Functional Heads Across Time: Syntactic Reanalysis and Change (Oxford Studies in Diachronic and Historical Linguistics) 9780198871538, 0198871538

Citation preview

Functional Heads Across Time

OX FORD STUDIES IN DIACHRONIC AND HISTOR IC AL LINGUISTICS General editors Adam Ledgeway and Ian Roberts, University of Cambridge Advisory editors Cynthia L. Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Oxford RECENTLY PUBLISHED IN THE SERIES 42 Phonetic Causes of Sound Change The Palatalization and Assibilation of Obstruents Daniel Recasens 43 Syntactic Features and the Limits of Syntactic Change Edited by Jóhannes Gı´sli Jónsson and Thórhallur Eythórsson 44 Romance Object Clitics Microvariation and Linguistic Change Diego Pescarini 45 The Diachrony of Differential Object Marking in Romanian Virginia Hill and Alexandru Mardale 46 Noun-Based Constructions in the History of Portuguese and Spanish Patrı´cia Amaral and Manuel Delicado Cantero 47 Syntactic Change in French Sam Wolfe 48 Periphrasis and Inflexion in Diachrony A View from Romance Edited by Adam Ledgeway, John Charles Smith, and Nigel Vincent 49 Functional Heads Across Time Syntactic Reanalysis and Change Edited by Barbara Egedi and Veronika Hegedűs For a complete list of titles published and in preparation for the series, see pp. 285–89

Functional Heads Across Time Syntactic Reanalysis and Change Edited by

B A R B A R A E GED I V ER O N I K A H E GED ŰS

Great Clarendon Street, Oxford, OX2 6DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Barbara Egedi and Veronika Hegedűs 2022 © the chapters their several authors 2022 The moral rights of the authors have been asserted Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2021950142 ISBN 978–0–19–887153–8 DOI: 10.1093/oso/9780198871538.001.0001 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

Contents Series Preface List of Abbreviations List of Contributors

1. The role of functional heads in syntactic change Barbara Egedi and Veronika Hegedűs 2. Changing patterns of clausal complementation in Latin: A parametric approach to ‘constructional’ changes Lieven Danckaert

vii viii xi

1

19

3. From split to remerged Fin in Romanian supine complements Adina Dragomirescu and Virginia Hill

56

4. The diachronic path of senão: From conditional subordination to exceptive coordination Ana Maria Martins, Sandra Pereira, and Clara Pinto

70

5. On Italian relative complementizers and relative pronouns: Rethinking grammaticalization Emanuela Sanfelici, Jacopo Garzonio, and Cecilia Poletto

90

6. Information structure, functional left peripheries, and the history of a Hungarian interrogative marker Julia Bacskai-Atkari

111

7. The recategorization of modals in English: Evidence from adverb placement Eric Haeberli and Tabea Ihsane

136

8. Tense recursion, perfect doubling, and the grammaticalization of auxiliaries Ida Larsson and Ellen Brandner

159

9. P-incorporation in the history of Icelandic Jóhannes Gı´sli Jónsson and Brynhildur Stefánsdóttir

186

vi

CONTENTS

10. From Old to Modern Icelandic: Dative applicatives and NP/DP configurationality Heimir F. Viðarsson

210

References Index of Subjects Index of Languages

247 271 283

Series Preface Modern diachronic linguistics has important contacts with other subdisciplines, notably first-language acquisition, learnability theory, computational linguistics, sociolinguistics, and the traditional philological study of texts. It is now recognized in the wider field that diachronic linguistics can make a novel contribution to linguistic theory, to historical linguistics, and arguably to cognitive science more widely. This series provides a forum for work in both diachronic and historical linguistics, including work on change in grammar, sound, and meaning within and across languages; synchronic studies of languages in the past; and descriptive histories of one or more languages. It is intended to reflect and encourage the links between these subjects and fields such as those mentioned above. The goal of the series is to publish high-quality monographs and collections of papers in diachronic linguistics generally, that is, studies focussing on change in linguistic structure, and/or change in grammars, which are also intended to make a contribution to linguistic theory, by developing and adopting a current theoretical model, by raising wider questions concerning the nature of language change or by developing theoretical connections with other areas of linguistics and cognitive science as listed above. There is no bias towards a particular language or language family, or towards a particular theoretical framework; work in all theoretical frameworks, and work based on the descriptive tradition of language typology, as well as quantitatively based work using theoretical ideas, also feature in the series. Adam Ledgeway and Ian Roberts University of Cambridge

List of Abbreviations 1 2 3 ABL ACC AcI ACT ADJ ADV ALL Appl ApplP Asp Aux BoolP C COMP COMPAR COND Cop CP D DAT DEF DEM DFCF DP dpf EA ECM ELA EModE EP EPP EVT F F FIN

first person second person third person ablative accusative Accusatiuus cum Infinitiuo active (Voice) adjective adverb allative case applicative applicative phrase aspect Auxiliary Boolean Phrase complementizer complementizer comparative conditional copula Complementizer Phrase determiner dative definite demonstrative ‘Doubly Filled COMP Filter’ Determiner Phrase double perfect (of languages) external argument Exceptional Case Marking elative case Early Modern English Extended Projection Extended Projection Principle evaluation time feminine focus finite

LIST OF ABBREVIATIONS FINAL FOFC FP FUT GEN HMC HPP I IA ILL IMP IND INF INS IP IPFV IPP LOC M ME MID MP N N NCA NcI Neg NegP NOM NPI O OE OI P PASS PAST PDE PL PLD PLPRF POSS POSSIB PP PPI PRES

final case Final-over-Final Condition focuse phrase; functional projection future genitive Head Movement Constraint Head Preference Principle inflection(al) internal argument illative case imperative indicative infinitive instrumental case inflectional phrase imperfective infinitives for participle morphology locative masculine Middle English middle voice modal phrase neuter noun Null Complement Anaphora Nominatiuus cum Infinitiuo negation Negation Phrase nominative negative polarity item object Old English Old Italian preposition passive past Present-Day English plural Primary Linguistic Data pluperfect possessive possibility suffix adpositional phrase participle for infinitive morphology present

ix

x

LIST OF ABBREVIATIONS

PRF PRS PRT PTCP PTS Q QP RC RED REFL S SBJV SCL SE SF SG Spec SUB SUBJ Subj(P) SUPER SUPESS T Top TP V voc VP ZP

perfect present particle participle perfect time span question particle quantifier phrase relative clause reduced reflexive subject subjunctive subject clitic reflexive particle Stylistic Fronting singular specifier sublative case subjunctive Subject (Phrase) superlative superessive case tense topic Tense Phrase verb vocative Verb Phrase ZeitP

List of Contributors Julia Bacskai-Atkari, University of Konstanz Ellen Brandner, University of Stuttgart Lieven Danckaert, CNRS—University of Lille Adina Dragomirescu, ‘Iorgu Iordan—Alexandru Rosetti’ Institute of Linguistics Bucharest Barbara Egedi, Hungarian Research Centre for Linguistics, Budapest Jacopo Garzonio, University of Padua Eric Haeberli, University of Geneva Veronika Hegedűs, Hungarian Research Centre for Linguistics, Budapest Virginia Hill, University of New Brunswick Tabea Ihsane, University of Geneva and University of Zurich Jóhannes Gı´sli Jónsson, University of Iceland Ida Larsson, University of Oslo and Østfold University College Ana Maria Martins, University of Lisbon, School of Arts and Humanities, Linguistics Center Sandra Pereira, University of Lisbon, School of Arts and Humanities, Linguistics Center Clara Pinto, University of Lisbon, School of Arts and Humanities, Linguistics Center Cecilia Poletto, University of Padua and Goethe University, Frankfurt am Main Emanuela Sanfelici, University of Padua Brynhildur Stefánsdóttir, Cornell University Heimir F. Viðarsson, University of Iceland

1 The role of functional heads in syntactic change Barbara Egedi and Veronika Hegedűs

1.1 Introduction The studies in this volume contribute to our knowledge about the role functional elements play in syntactic change and the semantic and functional features that are the driving force behind the changes examined here. These chapters emerged from the 16th Diachronic Generative Syntax conference, where—despite the broad range of topics and languages covered—the leading role was taken by (changes in) functional elements, thus providing the theme of this volume. The chapters are organized in an order that follows the hierarchy of projections in the clausal functional sequence and the other distinguished ph(r)asal projections. It is important to highlight the role the complementizer and inflectional fields, often split into several projections, play in structural changes cross-linguistically. In this introductory chapter we aim to present how formal syntactic studies interact with grammaticalization theory (Section 1.2) making a clear distinction between changes in the lexical and in the functional domains (Section 1.3). Extended projections, functional hierarchies, and the role of phases will be discussed in Section 1.4, connecting them to the case studies addressed in the chapters throughout the volume, while Section 1.5 is devoted to the problem of what syntactic changes are possible, predictable, or impossible. Finally, Section 1.6 will present the individual chapters of the volume, summarizing their major claims and assumptions, and then Section 1.7 will conclude our discussion.

1.2 Grammaticalization and formal theories of syntax Modern studies in grammaticalization began as early as the 1970s and developed as a truly independent research field in the late 1980s and in the Barbara Egedi and Veronika Hegedú´s, The role of functional heads in syntactic change. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Barbara Egedi and Veronika Hegedú´s (2022). DOI: 10.1093/oso/9780198871538.003.0001

2

BARBARA EGEDI AND VERONIKA HEGEDŰS

1990s, but they have grown out of the frames of functional linguistics only at the beginning of our century, when formal linguistics also discovered the importance of studying language change. (For a concise history of research, see Narrog and Heine 2011.) Works on grammaticalization are usually concerned with the nature and characteristics of the process that leads from lexical or less grammaticalized entities to more grammaticalized structures, in accordance with the original observations that made this subfield arise as an individual branch in linguistics. The process always has a semantic aspect; at the same time, it is the morphological aspect of changes (if there is one) that is the most easily observable. Grammaticalization often goes hand in hand with recategorization and loss in phonological material (cf. Heine et al. 1991; Hopper and Traugott 1993). Syntax and syntactic change have also become central research topics within studies on grammaticalization, and this is the point of interest where formal linguistics could join the overall discussion and also create its own model of grammatical change in general, and grammaticalization in particular. While the diachronic aspect of language used to be neglected by generative linguists, whose primary aim was to account for the native speaker’s innate competence and to model the language faculty, the overall attitude changed to a large extent with the introduction of the hypothesis that the trigger of language change can be found in language acquisition. The view itself goes back to Klima (1964: 83), but has been adopted by many. Most influentially for generative syntacticians, a formal model of language change was put forth by Lightfoot (1979). Early works on syntactic change do not focus on grammaticalization, what they have in common is the emphasis on the role of reanalysis that happens when new generations acquire a language. This change in the interest of generative linguists might also be related to the introduction of functional categories in the theory and, at a later stage, to the introduction of features to systematically account for linguistic variation (cf. van Gelderen 2011a). The locus of change therefore can be found in parameters, namely, in parameter settings. In the framework of Minimalism (cf. Roberts and Roussou 2003), parameter setting has been transferred to feature specification and to the way these feature specifications can change when children acquire the lexicon. Although there is a great variety of definitions and approaches to grammaticalization, the most frequently studied conceptual mechanism within formal approaches is grammaticalization as reanalysis. Technically speaking, language change is due to the learner’s reanalysis of the features encoded in lexical items that served as an input during the process of language acquisition.

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

3

Traditionally, grammaticalization is also meant to refer to the categorial change and semantic weakening of certain lexical items. It might be important, however, to make a clear distinction between purely (lexical) semantic shifts on the one hand, and novel feature specifications on the other hand. The latter are also acquired by a new generation of speakers through learning lexical items, but such changes are more abstract and may also have crucial effects on the possible syntactic configurations in the language.

1.3 Changes in the lexical and functional domain Language change is most easily detected in the changing lexicon of a language. New words and expressions often appear (and sometimes quickly disappear) in any language, especially if it is in close contact with other languages. A narrower interpretation of lexical change concerns the change in the lexical semantics of words, and while such lexical semantic changes may be without larger consequences outside the lexicon itself, semantic changes often have an effect on syntactic structure as well. A simple meaning change of Old English deor ‘animal’ > deer has no effect on the grammar of the language, and similarly there are cases of metaphorical extensions and metonymization the effects of which are equally restricted to the lexicon. However, meaning changes, sometimes based on metaphors and metonyms, also occur under reanalysis, and that has an effect on the syntactic distribution of the affected lexical items, so these changes concern us when dealing with structural change. Semantic change is often characterized by the term semantic bleaching or weakening, the process whereby a lexical item loses (some of ) its original lexical meaning. At the same time, some functional meaning is often gained, that is, the name ‘bleaching’ only reflects the process as a loss of descriptive content, while a new, more functional meaning becomes lexicalized. As Traugott and Ko¨nig (1991) state, the early stages of grammaticalization can be described less as bleaching and more in terms of (lexical) meaning specification, often through the process of metonymization that involves meaning inferences. The development of discourse markers can be illustrative of a multi-step change like this, as English actually has been shown to have undergone the change from a manner adverb meaning ‘actively, effectively’ (in the 15th century) to an epistemic adversative adverb (by the mid-18th century) to a discourse marker (in the early 19th century; see Traugott and Dasher 2002 for details). While the changes in these steps alter the lexical meaning of a single item, the consequence is that it will appear in different syntactic contexts after the change

4

BARBARA EGEDI AND VERONIKA HEGEDŰS

has happened. Similarly the development of complementizers often starts out with a nominal lexical item that develops new senses and thus new functions. An example for such a change is the English noun while becoming a temporal connective and later developing the concessive sense of ‘although’ (Traugott and Ko¨nig 1991). Martins, Pereira, and Pinto also examine the development of a complementizer in their chapter (Chapter 4); however, their case study involves the merging of a complementizer with negation to yield an exceptive connective. Among the lexical semantic changes, changes in the verbal domain can have far-reaching consequences. Whenever a change in the lexical meaning of a verb affects its valency, it has a syntactic impact: it can be represented as a change in the lexical and functional layers of the VP. In other cases, changes in an embedded clause might affect the subcategorization properties of the matrix verb. Danckaert’s chapter in this volume (Chapter 2) examines a change that occurred in the Latin CP field arguing that it modified the clausal complementation patterns in the language. Besides changes in their complementation, verbs may also undergo a change in their category, that is, recategorization. This is best illustrated with the wellstudied empirical domain of the change of English modal verbs into modal auxiliaries (see e.g. Lightfoot 1979 and much subsequent work), a categorial change from V to Infl/T, from a lexical head (with its selectional properties) to a functional head. Modal verbs of the Old English period were just like other lexical verbs, however, they underwent a change that resulted in the distinct behaviour of modal auxiliaries at the beginning of the Early Modern English period. This lexical change went hand in hand with the loss of V-to-I movement, as well as the development of do-support in English. The development of auxiliaries is a process of recategorization (V > T/Infl) and grammaticalization and has been studied along with the changes in English word order and do-support. A less-studied aspect of the empirical consequences of this change is taken up by Haeberli and Ihsane in this volume (Chapter 7). They look at the word order properties of adverbs with respect to the modal elements undergoing recategorization in order to determine when they switch to being auxiliaries. Various cases of lexical change actually affect already (semi-)functional elements, which may also become ‘more functional’ in the sense that they may lose some of their lexical meaning components; this has been part of the definition of grammaticalization. In such cases, there may be no recategorization anymore, but the semantic feature composition of an element may change, and thus, it may be reanalysed as the lexicalizer of a different

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

5

functional projection, or it may change from a specifier into a syntactic head. In recent accounts of syntactic change, many cases have been explored where the change affected functional projections in the CP, TP, or DP domain and these have often exemplified grammaticalization in the late stages of the process. In this volume, Dragomirescu and Hill (Chapter 3) argue for a diachronic split and remerge of the features lexicalized by the Fin head in the Romanian CP domain, where the syntactic heads they examine take part in the checking of features related to (non-)finiteness and modality in different ways; the changes they study relate to the changing feature specification of a functional projection. The chapter by Bacskai-Atkari (Chapter 6) similarly discusses changes in the semantic features of functional heads, specifically changes in the position of the [Q] feature and related phenomena in embedded interrogatives through the history of Hungarian. Looking at changes in the inflectional projections of the clause, Larsson and Brandner chapter (Chapter 8) discusses the various stages of grammaticalization of the perfect auxiliary in Germanic languages and argues that German (and other languages where there is auxiliary selection in perfect aspect) developed a simple auxiliary head and the recursion of T.

1.4 Functional categories and hierarchies In The Minimalist Program, Chomsky (1995: 131) adopted the suggestion by Borer (1984) that only parts of the lexicon, namely, only functional elements within the lexicon are parameterized, and the computational component of language is invariant. Baker (2008a) dubbed this the Borer–Chomsky Conjecture, and formulated it in the way given in (1). (1)

The Borer–Chomsky Conjecture All parameters of variation are attributable to differences in the features of particular items (e.g., the functional heads) in the lexicon. (Baker 2008a: 156)

Under this assumption, language acquisition is basically the acquisition of lexical idiosyncrasies. Since, from the beginning of generative theories of language change it has been assumed that change occurs in the acquisition of language (see e.g. Lightfoot 1991, 1999), this conjecture necessarily implies that change only concerns the lexicon and thus grammatical change amounts to new developments in the properties of functional items.

6

BARBARA EGEDI AND VERONIKA HEGEDŰS

The inventory of functional categories and the properties thereof have been in the focus of syntactic research for the past decades. Two aspects of research are highlighted in this section as they both play a significant role in research on syntactic change in general and in the case studies presented in the chapters of this volume in particular.

1.4.1 Functional hierarchies Starting with Rizzi (1997) and Cinque (1999), detailed studies on the functional projections built on the lexical core have been developed, attempting to identify the fine-grained structure of those projections. The approach initiated by these studies hypothesizes that if there is evidence for certain functional heads and projections and their hierarchical relations in one language, then these heads are present in any other language as well, with the same dominance relations, that is, it is possible to map out a cartography of syntactic heads and projections that are universally present (Cinque and Rizzi 2010b). The project of identifying such functional hierarchies has been very fruitful; the details of functional hierarchies projected on top of lexical categories have been an essential part of syntactic analyses, resulting in rigorous examinations of individual languages, as well as cross-linguistic comparisons. Cinque and Rizzi (2010a), for example, consider studies on the extended structure of adpositional phrases (PPs), with great detail on the lexical and functional layers of projections in various languages. Soon after the synchronic attempts at cartographic descriptions, diachronic paths for syntactic change have been proposed, where the decomposed functional heads and their hierarchies also map out possible grammaticalization paths. Taking a close look at functional heads and their features, the maximal decomposition of these structures provides functional sequences where each projection encodes exactly one feature (Cinque and Rizzi 2010b); this is the assumption Sanfelici, Garzonio, and Poletto adopt in their chapter (Chapter 5), where they discuss the feature specification and categorial status of Old Italian relativizers and argue that grammaticalization can be characterized by the loss of semantic or morphological features until only one feature remains in the projection. An alternative to the cartographic approach does not assume an omnipresent functional sequence of the above-mentioned sort but assumes that languages may vary as to which functional features are present in their syntactic representations and may also vary in how exactly these are

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

7

represented (an assumption in line with the Borer–Chomsky Conjecture), and much work has been done on the uniformity or the parametric variation in the properties of extended projections across languages.

1.4.2 Extended projections and phases What seems common in any recent approach to syntactic structure is that functional heads and their projections have played a central role, and some of them have played a more crucial role in theoretical proposals than others. The notion of phases is of key importance in minimalist syntax (see Chomsky 2001, 2008, among others). Phases are structural domains that exhibit syntactic, semantic, and phonological independence in the structure in that they are impenetrable for movement and other syntactic operations, except for the edge of the phase, which serves as an escape hatch and which is visible from outside the local domain. Phases are units of cyclic spell-out that get transferred to the interfaces. The two projections that have generally been assumed to be phases are C and v, the highest layers of the clause and of the thematic field, respectively. Furthermore, D has sometimes been assumed to be a phase in the nominal extension (also suggested by Chomsky 2008); and, more recently, pP—an equivalent of vP in the verbal domain—has been proposed to be a phase in the extended adpositional phrase (PP). If they are central in synchronic analyses and potentially in language acquisition, they are likely to play a role in syntactic change. While the notion of phases is not central in most chapters in this volume, the local domains that phase theory also relies on are highly relevant in these chapters as they deal with changes related to the complementizer field (C and lower functional heads in the domain) and the inflectional field (TP), as well as the functional extension of the PP and developments related to the DP. Diachronically, the investigation of the higher clausal projections, that is, the complementizer field and the inflectional field, has played an important role in generative analyses of syntactic change and they are essential in this volume as well. The grammaticalization of complementizers, tense, and aspect markers is an empirically rich area of investigation. Martins, Pereira, and Pinto (Chapter 4) argue for a grammaticalization path into C in Portuguese exceptive clauses, made possible by changes in the functional domain below the CP, including the loss of left-peripheral focus. Dragomirescu and Hill’s chapter (Chapter 3) on the diachronic developments in the CP field in Romanian makes use of the further distinction between phasal and non-phasal

8

BARBARA EGEDI AND VERONIKA HEGEDŰS

CPs with the former projecting a larger structure than the latter. In Chapter 6, Bacskai-Atkari examines changes affecting the CP field and the edge of the vP in Hungarian. Among the non-clausal functional layers, the DP may be considered to be a phase (see e.g. Chomsky 2008; Boškovic´ 2008, 2013; Jiménez-Fernández 2009). Changes concerning the functional projections within the DP, especially those related to the determiner itself, are well-studied in languages with long written traditions: for example, the development of D elements and their grammaticalization patterns have gained considerable attention both typologically and in the generative tradition (cf. Greenberg 1978; Giusti 2001; Van Gelderen 2007). In this volume, Heimir F. Viðarsson argues (in Chapter 10) for the grammaticalization of the definite article into a D in the history of Icelandic and claims this change to be the trigger behind the disappearance of some dative applicatives in the language (as compared to Old Norse). A much less-studied functional projection that has occasionally been claimed to be a phase is the extended projection of PPs, namely, the pP (on its phasal status, see e.g. Real Puigdollers 2013; cf. also Abels 2012 on the phasal status of PP). The extended projection of the (semi-)lexical adpositional phrase has received some attention in the past decades and several semantically and morphologically motivated extended projections have been proposed to make up the functional layers, with the projection of particles among them, as a pP on par with the vP in the verbal domain (Svenonius 2003, 2010). Jónsson and Stefánsdóttir’s chapter (Chapter 9) makes the case for a relatively recent change in the history of Icelandic whereby bare lexical PPs are no longer licensed as complements and a pP needs to be projected with its head lexicalized.

1.5 Possible/impossible syntactic changes It is part of the generative linguistic perspective that universal grammatical principles can only be expressed within a limited variability, along a certain set of possible parameters. If it is true for synchronic systems, it must be true for diachronic processes as well, since language change can be considered as a transition from one state into another via the resetting of certain parameters, and often as the outcome of a competition between parallel grammars with only small differences. Based on purely theoretical grounds, there is no meaningful way to predict what syntactic changes are possible and what are impossible (cf. Lightfoot 1999). Nevertheless, typical directions of

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

9

grammaticalization have long been observed allowing for the formulation of the principle of unidirectionality (see Section 1.5.1 below), an observation suggesting that it is not the case that everything is possible as far as morphosyntactic change is concerned. It is to be noted that in the absence of external forces, language transmission is rather conservative, that is, grammatical systems remain inert if there is no specific cause for them to change (Keenan 2002, 2009). It is far from being clear, however, whether change ever happens due to an internal principle or following from the internal characteristics of a given language. On the one hand, it might be assumed that the primary input is hardly ever homogeneous in the course of language acquisition (the problem of corpora heterogeneity); on the other hand, analogical generalization might be in operation as well (this is what Dragomirescu and Hill also refer to in Chapter 3). Furthermore, a tendency for the computational part to be as economical as possible might also play a crucial role in diachronic processes (see Section 1.5.2 below).

1.5.1 Unidirectionality A core property of grammaticalization appears to be unidirectionality. As was mentioned above, grammaticalization proceeds from lexical to grammatical, or from grammatical to even more grammatical forms. According to the hypothesis of unidirectionality, this process is irreversible and only works in one direction. Bo¨rjars and Vincent (2011) collects several claims that aimed to challenge this hypothesis, showing that the notion of degrammaticalization (first suggested by Ramat 1992, cf. also Norde 2009, as a concise monograph on the subject) can plausibly be verified in isolated cases only. (For the same conclusion, see Traugott and Dasher 2002: 87.) At the same time, they review the explanations provided in the literature to account for the directional asymmetry without any conclusive results. Generative syntax (following the proposal found in Roberts and Roussou 2003) has a rather simple answer to the dilemma, as unidirectionality follows from the theoretical and technical apparatus of the framework itself. Reanalysis of a lexical head as a functional head implies loss of movement and an ‘up the tree’ direction, which means that grammaticalization necessarily takes place upwards in the structural hierarchy. Unidirectionality seems to hold at another level as well. The Final-overFinal Condition (FOFC) was originally an observation made about the universally available structural configurations in languages (Holmberg 2000: 124),

10

BARBARA EGEDI AND VERONIKA HEGEDŰS

and has been reformulated as a synchronic generalization that provides a constraint on possible word order changes (see Biberauer et al. 2007, 2014, among others). The FOFC makes the prediction that head-final to head-initial changes proceed top-down, that is, from the higher functional categories to the lower ones, and the opposite direction is impossible (in the case of a change from head-final to head-initial) as that would involve stages where the FOFC is violated. This generalization emphasizes the dominant role that the hierarchical organization of functional heads plays in syntax when it comes to word order in any domain (cf. Martins and Cardoso 2018, a previous volume in the Oxford Studies in Diachronic and Historical Linguistics series).

1.5.2 Universal economy principles According to the Minimalist Program, developing from the Principles and Parameters theory, economy conditions are at work to restrict syntactic operations; for example the operation Merge is preferred over Move as it needs ‘least effort’ during the computational process. Economic principles are also assumed to help language learners to construct as ‘economical’ a grammar as possible. Such principles have been proposed and formulated by Elly van Gelderen (2004, 2008), who integrated this proposal into her theory of cyclical change in historical syntax. Among the three main mechanisms she introduced, the Head Preference Principle has been adopted by some authors of this volume as well. (For instance, Viðarsson refers to the Head Preference Principle when analysing how a more configurational DP emerged in Icelandic in Chapter 10, while Sanfelici and her co-authors in Chapter 5 argue that the shift from pronoun to complementizer they describe in Italian varieties cannot be accounted for in terms of this principle.) It is important to note that van Gelderen’s (2008) principles have been claimed to be preference principles rather than absolute ones that cannot be violated. This means that prescriptive and innovative tendencies can always counteract them.

1.5.3 Appearance and disappearance of functional material As was mentioned above, it follows from the nature of reanalysis that grammaticalization necessarily takes place upwards, and might necessarily correlate with the creation of new functional material in the structure (or the novel, lexical spell-out of already existing material).

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

11

New functional material has been claimed to emerge in favour of structural simplicity (Roberts and Roussou 2003), but the picture seems to be more complicated. As a matter of fact, after the main principles and the essential mechanisms of syntactic change had been established, the latest studies have revealed interesting, or even unexpected details providing a more fine-tuned account of various phenomena that can be observed through diachronic processes. The chapters in this volume are all concerned with the functional domains of the clause, describing apparent rearrangements in parameter settings or in feature specification. The reanalyses the authors study all seem to be internally motivated, which means that they often take part in a sort of chain reaction. In many cases, the phenomena under investigation are the consequence of previous changes in related functional layers; for instance, they can be due to a loss in functional complexity (disappearance of potential syntactic positions, cf. Chapter 4) or to the development of new functional material (which is proposed in BacskaiAtkari’s chapter for the Hungarian vP-periphery (Chapter 6) and in the chapter by Jónsson and Stefánsdóttir about the structure of the Icelandic PP (Chapter 9)).

1.5.4 The role of parsed corpora With the ongoing development of digital parsed corpora for ancient sources and historical linguistic periods, more and more electronic tools are available to examine the texts to a considerable amount and depth. Without grammaticality judgements on behalf of native speakers, digital corpora guarantee the only way to retrieve enough data to reconstruct subsequent language stages. Beyond attestations, the non-attestations of linguistic forms might have a decisive role in these studies. The latter, however, can only be considered as reliable data, if the relevant corpus is large enough to have a statistically significant number of possible contexts in which the phenomena under investigation would occur. What is even more remarkable, electronic corpora for minor languages and for non-Indo European languages have also emerged and are constantly being developed. The increasing quantity of available data and the linguistic diversity these corpora offer may advance more and more fine-grained hypotheses that can be formulated and tested about structural changes, as more and more individual states of languages of the past can be identified via historical evidence.

12

BARBARA EGEDI AND VERONIKA HEGEDŰS

The chapters in this volume heavily rely on such digital corpora. Some of them are huge and well known to the linguistic community (such as the PennHelsinki Corpora), while others have been published recently (e.g. the Old Hungarian Concordance). Readers of the chapters that have been collected in this volume will meet parsed corpora for Italian dialects, Old Portuguese, Old Icelandic, Old and Middle Hungarian, Old, Middle, and Early Modern English.

1.6 Overview of this volume: changes affecting functional domains This section provides an overview of the chapters collected in the present volume. They have already been mentioned in the previous sections in relation to the overarching theme that emerges from these studies, namely, that changes affecting functional domains play a central role in diachronic syntax. Here, a slightly more detailed summary of each individual study is presented, in order to allow a glimpse at the empirical diversity and functional layeredness of the changes concerned. The chapters are ordered to proceed from the top downwards on the clausal spine but as we have already mentioned non-clausal functional extensions are highly relevant as well. In his chapter, Changing patterns of clausal complementation in Latin: A parametric approach to ‘constructional’ changes, Lieven Danckaert discusses two seemingly independent changes in the clausal complementation patterns in Latin and argues that they are related to a single parametric change. On the one hand, infinitival complement clauses are gradually replaced by finite declarative clauses introduced by complementizers. On the other hand, finite complements of the causative verb facio ‘make’ are replaced by infinitival complements. The change is argued to impact the way the clausal EPP-requirement is satisfied: in Classical Latin, VP-movement satisfies the EPP, while in the innovative Late Latin system, it is replaced by V-movement followed by optional subject-movement. Danckaert assumes a functional projection (FP) on the clausal spine, whose head is endowed with an EPP-feature. In the older grammar, FP attracts VoiceP to its specifier, which also means that argument DPs do not undergo case-driven A-movement, case is checked in situ. After the resetting of the EPP parameter, the EPP is satisfied by head movement, namely, V-to-F movement. Along with this change, Exceptional Case Marking came about by reanalysis of the subject as being on the causal spine of the embedded TP, as opposed to a more embedded structural position on the left

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

13

branch of the embedded clause. The chapter thus ties various changes together as consequences of a more general parametric development. Adina Dragomirescu and Virginia Hill, in their chapter From split to remerged Fin in Romanian supine complements, examine the structure of supine clauses in Romanian, clauses that involve the verbal use of the supine form. These clauses can be adverbial or lexically selected complement clauses, and while the former involve no complementizer, complement supine clauses are introduced by the C element de. Old Romanian used de in infinitival and subjunctive clauses as well, but its distribution has become more restricted. The chapter argues that in selected supine clauses, the Fin projection was split, with de checking [−finite] in a higher Fin1 and the verb checking [modal] in Fin2 by V-to-Fin2 movement. De is also used to check the [−finite] feature in infinitival and subjunctive subordinate clauses but there the subordinators a and să, respectively, checked the [modal] feature, so the split Fin involved two lexical complementizers. A reanalysis took place in these clauses, which made a/să syncretic for these two features, and thus de was lost in infinitives and subjunctives, however, it remained constant in supines. This split Fin in supines remains in Standard Modern Romanian. However, in dialectal (Northern) Romanian, the structure is different, showing a remerging of the split Fin, whereby de checks both the [−finite] and the [modal] feature (and there is no V-movement to Fin2). The authors propose that this reanalysis of de is downwards in the syntactic structure (the opposite of what is generally the case as argued by e.g. Roberts and Roussou 2003) and accordingly it involves resemantization, gaining the ability to check the [modal] feature. This marked change is argued to have taken place as a result of paradigmatic pressure, an analogical effect, which makes de more similar to a and să. The next chapter is entitled The diachronic path of senão: From conditional subordination to exceptive coordination. Ana Maria Martins, Sandra Pereira, and Clara Pinto study the development of exceptive constructions in the history of Portuguese and argue for a reanalysis of the conditional conjunction se ‘if ’ and the Neg head nom into a lexical unit senão ‘except’. The main claim is that 13th-century Portuguese had a high clausal position for information focus above T, where fronted information focus appeared and resulted in the presence of strings where a focussed constituent (often the object) intervened between se and Neg with the ellipsis of everything else in the clause. These clauses were free exceptives, involving a compositional exceptive meaning in a conditional clause. However, this focus construction was lost by the 14th century: information focus was no longer fronted. Once this option for fronting constituents for information structural reasons was lost, the conjunction and

14

BARBARA EGEDI AND VERONIKA HEGEDŰS

Neg always appeared string adjacent, and a morphological reanalysis was possible, hence the exceptive senão was grammaticalized. With the loss of the higher focus position, only the clause-final position remained for information focus, which is still the available option now. It is also proposed that this change accompanied the decrease of OV orders in Portuguese in the 14th century. Emanuela Sanfelici, Jacopo Garzonio, and Cecilia Poletto present a study On Italian relative complementizers and relative pronouns: Rethinking grammaticalization. The chapter discusses the Italian relativizer che and its diachronic development. Contrary to previous proposals that treat Modern Italian che as a complementizer head, the chapter argues that it is a pronoun, a specifier. According to the proposal, the change from Old Italian to Modern Italian does not lie in a reanalysis from a pronoun to a head (i.e. it is not a realization of the Head Preference Principle; van Gelderen 2004) but can be treated as a change in the feature specification of che. In Old Italian, agreeing ‘complementizers’ and non-agreeing ‘pronouns’ can both be found, thus the ability to agree does not seem to be a systematic distinguishing factor. Looking at diachronic dialectal data, the authors show that relativizers display agreement even if they do not occur with prepositions, and the agreement is in terms of animacy and gender, which is typical in the nominal domain. The change affects the feature specification of the pronoun, and is to be considered a kind of weakening but not a reanalysis into a head. Arguing against the Head Preference Principle being at play, the authors propose that the preference is towards unmarked elements that carry a single semantic feature in accordance with the general assumptions of the cartographic approach (e.g. Cinque 1999), but these are still specifiers. This also concerns the difference between che and (il) quale, the latter of which has been analysed as a relative pronoun, a specifier, due to its distribution and morphological properties. Since che is also a pronoun, there is no categorial difference between the two. In Julia Bacskai-Atkari’s chapter, Information structure, functional left peripheries, and the history of a Hungarian interrogative marker, the diachronic development of the Hungarian interrogative marker -e is investigated in embedded questions. The chapter relates changes concerning its word order to a large-scale change in the word order properties of the language, especially developments on the left periphery of clauses (see É. Kiss 2014). It is proposed that -e is the head of a functional projection (FP) on the lower periphery of the clause, that is, below the CP field but above TP and the change that took place in the Old Hungarian period concerns the grammaticalization of the encoding of [Q] in this position rather than on a C head. In Modern Hungarian

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

15

there is a split: the marking of subordination is linked to the CP-periphery, while the marking of [Q] is linked to the lower functional projection FP. It is shown that Old Hungarian embedded polar questions were introduced by the subordinator ha ‘if ’, which encoded [Q] in C, then in Middle Hungarian the interrogative marker -e was also present and the complementizer was still ha ‘if ’ in this clause type. Finally, in Modern Hungarian -e is obligatory in embedded polar questions, and these clauses, as well as embedded wh-questions and non-interrogatives, are introduced by the general finite complementizer hogy ‘that’. The chapter argues that the steps in the change went from single encoding of subordination and [Q] on C to double encoding of subordination and [Q] in C and additionally [Q] being encoded in FP by -e and finally to split encoding of subordination on C and [Q] on F. The interrogative marker was already present in Old Hungarian main clauses, as a clause-final C head in a head-final system. With the grammaticalization of [Q] in the FP, clause-final -e disappeared. The structural changes in Hungarian that led to the strengthening of a head-initial CP layer and the rise of the functional layer above vP and the establishment of obligatory preverbal focus reinforced the change that was happening in embedded interrogatives. Eric Haeberli and Tabea Ihsane, in their chapter entitled The recategorization of modals in English: Evidence from adverb placement, present quantitative data concerning the word order of finite main verbs and finite modals with respect to adverbs from Old English to Early Modern English. The discussion contributes to the major issue of the categorical identity or difference of English modals and main verbs in the early stages of grammar and the diachronic development of modals, about which there have been various proposals in earlier literature and has been in the focus of generative studies on English historical syntax since its beginnings (Lightfoot 1979). The authors cover new empirical ground by looking at word order possibilities of clauses containing adverbs, focussing first on lexical verbs and then on modals. Since the position of adverbs is considered to be one of the main diagnostics for verb movement from V to higher functional positions, for example to T, in generative grammar (since Pollock 1989) and this property distinguishes main verbs from auxiliaries in Present-Day English, word order possibilities observed in earlier stages are important in establishing changes in V-movement and in the steps of recategorization of modals. The data under discussion do not provide evidence that modals and main verbs are categorially different before the 16th century (modals have been claimed to be just like regular main verbs by Lightfoot 1979, or comprising a subclass of Vs by Warner 1993), but they reinforce their difference after that period, thus, the timing of the change proposed

16

BARBARA EGEDI AND VERONIKA HEGEDŰS

earlier is supported by independent evidence here. The chapter also discusses the reasons for the high frequency of SAdvV order and the lower frequency of SAdvMV order in Old English, considering the role of head-final TP and of a biclausal analysis of sentences containing modals may play, and concluding that these factors influence greatly what we can establish about V-movement based on adverb placement in early periods. In Tense recursion, perfect doubling, and the grammaticalization of auxiliaries, Ida Larsson and Ellen Brandner present an analysis of sentences containing double perfect in Germanic and the grammaticalization process of the perfect auxiliary that has led to its present semantic make-up. Two semantic types of perfect doubling are distinguished (anterior and superperfect double perfects) and their cross-linguistic distribution is explained by the various stages of grammaticalization and internal structure of the perfect auxiliary. The grammaticalization process presented here involves changes in the internal complexity of a lexical item and in argument structure. It is claimed that the development of double perfect is not a consequence of the loss of preterite; however, double perfect spreads when the preterite is lost, since at that time the finiteness of the verb is disconnected from tense. Syntactically, perfect doubling is argued to involve tense recursion, that is, multiple T heads, which involve multiple topic times. Building on Kayne’s (1993) proposal on the composition of HAVE as BE + P, the difference between English (which does not have perfect doubling) and for example German (which does) is claimed to be due to the difference of the auxiliary. In German—and other languages with auxiliary selection in the perfect—HAVE does not have a complex internal structure and only spells out BE (or Cop as the authors represent it), contrary to English, where it does spell out P as well, namely, a temporal preposition. The English-type perfect developed first, and some Germanic languages later developed auxiliary selection and simple HAVE, a grammaticalization process that led to a simpler internal structure in the composition of the auxiliary and thus to different selectional properties. Jóhannes Gı´sli Jónsson and Brynhildur Stefánsdóttir present a study of P-incorporation in the history of Icelandic. The chapter explores types of preposition movement, especially the process of preposition incorporation in Icelandic, showing that P-incorporation was possible until the 19th century, disappearing at the same time as Stylistic Fronting of Ps. This fits in with the general tendency that light elements tended to be left-adjacent to the verb in Icelandic up to that period. The authors argue that only heads of complement PPs could incorporate, adjuncts could not, and incorporation involved head movement into the verb or participle. Diachronically, they connect the loss

THE ROLE OF FUNCTIONAL HE ADS IN SYNTACTIC CHANGE

17

of P-incorporation to a change in the structure of complement PPs: while in earlier stages bare (lexical) PPs could fill the position of complements to verbs and participles, in Modern Icelandic the categorizing functional head p is always lexicalized, pP is always projected (see especially Svenonius 2003, 2010 on p). Therefore, the P head, where prepositions are inserted into the structure in Icelandic, cannot move across the p head into V as it would violate the Head Movement Constraint. This change in the lexicalization of PPs and the loss of P-incorporation is tied in with the loss of OV-orders in Icelandic. In general, the findings of this chapter also support previous proposals with respect to an underlyingly head-initial VP, from which an OV order is derived (in the spirit of Kayne 1994). The chapter From Old to Modern Icelandic: Dative applicatives and NP/DP configurationality by Heimir F. Viðarsson considers the case of dative applicatives from Old Norse to present-day Icelandic and argues that Old Norse had bare dative applicatives but they were lost and Icelandic uses different strategies for the encoding of such arguments. Modern Icelandic makes use of oblique subject constructions and prepositional phrases to encode constituents that were licensed as datives in Old Norse. Old Norse allowed for affected datives, datives with unaccusative verbs (only preserved idiomatically today) and datives with the copula. What have been analysed as High Applicatives and Low Applicatives can also co-occur in Old Norse. The change that led to the loss of this system is connected to the developments in the determiner system, which turned an NP-type language into a DP-type one. (This is different from previous assumptions about the loss of dative applicatives being related to the loss of morphological case systems.) This change and grammaticalization process is argued to be a case for the Head Preference Principle (Van Gelderen 2004) at work, manifesting in the reanalysis of phrases—mostly adjuncts—into functional heads, in this case within the newly configurational DP, which also developed a fully grammaticalized definite article. The development of a configurational DP resulted in the loss of discontinuous phrases and the loss of ‘floating’ datives as they have become integrated into a rigid DP structure. The chapter thus connects a change in the functional domain of one lexical category (N) to a change in the functional domain of another (V ).

1.7 Summary The chapters in this volume all discuss changes that concern functional elements: changes in parameter settings and feature specifications are explored in

18

BARBARA EGEDI AND VERONIKA HEGEDŰS

the functional sequence of the clausal as well as the nominal and adpositional domains. The split CP, the TP, and various functional projections in-between are all called on to account for the phenomena under consideration. In a sense this collection of chapters underscores Biberauer and Walkden’s (2015) summary of the changing theoretical views on syntactic change and their emphasis on the growing share of what they call ‘microdiachronic syntax’, which results in more focus on grammaticalization, functional categories, and features thereof. The changes proposed have implications for large-scale syntactic effects, many of which are explicitly discussed: word order variation and change, emerging (lexicalization of ) syntactic projections, changing information-structural properties are touched upon as starting points or end results of the ‘micro’ changes that are in focus throughout.

Acknowledgements We would like to thank the Hungarian Research Centre for Linguistics (former Research Institute for Linguistics) and the Hungarian Academy of Sciences for hosting and financially supporting the 16th Diachronic Generative Syntax conference, the papers of which gave rise to the chapters that constitute the volume in its present form. We are indebted to many colleagues all around the world for reviewing the chapters anonymously as well as to the reviewers from Oxford University Press, whose comments definitely helped the earlier versions to improve. Finally, we are especially grateful to Katalin É. Kiss for her constant encouragement and support.

2 Changing patterns of clausal complementation in Latin A parametric approach to ‘constructional’ changes Lieven Danckaert

2.1 Two apparently unrelated changes I will start by introducing the empirical data that constitute the focus of this chapter.¹ I first discuss the rise of finite declarative complement clauses, and I then turn to clausal complements of causative facio.

2.1.1 From infinitival to finite embedded declaratives In Early and Classical Latin (informally defined as the period from 200 BC until 200 AD), (non-factive) declarative complement clauses typically take the shape of an Accusatiuus cum Infinitiuo (henceforth AcI, cf. (1)), which as the name suggests feature an infinitive as the main verb, and an accusative noun phrase, which is typically (but not always) expressed overtly, as the subject.² (1)

cred-o [de-os inmortal-es spar-sisse anim-os believe-PRS.1SG gods-ACC immortal-ACC plant-PRF.INF souls-ACC

¹ The research reported on in this chapter was funded by the Research Foundation—Flanders (postdoctoral grant number FWO13/PDO/024). Parts of this chapter were presented at DiGS 16 in Budapest, LVLT 11 in Oviedo and at a research seminar at the University of Geneva. I would like to thank the audiences at these events as well as Liliane Haegeman and Luigi Rizzi for valuable feedback. Finally, I am grateful to two anonymous reviewers for some very constructive comments and suggestions. ² When glossing verbal morphology in Latin examples, I adopt the following conventions: as a rule, for all active and/or indicative verb forms, no voice and/or mood specification is given in the glosses, unless this information is really relevant. Passives, subjunctives, and infinitives are always explicitly glossed as such.

Lieven Danckaert, Changing patterns of clausal complementation in Latin. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Lieven Danckaert (2022). DOI: 10.1093/oso/9780198871538.003.0002

20

LIEVEN DANCKAERT

in corpor-a human-a]. in bodies-ACC human-ACC ‘I believe that the immortal gods have planted souls in human bodies.’ (Cicero, De senectute 77) In Late Latin (c.200–600 AD), declarative complements more and more often appear as finite clauses introduced by that-complementizers such as quoniam (2), quod, and quia, although AcIs are still well attested in all Late Latin texts: see among others Mayen (1889); Cuzzolin (1991, 1994); and Adams (2005). (2) non cred-is [quoniam sic pot-est not believe-PRS.2SG that so be.able-PRS.IND.3SG pasc-ere de-us […]]? feed-PRS.INF god-NOM ‘Don’t you believe that God can nourish in this way?’ (Augustine, Sermo 104) The earliest reliable example of a non-factive finite declarative is found in the anonymous Bellum Hispaniense (c.40 BC, cf. (3)), but the finite structure becomes productive only in the 3rd and 4th centuries AD. (3) Dum haec ger-untur legat-i while these.NOM.N.PL do-PASS.PRS.3PL ambassadors-NOM Carteiens-es renuntia-u-erunt [quod Pompei-um in from.Carteia-NOM report-PRF-3PL that Pompey-ACC in potestat-e habe-re-nt]. power-ABL have-IPFV.SBJV-3PL ‘When this was being done, ambassadors from Carteia announced that they had Pompey in their power.’ (Bellum Hispaniense 36.1) The elements quoniam and quia are originally adverbial conjunctions introducing (various types of ) because-clauses. Quod can have this function too, but importantly, even in very early Latin it was also used to introduce a range of (emotive) factive complement clauses. Examples include complements to predicates like paenitet ‘be sorry’ (4), gaudeo ‘rejoice, be happy’, and queror ‘complain’ (5): (4) Ait enim […] se paenite-re [quod say.PRF.3SG PRT REFL.ACC be.sorry-PRS.INF that anim-um tu-um offend-eri-t]. mind-ACC your-ACC offend-PRF.SBJV-3SG ‘He says that he is sorry that he has offended you.’ (Cicero, Ad Atticum 11.13.2)

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

21

(5) Saep-ius=que mihi ueni-eba-t in ment-em often-COMPAR=and me.DAT come-IPFV-3SG in mind-ACC quer-i [quod ita uiu-ere-m] quam complain-PRS.INF that like.this live-IPFV.SBJV-1SG than gaude-re [quod uiu-ere-m]. rejoice-PRS.INF that live-IPFV.SBJV-1SG ‘I used to think more often about complaining about the way I live than to be happy about the fact that I’m alive at all.’ (Cicero, Ad familiares 4.13.1) It is standardly assumed that the pattern illustrated in (4) and (5) later spread to non-factive environments, eventually taking over from AcIs and ut-clauses (neither of which survive in the Romance languages). In what follows, I will assume that this is correct.

2.1.2 From finite to infinitival causatives Interestingly, in the same period we witness a second shift which is to a large extent the mirror image of the previous development. More specifically, finite complements to causative facio ‘make’ (such as (6)), which feature a verb in the subjunctive mood and—optionally—the complementizer ut (roughly ‘that’), are fully productive in Classical Latin, but in later times they are gradually replaced by infinitival clauses which at first sight look the same as the Classical Latin AcI (7) (but which I will argue are underlyingly quite different): (6)

aut fortasse fec-isse-t [ut tu ex or maybe make-PLPRF.SBJV-3SG that you.NOM.SG out.of popul-i sermon-e excid-ere-s]. people-GEN speech-ABL fall.out-IPFV.SBJV-2SG ‘Or maybe he would have made you escape people’s criticism.’ (Cicero, Pro Flacco 82)

(7)

et in ips-a part-e faci-e-s iace-re and in self-ABL.F.SG part-ABL make-FUT-2SG lie-PRS.INF laborant-es. suffering-ACC.M.PL ‘And you will make the patients lie on the same side.’ (Cassius Felix, De medicina 44.8)

The finite causative strategy with ut is not preserved in any present-day Romance variety, but this is clearly because the complementizer/conjunction ut, which in (Classical) Latin introduces a variety of complement and adverbial

22

LIEVEN DANCKAERT

clauses, has completely disappeared. On the other hand, in most modern Romance languages (Daco-Romance being a notable exception), successors of Latin facio ‘make’ or mando ‘send’ can act as causative verbs selecting an infinitival clause.³ In many cases, the same predicates can also appear with a finite clause introduced by a KE-complementizer followed by a subjunctive verb form. This last construal is, however, a secondary innovation which I will not be concerned with here. The rise of the ‘facio + infinitive’ pattern is fairly well described: see among others Thielmann (1886); Muller (1912); Norberg (1945); Chamberlain (1986); Robustelli (1993, 2000); Biville (1995); Iliescu (1995); and Vincent (2016). As was the case with finite non-factive declaratives, there is a handful of very early attestations (mainly in poetry), the earliest (shown in (8)) dating from the 2nd century BC. This pattern too becomes productive from c.200 AD onwards. (8)

purpure-am=que uu-am fac-it alb-am purple-ACC=and grape-ACC make-PRS.3SG white-ACC pampin-um habe-re vine.shoot-ACC have-PRS.INF ‘And it (the sun ld) makes the pale vine-shoot have purple grapes.’ (Lucilius, fragment 1224 (Warmington))

As can be deduced from (8), in the earliest type of Late Latin infinitival causatives the Causee⁴ argument (agentive or otherwise) surfaces as an accusative. As is well known, in certain types of Romance causatives, agentive Causees are realized as datives (formally PPs), yielding the much-discussed faire à causative illustrated in (9) (cf. Kayne 1975 and subsequent literature): (9)

Marie fait mang-er la tarte [à Pierre]. Marie make.3SG eat-INF the cake to Pierre ‘Marie makes Pierre eat the cake.’

As will be elaborated on in Section 2.5.2.3, dative Causees only originated long after infinitival causatives with facio had become productive.

³ The literature on Romance causatives is particularly rich (accurately reflecting the complexity of the empirical landscape). Key references include Kayne (1975) and Guasti (1993); see Guasti (2006) and Sheehan (2016) for general overviews. In particular on varieties of Old Romance, see Robustelli (1992, 1994, 2000); Egerland and Cennamo (2010) (on Old Italian); Martineau (1990a, 1990b) and Pearce (1990) (on Old French); and Martins (2006) (on Old Portuguese). ⁴ Throughout this chapter, I (informally) use the term ‘Causee’ to refer to the highest (external or internal) argument DP associated with the embedded lexical verb of a given causative configuration, without prejudging anything more about the syntactic and thematic properties of this DP (see Section 2.5.2 for further discussion of the syntax and argument structure of various causative patterns).

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

23

2.1.3 The proposal in a nutshell The facts reviewed in the previous two sections are summarized in Table 2.1, in which the shading highlights the apparently antagonistic nature of the two changes. As indicated, I assume the combination of AcIs and finite causatives to be part of one single grammar, which I call ‘Grammar A’, and which is most prevalent in Early and Classical Latin (c.200 BC–200 AD). Correspondingly, I take it that both finite declarative complements and infinitival causatives are part of a second coherent system, Grammar B, which steadily gains ground during the Late Latin period (c.200–600 AD). The main proposal of this chapter is that except for the loss of ut-clauses after facio (which as we saw is related to the loss of the lexical item ut), the developments summarized in Table 2.1 are part of a single integrated change from one major grammatical system (Grammar A) to another (Grammar B). The key feature that differentiates the two grammars will be argued to be the way in which the clausal EPP-requirement is satisfied. Assuming a parameterized approach to EPP-checking in the spirit of Alexiadou and Anagnostopoulou (1998) and Biberauer (2003), I will propose that in Classical Latin, the entire VP moves to the functional projection where the EPP-requirement is satisfied. Concomitantly, subject DPs do not A-move out of the verb phrase, but are case-licensed in situ. Whenever they do evacuate the thematic layer (for instance for reasons related to information structure), the relevant operation is not an instance of (case-driven) A-movement. In contrast, in the innovative grammar, EPP-driven VP displacement is replaced by verb movement (coupled with optional A-movement of the subject DP). As a consequence, in Grammar A VP-internal embedded subjects do not canonically appear in a local configuration with the matrix predicate: this state of affairs will be shown to be compatible with the syntactic properties of the AcI. Next, I will argue that in the new grammar matrix verbs and embedded subject DPs are actually in a sufficiently local configuration: as a result, from Late Latin onwards (causative) ECM structures become possible and indeed available. Table 2.1 Changing patterns of clausal complementation in Latin: the very basic picture

24

LIEVEN DANCKAERT

The chapter is structured as follows. In the following section, I will further elaborate on the model of parameterized EPP-satisfaction. I proceed to discuss the most important properties of Classical Latin clause structure (or more precisely, of what I call ‘Grammar A’), and I detail my assumptions about the syntax of the AcI (Section 2.3). Special attention is paid to the way in which subjects (and objects) are assigned structural case. In Section 2.4, I present my analysis of Late Latin clause structure (‘Grammar B’). Next, I show how the constructional changes reviewed earlier fall out from the new clausal architecture. Section 2.6 concludes.

2.2 Background: parameters of EPP-checking I will assume that the basic structure of the (Latin) clause consists of a number of projections, which come in a fixed hierarchical order. The basic sequence is shown in (10) (in which no left-peripheral projections are represented):

(10)

SubjP Subj°

FP NegP

F°[EPP] Neg°

TP T°

VoiceP DPEA

Voice' Voice°

vP v°

√P √°

DPIA

Let me briefly comment upon the properties of the various heads in this representation. First, I take it that the whole structure is an extended projection of an a-categorial root (√), which can select a phrasal complement (an internal argument, IA), and on top of which a series of functional projections is merged. The first of these is a verbalizing head v (which characterizes the entire extended projection as verbal rather than, for instance, nominal). Next, there is a Voice head, which determines whether or not an (agentive) external argument (EA) is added to the structure. Higher up, I assume a Tense phrase (TP), a NegP (optionally) and most importantly in the present context, a functional projection

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

25

FP whose head is endowed with an EPP-feature, which, simply put, requires FP to be lexicalized overtly by some element bearing ϕ-features.⁵ Finally, high in the articulated inflectional layer there is a designated subject position, which can host ‘subjects of predication’ (in the sense of Cardinaletti 2004). I will adopt a parameterized approach to EPP-checking along the lines of Alexiadou and Anagnostopoulou (1998), which says that the EPP-requirement is universal, but that the way in which it is met varies cross-linguistically. An explicit system to model cross-linguistic variability in this domain was proposed in Biberauer (2003), Biberauer and Roberts (2005), Richards and Biberauer (2005), and Biberauer and Richards (2006).⁶ These authors propose a system with two independent parameters, one specifying the nature of the category satisfying the EPP-requirement (a head or a phrase), and one specifying whether or not movement to the functional projection carrying the EPP-feature (F(P) in my system) is coupled with pied-piping of additional material. This gives rise to a four-way typology which is summarized in Table 2.2. In all cases, the probing head F is in need of a Goal endowed with ϕ-features (a finite verb or a DP; Person-features are probably the most crucial for the purpose of EPP-checking). This system makes available three options to satisfy the EPP-requirement, namely (i) verb movement to F°, (ii) VP movement to SpecFP, and (iii) insertion of a(n expletive or otherwise) DP in SpecFP. As indicated, in certain circumstances the system allows for some degree of optionality. Languages that choose the verb movement option, such as Italian and Modern Greek, typically display rich agreement (and pro drop), and productively allow for postverbal subjects. In contrast, DP movement languages such as English have preverbal subjects and tend to have expletives. Finally, two types of languages have VP movement at their disposal to check F’s EPP-feature. In languages like German (‘head pied-piping’ languages), this Table 2.2 Parameters of EPP-checking (adapted from Biberauer and Roberts 2005; Biberauer and Richards 2006)

⁵ The reason why I assume the EPP-requirement not to be associated with T is related to the fact that in Latin Tense-final clauses, negation canonically intervenes between the verb phrase and an auxiliary in T. See the analysis of example (11). ⁶ The same system was also applied to Latin in Mackenzie and van der Wurff (2012): see Danckaert (2017: 236–8) for discussion of how the implementation of these authors differs from my own.

26

LIEVEN DANCKAERT

operation is argued to be the only option (with a fixed VPAux order as a result), whereas in a ‘spec pied-piping’ language like (colloquial) Afrikaans, VP movement freely alternates with DP movement to SpecFP. I refer to the original papers as well as to Danckaert (2017: 230–3) for additional discussion and illustration.

2.3 Clause structure in Grammar A (‘Classical Latin’) I will now apply the system outlined in the previous section to Classical Latin, which will give us a fair idea about the overall constituency of finite clauses and AcIs. I will also pay attention to the way in which subject DPs are assigned structural case in both of these environments.

2.3.1 Finite clauses 2.3.1.1 VoiceP movement to SpecFP Following proposals made in Danckaert (2017, 2018), I will assume that the grammar of Classical Latin only has VP movement (or more precisely, VoiceP movement (see below)) at its disposal as a means to satisfy the EPPrequirement. Importantly, however, contrary to what one might think on the basis of the discussion in the previous section, there are reasons to assume that in terms of the typology summarized in Table 2.2, Latin is better considered a ‘spec pied-piping’ than a ‘head pied-piping’ language (i.e. that what is probed for is not the verb, but rather the highest VP-internal argument DP), with the proviso that VP movement cannot optionally alternate with DP movement, resulting in obligatory pied-piping. In all cases, F probes down the tree and attracts the highest DP argument (the external argument in SpecVoiceP in the case of clauses with a transitive predicate, or the internal argument (complement of the root) in the case of passives), which pied-pipes the entire verb phrase to SpecFP. To be more precise, I take it that the displaced constituent corresponds to the VoiceP node. One reason why I prefer this slightly altered system is that it does not run into difficulties in non-finite contexts, where it is hard to see which verbal head endowed with φ-features would be available to satisfy F’s EPP-requirement (see also Danckaert 2017: 236–8).⁷ Infinitives ⁷ A reviewer points out that it might be conceptually attractive to retain the option of ‘free variation’ between VP and DP movement in finite clauses. However, as shown in Danckaert (2017: 243–7, 262– 4), there are empirical reasons to doubt whether this is on the right track. Specifically, in the period from 200 BC until 200 AD internal arguments in active and passives cannot be differentiated on the basis of their word order preferences, suggesting that ‘derived subjects’ are not more likely to evacuate the verb phrase than direct objects in transitive clauses (contrary to what an ‘optional DP movement’ hypothesis would predict). Interestingly, such a contrast between internal arguments in active and passive clauses

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

27

especially are a point in case, as these do not display any φ-agreement at all.⁸ No such problems arise if one assumes that a DP is probed for, (i) as (overt as well as covert) DPs presumably always have a set of φ-features, and (ii) because each VP contains at least one DP. I will come back to the diachronic implications of EPP-checking in infinitival clauses in Section 2.5.1. An obvious correlate of a derivation involving EPP-driven VoiceP movement is the characteristic ‘VP(-Neg)-Aux’ word order of Classical Latin, which we see in (11): (11)

[VoiceP

Roman-us equitatu-s ips-um quidem reg-em Roman-NOM cavalry-NOM self-ACC PRT king-ACC Elati-ae adsecut-us] non est. Elatia-LOC reached-NOM.M.SG not be.PRS.3SG ‘The Roman cavalry did not manage to find the king himself in Elatia.’ (Livy, ab Urbe condita 36.19.10)

The structural representation that I will assume for examples featuring the word order ‘VP-Neg-Aux’ is as in (12) (overt terminals in boldface), where the VP-internal operation displacing √P to SpecvP is a case of roll-up movement (Biberauer et al. 2014), which gives rise to the order OV:

(12)

CP C°

FP F'

VoiceP Voice'

DPEA

Voice°

F°[EPP] vP

Neg° vP

√P t√°

DPIA

NegP

v°

TP Aux

tVoiceP

t√P

√° v°

does in fact arise in later centuries, in line with the characterization of Grammar B that will be offered below. ⁸ Past participles, on the other hand, do in fact agree with the subject in Gender, Number, and Case (but not Person), as can be seen in the example in (11).

28

LIEVEN DANCKAERT

In this structure, the position of non-finite verbs (the complex head √/v) with respect to negation (Neg°) can be taken to be indicative of a high landing site of the moved verb phrase (Danckaert 2014), as well as of the phrasal nature of the movement operation that gives rise to the surface order V-Aux (head movement of the non-finite verb past negation being ruled out by the Head Movement Constraint, see Danckaert 2017: 45–62). In other words, movement of VoiceP in VPAux clauses is not of the strictly local ‘roll-up’ type which in the tree in (12) is responsible for bringing about the complement-head orders inside the extended VP (cf. the OV order). In Danckaert (2017, 2018), this set of assumptions is shown to be compatible with the observation that Classical (but not Late) Latin allows for the order ‘VOAux’, which, following Biberauer et al. (2014), can be assumed not to be possible in a grammar where roll-up movement is the canonical way to generate head-final surface orders (VPAux, in this case). 2.3.1.2 Case assignment An important consequence of the analysis just sketched is that argument DPs never undergo A-movement, EPP-driven or otherwise (see also Julien 2002 for a similar approach to argument licensing in Tense-final languages). Instead, I will take it that all arguments are case-licensed in situ, by virtue of an Agree relation between the relevant phrase and a c-commanding functional head, which is established as soon as the latter is merged. More specifically, I assume that if Voice hosts an external argument in its specifier it also assigns accusative case to the internal argument (cf. Burzio’s Generalization). On the other hand, nominative case is assigned by T to the closest caseless DP in its c-command domain. For a finite clause with a transitive verb, the picture we thus get looks like that in (13) (matters of word order (roll-up movement) aside):

(13)

TP T' Tfin° [K: NOM]

VoiceP DPEA [K: __ ]

Voice' Voice° [K: ACC]

vP √P

v° √°

v°

t√°

DPIA [K: __ ]

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

29

I assume that structural case assignment involves an Agree relation between a probing functional head and some Goal DP (cf. the dashed arrows in (13)). Importantly, I take it that this Agree relation does not have to involve a strictly local (spec-head or head-complement) configuration, but that it is subject to the following constraints:⁹ (14)

Locality conditions on structural case assignment ( first version) A case assigning Probe X can case mark a Goal YP iff (i) X c-commands YP, (ii) YP is categorially a DP, (iii) YP does not bear inherent case (i.e. is caseless), (iv) no CP-boundary intervenes between X and YP, and (v) there is no caseless ZP, also of category D, such that X c-commands ZP and ZP dominates or c-commands YP.

Put differently, I assume that case assignment involves a ‘sufficiently local’ configuration, in which the Probe and the Goal are not separated by any clause boundaries or intervening competing DPs.¹⁰, ¹¹

2.3.2 The syntax of the AcI 2.3.2.1 Constituency I will now say a couple of words about the constituency of infinitival clauses. First of all, assuming that the proposed mechanism of EPP-checking applies to all non-defective clauses (i.e. clauses with full functional structure) indiscriminately, the same operation of VoiceP movement to SpecFP should ⁹ On non-local case assignment, see Alexiadou, Kiss, and Müller (2012) and references cited there. Interestingly for the upcoming discussion (cf. Section 2.5), some ECM environments seem to constitute cases where structural case is assigned at a distance. One potential such case would be Icelandic examples such as (i) (from Sigurðsson 2000; cf. McFadden 2009: 113–14, fn. 11), where the embedded subject fjóra laxa ‘four salmons’ is case marked by the believe-predicate in the matrix clause, without being linearly adjacent to it (of course it remains to be seen what the underlying syntax of this type of example is). (i)

Þeir to¨ldu [líklega hafa verið veidda [fjóra laxa]]. they believed probably have been caught four salmon.ACC.PL (approx.) ‘They believed there to have probably been caught four salmon.’

For related discussion, see also Raposo and Uriagereka (1990); Ura (2007); and McFadden (2009). ¹⁰ On the fourth clause of the definition in (14), see Sigurðsson (2012: 207): ‘CPs are A-islands; that is, A-relations, including T-licensing, are blocked from being established across C-boundaries.’ This requirement can presumably be reformulated in terms of phase theory (see Chomsky 2001 and subsequent literature). See also Danckaert et al. (2016) (and references cited there) for potential counterexamples to this generalization. ¹¹ Alternatively, if CPs can bear (structural) case (Picallo 2002; Nunes 2008), clause (ii) of (14) should be reformulated as ‘YP is categorially a DP or a CP’; clause (v) would also have to be updated accordingly. Same remark for (27) below.

30

LIEVEN DANCKAERT

take place in infinitival clauses too. In relation to this, note that there is good evidence that AcIs are categorially CPs rather than TPs. Concretely, given the discussion of Latin clausal pied-piping in Danckaert (2012: chs 4 and 5), it seems to be the case that Latin AcIs are endowed with a left-peripheral space which can host syntactic operators. The relevant data involve pied-piping of an AcI under wh-movement, whereby a wh-moved operator extracted from within the AcI first undergoes what is known as ‘internal’ (or secondary) whmovement to the edge of the AcI, an operation which is followed by movement of the entire embedded clause to the left periphery of the next higher CP (see Danckaert 2012: 169–70, and especially the tree in (413)''). The structure is such that the actual operator surfaces in the left periphery of the AcI, but crucially is not interpreted there, its scope position being located in the left periphery of the clause that ends up hosting the pied-piped AcI. Given the standard assumption that CPs but not TPs constitute cyclic domains for the purpose of long distance phrasal movement, these clausal pied-piping facts strongly suggest that AcIs are indeed CPs. Let us now consider how the subject of an AcI is case marked. Importantly, there is a broad consensus that the characteristic accusative case of the subject is not to be explained in terms of a structure whereby the predicate of the matrix clause assigns case to the embedded subject, yielding what one could call an ‘Exceptional Case Marking’ (ECM) configuration (see Bolkestein 1976, 1979; Pillinger 1980; Cecchetto and Oniga 2002; Jøhndal 2012). The arguments in favour of this position are numerous and indeed compelling. For instance, whenever an AcI appears as the complement of a predicate which canonically assigns dative case to DP complements (like e.g. confido ‘trust’, cf. (15)), the subject of the AcI still bears accusative case (16): (15)

ne quis [fide-i Roman-ae aut not anybody.NOM.M.SG promise-DAT Roman-DAT or societat-i] confid-a-t alliance-DAT trust-PRS.SBJV-3SG ‘No man should ever trust a Roman promise or alliance.’ (Livy, ab Urbe condita 21.19.10)

(16)

Sed confid-o [te / but trust-PRS.1SG you.ACC.SG fac-tur-um]. do-PTCP.FUT-ACC.M.SG ‘But I trust that you will do this.’

*tibi you.DAT.SG

esse be.PRS.INF

(Cicero, Ad Atticum 3.3)

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

31

Second, AcIs readily appear in other contexts where there is no category that could assign accusative case. For instance, they can be a complement to a noun (17), and they can appear as arguments of one-place predicates such as expedit ‘be useful’ (18) and adjectives like manifestum ‘clear, obvious’ (19): (17)

haec autem opinatio est iudicatio DEM.NOM.F.SG PRT opinion.NOM be.PRS.3SG judgment.NOM [se sci-re qu-od nesci-a-t]. REFL.ACC know-PRS.INF what-ACC.N.SG not.know-PRS.SBJV-3SG ‘However, this opinion amounts to thinking that one knows what one does not know.’ (Cicero, Tusculanae disputationes 4.26)

(18)

omn-ibus enim bon-is expedi-t [salu-am all-DAT.M.PL PRT good-DAT.M.PL be.useful-PRS.3SG safe-ACC esse re-m public-am]. be.PRS.INF cause-ACC public-ACC ‘It is to the benefit of all good citizens that the state is safe.’ (Cicero, Philippicae 13.16)

(19)

[Hunc uepr-em]i manifest-um est [ti DEM.ACC.M.SG thorn.bush-ACC clear-NOM.N.SG be.PRS.3SG interim-i non pos-se, nisi radicitus kill-PASS.PRS.INF not be.able-PRS.INF unless completely effod-ere uel-i-s]. dig.out-PRS.INF want-PRS.SBJV-2SG ‘It is clear that this type of thorn-bush cannot be destroyed unless you dig it out completely.’ (Columella, De agricultura 11.3.7)

Finally, note that the existence of subject-to-subject raising with passives such as dicitur (‘X is said to …’), yielding a so-called Nominatiuus cum Infinitiuo (NcI), alongside AcIs selected by the same predicate (‘they say that X …’), cannot obviously be construed as evidence in favour of an ECM analysis of the latter. The logic of the argument would be that in both the AcI and the NcI, the embedded subject is actually the internal argument of the matrix predicate (assuming a subject-to-object raising analysis of what is here called ECM, as in Postal 1974). However, even in English the availability of raising across passive predicates with an infinitival complement does not always depend on the availability of an ECM ‘input’ structure. For instance, whereas a predicate like expect allows for ECM when active and subject raising when passive, say and hope can never appear in an ECM configuration, which does not preclude the availability of raising with be said (and more marginally also with be hoped,

32

LIEVEN DANCKAERT

in particular with wh-subjects (see Danckaert et al. 2016: 146, fn. 1)). On the other hand, want allows for ECM, but be wanted not for passive raising (Kayne 1984: 37, 45, fn. 33). In other words, at least in English there is a double dissociation between ECM and passive raising. Whether or not a given verb qualifies as a (passive) raising and/or ECM predicate seems to be lexically (and thus idiosyncratically) specified, on a case to case basis.¹² To conclude, it seems clear that an ECM analysis of AcIs is not on the right track. This is a welcome result, given the constraints on structural case assignment listed in (14): recall that by assumption, case assignment cannot take place across a CP-boundary. Given that there is robust evidence that AcIs are CPs, any account whereby an argument inside the AcI receives case from a functional head in the higher clause would have to be rejected on independent grounds. 2.3.2.2 Case assignment revisited: a brief excursion But then how does the embedded subject end up being case marked as an accusative? To answer this question, consider first the structure in (20).¹³ As indicated, I take it that there are two potential case assigners, viz. (i) the matrix verb and (ii) a (phonologically null) case assigner in the embedded left periphery (identified here as C°):

(20)

Voice° matrix [K: ACC] CP C° [K: ACC] ??

FP

VoiceP DPEA Voice' [K: __ ] Voice° vP

F' NegP

F°[EPP] Neg°

TP Tinf°

tVoiceP

¹² Compare also Jøhndal (2012: 77): ‘Some evidence suggests a third solution, that two distinct lexical entries are involved for the NcI-passive and the AcI-passive.’ ¹³ In this structure I only consider case assignment of an external argument in SpecVoiceP. The same logic can be applied to passives/unaccusatives where the subject to be case marked would be an internal argument lower in the structure.

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

33

The first possibility can be discarded on the basis of the discussion in the previous section. Let us then consider whether an analysis involving a null case assigner in the left periphery of the AcI—say the covert counterpart of English for in a structure like For John to leave would be a good idea—could fare better.¹⁴ However, although I do not have any principled objection against a solution along these lines, there are reasons to doubt whether it could work in the particular case of the Latin AcI. More specifically, it is not clear whether in a structure like (20) an Agree relation can be established between a Probe in the main projection line and a Goal which is embedded inside a left branch, even though all the requirements listed in (14) are met. Consider the following pair of examples, which illustrate two types of English infinitival clauses with an overt subject. (21)

a. I consider [TP [DP John/him] to be foolish]. b. We all hope [CP for [TP [DP John/him] to win the elections]].

ECM for-to infinitive

In both structures, the embedded subject receives accusative case from a c-commanding functional head, viz. (i) matrix Voice° (lexicalized by the predicate consider) in the ECM-configuration in (21a) (where I assume that the infinitive is a TP rather than a CP, and that the subject DP is in SpecTP (Stowell 1982)), and (ii) the ‘dummy’ preposition for in (21b). The relevant structural configurations are given in (22):

(22) a.

VoiceP Voice° [K: ACC] DP [K: __ ]

b.

C° [K: ACC]

TP

TP

DP [K: __ ]

T' T°

CP

T°

T' VoiceP

VoiceP

Interestingly, as suggested by the examples in (23), embedded subjects that sit in the specifier (Spec1) of another specifier (Spec2) can apparently not be case marked by a c-commanding functional head located higher than Spec2 (on cases like (23a), see also Munn 1999: 662–3, who refers to Stowell 1981).

¹⁴ On null case assigners in AcI-like environments, see Kayne (1984: 38) and Sevdali (2013). A variant of this line of analysis (involving covert V-to-C movement) is applied to Latin AcIs in Cecchetto and Oniga (2002), who base themselves on Longobardi (1996).

34

LIEVEN DANCKAERT

In both examples, the target for case assignment is the subject of a subject infinitive: (23)

a. * I consider [TP [TP [DP John] to be foolish] to be obvious] (compare I consider John’s foolishness to be obvious). b. * We all expect [CP for [TP [TP [DP John] to win the elections] to create new opportunities for our country] (compare We all hope for John’s (contingent) election to create new opportunities for our country).

I hasten to add that it is not immediately obvious that a failure of case assignment (rather than some other factor) is indeed the cause of the ungrammaticality of these examples. However, note that the unacceptability of (23a) is probably not the result of a general ban on to-infinitives to appear in the canonical subject position. Native speakers that I consulted find the examples in (24), which arguably involve an infinitive with some sort of ‘arbitrary PRO’ subject, markedly better than those in (23) (note in particular the contrast between (23a) and (24b)), although it has to be said that speakers clearly prefer an -ing form (being foolish) as the subject:¹⁵ (24)

a. ? To be foolish is a bad thing. b. ? I consider to be foolish to be a bad thing.

In addition, observe that in both examples in (23), no phonologically overt material intervenes between the case assigner and the embedded subject: in other words, the well-known requirement that in English, accusative subjects need to be string adjacent to their case assigner (Stowell 1981) is met. On the other hand, on the basis of what we said earlier, the ungrammaticality of these examples is unexpected, given that none of the constraints listed in (14) above seems to be violated. In other words, (14) will have to be updated in such a way that it rules out configurations like (25a), in which the circled WP can be considered the ‘offending’ node.¹⁶ However, we don’t want to be overly restrictive,

¹⁵ On -ing forms acting as subjects, see Chierchia (1985). One could object that it is not clear whether in (24a) the infinitive does indeed sit in the canonical TP-internal subject position rather than in some left-peripheral topic position. This problem does not arise for (24b) however, as ECM-complements are in all likelihood not endowed with a left-peripheral space of their own. Thanks to Ji Young Shim (p.c.) for pointing this out to me. ¹⁶ I am—perhaps not entirely innocuously—assuming that it is immaterial whether WP in (25a) is an internally or an externally merged specifier.

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

35

as (25b), where the path from X° to YP also goes over the same two maximal projections ZP and WP, should still be fine.¹⁷

(25) a.

b.

XP X°

XP X°

ZP

ZP

WP YP

Z' W'

W°

Z° ...

Z° ...

WP YP

W' W°

...

Ideally, we will want to offer a solution without making direct reference to notions like ‘left’ and ‘right’, which I assume are irrelevant for core syntactic operations like Agree. Before I proceed, let me just point out that phase theory does not seem to have anything to say about the difference between (25a) and (25b). Imagine for instance that W is a phase head and that none of Z, X, or any of the heads in between those two are phase heads: under such a scenario, the edge of WP should be equally accessible for a probing head located outside WP when the latter c-commands Z (25a) as when it is a sister to it (25b). Alternatively, if Z is the only phase head in the structure, the establishment of a Probe–Goal relation between X and YP is blocked only in (25b), not in (25a) (which is the opposite of what we are trying to derive). The solution that I would like to propose builds on the notion of ‘Extended Projection’ (henceforth EP), as developed in Grimshaw (1991, 2000, 2005). More specifically, I will update the set of conditions given in (14) with an additional clause stating that although the Probe and the Goal need not be part of the same EP, all nodes intervening between the two have to be a member of the same EP as the Probe. Before going there, let me offer a number of definitions. First, Grimshaw (2005: 4) defines the notions of ‘head’ and ‘projection’ as follows (italics in original):¹⁸ ¹⁷ The issue is arguably related to the broader question as to whether it is possible for a Probe to enter into an Agree relation with a Goal embedded deeply in a left branch (cf. the very rich literature on (the ban on) phrasal movement out of subjects, or left branches more generally; see Haegeman et al. 2014 for recent discussion): in a nutshell, in some cases it clearly is possible to subextract phrasal material from within subjects, suggesting that the locality conditions governing the operations ‘Move’ and ‘Agree’ are not entirely equal. ¹⁸ Note that I am abstracting away from Grimshaw’s (2005: 3–5) discussion of ‘F-values’: in particular I am not adopting the additional requirement that YP can only be a(n extended) projection of X if

36

(26)

LIEVEN DANCKAERT

X is a head of YP, and YP is a projection of X iff: a. YP dominates X b. The categorial features of YP and X are consistent c. There is no inconsistency in the categorial features of all nodes intervening between X and YP (where a node N intervenes between X and YP if YP dominates X and N, N dominates X).

Next, the author makes a further distinction between ‘perfect’ and ‘extended’ projections. Simply put, the perfect head of a phrase XP is the closest head dominated by XP; all of its other (i.e. lower) heads are extended heads. Similarly, a perfect projection of a head is the closest phrasal node dominating it; all of its other (i.e. higher) projections are extended projections. Here I will—slightly informally, but in accordance with common usage—call the set of extended heads and projections of a lexical head L ‘the extended projection of L’. Crucially, Grimshaw (2000: 117–18) adds that ‘it may be necessary to add the requirement that only complements, and not specifiers, participate in extended projections, by requiring that all maximal projections intervening between Y and X be complements’. If we assume that this is indeed correct, we can reformulate (14) as in (27). (27)

Locality conditions on structural case assignment ( final version) A case assigning Probe X can case mark a Goal YP iff (i) X c-commands YP, (ii) YP is categorially a DP, (iii) YP does not bear inherent case (i.e. is caseless), (iv) no CP-boundary intervenes between X and YP, (v) there is no caseless ZP, also of category D, such that X c-commands ZP and ZP c-commands¹⁹ YP, and (vi) there is no WP such that (a) X c-commands WP and WP dominates YP, and (b) WP and X are not part of the same extended projection.

Crucially, (27) does not stipulate that X and YP themselves have to be part of the same EP, which correctly allows for a YP in a specifier to be case marked by X (as long as all requirements in (27) are met). the F-value of the latter is not higher than that of the former. This conception of ‘extended projection’ is in line with that assumed in Biberauer et al. (2014). ¹⁹ Note that given (26c), cases in which a caseless DP which is c-commanded by X and which dominates (rather than c-commands) YP are subject to clause (vi) of (27).

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

37

With this in place, we are in a position to account for the different status of WP in (25a) and in (25b): only in the latter configuration can WP belong to the same EP as the probing head X (namely when all requirements listed in (26) are met), whereas in the former, it always acts as a ‘barrier’ which prevents X from case marking YP. As a result, if the reasoning developed above is on the right track, we should rule out both patterns schematized in (22) as possible case marking configurations involved in the derivation of the Latin AcI, namely (i) the ECM-pattern (which we had already discarded on independent grounds) and (ii) the null case assigner in the embedded C layer. More generally, all derivations in which a(n embedded) subject has to receive structural case after VoiceP is moved to SpecFP (and thus where the case assigning head sits higher than the embedded FP) are expected to lead to ungrammaticality. Two alternatives come to mind. The first—which I will adopt here—involves accusative case assignment by the embedded non-finite T, essentially following Jøhndal (2012: 79–81) (see also Bolkestein 1979 and Miller 1993: 259 fn. 1, among others).²⁰ Although it definitely remains to be seen why and under which conditions non-finite T can assign accusative case,²¹ this solution does have the advantage that the case assigner (T) and the embedded subject (in SpecVoiceP) are in the right structural configuration at the point where the former is first merged. The structure would be exactly as in (13), the one difference being that the case assigned is accusative rather than nominative. The second alternative is to say that the embedded accusative is actually not a structural but rather a ‘default’ case (Pillinger 1980), which occurs (as a kind of ‘last resort’) whenever for a given argument DP no structural case assigner is available. One could object that the fact that the subject of an AcI systematically appears as an accusative casts doubt on such a last resort scenario (see Jøhndal 2012 for further discussion). I will not further pursue this second alternative here. In any event, what matters for the present purposes is that there are good reasons to assume that the source of the accusative case of the subject of an AcI is not a functional head situated higher than the embedded FP. Once this is agreed upon, the question as to why subjects of AcIs surface as they do is to some extent orthogonal to the main concerns of this chapter.

²⁰ A similar pattern of case assignment has been argued to be available in Irish non-finite clauses (Chung and McCloskey 1987). ²¹ Note that it is not a general property of Latin infinitival T to assign accusative case to its subject, as overt subjects of historical infinitives (which canonically appear in root clauses) are always marked for nominative case.

38

LIEVEN DANCKAERT

2.3.2.3 Intermediate summary At this point we can conclude that three factors contribute to the fact that the grammar of Classical Latin (‘Grammar A’) can generate AcIs. The first is the CP status of infinitival clauses, which is one of the reasons why case assignment to the embedded subject from within the matrix clause is not an available option. The second is the fact that the ‘Goal parameter’ in the parameterized EPP system summarized in Table 2.2 above is set at the ‘XP’ value: as a consequence, the non-finite character of the hierarchically highest verb of an AcI (and concomitantly, its lack of ϕ-features) does not prevent the EPP-requirement from being satisfied. Third, the fact that the second EPP parameter (± pied-piping) is set at the [+ pied-piping] value entails that the embedded subject is removed from the main projection line. As a result of this, subjects in Grammar A occur deeply embedded inside a complex specifier, a position which is inaccessible for a case assigning Probe c-commanding the embedded FP. To conclude, the availability of AcIs can be considered a result (not to say a ‘conspiracy’) of a number of independent factors. As we will now see, the parameters involved are set at different values in Late Latin, which has repercussions for—among other things—the system of clausal complementation.

2.4 Clause structure in Grammar B (‘Late Latin’): new ways of EPP-checking 2.4.1 From Grammar A to Grammar B: structural reanalysis and its consequences Building on Danckaert (2017, 2018), I assume that the main difference between the grammar of Classical and Late Latin is a resetting of the EPP parameters, which entails a major reorganization of the basic structure of the clause. In particular, I will assume (i) that in the new grammar (Grammar B, which I take to be the most prevalent one in the Late Latin period (c.200–600 AD)), F probes for a verbal head rather than for a DP, and (ii) that this verb does not pied-pipe any material when it moves to F. The result is a grammar with EPP-driven V-to-F movement, an operation which can optionally be accompanied by movement of the subject DP across the verb (for reasons related to information structure rather than the EPP proper). As argued elsewhere (see again Danckaert 2017: 248–54, 2018), the innovations just sketched are themselves the result of a series of independent developments. The key change that sets the whole process in motion is

39

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

weakening of the preverbal negator non, which loses its status as a free standing morpheme and becomes proclitic to the T-head. Syntactically, this process of procliticization involves head movement from T to Neg, resulting in the formation of a complex head ‘Neg-T’. This change can be considered an entirely regular stage of Jespersen’s cycle (‘weakening of a preverbal negator’). Importantly, this development has two major consequences. First, the preverbal negator now no longer constitutes a locality barrier for verb movement to F: in the new grammar, the verb in T first moves to Neg and then to F, where it can effectively serve to satisfy the EPP-requirement.²² Crucially, movement of the complex ‘Neg-T’ to F and absence thereof result in the same surface string. Put differently, learners now have to decide whether they posit a grammar where the highest verb only moves to T (and further to Neg in negated clauses), or a grammar where it moves a bit higher to F. With Danckaert (2017: 252–3) I take it that given that, in a Grammar B setting, a finite verb is no longer just a potential EPP-satisfier (by virtue of its having ϕ-features), but also in all cases the closest possible goal for F, the child learner should postulate verb movement all to the way to F, which effectively makes further phrasal movement to SpecFP superfluous. The new structure is detailed in (28), which shows the tree for an ‘S-Neg-AuxOV’ clause in Grammar B.

SubjP

(28)

DPEA

Subj' Subj°

FP NegP

F°[EPP] Neg° F° Neg°

T°

tNeg°

TP tT°

VoiceP Voice'

tDPEA Voice°

vP v'

√P √° DPIA

v°

t√P

t√° v°

²² Recall that no such thing is possible in Grammar A, as movement from T to F across the free standing head negator non results in a violation of the Head Movement Constraint.

40

LIEVEN DANCKAERT

A second result of Neg-incorporation is reanalysis of VoiceP movement to SpecFP as a ‘roll-up’ structure, where movement of VoiceP is entirely unrelated to EPP-checking, despite giving rise to the same surface order ‘VPAux’ as in Grammar A. I take it that this type of ‘snowballing movement’ always sets out at the bottom of an extended projection, and that it can then optionally apply to higher nodes, with the proviso that no projection can ever be skipped (cf. Biberauer et al. 2014). As mentioned, this analysis correctly predicts that clauses featuring the order ‘VOAux’ should be very rare in Late Latin texts. In later stages, the ‘height’ of roll-up movement was gradually reduced, leading to the eventual loss of complement-head sequences in the Romance clause. This scenario correctly explains why VPAux orders were lost earlier than OV orders (see again Danckaert 2017 for full details). To conclude, a Late Latin clause like (29) which features the order ‘SOV(Neg-)Aux’ (which I assume for the sake of the argument to be the output of Grammar B) seems at first sight very similar to its Classical (Grammar A) Latin counterpart (cf. (11)). However, both structures can actually be assumed to have very different properties (see Danckaert 2017: 254 for a detailed tree structure of the example in (29)). (29)

postquam Maroue-us episcopatu-m urb-is after Maroueus-NOM bishophood-ACC city-GEN est adept-us obtained-NOM.M.SG be.PRS.3SG ‘after Maroueus had become bishop of the city’ (Gregory of Tours, Historia Francorum 9.40)

2.4.2 Grammar A and Grammar B compared Table 2.3 offers a summary of the main properties of the two systems just discussed. Most importantly, the new parameter settings related to the EPPrequirement (‘X° attracted’ and ‘no pied-piping’) give rise to a grammar

Table 2.3 Classical and Late Latin clause structure compared

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

41

where a verbal head moves to F to check the EPP. As a corollary, subject DPs canonically stay inside the main projection line, unless some optional operation of for instance roll-up movement displaces VoiceP (or a phrasal category containing VoiceP) to a left branch.

2.5 Explaining the changes With all this in place, we are now ready to solve the main puzzle of this chapter, namely the changes in clausal complementation introduced in Section 2.1. I will first show how the loss of the AcI can be made to follow from the more general reorganization of Latin clause structure discussed in Section 2.4. I then turn to causatives.

2.5.1 The loss of the AcI: the EPP in non-finite environments Recall that the EPP is conceived of here as the need for the functional projection FP to be lexicalized by an element bearing φ-features. As pointed out above, in Classical Latin the finiteness of (the highest verb of ) a clause is of no importance to matters of EPP-checking, as the element probed for is one of the verb’s arguments. Things are very different in the new grammar, which sharply differentiates finite verbs (displaying full subject–verb agreement) and non-finite ones, as only the former can satisfy the EPP. Given that the grammar does not make available any alternative strategy for EPP-checking (assuming optional DP movement to SpecSubjP not to be driven by the EPP), the loss of the AcI follows trivially.²³ Needless to say, the process of one grammar replacing the other is very slow and gradual. It is therefore not surprising that AcIs are still regularly attested after 200 AD. However, in Late Latin they clearly start to decline, witness the fact that in this period finite complement clauses introduced by quod, quia, and quoniam become productive also in non-factive environments. I take it that the increased productivity of the finite complementation strategy is a direct result of the loss of the AcI.

2.5.2 From AcI to ECM The second instance of syntactic change that we still have to account for is the genesis of non-finite (infinitival) causatives, as well as the loss of ut-clauses ²³ For some recent alternative analyses, see Goldbach (2003, 2008) and Roberts (2007: 161–75).

42

LIEVEN DANCKAERT

after facio. As mentioned in the introduction, the second development is arguably related to the complete loss of the lexical item ut, and is probably unrelated to the shift from Grammar A to Grammar B, not least because it is very common for one predicate to appear with both finite and non-finite complement clauses. As to the new type of infinitival complement, what I would like to propose is that the old AcI was so to speak ‘recycled’ and transformed into a new structure, which has a more limited distribution than its historical source.²⁴ More specifically, I take it that the new pattern is of the ECM type, which requires the presence of a(n accusative) case assigner in the matrix domain. Two structures in which such a matrix case assigner is present are (i) causatives with facio and (ii) complements to perception verbs. Both of these survive as ECM structures in present-day Romance.

2.5.2.1 The syntax of (early) ‘facio + infinitive’ causatives Before looking at the details of the proposed process of syntactic reanalysis, let us first consider what the syntactic status is of ‘facio + infinitive’ causatives. First of all, there are reasons to doubt that this pattern is to be analysed as a genuine AcI. For one thing, it seems very unlikely that an AcI-type ‘facio + infinitive’ causative should become productive at exactly the same time at which the unambiguous AcI pattern (which can occur with verbs of saying/believing, nouns, adjectival and impersonal predicates, etc.) itself starts to decline (recall that the earliest non-finite causatives date from roughly the same period as the first attestations of non-factive finite embedded declaratives). Second, given the syntactic analysis just outlined, all genuine AcIs are predicted to disappear once the grammar which employs V-to-F movement to satisfy the EPP has become fully established. We know, however, that infinitival causatives did not die out, which suggests that they are not AcIs. Imagine now a situation in which a language learner who has access to both Grammar A and Grammar B (which we can assume to have been in competition during an extended period of time): a number of things can happen when a child is exposed to an AcI in the Primary Linguistic Data (PLD). First of all, whenever Grammar A is put to use, the relevant string could still be

²⁴ The idea that the Latin AcI and certain Romance infinitival clauses are historically related is of course not new: see e.g. Bourciez (1956: 110).

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

43

analysed as an AcI, with exactly the same structure as the one discussed earlier. On the other hand, when Grammar B is selected, it is no longer possible to postulate the structure schematized in (20), and the child can only impose an alternative parse on the same string (or reject it altogether). Three such alternative configurations featuring an infinitive and an accusative DP come to mind, namely (i) ECM, (ii) a monoclausal structure in which the infinitive and a matrix verb form a complex predicate, and (iii) object control.²⁵ The exact properties of these structures will be elaborated on below. What I will propose is that in Grammar B some AcIs were reanalysed as ECM structures, a pattern that is still available in (some) present-day Romance varieties (but see Section 2.5.2.3 for some thoughts on the diachrony of ECM in Romance). The reasons why I think this scenario is viable are the following. Let me start by considering whether an analysis in terms of object control is appropriate for ‘facio + infinitive’ causatives.²⁶ I will define object control as a syntactic configuration in which a three-place verb selects both a nominal and a clausal internal argument (the exact category (and thus ‘size’) of the latter need not concern us here), and in which the nominal argument receives a theta role from the control predicate rather than from the infinitive (Jøhndal 2012: 54–6). In Latin, a likely candidate to act as a matrix predicate in an infinitival²⁷ object control configuration is cogo ‘force’. This being said, in actual practice it is often quite difficult to tell apart object control from other structures featuring an accusative and an infinitive, and some of the diagnostics that have been discussed in the literature on Latin (most prominently in Bolkestein 1976, 1979) are unlikely to be reliable. Specifically, to illustrate syntactic and semantic restrictions on object control (‘nominal object + an infinitival complement’, in her terms), Bolkestein (1979: 21) reports a number of contrasts between a predicate like dico ‘say’ followed by an AcI and cogo ‘force’ with a DP complement and a control infinitive. Specifically, the author claims that in cases of object control, the Causee must be sentient (or it should at least be the case that it can plausibly be construed as such), and able to influence/control the course of the event expressed by ²⁵ Note that this taxonomy of (Latin) ‘ACC + INF’ structures (to which we have to add genuine AcIs) is slightly richer than the two-way classification assumed in, for instance, Bolkestein (1976, 1979) and Vincent (2016), which does however seem to be sufficient to capture the Classical Latin facts (my Grammar A). ²⁶ On object control structures as the historical source of ECM causatives in English, see Miller (2002: 159, 183). ²⁷ In all likelihood, with predicates like suadeo ‘advise’ or persuadeo ‘convince’ (and many more) Latin allowed for a type of finite object control (featuring an embedded clause introduced by ut or ne, cf. example (35b)).

44

LIEVEN DANCKAERT

the embedded infinitive, whereas no such restriction holds for the subject of an AcI (cf. (30a)). She also claims that in the AcI the infinitive can be either active or passive, but that it must be active in the control structure (30b). Finally, there are also restrictions on the tense properties of the infinitive in the object control pattern, to the effect that a perfective infinitive is not allowed there (30c). (30)

a. dic-o / *cog-o port-am pate-re say-PRS.1SG force-PRS.1SG door-ACC be.open-PRS.INF ‘I say that the door is open / I force the door to be open.’ b. dic-o / *cog-o te lauda-ri say-PRS.1SG force-PRS.1SG you.ACC.SG praise-PASS.PRS.INF ‘I say that you are (being) praised / *I force you to be praised.’ c. dic-o / *cog-o te uen-isse say-PRS.1SG force-PRS.1SG you.ACC.SG come-PRF.INF ‘I say that you have come *I force you to have come.’

However, it is clear that Bolkestein’s judgements for the constructed examples in ((30a), (30b)) are inaccurate, given that cogo can in fact appear with inanimate Causees (such as pacem ‘peace’ in (31)), with passive infinitives (like condemnari ‘be condemned’ in (32)), and with both simultaneously (cf. tot et tanta uitia ‘so great and so many vices’ and superari ‘be overcome’ in (33)):²⁸ (31)

od-ere uent-os et imbr-es, qu-i hate-PRF.3PL winds-ACC and showers-ACC which-NOM.M.PL inter ill-os pac-em esse cog-unt between DEM-ACC.M.PL peace-ACC be.PRS.INF force-PRS.3PL ‘They hate wind and rain, which force peace upon them.’ (Plinius, Naturalis Historia 16.159)

(32)

sed indict-a caus-a ciu-em Roman-um but not.heard-ABL cause-ABL citizen-ACC Roman-ACC capit-is condemna-ri coeg-it head-GEN condemn-PASS.PRS.INF force-PRF.3SG ‘But without a hearing, he had a Roman citizen condemned to death.’ (Cicero, Pro Rabirio perduellionis reo 12)

(33)

nonne […] Q

di gods.NOM

ips-i immortal-es self-NOM.M.PL immortal-NOM

²⁸ On the other hand, object control infinitives do in fact always seem to display present tense morphology (arguably for semantic reasons, cf. Vincent 2016).

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

45

cog-a-nt ab his praeclar-issim-is force-PRS.SBJV-3PL by DEM.ABL.F.PL outstanding-SUPER-ABL uirtut-ibus tot et tant-a uiti-a virtues-ABL so.great and so.many-ACC vices-ACC supera-ri? defeat-PASS.PRS.INF ‘Would not the immortal gods themselves ensure that such a display of depravity were defeated by these outstanding virtues?’ (Cicero, In Catilinam 2.25) It is not immediately clear how exactly these data are to be interpreted: one possibility is that the tests proposed by Bolkestein (1979) can simply not be used to identify object control. Alternatively, it may be the case that cogo was at least for some speakers ambiguous, in that it could appear with more than one type of clausal complement (object control infinitive, AcI, and/or ECM). Finally, it is also conceivable that Latin did not have infinitival object control, and that all ‘accusative + infinitive’ patterns are either AcIs or ECM configurations (i.e. two structures in which the embedded accusative is not thematically related to the matrix predicate). Whatever the correct interpretation may be, from the above discussion it follows that the lack of animacy restriction on the Causee in (Late) Latin ‘facio + infinitive’ causatives (cf. (34)) does not directly inform us about the syntactic structure of the relevant pattern; nor does the often-noted fact that, especially in the earlier stages of the development of the infinitival causative, the embedded infinitive can either be morphologically active or passive (see Chamberlain 1986 for detailed discussion of this last point; compare again the example in (34)). (34)

uas pica-ri et gypsa-ri vessel.ACC smear.with.pitch-PASS.PRS.INF and plaster.up-PASS.PRS.INF faci-e-s make-FUT-2SG ‘You will seal the vessel with pitch and plaster.’ (Apicius, De re coquinaria 1.17.2)

A better argument against an object control analysis of ‘facio + infinitive’ structures is the fact that there was otherwise (i.e. outside putative object control configurations) no unambiguous evidence which could lead language learners to conclude that facio could ever be used as a bona fide three-place verb. In this respect, it contrasts with, for example, suadeo ‘advise’, which does in fact allow

46

LIEVEN DANCKAERT

for, on the one hand, a construal with two nominal internal arguments (‘advise something (ACC) to somebody (DAT)’, cf. (35a)), and, on the other hand, one with a nominal Recipient (marked for dative case) and a Theme argument realized as an embedded clause whose null subject is co-referential with (‘controlled by’) the dative argument in the higher clause (35b). It should go without saying that object control structures of this last type—assuming the analysis to be correct—are very different from the infinitival causatives with facio. (35)

a. si inutil-iter aliquid senatu-i if useless-ADV something.ACC.N.SG senate-DAT suas-eri-m advise-PRF.SBJV-1SG ‘if I have given useless advice to the senate’ (Cicero, Ad Brutum 2.1.2) b. suade-bi-t tibii [ut __i hinc advise-FUT-3SG you.DAT.SG that from.here disced-a-s] go.away-PRS.SBJV-2SG ‘He will advise you to go away from here.’ (Cicero, Diuinatio in Q. Caecilium 52)

In sum, given the absence of unambiguous evidence for a three-place usage of facio, perhaps compounded with the fact that inanimate Causees and passive infinitives are readily found in the pattern under discussion, it seems safe to discard the object control analysis. It then remains to be decided whether the Late Latin ‘facio + infinitive’ causatives are of the ECM or of the complex predicate type. We can define an ECM structure as a biclausal configuration in which an embedded subject is case marked but not theta marked by a matrix predicate (see the tree in (43) below, and recall the discussion in Section 2.3.2.1). In contrast, as the term suggests, a complex predicate is a ‘restructured’ monoclausal domain in which a functional (in casu causative) verb and a lexical non-finite verb together act as one single predicate. According to Ciutescu (2013b), (present-day) Catalan, Italian, and French²⁹ only have complex predicate causatives, Romanian only has the ECM pattern, and Spanish and European Portuguese have both. The syntactic differences between the two patterns are numerous (for recent discussion, see Ciutescu 2013a, 2013b and Sheehan 2016; see also the many ²⁹ But see Kayne (2004 : 197–9) for an ECM analysis of agentive datives in French faire-infinitive causatives.

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

47

references cited there). One difference (again) concerns the availability of passive infinitives (which is of course related to whether or not the lexical verb sits in a clause of its own, and thus whether it is endowed with some amount of functional superstructure): as we will see in more detail in Section 2.5.2.3, in the case of complex predicates only (morphologically) active infinitives are licit, even if the relevant verb is interpreted as a passive, but in ECM configurations, passives tend to be acceptable.³⁰ In other words, example (34) does in fact suggest that the earliest ‘facio + infinitive’ causatives are not complex predicates. A second very conspicuous difference between the two types of causatives is that it is only possible in the ECM pattern for the embedded verb to be independently negated (see e.g. Sheehan 2016: 989). For instance, in Standard Italian, where only complex predicate causatives are available, the following example is strongly unacceptable (Guasti 1993: 36, her example (38)): (36)

*Ciò ha fatto non mangia-re (più) la that have.PRS.3SG made not eat-INF anymore the pappa al bambino. soup to.the child ‘That made the child not eat his soup (anymore).’

Similar facts have been reported for Catalan and French (albeit in the latter case with some complications, cf. Labelle 1996 and Rowlett 2007). Importantly, however, in those Romance varieties where ECM causatives are available, the embedded infinitive can freely be negated, which suggests that in the relevant pattern the infinitival domain is endowed with a certain amount of functional superstructure of its own (examples from Torrego 2010: 451, her (9) and Martins 2006: 328, her (3b): (37)

a. El jefe hizo a sus clientes no divulga-r la noticia. Sp. the boss made to his clients not spread-INF the news ‘The boss made his clients not spread the news.’

³⁰ For instance, passives are fine in English ECM configurations with believe, as illustrated by the following attested example: (i)

Fugro later clarified in a statement that it believed the search area to have been well defined based on the data available. (, last accessed 30 March 2020)

In addition, as pointed out in Guasti (1993: 118–19), in Italian complements of perception verbs of the ECM type (with accusative subjects), passives with the auxiliary venire (lit. ‘come’) are completely acceptable (but with the auxiliary essere ‘be’ they are only marginal).

48

LIEVEN DANCKAERT

b. O médico mandou=o não beb-er vinho. E. Prt. the doctor sent=him.ACC not drink-INF wine lit. ‘The doctor made him not drink wine.’ Importantly, it is not difficult to find Late Latin examples with causative facio and a negated infinitive. In (38) and (39), the relevant pattern (in boldface) is accompanied by an overt Causee (underscored): (38)

Qu-is enim neg-e-t ita homin-es who-NOM.M.SG PRT deny-PRS.SBJV-3SG so people-ACC natur-a mort-em time-re [ut eam nature-ABL death-ACC fear-PRS.INF so.that DEM.ACC.F.SG uix rar-a anim-i magnitudo nonnull-os nonnull-os some-ACC.M.PL hardly rare-NOM mind-GEN magnitude.NOM faci-a-t non time-re]? make-PRS.SBJV-3SG not fear-PRS.INF ‘For who would deny that people naturally fear death to such an extent that it only rarely happens that great strength of mind makes some people abandon their fear (lit. not to fear it)?’ (Augustine, Contra Iulianum opus imperfectum 6.16)

(39)

quod plerumque fac-it non nos nos which-NOM.N.SG us.ACC often make-PRS.3SG not dormi-re sleep-PRS.INF ‘what usually causes us not to sleep’ (Tiberius Claudius Donatus, Interpretationes Virgilianae 1.6 (p. 618 l. 11 Georgii))

Two additional examples in which the Causee (a generic third person in (40), and an imaginary pious man introduced earlier in the discourse in (41)) is left unexpressed are given below: (40)

Ceruin-a pell-is substrat-a fac-it non of.a.deer-NOM skin-NOM spread.out-NOM make-PRS.3SG not time-re serpent-es. fear-PRS.INF snakes-ACC (lit.) ‘A deer skin as bed spread makes one not fear snakes.’ (Cassius Felix, De medicina 70)

(41)

ut faci-a-t non cred-ere so.that make-PRS.SBJV-3SG not believe-PRS.INF fili-us uer-e hanc nostr-am son-NOM real-ADV DEM.ACC our-ACC

quod that

de-i god-GEN

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

49

suscep-eri-t carn-em assume-PRF.SBJV-3SG flesh-ACC ‘so that he makes him not believe that the son of God really assumed this flesh which is also ours’ (Ambrosius, Expositio euangelii secundum Lucam 4.10 (Tissot)) It therefore seems safe to conclude that the earliest infinitival causatives in Latin were of the ECM type, and that complex predicate causatives are later formations. I will come back to this last point in Section 2.5.2.3. For now, suffice it to say that one advantage of the present analysis is that of the three ‘accusative + infinitive’ structures that can be generated by Grammar B, the ECM pattern is the one that structurally resembles the AcI most closely, in that (i) both structures are biclausal, and (ii) the subject DP marked for accusative case is not thematically related to the matrix verb in either configuration. 2.5.2.2 More reanalysis Turning to the actual genesis of the ECM pattern, I assume that it came about as the result of a process of reanalysis, whereby the old (‘Grammar A’) AcI structure (42a) was assigned a structurally simpler parse (42b), which is compatible with Grammar B, but which crucially requires the presence of a case assigner in the matrix domain. In other words, a subclass of pre-existing AcIs is preserved, albeit in a ‘transformed’ shape.³¹ We can schematically represent this process of structural simplification as in (42) (overt terminals (causative Vmatrix , embedded subject, embedded Vinf ) in boldface, linear word order immaterial):

(42) Reanalysis (structural simplification): from CP to TP

a.

Input structure (Grammar A)

AcI

[VP VCAUS/PERC [CP [FP [VoiceP DPACC [Voice' ... VINF ... ]] [F' F[EPP] [TP [T' T tVoiceP ]]]]]]

b.

Output structure (Grammar B)

ECM

[VP VCAUS/PERC [TP [T' T [VoiceP DPACC [Voice' ... VINF ... ]]]]] ³¹ There is no need to assume that any sort ‘grammaticalization’ of the verb facio is involved in this process (compare Butt and Lahiri 2013 on the ‘diachronic pertinacity of light verbs’).

50

LIEVEN DANCKAERT

A more detailed characterization of the newly created ECM configuration (and of the case marking mechanism therein) is given in (43):³²

(43)

VoiceP vP

Voice° [K: ACC] v°

√P TP

√°

VoiceP

T° DPEA [K: __]

Voice' Voice°

vP

It should be clear that the AcI and the ECM pattern are very different in two important ways. First, the new structure is not a full clause but rather a structurally deficient domain which is not endowed with an EPP-requirement of its own. Simplifying somewhat, let us say that it is a TP rather than a CP. Second, given that in the newly created structure (i) no node intervenes between the matrix Voice head and the embedded subject which is not part of the same EP as the case assigning Probe (by virtue of the new EPP-parameter settings), and (ii) given the absence of a CP-boundary (structural simplification)),³³ a new pattern of case assignment becomes available, namely one in which the matrix predicate (more accurately, matrix Voice) case marks the embedded subject. I will assume that as soon as the language learner has postulated an ECM parse, the embedded T head loses its ability to assign accusative case.³⁴ A crucial result of this analysis is that the old AcI could only be ‘recycled’ in cases where a matrix predicate is present which can assign accusative case to the embedded subject. As indicated in (42), environments featuring a perception or a

³² Note that in (43) we see ECM of an external argument: in passive/unaccusative contexts the same operation can also apply to an internal argument lower in the structure. ³³ The only additional point to note is that I assume that the restriction in clause (d) of (26) is lifted through head movement of the matrix lexical verb (viz. a ‘√+v’-complex) to Voice°, to the effect that there is no longer any lexical node intervening between Voice° and the embedded subject. ³⁴ Alternatively, if one adopts the scenario in which a null case assigner in C° is the source of the accusative case marking of the subject of the AcI (despite the objections raised in Section 2.3.2.2), structural simplification from CP to TP would automatically entail the loss of the case assigner, and thus the loss of the AcI as a grammatical option. Many thanks to Ian Roberts for pointing this out to me.

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

51

causative predicate in the matrix clause meet this requirement.³⁵ Let us then briefly consider some additional details of this process of reanalysis. In the case of perception verbs, the scenario is rather straightforward, as verbs of visual and auditory perception frequently appear with AcIs in Classical Latin, and as a result, we can be confident that the input structure for the proposed reanalysis was readily available in the PLD. It only needs to be added that the input and the output structure are not entirely semantically equivalent, in that the newly formed ECM configuration typically involves direct perception (see for instance Rizzi 2000), whereas there is no such requirement for Classical Latin AcIs after perception verbs.³⁶ For the case of facio-causatives, things are slightly more complex, as the input structure one would have to assume differs more strongly from the output structure. Two factors can be thought to have contributed to the creation of ‘facio + infinitive’ ECM structures. The first is the existence of infinitival complements to the verb iubeo ‘order’, which in the classical language can also be construed with a finite ut-clause.³⁷ As pointed out in Jøhndal (2012), the syntactic status of infinitival complements to iubeo is not immediately clear, as both an AcI and an object control analysis seem possible.³⁸ In cases were iubeo appears with a passive infinitive (44), an AcI analysis seems most plausible (but recall the discussion in Section 2.5.2.1): (44)

itaque [infirm-ior-es milit-es ex omn-ibus and.so weak-COMPAR-ACC soldiers-ACC from all-ABL centuri-is delig-i] iube-t […]. century-ABL select-PASS.INF.PRS order-PRS.3SG ‘And so he ordered that the weakest soldiers of all units be picked out.’ (Caesar, De bello ciuili 1.64.4)

The idea is then that the causative ‘facio + infinitive’ structure is modelled on the older ‘iubeo + infinitive’. This is in essence the line of analysis pursued ³⁵ In addition, in some early Romance varieties ECM-like constructions also appear with verbs of saying and believing (see among others Martineau 1990b: 297–320 and Pountain 1998), yielding a construction which according to many is a learned borrowing from Latin. As is well known, in presentday Romance languages an ECM configuration with believe-predicates is typically only available under wh-movement of the subject (see Postal 1974: 53 for an early statement). ³⁶ But see among others Labelle (1996) for some qualification of this direct perception reading of infinitival complements. ³⁷ For recent discussion on causative iubeo in Classical Latin, see Costantini (2012), who suggests that in some cases, the relevant predicate does not behave as a fully lexical category. ³⁸ The following often-quoted example from Plautus might well instantiate object control (with a covert direct object of iubeo): (i)

iube proi [PROi ocul-os order.IMP.2SG eyes-ACC ‘Tell them to gouge out his eyes.’

elid-ere] gouge.out-ACT.PRS.INF (Plautus, Rudens 659A)

52

LIEVEN DANCKAERT

in Norberg (1945), who proposes that the rise of the facio-causatives is to be understood as an analogical extension of the pattern with iubeo. It is in this respect tempting to hypothesize that there is a correlation between the fact that the lexical item iubeo does not survive in the Romance languages (despite being well represented in Late Latin), and the eventual success of causative facio (in its many guises). On the other hand, in Classical Latin it is in fact possible for facio to appear with an AcI, but the relevant structure clearly does not have the same semantics as the later causatives. More precisely, as noted as early as Thielmann (1886), in Classical Latin one finds examples such as (45), where the verb facio means something like ‘portray as’: (45)

at uero Polyphem-um Homer-us […] cum ariet-e etiam but PRT Polyphemus-ACC Homer-NOM with ram-ABL also conloqu-ent-em fac-it eius=que speak.with-PTCP.PRS-ACC.M.SG make-PRS.3SG DEM.GEN.M.SG=and lauda-re fortun-as […]. praise-PRS.INF fortunes-ACC ‘But Homer also portrays Polyphemus as talking with a ram and praising its fortune.’ (Cicero, Tusculanae disputationes 5.115)

Note however that in order for this type of ‘facio + infinitive’ collocation to be able to be the historical source of the later genuine causatives, there would have to be a rather strong semantic difference between the putative input and output structure. In other cases, such as (46), facio complemented by an AcI means something like ‘pretend’ (rather than ‘portray myself as’), but here too we are still quite far away from the desired causative meaning. (46)

itaque fac-io me ali-as re-s therefore make-PRS.1SG me.ACC other-ACC things-ACC ag-ere do-PRS.INF ‘So I pretend I’m doing other things.’ (Cicero, Epistulae ad familiares 15.18.1)

I therefore (tentatively) conclude that the input structure for the reanalysis of causative structures features a form of iubeo rather than facio, and that the latter verb was only later ‘inserted’ into the newly formed ECM causative.

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

53

2.5.2.3 Later developments: variation between ECM and complex predicates It should be pointed out that the scenario outlined in the previous section is apparently at odds with a number of well-documented developments in Old Romance varieties, which suggest that complex predicate causatives are older than ECM causatives (see among others Pearce 1990; Davies 1995a, 1995b, 1996, 2000; Martins 2006; Soares da Silva 2012). In particular, although the ECM configuration is attested in most of the earliest texts, corpus data concerning (i) case marking patterns of the Causee (dative (complex predicate) vs. accusative (ECM)), (ii) presence or absence of clitic climbing (ex situ (complex predicate) vs. in situ (ECM)), (iii) presence or absence of reflexive SE (absent (complex predicate) vs. present (ECM)), and (iv) word order (V(O)S (complex predicate) vs. SV(O) (ECM)) suggest that the incidence of the ECM pattern increases over time, which in turn makes it a likely candidate to be a Romance innovation (see especially Davies 1995a (on Old Spanish) and 1996 (on Old Portuguese)). Although I certainly do not wish to dispute the relevant observations, I do think that they should not be interpreted to mean that the ECM pattern was not inherited from Late Latin, and that it did not come into being earlier than the complex predicate type. Let me briefly outline why I think this is so. As mentioned, there are good reasons to assume that the Late Latin ‘facio + infinitives’, which become fully productive from about the 3rd century onwards, are neither AcIs (unlikely chronology) nor complex predicates (availability of embedded negation and passive infinitives). On the other hand, there is also evidence that in the very late stages of the Latin language, the complexpredicate-type causative had in fact emerged (plausibly as a result of another round of structural reanalysis). The crucial pieces of evidence that support this conclusion are (i) the earliest attestations of the agentive dative subject and (ii) the spread of morphologically active infinitives with passive meaning, both of which we can take to be defining properties of the complex predicate construction. As to the agentive dative Causee, the oldest attested token seems to date from the 6th century AD (Norberg 1945: 88): (47)

ut faci-a-m ei, qu-i so.that make-PRS.SBJV-1SG DEM.DAT.M.SG who-NOM.M.SG surrex-it, inueni-re merced-em stand.up-PRF.3SG find-PRS.INF reward-ACC ‘so that I will make him who has stood up find his reward’ (Vitae Patrum 5.10.28)

54

LIEVEN DANCKAERT

As documented in Chamberlain (1986), at around the same time we see that infinitives complementing facio that are interpretively passive are more and more frequently morphogically realized as active. This phenomenon, illustrated in (48), is very common in texts from the 7th and 8th centuries. (48)

omn-es Goth-os ad christian-am leg-em all-ACC Goths-ACC to Christian-ACC law-ACC baptiza-re fec-it baptize-ACT.PRS.INF make-PRF.3SG ‘He had all Goths baptized in accordance with the Christian customs.’ (Fredegarius, Chronicarum libri 4.8)

Importantly, this last development is not to be ascribed to the disappearance of the synthetic passive (which only happens later, see Danckaert 2017: 289– 91 and references cited there). Nor can it plausibly be due to phonological levelling of the endings of (certain) active and passive infinitives which in the classical language were differently realized as -re˘ and -rī respectively. As is well known, the phonological distinction between these two endings had become neutralized in Late Latin, rendering infinitives such as baptizare in (48) ambiguous. This phonological ambiguity was argued in Muller (1912) to be the driving force behind syntactic changes affecting ‘facio + infinitive’ constructions (see also Davies 1995a, 1995b, 1996), but this idea was strongly (and rightly) criticized in Norberg (1945). For one thing, the relevant phonological development does not lead to any morphological overlap in the case of infinitives of verbs of the third conjugation, but here too the active form (with passive meaning) appears in causative constructions: (49)

du-os eiusdem german-os capit-e truncat-o two-ACC same.GEN.M.SG brothers-ACC head-ABL cut.off-ABL in pute-um fec-isti proiec-ere in well-ACC make-PRF.2PL throw-ACT.PRS.INF ‘You had two of his brothers beheaded and thrown in a well.’ (Fredegarius, Chronicarum libri 3.19)

All this strongly suggests that causative complex predicates only came about in the very last stages of Late Latin, and thus much later than the earlier ‘facio + infinitive’, which (by elimination) we have identified earlier as ECM configurations. An additional advantage of this scenario is that it involves a ‘logical’ development with two successive steps of structural reduction, going from a biclausal structure with an infinitival CP to a biclausal structure with an infinitival TP, and finally to a monoclausal structure with a complex predicate. In

CHANGING PATTERNS OF CL AUSAL COMPLEMENTATION IN L ATIN

55

contrast, if one were to assume that it was the complex predicates that were created first, one would either have to say that the relevant structures are entirely unrelated to the Latin AcI (which would leave the question as to the origin of the facio-causatives again entirely open), or it would be necessary to assume a direct transition from the AcI to a monoclausal structure. The radical character of this last process does not make this an appealing option. To return to the later development of the ECM and complex predicate structures in Romance, I take it to be uncontroversial that both patterns still exist in many varieties (see again Ciutescu 2013a, 2013b and Sheehan 2016). Note, however, that I do not want to say that all of the ECM structures in present-day Romance are continuations of the pattern that came about as a by-product of the shift from Grammar A to Grammar B; rather, some ECM patterns may very well be (much) later innovations. For instance, according to Martins (2006) the Portuguese ECM configuration with a perception or causative verb and an inflected infinitive was only innovated in the 15th century. I have no objection to this analysis, which strikes me as fully compatible with my own account of the (Late) Latin data.

2.6 Conclusion In this chapter, I have tried to derive a number of phenomena concerning the availability and distribution of infinitival complements in Classical and Late Latin from a set of independently motivated assumptions on clause structure in the two periods. Crucial reference was made to a major parametric change concerning the way in which the EPP-requirement is satisfied. The proposed analysis neatly explains why the Classical Latin AcI was lost, and how an ECM configuration could emerge. At a more general level, this chapter shows how at least some ‘constructional’ changes can be made to follow from more abstract, non-construction-specific parametric changes.

3 From split to remerged Fin in Romanian supine complements Adina Dragomirescu and Virginia Hill

3.1 Introduction Traditional Latin grammars (e.g. Allen and Greenough [1888] 2001: 72–85, Palmer 1977: 324–5) define the supine as an invariable verbal noun based on a separate stem (i.e. different from perfectum and infectum stems) and with limited distribution (only in purpose and ‘in regard to’ clauses/phrases); for example, Lat. mirabile dictu ‘amazing to say’. It is unclear whether the Romanian supine root is or is not inherited from Latin,¹ but the fact is that Old Romanian (i.e. 16th–18th centuries) displays supines as nouns in the 16th-century texts (e.g. sculatul ‘rise.SUP.DEF’ / ‘the rising’), and then concurrently as verbs from the 17th century on (Dragomirescu 2013) (e.g. dobaˆndit ‘obtain.SUP’ / *dobaˆnditul ‘obtain.SUP.DEF’ / ‘the obtaining’). The concurrent nominal and verbal use of the supine is preserved in Modern Romanian, where the categorial distinction becomes visible in the treatment of DP complements: a verbal supine assigns structural (accusative) case to its complement, as in (1a), whereas the same complement must either come with morphological (genitive) case or with a de-linker when the supine is nominal, as in (1b).²

¹ Romanian supines are homophonous to masculine singular past participles, so although some etymons are Latin (e.g. Lat. dictum > Rom. zis ‘said’), they are common to both past participle and supine forms in Romanian. Thus, it is unclear whether they arise from the Latin perfectum or the supine roots, and therefore, whether the Romanian supine was inherited (Grandgent 1908: 49; Bourciez 1946: 250; Joseph 1983: 170, 172; Bauer 2000: 250) or reinvented (Densusianu 1961: 152; Caragiu-Marioțeanu 1962; Braˆncuș 1967 [2007]). ² See the cross-linguistic generalization in Kayne (1994) regarding the inability of nouns to assign accusative case.

Adina Dragomirescu and Virginia Hill, From split to remerged Fin in Romanian supine complements. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Adina Dragomirescu and Virginia Hill (2022). DOI: 10.1093/oso/9780198871538.003.0003

ROMANIAN SUPINE COMPLEMENTS

57

(1) a. I-au chemat la cules porumbul. them-have.3PL called at gathered.SUP corn.DEF ‘They called them to gather the corn.’ b. Trecem la cules de porumb.// pass.1PL at gathered.SUP of corn culesul porumbului. gathered.DEF.SUP corn.DEF.GEN ‘We transition to the gathering of the corn.’ There are several studies focussing on the structure of supine DPs (Soare 2002; Dragomirescu 2015), the emergence of the verbal supine (Hill 2013a; Dragomirescu 2013) and its distribution in the clause (Hill and Alboiu 2016). One of the main observations in these studies is that supine clauses spell out the CP differently according to whether they are selected or unselected: there is no complementizer in non-selected supine clauses, but only selecting prepositions that occur in complementary distribution. On the other hand, the complementizer de is obligatory in the CP of selected supines, as shown in (2c).³ The same studies bring evidence that de in supine complements is not a preposition but a C element that is functionally equivalent to infinitive a and subjunctive să (see also Giurgea and Soare 2010), especially because these clausal complements appear in free alternation after the same selecting verb, as in (2). • a-infinitive (2) a. au început a învăța învăța cărți a scrie has began INF learn letters INF write ‘he began to learn to write letters’

(AB: 278)

• să-subjunctive b. de oaste au învățatu grijască învățatu să să of army had learned SUBJ REFL take.care ‘he learned to take care of the army’

(Ureche: 73r )

• de-supine c. că le-au învățat învățat toate de covaˆrs¸it for them=has learned all DE improved.SUP ‘for she learned to improve everything’

(Filerot: 213)

³ In the class of clausal complements we include the non-finite relatives, which also display obligatory de (Soare 2002; Giurgea and Soare 2010), under the assumption that relatives are selected D/CPs, as in Donati and Cecchetto’s (2011) analysis, which was validated for Old Romanian in Sevcenco (2015).

58

ADINA DRAGOMIRESCU AND VIRGINIA HILL

Nevertheless, despite the free alternation in (2), the internal organization of the supine clause is different from the other two alternates insofar as it disallows clitics, it requires a different negation (i.e. affixal ne- instead of nu ‘not’), and it cannot license lexical subjects.⁴ Thus, at first sight, there seems to be an asymmetry within the paradigm of non-finite clausal complementation in the language, as the word order in supine clauses suggests a different derivational pattern when compared to their infinitive and subjunctive counterparts. This contrast applies to both Old and Standard Modern Romanian. In this chapter, we point out that the above mentioned contrast is superficial, as the same derivational pattern underlies all three types of non-finite complements. Two empirical facts are relevant in this respect and will be the focus of our investigation: (i) In Old Romanian, de was not specialized for supines, as it could also appear in infinitives and subjunctives, as shown in (3). (3) a. iară vicleanul diavol nu mai părăsiia and malicious.DEF devil not more stopped.3 de-a o supărarea DE=INF her=upset.INF ‘and the malicious devil did not stop upsetting her’ (Varlaam: 13v ) b. Ce veți de să facu voao? what want.2PL DE SUBJ make.1SG you.DAT ‘What do you want me to do for you?’ (Coresi, CC2 : 91) Modern Romanian lost the use of the complementizer de in subjunctives and restricted it in infinitives (i.e. only in complements to nouns), while it remained obligatory with supines. This contrast in the use and productivity of de needs explaining. (ii) In the Northern varieties of Modern Romanian, there are no grounds to assume structural asymmetry, as de-supine complements display the same internal options as the infinitives and subjunctives; that is, clitics may precede the verb, and the negation is nu ‘not’, as shown in (4). (4)

Așa că trebuie de le le lăsat so that must.3 DE them=leave.SUP

poarta door.DEF

deschisă spre open.F for

⁴ Some lexical subjects have been signalled, but in fragments, not in clauses (Pană Dindelegan 2011; Dragomirescu 2011).

ROMANIAN SUPINE COMPLEMENTS

59

plecare și de nu-i mai netezit pe bășcălie nu-i leaving and DE not=them =more= protect.SUP on.mockingly ‘One must leave the gate open for them to leave and not to mockingly protect them anymore.’ (m.publica.md) The data presented so far suggest that, in Old Romanian, supine de was the same de that occurs in (3), but the absence of another complementizer (i.e. equivalent to a and să) somehow triggered a structural asymmetry between supine versus infinitive and subjunctive complements. Accordingly, the cancellation of this asymmetry in (4) must arise either from the change in the status of de (which takes place in other contexts as well) or/and from the change in the mapping of functional features to C. Notably, Modern Romanian also displays only one complementizer in subjunctive and infinitive complements (i.e. să or a, respectively) in contexts with obligatory control similar to (4). Therefore, something has changed in these constructions as well, so that double complementizers as in (3) are discontinued. The analysis proposed in this chapter follows in the path of Hill (2013b), where the co-occurring complementizers in (3) reflect a separate mapping of the features of Fin: considering that in cartography (Rizzi 1997, 2004) Fin is the C head associated with finiteness and sentence modality, in addition to phifeatures, de checks [−finite], whereas a/să checks [modal]. The phi-features are transferred to T, so their checking does not concern Fin in subjunctive and infinitive clauses. Accordingly, Fin has been split, and the disappearance of de from the contexts in (3) in Modern Romanian follows from the reanalysis of a and să as syncretic markers for both [−finite] and [modal] (i.e. Fin remerged, as pointed out in Hill and Alboiu 2016). Along the same lines, the working hypothesis in this chapter is that the supine CP complement also projects a split Fin in (2c), where de checks [−finite] only. The first question (addressed in Section 3.2) concerns the operation that ensures checking/valuation for the [modal] feature of Fin, since there is no second complementizer to match the de a or de să doubles in infinitives and subjunctives, respectively. The second question (addressed in Section 3.3) concerns the change we see in example (4): Hill (2013b) argues that a and să took over the function of de in the respective CPs and became the only spellout for Fin. In other words, Fin remerged and its features are checked by one element. If we assume that the change in (4) is due to a change in the internal structure of the supine CP, so that it matches the internal structure of its infinitive and subjunctive counterparts, then we must also assume that Fin in (4) is remerged. The question is how, since there was no modal complementizer that

60

ADINA DRAGOMIRESCU AND VIRGINIA HILL

could take over the function of de, and in fact, de is still present. In addition, the checking of phi-features in a supine clause is an issue, since the supine form does not display agreement morphology. The answer we provide is that Old Romanian and Standard Modern Romanian supine complements preserve a split Fin, where de checks [−finite], whereas [modal] and phi-features are checked by V-to-C. On the other hand, the modern Northern varieties display the change in (4), which arises from the functional reanalysis of de, through feature enrichment, and through exemplar-based analogy (as in Kiparsky 2011) with infinitive a and subjunctive să. We suggest (in Section 3.4) that this happened under paradigmatic pressure, having the effect of eliminating the structural asymmetry between supine CPs and the other non-finite CPs that occur in the same selected environment. The theoretical framework for this chapter consists of the cartographic representation of clauses (Rizzi 1997, 2004), in which feature checking/validation applies according to minimalist assumptions (Chomsky 2008). In cartography, CP is articulated over several projections, as in (5), which allows us to define the phasal property of CPs in terms of the CP field size. In (5), Force is the phasal head and types the clause, whereas Fin types the verbal inflection, through finiteness, modality, and phi-features. (5)

ForceP > TopP > FocusP > FinP > (inflectional field)

The inflectional field in (5) has the hierarchy in (6) in Old and Modern Romanian, which is a template that comes out of several studies (Cornilescu 2000; Alboiu 2002 a.o.). The actual implementation of (6) depends on the values of the features in Fin. (6)

NegP > TP > AspP > vP

Crucially, in (5), a phasal CP projects up to Force, which is the default site for merging că ‘that’ in Romanian (Hill and Alboiu 2016), as it is also the case in most Romance languages (Rizzi 1997). On the other hand, a non-phasal CP, which allows for matrix features to probe inside the embedded clause (Chomsky 2008), must lack the ForceP level, which would limit the probing from the matrix to the edge position only (Spec, ForceP).⁵

⁵ The presence of TopP and FocusP is irrelevant for this issue as these projections are orthogonal for the computation of phi and finite/modal features.

ROMANIAN SUPINE COMPLEMENTS

61

3.2 De as [−finite] Verbal supines emerged in adverbial adjuncts, with subsequent spread to relatives and then to clausal complements (Dragomirescu 2013; Hill 2013a).⁶ This section shows that adverbial supine clauses cannot display de, whereas de is obligatory in the selected supine clauses. This contrast is related to the function of de as a [−finite] marker in C/Fin. The account is that the respective Fin is split, as de checks [−finite] but is unable to do the same for [modal]. This is not the case for adverbial supine clauses, where the CP is collapsed (i.e. no articulation over Fin and Force). We start with a test showing that adverbial supine clauses are peripheral (hence, anchored to the matrix but propositionally independent) versus central (part of the matrix event), in the sense of Haegeman (2010a). In (7), the preposition precedes a verbal (versus nominal) supine because the direct object has unmarked (accusative) case (objects to N come in genitive, see (1b)). The adjunct clause structure consists of a full-fledged CP, because the supine verb can be modified by an adverb that indicates a separate event in relation to the matrix clause, as in (7b). (7) a. caˆnd pentru făcut [folosul de obs¸te] when for do.SUP deed.DEF of community cineva să ferices¸te someone REFL contents ‘when someone is content about having done a good deed for the community’ (Cantemir I: 106) b. A fost modificată pentru ieri ieri has been modified.F yesterday for calculat multe probabilități acum acum mai calculate.SUP now more much probabilities ‘It was modified yesterday in order to calculate more probabilities now.’ The adverbs in Modern Romanian (7b) indicate events with different time frames in the matrix and the supine clauses; a time frame contrast is also implied in Old Romanian (7a). Therefore, the adverbial supine clauses are full-fledged CPs. ⁶ In the texts of the late 16th to the early 17th century most adjunct phrases with supines are ambiguous as to the nominal or verbal status of the supine: there is a preposition followed by the supine item, but around the supine, there is no determiner, modifier, complementizer, or direct object that would indicate its categorial status. We base our analysis on examples where such indications are available, as in (1).

62

ADINA DRAGOMIRESCU AND VIRGINIA HILL

In the same type of clauses, de is disallowed—see (8a), whereas indicative că ‘that’ can co-occur with the same preposition—see (8b). (8) a. A fost modificată ieri pentru (*de) ieri has been modified.F yesterday for DE calculat acum acum mai multe probabilități calculate.SUP now more much probabilities Intended: ‘It was modified yesterday in order to calculate more probabilities now.’ b. A fost modificată ieri ieri, pentru că ieri has been modified.F yesterday for that mai multe probabilităt¸i acum acum poate calcula now can calculate more much probabilities ‘It was modified yesterday since now it can calculate more probabilities.’ We attribute the contrast between de and că ‘that’ in this context to their properties for finiteness marking: că ‘that’ is associated by default with independent CP domains and finite verbs (it is in Force in cartographic representations), whereas de signals an anaphoric T, and checks and values the finiteness of C as [−finite] in selected clauses with non-finite verbs, in general (Hill 2013b). Therefore, the preposition pentru ‘for’ selects a full-fledged CP in both (7) and (8b); the difference concerns the spell-out of C, which is că ‘that’ in (8b) but the verb in the supine CP in (7).⁷ We know that V-to-C takes place in (7) because of the obligatory adjacency between the preposition and the verb. That is, the temporal adverb is obligatorily postverbal for the supine in (8a), which signals high verb movement, by contrast to (8b), where the same adverb is preverbal. Furthermore, constituent fronting to topic or focus fails with the supine but not with the că-CP, which signals a collapsed C with the supine, whereas the CP is articulated with the indicative. The contrast is shown in (9). (9) a. A fost modificată ieri, pentru că has been modified.F yesterday for that acum altfel de calcule propune Mihai. now other of calculations proposes Mihai. ‘It was modified yesterday because now Mihai proposes other kinds of calculations.’ ⁷ The complementizer de is specialized for non-finite CPs (be they infinitive, subjunctive, or even indicative) that display anaphoric tense, i.e. the embedded tense is restricted by the matrix tense (obligatory tense concordance).

ROMANIAN SUPINE COMPLEMENTS

63

b. A fost modificată ieri ieri, pentru (*acum) ieri has been modified.F yesterday for now (*altfel de posibilități) calculat acum altfel de posibilități. other of possibilities calculate.SUP now other of possibilities ‘It was modified yesterday in order to calculate other probabilities now.’ Thus, the data in (7) to (9) support the following analysis: Supine clauses always involve V-to-C, whereby the verb checks the entire feature cluster of C when CP is not split. When CP is split, V in C shares the feature checking task with de. The exact matching between lexical item and C-features is detailed in the remainder of this section. The crucial observation at this time is that there is an asymmetry between adjunct and complement supine clauses, on the one hand, and between adjunct indicative and adjunct supine clauses, on the other hand. Two contrasts are relevant in this asymmetry: the split versus unsplit projection of C, and the way the features of C are checked. More precisely, C is split with indicative adjuncts and a complementizer checks the features of Force, whereas the features of Fin are checked by the verb through long distance Agree. On the other hand, C is collapsed for supine adjuncts and its features are checked through V-to-C. The complementizer de does not qualify for adjunct supine clauses since it is specialized for Fin (which is not projected separately) and for anaphoric Fin/T, which is incompatible with peripheral adverbial clauses. Such clauses are anchored and have independent tense, irrespective of the finite or non-finite inflection of their verb. Conversely, with selected de supines, the matrix and the embedded verbs describe a single event with one time frame, as in (10), where the insertion of contrastive time adverbs would rule out the sentence. These are typical configurations with obligatory control and anaphoric T in the embedded clause. (10)

După ce deci o au isprăvit after what thus it= have.3 finished (*azi) [de zugrăvit] (*ieri/*maˆine) today DE paint.SUP yesterday/tomorrow ‘Thus, after they finished painting it …’

(Greceanu: 151)

The contrast between (7) and (10) is then captured in terms of tense properties in the embedded clause: following Chomsky (2008), independent tense pairs with a phasal CP, whereas anaphoric tense pairs with a non-phasal CP. For the data presented so far, it follows that the presence of că ‘that’ or V-to-C signal a phasal CP, whereas the presence of de signals a non-phasal CP. Next, we zoom in on the non-phasal supine CP complement and identify the way feature checking takes place for the features of Fin. Starting with [finite],

64

ADINA DRAGOMIRESCU AND VIRGINIA HILL

the association between de and anaphoricity in other non-finite complement clauses has already been formalized in the literature (Dobrovie-Sorin 1994; Hill 2013b; a.o.), and it entails that de is a marker for the [−finite] feature, so it is located in Fin. According to the hierarchy in (5), in a split CP field Fin allows for Topic and Focus constituent to precede it, and that is indeed the case with de; that is, Topic and Focus constituents (bracketed) can precede it in infinitive and subjunctive clauses, as in (11a), (11b). (11)

a. Ce să fimu cu usaˆrdie, frațiloru but SUBJ be.1PL with fervour brothers.VOC ¸si [cu cinste dragă] de-a primirea and with honour dear DE=to accept ‘But we should be eager, brothers, and we should accept it with great honour.’ (Coresi, CC2 : 472) b. aceasta nu-ți ajunge ¸tie, [a this not=you.DAT suffices you.DAT GEN legiei învățătură] de să isprăves¸ti law learning DE SUBJ finish ‘it does not suffice for you to finish learning the law’ (Coresi, CC2 : 324)

An important observation is that de shares the Fin head with the second complementizer, since both items precede the negation. Note that in (6), Neg marks the border between the inflectional and the CP fields. The data in (12) confirms that word order: the double complementizers in (3), namely de a or de să, are higher than NegP spelled out as nu ‘not’ (a free versus clitic morpheme; Isac and Jakab 2004), therefore both items are in Fin. (12)

a. că nu-l opri pri însu avuțiia that not=him stop DOM him fortune.DEF de-a rugarea lui Dumnedzău nu nu să DE=INF not REFL pray.INF GEN God ‘his fortune did not stop him from praying to God’ (Varlaam: 277r ) b. ¸si nici unui om pizmit-amu and not a man hate=have.1 de să învățătura nu nu spuiu DE SUBJ not tell.1SG teaching.DEF ‘and I have never hated a person so much that I would not tell him/her the teachings (of God)’ (Coresi, CC2 : 203)

ROMANIAN SUPINE COMPLEMENTS

65

The argument is then that Fin is split, because de is semantically bleached and unable to check the [modal] feature, which is separately checked by a or să, as in (13).⁸ (13)

Fin1[-finite] de > Fin2[modal] a/să > NegP

Accordingly, we expect the supine de to yield the same results in tests. Indeed, the low merge location of de in Fin can be verified in (14): de is preceded by a fronted constituent in (14a), and it is higher than the negation in (14b).⁹ (14)

a. caˆnd vei face nodurile la fies¸tecari nod when will.2SG= make knots.DEF at each knot vei înnoda să zici ¸si isprăvind will.2SG= tie SUBJ say.2SG and finishing Evanghelii de cetit,] [[12 Evanghelii] vor fi ¸si pre ață 12 Evanghelii 12 gospels DE read.SUP will.3PL be also on thread 12 noduri 12 knots ‘when you will make the knots, for each knot you are tying you must pray and when you finish reading 12 gospels there will also be 12 knots on your thread’ (Timotin: 311) b. ¸si de lucrurile de ne neputut să ne and of things.DEF DE not.can.SUP REFL ‘and he starts making impossible things’

apucă starts (Cantemir I: 40)

There are, however, two points of dissimilarity between de supines and the infinitive and subjunctive alternates: first, there is no other complementizer to check the [modal] feature in the supine clause; and second, the negation is different: affixal ne-, as in (14b), instead of the free morpheme nu, although the reading is equivalent (i.e. propositional scope for the negation Cornilescu and Cosma 2011). We relate both contrasts to the application of V-to-Fin2 in supines, whereas only V-to-T applies in infinitives and subjunctives (see arguments in Hill and Alboiu 2016). That is, in the absence of a dedicated complementizer for ⁸ Evidence for the inability of de to check [modal] is that there are no de-infinitives or de-subjunctives in the language; the presence of a or să is obligatory to supplement de. ⁹ (14b) is a non-finite relative, not a complement, but qualifies as selected (Giurgea and Soare 2010; Sevcenco 2015). Negation is not possible in supine clausal complements because they are truncated (ForceP is missing), and in such configurations the negation is always in the matrix for scope requirements (Progovac 1994). The relative is a ForceP, and thus a separated domain that ensures propositional scope for the negation. In (14b), ForceP contains the wh-operator, while de occupies Fin1 (Hill and Alboiu 2016).

66

ADINA DRAGOMIRESCU AND VIRGINIA HILL

Fin2[modal] , this feature is checked through verb movement (i.e. V-to-C translates to V-to-Fin2 in this configuration). It is well known that nu ‘not’ blocks V-movement in general (Rivero 1994; Isac and Jakab 2004; a.o.), so only an affixal negation would allow the derivation to converge, because it moves with the verb head. Hence, the merging of de in supine CPs replicates (13) with the variation in (15).

(15) Fin1[–finite] de > Fin2[modal]V > TP The difference between (13) and (15) is that the split Fin in (13) is checked by two dedicated complementizers, and does not probe the verb (hence, only V-to-T applies in infinitive and subjunctive complements), whereas the split Fin in (15) has a dedicated complementizer for only one feature, and thus the other feature probes the verb (hence, V-to-Fin2). Since de-supine complement clauses project a truncated CP (i.e. ForceP is missing), they are compatible with obligatory control, as in (2c) and (7). More precisely, obligatory control configurations require truncated CPs in all the Balkan languages (Hill and Alboiu 2016), and in Old Romanian this requirement applies irrespective of the grammatical mood of the embedded verb (i.e. it equally applies to infinitive, subjunctive, or supine complements). Thus, we can see that supine complements are integrated in the general clausal complementation pattern in the language. This generalization is illustrated in our data by the match between the configurations in (13) and (15). However, the results of these derivations show some differences that arise from different implementations of the [modal] checking in Fin2. That is, feature checking of Fin1 and Fin2 through two dedicated complementizers (as with infinitives and subjunctives), as opposed to feature checking through one complementizer in combination with V-to-Fin2 (with supines) has derivational consequences. First, the contrast in negation type, which is nu ‘not’ for infinitives and subjunctives, but affixal ne- ‘not’ for supine, responds to the locality constraint in head-to-head movement. That is, V-to-Fin2 cannot take place if the free morpheme nu is in Neg. However, the affixal ne- can move together with the verb. Second, the checking of phi-features is implemented differently. In particular, V-to-T in infinitive and subjunctive complements allows for clitic pronouns, auxiliaries, and the free morpheme negation, whereas V-to-Fin2 in the supine complements blocks the same elements. This peculiarity is directly related to the endowment of Fin with the phi-features deemed to be

ROMANIAN SUPINE COMPLEMENTS

67

generated at C in the minimalist theory (Chomsky 2008). The feature transfer/inheritance assumption entails cross- and intra-linguistic variation with respect to the phi-feature transfer to T. This transfer takes place by default in Romanian clauses (be they finite or non-finite), and it correlates with obligatory V-to-T, where T with phi-features allows for subject–verb agreement and object/subject clitic pronouns (which stand for agreement markers; Delfitto 2002; Pancheva 2005). However, when V-to-Fin2 takes place in addition to complementizer merge, all the features of Fin, including phi, are checked locally in the C field, so no phi-feature transfer takes place. Hence, supine T is not compatible with agreement and disallows clitics.¹⁰ To conclude this section, the structure that underlies clausal complements is the same for infinitive, subjunctive, and supine clauses, involving a split Fin in (13) and (15). Variation arises in the way feature checking is performed, which entails further variation in the type of syntactic operations that may occur internally in the clausal complement (i.e. the type of compatible negation and the possibility of merging clitics). Crucially, infinitive and subjunctive Fin could be split or remerged in Old Romanian, and is systematically remerged for subjunctives in Modern Romanian, whereas supine clauses maintain a split Fin at all times, in both Old and Standard Modern Romanian.

3.3 Reanalysis of de In Old Romanian, split Fin in infinitive and subjunctive complements, as in (3), occurs in free alternation with unsplit Fin (i.e. de is absent) in the same text, and frequency of occurrence (Fraˆncu 2009) indicates that the latter is the preferred option. Unsplit Fin signals that a or să are able to check the features of Fin syncretically, and this is the option preserved in Modern Romanian. Hill and Alboiu (2016) argue that de was instrumental in the transition of these structures from adjunct to selected contexts, but once they became established as non-finite complements, a and să were further grammaticalized through the bleaching of their modal value and their merging at the next hierarchical level in the clause structure (i.e. they were reanalysed from Fin2 to Fin1, upwards in the clause hierarchy; Roberts and Roussou 2003). In terms of Kiparsky (2011), ¹⁰ V-to-C instead of a dedicated complementizer is also seen in imperative and gerund clauses, which, however, display clitic pronouns, i.e. the phi-feature transfer to T is not blocked. Crucially, V-to-C in those configurations does not concern the basic features of Fin, but is triggered by extrafeatures involved in clause typing (e.g. [jussive] with imperatives, as in Zanuttini 2008; or [assertion] with gerunds, as in Hill and Alboiu 2016).

68

ADINA DRAGOMIRESCU AND VIRGINIA HILL

de and a/să qualified as competitors at a certain point, and the former was eliminated when the latter underwent grammaticalization through analogy. Along the same lines, supine de had no competitors in a structural configuration similar to the one that underlies infinitive and subjunctive complements. While V-to-Fin2 successfully replaced a second complementizer, the economy principle is expected to intervene favouring a remerged versus split Fin. Indeed, the remerging process is on the rise in the Northern varieties of Modern Romanian, as illustrated in (4), and it involves the reanalysis of de in the opposite direction to a/să. That is, whereas a/să were bleached of their strong modal feature, supine de is enriched with the same feature. The result is a paradigmatic identity between supine de, infinitive a, and subjunctive să. The acquired modality of supine de is confirmed by its distribution: constructions as in (4) occur almost exclusively after modals, as further shown in (16). (16)

avem de ne plătit impozitele have DE us=pay.SUP taxes.DEF ‘we have to pay our taxes’

(www.bistriteanul.ro)

Importantly, V-to-Fin2 is dispensed with in these constructions, being replaced by V-to-T. Predictably, the negation changes from affixal to the free morpheme nu ‘not’, and the phi-feature transfer from C/Fin to T takes place, thus allowing for object clitic pronouns, as seen in (16). Hence, the underlying structure of the supine clause remains the one in (15), but the feature checking implementation switches to the options in (13). The diachronic change entailed by the remerging of Fin in supine complement clauses is typical of grammatical optimization, in Kiparsky’s terms, whereby the reanalysis involves an exemplar-based proportional analogy. The exemplar base line, in this case, consists of the similar syntactic pattern that underlies the infinitive and subjunctive complements, and which was still relatively unstable when supine complements emerged. Once the infinitive and subjunctive complementation pattern became stabilized (through the systematic remerging of Fin), learners could use that pattern as a clue for analysing the third type of non-finite complement clause, which contains supine verbs.

3.4 Conclusions This chapter proposed an analysis of supine clauses and the changes they display from Old to Modern Romanian, by taking into account the wider context of clausal complementation in the language. In particular, it was argued

ROMANIAN SUPINE COMPLEMENTS

69

that supine CP in complement clauses had the same structure as their infinitive and subjunctive counterpart insofar as Fin was split and its features were checked by different items. Split Fin has been shown to remerge in Standard Modern Romanian infinitive and subjunctives, but not in supine complements. However, the remerging process is seen in the regional varieties of the North, which are considered less conservative with respect to certain features (Ghet¸ie 1975: 197). The paradigmatic view we adopted in this chapter allows us to conclude that the remerging of Fin in supine complements is on a rising path in the language, and that the internal structure of the supines with remerged Fin will allow this construction to spread beyond the contexts with obligatory control, to which it is restricted at this time. That is, considering that a remerged Fin allowed infinitive and subjunctive clauses to appear in contexts with non-obligatory control and subject clauses, the same is predicted for supine clauses, whose activated TP provides the possibility of phasal projections with temporal independence and lexical subjects.

Sources AB Gregorian, M. (ed.). 1961. Cronica Anonimă despre Braˆncoveanu, in Cronicari munteni, 273–352. Bucharest: Editura pentru Literatură. Cantemir I Panaitescu, P.P. and I. Verdeș (eds). 1965. Dimitrie Cantemir, Istoria Ieroglifică, I. Bucharest: Editura pentru Literatură. Coresi, CC2 Pus¸cariu, S. and Al. Procopovici (eds). 1914. Diaconul Coresi, Carte cu învăt¸ătură (1581), vol. I, Textul. Bucharest: Socec. Filerot Tarantino, A. (ed.). 1996. La storia di Filerot e Anthusa. Istoria lui Filerot și a Anthusei (ms. 1374 BAR—Bucarest). Roma: Bagatto Libri. Greceanu Gregorian, M. (ed.). 1961. Radu Greceanu, Începătura istoriii, in Cronicari munteni II, 5–272. Bucharest: Editura pentru Literatură. Timotin, E. 2010. Descaˆntecele manuscrise romaˆnes¸ti (secolele al XVII-lea–al XIX-lea). Bucharest: Editura Academiei Romaˆne. Ureche Panaitescu, P.P. (ed.). 1955. Grigore Ureche, Letopisețul Țărîi Moldovei. Bucharest: Editura de Stat pentru Literatură și Artă. Varlaam Byck, J. (ed.). 1964. Varlaam, Cazania. Bucharest: Editura Academiei.

4 The diachronic path of senão From conditional subordination to exceptive coordination Ana Maria Martins, Sandra Pereira, and Clara Pinto

4.1 Introduction This chapter deals with exceptive constructions focussing on the genesis of the Portuguese exceptive word senão (if+NEG, i.e. ‘except’), which originated in earlier conditional structures associated with negation.¹ The relevant conditionals are exemplified in (1) and (2) with sentences taken from 13th-century Portuguese texts.² These exceptive conditionals ceased to be attested as early as the beginning of the 14th century. From then on, senão constitutes an indivisible lexical unit. Hence, in contemporary Portuguese only (3a) is a grammatical option, while (3b), which parallels (1), is excluded. (1) Nehu˜u˜ outro nõ no ousaria dizer se hu˜u˜ destes nõ. no.one other not it would.dare say.INF if one of.these NEG ‘Nobody would dare saying it except one of them.’ (13th century, DSG635,.25, WOChWEL) (2)

na terra nõ achou el se duros corações nõ. in.the earth not found he if hard hearts NEG ‘He couldn’t find anything in the earth except people hard at heart.’ (13th century, DSG60,.3, WOChWEL)

¹ Research for this work was funded by FCT—Fundação para a Ciência e a Tecnologia—within the projects WOChWEL (PTDC/CLE-LIN/121707/2010) and ARQUÍDIA (SFRH/BPD/99678/2014). We thank the two anonymous reviewers for their suggestions and comments, which helped us to improve the manuscript. ² Similar exceptive conditionals are attested in Old French, Old Occitan, Old Spanish, and Old Galician (see Pietsch 1915: 377–8 [73–4]; Jensen 1986: 311–12).

Ana Maria Martins, Sandra Pereira, and Clara Pinto, The diachronic path of senão. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Ana Maria Martins, Sandra Pereira, and Clara Pinto (2022). DOI: 10.1093/oso/9780198871538.003.0004

THE DIACHRONIC PATH OF SENÃO

(3)

a. Nenhum outro o ousaria dizer no.one other it would.dare say.INF b. *Nenhum outro o ousaria dizer no.one other it would.dare say.INF ‘Nobody would dare saying it except him.’

senão SENÃO se ele if he

71

ele. he não. NEG

It will be proposed that the change leading to the emergence of senão is motivated by an independent change bearing on the relation between word order and information structure. To be more specific, in 13th-century Portuguese there were two structural positions available for a constituent carrying information focus prominence, namely the clause-final position that can be found from the earlier texts to contemporary Portuguese (Ambar 1992; Costa 1998, 2004; Costa and Figueiredo Silva 2006; Cardoso 2010; Martins 2011; Kato and Martins 2016; Martins and Costa 2016; Lobo and Martins 2017; among others), and a left-peripheral position that ceased to be available from the turn of the 13th to the 14th century. In fact, the diachronic development of senão pinpoints this other instance of syntactic change in Early Old Portuguese, which had not been identified previously but is supported by comparative evidence when the different strategies available across the Romance languages to set appropriate information focus configurations are taken into consideration (cf. Cruschina 2015). The unifying factor behind exceptive constructions in the world’s languages is of a semantic nature and leaves space for different syntactic/lexical strategies. As Von Fintel (1993) puts it, ‘the semantics of exceptives is primarily one of subtraction from the domain of a quantifier’.³ This chapter will not address semantic issues (cf. Breitbarth 2015). It deals instead with aspects of the permitted syntactic variation. Let us thus exemplify how coordination and subordination are both available as structural strategies to build exceptive constructions. Pérez-Jiménez and Moreno-Quibén (2012) analyse Spanish sentences with excepto (‘except’) as illustrated in (4). The authors frame their analysis in the Boolean Phrase (BoolP) Hypothesis (Munn 1993; among others) and treat excepto as the coordinating conjunction. In the second member of the coordination structure there is deletion under identity, after one or more constituents are extracted from the deleted phrase.⁴ ³ Cf. Hoeksema (1995) who argues in favour of a non-uniform semantics for connected and free exceptives. The distinction between the two types of exceptive structures will be clarified in Section 4.2. ⁴ Cf. Pérez-Jiménez and Moreno-Quibén (2012: 591): We claim that in F[ree] E[xceptive]s … the exceptive conjunction selects for a CP whose head is null and is endowed with a feature that triggers and licenses the ellipsis process (the

72

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

Spanish excepto (4) [CP1 Los asistentes recibieron un regalo de recuerdo [BoolP excepto the attendees got.3PL a gift of keepsake except [CP2 Juani [TP ti recibió un regalo de recuerdo]]]] Juan got.3PL a gift of keepsake ‘A ll those present received a gift as keepsake, except Juan.’ (Example taken from Pérez-Jiménez and Moreno-Quibén 2012: 590; adapted) Also contemporary Portuguese senão has been analysed as a coordinating conjunction (Matos 2003; Colaço 2005). It differs from other exceptive coordinating conjunctions (such as excepto ‘except’) in that it imposes the presence of negation in the first conjunct of the coordinate structure, as illustrated by the grammaticality contrast between (a) and (b) in (5).⁵ In that respect, the Portuguese construction with senão resembles the French ne … que exceptive construction, analysed by O’Neill (2011) as shown in (6). Portuguese senão (5) a. Não houve queixas (de ninguém) senão do João. not was complains of anybody SENÃO of.the João ‘Nobody complained except João.’ b. *Houve queixas de toda a gente senão do João was complains of all the people SENÃO of.the João ‘Everybody complained except João.’ French ne … que (6) [IP Je n’ai [VP vu [QP [QP (personne d’autre)] [CP OP que I NE have seen anyone of other QUE [le professeur]i [IP j’ai vu ti ]]]]] the professor I have seen ‘I haven’t seen anyone else than/except the professor.’ (Example taken from O’Neill 2013: 216; adapted)

E-feature). This feature, on the one hand, attracts one or more constituents internal to CP2 … to the left periphery of the elliptical sentence; these constituents are thus the remnants of the ellipsis process. On the other hand, the E-feature on C also forces the deletion/nonpronunciation of the phonological features of the syntactic complement of C, that is, TP. ⁵ While except acts as a polarity-reversal operator (Holmberg 2003) that is compatible with a positive or a negative polar value in the antecedent of the exception phrase, senão requires negation in the antecedent. It thus behaves as a strong NPI.

THE DIACHRONIC PATH OF SENÃO

73

Under O’Neill’s (2013) account, French ne … que displays a reduced clausal comparative. The exception phrase following que is the remnant of an elliptical clause adjoined to an optionally covert negative polarity item (NPI). The exception phrase is moved out of the ellipsis site to satisfy the licensing requirements of an exhaustive identification operator. In both (4) and (6) above, the exceptive interpretation is compositional, but the coordinate structure in (4) includes an element lexically associated with the exceptive meaning (i.e. the exceptive word excepto) whereas the subordinate structure in (6) does not. However, an exceptive word does not necessarily integrate a coordinate structure; it can also be part of a subordinate structure. Pérez-Jiménez and Moreno-Quibén (2012) treat Spanish excepto que (‘except that’) as a complex subordinating conjunction. English unless-clauses display conditional subordination (Von Fintel 1991; among others). Breitbarth (2014) draws attention to the Dutch subordinating conjunction tenzij ‘unless’, which historically emerged from a former exceptive construction not headed by an exceptive word (see (7)). Differently from Portuguese senão, Dutch tenzij does not require the presence of negation in the antecedent clause, as exemplified in (8). (7) Maer dat en mach niet siin het en waer een sempel wonde. but that NE can NEG be it NEG were.SUBJ a simple wound ‘But that cannot be unless that was a simple wound.’ (Year 1350; Burridge 1993, quoted by Breitbarth 2014) (8) a. Wij zullen het hallen, tenzij de trein te laat aankomt. we will it make unless the train too late arrives ‘We will make it unless the train arrives too late.’ b. Wij zullen de trein niet halen, tenzij ere en wonder gebeurd. we will the train not catch unless there a miracle happens ‘We will not catch the train unless a miracle happens.’ (Breitbarth 2014) The change studied in this chapter caused a coordinating conjunction to emerge from a former conditional structure, where the exceptive meaning was strictly compositional, not lexically based. The chapter is organized in five sections. In Section 4.2, the distinction between free exceptives and connected exceptives is clarified and the Old Portuguese conditional exceptives are shown to display the characteristic traits of free exceptives. In Section 4.3, the syntax of 13th-century Portuguese exceptives with se … não (‘if … NEG’) is described in detail. Section 4.4 accounts for

74

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

the emergence of the exceptive word senão ‘except’, which is concomitant to the loss of the structures displaying discontinuity between se ‘if ’ and negation. This change will be correlated with the loss of the availability of a left-peripheral functional position for constituents expressing information focus. As early as the early 14th century, Portuguese came to limit left-peripheral focalization to instances of contrastive/emphatic focus (Rizzi 1997; Belletti 2004, 2008; Costa and Martins 2011; Lobo and Martins 2020). Section 4.5 concludes the chapter. Most of the diachronic data presented throughout the chapter come from the 13th-century Portuguese translation of the Arthurian Post-Vulgate Cycle, originally written in French in the same century. There are also examples taken from the profane Galician-Portuguese poetry, with the same chronology.⁶

4.2 Old Portuguese exceptive conditionals are free exceptives, not connected exceptives The distinction between free exceptives and connected exceptives is well established in the literature (see Von Fintel 1993; Hoeksema 1995; Pérez-Jiménez and Moreno-Quibén 2012). It is important to consider it here because only free exceptives feature biclausal structures, while connected exceptives combine lower level constituents.⁷ Hence, ellipsis is part of the structure of free exceptives, but not connected exceptives, a non-trivial proviso to understand the diachronic path of senão. The main differences between the two types of exceptive structures are summarized under (i) to (iv). ⁶ The Post-Vulgate Cycle is one of the major Old French prose cycles of Arthurian literature. Written probably between 1230 and 1240, it is a rehandling of the earlier Vulgate Cycle (also known as the Lancelot-Grail Cycle). The Post-Vulgate Cycle does not survive complete, but has been reconstructed from French, Spanish, and Portuguese fragments (Bogdanow (1991–2006). The Portuguese translation dates from the 13th century but was only preserved by two later manuscripts (see Castro 1983, 1988), namely a 15th-century manuscript of the Queste del Saint Graal (Portuguese Demanda do Santo Graal—DSG in the examples) and a 16th-century manuscript of the Joseph of Arimathea (Portuguese José de Arimateia—JAR in the examples). Because the conditional exceptives that we will be dealing with are not attested in Portuguese texts from the 14th century on, it is clear that the examples found in the Arthurian texts and the profane poetry are rooted in the original 13th-century manuscripts and not in later copies. It is also important that attestations come from two different textual genres, i.e. narrative prose translations and poetry originally written in Galicien-Portuguese (a written language variety showing minor differences between texts attributed to Galician and Portuguese poets; crucially such differences do not include the conditional exceptives discussed in this chapter). ⁷ An example of a connected exceptive (taken from Pérez-Jiménez and Moreno-Quibén 2012: 590) is given in (i), to be compared with (4) above: (i)

a. Recibí regalos de todos los asistentes excepto Eva. got.1PL gifts of all the attendees except Eva ‘I received gifts from all those present except Eva.’ b. … [PP de [DP [DP todos los asistentes [BoolP [conjunction excepto] [DP Eva]]]]]

THE DIACHRONIC PATH OF SENÃO

75

(i) Connected exceptives are DP-internal constituents and usually follow the QP they operate on. Free exceptives are clause-level constituents and appear in both left- and right-peripheral sentential positions. They can also be parenthetical. (ii) Connected exceptives are licensed only by a restricted set of quantifier phrases, in particular DPs headed by the universal determiners every, all, and no, and their correspondents in different languages. They are not licensed by other quantifiers, definite DPs, indefinite polarity items in a negative context or a generic (null) operator in the antecedent clause. Free exceptives do not display such restrictions. (iii) Null QPs are allowed in free exceptives, but not in connected exceptives. (iv) Only connected exceptives have as part of their lexical meaning the Uniqueness Condition (Von Fintel 1993), requiring a unique exception set. Under all the criteria above identified, the exceptive conditionals of 13thcentury Portuguese are free exceptives. First of all, they usually surface in right-peripheral position (not following a QP), as illustrated in (9). Fronting is not permitted because the conditional clause must be preceded by negation, which obligatorily occurs in the main clause (cf. Section 4.3). They can, however, be parenthetical, like (10), which is compatible with their being free exceptives. (9) Nehu˜u˜ outro nõ no ousaria dizer se hu˜u˜ destes nõ. no.one other not it would.dare to.say if one of.these NEG ‘Nobody would dare saying it except one of them.’ (13th century, DSG635,.25, WOChWEL) (10)

tu nunca fizeste, se poucas vezes não, cousa que you never did, if few times NEG, thing that aprouvesse a Nosso Senhor Jesu Cristo would.please to Our Lord Jesus Christ ‘You haven’t ever done anything pleasant to Our Lord Jesus Christ, except in a few occasions.’ (13th century, JAR77,.101, WOChWEL)

Moreover, the licensing QP on which the exceptive clause operates can be null, as shown in (11). When overt, it can be an indefinite polarity item under the scope of negation, like al (‘something, anything’) in (12a), ren (‘anything’) in (12b), and outro (‘any other (one)’) in (12c). Both null QPs and indefinite

76

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

polarity items are barred from connected exceptives, which demonstrates that Old Portuguese exceptive conditionals are free exceptives. (11)

na terra nõ achou el se duros corações nõ. in.the earth not found he if hard hearts NEG ‘He couldn’t find anything in the earth except people hard at heart.’ (13th century, DSG60,.3, WOChWEL)

(12) a. se lhes Deus ora non val, | non jaz i AL if them God now not helps not stays there anything se morte non if death NEG ‘If God doesn’t help him, he can’t expect anything except death.’ (13th century, Poetry, TMILG)⁸ b. u˜a cousa vus direi: | non me quitará one thing you.DAT will.say.1SG not me will.free REN, eu bem no sei, | de vus querer anything I well it know from you-DAT wanting assi, se morte non. so if death NEG ‘I will tell you one more thing: nothing except death will free me from loving you.’ (13th century, Poetry, TMILG)⁹ c. nom havia OUTRO na camara, se el-rei e ele nam. not was other in.the room if the-king and he NEG ‘There wasn’t anybody else in the room except the king and him.’ (13th century, JAR60,.15, WOChWEL) Finally, the exceptive conditionals of 13th-century Portuguese do not seem to display the Uniqueness Condition (Von Fintel 1993). The presence of the exclusive focus-marker só/somente (‘only’) in (13a) indicates that the restriction to a ‘unique exception set’ must be constructed syntactically as it is not an intrinsic semantic feature of free exceptives. This is confirmed by (13b) where the right-hand coordinated constituent introduces an additional exception to that signalled by the conditional, that is, por morte (‘by death’). (13) a. nõ a ora tam alto omẽ do mundo que lhis not is now so high man of.the world that them ⁸ Fernan Padrom, Os meus olhos, que mia senhor (cf. Brea 1996: 308); a similar example can be found in Pero da Ponte, Por Deus, amigu’, e que será de mi (cf. Brea 1996: 784). ⁹ Vasco Fernandez Praga de Sandin, Senhor fremosa, par Deus, gran razon (cf. Brea 1996: 951).

THE DIACHRONIC PATH OF SENÃO

77

podesse muyto nuzir … se vós SOLAMẼTE SOLAMẼTE nõ could much threaten if you only NEG ‘There is not in the world a man powerful enough to challenge them, except you.’ (13th century, DSG651,.7, WOChWEL) b. e disse que des ora querya comecar a demãda and said that from now wanted start.INF the quest da Bescha Ladrador e que iamais nõ na leyxarya, of.the beast barking and that never not it would.quit ou se por morte nõ se ou por cõpanha onde if by death NEG or by company where SE pagasse, ata que lhi desse cima. would.pay until that it.DAT would.give end ‘And he said that he was ready to start the quest of the Barking Beast and he would never quit it unless he would find death, or some treasured friendship.’ (13th century, DSG477,.7, WOChWEL) Assuming that free exceptives involve a biclausal structure and ellipsis (PérezJiménez and Moreno-Quibén 2012), we find the necessary link between the former conditional exceptive construction and the current coordinate construction with senão (Matos 2003; Colaço 2005).

4.3 General characterization of Old Portuguese se … nom (if … NEG) exceptives In this section we will focus our attention on the conditional structures exhibiting discontinuity between the two items that will end up forming the word senão. In examples (14) to (19), the two items that will end up forming the word senão are signalled in bold, while the constituent surfacing between them is underlined. Let us first identify which constituents can intervene between the conditional conjunction se ‘if ’ and negation (NEG). Example (14) shows that it can be the subject. But more often it is a fronted object of different kinds, as exemplified in (15) to (17). In (15) the intervener is the direct object, in (16) it is a prepositioned object, and in (17) it is an adverb. In all cases there is predicate ellipsis in the conditional clause, which includes the verb and all constituents except the one that intervenes between the conditional conjunction and negation.

78

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

destes nõ. (14) a. Nehu˜u˜ outro nõ no ousaria dizer se hu ˜˜u˜u˜ destes hu no.one other not it would.dare say.INF if one of.these NEG ‘Nobody would dare saying it except one of them.’ (13th century, DSG635,.25, WOChWEL) b. Mas eu creio que nhum homem vio a besta em but I think that no man saw the beast in todo aquele caminho, se aquele aquele não. all that way if that.one NEG ‘But I believe that no one saw the beast anywhere, except that one.’ (13th century, JAR14,.7, WOChWEL) (15) a. na terra nõ achou el se duros duros corações corações nõ. in.the earth not found he if hard hearts NEG ‘He couldn’t find in the earth anything except people hard at heart.’ (13th century, DSG60,.3, WOChWEL) b. os que hi ham de ir nõ atendem se uos uos nõ those that there have to go not praise if you NEG ‘The knights that must go there do not praise anyone else except you.’ (13th century, DSG38,.12m WOChWEL) c. nunca devedes fazer en nulha cousa se ben ben non. never should.2PL do of.it any thing if good NEG ‘You should never do anything about it except good.’ (13th century, Poetry, TMILG)¹⁰ (16) a. Tu jamais a ela tornarás, se em sonhos não you never to it will.return if in dreams NEG ‘You will never return there except in your dreams.’ (13th century, JAR89,.7m WOChWEL) b. çercamo-lla de todallas partes assi que nõ podia sair circled-her.1PL of all.the sides so that not could get.out se por ˜˜u˜u˜ de por hu hu de nos nos nom if by one of us NEG ‘We circled it all around, so that it couldn’t get away, except passing by one of us.’ (13th century, DSG125,.8, WOChWEL) (17) a. nunca cuidou haver morte se entam entam nam never thought to.have death if then NEG ‘He never thought about getting killed except in that occasion.’ (13th century, JAR101,.31, WOChWEL) ¹⁰ Vasco Fernandez Praga de Sandin, Se vus prouguesse, mia senhor (cf. Brea 1996: 952).

THE DIACHRONIC PATH OF SENÃO

79

b. E assi andou quatro dias que nom comeo nem bebeo, and so went.on four days that not ate nor drank se pouco pouco nam if little NEG ‘And he kept going for four days without eating or drinking, except very little.’ (13th century, JAR109,.23, WOChWEL) Although, in general, the ellipsis displayed by the if -clause includes the verb, some examples can be found where the copular verb ser ‘be’ surfaces instead, as shown in (18). Sentences like (18a), (18b) make it clear that the constituent spelled out between se ‘if ’ and NEG has undergone leftward movement. (18)

a. jamais coraçom mortal nom as podera conhoçer, never heart mortal not them will.be.able know.INF se pello he. pello santo santo spiritu spiritu nom he he if by.the Holy Spirit NEG is ‘Nobody will ever be able to know them except through the Holy Spirit.’ (13th century, DSG161,.13, WOChWEL) b. Eu nõ sey caualeyro no mu˜do por que a I not know knight in.the world for whom her leixase, se da Mesa Rredonda nom fosse. leave. SUBJ.1SG if of.the Table Round not be.SUBJ.3SG ‘There is no knight in the world for whom I would leave her except the knights of the Round Table.’ (13th century, DSG122,.10, WOChWEL)

The root clause is always negative and includes an optionally covert quantificational element that licenses the exceptive clause. The alternation between the two possibilities is attested in (19). Whereas in (19a) the QP is overtly expressed by the NPI al ‘anything’, in (19b) it is covert. The two sentences are taken from the same prose text and are minimally distinct. Similar data are found in the poetic texts. (19)

morte nom a. Nom ha hi AL se morte not is there anything if death NEG (13th century, DSG176,.21, WOChWEL)

80

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

b. Nom ha hi se morte morte nom not is there if death NEG ‘There isn’t anything there except death.’ (13th century, DSG197,.7, WOChWEL) Although the requirement that negation be associated with the antecedent of the conditional clause (i.e. the root clause) appears to be mandatory, negation does not have to be overtly realized. It can be implied, as shown by the rhetorical question in (20). On the semantics of rhetorical questions and its ability to license strong NPIs, see Han (2002).¹¹ (20) Senhor, diserom os que hi stauam, quẽ uo-llo Sir said the-ones that there were who you.DAT-it.ACC poderia fazer se estes caualeiros nom que aojtẽ cõujdastes?¹² could tell if these knights NEG that yesterday invited ‘Who else could help you, sir, said those who were there, except the knights that you have invited to your house yesterday?’ (13th century, DSG118,.6, WOChWEL) The structure of the conditional clauses described in this section can be analysed as sketched in (21), with TP ellipsis licensed by the polarity-head nõ.¹³ Since the negation marker is a head (see (21) and also (22)), it does not give rise to intervention effects (cf. Rizzi 1990:15ff.) relative to the phrasal constituent undergoing movement to the left periphery of the conditional, that is, to the functional projection in the CP field encoding information focus (FocP in (21)).¹⁴ ¹¹ Also in contemporary Portuguese, senão can have as antecedent a rhetorical question: (i) Quem havia de querer ir lá senão ele? who would of want to.go there SENÃO he ‘Who would want to go there except him?’ ¹² The sentence in (20), displaying extraposition of the sentence-final relative clause, is not problematic for the analysis proposed here because in Old Portuguese relative clause extraposition could be derived by specifying coordination, in the sense of De Vries (2006). See on this subject, Cardoso (2010, 2017). ¹³ We are not working here with a Rizzian cartographic model. Thus we leave open the issue of what other specific functional categories might be activated in the CP field in Old Portuguese exceptive conditionals. ¹⁴ A reviewer rightly points out that conditionals display restrictions with respect to argument fronting and other instances of movement into the clausal left periphery, as demonstrated by Haegeman (2006, 2010b, 2012). In fact, according to Haegeman’s findings, Old Portuguese exceptive conditionals must have been ‘peripheral’ conditionals (of a distinct type from the ‘performative’ peripheral conditionals discussed by Haegeman), which in contrast to ‘central’ conditionals do not display such restrictions. If, as suggested by Haegeman (2010b: 616, footnote 16), ‘one way of integrating such clauses [i.e. peripheral conditionals] with the associated clause would be to adopt the “paratactic” projection (πP) as in Ga¨rtner 2001’, we would have identified further motivation for the reanalysis of the former exceptive conditionals as coordinate structures.

THE DIACHRONIC PATH OF SENÃO

(21)

81

na terra nõ achou el [CP se [FocP duros corações]i in.the earth not found he if hard hearts [ΣP [Σ' nõ [TP achou el ti ]]]] NEG found he ‘He couldn’t find in the earth anything except people hard at heart.’ (13th century, DSG60,.3, WOChWEL)

It has been proposed by different authors that TP ellipsis in Romance languages is licensed by the polarity-encoding category ΣP (Laka 1990), under the condition that its Spec, its head, or both be filled with polarity particles (cf. Matos 1992, for Portuguese; López 1999, Vicente 2006, 2010, for Spanish; Busquets 2006, for Catalan; Morris 2009, for French; among others).¹⁵ The representation in (22) below is taken from Busquets (2006) and illustrates the intended licensing configuration for TP Ellipsis. The different polarity particles usually exclude each other.¹⁶ In Old and contemporary Portuguese (as more generally in Romance) the verb moves to T, therefore the verb is deleted under TP Ellipsis whereas it is realized under VP Ellipsis, which allows us to clearly distinguish between the manifestations of the two types of predicate ellipsis (cf. Matos 1992; Costa et al. 2012; Martins 2016).¹⁷

¹⁵ Lipták (2013) also makes a similar proposal to account for TP Ellipsis in Hungarian. For a different type of TP ellipsis, referred to as Modal Ellipsis or Null Complement Anaphora (NCA), see Dagnac (2010) for French, Spanish, and Italian; Brucart (1999) for Spanish; Depiante (2000, 2001) for Spanish and Italian; Cyrino and Matos (2006), and Gonçalves and Matos (2009), for Portuguese. See also Saab (2008, 2010) for a thorough discussion of TP Ellipsis and references. ¹⁶ But a double filled ΣP is not excluded, as shown by the availability of [ΣP também [Σ' não]] (‘also NEG’) in Portuguese. On the other hand, *também sim (‘also AFF’) is excluded in Portuguese, presumably because Portuguese sim, in contrast with Spanish/Catalan sí is not a head (see Martins 2006b, 2013). Adapted to Portuguese, Busquet’s diagram in (22) will therefore take the form of (i): (i)

[ΣP também/sim (‘also’/AFF) [Σ' [Σ[±neg] não (NEG)] [TP [e]]]]

¹⁷ The examples below show that contemporary Portuguese permits the ellipsis site in the second member of a coordination structure to include or exclude the verb, because it allows both TP Ellipsis (where the verb in T is deleted) and VP Ellipsis (where the verb in T is phonologically realized). (i)

Portuguese: a. Bach é difícil de interpretar e Bach is hard to play, and b. Bach é difícil de interpretar e Bach is hard to play and ‘Bach is hard to play, and Mozart (is) too.’

Mozart Mozart Mozart Mozart

também. also. também é. also is.

(TP Ellipsis) (VP Ellipsis)

82

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

(22)

ΣP Σ'

SPEC també/tampoc

Σº

TP

[±neg] sí/no

[e]

The configuration in (22) is compatible with the proposal that the exceptive conditionals described in this section display TP Ellipsis. In Section 4.4 further arguments will be given in support of the view that the exception phrase that escapes ellipsis in the se … nom conditionals of 13th-century Portuguese is in the CP field.

4.4 The loss of se … nom exceptives and the syntax of information focus Two central issues must be discussed at this point: first, the identification of the structural position occupied by the constituent intervening between se ‘if ’ and NEG in the Old Portuguese sentences displaying discontinuity between se ‘if ’ and nõ (NEG); second, the reason why movement to that position ceased to be available (with the consequence that se ‘if ’ and nõ started to appear always adjacent and eventually came to be reanalysed as an exceptive conjunction). The sentences where a verbal complement precedes the negation marker nõ, like (21) above for example, are clearly OV sentences, as far as it is assumed that nõ is the head of ΣP, which dominates TP. Moreover, sentences like (18a), (18b) above, here repeated as (23a), (23b), where the copula ‘be’ surfaces, constitute further evidence of the OV order. (23)

a. jamais coraçom mortal nom as podera conhoçer, never heart mortal not them will.be.able know.INF se pello he. pello santo santo spiritu spiritu nom he he if by.the Holy Spirit NEG is ‘Nobody will ever be able to know them except through the Holy Spirit.’ (13th century, DSG161,.13, WOChWEL) b. Eu nõ sey caualeyro no mu˜do por que a I not know knight in.the world for whom her

THE DIACHRONIC PATH OF SENÃO

83

leixase, se da Mesa Rredonda nom fosse. leave. SUBJ.1SG if of.the Table Round not be.SUBJ.3SG ‘There is no knight in the world for whom I would leave her except the knights of the Round Table.’ (13th century, DSG122,.10, WOChWEL) Since Old Portuguese was an SVO language displaying verb movement to T (Martins 1998, 2011), both (21) and (23) are sentences with fronting of the object to some position above the verb position in the IP field. Martins (2011) analyses the Old Portuguese SOV word order mostly found in subordinate clauses as the outcome of middle scrambling. Old Portuguese object scrambling would move the object to the periphery of the IP space as a strategy to allow the constituent left in clause-final position to receive focus prominence.¹⁸ But this strategy to establish appropriate information focus configurations cannot be what we see in the Old Portuguese sentences above. As a matter of fact, in those sentences the fronted constituent that surfaces between se ‘if ’ and NEG is precisely the element informationally more important. Another piece of evidence indicates that the order OV exhibited by the relevant exceptive conditionals is not the type of OV word order studied by Martins (2011). While the latter can be found in Old Portuguese texts up to the 16th century, the sentences displaying discontinuity between se ‘if ’ and NEG (with the object being the most common intervener) totally disappear as early as the early 14th century.¹⁹ So we need a different approach to the word order pattern exhibited by the 13th-century Portuguese exceptive conditionals. We will pursue the hypothesis that in 13th-century Portuguese there were two positions available for a constituent bearing information focus prominence. One of them is the clause-final position that can be found from the

¹⁸ A similar strategy to establish appropriate information focus configurations is found in contemporary Portuguese (Ambar 1992; Costa 1998, 2004; Lobo and Martins 2017; among others). But in contemporary Portuguese only short scrambling is available. Hence, in contemporary Portuguese, object scrambling does not derive OV sentences. ¹⁹ All the examples of conditional exceptives we could find in the Old Portuguese prose texts are restricted to the 13th century. An isolated example appears in a Galician chronicle (the Galician translation of Crónica General and Crónica de Castilla) at the very start of the 14th century. As for the Galician-Portuguese poetry, we generally do not know the dates of composition of particular poems. So, for example, the king of Portugal and poet Dom Dinis (1261–1325) may have written the poem from which the sentence in (i) is extracted in the 13th or the early 14th century. But only poets born and (already) productive in the 13th century appear to display the relevant exceptive conditionals. (i)

nem vej’ i, amiga, morte non se morte nor see.1SG in.it friend-F.SG if death NEG ‘I can’t envision anything, my friend, but death.’ (Dom Dinis, Vai-s’ o meu amig’ alhur sem mi morar, cf. Brea 1996: 239).

84

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

earlier texts to contemporary Portuguese. The other is a left-peripheral position as identified in Latin (Devine and Stephens 2006) and different Romance languages at different historical stages (Cruschina 2008; Sitaridou 2011, 2012; among others), but which in Old Portuguese did not survive beyond the 13th century (the notions of ‘information focus’ and ‘contrastive focus’ are to be kept clearly separated). Once the choice for marking information focus prominence to the left was lost, the availability of the word order displayed by the if-clauses with discontinuity between se and NEG was also lost with the result that the adjacency between the two elements (se+NEG) in the relevant exceptive conditionals became mandatory. This fixed word order (associated with focus prominence to the right) in tandem with the characteristic predicate ellipsis displayed by the relevant if-clauses set the scene for the reanalysis of [se+NEG] as the exceptive conjunction senõ/senão.²⁰ A number of empirical facts provide evidence to support this hypothesis. The exceptive if-clauses allowed, in the 13th-century texts, an alternative word order with the focus-prominent constituent surfacing to the right of NEG, in clause-final position. That is to say, in the same texts where the discontinuous se … nom sequence is attested, sentences with adjacency between the two elements are also found, which supports the idea that two different strategies could be used to establish appropriate information focus configurations. The fact that, in the relevant data, the informationally prominent constituent surfacing at the right edge of the sentence is in most cases a heavy constituent (including a relative clause, as in (24), or some other type of clausal modifier, as in (25)) can easily be accommodated under the present account.²¹ At the same time, it makes it unlikely that the sentences in (24) and (25) merely represent the innovative structures with the exceptive word senão (signalling a situation of grammar competition or different chronological strata in the manuscripts).²² (24)

a. E nom ouue de comer senã heruas que and not had to eat if.NEG herbs that colhera hu˜u˜ homem ẽ sua orta gathered a man in his garden ‘And he didn’t eat anything except some herbs that a man had gathered in his garden.’ (13th century, DSG97,.5, WOChWEL)

²⁰ In Old Portuguese texts, the negation marker is spelled as nom/non/nõ, mam/nan/nã, or não. The latter spelling is found in current orthography. ²¹ The data appear to signal a prosodic bias towards the clause-final position for heavy constituents. ²² The fact that the sequence se plus NEG is spelled as one word or two words in the manuscripts (or in the editions of the manuscripts) has no syntactic relevance.

THE DIACHRONIC PATH OF SENÃO

85

b. em vaao emtraste na demanda do Santo Graal in vain entered in.the quest of.the Holy Grail e tu nom acharas hi senam and you not will.find there if.NEG onta que sobre ti ujnrra se te shame that over you will.come if yourself nam quitas deste pecado not free of.this sin ‘You entered the quest of the Holy Grail in vain and you will not find anything in it except the shame that will fall on you if you do not free yourself of that sin.’ (13th century, DSG201,.39, WOChWEL) c. nom tiverão outra guia senão u˜a estrela que not had another guide if.NEG a star that pareceo tão asinha como ele naceo e appeared so soon as he was.born and nunca ante homem vira. never before man had.seen ‘They did not have any guide except for a star that appeared in the sky as soon as he was born and had never been seen before.’ (13th century, JAR32,.15, WOChWEL) (25) a. por al nom fiz eu vijr tanta gente for anything not made I come.INF so.many people aa mjnha corte senam por ueerem as maravilhas to my court if.NEG for see-INF.3PL the wonders que auerram a esta mesa that will.experience at this table ‘I haven’t made so many people come to my court except for having them experience the wonders that will be revealed to them by sitting at this table.’ (13th century, DSG15,.12, WOChWEL) b. Uerdade he, mas se outro cavaleyro o começa nõ truth is but if other knight it starts not no devo ende estorvar, senã se ho caualeyro him should.1PL of.it obstruct if.NEG if the knight ffaleçe do que começa gives.up of.the what starts ‘That’s true, but I shall not get in the way of another knight unless he abandons what he started.’ (13th century, DSG229,.25, WOChWEL)

86

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

The structure that we propose to derive the word order of the exceptive conditionals in (24) and (25) is given in (26). The exceptive phrase moves to Belletti’s (2004) VP-peripheral FocP (a dedicated functional projection for a low/clause-final information focus) and the Σ-head still licenses TP Ellipsis, but the focussed material escapes PF-deletion (yielding an instance of ‘scattered deletion’ to avoid a phonological violation; see Nunes 2004; Boskovic´ and Nunes 2007; Sheehan 2011; among others).²³ (26)

[CP se [ΣP [Σ' nõ [TP verbi [FocP [exceptive phrase]j [VP ti tj (remnant material)]]]]]]

In Section 4.2, it was shown that exceptive conditionals displaying the copula ‘be’ instead of verb ellipsis confirm that the structures with discontinuity between se and NEG have an OV pattern (whenever the exception phrase is a verbal complement). As expected under our proposal, conditional exceptives with the copula ‘be’ also reveal that the adjacency between se and NEG correlates with VO structures. This is illustrated in (27). (27) a. quando elles hi aportarõ, vjrõ outra naue entre when they there landed saw other boat between outras penas hu nõ podoriam entrar, pero other rocks where not could enter even.if quisessem, se nõ fosse pena. fosse per cima de toda a wanted if NEG was by over of all the rock ‘When they landed there, they saw another boat hidden between the rocks, where they couldn’t get, even if they wanted, except going over the rocks.’ (13th century, DSG413,.6, WOChWEL) b. Eu nõ me partira de Corberique I not myself would.have.left from Corberique se nõ polo alcalçar fosse fosse if NEG was for.him reach.INF ‘I wouldn’t leave from Corberic (for any reason) except to reach him.’ (13th century, DSG552,.22, WOChWEL) Sentence (28) displays a recursive exceptive conditional. Hence, recursion was allowed in 13th-century Portuguese, while it is totally excluded in exceptive sentences with senão in contemporary Portuguese (i.e. the contemporary ²³ An analysis in terms of ‘scattered deletion’ is also compatible with the focussed material being not moved but left in situ (see Costa 1998, 2004; Costa and Figueiredo Silva 2006)

THE DIACHRONIC PATH OF SENÃO

87

equivalent of (28), featuring two instances of exceptive senão, is a fully ungrammatical sentence). Sentences like (28) support the hypothesis that two different structures were possible for exceptive conditionals. The first exception phrase in (28) is in the clausal left periphery, whereas the second one is clause-final.²⁴ Together, they represent the two different strategies available in 13th-century Portuguese, but not afterwards, to assign information focus. The loss of the left-peripheral option made unavailable the exceptive conditionals displaying discontinuity between se and NEG and prompted the emergence of the exceptive word senão.²⁵ (28) nom vio de nhu˜a parte se agoa nam, senam u˜a mui not saw of any part if water NEG if.NEG a very pequena pena donde ele estava small rock where he was ‘He couldn’t see (all around) anything but water, except for a small rock where he was staying.’ (13th century, JAR61,.7, WOChWEL) Typologically, 13th-century Portuguese would not be an oddity. Belletti (2004, 2008) hypothesizes that the languages that display two different structural positions for focus constituents reserve left-peripheral focalization for contrastive focus (Rizzi 1997) whereas the lower structural position specializes for information focus. This is indeed the pattern found in most present-day Romance languages, including Portuguese. Nevertheless, an experimental work conducted by Cruschina (2015) shows that Sicilian has two positions for foci, a left-peripheral and a lower one, but crucially it can use both for information focus (as well as for contrastive focus).²⁶ So, typologically, 13th-century Portuguese would be like Sicilian but early on it evolved towards a more common typological pattern.

²⁴ The sentence in (28) also gives evidence that the exceptive conditionals found in 13th-century Portuguese were peripheral (not central) conditionals. In contemporary European Portuguese, a sentence parallel to (28) would necessarily display two different exceptive conjunctions, for example senão and à exceção de (‘to the exception of ). (i)

não not

viu saw

em in

nenhum any

lado side

senão SENÃO

água, water

à to.the

exceção exception

de/*senão of/*SENÃO

uma pequena rocha onde ele estava a small rock where he was ‘He couldn’t see (all around) anything but water, except for a small rock where he was staying.’ ²⁵ Focus fronting (to the CP field) is available in contemporary European Portuguese but only as an instance of contrastive focus (see Costa and Martins 2011; Martins and Costa 2016). ²⁶ It is irrelevant for our current purposes to decide whether, under a cartographic approach, the CP field should contain one single FocP, with the option of being associated with different features and content or include a series of Focus projections for different types of foci.

88

ANA MARIA MARTINS, SANDRA PEREIRA, AND CL ARA PINTO

Table 4.1 OV order in three Old Portuguese texts, with O restricted to full nominal object DSG 13th century (transmitted by a 15thcentury manuscript) 212,145 words 1294/12,432 10.4%

JAR 13th century (transmitted by a 16thcentury manuscript) 140,130 words (integral text) 659/6438 10.2%

CGE 14th century (transmitted by 14th/15thcentury manuscripts) 403,580 words 608/7010 8.7%

An expected indicator of the change is a decrease in the frequency of OV order in the 14th-century texts, as one type of object fronting was lost, that is, fronting related to assignment of narrow information focus (whereas contrastive focus fronting, clitic left dislocation, left dislocation, and middle object scrambling continued to be possible sources of OV orders). Table 4.1 indicates that we may be making the right prediction, as the 14th-century text Crónica Geral de Espanha de 1344 (CGE) shows a decrease in the frequency of OV relative to the 13th-century texts Demanda do Santo Graal (DSG) and José de Arimateia (JAR).²⁷ The extent of the decrease in OV order between the 13th and 14th centuries may actually have been higher than shown in Table 4.1. Because the two 13th-century texts only survived through later manuscripts, we cannot be sure that all instances of the relevant OV order in the original texts were preserved in these later copies.

4.5 Conclusion The study of the Portuguese medieval Arthurian novels, translated from French in the 13th century, revealed the existence of a type of exceptive construction that had not been identified previously. This chapter analyses the Early Old Portuguese exceptive conditionals allowing discontinuity between se (‘if ’) and nom (NEG) and attempts to explain how the string formed by the conditional conjunction plus the negation marker originated the exceptive coordinating conjunction senão (‘except’). It is proposed in the chapter that this particular change correlates with a more general change bearing on the relation between word order and information structure. If our hypothesis ²⁷ The search in the corpus WOChWEL was restricted to root clauses displaying a non-clitic direct object.

THE DIACHRONIC PATH OF SENÃO

89

is on the right track, the loss of exceptive conditionals and concomitant emergence of senão signals that 13th-century Portuguese allowed focus fronting as a strategy to assign information focus, but this option was already unavailable by the beginning of the 14th century. Thereafter, only the clause-final position was available for constituents assigned information focus (an option uninterruptedly found from the earlier texts until nowadays). By focussing on the diachronic development of senão (if+NEG, i.e. ‘except’), the chapter should also contribute to a better understanding of the extent of possible syntactic variation in the domain of exceptive constructions.

Sources of the Historical Data—Corpora CGE Parsed Crónica Geral de Espanha de 1344, in WOChWEL (Sandra Pereira. 2016–18. Parsed Crónica Geral de Espanha. CC licensed: WOChWEL by Centro de Linguı´stica da Universidade de Lisboa). DSG Parsed Demanda do Santo Graal, in WOChWEL (Ana Maria Martins, Sandra Pereira, and Adriana Cardoso. 2013–15. Parsed José de Arimateia. CC licensed: WOChWEL by Centro de Linguı´stica da Universidade de Lisboa). JAR Parsed Livro de José de Arimateia, in WOChWEL (Ana Maria Martins, Sandra Pereira, and Adriana Cardoso. 2013–15. Parsed Demanda do Santo Graal. CC licensed: WOChWEL by Centro de Linguı´stica da Universidade de Lisboa). TMILG Tesouro Medieval Informatizado da Lingua Galega (Xavier Varela Barreiro, coord. ILG). . WOChWEL WOChWEL’s POS-tagged and Parsed Old Portuguese Texts. A. M. Martins, coord. CLUL/FCT. .

5 On Italian relative complementizers and relative pronouns Rethinking grammaticalization Emanuela Sanfelici, Jacopo Garzonio, and Cecilia Poletto

5.1 Introduction This chapter explores the syntactic status of relativizers, that is, what are standardly referred to as relative complementizers and relative pronouns, in Old and Modern Italian varieties, and proposes a unified analysis for both types of items. It takes into account the ongoing debate regarding the categorial status of relativizers (Kayne 1975, 2008, 2010; Lehmann 1984; Manzini and Savoia 2003, 2011; among many others) and aims to show that what we call complementizers are not C° heads, as commonly assumed. Instead, we propose that both relative ‘complementizers’ and ‘pronouns’ have the same categorial status, that is, they are wh-items.¹ The reason why the traditional view claimed that two distinct types of categories, that is, (i) relative pronouns and (ii) complementizers, can introduce a relative clause, are the following three main properties (e.g. Klima 1964; Kayne 1975): (1) a. Case marking: complementizers do not inflect for case, pronouns do. b. Sensitivity to animacy: complementizers do not carry animacy features, pronouns do. c. Compatibility with prepositions: complementizers cannot be combined with prepositions, pronouns can. ¹ For a proposal on the syntactic derivation of relative clauses we refer to Poletto and Sanfelici (2014): capitalizing on Cinque’s (2008, 2013) analysis according to which all relative clauses have an internal as well as an external head, they propose that both pronouns and complementizers are syntactically determiner-like elements of the internal head noun.

Emanuela Sanfelici, Jacopo Garzonio, and Cecilia Poletto, On Italian relative complementizers and relative pronouns. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Emanuela Sanfelici, Jacopo Garzonio, and Cecilia Poletto (2022). DOI: 10.1093/oso/9780198871538.003.0005

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

91

These three characteristics have been considered to depend on a single abstract property, namely the fact that complementizers are heads, and as such, cannot be selected by prepositions, and do not have typical pronominal features like case and animacy. More recently, however, many proposals have been advanced to disclaim that pronouns and complementizers differ in terms of their categorial status. For instance, Kayne (2008, 2010) has argued that there is no distinction between relative and declarative complementizers as both are relative pronouns (and not complementizers).² A similar proposal has been advanced by Manzini and Savoia (2003, 2011) for Italian che, which, in their view, is not a complementizer but a wh-item in all the contexts where it appears.³ In this work we provide empirical evidence that complementizers and pronouns have the same categorial status and are subject to the same derivation. We show that properties (a) and (b) do not hold across Romance for all socalled complementizers. The testing ground for our claim is a diachronic investigation of three Old Italian varieties, namely Old Ligurian, Old Piedmontese, and Old Neapolitan. These varieties show that what should be considered as complementizers display agreement patterns with the lexical head, a property typical of pronouns. However, they still cannot be selected by prepositions. This means that the three characteristics that single out relative complementizers from actual pronouns cannot be derived from the same abstract property and show that complementizers can have at least some of the typical pronominal properties. At the same time, typical relative pronouns like Old Neapolitan (il) quale do not display overt agreement with the head noun, that is, there exist relative pronouns that behave like complementizers according to properties (a) and (b), although they are compatible with prepositions (c). This again shows that the dichotomy between complementizers and relative pronouns does not hold. We interpret this conclusion in diachronic terms by arguing that relativizers never change their categorial status but only lose their feature specifications in Modern Standard Italian in which che only has one, that is, that of the wh-morpheme. This means that there has never been a change from pronouns to complementizers at least throughout the history of Italian (contrary to the usual claim Roberts and Roussou 2003; van Gelderen

² Under Kayne’s (2008) view, complementizers do not exist, i.e. there is no such category that is merged in the head of C, since there are no overt heads. This implies that all complementizers, including those of complement clauses have to be reanalysed as phrasal categories. ³ For a similar proposal, see Sportiche (2011) on qui/que in French. The idea that complementizer que and interrogative que in French are the same element was already present in Obenauer (1976).

92

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

2004; among many others). In other words, the shift from pronoun to complementizer cannot be seen as an instance of a grammaticalization process in the terms of van Gelderen’s Head Preference Principle, which assumes that heads are preferred over full XPs for reasons of economy. On the contrary, the grammaticalization process that the relativizer undergoes can be considered as the reflex of the well-known cartographic assumption that each projection expresses a single feature (see Cinque and Rizzi 2010b) but this feature can be realized by its specifier and not only by its head. The chapter is organized as follows. In Section 5.2 we will briefly summarize the state of the art concerning relativizers in Old and Modern Standard Italian, with the focus on che. Section 5.3 is devoted to providing empirical evidence from diachrony for the claim that che has the same categorial status as quale. In Section 5.4 we will outline the implications derived from Section 5.3 for the syntactic theories on language change. We will cast doubts on their empirical and theoretical adequacy and reformulate them in a different perspective. Section 5.5 will draw the conclusions.

5.2 From relative pronouns to complementizers: Old and Modern Standard Italian Modern Standard Italian, as well as other Romance languages, displays a mixed relativizer system. Two types of elements can introduce a relative clause: wh-elements, namely (il/la) quale lit. ‘the which’; an element that also serves to introduce complement clauses, namely che ‘that’.⁴ Their distribution is sensitive to two parameters: the type of relative clause, that is, whether restrictive or appositive, and the argumental function of the relativized phrase. In restrictive relative clauses only che can relativize subjects and complements not selected by prepositions as in (2a), (2b), while forms (il) quale are only admitted with PPs, cf. (2c), (2d). On the contrary in appositives (il) quale can always be used independently from the thematic role and case of the relativized element, cf. (3a–d). (2) a. La ragazza che / *la quale ho incontrato ieri… the girl that / the which have.1SG met yesterday ‘The girl that I met yesterday …’ ⁴ Italian also has another wh-pronoun, which is only found in relative clauses, namely cui. Since its historical development and its analysis are very complex, we leave this element out of the present work and refer to Rohlfs (1966) for a discussion of the etymology and to Cinque (1978) for an analysis of its syntactic properties.

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

93

b. Il libro che / *il quale è sul tavolo … the book that / the which is on.the table ‘The book that is on the table …’ c. La ragazza con la quale / *che ho parlato ieri … the girl with the which / that have.1SG talked yesterday ‘The girl with whom I talked yesterday …’ d. Il libro del quale / *che tutti parlano … the book of.the which / that everybody talks ‘The book about which everybody talks …’ (3) a. Maria, che / la quale non vedo da oltre tre anni … Maria that / the which not see.1SG from over three years ‘Maria, whom I haven’t seen since three years …’ b. Roma, che / la quale è la capitale d’Italia … Rome tha / the which is the capital of Italy ‘Rome, which is the capital of Italy …’ c. Maria, con la quale / *che ho parlato ieri … Maria with the which / that have.1SG talked yesterday ‘Maria, with whom I talked yesterday …’ d. Roma, della quale / *che tutti si innamorano … Rome of.the which / that everybody cl love.3PL ‘Rome, which everybody falls in love with …’ To account for the pattern in (2) and (3), Cinque (1978, 1982, 2013) concludes that whereas (il) quale is a wh-pronoun, che is a complementizer, and as such it cannot be combined with prepositions and is insensitive to the animacy of the head noun, as (2b) and (3b) shows. Similarly to Modern Standard Italian, in Old Italian (OI) headed relative clauses can be introduced by both che ‘that’ and (il) quale ‘the which’. Benincà and Cinque (2010: 472) claim that whereas in the modern stage of Italian che is simply a subordinator marker, in OI che can in some cases be a complementizer but in some others it can also be a relative pronoun. The arguments they provide in favour of this hypothesis are the following: I. In OI che can be combined with prepositions as in (4), which is impossible in Modern Standard Italian (cf. (2c), (2d) and (3c), (3d)). (4)

Uno bastone con che s’apogiava perch’era debole A stick with that CL rested because was weak ‘A stick on which he rested because he was weak’ (Fiori e vita di filosafi, 9, 4–5)

94

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

II. In OI che is sensitive to the human/animate distinction: when preceded by prepositions it is exclusively referred to a [−animate] antecedent, as exemplified in (4). Nevertheless, in other cases, che clearly also behaves as a complementizer in OI. If the relativized element is the grammatical subject or direct object, or other complements without prepositions, che is insensitive to the semantic distinction and can also introduce a relative clause on a [+human] antecedent. This is shown in (5), where although the antecedent is [+human] che introduces the relative clause. (5)

Andò alli altri giovani che stavano a ricevere l’acqua Went to.the other young that were at receive the water ‘He went towards the other young people that were receiving the water’ (Novellino, 4, l. 16–17)

This means that postulating a dichotomy between complementizers and pronouns comes at the cost that in OI the same element must have both specifications in the lexicon, which is clearly not very elegant. Adopting van Gelderen’s (2004, 2009a, 2009b) account of grammaticalization, the change in the system of relativizers in the history of Italian that Benincà and Cinque observed could be interpreted as an instance of reanalysis from a specifier into a head, under the Head Preference Principle. However, this is not a necessary conclusion. The only difference between Old and Modern Italian che lies in its feature specification. All we need to explain the change illustrated by Benincà and Cinque (2010) is to assume that from being fully specified for nominal features such as case, gender, number, and animacy che turned into a morphologically unmarked element with the only semantic content of being a variable ranging over individuals or properties.

5.3 Complementizers and relative pronouns in old Italian varieties The fact that ‘relative pronouns’, for example, Italian type (il) quale, and ‘relative complementizers’, che, do not differ in categorial terms and hence relativizers come in one single fashion, that is, as specifiers, is shown by two sets of phenomena. The first phenomenon is the one of agreeing complementizers: in some OI varieties, what is usually referred to as a relative complementizer actually carries subject agreement features, which should only be found with pronouns

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

95

according to the properties mentioned in (1) (see Section 5.3.1). The second phenomenon is the converse case, namely the one of ‘uninflected’ relative pronouns: what is usually labelled as a relative pronoun of the quale type can be invariant like a complementizer (Section 5.3.2).

5.3.1 Agreeing ‘complementizers’ In Old Neapolitan and Medieval North-western Italian, supposed ‘complementizers’ behave just like pronouns and carry agreement morphology with the head noun in restrictive relatives. The agreement features found on the relativizer are different from those expressed inside TP since these are typically found in the nominal domain, as they are either case and animacy, as in Medieval North-western Italian dialects (Section 5.3.1.1) and or case, gender, and animacy, as in Old Neapolitan (Section 5.3.1.2). Differently from Old Florentine, in the relativizer paradigm old North-western varieties, together with other Northern varieties such as Lombard and Emilian (but not the Northeastern Veneto varieties), and Old Neapolitan maintained the case opposition that distinguishes relatives on the subject from relatives on the object in a way which looks prima facie similar to the modern French system (Rohlfs 1966; Formentin 1996; Parry 2007; Ledgeway 2009). However, on closer scrutiny, the historical data reveal a more complex and interesting picture. 5.3.1.1 Case and animacy on the ‘complementizer’: Medieval Northern Italian dialects Medieval Northern-western Italian dialects display a different form of the relativizer in subject relatives when the head noun is animate. Since this phenomenon is only found in subject relatives, we also assume that the relativizer is sensitive to case, or at least to the distinction between nominative and all the rest. We chose two varieties to illustrate the system, namely Old Ligurian and Old Piedmontese. In these dialects the form of the complementizer in complement clauses is che/que (6). (6)

saveien que Deus avea recue lor sacrifice know.3PL that God has received their sacrifice ‘They know that God received their sacrifice.’ (Sermoni Subalpini 1, 220)

Parry (2007) noticed that in the 13th-century texts the same unmarked form of the complementizer is also found in object relative clauses (7a) for Old Ligurian and (7b) for Old Piedmontese, whereas a different form, chi, introduces

96

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

subject relatives (8a) and (8b) respectively, in a way which strongly recalls the modern French system. (7) a. Semo contenti che ogni terra che tenen li Venecian are.1PL happy that each land that keep the Venetian sea de lo Rey is.SUBJ of the king ‘We are happy that each land that the Venetian occupy is owned by the king.’ (Proposizioni fatte dal Comune di Genova al re d’Ungheria, 24:15) b. era figura de la novela que Christ fis e comandè was image of the novel that Christ did and ordered ‘It was image of the novel that Christ did and ordered.’ (Sermoni Subalpini 10, 252, 11–14) (8) a. ogni lavorao chi lavore a lo Ponte de lo Peago every worker that works at the Ponte of the Peago ‘Every workers that works at Ponte of Peago […].’ (Statuti della Compagnia dei Caravana, 9:17–18) b. Car no serà neun qui sia there not be.FUT nobody that is.SUBJ blanquì per baptisme, qui no veigna purified for Baptism that not comes.SUBJ ‘There will be nobody that is purified for the Baptism and that would not come.’ (Sermoni Subalpini 5, 233:36–7) However, already in the 13th-century texts and more frequently in the 14th– 15th-century texts, the marked form chi appears only when the subject antecedent is animate as in the examples in (9). Otherwise, when the subject antecedent is inanimate, che is found as in (10) and (11). (9)

questa femena chi m’ à spanyunto questo this.FEM woman REL to.me has spread this inguento adosso unguent on.me ‘This woman that spread this unguent on me.’ (Passione 28)

(10)

questa cità que avea num Iherico this city REL had name Gericho ‘This city that was named Gericho […].’ (Sermoni Subalpini 9, 246, 12–13)

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

(11)

97

receveyva tuto zo che era dayto a Criste received.3SG all that REL was given to Christ ‘He received all that was given to Christ.’ (Passione 28)

We can conclude that in the 14th-century texts, the complementizer in subject relatives is simultaneously marked for (nominative) case and for animacy. We conducted a survey of two texts for the two dialects, available on the lemmatized database Opera del Vocabolario Italiano: We chose Passione (1356) for Old Ligurian and Sermoni Subalpini (around 1300) for Old Piedmontese. As for the Passione, the total of occurrences of chi was 74, among which we excluded the following instances: 12 with interrogative pronouns; 5 instances containing the relative chi selected by prepositions. Hence, we analysed 57 instances of chi. We searched for che and the total of occurrences we gathered was 197. Among these, we excluded 131 with interrogative pronouns and complementizers selecting a complement clause. We further excluded 12 instances containing the relative che selected by prepositions. That left 54 occurrences of che to be analysed. We manually counted and classified the occurrences according to the animacy of the head noun and the type of extraction, subject vs. object. Table 5.1 summarizes the results, confirming that the vast majority of chi was found with animate subject antecedent. We then made the same survey of Sermoni Subalpini. We found 389 occurrences of chi, from which we excluded 22 containing interrogative pronouns, 9 with the chi relative selected by prepositions. As for che, we found 430 occurrences. From these we excluded 206 instances with interrogative pronouns and complementizers selecting a complement clause, as well as 36 instances containing the che relative selected by prepositions. Again, the vast majority of chi was found in subject extraction with animate antecedents (Table 5.2).⁵ In these old dialects the ‘complementizer’ carries nominal features and agrees with the antecedent, although it is rarely selected by prepositions: in Table 5.1 Old Ligurian distribution of chi/che in Passione (1356) Chi (N=57)

NOM ACC

Che (N=54)

[+ANIMATE]

[−ANIMATE]

[+ANIMATE]

[−ANIMATE]

54 0

3 0

0 23

18 13

⁵ The rather high number (62) of chi in subject extractions with a [−animate] antecedent may suggest that the pattern is hybrid. Whether these instances are the reflex of an older or of an innovative system is still to be detected.

98

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

Table 5.2 Old Piedmontese distribution of chi/che in Sermoni Subalpini (1300)

NOM ACC

Chi (N=367)

Che (N=188)

[+ANIMATE]

[−ANIMATE]

[+ANIMATE]

[−ANIMATE]

305 0

62 0

3 91

32 62

Passioni, we found 5 instances with chi and 12 with che; in Sermoni Subalpini 9 with chi and 36 with che.⁶ Our findings are compatible with the results reported in Fiorentino (1998), according to which prepositions combine with (il/lo) qua(l/r) and in the majority of the 14th–15th-century texts never combine with chi/che. As a final note, Fiorentino (1998) shows that in the 16th-century texts che also generalized over chi in subject extractions. 5.3.1.2 Case and gender on the ‘complementizer’: Old Neapolitan (circa) 1300–1350 The same type of phenomenon noted above is found in Old Neapolitan, although in a more complex fashion: the complementizer of restrictive subject relatives displays two different forms, one for the masculine animate antecedent and a different one for the masculine inanimate and for the feminine (animate and inanimate), which means that the system is sensitive to at least three features: (a) case, (b) gender, and (c) animacy, none of which is also marked in the TP. Formentin (1996) and Barbato (2001) note that Old Neapolitan, like several other Old Southern varieties, has two forms of the complementizer in complement clauses: che and ca (also attested as ), whose distribution is related to the properties of the main verb.⁷ (12)

a. prego che li boni lecture piglieno da beg.1SG that the good readings take.3PL from lor consiglio them counsel ‘I beg that the good readings would take advice from them.’ (Brancati 189 v.34) b. respondimo-ve cha ne par(e) reply.1PL-to.you that SCL seems ‘We reply to you that it seems […].’ (Lettera 126.15-16)

⁶ It is worth noticing that the great majority of instances containing che combined with a preposition is of the type per che ‘for which’, which could be argued to be the subordinate conjunction ‘because’. ⁷ The different forms of complementizer are related to modality (cf. Ledgeway 2009).

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

99

According to Formentin (1996), the ‘complementizer’ introducing relative clauses also displays two different forms: chi, in subject extractions (13a), and che, in object extractions (13b). (13)

a. quilli Grieci, chi navegavano in altra parte those Greeks that sailed in other area ‘Those Greeks, who have sailed in other areas […].’ (Libro de la destructione de Troya 76. 21–2) b. le iniuriose parole che avea in bocha the insulting words that had.3SG in mouth ‘The insulting words that he pronounced […].’ (Libro de la destructione de Troya 111. 23)

As both Formentin (1996) and Barbato (2001) conclude, this shows that Old Neapolitan relative pronouns exhibit case distinctions: chi for nominative (13a) and che for accusative (13b), in both restrictive and non-restrictive RCs. A closer look at the texts reveals that the distribution of the two forms is not only sensitive to the distinction subject/object relatives (i.e. to case) but also at least to the gender of the extracted subject (Ledgeway 2009). Indeed, when the subject is masculine, the relativizer is chi as in (14), while, when feminine, it is che (< lat. quae Formentin 1996: 139), as in (15). (14)

Lo re de Cipre chi se clamao Eneo the king of Cyprus that CL called Enea ‘the king of Cyprus who was named Enea.’ (Libro de la destructione de Troya 153. 14–15)

(15)

Questa Medea che desiderava tanto la soa dolce this Medea that desired a.lot the her sweet partenza departure ‘Medea, who really desired her sweet departure […].’ (Libro de la destructione de Troya 67.24)

In order to provide a quantitative analysis of the two forms of relativizer in subject relative clauses, we made a survey of one text, Libro de la destructione de Troya (LDT) (1330–40, Paris manuscript) available on the lemmatized database Opera del Vocabolario Italiano. The total number of occurrences of the element chi was 656. From these, we excluded 90 occurrences: 76 with chi as an interrogative pronoun; 14 instances with chi selected by prepositions— from these 9 occurrences had an interrogative pronoun and 6 contained the

100

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

Table 5.3 Distribution of chi/che in subject extraction (LDT)

MASCULINE FEMININE

Total

Chi

Che

548 18 566

147 268 415

relative one. We were left with a total of 566 occurrences, which were instances of chi in subject extraction in relative clauses. We also made the same survey for che. Among the total amount of 3771 occurrences, we excluded 3356 either because they were instances of che in various subordinate clauses (N = 1690), or che as interrogative pronoun (N = 951), or che as relativizer in object (N = 679) and che relative combined with prepositions (N = 36).⁸ Hence, we analysed 415 occurrences containing che in subject relative clauses. We manually counted the instances and classified them according to the gender of the head noun. Table 5.3 summarizes the results, confirming that the vast majority of chi was found with masculine. We conclude that in Old Neapolitan the agreeing ‘complementizer’ in subject relatives is specified for both case and gender, which is a property only expected of relative pronouns. Nevertheless, it is usually not selected by prepositions, as would be anticipated from relative pronouns.⁹ Looking more closely at Table 5.3, it emerges that whereas the 18 cases of feminine can be interpreted as the result of an analogical generalization of chi as a relativizer for subject (as proposed by Formentin 1996: 140), the high number of che with masculine antecedent is suspicious. Checking the examples, we can conclude that 123 of the 147 instances which contain che in subject extraction, all have a masculine head noun specified for [−animate]. Compare (16) with (14). (16)

multi cirii de cera che faceano la camera many candles of wax REL make.3PL the room tutta lucent all bright ‘Many candles of wax that lit up the whole room […].’ (Libro de la destructione de Troya 61.28)

⁸ Our finding concerning the small amount of che combined with the preposition (6 instances with chi and 36 with che) is in line with the claim in Barbato (2001: 92) that the preposition in Old Neapolitan is usually combined with the relative pronoun (lo) qual. ⁹ As in the case for Old Ligurian and Old Piedmontese, the great majority of instances with che selected by the preposition is of the type per che ‘for which, because’.

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

101

This suggests that in the Old Neapolitan system there are three features that can play a role in the distribution of che/chi, namely case, animacy, and gender. 5.3.1.3 Conclusions on the agreeing ‘complementizers’ Up to now, we have provided evidence that when looking at diachronic dialectal data, there are clear cases in which relativizers are not selected by prepositions, but do display agreement with the extracted subject. The variation we observe shows that this agreement crucially does not replicate the same features found in the TP area, such as number or person, but expresses (i) an animacy, and/or (ii) a gender feature. One way to account for these data would be to conclude that subject relatives are special in some sense and the relativizer introducing them is a relative pronoun, whereas it is a complementizer in the case of object extractions (on this, see Sanfelici et al. 2014), similarly to what Benincà and Cinque (2010) concluded for OI. The authors propose that che is a relative pronoun when selected by prepositions, and a complementizer when the head noun is a bare NP. This analysis implies that the element che has two lexical entries, one as a complementizer and one as a pronoun, which is not very elegant and methodologically undesirable (cf. Cinque 2007: 93 on the methodological aim to unify in a single structure various and different instances of the same item/structure). Furthermore, in Old Ligurian, Piedmontese, and Neapolitan subject relatives, chi and che are part of an opposition in terms of nominal features which are not usually marked in the CP domain, such as gender and animacy, so even when they should be complementizers, they behave like pronouns.

5.3.2 Uninflected relative pronouns The second set of phenomena which shows the categorial unity of relativizers is represented by the converse case of agreeing complementizers: that is, what is usually labelled as relative pronoun of the quale type can be invariant just like a complementizer. For instance, in Old Neapolitan qual- exhibits no overt agreement although it can be combined with prepositions. This is unexpected under the standard view that there is a dichotomy between complementizers and relative pronouns according to which the three properties illustrated in (1), repeated here as (17), always go together:

102

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

(17) a. Case marking: complementizers do not inflect for case, pronouns do. b. Sensitivity to animacy: complementizers do not carry animacy features, pronouns do. c. Compatibility with prepositions: complementizers cannot be combined with prepositions, pronouns can. Qual- displays properties (a) and (b) but not property (c), which shows that the three properties have to be dissociated and cannot depend on one single property. Ledgeway (2009) notes that in Old Neapolitan, qual- exhibits no overt agreement, similarly to the Standard Italian che. Relative qual- shows the adjectival agreement pattern of a third declension adjective: -e in both singular and plural, both feminine and masculine (Ledgeway 2009).¹⁰ (18)

(19)

Tutte queste parole le quale all.PL.F this.PL.F word.PL.F the.PL.F REL re Peleo king Peleo ‘A ll these words which the king Peleus said’

le disse CL said

lo the

(LTD 51.17–18)

glora de Iesu Christo et dela Vergene matre, glory of Jesus Christ and of.the Virgin mother, li quale illumenenno lu intellectu the.PL.M REL bright the mind ‘Glory of Jesus Christ and of the Virgin Mother, who bright the mind’ (SDM 65.14–16)

This is also true when qual- is not preceded by the article. (20) a. Amico quale te friend REL you ‘Friend that you are’

si’ are (LTD 76.20–2)

¹⁰ We abstract here from other details. Qual- also exhibits another agreement pattern, i.e. the one attested for determiners and quantifiers: -u (m.sg); -a (f.sg.); -i/-a (pl). This is true both when qualis found in interrogatives (quala nziria ‘which danger’) as well as when it is found in relatives, both with qual- without the article (da certe brigate… con quali familiarmente se accompagnò ‘From some groups with whom usually he was together’ Masuccio XLIII.14) and qual- preceded by the article (certe alter provincie le quali ‘some other provinces which’ LAFGP 6v.b. 19–21). It is enough here to note that the variation in the agreement patterns, between an agreement determiner-like and one adjectival-like, attested for qual- is the one attested for prenominal adjectives in general (Iandolo 2001).

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

103

b. Haverno facte cose quale mai tentarono fare have.3PL done things REL never tried do ‘They did things that they never tried to do.’ (Brancati 162. 10–11) Another relevant aspect is that in the same vein as qual- in relative clauses, there are instances in which the interrogative adjective qual- follows the same declension pattern found in relatives, that is, the pattern of a third declension adjective that does not display either gender or number agreement: quale ri ‘which kings’, quale peccato ‘which sin’. This suggests that qual- as relativizer and qual- as interrogative determiner have to be treated in a similar way.

5.3.3 Conclusion ad interim On the basis of data coming from Old Italian dialects, we have shown that the standard dichotomy between complementizers and pronouns cannot be maintained, being contradicted by the two sets of phenomena we have investigated. In the first part of Section 5.3 we have seen that what is usually called a complementizer can carry subject agreement features of a different type with respect to phi-features that is, features typically found in the nominal domain, such as animacy and/or gender. Furthermore, we have seen that although the ‘complementizers’ agree with the extracted subject, just as pronouns do, they are only very rarely found with prepositions. On the other hand, what is usually labelled as a ‘relative pronoun’ can behave as a complementizer and avoid carrying overt agreement with its antecedent— although it is compatible with a preposition.¹¹ We conclude that the three characteristics listed under (1), repeated as (17), cannot stem from a single abstract property, since they can be dissociated, and that the dichotomy between complementizers and pronouns based on (1) has no reason to exist. If the change documented for Old Ligurian, Piedmontese, and Neapolitan concerning che is not to be analysed as a change from pronoun to complementizers, since there exist no relative complementizers, then it has to be attributed to a different cause. We propose that the loss of agreement which occurred around the 16th century is nothing else than, precisely, a loss of feature specifications on the element so that from being inflected it turns into an uninflected ¹¹ Indeed, to understand why che in Old Italian was compatible with prepositions, contrary to Modern Italian, we believe that other properties of the nominal domain must be addressed. Hence, given the space limitations here, we leave this topic aside and we refer to Poletto and Sanfelici (2014) for discussion.

104

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

che. The progressive loss of nominal features started from the Latin system, in which the relativizer was marked for case, gender, and number, to a system which only displays overt features in cases of subject extraction—in the same way as the Old Ligurian, Old Piedmontese, and Old Neapolitan system or simply nominative case as in Modern French—is complete in Modern Italian, where the relativizer che is invariable, as it introduces a variable ranging over individual instances or properties in all contexts in which it appears.¹² However, this does not necessarily imply that che has changed its categorial status, that is, from a pronoun it has become a head. We can then sketch the diachronic change che underwent according to the path formulated in Table 5.4. As Noordhof (1937: 84), Fiorentino (1998), Parry (2007), and De Roberto (2008) pointed out, when the ‘complementizer’ reached the stage represented by Modern Italian, we observe the spread of resumptive clitics which can be

Table 5.4 The feature specification on the ‘complementizer’ in diachrony Latin

Old Neapolitan

Old Piedmontese/ Ligurian

French*

Old Italian / Modern Italian

Case

Case

Case

Case

Wh-variable

Gender

Gender

Animacy

Wh-variable

Number

Animacy

Wh-variable

Wh-variable

Wh-variable

* It is still controversial whether Old French patterns with early Italian varieties or not. On the one hand, Old French has been suggested to pattern with Old Italian varieties. For instance, Ledgeway (2012: 306) argues that subject relative clauses are marked by qui when the subject is animate and agentive, whereas que, even if it is used for human antecedents, generally denotes a non-agentive subject. The author proposed that this pattern is typical also of Northern Romance languages. On the other hand, however, these observations seem to contrast with the traditional descriptions of Medieval Occitan and Old French. According to Jensen (1986, 1990), both qui and que are used indifferently and no animacy/agentive features are involved. However, the picture is more intricate. Jensen (1986: 140) further noticed that in Medieval Occitan there is a tendency to also use que in subject relative clauses and when looking at the instances the author offers, no examples of que used with an animate antecedent appear. This may suggest that also in Old French varieties the relativizer was specified for both case and animacy, similarly to Old Italian varieties. Further studies are in order to determine whether French instantiates a parallel path alternative to that of Old Italian varieties or whether French represents stage four as in Table 5.4, where only an opposition between subject and object is marked. Hence, we leave open the exact diachronic path and, more precisely, the exact position of French.

¹² We adopt Manzini and Savoia and Kayne’s conclusion that che is always the same element in interrogatives, relatives, and complement clauses. This is so because its feature specification is so poor that it is compatible with several different contexts.

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

105

considered as an alternative strategy to mark case, crucially the case previously carried by the relativizer itself.¹³ Although an investigation on the rise of resumptives is beyond the scope of this chapter, it is interesting to underline the fact that the features previously specified on a single element tend to be spelled out on two different items, as expected under the ‘One feature—one projection Principle’ generally adopted in the cartographic approach (see Cinque 1999; and Cinque and Rizzi 2010b on this).¹⁴, ¹⁵

5.4 New insights into grammaticalization in the light of relativizers The conclusion reached in Section 5.3.3 brings us to reconsider the process of Grammaticalization in its essence. Grammaticalization is generally defined as a process whereby a lexical item develops into a grammatical marker: ‘[g]rammaticalization consists in the increase of the range of a morpheme advancing from a lexical to a grammatical and from a less grammatical to a more grammatical status’ (Kuryłowicz 1965: 69). This process is usually accompanied by phonetic and phonological reduction, loss of syntactic independence and lexical meaning, and increase of grammatical function (see e.g. Heine and Reh 1984; Traugott and Heine 1991). Well-known examples include lexical verbs changing to auxiliaries and prepositions to complementizers. Most accounts have been developed in a functionalist framework but recently there have been attempts to formalize grammaticalization in structural terms. The latter view is shared by Roberts and Roussou (2003), and van Gelderen (2004, 2009a, 2009b) in their discussion about changes of lexical heads to functional ¹³ In a corpus study of 29 Old Florentine (c. 1200) texts, the pattern complementizer plus resumptive clitics appears in object relative clauses first (De Roberto 2008: 314–16). The situation is replicated in the 14th-century texts. This finding may suggest that the use of clitics in relative clauses is not only reserved to the cases in which the argument relativized is in a low position on the Keenan and Comrie’s (1977) Accessibility Hierarchy, e.g. indirect object. ¹⁴ This principle seems to be in competition with the Superset Principle of the nanosyntactic approaches (Starke 2009), i.e. bigger lexical items win over smaller competitors such that mice, which stores the entire tree [N plural], wins over *mouse-s, which spells out the tree [N plural] by two different lexical items mouse=N and s=plural. We leave for further research the investigation of the interaction between these two principles in both synchrony and diachrony. ¹⁵ A reviewer pointed out that given our systematization, the cartographic ‘One feature—one projection’ principle must be violable and interacts with other principles whose effects may emerge diachronically: as a result, a head can be specified for more than one feature. Although this issue deserves a proper discussion, we here note that not only heads, but more generally lexical items can be specified for more than one feature. Featurally complex heads and phrases, arising via movement, are in fact contemplated in the cartographic model (see Cinque and Rizzi 2009: 50). Rizzi (1997: 312–15) suggests that under specific circumstances, the features of a head can be transferred to other heads. One case is represented by the fusion of Force0 and Fin0 heads when topic and focus elements do appear in the CP layer, thereby resulting in a single head carrying features of both force and finiteness. The same logic applies to our case study: the wh-pronoun is at the end of the derivation specified for all the features in Table 5.4.

106

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

heads (e.g. verbs to modals). Notice, however, that the notion of Grammaticalization has primarily evolved and was first proposed to capture the change that concerns lexical elements which progressively assume a functional value, and not the evolution of elements that are per se already functional, as in the case of relativizers. In the formal approach to syntactic variation, the notion of Grammaticalization has been extended to elements that are already functional and get reduced in terms of the morphological features they express but appear at the same time to extend to more contexts. In order to explain this type of change from functional to functional, van Gelderen (2004, 2009a, 2009b) proposes economy principles, which in her account explain several changes. The author classifies the change inside the relativizer system as an instance of change from a specifier into a head, which is the result of the economy principle in (21) (van Gelderen 2004: 11). (21)

Head Preference Principle: Be a head, rather than a phrase.

This has as a consequence that whenever possible, that is, in the absence of evidence of the opposite, the speaker will always project a head and not a specifier. The problem here is precisely what this ‘evidence’ is. Take the element che, which in van Gelderen’s account needs to be ambiguous between a head (i.e. a complementizer) and a specifier (i.e. an interrogative wh-pronoun): according to which evidence does the speaker treat it as a head or as a specifier in the various constructions? From the point of view of the lexicon, it would instead be more economic to say that che is always a specifier, both in interrogatives and in relatives (and even in declaratives). The evidence on the basis of which the speaker decides whether a given element is a specifier or a head clearly cannot be the number of features the element expresses, as che has the same form when it is treated as a complementizer and when it is treated as an interrogative pronoun. While in French there is empirical evidence that interrogative que behaves as a clitic, and this has been proposed already in Bouchard and Hirschbu¨hler (1986), this evidence does not extend to the varieties of Italian which use che as the interrogative pronoun meaning ‘what’. Hence, adopting the principle in (21) in order to explain the change inside the relativizer system and assume that che used to be a pronoun and is now a complementizer is not supported by any empirical evidence for Italian. Assuming that the loss of agreement features in relativizers directly corresponds to a loss of structure is not empirically justified, hence, we will simply propose that the only change in the relativizer system is in the number of features that the element expresses and not in its categorial status.

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

107

Another general aspect that is problematic with the Head Preference Principle concerns its concrete reality as a preference principle. Van Gelderen (2009a, 2011b) claims that the Head Preference Principle is at work in acquisition, and in the acquisition of relativizers, in particular. English speakers, both adults and children, prefer a that complementizer over a wh-pronoun in relative clauses (e.g. Montgomery and Bailey 1991; Diessel 2004). This is expected under the Head Preference Principle, since that is in the head C but the whpronoun is in the specifier position. Thus, the author concludes that ‘the acquisition data shows that children avoid phrases completely’ (van Gelderen 2009b: 101). As follows, we would like to reconsider her evidence and rethink her conclusion. Looking more closely at the acquisition data, again it is not an obvious conclusion that children prefer heads over specifiers. Rather, we argue that the data used by van Gelderen provide evidence only for a preference towards elements that are morphologically unmarked and that carry one single morpho/semantic feature, but which can structurally still be specifiers. The earlier and more frequent production of that over who/which in children should be captured in terms of feature specification: (a) that is unmarked for case contrary to who/whom—and children deal problematically with case marking, and (b) that has only one semantic feature contrary to who/which which are discourse-linked elements marked for human feature—and children acquire different features at different stages. Accounting for the preference of that over who/which in terms of feature complexity is more accurate than the Head Preference Principle not only because both that and who/which are specifiers as clearly demonstrated by Kayne (2010) but also because it fits the acquisition processes more properly. Indeed it is well known from other phenomena that children deal problematically with overt morphological agreement, case marking, and they start with and opt for the unmarked version of the inflected elements in both the verbal and the nominal domain, if available (‘underspecified elements’ in the terms of Clahsen et al. 1996; Hyams 1996; but see Marinis 2004 for a critique of the label ‘underspecified’). In the verbal domain, children—from the beginning of their linguistic production until about the middle of their third year of life—allow infinitives in declarative clauses, contrary to the adult grammar (English, French, German, Swedish, Danish; Poeppel and Wexler 1993, Wexler 1994; Rizzi 1994, 2000; Hyams 2012; among many others).¹⁶ ¹⁶ The extensive literature on the topic convincingly shows that this is not just a morphological problem (cf. Rizzi 2000 and references cited there). For the rarity of root infinitives in pro drop languages, we refer to Rizzi (2000: ch. 12).

108

(22)

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

Michel dormir là Michel sleep there ‘Michel sleeps there.’

(French, Philippe 2;2: from Rizzi 2000: 13)

As for the nominal domain, it has been shown that determiners, both within the nominal phrase as well as when used pronominally, most frequently occur as neutralized invariant forms without case, number, and gender distinction in children until the age of c. 2;5 (German: Clahsen et al. 1994; Italian: Bottari et al. 1998; Greek: Marinis 2004; among others). (23)

Wo’s de lala where’s the.RED pacifier ‘Where is the pacifier?’ (German, Simone 1;10: from Clahsen et al. 1994: 99)

Far from saying that the presence of overtly uninflected forms in child language is only a morphological issue, we limit our discussion to highlight that uninflected forms are the ‘starting point’, which of course has a syntactic ground but we do not have the space to address this here (e.g. truncation hypothesis, cf. Rizzi 1994, 2000; underspecification in the functional domain, cf. Hyams 1996; among others). What is clear is that elements with a complex feature specification are mastered later than elements specified for fewer/single features. This observation is well captured in the proposal that the acquisition process is incremental (Clahsen et al. 1996; Hyams 1996; Penner and Weissenborn 1996; Roeper 1999; among many others): not all the features are in place at the same time and the acquisition of each feature may depend on the status of that feature in a given language and on the hierarchical order of such features (see Belletti et al. 2012).¹⁷ Crucially, the acquisition pattern from uninflected to fully fledged specified elements does not necessary imply that elements that have fewer morphological specifications must be heads or that children avoid phrases completely, as van Gelderen claimed. Not only is this implication not necessary, it is also not confirmed empirically when looking at both language acquisition and dialectal variation. An argument that casts doubts on the reality of a preference for heads over specifiers is the acquisition of negation. According to van Gelderen (2009a, ¹⁷ A concrete example of the incremental nature of the language acquisition process is found in Marinis (2004), who investigates the acquisition of the DP features related to definiteness, specificity, case, and number in two longitudinal corpora of five monolingual Greek-speaking children. He showed that grammaticalization of these features does not take place all at once: (a) definite vs. indefinite marking emerges earlier than specific vs. non-specific (in the case of indefinites), (b) case marking on nouns emerges earlier than case marking on definite articles, and (c) in the speech of one child, case marking emerges earlier than number marking.

ITALIAN REL ATIVE COMPLEMENTIZERS AND PRONOUNS

109

2009b, 2015) the change from negative adverb to negation marker results from the application of that principle. Applying the same reasoning she used in the case of pronouns vs. complementizers to acquisition data, we would expect children to treat negation as a head rather than as a specifier, if possible. However, this prediction is not borne out. Indeed, it is fairly uncontroversial that children also first treat negation as adverbial, hence as a specifier and not as a head, in languages where there is evidence in the input for treating it as a head: this finding is reported for various languages, for example German, English, French, Dutch (Pierce 1989; Hoekstra and Jordens 1994; Hamann 1996; Thornton and Tesan 2007; Thornton and Rombough 2014). Along the same lines, we would like to revise the last piece of evidence coming from dialectal variation that van Gelderen (2011b) quotes to support the Head Preference Principle, that is, the behaviour of interrogative wh-elements in Norwegian dialects. She proposes that short wh-items, such as kven ‘who’, are situated in the head as evidenced by the absence of verb movement to C, as in (24), contrary to phrasal wh-elements that are specifiers and hence, verb second appears (25). (24)

Kven du såg Norwegian variety Who you saw ‘Who did you see?’ (from Åfarli 1986: 6, quoted in van Gelderen 2011b)

(25)

Ka for nokka sa dokker Norwegian variety what for something said you ‘What kind of thing did you say?’ (from Taraldsen 1985: 21, quoted in van Gelderen 2011b)

However, Westergaard and Vangsnes (2005) provide a different account for the difference between (24) and (25). Leaving aside the details of their proposal, the crucial point here is that the authors convincingly demonstrate that short wh-elements are non-projecting elements in the specifier of IntP and not Int0 heads. In conclusion, we interpret the same data van Gelderen uses to support her Head Preference Principle as evidence for a preference towards elements that are unmarked and that carry one single semantic feature, but which are structurally still specifiers. This means that van Gelderen’s Head Preference Principle can simply be seen as an instance of the well-known assumption made in cartographic work, namely the ‘One feature—one projection Principle’ which states that each functional projection expresses a single semantic value either on the head or on the spec (or on both) and the possible values

110

EMANUEL A SANFELICI, JACOPO GARZONIO, AND CECILIA POLETTO

of the feature are only + or −, that is, there is no third value which states an unmarked form.¹⁸

5.5 Conclusion In this chapter we have argued that there is no dichotomy between relative pronouns and complementizers, since in Old Italian varieties there exist ‘agreeing complementizers’ that display features typical of pronouns and uninflected ‘pronouns’ which behave like complementizers. Moreover, the co-occurrence with prepositions is not related to the presence of overt agreement. We have concluded that both ‘complementizers’ and ‘relative pronouns’ are of the same categorial type. What the diachronic path shows in the evolution of Italian varieties is a change in overtly marking the features specified on the relativizer: from a fully specified relativizer to an unmarked one. This has brought us to reconsider the type of change that we observe across the history of Romance. Instead of a change from specifier to head according to the Head Preference Principle, we have argued that it is possible to recast the grammaticalization process of relativizers simply in terms of a loss of morphological/semantic features without necessarily having to admit that the element reduces to a head. This means that relativizers do not change their categorial status through time. What changes is only the feature specification of those items under the ‘One feature—one projection Principle’ adopted in cartography.

Databases ASIt Atlante sintattico d’Italia. University of Padua: . OVI Opera del vocabolario italiano. University of Chicago: .

¹⁸ We tentatively speculate on some predictions that follow from this principle. We may expect speakers, both adults and children, to break the feature bundle specified on an element and assign to each feature a single item, if available in the grammar. The degree of probability according to which we find splitting can depend on various factors, e.g. the number, the type, and the position of features involved. Evidence that the expectation is met comes from the distribution of resumption in relative clauses, where the case of the relativized element is spelled out on the resumptive and the wh-variable on the morphologically unmarked relativizer. Indeed, breaking the feature bundle on at least two items has an amelioration effect in both production and comprehension (e.g. Spanish and English: Pérez-Leroux 1995; French: Guasti and Cardinaletti 2003; Hebrew: Friedmann and Lavi 2006; Italian: Utzeri 2007; Mandarin: Hu, Gavarrò, and Guasti 2015). Although further studies are in order to classify which features can be split on two or more items and to determine the mechanism at work underlying the ‘One feature—one projection Principle’, e.g. syntactic and/or processing, our account seems to be empirically more adequate and theoretically desirable than an application of the Head Preference Principle.

6 Information structure, functional left peripheries, and the history of a Hungarian interrogative marker Julia Bacskai-Atkari

6.1 Introduction The present chapter provides an analysis for the diachronic development of the Hungarian interrogative marker -e in embedded questions. I show that its appearance and grammaticalization in a verb-adjacent position is the direct result of the evolution of functional left peripheries in Old Hungarian (c.9th–16th centuries), and the grammaticalization of marking the interrogative nature of the clause, given here as [Q],¹ at the left edge of a lower functional periphery (located below the CP-periphery). ¹ The feature specification of interrogative clauses—either constituent questions or polar questions—is slightly more complex: while the feature [Q] is responsible for typing the clause as interrogative, there is a [wh] feature that is responsible for marking the scope of the wh-element (see Ginsburg 2009). As pointed out by Bayer (2004), while the overt marking of the two features coincides in several languages (including English and German), there are also many languages where the two are separated, resulting in doubling patterns in constituent questions (this applies to Japanese and to certain non-standard dialects of Dutch). The point is that in languages like English and Hungarian, the wh-element in constituent questions is both [wh] and [Q]. That is, it not only takes scope over the clause as an indefinite but it is also able to type the clause as interrogative overtly; hence, no additional element is needed. In constituent questions, the operator is necessarily overt since it carries non-recoverable information, and it is also focussed. Polar questions also contain a wh-type operator corresponding to whether: this operator turns the proposition into a ‘proto-question’ (Bayer 2004, following Karttunen 1977), and it marks the scope of a covert or (Larson 1985). It is recoverable; therefore, in many cases it is not overt at all, since the type of the clause can be marked by the complementizer as well. Consequently, constituent questions and polar questions differ primarily in the semantic properties of the operator and not in the presence or absence of the morphosyntactic features [Q] and [wh]. In this chapter, I will concentrate on the issue of marking the clause-type interrogative and will not be dealing with the exact role of the operator in polar questions; therefore, I will merely use the [Q] feature in all the representations. (Alternatively, if one has to deal with a single interrogative clause-typing feature, one could naturally adopt [wh] as the general interrogative feature, as is customary in the literature, see for instance Kenesei 1994 and Zimmermann 2013.) Note that ‘Q’ in the glosses merely refers to ‘question particle’ and should not be confused with the clause-typing feature [Q].

Julia Bacskai-Atkari, Information structure, functional left peripheries, and the history of a Hungarian interrogative marker. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Julia Bacskai-Atkari (2022). DOI: 10.1093/oso/9780198871538.003.0006

112

JULIA BACSKAI-ATKARI

The example in (1) illustrates Modern Hungarian embedded yes–no interrogatives: (1)

Azt kérdeztem, (hogy) megérkezett-e that.ACC asked.1SG that PRT.arrived.3SG=Q ‘I asked if Mary had arrived.’

Mari. Mary

As can be seen, the subordinate clause is introduced by the complementizer hogy ‘that’, the overt presence of which is optional (see Kenesei 1994: 341); furthermore, the subclause contains the interrogative marker -e, which is obligatory. By contrast, Old Hungarian embedded yes–no interrogatives show the following configuration (cf. Kenesei 1994: 341):² ¯ (2) mōgadmg nèko̗nc ha te vag xc istènnc fia tell.IMP.2SG.PRT we.DAT if you are Christ God.DAT son.POSS ‘tell us whether thou be the Christ, the Son of God’ (Munich Codex 33va [Matthew 26:63]; from 1466)³ The example in (2) shows, on the one hand, that the presence of the interrogative marker -e is not attested in embedded yes–no interrogatives in Old Hungarian; on the other hand, the complementizer is ha ‘if ’, which shows no alternation with a zero counterpart. Hence, the change from structures like (2) to structures like (1) has two main aspects. First, there is a change in the complementizer from ha ‘if ’ to an optional hogy ‘that’. Second, there is a change in that the element -e appears. Regarding the status of the element -e, it has frequently been analysed as a Focus head, for instance by Van Craenenbroeck and Lipták (2008). This is problematic inasmuch as the element -e cannot be directly tied to the notion of focus since it can appear in sentences not containing a focussed constituent, see (1) above, and thus the generation of a designated FocP is not expected. Another possibility would be to analyse -e as the head of an IntP in the sense of Rizzi (2001): in that case, the position of -e would be appropriately linked to that of other interrogative elements, but at the same time one would be forced to assume that non-interrogative elements (focussed constituents) can move to the specifier of IntP. I am going to propose that -e is the head of a functional projection (FP) on a lower functional periphery: as such, this position is not directly tied to ² The Old Hungarian data are taken from the searchable (normalized) database of the Old Hungarian Concordance, available at . The metadata given in this chapter—date, token numbers—for the individual texts are also taken from there. ³ The English translations of quotes taken from Old Hungarian Bible translations are from the King James Bible.

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

113

either focussing or [Q]; this projection is always generated in interrogative clauses in Hungarian, but its specifier regularly hosts the focussed constituent in non-interrogative clauses as well. In this way, the generation of -e in F and the movement of a focussed constituent to [Spec, FP] are independent from each other. I argue that the necessity of generating the FP in interrogatives is due to the fact that the overt marking of [Q] has grammaticalized on the lower periphery in Hungarian: that is, a [Q] C head passes on its features to the F head, and the [Q] feature is checked off by overt elements in this lower projection.⁴ The relevant changes are parallel with the evolution of functional left peripheries: this can be linked to the major word order change from SOV to ‘Top Foc V X’ (É. Kiss 2013), which was chiefly responsible for the appearance of head-initial functional projections and the establishment of the FP as a landing site for left-peripheral foci. The chapter is structured as follows. Section 6.2 briefly presents interrogatives in Modern Hungarian; Section 6.3 describes the most important diachronic changes affecting embedded yes–no interrogatives between Old and Modern Hungarian, to be followed by a detailed investigation of Old Hungarian interrogatives in Section 6.4. Finally, Section 6.5 presents a theoretical analysis for the change, accommodating it into a cross-linguistic framework on the relation between clause typing and functional left peripheries.

6.2 Clause typing and interrogatives in Modern Hungarian In Modern Hungarian, main clause questions are marked by a distinctive intonation pattern. In addition, in wh-interrogatives a wh-element is present, which is also an overt marker of the [Q] nature of the clause. Consider: (3)

Ki érkezett meg? who arrived.3SG PRT ‘Who has arrived?’

As can be seen, in (3) the wh-element ki ‘who’ appears at the front of the clause. In yes–no interrogatives, the interrogative marker -e appears optionally (see Kenesei 1994: 339–40),⁵ as demonstrated by (4): ⁴ This means that -e is base-generated as an F head and not as a C head, contrary to Kenesei (1994), who assumes that the Q element has to lower from the C to some clause-internal functional projection to adjoin to the verb. In my analysis, no lowering is necessary, which is favourable on general theoretical grounds. On the other hand, I do adopt the idea of Kenesei (1994: 341–3) that the element -e has to be a head. ⁵ See also Gyuris (2017) on the pragmatic differences between clauses with and without -e.

114

JULIA BACSKAI-ATKARI

(4)

Megérkezett(-e) Mari? PRT.arrived.3SG=Q Mary ‘Has Mary arrived?’

The chief difference between main clause questions and embedded questions is that embedded questions do not have a distinctive [Q] intonation pattern in Hungarian, and hence the [Q] nature of the clause has to be marked in a different way. In wh-interrogatives, an overt wh-element is present anyway; in addition, the complementizer hogy ‘that’ may appear optionally. Consider: (5)

Azt kérdeztem, (hogy) ki that.ACC asked.1SG that who ‘I asked who had arrived.’

érkezett arrived.3SG

meg. PRT

In yes–no interrogatives, the appearance of the interrogative marker -e is obligatory, and the complementiser hogy is optionally present: (6)

Azt kérdeztem, (hogy) megérkezett-e that.ACC asked.1SG that PRT.arrived.3SG=Q ‘I asked if Mary had arrived.’

Mari. Mary

The patterns represented by (5) and (6) show a great deal of similarity. It should be noted that the complementizer hogy in Hungarian is a general subordination marker and it is not sensitive to [±Q], see É. Kiss (2002), hence its availability in [Q] environments. Considering all this, the embedded interrogative patterns can be described as follows: the optional complementizer hogy is a subordination marker, and [Q] has to be marked overtly by a different element, which is a wh-phrase in wh-interrogatives and -e in polar questions. Regarding the position of these various elements, hogy as a complementizer is located in a C head; wh-operators, however, are located considerably lower in the structure. Operator movement in Hungarian targets different positions for relative operators and wh-operators (Horvath 1986); relative operators target a [Spec, CP] position (see also Kántor 2008 and Bacskai-Atkari 2014a), while wh-operators target a lower functional projection, namely the ‘focus position’ (see É. Kiss 2002). In sum, Hungarian operator movement differs crucially from languages like English and German, where both relative operators and wh-operators target a [Spec, CP] position. The position of wh-phrases in Hungarian can be easily detected if one considers their similarity with focussed constituents. As is known, focus

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

115

movement in Hungarian targets a preverbal position adjacent to the verb.⁶ Consider the following examples: (7)

a. Mari megérkezett. Mary PRT.arrived.3SG ‘Mary has arrived.’ b. MARI érkezett meg. Mary arrived.3SG PRT ‘It is Mary who has arrived.’ c. Tudom, hogy Mari megérkezett. know.1SG that Mary PRT.arrived.3SG ‘I know that Mary has arrived.’ d. Tudom, hogy MARI érkezett meg. know.1SG that Mary arrived.3SG PRT ‘I know that it is Mary who has arrived.’

The clause in (7a) represents the neutral ‘verbal particle + verb’ order, whereby the verbal particle (meg) is located immediately before the lexical verb (érkezett); the subject (Mari) is in a left-peripheral topic position. By contrast, in (7b) the subject is focussed and therefore it occupies the preverbal position, which thus cannot be filled by the verbal particle at the same time: this results in the inverted ‘verb + verbal particle’ order. The same differences apply to the embedded versions given in (7c) and (7d). The structures for the subclauses in (7c) and (7d) are given in (8) below:⁷ ⁶ Note that not all foci have to undergo leftward movement in Hungarian: in situ foci are also possible. Horvath (1986) identifies exhaustivity as the key property of preverbal foci. The cross-linguistic study of É. Kiss (1998) argues that foci undergoing leftward movement are identificational (and therefore exhaustive; this type is also referred to as structural focus as it has a syntactic reflex and the focussed constituent occupies a scope position), while in situ foci are instances of information focus. For the present investigation, the exact nature of fronted foci is not relevant; what matters is that if a focussed constituent moves to the left periphery, then it lands in a relatively low functional projection and is adjacent to the finite verb, while topics and elements in the CP obligatorily precede such focussed constituents. In the present discussion, the notions ‘focus’ and ‘focussed constituent’ refer to this particular type of focussing. ⁷ The representation in (i) shows the schematic structure of the Hungarian clause: (i)

CP* topics FP* TP PredP VP

I follow É. Kiss (2008a) in assuming that the constituent in [Spec, FP] (her FocP) moves from VP, via moving to [Spec, PredP] and [Spec, TP]; the verb moves along into the respective heads (verb movement occurs generally in finite clauses, not just interrogatives, see also Brody 1990, 1995). See also Surányi (2011) on the relation between TP and focussing; cf. Kenesei (1989), Brody (1990), Surányi (2009) on the role of TP regarding finite verb movement and elements such as verb modifiers appearing in [Spec, TP], showing that [Spec, TP] is not a canonical subject position. Since in this chapter the projections below TP are not relevant, I will consistently use a triangle for TP. The iterable FP constitutes the lower functional periphery immediately above the TP; the iterable CP constitutes the higher periphery. Optional topics may occur in between the two. One might wonder why assuming

116

JULIA BACSKAI-ATKARI

(8)

CP C' C hogy

TopP Top'

Marik Top

FP F'

megi Marii F

TP

érkezettj érkezettj

ti tj tk ti tj meg

The topic projection (given here as TopP) is generated only in (7c) and its main clause counterpart. If there is no preverbal focus, the verbal particle (here: meg) moves to the [Spec, FP] position. If there is a focussed constituent moving to this position, the verbal particle remains in TP. Further, the representation in (8) indicates the position of the FP in the clause with respect to other projections: it is above the TP and below the CP, and while the FP will always be immediately generated above TP, there are various constituents, such as topics, which can move in between the FP and the CP, and which are not themselves related to clause typing.⁸ a lower CP instead of FP is not an option, involving a structure reminiscent of the split CP of Rizzi (1997). This would be problematic for several reasons, of which I would like to mention two here (see Bacskai-Atkari 2018 for details). First, the historical pattern in Hungarian to be discussed here involving the interrogative C ha ‘if ’ co-occurring with a clause-internal -e would be problematic for a single periphery, since the [wh]/[Q] feature should be checked off only once in the CP, and there would be no reason to generate a second projection with the same feature. The problem does not arise if the relevant feature is copied from C to F, as the CP is not an extension of the FP. Second, as shown by Lipták and Zimmermann (2007), a Hungarian clause may host a wh-element clause-internally and a relative operator in the CP, and the wh-operator can be extracted without triggering an island violation effect, indicating that the CP is not a landing site for the wh-element. ⁸ As should be clear, the relative position of the FP is fixed; therefore, preposed foci cannot move higher than topics. This is demonstrated in example (i) below, where the sentence is fine with the first interpretation where Mari ‘Mary’ is focussed but not with the second interpretation where tegnap ‘yesterday’ should be focussed: (i)

Tegnap MARI érkezett meg. yesterday Mary arrived.3SG PRT ‘It was Mary who arrived yesterday.’ # ‘It was yesterday when Mary arrived.’

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

117

As shown by (9), wh-elements target the same position as focussed constituents: (9)

a. *Ki megérkezett? who PRT.arrived.3SG ‘Who has arrived?’ b. Ki érkezett meg? who arrived.3SG PRT ‘Who has arrived?’

The wh-element has to be in the position immediately preceding the lexical verb, as in (9b): a higher position is not permitted, as shown by the ungrammaticality of (9a).⁹ If wh-elements in Hungarian were located in the [Spec, CP] position, (9a) should be well-formed; since this is obviously not the case, it can be concluded that the overt marking of [Q] is associated with a lower functional projection, which is identified here as FP.

In order to derive the second interpretation, the order of tegnap and Mari has to be changed: (ii)

Mari TEGNAP érkezett meg. Mary yesterday arrived.3SG PRT ‘It was yesterday when Mary arrived.’

Naturally, the sentence in (ii) would not be compatible with an interpretation where the subject is focussed. ⁹ Apart from the interaction with the preverbal element, the strictly verb-adjacent position of the wh-phrase can be detected in clauses containing an additional focussed element. Consider: (i)

TEGNAP ki érkezett yesterday who arrived.3SG ‘Who arrived YESTERDAY?’

meg? PRT

(ii)

*Ki TEGNAP érkezett who yesterday arrived.3SG ‘Who arrived YESTERDAY?’

meg? PRT

As can be seen, the wh-constituent must follow the focussed constituent. The data indicate that the FP is iterable; however, if the F head encodes [Q], it must attract an interrogative constituent to its specifier (if there is no interrogative element in the head itself ) to check off this feature, and a non-interrogative focussed constituent cannot do this job. Once the feature is checked off, the FP can naturally project further and be iterated; the necessity of operator movement to the lowest projection can also be observed in the CP domain (cf. Bacskai-Atkari 2016a), and it follows from the Minimal Link Condition (see Fanselow 1990, 1991; Chomsky 1995). The point is that wh-phrases in Hungarian do not merely move below the CP domain, they move specifically to the lowest functional projection above TP.

118

JULIA BACSKAI-ATKARI

The structure of embedded wh-clauses, see (5), is sketched in (10):

(10)

CP C' C[Q]

… FP

(hogy)

F'

kii F[Q]

TP

érkezettj

ti tj meg

As can be seen, [Q] is marked on the F head immediately above the TP, and the specifier of the FP hosts the wh-element (e.g. ki ‘who’); the verb moves up to the F head. In addition, the subordinating complementizer (hogy ‘that’) may be overt in the topmost CP layer; note that other projections may appear in between the CP and the FP (such as topics), as was seen above, but these are not relevant in terms of clause typing. Since the [Spec, FP] position is occupied by the wh-element, the verbal particle (meg) remains in the VP in a postverbal position. Regarding the patterns with the interrogative marker -e, there is either a focussed XP or verum focus. Consider: (11)

a. Azt kérdeztem, (hogy) megérkezett-e that.ACC asked.1SG that PRT.arrived.3SG=Q ‘I asked if Mary had arrived.’

Mari. Mary

b. Azt kérdeztem, (hogy) MARI érkezett-e meg. that.ACC asked.1SG that Mary arrived.3SG=Q PRT ‘I asked if it was Mary who had arrived.’ In (11a), there is verum focus with the neutral ‘verbal particle + verb’ word order, while in (11b) the subject Mari is focussed, and hence the word order is the inverse ‘verb + verbal particle’ order. In both cases, the complementizer hogy is optionally present, and the interrogative marker -e is attached to the lexical verb. The element -e as an interrogative marker occupies the F head position, and the lexical verb moves up to adjoin it: adjunction takes place regularly in that the element moving to a lower position adjoins the higher

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

119

one from the left (see Kayne 1994 for the Linear Correspondence Axiom; and Baker 1985, 1988 for the Mirror Principle). The structure for the subclause in (11a) can thus be represented as follows:

(12)

CP C' …

C[Q]

FP

(hogy) megi

F' F[Q]

TP

érkezettj-e

ti tj Mari

As there is no focussed constituent, the verbal particle can move to the [Spec, FP] position. The C head can be filled by the overt complementizer hogy, just as in (10). The structure for the subclause in (11b) is given in (13):

(13)

CP C' …

C[Q]

FP

(hogy)

F'

Marii F[Q]

TP

érkezettj -e

ti tj meg

Just as in (12), the F head is filled by -e and the lexical verb moves up to it via head movement; the overt complementizer hogy is licensed again. The

120

JULIA BACSKAI-ATKARI

difference from (12) is that the subject DP is focussed and hence it occupies the [Spec, FP] position. Based on the data presented in this section, it can be concluded that in Modern Hungarian, the marking of subordination is linked to the CPperiphery, while the overt marking of [Q] is associated with a lower functional periphery.

6.3 Diachronic changes in Hungarian embedded yes–no interrogatives In this section, I am going to review the typical patterns attested in embedded polar questions throughout the history of Hungarian. The three major periods are as follows: Old Hungarian (c.9th–16th centuries), Middle Hungarian (c.16th–18th centuries), and Modern Hungarian (from the end of the 18th century onwards). As was mentioned before, Old Hungarian embedded polar questions were introduced by the complementizer ha ‘if ’, as illustrated by the examples in (14): ¯ (14) a. Kèrdlec teged az elo̗ istènè hog mōgadmg ask.1SG you.ACC the living God.SUB that tell.IMP.2SG.PRT nèko̗nc ha te vag xc istènnc fia we.DAT if you be.2SG Christ God.DAT son.POSS ‘I adjure thee by the living God, that thou tell us whether thou be the Christ, the Son of God.’ (Munich Codex 33va [Matthew 26:63]; from 1466) b. & a vacnac kezet foguan viue o̗tet and the blind.DAT hand.POSS.ACC holding took.3SG he.ACC az vc´an kuuo̗l Es o̗ zemèibè The street.SUPESS outside and he eyes.POSS.ILL hagapuan o̗ kezeit reia vètuen spitting he hands.POSS.ACC he.SUB putting kèrde o̗tèt ha mit latna asked.3SG he.ACC if what.ACC see.COND.3SG ‘And he took the blind man by the hand, and led him out of the town; and when he had spit on his eyes, and put his hands upon him, he asked him if he saw ought.’ (Munich Codex 44ra [Mark 8:23]; from 1466)

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

121

The Middle Hungarian pattern is demonstrated in (15);¹⁰ as can be seen, Middle Hungarian still used the complementizer ha ‘if ’ but the interrogative marker -e was already present: (15) a. kérdette tu¨lle ha nyughatike asked.3SG (s)he.ABL if rest.POSSIB.3SG.Q ‘(s)he asked him/her whether (s)he could rest’ (Witch Trial 13; from 1724) b. el off eszi eats

hozvan bringing é de Q but

a the a the

vajat Macskával probáltatta ha meg butter.ACC cat.INS tried.3SG if PRT Macska nem is nyúlt hozzá cat not even touched.3SG it.ALL (Witch Trial 1a; from 1736)

Finally, in Modern Hungarian the interrogative marker -e is obligatory in embedded polar questions; the complementizer hogy may be optional, as in (16a), but certain matrix predicates (such as adjectives) may require its overt presence, as in (16b):¹¹ (16) a. Azt kérdeztem, (hogy) Péter szereti-e a lazacot. that.ACC asked.1SG that Peter likes=Q the salmon.ACC ‘I asked if Peter likes salmon.’ b. Kérdéses, *(hogy) Péter szereti-e a lazacot. questionable that Peter likes=Q the salmon.ACC ‘It is questionable whether Peter likes salmon.’ The changes in embedded interrogatives can be summarized as follows: (17)

ha[Q] → ha[Q] + -e[Q] → (hogy) + -e[Q]

In the first stage, [Q] is marked by the complementizer ha; in the second stage, [Q] is presumably marked both by the complementizer and by the interrogative marker -e; finally, [Q] is exclusively marked by -e, as the optional complementizer hogy is insensitive to [±Q]. Hence, there is a major change from overtly encoding [Q] in the CP to overtly encoding [Q] in the FP.

¹⁰ The Middle Hungarian data are taken from the searchable (normalized) database of the Historical Corpus of Private Correspondence (‘To¨rténeti Magánéleti Korpusz’), available at . ¹¹ The factors determining whether hogy has to be overt are independent of the [Q] nature of the clause; see Kenesei (1994) for an analysis of hogy-deletion in Hungarian.

122

JULIA BACSKAI-ATKARI

6.4 The Old Hungarian pattern Regarding Old Hungarian, so far I have discussed only embedded polar questions; in this section I am going to show how these fit into the system of marking [Q] in Old Hungarian. Let us start with main clause questions. In Old Hungarian main clause whinterrogatives, the wh-element is regularly present as the overt marker of [Q]. An example is given in (18) with the wh-element mi ‘what’: (18) Auag mi valtſagot ad ember o̗ lèlkèiert or what exchange.ACC gives human (s)he soul.POSS.FINAL ‘or what shall a man give in exchange for his soul?’ (Munich Codex 23a [Mark 8:37]; from 1466) Main clause polar questions in Old Hungarian very often contained a clauseinitial interrogative marker (such as nemde ‘isn’t it’ and minemde ‘isn’t it’, both containing the negative morpheme nem ‘not’, cf. the invariable question particle innit in non-standard varieties of English), and the clause-final interrogative marker -e: both of these were optional on their own, but whenever -e appeared clause-finally (unambiguously as a C head), the operator was always present and is assumed to have been obligatory in these cases (É. Kiss 2014: 16). Consider: (19) nēde tu˙ incab nagobbac vattoc aʒocnal e˙ Q you rather greater.PL are.3PL those.DAT Q ‘Are ye not much better than they?’ (Munich Codex 12vb [Matthew 6:26]; from 1466) As can be seen, the interrogative marker -e appears in a clause-final position (cf. Kenesei 1994: 341–2): it is not adjacent to the verb, and thus it cannot be located in an FP in the same way it is in Modern Hungarian. Note that as soon as the FP is generated, the finite verb moves up to the F head (see Section 6.2; presumably due to a [fin] feature, see Bacskai-Atkari 2018, 2020); therefore, -e cannot be in the F head without the verb moving there, which rules out the possibility of a head-final FP.¹² ¹² In addition, there are several reasons to believe that there was no head-final FP at all in the history of Hungarian. As shown by É. Kiss (2014), there are various factors indicating that Proto-Hungarian was a language with head-final CP, TP, and VP projections, inherited alongside the basic SOV order from the Proto-Ugric stage. However, by the end of Proto-Hungarian a major word order change took place, which resulted in the reinterpretation of surface SOV orders as ‘Top Foc V X’, and this process involved the emergence of the FP position as a landing site for preposed foci, including the movement of the verb to the F head position. That is, the FP projection arose as a head-initial projection, fitting into the context of the general change from head-final to head-initial functional projections in the language. In cases like (19), then, if one were to pursue an analysis of -e as an F head in a head-final FP, this would be highly problematic for the following reasons. First, the FP appeared in the language

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

123

Turning now to embedded questions, in embedded wh-interrogatives the wh-element is naturally present; in addition, the complementizer hogy ‘that’ may optionally appear, similarly to the Modern Hungarian pattern. The example in (20) shows the overt presence of hogy and the wh-element mi ‘what’: (20) kèrdeʒkeduē / hog mi volna micor halottaibol felkèlend asking that what be.COND.3SG when dead.ELA up.rises ‘questioning what the rising from the dead should mean’ (Munich Codex 44vb [Mark 9:10]; from 1466) Finally, as has already been mentioned, embedded yes–no questions were introduced by the complementizer ha ‘if ’: ¯ (21) mōgadmg nèko̗nc ha te vag xc istènnc fia tell.IMP.2SG.PRT we.DAT if you are Christ God.DAT son.POSS ‘tell us whether thou be the Christ, the Son of God’ (Munich Codex 33va [Matthew 26:63]; from 1466) Regarding the system of marking [Q] in Old Hungarian, there are two important observations to be made here. On the one hand, it should be obvious that the double encoding pattern attested in Modern Hungarian embedded interrogatives was already present in Old Hungarian embedded wh-questions: that is, the marking of subordination and the marking of [Q] are associated with distinct functional projections. On the other hand, the interrogative marker -e was already available in Old Hungarian in main clauses, as the head of a head-final CP, as will be discussed in Section 6.5 in more detail. The changes attested from Old Hungarian to Middle Hungarian thus involve a change in the position of -e (from clause-final C into a clause-internal F), and the appearance of -e in embedded clauses (as opposed to main clauses, where -e was present anyway). Neither of these changes was abrupt: as was shown in Section 6.3, the appearance of -e in embedded clauses started in clauses still introduced by ha, leading to a double overt marking of [Q], and the marker in the CP-periphery disappeared only after the grammaticalization of marking [Q] in the FP. Similarly, the change in the position of -e in main clauses was presumably not the result of the original C head of a head-final CP directly grammaticalizing into an F head (of a head-initial FP): there was an intermediate stage involving doubling. Consider (example from É. Kiss 2014: 16, ex. (17)): already as a head-initial projection. Second, the analysis of clause-final -e as a C in a head-final CP is well supported by the closest related languages, where its cognates are likewise clause-final interrogative C heads (see É. Kiss 2014). Third, even a head-final FP would require verb movement to the F head, hence the verb should be adjacent to -e.

124

JULIA BACSKAI-ATKARI

(22) Mı´nemde elfelethethı´-e az an´a v kis Q off.forget.POSSIB.3SG=Q the mother she small germo̗ket-e child.POSS.ACC=Q ‘Can the mother forget her small child?’ (Nádor Codex 26r; from 1508) In this case, there is one -e in the F head, resulting in -e appearing right next to the finite verb, and there is also a second -e in a clause-final position, in the C head of a head-final CP. In such constructions, the [Q] property is marked twice overtly: both [Q] functional heads, the C and the F, are lexicalized. The grammaticalization of [Q] in the FP contributed to the ultimate disappearance of clause-final -e, and cases like (22) were crucial in terms of reanalysis in that the shift from the C head as the primary marker into the F head as the primary marker could take place in a gradual fashion. On the other hand, ambiguous constructions (verb-final TPs and elliptical clauses) were also crucial in terms of reanalysis (see the discussion in Section 6.5). In these instances, it is impossible to tell from the surface patterns whether -e is underlyingly a C or an F head, and the lack of environments where -e is unambiguously the head of a head-final CP contributed to the gradual disappearance of the corresponding structures.

6.5 Clause typing and functional left peripheries In this section, I am going to present my analysis for the diachronic changes attested in Hungarian polar interrogatives. Section 6.5.1 briefly describes the relevant theoretical background; Section 6.5.2 presents the results of a corpus study of Old Hungarian texts; finally, Section 6.5.3 provides an analysis based on the empirical data.

6.5.1 The theoretical background In order to provide a theoretical account for the changes described above, let us first overview some basic assumptions concerning the marking of [Q] and of subordination cross-linguistically. A fairly traditional assumption is that clause typing and the marking of subordination alike are associated with the CP-periphery (Rizzi 1997); with respect to overt marking, however, there is considerable parametric variation. As far as embedded interrogatives are

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

125

concerned, there are two major types of encoding these properties: single encoding and double encoding. In single encoding, there is one element responsible for marking clause type and subordination overtly. This can be observed, for instance, in the case of German ob ‘if ’ (or English if ): (23)

Ich weiß nicht, ob er kommt. I know.1SG not if he comes ‘I don’t know if he is coming.’

In double encoding, there are different elements responsible for the overt marking of clause type and of subordination. Note also that the element overtly marking the type of the clause may also be an operator (e.g. a wh-operator or a relative operator). This can be attested in embedded wh-interrogatives in certain German dialects (see Weiß 2013: 777–8), and similar patterns can be observed in many non-standard varieties of English, and also Middle English. The example in (24) shows the co-occurrence of the wh-element wer ‘who’ and the complementizer dass ‘that’ in Bavarian (original example from Weiß 2013: 778, ex. (15a), written here in Standard German spelling and with English glosses): (24)

Ich weiß auch nicht, wer dass I know.1SG too not who that ‘I don’t know who was there either.’

da gewesen there been

ist. is

The example in (24) is similar to Modern Hungarian in that the wh-element co-occurs with a general subordination marker. The chief difference is that while in German they are located in the same CP projection (the wh-element in the specifier, the complementizer dass in the head, resulting in a doubly filled COMP),¹³ in Hungarian the two elements are associated with different projections, moreover, with different functional layers. Considering the question of single vs. double encoding throughout the history of Hungarian, the following can be observed in embedded interrogatives. In wh-interrogatives, double encoding is attested in all periods: an optional ¹³ Patterns like (24) challenge the universal applicability of the so-called ‘Doubly Filled COMP Filter’ (DFCF) of Chomsky and Lasnik (1977). There have been several attempts in the literature to analyse constructions like (24) in a way that avoids the violation of the DFCF, including minimalist approaches such as Baltin (2010). However, there are a number of problems with these approaches. In particular, they are unable to handle certain asymmetries attested in several dialects (see Bayer and Brandner 2008 on these), and they are unable to exclude the possibility of doubling structures in polar interrogatives in the same dialects in a principled way (see Bacskai-Atkari 2016a, 2016b, 2020 on these problems and their resolution in a feature-based approach). On the other hand, the very notion of the DFCF is highly problematic as its application domain within a single language should also be restricted (see Bacskai-Atkari 2016a, 2016b, 2020), and at best it can be thought of as the reflex of general economy principles. In short, the analysis of (24) as involving a single CP is favourable both on theoretical and empirical grounds.

126

JULIA BACSKAI-ATKARI

overt C head (hogy ‘that’) marks subordination, and the marking of [Q] is carried by the wh-element itself. Double encoding can also be observed in yes–no interrogatives in Modern Hungarian: the optional overt C head (hogy) marks subordination, and the [Q] property is marked by -e. Single encoding was present in Old Hungarian yes–no interrogatives: the C head ha ‘if ’ marked both subordination and [Q], similarly to ob ‘if ’ in German. Middle Hungarian thus represents an intermediate stage in the transition from single to double encoding. Let us now examine how this is related to functional left peripheries. In embedded interrogatives, the overt marking of subordination is associated with the CP-periphery in all periods. The overt marking of [Q] is linked to the FP in Modern Hungarian entirely: this is the result of the evolution of a lower functional periphery during Old (and partially Middle) Hungarian. Initially, however, [Q] was marked by ha ‘if ’ in the CP domain, and the clause-final -e was a head of a head-final CP, hence this element was also initially associated with the CP layer. As discussed in Sections 6.3 and 6.4, there is evidence for -e as the head of a head-final CP in Old Hungarian, the period which also saw the appearance of the clause-internal -e as an F head (see also the arguments against a head-final FP in the same sections). The changes are attested not only merely within the period of late Old Hungarian, but potentially also within a single text. Before turning to the results of my corpus study, let me first briefly highlight what the exact structures are, and which surface patterns reflect them. Consider first the minimal tree representations in (25) below:

(25) a.

CP Op.

b. C'

CP

(Op.)

TP

C

(…) V (…)

-e

C' C

FP (…)

F' F

TP

Vi-e

ti (…)

The trees are minimal in the sense that they contain only the projections obligatorily present in the relevant structures: a CP and a TP in (25a), and an additional FP in (25b); of course, as was discussed in Section 6.2, there are

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

127

other possible projections between the CP and the FP, but these do not matter here. The denotation ‘(…)’ refers to optionally realized additional phonologically visible material that may appear in the constructions. In both cases, an operator may appear in [Spec, CP]: this is optional in (25b), while it is assumed to have been obligatory in (25a), see the discussion in Section 6.3. However, since there is no direct evidence for the ungrammaticality of clauses with a clause-final -e and with no overt operator, I will not consider the absence of the overt operator to be a diagnostic of identifying structures like (25b). In (25a), -e is a clause-final C head and it cliticizes onto the last element in the TP: this may be the verb itself, but it may be another constituent following the verb as well. In (25b), -e is a clause-internal F head and it cliticizes onto the verb. Essentially the following four surface strings have to be considered: (26)

a. b. c. d.

(…) V … -e (…) V -e … (…) V -e … -e

Again, ‘…’ denotes overtly realized material that is not the finite verb (V ) or the interrogative marker (-e): if this is given in brackets, it indicates that the presence or absence of such a string is not relevant for the properties of the given type in (26). The presence or absence of this string is crucial after the verb since it determines whether the verb and the interrogative marker are syntactically adjacent or not. By contrast, it is usually immaterial before the verb, unless there is no overt verb of course (in which case some other element is obligatory, since the interrogative marker is a clitic that has to be attached to some phonologically visible element). The surface pattern in (26a) indicates an unambiguous head-final C status for -e, which is clearly not adjacent to the verb: this pattern is excluded in Modern Hungarian, while all the other three are permitted. This pattern can only arise in structures like (25a), since in this case there is no F head to hold both the verb and -e. The surface pattern in (26b) indicates an unambiguous head-initial F status for -e, which cannot be clause-final since there is material following it. This pattern can only arise in constructions like (25b) since the TP would not follow -e otherwise. Note that there is also no reason to believe that the string following -e would be extraposed either: there are several examples where the material following the verb cannot be assumed to be extraposed, as illustrated in (27) below. Such instances include bare nouns, as in (27a), pronouns, as in (27b), and verbal particles, as in (27c). Consider:

128

JULIA BACSKAI-ATKARI

(27)

a. Wagene bor is.Q wine ‘Is there wine?’ (Hungarian Manual of the Scribe János Rotenburg 62b; from 1418–22) b. Minēdè hirtèlèn tamadnac e˙ fel Q suddenly arise.3PL Q up ‘Shall they not rise up suddenly?’ (Vienna Codex 272 [Habakkuk 2:7]; from the middle of the 15th century) c. zẃz vagÿe teh virgin be.2SG.Q you ‘Are you a virgin? (Booklet on the Dignity of the Apostles 18v; from 1521)

Turning back to (26), the pattern in (26c) is ambiguous. On the one hand, it is possible that the interrogative marker would be attached to the verb anyway, corresponding to the structure given in (25b). On the other hand, it is also possible that the adjacency of the verb and the interrogative marker is the result of the verb being the clause-final element in a construction like (25a), where the string following the verb (V ) is optional. Similarly, the pattern in (26d) is ambiguous: these are cases where the verb is not overt (mostly elliptical constructions but note that the third person copulas are also zero in the present tense). On the one hand, it is possible that the interrogative marker would attach to the verb in the full structure and cliticizes to the material in [Spec, FP] only as a last resort, in line with the representation in (25b). On the other hand, it is also possible that it would be clause-final as in (25a), where the presence or the absence of V does not change the fact that -e attaches to the last phonologically visible element in TP.

6.5.2 The data Let me now turn to the results of my corpus study. I conducted a corpus search based on the normalized part of the Old Hungarian Concordance corpus (the token numbers given in this section always refer to the normalized version without punctuation marks). As of 14 September 2016, the search gave a total of 300 hits for the interrogative marker in eight texts (many other texts in the normalized corpus seem to contain no instances of -e, but this is in line with

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

129

the fact that -e was only used in main clauses in Old Hungarian, and it was also optional). Note also that the corpus is restricted to Old Hungarian only. The highest number of examples is attested in the Munich Codex (1466; token number: 63,990), and all the possible surface patterns are attested, as given in Table 6.1. Table 6.1 Patterns of -e in the Munich Codex Surface pattern

Underlying structure

Number of occurrences

(…) V … -e (…) V -e … (…) V -e … -e TOTAL

unambiguous → head-final C unambiguous → (head-initial) F ambiguous → head-final C or (head-initial) F ambiguous → head-final C or (head-initial) F

14 (8.97%) 65 (41.67%) 61 16 156

The Vienna Codex (middle of the 15th century; token number: 55,346) also has a high number of examples, and again all the possible surface patterns are attested, as given in Table 6.2. Table 6.2 Patterns of -e in the Vienna Codex Surface pattern

Underlying structure

Number of occurrences

(…) V … -e (…) V -e … (…) V -e … -e TOTAL

unambiguous → head-final C unambiguous → (head-initial) F ambiguous → head-final C or (head-initial) F ambiguous → head-final C or (head-initial) F

7 (10.61%) 32 (48.48%) 18 9 66

As shown by Tables 6.1 and 6.2, the number of unambiguous (head-initial) F occurrences for -e is quite high, and amounts to slightly more than 40 per cent of all the cases attested in the Munich Codex and to almost 50 per cent in the Vienna Codex. This suggests that -e by the middle of the 15th century was predominantly used as the F head (contrary to the assumption of Kenesei 1994: 342). Still, the previous stage was co-present for a while, as also indicated by the relatively high number (amounting to around 10 per cent) of -e as a clause-final C head. Apart from the Munich Codex and the Vienna Codex, there was only one text in the normalized corpus that contained instances both of unambiguous

130

JULIA BACSKAI-ATKARI

final C and of F: this is the Guary Codex (before 1508; token number: 20,126). The data are given in Table 6.3. Table 6.3 Patterns of -e in the Guary Codex Surface pattern

Underlying structure

Number of occurrences

(…) V … -e (…) V -e … (…) V -e … -e TOTAL

unambiguous → head-final C unambiguous → (head-initial) F ambiguous → head-final C or (head-initial) F ambiguous → head-final C or (head-initial) F

1 (14.29%) 1 (14.29%) 3 2 7

Since the total number of occurrences is low, there is not much to say about this particular data set; however, it can be concluded that the appearance of both patterns is not restricted to the Munich Codex and the Vienna Codex (the two, together with the Apor Codex, are frequently referred to as the ‘Hussite Bible’ in the literature, indicating a historical and linguistic relatedness). The remaining texts in the normalized corpus do not show any instances of unambiguous head-final C uses of -e: the texts contain unambiguous F uses and a number of ambiguous patterns, hence the absence of clear cases of headfinal C does not necessarily mean that this pattern was excluded from the given texts. The results for the Booklet on the Dignity of the Apostles (‘Ko¨nyvecse’, from 1521; token number: 9757) and the Jókai Codex (around 1140; token number: 23,194) are given in Table 6.4. Table 6.4 Patterns of -e in the Booklet (‘Ko¨nyvecse’) and the Jókai Codex Surface pattern

Underlying structure

Booklet

Jókai Codex

(…) V -e … (…) V -e

unambiguous → (head-initial) F ambiguous → head-final C or (head-initial) F ambiguous → head-final C or (head-initial) F

23 (53.49%) 9

3 (37.50%) 3

11

2

43

8

… -e TOTAL

The results for the Kazinczy Codex (1526–41; token number: 20,027) and the Miskolc Fragment (1525; token number: 886) are given in Table 6.5.

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

131

Table 6.5 Patterns of -e in the Kazinczy Codex and the Miskolc Fragment Surface pattern

Underlying structure

Kazinczy Codex

Miskolc Fragment

(…) V -e … (…) V -e

unambiguous → (head-initial) F ambiguous → head-final C or (head-initial) F

2 (28.57%) 5

1 (50.00%) 1

7

2

TOTAL

Finally, there are texts that contain instances of unambiguous F head uses of -e only: these are the Bod Codex (first half of the 16th century; token number: 8754) and the Hungarian Manual of the Scribe János Rotenburg (1418–22; token number: 95). The results are summarized in Table 6.6. Table 6.6 Patterns of -e in the Bod Codex and the Hungarian Manual Surface pattern

Underlying structure

Bod Codex

Hungarian Manual

(…) V -e … = TOTAL

unambiguous → (head-initial) F

6

3

It seems that the head-final C use of -e was quite infrequent in Old Hungarian, and the text that contains the most instances thereof (in the normalized corpus) is considerably earlier than the other ones. Note that texts outside the scope of the searchable normalized corpus may contain relevant examples: the Apor Codex (end of the 15th century to the beginning of the 16th century) can be mentioned, which is standardly grouped together with the Munich Codex and the Vienna Codex, and the Nádor Codex (1508) also belongs here, see the examples given by É. Kiss (2014: 16). Of course, the development in this period involves the co-existence of both major patterns for a while and thus the process cannot be modelled in a purely linear fashion. Based on the approximate dates of the individual texts given above, the chronological order is reflected in Table 6.7 below. In order to represent the differences more clearly, this table includes only the two unambiguous patterns under scrutiny, while the two ambiguous patterns (discussed above) are subsumed as ambiguous. The differences are visualized in Figure 6.1. As can be seen, the development shows considerable variation in this period and a substantial portion of the available data contains ambiguous patterns. The percentage of unambiguous head-final C patterns is very low and occurs

132

JULIA BACSKAI-ATKARI

Table 6.7 The distribution of unambiguous patterns Text

Unambiguous C (%)

Unambiguous F (%)

Ambiguous (%)

Jókai Codex Hungarian Manual Vienna Codex Munich Codex Guary Codex Booklet Miskolc Fragment Bod Codex Kazinczy Codex

0.00 0.00 10.61 8.97 14.29 0.00 0.00 0.00 0.00

37.50 100.00 48.48 41.67 14.29 53.49 50.00 100.00 28.57

62.50 0.00 40.91 49.36 71.42 46.51 50.00 0.00 71.43

100,00% 90,00% 80,00% 70,00% 60,00% 50,00% 40,00% 30,00% 20,00% 10,00% 0,00%

Jókai Hungarian Vienna Codex Manual Codex

Munich Codex

Unambiguous C

Guary Codex

Booklet

Unambiguous F

Miskolc Fragment

Bod Codex

Kazinczy Codex

Ambiguous

Fig. 6.1 The distribution of unambiguous patterns

only in texts dating before 1508. It seems to be a fair conclusion that late Old Hungarian is precisely the period when the end of the change from head-final C into head-initial F can be observed.

6.5.3 Discussion The question arises why the change took place. In Old Hungarian, a major typological change in word order can be observed from Proto-Hungarian SOV

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

133

to ‘Top Foc V X’ (see É. Kiss 2013). The change can be partially observed within Old Hungarian, too. On the one hand, this involved a change from predominantly head-final to predominantly head-initial projections. On the other hand, an increased preference of finite over non-finite subordination can be detected (see Bacskai-Atkari and Dékány 2014), which was also crucial in the evolution and reinforcement of functional left peripheries (in terms of both the CP and the FP). The effects of these major changes can also be detected in embedded yes– no questions. The main triggers of the change from marking by ha ‘if ’ to the combination of hogy ‘that’ + -e are hence the following. First, an increased role of the general finite subordination marker (hogy ‘that’) can be observed during Old Hungarian, in line with the increased role of finite subordination and the evolution and strengthening of the (head-initial) CP-periphery (see Bacskai-Atkari 2012, 2014a, 2014b). This was a crucial factor in the appearance of hogy in embedded [Q] clauses, first in wh-interrogatives and later in yes–no questions. Second, the overt marking of [Q] grammaticalized on the lower functional periphery (FP): in this respect, an analogy from wh-questions into yes–no questions can be observed. This caused the reanalysis of -e into an F head: as described at the end of Section 6.4, it is supposed that -e was reinterpreted from a head-final C into a head-initial F head in two facilitating environments: doubling constructions and elliptical constructions. The change happened in parallel with the evolution of the lower functional periphery. Doubling is illustrated in (22), repeated here as (28) for the sake of convenience (É. Kiss 2014: 16, ex. (17)): (28)

Mı´nemde elfelethethı´-e az an´a v Q off.forget.POSSIB.3SG=Q the mother she kis germo̗ket-e small child.POSS.ACC=Q ‘Can the mother forget her small child?’ (Nádor Codex 26r; from 1508)

As can be seen, one -e is attached to the verb, and another one appears clausefinally, that is, as a head of a head-final C; furthermore, the interrogative marker minemde ‘isn’t it’ appears clause-initially, just as nemde ‘isn’t it’ did, see Section 6.4.¹⁴ ¹⁴ At this point, one might wonder whether the reanalysis from C to F constitutes a counterexample for the upward grammaticalization theories of Roberts and Roussou (2003) and van Gelderen (2004, 2009a) as the element -e clearly changes to a lower position. In my view, this is not necessarily the case.

134

JULIA BACSKAI-ATKARI

Third, the changes in the expression of focus also contributed to the change observed in embedded interrogatives. The major typological change from SOV to ‘Top Foc V X’ resulted in the establishment of a preverbal focus position: sentential stress regularly falls on the left edge of the relevant Intonational Phrase, which is either the verb, or the element located in the immediate preverbal position, as described by Szendrői (2001). However, the [Spec, FP] position can also be occupied by elements other than focus: for instance, by negatives or by verbal particles (see É. Kiss 2008b), thus there is no reason to believe that this position is directly linked to the notion of focus. The element -e is the head of this FP, and as it is clearly not a focus marker, it should be treated as an interrogative marker in polar questions.

6.6 Conclusion This chapter has examined the diachronic changes in embedded yes–no interrogatives in Hungarian, with particular attention paid to the typological differences between Old Hungarian and Modern Hungarian. While in Old Hungarian, embedded yes–no interrogatives were marked by the complementizer ha ‘if ’, in Modern Hungarian they are marked by an optional complementizer hogy ‘that’ for subordination and by the interrogative element -e for [Q]. It was shown that embedded polar questions in Hungarian underwent a change from single encoding to double encoding, with respect to the overt markers of subordination and the [Q] property. The particular change is also linked to a number of more general factors in the language. Among others, it is related to the issue of clause typing, and the grammaticalization of [Q] at the lower functional periphery in particular. In It should be kept in mind that the reanalysis of -e does not involve actual grammaticalization in the sense of feature loss (which would predictably be coupled with upward reanalysis, as also discussed by van Gelderen 2004, 2009a). Rather, as demonstrated by examples like (28), there were two candidate positions for the same function (marking interrogative clause type overtly), and the change in the preference towards the lower position (and the ultimate loss of the higher position) presumably went along with the changes in where main stress was assigned. The two positions are functionally equivalent and belong to two different peripheries (note also that clause type is still encoded by C), that is, they are not adjacent projections where a classical pattern of elements climbing upwards during grammaticalization could be expected in the first place. Roberts and Roussou (2003: 207) suggest that while upward reanalysis can change the category, downward reanalysis normally only involves the loss of movement but no categorial change. As shown by Quinn (2009), such a stance is clearly too restrictive as in some cases downward reanalysis involves category change and de-grammaticalization. She argues that this is the case in New Zealand English, where stative have has undergone reanalysis from a Pred into a V head, most probably due to analogy (with canonical possessive constructions). While the de-grammaticalization part does not apply to -e, analogy seems to have played an important role here as well: as discussed previously, the overt marking of the interrogative nature of the clause was already tied to the FP in constituent questions.

THE HISTORY OF A HUNGARIAN INTERROGATIVE MARKER

135

addition, the evolution of functional left peripheries also played a crucial role, in that it strengthened the head-initial CP layer, and saw the rise of the FP. Finally, the role of information structure must also be considered: the establishment of the focus position at the FP contributed to the reinforcement of that periphery.

7 The recategorization of modals in English Evidence from adverb placement Eric Haeberli and Tabea Ihsane

7.1 Introduction In Present-Day English (PDE) clauses containing a finite auxiliary, adverbs can occur in a position immediately following the auxiliary or immediately preceding it. This variation is illustrated in (1).¹ (1) a. She will probably read this book. b. She probably will read this book.

(SAuxAdvV ) (SAdvAuxV )

As Huddleston and Pullum (2002: 780) point out, however, ‘the order shown in [1a] is not only possible but quite strongly preferred over that shown in [1b]’. Similarly, Quirk et al. (1985: 494) add the diacritic ‘(?)’ to an example that is comparable to (1b) whereas the equivalent of (1a) is treated as fully acceptable. Several authors have examined the nature of this variation in PDE in some detail (e.g. Jacobson 1975; Granath 2002; Waters 2011, 2013). However, very little work has focussed on the diachronic development. The main exception is Jacobson (1981), but his study is based on an extremely limited corpus. There are at least two reasons why the diachrony of the variation in (1) is of interest. First, besides negation, inversion, ellipsis, and emphatic polarity, adverb placement is one of the diagnostics that distinguish main verbs from ¹ Earlier versions of this material were presented at the 8th Days of Swiss Linguistics (University of Zurich), the 16th Diachronic Generative Syntax Conference (DiGS16, Hungarian Academy of Sciences, Budapest), and the 18th International Conference on English Historical Linguistics (ICEHL18, KU Leuven). We thank the audiences at these conferences as well as two anonymous reviewers for their comments and suggestions. This work was supported by the Swiss National Science Foundation under grant no. 143302.

Eric Haeberli and Tabea Ihsane, The recategorization of modals in English. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Eric Haeberli and Tabea Ihsane (2022). DOI: 10.1093/oso/9780198871538.003.0007

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

137

auxiliaries in PDE. Whereas the use of the pre-auxiliary position of adverbs is rather restricted, preverbal placement is very common (She probably read this book). On the other hand, immediate post-auxiliary placement is common, while immediate postverbal placement is restricted as, for example, when the main verb takes a direct object (*She read probably this book). In earlier English, however, these differences did not exist. The emergence of adverb placement as a diagnostic for the auxiliary/main verb distinction may therefore shed new light on the diachronic development towards the modern auxiliary system that characterizes PDE, particularly also the emergence of the distinctive class of modals. A second reason why the diachrony of the variation in (1) is of interest is the fact that it has been maintained for centuries. It therefore provides the basis for a case study in long-term syntactic variation and may shed some light on the question why some cases of syntactic variation lead to change whereas others seem to remain stable. For reasons of space, we will not be able to address the second issue mentioned above but we will return to it in future work. Instead, what we will focus on in this chapter is the first aspect, and more specifically the diachronic development of the distribution of adverbs with respect to modals. The history of modals in English has played a very prominent role in the literature in diachronic generative syntax over the last 40 years (cf. Lightfoot 1979 and much subsequent work). By examining the distribution of modals with respect to adverbs, we will provide new empirical evidence to evaluate proposals that have been made with respect to the historical development of modals in the earlier literature. There is a general consensus that modals underwent a reanalysis at the beginning of the Early Modern English (EModE) period, which changed their categorial status from lexical to functional. Accounts differ with respect to whether this recategorization involved a radical change from V to T or whether this transition was less abrupt in that modals were already categorially distinct in some way as compared to regular main verbs before they became purely functional items. These two options make somewhat different predictions with respect to adverb placement. Presenting quantitative data concerning verb movement and, more specifically, the distribution of finite main verbs and finite modals with respect to adverbs from Old English (OE) to EModE, we will show that, while the behaviour of the two types of elements up to the reanalysis of the modals is not entirely identical, their overall diachronic development is. Given, furthermore, that the observed differences can be accounted for independently of a categorial distinction, we conclude that adverb placement does not provide any evidence against the hypothesis that modals and main verbs are categorially identical in the early history of English. Finally, our data

138

ERIC HAEBERLI AND TABE A IHSANE

will provide new evidence for the dating of the reanalysis of the modals and confirm proposals that have been made in the earlier literature. The structure of this chapter is as follows. In Section 7.2, we will briefly present the main aspects of previous work on the development of modals in the history of English, and we will explore the predictions made by this work with respect to adverb placement. In Section 7.3, we will test these predictions on the basis of quantitative evidence drawn from the syntax of adverbs. As a starting point, we will introduce previous research of ours on the development of adverb placement with respect to main verbs. We will then compare those findings with data involving modals and we will explore the implications for the analysis of the categorial change affecting modals. The comparison of main verbs and modals will also allow us to evaluate earlier proposals concerning the dating of the recategorization of modals in English. Section 7.4 concludes the chapter.

7.2 The recategorization of modals 7.2.1 Earlier work on the history of modals in English Ever since Lightfoot’s (1979) seminal work, the historical development of the English modals has attracted a lot of interest in the literature. In PDE, modals have a wide range of properties that distinguish them from main verbs. They lack non-finite forms, do not take complements (except bare infinitives), have no agreement morphology, cannot be iterated, precede not in negative clauses and the subject in interrogatives, can license VP fronting and VP ellipsis, and can be phonologically reduced. Some of these properties are illustrated in (2) (cf. also e.g. Lightfoot 2006: 27–8). (2) a. Non-finite forms—e.g. perfective Modals: *to have might play Main verbs: to have played b. Complements other than bare infinitives Modals: *He can music. Main verbs: He understands music. c. Iteration Modal + Modal: *He will can understand. Modal + Main verb: He will understand. The ancestors of the PDE modals, that is, what Lightfoot calls the pre-modals, generally do not have distinctive properties. Instead, they behave to a large

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

139

extent like regular main verbs in earlier English. Example (3) shows that the ungrammatical options in (2) can be found in earlier English ((3a)/(3c) from Roberts 1985: 23–4). (3) a. Non-finite forms—e.g. perfective if we had mought conuenyently come togyther ‘if we had conveniently been allowed/able to come together’ (c.1528 St. Thomas More) b. Complements other than bare infinitives I can a noble tale ‘I know a noble tale.’ (c.1386 Chaucer) c. Iteration I shall not konne answere ‘I will not be able to answer.’ (c.1386 Chaucer) The historical development towards a distinctive class of modals has been summarized as follows (cf. Aitchison 1980: 141–2, cited in Denison 1993: 336): (4)

Stage 1. A slow moving away of the pre-modals from other verbs between Old English and the end of the 15th century. Stage 2. A much faster withdrawal of (pre-)modals from other verbs in the 16th century. Stage 3. A continued, slowed-down moving away of (pre-)modals from other verbs after the 16th century.

Various aspects of this development have remained controversial, however, in particular certain empirical details and the exact analysis of the change affecting the grammatical system. With respect to the analysis, two main positions can be distinguished. The first one is to assume that pre-modals are regular main verbs in OE and Middle English (ME), that is, that the ancestors of modals belong to the word category V. The other one is to assume that in OE and ME pre-modals already represent a categorially distinct subgroup of verbs. The first position is argued for most prominently by Lightfoot (1979, 1999, 2006) (but cf. also e.g. Allen 1975; Roberts 1985, 1993; Roberts and Roussou 2003). As pre-modals have non-finite forms, full person-number agreement, and nominal complementation (direct objects), Lightfoot argues that they should be assimilated to lexical verbs. According to him, the transition from OE full verb to PDE modal auxiliary is a syntactic phenomenon. More precisely, he suggests that the syntactic categories of verbs and modals restructure

140

ERIC HAEBERLI AND TABE A IHSANE

radically in EModE so that modals change word category and become elements of the category I, that is, inflectional elements. This change corresponds to Stage 2 in (4) above and it manifests itself mainly in the loss of non-finite forms and the loss of the ability to occur with a direct object. The recategorization of the modals is caused by changes occurring in the ME period (cf. Lightfoot 2006: 30f.). One of these changes is related to the morphology of modal auxiliaries. Modals become the only surviving members of the preterite-present category of verbs, that is, verbs whose present tense forms have past morphology (and therefore no third person singular inflection). Other preterite-present verbs are assimilated to more regular verb classes or drop out of the language altogether. In addition, the past forms of the modals become opaque from a semantic point of view because they no longer indicate past time reference but survive with subjunctive-type meanings (cf. e.g. They should stay at home). This accentuates their morphological distinctiveness further, and it is following these morphological changes that modals are assigned a new category, a development that has been situated in the early 16th century (cf. e.g. Roberts 1993: 310f.; Lightfoot 2006: 31). Aspects of Lightfoot’s original (1979) account have been criticized by various authors (e.g. Warner 1983; Plank 1984; Denison 1990) and several of these criticisms have been taken into account in some of Lightfoot’s later work (e.g. 1999, 2006). Nevertheless, some issues remain controversial. The most substantial re-examination of the development of auxiliaries in the history of English can be found in Warner (1993), where, on the basis of detailed empirical work, the author proposes a lexical account of these changes. Warner’s approach to syntactic categories draws on elements of traditional structural linguistics and on insights from a theory of human categorization developed in psychology. The main idea is that the lexicon is a point of contact between general principles of cognitive organization and generative grammar. Human categorization is taken to involve basic and subordinate categories, with clusters of associated properties which oppose categories. These classes are not homogeneous and do not have clear-cut boundaries. Basic categories are most internally coherent and have most distinctive properties whereas subordinate levels have fewer distinctive properties and are less clearly defined. Warner agrees that OE auxiliaries, modals and non-modals, have to be identified as verbs. This is because they share important properties with the rest of this class, including the complementation characteristics and their morphosyntactic properties. However, Warner also observes that auxiliaries

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

141

already have distinctive characteristics in OE, for example with respect to ellipsis. Furthermore, certain modals (shall, mot) seem to have a restriction to finite use already in OE, which would be another unexpected property of simple lexical verbs. Warner therefore argues that there is already a subordinate ‘auxiliary’ word class in OE with distinctive formal characteristics. In other words, he proposes that pre-modals/auxiliaries should be distinguished from lexical verbs, as a subcategory of verbs, and this well before the 16th century, contrary to what is suggested by Lightfoot. However, the 16th century also plays a central role in Warner’s account, and more specifically a ‘period whose centre is 1500’ (Warner 1993: 211). Although not postulating a ‘cataclysmic’ change of the type proposed by Lightfoot, Warner agrees that there is a certain acceleration in the development of modals during this century in that earlier properties are generalized (e.g. loss of non-finite forms, full correspondence of the preterite-present morphology with the modal group) and new properties emerge (difference between auxiliaries and lexical verbs concerning the placement of adverbs, appearance of tag questions, appearance of clitic forms, appearance of contracted negatives). According to Warner, these developments have the cumulative effect of giving rise to a new basiclevel category of auxiliary. Warner argues that one advantage of his approach compared to Lightfoot’s is that while for Lightfoot the changes that auxiliaries undergo in OE/ME are ‘a chance piling up of irregularities’ (Warner 1993: 218), they can be seen as a movement towards a steadily more coherent and distinct status within his approach. It also accounts for certain differences between modals and lexical verbs before the 16th century. In summary, Lightfoot and Warner agree on a change affecting the categorial status of modals in the 16th century. What distinguishes their analyses are the assumptions concerning the situation before this event. In Lightfoot’s analysis, modals are regular lexical verbs, in Warner’s framework they already form a subcategory among verbs. In principle, these two hypotheses could lead to different predictions with respect to the syntactic development of modals. Since modals and main verbs are not different from a categorial point of view before the 16th century in Lightfoot’s approach, they would be expected to have exactly the same syntactic properties throughout the OE and ME periods. In Warner’s analysis, however, modals are already categorially distinct before the 16th century, and this would make it possible for them to have syntactic properties that differ from those of main verbs. As mentioned above, Warner indeed argues that such properties can already be identified in OE (ellipsis, restriction to finite forms for some modals). Here, we will examine an

142

ERIC HAEBERLI AND TABE A IHSANE

entirely different empirical domain to test the predictions made by the two approaches, namely the behaviour of main verbs and modals with respect to verb movement (V-movement).

7.2.2 Predictions for the development of V-movement The distributional properties of negation and adverbs provide the main diagnostics to determine whether a language has V-movement out of the VP or not (cf. Pollock 1989 among many others). A language has V-movement if verbs occur to the left of negation and certain adverbs, and it does not have such movement if the verb occurs to the right of negation and adverbs.² As often observed in the literature (e.g. Roberts 1985, 1993; Kroch 1989), English undergoes a change in this domain of the grammar during its history. Whereas early English shows clear evidence for movement of finite main verbs, such movement cannot be found in PDE. This contrast is illustrated in (5) with data from late ME (cf. Haeberli and Ihsane 2016: 499). (5) a. þerfor I aske now mercy Therefore I ask now mercy ‘Therefore, I now ask mercy.’ b. Bott I sawe noght synne. But I saw not sin ‘But I did not see sin.’

(VAdv) (CMKEMPE, 141.3272) (VNeg) (CMJULNOR, 60.289)

The ME examples show the word order V{Adv/Neg}O, which is characteristic for V-movement. The same word orders would be ungrammatical in PDE (*I ask now mercy; *I saw not sin). Instead, as the translations in (5) show, the main verb must follow adverbs and negation. This indicates that V-movement is no longer possible in PDE and that this process must have been lost in the history of English. As for the timing of this change, we show in Haeberli and Ihsane (2016) that the loss of V-movement in English is a two-step process. V-movement past adverbs starts declining in the middle of the 15th century and is to a large ² Language-specific properties may then make the picture somewhat more complex. For example, despite the absence of V-movement in PDE, certain adverbs can occur to the right of the verb. This can be argued to be due to an alternative structural option that does not involve V-movement (cf. Section 7.3.1). Similarly, even though a language may show signs of V-movement, it is possible for sentential negation to be found to the left of the verb (e.g. Italian, Hungarian). Such a situation may arise when the negative marker is a head that either blocks V-movement or procliticizes onto the verb (cf. also footnote 3).

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

143

extent lost by the middle of the 16th century while V-movement past negation begins a rather slow decline in the 16th century that is then completed over 200 years later. In Haeberli and Ihsane (2016), we analyse this mismatch between the two contexts in terms of a stepwise decline of V-movement. Assuming an inflectional domain of the type TP-MP-AspP-NegP for the periods of the history of English that are relevant for our purposes, we initially have a situation where finite main verbs move to T. In the first phase of the change, Vmovement past adverbs to a high inflectional head (T and M) starts being lost, but V-movement past negation to Asp remains possible. Then, in the second phase, V-movement to Asp is lost as well, and we obtain a grammar in which V-movement is no longer possible at all. The two approaches to the diachrony of modals in English discussed earlier make somewhat different predictions with respect to the behaviour of modals in connection with the decline of V-movement. Since modals cannot be distinguished categorially from main verbs up to the 16th century in Lightfoot’s analysis, we expect the diagnostics for V-movement not to allow us to detect any differences between the two before the 16th century. Contrasts should then start emerging only after the categorial reanalysis. In Warner’s approach, modals are already distinct at the subordinate level before the 16th century, and, assuming that subordinate-level distinctions may have an impact on the syntax, modals could in principle show distinctive properties with respect to the diagnostics of V-movement. However, such a difference is not absolutely necessary as it could be the case that movement is driven by the basic-level category only (i.e. V ). In that case, we would observe no differences between modals and main verbs before the 16th century, either. The above observations mean that there would be one diachronic Vmovement scenario that can distinguish Lightfoot’s and Warner’s approaches and there is one that cannot. A distinction is not possible if modals and main verbs behave identically with respect to the diagnostics for V-movement up to the 16th century. However, if modals and main verbs behave differently before the 16th century, we would have to conclude that modals and main verbs can already be distinguished by the syntax and that therefore Warner’s proposal is supported. The two main diagnostics for V-movement (negation, adverb) are not equally useful to test the predictions made by Lightfoot’s and Warner’s approaches. Before 1500 there is virtually no variation in the syntax of negation.³ ³ The only major area of variation in the period before the 15th century concerns the status of the preverbal negative marker ne as the frequency of its use declines during the ME period. However, since

144

ERIC HAEBERLI AND TABE A IHSANE

Not systematically follows any finite verbal elements (main verbs as well as premodals).⁴ Furthermore, do-support, which ultimately replaces V-not order, is very rare before the 16th century,⁵ and its earliest uses have been argued to be influenced by pragmatic or semantic factors (cf. e.g. Garrett 1998; Warner 1993: 224). Periphrastic do in negative clauses only starts its rise in the 16th century, that is, when modals are already being recategorized. Hence, the syntax of negation does not allow us to determine whether modals already behave differently from main verbs before 1500. The situation is different for the syntax of adverbs. According to the findings in Haeberli and Ihsane (2016) mentioned earlier in this section, evidence for V-movement past adverbs starts declining from the middle of the 15th century onwards. We can therefore compare the development of main verbs and modals in the short period of about 50 years that leads up to the recategorization of modals. Two possible scenarios are conceivable then. First, movement declines with main verbs but not with modals. Evidence for such a scenario would come from a decrease of SVAdv order at the expense of SAdvV order with main verbs but not with modals. In this case, we would have support for Warner’s approach. Alternatively, modals and main verbs initially develop in parallel with respect to adverb placement, with a split occurring only once the recategorization of the modals has taken place. Such a scenario is compatible with both approaches as they both consider early English modals to be fundamentally verbal in nature until their recategorization. The split would allow us to evaluate the dating of the recategorization that was proposed in the earlier literature on the basis of entirely independent evidence. The aim of Section 7.3 is to examine the historical development of the distribution of main verbs and modals with respect to adverbs in order to determine which of the two scenarios described above corresponds to the situation in the late 15th century.

ne is a proclitic on the verb, it does not allow us to draw any conclusions with respect to the placement of the verb in the syntactic structure, and its status is therefore of no interest for our purposes. ⁴ Occurrences of preverbal not are found with a certain frequency in early ME, but they can be related to residues of head-final structure rather than to the lack of V-movement (cf. Haeberli and Ingham 2007: 18f.). ⁵ In the data examined by Haeberli and Ihsane (2016), there is only one out of 217 negative declarative clauses (0.5%) that contains do in the period 1475–1500. Ellegård (1953) gives a frequency of 4.8% for the same period (33/693). But, as pointed out by Warner (1993: 220f.), this frequency does not show the overall use of periphrastic do but only the use in those texts that have instances of do. Warner estimates that, if all texts examined by Ellegård were included, the frequency would rather be around 3.6%.

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

145

7.3 Main verbs, modals, adverbs: historical development 7.3.1 Main verbs and the distribution of adverbs In Haeberli and Ihsane (2016), we examine the distribution of adverbs with respect to finite main verbs in five parsed corpora: The York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE; Taylor et al. 2003), The PennHelsinki Parsed Corpus of Middle English 2 (PPCME2 (1150–1500); Kroch and Taylor 2000), The Parsed Corpus of Early English Correspondence (PCEEC (c.1410–1695); Taylor et al. 2006), The Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME (1500–1700); Kroch et al. 2004), and The Penn Parsed Corpus of Modern British English (PPCMBE (1700–1910); Kroch et al. 2010). Different contexts such as clauses with objects or clauses with specific adverbs are considered in order to determine the status of V-movement past adverbs over time. Here we will focus on the most general distributional evidence, namely the contrast between SAdvV order as opposed to SVAdv order as illustrated in (6) and (7).⁶ (6) SAdvV a. In this temple he often prayed. b. as smoke kyndly signifieþ fier, ‘as smoke naturally signifies fire,’ (7) SVAdv a. Thou seist wel, You say well ‘You speak well,’

(CMFITZJA, A6R.88; ?1495) (CMWYCSER, 347.2156; c.1400)

(CMAELR4,16.456; a1450)

b. that hyt causyd aftyr many mannys dethe. that it caused after many men’s death ‘that it caused the death of many men afterwards.’ (CMGREGOR, 196.1547; c.1475) c. if I leve hym now, ‘if I leave him now,’ (CMMALORY, 63.2117; a1470) It must be stressed that a quantitative analysis of the contrast in (6) and (7) does not directly reflect the development of V-movement since it is not ⁶ In the two basic word order patterns we distinguish, additional constituents may occur in any position within the clause. Thus, SAdvV includes examples in which the verb is not in absolute final position but is followed by other elements like an object (cf. (6b)) or other adjuncts. Similarly, with SVAdv, an additional constituent such as the object may follow the adverb (7b) or immediately precede it (7c). What is crucial for a clause to be classified as SAdvV or SVAdv is simply the relative order of the three elements, independently of the presence or absence of other constituents.

146

ERIC HAEBERLI AND TABE A IHSANE

the case that one word order corresponds to movement and the other one to the absence of movement. Whereas cases like (6) can be considered as evidence for the lack of V-movement, the order SVAdv is ambiguous. It could involve V-movement, but since cases like (7a) and (7c) remain possible in a language without V-movement like PDE, it is a word order that must be derivable independently as well (i.e. through adjunction of the adverb to the right according to the traditional, non-Kaynian analysis). A comparison of SAdvV and SVAdv orders is nevertheless useful for our purposes as a rise in the frequency of SAdvV can be considered as a symptom of the decline of Vmovement. Quantitative developments and the absence thereof then allow us to identify the beginning and the end of the change.⁷ Furthermore, the SAdvV– SVAdv contrast has the advantage that it can be straightforwardly extended to data involving modals as will be shown in Section 7.3.2. Table 7.1 gives an overview of the development of SAdvV order in the periods that are relevant for our later comparison with modals, that is, from OE to EModE. Included in the data are all affirmative main and subordinate clauses that contain an overt subject, a finite main verb, and a one-word adverb of any semantic type, with the latter two in a position to the right of the subject. For the purposes of diachronic analysis, the data are divided as follows. For OE, the figures for the entire period are given. For ME, we follow the divisions of the PPCME2 up to 1420 (PPCME2 periods M1 to M3). Finally from 1420 onwards (late ME and EModE), we adopt the periodization introduced by Ellegård (1953) in his study of the rise of do-support.⁸ Only texts that can be assigned clearly to one of the periods are included.⁹ SAdvV order undergoes substantial change in the history of English. Its frequency of occurrence is very high in OE. Then, SAdvV gradually declines and reaches a low point in the middle of the 15th century before becoming much more frequent again in the second half of the 15th century and the beginning of the 16th century.

⁷ As discussed in Haeberli and Ihsane (2016), certain word orders allow us to trace the loss of Vmovement in a more precise way because they disappear as a consequence of the loss of V-movement. Relevant contexts are clauses with a non-heavy direct object (loss of SVAdvO order) and clauses with never (loss of SVnever order). The diachronic conclusions derived from these contexts converge to a large extent with those obtained from the SAdvV/SVAdv distinction. ⁸ The only difference compared to Ellegård concerns period 8 (1525–50), which is further subdivided by him into 1525–35 and 1535–50. For our data, there are no grounds to treat this period differently from the others. ⁹ This means that all PPCME2 files with the extensions m23, m24, m34, mx4 are not included. These texts have composition dates and manuscript dates falling in different periods. The data for periods 4 to 12 are taken from Haeberli and Ihsane (2016: 512). The data for the earlier periods are not explicitly mentioned in Haeberli and Ihsane (2016) but were obtained in the context of the same study.

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

147

Table 7.1 The distribution of finite main verbs and adverbs following an overt subject in Old, Middle, and Early Modern English (YCOE, PPCME2, PCEEC, PPCEME) Period

SAdvV

SVAdv

Total

%SAdvV

1. OE 2. 1150–1250 3. 1250–1350 4. 1350–1420 5. 1420–1475 6. 1475–1500 7. 1500–1525 8. 1525–1550 9. 1550–1575 10. 1575–1600 11. 1600–1625 12. 1625–1650

9410 299 25 164 161 123 211 432 370 460 641 466

4000 483 160 1486 1744 622 355 844 690 891 925 705

13,410 782 185 1650 1905 745 566 1276 1060 1351 1566 1171

70.2 38.2 13.5 9.9 8.5 16.5 37.3 33.9 34.9 34.0 40.9 39.8

The high frequency of SAdvV order in OE is partly due to the occurrence of head-final structure (both in main and subordinate clauses but with a higher frequency in subordinate clauses). However, even in clauses that are clearly head-initial, SAdvV order can be found with a certain frequency (28.7 per cent in the data presented by Haeberli and Ihsane 2016: 506). A striking aspect of the latter type of SAdvV order is that in a large majority of these cases (71.7 per cent) we find the adverbs þa or þonne (‘then’), which, apart from their temporal role, have been argued to have the function of discourse particles (van Kemenade and Los 2006). In Haeberli and Ihsane (2016: 506–7), we propose that, although the finite verb can also occur in higher positions in OE main clauses (C and Fin), SAdvV order generally involves V-to-T movement. Under the assumption that TP shows variable directionality (cf. Pintzuk 1999 among others), SAdvV can be derived either through head-final TP and adverb placement anywhere below the subject position, or through head-initial TP, adverb placement in a position above TP, which mainly attracts discourse particles, and subject placement above the adverb.¹⁰ ¹⁰ The assumption that there are two subject positions above the lowest V-movement position is in line with much of the recent literature on OE. As the higher position is typically occupied by subject pronouns, some authors propose that this variation is driven by discourse factors: the high position hosts discourse-given subjects whereas non-given ones occur in the lower subject position (e.g. van Kemenade 2012).

148

ERIC HAEBERLI AND TABE A IHSANE

After the OE period, SAdvV order declines rapidly. According to Haeberli and Ihsane’s (2016) account, various factors contribute to this development. The first important factor is the loss of head-final TP as it eliminates one of the two OE options for deriving SAdvV order. The only remaining option is then weakened by two further changes: the decline of high adverbial discourse particles like þa/þonne, and the decline of the distinction between two subject positions above T. These developments cause ME to move towards a system of the type found in French, where the verb moves to a high inflectional head (T) and the subject occupies the specifier position of that head (SpecTP). SAdvV order therefore becomes less and less common. The low point of this development is reached in the middle of the 15th century with a frequency of 8.5 per cent of SAdvV. Since this frequency is still not entirely negligible at this point, language learners maintain SAdvV in their grammar by postulating a residue of the second OE option mentioned above. The main difference is that the role of the high adverb position is extended as it now also attracts adverbs other than discourse particles. However, given that there is no independent motivation for adverb fronting apart from the fact that it maintains a word order that occurs in the language learner’s input and given that the higher subject position is in decline, the structure deriving SAdvV becomes marginal (cf. Haeberli and Ihsane 2016: 509f., 532 for details). After the middle of the 15th century, the downward trend with respect to SAdvV is inverted, and the frequency of SAdvV rises from 8.5 per cent to 37.3 per cent within about half a century. In the following periods, we can observe some minor variation but the overall effect is stability (37.3 per cent SAdvV for 1500–25, 39.8 per cent SAdvV for 1625–50). As shown in Haeberli and Ihsane (2016), the short phase of increase of SAdvV order at the end of the 15th century signals the loss of V-movement past adverbs. At the end of this phase, the verb still moves to an inflectional head above negation (hence the predominance of SVnot order in negative clauses), but movement across adverbs to the high heads T and M is lost.

7.3.2 Modals and the distribution of adverbs Let us now consider modals. According to Lightfoot’s hypothesis that main verbs and modals are categorically identical in OE and ME, we would expect the two to have the same properties with respect to V-movement and the quantitative development of adverb placement with modals to follow the pattern of main verbs in ME. Furthermore, if the dating of the recategorization

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

149

of modals is correct, the separate developments should start at the beginning of the 16th century. These developments could also be captured by Warner’s analysis under the assumption that it is the basic-level category that determines movement. However, if the diachronic path of adverb placement with modals is already substantially different before 1500, Warner’s approach would be supported as it is only in his analysis that modals can be distinguished syntactically before 1500. 7.3.2.1 Modals and adverbs: quantitative data For our quantitative analysis of adverb placement with modals, we distinguish three word order patterns: SAdvMV, SMAdvV, and SMVAdv.¹¹ These are illustrated in (8) to (10). (8) SAdvMV a. And I always schall be yowre herault And I always shall be your herald ‘And I shall always be your herald.’ (PASTON, I,397.129.3924; 1467) b. as ye hiere may see as you here may see ‘… as you may see here’ (CMREYNAR, 8.66; 1481) (9) SMAdvV a. and that ye shall well vnderstand and that you shall well understand ‘and you shall understand that well’ (PASTON, I,145.039.987; 1465) b. but he wolde nevyr cope whithe no man but he wanted never engage.in.combat with no man ‘but he never wanted to engage in combat with any man’ (CMGREGOR, 219.2122; c.1475) (10) SMVAdv a. they shall dwell there, ‘they shall dwell there’

(PASTON, I,461.144.4497; 1473)

¹¹ Again, these labels refer to the relative order of the four elements, but additional constituents, including a non-finite auxiliary, may occur elsewhere within the clause. Cf. also fn. 6. Covered by our counts are nearly all elements labelled as modals in the parsed corpora such as can, could, may, might, must, mot, shall, should, will, would, as well as modal uses of dare and need. The only exception is ought, which we have left aside due to its distinct syntactic behaviour (selection of a to-infinitive rather than a bare infinitive).

150

ERIC HAEBERLI AND TABE A IHSANE

b. ʒyf þei woldyn receyuen hir a-geyn if they wanted receive her again ‘if they wanted to receive her again’ (CMKEMPE, 65.1474; c.1450) To compare the data involving modals with those presented in Table 7.1 for main verbs, we will determine the frequency of SAdvMV order as opposed to SMAdvV and SMVAdv. The frequencies obtained in Table 7.1 are included in the last column of Table 7.2. If modals and main verbs behave identically with respect to adverb placement, we would expect the frequencies of SAdvMV and SAdvV to behave in similar ways. The most striking aspect of the figures in the last two columns of Table 7.2 is that SAdvMV order is considerably less frequent than SAdvV in all the periods examined. At first sight, this contrast could be interpreted as meaning that from the earliest attested periods of English onwards modals move higher in the structure and therefore remain less often to the right of an adverb than main verbs. This would imply that a categorial difference between modals and main verbs must have existed in OE and ME already as, otherwise, the syntax could not make the necessary distinction with respect to which items are attracted to a higher head and which remain lower. However, such a conclusion is too hasty. Although an analysis in terms of a categorial distinction would provide a simple explanation of the quantitative differences shown in Table 7.2, it is by no means the

Table 7.2 The distribution of finite modals and adverbs following an overt subject in Old, Middle, and Early Modern English (YCOE, PPCME2, PCEEC, PPCEME) Period

SAdvMV

SMAdvV

SMVAdv

Total

%SAdvMV

%SAdvV

1. OE 2. 1150–1250 3. 1250–1350 4. 1350–1420 5. 1420–1475 6. 1475–1500 7. 1500–1525 8. 1525–1550 9. 1550–1575 10. 1575–1600 11. 1600–1625 12. 1625–1650

327 19 2 26 10 14 4 19 28 20 18 21

873 124 34 312 419 159 177 553 453 661 706 686

250 65 40 266 484 185 114 375 316 386 386 475

1450 208 76 604 913 358 295 947 797 1067 1110 1182

22.6 9.1 2.6 4.3 1.1 3.9 1.4 2.0 3.5 1.9 1.6 1.8

70.2 38.2 13.5 9.9 8.5 16.5 37.3 33.9 34.9 34.0 40.9 39.8

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

151

only possible way to account for them. In Section 7.3.2.2 we present some alternatives. 7.3.2.2 Accounting for the frequency differences between modals and main verbs A first observation that is relevant in connection with the contrasts shown in Table 7.2 is that clauses involving modals differ from those involving finite main verbs in an important way with respect to how adverb placement can be analysed. If we assume that modals are of the category V in early English and, furthermore, that non-finite verbs have their own functional structure, we must conclude, as Roberts and Roussou (2003) do, that a sentence like (11a) has a structure as given in (11b) (from Roberts and Roussou 2003: 40).¹² (11) a. Sone hit mæi ilimpen Soon it may happen b. [TP Sone [TP hit mæi [VP tmæi [TP T [VP ilimpen ]]]]] With a biclausal structure like (11b), adverbial modification can occur in two domains, either in the lower clause or in the higher clause. This gives rise to potential structural ambiguities with word orders of the type SMAdvV in early English as shown in (12). (12) a. [TP SU modal AdvP [VP tmodal [TP T [VP V ]]]] b. [TP SU modal [VP tmodal [TP T AdvP [VP V ]]]] In (12a), the order SMAdvV is derived through movement of the modal to T. Absence of movement would lead to SAdvMV. With a structure like (12b), however, SMAdvV is obtained regardless of whether the modal moves to T or whether it does not. The adverb always follows the modal. There are indeed many examples for which (12b) seems to be the most appropriate analysis. (13) a. þt [TP ich mote [VP tmote [TP T þe that I might you ‘that I might truly love you’

trewoliche [VP luuien ]]]] truly love (CMJULIA, 104.150; c.1225)

b. ðat [TP he wile [VP twile [TP T ðar mid ðe [VP wuniʒen]]]] that he wants there with you live ‘that he wants to live with you there’ (CMVICES1, 91.1086; a1225) ¹² The structures in the following examples leave aside additional inflectional heads that may occur between T and V. For our purposes here, the presence or absence of such additional heads is not essential. We will simply assume that the domain between TP and VP is a domain that can host adverbs.

152

ERIC HAEBERLI AND TABE A IHSANE

c. and [TP I schal [VP tschal [TP T ‘and I shall laste dai ]]]] last day’

aʒen [VP again

reyse hym in the raise him on the

(CMNTEST, VI,40.522; c.1388)

d. that [TP it myghte [VP tmyghte [TP T greetly [VP harme me ]]]] ‘that it might greatly harm me’ (CMCTMELI, 232.C2.597; c.1380) In all the examples in (13), as well as in (9a) above, the adverb is best interpreted as modifying the main verb rather than the modal. This suggests adverb attachment in the lower clause in a biclausal structure. A plausible conclusion based on these observations would be that, due to the additional low adverb placement option, the occurrence of an adverb in the inflectional domain of finite modals is less common than the occurrence of an adverb in the inflectional domain of a finite main verb. The lower frequencies of SAdvMV compared to SAdvV could then simply be argued to be a manifestation of this contrast. As there are fewer instances of adverbs in the higher inflectional domain, there are also fewer instances of SAdvMV order. The above considerations suggest that it is not legitimate to compare the frequencies in Table 7.2 directly to draw any conclusions as to the syntactic status of modals and main verbs. In order to do so, we would have to examine data that unambiguously identify movement of the finite element. For finite main verbs, evidence of this type can be found with (non-heavy) direct objects, as the order SVAdvO necessarily involves V-movement past the adverb. SVAdvO order can then be measured against SAdvVO order. As for modals, we would need all clauses in which the adverb modifies the modal and therefore merges in the higher clause. In such clauses, SMAdvV order must be derived, as in (12a), through movement of the modal from V to T in the periods when the modal is of the category V. If we then measured SMAdvV order against SAdvMV, we would have a frequency that should be more directly comparable to the main verb data. The problem with this method is that it does not seem to be possible to obtain exact data for the modals. This is because in many cases it can be difficult or impossible to conclusively determine whether the adverb is indeed in the higher clause or whether a low attachment would be possible as well. Consider for example the clause in (9b) (but he wolde nevyr cope whithe no man). Although our translation suggests that the adverb modifies the modal (but he never wanted to engage in combat with any man), the alternative option cannot be entirely excluded (but he wanted to never engage in combat with any man). Such uncertainties of interpretation make an exact measurement of

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

153

high attachment impossible. Therefore, the method of comparison between modals and main verbs suggested in this paragraph lacks precision as well. We thus conclude that simple comparisons of frequency figures of any kind do not allow us to evaluate whether modals and main verbs have the same syntactic status or not because semantic factors have an impact on the placement of the adverb as well.¹³ Besides the possible effects of the biclausal structure with modals, there is another factor that is likely to play a role in the frequency contrasts between SAdvMV and SAdvV order. As Table 7.2 shows, the contrast is already fully in place in the OE period. SAdvMV is a clear minority option whereas SAdvV is very frequent. A plausible assumption to make then is that what follows in the ME period is simply a natural continuation of the OE situation, with SAdvV being maintained at a higher frequency level than SAdvMV throughout the decline of these word order options. As for the way language learners acquire this contextual distinction, the syntactic environment alone (non-finite complementation with biclausal structure with modals but not (or at least not systematically) with main verbs) could be argued to be sufficient, and a categorial distinction between modals and main verbs would therefore not be necessary. But if the situation in OE is crucial for the subsequent diachronic developments, the question we have to address is why there is such an important quantitative contrast between main verbs and modals with respect to adverb placement in OE in the first place. Although the structural aspects discussed above (additional adjunct position(s) due to biclausal structure) are relevant here as well, there is an additional factor that can be identified specifically for OE. In OE, there is an important difference between the way SAdvV is derived and the way SAdvMV is derived. As pointed out earlier, there are two ¹³ An anonymous reviewer suggests that an analysis of clauses containing verbs other than modals that take an infinitival complement (e.g. control verbs) may shed some further light on the quantitative impact a biclausal structure has on the distribution of adverbs. However, it is not entirely clear what the predictions for such clauses would be. At first sight, one might expect the frequency of preverbal adverb placement to be identical in both biclausal contexts. However, in light of the discussion in the text, the frequencies could also be different if the likelihood with which a modal is semantically modified by an adverb is different from that for other verbs taking an infinitival complement. Given the much more constrained semantics of modals compared to other lexical verbs selecting a non-finite clause, a difference would indeed not be entirely surprising. An analysis of clauses with a non-modal verb taking an infinitival complement in our corpora in the period up to 1500 (i.e. when modals change their categorial status, cf. Section 7.3.2.3) gives the following results: the order SAdvV is systematically less frequent with lexical verbs taking an infinitival complement than with other lexical verbs. This would support the claim that the increase in the number of possible adverb positions with a biclausal structure leads to a reduction of adverb placement before the finite verb. However, the frequencies of SAdvV with lexical verbs are not as low as for SAdvMV with modals. Given the observations made above, this contrast could be related to a higher likelihood of adverbial modification of lexical matrix verbs as compared to modals.

154

ERIC HAEBERLI AND TABE A IHSANE

possible options to obtain SAdvV order in OE: either through head-final structure, a very frequent option in particular in OE subordinate clauses, or through high adverb placement above a head-initial TP, a less frequent option. The latter option is also available for SAdvMV order. Head-final structure, however, generally does not lead to SAdvMV order but to SAdvVM order. SAdvMV is only possible if the non-finite main verb undergoes rightward movement (Verb Raising, Verb Projection Raising). Although such movement can indeed be found in OE, it is a minority option (cf. Haeberli and Pintzuk 2012). Thus, the contribution of head-final structure to the two word orders we are interested in differs considerably. While only a small part of the head-final clauses gives rise to SAdvMV order in Table 7.2, most clauses of this type contribute to the total number of SAdvV orders in Table 7.1.¹⁴ This contrast can then be argued to be, at least partly, responsible for the observed frequency differences between the two contexts. If we add the clearly head-final order SAdvVM (570 cases) to the OE data in Table 7.2, we obtain a considerably higher frequency of adverb placement to the left of the modal. SAdvVM and SAdvMV together correspond to 44.4 per cent of all clauses containing an adverb. This figure is still not quite as high as the frequency of SAdvV order (70.2 per cent). This suggests that adverb placement in the lower clause in a biclausal structure contributes to the frequency contrast in OE as well. Let us consider one final issue in connection with the systematic frequency differences between main verbs and modals in Table 7.2. An additional possibility to account at least partially for these is suggested by Roberts and Roussou (2003: 43ff.). Based on some of the evidence discussed by Warner (1993), they propose that ‘at least some pre-modals may have started “leaking” into the functional domain from much earlier than the sixteenth century’ (Roberts and Roussou 2003: 43). This could be considered as an intermediate position between Warner’s and Lightfoot’s. The systematic recategorization of modals may indeed have occurred in the 16th century, but before that some modals or some modals in certain uses may already have been reanalysed as elements merged directly in the functional domain. Thus, the recategorization would have taken place in a gradual fashion. There is nothing in our data that would clearly go against such a hypothesis, and it could potentially provide an additional element towards the explanation of why SAdvV is considerably more frequent than SAdvMV, if we assume that insertion of a modal in a high position reduces the likelihood for an adverb to occur to its left. However, ¹⁴ Adverb extraposition may sometimes lead to SVAdv order despite underlying head-final structure. We assume here that this option is a marginal one and that in the large majority of cases, the adverb occurs preverbally with a head-final structure.

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

155

Table 7.3 The distribution of may, shall, and will and adverbs between an overt subject and the main verb in Middle English (PPCME2, PPCEME) Period

may

2. 1150–1250 3. 1250–1350 4. 1350–1420 5. 1420–1475 6. 1475–1500

SAMV SMAV 8 28 0 7 8 99 3 67 6 27

shall

will

%SAMV SAMV SMAV %SAMV SAMV 22.2 4 50 7.4 5 0.0 0 20 0.0 1 7.5 14 124 10.1 3 4.3 4 178 2.2 1 18.2 3 55 5.2 4

SMAV 24 3 58 93 51

%SAMV 17.2 25.0 4.9 1.1 7.3

we also do not find any clear support for this proposal. Roberts and Roussou (2003: 43) introduce the hypothesis of ‘leakage’ into the functional domain in connection with Warner’s (1993: 145, 186) observation that the modals shall and mot never seem to have occurred in a non-finite form in OE and ME. A plausible assumption to make then would be that these modals have always been directly merged in T in (12) rather than under V. Assuming that this is correct, one may wonder whether this property had an impact on the distributional properties with respect to adverbs. To examine this question, we will have to focus on shall as mot occurs only rarely in our data. Similar limitations hold for the modals with which we can compare the properties of shall. Only may and will occur with substantial numbers in each of the early periods. In Table 7.3, we compare the numbers of occurrences of SAdvMV and SMAdvV order with the three modals may, shall, and will (and their different forms including might, should, would) in the different periods of ME. At first sight, there do not seem to be any substantial differences between individual modals with respect to adverb placement. A closer statistical analysis confirms this initial impression. If, for each period, we compare may and shall on the one hand and shall and will on the other, we obtain only one statistically significant difference among the ten possible cases. In the period 1150–1250, may has a significantly higher frequency of SAdvMV order than shall (onetailed Fisher’s exact test, p = 0.045). All the other comparisons yield results that are statistically not significant. There is therefore no clear evidence in our data suggesting that shall behaves differently from may and will with respect to adverb placement.¹⁵ This conclusion is confirmed if we collapse all the data ¹⁵ In terms of the analysis proposed in Haeberli and Ihsane (2016), this result would not be unexpected, even if Roberts and Roussou’s hypothesis of some ‘leakage’ into the functional domain before the 16th century were correct. Our proposal is that, until the middle of the 15th century, finite main verbs generally do not occur below the highest inflectional head (T) and preverbal adverbs move to a

156

ERIC HAEBERLI AND TABE A IHSANE

from Table 7.3 for the period 1150–1500. Whereas shall and will pattern almost alike (5.5 per cent SAMV for shall, 5.8 per cent SAMV for will), it is may, rather than shall, that has the highest rate of SAMV order (9.9 per cent). 7.3.2.3 Diachronic developments So far, our discussion of the results in Table 7.2 has focussed on the frequency differences between modals and main verbs, and we have seen that they are not necessarily signs of a different categorial status of the two elements in early English. To conclude our discussion, let us turn to another aspect of the data shown in Table 7.2, namely the diachronic trajectories of the two contexts. Even if there are independent factors that influence the exact frequencies of SAdvV and SAdvMV, one would nevertheless expect the two contexts to develop identically over time if main verbs and modals were of the same syntactic category. On the other hand, if they were categorially distinct, the diachronic developments could be independent of each other. As pointed out in our discussion of Table 7.1, SAdvV order steadily declines from OE to late ME, reaching its lowest point in the period 1420–75. After that, SAdvV order rapidly increases again within approximately half a century before reaching a certain stability. As for SAdvMV order in Table 7.2, it declines after OE and reaches a first low point in the period 1250–1350 with 2.6 per cent. There is then a small rise to 4.3 per cent in 1350–1420, and a decline back again to 1.1 per cent in 1420–75. This pattern can be considered as parallelling that of main verbs. The low point with SAdvMV in the period 1250–1350 is unlikely to be significant for two reasons. On the one hand, the evidence for this period in the PPCME2 is somewhat problematic as it is very restricted and only consists of two texts of a substantial size for the entire period. On the other hand, the rise from period 1250–1350 to period 1350–1420 is not significant (Fisher’s exact test, p > 0.5). This suggests that the period 1250–1350 does not clearly fall outside a downward trajectory that starts after OE and reaches its lowest point in the period 1420–75, the decline from 1350–1420 to 1420–75 being highly significant (chi-square = 16.16, p < 0.0001). The parallelism between main verbs and modals then continues in the period 1475–1500. We can observe a significant increase of SAdvV order (chi-square = 36.35, p < 0.001) as well as a significant increase of SAdvMV (chi-square = 11.00, p < 0.001). As discussed earlier, the increase in SAdvV order is a manifestation of the decline of V-movement. The fact that SAdvMV increases in parallel suggests that the position above TP. Thus, even if some modals were inserted directly in the inflectional domain, they would not be expected to occur higher in the structure and therefore to be more likely to precede adverbs than finite main verbs.

THE RECATEGORIZ ATION OF MODALS IN ENGLISH

157

beginning of the decline of V-movement affects both modals and main verbs and that therefore the two types of elements are not categorially distinguished, at least with respect to this aspect of the syntax. Hence, our data are compatible with both Warner’s and Lightfoot’s approaches, as both treat modals as belonging to the same main syntactic category V up to around 1500. After 1500, main verbs and modals take different diachronic paths. SAdvV order again increases in a highly significant way in the period 1500–25 (chisquare = 73.08, p < 0.001). This can be considered as a consequence of the further decline of V-movement. As for SAdvMV order, its frequency stops rising and declines instead. The decline from period 1475–1500 to 1500–25 is borderline significant (chi-square = 3.94, p = 0.047). But one text sample alone (John Fisher’s sermons) contains three out of the four cases of SAdvMV for the period 1500–25, and if that text were left aside, the decline would be much more clearly significant (chi-square = 8.27, p < 0.005).¹⁶ The weakness of SAdvMV in this period can plausibly be analysed in terms of the categorial reanalysis of modals. As modals are no longer members of the category V, the decline of V-movement does not affect their distribution any more. Instead modals are now merged directly in the inflectional domain, either in a position which is to the left of adverbs or from which they can move to the left of adverbs.¹⁷ Thus, by identifying the beginning of the 16th century as the period when main verbs and modals start having clearly distinct diachronic trajectories with respect to adverb placement, the data in Table 7.2 confirm the dating of the recategorization of modals proposed in the earlier literature on the basis of novel and entirely independent evidence.

7.4 Conclusions The history of English modals has been discussed extensively in the diachronic literature. In this chapter, we have presented new data with the aim of evaluating certain proposals made in earlier work. Our focus has been on two accounts of English modals, one by Lightfoot (1979, 1999, 2006) and one by Warner (1993). Both accounts agree that modals undergo a categorial change at the beginning of the 16th century. What distinguishes them is that for Lightfoot the change consists of a radical reanalysis from V to I whereas for ¹⁶ The raw figures for period 1500–25 without John Fisher’s sermons are as follows: SAdvMV: 1; SMAdvV: 162; SMVAdv: 107. ¹⁷ The fact that SAdvMV order can still be derived beyond 1500 suggests that, at least in some contexts, the second option must be available as instances of that order could be analysed as structures lacking movement of the modal.

158

ERIC HAEBERLI AND TABE A IHSANE

Warner the change is more gradual in that early English modals, although being fundamentally verbal, are already categorially distinct from regular main verbs at a subordinate level and they then develop into an entirely separate category later. Focussing on the history of V-movement in English and, more specifically, the distributional properties of modals and main verbs with respect to adverbs, we have examined whether there is evidence for a categorically distinct status of modals in early English or not. Three main observations can be made on the basis of our quantitative analysis of adverb placement. First, there are considerable frequency differences throughout early English between SAdvV order and SAdvMV order. Second, the overall diachronic trajectories of main verbs and modals are identical up to the recategorization of the modals. Third, our data provide new and entirely independent evidence for the dating of the recategorization in the early 16th century. Whereas the second and the third findings are entirely compatible with Lightfoot’s position, the first one seems to be problematic at first sight. However, we have shown that the frequency differences between main verbs and modals can be related to factors that are independent of the categorial status of the two elements, namely additional adjunct placement options with modals due to the biclausal structure they occur in, and the general clausal syntax in OE which favours SAdvV as compared to SAdvMV. We therefore conclude that adverb placement and, hence, V-movement do not provide any evidence against Lightfoot’s hypothesis that modals and main verbs are categorially fully identical in the early history of English. However, given that in Warner’s analysis main verbs and modals share their basic-level category (i.e. V ) but are distinct at the subordinate level, our findings can also be made compatible with this alternative approach under the assumption that it is the basic-level categorial feature that drives movement past adverbs in early English. Conclusive evidence in favour of one approach or the other must therefore be found outside the syntax of V-movement.

8 Tense recursion, perfect doubling, and the grammaticalization of auxiliaries Ida Larsson and Ellen Brandner

8.1 Introduction The possibility of so-called perfect doubling (1) is widespread in varieties of German, Italian, and French, and it occurs to a more limited extent also in Dutch and Danish.¹ In perfect doubling structures, the temporal auxiliary occurs twice: a finite form of the temporal auxiliary HAVE (or sometimes BE) combines with a perfect participle of temporal HAVE (or sometimes BE) and a participle of a lexical verb.² (1)

Ich habe I have.PRES

das ganz vergessen gehabt that completely forget.PTCP have.PTCP

A widespread assumption (see already Behaghel 1924) is that the doubling of the auxiliary is a compensation strategy in the context of what is widely known as Pra¨teritumschwund, that is, the loss of the preterite verb forms in Upper German dialects: the pluperfect (normally participle plus preterite form of HAVE) can now only be expressed via an ‘analytic’ form of HAVE, resulting in something like (1). The connection between Pra¨teritumschwund and double perfect is surely not wrong—but it is not complete. As has been shown by several researchers in recent times (see Ro¨del 2007 among others), the assumed causal chain (first the loss of the preterite form and then the compensation via doubling) does not hold, as will be illustrated in this chapter in some more ¹ So far there have not been any large-scale investigations of dialect speakers, but there is ongoing work by Ellen Brandner, which we draw from here. In the analysis of perfect doubling in the following, we focus on German and particularly on data from Bodensee-Alemannisch (where we have access to native speakers). For more details on the geographical distribution of double perfects in German, Dutch, French, and Italian, see Barbiers et al. (2011) and Koeneman et al. (2011) and references cited there. ² Capitalized BE and HAVE are used to cover forms of be and have in English, sein and haben in German, and so on, abstracting away from cross-linguistic and diachronic differences in realization.

Ida Larsson and Ellen Brandner, Tense recursion, perfect doubling, and the grammaticalization of auxiliaries. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Ida Larsson and Ellen Brandner (2022). DOI: 10.1093/oso/9780198871538.003.0008

160

IDA L ARSSON AND ELLEN BRANDNER

detail. Moreover, perfect doubling has a wider range of interpretations than the pluperfect alone. Double perfects like (1) can have two different interpretations. On the first reading, the double perfect corresponds closely to a simple pluperfect (English I had completely forgotten that)—we refer to this as the anterior double perfect. On this reading, the event of forgetting is anterior to a reference time in the past of the speech time. The anterior double perfect has a pluperfect interpretation, but the finite auxiliary is never in the preterite form. We will argue in this chapter that the finite auxiliary is in fact tenseless in the anterior double perfect, and that tense is instead contributed by the participial form of the auxiliary. The second reading is more specific to double perfects, and unlike the anterior it makes a statement about the present, namely (roughly) that the target state no longer holds at the speech time. On this reading, (1) states that something that was forgotten in the past has been remembered again. We follow Barbiers et al. (2011) and refer to this as the superperfect reading, but it is sometimes also referred to as the reversed-result reading. Unlike the anterior double perfect, it can have a finite auxiliary in the preterite, yielding what looks like a pluperfect of auxiliary HAVE together with the lexical participle. (2)

Ich hatte I have.PAST

das ganz vergessen that completely forget.PTCP

gehabt have.PTCP

Here, the combination of preterite HAVE with a second HAVE inevitably leads to a superperfect reading: something that was forgotten before at some time in the past was remembered again at a later time, also in the past of the time of speech. Not all varieties that have perfect doubling have both the anterior double perfect and the superperfect (see Barbiers et al. 2011 and Koeneman et al. 2011). However, in varieties that have lost the simple preterite, and which have the superperfect, (1) is systematically ambiguous between the two interpretations, and only context or the addition of adverbials may disambiguate between the two.³ One such variety is Alemannic. In Alemannic, the simple preterite has been lost, and forms of HAVE like hatte as in (2) are therefore not available. To express the situation in (2), a double perfect is instead used ³ This of course makes quantitative study difficult. In the following, we use previous corpus studies for the discussion of the historical development, but we focus on the unambiguous cases. Another methodological difficulty is that doubling is generally non-standard, and speakers’ judgements can therefore be uncertain and unsystematic. Here, we focus on the systematic patterns and variation that appear to have grammatical underpinnings.

TENSE RECURSION AND PERFECT DOUBLING

161

together with a temporal adverbial that refers to an additional point in time in the past, as illustrated in (3), where ‘until yesterday’ introduces this additional point in time: (3)

I ha des bis gescht ganz vergesse I have.FIN that until yesterday completely forget.PTCP gha have.PTCP ‘Until yesterday, I had completely forgotten that.’

The distinction between the two interpretations of the double perfect is also visible outside the German dialects. The anterior double perfect is possible only in varieties that have Pra¨teritumschwund. It occurs in some Italian and French dialects, but not in, for example, Dutch and Danish. The superperfect on the other hand is not tied to the loss of the preterite. It occurs in Dutch and Danish, but not in all Italian and French dialects that have the anterior double perfect. As we have already noted, perfect doubling is not semantically vacuous.⁴ Rather, both auxiliaries make a semantic contribution—and this is true in both the superperfect and the anterior double perfect. In this chapter, we will argue that perfect doubling involves tense recursion. This means that seemingly monoclausal structures under certain conditions allow recursion of the functional element T. As we will see, there are good reasons to assume that the two types of double perfects have partially different structures; we will argue that the anterior double perfect involves two tense elements, while the superperfect has three. The analysis we will propose for double perfects lends additional support to an analysis of the simple perfect as a complex tense (as suggested by Julien 2001; Larsson 2009; and others). The conditions for tense recursion is discussed further below, where we consider the historical emergence of double perfects in German and look more closely at the difference between languages that allow perfect doubling (henceforth dpf-languages) and languages that do not. The distribution of double perfects correlates with well-known differences in the properties of perfects. The languages that allow (superperfect or anterior) double perfects all have the possibility of positional past adverbials (which like e.g. yesterday place an event in a specific time in the past) in the ⁴ Perfect doubling should not be confused with the type of morphological tense copying discussed e.g. by Wiklund (2007). A reviewer asks what the consequences of our analysis are for IPP (infinitives for participle morphology) and PPI (participle for infinitive morphology), but a discussion of these phenomena would lead too far afield. We believe that our analysis is compatible with several different approaches to both IPP and PPI.

162

IDA L ARSSON AND ELLEN BRANDNER

present perfect, as in the Danish and German examples in (4a) and (5a), and they all have auxiliary selection (i.e. an alternation between HAVE and BE typically depending on the lexical verb), as in (4b) and (5b). An example of a Danish double perfect, with a superperfect reading corresponding to the German examples above, is given in (4c). In contrast, languages like English or Swedish, which disallow perfect doubling, also disallow positional past adverbials in the present perfect (the so-called present perfect puzzle in Klein 1992) and have HAVE as the only temporal auxiliary; see (5) and (6). BE is sometimes marginally possible with active participles in Swedish and (archaic) English, but the construction does not have the temporal properties of a perfect (see Larsson 2009, 2015 and McFadden and Alexiadou 2010 for discussion and diagnostics).⁵ (4) a. Jeg har set I have.PRES see.PTCP

ham i går. him yesterday

(Danish) (cf. Jespersen 1924: 271)

b. Han er/*har kommet. he be.PRES/have.PRES come.PTCP ‘He has come.’ c. Jeg har haft skrevet I have.PRES have.PTCP write.PTCP

til dig. to you

(5) a. Ich habe ihn gestern gesehen. I have.PRES him yesterday see.PTCP b. ER ist/*hat gekommen. he be.PRES/have.PRES come.PTCP ‘He has come.’ c. Ich habe an dich geschrieben gehabt. I have.PRES to you write.PTCP have.PTCP (6) a. *I have seen him yesterday. b. *John is come. c. *He has had built the church.

(German)

(Not a perfect)

⁵ It is important to distinguish double perfect from perfects of stative participial constructions— the latter (which also occur in English, at least to a limited degree) involve a stative passive or an unaccusative participle: (i)

a. She has been gone for several hours. b. She has had it hidden.

The double perfects under discussion are also different from English have-doubling, as in (ii), which necessarily has a counterfactual reading (Kayne 1997; Stowell 2008): (ii)

If I had’ve done it…

TENSE RECURSION AND PERFECT DOUBLING

(7) a. *Jag har I have.PRES b. ??Han a¨r he be.PRES c. *Han har he have.PRES

163

sett honom igår. (Swedish) see.PTCP him yesterday kommen (Not a perfect) come.PTCP haft byggt kyrkan. have.PTCP build.PTCP the.church

In the following, we propose that these differences all have the same source: the different stages of grammaticalization and the varying semantic contribution of the perfect auxiliary. We will argue that the dpf-languages have an impoverished HAVE without complex structure, whereas the auxiliary HAVE in languages such as English always spells out BE + a prepositional element (cf. Kayne 1993) which restricts the semantics of the perfect. In this way, we relate the variation and change in the possibility of double perfects to changes in the properties of the perfect auxiliary, where simple perfects are also affected. The development of this impoverished HAVE will be tied to the development of a BE-perfect and the possibility of past adverbials in the present perfect. Note that in our account, the grammaticalization of the temporal auxiliary is not understood as ‘climbing up the tree’, as in, for example Roberts and Roussou (2003), where a grammaticalized element is base-generated higher up the tree than its less grammaticalized relative. Instead, grammaticalization here involves changes in argument structure, and loss of structure. Section 8.2 gives an overview of the diachrony of simple and double perfects in German. Section 8.3 presents an analysis of the perfect (based on Larsson 2009) that places the variation in the properties of the auxiliary. We argue that the lexical participle in double perfects is an ordinary perfect participle, which always comes with a non-finite past tense (Julien 2001; Eide 2009; Larsson 2009). Thus, all Germanic perfects have a partly unified structure. In Section 8.4, we present our analysis of the double perfect and discuss the restrictions on tense recursion. Section 8.5 gives a short summary and conclusion.

8.2 Diachronic development of simple and double perfects in German Double perfects emerge early on in the history of the perfect; this has already been established in previous studies, which we draw data from here (see e.g. Litvinov and Radcenko 1998 and Buchwald-Wargenau 2012 for additional data). As has also previously been made clear (by Ro¨del 2007 and

164

IDA L ARSSON AND ELLEN BRANDNER

others), the emergence of doubling cannot be a direct consequence of the loss of the simple preterite (Pra¨teritumschwund). Doubling is widespread in modern German dialects, irrespective of whether they belong to the dialects that have Pra¨teritumschwund or not. Here, we want to tie the development of doubling (anterior or superperfect) to the emergence of the BE-perfect. Since both HAVE-perfects and BE-perfects have developed from often morphologically identical resultative constructions (see e.g. Larsson 2009 for the details of this development), we need diagnostics that distinguish perfects from resultatives. One such diagnostic is past counterfactuals: a past perfect but not a past tense resultative can get a past counterfactual reading (see Larsson 2009, 2015 and McFadden and Alexiadou 2010 for discussion). In addition, resultatives are generally not possible with stative and unergative lexical verbs. When we track the development of the perfect, these contexts are of particular interest. With respect to the development of double perfects, we want to distinguish between the two types of doubling (anterior and superperfect). As said above, the distinction between the two readings is sometimes hard to draw without a disambiguating context. Nevertheless, we will see that the superperfect as well as the anterior reading is already present as soon as double perfects start occurring. The clearest examples of the superperfect has the auxiliary in the preterite.

8.2.1 The development of the German simple perfect with HAVE and BE The HAVE-perfect starts appearing already in the early Germanic sources, and unambiguous examples like (8) can be found in 10th century Old High German (Alemannic), with one even earlier attestation in the Old Saxon Heliand text (see e.g. Larsson 2009; Öhl 2009). (8)

so hábet er gelógen so has he lied ‘so has he lied’ (Notker *c.950:I 544; from Dieninghoff 1904: 56)

The Germanic BE-perfect emerges somewhat later, and it takes time before the modern system is fully established. According to Grønvik (1986: 44), stative verbs including sein ‘be’ and position verbs like liegen ‘lie’, stehen ‘stand’ form perfects with the auxiliary BE in Southern German dialects from the 12th century and onwards. The earliest example is from c.1130, and has sein; see example (9) from Grønvik (1986: 44).

TENSE RECURSION AND PERFECT DOUBLING

(9)

165

si waren da zedem zagele alle gerne gewesin they be.PAST there due.to.that weak.hearted all really be.PTCP ‘They were all really weak-heartened because of that.’ (Rolandslied, Bayern 3999)

According to Paul (1902: 175), Southern German BE-perfects with liegen ‘lie’ do however not become common until the 16th century. Unaccusative change of location verbs can occur in HAVE-perfects in Notker, as in (10a), and this continues well into Early New High German. Paul (1902) also gives examples with change of state verbs like those in (10b) and (10c) from texts up until as late as the 17th century: (10)

a. sô sie dô gefaren habetôn when they there travel.PTCP have.PRES (Mod. German: BE) ‘When we have travelled there.’ (Notker, Marc. Cap. II, 35, cited after Paul 1902: 182) b. des sun wol geraˆten his son well develop.PTCP ‘His son had developed well.’ c. [derhalb hat because-of-that have.PRES

haˆt have.PRES

[er] [he]

(Mod. German: BE) (Erec 2914, MHG)

die that unterwunden,] the deed take-over.PTCP,

und hat im auch glu¨klich and have.PRES him also well geraten succed.PTCP (Mod. German: BE) ‘He took over the deed because of that and he succeded.’ (16th century, from Paul 1902: 168) In Middle Low German and Middle Dutch, there is a strong tendency of past counterfactuals also having auxiliary HAVE with unaccusatives, but the ban on BE is not categorical, as it is in English and Scandinavian where the BEperfect is either very late (17th–18th-century Danish) or does not emerge at all (Shannon 1995, cf. Kern 1912; Larsson 2009; McFadden and Alexiadou 2010). This shows that there is a BE-perfect early on in German and Dutch, but that there is also an old possibility of having a HAVE-perfect with unaccusatives, which is no longer an option (but see Gillmann 2011 for extensive discussion). There are other changes in the distribution of the German perfect during this early period. According to Oubouzar (1974: 27), positional past time adverbials start appearing in the German present perfect around the 13th century. As noted by Sapp (2009), the temporal adverb dann strongly favours the preterite tense in Early New High German. There is a shift between the 15th

166

IDA L ARSSON AND ELLEN BRANDNER

and 16th century from 5.3 per cent dann/da/do in the present perfect in the 15th century to 15.3 per cent in the 16th century (Sapp 2009: 438, table 11). We interpret the historical data in the following way. In early Germanic, the resultative with HAVE is reanalysed as involving a temporal auxiliary and a tensed participial clause, and thus the simple HAVE-perfect emerges. This is the stage that is attested for example in the texts by Notker. The same change can be found in all of the Germanic languages (as argued by Larsson 2009)— what emerges at this point is in fact a perfect of the English type, which has HAVE as the only temporal auxiliary and which disallows positional past time adverbials in the present perfect. At a second stage, a new, partly different type of perfect emerges in some of the Germanic languages (German, Dutch, and later Danish, but not English and the other Scandinavian languages). This new perfect has (i) an alternation between HAVE and BE, and (ii) it allows past adverbials in the present perfect. We will assume that this new perfect competes with the old perfect (and to some extent also with the preterite) for some time. What we can observe as variation and changes in the distribution of HAVE vs. BE with unaccusatives and the increased use of past adverbials in the present perfect is a consequence of the gradual establishment of the new perfect, which follows the grammatical reanalysis. It seems that while the new perfect can already be attested in the 12th–13th century, it gains some ground around the 15th–16th century. We return to the reanalysis in Section 8.4 below. First, we take a look at the emergence of perfect doubling.

8.2.2 The emergence of double perfects Examples of the double perfect are rare in the historical records. The first attestations of the double perfect in German are from the same period as we find an increase in the use of past adverbials in the present perfect and the BE-perfect (see e.g. Litvinov and Radcenko 1998; Buchwald-Wargenau 2012). Most of the few early examples in Buchwald-Wargenau are from Pontus and Sidonia, a text written in the 15th century in the Alemannic area (author unknown). The number of attestations increases from the 16th and 17th century and onwards. As in the present day, many of the historical examples are ambiguous between the superperfect and the anterior double perfect. However, we can still conclude that the superperfect is possible at least from around the 15th century and onwards. An indication is the fact that perfect doubling occurs with the finite auxiliary in the preterite, as in (11). As noted in Section 8.1, the

TENSE RECURSION AND PERFECT DOUBLING

167

anterior double perfect—as a compensation strategy for the pluperfect—can never have an auxiliary in the preterite.⁶ (11)

und nach grossen eren gelebt hatte gehabt and after big honours live.PTCP have.PAST have.PTCP ‘…and had lived in great honour.’ (said about someone who had been slain) (Pontus und Sidonia, 15th century, Buchwald-Wargenau 2012: 254)

In the same text, we find anterior double perfects with the auxiliary in the present tense: (12)

ir hant mir me eren gethan gehabt… you.PL have.PRES me more honour done.PTCP have.PTCP (Pontus und Sidonia, 15th century, Buchwald-Wargenau 2012: 256)

The early examples are often subjunctive, as in (13) where the subjunctive is used for marking reported speech. (13)

a. …hette Sie sich in einen Knaben verliebt have.SUBJ she REFL in a boy fallen.in.love.PTCP gehabt have.PTCP ‘She had fallen in love with a boy.’ (reported speech) (Hexenverho¨re P Bamberg, 1630, from Buchwald-Wargenau, 2012: 263) b. iedoch hette sich die Ertmansche sehr geeifert but have.SUBJ REFL the E. much rant.PTCP gehabt have.PTCP ‘but she had ranted against the E. much’ (reported speech) (Hexenverho¨re P Crivitz, 1642, from Buchwald-Wargenau, 2012: 264)

As noted, it has often been assumed that the emergence of double perfects was a consequence of the loss of the preterite (e.g. Behaghel 1924): a double perfect was ‘necessary’ to express a pluperfect when the preterite had been lost. However, it has since been established that this cannot be the right chain of ⁶ In Section 8.1, we illustrated the superperfect with a sentence that has a ‘reversed result’ reading, i.e. the resultant state no longer holds at the time of utterance. This is the most typical interpretation of a superperfect, but it is of course hardly possible with the lexical meaning of living (or dying). However, the precise interpretation of the superperfect depends on lexical meaning and world knowledge. See Section 8.4.2 for further discussion and illustrations.

168

IDA L ARSSON AND ELLEN BRANDNER

events, for several reasons (cf. Ro¨del 2007; Buchwald-Wargenau 2012). First, the first attestations of the double perfect predates the loss of the preterite.⁷ Second, the superperfect also occurs in varieties that have retained the simple past, and the matrix auxiliary can even be in the past tense, as we have seen. This, however, does not mean that the use of the double perfect is completely independent of Pra¨teritumschwund. Rather, the anterior double perfect seems to spread when the preterite is lost. According to Lindgren (1957), the present perfect is extended to narrative past contexts in the 16th century, and there are Southern German letters from the end of the 16th century where the preterite is already rather uncommon (1957: 65, 95). At this point, the anterior double perfect too is not uncommon in the Southern German varieties (cf. Buchwald-Wargenau 2012). It seems that from then on it is fully established in this variety. In varieties that retain the preterite, the anterior double perfect is generally restricted to certain contexts, namely subjunctives. This is the case in Standard German, as we will see in Section 8.4.1 below. The anterior is found more generally in varieties that have lost the preterite: it occurs in Alemannic, Swabian, and Northern Italian, but not in, for example, Dutch (Barbiers et al. 2011, table 1). In Section 8.4, we argue that the anterior double perfect is blocked by the pluperfect in varieties that have both. The use of the superperfect, on the other hand, we take to be completely independent of the availability of simple past tense forms.

8.2.3 Summary There are several difficulties in tracing changes in the perfect tense system in the historical records. Many examples are ambiguous. Perfect doubling is non-standard and infrequent. Moreover, we can note several different changes and gradual shifts in the frequency of the different constructions: (i) a HAVEperfect emerges and spreads, (ii) a BE-perfect emerges with unaccusatives and alternates with the HAVE-perfect for a considerable time, (iii) perfect doubling emerges and becomes more common, (iv) the perfect starts to be used with ⁷ Sapp takes the early spread of the present perfect in German to be part of Pra¨teritumschwund. However, the fact that the frequency of the present perfect (relative to the simple past) rises from 11.9% in the 14th century to 30.2% in the 16th century (Sapp 2009: 425, table 1) is at least partly due to the establishment of the present perfect. We can see a parallel development in English. In English, the frequency of perfects with HAVE appears to have risen steadily throughout the period 1150–1710 (McFadden and Alexiadou 2006: 277, table 1; cf. Elsness 1997: 264, table 4.1). As far as we understand, the assumption that the double perfect emerges as a strategy when pluperfect morphology is no longer an option, presupposes a substantial restriction of preterite morphology and not only a general increase in the use of the perfect (which, as we have noted, can be observed in all of the Germanic languages).

TENSE RECURSION AND PERFECT DOUBLING

169

positional past adverbials, and (v) finally spreads to narrative past contexts as the preterite starts to disappear. These developments are clearly intertwined. Focussing on early attestations and unambiguous cases, we have suggested that the development of the simple perfect in German takes place in two steps. In the first step, a common Germanic HAVE-perfect emerges; in the following we refer to this as the English-type perfect. At a later time, a new perfect with both HAVE and BE emerges, which allows positional past adverbials in the present perfect; we refer to this as the (Modern) German-type perfect. However, the old perfect is retained, and competes with the new perfect in certain contexts (which gives variation and gradual shifts in the historical data). The superperfect and the anterior double perfect both seem to appear at around the time when the new perfect spreads. In the following, we will argue that this is not a coincidence. Instead, we suggest that the possibility of doubling is a direct consequence of the development of a German-type perfect. As noted in the introduction, this hypothesis is supported by the synchronic cross-linguistic patterns: only languages that have auxiliary selection, and which allow positional past adverbials in the present perfect, have the possibility of double perfects. Like previous work, we conclude that perfect doubling cannot be a direct consequence of the loss of the preterite, since it predates Pra¨teritumschwund historically, and since doubling also occurs in varieties that retain the preterite. At the same time, the use of double perfects is not completely independent from Pra¨teritumschwund. The anterior double perfect becomes more frequent (and occurs in new contexts) when the preterite disappears. As we will argue below, this is because the anterior double perfect competes with the simple pluperfect (which requires past tense morphology on the auxiliary). Before we proceed to the analysis of double perfects, we present an analysis of the German and English simple perfects.

8.3 Variation in the Germanic Perfect This section outlines the analysis we are assuming for the simple perfect. There is a long discussion of how the perfect should be analysed, with two main alternatives. First, the perfect is sometimes taken to place an event (E) before a reference time (R), which in turn is placed at the speech time (S) in a present perfect, or in the past in a pluperfect. This is the analysis of Reichenbach (1947) and many others, and it intuitively captures perfects such as I have already eaten lunch where the past event of eating is relevant at the

170

IDA L ARSSON AND ELLEN BRANDNER

reference/speech time. As we will see in this section, however, the Reichenbachian account runs into problems with the English-type perfect. The main alternative is referred to as the extended-now account (McCoard 1978; Iatridou et al. 2001; Rothstein 2008). On this view, the perfect involves a temporal interval that starts in the past and continues into the present. The extendednow approach fares better with examples such as She has been sleeping for an hour now, where the sleeping event holds both in the past and in the present. For the English-type perfect, we will assume a version of the extended-now analysis. Importantly, even the English present perfect includes a past tense in addition to a present tense—this is how we can account for the fact that She has been sleeping for an hour makes a statement both about the present and about the past. Syntactically, we suggest that all Germanic perfects have a partly unified structure that includes a non-finite past T, and we argue that variation in the properties of the perfect tense in Germanic depends on the structure that is spelled out by the auxiliary HAVE. In the English-type perfect, the auxiliary restricts the interpretation of the non-finite past tense and establishes the extended-now interval. As we will see further in Section 8.4, the properties of double perfects give additional support to the analysis of the perfect as a complex tense.⁸

8.3.1 The German Perfect Following Larsson (2009) and others, we will assume that a simple present or past perfect has the biclausal structure provided in (14) below (cf. e.g. Julien 2001). The perfect participial clause always comes with a past tense T, whereas the matrix tense varies with the tense of the auxiliary. In this analysis, a past perfect involves a past under a past, and a present perfect past under a present tense. We take tense and aspect to be predicates that establish a relation of order or inclusion between their time-denoting arguments, represented as ZP (for ZeitP) in (14) (Zagona 1995, 2007; Stowell 1996; Demirdache and UribeEtxebarria 2000, 2007). Tense (T) establishes a relation between an evaluation time (EVT; the speech time in a root clause) and topic time (TOP; see Klein 1992). The aspectual head below T relates the assertion time to the event time, which we assume is introduced in VoiceP. ⁸ The discussion in Section 8.3.1 is kept brief, for reasons of space. For additional discussion, we refer to Iatridou et al. 2001; Rothstein 2008; and ch. 3 in Larsson 2009.

TENSE RECURSION AND PERFECT DOUBLING

171

(14) The present or past perfect: CP T1P

C ZP EVT1

T1 ±PAST ZP TOP1

Asp1P

Asp1

… HAVE

T2P ZP EVT2

T2 +PAST

Asp2P

ZP TOP2 Asp2 Voice2P

Given the analysis of the perfect as a complex tense, a present and past perfect involves two topic times (TOP1 and TOP2 ). In the present perfect, the matrix topic time (TOP1 ) lies in the present, whereas the embedded (non-finite) topic time (TOP2 ) lies in the past. An advantage of this analysis is that it only involves elements that are independently motivated; that is, there is no perfect phrase in the structure. The challenge is to derive the right semantics. As we will see, the embedded topic time corresponds to the extended-now interval or the perfect time span in previous analyses of the perfect. Assuming that positional temporal adverbs always relate to some topic time (see e.g. Pancheva and von Stechow 2004), the analysis rather straightforwardly accounts for the behaviour of present and past perfects in languages like Dutch or Standard German. The present perfect, which involves a present tense embedding a past tense, allows both present (and future) tense adverbials (relating to the higher topic time) and past tense adverbials (relating to the lower topic time): (15)

a. Morgen hat er die Pru¨fung bestanden (German) tomorrow have.PRES he the exam pass.PTCP ‘Tomorrow, he will have passed the exam.’

172

IDA L ARSSON AND ELLEN BRANDNER

b. Er hat gestern gearbeitet. he have.PRES yesterday work.PTCP ‘He worked yesterday.’ The past perfect has two topic times that lie in the past, and past adverbials are therefore ambiguous. The example in (16) can either mean that the eating took place at nine o’clock (when the adverbial modifies the embedded topic time), or that the eating took place before nine o’clock (when the adverbial modifies the matrix topic time):⁹ (16)

Er hatte gestern um neun Uhr einen Apfel gegessen he have.PAST yesterday at 9 o’clock an apple eat.PTCP (i) E(eat) < 9 o’clock (ii) E(eat) at 9 o’clock

(TOP1 at 9 o’clock) (TOP2 at 9 o’clock)

Now, something additional must be said about the English (Swedish, Icelandic, and Norwegian) present perfect, which does not allow past adverbials.¹⁰

8.3.2 Extended-now and the present perfect puzzle As noted above, English and the Scandinavian languages apart from Danish disallow past adverbials like yesterday in the present perfect; cf. (17). However, past adverbials are still ambiguous in the past perfect, as shown by (18). This is what Klein (1992) refers to as the present perfect puzzle. (17)

*Frida has eaten an apple yesterday / last week.

(English)

(18)

a. Context: Yesterday Frida came to Peter’s office at six o’clock. But Peter had left the office at six o’clock. E (leaving) < six o’clock b. Context: Yesterday Frida came to Peter’s office at seven o’clock. But Peter had left the office at six o’clock. E (leaving) at six o’clock

⁹ To be precise, the adverbial specifies one of two topic times that both lie in the past. On the first reading, the adverbial gives the time of the most embedded topic time (corresponding to the perfect time span, which is directly related to the event time of eating). On the second reading, the adverbial specifies a higher topic time, which is interpreted as the right boundary of the lower topic time (i.e. it gives the right boundary of the perfect time span). See Larsson (2009) for some discussion. ¹⁰ Differences in the properties of the present tense cannot account for the cross-linguistic patterns (as proposed by Pancheva and von Stechow 2004). All of the Scandinavian languages have a present tense which is identical to the German present tense, but only Danish allows past adverbials in the present perfect.

TENSE RECURSION AND PERFECT DOUBLING

173

The ban on positional past time adverbials in the present perfect is one of the motivations for the extended-now analysis of the present perfect (see e.g. Iatridou et al. 2001; Pancheva and von Stechow 2004). On an extendednow analysis, the present perfect has a topic time that starts in the past and continues into the present. Following Iatridou et al. (2001), we refer to this extended-now as the perfect time span (PTS). The difference between the simple past and the present perfect can thus be represented as in (19). The Reichenbachian reference time (R) corresponds to the right boundary of the perfect time span (= S in the present perfect).

(19)

a. The simple past: [

]

S

(The brackets represent the topic time, S the speech time.)

b. The Scandinavian and English present perfect: =[

S=]

(The barred brackets represent the perfect time span.) The difference between the simple past and the present perfect is perhaps particularly clear in sentences where the eventuality holds throughout the topic time, as in the examples in (20). In the so-called universal perfect in (20a), the topic time necessarily includes the speech time, and the eventuality must therefore hold at the speech time—(20a) is only possible if Lisa is still alive at the speech time. In the simple past in (20b), on the other hand, the topic time excludes the speech time. The interpretation is therefore that Lisa is no longer alive. (20) a. Lisa has lived in Stockholm her whole life. #But she died a few years ago. b. Lisa lived in Stockholm her whole life, (but she died a few years ago). What we need is a way to derive the perfect time span (the extended-now) in the English-type perfect (19b). In the following, we argue that the basic structure of the present and past perfect in English is the same as in German. Also in English, the perfect involves a non-finite past tense. The difference between English and German we argue is the consequence of the

174

IDA L ARSSON AND ELLEN BRANDNER

auxiliary. In Section 8.3.3 we will see how we can use the auxiliary to derive the English-type perfect time span, while maintaining the structure in (14) above.

8.3.3 The contribution of the auxiliary Starting from a Kaynean analysis of HAVE as BE + P (Kayne 1993), we suggest that temporal HAVE in languages like English spells out the copula + a temporal preposition that restricts the interpretation of the perfect and establishes a perfect time span which necessarily abuts the time set by the matrix tense (cf. Larsson 2009). Following Demirdache and Uribe-Etxebarria (2007), we assume that locative and temporal prepositions differ with respect to the type of arguments: locative in in (21a) takes DP arguments, whereas temporal in in (21b) takes temporal arguments.¹¹ The difference between possessive and temporal HAVE can be understood in the same way; cf. (22). (21)

a. Locative in: (Kim is) in Stockholm. [PP DP P [DP Stockholm]] b. Temporal in: (Kim left) in December. [PP TOP P [ZP Z [NP SIT December]]] (cf. Demirdache and Uribe-Etxebarria 2007: 336)

(22)

a. Possessive HAVE: [DPi Cop [PP DPi Pposs [XP]]] b. Temporal HAVE in the English-type perfect: [Cop [PP TOP1 PT [TP T [AspP TOP2 Asp [VoiceP]]]]] (modified from Larsson 2009)

In this analysis, possessive HAVE has a possessive preposition (cf. e.g. Harley 2002) that takes DP arguments, whereas temporal HAVE has a temporal preposition (largely corresponding to temporal at) that relates matrix to the embedded topic time: it places the matrix topic time at the participial topic time. This means that the auxiliary will make the matrix topic time the right boundary of the participial topic time, which also stretches into the past. In this way, the prepositional element of HAVE in combination with the embedded past T derives the perfect time span, with RB=S as in the representation in (19b) above. We can describe the semantics informally as in (23): ¹¹ For simplicity, we assume that the DP/topic time is the external argument of the locative/temporal preposition, leaving aside the possibility of an internally complex PP.

TENSE RECURSION AND PERFECT DOUBLING

(23)

I have seen him. Asp2 : E(see) ⊂ TOP2 T2 : TOP2 < EVT2

175

(TOP2 = PTS) (EVT2 ⊆ E1 )

PT : TOP1 at TOP2 Asp1 : TOP1 ⊂ E1 T1 : TOP1 = EVT1 , EVT1 = S →RB of PTS at S Here, the (perfective) embedded Asp (Asp2 ) places the event time within the participial topic time (TOP2 ; corresponding to the perfect time span). The embedded past tense places TOP2 before an evaluation time, which we for simplicity assume will be interpreted as lying within the event time of HAVE (see further Larsson 2009).¹² What is important here is the contribution of the prepositional element of the auxiliary (PT ), which places the matrix topic time (TOP1 ) at TOP2 . Assuming that the auxiliary comes with imperfective aspect, the matrix topic time includes the matrix event time. Since (23) is an English present perfect, the matrix T (T1 ) will say that TOP1 corresponds to the speech time. This means that the TOP2 will be at the speech time. We can now account for the restriction on past adverbials in the present perfect. Since the participial topic time abuts the present, and the matrix topic time is in the present, neither topic time can therefore be placed completely in the past by a positional past time adverbial. In the past perfect, on the other hand, both topic times lie completely in the past (just as in German), and positional past time adverbials are therefore ambiguous. In this way, the fact that TOP2 lies at TOP1 will only have an effect on the possibility of adverbials in the present perfect. (See Larsson 2009: ch. 3 for additional discussion.) Contra Kayne (1993), we assume that the prepositional element of HAVE makes a semantic contribution, and that it is never spelled out by BE. In a system with auxiliary selection, PT is therefore not necessarily included in the structure of the perfect, as is shown for the BE-perfect in (24). Informally we can describe a German BE-perfect in the following way: (24)

Er ist (gestern) gekommen. ‘He has arrived (yesterday).’ ASP2 : E(arrive) ⊂ TOP2 T2 : TOP2 < EVT

(TOP2 = PTS) (EVT ⊆ E1 )

¹² There is of course more to say about the evaluation time of embedded (non-finite) tenses. It seems, however, that the evaluation time can generally be understood as a higher event time—here, the time of HAVE, in root clauses the time of the speech event.

176

IDA L ARSSON AND ELLEN BRANDNER

Asp1 : TOP1 ⊂ E1 T1 : TOP1 = EVT1 , EVT1 = S → RB of PTS ≤ S Since PT (which in the English-type perfect places TOP1 at TOP2 ) is not included in the structure of the German-type perfect, TOP2 can lie completely in the past of TOP1 . This derives the German perfect time span, with largely the interpretation proposed by Pancheva and von Stechow (2004) and Rothstein (2008). Now, the German HAVE-perfect and BE-perfect appear to be semantically identical. We will therefore assume the same derivation for German HAVEperfects as for BE-perfects. Thus, HAVE does not restrict the interpretation of the embedded past tense in German, and past adverbials are therefore possible in the present perfect. We thus propose that in a German-type perfect, the auxiliary (HAVE/BE) merely spells out verbal inflectional features and the predicational verbal element that we for simplicity represent as Cop (for copula).¹³

8.3.4 The grammaticalization of the temporal auxiliary In the account sketched in Section 8.3.3, the difference between the Englishtype perfect and the German-type perfect boils down to the structure spelled out by the auxiliary. In a system with auxiliary selection, we expect the structure of the perfect to lack the prepositional element, and we expect positional past time adverbials to be possible in the present perfect. This is also what we find cross-linguistically. In Germanic, the languages that lack auxiliary selection (English, Swedish, Norwegian, Icelandic) disallow past adverbials in the present perfect. The same appears to hold also for Romance: Spanish does not have auxiliary selection, and has a perfect of the English-type (Larsson 2009).¹⁴ The languages that have auxiliary selection all allow past adverbials in the present perfect. Given the historical data discussed briefly in Section 8.2 above, the pattern seems also to hold diachronically. Recall that we noted a shift in the distribution of the present perfect around the 15th century, which we ¹³ Note that in this analysis, there is no obvious structural difference between the HAVE-perfect and the BE-perfect, and a Kaynean account of auxiliary selection is not available. See Larsson (2015) for an analysis where auxiliary selection is seen as a morphological agreement phenomenon. ¹⁴ There is additional variation in the distribution of the present perfect relative to the simple past, depending only partly on Pra¨teritumschwund. For instance, the distribution of the present perfect is not completely identical in English and Swedish, or in all varieties of English. However, as far as we know, nothing suggests that this variation is structural.

TENSE RECURSION AND PERFECT DOUBLING

177

took to be a consequence of the development of a new perfect, with auxiliary selection. We can now interpret the historical development of the perfect sketched in Section 8.2 above in the following way (cf. Larsson 2009: ch. 10). The HAVEperfect first emerges when the possessive preposition of possessive HAVE is reanalysed as a temporal preposition. This gives us the English-type system without auxiliary selection, and without past adverbials in the present perfect. At a later stage, temporal HAVE is reanalysed in German as a semantically weaker element that only spells out a predicational element, not a temporal preposition. As a consequence, German gets a system with auxiliary selection, and with the possibility of having past adverbials in the present perfect. We have now placed the observed variation in the properties of the perfect auxiliary, and we view the variation in the properties of the perfect in Germanic as a consequence of the different stages in the grammaticalization of HAVE. In this way, the analysis can account for the differences in the present perfect between English, on the one hand, and German, on the other, without relying on a notion of a parameterized semantics of tense. Instead, we assume that three different kinds of HAVE can be found in Germanic, corresponding to three different stages in the grammaticalization process:¹⁵ (i) Possessive HAVE spells out Cop + Pposs (ii) English temporal HAVE spells out Cop + PT (iii) German temporal HAVE/BE spells out Cop Here the grammaticalization involves a change in the structure of HAVE rather than a shift from lexical to functional element, or from one functional element to a higher functional element. Of course, the complementation of HAVE also changes—temporal HAVE necessarily takes a tensed (clausal) complement, whereas possessive HAVE does not. (In the resultative construction, the participle is tenseless; see Larsson 2015 for discussion.) As in many cases of grammaticalization, reanalysis here leads to a new temporal HAVE without a necessary loss of the older types of HAVE. We suggested above that the new German-type HAVE competes with the older English-type HAVE in the history of German—this can be observed in variation in the choice of auxiliary

¹⁵ We assume that the Cop takes Pposs /PT as its complement, but leave open whether the preposition incorporates into Cop to spell out HAVE, or if HAVE is a spell-out through spanning (e.g. Svenonius 2016).

178

IDA L ARSSON AND ELLEN BRANDNER

with unaccusatives. The fact that the old auxiliary HAVE is retained (in some varieties) will also be crucial for our analysis of the superperfect.¹⁶ In Section 8.4, we turn to the double perfects. We will propose that the superperfect involves the combination of a less grammaticalized English-type HAVE with the more grammaticalized German-type HAVE. The anterior double perfect, on the other hand, involves doubling of German HAVE (or possibly BE). In this way, the possibility of double perfects is also tied to the grammaticalization of the auxiliary—only languages that have the more grammaticalized German-type HAVE allow perfect doubling, and this is all that is needed for doubling to be possible. This is also what the historical data reviewed in Section 8.2 suggest.

8.4 The interpretation and structure of double perfects In this section, we argue that while the anterior double perfect has a structure and interpretation largely corresponding to an ordinary (simple) German pluperfect, the superperfect involves additional tense recursion.

8.4.1 The anterior double perfect As noted in the introduction, the anterior double perfect has the interpretation of a pluperfect. It also allows for the same range of interpretations as a simple pluperfect: it can have a resultative (25a), recent past (25b), experiential (25c), or universal (25d) reading, in the same way as the English pluperfect in (26). (For a discussion of the different readings of the perfect see, for example, Iatridou et al. 2001 and Pancheva 2003.) (25)

a. dazmol ha¨t se ihre Brille at-that-time have.FIN she her glasses gha have.PTCP ‘at that time, she had lost her glasses’

verlore lost.PTCP (Resultative)

¹⁶ We do not believe that there is a principled explanation to why the less-grammaticalized elements are sometimes (but not always) preserved when auxiliaries grammaticalize. In the case of HAVE, it is clear that possessive HAVE remains a useful element, much as before the emergence of temporal HAVE. In the case of the older and newer temporal HAVE, the case is less clear, and it is perhaps not surprising that the uses of the older HAVE diachronically become more and more restricted, and are often non-standard.

TENSE RECURSION AND PERFECT DOUBLING

179

b. sie ha¨t do grad ihren Abschluss gmacht she have.FIN then just her degree make.PTCP gha (Recent past) have.PTCP ‘she had just finished her degree’ c. mir sind scho oft uff Mallorca gfahre we be.FIN already often on M. drive.PTCP gsi (Experiential) be.PTCP d. ich ha immer in Konstanz gwohnt gha (Universal) I have.FIN always in K. live.PTCP have.PTCP (26)

a. b. c. d.

At that time, she had lost her glasses. At that time, she had just finished school. She had already been to Mallorca many times. She had always lived in Konstanz.

(Resultative) (Recent past) (Experiential) (Universal)

In the anterior double perfect, there is no present tense feature involved, but two past tenses—as in a simple pluperfect. (The gloss of the finite aux in (25) is therefore FIN and not PRES.) Based on the semantics, we conclude that the anterior double perfect has the same structure as a Standard German simple pluperfect. What is particular about the anterior double perfect is that the finiteness + agreement features and the past tense feature are spelled out separately. This explains our observation above that the anterior double perfect is restricted to varieties that have lost the simple preterite, and that it spreads when the preterite is lost. In the dialects that have lost the preterite, the finiteness of verbs is disconnected from tense, so that finite verbs do not carry tense morphology. Instead, the past tense (the marked tense value) requires a participle. We saw that there is some early evidence of an anterior double perfect, and it seems that the possibility of the anterior appears before the preterite is lost and then gains ground as the preterite disappears. Given that the anterior double perfect and the simple pluperfect involve the same structure, we can assume that the former is blocked by the latter, due to a principle like Minimize exponence (Siddiqi 2006). This means that whenever there is a choice of having a simple perfect with the finite auxiliary in the past tense, this will be used. Now, there is one context where Standard German also has a tenseless auxiliary, namely in subjunctives. Subjunctives are tenseless in Standard German (even though they are built on the present and preterite stems respectively), and they cannot even be used to mark tense agreement in sequence of tense

180

IDA L ARSSON AND ELLEN BRANDNER

contexts. In order to get a past shifted reading in a subjunctive sequence of tense context, perfect doubling is therefore obligatory, even in Standard German (cf. Ro¨del 2007: 110): (27)

Er sagte, er habe schon 600 Kilometer zuru¨ckgelegt he say.IND.PAST he have.SUBJ1 already 600 km cover.PTCP *(gehabt,) als er ins Schleudern geraten sei. have.PTCP when he into sliding get.PTCP be.SUBJ1 ‘He said that he had already covered 600 kilometers when he started sliding.’ E (say) > E (start sliding)

In examples like (27), the finite auxiliary spells out finiteness and mood, but participle morphology is required for the past tense. The fact that examples like (27) occur in varieties that do not otherwise have the anterior double perfect, and which have tensed indicative forms, suggests that the anterior double perfect does not involve any special structure, but is a direct consequence of the tenselessness of the subjunctive and the properties of the ordinary German perfect. The distribution of the anterior double perfect thus seems to depend directly on the availability of tensed finite forms. We assume that in varieties that retain the preterite, indicative preterite forms block anterior doubling—a simple pluperfect must be used when possible. As noted, the historically early examples are often subjunctive—this is a context where the double perfect is not blocked by the pluperfect. In varieties where the preterite is lost, the indicative finite auxiliary will also lack tense morphology. Doubling is then possible, and required, for an indicative pluperfect. To summarize, we assume therefore that the anterior double perfect has the structure of an ordinary German perfect. The finite auxiliary spells out finiteness and agreement (like an ordinary German auxiliary but without tense), whereas the participial form of the auxiliary spells out past tense. As in a normal simple perfect, the participle of the lexical verb comes with a past tense value, as well. This gives a past under past reading, that is, a pluperfect. The anterior double perfect is never possible in a language with an Englishtype perfect, since it requires that the highest auxiliary does not contribute to the temporal interpretation. The English-type auxiliary always necessarily does, since it includes a temporal preposition. We therefore only find the anterior double perfect in varieties that have the more grammaticalized German-type temporal auxiliary.

TENSE RECURSION AND PERFECT DOUBLING

181

8.4.2 The superperfect Unlike the anterior double perfect, the superperfect has a specific meaning that we do not find in simple perfects. On a superperfect reading, (28) (=(1)) states that the target state of the event of forgetting does not hold at the speech time. This is sometimes referred to as the reversed result reading of the superperfect. (28)

Ich habe I have.PRES

das ganz vergessen gehabt that completely forget.PTCP have.PTCP

Unlike a simple pluperfect (which can have implications of a reversed result as well) or an anterior double perfect, the superperfect here involves a finite auxiliary with a present tense value, and the superperfect relates a past time event to the present—like a present perfect. Morevoer, the superperfect is not restricted to varieties where the finite verb does not spell out tense, unlike the anterior double perfect. As we have seen, the finite auxiliary can also be in the past tense in the superperfect, cf. (2) above. Put differently, the question of whether the finite verb is tensed or not is a completely orthogonal issue. Note that the superperfect is not restricted to participles of verbs that have a target state. On the contrary, the superperfect is possible with telic (29a), atelic (29b–c), and stative verbs (29d).¹⁷ The participle has an active reading. (29)

a. Ich I b. das the c. wir we d. ich I

habe meine Brille verloren gehabt have.PRES my glasses lose.PTCP have.PTCP Baby hat die ganze Nacht geschrien gehabt baby have.PRES the whole night cry.PTCP have.PTCP haben die ganze Nacht getanzt gehabt have.PRES the whole night dance.PTCP have.PTCP habe sie geliebt gehabt have.PRES her love.PTCP have.PTCP

With verbs that have a target state (29a), the superperfect states that the target state lies in the past (i.e. reversal). In (29b–d), on the other hand, it rather places the eventuality in a distant past, and implies that it is over at the speech time. The superperfect can be found in Dutch and French varieties, with similar interpretations. The Dutch example in (30a) states that a bike that was stolen in ¹⁷ As already mentioned, double perfects have the flavour of non-standard speech and thus speaker judgements are often biased. Also, context can play an important role for acceptance. Nevertheless, the sentence in (29a) has an acceptance rate of 71% of 516 Alemannic speakers. (29c) on the other hand was accepted by only 27% of the speakers. Since these results were obtained via written questionnaires, it must be left open at the moment whether these diverse results are due to badly designed contexts or whether further factors play a role. In direct inquiries with hints given for the readings described, informants were much less reluctant. But naturally, this cannot be quantified.

182

IDA L ARSSON AND ELLEN BRANDNER

the past is back with its owner at the time of speech. The French example in (30b) states that the event of voting for the socialists lies in the past, and is no longer relevant in the present. (30)

a. Jan heeft mijn fiets gestolen Jan have.PRES my bike steal.PTCP gehad (South-eastern Dutch) have.PTCP ‘Jan had stolen my bike (but now it’s back).’ (Barbiers et al. 2011, ex. (14)) b. J’ai eu voté socialiste. (Colloquial French) I=have.PRES have.PTCP vote.PTCP socialist ‘I once voted for the socialist party (but that was long ago, and I will certainly not do it again).’ (Barbiers et al. 2011, ex. (5))

We take the most embedded participle to be an ordinary perfect participle—it is active, and there are no (systematic) lexical restrictions like the ones we find in stative participles. In the present analysis, this means that the participle of the lexical verb carries a past tense value. We also suggest that the participle of the auxiliary carries a past tense value. Together with the tensed finite verb, this means that the superperfect involves three layers of tense morphology. We can see this in the behaviour of temporal adverbials. Examples like (31) make a statement about the present, but past adverbials are still ambiguous, as in a past perfect. The example in (31) states either that the losing event took place at 9 o’clock or that it took place before 9 o’clock. (31)

ich ha d’bru¨lle wohl schon um nu¨ni I have.PRES the.glasses obviously already at nine verlore gha lose.PTCP have.PTCP ‘I had obviously lost my glasses at nine o’clock (but I have found them now).’ (i) E(lose) < 9 o’clock (ii) E(lose) at 9 o’clock

In other words, the superperfect involves embedding a perfect under additional temporal structure. In order to make the temporal interpretation of the superperfect clearer, we can look at a superperfect of a universal perfect. Again, the universal perfect states that the eventuality holds throughout the perfect

TENSE RECURSION AND PERFECT DOUBLING

183

time span (= the most embedded topic time). Now, with a superperfect of a universal perfect, the eventuality must not hold at the speech time. Unlike a simple universal perfect, examples like (32) thus imply that the eventuality of admiration lies in the past. (32)

ich hab ihn immer bewundert gehabt— I have.PRES him always admire.PTCP have.PTCP zum Tag X, als…. to.the day X, when ‘I always used to admire him—until the day when…’

bis until

To derive the semantics of the superperfect, we propose that the superperfect involves a German-type perfect auxiliary embedding an English-type perfect. A partial structure of the superperfect is sketched with the relevant morphology in (33) (abstracting away from word order).

(33)

Ich

habe

das

ganz

vergessen

gehabt

I

have.PRES

that

completely

forget.PTCP

have.PTCP

T1 PRES … Cop T2 PAST … Cop PT habe1

gehabt2

T3 PAST

… Voice…

vergessen3

In this analysis, the lexical participle and the participle of the auxiliary establishes the perfect time span that we know from the English simple perfect. The participial auxiliary includes the temporal preposition that sets the right boundary of the perfect time span to the topic time of the intermediate tense. Since the tense (T2 ) of the participial auxiliary is past, the right boundary of the perfect time span is necessarily in the past. The highest tense adds a topic time which lies in the present, just as in the German simple present perfect. This derives the interpretation of the superperfect: the matrix present perfect places the embedded perfect time span in the past of the speech time. We thus propose that the superperfect requires a system that has both an English-type auxiliary HAVE (including the temporal prepositional element) and the more grammaticalized German auxiliary HAVE (that lacks the temporal preposition). The question is why this should be, and why languages

184

IDA L ARSSON AND ELLEN BRANDNER

like English cannot have doubling. One possibility is the (c-)selectional requirements of the auxiliary: an auxiliary cannot combine with itself. Another possibility is semantics, and this is what we want to suggest. The combination of an English-type perfect with another English-type perfect would involve vacuous tense recursion: the participial auxiliary would give the right boundary of the most embedded topic time, and the right boundary would then be placed in the present by the matrix auxiliary, yielding the same interpretation as a simple present perfect. With a ban on vacuous embedding (as proposed by Demirdache and Uribe-Etxebarria 2000; Julien 2001), this is necessarily ruled out. The option of having a German perfect embedding another German perfect is, on the other hand, less easily excluded on semantic grounds, and in fact this is what we found in the anterior double perfect. This analysis makes two predictions. First, we only expect to find the superperfect in varieties that have developed the more grammaticalized Germantype auxiliary, that is, varieties (like Danish and Dutch) that have auxiliary selection and which allow past adverbials in the present perfect. As we have seen, this prediction appears to be borne out. Second, since BE never spells out PT , we don’t expect to find BE in superperfects. This prediction too appears to be borne out; see the examples in (34) which show that superperfects are degraded or ungrammatical with the auxiliary BE. (34)

a. ?*der schlu¨ssel isch doch immer dort glega the key be.PRES PRT always there lie.PTCP gsi be.PTCP Intended: ‘The key always used to lie there.’ (But now it’s gone.) b. ??der Schtuel isch immer i der Eck gschtande the chair be.PRES always in this corner stand.PTCP gsi be.PTCP Intended: ‘The chair always used to stand there.’

8.5 Conclusion In this chapter, we have suggested that the possibility of perfect doubling found in some varieties of Germanic is due to the grammaticalization of a temporal auxiliary HAVE that does not spell out a temporal preposition but which corresponds more directly to BE. We have proposed that when the temporal preposition is lost from the structure of the perfect, a system with auxiliary

TENSE RECURSION AND PERFECT DOUBLING

185

selection, the possibility of past adverbials in the present perfect, and perfect doubling emerges. We have argued that the anterior double perfect is built on the ordinary grammaticalized German HAVE. In principle, the anterior double perfect should therefore also be available as soon as the present perfect puzzle is lost. However, the anterior double perfect is blocked by the simple pluperfect and is therefore expected to spread when the preterite is lost. In other words, the spread of the anterior double perfect is directly tied to Pra¨teritumschwund. The so-called superperfect, on the other hand, we have assumed involves both the old auxiliary HAVE, which includes a prepositional element which restricts the interpretation of the perfect time span, and the new HAVE. (Languages might then lose the English-type auxiliary, in which case superperfects will no longer be possible.) Thus, we have argued that the emergence of simple as well as double perfects is tied to the grammaticalization of the auxiliary HAVE, and we have assumed that the variation in Germanic is placed in the functional makeup of lexical items (auxiliaries), rather than in the syntactic or semantic component. The grammaticalization involves a change in argument type and selectional properties, and in a second step loss of structure. We have tied the (im-)possibility of tense recursion to the inventory of different grammaticalized auxiliaries, allowing iteration or not. We suggest that the superperfect is possible only iff the language has more than one type of temporal HAVE. Also the anterior double perfect requires a more grammaticalized HAVE—but nothing else. We take the impossibility of doubling in English to be due to a general ban on vacuous tense recursion (as has been previously suggested by Demirdache and UribeEtxebarria 2000 and Julien 2001). We also take the contribution of the perfect participle morphology to be constant across the constructions—in both simple and double perfects, the participial clause has a non-finite past tense value.

9 P-incorporation in the history of Icelandic Jóhannes Gı´sli Jónsson and Brynhildur Stefánsdóttir

9.1 Introduction This chapter reports the results of an investigation of preposition incorporation (henceforth, P-incorporation) in the history of Icelandic. We will illustrate the basic properties of P-incorporation into verbs and participles as they can be determined from the available textual sources. It will be shown that P-incorporation is subject to two well-known restrictions on incorporation, the Head Movement Constraint (HMC) and the Adjunct Condition. Moreover, since P-incorporation involves left adjunction to a verb or participle, strict adjacency between the incorporated preposition and its host is required so long as there is no excorporation from the [P+V/Part] complex. Our results are mainly based on data from the IcePaHC corpus of Icelandic Texts (). We have also searched the Icelandic Gigaword Corpus (Risamálheild; see ) for further examples from Old Icelandic. The IcePaHC corpus contains 67 examples of P-incorporation from the 12th century until the mid-19th century when P-incorporation disappeared as a productive process. We will argue that this is due to a structural change in Icelandic which made bare PP complements impossible. The chapter is organized as follows. Section 9.2 reviews three types of preposition movement attested in the history of Icelandic, topicalization, Stylistic Fronting, and P-incorporation, including cases where the preposition is reduplicated. The basic restrictions on P-incorporation are outlined in Section 9.3. Section 9.4 discusses the diachronic development of P-incorporation and Section 9.5 reviews some remaining issues. Finally, the main results of the chapter are summarized in Section 9.6.

Jóhannes Gísli Jónsson and Brynhildur Stefánsdóttir, P-incorporation in the history of Icelandic. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Jóhannes Gísli Jónsson and Brynhildur Stefánsdóttir (2022). DOI: 10.1093/oso/9780198871538.003.0009

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

187

9.2 Moving prepositions 9.2.1 Introduction Leftward movement of prepositions was possible in Older Icelandic, a term which we will use in this chapter to refer to Icelandic from its earliest attestations until the end of the 19th century.¹ Preposition movement in Older Icelandic can be divided into three subtypes: (i) topicalization, (ii) Stylistic Fronting (SF), and (iii) P-incorporation. The movement of the preposition was optional in all cases. Moreover, (ii) and (iii) are closely linked since they involve roughly the same prepositions and disappear at approximately the same time, that is, in the middle of the 19th century. It is more difficult to determine how topicalization fits into the big picture because all the available evidence is a handful of examples from Old Icelandic.

9.2.2 Topicalization Ro¨gnvaldsson (2005: 621) gives the following three examples of topicalization of prepositions in Old Icelandic. For ease of exposition, the moved preposition and its complement are boldfaced here and elsewhere in the chapter: (1)

a. Á mun eg gera kosti að þér séuð hér… of will I make options that you are here ‘I will offer you to stay here…’ (Bjarnar saga Hı´tdælakappa, 110) b. Á ho¨fðu þeir unnið fleirum mo¨nnum on had they worked more.DAT men.DAT ‘They had wounded more men.’ (Þórðar saga kakala, 496) c. Við hefur Steingrı´mur enn leitað að þvo af at has Steingrı´mur again tried to wash off sér svı´virðinguna REFL the.disgrace ‘Steingrı´mur has again tried to wipe out the disgrace.’ (Reykdæla saga, 1757)

In addition to á ‘on’ and við ‘at’, as in (1a–c), it was also possible to topicalize af ‘from, off ’. This can be seen in (2a) (from Faarlund 1990: 98) and (2b) (from Haugan 2001: 504): ¹ The term Old Icelandic covers a shorter time period, i.e. Icelandic until the mid-16th century.

188

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

(2)

a. Af hefur þú ráðið mig brekvı´si við þig from have you advised me trouble.DAT to you ‘You have taught me not to give you trouble.’ (Laxdæla saga, 1582) b. En af verður að ráða nokkuð hverju vandræði but off must to solve somewhat each.DAT trouble.DAT ‘But each problem must be solved in some way.’ (Ljósvetninga saga, 1675)

Haugan (2001) found only two examples of topicalization in a corpus with 5477 tokens of the preposition af, namely (2a–b), and the IcePaHC corpus has only one example of preposition topicalization, shown in (3) below. These numbers indicate how infrequent topicalization of prepositions was in Old Icelandic. (3)

En við mun ég geta gert o¨llum þessum draumum but against will I be.able fight all.DAT these.DAT dreams.DAT er þig hefur dreymt fyrir hallærinu that you have dreamt for the.hard.times ‘But I will be able to counteract all the dreams…’ (mid-13th century, Jómsvı´kinga saga, 162)

Prepositions were not the only heads that could undergo topicalization in Old Icelandic. As shown by Ro¨gnvaldsson (2005: 620–1), it was also possible to topicalize non-finite main verbs. It is implausible to analyse such cases as movement of a verbal complement followed by remnant VP movement because Old Icelandic did not allow VP fronting (Ro¨gnvaldsson 1995). Particles could also undergo topicalization in Old Icelandic (Ro¨gnvaldsson 2005: 621) but since they do not take any complements, fronted particles can be analysed as full phrases rather than heads.

9.2.3 Stylistic Fronting The IcePaHC corpus has almost 40 examples where prepositions undergo SF. Some examples of this are shown in (4):² (4)

a. heim til fagurrar hallar er ´ı var borginni home to beautiful palace that in was the.city.DAT ‘home to a beautiful palace that was in the city’ (1861, Sagan af Heljarslóðarorrustu, 1152)

² We assume that fyrir ofan ‘above’ in (4c) is syntactically one preposition because ofan cannot assign case on its own.

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

189

b. að ekki sé það, er á dragi vora sæmd that not is it that from subtracts our.ACC honor.ACC ‘that it should not diminish our honor’ (early 14th century, Alexanders saga, 1355) c. En það hvolfið sem fyrir ofan er ho¨fuðið but that the.concave which above is the.head.ACC ‘But the concave which is above the head.’ (late 15th century, Jarlmanns saga og Hermanns, 409) In all these examples, the preposition moves to the left of the finite verb in a subject relative. Since SF requires a subject gap, it is not surprising that most of the fronted prepositions in IcePaHC are in subject relatives. A vast majority of these examples contain one verb; only three examples have a non-finite verb in addition to a finite auxiliary. One of these examples is (5): (5)

hver eð yfir var settur allar hennar fjárhirslur who that over was placed all.ACC her coffins.ACC ‘who was in charge of all her coffins’ (1540, Nýja Testamenti Odds Gottskálkssonar, 253.462)

SF obeys locality conditions which were originally discussed in terms of an accessibility hierarchy (Maling 1990). Jónsson (1991) argues that if more than one element in a particular clause is a possible candidate for SF, only the highest one, defined by asymmetric c-command, is able to move. This is unproblematic in (4a–c) because finite verbs do not move by SF and thus do not block SF of the preposition. By contrast, since non-finite main verbs and participles undergo SF and asymmetrically c-command a preposition inside a PP complement, they are expected to prevent SF in (5). To account for the grammaticality of examples like (5), we propose that the preposition incorporates into the verb/participle and subsequently excorporates out of the [P+V/Part] complex. P-incorporation creates a structure where the preposition and the verb/participle c-command each other and this enables the preposition to move further by SF.³ This yields examples like (5), as shown in the following derivation:

³ Presumably, the verb/participle can also undergo SF in this configuration but this will in most cases create a string that is indistinguishable from a string where the preposition stays in situ.

190

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

(6)

a. sem var settur [PP yfir allar hennar fjárhirslur] P-incorporation b. sem var [yfiri settur] [PP ti allar hennar fjárhirslur] SF of the preposition (excorporation) c. sem yfiri var [ti settur] [PP ti allar hennar fjárhirslur]

The claim that incorporation into a verb/participle is required as a first step for prepositions undergoing SF is supported by the fact that this step is constrained by the HMC as well as the Adjunct Condition (see further in Section 9.3). Linking P-incorporation and SF of prepositions in this way is also supported by the fact that both kinds of movement involve roughly the same prepositions and disappear at approximately the same time. A potential problem for our analysis is that excorporation is widely assumed to be excluded (see Baker 1988: 73 and much subsequent work; but see Roberts 2010 for a somewhat different view). The main motivation for this view is that if excorporation is freely available, it can be used to evade the HMC of Travis (1984) in cases where the HMC must hold, for example in verb movement to T or C. This is exemplified in (7): (7)

a. Hann hefuri ekki ti lesið bókina he has not read book.the ‘He has not read the book.’ b.* Hann lesi ekki hafa/haft ti bókina he reads not have/had book.the

The auxiliary moves to T without problems in (7a) but (7b) is ruled out as a violation of the HMC as the main verb crosses the auxiliary, thereby skipping an intervening head. However, if the main verb can incorporate into the auxiliary and then excorporate by moving to T, (7b) is expected to be grammatical, contrary to fact. The derivation in (6) is crucially different from this hypothetical derivation in that the intermediate step in the first derivation, the incorporation of the preposition, is independently attested. By contrast, there is no independent evidence that main verbs incorporate into auxiliaries in Icelandic. Thus, allowing excorporation in cases like (6) does not create any empirical problems.

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

191

Table 9.1 The diachronic development of P-incorporation in Icelandic Century

Examples

Examples per 10,000 words

12th 13th 14th 15th 16th 17th 18th 19th 20th

1 10 18 16 3 11 6 2 0

0.22 1.00 1.53 1.33 0.33 0.86 0.55 0.16 0.00

9.2.4 Incorporation The IcePaHC corpus has 67 examples of P-incorporation. Table 9.1 shows how these examples distribute across centuries. The number for the 20th century is 0 because contemporary Icelandic has no examples showing productive use of P-incorporation. Note that the numbers in Table 9.1 only include prepositions with nominal objects but prepositions taking clausal complements could also incorporate in Older Icelandic as shown by Hróarsdóttir (2000). The numbers presented here include examples where an incorporated preposition is written as prefixes on its host.⁴ As discussed in Section 9.5.1, P-incorporation is essentially the same phenomenon whether the preposition forms a morphological word with the verb/participle or not. The numbers in Table 9.1 show that P-incorporation with nominal objects was rather uncommon throughout the history of Icelandic. This is consistent with Ro¨gnvaldsson’s (1995) claim that five very common prepositions, namely ´ı ‘in/into’, á ‘on’, til ‘to’, með ‘with’, and við ‘by, at’, are adjacent to their objects in more than 99 per cent of the cases in Old Icelandic. Let us now examine some representative examples of P-incorporation from different times in the history of Icelandic: (8)

a. kvaðst eigi til kominn þess ´ı heim þenna said not for come it.GEN in world this ‘(He) said he had not come to this world for that’ (mid-12th century, Hómilı´ubók, 113)

⁴ P-incorporation was also possible into a handful of adjectives in Old(er) Icelandic but we have no example where the incorporated preposition is written as an adjectival prefix.

192

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

b. Þetta muntu af hljóta fénu this will.you of get money.the.DAT ‘You will get this (cut) of the money.’ (mid-15th century, Bandamanna saga, 635) c. Illugi bað hann þo¨kk fyrir hafa góð boð Illugi told him thank.ACC for have good.ACC offers.ACC ‘Illugi thanked him for his good offers.’ (around 1650, Illuga saga Tagldarbana, 1879) d. þar sem aðrir voru frá gengnir leitinni where others were from walked search.the.DAT ‘where others had stopped searching’ (1850, Piltur og stúlka, 109) The host of P-incorporation in these examples is either a non-finite verb (8b–c) or a participle (8a), (8d). As discussed in Section 9.5.2 below, P-incorporation into finite main verbs is theoretically possible but, due to verb movement to T, this can only be detected when something intervenes between the incorporated preposition and the object of the preposition. P-incorporation in Older Icelandic has no effect on case marking and it is independent of the transitivity of the relevant verb/participle. A preposition can incorporate into intransitive verbs/participles as well as transitive verbs and the object stranded by the incorporation bears accusative, dative, or genitive case depending on the case selected by the preposition. All of this is exemplified in (8a–d). The case properties associated with P-incorporation in Older Icelandic are replicated in Classical Greek (see Miller 1993: 118–19 for examples) but the polysynthetic languages discussed by Baker (1988) behave differently in that prepositions can only incorporate into verbs that assign structural case. Thus, the stranded object gets structural case from the verb as a result of incorporation and thereby acquires many properties of a direct object. The fact that these languages have no morphological case, unlike Icelandic and Classical Greek, may explain why P-incorporation works differently in these two sets of languages but this issue will not be explored here. Since P-incorporation involves left adjunction into a verb or participle, an incorporated preposition should immediately precede the verb/participle. Hence, a phrasal category cannot intervene between an incorporated

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

193

preposition and its host, for example the direct object (þo¨kk) in (8c), as shown in (9).⁵ (9)*

Illugi bað hann fyrir þo¨kk hafa góð Illugi told him for thanks have good.ACC

boð offers.ACC

Interestingly, the direct object in (8c) moves to the immediate left of the incorporated preposition and the direct object in (8b) is topicalized, that is, neither of the objects stays in base position after the verb. These and similar examples with transitive verbs suggest that there is a strong tendency for direct objects not to intervene between the verb/participle and the object stranded by P-incorporation. In fact, we have only found one such example in Old Icelandic: (10)

báðu menn af reka hestana vellinum asked people from drive horses.the.ACC field.the.DAT ‘(They) asked people to drive the horses away from the field’ (Þorgils saga skarða, 684)

As shown in Section 9.5.2, a direct object can occur after an incorporated preposition, provided that the (finite) verb excorporates to T. Importantly, there is only one item between the incorporated preposition and its object in this case, namely the direct object. Apart from exceptional cases like (10), the generalization that emerges from our study is that an incorporated preposition can only be separated from its object by one item, the host of the incorporation or the direct object of the verb (if the host excorporates).

9.2.5 Reduplication A preposition that undergoes leftward movement can be reduplicated in Old Icelandic (Jónsson 2008) but this is so infrequent that IcePaHC has no example of this. It appears that P-reduplication has no semantic or pragmatic import. This can be seen in triplets like (11a–c) where the same expression (kenna kulda af ‘sense animosity from’) features a preposition in situ (11a), an incorporated preposition (11b), and a preposition that is both ⁵ Hróarsdóttir (2000: 146, 244) makes the same claim about incorporated particles in Older Icelandic. Thus, out of 186 examples of particle incorporation, she only found one where something intervenes between the particle and the main verb.

194

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

incorporated and reduplicated (11c) without any observable difference in meaning: (11)

a. Álfur þóttist og kenna kulda af Óspaki Álfur thought also sense animosity from Óspakur.DAT ‘Á lfur also felt that he sensed animosity from Óspakur.’ (Eyrbyggja saga, 611) b. óvinir hans þóttust heldur kulda af kenna enemies his thought rather animosity from sense ráðum hans advice.DAT his ‘His enemies sensed some animosity from his advice.’ (Eyrbyggja saga, 547) c. hann þykist kulda af kenna af skiptum he feels animosity from sense from dealings.DAT yðrum your.DAT ‘he feels animosity from his dealings with you’ (Laxdæla saga, 1591)

The Icelandic Gigaword Corpus has at least 30 examples of P-reduplication in Old Icelandic; most of them involve P-incorporation but some feature topicalization or SF as in (12): (12)

a. og af gekk skinnið af knúnum and off went the.skin off knuckles.the.DAT ‘and the skin was peeled off the knuckles’ (Gı´sla saga Súrssonar, 867) b. til gluggs er á var á húsinu to window that on was on house.the.DAT (Grettis saga, 1089) ‘to a window that was on the house’

Reduplication is conditioned by the movement of the preposition out of its base position which creates a copy inside PP that can be overtly realized. Movement is not a sufficient condition, however, although only moved prepositions can be reduplicated. In fact, only prepositions taking nominal objects can undergo reduplication in Old Icelandic as this is not attested with prepositions taking clausal or null objects. This suggests that mophological case is a crucial factor here (but see Jónsson 2008 for a different analysis). Hence, it may not be a coincidence that P-reduplication is also found in a case-rich language like Classical Greek (see Miller 1993: 119).

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

195

9.3 Restrictions on P-incorporation In this section, we will take a closer look at the restrictions that hold of P-incorporation in Older Icelandic. As will be discussed in more detail in this section, there are two important syntactic principles that constrain P-incorporation, the HMC discussed in Section 9.3.1 and the Adjunct Condition discussed in Section 9.3.2.

9.3.1 The HMC As argued at length by Baker (1988), incorporation across languages is subject to the HMC. This means that a head can only incorporate into the next ccommanding head. For instance, a preposition cannot move out of a PP inside a DP because such a movement would inevitably involve an intervening Dhead skipped by the preposition. As expected, we have not found any example of this kind in Older Icelandic. The HMC also rules out P-incorporation in Older Icelandic if the preposition has to cross intervening heads inside an extended PP. It has become standard in generative syntax to assume that PPs may form an articulated structure with various functional heads (see Koopman 2000; den Dikken 2010; and Svenonius 2010; among others), including Path hosting directional particles like upp ‘up’, niður ‘down’, út ‘out’, and inn ‘in’. P-incorporation across such items is ruled out by the HMC. This is exemplified in (13c) below where the preposition af moves across upp. The example in (13a) shows the complex prepositional expression upp af ‘up from’ following a non-finite verb whereas (13b) illustrates that upp can incorporate into the non-finite verb. (13)

a. vil ég ekki að þú rı´sir upp af stólinum want I not that you rise up from chair.the.DAT ‘I don‘t want you to rise up from the chair.’ (Fóstbræðra saga, 831) b. Þá vildi Þorgrı´mur upp rı´sa af stólinum then wanted Þorgrı´mur up rise from chair.the.DAT ‘Then Þorgrı´mur wanted to rise up from the chair’ (Fóstbræðra saga, 824) c.* Þá vildi Þorgrı´mur af rı´sa upp stólinum then wanted Þorgrı´mur from rise up chair.the.DAT

196

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

Importantly, (13c) cannot be ruled out as a violation of the Adjunct Condition because the complex PP here is not an island for extraction. This is shown by the possibility of moving the directional particle upp out of the PP, as in (13b). Hróarsdóttir (2000: 243–4) claims that, unlike full PPs, prepositions cannot occur immediately to the left of non-finite auxiliaries in Older Icelandic. This is expected since the HMC prevents a preposition from bypassing a main verb by incorporating into a non-finite auxiliary. By contrast, movement of a PP to the left of a non-finite auxiliary is not blocked by the HMC which does not hold of phrasal movement. There are some potential counterexamples to this generalization. The IcePaHC corpus has one example of a preposition preceding the infinitive hafa ‘have’, namely (14a). Additional examples can be found in The Gigaword Corpus, for example (14b): (14)

a. að konungsdóttir myndi hann that king‘s.daughter would he

laust þykja á hafa loosely find on have

haldið merkinu Magnúss konungs held the.banner.DAT Magnus.GEN king.GEN ‘that the kings’ daughter would think that he had not carried the banner of king Magnús with enough vigor’ (late 13th century, Morkinskinna, 925) b. og sagðist af hafa brugðið hans boði and said from have deviated his command.DAT ‘and said that he had disobeyed his commands’ (Þorsteins þáttur skelks, 2292) In both of these examples, the preposition is within an infinitival clause selected by the matrix verb, þykja ‘find, think’ in (14a) and segjast ‘say, claim’ in (14b). A preposition can also incorporate into the main verb in such examples as in (15): (15)

þykist hann þungt hafa af beðið yðrum claims he heavy have from suffered your.DAT skiptum exchanges.DAT ‘he claims to have suffered greatly from his dealings with you’ (Finnboga saga ramma, 666)

It looks like the examples in (14) involve SF. If this is the case, the preposition incorporates first into the main verb and then moves to a higher position in

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

197

the clause. The subject gap condition for SF is fulfilled in (14a) as the subject position has been vacated by movement of the pronoun hann ‘he’ into the matrix clause. Things are less clear in (14b) because the infinitival clause appears to have a controlled null subject (PRO), but this may have been a sufficient condition for SF in Old Icelandic. Another kind of potential counterexample can be seen when a preposition ends up in front of the infinitive marker að as in (16): (16)

a. að hann girnist… fyrir að ráða sjálfu himinrı´ki that he desires for to decide self.DAT heaven.DAT ‘that he desires to rule heaven itself ’ (early 14th century, Alexanders saga, 1127) b. er ég hlýt frá að skýra sorg þeirri sem… as I must about to tell sorrow.DAT that.DAT which ‘as I must recount the sorrow that…’ (late 17th century, So¨guþáttur af Ármanni og Þorsteini gala, 113.775)

The IcePaHC corpus has only four examples where an incorporated preposition precedes the infinitive marker. Because of this low number, it may be a coincidence that the IcePaHC corpus has no examples of a preposition intervening between the infinitive marker and the infinitive verb. However, in view of Hróarsdóttir’s (2000: 250) observation that incorporated particles always precede the infinitival marker in Older Icelandic, it is highly likely that the unattested order [að–preposition–infinitive] is excluded in Older Icelandic. Moreover, since there is a strict adjacency requirement in Icelandic between að and the infinitival verb, as seen by the near absence of split infinitives, this is clearly what is expected.⁶ To account for the grammaticality of examples like (16a–b), it is reasonable to assume that the infinitive marker and the infinitive verb form a complex head that cannot be broken up by another syntactic element. A preposition can incorporate into this complex head but that yields the order [preposition– að–infinitive] as in (16a–b).

⁶ Possibly the most common adverb that can split an infinitive in Modern Icelandic is bara ‘just’, but this adverb has an unusally free distribution compared to other adverbs in the language.

198

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

9.3.2 The Adjunct Condition Since adjuncts are islands for extraction, P-incorporation out of a PP-adjunct should be excluded. As Baker (1988: 236) notes, it is not easy to test this prediction since the distinction between PP complements and PP adjuncts is often hard to make. Still, there are many undisputed cases and our results from the IcePaHC corpus show that there are no clear examples of P-incorporation out of adjunct PPs. Note also that since preposition stranding is not possible out of adjunct PPs, the expectation is that P-incorporation in Older Icelandic should be possible in roughly the same environments as preposition stranding is in contemporary Icelandic.⁷ As far as we know, this is borne out by the data. The following examples show adjunct PPs in Old Icelandic, that is, a temporal PP, a locative PP that is not accompanied by a locative verb, and a comitative PP: (17)

a. að langskip hafði komið um haustið til eyjanna that war.ship had come in the.fall. ACC to the.islands ‘that a war ship had come to the islands that fall’ (Egils saga, 408) b. Var hann veginn við was he killed by

húsin sjálf the.houses.ACC self.ACC (Þórðar saga kakala, 506)

c. Hákon jarl hafði þar setið með fjo¨lmenni miklu Hákon earl had there sat with crowd.DAT big.DAT ‘Earl Hákon had been there with a lot of people’ (Haralds saga gráfeldar, 147) The PPs in these examples are not complements of the verb. Morever, judging by the examples found in the IcePaHC corpus, P-incorporation is ruled out in all these cases: (18)

a.* að langskip hafði um komið haustið til eyjanna that war.ship had in come the.fall.ACC to the.islands b.* Var hann was he

við by

veginn killed

húsin sjálf the.houses.ACC self.ACC

c.* Hákon jarl hafði þar með Hákon earl had there with

setið sat

fjo¨lmenni miklu crowd.DAT big.DAT

⁷ As discussed in Section 9.5.3, preposition stranding was extremely rare in Old Icelandic outside of relative clauses. Thus, we have to compare here Older Icelandic and contemporary Icelandic.

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

199

The distinction between argument and adjunct PPs cuts across different uses of the same preposition. For instance, although af ‘from’ is the most common preposition in P-incorporation in Older Icelandic, it has many uses where P-incorporation seems to be impossible. This is exemplified in (19) and (20): (19)

a. Ertu are.you

mjo¨g lofaður af mo¨nnum highly praised by people.DAT (Bandamanna saga, 2)

b. Hafði hann sprungið af harmi eftir sonu sı´na had he blown.up from sorrow.DAT after sons REFL ‘He had died from sorrow after his sons’ death.’ (Gunnars saga Keldugnúpsfı´fls, 1157) c. að that eftir by (20)

þetta væri this was lo¨gum law

meir gert af kappi en more done by aggression.DAT than (Grettis saga, 1023)

a.* Ertu mjo¨g af lofaður mo¨nnum are.you highly by praised people.DAT b.* Hafði hann had he c.* að that eftir by

af sprungið harmi eftir sonu sı´na from blown.up sorrow.DAT after sons REFL

þetta væri this was lo¨gum law

meir af gert kappi en more by done aggression.DAT than

Thus, incorporation out of adjunct PPs denoting a passive agent, reason PPs, and manner PPs is excluded. This ban is completely general and independent of the presence of any particular preposition. As discussed in Section 9.2.3, prepositions that undergo SF must first incorporate into a verb/participle, but the Adjunct Condition prevents prepositions heading adjunct PPs from moving by SF. By contrast, adjunct PPs can undergo SF, as shown by the following examples from Old Icelandic: (21)

a. hurðunni sem um aftaninn hafði verið um búið door.the which in evening.the had been prepared ‘the door which had been left open that evening’ (Gı´sla saga Súrssonar, 868)

200

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

b. svo sem með spjóti hefði lagt verið as if with spear.DAT had attacked been ‘as if a spear had been used as a weapon’ (Gunnars þáttur Þiðrandabana, 2141) Since both of these examples have a subject gap, there is no doubt that the fronting of the PPs instantiates SF rather than topicalization.

9.4 The loss of P-incorporation In this section, the diachronic aspect of P-incorporation in Older Icelandic will be discussed. Possible triggers for the loss of P-incorporation in the 19th century will be examined as well as the nature of this change within a generative framework. We will put forward a particular hypothesis for the second issue in Section 9.4.1 but our remarks about the first issue in Section 9.4.2 will be rather speculative.

9.4.1 The nature of the change To understand the nature of the change under discussion, we believe it is necessary to consider the structural architecture of PPs across languages, as shown by the sequence of functional heads like Path, Place, and Axial Part above the preposition (Koopman 2000; den Dikken 2010; Svenonius 2010) as in (22) below: (22)

[PathP [PlaceP [AxPart [PP …]]]]

As discussed in Section 9.3.1, P-incorporation is excluded if the preposition moves across a directional particle in Path. More generally, an incorporating preposition cannot bypass any functional head inside the extended PP as this would run foul of the HMC. This raises the question if a preposition can incorporate into a higher functional head X inside PP, forming a complex head [P+X], which then incorporates into a verb/participle. According to the theoretical literature, the answer is no. As den Dikken (2010) points out in his discussion of incorporation of P-elements in Dutch (see also Koopman 2000), this type of derivation violates improper movement which bans movement from an A-position to an Ā-position and then back to an Aposition. The underlying assumption behind this is that lexical heads count as A-positions but functional heads as Ā-position (Li 1990). Baker (1996: 284) refers to this as the Proper Head Movement Generalization, accepting

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

201

its empirical validity while also expressing doubts about its theoretical status. The main empirical motivation for this restriction comes from the observation that verb incorporation always features a verb root rather than an inflected verb displaying tense, aspect, or agreement morphology (Li 1990). This follows straightforwardly from improper movement because a verb cannot inflect unless the functional head(s) licensing the relevant inflectional features are present. Once a verb has moved to an inflectional head it cannot incorporate into a higher verb without violating improper movement. Importantly, this restriction to bare roots is not limited to verb incorporation. It is also a robust cross-linguistic generalization that nouns undergoing incorporation tend to be bare roots. It is only in some languages that inflected nouns can incorporate (see Olthof 2020 and references cited there). The conclusion then is that prepositions cannot incorporate from a structure like (22) or an extended PP of any kind. P-incorporation is only possible out of bare PPs, that is, PPs where the preposition occupies the highest head position in the absence of any functional projections. This means that bare PPs were possible as complements of verbs or participles in Older Icelandic. Moreover, under this view, the loss of P-incorporation in the history of Icelandic is naturally analysed as a loss of bare PPs as a grammatical option. While there are different ways of technically implementing this basic idea, we hypothesize that the crucial component of this change is the categorizing head p, taking PP as a complement. This functional head was optionally present until the 19th century when it became an obligatory part of all PPs. This change in the grammar of Icelandic made P-incorporation impossible for reasons that we discussed above. Until that point in time, it was possible to generate PPs without p. These two options are illustrated in (23): (23)

a. [PP [P' P [DP ]]] (old grammar) b. [pP [p' p [PP [P' P [DP ]]]]] (old and new grammar)

This analysis treats prepositions in Icelandic as lexical categories on a par with nouns, verbs, and adjectives, even though they have a lot in common with functional categories, for example as closed class items that do not inflect in any way even though Icelandic is a highly inflectional language. The status of prepositions across languages is a highly complex and controversial issue (see Baker 2003: 303–25, Botwinik-Rotem 2004, and den Dikken 2010 for relevant discussion), but we will focus on Icelandic here and bring out three arguments that point to strong similarities between verbs and prepositions in Icelandic, suggesting that the latter are lexical rather than functional.

202

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

First, as discussed in Section 9.5.3, Icelandic has allowed preposition stranding outside of relative clauses since the 16th century. This is unexpected under a functional view of prepositions because functional heads are known to prohibit stranding by their complements. In fact, there is no clear example of such stranding in Icelandic that we are aware of. Second, prepositions are like transitive verbs in many ways with respect to case assignment. Both classes can take accusative, dative, or genitive complements and, as in other case-rich languages, there is very little evidence for prepositions as pure case markers in Icelandic. In view of this, it is plausible to assume that the category-defining head p plays a role in the case marking of prepositional objects, similar to ν in extended VPs (Svenonius 2003). Third, as discussed by Hjartardóttir (1993), Icelandic allowed null objects of verbs and prepositions until the 18th century when both types of objects more or less disappeared. This is a strong argument in favour of prepositions as a lexical category because it is unusual for a functional head to take a null complement.

9.4.2 Contributing factors Under the analysis presented above, the loss of P-incorporation is a direct consequence of the loss of bare PP complements in the grammar of Icelandic, that is, the obligatory presence of categorizing p in all PPs. This is clearly a change towards more uniformity between prepositions and other lexical categories, especially verbs. Thus, it is interesting to note, as Baker (2008b) does, that there is a strong tendency for verbs and adpositions across languages to be consistently head-final or head-initial with respect to their complements. This cross-categorial harmony can be explained through parameter hierarchies if language learners adopt a learning strategy whereby the most general option consistent with the primary linguistic data is chosen (Roberts 2016). For the case at hand, this means that language learners will assume that all PPs have a p-head, just like all VPs have a ν-head, until they are exposed to clear evidence (in the form of P-incorporation) to the contrary. Be that as it may, P-incorporation survived for seven centuries in Icelandic as there must have been sufficiently robust evidence for incorporating prepositions in the linguistic input that language learners were exposed to. Still, as shown in Table 9.1, attested examples of P-incorporation with nominal objects throughout the history of Icelandic are quite few. The numbers provided Hróarsdóttir (2000: 145) are higher, that is, 102 examples from the 14th to the

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

203

19th century in a corpus of 4875 sentences, but only two of these examples involve a nominal object. The number 102 includes cases where the incorporated preposition takes a clausal or a null object, although the inclusion of examples of the latter kind is potentially problematic because they could be analysed as full PPs undergoing leftward movement. P-incorporation was lost in the 19th century, a time that coincides with the general loss of OV-orders in Icelandic. In fact, Hróarsdóttir (2000) treats P-incorporation as a kind of OV-order even though the preposition itself is not a complement of the verb and undergoes incorporation rather than phrasal movement. Still, these two phenomena are interconnected if Hróarsdóttir (2009) is correct that a crucial part of OV-loss in Icelandic was the tendency for objects denoting new information to occur postverbally towards the end of the OV-period. Although terms like old and new information are rarely used for prepositions, it seems clear that a postverbal position associated with new information will hamper P-incorporation since prepositions typically express information that is not known or inferable from the context.

9.5 Some remaining issues This section addresses some remaining issues relating to P-incorporation in Older Icelandic. We begin in Section 9.5.1 by looking at examples where Pincorporation results in word formation and then we move on in Section 9.5.2 to the issue of excorporation of the finite verb, leaving behind an incorporated preposition and a direct object. Finally, we argue in Section 9.5.3 that P-incorporation can only take place from a VO-base, thus providing evidence for the view that OV-orders in Older Icelandic are derived by leftward movement.

9.5.1 Incorporation as word formation An incorporated preposition is usually written as a whole word in the IcePaHC corpus but there are 11 examples where it is written as part of the following verb or participle. These will be referred to here as independent prepositions and prefixal prepositions, respectively. The following prepositions can be prefixal in the IcePaHC corpus: að ‘at’, af ‘from’, á ‘on, of ’, frá ‘from’, með ‘with’, and til ‘to’. The oldest example of a prefixal preposition is from the 13th century but the rest are from the 17th and 18th centuries:

204

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

(24)

a. eigi vera verr tilkominn rı´kis en Haraldur not be worse to.come throne.GEN than Haraldur bróðir hans brother his ‘not to be less deserving of the throne than his brother Haraldur’ (late 13th century, Morkinskinna, 1850) b. þá menn skulu farminn afleggja skipunum when people should freight.the.ACC off.put ships.the.DAT ‘when the freight should be loaded off the ships’ (1628, Ólafur Egilsson, 698) c. ef hann áhangir honum if he onto.clings him.DAT ‘if he adapts to him’ (1720, Vı´dalı´nspostilla, 180)

As expected, there is no distinction between prefixal and independent prepositions in that the incorporation is in both cases subject to the same syntactic restrictions. Thus, the set of prefixal prepositions found in the IcePaHC corpus is a subset of the independent prepositions. Moreover, the IcePaHC corpus has minimal pairs exhibiting prefixal vs. independent prepositions like (25a) vs. (25b): (25)

a. og varst frá skilinn o¨llum o¨ðrum syndugum and were from separated all.DAT other.DAT sinful.DAT ‘and you were separated from all other sinful people’ (1593, Eintal sálarinnar við sjálfa sig, 437) b. þótt hann væri… og fráskilinn syndurunum although he was and from.separated sinners.the.DAT (1720, Vı´dalı´nspostilla, 275)

One may wonder if this is simply a spelling convention that does not reflect a linguistic distinction between prefixal and independent prepositions.⁸ However, verb movement to T shows that a prefixal preposition forms a morphological word with its host. Thus, if such a preposition incorporates into a verb that subsequently moves to T to check its tense features, the preposition moves with the verb, as in (24c). This is not the case with independent prepositions as the IcePaHC corpus has no examples where they move with the finite verb in a subject-initial clause like (24c). ⁸ We have ascertained that the spelling of the examples of P-incorporation in the IcePaHC corpus matches the spelling of the relevant Old Icelandic manuscripts. Hence, the data presented here cannot be challenged on philological grounds.

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

205

The conclusion is that P-incorporation in Older Icelandic involves morphological word formation whenever the preposition is written as a prefix. This can also be seen in Modern Icelandic with prefixed adjectives like viðstaddur ‘present (at)’ (from the preposition við ‘at’ and the adjective staddur ‘present’). This word looks like a case of obligatory P-incorporation because viðstaddur assigns accusative case, a syntactic property that is due to the presence of við, as all the adjectives taking accusative objects in Icelandic have this prefix.⁹

9.5.2 Excorporation The IcePaHC corpus has some examples where a direct object of the verb occurs between an incorporated preposition and its object. All of these examples have a finite main verb which moves to T, leaving the preposition behind. This is exemplified in (26):¹⁰ (26)

a. og hefir Oddur af virðing málunum and has Oddur from respect.ACC affairs.the.DAT ‘and Oddur gains respect from the affairs’ (mid-15th century, Bandamanna saga yngri, 555) b. Kom hann á ho¨ggi einn fylgdarmann managed he on blow.DAT one.ACC follower.ACC Gunnars Gunnar.GEN ‘He was able to hit one of Gunnar‘s men.’ (Grettis saga, 991) c. og taki af spjótið af skaftinu and take off spear.the.ACC off shaft.the.DAT ‘and take the spear off the shaft’ (Fljótsdæla saga, 703)

This suggests that the preposition incorporates into the finite verb, followed by excorporation of the finite verb to T. This is shown in the following derivation for (26a):

⁹ In addition to viðstaddur, there is also viðriðinn ‘involved in’ and viðloðandi ‘associated with’. ¹⁰ As far as we know, the P-reduplication in (26c) has no special significance in examples of this kind.

206

(27)

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

a. Oddur hefir [DP virðing] [PP af málunum] P-incorporation (into the finite V ) b. Oddur [af i hefir] [DP virðing] [PP ti málunum] Finite verb excorporates to T c. Oddur hefirj [af i tj ] [DP virðing] [PP ti málunum] Finite verb moves to C d. Hefirj Oddur tj [afi tj ] [DP virðing] [PP ti málunum]

The assumption here is that a preposition cannot incorporate into a direct object of the verb. This is correct as shown by the fact that examples like (28), which can only be analysed as P-incorporation into the direct object, are not attested in Older Icelandic: (28)*

og mun Oddur hafa af virðing and will Oddur have from respect.ACC

málunum affairs.the.DAT

This is excluded because the trace of the incorporated preposition is not ccommanded by the preposition. However, we are not aware of any principle of syntax that rules out examples like (29), where a preposition has moved across an object to incorporate into a non-finite verb: (29)

og mun Oddur af hafa virðing málunum and will Oddur from have respect.ACC affairs.the.DAT

As discussed in Section 9.2.4, there is a very strong tendency for direct objects not to intervene between the verb/participle and the object stranded by Pincorporation. We suspect that the problem with examples like (29) is that the string between the incorporated preposition and its object, that is, the verb + direct object, is too heavy prosodically, but this merits further investigation.

9.5.3 Incorporation from a VO-base Working within the anti-symmetry framework of Kayne (1994), Hróarsdóttir (2000, 2008b) argues for a strictly head-initial VP in Older Icelandic despite the fact that various OV-orders are fairly common.¹¹ This means that all OVorders are derived by leftward movement of the verbal complement. The result is that prepositions cannot incorporate from a preverbal position because they would inevitably move from a PP in a specifier position c-commanding the verb/participle. Consequently, the incorporated preposition would fail to ¹¹ Similar arguments have also been made for Old and Middle English; see Roberts (1997) and van der Wurff (1997).

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

207

c-command its trace. By contrast, incorporation from a postverbal position should be fine if the PP is a complement of the verb/participle and this is consistent with all the examples of P-incorporation that we have found. Applying P-incorporation from an OV-order would create a string where the stranded object precedes a [P+V/Part] complex. Searching for examples with the order [DP–P–V/Part] reveals that IcePaHC has none and the Tagged Icelandic Corpus has only a handful, including (30): (30)

að hann hefði kú við tekið af þeim that he had cow.ACC with taken from them ‘that he had accepted a cow from them’ (Reykdæla saga, 1743)

Although this example looks superficially like P-incorporation from a PP–V order, examples like (30) can be derived from a V–PP order in two steps, as shown in (31) below, that is, P-incorporation and object scrambling out of the PP: (31)

a. að hann hefði tekið [PP við [DP kú]] af þeim (V–PP) P-incorporation b. að hann hefði [viði tekið] [PP ti [DP kú]] af þeim object scrambling (P-stranding) c. að hann hefði kúj [viði tekið] [PP ti [DP tj ]] af þeim

The last step here may look suspicious as P-stranding was extremely rare in Old Icelandic outside of relative clauses (Haugan 2001; Delsing 2003; Jónsson 2008). The crucial point here is that P-incorporation makes P-stranding possible (see Baker 1988: 260 on Chichewa) as islands cannot be headed by traces (Boškovic´ 2011; Stepanov 2012). In fact, all the examples of Pstranding outside of relative clauses in Old Icelandic that we know of involve P-incorporation. Faarlund (2004: 233) provides (32a) as an example of Pstranding with án ‘without’. Further examples with án, as in (32b–c), can be found by searching The Gigaword Corpus: (32)

a. Þess máttu Gautar illa án vera this.GEN could Gauts badly without be ‘The Gauts could hardly be without it.’ (Heimskringla, Ólafs saga helga, 306) b. mættum vér þess vel án vera could we that.GEN well without be ‘We could easily be without that.’ (Heiðarvı´ga saga, 1373)

208

JÓHANNES GÍSLI JÓNSSON AND BRYNHILDUR STEFÁNSDÓTTIR

c. aldrei má eg þı´n án vera never can I you.GEN without be ‘I can never be without you.’ (Heimskringla, Magnússona saga, 742) All these examples display incorporation of the preposition án into the infinitive vera ‘be’. Moreover, the object of án is a pronoun in all cases, suggesting that P-stranding in Old Icelandic is confined to prosodically light elements like pronouns and bare nouns as in (30).¹² This restriction fits nicely with the view that the placement of the object of the preposition is due to leftward movement as light elements tend to move to the left in Old Icelandic (Hróarsdóttir 2000: 130–4).

9.6 Conclusions In this chapter, we have shown that P-incorporation into verbs and participles in Older Icelandic obeys the same syntactic principles as incorporation in other languages, the HMC, and the Adjunct Condition. These restrictions are independent of whether the [P+V/Part] complex forms a morphological word or not. We have also argued that an incorporated preposition can excorporate from the [P+V/Part] complex by undergoing SF. A finite verb can excorporate and leave behind an incorporated preposition followed by a direct object. However, if the finite verb forms a morphological word with the preposition, the prefixed verb moves as one unit to T to check its tense features. We hypothesize that P-incorporation in Older Icelandic was possible by virtue of the fact that verbs and participles could take bare PP complements. This option was lost in the 19th century and the old grammar was replaced by a new grammar where the presence of the functional head p was required in all PPs. As a result, P-incorporation became impossible as it would inevitably violate the HMC or improper movement. Finally, we have shown that P-incorporation cannot be derived from an OVstructure, that is, a configuration where the PP precedes the verb or participle. This restriction can be derived if OV-orders involve leftward movement of the complement to a position c-commanding the verb or participle.

¹² The leftward movement of the pronouns in (32) is probably strongly preferred to no movement given the paucity of examples where P-incorporation strands a pronoun.

P-INCORPORATION IN THE HISTORY OF ICEL ANDIC

209

Texts and corpora Heimskringla (1991). Bergljót S. Kristjánsdóttir, Bragi Halldórsson, Jón Torfason, and Örnólfur Thorsson (eds). Reykjavı´k: Mál og menning. (Haralds saga gráfeldar, Ólafs saga helga, Magnússona saga.) Íslendinga so¨gur (1985–6). Bragi Halldórsson, Jón Torfason, Sverrir Tómasson, and Örnólfur Thorsson (eds). Reykjavı´k: Svart á hvı´tu. (Bandamanna saga, Bjarnar saga Hı´tdælakappa, Egils saga, Eyrbyggja saga, Finnboga saga ramma, Fljótsdæla saga, Fóstbræðra saga, Gı´sla saga Súrssonar, Gunnars saga Keldugnúpsfı´fls, Gunnars þáttur Þiðrandabana, Grettis saga, Heiðarvı´ga saga, Laxdæla saga, Ljósvetninga saga, Reykdæla saga, Þorsteins þáttur skelks.) Steingrı´msson, Steinþór, Sigrún Helgadóttir, Eirı´kur Ro¨gnvaldsson, Starkaður Barkarson, and Jón Guðnason (2018). Risamálheild: A very large Icelandic text corpus. [The Icelandic Gigaword Corpus.] Proceedings of LREC 2018, 4361–6. Myazaki, Japan. Sturlunga saga (2010). Bergljót S. Kristjánsdóttir, Bragi Halldórsson, Gı´sli Sigurðsson, Guðrún Ása Grı´msdóttir, Guðrún Ingólfsdóttir, Jón Torfason, Sverrir Tómasson, and Örnólfur Thorsson (eds). Reykjavı´k: Mál og menning. (Þorgils saga skarða, Þórðar saga kakala.) Wallenberg, Joel C., Anton Karl Ingason, Einar Freyr Sigurðsson, and Eirı´kur Ro¨gnvaldsson (2011). Icelandic Parsed Historical Corpus (IcePaHC). Edition 0.9. . (Alexanders saga, Bandamanna saga yngri, Eintal sálarinnar við sjálfa sig, Hómilı´ubók, Illuga saga Tagldarbana, Jarlmanns saga og Hermanns, Jómsvı´kinga saga, Morkinskinna, Nýja Testamenti Odds Gottskálkssonar, Ólafur Egilsson, Piltur og stúlka, Sagan af Heljarslóðarorrustu, So¨guþáttur af Ármanni og Þorsteini gala, Vı´dalı´nspostilla.)

10 From Old to Modern Icelandic Dative applicatives and NP/DP configurationality Heimir F. Viðarsson

10.1 Introduction ‘Free’ dative benefactives have long been considered rather elusive compared to more easily definable elements of argument structure, referred to by Grimm (1837: 705), for instance, as datives ‘hovering’ somewhere in-between.¹ Applicative Theory has made this notion of elements that do not clearly belong to the obligatory argument structure of the verb more precise by identifying certain cross-linguistically attested readings with specific structural positions, distinguishing between datives merged in a structurally high vs. low position (see Pylkka¨nen 2008; Marantz 2013). In this chapter I attempt to combine this theory in the context of diachronic change in Icelandic with recent discussion about NP/DP configurationality and the absorption of benefactives and external dative possessors into the nominal domain (Van de Velde 2010; Van de Velde and Lamiroy 2017). The licensing of datives has undergone major changes since the Old Norse period (including but not limited to Old Icelandic), an observation which has received little attention in the literature on historical developments in Icelandic. Old Norse allowed for dative applicatives to be spelled out in a ‘high’ or a ‘low’ position, denoting an affected or possessive reading, respectively, ¹ An earlier version of this chapter has appeared in Working Papers in Scandinavian Syntax (Viðarsson 2017a). Parts of this chapter have been presented at the 25th Scandinavian Conference of Linguistics at a workshop on Morphosyntactic Variation and Change in Germanic in Reykjavík 13–15 May 2013, the 16th Diachronic Generative Syntax Conference at the Research Institute for Linguistics— Hungarian Academy of Sciences in Budapest on 3 July 2014, 29. Rask-ráðstefnan um íslenskt mál og almenna málfræði in Reykjavík on 31 January 2015, and the 39th Penn Linguistics Conference at the University of Pennsylania on 22 March 2015. I would like to thank the organizers, audiences, and anonymous reviewers for valuable questions and comments. The usual disclaimers apply.

Heimir F. Viðarsson, From Old to Modern Icelandic: Dative applicatives and NP/DP configurationality. In: Functional Heads Across Time. Edited by Barbara Egedi and Veronika Hegedú´s, Oxford University Press. © Heimir F. Viðarsson (2022). DOI: 10.1093/oso/9780198871538.003.0010

ICEL ANDIC DATIVE APPLICATIVES

211

although the dative is frequently ambiguous between these readings. In neither case does the free dative actually belong to the obligatory argument structure of the verb: (1)

Old Norse dative applicatives (affected and/or possessive reading) a. þaa kleyiaði honum hinn minnzti fingr then itched him.DAT the.NOM smallest.NOM finger.NOM aa hinni hægri hendi framanverðr on the right hand anterior ‘Then the front of his right hand little finger itched.’ (MAR 153) b. konungr sagði at ecki hafði Þórólfr veitt þat sar. king said that not had Þórólfr inflicted that wound ok allt bitu honum annan veg vápnin and all bit.SUBJ him.DAT different way weapons.the.NOM ‘The king said that Þórólfr had not inflicted that wound because the weapons would cut completely differently for him.’ Or: ‘… his weapons would cut completely differently.’ (EG 31)

In present-day Icelandic, the datives in these constructions have been replaced by a different strategy (or strategies), for example involving possessive pronouns or prepositional marking: (2) Modern Icelandic renderings of (1b) a. Vopnin bitu allt o¨ðruvı´si hjá honum weapons.the.NOM bit.SUBJ all differently at him ‘The weapons would cut completely differently for him.’ b. Vopnin hans bitu allt o¨ðruvı´si weapons.the.NOM his bit.SUBJ all differently ‘His weapons would cut completely differently.’ The same carries over to dative benefactives in ditransitive (or ‘double object’) constructions (cf. also Holmberg and Platzack 1995), which typically require prepositional marking in Modern Icelandic. Since Icelandic thus lost ‘free’ dative benefactives while retaining its morphological case system, deflection arguably cannot be invoked as an explanation (see also Allen 2019: 200–9 for critical discussion of case-related explanations for English and other languages, including Icelandic). Note that dating this change (or changes) is difficult due to structural ambiguities that are typically hard to resolve. Based on evidence from benefactive ditransitive constructions, structurally high affected datives appear to be confined to the Old Norse period, whereas the structurally low possessed dative is still found in benefactive ditransitives

212

HEIMIR F. VIÐARSSON

up until and perhaps even beyond the Early Modern period (see Viðarsson 2017b). Some of these structures have persisted as idiomatic constructions or where datives can be licensed as oblique subjects in DAT-NOM contexts like (1a), the structural representation of which as regards the distinction between high vs. low applicatives can be rather unclear (see Wood and Sigurðsson 2014 and discussion in Section 10.2). Based on proposals recently advanced in the literature, I will argue for a possible explanation for the observed changes from Old Norse to present-day Icelandic in terms of increasing NP/DP configurationality (cf. also Lander and Haegeman 2013; Van de Velde and Lamiroy 2017). The proposal in a nutshell involves the grammaticalization of determiners, giving rise to tighter structures whereby clause-level elements, such as various kinds of non-thematic or ‘free’ datives, get absorbed into the nominal domain (cf. Van de Velde and Lamiroy 2017). That particular process of increasing configurationality will be treated here as an instance of the Head Preference Principle (see e.g. Van Gelderen 2009c, 2011b), where the rise of a fully grammaticalized D head from a phrasal modifier led to tighter, more configurational and hierarchical structures in the extended NP projection. As a result, a number of displacement processes were lost, including Left Branch Extraction of nominal modifiers (see esp. Platzack 2008; Lander and Haegeman 2013) and displaced or floating datives denoting possession (see Skard 1951; Bjarnadóttir 2011). This led to an overall increasingly rigid word order along the lines argued for cross-linguistically by Boškovic´ (2009, 2012) and Ledgeway (2012) for the development from Latin to the modern Romance languages. The chapter is structured as follows: Section 10.2 provides a theoretical background to dative applicatives where it will be emphasized that ‘free’ datives may be ambiguous between affectedness and possession, the latter arising from the context or by properties of the verb or the argument. In Section 10.3 Old Norse and Modern Icelandic applicatives are contrasted, indicating that fundamental changes have occurred in the licensing of overtly marked morphological datives, especially those associated with high (=eventive) readings. Section 10.4 outlines a possible account of the changes observed based on increasing configurationality in the NP/DP domain. Both quantitative and qualitative evidence obtained from the Icelandic Parsed Historical Corpus (Wallenberg et al. 2011) will be used to argue that Icelandic has been moving from an emergent article or ‘hypodetermining’ system with a flexible word order towards a rigid system with a full-fledged definite article. Section 10.5 concludes the chapter with a brief summary.

ICEL ANDIC DATIVE APPLICATIVES

213

10.2 Theoretical background The means by which affectedness is marked cross-linguistically varies considerably (see e.g. Radetzky and Smith 2010; Zúñiga and Kittila¨ 2010). From a typological perspective, the formal realization of beneficiaries varies both across and within languages, the major mechanisms being (i) case marking, (ii) adpositions, (iii) serial verb constructions, and (iv) applicativization (Zúñiga and Kittila¨ 2010: 7–10). From a generative perspective, it could be suggested that these strategies are not fundamentally different but rather varying outcomes depending on where and how an Appl(icative) head, denoting affected (or ‘applied’) readings, is spelled out (see Marantz 1993, 2013; Pylkka¨nen 2008; Wood 2013; Wood and Sigurðsson 2014). As regards the elements exemplified in (1) above, there is no general consensus in the literature as to how such datives are to be analysed in Old Norse (see also Viðarsson 2009, 2017b). Datives of this sort are often considered to be benefactives (or malefactives), but scholars have also assumed that they are experiencers or that they denote possession. Zúñiga and Kittila¨ (2010) point out that the definition is often circular, such that the role or function of a benefactive is defined in terms of whether or not an action or a situation is to the benefit of a participant. These authors propose the following working definition: The beneficiary is a participant that is advantageously affected by an event without being its obligatory participant (either agent or primary target, i.e. patient). Since normally only animate participants are capable of making use of the benefit bestowed upon them, beneficiaries are typically animate. (Zúñiga and Kittila¨ 2010: 2)

Grimm (1837) already observed that certain datives, which may or may not be directly associated with specific verbs, are notoriously difficult to analyse, appearing to hover somewhere in-between:² Such datives, hovering in the middle between case which is dependent on the verb, abound in the old and new languages in a variety of guises. (Grimm 1837: 705, my translation)

Van Valin and LaPolla (1997: 383), in fact, do not recognize benefactives as thematic relations, as these are ‘not a thematic relation which is a part of a verb’s logical structure’. The benefactive sense is either due to prepositions, for ² ‘Solcher dative, die zwischen dem von verbum abha¨ngigen casus in der mitte schweben, gibt es in der alten und neuen sprache eine menge, und der verschiedensten abstufung’ (Grimm 1837: 705).

214

HEIMIR F. VIÐARSSON

example for in English, or applied verb forms, for example in Chicheŵa (Van Valin and LaPolla 1997: 384). Similarly, Allen (2019: 225) assumes that ‘sentence datives’ in Old and Early Middle English were non-thematic and while they could be added to the valency of a verb, Allen maintains that they were interpreted pragmatically rather than syntactically. Van Valin and LaPolla (1997) distinguish at least three types of beneficiaries: (i) recipient beneficiaries, (ii) ‘plain’ beneficiaries, and (iii) deputative/substitutive beneficiaries, as exemplified in (3) (based on Van Valin and LaPolla 1997: 383–4): (3)

a. Robin baked Sandy a cake (recipient beneficiary) b. Robin baked a cake for Sandy (plain beneficiary) ‘[i.e. to show her she could do it, to amuse her, etc.]’ c. Robin baked a cake for Sandy (deputative beneficiary) ‘[i.e. so that she wouldn’t have to]’

Languages may also vary with respect to the nature of beneficiary markers, that is, whereas some languages allow beneficiaries to mark only a specific type, others may employ more general beneficiary markers, for example benefactive vs. malefactive; plain benefactive, deputative benefactive, and/or recipient benefactive (see e.g. Van Valin and LaPolla 1997; Zúñiga and Kittila¨ 2010; Colleman 2010). A unified theory of introducing arguments into the syntactic structure has been proposed under the heading of Applicative Theory (see e.g. Marantz 1993, 2013; Pylkka¨nen 2008; Wood 2013; Wood and Ármann Sigurðsson 2014; and many others). On these generative approaches to argument structure, both ‘core’ arguments (traditional indirect objects) and ‘non-core’ arguments (benefactives) are argued to project into Appl(icative) phrases. Applicative datives are associated with an ‘applied’ (or affected) meaning, which depends mainly on the structural position of the Appl head. Syntactically, Appl heads come in two guises, High and Low. High applicatives are typically elements negatively or positively affected by the event denoted by the verb, whereas low applicatives are in a relation with other arguments, often being possessors or recipients, for example of the theme in the traditional double object construction (cf. Pylkka¨nen 2008: 13): (4)

a. High applicatives denote a relation between an event and an individual. b. Low applicatives denote a relation between two individuals.

ICEL ANDIC DATIVE APPLICATIVES

215

Applicatives are merged in ApplP and the two types are structurally distinct projections which vary in height. For concreteness, I assume that these projections are interleaved with the rest of the clausal structure, where ApplPhigh attaches above vP/VP whereas ApplPlow is the complement of vP:

(5)

Two syntactic functional projections: High vs. Low ApplP ApplPhigh ApplP'high

DAThigh

ApplP'low

ApplPhigh

DATlow

ApplPlow

As we will see in Section 10.3.2, both structural positions may be filled simultaneously by a dative, providing evidence for a syntactically distinct high vs. low Appl and resulting in what appear to be ‘third’ object constructions. In this analysis, an applicative phrase is regarded as a thematic relation despite the fact that it does not belong to the ‘core’ argument of the verbal argument structure. Unlike Van Valin and LaPolla (1997) above, where applicatives would be analysed syntactically as adjuncts, both high and low applicatives are treated as arguments in the applicative analysis of Pylkka¨nen (2008), among others. The reasoning behind this derives from the analysis of subjects as external arguments, which are seen as arguments despite the fact that they do not belong to the ‘core’ arguments of the verb. Instead, these participants are introduced as arguments of functional projections, independently from the core argument structure of the VP: VoiceP (or vP) and ApplP for external and applied arguments, respectively (see e.g. Pylkka¨nen 2008: 6f.). Although two main configurations are usually distinguished depending on whether Appl relates an argument to an event or whether it relates two individuals as in (4a), (4b), respectively, more combinations are available (cf. Marantz 2013; Wood 2013): (6)

Type a. High b. High–Low c. Low

Complement Category vP DP DP

Syntactically High Low Low

Semantically High High High

216

HEIMIR F. VIÐARSSON

The middle High–Low type refers to so-called eventive DPs such as trip as opposed to shirt, which can be conceived of as events: The trip only took 10 minutes vs. the shirt only took 10 minutes. The eventive reading of shirt is only possible if it refers to an event (e.g. the making of the shirt), whereas a trip is naturally eventive as something that takes time (see e.g. Wood 2013; Marantz 2013). Thus, eventive DPs are generated in a syntactically Low Appl position but have High Appl semantics. From the present perspective, datives such as the ones exemplified in (1) and (3) above are all to be regarded as applicatives. However, the syntactic status of these elements, for example whether they behave like indirect objects or raise to subject, is subject to variation. At least traditionally, datives like (1) are often seen as indirect objects and are sometimes analysed as dative external possessor constructions as opposed to NP-internal possessive pronouns (see e.g. Van de Velde and Lamiroy 2017). In the literature on Icelandic, however, they would tend to be considered oblique subject constructions involving a lexically case-marked dative which raises to subject position, selected in (1a) by the Old Norse verb kleyja (Icelandic klæja) ‘to itch’ (see Viðarsson 2009, 2017b for critical discussion). In that case, the dative is usually treated as an ‘experiencer’, either solely or interchangeably with ‘benefactive’ (for further discussion, see e.g. Jónsson 1997–8; Jónsson and Eythórsson 2005; Holmberg and Platzack 1995: 196–200, 207–8 on the double object construction). Van de Velde and Lamiroy (2017) propose an account for the loss of this family of constructions cross-linguistically, focussing on the West-Germanic and Romance languages. As will be discussed in more detail in Section 10.4.1, these authors suggest that the loss of dative external possessor constructions in many of the modern Indo-European languages is not due to changes in the morphological case systems, as is often claimed to be the case. Rather, they propose that these languages have been drifting from non-configurational NP structures towards tighter, hierarchically structured NPs with grammaticalized determiners which have taken over the possessive uses of the dative. Van de Velde and Lamiroy (2017) suggest an account of these grammaticalization patterns from a constructional view where grammaticalization is seen as the rise of abstract, lexically underspecified constructions with specialized slots for determination and modification. This change in configurationality can be captured by assuming, on the one hand, a less configurational structure as in (7a) that is pragmatically motivated and a rigid functional structure as in (7b), on the other (cf. Ledgeway 2012).

ICEL ANDIC DATIVE APPLICATIVES

(7)

XP

a. X

X

XP

b. X

217

X

X X

…

From a generative view point, it appears that what is at issue here is basically that phrases (presumably adjuncts) denoting definiteness have been reanalysed as the D head of a designated DP functional projection. This grammaticalization process has been referred to as the Head Preference Principle (see Van Gelderen 2009c: 232): (8)

Head Preference Principle (HPP): Be a head, rather than a phrase

The HPP captures a bias operative in language acquisition that favours heads over phrases when the evidence presented to the learner is compatible with either analysis. Grammaticalization as a process is unidirectional and constitutes a linguistic cycle, meaning that an element starts out as an independent phrase or adjunct, gradually takes on a grammatical function, and is reanalysed as a head. The result of this kind of reanalysis can be morphosyntactic changes where a phrase becomes an affix or when an adjunct or argument becomes an agreement marker (see e.g. Van Gelderen 2011b: 6). In Section 10.4, I will sketch an account employing the term ‘construction’ in a loose sense, the building blocks of which I take to be heads and phrases, in an attempt at incorporating the basic insight of Van de Velde and Lamiroy’s account and applying it to the history of Icelandic. However, before doing so, a brief overview of the basic facts concerning the status of applicatives are in order.

10.3 Descriptive overview 10.3.1 Applicatives in Germanic Among the modern Germanic languages, German stands out in its use of dative case to denote a wide variety of relations, similar to those above, whereas, for example, Dutch, English, and the Scandinavian languages are much more restricted (cf. Hole 2005; McFadden 2006; Tungseth 2007; Colleman 2010; Scholten 2018). In German, a benefactive dative ditransitive construction can be formed productively with verbs to denote an affected meaning:

218

HEIMIR F. VIÐARSSON

(9)

Ich repariere ihm das Auto I repair him the car ‘I repair the car for him.’

(German)

The dative can also be interpreted possessively (‘repaired his car’) but this is not necessarily the case (for extensive discussion, see Hole 2005; Boneh and Nash 2012).³ The affected dative can also be used in contexts where there is no external (agentive) argument, where the unmarked order appears to be DAT > NOM (cf. Hole 2005: 226): (10) a. Ihm juckt die Kopfhaut him.DAT itches the.NOM scalp ‘His scalp itches.’ b. Ihm ist die Mutter gestorben. him.DAT is the.NOM mother died ‘His mother died on him.’

(German)

The interpretation of these affected datives varies depending on the context, another reading being the ‘unintentional causer’ (cf. Wood 2013): (11)

dem Hans zerbrach die Vase the.DAT Hans broke the.NOM vase ‘The vase broke on Hans (=affecting him)’

(German)

Affected datives in German are strictly speaking not ‘free’ because the presence of an argument embedded more deeply in the structure is required (cf. Hole 2005: 227): (12)

a. Ed hat ihr die Wa¨sche Ed has her.DAT the laundry ‘Ed did her laundry for her.’

gewaschen. washed

³ Boneh and Nash (2012) suggest that the possessive reading depends entirely on the nature of the theme. A native speaker of German confirms that the car in (9), indeed, does not have to be ‘his car’, as seen by the fact that it is still grammatical if the car is replaced by the neighbour’s car. Hole (2005: 220) provides the following contrast indicating that what is sometimes claimed to be a possessor dative is actually more like a perceiver or experiencer: (i) [Paul died first.] a. Dann starb auch seine Mutter. then died also his mother ‘Then his mother died, too.’ b. # Dann starb ihm auch seine Mutter. then died him.DAT also his Mother ‘Then his mother died on him, too.’ Despite the fact that Paul is dead, one can still refer to Paul’s mother using an internal possessor (seine Mutter), whereas this is not the case when the affected dative is used in (i-b). This an argument against treating ‘free datives’ as denoting possession.

ICEL ANDIC DATIVE APPLICATIVES

219

b. Ed hat (*ihr) gewaschen. Ed has her.DAT washed.laundry ‘Ed did the laundry (*for her).’ Thus, the intransitive verb waschen ‘do/wash laundry’ does not license an affected dative, whereas the corresponding transitive structure does. Hole (2005) develops an account in terms of variable binding to account for this contrast. As will be discussed below, there is some potential evidence that Old Norse affected datives could be completely free in this sense, raising the question whether the same held for Old Germanic in general. Note in this regard that evidence for free datives of this type is rather scarce in Old English (see Allen 2019: 38–43). Based on the discussion above, we should be careful when referring to affected datives as ‘possessives’ as is often done in the literature (see e.g. Hole 2005 and Boneh and Nash 2012 for a critical discussion). However, they clearly do participate in ‘external possessor constructions’ as an alternative to NPinternal possessive pronouns. The following examples from some of the Old Germanic languages appear to correspond closely to (10), the possessive sense, then, arguably being due mostly or wholly to properties of the theme (examples (13)–(14), also cited in Van de Velde and Lamiroy 2017): (13)

(14)

(15)

Old High German So riuzit thir thaz then mourns you.DAT the ‘Then your heart will mourn.’

herza heart

Old Saxon Thiu hlust uuarð imu the ear was him.DAT ‘His ear was cut off.’

farhauuan hewn

(Havers 1911: 285)

Old Norse svát þer brotnar beina so-that you.DAT break bones.GEN ‘So that all your bones will break.’

(Havers 1911: 293)

hvat each.NOM (Havers 1911: 268)

While it is often observed that the case system of Old Norse is still preserved intact in Modern Icelandic in all the relevant respects, the same cannot be said about the licensing of dative case on applicatives. In present-day Icelandic, dative applicatives now usually require some means of marking other than bare morphological case. To render an applicative like (10) or (11) above, Modern

220

HEIMIR F. VIÐARSSON

Icelandic may in some cases employ the oblique subject construction, as in (16a), but otherwise typically patterns with English rather than with German: (16)

a. Hann/honum klæjar ´ı ho¨fuðið. (Modern Icelandic) him.ACC/him.DAT itches in head.the.ACC ‘His head itches.’ b. Móðir hans er látin. mother his.GEN is diseased ‘His mother is dead.’ c. Vasinn brotnaði hjá honum. vase.the.NOM broke at him. ‘The vase broke on him.’

The dative found in (16a) is considered to be an instance of Dative Substitution, variation where accusative experiencer subjects of psych verbs tend to become dative (see e.g. Viðarsson 2009 and Barðdal 2011 for discussion). Importantly, note that (16a) is not about the productive use of the accusative/dative case to realize experiencers or applied arguments but rather the idiosyncrasy of a limited class of verbs taking oblique subjects. The oblique subject in (16a) is thus an obligatory internal argument of klæja in Modern Icelandic. Its use is, therefore, very different from the possessive pronoun in (16b) and the prepositional argument in (16c) which are not associated with any particular verb-dependent features in the lexicon. Modern Icelandic also does not allow fully fledged benefactive dative ditransitives except with a handful of verbs (see e.g. Holmberg and Platzack 1995; Maling 2001, 2002; Barðdal 2007; Radetzky and Smith 2010), in fact much like the situation in present-day Standard Dutch (cf. Colleman 2010). Thus, the only way to produce a ditransitive construction like (3) or (9) above is in the form of a prepositional ditransitive construction as shown in (17b) and (18b): (17)

a. *Páll bakaði Eirı´ki ko¨ku (Modern Icelandic) Paul.NOM baked Eric.DAT cake.ACC b. Páll bakaði ko¨ku *(handa) Eirı´ki Paul.NOM baked cake.ACC for Eric.DAT ‘Paul baked Eric a cake.’

(18)

a. *Páll lagaði Eirı´ki bı´linn (Modern Icelandic) Paul.NOM repaired Eric.DAT car-the.ACC b. Páll lagaði bı´linn fyrir Eirı´k Paul.NOM repaired car-the.ACC for Eric.ACC ‘Paul repaired the car for Eric.’

ICEL ANDIC DATIVE APPLICATIVES

221

Maling (2002: 53f.) observes that the benefactive ditransitive construction was still productive in Icelandic in the 19th century, although this observation rests on a single attestation. Based on the IcePaHC Corpus, Viðarsson (2017b) provides an overview of benefactive ditransitives with a bare dative, that is, corresponding to (17a)/(18a) above, which suggests that the construction was gradually falling into disuse during the Early Modern period. During the 19th century onwards, the examples are of an increasingly ‘figurative’ nature and tend to involve idiomatic expressions. An example of this is the Modern Icelandic expression baka e-m vandræði ‘get sb into trouble’ (lit. ‘bake sb trouble’), whereas baka e-m ko¨ku ‘bake sb a cake’ is completely out (see also Maling 2002: 43, 53). As mentioned in Section 10.2, it is often suggested that the replacement of the case marking strategy by prepositions is a direct consequence of the collapse of the morphological case systems in Dutch, English, and Mainland Scandinavian, where most of these constructions are ungrammatical with a bare dative (see e.g. Tungseth 2007). However, Icelandic could be argued to have retained even more of its case morphology than German (see e.g. Barðdal 2009), yet bare dative benefactives have, since the Old Norse period, become extremely restricted if not confined to idiomatic expressions. Just how different Old Norse applicative structures were from their Modern Icelandic counterparts will be explored in more detail in Section 10.3.2.

10.3.2 Old Norse applicatives 10.3.2.1 The ambiguity of single datives As shown in (15) above, applicatives corresponding to (10) were grammatical in Old Norse, in stark contrast to Modern Icelandic. According to the possessive tradition (see e.g. Skard 1951), these datives are not of the ‘free’, nonthematic type found in German but datives licensed in PPs denoting (mostly inalienable) possession. Example (19) is a case in point: (19) ofarla bı´ta ek sá einum hal | orþ high bite I saw one.DAT man.DAT words.NOM illrar konu (Old Norse, Poetic Edda) evil.GEN woman.GEN ‘The evil words of the woman bit one man high, I saw.’ (Havers 1911: 268) The dative in (19) does not belong to the argument structure of the verb bı´ta ‘to bite’ as can be seen by the fact that the transitive verb bı´ta ‘to bite’ takes an object in the accusative case. Skard (1951: 10) suggests that a prepositional

222

HEIMIR F. VIÐARSSON

phrase has been understood here, for example ´ı ho¨fuð ‘in head’, corresponding roughly to ‘bites in one’s head’, as the adverb ofarla ‘high’ might be taken to suggest. Skard’s study clearly demonstrates that Old Norse had a robust system of datives usually occurring with (or dependent) on PPs (see also Bjarnadóttir 2011). So the question is whether a PP is really necessary to license these datives in Old Norse. I would like to argue that it is not. From the perspective of Applicative Theory, there is no particular reason to assume that (19) is necessarily any different from the sorts of Appls we find in the German-style system, although the licensing mechanism or the merging order may be different (see e.g. Wood 2013, Wood and Sigurðsson 2014 on the cross-linguistic differences this may be argued to give rise to). However, it can be demonstrated that Old Norse datives truly are ‘free’ in the relevant sense, much as in German. The following Old Norse prose examples, again with bı´ta ‘bite’ as in (19), illustrate this point: (20) hvárt reiðið þér svá slæliga sverðin, er ek whether brandish you so poorly swords.the that I sé, at ekki bı´ta yðr? see that not bite you.DAT ‘Do you brandish the swords so poorly, because I see they do not bite for you?’ (HKR 449) (21) konungr sagði at ecki hafði Þórólfr veitt þat sar. king said that not had Þórólfr inflicted that wound ok allt bitu honum annan veg vápnin and all bit.SUB him.DAT different way weapons.the.NOM ‘The king said that Þórólfr had not inflicted that wound because the weapons would cut completely differently for him.’ Or: ‘… his weapons would cut completely differently.’ (EG 31) Note that the applicative yðr in (20) is formally ambiguous between an accusative patient and a dative benefactive, but the context implies that it is indeed the latter affected reading we are after, rather than the patient one. The affected reading is also the only one possible in (21) and honum is unequivocally a dative form. It seems, then, that these datives are indeed similar to the ones we find in German. We also find dative applicatives with unaccusative verbs such as eyðask ‘erode’, fallask ‘fall’, hverfa ‘vanish’, and koma ‘come’, as in (22), as well as with the copula vera ‘to be’ and verða ‘become’, as shown in (23):⁴ ⁴ The example in (23b), from Hrafnkels saga, was pointed out to me by Thórhallur Eythórsson.

ICEL ANDIC DATIVE APPLICATIVES

223

(22)

a. Geirr fann af skynsemi sinni at honum eyddusk Geirr felt of reason his that him.DAT eroded skotin shots.the.NOM ‘Geir sensed that his shots were being wasted.’ (EB 222) b. Skopta hvarf skyrta Skopti.DAT vanished shirt.NOM ‘Skopti’s shirt vanished.’ (STU 469–470) c. blicnaði hann oc varð faulr sem nár oc paled he and became pale as corpse and felluz honom hendr fell him.DAT hands.NOM ‘He became pale as a corpse and his hands fell motionless.’ (ÓH 173) d. litlv siðar com diacnanvm las-avr … i brvnina little later came deacon.the.DAT arrow.NOM in edge.the.ACC ‘A little later, an arrow came for the deacon, hitting the edge.’ (STU 217) e. þaa kleyiaði honum hinn minnzti fingr then itched him.DAT the.NOM smallest.NOM finger.NOM aa hinni hægri hendi framanverðr on the right hand anterior ‘Then the front of his right hand little finger itched.’ (MAR 153)

(23)

a. Þér er tungan lǫng orðin you.DAT is tongue.NOM long become ‘Your tongue has become long.’ (POST 175) b. En hánum varð þar eptir ein geit ok hafr but him became there after one goat and buck ‘One goat and a buck of his were left behind.’ (HRAFN 3)

Some such cases are still preserved in Modern Icelandic, usually in an idiomatic and/or figurative sense. The phrase e-m fallast hendur ‘sb is overwhelmed by sth’ survives as an idiomatic expression, unlike the more literal sense expressed in (22c). Naturally, the expression is not confined to fallast hendur ‘fall hands’ in Old Norse but combines with a variety of phrases, including andsvǫr ‘answers’, kveðjur ‘greetings’, læknidómr ‘healing’, orðtǫk ‘expressions’, and so on (see ONP: falla). The dative in Modern Icelandic is, therefore, a matter of learning an idiomatic expression, whereas in Old Norse, I argue, the dative applicative is a part of a productive system of expressing affectedness, which is licensed independently from the argument structure of the verb.

224

HEIMIR F. VIÐARSSON

It is often far from straightforward to determine whether a dative applicative at any given time denotes possession, possession and affectedness, or affectedness alone. Still, there are arguably at least two possible ways to distinguish these different senses and (arguably) structures. We can look for contexts where there is nothing to be possessed to begin with or where the possessive relation is already expressed somehow. In the case of the latter, this can either be distinguished based on the overtness of an NP-internal pronoun or, as will be discussed in Section 10.3.2.2 focussing on ditransitive structures, where the possessee lower in the structure is distinct from reference of another higher applicative dative. While there is no shortage of dative benefactive ditransitive constructions in Old Norse, they tend to involve recipient benefactives, that is, resulting in caused possession of something (see Viðarsson 2017b). Possible candidates for relations other than possession, where there is no obvious transfer of possession involved, include cases like the following: (24)

a. Kona ein spurði hvat eldrinn skyldi. „Til Bœjar,“ Woman one asked what fire.the should to Bær segir hann, „at elda Þorvaldi bað.“ says he to fire Þorvaldr.DAT bath.ACC ‘A woman asked what the fire was meant for. “To (the farm) Bær,” he says, “to warm a bath for Thorvald with fire.” ’ (STU 395) b. … er þar firir iarn hurð. oc ængi maðr er nu þar … is there for iron door and no man is now there honum upp at luka. him.DAT up to open ‘It was shut with an iron door and no one there to open (it) for him.’ (ÞIÐR 315)

In (24a) there is nothing to suggest that to warm sb a bath results in any transfer of possession to Thorvald, who instead appears to be the plain beneficiary of the event. This is perhaps even clearer in (24b), where opening ( for) sb (the door) obviously does not result in honum ‘him’ becoming the owner of the door, neither literally nor figuratively. Similarly, the applicative in (25) really appears to denote affectedness rather than possession, as the possessive relation is expressed explicitly by the possessive pronoun:

ICEL ANDIC DATIVE APPLICATIVES

(25)

225

‘Viti þat sá ungi maðr er sat næst kónginum, Know that the young man who sat next king.the at eptir varð honum yfirklæði sitt.’ that after became him coat his.REFL ‘May the young man, sitting next to the king, know that his coat was left behind.’ (ÆV 61)

Examples of this sort have also been reported in Old and Middle English, where it is referred to as the ‘blended’ construction (see e.g. Allen 2019: 39f., with references). The only way to interpret the dative in (25) possessively would be to treat it as an instance of doubling, where the two elements are still clearly structurally/morphologically distinct. Instead of assuming the same function for both these elements, I tentatively assume that the dative in this construction denotes a relation other than possession.⁵ However, the fact that the reference of the dative can be ambiguous as well as the fact that the dative can co-occur with NP-internal possessives may very well have served to pave the way for the sort of absorption of the dative into an NP-internal position. Before turning to the account of Van de Velde and Lamiroy (2017), let us first home in on the availability of a high applicative in Old Norse where there is already a low applicative present in syntactic structure—the former being the event argument, the latter arguably fulfilling the role of the internal argument and the possessor of the theme. 10.3.2.2 Resolving ambiguities structurally: co-occurring high vs. low dative applicatives Recall now that according to Applicative Theory, there are two functional projections present in the syntax, a low Appl and a high Appl. These are responsible for the different semantics available to applicative constructions cross-linguistically. Presumably, the datives in (24) denote a sense of plain or deputative benefaction (see (3b), (3c) above), associated above with the high Appl. The split high and low ApplP structure gives rise to the following testable prediction. Since these two readings do not reside in the same functional projection, they ought in principle to not be mutually exclusive. This is stated in (26): ⁵ This is an option that is also seen as a possible treatment of the ‘blended’ construction in Old English, where ‘the coreference of the dative and the possessor would be inferred from the context’ (cf. Allen 2019: 40).

226

(26)

HEIMIR F. VIÐARSSON

Applicative Theory predicts possible co-occurrence of high and low applicatives.

Accumulating two identical applicatives should be impossible as they would be competing for the same structural position or, alternatively, violate Chomsky’s (1981) θ-criterion. Despite any potential ambiguity, therefore, one of the phrases must be syntactically high and the other syntactically low. An example corresponding to (26) would thus contain a dative that is referentially distinct from the possessor of the theme. Boneh and Nash (2012) indeed demonstrate for French that only distinct types of datives can co-occur (or be accumulated), distinguishing between core vs. non-core datives, which largely coincides with the low vs. high distinction above. The French example in (27) exemplifies this property: (27)

Ce matin, j’ai juste à me repasser quelques This morning, I’ve only to 1SG iron several chemises à ma femme. shirts for my wife ‘This morning, I only have to iron some shirts for my wife.’

The non-core argument à me ‘to me’ does not interfere with the core argument à ma femme, the former of which establishes a relation between the event and the ironing, whereas the latter establishes a relation between the shirt and the wife. Co-occurring high and low Appls are relatively hard to find in linguistic corpora unless the corpus is annotated syntactically. The Icelandic Parsed Historical Corpus (cf. Wallenberg et al. 2011) codes for co-occurrences of these sorts as NP-OB3, as opposed to the direct object NP-OB1 and indirect object NP-OB2 relations, making it possible to test this prediction with minimal effort. Such examples occur and they appear to be confined to the Old Norse period. Unfortunately for our purposes, however, all instances of NP-OB3 in IcePaHC turn out to be datives that are co-referential with the subject: (28)

Vér we með with

skulum oss shall us.DAT tárum tears

(29)

þá bað then asked fara … go

biðja drottin ask lord.ACC

várn our.ACC

Bárðr sér orlofs Bárðr REFL.DAT permission.GEN

miskunnar mercy.GEN

konung king.ACC

at to

ICEL ANDIC DATIVE APPLICATIVES

(30)

227

… at þeir gleymdu at æsta sér guð … that they forgot to ask REFL.DAT god.ACC undankvámu … escape.GEN

Co-referential datives are known in the literature on Old Norse but they have been dismissed / explained away by scholars as only involving two rather than three arguments, either by suggesting an amendment to them by adding a (supposedly understood) dative-assigning verb responsible for the ‘third’ object, for example fá ‘give’ (Haugan 2000: 168) or by assuming that reflexives do not function as objects in some languages (Maling 2001: 432f., fn. 9). Despite the obvious fact that configurations of this sort are not very likely to be richly attested in any corpus, let alone a historical corpus of a limited size, it would be quite unfortunate if all potential cases of a third argument were always amenable to either of the conditions mentioned by Haugan (2000) and Maling (2001). Indeed, they are not. A closer scrutiny of Old Norse sources reveals that co-occurring or accumulated applicatives did not always involve either co-referential or reflexive datives. A putative example of this sort happens to be discussed by the IcePaHC parsing team on their forum, attested in Gı´sla saga Súrssonar, a late 13th-century text. However, the proper analysis in terms of the different applicative relations is not particularly clean-cut. Note that this example is not found in the IcePaHC corpus but is provided by Eirı´kur Ro¨gnvaldsson to demonstrate that they did not always involve reflexives in Old Norse: (31)

En eigi mun eg biðja Gı´sla ykkur bjargar but not will I ask Gı´sli.ACC/DAT you.DAT rescue.GEN héðan af. here of ‘But I will not ask Gı´sli for rescue for you now.’ (, cf. Wallenberg et al. 2011)

The expression biðja e-n e-s ‘ask sb sth’ involves the ACC-GEN verb biðja ‘ask, beg’ but the verb can also occur with an indirect object in the dative (DATGEN), biðja e-m e-s ‘ask sth for sb’. The dative ykkur ‘you’ poses a problem since it is not obvious whether that phrase is (positively) affected by the asking (the high, eventive reading) or whether it is the recipient/recipient benefactive of the help (the low, complement-complement reading), or both. In the same text as in (31) there is also an example featuring two datives that appears to be more straightforward in terms of the High/Low readings and,

228

HEIMIR F. VIÐARSSON

importantly, does not involve a reflexive pronoun—an example par excellence conforming to (26): (32)

veittu mér þat, at þú sker mér skyrtu, provide me that that you.NOM cut me.DAT shirt.ACC, Auðr, Þorkeli bónda mı´num Auðr.NOM, Þorkell.DAT husband.DAT my.DAT ‘Please do this for me, Auður, that you cut me a shirt for my husband Þorkell.’ (GÍSL 11)

In (32), we now have all the positions filled that we need, the high/eventive Appl by a dative third person pronoun disjoint from the second person subject, again disjoint from the low/complement Appl third person dative noun phrase which is the canonical indirect object (recipient or recipient benefactive).

(33) (Simplified) syntactic structure of example (32) ApplPhigh ApplP'high

Spec mérhigh

NP

ApplPhigh N skyrtu

ApplP'low Spec

ApplPlow

Þorkeli bónda mínumlow In Modern Icelandic, the only way to render an example like (32) is by resorting to a distinct PP for each dative, with fyrir vs. handa (both meaning ‘for’) to denote the high/purely benefactive vs. low/recipient reading, respectively: (34) Gerðu það fyrir mig að skera fyrir mig skyrtu, do.you it for me to cut for me.ACC shirt.ACC, Auður, handa Þorkeli manninum mı´num Auður.NOM, for Þorkell.DAT husband.the.DAT my.DAT ‘Please do this for me, Auður, to cut me a shirt for my husband Þorkell.’ It may be no accident that the ‘third’ argument, the high applicative, in (32) is a pronoun rather than a full NP. It is known cross-linguistically that nonthematic datives tend to be pronouns, often restricted only to first and second person pronouns. Ethical datives, for instance, generally only allow first and

ICEL ANDIC DATIVE APPLICATIVES

229

second in French, usually having the pragmatic effect of addressing or reflecting the views of the speaker or hearer (see e.g. Boneh and Nash 2012). Nonetheless, the high Appl in (32) really does appear to be a fully fledged participant, as witnessed by the fact that it is not co-referential with the subject. What (28)–(32) all show beyond reasonable doubt is that merging a high Appl dative was a possibility in Old Norse, suggesting that Old Norse patterns more with German, and perhaps with Old Germanic in general (cf. Van de Velde and Lamiroy 2017), than with Modern Icelandic. This also lends support to the view that the Old Norse case system is different from the one found in Modern Icelandic in a fundamental way. An important question that remains is what may have triggered these changes. In Section 10.4, I would like to explore a proposal made by Van de Velde and Lamiroy (2017) that the loss of these dative constructions correlates with changes at the level of the NP.

10.4 Towards an explanation 10.4.1 The absorption of dative external possessors into the NP and the rise of DP The fuzzy borders between affected datives, experiencers, and possessors have already figured a number of times in the discussion above. Van de Velde and Lamiroy (2017) make use of this opacity in their account which can be roughly summarized as follows. Ancient Indo-European languages had an extremely flexible word order and seem to lack the extended NP structure typically found in the modern European varieties (see also e.g. Ledgeway 2012). Over time, ‘clause-level elements’ such as adjectives, quantifiers, and pronouns modifying the noun grammaticalized into determiners, giving rise to a hierarchically structured NP constituent with designated determiner slots. This move towards greater configurationality resulted in NP-external material getting absorbed in the NP, whereby the dative external possessors were replaced with NP-internal possessors. Van de Velde and Lamiroy (2017) suggest that the rise in configurationality in the NP resulted in an increasingly rigid word order so that the datives which had had ‘floating’ properties became an integral part of the NP. Allen (2019: 213f.) voices criticism over linking this change and increased NP configurationality with the loss of the external dative possessor. She notes that the grammaticalization of the definite article in Old and Early Middle English

230

HEIMIR F. VIÐARSSON

does coincide with a significant decrease of some external dative possessor types. However, her results also show that the changes in the datives are more gradual and piecemeal in nature than we would expect if it were tied to NP/DP configurationality alone. The same is probably true for Icelandic but this does not automatically entail that the hypothesis is without merit in my view. As early as Havers (1911), changes in the use of the dativus (in)commodi or dativus sympatheticus constructions have been associated with pronouns and nominal possessive genitives. Thus, Havers (1911: 273–4) claims that pronominal sympathetic datives are usually preposed in the Poetic Edda, which generally predates the oldest Old Norse prose by a couple of centuries, whereas the dative tends to occur in a post-nominal position in the prose (var hann senn ór augliti mér ‘he was soon out of my sight’). Havers suggests that this is due to the possessive pronouns which also follow the noun. The postposing of nominal sympathetic datives is similarly considered to be related to the postposing of nominal possessive genitives (sneiþk af haufuþ húna þinna ‘I cut off the head of your sons’). These observations thus arguably point in the same direction, viz. that elements outside the NP get absorbed into the nominal domain on the model of NP-internal possessives. It appears that Havers’ (1911) claims regarding the Old Norse prose are more or less confirmed by Bjarnadóttir’s (2011) study, although it should be borne in mind that it focuses on Icelandic sagas, viz. texts dating from the 13th century and later. When datives are used possessively with a prepositional phrase, the dative is usually found following the noun it modifies in much the same way as possessive pronouns do. However, when the dative is a pronoun, it is separated from the noun and moves out of the PP about 22 per cent of the time, 86 per cent of the datives that move being personal pronouns and reflexives (cf. Bjarnadóttir 2011: 27). The following examples illustrate the options we find during this classic period where the dative is a nominal (examples from Bjarnadóttir 2011: 33; translations and linguistic annotation added here): (35) a. PP>DAT, Old Norse þá seldi hann [ı´ hendur [ Eirı´ki syni then sold he.NOM in hands Eric.DAT son.DAT sı´num]DAT rı´ki] his.REFL.DAT kingdom.ACC ‘Then he passed the kingdom over to / into the hands of his son, Eirı´kr.’

ICEL ANDIC DATIVE APPLICATIVES

231

b. DAT>PP, Old Norse seldi hún [sonum sı´num]DAT [ı´ hendur [ ]DAT sold she.NOM sons.DAT her.REFL.DAT in hands bú sitt] household.ACC her.REFL.ACC ‘She passed the estate over to / into the hands of her sons.’ The system of PP dative possessives is already in competition with the possessive pronoun in Old Norse. Inalienable possession with a PP is denoted by a dative in over 90 per cent of the cases (Bjarnadóttir 2011: 27). It appears that the dative possessive with inalienable possessions in a PP begins its decline as early as the 16th century (see Bjarnadóttir 2011: 9–10, with references). Hence, the loss of the applicative datives is presumably a gradual affair where the morphological datives are reanalysed as NP-internal elements. Importantly, the dative was often ambiguous with regard to an affected vs. a possessed reading, as we have seen, and the dative external possessor could even co-occur with an NP-internal possessive in Old Norse as in (25) above, repeated in (36) below: (36) Viti þat sá ungi maðr er sat næst kónginum, Know that the young man who sat next king.the at eptir varð honum yfirklæði sitt.’ that after became him.DAT coat.NOM his.REFL ‘May the young man, sitting next to the king, know that his coat was left behind.’ Bjarnadóttir (2011: 25) mentions similar cases of doubly-marked (or ‘blended’) possession involving inalienable possession with the predicates verða dofinn/sár ‘become numb/hurt’ and the NP fótr sinn ‘his(refl) foot’. An important motivation for the absorption of the external possessor into the internal NP structure is the gradual loss of ‘floating’ properties and flexible word order giving rise to discontinuous structures (cf. Van de Velde and Lamiroy 2017).⁶ The rise of a grammaticalized determiner system is seen as

⁶ Note in this respect that word order during the Old Norse period was generally speaking much more flexible than was later the case in Icelandic. Old Norse allowed for a variety of other discontinuous syntactic structures, including long, non-clause-bounded scrambling (see Lander and Haegeman 2013: 296–301, with references), exhibited pragmatically motivated Verb-Object order, i.e. OV, VO, and mixed OVO or ‘T3’ orders (see e.g. Hróarsdóttir 2008a, 2009; Haider 2013) and allowed for the reversal of the order of nominal direct and indirect objects akin to clause-bounded scrambling (see Viðarsson 2017a: 48f.), all of which is either ungrammatical or subject to severe constraints in Modern Icelandic.

232

HEIMIR F. VIÐARSSON

having led to the loss of discontinuous structures where elements could be separated from the phrases they modified, citing cases such as (37) from Latin: (37)

a. magno cum dolore great with grief ‘with great grief ’ b. nostram … ridebant inuidiam our … they.laughed unpopularity ‘they mocked at our unpopularity’

(Ledgeway 2012: 393)

(Ledgeway 2012: 394)

Discontinuous structures used to be features of both the Germanic and Romance languages but these were gradually lost as these languages developed a definite article and an elaborate DP structure (see e.g. Lander and Haegeman 2013 on Old Norse). As discussed at length by Van de Velde and Lamiroy (2017), there appears to be an inverse correlation between the extent of the grammaticalization of the article and the retention of the external possessor. Thus, the external possessor is least retained in languages where the definite article has progressed the most, NP configurationality following an English > Dutch > German cline in West-Germanic and a French > Italian > Spanish cline in Romance. This is demonstrated on the basis of a number of properties, one of which is the ability for possessives to co-occur with the article (cf. also Van de Velde 2010 on the rise of the article in Dutch): (38)

a. (*ce/*le) mon livre this/the my book ‘my book / this book of me’ b. il mio libro the my book ‘my book’ c. el libro mı´o the book my ‘my book’

(French)

(Italian)

(Spanish)

Similar claims concerning the status of determiners and co-occurrence of demonstratives, (alleged) definite articles, and possessives in Old Norse are made by Lander and Haegeman (2013: 291–2), among others. They argue that Old Norse lacked a fully grammaticalized definite article, being an ‘NP’ rather than ‘DP’ language (see further Boškovic´ 2009, 2012 on this distinction). According to these authors, less configurational NP structures correlate with having a very flexible word order, for example in terms of discontinuous phrases as in (37). Some of these properties are exemplified in (39) from the Old Norse

ICEL ANDIC DATIVE APPLICATIVES

233

MÍM corpus, the left branch extraction of quantifiers, numerals, adjectives, head nouns, and demonstratives annotated syntactically for clarity:⁷ (39) Discontinuous phrases in Old Norse a. Hversu margai munum vér [NP ___i menn ] þurfa […]? how many.ACC will we.NOM men.ACC need ‘How many men will we need?’ (MÍM: Brennu-Njáls saga) b. Tvo¨i hafði hann [NP ___i lo¨g ] ok mo¨rg two.ACC had he.NOM stabs.ACC and many.ACC sár o¨nnur ok stór wounds.ACC other.ACC and large.ACC ‘He had two stab wounds and many other big wounds.’ (MÍM: Sturlunga sga) c. Góðai áttu þeir [NP ___i ferð ] um heiðina … good.ACC had they.NOM trip.ACC over heath.the.ACC ‘They had a good trip over the heath.’ (MÍM: Sturlunga saga) d. Vinuri var hann [NP ___i Eirı´ks jarls ] friend.NOM was he.NOM Eric.GEN earl.GEN ‘He was a friend of earl Eirı´kr.’ (MÍM: Bjarnar saga Hı´tdælakappa) e. Þanni vissi ek [NP ___i mann ] bestan ´ı heimi that.ACC knew I.NOM man.ACC best.ACC in world ‘That man was the best one I knew in this world.’ (MÍM: Heimskringla) Whether the structure of the Old Norse NP is to be regarded as entirely ‘flat’ and non-configurational (cf. Bo¨rjars et al. 2016) is not immediately obvious. Here, I follow Stroh-Wollin (2016), partly also Platzack (2008), in assuming slightly less dramatic differences between Old Norse and Modern Icelandic than the truly non-configurational structure adopted by Bo¨rjars et al. (2016). To account for syntactic discontinuity as in (39) above, I assume for (Early) Old Norse an NP structure lacking DP as a functional projection, where the modifier and crucially also the emerging definite article originate as adjuncts to the NP (along the lines of Lander and Haegeman 2013; Stroh-Wollin 2016). In Modern Icelandic, in contrast, there is rigid DP structure where the modifiers each project as a functional projection, taking the NP as their complement and

⁷ The examples in this section are cited from the tagged, historical corpus of Old Norse, Mo¨rkuð íslensk málheild (MÍM, ). The corpus mostly consists of the Icelandic sagas, thus representing 13th–14th-century Old Norse.

234

HEIMIR F. VIÐARSSON

thus preventing extraction (see e.g. Platzack 2008: 363).⁸ The tree diagram in (40a) provides the assumed structure for Old Norse, cf. example (39c), prior to extraction of the AP. The structure in (40b) provides the corresponding DP analysis of Modern Icelandic, where the extraction of AP cannot target A to the exclusion of NP:

(40) a. Old Norse NP structure

b. Modern Icelandic DP structure

NP

DP

AP

NP

A

N

góða ferð

D

AP A

NP

góða

N ferð

I follow Bo¨rjars et al. (2016) in not assuming a unified category D in Old Norse as long as the overt marking of definiteness is optional and there still is no complementary distribution between elements such as definite markers, demonstratives, and possessive pronouns. Nonetheless, there is evidence that a structural D position was an emerging property in (Late) Old Norse. Based on Stroh-Wollin’s (2016) systematic study of the use of the definite marker in Runic Norse and Old Norwegian in a range of different contexts, it appears that the use of the definite marker has increased dramatically in the 13th century in comparison to a century earlier but is still not obligatory in all contexts. With regard to most of the features discussed by Van de Velde and Lamiroy (2017), Old Norse patterns with the languages which have least NP configurationality and most productive use of dative applicatives (or dative external possessors). Conversely, Modern Icelandic shares most of its features with languages with most NP configurationality and least productive use of dative applicatives. As Van de Velde (2010) discusses in detail, each element (or construction) becomes configurational gradually, the determiner being diachronically emergent through lexical diffusion. The same point is also argued extensively by Ledgeway (2012) with regard to developments from Latin to the Romance languages. Thus, these languages as a whole did not develop

⁸ Platzack (2008) assumes a DP analysis for both Old Norse and Modern Icelandic. The only difference is that nominal modifiers such as adjectives are adjoined in Old Norse (‘NP-over-MP type’), but project as MPs (APs) in Modern Icelandic (‘MP-over-NP type’).

ICEL ANDIC DATIVE APPLICATIVES

235

from ‘non-configurational’ to ‘configurational’ but rather individual constructions did. This is, therefore, a much weaker claim than the sort of system-wide non-configurationality originally argued for by Hale (1983) on languages like Warlpiri, which Faarlund (1990) originally adopted in his analysis of Old Norse. This analysis was rightly criticized by Ro¨gnvaldsson (1995) in certain important respects. By the criteria discussed by Van de Velde and Lamiroy (2017), Old Norse does not have a fully grammaticalized article system (cf. also Lander and Haegeman 2013). What later develops into an article can co-occur at least with demonstratives and possessive pronouns, indicating furthermore that the latter two elements, too, have not become D heads themselves (see Van de Velde 2010: 268–9). Before turning to these properties in more detail, observe first that the definite adnominal article is a late innovation in Old Norse, lacking in Runic and Eddic Old Norse, save the pre-adjectival one (cf. Nygaard 1867: 47–8, 1905: 33–4; Stroh-Wollin 2009): (41) ÞioðrikR hinn þurmoði Theoderic the bold (Runic, 9th century, cf. Stroh-Wollin 2009: 6) The first instances of hinn ‘the’ without an adjectival attribute, the precursor of the definite suffix, are considered to stem from the 11th century (see Stroh-Wollin 2009: 6). Nygaard (1905: 35) furthermore points out that even in prose usage (12th century onwards), the definite article is not yet systematically found (‘ikke … gjennemført’) where one expects to find definite forms (see Nygaard 1905: 35–47; Lander and Haegeman 2013: 287–91). I assume that at this point, definite markers in Old Norse were adjuncts as indicated for hinn minnzti fingr ‘the smallest finger’ (from example (1)) in (42a):⁹

(42) a. Old Norse NP structure

b. Later DP structure

NP

DP

DemP

NP

D

Spec Dem

AP

NP

hinn

A

N

minnzti fingr

hinn

AP A

NP

minnzti N fingr

⁹ Bo¨rjars et al. (2016) assume a flat NP-structure where Det is a part of AP. This makes sense while the definite marker was mainly confined to (and given its origin in) adjectival contexts.

236

HEIMIR F. VIÐARSSON

I take the grammaticalization of the definite article, demonstratives, and possessive pronouns as mutually excluding determiners to be an instance of the Head Preference Principle (cf. Section 10.2 above). By the HPP, a phrasal modifier in a ‘non-configurational’ NP is reanalysed as a functional head, in this case a D(eterminer). Before the reanalysis takes place, these modifying elements can co-occur, for example as in Italian, cf. (38b). Once a modifier has been reanalysed as a D head, it will be blocked by any other existing D head in the same phrase, thus ruling out the co-occurrence of a demonstrative/definite article and a possessive pronoun. The status of these elements is definitely not a matter of setting an NP/DP parameter globally for the whole language. Since grammaticalization is unidirectional, we predict that while the step from a phrasal modifier to a head is made possible through the HPP, the opposite change where the D head is subsequently reanalysed back to being a phrasal modifier will be impossible.

10.4.2 A closer look at historical changes in the use of definite determiners in Icelandic Let us now turn more closely to Van de Velde’s (2010) criteria, which include the following distributional properties (cf. Van de Velde 2010: 268f.):¹⁰ (43)

a. Exclusiveness When a language has a specialized part of speech (an article), exclusively expressing definiteness, referentiality, specificity, or identifiability, any item of this class is a determinative. b. Obligatoriness When a language uses an obligatory element for the expression of definiteness, etc., these elements may be considered determinatives. c. Complementary distribution When a language has various adnominal elements entailing (in)definiteness, (un)specificity, referentiality, or identifiability, and these elements stand in complementary distribution, they may be considered determinatives. d. Subjacency If an adnominal dependent occurs to the right of an element that is not a determiner, but an adjunct, it is not a determiner itself.

¹⁰ See Sommerer (2018: 62–8) for further discussion and elaboration of these criteria in the context of the rise of the definite article in the history of English.

ICEL ANDIC DATIVE APPLICATIVES

237

In relation to the first criterion, scholars have pointed out that Old Norse had multiple definite and/or demonstrative markers that could be used for the purpose of constraining reference (see e.g. Stroh-Wollin 2009: 20–2; Lander and Haegeman 2013: 286f., 291f.; Bo¨rjars et al. 2016). These markers appear to have been used somewhat interchangeably, thus potentially failing the exclusiveness criterion. Similarly, it is frequently observed that overt definite marking is not used consistently in definite contexts where expected based on its obligatory use in the modern descendant languages. I will briefly return to both these points below. The two remaining criteria refer to structural aspects in terms of the use of the definite markers relative to other elements present (or absent) in the NP/DP. These criteria also strongly suggest that the definite marker is not to be considered a fully grammaticalized definite article. Of these two, the cooccurrence of definite determiners, demonstratives, and possessives in Old Norse is probably most cited (see e.g. Lander and Haegeman 2013: 291f.), but scholars also point out that the order in which the elements occur is not firm (cf. Bo¨rjars et al. 2016: 10f.). This last property is at least partly related to the subjacency criterion. Subjacency sets out to determine whether or not the definite marker can be subjacent to modifiers of NP (for further discussion, see also Sommerer 2018: 65f.). The gist of the idea is thus that the definite article should demarcate the left edge and if the definite marker appears more deeply embedded in the NP/DP than modifiers, it is not yet a definite article. Based on the above criteria and discussion, I would like to highlight the following properties, which together provide a good argument for the different status of determiners in Old Norse vis-à-vis the modern varieties: • The supposed article and possessives were not yet in complementary distribution. • An adjective could precede the possessive pronoun. • Possessives can occur to the right of modifiers. • Possessives did not entail definiteness and could co-occur with demonstratives in either order (DEM > POSS or POSS > DEM). The examples in (44)–(48) below, all taken from the Old Norse MÍM corpus, serve to illustrate these points. (44)

a. þı´n hin mesta gæfa (POSS and ART not in complementary your the greatest fortune distribution) ‘The greatest fortune of yours.’ (MÍM: Brennu-Njáls saga)

238

HEIMIR F. VIÐARSSON

b. sı´na hina ágæstu menn their the most.excellent men ‘The most excellent men of theirs.’

(MÍM: Heimskringla)

(45)

þessir hinir góðu mı´nir félagar ok these the good my companions and fóstbræður (ADJ can precede POSS) foster.brothers ‘These good companions and foster brothers of mine.’ (MÍM: Sturlunga saga)

(46)

a. þrjá sı´na men three his men ‘His three men.’ b. fjórir mı´nir félagar four my companions ‘My four companions.’ c. það eitt sitt efni that one his solution ‘That one solution of his.’

(POSS to the right of modifiers) (MÍM: Grænlendinga saga)

(MÍM: Þórðar saga hreðu)

(MÍM: Sturlunga saga)

Demonstrative and possessive pronouns are not mutually exclusive and can appear in either order: (47)

(48)

a. þessa sı´na dóttir, Droplaugu this her daughter Droplaug ‘This daughter of his, Droplaug.’ b. sá þinn bóndi that your husband ‘That husband of yours.’ a. sı´na þá heimanferð his that departure ‘That departure of his.’ b. skaða sı´num þessum harm his this ‘This harm of his.’

(DEM > POSS) (MÍM: Fljótsdæla saga)

(MÍM: Laxdæla saga) (POSS > DEM) (MÍM: Egils saga)

(MÍM: Brandkrossa þáttur)

The fact that the boldfaced elements could co-occur and/or did not necessarily occur in a fixed structural position is suggestive of their being at a different stage in the linguistic cycle than in Modern Icelandic, where definite determiners have become fully syntacticized.

ICEL ANDIC DATIVE APPLICATIVES

239

100% 90% 80% 70%

DAN

60%

D A N-D

50%

NDA

40%

N-D A

30%

A N-D

20% 10% 0%

EON

LON

LI

MI

Fig. 10.1 Proportion of each definiteness pattern found within each period. Periods: Early Old Norse-Icelandic (EON) 1150–1350 (N=443), Late Old Norse-Icelandic (LON) 1350–1550 (N=177), Later Icelandic (LI) 1550–1800 (N=274), Modern Icelandic (MI) 1800–2008 (N=850). Notation: D ‘definite article hinn (free form)’, A ‘adjective’, N ‘noun’, N-D ‘definite noun (-inn, bound form)’.

I am not aware of any systematic study documenting the grammaticalization of (h)inn as a definite determiner in the history of Icelandic but the results reported in Figure 10.1 obtained from IcePaHC of the major definiteness patterns are suggestive of fundamental changes in this domain diachronically.¹¹ What Figure 10.1 reveals is that, in the presence of both an adjective and a noun, the suffixed article (A N-D) gradually replaces the free-standing article (D A N) as the dominant strategy for marking definiteness. As no attempt was made to distinguish between weakly and strongly inflected adjectives, A N-D (as well as N-D A; cf. footnote 11) conflates two distinct patterns, viz. the weak NP-internal and the strong NP-external one (see Pfaff 2015 on this distinction in Modern Icelandic).¹² The existence in Early and Late Old Norse ¹¹ The data shown in Figure 10.1 were obtained by extracting NPs immediately dominating a Dand an ADJ-element, where the D-element immediately dominates the lemma hinn ‘the’ and ADJ (weak/strong) either precedes or follows N. The results were manually checked for consistency and coding errors. The determiner sá ‘that/the’ was left out of consideration (see Figure 10.2), meaning that the ‘double definiteness’ pattern, labelled D A N-D, always featured hinn as a free article and as a definite noun with the bound -inn form, i.e. both simultaneously. These cases did not include demonstrative uses of hinn that select the bound form (meaning ‘the other’). Double definiteness in the N D A and N-D A patterns (i.e. N-D DA) was conflated with N D A. The N-D A pattern was often difficult to distinguish from N-Ds occurring with depictives, ‘afterthoughts’, and various other NP-external elements. ¹² I would like to thank Alexander Pfaff for valuable discussions about this/these pattern(s). The weak vs. strong contrast is clearly an important distinction to make, although it does not change the

240

HEIMIR F. VIÐARSSON

of the post-nominal adjectival article pattern (N D A; cf. (41) above), albeit not strictly confined to epithets or name-like designations, indicates that the free-standing article modifies the adjective rather than the noun. The fact that the post-nominal adjective with a suffixed article pattern (N-D A) survives into the modern period suggests that it was reanalysed on par with the A ND pattern as involving a true adnominal definite determiner. These results largely confirm the above claims according to which the adnominal article is an emergent property in Early Old Norse. They do not show quantitatively, however, how Old Norse changed from a hypodetermining language, that is, ‘expressions which are inherently definite are not marked by an article’ (cf. Leiss 2007: 88; see also Stroh-Wollin 2009; Lander and Haegeman 2013), to a language with a fully fledged, obligatory definite article. According to Leiss (2007: 88–9), systematically marked thematic arguments are definite but not thematic arguments even when they were semantically definite. To address quantitatively the overall rise of an obligatory determiner, now regardless of whether nouns are modified by adjectives, we can study its frequency of occurrence. Since definiteness was not only marked by the emergent definite article hinn/-inn ‘the’ but could alternatively be realized with the demonstrative pronoun sá ‘that’, Figure 10.2 includes hinn, -inn, and sá for comparison, normalized per 100,000 words. The results in Figure 10.2 suggest that the definite article in the oldest prose was used much less frequently than in later periods. The near identical frequency of sá and -inn could also be taken as evidence that the definite article fails Van de Velde’s (2010: 268) exclusiveness criterion, where the demonstrative sá denotes a similar function. Further developments indicate that sá and the bound marker -inn become more divergent and that the latter is increasing in frequency, again, in line with the above claims that it changes from being optional to being obligatory in most contexts. The sharp difference in frequency between sá and -inn from the 13th century onwards (cf. Figure 10.2) and the sharply increasing use of the bound definite marker -inn in contexts with adjectival modification (at the expense of the free form hinn) throughout the documented history of Icelandic (cf. Figure 10.1) point towards an analysis where what used to be an adjunct of NP is becoming an integral part—the head—of a functional projection for definiteness, the DP.

fact that the D A N pattern gave way to a pattern featuring the bound form of the article. Interestingly, A N-D in both Old Norse periods feature the strong form of the adjective, whereas the N-D A pattern occurs with strongly and weakly inflected adjectives.

ICEL ANDIC DATIVE APPLICATIVES hinn (free)

–inn (bound)

sá

17th

19th

241

4500 4000 3500 3000 2500 2000 1500 1000 500 0

12th

13th

14th

15th

16th

18th

20th

21st

Fig. 10.2 Frequency of use normalized per 100,000 words. Demonstrative/article sá ‘that’ vs. demonstrative/article hinn ‘the (free)’ and -inn ‘the (bound)’.

10.4.3 Double genitives with transitive deverbal nouns (and their equivalents) One aspect that is of potential interest with regard to NP/DP structure is the absence of double adnominal genitives with transitive deverbal nouns or a corresponding structure headed by a dummy preposition (see Lander and Haegeman 2013 for discussion and references). The relevant construction is exemplified on the basis of two DP languages in (49), German and Modern Icelandic, respectively (cf. Lander and Haegeman 2013: 301f.): (49)

a. Hannibals Eroberung Roms (German) Hannibal.GEN conquest Rome.GEN ‘Hannibal’s conquest of Rome’ (Boškovic´ 2012: 188) b. Lausn Péturs á vandamálinu (Icelandic) solution Peter.GEN of problem.the.DAT ‘Peter’s solution of the problem’ (Sigurðsson 2006: 15)

According to Lander and Hageman (2013: 303), with references to Faarlund (2004, p.c.), Old Norse lacked both of the structural options in (49). Instead, these structures would typically be realized as compound nouns (see Lander and Haegeman 2013: 303, ex. (57)). The reason for this is that (49) would require a more elaborated DP structure which is claimed to be absent in NP

242

HEIMIR F. VIÐARSSON

languages (cf. e.g. Boškovic´ 2012). As it turns out, this claim is not entirely unproblematic even as far as Old Norse is concerned, although it may be no coincidence that potential attestations of (49b) in Old Norse are found in texts that date to the 13th century (or later). Some examples are shown in (50) below:¹³ (50) a. Bauð Egill dóm Sturlu á málum ordered Egill.NOM judgment.ACC Sturla.GEN of/on matters.DAT þeira. their.GEN ‘Egill ordered for the verdict of Sturla of/on their matters.’ (STU 252, ms. c.1696) b. Ok bauð Þorgils þá eindæmi biskups á and ordered Þorgils then sole.judgment bishop.GEN of/on ǫllum þeim málum, sem … all these cases REL … ‘And Þorgils ordered for the ultimate verdict of/on the bishop (i.e. his right to dictate terms on all these matters, which …)’ (STU 283, ms. c.1696) c. Þórormr bauð dóm Þórdı´sar á málinu Þórormr ordered judgment Þórdı´s.GEN of/on matter.the ‘Þórormr ordered for the verdict of Þórdís of/on the matter.’ (VATN 103, ms. c.1700) If á in these examples is indeed to be regarded as the dummy preposition of a transitive deverbal noun, as opposed to a contentful preposition (meaning ‘on’) that presumably would not fall within the scope of Boškovic´’s hypothesis, ¹³ Note that although the examples in (50) are found in manuscripts from around 1700, the text originally stems from the 13th century. The following Sturlunga saga examples, preserved in an older manuscript dating from the 14th century, are presumably not of this type: (i)

a. oc cveds vilia hans dom a sinv mali and says.MID want his judgment.ACC of/on his.REFL case ‘… and says hei wants hisj judgment of/on hisi case.’ (STU 75, ms. c.1350–1370) b. qvaz illt þickia at hafa forboð biskupa a sér said.MID bad feel to have interdiction bishops of/on him.REFL ‘… said (hei ) disliked having the interdiction of bishops on/of him(self )i .’ (STU 217, ms. c.1350–1370)

While the preposition a (=á ‘on, of ’) in (i-a) might be regarded as a dummy preposition of the transitive deverbal noun (=‘of ’), the form hans, the genitive of hann ‘he’, here serves as a possessive pronoun and is thus hardly the equivalent of a double genitive. In (i-b), it is more likely that a is a contentful preposition (meaning ‘on’) rather the dummy marker we’re after.

ICEL ANDIC DATIVE APPLICATIVES

243

it might hint at the innovation of uses that suggest an emerging DP structure already in Late Old Norse. More research is clearly needed to establish a link between the rise of the article in Old Norse and the loss of non-thematic datives. However, what I hope to have shown is that the Old Norse case system really is different from the Modern Icelandic system in a fundamental way. Old Norse appears to provide empirical support for two separate Appl projections, a high and a low Appl, each with different semantics, which can be spelled out morphologically. Whereas Old Norse had a rich system of morphological case marking on applicatives, Modern Icelandic mostly spells out these relations as prepositions or resorts to possessive pronouns. Increasing NP configurationality was suggested as a potential trigger for this change. Although much is still unclear regarding the timing of the reanalysis argued to have taken place within the NP, this approach clearly fares better and is superior to the common alternative to relate the changes in question to deflection. Moreover, although there is an interesting overlap, the claim that Icelandic has been becoming ‘more configurational’ should not be equated with Faarlund’s (1990) stronger claim that Old Norse was non-configurational or that oblique subjects are purely a modern phenomenon (Faarlund 1990, 2001, 2004). The way I see it, the partial fusion of a productive system of applied datives and the oblique subject construction may have been facilitated, or made possible, by the fact that oblique subjects already existed as a construction in Old Norse (cf. e.g. Ro¨gnvaldsson 1995; Eythórsson and Barðdal 2005). The structural ambiguity often observed between experiencers, benefactives, and possessives likely paves the way for potential reanalysis in these contexts.

10.5 Concluding remarks In this chapter it has been argued that Old Norse could license ‘free’ datives in two distinct structural positions, a high position and a low position. The high applicative position is associated directly with an event denoting an affected reading without implying possession, whereas the low applicative position typically denotes a possessive relation between the dative and the theme. Old Norse provides empirical evidence for both of these Appl positions in the syntactic structure in that both positions could be filled at the same time with a bare dative, where each dative has a distinct reference and need not be coreferential with the subject. These complex constructions thus reveal not only that theoretical constructs in terms of two distinct structural positions (rather

244

HEIMIR F. VIÐARSSON

than merely different senses) are empirically motivated, but also that the licensing of datives in Old Norse is very unlike that of Modern Icelandic. In Modern Icelandic, these high applicatives must be spelled out as a prepositional phrase, whereas morphological (bare) datives are mostly confined to specific constructions where they are licensed as (oblique) subjects. Moreover, the external possessive dative constructions found in Old Norse are realized as a part of the NP/DP in Modern Icelandic, with an NP-internal possessive pronoun much as is typical in the rest of the Scandinavian languages, English, and Dutch. The loss of morphological case marking has traditionally been regarded as a potential explanation for the limited scope of dative benefactives in various related languages. The hypothesis that increasing NP/DP configurationality resulted in the absorption of datives into the nominal domain (cf. Van de Velde and Lamiroy 2017) is a potentially promising alternative to deflection, which is hardly applicable in the case of Icelandic. Regardless of whether these changes in the licensing of datives and NP/DP configurationality can be linked, it is a fact that Old Norse exhibits various non-configurational traits that were later lost. The temporal increase in the use of definite marking during the Old Norse period is also consistent with the claim that the definite marker was not yet a fully grammaticalized definite article in Old Norse. Based on the approach of Van de Velde (2010); Van de Velde and Lamiroy (2017), and data obtained mainly from the IcePaHC and MÍM corpus of Old Icelandic, it is therefore argued that the rise of obligatory definite determiners in D in the history of Icelandic may at least go some way towards accounting for these changes. One aspect that has not been addressed yet is that the loss of ‘free’ datives in the history of Icelandic did not only involve possessive uses of the dative, but also the affected (=high) dative applicative. It is not immediately obvious on this account that this should necessarily follow since the rise in NP/DP configurationality and subsequent absorption of the dative into the extended NP was facilitated by the shared meaning of the NP-internal and NP-external possessors. However, a solution to this problem might be found in Modern Icelandic NP-internal applicatives of the kind discussed by Ingason (2016), exemplified in (51) (cf. Ingason 2016: 86): (51) Þeir do¨nsuðu [ til skemmtunar ]. they danced [ girls.the.DAT for entertainment.GEN girls.the.DAT] ‘They danced for the girls’ entertainment.’ Ingason (2016: 83–8) argues that the dative applicative in (51) is a part of the noun phrase (his nP), although it can move to the specifier of PP til ‘for’

ICEL ANDIC DATIVE APPLICATIVES

245

(as indicated by ‘< >’). The applicative is analysed by Ingason (2016: 78) as an applicative experiencer (Applexp ), but is considered to have the semantics of the high event applicative of Pylkka¨nen (2008). Assuming that this is so, the clause-level high applicative datives can now also potentially be subsumed under a similar absorption account, that is, into an NP-internal position, but where it retains its morphological dative form. More research is clearly needed to evaluate these observations and claims, as well as the intermediate steps of the linguistic cycle, to be able to develop this account further. An important aspect is how to evaluate the scope of the variation discussed in this chapter. While I believe the discussion has demonstrated beyond doubt that the licensing of datives in Old Norse goes a long way towards that which has been shown for German, it is still not certain at this point how to define it in terms of the extent (cf. also Allen 2019 in the context of Old English). As regards ‘free’ datives, are we dealing with a fully productive phenomenon (or phenomena) in Old Norse or were some of these patterns already a part of the thematic grid of the lexical verbs with which they occur? Because of the semantic and structural ambiguity present in many of the examples, this is not as straightforward to evaluate as one would wish. Despite these obvious shortcomings, it is my hope the data discussed in this chapter will help pave the way for a systematic analysis of dative applicatives in the history of Icelandic and other languages.

Texts ÆV Íslendzk æventýri: Isla¨ndische Legenden, Novellen und Ma¨rchen I. 1882. Hugo Gering (ed.). Halle. EB Eyrbyggja saga: The Vellum tradition. 2003. Forrest S. Scott (ed.). Editiones Arnamagnæanæ A 18. C.A. Reitzels forlag, København. EG Egils saga Skallagrı´mssonar I. 2001. A-Redaktionen. Bjarni Einarsson (ed.). Editiones Arnamagnæanæ A 19. C.A. Reitzels forlag, København. Gı´SL Gı´sla saga Súrssonar. 1929. Finnur Jónsson (ed.). Det kongelige nordiske oldskrift-selskab, København. HKR Heimskringla I–IV. 1893–1900. Nóregs konunga sǫgur af Snorri Sturluson. Finnur Jónsson (ed.). Samfund til udgivelse af gammel nordisk litteratur 23. S. L. Møllers bogtrykkeri, København. HRAFN Sagan af Hrafnkeli Freysgoða. 1847. Konráð Gı´slason (ed.). Det nordiske Literatur-Samfund. Brødrene Berlings Bogtrykkeri, Kjøbenhavn. [2nd edition.]

246

HEIMIR F. VIÐARSSON

ÓH [Ólafs saga helga]: Den store saga om Olav den hellige I–II. 1941. Efter pergamentshåndskrift i kungliga biblioteket i Stockholm nr. 2 4to med varianter fra andre håndskrifter. Kjeldeskriftfondet. Oscar Albert Johnsen & Jón Helgason (eds.). Jacob Dybwad, Oslo. MAR Mariu saga. 1871. Legender om Jomfru Maria og hendes Jertegn. C. R. Unger (ed.). Det norske Oldskriftselskabs Samlinger 11–16. Christiania. POST Postola so¨gur. 1874. Legendariske Fortællinger om Apostlernes Liv, deres Kamp for Kristendommens Udbredelse samt deres Martyrdød. C. R. Unger (ed.). B. M. Bentzen, Christiania. STU Sturlunga saga I–II. 1906–11. Efter membranen Króksfjarðarbók udfylt efter Reykjarfjarðarbók. Kr. Kålund (ed.). Det kongelige nordiske oldskrift-selskab. Gyldendanske boghandel—Nordisk forlag, København og Kristiania. ÞIÐR Þiðriks saga af Bern I–II. 1905–11. Henrik Bertelsen (ed.). Samfund til udgivelse af gammel nordisk litteratur 34. S. L. Møller, København. VATN Vatsdælasaga. 1934. Finnur Jónsson (ed.). Samfund til udgivelse af gammel nordisk litteratur 58. J. Jørgensen, København.

References Abels, Klaus (2012). Phases: An Essay on Cyclicity in Syntax. Berlin: De Gruyter. Adams, James (2005). ‘The Accusative + Infinitive and Dependent quod-/quia-Clauses: The Evidence of Non-Literary Latin and Petronius’, in Sándor Kiss, Luca Mondin, and Giampaolo Salvi (eds), Latin et langues romanes: Etudes de linguistique offertes à József Herman à l’occasion de son 80ème anniversaire. Tu¨bingen: Niemeyer, 195–206. Åfarli, Tor (1986). ‘Some Syntactic Structures in a Dialect of Norwegian’, Trondheim Working Papers in Linguistics 3: 93–111. Aitchison, Jean (1980). ‘Review of Lightfoot (1979)’, Linguistics 18: 137–46. Alboiu, Gabriela (2002). The Features of Movement in Romanian. Bucharest: Editura Universității din București. Alexiadou, Artemis and Elena Anagnostopoulou (1998). ‘Parametrizing AGR: Word Order, Movement and EPP-checking’, Natural Language and Linguistic Theory 16: 491–539. Alexiadou, Artemis, Tibor Kiss, and Gereon Mu¨ller (2012). ‘Local Modelling of NonLocal Dependencies in Syntax: An Introduction’, in Artemis Alexiadou, Tibor Kiss, and Gereon Mu¨ller (eds), Local Modelling of Non-Local Dependencies in Syntax. Tu¨bingen: Niemeyer, 1–48. Allen, Cynthia (1975). ‘Old English Modals’, in Jane Grimshaw (ed.), Papers in the History and Structure of English (Occasional Papers in Linguistics 1), 89–100. Amherst: University of Massachusetts. Allen, Cynthia L. (2019). Dative External Possessors in Early English. Oxford: Oxford University Press. Allen, J. H. and J. B. Greenough (2001). [1888] New Latin Grammar. Boston: Gin & Company. Ambar, Manuela (1992). Para uma Sintaxe da Inversão Sujeito-Verbo em Português. Lisboa: Edições Colibri. Bacskai-Atkari, Julia (2012). ‘The Diachronic System of the Left Periphery of Subordinate Clauses in Hungarian’, in Balázs Surányi (ed.), Proceedings of the Second Central European Conference in Linguistics for Postgraduate Students. Budapest: Pázmány Péter Catholic University, 3–23. Bacskai-Atkari, Julia (2014a). The Syntax of Comparative Constructions: Operators, Ellipsis Phenomena and Functional Left Peripheries. Potsdam: Universita¨tsverlag Potsdam. Bacskai-Atkari, Julia (2014b). ‘Cyclical Change in Hungarian Comparatives’, Diachronica 31.4: 465–505. Bacskai-Atkari, Julia (2016a). ‘Towards a Cross-linguistic Typology of Marking Polarity in Embedded Degree Clauses’, Acta Linguistica Hungarica 63.4: 389–409. Bacskai-Atkari, Julia (2016b). ‘On the Diachronic Development of a Hungarian Declarative Complementiser’, Transactions of the Philological Society 114.1: 95–116.

248

REFERENCES

Bacskai-Atkari, Julia (2018). ‘Marking Finiteness and Low Peripheries’, in Huba Bartos et al. (eds), Boundaries Crossed, at the Interfaces of Morphosyntax, Phonology, Pragmatics and Semantics. Dordrecht: Springer, 183–98. Bacskai-Atkari, Julia (2020). ‘German V2 and Doubly Filled COMP in West Germanic’, The Journal of Comparative Germanic Linguistics 23: 125–60. Bacskai-Atkari, Julia and Éva Dékány (2014). ‘From Non-finite to Finite Subordination: The History of Embedded Clauses’, in Katalin É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax. Oxford: Oxford University Press, 148–223. Baker, Mark (1985). ‘Mirror Theory and Morphosyntactic Explanation’, Linguistic Inquiry 15: 373–415. Baker, Mark (1988). Incorporation: A Theory of Grammatical Function Changing. Chicago/London: The University of Chicago Press. Baker, Mark (1996). The Polysynthesis Parameter. New York: Oxford University Press. Baker, Mark (2003). Lexical Categories: Verbs, Nouns and Adjectives. Cambridge: Cambridge University Press. Baker, Mark (2008a). The Syntax of Agreement and Concord. Cambridge: Cambridge University Press. Baker, Mark (2008b). ‘The Macroparameter in a Microparametric World’, in Theresa Biberauer (ed.), The Limits of Syntactic Variation. Amsterdam/Philadelphia: John Benjamins, 351–74. Baltin, Mark (2010). ‘The Nonreality of Doubly Filled Comps’, Linguistic Inquiry 41.2: 331–5. Barbato, Marcello (2001). Il libro viii del Plinio napoletano a cura di Giovanni Brancati. Napoli: Liguori. Barbiers, Sjef, Ellen Brandner, Olaf Koeneman, Marika Lekakou, Cecilia Poletto, Martin Salzmann, and Gerhard Schaden (2011). Mesocomparative Syntax of Perfect Doubling. Ms. Barðdal, Jóhanna (2007). ‘The Semantic and Lexical Range of the Ditransitive Construction in the History of (North) Germanic’, Functions of Language 14.1: 9–30. Barðdal, Jóhanna (2009). ‘The Development of Case in Germanic’, in Jóhanna Barðdal and Shobhana Chelliah (eds), The Role of Semantic, Pragmatic and Discourse Factors in the Development of Case. Amsterdam: John Benjamins, 123–59. Barðdal, Jóhanna (2011). ‘The Rise of Dative Substitution in the History of Icelandic: A Diachronic Construction Grammar Approach’, Lingua 121.1: 60–79. Bauer, Brigitte (2000). Archaic Syntax in Indo-European: The Spread of Transitivity in Latin and French. Berlin/New York: Mouton de Gruyter. Bayer, Josef (2004). ‘Decomposing the Left Periphery: Dialectal and Cross-linguistic Evidence’, in Horst Lohnstein and Susanne Trissler (eds), The Syntax and Semantics of the Left Periphery. Berlin/Boston: Mouton de Gruyter, 59–95. Bayer, Josef and Ellen Brandner (2008). ‘On Wh-Head-Movement and the Doubly-FilledComp Filter’, in Charles B. Chang and Hannah J. Haynie (eds), Proceedings of the 26th West Coast on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project, 87–95. Behaghel, Otto (1924). Deutsche Syntax III. Heidelberg: Carl Winter. Belletti, Adriana (2004). ‘A spects of the Low IP Area’, in Luigi Rizzi (ed.), The Structure of CP and IP: The Cartography of Syntactic Structures. Vol. 2. Oxford: Oxford University Press, 16–51. Belletti, Adriana (2008). ‘The CP of Clefts’, Rivista di Grammatica Generativa 33: 191–204.

REFERENCES

249

Belletti, Adriana, Naama Friedmann, Dominique Brunato, and Luigi Rizzi (2012). ‘Does Gender Make a Difference? Comparing the Effect of Gender on Children’s Comprehension of Relative Clauses in Hebrew and Italian’, Lingua 122: 1053–69. Benincà, Paola and Guglielmo Cinque (2010). ‘La frase relativa’, in Renzi Lorenzo and Giampaolo Salvi (eds), Grammatica dell’italiano antico. Vol. I. Bologna: Il Mulino, 469–507. Biberauer, Theresa (2003). Verb Second (V2) in Afrikaans: A Minimalist Investigation of Word Order Variation. Ph.D. dissertation. University of Cambridge. Biberauer, Theresa and Marc Richards (2006). ‘True Optionality: When the Grammar Doesn’t Mind’, in Cedric Boeckx (ed.), Minimalist Essays. Amsterdam: John Benjamins, 35–67. Biberauer, Theresa and Ian Roberts (2005). ‘Changing EPP Parameters in the History of English: Accounting for Variation and Change’, English Language and Linguistics 9: 5–46. Biberauer, Theresa and George Walkden (2015). ‘Introduction: Changing Views of Syntactic Change’, in Theresa Biberauer and George Walkden (eds), Syntax over Time: Lexical, Morphological and Information-structural Interactions. Oxford: Oxford University Press, 1–13. Biberauer, Theresa, Anders Holmberg, and Ian Roberts (2007). ‘Disharmonic Wordorder Systems and the Final-over-Final-Constraint (FOFC)’, in Antonietta Bisetto and Francesco Barbieri (eds), Proceedings of XXXIII Incontro di Grammatica Generativa, Università di Bologna, 86–105. Biberauer, Theresa, Anders Holmberg, and Ian Roberts (2014). ‘A Syntactic Universal and its Consequences’, Linguistic Inquiry 45: 169–225. Biville, Frédérique (1995). ‘Énoncés factitifs en latin. Syntaxe et sémantique’, in Dominique Longrée (ed.), De usu. Etudes de syntaxe latine offertes en hommage à Marius Lavency. Louvain-la-Neuve: Peeters, 31–44. Bjarnadóttir, Kristı´n (2011) [1989]. Dativus sympatheticus. Ms., University of Iceland. URL: . [Original version 1989, published online with updated layout in 2011.] Bogdanow, Fanni (ed.) (1991–2006). La version Post-Vulgate de la Queste del saint Graal et de la Mort Artu. Troisième partie du Roman du Graal, 5 vols. Paris: Picard/Société des Anciens Textes Français. Bolkestein, A. Machtelt (1976). ‘AcI and ut-Clauses with Verba Dicendi in Latin’, Glotta 54: 263–91. Bolkestein, A. Machtelt (1979). ‘Subject to Object Raising in Latin?’, Lingua 48: 15–34. Boneh, Nora and Léa Nash (2012). ‘Core and Noncore Datives in French’, in Beatriz Fernandez and Ricardo Etxepare (eds), Variation in Datives: A Microcomparative Perspective. Oxford/New York: Oxford University Press, 22–49. Borer, Hagit (1984). Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris. Bo¨rjars, Kersti and Nigel Vincent (2011). ‘Grammaticalization and Directionality’, in Heiko Narrog and Bernd Heine (eds), The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, 163–76. Bo¨rjars, Kersti, Pauline Harries, and Nigel Vincent. (2016). ‘Growing Syntax: The Development of a DP in North Germanic’, Language 92.1: e1–e37. Boškovic´, Željko (2008). ‘What Will You Have, DP or NP?’, Proceedings of NELS 37: 101–14.

250

REFERENCES

Boškovic´, Željko (2009). ‘More on the No-DP Analysis of Article-less Languages’, Studia Linguistica 63.2: 187–203. Boškovic´, Željko (2011). ‘Rescue by PF Deletion, Traces as (Non)interveners, and the Thattrace Effect’, Linguistic Inquiry 42: 1–44. Boškovic´, Željko (2012). ‘On NPs and Clauses’, in Gu¨nther Grewendorf and Thomas Ede Zimmermann (eds), Discourse and Grammar: From Sentence Types to Lexical Categories. Berlin/Boston: Mouton De Gruyter, 179–242. Boškovic´, Željko (2013). ‘Phases beyond Clauses’, in Lilia Schu¨rcks, Anastasia Giannakidou, and Urtzi Etxeberria (eds), The Nominal Structure in Slavic and beyond. Berlin/Boston: de Gruyter, 75–128. Boškovic´, Željko and Nunes, Jairo (2007). ‘The Copy Theory of Movement: A View from PF’, in Norber Corver and Jairo Nunes (eds), The Copy Theory of Movement. Amsterdam/Philadelphia: John Benjamins, 13–74. Bottari, Piero, Paola Cipriani, Anna Maria Chilosi, and Lucia Pfanner (1998). ‘The Determiner System in a Group of Italian Children with SLI’, Language Acquisition 7: 285–315. Botwinik-Rotem, Irena (2004). The Category P. Features, Projections, Interpretation. Ph.D. dissertation. Tel-Aviv University. Bouchard, Denis and Paul Hirschbu¨hler (1986). ‘French Quoi and Its Clitic Allomorph Que’, in C. Neidle and R. A. Nuñez Cedeño (eds), Studies in Romance Languages. Dordrecht: Foris Publications, 39–60. Bourciez, Édouard (1946). Éléments de linguistique romane. Paris: Librarie C. Klincksieck. Bourciez, Édouard (1956). Éléments de linguistique romane (quatrième édition, revisée par l’auteur et par les soins de Jean Bourciez). Paris: Librarie C. Klincksieck. Braˆncus¸, Grigore (2007). [1967] ‘O concordanță gramaticală romaˆno-albaneză. Modul supin’, in Studii de istorie a limbii romaˆne. Bucharest: Editura Academiei Romaˆne, 167–73. Brea, Mercedes et al. (eds) (1996). Lı´rica profana galego-portuguesa. Santiago de Compostela: Xunta de Galicia / Centro de Investigacións Lingu¨´ısticas e Literarias Ramón Piñeiro. Breitbarth, Anne (2014). ‘Exceptive Negation in Historical Low German’. Paper presented at GLOW 37 (Brussels, April), and DiGS 16 (Budapest, July). Breitbarth, Anne (2015). ‘Exceptive Negation in Middle Low German’, in Ellen Brandner, Anna Czypionka, Constantin Freitag, and Andreas Trotzke (eds), Webschrift for Josef Bayer. Konstanz: Universita¨t Konstanz, 11–15. Brody, Michael (1990). ‘Remarks on the Order of Elements in the Hungarian Focus Field’, in István Kenesei (ed.), Approaches to Hungarian 3: Structures and Arguments. Szeged: JATE Press, 95–112. Brucart, José Marı´a (1999). ‘La ellipsis’, in Ignacio Bosque and Violeta Demonte (eds), Gramática descriptiva de la lengua española. Madrid: Espasa Calpe, 2787–863. Buchwald-Wargenau, Isabel (2012). Die doppelten Perfektbildungen im Deutschen. Eine diachrone Untersuchung. Berlin: De Gruyter. Burridge, Kate (1993). Syntactic Change in Germanic: Aspects of Language Change in Germanic with Particular Reference to Middle Dutch. Amsterdam/Philadelphia: John Benjamins. Busquets, Joan (2006). ‘Stripping vs. VP-Ellipsis in Catalan: What Is Deleted and When’, Probus 18: 159–87. Butt, Miriam and Aditi Lahiri (2013). ‘Diachronic Pertinacity of Light Verbs’, Lingua 135: 7–29.

REFERENCES

251

Caragiu-Marioțeanu, Matilda (1962). ‘Moduri nepersonale’, Studii S¸i Cercetări Lingvistice 13: 29–43. Cardinaletti, Anna (2004). ‘Towards a Cartography of Subject Positions’, in Luigi Rizzi (ed.), The Cartography of CP and IP. Oxford: Oxford University Press, 115–65. Cardoso, Adriana (2010). Variation and Change in the Syntax of Relative Clauses: New Evidence from Portuguese. Ph.D. dissertation. Universidade de Lisboa, Faculdade de Letras. Cardoso, Adriana (2017). Portuguese Relative Clauses in Synchrony and Diachrony. Oxford/New York: Oxford University Press. Castro, Ivo (1983). ‘Sobre a data da introdução na Penı´nsula Ibérica do ciclo arturiano da post-vulgata’, Boletim de Filologia 28: 81–98. Castro, Ivo (1988). ‘Remarques sur la tradition manuscrite de l’Estoire del Saint Graal’, in Dieter Kremer (ed.), Homenagem a Joseph-Maria Piel por ocasião do seu 85º aniversário. Tu¨bingen: Max Niemeyer Verlag, 195–206. Cecchetto, Carlo and Renato Oniga (2002). ‘Consequences of the Analysis of Latin Infinitival Clauses for the Theory of Case and Control’, Lingue e Linguaggio 1: 151–89. Chamberlain, Jeffrey (1986). Latin Antecedents of French Causative Faire. Frankfurt: Lang. Chierchia, Gennaro (1985). ‘Formal Semantics and the Grammar of Predication’, Linguistic Inquiry 16: 417–43. Chomsky, Noam (1981). Lectures on Government and Binding: The Pisa Lectures. Dordrecht: Foris. Chomsky, Noam (1995). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam (2001). ‘Derivation by Phrase’, in Michael Kenstowicz (ed.), Ken Hale: A Life in Language. Cambridge, MA: MIT Press, 1–52. Chomsky, Noam (2008). ‘On Phases’, in Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizaretta (eds), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: MIT Press, 133–66. Chomsky, Noam and Howard Lasnik (1977) ‘Filters and Control’, Linguistic Inquiry 8.3: 425–504. Chung, Sandra and James McCloskey (1987). ‘Government, Barriers and Small Clauses in Modern Irish’, Linguistic Inquiry 18: 173–237. Cinque, Guglielmo (1978). ‘La sintassi dei pronomi relativi “cui” e “quale” nell’italiano contemporaneo’, Rivista di Grammatica Generativa 3: 31–126. Cinque, Guglielmo (1982). ‘On the Theory of Relative Clauses and Markedness’, The Linguistic Review 1: 247–94. Cinque, Guglielmo (1999). Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford/New York: Oxford University Press. Cinque, Guglielmo (2007). ‘A Note on Linguistic Theory and Typology’, Linguistic Typology 11: 93–106. Cinque, Guglielmo (2008). ‘More on the Indefinite Character of the Head of Restrictive Relatives’, Rivista di Grammatica Generativa 33: 3–24. Cinque, Guglielmo (2013). Typological Studies: Word Order and Relative Clauses. New York: Routledge. Cinque, Guglielmo and Luigi Rizzi (2009). ‘The Cartography of Syntactic Structures’, in Bernd Heine and Heiko Narrog (eds.), The Oxford Handbook of Linguistic Analysis. Oxford Handbooks Online. Cinque, Guglielmo and Luigi Rizzi (eds) (2010a). Mapping Spatial PPs: The Cartography of Syntactic Structures. Vol. 6. Oxford/New York: Oxford University Press.

252

REFERENCES

Cinque, Guglielmo and Luigi Rizzi (2010b). ‘The Cartography of Syntactic Structures’, in Bernd Heine and Heiko Narrog (eds.), The Oxford Handbook of Linguistic Analysis. Oxford/New York: Oxford University Press, 51–65. Ciutescu, Elena (2013a). ‘Micro-Parametric Variation in Romance Causative Constructions’, Bucharest Working Papers in Linguistics 15: 45–60. Ciutescu, Elena (2013b). ‘Remarks on the Infinitival Subject of Perception Verb Complements: Evidence for Two Syntactic Configurations’, Revue Roumaine de Linguistique 58: 299–312. Clahsen, Harald, Sonja Eisenbeiss, and Anne Vainikka (1994). ‘The Seeds of Structure: A Syntactic Analysis of the Acquisition of Case Marking’, in Teun Hoekstra and Bonnie Schwartz (eds), Language Acquisition Studies in Generative Grammar. Amsterdam: John Benjamins, 85–118. Clahsen, Harald, Sonja Eisenbeiss, and Martina Penke (1996). ‘Lexical Learning in Early Syntactic Development’, in Harald Clahsen (ed.), Generative Perspectives on Language Acquisition. Amsterdam: John Benjamins, 129–59. Colaço, Madalena (2005). Configurações de coordenação aditiva: Tipologia, concordaˆncia e extracção. Ph.D. dissertation. Universidade de Lisboa, Faculdade de Letras. Colleman, Timothy (2010). ‘Lectal Variation in Constructional Semantics: “Benefactive” Ditransitives in Dutch’, in Dirk Geeraerts, Gitte Kristiansen, and Yves Peirsman (eds), Advances in Cognitive Sociolinguistics. Berlin/New York: Mouton de Gruyter, 191–221. Cornilescu, Alexandra (2000). ‘The Double Subject Construction in Romanian’, in V. Motapanyane (ed.), Comparative Studies in Romanian Syntax. Oxford: Elsevier, 83–133. Cornilescu, Alexandra and Ruxandra Cosma (2011). ‘The Structure of Supine Relative Clauses’, The Annual Conference of the English Department, 3–4 June. Costa, João (1998). Word Order Variation: A Constraint-based Approach. The Hague: Holland Academic Graphics. Costa, João (2004). Subject Positions and Interfaces: The Case of European Portuguese. Berlin/New York: Mouton De Gruyter. Costa, João and Ana Maria Martins (2011). ‘On Focus Movement in European Portuguese’, Probus 23: 217–45. Costa, João and Maria Cristina Figueiredo Silva (2006). ‘On the (In)Dependence Relations between Syntax and Pragmatics’, in Valéria Molnar and Susanne Winkler (eds), The Architecture of Focus. Berlin/New York: Mouton De Gruyter, 83–104. Costa, João, Ana Maria Martins, and Fernanda Pratas (2012). ‘VP Ellipsis: New Evidence from Capeverdean Creole’, in Irene Franco, Sara Lusini, and Andrés Saab (eds), Romance Languages and Linguistic Theory. Vol. 4. Amsterdam/Philadelphia: John Benjamins, 155–75. Costantini, Francesco (2012). ‘Iubeo and Causative Structure’, in Renato Oniga, Rossella Iovino, and Giuliana Giusti (eds), Formal Linguistics and the Teaching of Latin. Newcastle upon Tyne: Cambridge Scholars Publishing, 101–15. Craenenbroeck, Jeroen van, and Anikó Lipták (2008). ‘On the Interaction between Verb Movement and Ellipsis: New Evidence from Hungarian’, in Charles B. Chang and Hannah J. Haynie (eds), Proceedings of the 26th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project, 138–46. Cruschina, Silvio (2008). Discourse-related Features and the Syntax of Peripheral Positions: A Comparative Study of Sicilian and other Romance Languages. Ph.D. dissertation. University of Cambridge.

REFERENCES

253

Cruschina, Silvio (2015). ‘Some Notes on Clefting and Fronting’, in Elisa Di Domenico, Cornelia Hamann, and Simona Matteini (eds), Structures, Strategies and Beyond. Studies in Honour of Adriana Belletti. Amsterdam/Philadelphia: John Benjamins, 181–208. Cuzzolin, Pierluigi (1991). ‘Sulle prime attestazioni del tipo sintattico dicere quod’, Archivio Glottologico Italiano 76: 26–78. Cuzzolin, Pierluigi (1994). Sull’origine della costruzione dicere quod. Aspetti sintattici e semantici. Firenze: La Nuova Italia. Cyrino, Sonia and Gabriela Matos (2006). ‘Null Complement Anaphora in Romance: Deep or Surface Anaphora?’, in Jenny Doetjes and Paz González (eds), Romance Languages and Linguistic Theory 2004: Selected Papers from ‘Going Romance’, Leiden, 9–11 December 2004. Amsterdam/Philadelphia: John Benjamins, 95–120. Dagnac, Anne (2010). ‘Modal Ellipsis in French, Spanish an Italian: Evidence for a TPdeletion Analysis’, in Karlos Arregi, Zsuzsanna Fagyal, Silvina A. Montrul, and Annie Tremblay (eds), Romance Linguistics 2008: Interactions in Romance. Selected Papers from the 38th Linguistics Symposium on Romance Languages (LSRL), Urbana-Champaign, April 2008. Amsterdam/Philadelphia: John Benjamins, 157–70. Danckaert, Lieven (2012). Latin Embedded Clauses: The Left Periphery. Amsterdam: John Benjamins. Danckaert, Lieven (2014). ‘The Derivation of Classical Latin Aux-final Clauses: Implications for the Internal Structure of the Verb Phrase’, in Karen Lahousse and Stefania Marzo (eds), Romance Languages and Linguistic Theory 2012. Amsterdam: John Benjamins, 141–59. Danckaert, Lieven (2017). The Development of Latin Clause Structure: A Study of the Extended Verb Phrase. Oxford: Oxford University Press. Danckaert, Lieven (2018). ‘The Decline of Latin VOAux: Neg-Incorporation and Syntactic Reanalysis’, in Ana Maria Martins and Adriana Cardoso (eds), Word Order Change. Oxford: Oxford University Press, 243–63. Danckaert, Lieven, Tijs D’Hulster, and Liliane Haegeman (2016). ‘Deriving Idiolectal Variation: English Wh-raising’, in Ermenegildo Bidese, Federica Cognola, and Manuela Caterina Moroni (eds), Theoretical Approaches to Linguistic Variation. Amsterdam: John Benjamins, 145–76. Davies, Mark (1995a). ‘The Evolution of the Spanish Causative Construction’, Hispanic Review 63: 57–77. Davies, Mark (1995b). ‘The Evolution of Causative Constructions in Spanish and Portuguese’, in Jon Amastae, Grant Goodall, Mario Montalbetti, and Marianne Phinney (eds), Contemporary Research in Romance Linguistics. Amsterdam: John Benjamins, 105–22. Davies, Mark (1996). ‘The Diachronic Evolution of the Causative Construction in Portuguese’, Journal of Hispanic Philology 17: 261–92. Davies, Mark (2000). ‘Syntactic Diffusion in Spanish and Portuguese Infinitival Complements’, in Steven Dworkin and Dieter Wanner (eds), New Approaches to Old Problems: Issues in Romance Historical Linguistics. Amsterdam: John Benjamins, 109–27. De Roberto, Elisa (2008). Le proposizioni relative con antecedente in italiano antico. Ms. Ph.D. dissertation. Università Roma Tre / Université Paris IV-Sorbonne. De Vries, Mark (2006). ‘The Syntax of Appositive Relativization: On Specifying Coordination, False Free Relatives, and Promotion’, Linguistic Inquiry 37: 229–70. Delfitto, Denis (2002). ‘On the Semantics of Pronominal Clitics and some of its Consequences’, Catalan Journal of Linguistics 1: 29–57.

254

REFERENCES

Delsing, Lars-Olof (2003). ‘Preposition Stranding and Case’, in Lars-Olof Delsing et al. (eds), Grammatik i fokus. Festskrift till Christer Platzack den 18 november 2003. Lunds: Institutionen fo¨r nordiska språk, Lunds universitet, 75–84. Demirdache, Hamida and Myriam Uribe-Etxebarria (2000). ‘The Primitives of Temporal Relations’, in Robert Martin, David Michaels, and Juan Uriagereka (eds), Step by Step: Essays in Minimalist Syntax in Honor of Howard Lasnik. Cambridge, MA: MIT Press. 157–86. Demirdache, Hamida and Myriam Uribe-Etxebarria (2007). ‘The Syntax of Time Arguments’, Lingua 117: 330–66. Den Dikken, Marcel (2010). ‘On the Functional Structure of Locational and Directional PPs’, in Guglielmo Cinque and Luigi Rizzi (eds), Mapping Spatial PPs: The Cartography of Syntactic Structures. Vol. VI. Oxford/New York: Oxford University Press, 74–126. Denison, David (1990). ‘Auxiliary + Impersonal in Old English’, Folia Linguistica Historica 9: 139–66. Denison, David (1993). English Historical Syntax. London: Longman. Densusianu, Ovid (1961). Istoria limbii romaˆne. Vol. II. Secolul al XVI-lea. Bucharest: Editura Științifică. Depiante, Marcela (2000). The Syntax of Deep and Surface Anaphora: A Study of Null Complement Anaphora and Stripping/Bare Argument Ellipsis. Ph.D. dissertation. University of Connecticut, Storrs. Depiante, Marcela (2001). ‘On Null Complement Anaphora in Spanish and Italian’. Probus 13: 193–221. Devine, Andrew M. and Laurence D. Stephens (2006). Latin Word Order: Structured Meaning and Information. Oxford/New York: Oxford University Press. Dieninghoff, Joseph (1904). Die Umschreibungen aktiver Vergangenheit mit dem Participium Praeteriti im Althochdeutschen. Bonn: C. Georgi. Diessel, Holger (2004). The Acquisition of Complex Sentences. Cambridge: Cambridge University Press. Dobrovie-Sorin, Carmen (1994). The Grammar of Romanian. Berlin: Mouton de Gruyter. Donati, Caterina and Carlo Cecchetto (2011). ‘Relabeling Heads: A Unified Account for Relativization Structures’, Linguistic Inquiry 42.4: 519–60. Dragomirescu, Adina (2011). ‘The Subject of the Supine Clause in Romanian and A-Chains’, Revue Roumaine de Linguistique 56: 571–92. Dragomirescu, Adina (2013). Particularităt¸i sintactice ale limbii romaˆne în context romanic. Supinul. Bucharest: Editura Muzeului Nat¸ional al Literaturii Romaˆne. Dragomirescu, Adina (2015). ‘Two DP Configurations for Supine-based Nouns’, in Virginia Hill (ed.), Formal Approaches to DP in Old Romanian. Leiden: Brill, 246–66. É. Kiss, Katalin (1998). ‘Identificational Focus versus Information Focus’, Language 74.2: 245–73. É. Kiss, Katalin (2002). The Syntax of Hungarian. Cambridge: Cambridge University Press. É. Kiss, Katalin (2008a). ‘Free Word Order, (Non)Configurationality, and Phases’, Linguistic Inquiry 39.3: 441–75. É. Kiss, Katalin (2008b). ‘The Structure of the Hungarian VP Revisited’, in Szilárd Szentgyo¨rgyi et al. (eds), Approaches to Hungarian 10. Budapest: Akadémiai Kiadó. 31–58. É. Kiss, Katalin (2013). ‘From Proto-Hungarian SOV to Old Hungarian Top Foc V X. Diachronica 30.2: 202–31.

REFERENCES

255

É. Kiss, Katalin (2014). ‘The Evolution of Functional Left Peripheries in the Hungarian Sentence’, in Katalin É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax. Oxford: Oxford University Press, 9–55. Egerland, Verner and Michela Cennamo (2010). ‘Frasi subordinate all’infinito’, in Giampaolo Salvi and Lorenzo Renzi (eds), Grammatica dell’italiano antico. Vol. 2. Bologna: Il Mulino, 817–79. Eide, Kristin Melum (2009). ‘Tense, Finiteness and the Survive Principle: Temporal Chains in a Crash-proof Grammar’, in Michael T. Putnam (ed.), Towards a Derivational Syntax: Survive-minimalism. Amsterdam: John Benjamins, 91–132. Ellegård, Alvar (1953). The Auxiliary Do: The Establishment and Regulation of its Use in English. Stockholm: Almqvist and Wiksell. Elsness, Johan (1997). The Perfect and the Preterite in Contemporary and Earlier English. Berlin/New York: Mouton de Gruyter. Eythórsson, Thórhallur and Jóhanna Barðdal. (2005). ‘Oblique Subjects: A Common Germanic Inheritance’, Language 81.4: 824–81. Faarlund, Jan Terje (1990). Syntactic Change: Toward a Theory of Historical Syntax. Berlin: Mouton De Gruyter. Faarlund, Jan Terje (2001). ‘The Notion of Oblique Subject and Its Status in the History of Icelandic’, in Jan Terje Faarlund (ed.), Grammatical Relations in Change. Amsterdam: John Benjamins, 99–135. Faarlund, Jan Terje (2004). The Syntax of Old Norse: With a Survey of the Inflectional Morphology and a Complete Bibliography. Oxford: Oxford University Press. Fanselow, Gisbert (1990). ‘On the Nature of Proper Government and Syntactic Barriers’, in Werner Abraham et al. (eds), Issues in Germanic Syntax. Berlin: Mouton De Gruyter, 33–48. Fanselow, Gisbert (1991). Minimale Syntax. Groningen: Rijksuniversiteit Groningen. Fiorentino, Giuliana (1998). ‘Clausola relativa debole e pronome relativo in italiano’, in Paolo Ramat and Elisa Roma (eds), Sintassi storica. Atti del XXX congresso internazionale della Società di Linguistica italiana. Pavia, 26–28 settembre 1996. Roma: Bulzoni. Formentin, Vittorio (1996). ‘Flessione bicasuale del pronome relativo in antichi testi italiani centro-meridionali’, Archivio Glottologico Italiano 81: 133–76. Fraˆncu, Constantin (2009). Gramatica limbii romaˆne vechi (1521–1780). Ias¸i: Casa Editorială Demiurg. Friedmann, Naama and Hedva Lavi (2006). ‘On the Order of Acquisition of A-movement, Wh-movement and V-C Movement’, in Adriana Belletti et al. (eds), Language Acquisition and Development. Cambridge: Cambridge Scholars Publishing, 211–17. Garrett, Andrew (1998). ‘On the Origin of Auxiliary Do’, English Language and Linguistics 2: 283–330. Ga¨rtner, Hans-Martins (2001). ‘On the Force of V2 Declaratives’, ZAS Papers in Linguistics 23: 103–9. Gelderen, Elly van (2004). Grammaticalization as Economy. Amsterdam: John Benjamins. Gelderen, Elly van (2007). ‘The Definiteness Cycle in Germanic’, Journal of Germanic Linguistics 19: 275–308. Gelderen, Elly van (2008). ‘Linguistic Cycles and Economy Principles: The Role of Universal Grammar in Language Change’, in Thórhalur Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers. Linguistik Aktuell 113. Amsterdam: John Benjamins, 245–64. Gelderen, Elly van (2009a). ‘Renewal in the Left Periphery: Economy and the Complementiser Layer’, Transactions of the Philological Society 107: 131–95.

256

REFERENCES

Gelderen, Elly van (2009b). ‘Feature Economy in the Linguistic Cycle’, in Paola Crisma and Giuseppe Longobardi (eds), Historical Syntax and Linguistic Theory. Oxford: Oxford University Press, 93–109. Gelderen, Elly van (2009c). ‘Grammaticalization from a Biolinguistic Perspective’, in Rudolf Botha and Chris Knight (eds), The Prehistory of Language. Oxford: Oxford University Press, 225–43. Gelderen, Elly van (2011a). ‘Grammaticalization and Generative Grammar: A Difficult Liaison’, in Heiko Narrog and Bernd Heine (eds), The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, 43–55. Gelderen, Elly van (2011b). The Linguistic Cycle: Language Change and the Language Faculty. Oxford/New York: Oxford University Press. Gelderen, Elly van (2015). ‘Forced Asymmetry and the Role of Features in Language Change’, Paper presented at the 37th DGfS, Leipzig University, 4–6 March. Ghet¸ie, Ion (1975). Baza dialectală a romaˆnei literare. Bucharest: Editura Academiei Romaˆne. Gillmann, Melitta (2011). ‘Die Grammatikalisierung des sein-Perfekts—Eine korpuslinguistische Untersuchung zur Hilfsverbselektion der Bewegungsverben im Deutschen’, Beitra¨ge zur Geschichte der deutschen Sprache und Literatur (PBB) 133. 2: 203–34. Ginsburg, Jason Robert (2009). Interrogative Features. Ph.D. dissertation. The University of Arizona. Giurgea, Ion and Elena Soare (2010). ‘Predication and the Nature of Non-finite Relatives in Romance’, in Anna Maria Di Sciullo and Virginia Hill (eds), Edges, Heads and Projections: Interface Properties, Amsterdam: John Benjamins, 191–214. Giusti, Giuliana (2001). ‘The Birth of a Functional Category: From Latin ILLE to the Romance Article and Personal Pronoun’, in Guglielmo Cinque and Giampaolo Salvi (eds), Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi. Amsterdam: Elsevier Science, 157–71. Goldbach, Maria (2003). ‘Repercussions of Morphology on Syntactic Checking: The Latin Accusativus cum Infinitivo’, Linguistische Berichte 193: 57–90. Goldbach, Maria (2008). ‘Morphological Developments Affecting Syntactic Change’, in Ulrich Detges and Richard Waltereit (eds), The Paradox of Grammatical Change: Perspectives from Romance. Amsterdam: John Benjamins, 85–106. Gonçalves, Anabela and Matos, Gabriela (2008). ‘Ellipsis and Restructuring in European Portuguese’, in Enoch O. Aboh, Elisabeth van der Linden, and Petra Sleeman (eds), Romance Languages and Linguistic Theory: Selected Papers from ‘Going Romance’, Amsterdam 2007. Amsterdam/Philadelphia: John Benjamins, 109–29. Granath, Solveig (2002). ‘The Position of the Adverb Certainly Will Make a Difference’, English Today 18: 25–30. Grandgent, Charles Hall (1908). An Introduction to Vulgar Latin. Boston: D. C. Heath & Co., Publishers. Greenberg, Joseph. (1978). ‘How Does a Language Acquire Gender Markers? Universals of Human Language’, in Joseph Greenberg (ed.), Universals of Human Language. Vol. 3, Stanford: Stanford University Press, 47–82. Grimm, Jacob (1837). Deutsche Grammatik. Vol. IV. Go¨ttingen: Dieterichsche Buchhandlung. Grimshaw, Jane (1991). Extended Projection. Ms. Brandeis University.

REFERENCES

257

Grimshaw, Jane (2000). ‘Locality and Extended Projection’, in Peter Coopmans, Martin Everaert, and Jane Grimshaw (eds), Lexical Specification and Insertion. Amsterdam: John Benjamins, 115–33. Grimshaw, Jane (2005). ‘Extended Projection’, Words and Structure. Stanford, CA: Center for the Study of Language and Information, 1–73. Grønvik, Ottar (1986). Über den Ursprung und die Entwicklung der aktiven Perfektund Plusquamperfektkonstruktionen des Hochdeutschen und ihre Eigenart innerhalb des germanischen Sprachraumes. Oslo: Solum Forlag. Guasti, Maria Teresa (1993). Causative and Perception Verbs: A Comparative Study. Torino: Rosenberg & Sellier. Guasti, Maria Teresa (2006). ‘Analytic Causatives’, in Martin Everaert and Henk van Riemsdijk (eds), The Blackwell Companion to Syntax. Vol. 1. Oxford: Blackwell, 142–72. Guasti, Maria Teresa and Anna Cardinaletti (2003). ‘Relative Clause Formation in Romance Child Production’, Probus 15: 47–89. Gyuris, Beáta (2017). ‘New Perspectives on Bias in Polar Questions: A Study of Hungarian -e’, International Review of Pragmatics 9.1: 1–50. Haeberli, Eric and Tabea Ihsane (2016). ‘Revisiting the Loss of Verb Movement in the History of English’, Natural Language and Linguistic Theory 34: 497–542. Haeberli, Eric and Richard Ingham (2007). ‘The Position of Negation and Adverbs in Early Middle English’, Lingua 117: 1–25. Haeberli, Eric and Susan Pintzuk (2012). ‘Revisiting Verb (Projection) Raising in Old English’, in Dianne Jonas, John Whitman, and Andrew Garrett (eds), Grammatical Change: Origins, Nature, Outcomes. Oxford: Oxford University Press, 219–38. Haegeman, Liliane (2006). ‘Conditionals, Factives and the Left Periphery’, Lingua 116: 1651–69. Haegeman, Liliane (2010a). ‘The Internal Syntax of Adverbial Clauses’, Lingua 120: 628–48. Haegeman, Liliane (2010b). ‘The Movement Derivation of Conditional Clauses’, Linguistic Inquiry 41: 595–621. Haegeman, Liliane (2012). Adverbial Clauses, Main Clause Phenomena and the Composition of the Left Periphery. Oxford/New York: Oxford University Press. Haegeman, Liliane, Ángel Jiménez-Fernández, and Andrew Radford (2014). ‘Deconstructing the Subject Condition in Terms of Cumulative Constraint Violation’, The Linguistic Review 31: 73–150. Haider, Hubert (2013). Symmetry Breaking in Syntax. Cambridge: Cambridge University Press. Hale, Ken (1983). ‘Warlpiri and the Grammar of Non-configurational Languages’, Natural Language and Linguistic Theory Linguistic Theory 1: 5–47. Hamann, Cornelia (1996). ‘Negation and Truncated Structures’, in Michelle Aldridge (ed.), Child Language. Clevedon: Multilingual Matters, 72–83. Han, Chung-hye (2002). ‘Interpreting Interrogatives as Rhetorical Questions’, Lingua 112: 201–29. Harley, Heidi (2002). ‘Possession and the Double Object Construction’, Linguistic Variation Yearbook 2: 29–68. Haugan, Jens (2001). Old Norse Word Order and Information Structure. Ph.D. dissertation. Norwegian University of Science and Technology. Havers, Wilhelm (1911). Untersuchungen zur Kasussyntax der indogermanischen Sprachen. Straßburg: Karl J. Tru¨bner. Heine, Bernd and Mechthild Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Busk.

258

REFERENCES

Heine, Bernd, Ulrike Claudi, and Friederike Hu¨nnemeyer (1991). Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hill, Virginia (2013a). ‘The Emergence of the Romanian Supine’, Journal of Historical Linguistics 3.2: 230–71. Hill, Virginia (2013b). ‘The Emergence of the Romanian Subjunctive’, The Linguistic Review 30.4: 1–37. Hill, Virginia and Gabriela Alboiu (2016). Verb Movement and Clause Structure in Old Romanian. Oxford: Oxford University Press. Hjartardóttir, Þóra Bjo¨rk (1993). Getið ´ı eyðurnar. Um eyður fyrir frumlo¨g og andlo¨g ´ı eldri ´ıslensku. [On null subjects and objects in Older Icelandic.] M.A. thesis, University of Iceland, Reykjavı´k. Hoeksema, Jack (1995). ‘The Semantics of Exception Phrases’, in Jaap Van Der Does and Jan Van Eijck (eds), Quantifiers, Logic and Languages. Stanford: CSLI Publications, 145–77. Hoekstra, Teun and Peter Jordens (1994). ‘From Adjunct to Head’, in Teun Hoekstra and Bonnie Schwartz (eds), Language Acquisition Studies in Generative Grammar. Amsterdam: John Benjamins, 119–49. Hole, Daniel (2005). ‘Reconciling “Possessor” Datives and “Beneficiary” Datives—Towards a Unified Voice Account of Dative Binding in German’, in Claudia Maienborn and Angelika Wo¨llstein (eds), Event Arguments: Foundations and Applications. Tu¨bingen: Max Niemeyer Verlag, 213–42. Holmberg, Anders (2000). ‘Deriving OV Order in Finnish’, in Peter Svenonius (ed.), The Derivation of VO and OV. Amsterdam: John Benjamins, 123–52. Holmberg, Anders (2003). ‘Questions, Answers, Polarity and Head Movement in Germanic and Finnish’, in Anne Dahl, Kristina Bentzen, and Peter Svenonius (eds.), Proceedings of the 19th Scandinavian Conference of Linguistics (= Nordlyd, University of Tomsø Working Papers on Language and Linguistics), 31.1: 88–115. Holmberg, Anders and Platzack, Christer (1995). The Role of Inflection in Scandinavian Syntax. Oxford/New York: Oxford University Press. Hopper, Paul J. and Elizabeth Closs Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Horvath, Julia (1986). FOCUS in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris. Hróarsdóttir, Þorbjo¨rg (2000). Word Order Change in Icelandic: From OV to VO. Linguistik Aktuell 35. Amsterdam/Philadelphia: John Benjamins. Hróarsdóttir, Þorbjo¨rg (2008a). ‘Types of DPs in OV Order’, Studia Linguistica 62.3: 261– 86. Hróarsdóttir, Þorbjo¨rg (2008b). ‘Verb Particles in OV/VO Word Order in Older Icelandic’, Working Papers in Scandinavian Syntax 82: 43–81. Hróarsdóttir, Þorbjo¨rg (2009). ‘OV Languages: Expressions of Cues’, in Roland Hinterho¨lzl and Svetlana Petrova (eds), Information Structure and Language Change: New Approaches to Word Order Variation in Germanic. Berlin/New York: Mouton De Gruyter, 67–90. Hu, Shenai, Anna Gavarrò, and Maria Teresa Guasti (2015). ‘Children’s Production of Head-final Relative Clauses: The Case of Mandarin’, Applied Psycholinguistics 37: 323– 46. Huddleston, Rodney and Geoffrey Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

REFERENCES

259

Hyams, Nina (1996). ‘The Underspecification of Functional Categories in Early Grammar’, in Harald Clahsen (ed.), Generative Perspectives on Language Acquisition. Amsterdam: John Benjamins, 91–128. Hyams, Nina (2012). ‘Eventivity Effects in Early Grammar: The Case of Non-finite Verbs’, First Language 32: 239–69. Iandolo, Antonio (2001). Parlare e scrivere in dialetto napoletano. Napoli: Tempolungo. Iatridou, Sabine, Elena Anagnostopoulou, and Roumyana Izvorski (2001). ‘Observations about the Form and Meaning of the Perfect’, in Michael Kenstowicz (ed.), Ken Hale: A Life in Language. Cambridge, MA: MIT Press, 153–204. Iliescu, Maria (1995). ‘Les racines latins du factitif roman’, in Louis Callebat (ed.), Latin vulgaire, latin tardif IV. Actes du 4e colloque international sur le latin vulgaire et tardif, Caen 2–5 septembre 1994. Hildesheim: Olms, 355–65. Ingason, Anton Karl (2016). Realizing Morphemes in the Icelandic Noun Phrase. Ph.D. dissertation. University of Pennsylvania. . Isac, Daniela and Edit Jakab (2004). ‘Mood and Force Features in the Languages of the Balkans’, in Olga Mišeska-Tomic´ (ed.), Balkan Syntax and Semantics. Amsterdam: John Benjamins, 315–338. Jacobson, Sven (1975). Factors Influencing the Placement of English Adverbs in Relation to Auxiliaries. Stockholm: Almqvist and Wiksell. Jacobson, Sven (1981). Preverbal Adverbs and Auxiliaries: A Study of Word Order Change. Stockholm: Almqvist and Wiksell. Jensen, Frede (1986). The Syntax of Medieval Occitan. Beihefte zur Zeitschrift fu¨r romanische Philologie, 208. Tu¨bingen: Max Niemeyer Verlag. Jensen, Frede (1990). Old French and Comparative Gallo-Romance Syntax. Tu¨bingen: Max Niemeyer Verlag. Jespersen, Otto (1924). The Philosophy of Grammar. London: Allen and Unwin. Jiménez-Fernández, Ángel L. (2009). ‘On the Composite Nature of Subject Islands: A PhaseBased Approach’, SKY Journal of Linguistics 22: 91–138. Jøhndal, Marius (2012). Non-finiteness in Latin. Ph.D. dissertation. University of Cambridge. Jónsson, Jóhannes Gı´sli (1991). ‘Stylistic Fronting’, Working Papers in Scandinavian Syntax 48: 1–43. Jónsson, Jóhannes Gı´sli (1997–8). ‘Sagnir með aukafallsfrumlagi’, Íslenskt mál og almenn málfræði 19–20: 11–43. Jónsson, Jóhannes Gı´sli (2008). ‘Preposition Reduplication in Icelandic’, in Sjef Barbiers et al. (eds), Microvariations in Syntactic Doubling. Bingley: Emerald Group Publishing Limited, 403–17. Jónsson, Jóhannes Gı´sli and Thórhallur Eythórsson (2005). ‘Variation in Subject Case Marking in Insular Scandinavian’, Nordic Journal of Linguistics 28.3: 223–45. Joseph, Brian D. (1983). The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General, and Historical Linguistics. Cambridge: Cambridge University Press. Julien, Marit (2001). ‘The Syntax of Complex Tenses’, The Linguistic Review 18: 125–67. Julien, Marit (2002). Syntactic Heads and Word Formation. Oxford/New York: Oxford University Press. Kántor, Gergely (2008). ‘On Hungarian Relative Operators’, The Even Yearbook 8: 1–12. Karttunen, Lauri (1977). ‘Syntax and Semantics of Questions’, Linguistics and Philosophy 1: 3–41.

260

REFERENCES

Kato, Mary A. and Ana Maria Martins (2016). ‘The Main Varieties of Portuguese: An Overview on Word Order’, in Leo Wetzels, Sergio Menuzzi, and João Costa (eds), Handbook of Portuguese Linguistics. Hoboken, NJ: Wiley-Blackwell, 15–40. Kayne, Richard S. (1975). French Syntax: The Transformational Cycle. Cambridge, MA: MIT Press. Kayne, Richard S. (1984). ‘Two Notes on the NIC’, in Richard Kayne, Connectedness and Binary Branching. Foris: Dordrecht, 23–46 (originally published in Adriana Belletti, Luciana Brandi, and Luigi Rizzi (eds), Theory of Markedness in Generative Grammar: Proceedings of the 1979 GLOW Conference. Pisa: Scuola Normale Superiore, 317–46). Kayne, Richard S. (1993). ‘Toward a Modular Theory of Auxiliary Selection’, Studia Linguistica 47: 3–31. Kayne, Richard S. (1994). The Antisymmetry of Syntax. Cambridge, MA: MIT Press. Kayne, Richard S. (1997). ‘The English Complementizer Of ’ , The Journal of Comparative Germanic Linguistics 1: 43–54. Kayne, Richard S. (2004). ‘Prepositions as Probes’, in Adriana Belletti (ed.), Structures and Beyond. Oxford/New York: Oxford University Press, 192–212. Kayne, Richard S. (2008). ‘Antisymmetry and the Lexicon’, Linguistic Variation Yearbook 8: 1–31. Kayne, Richard S. (2010). ‘Why Isn’t This a Complementizer?’ in Richard Kayne (ed.), Comparison and Contrasts. Oxford: Oxford University Press, 190–227. Keenan, Edward L. (2002). ‘Explaining the Creation of Reflexive Pronouns in English’, in Donka Minkova and Robert Stockwell (eds), Studies in the History of English: A Millennial Perspective. Berlin/New York: Mouton de Gruyter, 325–55. Keenan, Edward L. (2009). ‘Linguistic Theory and the Historical Creation of English Reflexives’, in Paola Crisma and Giuseppe Longobardi (eds), Historical Syntax and Linguistics Theory. Oxford: Oxford University Press, 17–40. Keenan, Edward L. and Bernard Comrie (1977). ‘Noun Phrase Accessibility and Universal Grammar’, Linguistic Inquiry 8: 63–99. Kemenade, Ans van (2012). ‘Rethinking the Loss of Verb Second’, in Terttu Nevalainen and Elizabeth Closs Traugott (eds), The Oxford Handbook of the History of English. Oxford: Oxford University Press, 823–34. Kemenade, Ans van and Bettelou Los (2006). ‘Discourse Adverbs and Clausal Syntax in Old and Middle English’, in Ans van Kemenade and Bettelou Los (eds), The Handbook of the History of English. Oxford: Blackwell, 224–48. Kenesei, István (1989). ‘Adjuncts and Arguments in VP focus in Hungarian’, Acta Linguistica Hungarica 45: 61–88. Kenesei, István (1994). ‘Subordinate Clauses’, in Ferenc Kiefer and Katalin É. Kiss (eds), The Syntactic Structure of Hungarian. San Diego: Academic Press, 275–354. Kern, J. H. (1912). De met het participium praeteriti omschreven werkwoordsvormen in ’t Nederlands. Amsterdam: Johannes Mu¨ller. Kiparsky, Paul (2011). ‘Grammaticalization as Optimization’, in Dianne Jonas, John Whitman, and Andrew Garrett (eds), Grammatical Change: Origins, Nature, Outcome. Oxford/New York: Oxford University Press, 15–51. Klein, Wolfgang (1992). ‘The Present Perfect Puzzle’, Language 68: 525–52. Klein, Wolfgang (1994). Time in Language. London: Routledge. Klima, Edward Stephan (1964). Studies in Diachronic Transformational Syntax. Ph.D. dissertation. Harvard University. Koeneman, Olaf, Marika Lekakou, and Sjef Barbiers (2011). ‘Perfect Doubling’, Linguistic Variation 11: 35–75.

REFERENCES

261

Koopman, Hilda (2000). ‘Prepositions, Postpositions, Circumpositions and Particles: The Structure of Dutch PPs’, in Hilda Koopman (ed.), The Syntax of Specifiers and Heads. London: Routledge, 204–60. Kroch, Anthony (1989). ‘Reflexes of Grammar in Patterns of Language Change’, Language Variation and Change 1: 199–244. Kroch, Anthony and Ann Taylor (2000). Penn-Helsinki Parsed Corpus of Middle English. Second edition. . Kroch, Anthony, Beatrice Santorini, and Ariel Diertani (2004). Penn-Helsinki Parsed Corpus of Early Modern English. . Kroch, Anthony, Beatrice Santorini, and Ariel Diertani (2010). Penn Parsed Corpus of Modern British English. . Kuryłowicz, Jerzy (1965). ‘The Evolution of Grammatical Categories’, Diogenes 51: 55–71. Labelle, Marie (1996). ‘Remarques sur les verbes de perception et la sous-categorisation’, Recherches Linguistiques de Vincennes 25: 83–106. Laka, Itziar (1990). Negation in Syntax: On the Nature of Functional Categories and Projections. Ph.D. dissertation. Massachusetts Institute of Technology. Lander, Eric T. and Liliane Haegeman (2013). ‘Old Norse as an NP Language: With Observations on the Common Norse and Northwest Germanic Runic Inscriptions’, Transactions of the Philological Society 112.3: 279–318. Larson, Richard (1985). ‘Bare NP-adverbs’, Linguistic Inquiry 16.4: 595–621. Larsson, Ida (2009). Participles in Time: The Development of the Perfect Tense in Swedish. Gothenburg: Gothenburg University. Larsson, Ida (2015). ‘The HAVE/BE Alternation in Scandinavian—Perfects, Resultatives and Unaccusative Structure’, in Rolf Kailuweit and Malte Rosemeyer (eds), Auxiliary Selection Revisited: Gradience and Gradualness. Berlin/Boston: De Gruyter, 145–82. Ledgeway, Adam (2009). Grammatica diacronica del napoletano. Tu¨bingen: Niemeyer Verlag. Ledgeway, Adam (2012). From Latin to Romance: Morphosyntactic Typology and Change. Oxford: Oxford University Press. Lehmann, Christian (1984). Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tu¨bingen: Narr. Leiss, Elisabeth (2007). ‘Covert Patterns of Definiteness/Indefiniteness and Aspectuality in Old Icelandic, Gothic, and Old High German’, in Elisabeth Stark, Elisabeth Leiss, and Werner Abraham (eds), Nominal Determination: Typology, Context Constraints, and Historical Emergence. Amsterdam: John Benjamins, 73–102. Li, Yafei (1990). ‘Xo -binding and Verb Incorporation’, Linguistic Inquiry 21: 399–426. Lightfoot, David (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, David (1991). How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press/Bradford Books. Lightfoot, David (1999). The Development of Language: Acquisition, Change, and Evolution. Oxford: Blackwell. Lightfoot, David (2006). ‘Cuing a New Grammar’, in Ans van Kemenade and Bettelou Los (eds), The Handbook of the History of English. Oxford: Blackwell, 24–44.

262

REFERENCES

Lindgren, Kaj B. (1957). Über den oberdeutschen Pra¨teritumsschwund. Helsinki: Academia Scientiarum Fennicae. Lipták, Anikó (2013). ‘The Syntax of Emphatic Positive Polarity in Hungarian: Evidence from Ellipsis’, Lingua 128: 72–94. Lipták, Anikó and Malte Zimmermann (2007). ‘Indirect Scope Marking Again: A Case for Generalized Question Formation’, Natural Language and Linguistic Theory 25.1: 103–55. Litvinov, Viktor P. and Vladimir Radcenko (1998). Doppelte Perfektbildungen in der deutschen Literatursprache. Tu¨bingen: Stauffenburg. Lobo, Maria and Ana Maria Martins (2017). ‘Subjects’, in Andreas Dufter and Elisabeth Stark (eds), Manual of Romance Morphosyntax and Syntax. Berlin/Boston: De Gruyter, 27–88. Lobo, Maria, and Ana Maria Martins (2020). ‘Estratégias de marcação de foco. Ordem dos constituintes frásicos e estruturas clivadas’, in Eduardo Paiva Raposo, Maria Fernanda Bacelar, Maria Antónia Mota, Luı´sa Segura, and Amália Mendes (eds), Gramática do Português. Vol. 3. Lisboa: Gulbenkian, 213–62. Longobardi, Giuseppe (1996). The Syntax of N-raising: A Minimalist Theory (OTS Working Papers in Linguistics 5). Utrecht: Research Institute for Language and Speech. López, Luis (1999). ‘VP-ellipsis in Spanish and English and the Features of Aux’, Probus 11: 263–97. McCoard, Robert W. (1978). The English Perfect: Tense-choice and Pragmatic Inferences. Amsterdam/New York: Elsevier. McFadden, Thomas (2006). ‘German Inherent Datives and Argument Structure’, in Daniel Hole, André Meinunger, and Werner Abraham (eds), Datives and Other Cases. Amsterdam: John Benjamins, 49–77. McFadden, Thomas (2009). ‘Structural Case, Locality, and Cyclicity’, in Kleanthes Grohmann (ed.), Explorations of Phase Theory: Features and Arguments. Berlin/New York: Mouton De Gruyter, 107–30. McFadden, Thomas and Artemis Alexiadou (2006). ‘Pieces of the Be Perfect in German and Older English’, in Donald Baumer, David Montero, and Michael Scanlon (eds), Proceedings of the 25th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Proceedings Project, 270–8. McFadden, Thomas and Artemis Alexiadou (2010). ‘Perfects, Resultatives, and Auxiliaries in Earlier English’, Linguistic Inquiry 41.3: 389–425. Mackenzie, Ian and Wim Van Der Wurff (2012). ‘Relic Syntax in Middle English and Medieval Spanish: Parameter Interaction in Language Change’, Language 88: 846–76. Maling, Joan (1990). ‘Inversion in Embedded Clauses in Modern Icelandic’, in Joan Maling and Annie Zaenen (eds), Modern Icelandic Syntax. San Diego, CA: Academic Press, 71–91. Maling, Joan (2001). ‘Dative: The Heterogeneity of the Mapping among Morphological Case, Grammatical Functions, and Thematic Roles’, Lingua 111: 419–64. Maling, Joan (2002). ‘Það rignir þágufalli á Íslandi: Verbs with Dative Objects in Icelandic’, Íslenskt mál og almenn málfræði 24: 31–105. Manzini, Rita and Leonardo Savoia (2003). ‘The Nature of Complementizers’, Rivista di Grammatica Generativa 28: 87–110. Manzini, Rita and Leonardo Savoia (2011). Grammatical Categories. Cambridge: Cambridge University Press. Marantz, Alec (1993). ‘Implications of Asymmetries in Double Object Constructions’, in Sam A. Mchombo (ed.), Theoretical Aspects of Bantu Grammar. Stanford, CA: CSLI Publications, 113–50.

REFERENCES

263

Marantz, Alec (2013). ‘Verbal Argument Structure: Events and Participants’, Lingua 130: 152–68. Marinis, Theo (2004). The Acquisition of the DP in Modern Greek. Amsterdam/Philadelphia: John Benjamins. Martineau, France (1990a). ‘La construction “accusatif avec infinitif ” avec les verbes causatifs et de perception en moyen français’, Revue Québécoise de Linguistique 19: 77–100. Martineau, France (1990b). La montée du clitique en moyen français. Une étude de la syntaxe des constructions infinitives. Ph.D. dissertation. University of Ottawa. Martins, Ana Maria (1998). ‘The Loss of IP-scrambling in Portuguese: Clause Structure, Word-order Variation and Change’, in David Lightfoot (ed.), Syntactic Effects of Morphological Change. Oxford/New York: Oxford University Press, 232–48. Martins, Ana Maria (2006a). ‘A spects of Infinitival Constructions in the History of Portuguese’, in Randall Gess and Deborah Arteaga (eds), Historical Romance Linguistics: Retrospective and Perspectives. Amsterdam: John Benjamins, 327–55. Martins, Ana Maria (2006b). ‘Emphatic Affirmation and Polarity: Contrasting European Portuguese with Brazilian Portuguese, Spanish, Catalan and Galician’, in Jenny Doetjes and Paz Gonzalez (eds), Romance Languages and Linguistic Theory 2004. Amsterdam/Philadelphia: John Benjamins, 197–223. Martins, Ana Maria (2011). ‘Scrambling and Information Focus in Old and Contemporary Portuguese’, Catalan Journal of Linguistics 10: 1–26. Martins, Ana Maria (2013). ‘Emphatic Polarity in European Portuguese and Beyond’, Lingua 128: 95–123. Martins, Ana Maria (2016). ‘VP and TP Ellipsis: Sentential Polarity and Information Structure’, in Susann Fischer and Christoph Gabriel (eds), Grammatical Interfaces in Romance. Berlin/Boston: De Gruyter, 457–85. Martins, Ana Maria and Adriana Cardoso (eds) (2018). Word Order Change. Oxford: Oxford University Press. Martins, Ana Maria and João Costa (2016). ‘Ordem dos constituintes frásicos. Sujeitos invertidos, objetos antepostos’, in Ana Maria Martins and Ernestina Carrilho (eds), Manual de Linguı´stica Portuguesa. Berlin/Boston: De Gruyter, 371–400. Matos, Gabriela (1992). Construções de Elipse do Predicado em Português: SV Nulo e Despojamento. Ph.D. dissertation. Universidade de Lisboa. Matos, Gabriela (2003). ‘Estruturas de coordenação’, in Maria Helena Mira Mateus et al., Gramática da Lı´ngua Portuguesa. Lisboa: Caminho, 549–592. Mayen, Georg (1889). De particulis QUOD QUIA QUONIAM QUOMODO UT pro accusativo cum infinitivo post verba sentiendi et declarandi positis. Kiel: Fiencke. Miller, D. Gary (1993). Complex Verb Formation. Amsterdam/Philadelphia: John Benjamins. Miller, D. Gary (2002). Nonfinite Structures in Theory and Change. Oxford: Oxford University Press. MÍM (Mo¨rkuð ´ıslensk málheild). Sigrún Helgadóttir (ed.). Fornrit: Árni Magnússon Institute for Icelandic studies. . Montgomery, Michael and Guy Bailey (1991). ‘In Which: A New Form in Written English’, American Speech 66.2: 147–63. Morris, Amanda (2009). Polarity Ellipsis and Negative Stripping. M.A. dissertation. University of California. Muller, Henri-François (1912). Origine et histoire de la préposition ‘à’ dans les locutions du type de ‘faire faire quelque chose à quelqu’un’. Poitiers: Masson.

264

REFERENCES

Munn, Alan (1993). Topics in the Syntax and Semantics of Coordinate Structures. Ph.D. dissertation. University of Maryland. Munn, Alan (1999). ‘First Conjunct Agreement: Against a Clausal Analysis’, Linguistic Inquiry 30: 643–68. Narrog, Heiko and Bernd Heine (2011). ‘Introduction’, in Heiko Narrog and Bernd Heine (eds), The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, 1–16. Noordhof, Harm (1937). La construction relative en italien. La Haye: van Haeringen. Norberg, Dag (1945). ‘Faire faire quelque chose à quelqu’un. Recherches sur l’origine latine de la construction romane’, Uppsala Universitets Årsskrift 12.2: 65–106. Norde, Muriel (2009). Degrammaticalization. Oxford: Oxford University Press. Nunes, Jairo (2004). Linearization of Chains and Sideward Movement. Cambridge, MA: MIT Press. Nunes, Jairo (2008). ‘Inherent Case as a Licensing Condition for A-movement: The Case of Hyper-raising Constructions in Brazilian Portuguese’, Journal of Portuguese Linguistics 7: 83–108. Nygaard, Marius (1867). Eddasprogets Syntax. Vol. II. Bergen: Giertsen. Nygaard, Marius (1905). Norrøn syntax. Kristiania: H. Aschehoug & Co. (W. Nygaard). O’Neill, Teresa (2011). ‘The Syntax of ne … que Exceptives in French’, NYU Working Papers in Linguistics 3: 199–230. Obenauer, Hans-Georg (1976). Études de syntaxe interrogative du français—Quoi, combien et le complémenteur. Tu¨bingen: Niemeyer. Öhl, Peter (2009). ‘Die Entstehung des periphrastischen Perfekts mit haben und sein im Deutschen—eine la¨ngst beantwortete Frage?’ Zeitschrift fu¨r Sprachwissenscshaft 28: 265–306. Olthof, Marieke (2020). ‘Formal Variation in Incorporation: A Typological Study and a Unified Approach’, Linguistics 58: 131–205. ONP = Ordbog over det norrøne prosasprog. Helle Degnbol, Bent Chr. Jacobsen, Eva Rode, Christopher Sanders, James E. Knirk, and Þorbjo¨rg Helgadóttir (eds), Den arnamagnæanske commission, København. . Oubouzar, Erika (1974). ‘Über die Ausbildung der zusammengesetzten Verbformen im deutschen Verbalsystem’, Beitra¨ge zur Geschichte der deutschen Sprache und Literatur 95: 5–96. Palmer, Leonard R. (1977). The Latin Language. London: Faber and Faber Limited. Pană Dindelegan, Gabriela (2011). ‘Din istoria supinului romaˆnesc’, in R. Zafiu, C. Us¸urelu, and H. Bogdan Oprea (eds), Limba romaˆnă—ipostaze ale variației lingvistice. Bucharest: Editura Universității din Bucures¸ti, I: 119–30. Pancheva, Roumyana (2003). ‘The Aspectual Makeup of Perfect Participles and the Interpretations of the Perfect’, in Alexis Alexiadou, Monika Rathert, and Arnim von Stechow (eds.), Perfect Explorations. Berlin/Boston: Mouton de Gruyter, 277–306. Pancheva, Roumyana (2005). ‘The Rise and Fall of Second-Position Clitics’, Natural Language and Linguistic Theory 23: 103–67. Pancheva, Roumyana and Arnim Von Stechow (2004). ‘On the Present Perfect Puzzle’, in Keir Moulton and Matthew Wolf (eds), NELS 34: Proceedings of the 34th Annual Meeting of the North East Linguistic Society. BookSurge Publishing, 469–483. Parry, Mair (2007). ‘The Interaction of Semantics and Syntax in the Spread of Relative che in Early Vernaculars of Italy’, in Delia Bentley and Adam Ledgeway (eds), Sui dialetti italoromanzi. Saggi in onore di Nigel B. Vincent. Norfolk: Bidles, 200–19.

REFERENCES

265

Paul, Hermann (1902). Die Umschreibung des Perfektums im Deutschen mit haben und sein. Mu¨nchen: Verlag der k. Akademie. Pearce, Elizabeth (1990). Parameters in Old French Syntax: Infinitival Complements. Dordrecht: Kluwer. Penner, Zvi and Ju¨rger Weissenborn (1996). ‘Strong Continuity, Parameter Setting and the Trigger Hierarchy: On the Acquisition of the DP in Bernese Swiss German and High German’, in Harald Clahsen (ed.), Generative Perspectives on Language Acquisition. Amsterdam: John Benjamins, 161–200. Pérez-Jiménez, Isabel, and Norberto Moreno-Quibén (2012). ‘On the Syntax of Exceptions: Evidence from Spanish’, Lingua 122: 582–607. Pérez-Leroux, Ana-Teresa (1995). ‘Resumptives in the Acquisition of Relative Clauses’, Language Acquisition 4: 105–38. Pfaff, Alexander Peter (2015). Adjectival and Genitival Modification in Definite Noun Phrases in Icelandic: A Tale of Outsiders and Inside Jobs. Ph.D. dissertation. University of Tromsø. Picallo, M. Carme (2002). ‘Abstract Agreement and Clausal Arguments’, Syntax 5: 116–47. Pierce, Amy (1989). On the Emergence of Syntax: A Crosslinguistic Study. Ph.D. dissertation. MIT. Pietsch, Karl (1915). ‘On the Language of the Spanish Grail Fragments. I’, Modern Philology 13.7: 369–78 [65–74]. Pillinger, O. Stephen (1980). ‘The Accusative and Infinitive in Latin: A Refractory Complement Clause’, Journal of Linguistics 16: 55–83. Pintzuk, Susan (1999). Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland. Plank, Frans (1984). ‘The Modals Story Retold (Review of Lightfoot 1979)’, Studies in Language 8: 305–64. Platzack, Christer (2008). ‘Left Branch Extraction of Nominal Modifiers in Old Scandinavian’, in Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory: The Rosendal Papers. Amsterdam: John Benjamins, 357–74. Poeppel, David and Kenneth Wexler (1993). ‘The Full Competence Hypothesis of Clause Structure in Early German’, Language 69: 1–33. Poletto, Cecilia (2014). Word Order in Old Italian. Oxford: Oxford University Press. Poletto, Cecilia and Emanuela Sanfelici (2014). ‘On the Nature of Complementizers: Insights from Italian Subject Relative Clauses’, Paper presented at the 28th Symposium on Romance Linguistics, Going Romance. University of Lisbon, 4–6 December. Pollock, Jean-Yves (1989). ‘Verb Movement, Universal Grammar, and the Structure of IP’, Linguistic Inquiry 20: 365–424. Postal, Paul (1974). On Raising: One Rule of English and its Theoretical Implications. Cambridge, MA: MIT Press. Pountain, Christopher (1998). ‘Learnèd Syntax and the Romance Languages: The “Accusative and Infinitive” Construction with Declarative Verbs in Castilian’, Transactions of the Philological Society 96: 159–201. Progovac, Ljiljana (1994). Negative and Positive Polarity: A Binding Approach. Cambridge: Cambridge University Press. Pylkka¨nen, Liina (2008). Introducing Arguments. Cambridge, MA: MIT Press. Quinn, Heidi (2009). ‘Downward Reanalysis and the Rise of Stative HAVE got’, in Paola Crisma and Giuseppe Longobardi (eds), Historical Syntax and Linguistic Theory. Oxford: Oxford University Press, 212–30.

266

REFERENCES

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik (1985). A Comprehensive Grammar of the English Language. London/New York: Longman. Radetzky, Paula and Tomoko Smith (2010). ‘An Areal and Cross-linguistic Study of Benefactive and Malefactive Constructions’, in Fernando Zúñiga and Seppo Kittila¨ (eds), Benefactives and Malefactives: Typological Perspectives and Case Studies. Amsterdam: John Benjamins, 97–120. Ramat, Paolo (1992). ‘Thoughts on Degrammaticalization’, Linguistics 30: 549–60. Raposo, Eduardo and Juan Uriagereka (1990). ‘Long-Distance Case Assignment’, Linguistic Inquiry 21: 505–37. Real Puigdollers, Christina (2013). Lexicalization by Phase: The Role of Prepositions in Argument Structure and its Cross-linguistic Variation. Ph.D. Dissertation. Universitat Autònoma de Barcelona. Reichenbach, Hans (1947). Elements of Symbolic Logic. London: Collier-Macmillan Ltd. Richards, Marc and Theresa Biberauer (2005). ‘Explaining Expl’, in Marcel Den Dikken and Christina Tortora (eds), The Function of Function Words and Functional Categories. Amsterdam: John Benjamins, 115–53. Rivero, Maria Luisa (1994). ‘Clause Structure and V-movement in the Languages of the Balkans’, Natural Language and Linguistic Theory 12: 63–120. Rizzi, Luigi (1990). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, Luigi (1994). ‘Early Null Subjects and Root Null Subjects’, in Teun Hoekstra and Bonnie Schwartz (eds), Language Acquisition Studies in Generative Grammar. Amsterdam: John Benjamins, 151–77. Rizzi, Luigi (1997). ‘The Fine Structure of the Left Periphery’, in Liliane Haegeman (ed.), Elements of Grammar. Dordrecht: Kluwer, 281–337. Rizzi, Luigi (2000). Comparative Syntax and Language Acquisition. London: Routledge. Rizzi, Luigi (2001). ‘On the Position “Int(errogative)” in the Left Periphery of the Clause’, in Guglielmo Cinque and Giampaolo Salvi (eds.), Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi. Leiden: Brill, 287–296. Rizzi, Luigi (2004). ‘Locality and Left Periphery’, in Adriana Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures. Vol. 3. New York: Oxford University Press, 223–51. Roberts, Ian (1985). ‘Agreement Parameters and the Development of English Modal Auxiliaries’, Natural Language and Linguistic Theory 3: 21–58. Roberts, Ian (1993). Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Roberts, Ian (1997). ‘Directionality and Word Order Change in the History of English’, in Ans van Kemenade and Nigel Vincent (eds), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 396–426. Roberts, Ian (2007). Diachronic Syntax. Oxford: Oxford University Press. Roberts, Ian (2010). Agreement and Head Movement. Clitics, Incorporation, and Defective Goals. Cambridge, MA: The MIT Press. Roberts, Ian (2016). ‘Some Remarks on Parameter Hierarchies’, in Luis Eguren, Olga Fernandez-Soriano, and Amaya Mendikoetxea (eds), Rethinking Parameters. Oxford: Oxford University Press, 170–99. Roberts, Ian and Anna Roussou (2003). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Robustelli, Cecilia (1992). ‘A lcune osservazioni sulla sintassi del costrutto fare + infinito nell’italiano dei primi secoli’, The Italianist 12: 83–116.

REFERENCES

267

Robustelli, Cecilia (1993). ‘Indagine diacronica sul costrutto latino facio + infinito’, Studi e Saggi Linguistici 33: 125–89. Robustelli, Cecilia (1994). ‘Il costrutto latino fare e infinito nell’italiano dal 1400 al 1800’, Studi e Saggi Linguistici 34: 151–203. Robustelli, Cecilia (2000). Causativi in italiano antico e moderno. Reggio Emilia: Il Fiorino. Ro¨del, Michael (2007). Doppelte Perfektbildungen und die Organisation von Tempus im Deutschen. Tu¨bingen: Stauffenburg Verlag. Roeper, Tom (1999). ‘Finding Fundamental Operations in Language Acquisition: Formal Features as Triggers’, in Bart Hollebrandse (ed.), New Perspectives on Language Acquisition. GSLA: Amherst: University of Massachusetts, 197–214. Ro¨gnvaldsson, Eirı´kur (1995). ‘Old Icelandic: A Non-configurational Language?’ NorthWestern European Language Evolution 26: 3–29. Ro¨gnvaldsson, Eirı´kur (2005). ‘Setningafræðilegar breytingar ´ı ´ıslensku’ [Syntactic Changes in the History of Icelandic]’, in Ho¨skuldur Þráinsson (ed.), Íslensk tunga III. Setningar. Reykjavı´k: Almenna bókafélagið, 602–35. Rohlfs, Gerhard (1966). Grammatica storica della lingua italiana e dei suoi dialetti. Torino: Einaudi. Rothstein, Bjo¨rn (2008). The Perfect Time Span. On the Present Perfect in German, Swedish and English. Amsterdam/Philadelphia: John Benjamins. Rowlett, Paul (2007). ‘Cinque’s Functional Verbs in French’, Language Sciences 29: 755–86. Saab, Andrès (2008). Hacia una teorı´a de la identidad parcial en la elipsis. Ph.D dissertation. Universidad Nacional de Comahue. Saab, Andrès (2010). ‘Spanish TP-ellipsis and the Theory of Island Repair’, Probus 22: 73– 116. Sanfelici, Emanuela, Irene Caloi, and Cecilia Poletto (2014). ‘Subject Object Asymmetries in Relative Clauses: An Investigation into Three New Empirical Domains’, Quaderni di lavoro ASIt. Padova: Atti della XIX Giornata di dialettologia, 127–160. Sapp, Christopher D. (2009). ‘Syncope as the Cause of Pra¨teritumschwund: New Data from an Early New High German Corpus’, Journal of Germanic Linguistics 21: 419–50. Scholten, Jolien (2018). The Ins and Outs of External Possession: A Micro-comparative Perspective. Utrecht: LOT. Sevcenco, Anca (2015). ‘Restrictive and Appositive Relatives in Old Romanian’, in Virginia Hill (ed.), Formal Approaches to DP in Old Romanian. Leiden: Brill, 329–364. Sevdali, Christina (2013). ‘Ancient Greek Infinitives and Phases’, Syntax 16: 324–61. Shannon, Thomas F. (1995). ‘Toward a Cognitive Explanation of Perfect Auxiliary Variation: Some Modal and Aspectual Effects in the History of Germanic’, American Journal of Germanic Linguistics & Literatures 7.2: 129–63. Sheehan, Michelle (2011). ‘Extraposition and Antisymmetry’, in Jeroen Van Craenenbroeck (ed.), Linguistic Variation Yearbook 2010. Amsterdam/Philadelphia: John Benjamins, 201–51. Sheehan, Michelle (2016). ‘Complex Predicates’, in Adam Ledgeway and Martin Maiden (eds), The Oxford Guide to the Romance Languages. Oxford: Oxford University Press, 981–93. Siddiqi, Daniel (2006). Minimize Exponence: Economy Effects on a Model of the Morphosyntactic Component of the Grammar. Ph.D. dissertation. University of Arizona. Sigurðsson, Halldór (2000). ‘The Locus of Case and Agreement’, Working Papers in Scandinavian Syntax 65: 65–108. Sigurðsson, Halldór (2006). ‘The Icelandic Noun Phrase: Central Traits’, Arkiv fo¨r nordisk filologi 121: 193–236.

268

REFERENCES

Sigurðsson, Halldór (2012). ‘Minimalist C/case’, Linguistic Inquiry 43: 191–227. Sitaridou, Ioanna (2011). ‘Word Order and Information Structure in Old Spanish’, Catalan Journal of Linguistics 10: 159–84. Sitaridou, Ioanna (2012). ‘A Comparative Study of Word Order in Old Romance’, Folia Linguistica 46.2: 553–604. Skard, Vemund (1951). Dativstudien. Dativus Sympatheticus und Dativus Comparationis in der norro¨nen Sprache. Oslo: Jacob Dybwad. Soare, Elena (2002). Participes, nominalisation et catégories mixtes: Le supin roumain. Ph.D. dissertation. Université Paris 7. Soares Da Silva, Augusto (2012). ‘Stages of Grammaticalization of Causative Verbs and Constructions in Portuguese, Spanish, French and Italian’, Folia Linguistica 46: 513–52. Sommerer, Lotte (2018). Article Emergence in Old English: A Constructionalist Perspective. Berlin: Mouton De Gruyter. Sportiche, Dominique (2011). ‘French Relative Qui’, Linguistic Inquiry 42.1: 83–124. Starke, Michal (2009). ‘Nanosyntax: A Short Primer to a New Approach to Language’, in Peter Svenonius, Gillian Ramchand, Tarald Taraldsen, and Michal Starke (eds), Special Issue on Nanosyntax. Tromsø: University of Tromsø, 1–6. Stepanov, Arthur (2012). ‘Voiding Island Effects via Head Movement’, Linguistic Inquiry 43: 680–93. Stowell, Tim (1981). Origins of Phrase Structure. Ph.D. dissertation. MIT. Stowell, Tim (1982). ‘The Tense of Infinitives’, Linguistic Inquiry 13: 561–70. Stowell, Tim (1996). ‘The Phrase Structure of Tense’, in Johan Rooryck and Laurie Zaring (eds), Phrase Structure and the Lexicon. Dordrecht: Kluwer Academic Publishers, 277–91. Stowell, Tim (2008). ‘The English Konjunktiv II’, in Jacqueline Guéron and Jacqueline Lecarme (eds), Time and Modality. Dordrecht: Springer, 251–72. Stroh-Wollin, Ulla (2009). ‘On the Development of Definiteness Markers in Scandinavian’, Working Papers in Scandinavian Syntax 83.1: 1–25. Stroh-Wollin, Ulla (2016). ‘The Emergence of Definiteness Marking in Scandinavian: New Answers to Old Questions’, Arkiv fo¨r nordisk filologi 131: 129–69. Surányi, Balázs (2009). ‘Verbal Particles Inside and Outside vP’, Acta Linguistica Hungarica 56: 201–49. Surányi, Balázs (2011). ‘An Interface Account of Identificational Focus Movement’, in Tibor Laczkó and Catherine O. Ringen (eds), Approaches to Hungarian 12: Papers from the 2009 Debrecen Conference. Amsterdam: John Benjamins, 163–208. Svenonius, Peter (2003). ‘Limits on P: Filling in Holes vs. Falling in Holes’, Nordlyd, Tromsø Working Papers on Language and Linguistics: Proceedings of the 19th Scandinavian Conference of Linguistics 31.2: 431–45. Svenonius, Peter (2010). ‘Spatial p in English’, in Guglielmo Cinque and Luigi Rizzi (eds), Mapping Spatial PPs: The Cartography of Syntactic Structures. Vol. 6. Oxford/New York: Oxford University Press, 127–60. Svenonius, Peter (2016). ‘Spans and Words’, in Daniel Siddiqi and Heidi Harley (eds), Morphological Metatheory. Amsterdam: John Benjamins, 201–22. Szendrői, Kriszta (2001). Focus and the Phonology–Syntax Interface. Ph.D. dissertation. University College London. Taraldsen, Tarald (1985). ‘Ka du trur?’ in Bull Tove and Anton Fjeldstad (eds), Heidersskrift til Kåre Elstad. Tromsø: University of Tromsø, 37–47.

REFERENCES

269

Taylor, Ann, Anthony Warner, Susan Pintzuk, and Frank Beths (2003). The York-TorontoHelsinki Parsed Corpus of Old English. University of York, York. Taylor, Ann, Arja Nurmi, Anthony Warner, Susan Pintzuk, and Terttu Nevalainen (2006). The Parsed Corpus of Early English Correspondence. York/Helsinki: University of York and University of Helsinki. Distributed through the Oxford Text Archive. Thielmann, Philipp (1886). ‘Facere mit dem Infinitiv’, Archiv fu¨r lateinische Lexikographie und Grammatik 3: 177–206. Thornton, Rosalind and Kelly Rombough (2014). ‘The Syntax-PF Interface in Children’s Negative Sentences’, Language Acquisition 22: 1–26. Thornton, Rosalind and Graciela Tesan (2007). ‘Categorical Acquisition: Parameter-Setting in Universal Grammar’, Biolinguistics 1: 49–98. Torrego, Esther (2010). ‘Variability in the Case-Patterns of Causative Formation in Romance and its Implications’, Linguistic Inquiry 41: 445–70. Traugott, Elizabeth Closs and Richard Dasher (2002). Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Bernd Heine (1991). Approaches to Grammaticalization. Amsterdam: John Benjamins. Traugott, Elizabeth Closs and Ekkehard Ko¨nig (1991). ‘The Semantics-Pragmatics of Grammaticalization Revisited’, in Elizabeth Closs Traugott and Bernd Heine (eds), Approaches to Grammaticalization. Vol. 1. Amsterdam: John Benjamins, 189–218. Travis, Lisa Demena (1984). Parameters and Effects of Word Order Variation. Ph.D. dissertation. MIT. Tungseth, Mai (2007). ‘Benefactive across Scandinavian’, Working Papers in Scandinavian Syntax 80: 187–228. Ura, Hiroyuki (2007). ‘Long-Distance Case-assignment in Japanese and its Dialectal Variation’, Gengo Kenkyu 131: 1–43. Utzeri, Irene (2007). ‘The Production and Acquisition of Subject and Object Relative Clauses in Italian’, Nanzan Linguistics Special Issue 3: 283–314. Van de Velde, Freek (2010). ‘The Emergence of the Determiner in the Dutch NP’, Linguistics 48.2: 263–99. Van de Velde, Freek and Béatrice Lamiroy (2017). ‘External Possessors in West Germanic and Romance: Differential Speed in the Drift toward NP Configurationality’, in Daniel Van Olmen, Hubert Cuyckens, and Lobke Ghesquière (eds), Aspects of Grammaticalization: (Inter)Subjectification and Directionality. Berlin/Boston: Mouton De Gruyter, 353–99. Van Valin, Robert D. and Randy J. LaPolla (1997). Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press. Vicente, Luis (2006). ‘Short Negative Replies in Spanish’, in Bettelou Los and Jeroen Van De Weijer (eds.), Linguistics in the Netherlands. Vol. 23. Amsterdam/Philadelphia: John Benjamins, 199–210. Vicente, Luis (2010). ‘On the Syntax of Adversative Coordination’, Natural Language and Linguistic Theory 28: 381–415. Viðarsson, Heimir F. (2009). ‘Tilbrigði ´ı fallmo¨rkun aukafallsfrumlaga: Þágufallshneigð ´ı fornı´slensku?’, Íslenskt mál og almenn málfræði 31: 15–66. Viðarsson, Heimir F. (2017a). ‘Grimm’s “Floating” Datives: Applicatives and NP/DP Configurationality in Icelandic from a Diachronic Perspective’, Working Papers in Scandinavian Syntax 99: 30–53.

270

REFERENCES

Viðarsson, Heimir F. (2017b). ‘Tilbrigði til forna’, in Ho¨skuldur Þráinsson, Ásgrı´mur Angantýsson, and Einar Freyr Sigurðsson (eds), Tilbrigði ´ı ´ıslenskri setningagerð, III. Sérathuganir. Reykjavı´k, Málvı´sindastofnun Háskóla Íslands, 139–79. Vincent, Nigel (2016). ‘Causatives in Latin and Romance’, in James Adams and Nigel Vincent (eds), Early and Late Latin: Continuity or Change? Cambridge: Cambridge University Press, 294–312. Von Fintel, Kai (1991). ‘Exceptive Conditionals: The Meaning of unless’, Proceedings of the Annual Meeting of the North Eastern Linguistics Society 22: 135–48. Von Fintel, Kai (1993). ‘Exceptive Constructions’, Natural Language Semantics 1: 123–48. Wallenberg, Joel C., Anton Karl Ingason, Einar Freyr Sigurðsson, and Eirı´kur Ro¨gnvaldsson (2011). Icelandic Parsed Historical Corpus. . Warner, Anthony (1983). ‘Review of Lightfoot (1979)’, Journal of Linguistics 19: 187–209. Warner, Anthony (1993). English Auxiliaries: Structure and History. Cambridge: Cambridge University Press. Waters, Cathleen (2011). Social and Linguistic Correlates of Adverb Variability in English: A Cross-varietal Perspective. Ph.D. dissertation. University of Toronto. Waters, Cathleen (2013). ‘Transatlantic Variation in English Adverb Placement’, Language Variation and Change 25: 179–200. Weiß, Helmut (2013). ‘Satztyp und Dialekt’, in Jo¨rg Meibauer et al. (eds), Satztypen des Deutschen. Berlin/Boston: De Gruyter, 764–85. Westergaard, Marit and Øystein Alexander Vangsnes (2005). ‘Wh-questions, V2 and the Left Periphery of Three Norwegian Dialect Types’, Journal of Comparative Germanic Syntax 8: 119–60. Wexler, Kenneth (1994). ‘Optional Infinitives, Head Movement, and the Economy of Derivation in Child Grammar’, in David Lightfoot and Norbert Hornstein (eds), Verb Movement. Cambridge: Cambridge University Press, 305–350. Wiklund, Anna-Lena (2007). The Syntax of Tenselessness: Tense/Mood/Aspect-Agreeing Infinitivals. Berlin: Mouton de Gruyter. Wood, Jim (2013). ‘The Unintentional Causer in Icelandic’, Proceedings of NELS 41: 193– 204. Wood, Jim and Halldór Ármann Sigurðsson (2014). ‘Let Causatives and (A)symmetric DAT-NOM Constructions’, Syntax 17.3: 269–98. Wurff, Wim van der (1997). ‘Deriving Object-Verb Order in Late Middle English’, Journal of Linguistics 33: 485–509. Zagona, Karen (1995). ‘Temporal Argument Structure: Configurational Elements of Construal’, in P. M. Bertinetto et al. (eds), Temporal Reference, Aspect and Actionality, Vol. 1: Semantic and Syntactic Perspectives. Torino: Rosenberg & Sellier, 397–410. Zagona, Karen (2007). ‘Some Effects of Aspect on Tense Construal’, Lingua 117: 464–502. Zanuttini, Raffaella (2008). ‘Encoding the Addressee in the Syntax: Evidence from English Imperative Subjects’, Natural Language and Linguistic Theory 26.1: 185–218. Zimmermann, Malte (2013). ‘Ob-VL-Interrogativsatz’, in Jo¨rg Meibauer et al. (eds), Satztypen des Deutschen. Berlin/Boston: De Gruyter, 84–104. Zúñiga, Fernando and Seppo Kittila¨ (2010). ‘Introduction: Benefaction and Malefaction from a Cross-Linguistic Perspective’, in Fernando Zúñiga and Seppo Kittila¨ (eds), Benefactives and Malefactives: Typological Perspectives and Case Studies. Amsterdam: John Benjamins, 1–28.

Index of subjects accusative 19, 22, 28, 30–31, 32, 33, 37, 43, 45, 49–50, 53, 56, 99, 205, 220, 222; see also case – accusative case adjective 31, 102–103, 191, 205, 233, 237, 239–240 inflected adjective 239–240 interrogative adjective 103 adjunct 61, 63, 67, 158, 198–199, 215, 217, 233, 235, 240 Adjunct Condition 186, 190, 196, 198–199, 208 adjoin(ed)/adjunction 73, 118, 186, 192 adpositional phrase 6–8, 18, adverb/adverbial 20, 61, 62, 63, 77, 109, 136–138, 142, 144, 145–155, 157–158, 160–163, 165–166, 169, 171–173, 175–177, 182, 184, 197 adverb fronting 148 adverb placement 136–138, 144, 147, 148–155, 157, 158 negative adverb 109 positional past time adverbial 161–162, 165–166, 169, 173, 175–176 temporal adverb/adverbial 62, 161, 165, 171, 182 affected(ness) 210–214, 217–219, 223–224, 229, 243–244 agentive 22, 24, 53, 104, 218 agentive dative 46, 53 non-agentive 104 Agree relation 28–29, 33, 35 agree(ment) 25, 27, 28–29, 33, 35, 63, 67, 91, 94, 95, 97, 100, 101–103, 106–107, 110, 138, 139, 179–180, 217 agreement features 94–95, 103, 106, 179 subject-verb agreement 41, 67 analogy/analogical 9, 52, 60, 68, 100, 133, 134 anaphora/anaphoric 62–64, 81 animacy 45, 90–91, 93, 94, 95, 97, 98, 101, 103–104

animacy feature 90, 102, 104 animate 94, 95, 96, 97–98, 100, 104, 213 inanimate 44, 46, 96, 98 antecedent 72, 80, 94, 96, 97, 98, 100, 104 antecedent clause 73, 75 anterior 160–161, 164, 166–169, 178–180, 181, 184–185 applicative 210–212, 214–216, 217, 219, 221–225, 227, 231, 234, 243–245 Appl(icative) head 213–214 applicative phrase/ApplP 215, 225, 228 dative applicative 17, 210–211, 219, 222, 223–224, 234, 244–245 High Applicative 212, 214–216, 225–226, 228–229, 243–245 Low Applicative 212, 214–216, 225–226, 243 argument 22, 24, 26, 28, 31, 32, 43, 46, 50, 170, 174, 199, 210, 214–215, 218, 220, 225, 226–227, 228, 240 argument fronting 80 argument structure 210–211, 214, 215, 221 external argument 24, 26, 28, 32, 50, 174, 215 internal argument 22, 24, 26, 28, 31, 43, 46, 220, 225 article 102, 108, 212, 229, 232, 235–237, 239, 240–241, 243–244 definite article 8, 17, 108, 212, 229, 232, 233, 235–237, 239, 240, 244 free-standing article 239–240 suffixed article 239–240 auxiliary 5, 25, 47, 136–137, 140, 141, 149, 159–160, 162–170, 173–185, 188, 190, 196 auxiliary selection 5, 162, 169, 175, 176–177, 184 modal auxiliary 4, 140 perfect auxiliary 5, 16, 163, 177, 183

272

INDE X OF SUBJECTS

auxiliary (Continued) temporal auxiliary 159, 162–163, 166, 176, 180, 184 benefactive 210–211, 213–214, 216, 217, 220–222, 224, 227, 228, 243, 244 Borer–Chomsky Conjecture 5, 7 cartography, cartographic 6, 14, 59, 60, 62, 87, 92, 105, 109, 110 case 23, 28–34, 36–38, 42, 49–50, 56, 61, 90, 95, 97, 98, 99, 102, 104–105, 107, 108, 192, 194, 202, 211, 213, 216, 217, 219, 220, 221, 229, 243–244 accusative case 19, 28, 30–34, 37, 42, 43, 45, 49–50, 56, 61, 99, 192, 205, 220, 221 case assignment 29, 32, 34, 36, 37–38, 50, 202 case-marking 37, 50, 53, 90, 102, 107, 108, 192, 202, 213, 221, 243 caseless 28, 29, 36 dative case 30, 46, 217, 219, 220; see also dative Exceptional Case Marking/ECM 23, 29, 30, 31–32, 33, 34, 37, 41–43, 45–47, 49–55 genitive case 56, 192; see also genitive inherent case 29, 36 loss of morphological case 221, 244 morphological case 17, 192, 211, 216, 219, 221, 243, 244 nominative case 28, 37, 94, 95, 97, 99, 104 null case assigner 32–33, 37, 50 structural case 24, 26, 29, 32, 36, 37, 192 change 1–18, 19, 23, 38–39, 41, 55, 59, 68, 71, 73–74, 88, 94, 103–104, 105–106, 110, 112–113, 121, 123, 124, 132–134, 137, 140–141, 142–143, 146, 148, 157–158, 163, 165–166, 168, 177, 185, 200, 201, 202, 210, 211–212, 216, 229–230, 236, 240 categorial change 3, 4, 134, 138, 157 cyclical change 10 lexical change 3–4 morphosyntactic change 9, 217 parametric change 12, 55 semantic change 3–4

structural change 1, 2, 11, 15, 186 syntactic change 2, 5–9, 11, 18, 41, 54, 71, 217 clause 4, 5, 7, 13–15, 19–24, 26, 28–30, 38, 40–42, 46–47, 57, 59–64, 66–69, 75, 80, 92, 111–118, 122–123, 133, 145, 151, 153, 197, 229 antecedent clause 73, 75 biclausal analysis/structure 46, 49, 54, 151–154, 158 clause final 71, 83–84, 86, 87, 89, 122–124, 129–129 clause type 111, 125, 134 complement clause 19, 20, 41–42, 64, 66, 68–69, 91, 92, 95, 97–98 conditional clause 13, 70, 73, 74–75, 77, 80 finite clause 12, 19–23, 26, 28, 41–43, 51, 67, 115 infinitival clause 12–13, 19, 21–23, 27, 29–31, 33, 38, 41–42, 59, 69, 107, 196–197 interrogative clause 111, 113, 134 main clause 15, 75, 113–114, 116, 122–123, 147 negative clause 138, 144, 148 passive clause 26 relative clause 57, 80, 84, 90, 92–95, 99–100, 103–105, 107, 110, 198, 202, 207 subjunctive clause 13, 64, 69 subordinate clause 13, 83, 100, 112, 146–147, 154 supine clause 13, 57–58, 60–61, 63, 65–69 clitic 53, 58, 67, 104, 106, 127, 128 clitic climbing 53 clitic left dislocation 88 clitic pronoun(s) 66, 67, 68 proclitic 39, 142, 144 complement 8, 12, 16–17, 19–20, 24, 30–31, 36, 40, 42, 43, 45, 51, 55, 56–60, 63–64, 65–69, 82, 92, 94, 97, 138–139, 153, 177, 186, 187–189, 191, 198, 201–203, 206–208, 215, 228, 233 complement PP/PP complement 16–17, 186, 189, 198, 202, 208 null complement 202

INDE X OF SUBJECTS supine complement 56–57, 59–60, 66, 68–69 complement clause/complementation 4, 12, 13, 19–20, 23, 38, 41–42, 58, 64, 66, 68–69, 92, 95, 97, 98, 139, infinitival complement clause 12, 31, 42, 51, 55, 153 see also embedded clause complementizer 1, 4, 7, 10, 12, 13, 14–15, 20, 21–22, 57–59, 63–68, 90–107, 110, 112, 114, 118–121, 123, 125, 134 C/C head 14, 15, 59, 90, 113, 114, 119, 122–124, 126–127, 129 CP 5, 7–8, 15, 27, 29, 30, 32–34, 36, 38, 49–50, 57, 59–664, 66, 69, 101, 111, 114–127, 133, 135 CP-field 4, 7–8, 14, 60, 64, 80, 82, 87 CP-periphery 15, 111, 115, 120, 123–124, 126, 133 finite complementizer 15 relative complementizer 90–92, 94, 103 split CP 18, 63–64, 116 conditional 70, 71, 73–77, 80, 82, 83–84, 86–89 conditional clause 75, 77, 80 conditional conjunction 77, 88 conditional subordination 70, 73 configurational(ity) 10, 17, 210, 212, 216, 229–230, 232, 234–236, 243–244 conjunction 13, 20, 21, 71–73, 77, 82, 84, 88 adverbial conjunction 20 conditional conjunction 13, 77, 88 coordinating conjunction 71–73, 88 exceptive conjunction 71, 82, 84, 87 subordinate conjunction 98 constituent 13, 17, 26, 65, 71, 74–75, 76, 80, 83–84, 87, 111, 117 constituent fronting/fronted constituent 13, 62, 65, 83 constituent question 15, 111, 113–114, 117, 123, 125, 133, 134 focus(ed) constituent/material 13, 64, 86–87, 112–120 heavy constituent 84 control 43–46, 51, 59, 63, 66, 69, 197 control predicate 43 object control 43–46, 51 obligatory control 59, 63, 66, 69

273

coordination/coordinating 70–73, 76, 81, 88 coordinate construction/structure 72– 73, 77, 80 coordinating conjunction 71–73, 88 copula 17, 79, 82, 174, 176, 222 counterfactual 162, 164, 165 cross-linguistic 16, 25, 56, 115, 124, 169, 172, 176, 201, 210, 213, 216, 225, 228 dative 8, 22, 30, 46, 53, 192, 202, 210–231, 234, 243–245 affected dative 211, 218–219, 229 agentive dative 53 core vs. non-core datives 226 dative applicative 210, 211, 219, 222, 223–224, 234, 244–245 dative benefactive 210, 211, 222, 224, 244 dative case 30, 46, 217, 219, 220 ‘floating’ datives 212, 229, 231 dative possessor 210, 216, 218, 225, 229–231 definite(ness) 75, 108, 217, 234–240, 244 definite article 8, 17, 108, 212, 229, 232–233, 235–237, 239–240, 244 definite determiner 236, 238–240, 244 demonstrative 232–241 demonstrative pronoun 235, 240 determiner 8, 17, 75, 102, 103, 108, 212, 216, 229, 231, 232, 234, 236–240, 244 D/D head 212, 217, 234–236 DP 5, 7, 8, 12, 17, 23, 25–30, 36, 43, 49, 56, 75, 108, 174, 195 DP movement 25–26, 41 hypodetermining 212, 240 universal determiner 75 discourse 107, 147–148 discourse marker 3 discourse particle 147–148 ditransitive construction/structure 211, 217, 220–221, 224 do-support 4, 144, 146 doubling 16, 111, 125, 133, 159–164, 166, 168–169, 178, 180, 183, 184–185 perfect doubling 16,159–164,166– 170, 178–181, 184–185 see also reduplication Doubly Filled COMP 125

274

INDE X OF SUBJECTS

economy/economical 9, 10, 68, 92, 106, 125 ellipsis 71–72, 73, 74, 77, 79–82, 84, 86 predicate ellipsis 77, 81, 84 TP ellipsis 80–82, 86 verb ellipsis 86 embedded clause 4, 13, 30, 43, 46, 60, 63, 123 embedded interrogative 5, 15, 114, 121, 123–126, 134 embedded subject 23, 29–34, 37–38, 46, 49–50 embedded question 14–15, 111, 114–121, 123, 133 see also complement clause/complementation emphatic 74, 136 emphatic focus 74 EPP 12, 23–29, 38–42, 50, 55 EPP-feature 12, 25 EPP parameter 12, 38, 50 EPP-requirement 12, 23, 25–26, 38–39, 50, 55 event 61, 63, 160, 161, 169–170, 175, 181–183, 213, 214–216, 243, 245 event argument 225 event time 170, 172, 175 eventive 212, 216, 227, 228 Exceptional Case Marking/ECM 23, 29, 30, 31–32, 33, 34, 37, 41–43, 45–47, 49–55 exceptive 7, 13–14, 70–77, 79–80, 82–89 connected exceptive 73, 74–76 exception phrase/exceptive phrase 72–73, 82, 86–87 exceptive clause 7, 75 exceptive construction 13, 70–73, 77, 88–89 exceptive meaning 13, 73 free exceptive 13, 73–77 experiencer 213, 216, 218, 220, 229, 243, 244 extended projection/phrase 1, 6–8, 24, 28, 35–36, 40, 195, 200–201, 212, 229, 244 feature(s) 2, 5–6, 60, 87, 92, 98, 101, 105–110, 111, 125, 134, 220 animacy feature 90, 102, 104 agreement features 94–95, 103, 106, 179

categorial features 36, 158 checking of feature(s) 60, 63, 66–69, 117 EPP-feature 12, 25 (…) feature specification 2–3, 5–6, 11, 14, 17, 91, 94, 103–104, 108, 110, 111 finite feature 5, 13, 59–61, 63–66, 122, 179 functional feature(s) 1, 6, 18, 25–27, 38–39, 41, 59–60, 63, 66–68, 101, 103, 111, 113, 116 inflectional features 176, 201 modal feature 7, 13, 59–61, 65–66, 68 morphological features 6, 106–107, 110 nominal/pronominal features 91, 94, 97, 101, 103–104, 108, 110 one feature - one projection principle 6, 105, 109, 110 phonological feature 72 semantic feature(s) 1, 4–6, 14, 76, 107, 109–110, tense feature 179, 204, 208 feminine 98–100, 102 Final-over-Final Condition 9 finite(ness) 16, 21, 41, 59–62, 105, 133, 141, 179–180 finite auxiliary/modal 15, 136–137, 150, 152, 159–160, 166, 179–181, 189 finite clause 12, 19–23, 26, 28, 41–43, 51, 67, 115 finite complementizer 15 finite feature 5, 13, 59–61, 63–66, 122, 179 finite verb 15, 25, 39, 41, 62–63, 115, 122, 124, 127, 142–147, 151–153, 155–156, 179–182, 189, 192–193, 203–206, 208 non-finite(ness) 5, 26, 28, 37–38, 41–42, 57–58, 60, 62, 64–65, 67–68, 133, 138–141, 149, 153–155, 163, 170–171, 173, 175, 185, 196 non-finite verb 28, 46, 62–63, 151, 154, 188–189, 192, 195, 206 focus 7, 15, 62, 64, 83–84, 105, 111–113, 115–116, 134 contrastive focus 74, 84, 87–88 emphatic focus 74 focus(ed) constituent/material 13, 64, 86–87, 112–120 focus construction 13, 115

INDE X OF SUBJECTS focus fronting 13, 87–89, 115 Focus head 112 focus marker 76, 134 focus movement 113–115 focus position 14, 114, 134–135 FocusP 60, 87 information focus 13–14, 71, 74, 80, 82–84, 86–89, 115 preverbal focus 15, 115, 116, 134 verum focus 118 fronting 75 adverb fronting 148 argument fronting 80 constituent fronting/fronted constituent 13, 62, 65, 83 focus fronting 13, 87–89, 115 fronted particle 188 fronted preposition/PP 16, 189, 200 object fronting/fronted object 77, 83, 88 VP fronting 138, 188 Stylistic Fronting 16, 186–190, 194, 196–197, 199–200, 208 functional 3–4, 7, 11–12, 60, 74, 106, 137, 185, 201–202 functional categories 2, 5–6, 10, 18, 80, 201–202 functional domain/layer/sequence/ structure 1, 3–8, 11–12, 15, 17–18, 29, 47, 108, 111–113, 120, 124–125, 126, 133–135, 151, 154–155, 216 functional element/item/material 1, 4, 10–11, 17, 137, 161, 177 functional feature(s) 1, 6, 18, 25–27, 38–39, 41, 59–60, 63, 66–68, 101, 103, 111, 113, 116 functional head 4–7, 9–10, 17, 28–29, 32–33, 37, 105–106, 112, 113, 115, 118–119, 122–124, 126–127, 129, 131, 133–134, 195, 200–202 functional hierarchies 1, 6 functional linguistics 2, 105 functional phrase functional projection 4–8, 12, 14–15, 18, 23–25, 41, 80, 86, 109, 112–115, 117, 122–123, 201, 215, 217, 225, 233, 240 functional verb 46

275

gender 27, 94, 95, 98, 99–101, 103, 104, 108 genitive 56, 61, 192, 202, 230, 241–242 double genitive 241–242 grammaticalization 1–10, 14–18, 49, 67–68, 90–92, 94, 105, 108, 110–111, 113, 123–124, 133–134, 159, 163, 176–178, 180, 183–185, 212, 216–217, 229, 231–232, 235–237, 239, 244 head 1, 5, 14, 24–26, 29, 35–37, 60, 66, 71, 80–82, 86, 90–95, 97, 100–101, 104–110, 113, 115, 142, 150, 170, 188, 190, 195, 201, 207, 217, 233, 236, 240 Appl(icative) head 213–214 C head 14–15, 59, 90, 91, 113–114, 119, 122–127, 129, 133 complementizer head 14 D head 195, 212, 217, 235–236 Fin head 5, 64, 68, 105 Focus head 112 functional head 4–7, 9–10, 17, 28–29, 32–33, 37, 105–106, 112, 113, 115, 118–119, 122–124, 126–127, 129, 131, 133–134, 195, 200–202 head-final 10, 15–16, 28, 122–124, 126–127, 129–133, 144, 147–148, 154, 202 head-initial 10, 15, 17, 113, 122–123, 127, 129–133, 135, 147, 154, 202, 206 head movement 12, 16, 39, 50, 66, 119, 200 Head Movement Constraint/HMC 17, 28, 39, 186, 190, 195–196, 200, 208 Head Preference Principle 10, 14, 17, 92, 94, 106–107, 109–110, 212, 217, 236 inflectional head 143, 148, 151, 155, 201 Int head 109, 112 lexical head 4, 9, 36, 91, 105, 200 multiple T head complex head 28, 39, 105, 197, 200 null head 71 p/P head 8, 16–17, 188, 199, 201–202, 208 head (Continued) syntactic head 5–6 T head 16, 39, 50, 148, 155 v/V head 24, 134, 202

276

INDE X OF SUBJECTS

verbal head 26, 38, 41, 66 Voice head 24, 28, 32–33, 50, 183, 215 incorporation 16, 186, 190–191 Neg-incorporation 40 particle incorporation 193, 197 preposition incorporation/Pincorporation 16–17, 177, 186–187, 189–208 verb incorporation 190, 196, 201 indefinite 108, 111 indefinite polarity items 75–76 infinitive 19, 22, 26, 33–34, 37, 42–55, 57–60, 62, 64–69, 107, 138–139, 149, 153, 161, 197 infinitival clause 12–13, 19, 21–23, 27, 29–31, 33, 38, 41–42, 59, 69, 107, 196–197 inflected infinitive 55 passive infinitive 44–47, 51, 53–54 inflect(ed), inflection(al) 5, 60, 63, 90, 102–103, 107, 140, 176, 201 inflected adjective 239–240 inflected infinitive 55 inflected noun 201 inflected verb 201 inflectional domain/layer/field 1, 7, 25, 60, 64, 143, 152, 156–157 inflectional features 176, 201 inflectional head 143, 148, 151, 155, 201 IP 72, 83 uninflected 95, 101, 103, 108, 110 information structure 23, 38, 71, 88, 111, 135 interrogative 91, 102–104, 106, 109, 111–117, 122–126, 133–134, 138 embedded interrogative 5, 15, 112–113, 120–121, 124, 126, 134 interrogative clause 111, 113, 134 interrogative marker 14–15, 111–114, 118, 121–123, 127–128, 133–134 interrogative pronoun 97–100, 106 IntP 109, 112 language acquisition 2, 5, 7, 9, 108, 217 last resort 37, 128 least effort 10

left branch 12–13, 33, 35, 41, Left Branch Extraction 212, 233 left periphery/peripheral 7, 14, 24, 30, 32–34, 71–72, 74–75, 80, 84, 87, 111–113, 115, 124, 126, 133, 135 lexical 8, 10, 13, 17, 50, 58, 69, 160, 163, 182, 183, 234 lexical change 3–5, 140 lexical category 6, 17, 51, 137, 201–202 lexical entity/form/item/unit 2–4, 9, 13, 16, 23, 32, 42, 52, 63, 70, 101, 105–106, 177, 185 lexical core/domain/layer 1, 3–4, 6 lexical head 4, 9, 36, 91, 105, 200 lexical meaning 3–4, 75, 105, 167, lexical semantic(s) 3–4 lexical verb 4, 15, 22, 46, 47, 50, 105, 115, 117–119, 139, 141, 153, 159, 162, 164, 180, 182, 245 (semi-)lexical 8 lexicalization 3–5, 8, 17, 18, 25, 33, 41, 124, license/licensing 8, 17, 23, 28, 29, 58, 71, 73, 75, 79–81, 86, 89, 119, 138, 201, 210, 212, 219, 221–223, 243–245 Linear Correspondence Principle 119 local(ity) 23, 28–29, 39, 67, local domain 7 locality conditions/constrain 29, 35–36, 66, 189 non-local 29 locative locative verb 198 locative preposition/PP 174, 198 masculine 56, 98–100, 102 Merge/(re)merge 10, 24, 28, 34, 37, 59, 65, 67–69, 91, 152, 154, 155, 157, 210, 215, microdiachronic syntax 18 Minimal Link Condition 117 Minimalist Program 5, 10 minimalist syntax/theory 7, 60, 67, 125 Minimize exponence (principle of ) 179 Mirror principle 119 modal(s) 4, 15–16, 59, 67–68, 81, 106, 136–158 modal verb 4 modal auxiliary 4, 140 modal feature 7, 13, 59–61, 65–66, 68

INDE X OF SUBJECTS modality 5, 59–60, 68, 98 pre-modal 139, 141, 144, 154 modifier 61, 115, 233, 236–238, clausal modifier 84 nominal modifier 212, 234 phrasal modifier 212, 236 mood 66, 180 subjunctive mood 21 movement 7, 25, 27–28, 30, 35, 39–41, 79–80, 82, 105, 113–115, 146, 149, 151, 154, 188, 190, 194, 196–197, 200–203, 206, 208, A-movement 12, 23, 28 DP movement 25–26, 41 head movement 12, 16, 39, 50, 66, 119, 200 Head Movement Constraint/HMC 17, 28, 39, 186, 190, 195–196, 200, 208 loss of movement 9, 134 operator movement 30, 114, 117 prepositional movement 16, 186–190, 193–195, 200, 206 subject-movement 12, 38, 51 verb movement 12, 13, 15–16, 23, 25, 39, 62, 66, 109, 115, 122–123, 137, 142–148, 152, 154, 156–158, 190, 192, 201, 204 V-to-C movement 33, 60, 62–63, 66–67, 190 V-to-F movement 12, 25, 38, 42, 118–119, 122–123 V-to-Fin2 movement 13, 65–68 V-to-I movement 4 V-to-T movement 65–68, 81, 83, 147, 148, 190, 192 VoiceP movement 26–29, 40 VP-movement 12, 23, 25–26, 188 wh-movement 30, 51 negation/negative 4, 25, 28, 39, 47–48, 53, 58, 64–68, 70, 72–75, 77, 79–80, 108–109, 134, 136, 141, 142–144, 148 Neg head 13, 28, 39 Neg-incorporation 40 negative adverb 109 negative clause 138, 144, 148 negative marker/morpheme 80, 82, 84, 88, 109, 122, 142–143,

277

negative polarity item/NPI 72–73, 79, 80, preverbal negator 39, 143, 144 nominal 4, 43, 46, 56, 61, 88, 139, 191, 194, 202–203, 230–231 nominal domain 14, 18, 95, 103, 107–108, 210, 212, 230 nominal extension 7, 24 nominal features 94, 97, 101, 104 nominal modifier 212, 234 noun 4, 31, 42, 56, 58, 90, 91, 93, 95, 97, 100–101, 108, 127, 201, 208, 229, 230, 233, 239–240, 241 definite noun 239 deverbal noun 241–242 inflected noun 201 noun phrase/nominal phrase/NP 17, 19, 101, 108, 210–212, 216, 219, 224–245 verbal noun 56 null 71, 75 null case assigner 32–33, 37, 50 null complement 202 Null Complement Anaphora 81 null head 71 null object 194, 202–203 null subject 46, 197 numeral 233 object 13, 24, 31, 43, 61, 67–68, 77, 83, 88, 95, 97, 99–101, 104, 105, 145, 191–192, 194, 202–203, 205–208, 211, 215, 221, 227, 231 direct object 26, 51, 61, 77, 94, 137, 139, 140, 146, 152, 192–193, 203, 205–206, 208, 226 double object construction 211, 214, 216 indirect object 105, 214, 216, 226–228, 231 null object 194, 202–203 object control 43–46, 51 oblique subject constructions 17, 212, 216, 220, 243 operator 30, 72–73, 75, 111, 122, 125, 127 operator movement 30, 114, 117

278

INDE X OF SUBJECTS

operator (Continued) relative operator 114, 116, 125 wh-operator/wh-type operator 65, 111, 114, 116, 125 parameter(s) 2, 5, 8, 10, 12, 23–25, 38, 92, 202, 236 parameter setting 2, 8, 11, 17, 40, 50 parametric change 12, 55 parametric variation 7, 124 participle 16–17, 27, 56. 159–163, 177, 179–181, 183, 186, 189–193, 199–201, 203, 206–208 perfect participle 159, 162–163, 170, 182, 185 stative participle 162, 182 particle 8, 81, 188, 193 directional particle 195–196, 200 discourse particle 147–148 particle incorporation 193, 197 question particle 111, 122 verbal particle 115–116, 118–119, 127, 134 passive 19, 26, 31–32, 47, 50, 54, 162 passive agent 199 passive clause 26 passive infinitive 44–47, 51, 53–54 perfect 5, 16, 159, 161–165, 168–171, 174–185 double perfect/perfect doubling 16, 159–164, 166–170, 178–181, 184–185 past perfect 164, 170–173, 175, 182 perfect auxiliary 5, 16, 163, 177, 183 perfect participle 159, 162, 170, 182, 185 perfect time span 171–176, 182–183, 185 pluperfect 159–160, 167–169, 178–181, 185 present perfect 162,163, 165–166, 168–173, 175–177, 181, 183–184 present perfect puzzle 162, 172, 185 superperfect 16, 160–162, 164, 166–169, 178, 181–185 universal perfect 173, 182–183 perfective 44, 138–139, 175 periphery 14, 75, 83, 111–113, 115–116, 120, 133–135, CP-periphery 15, 111, 115, 120, 123–124, 126, 133

VP/vP-periphery 11, 86 left periphery 7, 14, 24, 30, 32–34, 71–72, 74–75, 80, 84, 87, 111–113, 115, 124, 126, 133, 135 phase(s)/phasal 1, 7–8, 29, 35, 60, 63, 69 phrase/phrasal 17, 24, 25, 28, 30, 35–36, 39, 41, 56, 71, 80, 91, 92, 105–108, 109, 171, 188, 192, 196, 203, 212, 217, 223, 226, 227, 232, 236 adpositional phrase 6–8, 18 applicative phrase/ApplP 215, 225, 228 Boolean Phrase 71 discontinuous phrase/structure 17, 231–233 exception phrase/exceptive phrase 72–73, 82, 86–87 Intonational Phrase 134 noun phrase/nominal phrase/NP 17, 19, 101, 108, 210–212, 216, 219, 224–245 phrasal modifier 212, 236 prepositional phrase/PP/pP 6–8, 11, 17, 22, 92, 174, 189–190, 194–196, 198–203, 206–208, 221–222, 228, 230–231, 244 tense phrase/TP 24, 30, 32–34, 39, 49–50, 54, 60, 69, 72, 80–82, 86, 95, 98, 101, 115–119, 122, 124, 126–128, 143, 147–148, 151–152, 154, 156, 171, 174 verb phrase/VP/vP 4, 7–8, 11, 12, 15, 17, 23, 25–28, 32, 39, 49–50, 60, 81, 86, 115, 118, 122, 138, 142, 151–152, 188, 202, 206, 215 pied-piping 25–26, 30, 38, 40, possessive 134, 210–211, 216, 218–219, 221, 224–225, 230, 232, 237, 243–244 possessive HAVE 174, 177–178 possessive preposition 174, 177 possessive pronoun 211, 216, 219–220, 224–225, 230–231, 234–238, 242–244 possessor 214, 225–226, 229 dative possessor 210, 216, 218, 225, 229–231 external possessor 210, 216, 219, 229–234, 244 internal possessor 218, 229, 231 postverbal postverbal position/placement 118, 137, 193, 203, 207

INDE X OF SUBJECTS postverbal subject 25 PP, see prepositional phrase pragmatic 113, 144, 193, 214, 216, 229, 231 Pra¨teritumschwund 159, 161, 164, 168–169, 176, 185 predicate 20, 22, 26, 29–33, 42–43, 46, 51, 170, 231, complex predicate, 43, 46–49, 53–55 control predicate 43 matrix predicate 23, 30–31, 43, 45–46, 50–51, 121 predicate ellipsis 77, 81, 84 preposition 14, 17, 33, 57, 62, 77, 90–94, 97–103, 105, 110, 163, 174–177, 185, 186–208, 211, 213, 220, 221, 241–242, 243 complement PP/PP complement 16–17, 186, 189, 198, 202, 208 locative preposition/PP 174, 198 P head 17, 188, 202 possessive preposition 174, 177 preposition incorporation/Pincorporation 16–17, 177, 186–187, 189–208 preposition stranding/P-stranding 198, 202, 207–208 prepositional movement 16, 186–190, 193–195, 200, 206 prepositional phrase/PP/pP 6–8, 11, 17, 22, 92, 174, 189–190, 194–196, 198–203, 206–208, 221–222, 228, 230–231, 244 temporal preposition 16, 174, 177, 180, 183–184 preterite 16, 140–141, 159–161, 164–169, 179–180, 185 preverbal 62, 117, 154–155 preverbal focus 15, 115, 116, 134 preverbal negator 39, 143, 144 preverbal position/placement 115, 134, 137, 153, 206 preverbal subject 25 Principles and Parameters theory 10 Probe/Probe-Goal relation 26–27, 29, 33, 35–36, 38, 41, 50, 60, 66 projection 1, 5–8, 13, 18, 24, 33, 35–36, 38, 40–41, 60, 63, 87, 92, 113, 115–118, 122–123, 125–127, 133–134, 215, 243

279

extended projection 1, 7–8, 24, 35–36, 40, 212 Extended Projection Principle, see EPP functional projection 4–8, 12, 14–15, 18, 23–25, 41, 80, 86, 109, 112–115, 117, 122–123, 201, 215, 217, 225, 233, 240 one feature - one projection principle 6, 105, 109, 110 phasal projection 1, 69 pronoun 11, 14, 66–68, 90–95, 101–104, 106, 109–110, 127–128, 147, 197, 208, 220, 228, 229, 230 demonstrative pronoun 235, 240 interrogative pronoun 97–100, 106 possessive pronoun 211, 216, 219–220, 224–225, 230–231, 234–238, 242–244 relative pronoun 14, 90–95, 99–101, 103, 110 wh-pronoun 92–93, 105–107 quantifier 71, 75, 102, 229, 233 question 111–114, 117–124, constituent question/wh-question 15, 111, 113–114, 117, 123, 125, 133, 134 embedded question 14–15, 111, 114–121, 123, 133 main clause question 113–114, 122 polar question/yes-no question 15, 111, 120–123, 133–134 question particle 111, 122 rhetorical question 80 tag question 141 reanalysis 2–4, 9–14, 17, 38, 40, 42–43, 49–53, 59–60, 67–68, 80, 82, 84, 91, 94, 124, 133–134, 137–138, 143, 154, 157–158, 166, 177, 217, 231, 236, 240, 243 recategorization 2, 4, 15, 136–140, 144, 148, 154, 157–158 recursion 5, 16, 86, 159, 161, 163, 178, 184–185 reduplicated/reduplication 186, 193–194, 205 relative 57, 61, 65, 90–110, 189, relative clause 57, 80, 84, 90, 92–95, 99–100, 103–105, 107, 110, 198, 202, 207

280

INDE X OF SUBJECTS

relative (Continued) relative complementizer 90–92, 94, 103 relative operator 114, 116, 125 relative pronoun 14, 90–95, 99–101, 103, 110 relativizer 6, 14, 90–95, 99–107, 110 resemantization 13 reversed result reading 160, 167, 181 scrambling 83, 88, 207, 231 semantic(s) 2–3, 7, 16, 43, 44, 51–52, 71, 80, 94, 111, 140, 144, 153, 161, 163, 171, 174–177, 179, 183–185, 215–216, 225, 243–245 semantic bleaching/weakening 3, 65 semantic change 3–4 semantic feature(s) 1, 4, 5, 6, 14, 76, 107, 109–110, shift 10, 21, 42, 55, 92, 124, 165, 168–169, 176, 177 specifier/Spec 5, 12, 14, 28, 33, 34, 36, 38, 81, 92, 94, 106–110, 112–113, 117–118, 125, 148, 206, 244 Spec, CP 114, 117, 127 Spec, ForceP 60 Spec, FP 25–26, 29, 37, 39–40, 113, 115–116, 118–120, 128, 134 Spec, PredP 115 Spec, TP 33, 115, 148 Spec, VoiceP 26, 32, 37 Spec, vP 27 spell-out 7, 10, 59, 62, 177, 181, 184 split 1, 5, 13, 15, 110, 144, 225 split CP 18, 63–64, 116 split Fin 13, 56, 59–61, 65–69 split infinitive 197 unsplit 63, 67 stative 134, stative participle 162, 182 stative verb 164, 181 stress 134 subject 12, 19, 23–27, 30–35, 37, 41, 44, 47, 49, 53, 58, 67, 69, 77, 92, 94–104, 115, 117–118, 120, 138, 146–148, 150, 155, 189, 197, 204, 215–216, 220, 226, 228–229, 243 embedded subject 23, 29–34, 37–38, 46, 49–50

oblique subject constructions 17, 212, 216, 220, 243 null subject 46, 197 preverbal subject 25 postverbal subject 25 subject-movement 12, 38, 51 subject-verb agreement 41, 67 wh-subject 32 subjunctive 13, 19, 57–60, 62, 65–69, 140, 167–168, 179–180, subjunctive clause 13, 64, 69 subjunctive mood 21 subjunctive verb form 21, 22 subordinate 73, 140–143, 158 subordinate clause 13, 83, 100, 112, 146–147, 154 subordinate conjunction 98 subordinator/subordination marker 13, 15, 93, 114, 125, 133–134 subordination 13, 15, 71, 120, 123–126, 133–134 conditional subordination 70, 73 supine 13, 56–69 supine clause 13, 57–58, 60–61, 63, 65–69 supine complement 56–57, 59–60, 66, 68–69 verbal supine 56–57, 61 Stylistic Fronting 16, 186–190, 194, 196–197, 199–200, 208 syntactic change 1–2, 5–7, 11, 18, 41, 71, syntactic variation 71, 89, 106, 137 target state 160, 181 temporal 69, 147, 162, 170, 180, 182 temporal adverb/adverbial 62, 161, 165, 171, 182 temporal argument 174 temporal auxiliary 159, 162–163, 166, 176, 180, 184 temporal connective 4 temporal HAVE 159, 174, 177–178, 184–185 temporal PP 198 temporal preposition 16, 174, 177, 180, 183–184 tense 7, 16, 44, 62, 63, 160–185, 201 past tense 163–164, 168–171, 173, 175–176, 179–182, 185

INDE X OF SUBJECTS present tense 44, 128, 140, 167, 170–172, 179, 181 preterite tense 165 T 24–25, 28, 32, 39, 50, 148, 151–152, 171, 174 170, 183 tense feature 179, 204, 208 tense recursion 159–185 tenseless 160, 177, 179–180 tense morphology 44, 140–141, 169, 179–180, 182 tense phrase/TP 24, 30, 32–34, 39, 49–50, 54, 60, 69, 72, 80–82, 86, 95, 98, 101, 115–119, 122, 124, 126–128, 143, 147–148, 151–152, 154, 156, 171, 174 thematic 7, 22, 23, 240, 245 non-thematic 212, 214, 221, 228, 243 thematic relation/role 45, 49, 92, 213, 215 time 61, 63, 140, 160, 165–167, 170, 172, 174, 181 evaluation time 170, 175 event time 170, 172, 175 perfect time span 171–176, 182–183, 185 positional past time adverbial 161–162, 165–166, 169, 173, 175–176 reference time 160, 169–170, 173 speech time 160, 169–170, 173, 175, 181–183 topic time 16, 170–175, 182–184 topic 62, 64, 105, 115–116, 118 topic position 34, 115 topic projection 116 topic time 16, 170–175, 182–184 topicalization 186–188, 193–194, 200 transitive 26, 219, 241–242 ditransitive construction 211, 217, 220–221, 224 intransitive verb 192, 219 transitive verb 28, 192–193, 202, 221 trigger 2, 8, 59, 67, 71, 133, 200, 229, 243 unaccusative 17, 32, 50, 162, 165–166, 168, 178, 222 unidirectional(ity) 9, 217, 236 uniform(ity) 7, 71, 202 universal(ly) 6, 9, 25, 125 universal determiner 75

281

universal perfect reading 178–179, 182–183 universal principle 8, 10, 125 unmarked 14, 61, 94, 95, 107, 109–110, 218 verb 4, 13, 16, 17, 21, 26, 32, 38–39, 41, 43, 47, 49, 51, 52, 54, 56–58, 62, 63, 77, 79, 81, 83, 86, 106, 113, 115, 118, 127–128, 133–134, 139–145, 148, 153, 159, 179, 181, 186, 189–193, 198, 199–208, 210–217, 220, 223, 227 causative verb 12, 21, 22, 46, 55 change of state verb 165 copular verb 79 inflected verb 201 finite verb 15, 25, 39, 41, 62–63, 115, 122, 124, 127, 142–147, 151–153, 155–156, 179–182, 189, 192–193, 203–206, 208 intransitive verb 192, 219 functional verb 46 lexical verb 4, 15, 22, 46, 47, 50, 105, 115, 117–119, 139, 141, 153, 159, 162, 164, 180, 182, 245 location verb 165 locative verb 198 loss of verb movement 4, 142, 146, 148 main verb 15, 19, 98, 136, 137–139, 141–148, 150–158, 188–190, 192–193, 196, 20 matrix verb 4, 23, 32, 43, 49, 50, 153, 196 modal verb 4 non-finite verb 28, 46, 62–63, 151, 154, 188–189, 192, 195, 206 perception verb 42, 47, 51 position verb 164 psych verb 220 stative verb 164, 180 supine verb 61, 62, 68 telic/atelic verb 181 three-place verb 43, 45 transitive verb 28, 192–193, 202, 221 unaccusative verb 17, 222 unergative verb 164 verb-adjacent position 111, 115, 117, 122 verb incorporation 190, 196, 201

282

INDE X OF SUBJECTS

verb (Continued) verb movement 12, 13, 15–16, 23, 25, 39, 62, 66, 109, 115, 122–123, 137, 142–148, 152, 154, 156–158, 190, 192, 201, 204 verbs of saying/believing 42, 51 verb phrase/VP/vP 4, 7–8, 11, 12, 15, 17, 23, 25–28, 32, 39, 49–50, 60, 81, 86, 115, 118, 122, 138, 142, 151–152, 188, 202, 206, 215 verb raising 154 verb root 201 verbal domain 4, 7, 8, 107 verbal particle 115–116, 118–119, 127, 134 V-to-C movement 33, 60, 62–63, 66–67, 190 V-to-F movement 12, 25, 38, 42, 118–119, 122–123 V-to-Fin2 movement 13, 65–68 V-to-T movement 65–68, 81, 83, 147, 148, 190, 192 V-to-I movement 4 voice 19 Voice head 24, 28, 32–33, 50, 183, 215 VoiceP 12, 24, 26–29, 32–33, 37, 39–41, 49–50, 170, 174 VoiceP movement 26–29, 40 VP/vP-periphery 11, 86 wh-constituent/element/item/morpheme/ phrase 90–92, 109, 111, 113–114, 116–118, 122–123, 125–126 wh-interrogative/wh-question 15, 113–114, 123, 125, 133 wh-movement 30, 51 wh-operator 65, 111, 114, 116, 125 wh-pronoun 92–93, 105–107

wh-subject 32 wh-variable 104, 110 word order 4, 10, 14–15, 26–28, 49, 58, 64, 71, 83–84, 86, 88, 115, 118, 136, 145–148, 197, 212, 229–232, 237 Dem Poss/Poss Dem order 237–238 neutral/unmarked order 118, 218 OV order 14, 17, 27–28, 40, 82–83, 88, 203, 206–208, 231 PP-V order 207 SAdvV order 16, 144–148, 150, 153–154, 156–158 SAdvMV order 16, 149–158 SAdvVM 154 SMAdvV order 149–152, 155, 157 SMVAdv order 149, 157 SAuxOV 39 SOV order 83, 122, 133–134 SOV(-Neg)-Aux 40 SVAdv 144–146, 154 SVAdvO order 146, 152 SVnever order 146 SVnot order 148 SVO/SV(O) order 53, 83 TopFocVX order 133–134 V{Adv/Neg}O order 142 VAux order 28 VOAux order 28, 40 VO order 231 V(O)S order 53 VPAux order 26–28, 40 VP-Neg-Aux order 27 V-PP order 207 word order change 10, 14, 18, 113, 122, 132–134 word order variation 18, 84, 151–158, 237–239

Index of languages Alemannic 159, 160, 164, 166, 168, 181 Bavarian 125 Catalan 46, 47, 81 Chicheŵa 207, 214 Danish 107, 159, 161, 165, 166, 172, 184 Dutch 73, 109, 111, 159, 161, 165, 166, 168, 171, 181–182, 184, 200, 217, 220, 221, 232, 244 Middle Dutch 165 English 3, 4, 15, 16, 25, 31–34, 43, 47, 73, 107, 109, 111, 114, 122, 125, 136–158, 162, 163, 165, 166, 168, 169–170, 172–178, 180, 183–185, 211, 217, 220, 221, 232, 236, 244 Early English 142, 144, 145, 151, 156, 158 Early Modern English 4, 12, 15, 137, 145, 147, 150, 214, 229 Middle English 12, 125, 139, 145, 150, 155, 156, 206, 225 New Zealand English 134 Old English 3, 4, 12, 15, 16, 137, 139, 145, 150, 156, 206, 214, 219, 225, 229, 245 Present Day English 15, 136 French 46, 47, 72–73, 74, 81, 88, 91, 104, 106, 107–108, 109, 148, 159, 182, 226, 229, 232 French dialects 161, 181 Modern French 95–95, 104 Old French 22, 70, 74, 104 Galician 74, 83 Old Galician 70 German 5, 16, 25, 107–108, 109, 111, 114, 125, 126, 159, 161–185, 217–218, 220–222, 229, 241, 245 Bavarian 125 Early New High German 165 German dialects 125, 159, 164, 165, 168 Middle Low German 165, Old High German 164, 219

West-Germanic languages 216, 232 Greek 25, 108 Classical Greek 192, 194 Hungarian 5, 8, 11, 14–15, 81, 111–135, 142 Middle Hungarian 12, 15, 120–121, 123, 126 Modern Hungarian 14–15, 112–113, 120–123, 125–127, 134 Old Hungarian 12, 14–15, 111–113, 120, 122–124, 126, 128–134 Proto-Hungarian 122, 132 Icelandic 8, 10, 11, 16–17, 29, 172, 176, 186–246 Modern/present day Icelandic 17, 197–198, 205, 211–212, 219–221, 223, 228–229, 231, 233–234, 238–239, 241, 243–244 Old Icelandic 12, 17, 186–188, 191–201, 203–208, 210, 239, 244 Italian 10, 14, 25, 46, 47, 81, 90–110, 142, 159, 161, 168, 232, 236 Italian dialects 12, 95, 103–104, 110 Medieval North Western Italian 95 Modern Standard Italian 47, 91–94, 102 Old Italian 6, 14, 22, 90–94, 103–104, 110 Japanese 111 Latin 4, 12, 19–55, 56, 84, 104, 212, 232, 234 Classical Latin 12, 19, 21, 23–24, 26–28, 38, 40–41, 43, 51–52, 55 Early Latin 19, 20, 23, Late Latin 12, 20, 22, 23–24, 38–41, 46, 48, 52–55 Norwegian 109, 172, 176, 234 Old French 22, 70, 74, 104 Old Galician 70 Old Ligurian 91, 95, 97, 100–101, 103–104 Old Neapolitan 91, 95, 98–104

284

INDE X OF L ANGUAGES

Old Norse 8, 17, 210–213, 216, 219, 221–235, 237–245 Eddic Old Norse 235 Runic Old Norse 234–235 Old/Medieval Occitan 70, 104 Old Piedmontese 91, 95, 97–98, 100–101, 103–104 Old Portuguese 12, 22, 53, 71, 73–77, 80–89 Old Romanian 13, 56–61, 66–68 Old Saxon 164, 219 Old Spanish 53, 70 Portuguese 7, 13–14, 55, 70–89 Contemporary (European) Portuguese 46, 71–72, 80–81, 83–84, 86, 87

Old Portuguese 12, 22, 53, 71, 73–77, 80–89 Romance languages 21–22, 42–43, 47, 51–53, 55, 60, 71, 81, 84, 87, 91, 92, 110, 176, 212, 216, 232, 234 Romanian 5, 7, 13, 46, 56–69 Modern Romanian 13, 56, 58–61, 67–69 Old Romanian 13, 56–61, 66–68 Scandinavian languages 165, 166, 172–173, 217, 244 Mainland Scandinavian 221 Spanish 46, 71–73, 81, 110, 176, 232 Old Spanish 53, 70 Swabian 168 Swedish 107, 162–163, 172, 176

OX FORD STUDIES IN DIACHRONIC AND HISTOR IC AL LINGUISTICS

General editors Adam Ledgeway and Ian Roberts, University of Cambridge Advisory editors Cynthia L. Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge PUBLISHED 1 From Latin to Romance Morphosyntactic Typology and Change Adam Ledgeway 2 Parameter Theory and Linguistic Change Edited by Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar 3 Case in Semitic Roles, Relations, and Reconstruction Rebecca Hasselbach 4 The Boundaries of Pure Morphology Diachronic and Synchronic Perspectives Edited by Silvio Cruschina, Martin Maiden, and John Charles Smith 5 The History of Negation in the Languages of Europe and the Mediterranean Volume I: Case Studies Edited by David Willis, Christopher Lucas, and Anne Breitbarth 6 Constructionalization and Constructional Changes Elizabeth Closs Traugott and Graeme Trousdale 7 Word Order in Old Italian Cecilia Poletto 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent

9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro 11 The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by Katalin É. Kiss 12 Syntactic Reconstruction and Proto-Germanic George Walkden 13 The History of Low German Negation Anne Breitbarth 14 Arabic Indefinites, Interrogatives, and Negators A Linguistic History of Western Dialects David Wilmsen 15 Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden 16 Syllable and Segment in Latin Ranjan Sen 17 Participles in Rigvedic Sanskrit The Syntax and Semantics of Adjectival Verb Forms John J. Lowe 18 Verb Movement and Clause Structure in Old Romanian Virginia Hill and Gabriela Alboiu 19 The Syntax of Old Romanian Edited by Gabriela Pană Dindelegan 20 Grammaticalization and the Rise of Configurationality in Indo-Aryan Uta Reino¨hl 21 The Rise and Fall of Ergativity in Aramaic Cycles of Alignment Change Eleanor Coghill

22 Portuguese Relative Clauses in Synchrony and Diachrony Adriana Cardoso 23 Micro-change and Macro-change in Diachronic Syntax Edited by Eric Mathieu and Robert Truswell 24 The Development of Latin Clause Structure A Study of the Extended Verb Phrase Lieven Danckaert 25 Transitive Nouns and Adjectives Evidence from Early Indo-Aryan John J. Lowe 26 Quantitative Historical Linguistics A Corpus Framework Gard B. Jenset and Barbara McGillivray 27 Gender from Latin to Romance History, Geography, Typology Michele Loporcaro 28 Clause Structure and Word Order in the History of German Edited by Agnes Ja¨ger, Gisella Ferraresi, and Helmut Weiß 29 Word Order Change Edited by Ana Maria Martins and Adriana Cardoso 30 Arabic Historical Dialectology Linguistic and Sociolinguistic Approaches Edited by Clive Holes 31 Grammaticalization from a Typological Perspective Edited by Heiko Narrog and Bernd Heine 32 Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou 33 Indefinites between Latin and Romance Chiara Gianollo

34 Verb Second in Medieval Romance Sam Wolfe 35 Referential Null Subjects in Early English Kristian A. Rusten 36 Word Order and Parameter Change in Romanian A Comparative Romance Perspective Alexandru Nicolae 37 Cycles in Language Change Edited by Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen 38 Palatal Sound Change in the Romance Languages Diachronic and Synchronic Perspectives André Zampaulo 39 Dative External Possessors in Early English Cynthia L. Allen 40 The History of Negation in the Languages of Europe and the Mediterranean Volume II: Patterns and Processes Anne Breitbarth, Christopher Lucas, and David Willis 41 Variation and Change in Gallo-Romance Grammar Edited by Sam Wolfe and Martin Maiden 42 Phonetic Causes of Sound Change The Palatalization and Assibilation of Obstruents Daniel Recasens 43 Syntactic Features and the Limits of Syntactic Change Edited by Jóhannes Gı´sli Jónsson and Thórhallur Eythórsson 44 Romance Object Clitics Microvariation and Linguistic Change Diego Pescarini 45 The Diachrony of Differential Object Marking in Romanian Virginia Hill and Alexandru Mardale

46 Noun-based Constructions in the History of Portuguese and Spanish Patrı´cia Amaral and Manuel Delicado Cantero 47 Syntactic Change in French Sam Wolfe 48 Periphrasis and Inflexion in Diachrony A View from Romance Edited by Adam Ledgeway, John Charles Smith, and Nigel Vincent 49 Functional Heads Across Time Syntactic Reanalysis and Change Edited by Barbara Egedi and Veronika Hegedűs IN PREPARATION Redevelopment of Case Systems in Indo-Aryan Miriam Butt Classical Portuguese Grammar and History Charlotte Galves, Aroldo de Andrade, Christiane Namiuti, and Maria Clara Paixão de Sousa Morphological Borrowing Francesco Gardani Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τoς and its Implications for Language Change and Structure Brian D. Joseph Reconstructing Pre-Islamic Arabic Dialects Alexander Magidow Iranian Syntax in Classical Armenian Robin Meyer