons. If this conjecture is correct, the lexicalist account of adjective classes introduces a high redundancy in the lexicon and misses out on an explanation for the systematicity of the phenomenon. Furthermore, in the semantic literature, there has been considerable disagreement about which adjectives belong to which of the classes mentioned above. Even though this is not a decisive point – different speakers may have different lexicons –, it could already put some initial doubt on the idea that adjectives belong to inferential classes qua their lexical entry.

13.2 Intersective and subsective occurrences of adjectives As first noted in Bolinger (1967), deverbal adjectives exhibit an ambiguity when combined with a noun: they can either be used as attributing a temporal or as attributing a non-temporal, that is intrinsic, property. This contrast seems to coincide with the distinction between stage-level and individual-level predicates, discussed in Carlson (1980).13 According to one reading of the adjective visible in (12), for example, Cappella is one of the stars that are in principle visible but perhaps not now. The sentences could in this sense truthfully be asserted at daytime 13

Carlson’s account is lexicalist. Furthermore, he argues for the existence of a third category which modifies kinds, which might correspond to the kindlevel/individual-level dichotomy discussed below.

Inference and Grammar

229

when actually no stars are visible at all. In this case, the adjective tells us something about a general property of the referent of the phrase; it concerns the individual referred to irrespective of the current situation and is thus an individual-level predicate. However, (12) can also be used to say something about the current properties of Cappella. According to this reading, the sentence can only be truthfully asserted if Cappella is visible at the time of the assertion. Used in this way, the speaker is not committed to the claim that the predicate denoted by the adjective holds of the referent of the phrase in general. In this reading, (12) is true even if Cappella was not generally visible but only at the time of the assertion.14 Used in this way, the adjective is therefore called a stage-level predicate. In Germanic languages, only the stage-level reading is available for postnominal adjectives, as shown in (12). Post-nominal and pre-nominal occurrences can be combined. In this case the pre-nominal adjective has to be an individual-level predicate (see 14). Furthermore, if the same (deverbal) adjective appears twice in pre-nominal position, the first occurrence will have a stage-level and the second an individual-level reading, as in (15).15 In the examples (13) to (15), the stage-level predicates have been italicized. 14 Alexiadou et al. (2007: 297) provide the following example (which in part goes back to Bolinger 1967): Assume the X is not always navigable but has become so recently due to a flood. In this case (i) would be true, but (ii) would not be true:

i ii

The rivers navigable include the Amazon, the Nile, the Danube, the X. The navigable rivers include the Amazon, the Nile, the Danube, the X.

)XUWKHUPRUH DV /DUVRQ DQG 0DUXãLþ 4: 274) note, (iii) is odd, because redundant, whereas (iv) is fine: iii List all the stars visible, whether we can see them or not. iv List all the visible stars, whether we can see them or not. It has to be noted though that for many English speakers postnominal adjectives like visible or navigable are not very good anyway which makes the distinctions very subtle if at all existent. The grammatical distinctions are much clearer in other languages (see section 3). Furthermore, heavy adnominal modifiers have to occur postnominally in English. It is not clear that all of them are stage-level predicates. Thus, (v) seems to be fine (thanks to Alex Drummond for pointing this out): v

List all the stars visible to the naked eye/at night/tomorrow, whether we can see them or not. 15 Larson (1999) concludes from this fact that “the relevant syntactic contrast is not strictly one of linear order, but rather one of relative closeness to N.” This has first

230

Chapter Thirteen

12 The visible stars include Cappella. 13 The stars visible include Cappella. 14 The visible stars visible include Cappella. 15 The visible visible stars include Cappella. Stage-level readings are taken to be intersective, but individual-level readings are not. Thus, the denotation of stars visible in (13) can be correctly described as the intersection between all stars and all visible things. However, this is not the case for the individual-level reading of (12). Interpreted in this way, (12) does not imply that Cappella is (currently) visible. If visible has an individual-level reading, the denotation of visible star can therefore not be a mere intersection. However, it can hardly be questioned that visible stars are stars. The individual-level readings of adjectives, thus, seem to follow the pattern of subsective adjectives.16 A close correlate to the individual-/stage-level distinction can also be found in adjectives which are not deverbal. In this case, the individuallevel reading of the deverbal adjectives corresponds to a kind-level reading of the non-deverbal adjective and the stage-level interpretation of the been noted in respect to data from Mokilese and Thai by Sproat and Shih (1988). The distinction between stage-level and individual-level occurrences allows us to make sense of sentences that otherwise would sound contradictory like (i) and (ii). Larson (1999) attributes the discovery of this phenomenon to B. Citko. ((ii) remains odd, though, since it attributes visibility to something genuinely invisible.) i ii 16

The invisible visible stars include Cappella. The visible invisible stars include Cappella.

There are, however, some problems with treating the stage-/individual-level distinction as an intensional phenomenon. It could be argued that the individuallevel reading of adnominal modifiers is as intersective as the stage-level reading: since individual-level readings are in general available in predicative position (cf. n. 28), we could say that a visible (in the individual-level sense) star is indeed visible (in the individual-level sense). The distinction may then, following Chierchia (1995), be analyzed in terms of genericity rather than intensionality. The same strategy could be applied to the individual-/kind-level distinction discussed below. Reichard (in prep.) argues that the analysis in terms of genericity reduces straightforwardly to the grammar of determiner phrases, and indeed follows from more general grammatical organization.

Inference and Grammar

231

deverbal adjective corresponds to an individual-level interpretation of the non-deverbal adjective. The most obvious interpretation of (16) is intersective. According to this reading, I assert that I own an object which is both a car and big (compared to other cars). And if I owned a Mercedes, for example, the sentence would be true. However, there is another reading such that I could truthfully assert (16) if the car I owned is just a small model (or version) of a Mercedes. Now big doesn’t modify the individual car I own, but the kind of car I own. In this reading, the truth of (16) is compatible with that of (17). Germanic languages don’t allow the intersective adjective to appear post-nominally in this case, as shown in (18). However, both uses of the adjectives can appear prenominally. In this case the first has an intersective and the second a subsective reading (see 19 and 20). This parallels exactly the case discussed above (15). 16 17 18 19 20

I have a big car. I have a small car. *I have a car small. I have a small big car. I have a big small car.

Subsective kind-level readings (or alternatively stage-level readings) can be found for most adjectives usually treated as intersective. When presented with a tepid cup of tea in a café, you may say to the waiter that you were hoping for a hot hot drink (if you are not too English an Englishman, anyway). And when you find tomatoes sorted in two boxes, one for the red and one for the green ones, you can ask for the reddest green tomato. Finally, when you find out that the truffles you have bought are made following a French recipe, but are actually produced in the UK, you may wish you had bought French French truffles.17 In sum, the distinction seems to be a genuine one, even if the situations in which a kind reading of these adjectives is required are rather pragmatically special.18

17

Larson (2000) notes that in expressions like “I missed the Thursday Thursday lecture”, the second instance of Thursday gets a generic interpretation and the first a deictic one: what I missed in this case is this Thursday’s Thursday lecture. This seems to be another instance of the kind discussed here. 18 It has to be noted, though, that the kind reading often exhibits a quotational character: when you were promised to see the fattest man ever, but when presented with him, you are not impressed, you may say to your friend afterwards: “Hopefully we are going to see a fat ‘fat man’ next time”.

232

Chapter Thirteen

If this is true, then most common adjectives can be used both intersectively and subsectively, although one or the other usage may be more common. And there is a further ambiguity which exhibits a pattern similar to the one just reviewed: when adjectives are combined with deverbal (and some other) nouns, they can exhibit two different readings. Thus beautiful in (21) can modify either the person or her dancing. Let’s call the first reading adjectival and the second adverbial. In case the person is modified, the interpretation is (or at least can be) intersective. A beautiful dancer in the adjectival sense is a beautiful person who is also a dancer. And if we independently know she is also a singer, she will also be a beautiful singer (in the adjectival sense). However, if beautiful modifies her dancing, the reading is not intersective but subsective, since not everyone who dances beautifully also looks beautiful. And even if we know that our beautiful dancer (in the adverbial sense) is also a singer, we cannot infer from this that she is a beautiful singer. As Kamp (1975: 125) notes, even if we assume that the singers and dancers are coextensive, we cannot infer from the fact that someone dances beautifully that she also sings beautifully.19 Note that the adverbial reading is not available if the adjective appears in postnominal position, like in (22).20 21 A beautiful dancer 22 A dancer, more beautiful than her instructor (Cinque 2010: ch. 2) 19 A problem for the intensional analysis of the phenomenon is that, as Davies (1991) observes, even if the dancers and singers are necessarily coextensive, it does not follow that someone who dances beautifully also sings beautifully or vice versa (cf. n. 9). In this article, I heuristically adopt the subsumption of adverbial adjectives to subsective ones. I am, however, not convinced that their description in terms of intensionality is correct. First, it seems that the two distinctions discussed (stage-/individual-level and adjectival/adverbial) are quite different in nature. It is therefore not clear why they should receive the same semantic and syntactic treatment. Second, McConnell-Ginet (1982) has provided arguments against the proposed intensionality of adverbial adjectives. According to her, the failure of the inferential relation can also be due to a hidden relation. Larson (1998) suggests that for adverbial uses of adjectives this relation may be one to events. For a discussion of Larson’s account and an idea about how to give a grammatical analysis of the phenomena when it is accepted that the two distinctions are different phenomena, see Reichard (in prep.). 20 In English, evaluative adjectives like beautiful can only occur postnominally when they are made “heavier” with the help of a complement. In general, “heavy” modifiers like adjectives with complements, or rather, as Michelle Sheehan (p.c.) points out, right branching modifiers, have to appear postnominally in English.

Inference and Grammar

233

In sum, for most adjectives, both intersective and subsective readings are available. The distinction between these readings is, furthermore, very systematic. This suggests that the distinction has a grammatical origin rather than a lexical one – otherwise we would have to assume a high redundancy in the lexicon. Furthermore, if we take the lexicalist view, we cannot account for the systematicity of the distinction. And finally, if the distinction is indeed grammatical rather than lexical, it is evident why there are many adjectives which are hard to classify.

13.3 Grammatical differences between intersective and subsective uses of adjectives It has already been noted in the second section that there are grammatical differences between the intersective and subsective uses of adjectives discussed: in English, both intersective and subsective adjectives can occur prenominally, but only intersective adjectives can occur postnominally.21 The distinction is not very clear cut in English, since not all intersective uses of adjectives can occur postnominally and some additional ones are marked for many speakers, but there is crosslinguistic evidence for a grammatical difference. As Cinque (2010) notes, in Romance languages, the pattern is complementary to that of English: prenominal adjectives can only get a subsective reading whereas postnominal adjectives are ambiguous between intersective and subsective readings.22 For example, (23) exhibits only an individual-level reading in Italian, but (24) is ambiguous between an individual- and a stage-level reading. 23 Le invisibili stelle di Andromeda esercitano un grande fascino. the invisible stars of Andromeda have a great fascination

21 Originally, Bolinger (1967: 3–4) proposed a rigid correspondence between interpretation and position for those modifiers that can occur both prenominally and postnominally. However, as Sadler and Arnold (1994) stress, and as is evident from the above examples, it is possible for a prenominal adjective in English to give rise to a stage-level interpretation. Bolinger seems to note this himself later in his article when he admits that “there is a way for referent modification to become attributive” (1967: 21). 22 In the literature, it is often suggested that the post-/pre-nominal distinction in Romance mirrors exactly the subsectivity/intersectivity distinction (cf. Alexiadou, et al. 2007).

234

Chapter Thirteen

24 Le stelle invisibili di Andromeda sono moltissime. the stars invisible of Andromeda are very many (Cinque 2010: ch. 2) A similar pattern is found in respect to the adjectival/adverbial distinction. As noted above, in English a prenominal occurrence is ambiguous between the two readings, but in the postnominal position, only the adjectival reading is available. In contrast, in Italian, the prenominal position only allows adjectival readings, whereas the postnominal position is ambiguous between the two readings. Thus, (25) only exhibits the adverbial reading in Italian: “A forward who is good at playing forward would never have done such a thing.” (26), however, is ambiguous between the adverbial and the adjectival reading: “A goodhearted forward would never do such a thing.” 25 Un buon attaccante non farebbe mai una cosa del genere. a good forward not would.do never a thing of the kind 26 Un attaccante buono non farebbe mai una cosa del genere a forward good not would.do never a thing of the kind (Cinque 2010: ch. 2) In sum, the data from English and Italian jointly suggest that adjectives can be used in (at least) two different ways and that this is reflected in grammar. When adjectives are used in one of these two ways, the interpretation is (typically) intersective, whereas if they are used in the other way, the interpretation is typically non-intersective. Adopting the terminology from Sproat and Shih (1988), let’s call the first use of adjectives indirectly modifying and the second directly modifying. In contrast to the semantic intersective/subsective distinction, I take direct/indirect modification to be an abstract grammatical phenomenon, which I, however, argue to coincide with the semantic distinction in the attributive domain. It has to be noted that even though both kinds of adjectives can occur in prenominal position in English, the indirectly modifying adjectives always precede the directly modifying ones, as is evident from (15). Again a similar pattern can be found in Italian for postnominal adjectives: the directly modifying adjectives are always closer to the noun than the indirectly modifying ones. There are additional ambiguities in Romance and Germanic which point into the same direction

Inference and Grammar

235

(for an overview see Cinque 2010: ch. 2).23 Furthermore, there is crosslinguistic evidence for this phenomenon from many languages24 involving different aspects of grammar25 that space does not allow to be reviewed here. Given that the assumption is correct that directly modifying and indirectly modifying adjectives differ grammatically and that this correlates with interpretation, there remains the question of what exactly the difference consists in and how it is related to interpretation. I shall address the first of these two questions in the remainder of this section and reserve the next section for a proposal about the latter question. There is the very old observation that often attributive adjectives can be paraphrased with the help of relative clauses. Thus, already Arnauld and Lancelot (1660: 68-69) note in their “Grammar of Port-Royal” that visible god created the invisible world can be paraphrased as god who is invisible created the world which is visible. They furthermore maintained that the former is a derivationally dependent abbreviation of the latter (cf. Chomsky 1966: 80), an idea that has been taken up in the generative tradition. Bolinger criticized this approach by showing mainly for nonintersective uses of adjectives that such a derivation is not plausible. Sproat and Shih, however, argue that the original intuition can be used in order to explain the difference of interpretation and grammatical behaviour between the two uses of attributive adjectives if only indirectly but not directly modifying adjectives are taken to originate in relative clauses. Furthermore, Alexiadou et al. (2007: 354) comment on Kayne’s (1994, ch. 23

It seems, however, that most of these additional ambiguities can be analyzed as interaction effects of the ambiguities discussed here, in addition to the fact that the indirect modifier always outscopes the direct modifiers and the effects of quantification, negation and distribution. 24 In addition to the ordering phenomena in Romance and Germanic, Cinque (2010) provides in his Appendix an overview and discussion of the relevant literature on Chinese, Maltese, Bosnian, Croatian, Serbian, Romanian, Greek, Russian and German. Sproat and Shih (1988) provide additional evidence for their thesis from Japanese, Kannada, Arabic, Thai, Mokilese and Irish. Cf. also Alexiadou, Haegeman and Stavrou (2007). 25 Sproat and Shih (1988) note that Chinese adjectives with the suffix -de do not obey the ordering restrictions of adjectives found in most languages (cf. Cinque 1999; Scott 2002). They argue that this finding is not restricted to Chinese but is characteristic of indirect modification in all languages which they review. Furthermore, Bosnian, Croatian and Serbian require different agreement morphology in the two cases (Leko 1988 and subsequent work). Moreover, Greek allows determiner “spread” only in cases of indirect modification (Kolliakou 2004).

Chapter Thirteen

236

8) recent revitalization of the old idea to derive all attributive adjectives from relative clauses that such an approach is strongly supported in respect to adjectives that appear postnominally in English “by the fact that APs in postnominal position can only be predicative, i.e. they typically allow a paraphrase with postcopular APs. This follows naturally if syntactically postnominal adjectives originate as predicates in some kind of clause.” Thus, (27a) can be paraphrased as (27b) and (28a) as (28b). However, Alexiadou et al. also stress that the same is not true in respect to all prenominal adjectives, as many of them do not allow for such a paraphrase.26 For example, we use expressions like (29a) but not (29b) and (30a) is fine, (30b) but is not. 27 a b

the students present the students (who) are present

28 a b

dancer more beautiful than her instructor dancer who is more beautiful than her instructor

29 a b

the present situation *the situation (which) is present

26 In parallel to Kayne, Struckmeier (2010) argues for German that all attributive adjectives are clausal. He begins his analysis with present-participles which he shows to comprise full argument structure. Furthermore, he shows that anaphors can be bound within the attributive present-participle which suggests the existence of a subject position:

(i) der sich sehende Mann the himself seeing man “the man who sees himself”

(Struckmeier 2010: example 22)

Struckmeier furthermore argues that the same structure can be used to account for other adnominal modifiers: past-participles differ from present-participles in that vP is unaccusative, adjectival modifiers contain an AP instead of vP. The difference between relative clauses and attributive present-participles is the finite T-head in relative clauses. It is telling, however, that the most complex predicative modifiers which initially motivate Struckmeier’s analysis, present-participles and relative clauses, cannot have a subsective reading in German whereas predicative adjectives can. Thus, even though there do not seem to be morphosyntactic differences between subsective and intersective modifiers in German, the question is whether the largely empty functional structure in the case of adjectival modification is always projected or whether it is only projected in the case of indirect modification – as the data of other languages suggest.

Inference and Grammar

30 a b

237

the mere detail *the detail is mere

Thus, the proposal, which has also been defended most recently by Cinque (2010), is that the meaning contrast in attributive adjectives is due to a difference in structure: indirectly modifying adjectives are reduced relative clauses whereas directly modifying adjectives do not have a clausal origin.27 The details of the syntactic proposals differ between authors. The reduced relative clauses may be taken to be generated postnominally, perhaps along the lines of Kayne’s theory of relative clauses, and then fronted if they occur prenominally. Alternatively, as Cinque most recently has argued, they may be generated prenominally like directly modifying adjectives. The order we find in the languages discussed can then be derived by cyclic NP movement. The main difference between English and Italian in this respect will then be the amount of obligatory and possible movement of NPs. I shall remain neutral in respect to these questions since nothing about my proposal hinges on these details. In the next section, I shall argue that if the proposal that indirectly modifying adjectives are clausal whereas directly modifying ones are not is indeed correct in any of its formulations, there may be an explanation for the intersectivity/subsectivity distinction which goes beyond the mere correlation that has often been argued for so far.28

27 An alternative proposed by Sadler and Arnold (1994) is that direct modifiers are X0 adjunct whereas indirect modifiers are phrasal. If their analysis is right, my proposal will only work in case APs are phasal expressions. 28 There is a potential problem with the idea just delineated: If indirect modifiers originate as relative clauses, they occupy a predicative position in the relative clause. We would then expect to find a close correlation between the distribution and possible readings of predicative adjectives and indirectly modifying attributive adjectives. Whereas the adjectival/adverbial distinction supports this conjecture (Larson 1999), judgements seem to differ as to whether the stage-/individual-level distinction does so too. In (0b), beautiful can only get an adjectival but not an adverbial reading. However, consider (ia, iia). According to Bolinger (1967), who is in accordance with my own judgement, both expressions are ambiguous between a stage-level and an individual-level reading. And although the individual-level reading is harder to get for relative clauses, the ambiguity does not go away (ib, iib). This phenomenon motivated Bolinger’s suggestion that the copula itself is ambiguous between a stage-level and an individual-level version – a hypothesis which was supported by the fact that Spanish indeed seems to have two different copulas (cf. e.g. Sera 1992). However, Larson (1999) argues that predicative uses and occurrences in relative clauses only allow for the stage-level reading: “it seems

Chapter Thirteen

238

13.4 A grammatical explanation for subsectivity and intersectivity There are two kinds of implicational relations which have been described in section 13.1. First, the denotation of the combination of adjective and noun may or may not imply the denotation of the noun. Second, the combination may or may not imply the adjective. Intersective and subsective uses of adjectives do not differ in that both allow for the first kind of implication. They do differ, however, in that intersective adjectives do, but subsective adjectives do not allow for the second kind. As noted above, in formal semantics the inferential behaviour of expressions involving adjectives is often taken to be part of the lexical entry of the adjective. If this is the case, then all kinds of implications of expressions involving the relevant adjective will trivially have the same source: the lexical entry of the adjective.29 However, if it is correct that the difference between intersective and subsective adjectives is grammatical, then it is possible that the different kinds of implications which adjectives allow for have different grammatical sources. I shall argue in the following that this is indeed the case. From a grammatical point of view, it is not surprising that (in the absence of expressions which involve negation30) the first kind of implication holds, that is, that it is possible to drop the modifier in both the intersective and the subsective cases, whilst preserving the truth value. A grammatical object is complete without an adjective added to it: it is not that when used as a predicate, […] visible can only have its “temporary property” meaning”. If this is true, that’s good news for the clausal hypothesis. i.

a. b.

ii. a. b. 29

This river is The rivers which are navigable include the Danube. This star is The stars which are visible include Cappella.

navigable. visible.

An alternative developed in Larson and Segal (1995) and Larson (1998) is to treat the ambiguity (at least as far as the adjectival/adverbial distinction is concerned) as a matter of the lexical structure of the noun modified. Note that this is still a lexicalist solution to the problem. The approach considered here takes the ambiguity to be one in grammatical structure. 30 As already observed in the Middle Ages, negation changes an environment from downwards entailing to upwards entailing or vice versa. Ludlow (2002) shows how this relatively simple assumption can handle all kinds of quantification and can be integrated into the syntax of natural languages.

Inference and Grammar

239

selected for (Svenonius 1994: 442-43).31 Moreover, adding an adjunct to a given syntactic object doesn’t change anything categorically in that object. It is thus natural to assume that whatever semantic value an object has after specification by an adjunct (or specifier), it has already had before the addition of the adjunct. Therefore, when the adjunct is deleted from the structure, the semantic value of the host object should be unchanged (cf. Hinzen and Reichard 2011). This rather trivial explanation relies on the grammatical asymmetry of adjunction (and in fact most if not all grammatical relations32). The host is what provides the semantic value, whereas the modifier only specifies this value and is thus always interpreted relative to its host. Such an asymmetry is unexpected if modification is essentially conjunction, which is a symmetrical relation. Given the asymmetry, however, subsectivity of the modifier is expected.33 What is puzzling from a grammatical point of view, then, is intersectivity, not subsectivity. But the assumed clausal nature of intersective adjectives may be taken to explain this fact in the light of current syntactic theory. Chomsky (2001) proposes that the computational device of human language constructs syntactic representations in units which he calls phases. The motivation for phases is largely grammar-external and concerns, for example, memory limitations: phases reduce the computational burden. When the edge of a phase is reached in the computational derivation, the complement of the phase head is “sent” to the conceptual-intentional interface, where it receives an interpretation, and to the phonological interface, where it is spelled out; the derivational device now can “forget” about the complex structure of the expression generated. In Derivation by Phase, Chomsky takes CP and vP to be phases. Grammatical evidence for the existence of phases mainly comes 31 Attributive adjectives therefore have often been taken to be adjuncts. As a result of Kayne’s (1994) antisymmetry-thesis which assimilates specifiers and adjuncts, Scott (2002) and Cinque (2010), for example, assume adjectival modifiers to have more structure: in the extended projection of the noun, APs feature in the specifier position of functional heads which ensures the proposed universal order in which adjectives occur. This, however, does not change anything about the optionality of these modifiers—in principle, a structure is complete without them. 32 Even in cases which seem symmetrical, a certain asymmetry seems to be necessary for a compositional interpretation. See Moro (2000) and Sheehan (2011) for small clauses and Bauke (2011) for compounds. 33 In a certain sense, a similar conclusion was already reached by Bolinger (1967: 3) who argues that “by itself the fact that many more adjectives are restricted to attributive position than to predicative position is suspicious; if anything the reverse should be true if we want to base attribution on predication.”

240

Chapter Thirteen

from island conditions: CP and vP both “are reconstruction sites and have a degree of phonetic independence” (Chomsky 2001: 12). Phases are in this sense “impenetrable”. What is important for our current purposes is that, as noted, phases are, inter alia, taken to be units of semantic valuation. If indirect modifiers are clausal and thus CPs, they will, thus, be such units of valuation. In this case, the semantic value of an indirect modifier is already fixed when it modifies its host. The host can, thus, not relativize the modifier in any way. And if the modifier is not interpreted relative to its modifee then intersectivity is the only possible interpretation left.34

13.5 Conclusion It has first been argued that most adjectives have both intersective and subsective readings, even if often one of the two readings is marginal for pragmatic reasons. The different readings are, furthermore, systematically available. This suggests that the distinction is grammatical rather than lexical. Otherwise, the lexicon has to be taken to be highly redundant and there is no explanation for the systematicity of the phenomenon. A further argument for this conclusion has been the consistent problems of classifying adjectives into the relevant semantic classes. Second, some grammatical evidence for the thesis that the phenomenon is grammatical has been delineated in section 13.3: crosslinguistically, the different interpretations seem to correspond to syntactic position and structure. Thus, grammatical considerations suggest that there are two different kinds of modifiers which have been called direct and indirect modifiers. In accordance with recent syntactic theories, it has been assumed that indirect modifiers are reduced relative clauses whereas direct modifiers don’t have a clausal origin. In a third step, it has been demonstrated that, if this thesis is indeed correct, there may be a grammatical explanation for why direct modifiers are subsective while indirect modifiers are intersective. Whereas, due to the inherent asymmetry of language, subsectivity is expected, intersectivity follows from the clausal nature of the relevant occurrences of the modifiers: in current syntactic theory clauses are taken to be phases and phases are units of semantic valuation. If, because it is already fixed at the time of modification, the semantic value of the

34

Struckmeier (2010), for example, directly argues for the thesis that attributive modifiers are phasal in German. I would want to restrict this thesis to indirect modifiers, cf. n. 26.

Inference and Grammar

241

modifier cannot be relativized to that of the modifee, the only available interpretation left is intersective. Most current theories of language, both in linguistics and in philosophy, take language to be in the periphery of our cognitive abilities: it is a device for the expression of our thoughts. However, if, as argued above, it is correct that the validity and invalidity of inferences may not only correlate with grammatical phenomena, but find an explanation in them, language may turn out to be more central for human cognition than this picture suggests (cf. Hinzen 2007; Hinzen 2006). Thus, the question arises which aspects of human-specific cognition can in principle be accounted for with the help of independently motivated grammatical assumptions. As noted, Ludlow (2002) has shown more generally how inferences involving quantification (but not non-intersective modification) can at least be integrated into current syntactic theory. In case this research programme turns out to be fruitful, further metatheoretical questions arise about the relation between the normativity of the validity and invalidity of inferences, which is intuitively non-psychological, and our psychological grasp of this normativity.

References Alexiadou, A., L. Haegeman and M. Stavrou. 2007. Nouns Phrase in the Generative Perspective. BerlIn Mouton de Gruyter. Aristoteles. 1967. Topiques. Tome I: Livres I–IV. Ed. by J. Brunschwig. Paris: Les Belles Lettres. Arnauld, A and C. Lancelot. 1660. General and Rational Grammar: The Port-Royal Grammar. Transl. by J. Rieux and B. E. Rollin. The Hague: Mouton 1975. Bauke, L. S. 2011. A Phase-Theoretic Analysis of Root Compounds (From a Crosslinguistic Perspective). Paper presented at Irish Network in Formal Linguistics Conference. University of Ulster at Jordanstown. 25–27 May 2011. Bolinger, D. 1967. Adjectives in English: Attribution and Predication. Lingua 18: 1–34. Bresnan, J. W. 1973. Syntax of the Comparative Clause Construction in English. Linguistic Inquiry 4: 275–343. Carlson, G. 1980. Reference to Kinds in English. New York: Garland. Chierchia, G. 1995. Individual-Level Predicates as Inherent Generics. In G. Carlson and F. J. Pelletier (eds.), The Generic Book, 176–223. Chicago: University of Chicago Press. Chomsky, N. 1966. Cartesian Linguistics. New York: Harper and Row.

242

Chapter Thirteen

—. 2001. Derivation by Phase. In M. Kenstowicz (ed.), Ken Hale: a life in language, 1–52. Cambridge, MA: MIT Press. Cinque, G. 1999. Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: Oxford University Press. —. 2010. The Syntax of Adjectives: A Comparative Study. Cambridge, MA: MIT Press. Clark, R. 1970. Concerning the Logic of Predicate Modifiers. Nous 4: 311–335. Davidson, D. 1967. The Logical Form of Action Sentences. In Essays on Actions and Events, 105–122. Oxford: Oxford University Press (2001). Davies, M. 1991. Acts and Scenes. In N. Cooper and P. Engel (eds.), New Enquiries into Meaning and Truth, 41–82. Hemel Hempstead: Harvester Wheatsheaf. Geach, P. T. 1956. Good and Evil. Analysis 17: 33–42. Higginbotham, J. 1985. On Semantics. Linguistic Inquiry 16: 547–593. Hinzen, W. 2006. Mind Design and Minimal Syntax. New York: Oxford University Press. —. 2007. An Essay on Names and Truth. New York: Oxford University Press. Hinzen, W and U. Reichard. 2011. The Event-Argument Hypothesis: In Search of Evidence. Ms. Kamp, H. 1975. Two Theories about Adjectives. In E. L. Keenan (ed.), Formal Semantics of Natural Language, 123–155. Cambridge: Cambridge University Press. Kamp, H and B. H. Partee. 1995. Prototype Theory and Compositionality. Cognition 57: 129–191. Katz, J. J and J. A. Fodor. 1963. The Structure of a Semantic Theory. Language 39: 170–210. Kayne, R. 1994. The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kolliakou, D. 2004. Monadic Definites and Polydefinites: Their Form. Meaning and Use. Journal of Linguistics 40: 263–323. Larson, R. K. 1998. Events and Modification in Nominals. In D. Stolovitch and A. Lawson (eds.), Semantics and Linguistic Theory VIII (SALT8), 145–168. Ithaka. NY: Cornell University. —. 1999. The Semantics of Adjectival Modification. Paper presented at The Dutch National Graduate School (LOT). Amsterdam. http:// semlab5.sbs.sunysb.edu/~rlarson/LOT(99)/Contents.htmld/index.html. —. 2000. Temporal Modification in Nominals. Paper presented at The Syntax of Tense and Aspect. Paris. 15–18.11.2000. http://semlab5. sbs.sunysb.edu/~rlarson/ParisHandout.pdf.

Inference and Grammar

243

Larson, R. K. and G. Segal. 1995. Knowledge of Meaning: An Introduction to Semantic Theory. Cambridge, MA: MIT Press. /DUVRQ5.DQG)0DUXãLþ2Q,QGHILQLWH3URQRXQ6WUXFWXres with APs: Reply to Kishimoto. Linguistic Inquiry 35: 268–287. Leko, N. 1988. X-bar Theory and Internal Structure of DPs. Lingua 75: 135–169. Ludlow, P. 1989. Implicit Comparison Classes. Linguistics and Philosophy 12: 519–533. —. 2002. LF and Natural Logic. In G. Preyer and G. Peter (eds.), Logical Form and Language, 132–168. Oxford: Clarendon. McConnell-Ginet, S. 1982. Adverbs and Logical Form: A Linguistically Realistic Theory. Language 58: 144–184. Montague, R. 1969. On the Nature of Certain Philosophical Entities. In R. H. Thomason (ed.), Formal Philosophy: Selected Papers of Richard Montague, 148–187. New Haven: Yale University Press (1974). —. 1970. English as a Formal Language. In R. H. Thomason (ed.), Formal Philosophy: Selected Papers of Richard Montague, 189–224. New Haven: Yale University Press (1974). Moro, A. 2000. Dynamic Antisymmetry. Cambridge, MA: MIT Press. Parsons, T. 1970. Some Problems concerning the Logic of Grammatical Modifiers. Synthese 21: 320–334. Partee, B. H. 2007. Compositionality and Coercion in Semantics: The Dynamics of Adjective Meaning. In G. Bouma. I. Krämer and J. Zwarts (eds.), Cognitive Foundations of Interpretation, 145–161. Amsterdam: Royal Netherlands Academy of Arts and Sciences. —. 2009. The Dynamics of Adjective Meaning. In A. E. Kibrik et al. (eds.), Computational Linguistics and Intellectual Technologies: Papers from the Annual International Conference “Dialogue 2009” (Bekasovo. May 27–31. 2009), 593-97. Moscow: Russian State Humanities University. —. 2010. Privative Adjectives: Subsective Plus Coercion. In R. Bäuerle. U. Reyle and T. E. Zimmermann (eds.), Presuppositions and Discourse: Essays Offered to Hans Kamp, 273–285. Bingley: Emerald Group. Reichard, U. In prep. Making Events Redundant: Adnominal Modification and Phases. To appear in P. Stalmaszczyk (ed.), Philosophical and Formal Approaches to Linguistic Analysis. Frankfurt a. M.: Ontos. Sadler, L and D. J. Arnold. 1994. Prenominal Adjectives and the Phrasal/ Lexical Distinction. Journal of Linguistics 30: 187–226. Schlenker, P. 2006. Ontological Symmetry in Language: A Brief Manifesto. Mind and Language 21: 504–539.

244

Chapter Thirteen

Scott, G.-J. 2002. Stacked Adjectival Modification and the Structure of the Nominal Phrases. In G. Cinque (ed.), Functional Structure in DP and IP, 91–120. New York: Oxford University Press. Sera, M. D. 1992. To Be or to Be: Use and Acquisition of the Spanish Copulas. Journal of Memory and Language 31: 408-27. Sheehan, M. 2011. Is the Argument/Predicate Distinction a Semantic Primitive. Ms. Sheehan, M. and W. Hinzen. 2011. Moving Towards the Edge. Linguistic Analysis 37: 405–458. Siegel, M. E. A. 1976. Capturing the Adjective. Amherst, MA: University of Massachusetts. Sigurðsson, H. Á. 2004. The Syntax of Person. Tense and Speech Features. Italian Journal of Linguistics 16: 219-51. Sproat, R. and C. Shih 1988. Prenominal Adjectival Ordering in English and Mandarin. In J. Blevins and J. Carter (eds.), Proceedings of NELS 18: 465–489. Amherst, MA: GSLA. Struckmeier, V. 2010. Attributive Constructions, Scrambling in the AP and Referential Types. Lingua 120: 673–692. Svenonius, P. 1994. The Structural Location of the Attributive Adjective. In E. Duncan. D. Farkas and P. Spaelti (eds.), Proceedings of the Twelfth West Coast Conference on Formal Linguistics, 439–454. Stanford. CF: Stanford Linguistics Association.

CHAPTER FOURTEEN WHAT WENT WRONG IN “PARKY” VS. RYAN? FORMAL AMBIGUITY IN THE EMERGENCE OF DISAFFILIATION CATRIN S. RHYS 14.1 Introduction In 2003, veteran celebrity chat show host Michael Parkinson interviewed the actor Meg Ryan in what he later referred to as his “most awkward TV moment”. The interview received extensive media attention as a notoriously unsuccessful interview. Meg Ryan was widely quoted as describing Parkinson as “a nut”, as well as “patronizing” and “disapproving”. Conversely, in the UK, newspapers like the Sun described Meg Ryan as “monosyllabic, obstructive and ill-mannered”. Despite the notoriety of the interview, there was no dramatic breakdown in the interaction: no refusal to answer, no overt display of anger or hostility and no “walk out”, the ultimate in breakdown in the interview format (Llewellyn and Butler 2011). Nonetheless, in a 2006 UK survey, the interviewed was ranked as the third “most shocking” TV chat show moment ever. What is interesting about this particular “failed” celebrity interview is the degree to which perceptions and opinions of the accountability for the failure of the interview are polarized. Media reporting at the time and online YouTube commentary reveal two highly polarized groups of viewers: those who place the blame for the failed interview with Parkinson’s interviewing and those who conversely see Ryan as an uncooperative interviewee. The interaction in this interview is thus interesting from the perspective of its unusually disaffiliative nature as a failed chat show interview but particularly because of the highly polarized perceptions of the accountability for the disaffiliation in the interview.

246

Chapter Fourteen

14.2 Aims This paper aims to examine the sequential properties of the emerging disaffiliation in the interview data drawing on the conversation analytic concepts of preference organization (Sacks 1987) and repair (Schegloff et al. 1977). Additionally the semi- institutional context of a chat show interview (Ilie 2001) will be considered for its impact on the perceptions of the disaffiliation. The overall aim of the analysis is to examine the relevance and import of the micro level understanding of the data provided by a conversation analytic approach to the macro level question of the highly contested viewer perceptions of the interview participants’ roles in the interaction. Note, however, that the aim is not to address the interpretative procedures adopted by the audience per se. Hence, the analysis does not address issues such as viewer demographics, for example, distinctions between American and British audiences1 or male and female viewers. Rather, as articulated by Schegloff (1988/89), the aim of the analysis is to examine the practices by which the participants themselves both constituted the interactional event in progress and displayed their understanding of its constitution. The point here is that these practices, though intended primarily for the interlocutor, are available to the viewing audience and hence are likely to form a significant element of any interpretive procedure (Schegloff 1988/89: 216). The focus of this paper is thus the practices of talk-in-interaction deployed by the interview participants in producing this occasion of disaffiliative interaction. The paper adopts a Conversation Analytic approach to the data involving micro level turn by turn analysis, focusing on turn design, the action(s) accomplished by each turn at talk and the relationship between the two. At the heart of analysis in CA is the organization of the sequence. Conversational actions are understood to be organized in sequences in which each action is contingent on the preceding action and makes relevant a subsequent action. The sequence is thus also the locus of intersubjective understanding (Schegloff 1990, 1992) as each participant displays in their own turn of talk their understanding of the preceding turn, in addition to projecting/making relevant subsequent action(s). It is this displayed understanding that provides the analyst with evidence for analytic claims about the data. 1

For example, it is intuitively likely that UK audiences for whom “Parky” is something of a “national treasure” were predisposed to see his role in a more favourable light. The question for this paper however is what it was in the talk that made such contrasting perceptions of both Parkinson and Meg Ryan possible.

Ambiguity in the Emergence of Disaffiliation

247

14.3 Analytical constructs Adopting a CA approach to the emerging disaffiliation in this data requires an understanding of the organization of argument sequences, disaligning actions and the display of disaffiliative stance in talk-ininteraction (Stivers 2008). Hence we turn now to the CA concept of preference/dispreference and its relationship with alignment/affiliation2.

14.3.1 Preference and alignment We have seen already that the concept of sequence in CA refers to a stretch of talk in which some course of action gets “initiated, worked through and brought to closure” (Schegloff 2007: 3). Within this action oriented perspective on the organization of talk, preference organization refers to certain aspects of the talk that are oriented to the smooth accomplishment of projected sequence of actions. Sacks (1992) writes of initial actions that “invite” default interpretations; that is to say that an initial action, a First Pair Part (FPP), is designed by a speaker to invite a particular type of response of Second Pair Part (SPP). Following Pomerantz (1984: 63), “a next action that is oriented to as invited will be called a preferred next action; its alternative, a dispreferred next action”. Thus a response produced as a preferred next action is aligned to the projected action sequence. On the other hand, departure from default interpretation projects disalignment from the projected activity and is indicated by “dispreference markers” such as delay, mitigation etc (Pomerantz 1984). Alignment is thus understood as a property of the actional content of the talk in progress. Schegloff (1988) reflects that preference may be understood either to be grounded in structure of the course of action in progress, in which case the actional content of a particular first pair part is understood to project a preferred type of response, or to be grounded in the practices which mark the design of the response as preferred or dispreferred. Schegloff (2007) also applies the practice-based conception of preference to the design of first pair parts, recognizing that a range of linguistic features and other resources of turn design can influence the projected preferences of a FPP. An important concomitant effect of preference is that participants are observed to orient to both the structural and practice based features of 2

Following Stivers (2008), alignment is understood as structural; it refers to talk which is aligned with the interactional activity in progress. Affiliation, on the other hand, is social; it is used to refer to the displayed stance of the participants.

248

Chapter Fourteen

preference and draw conventional inferences based on those features about the course of action that is ongoing and what it projects. Another relevant feature of dispreferred responses is that they are “expansion relevant” (Schegloff 2007: 117), that is to say that while a preferred response is associated with closure of the sequence, a dispreferred response regularly leads to post expansion of the sequence (i.e. expansion following the dispreferred second pair part). Such expansions may involve elaboration or reworking of either first or second pair part (Schegloff 2007: 162). In both cases, the post expansion is displays an orientation to the disaligned nature of the dispreferred second pair part. Disagreement is an example of one such disaligned action. It is a typically a departure from default interpretation and hence typically a dispreferred response. That is to say that initiating actions such as assessments typically invite agreement, hence agreement is structurally projected as preferred and disagreement is typically produced with a variety of dispreference markers. However, Pomerantz (1978) also points out that it is not that disagreement is inherently dispreferred, since the preferred response after a self-deprecation, for example, is a disagreement. Hence, in the context of self-deprecation, it is an agreement that is produced with dispreference markers. For this paper, it is perhaps more important to also note that whether or not a disagreement is produced as preferred or dispreferred depends also on the nature of the overall opposition sequence involved. Hutchby (1996) observes that disagreements may be aggravated or mitigated, that is to say produced as preferred or dispreferred respectively, and that where a disagreement is produced as aggravated, it displays an orientation to the context as one of aggravated opposition (M.H. Goodwin 1983). We return to this idea when we examine the overtly argumentative sequences in our data below.

14.3.2 Repair and Alignment The notion of repair refers to sequences that are addressed to problems in speaking, hearing or understanding in the preceding talk (Schegloff 2007). It has long been noted, following Schegloff et al. (1977), that conversational repair is accomplished in phases: initiation; repair; receipt. The initiation of repair orients to some aspect of the preceding talk as being in need of repair and may or may not be co-terminus with the actual production of the repair itself. Initiation of repair and the actual repair may be carried out by either the producer of the repairable (self initiation, self repair) or by the hearer of the repairable (other initiation, other repair).

Ambiguity in the Emergence of Disaffiliation

249

However, there is also a preference hierarchy operative in contexts of repair such that self repair is preferred, and other repair dispreferred (Schegloff et al. 1977). Of particular interest for the analysis presented here are observations about patterns of ordering of other-initiation techniques. Svennevig (2008) points out that there are two very different parameters that have been used to reveal a pattern of ordering of other-initiation devices. One hierarchy relates to how precisely the Next Turn Repair Initiator (NTRI) identifies which part of the preceding talk is the repairable (e.g. Schegloff et al. 1977; Clark and Schaefer 1987; Drew 1997). Schegloff et al. (1977: 369) write of a “preference for stronger over weaker initiators”, where strength is understood as the “capacity to locate a repairable”. The second hierarchy focuses on the hearer’s orientation to the nature rather than location of the repairable. As Schegloff (2007: 102) points out: “other initiated repair sequences can operate as pre-rejections and predisagreements – as harbingers of dispreferred base second pair parts”. In other words, repair is closely linked to alignment in that it provides a sequential opportunity to amend an utterance that is threatened with nonalignment. This second hierarchy of NTRIs is thus related to the interaction between repair and alignment and reveals an ordering between practices of repair that are oriented to delimiting the nature of the trouble as one of hearing, understanding or disagreement (Mazeland and ZamanZadeh 2004; Pomerantz 1984; Schegloff 2007; Sorjonen 1996). Svennevig (2008) examines how such hierarchies are manifested in talk and provides evidence specifically for a preference for interpretation of an NTRI as repair of hearing.

14.3.3 Repetition, repair and alignment One particular conversational resource that gets employed in the service of repair is repetition. A participant may produce either a full or a partial repeat of the preceding turn as a device to initiate repair. Partial repeats function as closed class repair initiators that locate a component of a trouble-source question as the repairable (Schegloff et al. 1977). On the other hand, full repeats of a preceding question have been argued to be a form of open repair that orients to the preceding questions as a whole as the repairable (Robinson and Kevoe-Feldman 2010). Robinson and Kevoe-Feldman (2010) further argue that the repetition itself forms a “claim to have heard” and hence, with respect to the ordering of NTRIs discussed above, argue that repetition based NTRIs are oriented to repair of understanding or acceptability. However, others (e.g. Bolden 2009;

250

Chapter Fourteen

Koshik 2005; Svennevig 2008) treat repetition not as a claim to have heard but as a “candidate hearing” which the trouble source speaker may then confirm or correct. These two positions will be examined in the analysis of the data below; the important observation at this point is that as a practice of other-initiated repair, repetition is a potential resource for projecting disalignment. Interestingly, repetition is also associated with disalignment in contexts other than repair sequences. In particular, Schegloff (1997: 535) observes that repeat-prefaces to answers “embody the beginning of a disaligning, or otherwise negatively valenced, response”. More specifically, Bolden’s (2009) examination of repeat prefaces in Russian conversation reveals that repeat prefaces are a strategy for resisting the interactional agenda of the preceding initiating action and act as a tacit claim that the initiating actions embody “inappropriate or massively incongruent assumptions” (Bolden 2009: 140). These repetition based practices are sequentially different in that repetition that initiates other repair initiates a sequence and projects talk by the speaker of the repairable, whereas the repeat-prefaced response projects further talk by the current speaker, the producer of the SPP in the current sequence. Yet as Bolden (2009: 138) points out, “the two practices are related: Both consist of repeating a part of another person’s talk and both index the speaker’s difficulty in responding to the preceding utterance”. Moreover, both may function to display disalignment.

14.3.4 Argument sequences Hutchby (1996) presents an action-opposition model of arguments in which an oppositional move retrospectively treats the preceding turn/action as arguable. Argument sequences are thus in this view “response-centred emergent events” rather than projected sequences of actions. In other words, overtly argumentative sequences emerge where turn by turn speakers treat the preceding turn as arguable and produce a relevant oppositional move.

14.4 Analysis Although the data contains examples of explicit argument sequences, the organization of such sequences does not directly shed light on the contested nature of the perceptions of the two participants in this interview, since in explicit argument, the context of aggravated opposition is a collaborative interactional achievement of both participants. Hence, the primary analysis in this paper focuses on more subtle aspects of

Ambiguity in the Emergence of Disaffiliation

251

alignment/disalignment in the data where there may be some potential ambiguity in the actional content of the talk. In addition, in order to understand how this context of aggravated opposition has emerged, the analysis moves to the very beginning of the interview and the very first question – answer sequence. In this sequence, Parkinson links to the previous interview with Trinny and Susanna, co-hosts of a BBC fashion show: Extract 1 15 16 17 18 19 20 21 22 23 24

P so: do you have a take on this fashion stuff th they do? P =I mean do you find fashion empowering? d'you feel goo:d o o (.) [ from the outside ] M ((pulls face))[Do I find it] empowerM [ing? ] P [yeah:.] in the [sense of when] you M [Not usually ] P =you don’tĻ [you] don’t care [about it] M [no ] [Did I Ĺ ]just ruin your whole se:gment?

The turn has the format of a multi-unit questioning turn (MUQT), which Linell et al (2003) show are widely associated with topic initiation or retopicalization. Here Parkinson links with the previous interview by retopicalizing from that interview. The internal sequential organization of the MUQT involves what Clayman and Heritage (2002b: 757) have termed a “question cascade” which involves “different versions of what is ostensibly the same question” . Linell et al (2003) label the structure Q+Q+Q and suggest that the first question is usually the basic unit (the head) and that the relationship between two adjacent questions is typically a particularizing one where Qn+1 incorporates a candidate answer (Pomerantz 1984) to the preceding question (Qn). A response to Qn+1 is thus a relevant response to Qn. Overall the MUQT has the characteristics of a topic-proffering sequence, for example, being implemented by a yesno question and preferring expansion over sequence closure (Schegloff 2007). From the outset, the design of Parkinson’s MUQT brings issues of preference and alignment to the fore. Specifically, the cascade from Q1 to Q2 here has the effect of changing the preference orientation of the projected SPP. The first variant of the question (Q1) invites an expression of opinion, preferring the positive “having a take” over a negative not “having a take” but the format of the question does not project a particular format or polarity for the opinion it invites. Put more informally, the format of the question does not specifically invite the expression of a

252

Chapter Fourteen

positive or negative “take” on fashion3. Parkinson then reformulates the question in lines 16-17 incorporating a candidate answer involving a positive “take” on fashion as “empowering” given the general preference for agreement (Sacks 1987). The particularizing reformulation in Q2 thus has the effect of shifting the projected aligned response from one which expresses an opinion to one which in order to be aligned should express a positive opinion. In other words, Parkinson appears to orient to the possibility of a potentially undesirable response to Q1and reformulates Q2 to project agreement in the SPP. Ryan’s response, however, defers the production of the SPP by producing an insertion sequence. Specifically Ryan initiates repair with a full repeat of Q2. Parkinson responds to this repair initiation by re-affirming Q2 in line 20: “yeah” and beginning to produce an expansion (clarification). This sequence has very interesting properties from the perspective of the emerging disalignment in the interview. Parkinson’s MUQT we have already seen, is oriented to achieving a particular form of alignment, namely an agreement with the assessment of fashion as “empowering”. However, the break in contiguity between FPP and SPP caused by the insertion sequence and the production of an other initiated repair sequence have both been shown to project the possibility of a dispreferred (i.e. nonaligned) second pair part (Sacks 1987, Schegloff 2007). However, as Schegloff (2007: 105) also observes, sometimes repair is simply repair and the break in contiguity is “merely incidental to the repair operation per se”. Recall also that Svennevig (2008) argues for a hierarchy of NTRIs that reveals a preference ordering from the preferred interpretation of an NTRI as repair of hearing to the dispreferred interpretation as one of disagreement. In this data, Parkinson does not orient to Ryan’s repair initiation as a pre-disagreement. Instead, he orients to the structurally preferred interpretation of the repair initiation as a problem of hearing or understanding, by first re-affirming and then elaborating or clarifying the FPP. Ryan however produces an overlapping and barely mitigated dispreferred disagreeing SPP in line 21. There are a number of interesting observations here. Firstly, Ryan’s overlap with Parkinson’s elaboration provides evidence that the repair initiation was originally produced to project impending disalignment since the disagreement is produced almost immediately after the reaffirmed FPP (“yeah”) rather than following the elaboration. As such Ryan’s disagreeing SPP can be understood as an 3

The only minor sense in which the format might be argued to project a particular kind of opinion lies not in the structure of the FPP but the word choice “stuff” which might be heard as relatively negatively connoted.

Ambiguity in the Emergence of Disaffiliation

253

oppositional move that retrospectively orients to Parkinson’s FPP as an arguable. Moreover, although deferred by the NTRI, once produced the disagreement is formulated on balance as an aggravated oppositional move, since it does not carry the usual markers of a dispreferred SPP within the turn itself: it is overlapping rather than delayed; it is not particularly mitigated or hedged4. Note also that Ryan’s repair initiation, understood as a predisagreement, provides an opportunity for self repair to avoid a disaligned response. Parkinson’s orientation to the NTRI as a repair of understanding simultaneously entails lack of uptake of the opportunity for repair of disagreement and is therefore potentially an accountable absence; a failure to avoid projected disalignment. This interpretation becomes more salient when the non-verbal displays by Ryan are included in the analysis since her immediate reaction to Parkinson’s initial MUQT was to pull a face, one that might be interpreted as expressing surprise or a challenge to the assumptions of the preceding questions (eyebrows raised, mouth turned down). When we move to the post completion part of this sequence, similar observations emerge. Schegloff (2007) demonstrates that dispreferred SPPs make relevant expansion/explanation of the sequence, hence it is unsurprising to see non-minimal expansion of the sequence in lines 22 -23 (repeated here for ease of reference): Extract 2 22 P =you don’tĻ [you] don’t care [about it] 23 M [no ] [Did IĹ ]just ruin your 24 whole se:gment?

Parkinson’s post expansion FPP in line 22 is potentially ambiguous between initiation of repair and simple receipt of the preceding turn (Svennevig 2004). However, it is oriented to by Ryan as a receipt in that she produces a minimal response which is usually oriented to as a sequence closing move. Ryan therefore opts not to expand in a context which is understood to be expansion relevant (given the dispreferred second pair part). Hence, Ryan’s non-uptake of this sequential expansion opportunity might be understood as an accountable absence. Indeed Parkinson then expands the post expansion, but again the post expansion is 4

Of course, it is possible to argue that “not usually” is mitigated since the alternative might well have been a simple “no”. This along with the observation already made that the dispreferred SPP was of course deferred by the repair sequence is why I claim that the response is “on balance” aggravated.

Chapter Fourteen

254

ambiguous, this time between a closure relevant assessment (Goodwin and Goodwin1992) and repair initiation. Again Ryan orients to the post expansion as closure relevant; observably orienting to Parkinson’s “you don’t care about it” (line 22) as a Sequence Closing Third (Schegloff 2007), which does not initiate a sequence and hence does not make relevant a response. Instead, Ryan produces a post completion musing (Schegloff 2007) that clearly orients to the sequence as complete. The PCM does not extend the sequence but provides meta-commentary on it and is actually directed at Trinny and Susannah5. What is so interesting about this whole question sequence is the recurringly ambiguous nature of the interactional devices employed by both participants that results in recurring failure of alignment in the interactional practices deployed by the participants. Hence, Ryan adopts an ambiguous repair initiation device to signal pre-disagreement, but Parkinson opts for the structurally preferred (but visually incongruous) interpretation of repair of hearing/understanding. Parkinson twice ambiguously initiates post expansion, but Ryan orients to both post expansion offerings as sequence closure. Each time, the disalignment in the sequence of actions is demonstrated in the next turn orientation by the recipient. Put informally, this is talking at cross purposes but at the level of sequential organization rather than propositional content. Moreover, a similar pattern emerges in the very next sequence shown in Extract 3: Extract 3 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 5

P

M P M P M P M P M P M P

but I mean you’d be somebody of course being the the who you are the star and celebrity and that sorta thing who gets under the microscope of th the me:dia (.) yeahĻ with what you wear and all that sor[ta ] [yeah] what’s your attitude to it (0.8) I: jus’ grin and bear it do you? yeah do you, doesn’t doesn’t bother you no good. o o no alright I see we’re not gonna get very far in this conversation are we=

Visually the PCM is clearly directed at Trinny and Susannah as Ryan turns her physical orientation round to face them, but the only audible response is Parkinson’s.

Ambiguity in the Emergence of Disaffiliation 44 45 46 47

M P M P

255

=yeah hehehe .hhh [ĻLet’s t’] [can’t] go there with you let’s talk about th the film(.)

In this sequence, Parkinson orients to Ryan’s overt closing moves in the preceding sequence as sequence closing but not necessarily topic closing and proceeds to initiate a “second try” (Schegloff 2007, p.173) at proffering the topic of fashion with an extended presequence (lines 28 – 32). Here both parties orient to their institutional roles as interviewer and interviewee with Parkinson producing talk designed to be prefatory to his actual question and Ryan producing minimal passing turns that display an orientation to the requirement that the interviewee not respond until an actual question has been produced (Schegloff 1988/9). This type of question is labelled by Linell et al (2003) as a framing question (S+Q). Clayman and Heritage (2002b) have shown how these prefatory statements in the interview setting serve to narrow down the focus of the final question. The important observation here is that the institutional context thus creates an ambiguity to Ryan’s responses in this sequence, since they have the form of “go-ahead” type responses in a pre-sequence which might be heard as projecting a preferred outcome to the topicproffering sequence (Schegloff 2007), but they might also be heard merely as the collaborative production of the setting as an interview (Schegloff 1988/9). When Parkinson’s question finally appears in line 34, Ryan’s response (line 35) is constructed to be minimal and serves to reject the proffered topic. Hence, it is also produced as a dispreferred second with turn initial delay and. The use of the minimizer adverb “just” (Abell 2010) by Ryan appears to display an orientation to the preference for expansion in the response to a topic-proffer6. At the same time, the minimizing response with “just” indicates that there is nothing more to say. In other words, it is hearable as an overt move to topic closure. As in Extract 1, Parkinson’s response (line 36) to Ryan’s dispreferred SPP is to produce a non minimal post expansion that is ambiguous between receipt and repair and as in Extract 2, Ryan (line 37) opts for the closure relevant interpretation rather than the expansion relevant interpretation and produces a minimal response. Again as in Extract 1, Parkinson expands this post expansion sequence with a further post expansion in line 38 that is ambiguous between a closure relevant assessment (Goodwin 1987) and repair 6 This is similar to the minimising use of “just” observed by Lee and Sheon (2008) in their analysis of accounts of reasons for seeking an HIV test.

Chapter Fourteen

256

initiation and again Ryan opts for the closure relevant interpretation and responds with another minimal response. After then a pair of passing turns (lines 40-41), this time it is Parkinson who produces the metacommentary on the sequence (line 42) orienting to the completion of the sequence before initiating a topic shift and introducing Ryan’s film. Thus in Extract 3, as in Extract 1, the interactional moves by both participants are recurringly ambiguous in their actional and interactional import. The ambiguity at the outset of the sequence in Extract 3 relates to the distinction between sequence closure and topic closure in the preceding sequence. Then Ryan’s passing turns in lines 31 and 33 are ambiguous with respect to their displayed orientation to the projected base adjacency pair as a consequence of the institutional context of the celebrity interview. Ryan’s dispreferred response in the base adjacency pair might be understood as expansion relevant given its production as a dispreferred second, but the minimizer adverb “just” is hearable as counteracting the expansion relevance of the preference marking. And again Parkinson’s post expansion FPPs are both ambiguous with respect to the goals of expansion or closure but oriented to by Ryan as closure relevant. Hence, in Extract 3 as in Extract 1, the ambiguities in the interactional practices drawn on by both participants to manage the preference structures in the ongoing sequence can be seen to result in repeated turn by turn disalignment of the actional moves in the sequence. Thus, what I am suggesting is that the disaffiliation in the data initially emerges as an outcome of these recurring instances of failure of alignment, but it is the ambiguity in the participant orientations and in the available interpretations of the specific interactional practices both for the participants themselves and for the onlooking audience that makes possible the highly contested and highly polarized perceptions of viewers. Extract 4 occurs much later in the interview after both the initial disaligning sequences and several subsequent overtly oppositional sequences and provides a revealing contrast with Extracts 3 and 5: Extract 4 348 349 350 351 352 353 354 355 356 357

P

now that you’re war (hh) Now that you’re wary of journalists: (.) > I mean< d d does it give you an insight in t’ what they’re after. (1.0) M now that I’m wary of them= P =yes (.) you are wary of journalists (.) you’re wary of me you’re wary of th’interview (.) you don’t like being interviewed, you can see it in the way that you sit in the way you are:that you don’t like [doing it] M [tru;e](hhh)

Ambiguity in the Emergence of Disaffiliation

257

What is interesting about this sequence is that we see Ryan in line 352 using the same repetition device in the SPP that we saw in Extracts 3 and 5, but by this stage the context is already one of aggravated opposition, since it follows a number of overtly oppositional sequences. This time, Parkinson does not respond with the structurally preferred interpretation as repair of hearing or understanding. Instead, Parkinson directly orients to Ryan’s use of repetition as an oppositional move, specifically as an other initiation of repair displaying disalignment. In orienting to the repetition in this way at this point in the interaction, Parkinson is not only orienting to Ryan’s utterance as disaligned but is also clearly orienting to the context as one of disalignment and disaffiliation and reproducing it as such. Since Parkinson’s descriptions are hearable as criticisms of his coparticipant, it is noteworthy that they are produced as preferred rather than with the expected dispreferred format (Pomerantz 1984). This again points to orientation to the context as one of aggravated opposition. Similarly, Ryan’s response in line 357 while clearly disaffiliative in stance is produced as an aligned preferred agreement to Parkinson’s series of critical descriptions in lines 353-356, again orienting to the context of aggravated opposition. Thus, the sequence is clearly disaffiliative, hence the interactional practices are not oriented to as ambiguous here, they are produced and oriented to as disaligned and both participants orient to the context as one of aggravated opposition.

14.5 The institutional context Extract 3 introduced the relevance of the institutional context to the interpretation of the interactional moves deployed by the interview participants. This raises the interesting question of where else and how else the institutional context might be relevant to perceptions of the interactional practices. Participants in institutional talk-in-interaction are observed to orient to three basic distinguishing elements of the institutional nature of the talk: the specific goals tied to their institutional identities (here Chat Show Host and Celebrity Interviewee); the specific constraints on the sequential and actional organization of participation; the particular inferential frameworks and procedures associated with specific institutional contexts(Drew and Heritage 1992). Institutional CA has been fairly extensively applied to the news interview context (e.g. Clayman and Heritage 2002a), but less so to other genres of televized interview such as the celebrity interview, which Ilie

258

Chapter Fourteen

(2001) suggests might better be understood as “semi-institutional”. As pointed out by Koskela (2005), the celebrity interview shares some organizational features with other televised interview genres, such as orientation to the centrality of question – answer sequences, and orientation to the production of the talk for an overhearing audience as evidenced for example in the organization of both opening and closing sequences of the interview. However, the institutional goals of the celebrity interview are different. Ilie (2001) describes the goal of talk shows as “infotainment”. Norrick (2010:526) suggests that the focus of the celebrity interview is “human interest” and that “the interviewer must try to coax the celebrity into divulging personal information which he or she may or may not choose to ‘confess’”. Crucially one well recognized goal of the celebrity talk show interview is the collaborative production of a FRQVHQVXDO SRLQW RI YLHZ E\ WKH WZR SDUWLHV LQYROYHG 0DUWÕQH]  Lauerbach 2007; Norrick 2010). Moreover, Llewellyn and Butler’s (2010) examination of on air celebrity walkouts is argued to reveal an “institutional bias against heated and personalized interaction” (Llewellyn and Butler 2010: 62). These different (semi-)institutional goals thus can be seen in the way the interview is organized. For example, orientation to the asymmetrical speaking rights of the formal question–answer structure sometimes gives way to the give-and-take of everyday conversation (Clayman and Heritage, 2002a:108–110) relaxing the constraints that associate the specific institutional identities of the participants (Host/Interviewer and Guest/Interviewee) with the specialized turn-taking system of the interview setting. Thus far I have examined the disalignment in the interaction between Ryan and Parkinson with minimal attention to the (semi-)institutional nature of the interview setting, and hence with minimal attention to their institutional roles and goals and the constraints on their participation in the interaction. However, it is clear that for the interview participants themselves the relevance of the interaction as a celebrity interview is not merely implicitly but also explicitly oriented to on several occasions and hence should be examined for its relevance to the perception of disalignment/disaffiliation in the interview. Starting with Extracts 1 and 3 as the initial sequences of the interview in which the disaffiliation begins to emerge, we see both Ryan and Parkinson orienting to their respective roles as interviewer (IR) and interviewee (IE) in their adherence to the constraints on asymmetrical participation in question answer sequences. Parkinson asks the questions and Ryan answers them. It might also be argued that the recurring use of

Ambiguity in the Emergence of Disaffiliation

259

ambiguous discourse practices such as repetition to manage the preference issues in the talk demonstrates an orientation to the normative institutional expectation of a collaboratively produced consensual point of view (in contrast to Extract 4). However, we also see that although Ryan adheres to her institutional role as interviewee and the associated constraints on participation, she does not always orient to the institutional goal of “infotainment”. This is particularly evident in her rejection of topic proffers and in her production of minimal responses in the expansion relevant context of post expansion following a dispreferred SPP. Rejection of topic proffers and minimal responses can clearly be heard as an evasion of the key institutional goal of talk show interaction, namely getting the celebrity to talk. This explains the recurring perception that Ryan was unwilling to answer questions, despite many lengthy responses, since the lack of expansion in expansion relevant contexts will of course be heard as an accountable absence of expansion. Extract 5 provides an example of an overtly oppositional sequence involving disalignment and disaffiliation by both parties rather than the normatively expected consensus: Extract 5 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103

P M

P M P M P P

M P P

(du du do’you) do you share that view, (0.2) Jane (Campion’s) view. (0.3) tch(.) actually yea I mean I think what she’s made a movie about really is about (0.2) a sorta de-bunking of uh the current western romantic mythology that’s in place which is that there is such a thing (.hhh) as happily ever after or your prince charming will come (.) y’know. She is saying that most people (.) have a frustrated relationship with that (0.3) with that myth ‘n that indeed it is a myth ‘n if you are present and looking for truth you’ll be more satisfied. eh that’s a bleak view of life isn’t it?= =I think it’s beau::tiful what beauty (???) from that? the [the search for]truth? It’s a beautiful thing [I mean you can’t] well no no that’s not a search for truth it’s a search for cynicism and disenchantment >I mean< are you saying she’s saying do you think romance has more validity and [more] honesty in it than than love? [whh]romance romance has inspired great great great movies great great poetry great great[music]

260 104 M 105 106 P

Chapter Fourteen [Well I think] it’s very (.) I think I can distinguish between [love and romance] [eh(0.3) great sex never did]

In this extract, Parkinson produces an overtly critical assessment of Ryan’s film director’s position on the difference between love and romance as articulated by Ryan and this retrospectively turns Ryan’s explanation into the first move in an Action-Opposition sequence. Ryan responds with a latched oppositional move, a second assessment (Pomerantz 1984). Parkinson produces an oppositional post expansion FPP both in line 93 and in line 96. This has the effect of creating an extended argumentative sequence with a chain of oppositional moves by both parties. Following Hutchby (1996), this is a fairly typical extended argument sequence. Moreover, the oppositional moves in this sequence are produced as aggravated moves, in other words, they are produced without the discourse markers that mark a conversational action as dispreferred in that sequence. As mentioned above, this absence of markers of dispreference displays an orientation to the context as one of “aggravated opposition” (M.H. Goodwin 1983) rather than the institutionally expected collaboratively produced consensual point of view. The sequence begins with a preferred expanded response by Ryan to Parkinson’s topic proffer in line 80-81, as such Ryan appears here to adhere to the institutional goal of extended talk. However, it is clear in this sequence that Parkinson departs from the expected consensus building goal of his institutional role in the production of a series of aggravated oppositional moves. Moreover there is an escalation in his oppositional moves from line 91 to line 96. He begins with an oppositional assessment in line 91 which is relatively mitigated with a turn initial delay (“eh”) and the turn final tag question “isn’t it?”. His next turn however is a direct, unmitigated challenge in line 93 and finally he produces extended unmitigated opposition in lines 96-98. In this context, there is a conflict in the institutional goals placed on Ryan, since local institutional interactional goals require consensus building and hence would require Ryan to agree with Parkinson. However, to do so would be counter to the overarching goal of the celebrity interview to promote, and in this context defend, her new film. It is interesting and worthy of further analysis that Ryan in this context, prefaces her oppositional moves in lines 92 and 104 with “I think” perhaps as form of footing (Goffman 1981) that claims personal and individual affiliation with the assessments expressed. This personalized footing perhaps displays an orientation to the conflicting institutional goals at this point. Ryan also departs from institutional constraints on interview

Ambiguity in the Emergence of Disaffiliation

261

participation by producing a question in line 99; however, as mentioned above, this is not unexpected in the semi-institutional organization of chat show talk. Thus whilst both participants depart from the collaborative consensus building expected in this institutional context, it is perhaps a more unequivocal departure from the institutional constraints of the celebrity interview on the part of Parkinson as interviewer than on the part of Ryan as interviewee.

14.6 Conclusion This paper set out to examine how the disaffiliation in this notoriously unsuccessful interview initially emerges in the talk of the participants, focusing on whether a micro level analysis of the interactional practices that produced the disalignment in the talk might also shed light on the contested perceptions of the accountability of the two participants in the interactions. What the analysis reveals is the subtle, context sensitive, interactional and co-produced nature of emerging disalignment and disaffiliation. It is suggested that it is the complex interplay between the sequential organization of the production of dispreferred responses, the forms of initiation of repair and orientation to expansion relevant sequences that contributes to the availability of such contested evaluations of the interview participants’ roles in the disalignment/disaffiliation. The disalignment in the initial sequences of the interview is primarily produced through interactional moves that have been shown to be ambiguous in their actional and interactional import. These ambiguous interactional practices are drawn on by both participants to manage the preference structures, including issues of repair initiation and expansion, in the ongoing sequence, but they are shown to result in repeated turn by turn disalignment. The displayed understanding by the recipient of each turn is oriented to an interpretation that the speaker in their subsequent turn demonstrates was not the intended interpretation. However, interestingly the disalignment is not explicitly oriented to by the participants in that they do not overtly orient to a failure of intersubjective understanding, for example by initiating repair. This leaves unaddressed the question of whether the disalignment is a product of resistance or a failure of understanding. Moreover, the question is unaddressed not merely for the participants, but also for the analyst and for the viewing audience. The disaffiliation in the data thus initially emerges as an outcome of these recurring instances of failure of alignment that ultimately produce a context of aggravated opposition.

262

Chapter Fourteen

What I am suggesting is that both participants deploy ambiguous interactional practices both to manage the interaction and to display participant orientation to the interaction. Moreover, the point is that these ambiguous practices are ambiguous not merely for the participants themselves but also for the onlooking audience. The highly contested and highly polarized evaluations of the participants by viewers can thus be understood at least in part as an effect of these ambiguities in the available interpretations of the specific interactional practices. The analysis also suggests that the specific properties of the semi-institutional context of a celebrity interview have a role to play in the perceptions of the participants’ roles in the emerging disaffiliation. Both participants are shown to breach specific institutional norms relating to the institutional goals of infotainment and consensus. Both participants are shown to produce overt oppositional moves which breach the normative expectation for consensus. This explains the impression of rudeness on both their parts but Parkinson's role as Host places his oppositional moves in first and third position and hence leaves him potentially more accountable for the failure to achieve consensus, perhaps explaining why “rudeness” was more widely attributed to him. On the other hand, the accountable absence of expansion by Meg Ryan in a range of expansion relevant contexts also provides a source for the impression that she was “monosyllabic, obstructive and ill-mannered” despite her frequently lengthy turns. Thus this single case analysis of a failed talk show interview shows the subtle interactional and co-produced nature of disalignment. It also shows how normative expectations created by both the sequential context and the semi-institutional nature of the interaction interact with the form, function and propositional content of participant responses to influence perceptions of the participants in interaction. This underlines the view that we need to engage in formal, micro-level turn-by-turn examination of situated interaction in order to understand macro issues of communication and perceptions of communication in institutional contexts.

References Abell, Jackie. 2011. ‘They seem to think 'We’re better than you'’: Framing football support as a matter of ‘national identity’ in Scotland and England. British Journal of Social Psychology 50(2). 246–264. Bolden, Galina. 2009. Beyond answering: Repeat-prefaced responses in conversation. Communication Monographs. 76(2). 121–143.

Ambiguity in the Emergence of Disaffiliation

263

Clark, Herbert H. and Edward F. Schaefer. 1987. Collaborating on contributions to conversations. Language and Cognitive Processes 2(1). 19–41. Clayman, Stephen E. and John Heritage. 2002a. The news interview: Journalists and public figures on the air. Cambridge: Cambridge University Press. —. 2002b. Questioning presidents: Journalistic deference and adversarialness in the press conferences of U.S. Presidents Eisenhower and Reagan. Journal of Communication 52(4). 749–775. Drew, Paul. 1997. 'Open' class repair initiators in response to sequential sources of troubles in conversation. Journal of Pragmatics 28(1). 69– 101. Drew, Paul and John Heritage. 1992. Talk at Work. Cambridge: Cambridge University Press. Goodwin, Charles and Marjorie H. Goodwin. 1992. Assessments and the construction of context. In A. Duranti and C. Goodwin (eds.), Rethinking Context: Language as an Interactive Phenomenon, 147– 189. Cambridge: Cambridge University Press. Goodwin, Marjorie H. 1983. Aggravated correction and disagreement in children’s disputes. Journal of Pragmatics 7(6). 657–677. Hutchby, Ian. 1996. Confrontation Talk: Argument. Asymmetries and Power on Talk Radio. New Jersey: Lawrence Erlbaum. Koskela, Heidi. 2005. Invoking Different Types of Knowledge in Celebrity Interviews. SKY Journal of Linguistics 18: 93–118. Koshik, Irene. 2005. Alternative questions used in conversational repair. Discourse Studies 7(2). 193–211. Ilie, Cornelia. 2001. Semi-institutional discourse: The case of talk shows. Journal of Pragmatics 33(1). 209–254. Lauerbach, Gerda. 2007. Argumentation in political talk show interviews. Journal of Pragmatics 39(8). 1388–1419. Lee, Seung-Hee and Nicolas Sheon. 2008. Responsibility and risk: accounts of reasons for seeking an HIV test. Sociology of Health and Illness 30(2).167–181. Linell, Per., Johan Hofvendahl and Camilla Lindholm. 2003. Multi-unit questions in institutional interactions: Sequential organizations and communicative functions. Text 23(4). 539–571. Llewellyn, Nick and Carly Butler. 2011. Walking out on air. Research on Language and Social Interaction 44(1). 44–64. 0DUWÕQH] (VSHUDQ]D 5DPD  $FFRPSOLVKLQJ FORVLQJV LQ WDON VKRZ interviews: a comparison with news interviews. Discourse Studies 5(3). 283–302.

264

Chapter Fourteen

Mazeland, Harrie and Minna Zaman-Zadeh. 2004. The logic of clarification. Some observations about word-clarification repairs in Finnish-as-a-lingua-franca interactions. In Rod Gardner and Johannes Wagner. J. (eds.), Second Language Conversations, 132–156. London: Continuum. Norrick, Neal. 2010. Listening practices in television celebrity interviews. Journal of Pragmatics 42(2). 525–543. Pomerantz, Anita. 1978. Compliment responses: Notes on the cooperation of multiple constraints. In J. Schenkein (ed.), Studies in the Organisation of Conversational Interaction, 79–112. New York: Academic Press. —. 1984. Agreeing and disagreeing with assessments: Some features of preferred/dispreferred turn shapes. In J. M. Atkinson and J. Heritage (eds.), Structures of Social Action: Studies in Conversation Analysis, 57–101. Cambridge: Cambridge University Press. Robinson, Jeffrey D. and Heidi Kevoe-Feldman. 2010. Using full repeats to initiate repair on others’ questions. Research on Language and Social Interaction 43(3). 232 – 259. Sacks, Harvey. 1987. On the preferences for agreement and contiguity in sequences in conversation. In G. Button and J.R.E. Lee. (eds.), Talk and Social Organisation. 54–69. Cleveland: Multilingual Matters. —. 1992. Lectures on Conversation. Oxford: Blackwell. Schegloff, Emanuel A. 1988. On an Actual Virtual Servo-Mechanism for Guessing Bad News: A Single Case Conjecture. Social Problems 35(4). 442–457. —. 1988/89. From interview to confrontation: observations of the Bush/Rather encounter. Research on Language and Social Interaction 22: 215–240. —. 1990. On the organisation of sequences as a source of 'coherence' in talk-in-interaction. In B. Dorval (ed.), Conversational organization and its development, 51–77. Norwood. N J: Ablex. —. 1992. Repair after next turn: The last structurally provided defence of intersubjectivity in conversation. American Journal of Sociology 97(5).1295–1345. —. 1997. Practices and actions: Boundary cases of other-initiated repair. Discourse Processes 23(3). 499–545. Schegloff, Emanuel A. . 2007. Sequence Organisation in Interaction: A Primer in Conversation Analysis Vol. 1. Cambridge: Cambridge University Press .

Ambiguity in the Emergence of Disaffiliation

265

Schegloff, Emanuel A., Gail Jefferson and Harvey Sacks. 1977. The preference for self-correction in the organization of repair in conversation. Language 53(2). 361–82. Sorjonen, Marja-Leena. 1996. On repeats and responses in Finnish conversations. In Elinor Ochs, Emanuel A. Schegloff and Sandra A. Thompson (eds.), Interaction and Grammar, 277–327. Cambridge: Cambridge University Press. Stivers, Tanya. 2008. Stance. alignment and affiliation during story telling: When nodding is a token of preliminary affiliation. Research on Language in Social Interaction 41(1). 29–55. Svennevig, Jan. 2004. Other-repetition as display of hearing. understanding and emotional stance. Discourse Studies 6(4). 489–516. —. 2008. Trying the easiest solution first in other-initiation of repair. Journal of Pragmatics 40(2). 333–348.

CHAPTER FIFTEEN “IS THAT THE WAY I LOOK LIKE”: HEGEMONIC FEMININITY AS AN INTERACTIONAL ACCOMPLISHMENT CATRIN S. RHYS AND GRACE BURKE 15.1 Introduction This paper1 examines how gender was made relevant in naturally occurring talk about smoking between female friends who smoke. Previous social science research relating to women smokers is dominated by three themes: body image as a motivating factor in smoking; the role of smoking in adolescent identity construction; the impact of motherhood on women who smoke. This emphasis is perhaps an inevitable reflection of public health policy concerns but the top down focus on public health issues reduces women to “bodies” and “mothers” and potentially leaves unexamined women’s own concerns. In particular, adult women smokers who are not mothers are rendered somewhat invisible to researchers, since they are neither adolescents whose motives for taking up smoking are of interest, nor mothers whose children need to be protected from smoking. The data in this study comes from a corpus of naturally occurring conversations among such adult women smokers. As such, the conversations are not only about smoking and the talk about smoking is not directed in any particular way. The issues that emerge are thus those that the participants themselves produce as relevant. One issue that emerges as relevant and highly gendered for the participants in these conversations is the question of appearance. However, it is neither the physical sense of appearance nor the identity construction sense of appearance addressed in previous research. Rather, in talk that is ostensibly about smoking and not about gender, the participants are shown 1

A version of this paper is also forthcoming in Discourse Studies (2014).

Hegemonic Femininity as an Interactional Accomplishment

267

to orient to the impact of smoking on their public performance of gender/femininity. The analysis of participant orientation to issues of gender in this data thus addresses a concern expressed by Speer (2001a, p.130 fn 9.) that research should address “how gender is made relevant and oriented to in materials ... where gender is not explicitly indexed by the researcher”. The ethnomethodological examination of this orientation to issues of appearance presented here lays bare tacit knowledge of gendered norms and ideologies that are interwoven and negotiated across turns at talk. The focus on participant concerns in the analysis reveals how shared gender ideologies impinge on the women’s categorizations of themselves and other women who smoke. Finally, the fine-grained, technical analysis of the moment-by-moment orientation to categories that is afforded by Membership Categorization Analysis is shown to provide support for a participant orientation approach to the concept of “hegemonic femininity” (Schippers 2007).

15.2 Membership Categorization Analysis The analysis presented here applies the ethnomethodological/ conversation analytic (EM/CA) perspective of Membership Categorization Analysis (MCA) (Sacks 1992) in order to examine the demonstrable relevance of the category “smoker” for participants in their categorization of themselves and others and reveals the implicit normative assumptions relating to the activity of smoking that speakers draw on as they negotiate category incumbency. MCA focuses on revealing how participants in interaction display shared, taken-for-granted cultural knowledge and common sense understanding through the discursive organization and production of social categories in talk. Sacks (1992) suggested that categorizations are conventionally organized into “natural collections” which he called Membership Categorization Devices (MCDs). In their “reconsidered” model of MCA, Housley & Fitzgerald (2002) draw on Watson (1978) and Hester (1994) to argue for an “occasioned” model of MCA in which “the notion of ‘natural collections’, and the notion of stable external bodies of clearly identifiable common-sense collections, is eschewed in favour of a sensitivity to the in situ and locally occasioned character of members’ category work” (Housley & Fitzgerald 2002, p.69). This focus on the locally occasioned organization of category work also allows for a more sensitive focus on how predicates are conversationally tied to categories and devices by members as a localized (i.e. in situ) practical interactional achievement.

268

Chapter Fifteen

At the heart of MCA is the recognition that social categories are not neutral labels but are inference rich in that “a great deal of the knowledge that members of a society have about the society is stored in terms of these categories” (Sacks 1992, p 40). This knowledge includes the various predicates that can be imputed on the basis of category membership, including category-bound activities (CBAs) (Sacks 1992), entitlements, obligations, knowledge etc. (Watson 1978). Moreover, it is “tacit knowledge” that is revealed through the normative expectations that are associated with a particular category and available as an inferential resource to interpret the actions of members of that category (Wooffitt 2006, p.100). Jayyusi (1991) furthermore points to how that knowledge is “both morally constituted and constitutive of moral praxis” (p. 240) MCA thus allows us to track how speakers produce situated categories as relevant at different times and negotiate the acceptability of some categories by invoking category attributes and pointing to violations of category bound activities to warrant their complaints. In this way, participants’ formulations of gender appropriate behaviour are interwoven and negotiated across their turns at talk and similarly breaches of gender norms are thus also potentially oriented to and displayed in turns at talk.

15.3 Hegemonic Femininity The concept of hegemonic femininity is often simply equated with gender norms, “the socially sanctioned and accepted views of femininity” (Gaucher 2011), or with the idea that there are dominant and subordinate forms of femininity (Pike & Johnson 2003). Schippers (2007), however, develops a more nuanced theoretical model of hegemonic femininity which “provides a definition of femininity that situates it, along with masculinity, in gender hegemony and allows for multiple configurations” and is “empirically useful for identifying how masculinity and femininity ensure men’s dominance over women” (p.89). Schippers’ model of hegemonic femininity extends and re-works Connell’s work on masculinities (Connell 1995, 2000; Connell & Messerschmidt 2005). Connell’s conception of hegemonic masculinity focuses on social practices and characteristics that are understood to be masculine and, when embodied by men, guarantee men’s domination over women. Connell’s model of hegemonic masculinity is thus intended as an account not merely of gender norms but of gender hegemony. Central to Connell’s account of gender hegemony is the idea of the dominance of one particular form of masculinity over other subordinate and marginalized masculinities. In other words, Connell suggests that there are alternative

Hegemonic Femininity as an Interactional Accomplishment

269

practices and characteristics that are identified as masculine but do not perpetuate male dominance and are held as inferior to the hegemonic masculine ideal. In Connell’s model, subordinate masculinities are associated with femininity and marginalized masculinities with subordinated social classes or racial/ethnic groups. Schippers (2007) points out that these distinctions along with Connell’s emphasis on masculinity and femininity as social practice, render it difficult to distinguish between different masculinities and different groups of men. In drawing on Connell’s model of hegemonic masculinity in order to develop a theoretical conceptualization of “femininity”, Schippers rejects Connell’s emphasis on the centrality of social practice and emphasizes instead the symbolic nature of gender categories. She differentiates between masculinity and femininity as “contextually and culturally specific sets of meanings for what men and women are and should be” and social practice as the “the mechanism by which those meanings come to shape, influence and transform social structure” (p.92). Drawing on Butler (1990), she argues that the hegemonic significance of masculinity and femininity relies on hierarchical complementarities in the symbolic relationship between the gender categories that provide a rationale for social relations between men and women. While Connell and Messerschmidt (2005) argue that a reduction of masculinity to the realm of the symbolic ignores the ways in which gender relations are constituted through non-discursive practices, Schippers argues that to do otherwise is to conflate gender relations with masculinity and femininity and thereby reduce their explanatory capacity as a “rationale for practice”. Schippers also counters Connell and Messerschmidt’s suggestion that a reduction of masculinity and femininity to the symbolic leads to a static conception that conflates behaviour and norms and ignores variation, arguing that masculinity and femininity are produced, contested and transformed through discursive processes” (p.94). Schippers (2007) thus arrives at a conceptualization of hegemonic femininity as “the characteristics defined as womanly that establish and legitimate a hierarchical and complementary relationship to hegemonic masculinity and that, by doing so, guarantee the dominant position of men and the subordination of women” (p.94) Moreover, she argues that this relationship of ascendancy holds not only for hegemonic masculinity over femininity, but also for hegemonic femininity over other forms of femininity. These other forms of femininity involve characteristics/ practices which, when embodied by women, deviate from those defined as feminine and so constitute a refusal to complement hegemonic masculinity in a relation of subordination. Schippers calls these pariah femininities

270

Chapter Fifteen

instead of subordinate femininities “because they are deemed, not so much inferior, as contaminating to the relationship between masculinity and femininity” (p.95) Schippers’ emphasis on the symbolic and discursive nature of femininity and masculinity points to a need for empirical research that locates their production in everyday interactional practices. In an exchange between Speer (2001a,b) and Edley (2001) that pits Wetherell and Edley’s critical discursive reformulation of Connell’s concept of hegemonic masculinity against Speer’s (2001) participants’ orientations approach, Speer rejects the concept of hegemonic masculinity “because it is an abstract and decontextualized notion, which does not capture the subtleties and fluidity of identity work” (2001b p.142). In Speer (2005), she adopts a slightly different stance: I am not ‘rejecting’ or ‘discarding’ the concept of hegemonic masculinity (cf. Edley 2001b: 136) or suggesting that ideology and oppressive gender norms do not exist. Nor am I suggesting that what hegemonic masculinity describes ‘is entirely absent from everyday talk (Edley 2001a: 3). Instead, I wish to problematize analytic approaches which reify such concepts, and which treat them as determining members’ talk or the discourses that are available to them in a top-down fashion. The concept of hegemonic masculinity is, by its very nature, an abstract ‘uncapturable’ yardstick, used to describe something which extends beyond members’ local practices and actions. (p.148-9)

In her analysis of interview data in which men talk about their own masculinity, Speer argues that hegemony and hegemonic masculinities are not participants’ categories. Instead, she emphasizes the variability in the constructions of masculinity that the participants in her data make use of for different rhetorical effects targeted at the local interactional context. However, as discussed above, work in MCA (including Speer’s own work) has clearly shown that restricting the analytical focus to participants’ orientations can provide a very powerful demonstration of how gender categories are not only produced for local rhetorical effects but are also oriented to by the participants as short cuts to shared cultural and moral knowledge of the norms for gender appropriate behaviour (Stokoe 2003, 2012). As such, it is not clear that one should a priori rule out the possibility of participant orientation to hegemony as part of that shared cultural and moral knowledge. In other words, the question is whether participants can be observed to orient to predicates associated with a gender category as hegemonic and what that would look like.

Hegemonic Femininity as an Interactional Accomplishment

271

In Schippers’ model, hegemonic femininity is articulated as an available rationale for social practice, constituted through the proliferation of a network of discourses and institutionalized in gender relations. This points to the kind of top-down post-structuralist conception of the discursive and hence of hegemonic femininity that Speer (2001a,b, 2005) objects to, conceived as available outside the discourse and constraining the production of discourse. Elsewhere, Schippers emphasizes the local, context bound nature of hegemonic femininity (contextually specific features, locally defined femininities) and the importance of empirical examination of practice to understanding what characteristics are deployed by participants as a rationale for practice. These aspects of her approach point to the possibility of a participant orientation based account of hegemonic femininity. Given Schippers’ definition of hegemonic femininity and pariah femininities, such an account would need to reveal participant orientation to hierarchical complementarity in gendered categories and predicates and the stigmatizing of women conversationally associated with predicates that are oriented to as masculine.

15.4 Data and Participants The data analysed here was drawn from the corpus of a larger project about moral accountability and the discursive actions that speakers produce in everyday talk about their smoking experiences (Burke, 2011). Although, the research was initially prompted by a preanalytical “noticing” that talk about smoking mundanely occurs, it is clearly impossible to predict when mundane (as opposed to institutional) “smoking talk” is going to happen2. This, of course, presents a conundrum to the EM/CA researcher interested in examining naturally occurring talk; how do you record naturally occurring talk about smoking when its occurrence is so unpredictable? The “research conversations” (Roulston et al, 2001: 749) examined in this paper represent a compromise between fully naturally occurring talk and research led data collection in that the second author, a smoker herself, recorded naturally occurring (i.e. non research elicited) gatherings of her own friends who were also smokers3. 2

Initial attempts at recording, for example, were carried in designated smoking areas, having overheard such talk in this context prior to the research. However, the talk recorded was more focussed on (the highly contextually relevant) weather conditions rather than on “smoking/smoker issues”. 3 The potential criticism that the identification of the participants as smokers constitutes an a priori analytic categorization was addressed by overtly applying the viewer’s maxim (Sacks, 1992) which stipulates that “if a Member sees a

272

Chapter Fifteen

The compromise was that the researcher as participant in the conversation initially introduced the topic of smoking but also made a point of patrolling her input into the interactions4. The analysis then aimed to minimize the effect of this compromise by close attention to the sequential properties of the interaction, focusing attention on those sequences in the interaction that are not initiated or directed by the researcher5. The interactions used in the analysis thus involve talk between two different groups of friends, and although “smoking” was sometimes focussed on as a topic, it was not pursued as the “be-all and end-all” of the interaction. The first interaction (Transcript: Night Out) involved four women in their mid-twenties (including the second author), later joined by a fifth, having a drink prior to a night out. The recording lasted approximately one and a half hours. The second recording (Transcript: Cat’s House) took place at the house of one of the participants (Cat). The second author and two friends, in their mid-to-late twenties, were present and the recording lasted approximately one hour. At the time of the recordings, all the participants were smokers or ex-smokers (who nonetheless smoked (see Rhys and Burke (forthcoming) for discussion of the complex relationships between activities, predicates and categorizations related to smoking)). In both cases, such gatherings were a regular occurrence for these groups of friends; hence they were not oriented to by the participants as organized for the purpose of research. The extracts selected for analysis here are not the only extracts in which gender is made relevant by the participants. They were selected because they revealed a particularly distinctive sequential pattern in which locally occasioned gender ideologies are produced and oriented to as a category-bound activity being done, then, if one can see it being done by a member of a category to which that activity is bound, see it that way” (p. 259). In other words, if someone was seen to be smoking, then for the purposes of data collection at least, they could be categorized as a smoker. This, of course, still yields an unexamined Members’ categorization and it should be noted that Rhys & Burke (forthcoming) reveals a considerably more complex relationship between the activities, predicates and categorizations associated with smoking than this simple application of the viewer’s maxim suggests. 4 The researcher does participate in the interaction since to remain silent amongst friends would be unnatural, but she aims to avoid directing the interaction. Other recordings were also made but are not analysed here since they did not come off as being particularly “natural” or “conversational” in nature despite also involving friends, probably because they involved only conversational dyads and hence inevitably required greater input from the researcher who ended up producing the majority of First Pair Parts as might be expected in an interview setting. 5 The researcher is identified as Dawn in the transcriptions

Hegemonic Femininity as an Interactional Accomplishment

273

sense making resource in the participants’ local negotiation of their own incumbency of the category “smoker”.

15.5 Analysis This first extract appears fairly early on in the conversation between the three friends Cat, Mag and Dawn (the researcher) and reveals clear orientation on the parts of both Mag and Cat to how they are seen as smokers; that is, the potential negative assessments associated with the category label “smoker”: Extract 1 Transcript: Cat’s House 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280

Cat Mag Cat

Mag Cat Mag Cat Mag

Cat

Sp? Cat Mag Cat Mag

Cat Mag Cat

love my wee smoke (.) but hate the fact that I do smoke aye .hhh °uh[uh° [and I hate (.)that(0.5)wile anti-social part °see th-° it’s got mĹore now wile embarrassin’ ((takin it out of your mouth???)) mmm naw [((??????)) [aw I know (.) Ĺaye I’ve felt that a coupla times too as many people now [((really??)) [aye (.)that don’t smoke(.)an’ yi’ be sorta worried in case they sorta (.) change their mind about yi’ or (.) [ don’t like yi’ as much [uhum uhum even before the smokin’ ban stcame in I used teĹhate now if I went to a restaurant >>I woulda asked for non- smokĻin’>yous never would-a smoked if yous went out for dinner?>john’d smoke after>imagine goin’ inte a restaurant now and seein’ somebody smokin’>I woulda asked for non- VPRNĻLQ¶”. Here Cat describes adopting an activity associated with non-smokers (“asking for nonsmoking”), in order to avoid at least her appearance in the problematic categorization of “smoker”. Both participants here are oriented to the inference-rich nature of the category of “smoker” that ties in their shared cultural knowledge (Jayyusi 1984) and enables them to make sense of this sequence as an elaboration of Cat’s initial problematizing of her categorization as “smoker”. We also begin to see in this sequence a hint towards evidence that this problematic categorization of smokers and smoking may be gendered, through Cat’s contrasting production of her male partner, John, who smokes with abandon in this public setting: “john’d smoke after VWĹDUWHU after main course after dessert an:” At this stage, the production of gender as “the relevant thing” (Edwards, 1998 p.15) is at most implicit in the contrast between Cat and John. However, in the sequence immediately following extract one, Dawn calls Cat to account for her embarrassment at smoking and in this sequence, the gendered nature of this negative assessment of smokers is made explicitly relevant. Extract 2 Transcript: Cat’s House 281 Mag [((??????????)) 282 Daw but [wĹhy would yi’ be embarrassed? 283 (1.5) 284 Cat it’s just a wile dirty auld habit = 285 Mag = aye (0.5) [I would] never walk about (.)

6

Wile is a Derry English intensifier equivalent to very in Standard English.

Hegemonic Femininity as an Interactional Accomplishment 286 287 288 289 290 291 292 293 294 295 296 297

Cat Mag Mag Cat Cat Mag Cat Mag Cat Mag Cat

275

[it is] see pe-girls that walk about the town or whatever [or walk into Derry wi’ fags [John’s mammy walked (.) walked along (.) we were in Belfast shoppin’ Ĺsmokin’ = =b’t yi’see bĹoys [I don’t think’s as bad(.)girls doin’ it [naw [I think’s(.)just really [ahhah Ĺre:ally hĹorrible lookin’ I think it’s wile (.) trampy lookin’ or °suh-hin° aye

Cat’s response to the question of embarrassment is produced as dispreferred, marked by a fairly lengthy pause and the minimizer adverb just (Abell 2010). The use of just here appears to display an orientation to an understanding that Dawn’s question is looking for a more extended response, similar to a minimizing use of just observed by Lee and Sheon (2008) in their analysis of accounts of reasons for seeking an HIV test. At the same time, the minimizing response with “just” indicates that there is nothing more to say. Mag, however, immediately latches on to her talk and proceeds to provide a more elaborate account: “I would never walk about (.)”. Here Mag produces an activity “walk about” with implied reference to public space and in the process contests her own categorization as member of the category of “embarrassin’” smoker through the contrast created by the extreme case formulation “I would never” (Pomerantz 1986). Mag thus orients to “smoker” as a membership categorization device encompassing both negatively appraised (“embarrassin’”) smokers, who smoke publically, and acceptable smokers, who by implication smoke privately and she aligns herself with the latter by rejecting incumbency of the former. This rather implicit category work is upgraded by Mag's immediate production of a categorial formulation: “see pe-girls that walk about the town or whatever [or walk into Derry wi’ fags”. Here Mag makes explicit the previously implied category of unacceptable smoker who walks about town smoking. Moreover the self-repair from the linguistically nongendered term “people” to the linguistically gendered term “girls” makes relevant the membership categorization device “gender” (Stockill and Kitzinger, 2007: 233). Stokoe (2006) highlights how males and females are vulnerable to “gender assessment” if they do not live up to normative conceptions of femininity or masculinity. Here the juxtaposition of the category “girl” and the activity “walk about town with fags” are produced in situ as disjunctive, as a breach of the normative constraints on gender

276

Chapter Fifteen

appropriate behaviour, reinforcing and refining the in situ production of the categorization of the socially unacceptable smoker. Stokoe (2012) points out that “if a category-feature formulation “works”, that is, it does not become the object of repair, then it works on the basis that speakers share category knowledge and unspecified inferences enough to progress the sequence underway”. Here, the categorial formulation is not merely not repaired, it is supported by Cat's production of a “real-life” example. She invokes her husband’s mother as exemplary of this type of smoker and stresses how she “walked” along in Belfast smoking. The incredulity prompted by this action is hearable in the stress placed on her repetition of the word “walked” and the marked pitch rise on “smokin” (line 290). Thus, participants treat each others’ descriptions as “category-resonant” (Stokoe 2012, p.287) and when Mag produces two clearly gendered category-feature formulations in lines 291 296, Cat collaborates in the formulation through the production of agreement markers “naw” (no), “ahhah” (yes) and “aye” (yes). What is particularly interesting about these agreement markers is that they are produced in overlap with Mag's production of the relevant features for each category. This reveals that Cat is able to anticipate the features that Mag produces on the basis of the category labels alone, displaying shared cultural knowledge about the gender distinction that Mag is referencing. In other words, Cat treats the gender categories (“boys”, “girls”) not merely as inferentially rich but as a short cut (Stokoe 2012) to shared social knowledge about the locally occasioned relevance of gender to the assessment of smokers who smoke in public. Moreover, the production of the contrasting categories leaves no ambiguity, for either participant or analyst, that it is the MCD “gender” that is invoked here by the category labels “boys” and “girls” (Gardner 2012). Thus, in this sequence, the speakers have collaboratively displayed shared knowledge of the gender-implications of smoking, and they end up with a shared understanding of the type of girls who smoke in the street, namely “wile trampy lookin” girls. Here the category work produced by Mag is constitutive of a highly gendered moral order (Jayyusi 1984, Stokoe 2003) through which the participants engage in what Nilan (1995, p.92) describes as “profoundly ordinary processes of social identity boundary maintenance work”. Thus Mag resolves the disjunctiveness of “girls that smoke in the street” by drawing on an alternative MCD for girls, the hierarchically organized MCD “moral types of women” (Wowk 1984:77), conversationally tying the predicate “smoking in the street” to the morally “bad” women (Stokoe 2003). Moreover, the moral divisions that the participants produce as relevant reproduce the long recognized

Hegemonic Femininity as an Interactional Accomplishment

277

tendency towards sexually derogatory descriptions of women (Munro, 1989, Attwood 2007, Schulz 1990). In other words, women who breach the normative constraints on gender appropriate behaviour in relation to smoking are produced as incumbents of an overtly sexualized subcategory of the device “moral types of woman”. A similarly gendered categorial formulation emerges in Extract 3 from the Night Out transcript. In this extract, we see how the speakers similarly orient to breaches of gender-specific norms and how they manipulate taken-for-granted knowledge to make some smokers more accountable than others. Here we can identify how the speakers manage their production of themselves as smokers while policing the boundaries of (un)acceptability for other female smokers. Again, what emerges as highly relevant to these speakers is third party gaze and the highly gendered problem of appearance as a smoker, revealed through the negative appraisal of women who smoke in public places. Extract 3 Transcript: Night Out 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192

Sar Dawn Sar Ann Dawn Sar

Shar

Sar Shar Sar Dawn

Ann Shar Dawn

Ann

No one knows I smoke in m’work = = y’see but whĹy (.) an’ then whĹy d’ye not tell your [ma an’ da? [I don’t wany tell them ((undistinguishable talk)) Cos then you’re [really a smĹoker [people dĹo think it’s (.) y’know I think uhhhh °they think uhh she smokes (.) but KĹane my [manager smokes out the back [but see if you see a girl I seen a girl the day (.) I was comin’ out the road an’ she was sittin’ waitin’ to pull out she was like at y’know the side a the car an I thought (.) is that the way I look like hehehe sittin’ wi’ a fag hangin’ outa my hand like hangin’ outa your mouth [it looks disgustin’ y’know [but some people do look different °y’know when you see people°like y’know [pushin’ Ļa prĻam [rough aye= =wi’ a tracksuit on an auld fuckin’ lambert hangin’ outa their mouth (.) God forgive me for sayin’ that but it’s true > as opposed to somebody sittin’ (.) dressed nicely Ĺin the bar (.) y’know [havin’ a fag like [exactly (1) it’s def- it is different (.) like I notice when parents come in too (.) an y’just Ĺknow the smokers y’just y’know we have t’ask does anybody smoke at home (.) an ye just Ĺknow like

278 193 194 195 196 197

Chapter Fifteen people who smoke and people who don’t ·hhhhhhhh and like the ones who smoke are just goin’ out all the time (.) y’know °gon look after my wean there for 5 minutes while I run out for a fa:g (.) an you’re like (.) fuck you shouldn’t be smokin’

Again this sequence begins with one of the participants, Sarah, problematising her own categorization as a smoker. Sarah’s utterance in line 169 self-repairs from what is hearable as an incomplete third party appraisal of smoking (“people do think it’s”) to reported third party appraisal that is ambiguous between a categorial formulation about women in general who smoke, and appraisal of herself specifically. However, given that it is produced sequentially as an account for not smoking at work, it is hearably produced as relevant to Sarah herself. The negative appraisal is produced simply with the category reference to women, the pronoun “she”, and the category bound activity “smokes”, revealing very starkly the taken-for-granted disjunctiveness of the category-activity pair “women” and “smoking”. This reflects Jayyusi’s (1991) observation that “the conceptual, moral, and practical are reflexively and irremediably bound up with, and embedded in, each other” (p 241). Sarah is engaged in the practice of accounting, but embedded in her account for why she does not smoke in work is the shared cultural knowledge of a highly gendered moral order in which the activity of smoking constitutes a breach of the normatively appropriate behaviour for the category “women”. It is only because this is part of the “knowledge context” that Sarah’s utterance works for the participants as a practical account for not smoking in work. Sarah then introduces a contrast with “but” between this third party negative appraisal of women/herself smoking and her manager, Kane’ who is produced as openly, publicly “smoking out back” and, by implication, immune from this third party condemnation. This contrast with a specific work colleague reinforces the hearing that the categorial formulation “people think she smokes” is interactionally produced in situ as specifically relevant to Sarah and specifically relevant to the work setting despite the generic subject “people”. The MCD gender is also made relevant here, albeit implicitly, through Sarah’s production of the contrast with her manager “Kane” who is male. This gender relevance is oriented to in Sharon’s talk which begins with a categorial formulation “if you see a girl”, which overtly establishes the relevance of the device “gender”. Here, Sharon appears to be accounting for Kane’s immunity from negative third party appraisal through the categorial formulation that now explicitly contrasts Kane and “a girl”.

Hegemonic Femininity as an Interactional Accomplishment

279

However, Sharon then switches to a specific narrative about a specific girl “I seen a girl the day”, where the relevance of the device “gender” is sustained via consistency (Sacks 1992). Drew (1998) observes that “descriptions are unavoidably incomplete and selective, they are designed for specific and local interactional purposes”. Interestingly, here Sharon’s narrative involves identifying the girl with a set of descriptive predicates but omitting the activity “smoking” despite that being key to the category that Sharon is working up as evidenced in her proposed description of herself “sittin’ wi’ a fag hangin’ outa my hand like” (l.178). In line 179 Sarah upgrades Sharon’s talk about having a “fag” in her hand to having it “hangin’ outa your mouth”. Furthermore she reformulates Sharon’s utterance and provides an assessment of the prior turn, by describing it as “disgusting”. It would seem that the action being achieved by Sarah’s talk is to create an environment of alignment with the previous speaker, Sharon. In this sequence, the participants’ orientation to appearance and their own assessment by third party observers is what emerges very explicitly as salient for these speakers. Sharon’s narrative about the girl that she saw leads up to the key relevance for her: “is that the way I look like”. Moreover, we see in Sharon’s explicit reflection on her own appearance as a smoker, support for Bergmann’s claim that moral engagement is “to a very high degree self-reflexive” (Bergmann, 1998: 288). While Sarah’s upgrading of Sharon’s category resonant description into a caricature of the unacceptable smoker is perhaps implicitly hearable as a humorous rejection of their own incumbency of this unacceptable category, Dawn’s post-expansion makes explicit the category distinctions in the device “smoker” based on predicates of “looks” (“some people do look different”) and simultaneously reinforces the production of the participants as incumbents of an acceptable category of smoker. Dawn produces a description of “people like y’know pushin’ a pram”. Although the term “people” is linguistically non-gendered, Stockill and Kitzinger (2007) have claimed that “a linguistically non-gendered term (‘people’) may be understood as gendered by the participants” (p. 225). In this sequential context, the participants are oriented to the disjunctiveness of the category “woman” and the activity “smoking” as topic, hence the device “gender” is highly salient. Moreover, “pushing a pram” is hearable to co-cultural members as an activity bound to the category “mother”. Hence Dawn’s use of the term “people” can be understood to be gendered in this context. Dawn is thus meaningfully and subtly negotiating normatively shared knowledge about the category “mother” and appropriate category-bound activities associated with this category to

280

Chapter Fifteen

achieve a disjunctive, or mismatched, category of “smoking mother” (Stokoe 2003). Dawn’s utterances achieve agreement from the other speakers manifest when Sharon shows explicit agreement with a minimal agreement token (line 184) and in the following turns, when Ann provides further on-topic elaboration. In lines 185-186, Dawn upgrades her description to include category resonant predicates of clothing type (“tracksuit”) and cigarette choice (“an auld fuckin’ Lambert”)7. We can claim that Dawn has specifically designed her talk to “suit” the recipients since they are not in a category of smokers who push prams; wear tracksuits or smoke Lamberts (a brand of relatively cheap cigarettes). By introducing these predicates (Watson, 1978) - the clothes, the prams, the type of cigarettes - Dawn trades on common-sense knowledge that can be taken by “the vernacular culture to be specially characteristic of a category’s members” Schegloff (2007: 470). We can maintain that this is shared between the co-participants by analysing how they orient, or react to Dawn’s descriptions: Ann’s upgrading of the description in line 183; Sharon’s agreement in line 184; and finally with Ann’s supportive story-telling in lines 189-197. Dawn invokes a contrast between these negatively appraised “prampushing track suit wearing smokers” and the other “nicely dressed smokers sitting in a bar”. Dickerson (2000) has examined how speakers construct contrasts between self and others in talk-in-interaction for the purpose of performing “different sorts of activity” such as enhancing or deprecating self. Jaworski and Coupland (2005) have also identified how speakers in “gossipy episodes” use this kind of contrastive work to achieve specific outcomes in their talk (e.g. group solidarity). They define “othering” as “an emergent category”, and this would seem an apt descriptor for the kinds of category emerging here and produced in situ to exclude the participants from incumbency of the negatively appraised category of smoker and establish them as incumbents of an acceptable category of smoker. This series of category resonant predicates designed to establish themselves and others as incumbents of hierarchically ordered categories of the device “smoker” also maintains the gendered orientation to “looks” that has been produced as relevant by all the participants. Similarly it also orients to the situatedness of smoking as significant: embedded in the 7 As co-cultural members we can infer that the clothing type and cigarette brand are intended to identify a particular working-class stereotype, however this is not overtly oriented to by any of the participants, hence resonance with the device social class, although available to the analyst is not explicitly evidenced in the data.

Hegemonic Femininity as an Interactional Accomplishment

281

predicates is the distinction between “smoking in a bar” and “smoking in the street” which maintains and produces further in situ negotiation of the category bound attributes of the unacceptable smoker by excluding “the bar” from the negatively appraised public contexts for smoking. The analysis here has similarities with Stokoe’s (2003) category analysis where she was able to explore the complex practices through which participants established themselves and others as incumbents of a particular categorization device. In particular, we see here that the participants are not merely “doing describing” but are subtly engaged in negotiating the boundaries around a subcategory of morally accountable smoker. As in extract 2, the speakers here orient to the categories “girl” and “smoker” as disjunctive and morally accountable (Stokoe 2003) and again, the participants resist incumbency of this morally accountable category of female smoker through the in situ production of the negatively positioned subcategory of the female smoker who smokes in specific public places.

15.6 Discussion There is a striking degree of similarity observed in the two interactions analysed here, not only in the taken-for-granted cultural gender norms that emerge in the interactions but also in the patterns of interactional behaviours through which these gender norms emerge. Participants move between descriptions and narratives which participants orient to as “category resonant” (Stokoe 2012) and which lead to categorial formulations. In both sequences, the device “gender” is initially made potentially relevant through the formulation of a contrast but is subsequently oriented to and made explicit by another participant in the production of a categorial formulation. Moreover in these categorial formulations, gender categories are oriented to by the participants as short cuts to shared cultural and moral knowledge of the norms for gender appropriate behaviour (Stokoe 2003, 2012) which are drawn on collaboratively by the participants as a resource to make sense of the preceding narrative contrast. What emerges clearly in both sequences is the taken-for-granted nature of the disjunctiveness of the category “woman” and the activity “smoking”. Both sequences begin with a participant problematising their own categorization as a smoker and, in both cases, it is their appearance as incumbents of the category “smoker” that the speakers orient to as problematic. Moreover, it emerges through a series of descriptions, narratives and eventually categorial formulations that it is their

282

Chapter Fifteen

incumbency of the category “woman” that makes their membership of the category “smoker” problematic. In other words in both sequences, their own membership of the category “smoker” produces for the participants a “category puzzle” (Watson and Weinberg 1982) since incumbency of the category “smoker” is oriented to as a moral breach of the normative constraints on appropriate behaviour for incumbents of the category “woman”. It is interesting to note that not one of the participants, who are all women who smoke, resists this gendered moral ordering. Instead, in both sequences, we see that the participants collectively and collaboratively orient to this “category puzzle” as a problem of appearance. In solving the category puzzle, the participants in both sequences renegotiate both the categories “woman” and “smoker” in situ as hierarchically ordered MCDs: “moral types of women” and “acceptable/unacceptable types of smoker”. The “in situ” nature of this renegotiation reflects the simultaneously indexical and occasioned nature of categories and predicates which is why they are so revealing of shared cultural knowledge (Hester and Eglin 1997). In these sequences, it is in the predicates through which these locally assembled devices are produced that we see the orientation to appearance for women and the influence of third party gaze. In both sequences, the participants invoke predicates of location, particularly smoking in the street, in order to resist incumbency of the negatively appraised category of women who smoke. These speakers are thus working within a local context of the talk to produce and negotiate “socially acceptable images of femininity” (Guendouzi, 2001: 29). Similar to this research, Guendouzi (2001) finds in her analysis of womens’ talk that the values assigned to women in the current social marketplace were based on “physical appearance”, “moral worth” and being seen to be a “good mother” (p 47). However, it is important to note that the speakers in our data are oriented not so much to physical appearance as to the situated social meanings and “moral worth” attached to being seen to smoke in particular settings. Stokoe (2006) has previously highlighted how MCA can illuminate the “taken-for-granted ‘facts’ about gender-appropriate behaviour” and how these “facts” are worked out, or negotiated in the mundane and routine interactions of talk (p 477) In these sequences we see that the speakers orient to a taken-for-granted knowledge context in which “moral worth” for women is attached not to the behaviours themselves but to the adroitness with which women manage their appearance linked to the behaviours. This reflects a very stark sense in which gender is a “performance”, since it is not what the

Hegemonic Femininity as an Interactional Accomplishment

283

women do that counts, but what the women are seen to do and where they are seen to do it. In these interactions, the speakers orient to how women look when they are smoking and the implications that this has for their perceptions of themselves, made manifest in their talk. What is particularly interesting about the participants’ in situ negotiation of the norms for gender-appropriate behaviour is that it provides empirical support from mundane, everyday talk for a participant orientation approach to Schippers' (2007) model of “hegemonic femininity”. Schippers argues that the “production, proliferation and contestation of the quality content of masculinity and femininity are ongoing, dynamic, social processes and include everyday practices” (p.93 emphasis added). An important site for the empirical examination of hegemonic femininity should therefore be in the mundane practices of ordinary, everyday conversation. It is important to note however that the research presented here did not set out to investigate gender hegemony or even gender. The research set out simply to examine in a bottom up fashion the categorization work in the talk of the speakers. The research process thus followed Schegloff's challenge that “rather than beginning with gender ideologies … the analysis might begin by addressing what the parties to the interaction understand themselves to be doing in it” (1998: 415). The analysis thus far presented has asked precisely that: what do the parties to the interaction understand themselves to be doing in it? The orientation to gender emerges directly from this ethnomethodological examination of the participant categories. Moreover, we suggest that the hegemonic character of those gender categories emerges and is oriented to in the situ negotiation of the category puzzle “women smokers”. As suggested above, a participant orientation based approach to hegemonic femininity would need to reveal participant orientation to complementarity in gendered categories and predicates and to the subordinating nature of that complementarity. Participant orientation to a pariah femininity would involve orientation to women as stigmatized when they are conversationally tied to predicates that are oriented to as masculine. Looking first at the question of complementarity, in the data analysed in this paper, the participants were shown to orient to the device “gender” initially implicitly through personal narratives that presented gendered contrasts. In the first transcript, Cat, who didn’t smoke in restaurants, is contrasted with her husband, John, who did. In the second transcript, Sarah, who didn’t smoke at work, is contrasted with her manager, Kane, who did. We have seen that contrasts are deployed to attend to local interactional issues (Dickerson 2000) and in both these conversations the participants are dealing with their own problematic

284

Chapter Fifteen

categorization as a smoker at the point at which the contrasts are produced. This would suggest that the contrasts are produced to attend to this problematic categorization as smoker. At the same time, the contrasts make gender relevant and orient to complementarity in the predicates associated with the gender categories “man” and “woman”, since men smoking is not oriented to as a breach of shared norms for masculine behaviour. Moreover, as pointed out above, none of the participants contests this complementarity that restricts the behaviours of women who smoke, revealing not only complementarity in the predicates tied to men and women but also conformity to the restrictions imposed by that complementarity. This conformity is produced more forcefully in the categorial formulations that follow the personal narratives. In our analysis of Extract 3, we showed that Sharon responds to the contrast that Sarah produces between Kane and herself with a categorial formulation that explicitly contrasts Kane and “a girl” and makes explicit a complementarity that leaves Kane immune from negative third party appraisal. In Extract 2, the implicitly gendered contrast between Cat and her partner John is followed by paired categorial formulations in which the complementarity of the behaviours oriented to as normatively appropriate for men and women are made explicit: “girls that walk about town… wi’ fags” are “really horrible lookin’”, whereas “boys, I don’t think’s as bad”. These gendered categorial formulations thus orient to the hegemonic character of gender categories by conforming with a complementarity that subordinates women by subjecting women but not men to the negative effects of third party appraisal. An additional outcome of these categorial formulations is that women who smoke in the street are produced as a pariah femininity, as carrying out an activity that is oriented to as legitimate for men but unacceptable for women. Schippers emphasizes that although pariah femininities involve women embodying characteristics otherwise understood as masculine, they are produced and oriented to as femininities. This is also clear in the gendered description of the women who smoke in the street as “wile trampy lookin’”. In our analysis, we showed that this pariah femininity is occasioned by the participants renegotiating the category “woman” as a hierarchically ordered MCD. This in situ production of a pariah femininity serves a local rhetorical effect of avoiding incumbency of a disjunctive category. However, in producing a pariah femininity, the participants also orient to multiple femininities ordered by a relationship of ascendancy. Moreover, in this locally produced relationship of ascendancy, hegemonic femininity is interactionally accomplished in the

Hegemonic Femininity as an Interactional Accomplishment

285

contrast with acceptable forms of masculinity that grants men access to public spaces but restricts women's access. Returning to Speer's argument that hegemonic masculinity is not a participant category, the analysis presented here demonstrates that the participants, in their in situ production of categories and MCDs, are oriented to the defining elements of Schippers’ (2007) model of hegemonic femininity. In particular, we have shown that the interactional production of contrasts, both in narratives and in categorial formulations displays an orientation to hierarchical complementarity with masculinity. At the same time, the conversationally gendered predicates of smoking orient to the local production of a pariah femininity through the renegotiating of the category “woman” as hierarchically ordered device. In producing “women who smoke in public” as a pariah femininity, the participants are also shown to be complicit in the hierarchical complementarity with masculinity. The participants are thus shown to draw on inference rich predicates of shared cultural knowledge (including knowledge of gender ideologies) to produce themselves as incumbents of a locally occasioned hegemonic femininity that inoculates them from the stigma of the disjunctive category “women who smoke”. Hegemonic femininity in our data can thus be seen as a local accomplishment of the category work in the data that is designed to achieve locally occasioned rhetorical effects but also demonstrates an orientation to a shared knowledge context that includes gender norms and ideologies. This realigning of the concept of hegemonic femininity in line with the EM/CA emphasis on participant orientation clearly draws on Housley and Fitzgerald’s reworking of membership categorization analysis to focus on the “in-situ and locally occasioned character of members’ category work”. Hegemonic femininities and pariah femininities in this perspective are understood as in situ practical interactional achievements that are nonetheless oriented to achieving relationships of ascendancy that legitimate gendered power relations.

15.7 Conclusion This analysis using MCA has shown how orientation to gender is embedded in the practices of everyday talk of adult women who smoke. However, the point of the analysis is not to claim that the talk of these women smokers is extensively gendered. Rather the paper focuses specifically on the particular way that gender is drawn on by participants in both interactions as a sense-making resource in the context of narrative contrasts produced by women smokers engaged in negotiating their own

286

Chapter Fifteen

identities as smokers. The analysis thus follows Schegloff’s advice that “rather than beginning with gender ideologies … the analysis might begin by addressing what the parties to the interaction understand themselves to be doing in it” (1998: 415). In particular, the focus is on the way that gender is paradoxically oriented to by the participants as both the source and the solution to their problematic categorization as smokers. We have shown that the interactional production of contrasts, both in narratives and in categorial formulations displays an orientation to hierarchical complementarity with masculinity. At the same time, the conversationally gendered predicates of smoking orient to the locally occasioned renegotiation of “woman” as hierarchically ordered device. However, in producing “women who smoke in public” as a stigmatized category of this device, the participants are also shown to be complicit in the hierarchical complementarity with masculinity. Thus, the analysis reveals how the participants orient to the local production of a hegemonic femininity as a resource to manage their own problematic categorization as a smoker. This addresses Speer’s criticism of “approaches which reify such concepts, and which treat them as determining members’ talk or the discourses that are available to them in a top-down fashion”. As such, we might extend Jayyusi’s (1991) observations on the moral character of category work in interaction to suggest that such category work is constitutive of moral and hegemonic praxis.

References Abell, J. 2011. ‘They seem to think “We’re better than you”’: Framing football support as a matter of ‘national identity’ in Scotland and England. British Journal of Social Psychology 50(2). 246–264. Attwood, F. 2007. Sluts and Riot Grrrls: Female Identity and Sexual Agency. Journal of Gender Studies 16(3). 233–247. Bergmann, J. 1998. Introduction: Morality in Discourse. Research on Language and Social Interaction 31(3 and 4). 279–294. Butler, J. 1990. Gender trouble: Feminism and the subversion of identity. New York: Routledge. Connell, R.W. 1995. Masculinities. Berkeley: University of California Press. —. 2000. The men and the boys. Berkeley: University of California Press. Connell, R.W. and J.W. Messerschmidt. 2005. Hegemonic masculinity: Rethinking the concept. Gender and Society 19: 829–859.

Hegemonic Femininity as an Interactional Accomplishment

287

Dickerson, P. 2000. 'But I'm different to them': Constructing Contrasts between self and others in talk-in-interaction. British Journal of Social Psychology 39( 3). 381–389. Drew, P. 1998. Complaints about Transgressions and Misconduct. Research on Language and Social Interaction 31(3 and 4). 295–325. Edley, N. 2001. Conversation analysis. discursive psychology and the study of ideology: A response to Susan Speer. Feminism and Psychology 11(1) 136–140. Edwards, D. 1998. The relevant thing about her: Social identitiy categories in use. In C. Antaki and S. Widdicombe (eds.), Identities in Talk, 15– 33. London: Sage. Gardner, R. 2012. Enriching CA through MCA? Stokoe's MCA keys. Discourse Studies 14(3). 313–319 Gaucher, J. 2011. Women and Parachuting: Writings of the Outsider's Experience. The International Journal of the History of Sport 28(14). 2031–2046. Guendouzi, J. 2001. 'You'll think we're always bitching': the functions of cooperativity and competition in women's gossip. Discourse Studies 3(1). 29–51. Hester, S. 1994. Les Catégories en Contexte. Raisons Pratiques 5:219–42 [Special Issue: B. Fradin, L. Quéré and J. Widmer (eds.), L'Enquête sur les Catégories: de Durkheim à Sacks. Paris: Éditions de l'École des Hautes Études en Sciences Sociales]. Hester, S. and P. Eglin. 1997. The Reflexive Constitution of Category. Predicate and Context in Two Settings. In S. Hester and P. Eglin (eds.), Culture in Action: studies in membership categorization analysis, 25– 48. Boston: International Institute for Ethnomethodology and Conversation Analysis and University Press of America. Housley, W. and R. Fitzgerald. 2002. The Reconsidered Model of Membership Categorisation Analysis. Qualitative Research 2(1). 59– 83. Jaworski, A. and J. Coupland. 2005. Othering in gossip: “you can go out you have a laugh and you can pull yeah okay but like..”. Language in Society 34(1). 667–694. Jayyusi, L. 1984. Categorization and the Moral Order. London: Routledge and Kegan Paul. —. 1991. Values and moral judgment: communicative praxis as a moral order. In G. Button (ed), Ethnomethodology and the human sciences, 227–251. Cambridge: Cambridge University Press.

288

Chapter Fifteen

Lee, S.-H and N. Sheon. 2008. Responsibility and risk: accounts of reasons for seeking an HIV test. Sociology of Health and Illness 30(2).167–181. Nilan, P. 1995. Membership Categorization Devices Under Construction: Social Identity Boundary Maintenance in Everyday Discourse. Australian Review of Applied Linguistics 18(1). 69–94. Pyke, K.D and D.L. Johnson. 2003. Asian American women and racialized femininities: ‘Doing’ gender across cultural worlds. Gender and Society 17(1). 33–53. Pomerantz, A.M. 1986. Extreme Case Formulations: a way of legitimatising claims. Human Studies 9 (Special Issue on Interaction and Language Use): 219–229. Roulston, K., C. Baker and A. Liljestrom. 2001. Analysing the researcher’s work in generating data: The case of Complaints. Qualitative Inquiry 7( 6). 745 – 772. Sacks, H. 1992. Lectures on Conversation. Oxford: Blackwell. Schegloff, E.A. 1998. Reply to Wetherell. Discourse and Society 9( 3). 457–460. —. 1999. “Schegloff's texts” as “Billig's data”: a critical reply. Discourse and Society 10(4). 558–572. —. 2007. A tutorial on membership categorization. Journal of Pragmatics 39: 462–482. Schippers, M. 2007. Recovering the feminine other: masculinity. femininity and gender hegemony. Theory and Society 36: 85–102. Schulz, M.R. 1990. The semantic derogation of woman. In D. Cameron (ed) The Feminist Critique of Language: A Reader, 134–147. London: Routledge. Speer, S.A. 2001a. Reconsidering the concept of hegemonic masculinity: Discursive psychology. conversation analysis and participants orientations. Feminism and Psychology 11(1). 107–135. —. 2001b. Participants Orientations. Ideology and the Ontological Status of Hegemonic Masculinity: A Rejoinder to Nigel Edley. Feminism and Psychology. 11(1). 141–144. —. 2005. Gender and Talk. London: Routledge. Stockill, C. and C. Kitzinger. 2007. Gendered 'People': How Linguistically Non-gendered Terms Can Have Gendered Interactional Relevance. Feminism and Psychology 17( 2). 224–236. Stokoe, E. 2003. Mothers. Single Women and Sluts: Gender. Morality and Membership Categorisation in Neighbour Disputes. Feminism and Psychology 13( 3). 317–344.

Hegemonic Femininity as an Interactional Accomplishment

289

—. 2006. On ethnomethodology. feminism and the analysis of categorial reference to gender in talk-in-interaction. The Sociological Review 54( 3). 467–494. —. 2012. Moving forward with membership categorisation analysis: Methods for systematic analysis Discourse Studies 14(3). 277–303 Sutton, L. 1995. Bitches and skanky hobags. In K. Hall and M. Bucholtz (eds.), Gender Articulated: Language and the Socially Constructed Self. London: Routledge. Watson, D.R. 1978. Categorization. Authorization and Blame-negotiation in Conversation. Sociology 12( 1). 105–113. Watson, R. and T. Weinberg. 1982. Interviews and the Interactional Construction of Accounts of Homosexual Identity. Social Analysis 11: 56–78. Wetherell, M. and N. Edley. 1999. Negotiating Hegemonic Masculinity: Imaginary Positions and Psycho-Discursive Practices. Feminism and Psychology 9(3). 335–56. Wooffitt, R. 2006. Conversation Analysis and Discourse Analysis (2nd edn). London: Sage. Wowk, M.T. 1984. Blame Allocation. Sex and Gender in a Murder Interrogation. Women’s Studies International Forum 7(1). 75–82.

CONTRIBUTORS Paolo Acquaviva is senior lecturer in linguistics at University College Dublin. He is author of the OUP monograph Lexical Plurals (2008). His main areas of interest are the lexicon and morphology and its interface with semantics. Leah S. Bauke works as ‘wissenschaftliche Mitarbeiterin in the Department of Linguistics at the Goethe-University in Frankfurt. Her publications are in the areas of phase theory, morphology, linearization and antisymmetry. Mariachiara Berizzi a postdoctoral researcher at the University of Padua, Her current research is on the morphosyntax of Italian dialects and she is presently involved in the ASIt project (Syntactic Atlas of Italy). Grace Burke completed doctoral research at the University of Ulster. Her interest was on the negotiation of social identity categories in the everyday conversations of smokers. Andrew Carnie is Professor of Linguistics at the University of Arizona. He is widely published, including his very popular textbook Syntax: A Generative Introduction (now in its 3rd edition). Most recently he has been leading a team investigating the syntax, phonology and phonetics of Modern Scottish Gaelic. Lieven Danckaert is a post-doctoral researcher at the University of Ghent with a particular interest in the syntax of the left periphery within a cartographic framework. His book Latin embedded clauses: the left periphery was published by Benjamins in 2012. Alessio S. Frenda completed doctoral research at Trinity College Dublin. His particular research interest is on variation and change in Irish and Welsh focusing on grammatical gender. Liliane Haegeman is Professor of Linguistics at the University of Ghent. She has published widely in generative linguistics since the 1980s including several very successful textbooks. She is currently leading a

Minority Languages, Microvariation, Minimalism and Meaning

291

team of twelve researchers on the Odysseus Project (Comparative syntax. Layers of structure and the cartography project). Alison Henry is Professor of Linguistics at the University of Ulster. She works on syntax in the Minimalist framework and is best known for her work on microvariation, particularly her book Belfast English and Standard English (OUP 1995). Mark Henry is a PhD candidate at SOAS, University of London working on the syntax of Akkadian. Marko Hladnik is a PhD candidate at Utrecht University/ UiL-OTS working on resumption and recoverability in relative clauses. Frances Kane is a PhD candidate at the University of Ulster working on the syntax of the Irish DP. Maųgorzata Krzek is a PhD candidate at University of Newcastle working on the syntax of impersonal constructions in Polish. Masakazu Kuno is lecturer in linguistics at the Waseda University with a particular interest in the syntax semantics interface in the minimalist framework. Francesca La Morgia is lecturer in linguistics at the University of Reading. Her research interests are monolingual and bilingual child language development, the acquisition of syntax, language processing at interfaces, heritage language description and Italian linguistics. Ulrich Reichard is a Doctoral Research Fellow on the AHRC/DFG funded project Un-Cartesian Linguistics at the University of Durham which tries to assess arguments in favour of the hypothesis that thought is inherently linguistic. Catrin S. Rhys is lecturer in linguistics at the University of Ulster. She employs conversation analysis and membership categorization analysis to examine social interaction in mundane, clinical and multilingual contexts. With a background in formal linguistics, she is also interested in the interface between the interactional properties of language in use and the formal properties of language structure. Silvia Rossi is a linguist at the University of Padua working on the syntax of Irish English.

INDEX Abruzzese ...............................36–49 adjective .... 107, 108, 109, 111, 112, 117, 118, 124, 132, 133 subsective vs intersective ....................................223–41 Agreement .............. 101–12, 117–25 Alignment ............................247–50 antisymmetry ......................165, 239 Arabic........................................ 186 Argument sequence ................... 250 aggravated opposition .. 251, 257, 259–60 Aspect ........................ 53–66, 81–84 After-perfect ......................53–66 Auxiliaries ................................. 199 passive ........................... 197, 200 Selection ............................36–49 binding ...................................... 192 Breton................................. 104, 105 Cartography ............................57–63 categorial formulation ....... 275, 276, 277, 278, 281, 284 Chinese ........... 4, 152, 153, 160, 235 complementizer ......................... 173 Compounding .......................150–68 control ................................ 178, 192 Conversation Analysis..........245–62 Copula ......................................... 76 Cornish ....... 104, 105, 111, 127, 129 Demonstratives.....................131–47 Disaffiliation .......... 245–62, 245–62 disalignment ............... 251, 252, 254 doubling phonosyntactic ...................47–49 possessor .................................. 3 Dutch .......................... 4, 57, 64, 104 English Appalachian............................ 30 British English ........................ 66

Hiberno English .... 24–34, 53–66 features .. 76–84, 107, 112, 119, 121, 123, 146 edge features (EFs) .........157–58, 161, 162, 165 Feature Inheritance ......... 206–21 ij-features.............. 158, 160, 161, 166, 206 Flemish .................................... 2–21 French ....... 5, 32, 61, 105, 131, 136, 141, 144, 146, 150, 151, 231 Functional Discourse Grammar ................................ 104, 119–25 Gender Grammatical ..... 7, 101, 102, 104, 105–12, 114, 117–25, 166– 68, 172, 188 norms .............268, 270, 281, 285 Semantic ............... 101, 102, 116 Social category ............... 266–86 Sociolinguistic variable .......... 43 German .... 5, 64, 146, 150, 151, 153, 154, 156, 163, 235, 236, 240 Hegemonic femininity .........268–71, 283–85 Holmberg’s Generalization ...24, 28, 30, 196 Hungarian ...............4, 131, 135, 146 Iambic feet ............................. 95–97 Icelandic ........................ 29, 30, 215 Old Icelandic ........................ 135 imperatives ............................ 24–34 Institutional talk........... 255, 257–61 Irish ....... 64, 66–88, 92, 94, 101–25, 131–47, 186, 196, 235 Connacht ....................... 107, 113 Munster ........................... 82, 120 Old Irish ............... 128, 129, 141, 142, 146

Minority Languages, Microvariation, Minimalism and Meaning Ulster ...................................... 84 Italian .......... 57, 141, 150, 186, 187, 189, 196, 208, 214, 215, 217, 233, 237 Altamurano ............................. 40 Maceratese.............................. 40 Trentino .................................. 40 Venetan .................................. 61 left periphery .................. 17–18, 145 lenition .... 70, 71, 79, 105, 107, 108, 110, 111 linearization............................91–99 Manx ........... 92, 104, 105, 111, 126, 127, 129 Membership Categorization Analysis ...........................258–86 Membership Categorization Device ..... 267, 275, 276, 277, 278, 279, 281, 282, 283, 284 multi-unit questioning turn ...251–52 passive ........................ 198–204, 220 Phase Theory....................... 150–68, 206–21, 238–41 Polish.......... 178, 180, 186–204, 227 Possessor External ...............................2–21 Preference.............................247–48 Pronoun postposing .........................92–94 realis and irrealis ....................56, 84 Relative Clause ............. 17, 171–84, 235–37 Repair ........... 248–49, 275, 276, 278

293

repetition ......................... 249–50 Resumption.......................... 171–84 Russian ...............178, 180, 235, 250 Scottish Gaelic........91–99, 104, 123 semantics ......................... 75, 83, 84 formal ............................. 223–28 syntax-semantics interface ............ 15–16, 70–88, 222–41 Serbo-Croatian................... 176, 180 Slovene ................................ 171–84 Spanish .......................... 31, 61, 105, 134, 135, 150, 187, 189 Spell Out ..... 76, 78, 79, 80, 83, 131, 143, 159–65, 178 subject ..................... 2, 3, 4, 5, 6, 14, 20, 55, 63, 70, 178 generic .......................... 186–204 impersonal constructions 188–94 null-subject ............. 186, 187–88 quirky-subject ....................... 178 relative clause ......... 174, 178–81 Suppletion.............................. 84–87 Swedish ........ 26, 136, 151, 152, 180 Tense ..................................... 74–75 transitive expletives ..................... 30 Variation Data collection............ 27, 32–34, 43–45, 113–16 Diachronic ........................ 36–49 Synchronic............ 24–34, 36–49, 53–66, 95, 117–25 Welsh................................... 101–25