The Evolution of Functional Left Peripheries in Hungarian Syntax (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated] 9780198709855, 0198709854

Citation preview

The Evolution of Functional Left Peripheries in Hungarian Syntax

OX F O R D S T U D I E S I N D IAC H R O N IC A N D H I S T O R I C A L L I N G U I S T I C S general editors Adam Ledgeway and Ian Roberts, University of Cambridge advisory editors Cynthia 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; David Willis, University of Cambridge recently published in the series 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 For a complete list of books published and in preparation for the series, see pp. 264–65

The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by KATA L I N É . K I S S

1

3

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 Katalin É. Kiss 2014 © the chapters their several authors 2014 The moral rights of the authors have been asserted First Edition published in 2014 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: 2013957965 ISBN 978–0–19–870985–5 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  Introduction Katalin É. Kiss . . . . . . . .

Goals The origins of Hungarian Periods in the history of Hungarian The sources of the history of Hungarian Theoretical assumptions Results Technical issues Acknowledgements

 The evolution of functional left peripheries in the Hungarian sentence Katalin É. Kiss . Introduction . The Proto-Hungarian sentence was SOV, with fused functional and thematic domains . The shift from head-final to head-initial . The evolution of a left-peripheral PredP . The evolution of a left-peripheral FocP . The evolution of a left-peripheral NegP . Conclusion  The DP-cycle in Hungarian and the functional extension of the noun phrase Barbara Egedi . . . . .

Introduction Marking of definiteness in Old Hungarian The DP-cycle Competing strategies Summary

vii viii xi                       

vi

Contents

 From A-quantification to D-quantification: Universal quantifiers in the sentence and in the noun phrase Ágnes Bende-Farkas . . . . .

Introduction From all to every Quantificational strategies Free pronouns as distributive operators The impact of D-quantification

 The cyclical development of Ps in Hungarian Veronika Hegedűs . . . . .

Introduction Spatial elements in Old Hungarian Structural changes in the PP The P cycle Conclusions

 From non-finite to finite subordination: The history of embedded clauses Júlia Bácskai-Atkári and Éva Dékány . . . . .

Introduction On the definition of finite and non-finite clauses The changes affecting non-finite subordinate clauses The development of finite subordinate clauses Conclusion

Appendix: Corpus building from Old Hungarian codices Eszter Simon      

Introduction Collecting the original linguistic material Corpus annotation The structure of the corpus The corpus query tool Final remarks

Primary Sources References Index

                           

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, i.e. 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 abl

ablative case

acc

accusative case

ade

adessive case

all

allative case

Art

article

AxPartP

AxialPart Phrase

C.

Codex

caus

causative

com

comitative case

cond

conditional

conj

conjunction

CP

complementizer phrase

dat

dative case

defo

agreement with definite object

del

delative case

Dem

demonstrative

dist

distributive case

DP

Determiner Phrase

ela

elative case

ep

epenthetic

fem

feminine

final

final case

FinP

finiteness phrase

foc

focus

FocP

focus phrase

ForceP

Force phrase

ill

illative case

imp

imperative

ine

inessive case

List of Abbreviations inf

infinitive

ins

instrumental case

IntP

interrogative phrase

lat

lative case

loc

locative case

mod

modal

N

noun

neg

negation

nom

nominative case

obj

object agreement

OCR

Optical Character Recognition

OH

Old Hungarian

Op.

operator

p

functional adposition

P

adposition

part

participle

pass

passive

pastpart

past participle

PathP

Path Phrase

perf

perfect

PF

phonological form

pl

plural

PlaceP

Place Phrase

PolP

polarity phrase

POS

part of speech

poss

possession

poss.pl

possessed, plural

possib

possibility

PP

Adpositional Phrase

PredP

predicative phrase

prt

verbal particle

pst

past

pst.part

past participle

ix

x

List of Abbreviations

Q

interrogative complementizer

qprt

interrogative discourse particle

r

recto

refl

reflexive

sbjv

subjunctive

sg

singular

SPEC

specific

sub

sublative case

sup

superessive case

term

terminative case

TopP

topic phrase

trans

translative case

v

verso

VM

verbal modifier

List of Contributors Júlia Bácskai-Atkári received her Ph.D. in General Linguistics at the University of Potsdam, Germany, where she currently works as a research fellow. Her research concentrates on the syntax of comparative subclauses, including general questions of a functional left periphery, obligatory and optional deletion phenomena, as well as the internal structure of degree expressions. Applying a cross-linguistic approach, her main interests are Germanic languages and Hungarian. Ágnes Bende-Farkas is a researcher at the Research Institute for Linguistics of the Hungarian Academy of Sciences in Budapest. She has a Ph.D. from Stuttgart University (2002). Her specialties are dynamic semantics and cross-linguistic semantics. She has worked on the semantics of complex predicates, Focus, quantification, and indefinites. Éva Dékány received a Ph.D. from the Center for Advanced Study in Theoretical Linguistics at the University of Tromsø. She is currently a postdoctoral researcher at the Research Institute for Linguistics of the Hungarian Academy of Sciences. She has written on various aspects of nominal and adpositional morphology and syntax, as well as non-finite subordination. Barbara Egedi is a research fellow at the Research Institute for Linguistics of the Hungarian Academy of Sciences. She attained her Ph.D. in Egyptology for studying Ancient Egyptian and Coptic nominal syntax. She has also published papers on the syntax of Hungarian adverbials as well as on Egyptian-Greek contact linguistics, and is currently concerned with noun phrase structure of Old and Middle Hungarian from both a comparative and a diachronic perspective. Katalin É. Kiss is professor of the Research Institute for Linguistics of the Hungarian Academy of Sciences and of Pázmány Péter Catholic University. She is a member of the Hungarian Academy of Sciences and of Academia Europaea. Her research areas include generative syntax, information structure, and the syntax–semantics interface. She has published 6 books and 170 articles. Veronika Hegeds is a researcher at the Research Institute for Linguistics of the Hungarian Academy of Sciences. She did her Ph.D. at Tilburg University on the syntax of non-verbal predicates and predicate movement in Hungarian. Her research interests include the structure of adpositional phrases, copular constructions, and secondary predicates, both synchronically and diachronically.

xii

List of Contributors

Eszter Simon is a research fellow at the Research Institute for Linguistics of the Hungarian Academy of Sciences. She did her Ph.D. at the Cognitive Science Department of Budapest University of Technology and Economics on Hungarian named entity recognition. Her research interests include corpus building and several computational linguistics subtasks, such as morphological analysis and machine learning algorithms for information extraction.

 Introduction K ATA L I N É . K I S S

. Goals This book summarizes the results of a four-year project devoted to the study of the diachronic syntax of Hungarian. The primary goal of the project has been empirical: to reveal the changes that have taken place at various levels of Hungarian syntax in the past millennium, and to find their triggers and the mechanisms of their progress. Naturally, the empirical results of these investigations are of theoretical relevance. Hungarian is among the few non-Indo-European languages with a documented history spanning more than 800 years. Hence it provides a good testing ground for the principles and constraints of language change that have been observed in other languages, first and foremost the Indo-European language family.

. The origins of Hungarian Hungarian belongs to the Uralic language family; it is one of the Ugric languages of its Finno-Ugric branch. The Uralic proto-language is assumed to have been spoken west of the Urals in the Neolithic Era. The Samoyedic and Finno-Ugric branches of the family split around 5000 BC, with the former dispersing northward and eastward, and the latter spreading westward. The Ugric branch is believed to have split off from the Finno-Ugric proto-language at around 3000 BC. The period of Ugric unity with a more or less homogeneous Ugric proto-language is believed to have lasted from 3000 to 2000 BC, when Hungarian split off from the Ugric branch (Róna-Tas 1999, 2011).1 The Ugric sister languages of Hungarian, the so-called Ob-Ugric languages, which later split into Khanty (Ostyak) and Mansi (Vogul), remained close to the 1 The earlier Finno-Ugric literature hypothesizes a shorter time frame for all this: the Uralic proto-language split into a Finno-Ugric and a Samoyedic branch at about  BC, the Finno-Ugric proto-language split into a Finnic and a Ugric branch at about  BC, and the Ugric proto-language split into a Hungarian and an Ob-Ugric branch at about  BC (cf. Hajdú , Sinor ).



Katalin É. Kiss

original habitat of the Uralic peoples; now they live on the eastern side of the Urals in Western Siberia. These languages are on the verge of extinction; Khanty has about 14,000 speakers, and Mansi has about 3,500 speakers, practically all of whom are Russian-dominant bilinguals. After the break-up of Ugric unity, Hungarian drifted far from her Ugric sister languages both geographically and linguistically. The linguistic divergence of Hungarian may have been speeded up by intensive contacts with Turkic tribes in the 5th–8th centuries A.D. The Hungarian tribes migrating westward on the European steppe settled in the Carpathian Basin in 896, where they are likely to have absorbed a local Slavic population. It is an open question what the syntax of the vernacular Slavic language spoken in Central Europe in the 9th century was like, and whether its substratum effect had a role in the shift of Hungarian syntax from head-final to head-initial.

. Periods in the history of Hungarian Hungarian language history is divided into periods on the basis of historical events. The Proto-Hungarian period began with the splitting of Hungarian off the Ugric proto-language, and ended in 896, the year of the Hungarians’ settlement in the Carpathian Basin. Its last phase, referred to as the Late Proto-Hungarian period, began in the 5th century, the time of intensive Hungarian–West Old Turkic (Khazar) contacts. The Old Hungarian era lasted from 896 to 1526, the year of the occupation of the major part of the Hungarian Kingdom by the Osman Empire. The first part of this period (between 896–1350), documented by linguistic fragments and short coherent texts, is called the Early Old Hungarian period. The Late Old Hungarian period between 1350–1526 is the period of codices. The Middle Hungarian era lasted from 1526 to 1772, the date marking the beginning of the era of Enlightenment in Hungarian literature. The Modern Hungarian period between 1772 and the present has an important internal landmark: 1920, the time of the Trianon (Versailles) Treaty, which broke up the Hungarian Kingdom (then part of the Austro-Hungarian Monarchy). Since then, the versions of Hungarian spoken inside and outside the remaining Hungarian state have started slowly drifting apart. Our diachronic investigations focus on the period between the end of the 12th century and the end of the 16th century, i.e., the second half of the Old Hungarian period and the beginning of the Middle Hungarian era. Naturally, when we have sufficient evidence, we trace back the changes to the Proto-Hungarian period, and we also follow some ongoing changes up to Modern Hungarian times. We focus on the Hungarian language of the 13th–16th centuries because this is the period of the documented history of Hungarian when the most fundamental changes took place. The fact that we analyzed primarily 13th–16th century Hungarian documents also has a practical reason: as part of our Hungarian diachronic syntax project, we started building an electronic database of Old and Middle Hungarian texts. We started

Introduction



digitalizing and annotating the Hungarian historical corpus in a chronological order, and by the end of the project, we processed the Old Hungarian and Early Middle Hungarian codices, a corpus of 2.2 million words.

. The sources of the history of Hungarian The first three centuries of Old Hungarian are documented by fragments: Hungarian phrases, mostly personal names and place names in Latin and Greek documents, among them deeds of gift and foundation, land registers, court decisions, etc. The place names are often complex expressions that provide useful information on the structure of noun phrases and postpositional phrases, for example: (1) feheruuaru rea mene hodu utu rea Feheruaru [white castle] to going army road to ‘to the army road going to Feheruaru’ (2) Petre zenaia hel rea Peter hay-poss place to ‘to the place (called) Peter’s hay’ The first surviving coherent Hungarian texts are the 50-clause Funeral Sermon and Prayer from around 1192–1195, and a beautiful poem, Old Hungarian Lamentations of Mary, from 1300. The most important sources of Old Hungarian are 47 hand-written codices, written often for—and by—nuns, who did not read Latin. The spreading of the Hussite movement promoted the use of Hungarian in male monasteries, as well. These codices mainly contain regulations concerning monastic life, sermons, prayers, contemplations, legends, and excerpts from the Bible. The earliest surviving codex is the Jókai Codex, containing the legend of Saint Francis, translated from a Latin original around 1370, copied in 1446. Several codices contain translations of various books of the Bible. Subsequent translations of the same biblical text facilitate the tracing of syntactic changes. Not all codices are translations though, and not all codices contain only ecclesiastical texts. There are also other sources of vernacular Old Hungarian: private letters, testimonies of witnesses, a verse-chronicle, etc. A question often raised in connection with texts translated from Latin is to what extent their syntax has been influenced by Latin. The in-depth investigations of Egedi (2014) have shown that translations from Latin do not use constructions that are unattested in original, non-translated texts. The translators often adopt certain conventions, but the effect of such conventions is manifest at most in the increased frequency of certain constructions; i.e., it is quantitative, not qualitative. Middle Hungarian is documented by an exponentially growing number of printed books, representing a great variety of genres, registers, and dialects. Book printing started in Hungary in 1472; the first Hungarian-language book, Saint Paul’s letters, appeared in 1533.



Katalin É. Kiss

The sources that we cite our examples from are listed at the end of this book, where we supply the year of their writing or translation, the year of their copying, and the data of the authoritative edition that we used.

. Theoretical assumptions The theoretical background of our analyses of subsequent stages of Hungarian syntax and the syntactic changes having taken place is generative linguistic theory, and its concept of diachrony (see Lightfoot 1979; 1991; 1999). We assume that syntactic changes are inseparable from the process of language acquisition. The grammar of adults is essentially stable; changes happen when those acquiring a language assign to linguistic expressions other representations than those who generated them. Reanalyses by new generations of speakers often lead to contextual extension, generalization, and abstraction, as well. Lexical elements are prone to morphological and phonological erosion and semantic bleaching, which may result in their grammaticalization as function words. These changes have been claimed to have a cyclic nature (see, e.g., van Gelderen 2011), which we have also observed repeatedly in Hungarian. The  spreading of linguistic changes usually follows an -curve. The supplanting of an old construction by a new one can be a prolonged process; obsolete patterns can survive in infrequent sentence types or as linguistic fossils for hundreds of years, which can enable us to reconstruct phenomena from undocumented stages of the language. We have observed linguistic fossilization on various levels of Hungarian syntax, and we have relied on linguistic fossils in syntactic reconstruction.

. Results The studies in this book focus on the restructuring of Hungarian syntax from headfinal to head-initial, which started in the Proto-Hungarian age and led to fundamental structural changes resulting in the evolution of functional left peripheries on various levels of syntactic structure by the 16th century. Chapter 2 analyzes the changes that took place on the clause level. It argues that—though Hungarian has been a head-initial language throughout its documented history—Proto-Hungarian was consistently head-final. The claim is based on archaisms of early Old Hungarian converging with facts about Khanty and Mansi, the sister languages of Hungarian. The head-final constructions surviving from ProtoHungarian include SOV non-finite clauses with a morphologically unmarked subject and object, a fossilized V–auxiliary order (indicative of a head-final TP), and a clausefinal interrogative complementizer (evidence of a head-final CP). In Proto-Hungarian, the subject also functioned as a topic, and the object functioned as a focus or secondary topic. In the latter case, it elicited verbal agreement, which was blocked by the Inverse Agreement Constraint if the secondary topic was inherently more topical than the

Introduction



primary topic. The resulting gaps in object–verb agreement have fossilized and are still part of Hungarian grammar. The process that resulted in the reanalysis of the head-final Proto-Hungarian clause as a head-initial structure is hypothesized to have started with the spreading of rightward extraposition, which led to the reinterpretation of the postverbal section of the sentence as the base position of arguments, and the reanalysis of the preverbal section as a purely functional domain. What happened in the left periphery of the Old Hungarian sentence was more than the preverbal constituents of Proto-Hungarian, associated simultaneously with thematic and discourse functions, losing their thematic restrictions. The Proto-Hungarian lexical node dominating an occasional verbal particle plus verb or nominal predicate plus copula complex has developed into a PredP projection, the locus of complex predicate formation, having a crucial role in encoding situation aspect. The Proto-Hungarian object-focus has been reanalyzed as a [+exhaustive]/[+maximal] constituent in the specifier of a FocP projection, triggering verb movement. The negative particle has developed from an adverb-like element adjoined to the verb into a Neg head projecting a NegP, eliciting negative concord. The topic projection has become available for arguments of any grammatical function. Finite embedded clauses have developed a CP layer. Chapter 3 is concerned with the noun phrase, especially with the marking of definiteness. Several stages of the DP-cycle are discussed as well as their structural consequences in Hungarian. Beginning with the first written sources, the chapter tracks and analyzes how the marking of definiteness changed in time and how the emergence of the definite article and its functional spreading reshaped the nominal left periphery. After the definite article has grammaticalized through the reanalysis of a demonstrative element, the demonstrative modification had to be renewed. This typically takes place either by reinforcing the old form through a locative adverb or by the incorporation of an additional demonstrative (e.g. an appositive pronoun) into the construction. The paper investigates this process in Hungarian showing that the renewal of the demonstrative system involved both types of strategies, reinforcement as well as determiner doubling, whereas the old construction was also preserved to some extent. The reinforcing strategy produced two new series of demonstrative pronouns, while the determiner doubling phenomenon entailed a long-term structural reanalysis at the nominal left periphery. The word order change that can be observed in cases when demonstratives are combined with dative-marked possessor expressions indicates that the demonstratives first linked with the noun phrase headed by the definite article by means of adjunction, but later were incorporated into the DPinternal specifier position. The competing strategies of demonstrative modification coexist from the Middle Hungarian period onward, and their descendant constructions survive in presentday Hungarian as well, though used with a different degree of markedness. Dialectal



Katalin É. Kiss

data are also considered to show that in a few varieties of Hungarian a further step of demonstrative integration can also be observed. The main focus of chapter 4 is the divergent syntax and semantics of Old Hungarian mind ‘all’ and minden ‘every’, ‘everyone’, respectively. According to the evidence, the ‘floating’ A-quantifier mind is an adverb, adjoined to an extended projection of the VP or to its associate DP. It is interpreted as a maximality operator which is not inherently distributive, and which enters into a semantic association with plural definites. Minden in Old Hungarian was not obligatorily distributive; nevertheless it is argued to have introduced a tripartite structure even in non-distributive cases. Old Hungarian data support a more conservative view on D-quantification, in that the proposal in Matthewson (2001) was found not to be viable for Hungarian. Hungarian determiners (or floating adverbs) do not project a QP that takes the DP as its complement. Instead, the adverb mind is adjoined to the DP, while determiners are below the D-layer, i.e. within the DP itself. Since very little is known about the early stages of minden as an independent lexeme, a type-driven reconstruction is offered, supplemented by empirical evidence. According to this scenario minden was initially an adjective or adverb meaning ‘entire(ly), complete(ly)’. From DP-internal modifier it became a determiner via type shifting. DP-minden is derived from determiner-minden by saturating its first argument with the universal predicate ‘thing’. The chapter discusses the impact of determiner quantification on syntactic structures and the syntax–semantics interface. These include the reinforcement of the DPinternal NumP/QuantP layer and the availability of preverbal positions for quantifier movement. A negative finding is that Old Hungarian appears to have lacked the Distributive Phrase proposed in Beghelli and Stowell (1997) and Szabolcsi (1997): Old Hungarian manuscripts contain several instances of preverbal and non-distributive DPs with minden. Old Hungarian quantifiers appear to have moved for reasons of scope; a special case was when minden moved in order to convey negated existential statements (expressing ‘No Xs are Ys’ with logically equivalent ‘All Xs are not Ys’). One of the fundamental assumptions of the chapter is that prior to D-quantification, Proto-Hungarian or Old Hungarian had propositional quantification over indeterminate pronouns (Kratzer and Shimoyama 2002). Two Old Hungarian distributivity operators containing the pronoun ki ‘who’ (ki-ki ‘who-who’ and ki mind ‘who all’) are said to be (refurbished and reinterpreted) remnants from this period. The chapter ends with a comparison of the two modes of quantification, and concludes that if Hungarian did have the ‘indeterminate’ mode of quantification, the impact of D-quantifiers could have been quite radical, exchanging unselective, locally unconstrained quantification with frozen scope relations for selective, locally constrained, and scopally flexible quantifiers.

Introduction



Chapter 5 discusses the grammaticalization process and cyclical development of adpositions in Hungarian and the changes within the adpositional phrase (PP). Postpositions often develop in possessive structures, and can go through several steps in grammaticalization by losing their nominal features and becoming more functional elements, and sometimes losing their spatial meaning by the end of the process. At this point, a new element may enter the chain of changes, and a new cycle of grammaticalization may begin. The chapter shows that as a result of grammaticalization, a functional projection on the left periphery of the PP developed, the head of this new projection being different from the more lexical postpositional head in that it is peripheral within the phrase and it does not have to follow its complement. This functional head came to be filled in more and more frequently by adpositional elements that function as (verbal) particles within the clause. The fact that these elements are no longer in a strictly head-final structure results in them moving in the clause freely, independently of their complement, contrary to the other group of postpositions, which are either suffixal or morphologically less dependent but syntactically always head-final. Chapter 6 surveys the changes that took place in the history of Hungarian embedded clauses. It is hypothesized that between the Proto-Hungarian and Old Hungarian periods, an SOV to SVO change took place. SOV languages strongly prefer non-finite embedding, while finite subordination is typical of SVO languages. It is shown that from Proto-Hungarian to Old Hungarian, the role of non-finites indeed gradually diminished, while finite subordination assumed an increasingly important role. On the one hand, this involved the loss of certain non-finite types, the narrowing of external distribution and the decreased productivity of others. In addition, the remaining types of non-finites became more prototypically non-finite, some of them losing the ability to license a referentially independent subject or to agree with the subject. On the other hand, the system of finite subordinators became richer by various elements grammaticalizing as complementizer heads and by way of these elements combining with each other, also accompanied by an increased need to mark finite subordination overtly. Hence in finite clauses the development of a functional left-periphery can be observed, which came to be head-initial and started to exhibit multiple layers overtly. The development of the CP domain displayed repeated rounds of the relative cycle: the complementizers generated in C were all pronominals preposed into Spec,CP from argument or adjunct positions originally. The book also contains an appendix, introducing the Old Hungarian Corpus that has been prepared in the framework of our project to facilitate the research. The appendix presents the workflow of building the corpus and supplying it with several kinds of metadata. It gives an overview of corpus annotation, the normalization of tokens, the morphological analysis, and morphosyntactic disambiguation. Finally, it describes the corpus query tool, which is freely available to the linguistic community and to the public.



Katalin É. Kiss

. Technical issues Old Hungarian is essentially comprehensible for the Modern Hungarian speaker; nevertheless, the reading of Old Hungarian documents can be very difficult because of their archaic spelling. Scribes used the Latin alphabet to spell Hungarian; however, whereas Latin has 5 characters for vowels, the number of Hungarian vowels is 14. Hungarian also has 8 palatal consonants absent from Latin. Different scribes used different solutions (letter combinations and diacritics) to represent the Hungarian phonemes missing in Latin. So as to make the reading of our examples easier, we have replaced certain unusual characters and symbols with better-known equivalents. We have employed the following paleographical simplifications: ÿ→y l→č →s →z →& eσ → &

. Acknowledgements The Hungarian Diachronic Syntax project, whose results we summarize in this volume, took place at the Research Institute for Linguistics of the Hungarian Academy of Sciences, financed by OTKA, the Hungarian National Scientific Research Fund, under grant 78074. We wish to acknowledge the financial support of OTKA, and the logistic and intellectual support of the Research Institute for Linguistics. We owe thanks to several colleagues who participated in the project temporarily in some way or another, among them Sylvia Blaho, Judit Farkas, Mátyás Gerőcs, Katalin Gugán, Andrea Kacskovics-Reményi, Gergely Kántor, Eszter Mihály, Iván Mittelholcz, Attila Novák, Márta Peredy, Katalin Pólya, Bálint Sass, Dániel Szeredi, Orsolya Tánczos, and Ildikó Tóth.

 The evolution of functional left peripheries in the Hungarian sentence K ATA L I N É . K I S S

. Introduction This chapter argues that—even though Hungarian has been a head-initial language with a preverbal functional domain throughout its documented history—the ProtoHungarian sentence was SOV, where the thematic and functional roles were fused: the subject also functioned as a topic, and the object functioned as a secondary topic or a focus, depending on whether or not it elicited verbal agreement. These claims are based on evidence provided by Old Hungarian archaisms and Modern Hungarian linguistic fossils, as well as converging phenomena of the genetically related languages, first and foremost Khanty and Mansi, the Ugric sister languages of Hungarian. The process that resulted in the reanalysis of the head-final Proto-Hungarian clause as a head-initial structure is hypothesized to have started with the spreading of rightward extraposition, which led to the reinterpretation of the postverbal section of the sentence as the base position of arguments, and the reanalysis of the preverbal section as a purely functional domain. This hypothesis explains such curious properties of the Modern Hungarian sentence as the freedom of postverbal word order, and the referentiality constraint on postverbal arguments, and is also supported by parallel developments in present-day Finno-Ugric languages. Whereas the left-peripheral functional positions of the Modern Hungarian sentence (an immediately preverbal predicative position, a focus slot, a Neg head, a landing site for quantifier raising, and a topic position) have been accessible since the first surviving Old Hungarian document, in early Old Hungarian the left periphery was utilized to a much lesser extent than it is today. The PredP projection had become the locus of the consistent encoding of telicity only by the end of the



Katalin É. Kiss

Old Hungarian period. The potential targets of focus movement also represented a gradually increasing set. For example, in early Old Hungarian, interrogative whadjuncts could remain in situ, and the wh-phrases of embedded questions surfaced in Spec,CP, instead of Spec,FocP. The negative particle first only appeared in the leftperipheral Neg head if the left edge of the scope of negation was not marked in another way (by a negative pronoun participating in negative concord). The disappearance of the old pattern of negation, with the negative particle adjoined to the V in an adverblike fashion, was a gradual process spanning several centuries. The chapter is structured as follows: Section 2.2 analyzes the evidence testifying that Proto-Hungarian was SOV. Section 2.2.1 argues for the head-finality of the VP, TP, and CP projections. Section 2.2.2 discusses the fusion of the subject and topic functions and the object and focus functions attested in various SOV Uralic languages, and the role of differential object–verb agreement in exempting objects from the focus function, and it points out the remnants of this system in Modern Hungarian. Section 2.3 deals with the shift from head-final to head-initial grammar. Section 2.3.1. shows that the first surviving Hungarian written document from the late 12th century is already head-initial, and section 2.3.2 speculates about how the shift from head-final to head-initial could happen. Section 2.4 is devoted to the evolution of a preverbal PredP projection. Section 2.5 compares the properties of the Proto-Hungarian information focus and the Old Hungarian focus operator. Section 2.6 analyzes the evolution of a left-peripheral NegP, and the evolution of negative concord in Old Hungarian. Section 2.7 is a conclusion.

. The Proto-Hungarian sentence was SOV, with fused functional and thematic domains .. Evidence for head-final VP, TP, and CP projections The Hungarian sentence has been a head-initial verbal projection preceded by functional (focus, quantifier, and topic) positions throughout its documented history, beginning at around 1195. However, evidence of various kinds suggests that at the end of the 12th century, the head-initial structure was a relatively recent development, which took place either in the last centuries of the Proto-Hungarian period, before the settling of the Hungarian tribes in the Carpathian Basin in 896, or in the first centuries of the Old Hungarian period, perhaps under the influence of the language of the Slavic population found there. When reconstructing Proto-Hungarian syntactic phenomena, we rely on the generalization that language changes typically follow an ∫-curve, i.e., the appearence of a new construction proceeds slowly at first, then it speeds up, and eventually it slows down again, with remnants of the original state lingering on for a long time (see Kroch 1989; Croft 2000). The process of the disappearance of the variant supplanted

The evolution of functional left peripheries in the Hungarian sentence



by the new construction is the mirror image of this ∫; it follows a reverse ∫. Old Hungarian abounds in archaic constructions soon vanishing from the language, or losing their flexibility and surviving as linguistic fossils. We hypothesize that these vanishing patterns represent the last phases of reverse ∫-curves, and assume that by the backward extension of these reverse ∫-curves, we arrive at the prevailing constructions of Proto-Hungarian. The plausibility of our reconstructions is greatly enhanced if the reconstructed patterns can also be pointed out in present-day Khanty and Mansi, the Ob-Ugric sister languages of Hungarian. Old Hungarian archaisms, absent or sporadic in later phases of the language, but prevailing in Khanty and Mansi, include (i) SOV order with a morphologically unmarked object; (ii) V–auxiliary order; (iii) a clause-final complementizer; and (iv) V-final non-finite embedded clauses. These phenomena, to be discussed below, are manifestations of the head-finality of the VP, TP, and CP layers of the Proto-Hungarian clause structure. ... SOV order with a morphologically unmarked object The majority of Uralic languages have SOV word order, and, in general, a head-final phrase structure. SVO is mostly attested in the western branches of the family, in languages that have come into contact with Indo-European languages, e.g., in Finnish, Estonian, Karelian, and Vepse (Vilkuna 1998, 178). SVO is the dominant word order pattern in Komi and Mordvin, as well, however, according to the evidence of archaic folklore texts, it is a new development in these languages, supplanting the former SOV (cf. Rédei 1978; Saarinen 1991). Khanty and Mansi, the sister languages of Hungarian, are also strictly SOV (see Nikolaeva 1999, Riese 2001). The rigidity of their word order may be related to the fact that the object is mostly unmarked morphologically in them (except for personal pronouns in Khanty, and objects of a secondary topic role in some Mansi dialects), i.e., grammatical functions are identified on the basis of structural position. Compare the interpretations of (1a) and (1b): (1) a. o:xsar ńa:wre:m mu:smәlt-әs wound-pst.3sg fox child ‘The fox wounded a child.’

(Khanty)

b. ńa:wre:m o:xsar mu:smәlt-әs wound-pst.3sg child fox ‘The child wounded a fox.’

(Nikolaeva 1999: 42)

As will be discussed in section 2.2.2, Hungarian had developed morphological object marking prior to its documented period. In the finite clauses of the early documents, objects are consistently marked by the accusative suffix -t. In the nonfinite clauses of the earliest codices, Jókai Codex, a 15th-century copy of a codex written around 1370, and the Vienna, Munich, and Apor Codices, containing 15th-century copies of parts of a Bible translation from around 1416, however, we still find 131 non-



Katalin É. Kiss

finite clauses with a morphologically unmarked object (see Károly 1956). The word order of these non-finite clauses is strictly (S)OV. For example: ¯ (2) ne fordol’l’on mg [o kontos-o feluènni] ´ ´ ´ not turn-sbjv-3sg back he gown-poss.3sg-Ø on-put-inf ‘he should not turn back [to put on his gown]’ (Munich C. 30ra)1 (3) Kiral lèuèl-i irokat king letter-poss.pl-Ø write-part-pl-acc ‘those writing the king’s letters’

(Vienna C. 65)

meg ymmar [bewn-e zantnak] (4) Agyad sin-poss.3sg-Ø repent-part-dat give-imp-defo.2sg2 back now ‘give it back now [to that having repented his sin]’ (Jókai C. 158) (5) ky zent fferenczet lewlteuala [egyhaz seprette] who St Francis-acc found church-Ø sweeping ‘who found St Francis [sweeping the church]’

(Jókai C. 97)

(6) [ky haluan] legottan el mene what-Ø hearing immediately away went ‘[which having heard], he immediately left’

(Jókai C. 98)

(For the structural analysis of these non-finite clause types, see Chapter 5.) Notice that these caseless objects cannot be elements incorporated into the verb, as they are often definite, and they often display a complex structure. In (2), (3), and (4), for example, the caseless object is a possessive construction. These sentences are the last examples of a construction representing Ugric heritage. The fact that it survived longer in embedded contexts than in main clauses can have two explanations. (i) As will be discussed in section 2.2.2, object marking in Hungarian originally served the purpose of indicating the secondary topic role of the object. Topic marking is likely to have been restricted to main clauses, and the extension of object marking to embedded contexts may have been a relatively recent, not fully grammaticalized development in the 12th century. (ii) Lightfoot’s (1991) degree-0 learnability theory predicts that children identify/reanalyze the grammar of their mother tongue on the basis of root clauses. This theory also predicts embedded clauses to be more conservative than root clauses. Morphologically unmarked objects soon disappeared from the language. Compare the decreasing rate of unmarked objects in some early codices:

1 In Munich Codex and many other codices, the folios are numbered. The letter r abbreviates recto (‘front side’) and v abbreviates verso (‘back side’), whereas a and b mark the columns on a page. 2 defo stands for ‘definite object agreement’ on the verb, marking the presence of an object pro.

The evolution of functional left peripheries in the Hungarian sentence



(7) The number and rate of morphologically unmarked objects in subsequent codices codex year number of unmarked Os token/unmarked O tokens Jókai C. 1370 22,733 42 540 69,589 78 892 Munich C. 1416/14663 Apor C. 1416/a. 1500 22,118 18 1,382 Vienna C. 1416/1450 54,423 24 2,268 Jordánszky C. 1516–19 200,185 16 12,511 Both the Munich Codex and the Jordánszky Codex contain a translation of the New Testament. The unmarked objects of the Munich Codex mostly appear with an accusative suffix in the Jordánszky Codex. Crucially, the appearance of the accusative suffix also brings about an OV→VO shift in the verb phrase: (8) Matthew 4:20 a. azoc [legottan halo-io-c meghagu¯a] kouetec ´ they immediately net-poss-3pl-Ø leaving followed ‘Leaving their net immediately, they followed him’

otet ´ him (Munich C. 10rb)

b. Azok kedyg [legottan el hagywan halo-yo-k-at es they conj immediately off leaving net-poss-3pl-acc and hayo-yo-k-at] kóweteek hewtet boat-poss-3pl-acc followed him ‘Leaving their net immediately, they followed him’ (Jordánszky C. 364) Interestingly, a few fossilized OV structures with an unmarked object have survived until now: (9) a. szav-a tartó ember word-poss.3sg-Ø keeping man ‘a man keeping his word’ tévő legyek? b. Mi what-Ø doing be-sbjv.1sg ‘What shall I do’ c. hit-e hagyott ember faith-poss.3sg-Ø lost man ‘a man having lost his faith’ d. kalap levéve hat-Ø off-taking ‘taking off the hat [hat in hand]’ 3 The first date is the year in which, or shortly after which, the text of the codex was translated, and the second date is the year in which, or shortly after which, the codex was copied.



Katalin É. Kiss

In sum: SOV clauses with a morphologically unmarked object, representing a declining pattern restricted to non-finite clauses in early Old Hungarian, seem to have preserved the Proto-Ugric sentence pattern, which still prevails in the conservative sister languages of Hungarian. ... V–auxiliary order Old Hungarian had complex tenses involving a finite verb supplied with an aspectual morpheme and an agreement suffix, and a temporal auxiliary bearing a tense or mood suffix. This type of tense system seems to have evolved under Turkic influence; it only exists in the Finno-Ugric languages that have come into close contact with Turkic languages during their history (cf. É. Kiss 2013c). Hungarian tribes belonged to Turkic tribal alliances in the 7th–8th centuries, whereas the Udmurts, the Komi, the Mari, and the Mordvins have shared a habitat with Turkic peoples, e.g., the Chuvash, since the 8th century (cf. Bereczki 1983, Róna-Tas 2007, Róna-Tas, Berta, and Károly 2011). In the Old Hungarian complex verb forms, i.e., in past continuous, past perfect, and perfect conditional, the auxiliary bearing the tense or mood suffix obligatorily follows the lexical verb bearing an aspectual suffix and an agreement marker, e.g.: (10) a. Past continuous: én énekel-ek val-a I sing-1sg be-pst ‘I was singing’

mi énekel-ünk val-a we sing-1pl be-pst ‘we were singing’

b. Past perfect: én énekel-t-em val-a I sing-prf-1sg be-pst ‘I had sung’

mi énekel-t-ünk val-a we sing-prf-1pl be-pst ‘we had sung’

c. Perfect conditional: én énekel-t-em vol-na I sing-prf-1sg be-cond ‘I would have sung’

mi énekel-t-ünk vol-na we sing-prf-1pl be-cond ‘we would have sung’

The strict V–auxiliary order must be the relic of a clause structure involving a headfinal VP subsumed by a head-final TP: (11) [TP . . . [VP . . . O V] T] The V–auxiliary order disappeared from Hungarian with the disappearance of complex tenses in the Middle Hungarian period. By the end of the Old Hungarian period in the early 16th century, Hungarian had developed a system of situation aspect marking by means of resultative and terminative verbal particles, which made viewpoint aspect marking redundant. This led to the reinterpretation of the -t suffix of perfect aspect on the verb as a past tense marker, and the elimination of the tense marked temporal

The evolution of functional left peripheries in the Hungarian sentence



auxiliary. Compare the 16th-century and 20th-century translations of the same biblical sentence, Matthew 2:9: (12) a. És imé az czillag, mellyet lát-t-ac val-a napkeleten, and lo the star which-acc see-prf-3pl be-pst east-in elöttöc mégy-en val-a before-3pl go-3sg be-pst ‘And, lo, the star, which they had seen in the east, was going before them’ (Károli II/2: 3r) b. És íme, a csillag, amelyet napkeleten lát-t-ak, előttük and lo the star which-acc east-in see-pst-3pl before-3pl halad-t procede-past.3sg (Neovulgata) Modern Hungarian has seemingly retained the complex perfect conditional of Old Hungarian, reinterpreted as a past conditional. However, the perfective lexical verb and the auxiliary bearing the conditional suffix have been reanalyzed as a single complex head, e.g., in constructions involving verb movement, the V+auxiliary complex moves as a single unit. Compare the past conditional clause in (13a) with its variant involving verb movement: (13) a. Ha fel hív-t-ad vol-na a rendőrséget, . . . if up call-pst-2sg be-cond the police-acc ‘If you had called up the police, . . . ’ b. [Hív-t-ad vol-na]i fel ti a rendőrséget, . . . call-pst-2sg be-cond up the police-acc ‘Had you called up the police, . . . ’ Auxiliaries other than the obsolete vala, e.g., the future auxiliary fog, precede the lexical verb in the unmarked case in Old, Middle and Modern Hungarian alike, as expected in an SVO language:4 (14) hogy ehsegtewl sok emberek fognak meg halny that hunger-from many people will-3pl prt die-inf ‘that many people will die from hunger’

(Jókai C. 62–63)

4 An auxiliary cannot bear the main stress assigned to the first constituent of the comment, which may necessitate the movement of the lexical verb into pre-auxiliary position. This is what happens in (i). If the lexical verb has a particle, only the particle is preposed (ii).

(i) János ÉNEKELNI fog. John sing-inf will ‘John will sing.’ (ii) János EL fogja énekelni a dalt. John prt will-defo.3sg sing-inf the song-acc ‘John will sing the song.’



Katalin É. Kiss

In sum: the ‘lexical verb – temporal auxiliary’ order attested in Old Hungarian was the relic of a head-final TP subsuming a head-final VP. The V–auxiliary order disappeared with the disappearance of complex tenses. ... A clause-final complementizer The SOV Uralic languages have clause-final complementizers. Udmurt, Khanty, and Mansi share the same clause-final interrogative particle, which is a descendant of the Uralic negative auxiliary: (15) a. tit χujew-ä b. nèηem tˇottε ù.tot-á here sleep.1pl-q wife-poss.1sg there was-q ‘Do we sleep here?’ (Mansi) ‘Was my wife there?’ (Khanty; Juhász 1991: 501) c. Maša, [coje öt’iskomi-a], juaz. ˜ Masha her invite.1pl-q ask-pst-3sg ‘Masha asked if we would invite her.’

(Udmurt)5

The Hungarian interrogative particle is also cognate with the Khanty, Mansi, and Udmurt interrogative element, and, whereas it is cliticized to the verb preposed into the left periphery in Modern Hungarian, sporadically it still occurs in clause-final position in Old Hungarian: (16) a. n¯ede t˙u incab nagobbac vattoc azocnal e˙ qprt you rather bigger be-2pl those-adess q ‘Are you bigger than those?’ (Munich C. 12vb) mondatic marianac e˙ b. nemde o ańńa ´ qprt he mother-poss.3sg say-pass.3sg Mary-dat q ‘Is his mother called Mary?’ (Munich C. 20va) c. Nemde istennek alazatot lezen en lelkem e qprt God-dat subjugate-part be.3sg I soul-poss.1sg q ‘Will my soul be subjugated to God?’ (Apor C. 7) In a few cases, the interrogative particle is spelled out twice: clause-finally and in verbadjacent position: (17) Mínemde elfelethethí-e az ańa v kis qprt prt-forget-possib-3sg-q the mother she small g´ ermoket-e ´ child-poss.3sg-acc-q ‘Can the mother forget her small child?’

5

The Udmurt data were provided by Orsolya Tánczos.

(Nádor C. 26r)

The evolution of functional left peripheries in the Hungarian sentence



Interestingly, a clause-final -e—whether doubled by an -e on the verb or not— is always accompanied by the clause-initial negative interrogative discourse particle (mi)nemde. Yes–no questions framed by a clause-final -e and a clause-initial (mi)nemde seem to represent an intermediate stage in the shift from head-final to head-initial complementizers, perhaps involving two CP-layers: both a head-final one and a headinitial one. A similar situation has been attested in several modern Finno-Ugric languages heavily influenced by Russian. E.g., as shown by Tánczos (2013), Udmurt finite object clauses often end with the native šuisa, and begin with stoby borrowed from Russian, both corresponding to the complementizer ‘that’. In the great majority of Old Hungarian yes–no questions, -e is cliticized to the verb, which presumably occupies the head position of a functional projection; i.e., -e appears shifted to the left periphery. (In the case of a V-adjacent -e, the presence of the particle (mi)nemde is optional.) E.g.: (18) Te vag e˙ sidocnac kiral’a you are q Jews-dat king-poss ‘Are you the king of Jews?’

(Munich C. 82rb)

The occasional clause-final -e particle attested in Old Hungarian yes–no questions must be a relic of the Proto-Hungarian/Proto-Ugric/Proto-Finno-Ugric clause-final interrogative complementizer; it is evidence of a former head-final CP. ... V-final non-finite embedded clauses Conservative Uralic languages, among them Khanty and Mansi, employ few overt complementizers; they overwhelmingly use non-finite subordination, where the clause-final subordinator is the non-finite suffix on the verb. The comparison of map 81 with map 125 and 126 of The World Atlas of Language Structures (Haspelmath et al. 2005) shows a strong correlation between SOV structure and the preponderance of non-finite adverbial clauses. As discussed in detail in Chapter 5 of this book, Old Hungarian still displays a great variety of nonfinite subordinate clause types, but their proportion is gradually decreasing. These non-finite clauses mostly preserve the SOV order of Proto-Hungarian, although SVO order (perhaps derived by clause-internal right dislocation) also occurs. Observe two illustrative examples, a non-finite adverbial clause in (19a), and a nonfinite prehead relative clause in (20a), from the first Bible translation from 1416. The archaic nature of non-finite complementation in Old Hungarian is shown by the fact that in 16th-century Bible translations these sentences are already translated by finite clauses introduced by clause-initial complementizers. (19) a. [te kezedet meg nituad:] mendennek be you hand-poss.2sg-acc prt open-part-2sg all prt tell’esednek fulfil ‘You having opened your hand, all are filled in’ (Psalms 104: 28, Apor C. 68)



Katalin É. Kiss

Compare with the corresponding finite adverbial clause of the Károli Bible (1590): b. [Mikor az te kezedet meg nitod] ackor mindenec when the you hand-poss.2sg-acc prt open-2sg then all be telnec az te ió vóltoddal. prt fill the you good be-gerund-poss.2sg-with ‘When you open your hand, then all are filled with your goodness.’ (Károli I: 582vb–583ra) (20) a. ot vala eg [keze meg aźott] èmber there was a hand-poss.3sg prt wither-part man ‘There was a man whose hand had withered.’ (Munich C. 38ra) Compare with the corresponding finite relative clause of the Jordánszky Codex (1516–19): b. vala ot egy ember, [kynek hew kezee meg be-pst.3sg there a man who-dat he hand-poss.3sg prt azot vala] wither-prf.3sg be-pst ‘There was a man whose hand had withered.’ (Jordánszky C. 460) The fact that non-finite complementation is gradually losing ground can be demonstrated by the decreasing numbers of non-finite clauses in subsequent Bible translations. Compare, for example, the numbers of -ván/vén clauses in three subsequent translations of St Matthew’s Gospel: (21) The number of -ván/vén clauses in 15th–16th-century translations of Matthew: Munich Codex 1416/1448: 486 Jordánszky Codex 1516–1519: 322 Károli Bible 1590: 286 In the case of relative clauses, the increasing number of relative pronouns indicates the decrease of prehead non-finite relatives involving gap relativization. (22) The number of the relative pronouns ki ‘who’, mi ‘what’ and mely ‘which’ in 15th– 16th-century translations of Matthew: Munich Codex 1416/1448: 225 Jordánszky Codex 1516–1519: 314 Károli Bible 1590: 330 Despite their gradual decrease for more than 600 years, most types of non-finite constructions still have not completely disappeared from the language. However, the existing types have lost their flexibility; they have fossilized in various respects—as will be shown in detail in Chapter 5. Both clause types illustrated above have lost their

The evolution of functional left peripheries in the Hungarian sentence



ability to occur with pronominal subjects, and to combine with agreement suffixes. Participles derived by -ván/vén are still productive with PRO subjects, and, when derived from unaccusative verbs, they marginally allow a lexical subject, as well, e.g.: (23) a. [PRO Kinyitván az ajtót], Péter beengedett minket. open-part the door-acc Peter in-let-pst.3sg us ‘Opening the door, Peter let us in.’ b. [Mari otthon lévén], Péter beengedett minket. Mary at-home be-part Peter in-let-pst.3sg us ‘Mary being at home, Peter let us in.’ Prehead relative clauses can only occur productively with transitive verbs and a lexical subject, e.g.: (24) az [apám építette] ház the father-poss.1sg build-part-3sg house ‘the house which my father built’ In sum: non-finite complementation, characteristic of SOV languages, was widespread in Old Hungarian, but it has been losing ground to finite complementation ever since. Modern Hungarian remnants of non-finite complement clauses have lost their flexibility, e.g., their ability to take pronominal subjects and agreement suffixes. .. Evidence for the fusion of discourse roles and grammatical functions. The role of differential object–verb agreement In the sister languages of Hungarian (see Nikolaeva 2001, Skribnik 2001, Virtanen 2012), and—as far as we can judge on the basis of the available information—in other conservative Uralic languages, e.g., in Tundra Nenets (Dalrymple and Nikolaeva 2011) and Selkup (Kuznecova et al. 1982), as well—the grammatical functions are fused with discourse roles. The topic must coincide with the subject. If the external argument is unfit for the topic role, e.g., if it is a wh-phrase or a negative pronoun, and the internal argument is given, the sentence must be passivized. (25) a. (luw) juwan re:sk-әs he Ivan hit-pst.3sg ‘He hit Ivan.’ b. juwan xoj-na re:sk-әs-a Ivan who-loc hit-pst-pass.3sg ‘Who was Ivan hit by?’ (26) a. tam xu:j xoj-na an wa:n-s-a this man who-loc not see-pst-pass.3sg ‘This man was seen by nobody.’

(Nikolaeva 1999: 58)



Katalin É. Kiss b. ∗ xoj tam xu:j an wa:nt-әs-li who this man not see-pst-obj.3sg ‘Nobody saw this man.’

(Nikolaeva 2001: (28a,b))

Passivization, i.e., NP-movement to subject position, can also involve oblique internal arguments (see Kulonen 1989), e.g.: (27) nare:-l oxeәs-na xu:j-l-a bench-3sg sable-loc lie-prs-pass.3sg ‘His bench was covered with sables.’

(Nikolaeva 1999: 31)

In the SOV Uralic sentence, the object can have two alternative discourse roles: it can be focus, or it can be secondary topic. In the latter case, it elicits agreement on the verb. The secondary topic is a familiarity topic with a contextually given referent, capable of binding and controlling (see Nikolaeva 2001). An object topic is “secondary” because it never occurs without a primary subject topic. Observe the following Khanty sentence: (28) ma tam kalaη we:l-s-әm /we:l-s-e:m I this reindeer kill-past-1sg/kill-past-obj.1sg ‘I killed this reindeer.’

(Nikolaeva 1999: 64)

If this sentence answers the question ‘What happened?’, i.e., if the object conveys new information, it elicits no agreement, i.e., the verb form we:lsәm is used. If, on the other hand, (4) answers the question ‘What did you do to this reindeer?’, i.e., if the object is a familiar secondary topic, the verb form also incorporating an object agreement morpheme (we:lse:m) is employed. The subject functions as a primary, aboutness topic in both cases. The presence versus lack of an object agreement morpheme on the verb encodes the following discourse structures: (29) a. [Topic S O] b. [Topic S]

[Comment VOAgr, SAgr ] [Comment [Focus O] VSAgr ]

Differential object–verb agreement of this type is attested, in addition to Khanty, in Mansi (Skribnik 2001, Virtanen 2012), as well as in Tundra Nenets, Selkup, and Nganasan (Dalrymple and Nikolaeva 2011). Some Mansi dialects display differential object marking instead of differential object–verb agreement, i.e., the secondary topic role of the object is marked by a morpheme on the object. Earlier descriptions of these dialects by Steinitz (1950: 75) and Collinder (1960) assumed that morphological object marking occurs optionally in the case of definite objects. In fact, as Marcantonio (1985) suggested, and as Virtanen (2012) and Klumpp (2009) have shown, differential object marking is not optional; it is dependent on the information status of the object; it marks given, topical objects (which are, indeed, mostly definite).

The evolution of functional left peripheries in the Hungarian sentence



Hungarian displays differential object–verb agreement elicited by definite objects. However, there is evidence that the definite object–verb agreement attested in Hungarian is the result of function change; originally, it encoded the secondary topic role of the object in SOV sentences of type (29a). This claim is supported by two arguments. First, as shown by Peredy (2009), a specific indefinite object still elicits object–verb agreement for the majority of speakers if it is topicalized (as in (30a) and (31a)). In (30b) and (31b), where the specific indefinite object is in postverbal position, verb– object agreement is ungrammatical. (30) a. Bizonyos gyerekeket a társasjátékok leköt-i-k /?leköt-nek. certain kids-acc the board-games absorb-defo-3pl /absorb-3pl ‘Certain kids are absorbed by board-games.’ (Peredy 2009) cf.

b. A társasjátékok lekötnek/∗ lekötik bizonyos gyerekeket.

a sötét ruhák öregít-i-k (31) a. Egyes nőket certain women.acc the dark clothes make.look.old-defo-3pl /?öregíte-nek. /make.look.old-3pl ‘Certain women, dark clothes make look older.’ (Peredy 2009) cf.

b. A sötét ruhák öregítenek/∗ öregítik egyes nőket.

The second argument supporting the claim that Hungarian definite object–verb agreement is the descendant of the topical object–verb agreement attested by the SOV clause type in (29a) is more complex. In several SOV languages displaying differential object–verb agreement, among them the Uralic Tundra Nenets, Selkup, Nganasan (Dalrymple and Nikolaeva 2011) and Mansi (Bíró and Sipőcz 2013), and the non-Uralic Koryak, Kamchadal and Chukchi (Comrie 1980), object–verb agreement is blocked if the object is an inherently more “animate”, more “active”, more “topical” participant of the event than the subject. Apparently, a secondary topic cannot be more topical than the primary topic. An object that is inherently more topical than the primary topic can only figure as a focus. This constraint was first observed by Comrie (1980) in Chukchi, Koryak, and Kamchadal, and was formulated as follows: (32) Inverse Agreement Constraint An object agreeing with a verb must be lower in the animacy hierarchy than the subject agreeing with the same verb. (33) Animacy hierarchy 1SG > 1PL > 2SG > 2PL > 3SG > 3PL As shown by Comrie (1980), Chukchi, Koryak and Kamchadal have two strategies to avoid a violation of the Inverse Agreement Constraint. In case the object of a verb is



Katalin É. Kiss

more “animate” than its subject, the verb only agrees with its subject, but not with its object, or, alternatively, an inverse morpheme is prefixed to the verb to indicate that the Inverse Agreement Constraint is suspended. The languages adopting the animacy hierarchy under (33) collapse certain segments so that the resulting hierarchy should consist of three or two degrees. A three-degree segmentation can distinguish the “speaker participant”, the “non-speaker participants”, and the “non-participants”, as follows: (34) Animacy hierarchy 1SG (speaker participant) > 1PL, 2 (non-speaker participants) > 3 (nonparticipants) The case when the subject and the object both represent the same level of animacy are handled differently by languages observing the Inverse Agreement Constraint. A possible solution is to lift the Inverse Agreement Constraint in that case, i.e., to employ the following caveat: (35) The Inverse Agreement Constraint is suspended if both the object and the subject represent the lowest level of the animacy hierarchy. The Inverse Agreement Constraint in (33) interpreted on the animacy hierarchy in (34), supplemented with the caveat in (35), blocks object–verb agreement in case of the following subject–object combinations: (36) a. 3rd person subject – 1st/2nd person object b. 2nd person subject – 1st person object c. 1st person plural subject – 2nd person object In the very same cases, the Modern Hungarian verb also fails to agree with its definite object. Thus there is no object–V agreement if the subject is 3rd person and the object is 1st or 2nd person (37a), if the subject is 2nd person and the object is 1st person (37b), and if the subject is 1st person plural, and the object is 2nd person (37c): engem /minket /téged /titeket. (37) a. János lát-Ø /∗ lát-ja-Ø /us /you-acc /you-pl-acc John see-3sg /see-defo.3sg me ‘John sees me/us/you.’ engem /minket. b. Te lát-sz /∗ lát-od /us you see-2sg /see-defo.2sg me ‘You see me/us.’ téged /titeket. c. Mi lát-unk/∗ lát-j-uk we see-1pl/see-defo-1pl you-acc /you-pl-acc ‘We see you.’

The evolution of functional left peripheries in the Hungarian sentence



Third person pronouns, as opposed to first and second person pronouns, always elicit the definite conjugation: (38) János ∗ lát-Ø /lát-ja őt /őket John see-3sg /see-defo.3sg him /them ‘John sees him/them.’ As predicted by the Inverse Agreement Constraint, object–verb agreement is not blocked if the subject is 1st person singular and the object is a less “animate” 2nd person. The -l- morpheme agreeing with a 2nd person object is different from the morpheme agreeing with a 3rd person object. (39) Én lát-l-ak téged /titeket. I see-2-1sg you-acc /you-pl-acc ‘I see you.’ The blocking of object–verb agreement in (37a–c) must be a fossilized relic of the Inverse Agreement Constraint, which is a concomitant of the SOV sentence structure in (29a) (see É. Kiss 2005a, 2013d). Marcantonio (1985) claims that the type of differential object marking that is attested in certain Mansi dialects, and the type of differential object agreement that is attested in Khanty, and whose relics are attested in Hungarian, represent two subsequent stages of a diachronic process. According to her, differential object agreement evolved in the Proto-Ugric dialects which had generalized object marking, originally restricted to secondary topics, to all objects. Some linguistic fossils indicate that this might have happened in Hungarian, as well. Proto-Hungarian may originally have employed the suffix -t, the general accusative morpheme of Modern Hungarian, only to mark topical objects. Presumably it was after -t had been generalized to all objects that the secondary topic role of the object came to be marked by verbal agreement. The relative animacy of the primary and secondary topics must have been regulated by an equivalent of the Inverse Agreement Constraint in the period of differential object marking as well. Since the secondary topic, represented by the object, had a dependent, subordinate role with respect to the primary topic, it had to be less animate than the primary topic. An object more animate than the primary topic could only be construed as a focus. Since a first or second person object is inherently more animate, more topical, than a third person object, a first or second person object could not function as a secondary topic, i.e., in the period when the secondary object status was marked on the object by a -t suffix, a first or second person object received no -t. Fossilized relics of this constraint, too, have been preserved in Hungarian. First and second person singular object pronouns still receive no accusative case ending; they only bear a possessive morpheme, which is a typical means of marking definiteness in the Uralic languages (see Nikolaeva 2003):



Katalin É. Kiss

(40) én – I ‘I-nom’

en-g-em-Ø I-ep-1sg6 ‘I-acc’

te – you ‘yousg -nom’

té-g-ed-Ø you-ep-2sg ‘yousg -acc’

Non-standard varieties of Hungarian have already eliminated these exceptional forms: (41) én – I-nom ‘I’

en-g-em-et I-ep-1sg-acc ‘I-acc’

te – you-nom ‘yousg ’

té-g-ed-et you-ep-2sg-acc ‘yousg -acc’

As shown by Nikolaeva’s Khanty examples, an object whose possessor is bound by the subject counts as given, and can license object–verb agreement (Nikolaeva 2001). In other words, the relevant referential properties of the object can be determined by those of its possessor. It follows that in languages where a version of the Inverse Agreement Constraint is active, it is elicited not only by first and second person objects but also by objects with a first or second person possessor. In Modern Hungarian, objects with a first or second person singular possessor are still fully grammatical without an accusative ending, and objects with a first or second person plural possessor are also marginally acceptable: (42) Keresem a kalapom /kalapod /?kalapunk /?kalapotok. seek-defo.1sg the hat-poss.1sg /hat-poss.2sg /hat-poss.1pl /hat-poss-2pl ‘I am looking for my hat/your hat/our hat/your hat.’ In every other case, the omission of the accusative suffix of the object is strongly ungrammatical: (43)

∗∗ Keresem

a kalapja /a kalap. seek-defo.1sg the hat-poss.3sg /the hat ‘I am looking for his hat/the hat.’

The unmarked objects in (40) and (42) are also manifestations of the Inverse Constraint, blocking the construal of an inherently topical element (the speaker or the non-speaker participant, or a referent anchored to them) as a secondary topic. Presumably it was after -t had been generalized to all objects that the secondary topic role of the object came to be marked on the verb by a suffix agreeing with the object. By the end of the 12th century, the time of the first surviving coherent text, the Hungarian sentence had changed from SOV with fused functional and thematic domains to topic–focus–V–X∗ , and the topic function came to be encoded by movement into a designated left-peripheral position. Agreement between the primary topic and the verb grammaticalized as obligatory subject–verb agreement, whereas secondary topic–verb agreement grammaticalized as obligatory definite object–verb

6

EP stands for ‘epenthetic’.

The evolution of functional left peripheries in the Hungarian sentence



agreement. The Inverse Agreement Constraint fossilized as a gap in definite object– verb agreement in the case of “3rd person subject/1st or 2nd person object”, and “2nd person subject/1st person object” combinations. Later the evolution of a system of articles also made the new function of object–verb agreement, marking the definiteness of the object, redundant. Then object–verb agreement assumed a new role again: the licensing of object pro-drop. .. Interim summary We have reconstructed a head-final grammar for Proto-Hungarian on the basis of archaic head-final structures of Old Hungarian converging with corresponding constructions of the sister languages. These included (i) a strictly SOV non-finite sentence structure with a morphologically unmarked subject and object, (ii) the obligatory V– auxiliary order of the lexical verb and the temporal auxiliary, indicative of a head-final TP, (iii) the occasional head-final occurrence of the interrogative complementizer -e, a cognate of the clause-final interrogative complementizer of the sister languages, and (iv) the preponderance of non-finite subordination, a typical attribute of SOV clause structure. There is evidence that the Proto-Hungarian sentence displayed the same fusion of functional and thematic domains that is attested in present-day Khanty. Namely, the lack of accusative marking on first and second person singular objects, and on objects with a first or second person possessor, and the lack of verbal agreement with first and second person objects are relics of a former SOV stage of the language where the subject functioned as the primary topic, and the object functioned either as a focus or as a secondary topic. In this language type, surviving in various present-day Uralic languages, the secondary topic role of the object is marked by an accusative suffix or by an agreement morpheme on the verb. First and second person pronouns, representing the most active, most topical participants of the discourse, tend to be grammaticalized as primary topics; that is why their secondary topic marking is blocked.

. The shift from head-final to head-initial .. The first Old-Hungarian documents are already ‘topic– focus– V–X∗ ’ Whereas the evidence surveyed above shows that the Proto-Hungarian sentence was SOV with fused thematic and discourse roles, the first surviving coherent Hungarian text, the Funeral Sermon and Prayer, written (or copied) around 1195, displays the same sentence structure that is attested in Modern Hungarian, with head-initial projections and split thematic and functional domains. Half of the 50 clauses of the Funeral Sermon and Prayer begin with the verb. Though many of these clauses contain an invisible pro subject, which can be assumed to occupy a clause-initial topic position,



Katalin É. Kiss

overt subjects and objects indicate that the position of verbal complements with no special discourse roles is behind the verb. The postverbal arguments include the subjects of unergative, unaccusative, and passive verbs (44a–c). (44b) is a statement about the possessor of the subject; accordingly, only the possessor is topicalized, with the possessum left in situ. (44) a. Horoguvec isten raged God ‘God raged.’ b. Es [oz gimilsnec]i wl keseruv uola ti vize water-poss7 and that fruit-dat so bitter was ‘And of that fruit, the juice was so bitter’ c. kinec odut hotolm ovdonia. es ketnie who-dat give-pass.3sg power loose-inf-3sg and bind-inf-3sg ‘who was given power to bind and to loose’ The subject is preverbal only if it functions as a topic or a focus, or if it is a quantifier Q-raised into scope position. The subjects in (45a–b) are topics of embedded clauses, following a complementizer and a relative operator in Spec,CP, respectively: (45) a. hug isten ív uimadsagucmia bulsassa w that God they prayer-poss-3pl-because.of forgive-sbjv-3sg he bunet sin-poss.3sg-acc ‘that God should forgive his sin because of their prayer’ b. kit vr ez nopun ez homus vilag timnucebelevl mente whom Lord this day-on this false world prison-poss-from saved ‘whom the Lord has saved this day from the prison of this false world’ Focus subjects immediately precede the verb or the nominal predicate: (46) Kic ozvc. miv vogmuc who-pl those we be-1pl ‘Who are those? We are those.’ The focus role of the subject pronoun miv in (46) is shown by the facts that it answers an interrogative wh-phrase, and it is not pro-dropped. Objects with no special discourse role follow the verb: (47) isa es num igg ember mulchotia ez vermut indeed even not one man pass-possib-3sg this pit-acc ‘Indeed, no man can pass this pit.’ 7

poss stands for a morpheme marking possessedness.

The evolution of functional left peripheries in the Hungarian sentence



Preverbal objects function as relative operators—see kit ‘whom’ in (45b), or topics (48a–b), or foci (49). (48) a. hug turchucat mige zocoztia vola that throat-poss-3pl-acc prt rive-3sg be-pst ‘that it was riving their throat’ b. hug ur uvt kegilmehel abraam . . . kebeleben that Lord him grace-poss-3sg-with Abraham bosom-poss-in helhezie put-sbjv-3sg ‘that the Lord should put him in the bosom of Abraham with his grace’ (49) kinec ez nopun testet tumetívc who-dat this day body-poss-acc bury-1pl ‘whose BODY we bury today’ The preverbal section of the clause could also contain a negative particle. In (50), it immediately precedes the verb, in (51) it precedes a focused indefinite: meret nu eneyc (50) Ge mundoa neki8 but tell-pst.3sg dat-3sg why not eat-cond-3sg ‘but told him why he should not eat’ (51) isa es num igg ember mulchotia ez vermut indeed even not one man pass-possib-3sg this pit-acc ‘Indeed, no man can pass this pit.’ Universal quantifiers, too, surfaced in the preverbal section of the sentence. (52a) is a clear example of overt Q-raising from an embedded non-finite clause into the matrix functional domain: neki (52) a. Es [mend paradisumben uolov gimilcictul]i munda and all Paradise-in be-part fruits-from tell-pst-3sg dat-3sg elnie ti live-inf-3sg ‘And he told him to live on all fruits in Paradise’ b. mend w foianec halalut evec all he race-poss.3sg-dat death-acc ate-3sg ‘and he ate death for his whole race’ 8 In fact, neki, a postposition meaning ‘to, for’, supplied with an agreement suffix, has a pro complement, whose person and number can be reconstructed from the agreement suffix, i.e.:

(i) (ő)-nek-i he-dat-sg



Katalin É. Kiss c. isa mend ozchuz iarov vogmuc indeed all that-to go-part be-1pl ‘indeed, we are all approaching it’

In Modern Hungarian, a non-referential complement, for example, a verbal particle or a predicative bare nominal appears immediately before the verb. It is an open question of Hungarian synchronic grammar whether the preverbal predicative element (often referred to as ‘verb modifier’) and the preverbal focus are alternative fillers of the same focus slot, or they are specifiers of two separate projections, a predicative projection (PredP) and a FocP. Some approaches (Brassai 1860–65, É. Kiss 1987, 1992, 2006b, Surányi 2011) assume that these two types of preverbal constituents are alternative fillers of the same preverbal focus slot. In this framework, the different semantics of foci and verb modifiers, e.g., the exhaustivity of the former, is derived from their different referential properties. Other studies, e.g., Brody (1990b), É. Kiss (2002, 2008a), Szabolcsi (1994a), Surányi (2002) assume that the positions of predicative and exhaustive preverbal constituents are distinct; the verb modifier occupies the specifier of a projection immediately subsuming vP (labelled as PredP, TP, or AspP), whereas the exhaustive focus occupies the specifier of a higher FocP, and its immediately preverbal position is due to verb raising. This book tentatively adopts the latter view. Non-referential, predicative elements land in the specifier of PredP in the Funeral Sermon and Prayer, as well—see the verbal particle in (53a), and the predicative NP in (53b). The preverbal non-specific object in (53c) may also be in Spec,PredP, where it forms a complex predicate with the verb semantically. vize hug (53) a. Es [oz gimilsnec]i wl keseruv uola ti and that fruit-dat so bitter be-pst.3sg water-poss that turchucat mige zocoztia vola throat-poss.3pl-acc prt rive-3sg be-pst ‘and of that fruit, the juice was so bitter that it was riving their throat’ b. ysa pur es chomuv uogmuc. indeed dust and ash be-1pl ‘indeed, we are dust and ash’ c. es oz gimilsben halalut evec and that fruit-in death-acc eat-pst-3sg ‘and he ate death in that fruit’ The claim that verb modifiers and foci, apparently sharing the same preverbal slot, occupy different structural positions is based on the assumption that the focus elicits verb raising across the verb modifier. The focus constructions of the Funeral Sermon and Prayer contain no verb modifier, i.e., if there was V-to-Foc movement at that time,

The evolution of functional left peripheries in the Hungarian sentence



it was vacuous movement. What can be shown is that verb movement existed; it is attested in imperative clauses. In (54a), the imperative verb precedes the verb modifier, whereas the indicative verb follows it in (53a–c). Similarly, in (54b), the imperative verb precedes the quantified expression, whereas the indicative verb follows it in (52a–c). (54) a. legenec neki seged uromc scine eleut be-imp-3pl him helper Lord-poss-1pl presence-poss before ‘and be his helpers in the presence of our Lord’ b. es bulsassa mend w bunet and forgive-imp-3sg all he sin-poss.3sg-acc ‘and he should forgive all his sins’ Like in Modern Hungarian, the order of postverbal arguments in the Funeral Sermon and Prayer is free, except that light, short, unstressed elements precede longer, heavier constituents, e.g.: (55) a. es odutta vola neki paradisumut hazóá and give-prf-3sg be-pst dat-3sg Paradise-acc house-poss.3sg-for ‘and he had given him Paradise for his house’ b. Heon tilutoa wt ig fa gimilcetvl only bar-pst-3sg he-acc one tree fruit-poss-from ‘He only barred him from the fruit of one tree.’ c. Es zoboducha wt urdung ildetuitvl and deliver-imp.3sg he-acc devil harassment-poss-pl-from ‘and he should deliver him from the devil’s harassments’ In sum: the structure of sentences in the first Old Hungarian text, the Funeral Sermon and Prayer, is essentially identical with the structure of the Modern Hungarian sentence. The preverbal section of the sentence is a functional domain including a CP, an iterable TopP, and a FocP projection (56a). The functional domain also contains a predicative projection harboring non-referential, predicative complements forming a complex predicate with the verb, and it can provide a landing site for overt quantifier raising (56b). (56) a. [CP hug [TopP ur [TopP uvt [TopP kegilmehel [FocP abraam . . . Lord him grace-his-with Abraham that kebeleben helhezie]]]]] bosom-poss-in put-sbjv-3sg ‘that the Lord should put him with his grace in Abraham’s bosom’ b. [PredP mend [PredP ozchuz iarov vogmuc]] all that-to going be-1pl ‘we are all approaching that’



Katalin É. Kiss

The negative particle can be inserted below FocP, as in (50), and above FocP, as in (51). We have also found evidence of verb raising. The base position of arguments is in the postverbal domain. .. How did the reanalysis take place? As argued in section 2.2, Proto-Hungarian must have inherited from Proto-Ugric an SOV sentence structure with fused thematic and functional domains, and it must have preserved it for a long time. At the same time, the first coherent written record of Old Hungarian from the late 12th century is already the manifestation of a head-initial language similar to Modern Hungarian, displaying a head-initial verb phrase preceded by left-peripheral functional projections. That is, the following basic change must have taken place in the late Proto-Hungarian or early Old Hungarian period, prior to the end of the 12th century: (57) [Top S] [Foc O] V ‡ Top Foc V X∗ The question arises how this change happened; what kind of sentence structure generated by the SOV grammar inherited from Proto-Ugric could be analyzed by a new generation of speakers as the output of a head-initial grammar generating independent lexical and functional projections. In the head-final sister languages of Hungarian, we occasionally attest an ‘SOVX∗ ’ order. According to Nikolaeva (1999), it is an ‘afterthought’ construction, an instance of right dislocation. Nowadays, it is becoming more and more common in the ‘Russified’ speech of the younger, Russian-dominant bilingual speakers. In earlier texts not affected by Russian, it was typically used to explicate pro arguments crossreferenced by agreement suffixes, and implicit adverbials. E.g.: (58) a. luw simәl ru:pitl u:lәpsi xośa to he little works life ‘He earns little, for life.’

(Khanty)

b. pa su:sm-ә-s joxәś xo:t-әl u:lә-m taxa pela itta again walk-pst-3sg back house-poss.3sg be-pastpart place to that maxim Maxim ‘Again he walked back to the place where his house was, that Maxim.’ (Nikolaeva 1999, 57) Since such postverbal constituents usually spell out implicit pronominal or proadverbial elements of the clause, they have a familiar referent, i.e., they are specific noun phrases or postpositional phrases functioning as familiarity topics. The right dislocation of familiar elements is common in other SOV Uralic languages, e.g., in Udmurt, as well (Surányi and Tánczos 2011), and it is also attested in

The evolution of functional left peripheries in the Hungarian sentence



non-Uralic SOV languages, e.g., in German and Korean. It presumably occurred in the SOV Proto-Hungarian, too. It must have been facilitated by the generalization of the -t topical object marker to all objects, which enabled Proto-Hungarian to encode grammatical functions consistently by morphological means. Slavic influence upon early Old Hungarian may also have played a role in its spreading. The language of the Old Church Slavonic documents of that period is halfway between SOV and SVO (cf. Pancheva 2008), but the vernacular language of the Slavic population in Central Europe could already be SVO. In late Proto-Hungarian or early Old Hungarian, the spreading of right dislocation may have reached a rate at which the postverbal arguments came to be analyzed by a new generation of speakers as arguments left in argument positions, and the preverbal elements came to be interpreted as arguments raised into functional positions. The dual functions associated with the preverbal slots in the Ugric proto-language split: the thematic domain was identified with the postverbal section of the sentence, and the preverbal section was reanalyzed as a functional left periphery. The subject-topic position was reinterpreted as a topic position accessible to topical constituents of any grammatical function. The object-focus position was reinterpreted as a pure focus position. That is, the SOV sentence structure inherited from the Ugric proto-language evolved into a ‘topic–focus–V–X*’ structure, as follows: (59)

Proto-Hungarian: subject-topic object-focus right dislocated elements ↓↓↓ SOV

→ → → →

Old Hungarian: topic focus arguments in situ ↓↓↓ (Topic) (Focus) V X∗

As shown by Nikolaeva (1999: 35, 61–63), in the Khanty sentence, the object-focus can be separated from the verb by a reduced adverbial or nominal complement forming a lexical unit with the verb. This kind of reduced complement may have been the Proto-Ugric/Proto-Hungarian antecedent of the Old Hungarian verb modifier. The lexical node dominating the reduced complement and the verb has been reanalyzed as a functional projection, where the reduced complement and the verb are moved for predicate incorporation to be possible (cf. Zwart 1993, Koster 1994): (60) S [VP . . . [V VM V]] → Topic [PredP VM V [VP . . . tV . . .tVM ]] If we assume that the projection harboring the verb modifier is other than the focus projection, then the focus construction involves a further step of verb movement: (61) [Topic [Focus V [VM tV [VP tV . . . ]]]] The assumption that the SOV → SVO shift happened via the reanalysis of right dislocated arguments as arguments base-generated postverbally explains two properties



Katalin É. Kiss

of Hungarian which are inexplicable from a synchronic point of view. Postverbal word order is free both in Old Hungarian and in Modern Hungarian. The optimal word order is phonologically constrained, with short, unstressed constituents ordered before longer, stressed elements—in accordance with Behaghel’s Law of Growing Constituents (Behaghel 1932). This property of the Hungarian sentence must have grammaticalized when postverbal elements were still analyzed as constituents in right dislocation, adjoined to the VP or to the clausal projection in an arbitrary order. A further curious property of Modern Hungarian is that non-referential elements must be located in the preverbal section of the sentence (cf. Alberti 1997). This property, too, is a relic of the period when the postverbal position of a constituent was the result of right dislocation, which could only target referential constituents. The view that the shift from SOV to SVO was initiated by the spreading of right dislocation has also been raised in connection with other languages, among them members of the Niger-Congo family (Hyman 1975), and Latin (Polo 2005). .. Interim summary; further questions This section has argued that the Uralic/Proto-Hungarian SOV sentence structure, with fused thematic and functional domains, had changed into the Modern Hungarian ‘topic–focus–V–X*’ sentence structure prior to the end of the 12th century. The change may have been initiated by the spreading of right dislocation. The growing number of postverbal constituents led to the reanalysis of postverbal arguments as arguments in situ, and the reanalysis of preverbal elements as elements preposed into the specifier positions of left-peripheral functional projections. The 50 clauses of the first Old Hungarian document already utilize every functional projection attested in Modern Hungarian. At the same time, the functional left periphery of the sentence was exploited to a lesser degree in Old Hungarian than it is in Modern Hungarian. The differences are mainly related to the PredP, FocP, and NegP projections. The changes in the use of these projections will be surveyed in sections 2.4, 2.5, and 2.6, respectively.

. The evolution of a left-peripheral PredP .. The PredP in Modern Hungarian As mentioned in section 2.3.1., the Modern Hungarian sentence contains a preverbal structural position for predicative complements, the so-called verb modifiers, including verbal particles accompanying telic (accomplishment and achievement) verbs (62a,b), bare nominal complements (63a,b), nominal predicates accompanying the copula (64a,b), and small clause predicates of various kinds (65a,b). (62) a. János el-olvasta a könyvet. John prt-read-pst-3sg the book-acc ‘John read the book.’

The evolution of functional left peripheries in the Hungarian sentence



b. János be-tette a pénzt a zsebébe. John in-put-pst-3sg the money-acc the pocket-poss.3sg-in ‘John put the money in his pocket.’ (63) a. Éva könyvet olvas. Eve book-acc reads ‘Eve is reading a book/books.’ b. Éva zsebre tette a pénzt. Eve pocket-in put-pst-3sg the money-acc ‘Eve pocketed the money.’ (64) a. Péter középiskolai tanár lesz Peter high.school teacher becomes ‘Peter will become a high school teacher.’ b. Boldog vagyok. happy be-1sg ‘I am happy.’ (65) a. Éva orvosnak tanul. Eve doctor-dat studies ‘Eve studies to be a doctor.’ b. Jánost okosnak tartják. John smart-dat hold-3pl ‘They hold John to be smart.’ c. Zsuzsa fél cm vékonyra nyújtotta a tésztát. Susan half cm thin-sub roll-pst-3sg the pastry-acc ‘Susan rolled the pastry half a centimeter thin.’ A predicative element of this type and the verb adjacent to it are assumed to be in specifier–head relation in a functional projection, identified in this book as PredP.9 Though in examples (62)–(65) there is a selectional relation between the verb and the nonverbal predicate, this is not necessarily so; the Spec,PredP position preceding a modal or temporal light verb is filled with a predicative element selected by the lexical verb: (66) a. János orvosi akar ti lenni. John doctor wants be-inf ‘John wants to become a doctor.’ kell, hogy ti menjen. b. János eli go-sbjv-3sg John away needs that ‘John needs to leave.’ 9 The term PredP is adopted from Zwart () and Koster (), In Hungarian syntax, the projection hosting the verbal particle and other nonverbal predicates has also been identified as AspP (see, e.g., É. Kiss ), and as TP (see Pinón , Surányi ).



Katalin É. Kiss

In imperative clauses (67a), in the presence of a focus (67b,c), and in the case of PredP negation (76d), the verb is preposed across the verb modifier: (67) a. Olvasd el a könyvet! read-imp-2sg prt the book-acc ‘Read the book!’ b. János melyik könyvet olvasta el? John which book-acc read-pst-3sg prt ‘Which book did John read?’ c. János CHOMSKY KÖNYVÉT olvasta el. John Chomsky book-poss-acc read-pst-3sg prt ‘It was Chomsky’s book that John read.’ d. János nem olvasta el a könyvet. John not read-pst-3sg prt the book-acc ‘John didn’t read the book.’ If the left periphery of the clause contains more than one functional projection eliciting verb movement, the verb does not move cyclically; it is stranded in the first functional head. (Some analyses, e.g., Olsvay (2006), capture this fact by analyzing this head as the head of a separate functional projection called Non-Neutral Phrase (NNP).) E.g.: [NegP nem (68) János [NegP nem [FocP CHOMSKY KÖNYVÉT not Chomsky book-poss-acc not John olvasta el tV ]]] read-pst-3sg prt ‘It wasn’t Chomsky’s book that John didn’t read.’ .. Sporadic use of PredP in early Old Hungarian As argued in section 2.3.1, the predicative position was already available in Old Hungarian; PredP may be the descendant of the lexical node dominating the ProtoUgric/Proto-Hungarian complex predicate. However, whereas in Modern Hungarian, Spec,PredP is filled in the great majority of clauses (e.g., the Spec,PredP of accomplishment and achievement clauses is always occupied by a resultative or terminative element, mostly a verbal particle), in the first Old Hungarian documents, Spec,PredP is only used sporadically. The fact that the proportion of verb-initial clauses is much larger in early Old Hungarian than in Modern Hungarian (cf. É. Kiss 2013b: 219) is mostly due to this difference. The Khanty descendant of the Proto-Ugric complex predicate is a verbal particle + verb complex, e.g., lap re:sk- ‘down hit = kill’, nox ka:wәr ‘up cook = cook (something) ready’, or a light/idiomatic nominal + verb complex, e.g., nomәs-na nu:mәs- ‘mind-

The evolution of functional left peripheries in the Hungarian sentence



in think’ (Nikolaeva 1999: 35, 62). The nominal predicate and the copula also form a clause-final complex of a similar kind (Nikolaeva 1999, 40). The sporadic preverbal predicative elements found in the first Old Hungarian texts are also of these types; they are verbal particles (69a), non-specific idiomatic nominal complements (69b), and nominal or adjectival predicates accompanying the copula (70a,b). (69) a. qui . . . angolucnoc corat hamar ele who angel-pl-dat chorus-poss-acc soon by mulhassa. . . nazaret varasat mege pass-possib-sbjv-3sg Nazaret city-poss-acc prt lelhesse find-possib-sbjv-3sg ‘who could pass by the angels’ chorus and could find the city of Nazaret’ (Königsberg Fragment) b. Scemem kunuel arad eye-poss.1sg tear-with floods ‘My eyes are flooding with tears.’

(Old Hungarian Lamentations of Mary)

(70) a. ysa pur es chomuv uogmuc indeed dust and ash be-1pl ‘Indeed, we are dust and ash.’

(Funeral Sermon and Prayer)

b. scuzsegnec tukere tistan maradhassun virginity-dat mirror-poss clear stay-possib-sbjv-3sg ‘virginity’s mirror could stay clear’ (Königsberg Fragment) .. The spreading of verbal particles The gradual increase of elements occupying Spec,PredP is partly due to the gradual spreading of telicizing verbal particles. Old Hungarian inherited from ProtoHungarian a complex tense system marking viewpoint aspect, where telicizing particles played no role yet except for cases involving the simultaneous presence of telicity and imperfectivity such as repeated telic actions (71a) and miscarried telic actions (71b) (see Gerőcs 2011: 165–167). (71) a. kikèt akar-uala meg-oluala kikèt ´ who-pl-acc want.3sg-be-pst prt-kill.3sg-be-pst who-pl-acc akaruala meg-uèruala es kikèt akaruala want.3sg-be-pst prt-beat.3sg-be-pst and who-pl-acc want.3sg-be-pst fel-magaztatuala prt-exalt.3sg-be-pst ‘whom he would he slew; whom he would he put down; and whom he would he set up’ (Vienna C. 143)



Katalin É. Kiss ¯ foit’auala b. Ki meg-foguan mg otet monduan ´ who prt-grab-part prt strangle-3sg-be-pst him saying Ad meg miuèl tartozol give-imp-2sg back what-ins owe-2sg ‘who, having grabbed, was throttling him, saying, Pay me that thou owest.’ (Munich C. 24va)

In the course of the Old Hungarian period, however, telicizing particles started spreading. Whereas in the first Old Hungarian documents, verbal particles were used to encode telicity only when the boundedness of the event was not marked otherwise, later they gradually became more or less obligatory attributes of telic predicates, which were also present if the viewpoint aspect was marked as perfective (72a), and if the verb was inherently telic (72b). (72) a. Gohetec fel a derebeket hogy el ne veszjenek collect-imp-2pl up the pieces-acc that prt not get.lost-sbjv-3pl mellec mg marattacuala which prt remain-prf-3pl-be-pst ‘Gather up the fragments that had remained, so that nothing be lost.’ (Munich C. 90rb) b. a· torueń zèrent kël meg-halni ´ the law according.to must prt die-inf ‘by the law one ought to die’

(Munich C. 105va)

The spreading of verbal particles is illustrated in a spectacular way by the socalled Munich Language Record, a Hungarian note written on an empty folio of a codex containing linguistic texts collected by Johannes von Grafing, a monk interested in languages and linguistics, in the Benedictine monastery of Ebersberg sometime between 1495 and 1527. The Hungarian text includes (among others) two versions of the Lord’s prayer: a version copied from a written Hungarian source, and a version recording in German orthography the oral prayer as the scribe heard it. The copy of the Hungarian written prayer does not contain a single verbal particle. In the recorded oral version, on the other hand, there are already three particle verbs: (73) ës meg-bozässät mi vëtkenkët. mikëpen ës mi and prt-forgive.imp.2sg we sin-poss-1pl-acc as also we mag-boczätunk vëtëtëknek . . . de säbädicz-mk mikët a gonostwl prt-forgive-1pl sinners-dat but free-imp.2sg-prt us the evil-from ‘and forgive us our debts, as we also forgive our debtors. . . but deliver us from evil’ (Haader 2005) In Károli’s translation from 1590, there are four more verbal particles, i.e., we find the same array of verbal particles that are used today:

The evolution of functional left peripheries in the Hungarian sentence



(74) . . . szenteltesséc meg a’ te neued. Iöijön el hallow-pass-imp-3sg prt the you name-poss.2sg come-imp-3sg prt az te országod: Légyen meg a’ te akaratod. . . Az the you country-poss.2sg be-imp-3sg prt the you will-poss.2sg the mi minden napi kenyerünket add meg nékünc ma. we every daily bread-poss.1pl-acc give-imp-2sg prt us today ‘Hallowed be your name. Your kingdom come, your will be done . . . Give us this day our daily bread’ (Károli II/2, 6r) The spreading of verbal particles to all telic verbs may have been initiated by Slavic influence. The Hungarian tribes settling in the Carpathian basin in 896 found and absorbed a Slavic population there, whose language must have been the source of considerable substrate effects. The gradual spreading of verbal particles led to the attrition of complex tenses encoding viewpoint aspect. The past continuous and past perfect tenses disappeared in the course of the Middle Hungarian period (1525–1772), and the present perfect, marked by a -t suffix, was reinterpreted as a past tense, supplanting the -a/e-marked past tense inherited from Proto-Ugric (see É. Kiss 2006a). Eventually, grammaticalized viewpoint aspect marking gave way to grammaticalized situation aspect marking. In the new system, telicity is marked by the presence of a terminative or resultative element, in most cases a verbal particle, in Spec,PredP. .. Predicative nominals in Spec,PredP In the earliest Old Hungarian documents, Spec,PredP harbored two types of [+N] constituents: nominal and adjectival predicates accompanying the copula, and nonreferential, non-specific, often idiomatic nominals semantically incorporated into the verb, as illustrated in (69b) and (70a–b). These types of nominals were not possible targets of right dislocation in the SOV grammar of Proto-Hungarian. The constraint that barred non-referential nominals from the postverbal section of the sentence grammaticalized, and has remained part of Hungarian syntax up till now (see Alberti 1997). Non-referential nominals confined to preverbal position came to be analyzed as the specifiers of a PredP projection, the locus of predicate incorporation. “Predicate incorporation” is a notion developed in semantics by de Hoop (1992), adopted in syntax by Koster (1994). De Hoop’s theory is based on the distinction between weak and strong NPs proposed by Milsark (1974). According to de Hoop, strong NPs are generalized quantifiers, which can be referential and appear as full arguments. Weak (e.g., non-specific, non-generic) NPs, on the other hand, must often be interpreted as parts of the predicate. Predicate incorporation gained special significance in Hungarian when morphological viewpoint aspect marking was supplanted by situation aspect marking, i.e., when aspect came to be determined by the inner structure of events. Change-of-state



Katalin É. Kiss

predicates could be bounded by incorporating a resultative expression in Spec,PredP, and movement predicates could be bounded by incorporating a terminative expression in Spec,PredP. In these predicate complexes, the resultative/terminative constituent does not need to be a verbal particle or a non-specific nominal; it can also be a definite noun phrase or a postpositional phrase with a DP complement. As resultative and terminative phrases can be referential, the postverbal section of the clause is also accessible to them. In the earliest texts, referential resultative and terminative phrases are still located in the postverbal section of the sentence (75), then in the codices we attest variation in their position (76a–b, 77a–b). Since the Middle Hungarian period, they must be moved to Spec,PredP. (75) Horoguvec isten. es veteve wt ez muncas vilagbele raged God and threw him this laborious world-into ‘God raged and threw him in this laborious world.’ (Funeral Sermon and Prayer) (76) a. Vr papa ment uolna zent Claranak monostoraba Lord Pope go-prf.3sg be-cond Saint Clara-dat monastery-poss-to ‘The Lord Pope would go to Saint Clara’s monastery’ (Jókai C. 42) b. Es hogy a cellaba ment uolna and as the cell-into go-prf.3sg be-cond ‘as he would go into the cell’ (77) a. az gonozokath vety pokolrah and wicked-pl-acc throws hell-to ‘and he throws the wicked into Hell’

(Jókai C. 75)

(Könyvecse 7v)

´ vetteth b. es az twzreh and the fire-into throw-caus.3sg ‘and he has you thrown into the fire’

(Könyvecse 5v)

Verbs expressing existence and spatial configuration form atelic complex predicates with location expressions in Spec,PredP. Originally, referential locatives appeared postverbally. This order is attested in the earliest version of the Lord’s Prayer (and this order has fossilized in the Protestant version of the prayer): (78) Mi at’¯ac ki vag meńńècbèn our father who are heaven-in ‘Our Father, which art in Heaven’

(Matthew 6:9, Munich C. 12rb)

In the early codices, the referential locative can appear either postverbally or in Spec, PredP (see (79–80)), but eventually the latter variant prevails. (79) a. mykoron azon frater ewlne az aztalnal when that frater sit-cond.3sg the table-at ‘when that frater would sit at the table’

(Jókai C. 66)

The evolution of functional left peripheries in the Hungarian sentence b. mykoron azon frater az aztalnal ewlne when that frater the table-at sit-cond.3sg ‘when that frater would sit at the table’



(Jókai C. 67)

(80) a. És vala ott Heródes megholtiglan and was there Herod prt-die-gerund.poss-until ‘And he was there until Herod died’ (Matt 2:15 in Munich C. 9ra) b. es ot len herodesnek halalayglan and there was Herod-dat death-poss-until ‘and he was there until Herod’s death’ (Matt 2:15 in Jordánszky C. 359) .. Summary The preverbal position identified as Spec,PredP originally harbored non-specific, non-referential complements unfit for right dislocation: verbal particles, nominal predicates accompanying the copula, and “light”, idiomatic nominal complements. In the earliest Old Hungarian documents, Spec,PredP was filled sporadically. By the end of the Old Hungarian period, the resultative/terminative verbal particle had become a regular constituent of accomplishment and achievement sentences, and the consistent marking of telicity rendered the morphological marking of viewpoint aspect redundant. Aspect was to be derived from the internal structure of events, and the PredP projection became the locus of combining the predicates representing subevents into a complex predicate.

. The evolution of a left-peripheral FocP .. The Ugric/Proto-Hungarian focus in situ If Proto-Hungarian inherited from Proto-Ugric the sentence structure that has been preserved in Khanty and Mansi, then its preverbal constituent functioned simultaneously as an object/internal argument and a focus. An object assigned a secondary topic role, eliciting agreement on the verb, was scrambled to a position external to the VP. In the more than 1000 Khanty sentences analyzed by Nikolaeva (2001), the non-agreeing focus object is always left-adjacent to the verb, whereas the agreeing secondary topic object can be separated from the verb by adjuncts. In Khanty sentences containing two internal arguments, either one can occupy the immediately preverbal focus position, the locus of pitch accent assignment (Filchenko 2011), and the [+focus] feature associated with it can be projected to the whole VP— see (81). The extraction of one of the two internal arguments into the position of a secondary topic results in the narrow focus interpretation of the preverbal argument— see (81b) and (81c), cited from Dalrymple and Nikolaeva (2011: 174). The topic position of the theme in (81b) and the goal in (81c) is shown by the fact that they elicit an object agreement morpheme on the verb.



Katalin É. Kiss

(81) a. What did you do ´ ma a:n Pe:tra e:lti ma-s-әm I cup Peter to give-pst-1sg ‘I gave the cup to Peter.’ b. What did you do to the cup? ma a:n Pe:tra e:lti ma-s-e:m I cup Peter to give-pst-obj.1sg ‘I gave the cup to PETER.’ c. What did you do to Peter? ma Pe:tra a:n-na ma-s-e:m /∗ ma-s-әm10 I Peter cup-loc give-pst-obj.1sg /give-pst-1sg ‘I presented Peter with a CUP.’ The same correlation between the immediately preverbal internal argument position and the focus role has been reported from various dialects of Mansi (Skribnik 2001, Virtanen 2012), and from the more distantly related Udmurt (Surányi and Tánczos 2011). In Khanty, the external argument can only be focused by the passivization of the sentence: (82) a. juwan xoj-na re:sk-әs-a? John who-loc hit-pst-pass.3sg ‘Who hit John? [lit. By whom was John hit?]’ cf.

/re:sk-әs-li? b. ∗ xoj juwan re:sk-әs who Ivan hit-pst.3sg /hit-pst-obj.3sg ‘Who hit John?’

(Nikolaeva 1999: 58)

Khanty interrogative wh-arguments always end up in the preverbal focus position, however, wh-adjuncts can remain external to the VP. The equivalent of how can be a VP-adjunct (83a) or a preverbal focus (83b): (83) a. Nawrem mujsirn askulaja manl? child how school-to goes ‘How does the child go to school?’ b. Nawrem askulaja mujsirn manl? The equivalents of ‘when’ and ‘where’, which can have either VP-scope or sentential scope, have three potential positions: they can be adjoined to the clause (84a) or to the VP (84b), or they can be a verb-adjacent focus (84c). 10 Notice that the goal argument turned into secondary topic, eliciting agreement on the verb, is presented as a morphologically unmarked object, whereas the theme assumes a locative case.

The evolution of functional left peripheries in the Hungarian sentence



(84) a. Xun iwan u:n xul wel-s? when Ivan big fish catch-pst.3sg ‘When did Ivan catch a big fish?’ b. Iwan xun u:n xul wel-s? c. Iwan u:n xul xun wels? These facts suggest that in Khanty, the preverbal focus position is not an A-bar position. It is the position where pitch accent is assigned, hence it is the default position of information focus. A non-preverbal internal argument is focused by the removal of the constituent that separates it from the verb. A wh-adjunct can appear preverbally, or it can assume a [+focus] feature in situ, presumably by stress shift. The D-structure subject can only be focused if the sentence is passivized. .. The Old Hungarian focus operator Assuming that Proto-Hungarian inherited from Proto-Ugric a Khanty-type sentence structure where the preverbal internal argument functioned as an in situ information focus, the reanalysis of right-dislocated elements as base-generated arguments brought about the reanalysis of the preverbal position as an A-bar position, the landing site of a focus operator. The pitch accent on the preverbal constituent, preserved up till now, must have remained unchanged. What has changed is that (i) the preverbal position has become the landing site of all types of focused constituents, including the external argument; (ii) movement to the preverbal position has been accompanied by verb movement; and (iii) the preverbal constituent has assumed a [+exhaustive], or [+maximal] feature; it expresses exhaustive identification. The preverbal focus position of the Old Hungarian sentence is open to internal arguments, external arguments, and adjuncts alike, as shown by the following set of wh-foci: Subject foci: (85) a. KIK ozuc. MIV vogmuc. who-pl those we be-1pl ‘Who are those? We are those.’

(Funeral Sermon and Prayer)

b. de QUI legen neky atia ozut nem but who be-sbjv-3sg dat-3sg father-poss.3sg that-acc not tudiuc know-1pl ‘But who should be his father, we do not know.’ (Königsberg Fragment) c. KY palantalta ez zerzetett who plant-prf-3sg this convent-acc ‘Who has planted this convent?’

(Jókai C. 118)



Katalin É. Kiss

Internal argument foci: (86) a. MYTT adhatok te nagysagos ayandokodnak what-acc give-possib-1sg you great present-poss.2sg-dat ‘What can I give as your great present?’ (Jókai C. 47) b. enny wezedelmekben HOWA fordwlyak this.many peril-pl-in where turn-sbjv-1sg ‘Where shall I turn amidst so many perils?’

(Festetics C. 183v)

Adjunct foci: (87) a. Atte hazad MIKEPPEN kel mastan te nekod ´ the-you house-poss.2sg how needs now you dat-2sg ‘How is your house needed for you now?’ (Nádor C. 159v) b. fyom MERTH hol byuntelen son-1sg why dies sinless ‘Why does my son die sinless?’ (Old Hungarian Lamentations of Mary) Evidence indicating that the focus position is filled by movement is provided by examples where the wh-expression originates in an embedded clause, e.g.: (88) a. HOL akarod hog alkossuc tenèked megenned a· where want-2sg that arrange-sbjv-1pl you-dat-2sg eat-inf-2sg the barant lamb-acc ‘Where do you want us to arrange for you to eat the lamb?’ (Munich C. 32rb) b. KIT akartoc hog èlèrèziec t˙unèctec whom want-2pl that set.free- sbjv-1sg you.pl-dat-2pl ‘Whom do you want me to set free for you?’ (Munich C. 34rb) The embedded verb in (88b) agrees with an indefinite object, hence its object cannot be a definite pro, it must be the indefinite kit ‘whom’. The focus elicits verb movement across Spec,PredP: (89) a. KI hèng´ ereiti èl _ nèkonc a· kouèt a· koporso aitarol ´ ´ who rolls off DAT-1PL the stone-ACC the coffin door-POSS-from ‘Who rolls off the stone from the door of the coffin?’ (Munich C. 53rb)

monnayk AZANNA EWREWMUEL tewltenekuala b. Mert because both that.much joy-INS fill-PERF-3PL-be-PST ‘Because they both had filled up with that much joy’ (Jókai C. 134) meg _ c. EZKEPPEN fordolotvala this-way turn-PRF.3SG-be-PST back ‘He had turned back this way’ (Jókai C. 2)

bel _ in

The evolution of functional left peripheries in the Hungarian sentence



In Modern Hungarian, certain types of constituents, among them interrogative wh-expressions, are obligatorily focused11 (see Horvath 1985). These focusing requirements grammaticalized gradually. As for interrogative wh-phrases, whadjuncts could still appear in the position of clausal adjuncts in Old Hungarian— see (90a–c), but they could access the focus position, too, where they elicited verb movement—see (91a–c). (90) a. mirè t˙u mèstertec eźec and izec a· ièles b˙unos ´ why you master-poss.2pl ate and drank the distinguished guilty muuèlkedètbèlieckèl doers-com ‘Why did your master eat and drink with publicans and sinners?’ (Munich C. 37va) b. de miert ennekem bolondnak engedel but why I-dat-1sg fool-dat give.in-2sg ‘but why do you give in to me, a fool’

(Bod C. 6v)

¯ aźa c. a· taneituańoc čudalkodanac m¯oduan Mik˙eppèn mg the disciples marvelled saying how prt withered-3sg ol’ hamar so soon ‘the disciples marvelled, saying: How did it wither away so soon?’ (Munich C. 26vb) ¯ (91) a. MIRÈ toric-mg a· te taneituańid a· v˙enècnc ´ why break-prt the you disciples-poss.2sg the old-pl-dat źèrzesekèt tradition-poss-acc ‘Why do your disciples break the tradition of the elders?’ (Munich C. 21va) b. S MIERT haborottok meg ez azzonialatra and why be.indignant-2pl prt this woman-at ‘And why are you indignant at this woman?’

(Weszprémi C. 7r)

c. MIKEPPÈN vèthèt -ki sathanas sathanast how cast-possib.3sg -out Satan Satan-acc ‘How can Satan cast out Satan?’ (Munich C. 38va) The wh-phrase of embedded interrogative clauses appeared in Spec,CP, preceding the topic, in the first documents: (92) a. hogy tudna [myt vr ystentewl kertuolna] that know-cond.3sg what-acc Lord God-from ask-perf.3sg-be-cond ‘so that he would know what he had asked of the Lord God’ (Jókai C. 137) 11

Miért ‘why’ is an exception; it can be either focused or adjoined to FocP.



Katalin É. Kiss b. . . . omlanac az aitora hog meglatnac [mi az uolna] fall-3pl the door-on that see-sbjv-3pl what that be-cond.3sg ‘they would throw themselves on the door so that they would see what that was’ (Vienna C. 170) c. . . . meg kymleny . . . [myt euel tenne] prt spy-inf what at.night do-cond.3sg ‘to spy out what he would do at night’

(Jókai C. 144)

However, Old Hungarian developed a general complementizer, hogy ‘that’, to be discussed in Chapter 5, which gradually entered embedded interrogative clauses, as well. In interrogative clauses introduced by hogy, the wh-phrase appears in the preverbal focus position: (93) a. meg kerde [hog ez evt jozagos myelkedetevk kezzevl mellyk prt ask-pst.3sg that this five good deeds out.of which job]. better ‘he asked which is better out of these five good deeds’ (Példák könyve 36) b. meg kerde [hog myt gondolt volna az aztalnal] prt ask-pst.3sg that what-acc think-prf.3sg be-cond the table-at ‘she asked what she had thought at the table’ (Margaret Legend 59r) c. meg kerduen [hogy mykent yewtuolna yly zerre] prt asking that how come-prf.3sg-be-cond such gear-ins ‘asking how she had come with such gear’ (Jókai C. 142) .. The interpretation of the Old Hungarian focus The question whether the Old Hungarian focus had the same semantic content as the Modern Hungarian focus is hard to answer. The Modern Hungarian focus—similarly to the English cleft and pseudo-cleft focus—need not be contrastive, as shown by focus constructions of the following type: (94) a. Jánosnak MÁRCIUS 3-ÁN ünnepeljük (meg) a John-dat March 3rd-on celebrate-1pl (prt) the születésnapját. birthday-poss-acc ‘We celebrate John’s birthday on March 3rd.’ b. A lovas nomád kultúra AZ EURÁZSIAI SZTYEPPÉN alakult the mounted nomad culture the Eurasian steppe-on developed ki. prt ‘Mounted nomad culture developed on the Eurasian steppe.’

The evolution of functional left peripheries in the Hungarian sentence



The Modern Hungarian focus expresses exhaustive identification, or, in a different terminology, it is [+maximal], i.e., it identifies the maximal unique individual that the background section of the sentence predicates about. Thus, (94a) excludes the possibility that we celebrate John’s birthday more than once (in the domain under discussion, e.g., this year), (94b) excludes the possibility that the mounted nomad culture developed in more than one area (in the domain under discussion), and (95) below means that John is the maximal individual who got an A+ in historical linguistics (in the domain under discussion). (95) Történeti nyelvészetből JÁNOS kapott jelest. historical linguistics-from John got A+-acc ‘It was John who got an A+ in historical linguistics.’ If exhaustive identification is a manifestation of the specificational predication relation between the focus and the background (as argued in É. Kiss 2006b, 2008b), and if specificational predication is the semantic correlate of a syntactic predication relation between the focus and the background, then we expect the same exhaustive focus interpretation in Old Hungarian that is attested in Modern Hungarian. For the Modern Hungarian intuition, Old Hungarian focus constructions, e.g. those in (96a,b), are clearly exhaustive: (96) a. KIK ozuc. MIV uogmuc. who those we are ‘Who are those? It is us.’

(Funeral Sermon and Prayer)

b. Ew lelkett TE KEZEDBELEWL keresem meg. he soul-poss.3sg-acc you hand-2sg-from seek-1sg prt ‘It is your hand where I search his soul.’ (Jókai C. 95) It is a more objective and more reliable indicator of the exhaustivity/maximality of the Old Hungarian focus that scalar elements representing a value in the negative domain of a bidirectional scale, which are obligatorily focused in Modern Hungarian, appear in focus position in Old Hungarian, as well. The reason why their focusing is obligatory is that they have an inherent ‘maximal’ feature, which can be shown by the following chain of thoughts. Opposing adverbs like gyakran ‘often’ and ritkán ‘rarely’, or nagyon ‘very much’ and alig ‘barely’, are involved in different entailment relations. A proposition like He is often late is entailed by a corresponding proposition containing a stronger frequency adverb, e.g., He is always late. However, the proposition He is rarely late is not entailed by the stronger proposition He is often late, i.e., rarely must be interpreted as the maximal unique frequency of the given event. Accordingly, gyakran ‘often’ need not be focused in Hungarian, but ritkán ‘rarely’ must land in focus owing to its inherent maximality. (For a detailed discussion of this issue, see É. Kiss 2010).



Katalin É. Kiss

The negative scalar elements consistently focused in Modern and Old Hungarian alike include the adverbs kevesen ‘in small numbers’, ritkán ‘rarely’, and alig ‘barely’. Their focus position in the following examples is demonstrated by the fact that they have attracted the verb from behind the verbal particle: (97) a. my oka hogy. . . emberek yghen kewannyaak az what reason-poss.3sg that people very.much wish-3pl the bólcchesseegók. . . De kewes-en leelhetyk megh wisdom-poss.3pl but few-adv find.possib-3pl prt ‘What is the reason that people very much wish their wisdom but few of them can find it’ (Érdy C. 516b) b. alig esmerhette meg ewttet ew haza neppi barely recognize-possib-prf-3sg prt him he home folks-poss.3sg ‘his home folks could barely recognize him’ (Példák könyve 10) c. Ha te bekesseget keresz celladbol ritkan juz ky if you peace-acc seek-2sg cell-poss.2sg-from rarely come-2sg out ‘If you seek peace, you rarely come out of your cell.’ (Példák könyve 52) .. Summary Old Hungarian inherited a preverbal focus position from Proto-Hungarian and ProtoUgric. However, the preverbal focus of Proto-Ugric/Proto-Hungarian was not more than the position of the internal argument, the locus of pitch accent assignment. It was not accessible to the external argument, and adjuncts could be focused by vacating the VP. In Old Hungarian, the preverbal focus position is an A-bar position, the landing site of focus movement, which can target all kinds of constituents. Focus movement is accompanied by verb movement. The preverbal focus has a [+exhaustive], or [+maximal] feature; the focus construction expresses exhaustive identification.

. The evolution of a left-peripheral NegP .. Introduction This section examines the process in the course of which the V-adjoined negative particle inherited from Proto-Ugric and Proto-Hungarian assumed the status of a leftperipheral functional head eliciting verb movement. As section 2.6.2 demonstrates, the negative pattern inherited from the SOV period of the language still seems to exist in Old Hungarian alongside the innovative construction involving verb movement; in fact, it is sporadically attested even at present. However, whereas in the ProtoHungarian S O NegV sentence, negation might have assumed scope by projecting together with the verb, in Old Hungarian, the sentence structure in which the

The evolution of functional left peripheries in the Hungarian sentence



V-adjoined negative particle occurred had a NegP projection, as well, whose specifier position harbored negative indefinites. Section 2.6.3 reconstructs the process in the course of which negated focused indefinites lost their negative force and came to be reanalyzed as fillers of the specifier of a NegP, whose head was associated with the V-adjoined negative particle via long-distance agreement. .. Two competing negative constructions in Old Hungarian In most Uralic languages, negation is expressed by a negative auxiliary, the descendant of the Proto-Uralic ∗ e ∼ ä ∼a (Rédei 1986–91, entry 1876). In Khanty and Mansi, this auxiliary has developed into a negative particle. In the case of predicate negation, the negative particle is left-adjoined to the verb; in the case of constituent negation, it is left-adjoined to the negated constituent. The negated constituent occupies the preverbal object-focus position. Hungarian also has a negative particle, nem, however, it is of a different origin: it is claimed to be cognate with the indefinite pronoun némi, comprising the indefinite morpheme né, meaning ‘some/any’, and the pronoun mi, meaning ‘what’. As argued by Gugán (2013), originally the indefinite pronoun must have had an intensifying role alongside the carrier of negation inherited from Proto-Uralic; later it completely supplanted the former negator (as happened with the indefinite iht/ieht ‘anything’ in some German dialects according to Jäger (2008)). Like the negative particle of the Ob-Ugric languages, the Old Hungarian nem, too, could negate either the predicate or the preverbal focused constituent. The latter case is illustrated in (98): (98) nem PAYZUAL fegyuerkedet auagy PYKONHOKUAL de ZENT not shield-with arm-refl-prf.3sg or picons-with but holy KERESTNEK YEGYUEL cross-dat sign-poss-with ‘It is not a shield or picons that he has armed himself with, but it is the sign of the holy cross.’ (Jókai C. 147) In the case of predicate negation, the negative particle had two alternative positions: a position sandwiched between the verbal particle and the verb (99), and a position preceding both the verb and the verbal particle (100): (99) a. hogy ezt senkynek meg nem yelentene that this-acc nobody-dat prt not report-cond.3sg ‘that he would not report this to anybody’

(Jókai C. 27)

b. Es sonha az baratoktol el nem menenek and never the fraters-from off not go-pst-3pl ‘And they never left the fraters’

(Jókai C. 145)



Katalin É. Kiss c. senkyt meg-nem sert-uala nobody-acc prt-not offend.3sg-be-pst ‘he would not offend anybody’

(Jókai C. 153)

(100) a. hogy kewsded fratereknek zerzete es essÿ nem that little fraters-dat order-poss and professions-poss not fogyadcoznak meg ytellet napiglan lessen prt doom day-until ‘that the order and professions of little fraters will not lessen until doomsday’ (Jókai C. 104) b. Vgy hogy az baratok ewtett nem vonhattyakuala el so that the fraters him not pull-possib-3pl-be-pst off labaytol feet-poss-pl.3sg-from ‘So that the fraters cannot pull him from his feet’ (Jókai C. 142) In the pattern illustrated in (99), the verbal particle, the negative particle, and the verb are always adjacent, which suggests that the negative particle is adjoined to the verb in a structure of the following type: PredP

(101) Spec meg

Pred’ Pred V

VP

Neg V nem yelentene

… tV …

Jäger (2008) derives a similar ‘verbal particle, negative particle, V’ string attested in the SOV Old High German, involving the verb an-sehen ‘look at’, lit. ‘on-see’, by rightadjoining the verb to the head of NegP, as shown in (102). The Neg+V complex moves on from Neg to T. T’

(102) NegP VP …

T

Neg ni-sehej V

Prt ana

V tj

The evolution of functional left peripheries in the Hungarian sentence



If this were the right structure for Old Hungarian, we would expect the negated verb to be right-adjoined to the temporal auxiliary in T; however, the temporal auxiliary vala/volna always follows the negated verb (see, e.g., 99c). In the pattern illustrated in (100), obligatory adjacency is only attested between the negative particle and the verb; the verbal particle can be separated from the verb by intervening material, e.g.: (103) a. hogy nem vernek vala-kyt meg that not beat-3pl somebody-acc prt ‘that they do not beat anybody’

(Jókai C. 105)

b. Te nemynemew kewekrel . . . nem fyzettel telyesseguel meg you some-kind stones . . . not payed fully pl ‘you have not completely payed for some stones’ (Jókai C. 7) c. De ezek nem alnak erewsen kyserteteknek ellene but these not stand strongly temptations-dat against ‘but these do not withstand temptations strongly’

(Jókai C. 111)

In neutral sentences involving no negation or focus, the verbal particle always immediately precedes the verb. Their inverted order in (100a,b) and (103a–c) must be the result of verb movement to the left, presumably elicited by a Neg head in the left-periphery. The intervening noun phrases and adverbs can be elements scrambled/adjoined to PredP, as shown in (104): NegP

(104) Neg nem

V vernek

PredP vala kyt

PredP Spec meg

Pred’ Pred vernek

VP

The two negative constructions have existed side by side up until recently, with the . . . Neg V. . . Prt . . . order slowly supplanting the . . . Prt Neg V. . . order. Their relative frequency at a given time may have been different in various dialects and registers. Whereas in Jókai Codex, the oldest surviving book-length document (written/translated around 1370, copied around 1448), their proportion is close to fifty-fifty, in private letters from the first half of the 16th century (those in Hegedűs and Papp 1991), still the . . . Prt Neg V. . . order prevails. Relics of the decreasing pattern have survived until now; it is still used in amíg ‘until’ and hacsak ‘unless’ clauses, e.g.:



Katalin É. Kiss

(105) a. Várok, amíg meg nem érkezik. wait-1sg until prt not arrives ‘I wait until he arrives.’ b. Ott leszünk, hacsak le nem késsük a vonatot. there be-1pl unless prt not miss-1pl the train-acc ‘We will be there unless we miss the train.’ The resemblance of the negative construction in (99) to Ob-Ugric negation, and its slow decline suggest that construction (99) represents the pattern inherited from Proto-Hungarian, and the construction in (100), involving V-movement, represents the emerging new structure. In Jókai Codex, the distribution of the two patterns is not completely arbitrary. The . . . Prt Neg V . . . order is typically used if the left periphery of the sentence contains a negative pronoun (i.e., an indefinite pronoun participating in negative concord), and the . . . Neg V . . . Prt . . . order is typically used in lack of an additional left peripheral negative element. Whereas only 13 of the sentences displaying the . . . Neg V . . . Prt . . . order have an additional left-peripheral negative pronoun, 60 of sentences displaying the . . . Prt Neg V . . . have a preceding negative element. This difference is also illustrated in the examples cited above: in (99a–c), the string . . . Prt Neg V . . . is preceded by senkinek ‘to nobody’ and sonha ‘never’, respectively, while in (100a,b), representing the . . . Neg V . . . Prt . . . order, we find no additional negative element. Apparently, the movement of the negated verb to the left periphery (i.e., to the left edge of the comment) was elicited if there was no negative pronoun there to mark the scope of negation. In sum: Old Hungarian had two negative sentence structures: a declining pattern, with the negative particle adjoined to the verb in situ, and an innovative pattern, with the negated verb preposed into the left periphery of the sentence. When the left periphery contained a negative pronoun, the latter, innovative pattern was usually not elicited. The evolution of the innovative pattern of negation is intertwined with the emergence of negative pronouns in the language. .. The emergence of negative pronouns Negative pronouns derive historically from the particle sem (itself the fusion of the focus particle es12 and the negative particle nem), and an indefinite pronoun (also used as an interrogative pronoun) moved to the preverbal focus position. For example: (106) senki ‘nobody’: semmi ‘nothing’: soha ‘never’: sehol ‘nowhere’: sehogy ‘in no way’: etc. 12

[es+nem]+ki [es+nem]+mi [es+nem]+ha [es+nem]+hol [es+nem]+hogy

The particle es/is had other (additive, distributive, etc.) functions, as well.

The evolution of functional left peripheries in the Hungarian sentence



Sem also tended to fuse with a subsequent egy ‘one, a’, and egyik ‘one of ’, yielding semegy ‘no’, semegyik ‘none’. The evolution of these se-phrases must have taken place around the beginning of the Old Hungarian period. Es, nem, and egy are still separate words in example (107) cited from Funeral Sermon and Prayer; and the negative particle and the indefinite pronoun are still non-adjacent, separate elements in example (108) from Jókai Codex: (107) isa es num igg ember mulchotia ez vermut indeed even not one man miss-possib-3sg this pit-acc ‘No man can miss this pit’ (Funeral Sermon and Prayer) (108) de az egyebekrewl nem tudok mytt but the rest-about not know-1sg what-acc ‘but I don’t know anything about the rest’

(Jókai C. 45)

Though the negative pronouns evolved by incorporating the negative particle, in Modern Hungarian they have no negative force any longer; they always require the presence of an additional negative particle (i.e., Modern Hungarian is a strict negative concord language). In the first codices, however, we find several examples where negative pronouns occur without an additional negative particle; where they themselves function as the negative operators: (109) a. mendenestewlfoguan semegyben megharaguuan auagy zuguan altogether nothing-in getting.angry or grumbling ‘not getting angry or grumbling about anything at all’ (Jókai C. 21) b. semegyk mendenestewlfoguan indoltatykuala none altogether leave-caus-3sg-be-past ‘none of them at all was made to leave’

(Jókai C. 139)

s¯emiuèl vag c. Es te bethlechem Iudeanac foldo ´ ´ and you Betlehem Judea-dat land-poss nothing-ins be-2sg k˙usseb smaller ‘And Betlehem, land of Judea, you are not smaller by anything’ (Munich C. 8vb) This type of negative sentence, where negation was expressed by a mere se-phrase, quickly disappeared from the language. Whereas it was still the prevailing pattern in the archaic non-finite SOV sentence types of the first codices in the 14th–15th centuries, its occurrence in finite clauses was sporadic even at that time. Since then, only linguistic fossils such as semmirekellő ‘good-for-nothing’, semmittevés ‘doing nothing, idleness’ have preserved the former negative force of se-pronouns. Se-phrases must have lost their negative operator status because the fusion of nem with the surrounding morphemes triggered phonological assimilation processes as a result of which the presence of the negative particle became less and less transparent. When the



Katalin É. Kiss

negative particle could not be recognized any longer, it had to be spelled out again. The most opaque se-pronouns, soha ‘never’ and senki ‘nobody’, preserving none, or just one, of the original phonemes of nem, always occur with an additional negative particle even in the earliest documents: (110) a. qui Sumha nym hyul which never not abates ‘which never abates’

(Old Hungarian Lamentations of Mary)

b. Es sonha meg-nem sert tyteket valamyben and never prt-not hurts you-acc anything-in ‘And never hurts you in anything’

(Jókai C. 151)

When the negative particle plus indefinite pronoun complex became obsolete, the negative particle was reintroduced in its alternative position, left-adjoined to the verb, as illustrated in (99a–c), (110a,b), and in (111a) below. In the resulting structure, the se-phrase, divorced from the carrier of negation, assumed the function of the scope marker of negation. This must have led to its reanalysis as the occupant of the specifier position of a NegP, as shown in (111b). The negative particle adjoined to the verb must have been associated with the Neg head via Agree. (111) a. hogy sonha touaba meg nem yelennek that never further prt not appear-3pl ‘that they never appear any more’ b.

(Jókai C. 67)

CP C hogy

NegP Spec sonha

Neg’ Neg

PredP PredP

Adv touaba Spec meg

Pred’

Pred nem jelennéki

VP V ti

In sentences with no se-phrase, like those in (100a,b) and (103a–c), NegP was lexicalized by the negated verb, as shown in structure (104). If the negative particle

The evolution of functional left peripheries in the Hungarian sentence



entered the derivation left-adjoined to the verb, as shown in structure (111b), then in sentences displaying the . . . Neg V . . . Prt . . . order, it was the negated V that moved from Pred to Neg, at least initially. The gradual disappearance of the former order suggests that after a while the negative particle came to be base-generated under Neg, where it elicited verb movement.13 I assume that in sentences displaying the . . . Prt Neg V . . . order without any sephrase in Spec,NegP, the negated verb was preposed to Neg in LF. Ürögdi (2012) has demonstrated that in the present-day relic of this order, occurring in amíg ‘until’ clauses, the negative particle indeed assumes scope in a left-peripheral position. Observe the following example: (112) Addig olvastam, [amíg hirtelen ki nem aludt a villany]. till.then read-pst-1sg until suddenly out not went the light ‘I was reading as long as it was not the case that the light suddenly went out.’ Amíg as a complementizer means ‘for which time/as long as’. Hence the above sentence is only meaningful if the negative particle takes scope over the adverb hirtelen ‘suddenly’, i.e., if it is LF-moved to the left periphery. .. Summary On the basis of these observations and considerations, the following historical process can be reconstructed. In the Proto-Hungarian SOV sentence, negation was presumably expressed by a negative particle adjoined to the verb, or, in the case of constituent negation, by a negative particle adjoined to the focused object. The reinterpretation of the preverbal section of the sentence as a functional domain did not affect the relative word order positions of the negative particle, nevertheless, in the long run it led to the reanalysis of nem as a Neg head projecting a NegP, eliciting V-movement. The evolution of NegP may have been initiated by the morphological fusion and phonological erosion of negated indefinite pronominal foci. The opaqueness of the “focus particle, negative particle, indefinite pronoun” complex necessitated the reintroduction of the negative particle in V-adjacent position, whereby the left-peripheral negative pronoun came to be interpreted as the scope marker of negation, occupying the specifier of a NegP. NegP was also projected in sentences containing no se-pronoun, where it tended to be lexicalized by the movement of the negated verb from V via Pred to Neg. In

13 In a slightly different approach, the landing site of the verb is not in Neg, right-adjoined to the negative particle, but in the head position of a functional projection below NegP—see, e.g., É. Kiss (a). The assumption that verb movement in sentences containing a focus and/or negation is not cyclic but stops in the lowest functional head makes it easier to derive the whole array of word order possibilities involving focus and/or negation, i.e.: ‘Focus V Prt tV . . . ’, ‘Neg V Prt tV . . . ’, ‘Focus Neg V Prt tV . . . ’, ‘Neg Focus V Prt tV . . . ’, ‘Neg Focus Neg V Prt tV . . . ’. These constructions all fit into the frame [NegP . . . [FocP . . . [NegP . . . [XP V [PredP Prt tV [VP . . . ]]]]]].



Katalin É. Kiss

Modern Hungarian, the negative particle is generated under Neg (except for certain syntactic fossils, e.g., until-clauses), where it elicits V-movement.

. Conclusion This chapter has argued that Hungarian underwent a fundamental change in the late Proto-Hungarian–early Old Hungarian period, prior to its documented history; its head-final, SOV sentence structure with fused grammatical functions and discourse roles changed into a (Topic) (Focus) V X* structure, displaying head–complement order, and a rich functional left periphery independent of the thematic domain. It is a matter of scholarly debate whether the syntax of an undocumented language stage can be reconstructed—see Lightfoot (2002) versus Campbell and Harris (2002) and Ferraresi and Goldbach (2008)—and if reconstruction is attempted, what kind of evidence and what methodology it should be based on. The first part of this chapter is an attempt at reconstructing the basic structure of the Proto-Hungarian sentence on the basis of two types of evidence: Old Hungarian archaisms, and corresponding constructions of Khanty and Mansi, the sister languages of Hungarian.14 If these two kinds of evidence, obtained from independent sources, converge, reconstruction yields a plausible result. It is assumed in historical linguistics that changes typically follow an ∫curve, i.e., the spreading of a new structure starts slowly, then it gains momentum, and eventually it slows down again. This means that a vanishing construction, supplanted by a new form in most contexts, may linger on for a long time, i.e., prevailing constructions of former periods may survive as linguistic relics (see Weinreich and Herzog 1968). What we have done is extended the ∫-curve representing the spreading of a new variant and the disappearance of the old construction backwards into the undocumented past of the language. If the sister languages have preserved the disappearing pattern, we have regarded it as independent evidence that it represents the prevailing construction of the past. The first Old Hungarian linguistic records display several constructions which become sparser and sparser in later documents, and either completely disappear by (or in) the Middle Hungarian period, or lose their flexibility and survive as linguistic fossils. We have checked whether these vanishing constructions exist in the sister languages, and if we have found them there, we have concluded that they represent Ugric heritage preserved in Proto-Hungarian. The constructions surviving from Proto-Hungarian include a strict SOV base order with morphologically unmarked subject and object constituents, a head-final TP, and a clause-final interrogative complementizer (evidence of a head-final CP). Proto-Hungarian used primarily non14 We used significantly more evidence from Khanty than Mansi (though Mansi is believed to be closer to Hungarian) because more information is available on the syntax of Khanty than Mansi (see the studies of Nikolaeva, e.g., Nikolaeva , and ).

The evolution of functional left peripheries in the Hungarian sentence



finite subordination. In Proto-Hungarian, the subject also functioned as a topic, and the object functioned as a focus or secondary topic. In the latter case, it elicited verbal agreement, which was blocked by the Inverse Agreement Constraint if the secondary topic was inherently more topical than the primary topic. The shift from SOV to Topic–Focus–V–X∗ could be initiated by the spreading of right dislocation, which was facilitated by the generalization of differential object marking (originally targeting only topical objects) to all objects. When the rate of right dislocated constituents achieved a certain threshold, they were reanalyzed by new generations of speakers as arguments in situ, and the preverbal constituents were reinterpreted as operators in A-bar positions. This hypothesis is based on the following pieces of evidence: First, right dislocation is common in all the SOV Uralic languages, among them Khanty and Mansi. Second, the Modern Hungarian sentence has two properties which seem to have grammaticalized/fossilized at the time when the postverbal constituents were the output of right dislocation, namely: (i) The order of postverbal constituents is free; it is completely independent of grammatical functions, which follows if postverbal constituent order is the result of iterated right dislocation. (ii) Only referential constituents are licensed postverbally, which follows if their postverbal position is due to right dislocation, which can only target referential elements. The preverbal section of the Old Hungarian sentence has developed into a complex, multi-layered functional left periphery. What happened was more than the preverbal constituents of Proto-Hungarian, associated with thematic and discourse functions, losing their thematic restrictions. The Proto-Hungarian lexical node dominating an occasional verbal particle plus verb or nominal predicate plus copula complex has developed into a PredP projection, the locus of complex predicate formation, having a crucial role in encoding event structure and, thereby, situation aspect. The ProtoHungarian object-focus, the carrier of pitch accent, conveying new information, has preserved its preverbal surface position and its pitch accent; however, its derivation and its interpretation have changed. In Old Hungarian (and ever since), the preverbal position of the focus has been due to focus movement to Spec,FocP, accompanied by verb movement. The focus-moved constituent assumes a [+exhaustive]/[+maximal] feature; it is interpreted as a specificational/identificational predicate. The negative particle has developed from an adverb-like element adjoined to the verb and projecting together with it into a Neg head inserted in scope position, projecting a NegP, eliciting negative concord. Universal quantifiers, not being possible targets of right dislocation, have also ended up in the left periphery, in scope positions. The topic projection has become available for arguments of any grammatical function. Finite embedded clauses have developed a CP layer, with C harboring a complementizer, and Spec,CP providing a landing site for relative pronouns.

 The DP-cycle in Hungarian and the functional extension of the noun phrase BA R BA R A E G E D I

. Introduction The first impression might be deceptive, and it is indeed so if one looks through the overall history of Hungarian nominal expressions. The inner structure of the noun phrase has not changed a lot, at least as compared to some other phenomena on the sentential level. The order of the major constituents seems to be the same, as far as the order of the modifiers and the modified elements is concerned, and the possessive constructions also appear to have been used in the first written sources in the same manner as in Modern Hungarian. The most significant change is the emergence of the definite article, the exact date of which cannot be determined. The article seems to be attested as early as the first continuous text records appear, admitting that its use was considerably different from the way it functions in Modern Hungarian. The gradual expansion of the definite article can, however, be well observed in the historical stages and it undoubtedly had important structural consequences at the nominal left periphery. The aim of this chapter is to follow, describe, and analyze these interrelated phenomena, namely the emergence of the article, its functional expansion, the various strategies for the renewal of the demonstrative system, and the rearrangement of the leftmost peripheral position within the noun phrase; to put it differently, the DP-cycle in Hungarian and its consequences at the nominal left periphery. In section 3.2 the marking of definiteness will be addressed beginning with a general discussion of the phenomenon. This is followed by a more specific survey of definiteness marking in Proto-Hungarian as well as in Early and Late Old Hungarian. Section 3.3 is concerned with the DP-cycle, more precisely with the problem as to how the demonstrative system was renewed after the definite article had grammaticalized.

The DP-cycle in Hungarian and the functional extension of the noun phrase



It will be shown that in Hungarian, the renewal of the demonstratives involves at least two types of strategies (reinforcement and determiner doubling). The structural reanalysis of determiner doubling (from adjunction to a DP-internal specifier position) that took place in Middle and Modern Hungarian is also discussed. Section 3.4 presents how the competing strategies of demonstrative modification coexist from the Middle Hungarian period onward and what their descendant constructions in present-day Hungarian look like. This part will be complemented by an inspection of some dialectal peculiarities. The chapter relies heavily on corpus queries carried out in the recently developed (and still developing) Old Hungarian corpus. The chapter is concluded by summarizing the results.

. Marking of definiteness in Old Hungarian .. About definiteness and its marking In this chapter, definiteness is conceived and used in terms of the basic pragmatic notion of identifiability. According to this notion, the speaker signals that the hearer is able to assign a referent to a certain noun phrase.1 The hearer can identify the referent either because it is already accessible in the context, or it can be associated with another discourse referent present in the context, or the reference is clearly identifiable or inherently unique by the shared knowledge of the interlocutors in a given speech situation.2 Although the category of identifiability and, as a consequence, the semantic and pragmatic notion of definiteness can be assumed to be universal, its grammatical realization is a language-specific property. The definite article is the grammaticalization proper of the semantic and pragmatic concept of definiteness, but many languages lack this type of grammatical element. Definiteness can be encoded in several alternative ways, for instance by positional ordering, special case-marking, or verbal conjugation. Even in languages that make use of the definite article, its relative frequency or the contexts in which the article appears are not necessarily identical. The grammatical encoding of referential identification may segment the semantic field at different points in different languages, i.e. there is great variation as to how extensive the ground covered by the category of definiteness is within the semantic/pragmatic field in a language. Accordingly, this semantic range may even change in time (Lyons 1999: 336–337). In a language documented long enough for a 1 For definitions and properties of definiteness (e.g., uniqueness, inclusiveness, familiarity), consult first of all chapters  and  of Lyons (), and Alexiadou et al. (: –). For different approaches and concepts characterizing definite descriptions, see also Abbott (). 2 As Hawkins () claims in a detailed pragmatic account, existence and uniqueness of a definite description must hold within the universe of discourse or a subset of it, which can be inferred by the hearer through specific pragmatic parameters. For similar approaches, see Westerståhl’s () context sets, and Roberts’ () informational uniqueness relative to the discourse situation.



Barbara Egedi

diachronic investigation, one can observe how the grammatical encoding of referentiality changes gradually and what factors influence the process. Gradualness is an important feature in the case of Hungarian, too. When the definite article emerged, it first only appeared in the constructions where the referent of the noun phrase was not anchored in another way. To put it differently, the early variant of the Hungarian definite article encoded pragmatic definiteness only—as will be explicated in section 3.2.4. When the definite article emerges in a language, as a result of grammaticalization, it is not uncommon that the first attestations remain uncertain or ambiguous for a longer period. The source category begins to fulfil a new function, but its formal properties do not necessarily change immediately, and the two related constructions may appear to remain structurally homophonic.3 As in so many other languages, the Hungarian definite article developed from the distal demonstrative, but the time of this functional split is uncertain. In Old Hungarian the article and the distal demonstrative look identical, share a phrase-initial prenominal position, and even overlap functionally (e.g., in anaphoric use). Relevant data only survived in writing, so even if the simple definite noun phrases and those modified by a demonstrative were distinguished by different intonational properties from the very beginning, in absence of its graphic marking, it is impossible to test. This kind of uncertainty in identifying early uses of the article shows up in describing the historical changes in other languages as well, where the formal and distributional criteria are not sufficient to distinguish between the canonical and the article-like uses of a demonstrative (for the same problem in Old English, see Sommerer 2011: 183–209). To identify the Old Hungarian definite article, special semantic and pragmatic contexts must be examined in which the determiner in question appears regularly and in which it can hardly be interpreted as a demonstrative. As Nikolaus Himmelmann (2001: 833–834) points out, building on Hawkins’ (1978) systematic presentation of article use in English, demonstratives must not be used in certain semantic and pragmatic contexts in which articles consistently appear. Such contexts are the larger situation use (“the first mention of entities that are considered to be unique, hence generally identifiable, in a given speech community”) and the associative-anaphoric use (“the first mention of an entity that is not unique per se but with respect to a previously mentioned referent”). This approach turned out to be quite feasible in identifying Old Hungarian determiners, but the method obviously needs satisfactory contexts to apply. Unfortunately, the earliest article-like uses of the distal demonstrative remain ambiguous due to the nature of the texts in which they appear, as will be shown in section 3.2.3.

3 It is a general observation made on grammaticalization processes that change in meaning and in use tend to precede a change in form (cf. Heine, Claudi, and Hünemeyer : –).

The DP-cycle in Hungarian and the functional extension of the noun phrase



.. What can be said about definiteness marking in Proto-Hungarian? Before the end of the 12th century we have no continuous written sources in Hungarian. Any hypothesis about Hungarian syntax before that date can only be formed by means of reconstruction based on the evidence that the sister languages provide— which themselves had not been documented before the 19th century. According to the generally accepted view in the descriptive literature, the Hungarian definite article is the result of an internal development (Simonyi 1914: 68–69; Klemm 1928: 317; D. Mátai 2003: 419–420), which reasonably conforms to the fact that Uralic languages in general have no definite article at all. Exceptions to this tendency are, besides Hungarian, the Mordvin language, where a definite vs. indefinite declension developed (Zaicz 1998: 191–193) and Finnish, in which the emergence of the article is a fairly recent phenomenon and seems to be taking place currently in the spoken language variety (Laury 1997). Although the exact time of grammaticalization of the Hungarian article is impossible to tell, it must have taken place at the beginning of the Early Old Hungarian period or even before, during Proto-Hungarian. The change in definiteness marking was probably not independent of other, more general structural changes on the sentence level, which finally led to the characteristic discourse-configurational word order patterns of Hungarian (see sections 2.2 and 2.3 in the previous chapter and É. Kiss 2013 on this matter). But what can be said about the determination system in Proto-Hungarian before this new strategy emerged? It has long been observed and described that in most of the Uralic languages possessive affixes (primarily the 3rd person singular form) are frequently and quite regularly used to mark the definiteness of the noun, without expressing any kind of possessive relationship. It seems to be generally agreed on that this non-possessive function of the possessive affixes (Px) corresponds to that of the definite article in Indo-European languages.4 Irina Nikolaeva (2003) argues for a more complex explanation of the function these possessive affixes fulfil. According to her analysis, Px may indeed express identifiability as far as the 3rd person singular form is concerned; its use, however, is not obligatory. At the same time, the 1st and 2nd person singulars in non-possessive meaning rather serve to link the referent of the noun phrase to the participants of the actual speech act. Somewhat similarly, Kari Fraurud’s insightful study (2001) concludes that associativity is more essential than referentiality when one aims to find the common feature of a wide range of non-possessive uses of Px in different languages. That is why these special possessive clitics/affixes show a remarkable formal and functional stability in time, and do not necessarily develop into an article. Moreover, nothing prevents demonstratives from appearing in articlelike uses in these languages; they may step on a grammaticalizational path that finally leads to the birth of a definite article, independently of the extended uses of 4 Cf. inter alia Collinder (: –, §); See also Künnap () for a general survey of the phenomenon in various Uralic languages. One may find a good source of examples in Nikolaeva ().



Barbara Egedi

the Px. This observation is essential considering that the Hungarian definite article indubitably developed from a demonstrative, but we do not want to exclude the possibility that possessive affixes were also used for determination (more precisely, for referential anchoring) in an earlier stage of the language. It is to be noted that the determiner function of possessive affixes has been suggested to be a feature already present in Proto-Uralic (Décsy 1990: 81). Taking into consideration that the Ob-Ugric languages, the closest relatives of Hungarian, presentday Khanty (Ostyak) and Mansi (Vogul) also share this property, it is highly probable that Proto-Hungarian Px inherited this feature but, during its separate life, lost it at a certain point. As a matter of fact, such a use of possessive suffixes cannot be proved for Hungarian. Only a few morphological remnants suggest that the function of Px once might have been more complex than it is today. For instance, it is remarkable that the accusative form of personal pronouns in first and second persons always includes an apparently redundant Px, as can be observed in the contrastive Table 3.1. This might be the relic of a more productive marking of definite objects in Proto-Hungarian. (For the absence of the accusative ending -t in certain cells of the paradigm, see section 2.2.2. in Chapter 2.) (1) engem, téged, minket, titeket ‘me, you, us, youpl ’ (accusative series) Table .. Personal pronouns in nominative and accusative nominative

accusative

singular

plural

singular

plural

first person

én

mi

en-g-em-Ø I-ep-1sg

mi-nk-et we-1pl-acc

second person

te

ti

té-g-ed-Ø yousg -ep-2sg

ti-tek-et youpl -2pl-acc

There are also some odd expressions, primarily occurring in spoken registers, in which a non-possessive Px is used5 to express an (often negative or degrading) emotional attitude of the speaker towards the referent of the noun phrase. Consider the following examples: (2) a. a hülyé-je the stupid-3sg ‘this stupid one’

5

b. a szemtelen-je the shameless-3sg ‘this shameless one’

This observation has been made by Katalin É. Kiss (p.c.).

c. a csóró-ja the blighter-3sg ‘this blighter’

The DP-cycle in Hungarian and the functional extension of the noun phrase



Nevertheless, the determiner use of possessive suffixes in a systematic and productive way remains merely hypothetical, since in the first linguistic records, e.g. in the Funeral Sermon and Prayer (ca.1195), there is no trace of the non-possessive use of possessive suffixes. At the same time, the determiner az (‘that’) already seems to function more like a definite article. .. Is there a definite article in Early Old Hungarian? The question formed in the above title is not a trivial one. The use and distribution of the article in the earliest codices from the Late Old Hungarian period show a considerable difference with respect to the modern system of determination, which needs some explanation, but the existence of a true, grammaticalized article in the period of the codices is undeniable (see the next section). At the same time, the article status of the definite determiners in the Early Old Hungarian sources cannot be justified with certainty, and the very first attestations of the article, or more precisely the earliest article-like uses of demonstratives, remain uncertain. As has already been pointed out, when the demonstrative functionally split into a deictic determiner and a simple definite article, the formal and distributional properties of the two elements did not diverge immediately. In many cases, the interpretation of the determiners is ambiguous for today’s reader since they have identical forms, share the same prenominal position, and they even functionally overlap, e.g., in their direct anaphoric use. Special semantic and pragmatic contexts in which only articles can appear may help to decide the question, but in the Early Old Hungarian records the textual contexts are simply not as ample as would be necessary for disambiguation. The majority of the Early Old Hungarian sources are short and not continuous. They are principally names of people and places and glosses embedded in Greek or Latin documents and charters.6 Isolated words or phrases are obviously inadequate for a syntactic analysis. The first continuous texts from the same period (listed in Table 3.2) are traditionally grouped together and labeled as “shorter text records from the age of the Árpád dynasty”, but these texts, in reality, do not form a uniform corpus either

Table .. Text records from the age of the Árpád dynasty Manuscript

Date

Funeral Sermon and Prayer The Königsberg Fragment and Ribbons The Old Hungarian Lamentations of Mary Gyulafehérvár Lines

ca.1195 end of 12th c. – beginning of 13th c. second half of 13th c. second half of 13th c.

6 The first charter that contains a considerable amount of Hungarian words and word-groups is the Letter of Foundation of the Tihany Abbey, which can be dated to .



Barbara Egedi

in time or in space. They are actually very short (tokens will be given in Table 3.3). One of them is very fragmentary (The Königsberg Fragment and Ribbons), as it was found reused in the binding of another codex. The Lamentations of Mary is written in verse. The Gyulafehérvár Lines cannot even be considered as a coherent text, being composed of 15 juxtaposed noun phrases which are not embedded in sentences. In these texts, the number of the cases in which definite determiners appear and their function might be examined is small. Lamentations of Mary has no such case at all. The phrase of the Gyulafehérvár Lines in (3) very probably has an article before the complex participial modifier of the noun, but the whole phrase is isolated, a member of a list, so the precise nature of its use cannot be defined for lack of context. The “articlehood” of the determiner in (4) cited from the Königsberg Fragment and Ribbons is more debatable: the function is clearly cataphoric, and as such appropriate for a demonstrative as well. Moreover, the lexeme for God (ysten) is inherently unique by nature and this noun type consistently resists being determined, at least in the period of the first codices. (3) oz kerezt fan figeu kepeben the/that cross wood-sub hanging picture-poss-ine ‘in his form of hanging on the cross’

(Gyulafehérvár Lines)

(4) de qui legen neky atia ozut nem but who be-sbjv dat-3sg father-poss.3sg that-acc not Ez oz ysten mynt, evt esmeríuc quit sceplev tudhot iuc him know-1pl whom spot know-possib-1pl this the/that god as nem illethet. not concern-possib.3sg ‘But who would be his father, we cannot say. This is the/that God as we know him who is immaculate.’ (The Königsberg Fragment and Ribbons) The determiner oz ‘that’ occurs four times in the Funeral Sermon and Prayer, and the contexts of its uses are clear. In the previous discourse, God barred Adam only from the fruit of a single tree in Paradise. The first mention of this particular fruit is embedded in a possessive structure. Thus, there is no need for any determiner (see the discussion below). At the second mention, the determiner oz appears (5), which can be interpreted either as a demonstrative or as a definite article because it indicates a direct anaphor. The third and fourth mentions of the same fruit (6), as well as the fifth (7), are still anaphoric, but the clear coherence of the text makes a demonstrative interpretation redundant. These latter uses are much closer to what a definite article is supposed to do: the determiners simply identify the referent of the noun phrases as present in the discourse without any deictic or locative force.7 7 For the detailed presentation of all these early cases with an overview of the previous literature on the matter, written in Hungarian, see Benkő (: –). Loránd Benkő himself admits that based on this scarcity of data one can hardly form a definitive opinion about the use of the definite article in the period as a whole.

The DP-cycle in Hungarian and the functional extension of the noun phrase



(5) ysa kí nopun emdul oz gimils twl halalnec indeed which day-sup eat.mod.2sg the/that fruit-abl death-dat halaláál holz death-poss-ins die-2sg ‘Indeed, the day when you eat from that/the fruit, you will die the death of deaths.’ (Funeral Sermon and Prayer) (6) es evec oz tiluvt gimilstwl. es oz gimilsben and eat-pst.3sg the/that forbidden fruit-abl and the/that fruit-ine halalut evec. death-acc eat-pst.3sg ‘and he ate from the/that forbidden fruit and in the/that fruit he ate death’ (Funeral Sermon and Prayer) (7) Es oz gimilsnec wl keseruv uola vize and the/that fruit-dat so bitter be-pst water-poss ‘and the/that fruit had such a bitter juice’ (Funeral Sermon and Prayer) Table 3.3 summarizes all the article-like occurrences of the demonstrative oz (>az) in the Early Old Hungarian period. With such a low quantity of data it does not make much sense to draw charts and present statistics, but the overall number of determiners in proportion to the tokens of the manuscripts may be interesting as compared to the corresponding ratio found in later codices (e.g., those figuring in Table 3.4).8 Table .. The proportion of article-like determiners in the Early OH manuscripts MS

Date

Tokens

oz



Funeral Sermon and Prayer The Königsberg Fragment and Ribbons The Old Hungarian Lamentations of Mary Gyulafehérvár Lines

ca.1195

281

4

1.42

end of 12th c. – beginning of 13th c.

389

1

0.25

second half of 13th c.

145

–

0

51

1

1.96

866

6

0.69

Total

second half of 13th c.

The first texts of considerable length and thus suitable for syntactic investigation come from the first half of the Late Old Hungarian Period in the form of codices written in Hungarian. It is remarkable that the proportion of the morpheme (oz >) az in the first 8 Token numbers are always given on the basis of the original spelling of the manuscript, but without punctuation marks.



Barbara Egedi

extant codex is already 2.52, and this number is consistently rising as time goes by (see Table 3.4 in the following section). .. The use of the definite article in Late Old Hungarian For reasons discussed so far, the reconstruction of the earliest coherent synchronic system of determiners must focus on the first half of the Late Old Hungarian Period. Nevertheless, this period as a whole (1370–1526) cannot be discussed uniformly, considering that the use of the definite article in the earliest codices, which can be dated to the end of 14th century and the first half of the 15th century, is somewhat different from that of manuscripts from the beginning of the 16th century. The early manuscripts, the Jókai Codex and two further codices containing translations of the so-called Hussite Bible, are already long enough, they are uniform, narrative texts, each forming a closed corpus of its own. The Jókai Codex is the first extant, handwritten book in Hungarian. It is about the life and deeds of Saint Francis of Assisi, and is a 15th-century copy of the original translation from around 1370. The Hussite Bible is the first Hungarian Bible translation made between 1416 and 1441. It is preserved in three codices from which only two contain relatively early copies of the original translation: the Vienna Codex from 1450 (Old Testament books and the twelve minor prophets) and the Munich Codex dating to 1466 (the four gospels). As has already been pointed out, due to the formal and positional equivalence and functional overlap (e.g. anaphoric use) between the Old Hungarian article and the demonstrative modifier which the article developed from, these determiners cannot be distinguished from each other merely on formal grounds—at least for today’s reader. Apparently, the same morpheme corresponds to two different functions: sometimes it seems to behave as a demonstrative, sometimes as a definite article, and often it is impossible to decide between the two alternatives. It is therefore not surprising that in the descriptive literature on Old Hungarian there is a strong tendency to consider these early articles as “pre-articles” or “pronoun-articles” that belong to a special transitional word class with dual nature (Bakró-Nagy 1999:7; I. Galassy 1992:721–722). Nevertheless, contrary to earlier records, Late Old Hungarian sources are appropriate for a more sophisticated investigation: the definite article can be argued to be a fully grammaticalized category encoding definiteness on a syntactic level, since one can easily search for specific semantic and pragmatic contexts in which only an article can appear. For instance, the associative-anaphoric use of the definite article can be well attested as early as in the first codex: (8) masod napon mykoron azon frater ewlne az aztalnal second day-sup when that/same frater sit-cond.3sg the table-ade neze zent ferencznek kepere look-pst Saint Francis-dat picture-poss-sup ‘on the second day, when the same frater was sitting at the table, he had a look at the picture of Saint Francis’ (Jókai Codex 66)

The DP-cycle in Hungarian and the functional extension of the noun phrase



In the textual context of (8), there was no mention of any table before, but the scene is one of the convents of the monks where the protagonist, a certain preaching monk, has just arrived, as is reported in the preceding lines. The presence of a table in a building used by monks, or people in general, is naturally assumed. Considering that the object itself was not concretely mentioned before, the determiner preceding the noun cannot be a demonstrative (see also Egedi 2013 for further examples). The other special context which has been claimed to be specific to articles and from which demonstratives are banned is the larger situation use. The Old Hungarian definite article, however, appears in this context only gradually, as inherently unique nouns (such as God, earth, moon, devil, etc.) resist being determined, similarly to proper names, which are also semantically definite. It can be generally observed that the use of the article in early codices had a more restricted use than in the subsequent stages of the language. The definite article is absent with nouns modified by a demonstrative, which encodes directly accessible reference, and in possessive constructions, in which case the referent of the possessed noun is existentially presupposed and is identified through its relation to the referent of the possessor.9 The presence of a demonstrative (9) or a possessor expression, either pronominal (10) or nominal (11), implies the definiteness of the noun phrase as a whole, and the use of the article is not yet required. It may be concluded that the definite article only appears when referential identification is not encoded otherwise. This means that what has to be obligatorily marked on the syntactic level by a definite article, already at the time of the first codices, is pragmatic definiteness. (9) ez kener-ek-re this bread-pl-sub ‘onto these breads’ (10) èn keńèr-i-m-èt I bread-poss.pl-1sg-acc ‘my breads’ (11) az e˙lèt-nc keńèr-è the life-dat bread-poss ‘the bread of life’

(Jókai C. 76)

(Vienna C. 182)

(Munich C. 91ra)

The system of determiners has been changing quite rapidly. The use of the definite article proportionally increased already within the Old Hungarian period, as the results of an automatic query carried out in five normalized codices from the Old Hungarian Corpus indicates:

9 The identification and detailed analysis of definite contexts in which the article is still absent is discussed in Egedi (), which also provides several original data illustrating the various contexts.



Barbara Egedi Table .. The proportion of definite determiners (a/az) in five Old Hungarian codices Codex

Date

Tokens

a/az



Jókai Codex Vienna Codex Guary Codex Könyvecse Kazinczy Codex

after 1370/ca.1448 after 1416/ca.1450 before 1495 1521 1526–1541

22,733 54,423 21,714 8,745 20,027

573 2,233 1,390 623 1,437

2.52 4.10 6.40 7.12 7.17

The drawback of automatic queries like the one that yielded the figures above is that, besides adnominal modifiers/determiners, independently used demonstrative pronouns in nominative case also turn up in the search results. However, the well observable proportional increase of the morphemes a/az can only be due to the gradually increasing use of the definite article, that is, by its spreading into syntactic contexts in which it was not obligatory before. Note that this increase does not make its way equally fast in all the possible contexts: we observed that the spreading is more characteristic before generics and pronominal possessors, while the contexts presented in (9) and (11), nouns with demonstratives and lexical possessives, resisted the determiner throughout the period.10 At the same time, the automatically generated figures cannot be distorted to a large extent, since it is only the nominative pronouns that coincide with the adnominal demonstrative modifiers in form, and their number is supposed to be relatively low in a given text, as will be pointed out with respect to Table 3.5 in the next section. The Modern Hungarian phrases that correspond to the Old Hungarian examples in (9)–(11), given below in nominative case for the sake of simplicity, look rather different, as the definite article has to be used obligatorily in these contexts. The article appears with the demonstrative modifier, before a pronominal possessor, and on the possessed noun in the case of a dative-marked possessor.11 (12) ez a kenyér this the bread ‘this bread’ (13) az én kenyer-e-i-m the I bread-poss-pl-1sg ‘my breads’ 10 The results of this research were presented on the th June , at the conference “Exploring Ancient Languages through Corpora” in Oslo, see Egedi and Simon (). The figures in Table ., however, have recently changed slightly, as the ongoing correction and improvement of the Old Hungarian Corpus continuously modify the results. Unfortunately, the proportional increase cannot be measured in the Middle Hungarian period, as the digitalization of Middle Hungarian records has just begun. 11 The variation between a and az in Modern Hungarian is regular and purely morpho-phonetic, depending on whether the subsequent word begins with a vowel or a consonant.

The DP-cycle in Hungarian and the functional extension of the noun phrase



(14) az élet-nek a kenyer-e the life-dat the bread-poss ‘the bread of life’ Despite these clearly observable differences, the definite article had undeniably grammaticalized to encode simple referential identification by the time of the first part of the Late Old Hungarian period, even if it did not expand immediately into all syntactic contexts in which it is used today. Moreover, for quite a long period, simple definite noun phrases and those modified by a distal demonstrative could not be distinguished merely on formal grounds (unless by intonation). Nevertheless, the article and the demonstrative can be assumed to occupy distinct structural positions, D and Spec,DP, respectively. The emergence of the definite article may be reconstructed similarly to what Giuliana Giusti (2001: 167) proposes for the reanalysis of Latin ille in the Romance languages. As is shown in (15), ille, originally located in the specifier of the Determiner Phrase as a demonstrative, is reinterpreted as an element in D. (15) a.

DP DemP (IL)LE

b. ⇒

D⬘ D

…

DP Spec

D⬘ D

…

(IL)LE

This kind of reanalysis perfectly conforms to the economy principles formulated by Elly van Gelderen (2004, 2008), the so called “Head Preference Principle”, according to which speakers prefer to build structures where an element is merged directly into the head position instead of moving it to the specifier from below.12 If we accept that demonstratives are base-generated in a position lower than DP (see Giusti 1994, Bernstein 1997, and first of all Szabolcsi 1994b for distinguishing a D and a lower DET position in Hungarian),13 the above economy principle can be assumed to have been working in Old Hungarian when the definite article emerged. Demonstratives, 12 It is to be noted that the economy principles, which are often responsible for diachronic changes, express preferences rather than absolute principles and can be violated by prescriptive or innovative forces, as Elly van Gelderen argues. 13 The existence of a lower DET position is easy to justify in Modern Hungarian for there are constructions in which both determiner positions are filled, see (a) in section ... In Old Hungarian this claim remains hypothetical since pronominal possessors exclude the use of the definite article. However, we may find examples where a pronominal possessor expression combines with a determiner quantifier, and the latter is lower in the structure. Assuming that determiner quantifiers are generated in the same position as demonstratives in Old Hungarian as well, we may get indirect evidence for DET in Old Hungarian: (i) es te menden èllènsegid èluèznèc and you all enemy-poss.pl-2sg perish-3pl ‘and all your enemies perish’ (Vienna C. )



Barbara Egedi

associated with the grammatical features [+definite] or [+referential], left their original lower position and moved into the specifier of the DP. Since this movement took place frequently and regularly, the demonstrative element could be reanalyzed as a head rather than a phrase, and merged directly into the D head. Note that it is less economical to merge an element early in the syntax and then move it to a higher position than to procrastinate merging as long as possible (see van Gelderen’s “Late Merge Principle”). (16) a.

DP DemP oz (>az)

b. ⇒

D⬘ D

… DetP

DP Spec

D⬘ D oz (>az)

…

These two constructions may coexist, and they actually did co-exist for several generations in Old Hungarian, until new strategies for demonstrative modification emerged. However, in more archaic expressions (to be presented in 3.3.1) and in the very old construction with proximal demonstrative (e könyv ‘this book’), this structural parallelism survives in Modern Hungarian as well.

. The DP-cycle According to a frequently cited generalization by Joseph Greenberg (1978), grammaticalization often takes the direction from demonstratives developing into articles, while the latter at a later stage might develop into a gender marker or a mere sign of nominality. It is true that the DP-cycle, in a typical case, involves demonstratives becoming articles, but at the same time, demonstratives must be renewed in some way. According to Elly van Gelderen’s (2011) insightful overview of the phenomenon in several languages of the world, this renewal takes place by one of the following two strategies: it is frequently done either by reinforcing the old form with a locative adverb or by the incorporation of an additional demonstrative (e.g., an appositive pronoun) into the construction. So, in one of the strategies the new element comes from below, while in the other it incorporates rather high in the DP structure. Sometimes there is evidence for the simultaneous application of both strategies, as was the case in Old Norse (van Gelderen 2007: 295 and 2011: 207–208). It will be shown in the following sections that Hungarian also made use of at least two types of strategies after the definite article had grammaticalized and, in the case of the proximal demonstrative modifier, even the original structure was able to survive into Modern Hungarian.

The DP-cycle in Hungarian and the functional extension of the noun phrase



.. The renewal of the demonstrative: reinforcement As has been shown in section 3.2.4, after the reanalysis of the demonstrative as a simple definite determiner, it can be merged directly into the D-head. The two constructions, the original one, still open for deictic elements, and the newer one with the directly merged article, may coexist for a longer period. In Romance, however, new lexical items developed to fulfill the demonstrative function. An adverbial reinforcer (eccum) has been added to the phonologically weakened demonstrative head resulting in a new series of deictic elements (Giusti 2001: 170):14 (17) a.

DemP

b.

DemP

DemP

Dem⬘

Spec

Dem⬘

Dem

Dem

ECCUMECCUM-

(I)STE (I)LLE

(EC)CU-E-STE (EC)CU-E-LLE

Old Hungarian can be assumed to have two homophonous structures, one for definite determination and one for demonstrative modification, as was presented in (16). If one looks at present-day Hungarian, it becomes obvious that the two constructions diverged in a different way than they did in Romance. The unmarked and most commonly used constructions with demonstrative modifiers involve determiner doubling, which suggests that demonstrative modification has been renewed by readding a deictic element to the phrase headed by the article. Nevertheless, this grammaticalization path was not smooth and linear: more than one strategy was promptly available from the beginning and what happened in reality is only a shift in proportions between the possible constructions rather than one following the other. In Old Hungarian, two pairs of demonstratives can be found, namely ezen/azon ‘this/that’ and imez/amaz ‘this/that’, which probably developed from simple demonstrative pronouns by a process of reinforcement, similarly to what has been presented in (17). Before going into the details, four examples are provided to illustrate the use of the individual morphemes. (18) Azert mert èzen èmber harom nèuèckèl nèuèztètic that-for because this man three name-pl-ins be.named-3sg ‘Because this man is called by three names’ (Vienna C. 234) (19) Azon idoben az kÿralnak uduaraba uala e´g uitez ´ that time-ine the king-dat court-poss-ill be-past one hero ‘In that time there was a hero in the king’s court’ (Kazinczy C. 45r) 14 The individual demonstrative systems of modern Romance languages vary considerably in several respects, so () should only be considered as a schematic representation to model how certain demonstratives developed in these languages.



Barbara Egedi

(20) veeseetok óketh imez palotaaban ´ throw-sbjv-2pl them this palace-ine ‘throw them into this palace’

(Sándor C. 12v)

(21) tawol legen te telled hogy amaz embert zeressed far be-sbjv-3sg you abl-2sg that that man-acc love-sbjv-2sg kyt míndenbe ellensegesth vallaz who-acc everything-ill enemy-like have-2sg ‘Beware of loving that man who is your enemy in everything’ (Lobkowicz C. 297) As for the morphemes ezen/azon ‘this/that’, neither of the standard etymological dictionaries of Hungarian15 can provide precise information as to how these demonstratives were created. The morpheme ezen/azon ‘this/that’ is claimed to have been formed by augmenting the basic ez with a pronominal derivational suffix -n (EWUng 62 and 345, TESz I. 819 also adds that this derivational suffix has an emphasizing role). The original function of this reinforced pronoun (and pronominal modifier) must have been strong identification, which is indicated by the fact that in the early Jókai Codex, out of 15 occurrences of azon, 13 undoubtedly mean ‘the same’ rather than simply ‘that’ and it is regularly used as the translation of the Latin identity marker idem. Its semantic property of identification, however, bleached quite rapidly and reduced to anaphoricity, as suggested by the emergence and use of ugyan ‘same’ in combination with pronouns and adverbs to express referential identity (e.g., ugyaz-azon, ugyan-az ‘the same’, ugyan-ott ‘the same place’ lit. ‘same-there’). According to the etymological dictionary (EWUng 32, 318, and 608), the morphemes imez/amaz ‘this/that’ are compositions formed by a syntactic fusion, in which the first member derives from íme and ám ‘look, behold’, and has a reinforcing function (Verstärkungselement).16 Note that the form imez ‘this’ is later replaced by eme(z) and only this newer form survives in standard Modern Hungarian, although forms beginning with em- are only attested from 1628 on. The morphemes ezen/azon ‘this/that’ are claimed to have already existed in ProtoHungarian (D. Mátai 2003: 217), while the compounds imez/amaz ‘this/that’ only emerged in Late Old Hungarian (G. Varga 1992: 489). Stipulations on their use before the written documents do not have much sense of course. What can be stated with certainty is that the latter pair is indeed a more recent formation, since ezen/azon are clearly attested in the Königsberg Fragment and Ribbons (see section 3.2.3 for the source), while no imez/amaz can be found either in the Early Old Hungarian records or in the first codices of Late Old Hungarian.

15 Their standard abbreviations are TESz (A magyar nyelv történeti etimológia szótára) and EWUng (Etymologischen Wörterbuch des Ungarischen), see Benkő (–) and Benkő (), respectively. 16 These latter reinforcing elements are themselves claimed to derive from more ancient reinforced demonstratives.

The DP-cycle in Hungarian and the functional extension of the noun phrase



In the grammatical descriptions cited, no concrete numerical data are given or even estimated as to the frequency of these reinforced determiners in the texts during the Old Hungarian period. As our research will demonstrate, their rate of recurrence is quite low at this time; they are almost exceptional, at least in comparison with the large mass of “original” demonstratives. In Table 3.5 the results of an automatic query are provided (carried out with the aim of exhaustive listing, thus also manually revised). The query involved ten already normalized codices of the Old Hungarian Corpus (comprising 175,540 tokens).17 Table .. Demonstrative determiners (including articles) in ten normalized codices of the Old Hungarian Corpus Codex

Date

Tokens

e/ez

a/az

ezen

azon

ime(z)

ama(z)

Jókai Codex Vienna Codex Birk Codex Festetics Codex Guary Codex Könyvecse Kazinczy Codex Czech Codex Bod Codex

a1370/c.1448 a1416/c.1450 1474 1492–1494 before 1495 1521 1526–1541 1513 first half of the 16th c. 1525

22,733 54,423 2,142 23,700 21,714 8,745 20,027 10,998 10,084

272 407 22 77 450 85 259 88 119

573 2,233 130 234 1,390 623 1,437 239 554

5 6 1 – 13 – 3 – –

15 32 6 11 2 2 12 – 4

– – – 1 – 5 – –

– – – – 3 1 3 – 2

974 175,540

3 1,782

38 7,451

– 28

– 84

– 6

– 9

Miskolc Fragment SUM

The problem mentioned in 3.2.4. shows up again, inasmuch as the pronominal uses of ez and az also turn up in the hits of the query, which slightly distorts the data. Luckily enough, only pronouns in nominative case are concerned. In the case of the reinforced demonstratives the problem could be solved, as the low number of the hits made it possible to sort out the pronominal occurrences manually; the table only includes the modifier uses as expected. To get a rough impression of how strongly the pronouns in nominative alter the proportions, another testing has been carried out in two codices that also have morphological annotations. According to this research, in the Guary Codex, out of all the occurrences of e/ez and a/az, only 6.21 can be analyzed as a pronoun instead of a determiner, while in Jókai Codex this proportion is even lower, namely 4.85. Accordingly, if we assume that the pronouns out of all the hits for e/ez and a/az in the above ten codices do not exceed a ratio of 5–6 in general, the

17 As the Székelyudvarhely Codex is only partially normalized, it has been ignored in this calculation. See the Appendix of this volume for more information about the structure of the Old Hungarian Corpus and the different text processing levels of the manuscripts.



Barbara Egedi

significant difference between the occurrences of old (e/ez, a/az) and new, reinforced demonstratives (ezen, azon, imez, amaz) remains nearly unaffected. The strikingly high number of a/az is, of course, due to the fact that this column comprises both the distal demonstrative modifiers and the articles. Furthermore, it can clearly be observed that among the reinforced determiners, azon ‘that’ was the most commonly used. It is quite understandable, since formal disambiguation was only needed in the case of constructions involving a distal demonstrative. What the overall picture shows us, however, is that the use of these reinforced demonstratives had not yet diffused. This situation will considerably change in Middle Hungarian, as data drawn from secondary literature suggests (see section 3.4.1). .. The renewal of the demonstrative: doubling The noun phrase construction involving a demonstrative modifier renewed in another way as well by the emergence of a new pattern during the 16th century (the beginning of the Middle Hungarian period). In this structure a sort of determiner doubling can be observed, viz. the simultaneous spelling out of the demonstrative and the article. This is the pattern that is commonly used in Modern Hungarian (see (12)). Did we thus arrive at the construction that is also used in the present-day language? The first impression suggests that this is the case: the new doubling configuration may be analyzed as the double filling of the functional projection for definiteness. The demonstrative is moved to and spelled out in the specifier position, while the determiner in article function is the head of the DP (22). DP

(22) DemP

D⬘ D

NP

az az that the ‘that door’

ajtó door

Demonstrative modifiers in these doubling constructions behave in a rather peculiar way in Middle as well as in Modern Hungarian, that is to say, they are not prototypical deictic determiners. Demonstratives co-occur with the definite article, and unlike other modifiers, they agree in number and case with the head noun. (23) a. ez-ek

az ajtó-k b. az-t az ajtó-t c. az-ok-at az ajtó-k-at this-pl the door-pl that-acc the door-acc that-pl-acc the door-pl-acc ‘these doors’ ‘that door’ ‘those doors’

Moreover, a doubly filled functional projection (DP in this case) is also atypical and, in a generative framework, considered to be against the basic economy principles

The DP-cycle in Hungarian and the functional extension of the noun phrase



of syntax. These facts would suggest that the demonstrative and the noun phrase beginning with the article are actually two juxtaposed DPs, which can be represented in the phrase structure as an adjunction structure. In the rest of this section it will be demonstrated that both analyses are justified: one will account for Modern Hungarian, while the other for the Middle Hungarian data. As regards Modern Hungarian, there are good reasons to assume that the demonstratives ez/az ‘this/that’ are in fact located in the specifier position of the DP projection, rather than being adjoined to it, even though the definite article is also spelled out in the head of the same phrase. Nothing can intervene between the demonstrative and the article, the two determiners also form a prosodic unit, and by dropping the article the construction becomes ungrammatical.18 Observing the historical data, however, we can conclude that the conditions were slightly different when this pattern arose in the Middle Hungarian period. In the 16th and 17th centuries, the newly added demonstrative appears to link more loosely with the noun phrase than it would do today. Demonstratives can be separated from the rest of the construction by a variety of elements, e.g., conjunctions, various particles, sometimes by the verb itself. In these cases, the sequence of the demonstrative and the definite article is evidently broken. As has been mentioned earlier, the digitalization of Middle Hungarian records has just begun, so the following data are all drawn from the descriptive grammar of Simonyi (1914: 122–123). The examples come from the first half of the 17th century, showing how the sequence of the determiners is interrupted, for instance, by a scalar particle (24a), by an interrogative particle (25a), or by a modal verb and a conjunction (26a). For explicitness, (24b)–(26b) demonstrate how these phrases would look in Modern Hungarian so that the difference between the two word orders can clearly be observed.19 az esküvések (24) a. azok is those also the swearings ‘also those swearings’ (From the letters of Péter Pázmány)

b. azok az esküvések those the swearings ‘also those swearings’

is also

(25) a. Mit választasz inkább? ezt-e a pénzt? what choose-you rather? this.acc-q the money-acc ‘What do you prefer to choose? This money?’ (Káldi Preachings) b. ezt a pénzt-e this.acc the money.acc-q 18 Hereby we follow the analysis of Huba Bartos (, ), also adapted by É. Kiss (); however, the placement of the demonstrative modifier in the specifier of the DP was first suggested by Kenesei (: ). 19 The peculiarities of the Middle Hungarian demonstrative are the central topic of a closely related paper (Egedi, forthcoming). In that study, my aim was to account for the word order variation at the nominal left periphery, while here the main focus is on the renewal of the demonstrative system. However, as the two complement each other, the data and the analyses necessarily overlap and some arguments are repeated.



Barbara Egedi

It is also remarkable that in Middle Hungarian the combination of this determiner doubling construction with dative-marked possessors may result in a word order (26a) different from what would be grammatical in Modern Hungarian (26b). (26) a. ar-rul is az bibliá-nak rész-é-rűl that-del also the Bible-dat part-poss-del ‘also about that part of the Bible’

(Péter Pázmány)

b. a Bibliá-nak ar-ról a rész-é-ről is the Bible-dat that-del the part-poss-del also It is exactly the combination of possessive constructions with demonstratives that indicates how the structure of determiner doubling changed in time. Hungarian has (and has always had in its documented history) two possessive constructions. In one of them the possessor is dative-marked, in the other the possessor is unmarked. Although the dative marked possessor can be extracted from the noun phrase, or else, generated outside of the noun phrase as an external possessor (for the conditions of the possible configurations in Modern Hungarian, see É. Kiss, forthcoming), the two possessive constructions show no difference in meaning. In Old Hungarian, if the dativemarked possessor is generated internally, it seems to end up in the same position as the unmarked form, that is, in the specifier of the DP. Both constructions exclude any further determination and do not combine with demonstratives that also target Spec,DP. What the Middle Hungarian data suggest altogether is that the determiner doubling construction is the outcome of an adjunction operation, where the demonstrative pronoun as a whole adjoins to the DP headed by an article, thus the structure differs from the one proposed in (22): DP

(27) DP

DP D

az az that the ‘that part’

NP rész part

That is why the combination of Middle Hungarian demonstratives with possessive constructions results in the word order presented in (26a). To account for the reverse word order of the nominal left periphery in Modern Hungarian, according to which dative-marked possessors always have to precede the demonstrative modifiers (26b), it is plausible to assume that the adjoined determiner phrase containing the demonstrative has again been integrated into the DP domain. Note that this is in accordance with Elly van Gelderen’s (2008: 250) third universal economy principle.

The DP-cycle in Hungarian and the functional extension of the noun phrase



The so-called “Specifier Incorporation” states that elements coming from outside tend to be a specifier rather than an adjunct. The leftmost adjunction site of the DP was open for both dative-marked possessors and demonstratives, but the fact that demonstratives were chosen for incorporation might have been facilitated by the already present patterns with extracted and external possessors. At a later point of the process, the adjunction of the regularly appearing dative-marked possessors may have been reanalyzed as an internal topicalization, producing a new layer of the extended noun phrase—in line with what É. Kiss (2000) proposes for present-day Hungarian. TopP

(28) Spec

DP Spec

a Bibliának-naki az the Bible-DAT that ‘that part of the Bible’

D⬘ D

PossP

a the

rész-e ti part-POSS

This chain of changes finally led not only to the present-day word order on the nominal left periphery but also to the rather atypical behavior of the demonstratives, viz. their appearance in a doubly filled DP projection and their somewhat exceptional morphology of being marked for case and plurality, in spite of their modifier status. It is to be noted that although the use of this doubling construction is the primary means of expressing demonstrative modification in present-day Hungarian, it practically does not appear before the Middle Hungarian period. As regards its spreading, the next section will mostly rely on the figures provided in a recent paper by Adrienne Dömötör (2008). However, for lack of a considerable large digitized corpus similar to the Old Hungarian Corpus, it is impossible to determine at this point of the research when the above proposed structural reintegration of demonstratives took place. (Manual search for such constructions is a hopeless task considering that combination of demonstratives and possessives is an overall infrequent phenomenon for obvious semantic reasons.)

. Competing strategies Following the grammaticalization of the definite article, the demonstrative system has been renewed at least by two kinds of strategies in Hungarian. On the one hand, two pairs of reinforced demonstratives could be identified, one of them appearing already in the Early Old Hungarian period. On the other hand, a new pattern also emerged at



Barbara Egedi

the beginning of the Middle Hungarian period, in which an additional demonstrative adjoined to the noun phrase headed by the definite article. As will be shown in the remainder of this chapter, the original strategy has also been preserved and is still operative to some extent. The new strategies emerged gradually, reinforcement being the first in Old Hungarian, followed by determiner doubling in Middle Hungarian. Their upcoming success cannot be presumed on the basis of the digitized Old Hungarian Corpus, as the crucial period with respect to their spreading seems to be the Middle Hungarian stage. Importing Middle Hungarian records into our corpus has just begun and, consequently, no quantitative data can be provided for the distribution and spreading of the various constructions involving demonstrative modification. Luckily enough, Dömötör (2008) addressed the question of these constructions, and her research can serve as a good basis for our investigations. The corpus, as she informed me in a personal communication, was not a digital one, and includes text excerpts from various types of genres. It contains 200,000 characters per century, which means that the corpus amounts to approximately 100,000 tokens altogether. .. Data from Middle Hungarian The point of view according to which Dömötör (2008) arranged and interpreted her data is different from our own: she was more interested in the opposition of agreeing vs. non-agreeing demonstratives and examined the data in this respect: whether or not there is (case and number) agreement on the demonstrative modifier. Table .. The proportion of constructions with demonstrative modifiers (Figures from Dömötör 2008:24)20

Agreeing Dem + Art + N Non-agreeing Dem + N

16th c.

17th c.

18th c.

For the entire period

5.9 94.1

19.6 80.4

34.9 65.1

20.3 79.7

Table 3.6 shows us quite explicitly how the determiner doubling construction (“Agreeing Dem + Art + N” in Dömötör’s table) consistently spread during these centuries. Note, however, that there is a very important piece of information hidden in this table. The rising percentage does not only contrast with the old constructions (ez/az N). The lower line of the table also includes the nouns modified by ezen, imez, etc., that is, all the demonstratives—original and reinforced—that show no agreement when used as a modifier. An additional table not cited here suggests that the frequency of

20 The table, originally compiled in Hungarian, is not only translated here but also slightly modified for convenience.

The DP-cycle in Hungarian and the functional extension of the noun phrase



the pair ezen/azon ‘this/that’ also significantly increased, but precise data concerning the proportion of reinforced demonstratives with respect to the original constructions and to determiner doubling cannot be obtained on the basis of Dömötör’s figures. There is another very interesting phenomenon that can be observed in connection with the doubling construction. The spread of the pattern does not show the same rate in the case of the proximal demonstrative as in the case of the distal one. As the figures in Table 3.7 indicate, spreading of the doubling constructions with proximal ez is definitely slower than that with the distal demonstrative, and the constructions are far less frequently attested if the entire period is considered. It is highly reasonable to assume that this difference of percentage can be due to the fact that the proximal demonstrative ez differs from the definite article in its form, causing no ambiguity at all. Table .. The spread of the constructions ‘ez/az az N’ (Figures from Dömötör 2008: 24–25) 16th c.

17th c.

18th c.

For the entire period

‘az az N’ ‘az N’

54.5 45.5

91.7 8.3

92.6 7.4

85.5 14.5

‘ez az N’ ‘ez N’

16th c. 3.8 96.2

17th c. 21.5 78.5

18th c. 40.8 59.2

For the entire period 20.4 79.6

As has been mentioned, Dömötör’s calculation has been built on a representative corpus of considerable size, but still, this collection of data cannot be large enough, due to the limits of manual work. Hopefully, by the extension and improvement of our digitized corpus, the percentages in Table 3.7 will be either more justified or will need to be modified. Furthermore, the proportional spreading of reinforced determiners should also be established. It would be really valuable to understand the mechanism of such competing strategies: what prevents a syntactic strategy (reinforcement) from triumphing over another one (doubling) in spite of the fact that it emerged much earlier and seems to have perfectly accomplished the task, viz. the disambiguation between demonstrative modification and simple definiteness. What made the younger, determiner doubling strategy become the winner? .. Modern Hungarian distribution of demonstrative modifiers In Modern Hungarian, all the constructions involving demonstrative modifiers that have been discussed so far are still available, but in a completely different proportion than in Old or Middle Hungarian. The only construction that has been lost is the type az ajtó ‘that door’, which is exclusively used to express simple definiteness today



Barbara Egedi

(‘the door’).21 In what follows, the uses and properties of the constructions in Modern Hungarian will be summarized and classified into two different groups. I. Determiner doubling construction: ez/az a vélemény ‘this/that opinion’ This is the most common type in Modern Hungarian, the unmarked construction to express demonstrative modification. The demonstrative modifier within the construction (ez/az ‘this/that’) is morphologically inflected for number and case in agreement with the lexical noun (29). Moreover, postpositions that do not take an oblique case are also repeated on them (30). (29) Nem ismerjük ez-ek-et a vélemény-ek-et. not know-1pl this-pl-acc the opinion-pl-acc ‘We do not know these opinions.’ (30) Nem aggódunk ez-ek miatt a vélemény-ek miatt. not worry-1pl this-pl about the opinion-pl about ‘We do not worry about these opinions.’ Spelling out the article is obligatory in this construction, and the two determiners (the demonstrative modifier and the definite article) are strictly juxtaposed. The fixed position of the demonstrative in the specifier of the DP is very likely, since the doubling construction cannot be combined with a nominative/unmarked possessor. Unmarked possessor expressions are assumed to occupy the same structural position, i.e., the specifier of the DP (see É. Kiss 2000: 134).22 (31) a. a miniszter the minister ‘the minister’s

vélemény-e opinion-poss opinion’

+ + +

ez a vélemény this the opinion ‘this opinion’

⇒

b. ∗ a miniszter ez a vélemény-e the minister this the opinion-poss c. a miniszter-nek ez a vélemény-e the minister-dat this the opinion-poss ‘this opinion of the minister’s’

21 Exceptions can be found, of course: the old demonstrative meaning of az survived in a few temporal expressions, such as az-nap ‘that day’ (instead of ‘the day’), and in other fixed expressions, e.g., a tekintet-ben ‘in that respect’ (instead of ‘in the respect’). 22 There are two types of possessor expressions in Hungarian: a dative-marked and a nominative-marked (or preferably unmarked), with no actual difference in meaning. They only differ in one respect, viz. the syntactic position of the possessor expression. With dative-marked possessors we have a more extended DP as dative-marked possessors co-occur with the definite article and they can even be extracted (see Szabolcsi : –, É. Kiss ). In both possessive constructions the head noun is always marked for being possessed by a suffix.

The DP-cycle in Hungarian and the functional extension of the noun phrase



The construction has been shown to have developed in the Middle Hungarian period and to have spread only gradually until it became dominant today. When the construction emerged, the demonstrative behaved more like the host of an appositive lexical DP, with the lexical DP adjoined to it. This structure resulted in word orders that would be ill-formed in present-day Hungarian with demonstratives preceding the possessor expression (see (26a) above). II. A stylistically marked type: e/ezen/azon/eme/ama vélemény ‘this/that opinion’ This type is stylistically marked in Modern Hungarian (official, mannered even), or sounds rather archaic, and is completely missing in colloquial registers. In this pattern, the demonstrative determiner (e/eme/ezen ‘this’ or ama23 /azon ‘that’) and the article mutually exclude each other (32). There is no plural and case marking on the demonstrative (33), and postpositions are not repeated either (34). (32)

∗ e/ezen/azon/eme/ama

this/that

a vélemény the opinion

(33) Nem ismerjük e/ezen/azon/eme/ama vélemény-ek-et. not know-1pl this/that opinion-pl-acc ‘We do not know these/those opinions.’ (34) Nem aggódunk e/ezen/azon/eme/ama vélemény-ek miatt. not worry-1pl this/that opinion-pl about ‘We do not worry about these/those opinions.’ These constructions can freely be combined with both the unmarked and the dativemarked possessors. (35) a. a miniszter e/ezen/azon/eme/ama vélemény-e the minister this/that opinion-poss ‘this opinion of the minister’s’ b. a miniszternek e/ezen/azon/eme/ama vélemény-e the minister-dat this/that opinion-poss ‘this opinion of the minister’s’ To account for all these properties, it must be assumed that the position of these archaic demonstratives is lower in the construction and they are only raised to D optionally. Anna Szabolcsi (1994b) has argued that Hungarian determiners fall into two distinct categories syntactically. Only the article belongs to the category D, the others originate in a lower DET position. Indeed, these archaic demonstratives seem to occupy the same position as the determiner quantifiers in (36a).

23 The longer forms ending in -z (emez/amaz) are only used as pronouns, as far as standard Modern Hungarian is concerned. Dialectal variation, of course, cannot be excluded.



Barbara Egedi

(36) a. a hercegnő minden/melyik/valamennyi/egyik kívánság-a the princess every/which/all/one+spec wish-poss ‘all/which/one of the princess’ wishes’ b. a hercegnő e/eme/ezen kívánság-a the princess this wish-poss ‘this wish of the princess’ The low position of the determiners can also be observed when they are combined with a pronominal possessor (37a). In this case a phrase-initial article also appears, but this can also be due to the fact that spelling out the personal pronoun (only used for contrast or emphasis) is impossible without the article. The same construction with a null pronominal is not acceptable (37b). e/eme/ezen kívánság-a-i-tok (37) a. a ti wish-poss-pl-2pl the youPL this ‘these wishes of yours’ b. ∗ a e/eme/ezen kívánság-a-i-tok the this wish-poss-pl-2pl It can be seen that these stylistically marked patterns derive from the Old and Middle Hungarian constructions involving a reinforced demonstrative. Unlike ez/az ‘this/that’ of the doubling constructions, which first appeared as adjoined constituents, these demonstratives are base-generated below the DP and may as well remain in situ. On the basis of its similar behavior today, the demonstrative e ‘this’ in its very short form can be grouped together with the reinforced derivatives, but it must be pointed out that e ‘this’ is the successor of the oldest constructions of the type e(z)/a(z) ajtó ‘this/that door’ and was only able to survive because of its unambiguous shape. In contrast, its distal pair a(z) ‘that’ did not survive as a demonstrative modifier but transformed to encode the definite article. .. Some dialectal peculiarities: the complete integration of demonstratives As reported in some dialectal grammatical descriptions, plural agreement in doubling constructions may be absent in certain varieties of Hungarian. The data on the left (marked by a letter a) show the dialectal records, while the corresponding examples on the right (b) indicate how these phrases would be formed in standard Modern Hungarian. The phenomenon has been attested in the north-eastern dialectal region of the country (Kálmán 1966: 85): a (38) a. ez-t this-acc the ‘these houses’

ház-ak-at house-pl-acc

b. ez-ek-et a this-pl-acc the ‘these houses’

ház-ak-at house-pl-acc

The DP-cycle in Hungarian and the functional extension of the noun phrase



This pattern is also characteristic in the so called Palóc dialect (Király 1991: 538; É. Kiss 2006: 530): (39) a. Haggyuk ez-t a gyerek-ek-et táncolni! let.sbjv.1pl this-acc the child-pl-acc dance-inf ‘Let these children dance!’ b. Hagyjuk ez-ek-et a gyerek-ek-et táncolni! let.sbjv.1pl this-pl-acc the child-pl-acc dance-inf ‘Let these children dance!’ Moreover, in a certain isolated dialect spoken in the village of Domokos (situated in the historic Szolnok-Doboka region, today’s Dămăcuşeni in Northern Romania) both plural agreement and case marking have been lost on the demonstrative, which may be a clear sign of a more perfect integration of the modifier into the noun phrase structure (the data are cited from Németh 1913: 67): az ember-ek nem jó hâré (40) a. az that the man-pl not good place-sub ‘those men are not going to the right place’ b. az-ok a ember-ek that-pl the man-pl ‘those men’ a gyermek-nek (41) a. add az give that the child-dat ‘Give (it) to that child.’

mennek go-3pl

b. add an-nak a gyermek-nek give that-dat the child-dat ‘Give (it) to that child.’

These isolated records, collected from secondary literature and dating to a period about a century ago, are only cited to show a possible step in the DP-cycle in Hungarian. In these dialects, a more advanced incorporation of the demonstrative modifier can be observed. In the dialect of Domokos, the two determiners seem to have completely fused, and as far as I see there is no point in analyzing them as two separate elements.

. Summary In this chapter several stages of the DP-cycle have been discussed as well as their structural consequences in Hungarian. It has been presented and analyzed how the marking of definiteness changed in time and how the emergence of the definite article and its functional spreading reshaped the nominal left periphery from the first written sources to the present-day. It has also been shown that the renewal of the demonstrative system involved at least two types of strategies in Hungarian, reinforcement as well as determiner doubling,



Barbara Egedi

whereas the old construction is also preserved to some extent. The reinforcing strategy produced two new series of demonstrative pronouns, while the determiner doubling phenomenon entailed a long-term structural reanalysis at the nominal left periphery. The word order change that can be observed between Middle and Modern Hungarian, in the cases where demonstratives are combined with dative-marked possessor expressions, indicates that demonstratives first linked with the noun phrase headed by the definite article by means of adjunction, but later were incorporated into the DP-internal specifier position. The competing strategies of demonstrative modification coexisted from the Middle Hungarian period onward and their descendant constructions survive in presentday Hungarian as well, though they are used with a different degree of markedness. Dialectal data were also considered to show that in a few varieties of Hungarian a further step of demonstrative integration can also be observed.

 From A-quantification to D-quantification: Universal quantifiers in the sentence and in the noun phrase Á G N E S B E N D E - FA R K A S

. Introduction In this chapter we discuss the syntax and semantics of two Old Hungarian quantifiers expressing maximality or universality, mind ‘all’ and minden ‘every’, ‘everything’, ‘everyone’. The main focus will be on the clearly delineated quantificational strategies embodied by these two expressions. Roughly, mind resembles English all or French tous, toutes, whereas minden is more like every. Beside providing a semantic analysis, we also investigate the impact of the presence of determiners like minden on the left periphery of the sentence and (to a lesser extent) the internal structure of the DP. We also reconstruct a possible scenario according to which the expression of quantification could have changed in Hungarian, from the prevalence of A-quantification, the long-distance (quasi-)binding of indeterminate pronouns, to a quantificational system in which determiners and quantifying DPs bind individual variables locally. We also propose that today’s peculiar distributivity operator, the reduplicated pronoun ki-ki (lit. ‘who-who’) is a residue of that earlier stage. This chapter is structured as follows: Section 4.2 discusses data from the earliest Old Hungarian manuscript, the Funeral Sermon and Prayer, which only contained mind ‘all’ (4.2.1); subsection 4.2.2 presents data from the first codex in which minden has been attested; subsection 4.2.3 provides a type-driven reconstruction of the change in category and type undergone by minden ‘every’. Section 4.3 discusses the two strategies represented by mind and minden, detailing syntactic and semantic differences. Section 4.4 discusses the distributive reduplicated pronoun ki-ki lit. ‘who-who’: Part 4.4.1 presents some (sporadic) examples of indeterminate pronouns, proposing that Proto-



Ágnes Bende-Farkas

Hungarian had long-distance operators binding the alternatives introduced by such pronouns. Part 4.4.2 discusses the pronominal construction itself. Part 4.5 examines the consequences of the presence of D-quantification, and concludes the chapter.

. From all to every From a logician’s point of view (or, indeed, from the point of view of anyone who has taken a logic course) the prototypical quantifiers in natural language are DPs or determiners like everyone and every. This assumption of prototypicality may have been a pretheoretical motivating factor behind the universal proposed in Barwise and Cooper (1981): according to this universal, all languages have quantifying NPs. This generalization is untenable, as shown, for instance, in work reported in Bach et al. (1995). Many languages employ so-called A-quantification instead of what has been dubbed by Barbara Partee as D-quantification (determiner quantification). The term ‘A-quantification’ stands for quantification by means of adverbs, affixes, argumentstructure adjusters, special distributivity operators, or other categories. Old Hungarian, where the expression of universal quantification is concerned, can be seen as a period when D-quantification (in the form of the determiners/DPs like minden ‘every’) supplemented the inventory of adverbs, indeterminate pronouns and affixes. (In fact, the examples in this chapter contain some A-quantifiers besides mind ‘all’, such as egyenként and fejenként ‘one by one’ in (43), in part 4.3.2) The role played by minden deserves special attention, for the following reason: in Old Hungarian codices minden and its derivatives were the sole representatives of strong, quantifying DPs. The first books written entirely in Hungarian (the Jókai, Vienna and Munich codices) contain an inventory of determiners, but, apart from minden (and its partitive or hyper-distributive variants, such as egyminden(-ik) ‘each and every one (of them)’), these are weak determiners. (A legtöbb ‘most’, for instance, was not present.) The quantificational nature of weak determiners has been disputed since Milsark (1977) onwards, so they do not provide an optimal testing ground for the study of the appearance and stabilization of D-quantification. What remains, then, is the study of minden. In this chapter minden will be discussed in tandem with its floating counterpart, mind ‘all’. Fortunately, mind is an A-quantifier, so the contrast between mind and minden can be taken as typical of these two modes of quantification. .. In the beginning there was mind The earliest surviving Hungarian text, the Funeral Sermon and Prayer (ca. 1195) contains one expression conveying universal quantification. This is mend (roughly, ‘all’). A first example is (1): (1) Es vimagguc mend szentucut. and worship-sbjv-1pl all saint-pl-acc ‘and let us pray to all the saints’

From A-quantification to D-quantification



The standard view on the morphology of mend (according to the the Hungarian Etymological Dictionary, Benkő (1993)) is that it is composed of the (indeterminate) inanimate pronoun mi ‘what’ and the suffixes -n-d. The suffix -n is presumably a manner suffix; the role of -d is unclear. We propose that the semantic contribution of Old Hungarian mind can be derived from the contribution of its component morphemes: mind can be analyzed as a maximality operator where the indeterminate pronoun mi contributes alternatives (variables) unspecified for number, and the suffix complex -nd contributes a functor which takes these alternatives as input and returns the entire set (or the greatest individual “built” from the members of that set). Mind resembles expressions from other languages that convey the meaning ‘all’, in that it could operate either on (all) the members of some collection, or on the entire collection itself (see Haspelmath 1995). What sets mind apart is that it was not an open class lexical entry with an acquired logical function; rather, its operator status was quite transparent and tailor-made, as it were.1, 2 In the Funeral Sermon and Prayer mend exhibits properties typical of its use in Old Hungarian: Mend could be a floating quantifier attached to the left of the main predicate; in (2) its associate is covert. (The term ‘floating quantifier’ will be used neutrally, without assuming movement for mind. In the context of this chapter ‘floating quantifier’ is used to describe an expression that can be interpreted as a quantifier, and which does not form a—minimal—surface constituent, e.g. a DP, with its associate.) (2) mend ozchuz iarov vogmuc all that-all walk-part be-1pl ‘All of us are to walk there (to our graves)’ ‘We are all to walk there’ Mend preceded (nominative) pronominal possessors, as seen in (3). In examples from later texts it will be seen that it preceded the definite article as well. It can be seen in later texts (in which the definite article is more frequently and consistently used) that mend associated with definite expressions. In the case of example (4) it could be argued that the property expressed by the participle helped establish a definite or at least a delimited collection, especially if the description was “anchored” by the proper name paradisum ‘Eden’. (3) bulsassa mend w bunet forgive-sbjv.3sg all he sin-poss.3sg-acc ‘May He forgive all his sins’ 1

One could say that what underwent reanalysis with mind was the adverbial suffix -n. The lexical entry for ‘entire’, ‘complete’—teljes—coexisted with mind, occasionally replacing it, without any apparent logical overtones. 2



Ágnes Bende-Farkas

(4) mend paradisumben uolov gimilcíctul munda nekí elnie all Eden-ine be-part fruit-pl-abl told dat-3.sg live-inf.3sg ‘He told him to partake of all fruit from Eden’ Mind could also mean ‘entire’, when its associate was, for instance, a collective expression. Thus in (5) mend operates on a collection and not on the individual members of that collection. (5) Num heon muganec. ge mend w foianec halalut evec. not only self-dat but all he kin-poss.3sg-dat death-acc eat-pst ‘(In the forbidden fruit) he ate death, not only for himself but for all his kin’ Mind could be used with non-count nominals (mass terms or abstract nouns), as shown in (6a), where it precedes the abstract noun (phrase) iov ‘good’. It was also compatible with readings that were neither fully collective, nor did they imply distribution over the atoms of a collection. This can be seen in (6b). (6) a. es oggun neki munhí uruzagbele utot. es mend and give-sbjv-3sg dat-3.sg heaven-ly country-ill way-acc and all iovben rezet. good-ine part-acc ‘and may He grant him passage to heaven and a share in everything that is good’ b. mend vv scentíí es unuttei cuzicun iov felevl all he saint-poss.pl.3sg and chosen-poss.pl.3g among right from iochtotnia ílezie vvt. arrive-caus-inf-3sg resurrect-sbjv.3pl he-acc ‘May He resurrect him to be sent to the right of God, among all His saints and His anointed’ In (6b) mind behaves in a manner that can be best paraphrased with colloquial Modern English all: (7) a. John was standing among all those students. b. ???John was standing among every student. In this subsection we have seen that according to the earliest data mind was a floating quantifier whose associate could remain covert; when overt, its associate was definite. Prenominal mind preceded the nominative possessor. Where the interpretation of mind is concerned, it could operate on the collection contributed by its associate, without necessarily accessing the members of that collection. It could also associate with abstract nouns, and it could also access non-atomic parts of a collection. It will be seen in section 4.3 that these are among the relevant properties of the mind-strategy as opposed to the minden-strategy.

From A-quantification to D-quantification



.. D-quantification: minden In this part minden ‘every’ is introduced, as attested in Old Hungarian manuscripts written after the Funeral Sermon and Prayer. According to the Etymological Dictionary of Hungarian (Benkő 1993), the final suffix in minden is a manner or cardinality suffix, presumably the same suffix that contributed to the derivation of mind from the pronoun mi ‘what’. That is, -n can be the suffix responsible for the derivation of manner and degree adverbs, and what might be termed as predicative numerals (the term is our own): (8) a. Mari szép-en énekelt Mary beautiful-ly sang ‘Mary sang beautifully’ b. Macbeth teljes-en megőrült Macbeth complete-ly prt-go.mad-pst.3sg ‘Macbeth went completely mad’ c. Egy kosárcsapatban öt-en vannak one basket-team-ine five-ly be-3pl ‘A basketball team has five members’ Lit.: ‘In a basketball team there are five’ The first book written entirely in Hungarian, the Jókai Codex (translated and copied around 1448), contains, beside mind, today’s Hungarian equivalent of every (MH minden, OH menden, mynden, etc.). In fact, the Jókai Codex contains an entire system of determiners and quantifying DPs. The following is a sample of sentences with D-quantifiers from the Jókai Codex. These contain valamy ‘something’, nemy ‘some, (a) certain’, minden, and an A-quantifier, the reduplicated numeral ketten-ketten ‘in twos’. (9) a. belmenuen varasba ezkeppen mezeytelenewl valamyt into-go-part town-ill this-like naked-ly something-acc predicaly neppeknek preach-imp.sg people-dat ‘when you go into town preach something to the people, naked as you are’ (Jókai C. 56–57) b. Es nemy zakadozt gyekenek ualanak alattak and some tattered rushes were under-3pl ‘And they had some tattered straw mats under them’

(Jókai C. 86)

ew tanoytuanyt ketten c. ewduezeytew ysten boczata redeem-part god send-pst.3sg he disciple-poss.pl.3sg-acc two-adv ketten menden varasba es helyre two-adv every town-ill and estate-sub ‘God the redeemer sent out his disciples in twos, to every town and estate’ (Jókai C. 128)



Ágnes Bende-Farkas

The Jókai Codex does not contain merely a few isolated examples of minden: the text exhibits the clearly and consistently diverging strategies of mind versus minden. In addition, the Jókai Codex contains a host of expressions (mostly adverbs) derived from minden, e.g. myndenestewl ‘entirely’, myndenkor ‘at all times’, or myndenhato ‘omnipotent’. The large number and frequent occurrence of these derivatives of minden, together with the consistency of what may be called the minden-strategy of quantification, indicate that minden was not a new addition to the language at the time when the Jókai Codex was written. From the Jókai Codex onward, mind and minden exhibit two distinct strategies, roughly corresponding to the contrast between English ‘all’ and ‘every’. What is remarkable in the Hungarian case is that these two strategies have been so clearly and persistently distinguished, from the middle of the Old Hungarian period onward, even though the representatives of the two strategies are morphologically related.3 .. From mind to minden: type-driven change This subsection contains a reconstruction of the change that yielded the determiner and quantifying DP minden. This is a type-driven analysis, which could be taken as a diachronic application of the type shifting principles of formal semantics (Partee 1987, Partee and Rooth 1983, Winter 2001; see also Chapter 7 in Eckardt (2006) for an application to historical formal semantics): transitions from one type to another will be judged according to their appropriateness with respect to a set of principled and regular changes in semantic type, accepted by the semantics community. Typetheoretic considerations will be complemented with empirical arguments. It is assumed that Old Hungarian mind was a modifier of type τ , τ  instantiated as the VP-modifier e, t, e, t. Today’s minden is taken to be a determiner of type e, t, e, t, t or a quantifying DP of type e, t, t. (Old Hungarian minden could in fact also mean ‘everybody’. This will be ignored here: Old Hungarian DP-minden is taken to be unspecified with respect to animacy.) The question that arises is whether minden was first a DP that came to be used as determiner as well, or whether it evolved from a determiner into a quantificational DP. We propose that minden was originally ambiguous between an adverb and an adjective, with the lexical meaning ‘entire(ly), full, fully’. This entails that initially it had the same type as mind, the chief difference being that minden could be attributive. (And, presumably, minden had access to the atomic parts of a collection.) An indirect argument is that a type change from the modifier type e, t, e, t to the quantifier type e, t, t is highly implausible. It is more in line with the principles of type change for an expression with the modifier type e, t, e, t to become reanalyzed as a determiner of type e, t, e, t, t. 3 Taken separately, neither mind nor minden have remained unchanged. What is unchanged is that mind is like floating ‘all’, and minden is like ‘every’, ‘every one’.

From A-quantification to D-quantification



There is also empirical evidence to indicate that minden could first have been a modifier with the lexical meaning ‘entire(ly)’: (i) There are cases from Old Hungarian codices where minden combined with a head noun like föld ‘earth, land’ (as in examples (37) or (38) in 4.3.2). (ii) In Eastern dialects of Hungarian the adjective egész ‘entire, whole’ has become a determiner meaning ‘every’ when preceded by the definite article; similarly, az egészen (‘the whole-n’) is like adverbial ‘all’. The following is an example in the Csángó regional variant of Hungarian, spoken in Moldavia (Romania).4 (10) Az egész politikus szereti a pénzt the whole politician like-3sg the money-acc ‘All politicians like money’ (iii) There is one example in the codices where minden modifies a predicative adjective: (11) ez velagon zegen legy evrevmest. es menyorzagban legy this world-sup poor be-imp.sg gladly and heaven-ine be-imp.sg menden kazdag. every rich ‘In this world be poor gladly, and in heaven be all-rich (full of riches)’ (Cornides C. 81v) The expression menden kazdag (lit. ‘every rich’) in (11) can only be interpreted as a predicate, and the sentence is definitely not an identity statement (‘You are to be every rich person’). This is reinforced by the contrast between the two sentences: the addressee is instructed to be poor and to rejoice in poverty in this world, and is promised to be full of riches in heaven. If menden kazdag is a predicate, it follows that menden in this example is an adverb, which is as it should be, since Modern Hungarian minden, just like English every cannot readily be used predicatively.5 The adverbial use of menden in this example is taken to be indicative of an intermediate stage in the process that yielded the quantificational determiner minden. Finally, there is a piece of indirect evidence suggesting that minden could be interpreted as a quantifying DP after it had become a DP-internal modifier. The contribution of the suffix -en has been consistently adverbial, yielding adverbs of manner or degree, and predicative numerals, as seen in the Modern Hungarian sentences from (8), or in Old Hungarian ketten-ketten ‘in twos’ from (9c). Given the invariable adverbial contribution of the suffix -en, it would be ad hoc to assume that in the one case of mind, it yielded a DP. If the direct change from adverb to quantifying DP is considered ad hoc, the change from DP to determiner appears to be even more arbitrary: Such a change would mean 4 Sentence () is from our informants at the Workshop on Csángó Syntax held at the Research Institute of Linguistics of the Hungarian Academy of Sciences in March . 5 () is in fact the only example of this kind that we could find in Old Hungarian codices.



Ágnes Bende-Farkas

that a quantifier Q of type e, t, t, which would need only a predicate to form a complete sentence, would suddenly need two arguments (having shifted to the type e, t, e, t, t). We propose therefore that it was the determiner minden that could be used as a quantifying DP: an expression of type e, t, e, t, t could also be used as a quantifier of type e, t, t, simply by saturating its first argument with, say, the universal predicate ‘thing’.6

. Quantificational strategies In this section we present and discuss the two quantificational strategies embodied by mind ‘all’ and ‘every’. The two expressions will be discussed in tandem, against a checklist of syntactic and semantic properties (in subsections 4.3.1 and in 4.3.2, respectively). .. Syntax ... Syntactic category In a nutshell, mind is analyzed as an adverbial, whereas minden could be a determiner of a full quantifying DP. This difference in syntactic category has clear implications for quantificational properties: as an adverbial, mind was generated in situ, and this determined its scopal properties once and for all (if it entered scope interactions). Minden, or the DP containing it, could move, overtly or covertly; hence, it had considerably more freedom where scope interactions were concerned. Since the categorial status of minden is uncontroversial, the bulk of this subsection will be devoted to arguments and empirical evidence supporting the adverbial analysis of mind. Morphology: inflection, agreement Mind was not inflected for case, not even when it had no overt associate, and it (more exactly, its covert associate) had a grammatical function that came with morphological marking. (Mind quite often associated with demonstratives, which could then be affixed when necessary.) When minden was a full DP it could be inflected for case and (quite often) for number. Minden as a full DP could fill an argument position in the sentence, whereas mind on its own could not.7 The examples in (12) show minden with the accusative suffix -t. (The examples under (14) also show uninflected mind, in addition to the phenomenon for which they were originally cited.) 6 Modern Hungarian minden (lit. ‘every’) works this way, since it means everything. In Old Hungarian minden as a DP could also mean ‘everyone, everybody’. 7 Complexes like mind ez ‘all this’ could of course have an argument role, but on its own mind could only be a modifier.

From A-quantification to D-quantification



(12) a. De vr Bernald mendent latuan egew lampasnal but master Bernard everything-acc see-part burn-part lamp-ade ‘But Master Bernard, seeing everything in the light of the burning lamp . . .’ (Jókai C. 5) b. menel sarwldel mendenedett kyket goimp.2sg-away and-sell-imp.2sg everything-poss.2sg-acc who-pl-acc vallaz own-2sg ‘go forth and sell everything you own’ (Jókai C. 6) Definiteness agreement: To put it simply, object DPs with minden did not trigger definiteness agreement on the verb, whereas the covert associate of mind did. This is shown in the contrast between (13) and (14). The sentences in (14a) and (14c) are particularly interesting, since mind has no overt associate in them, yet the verb agrees in definiteness with the covert object. Definiteness agreement with mind is taken as evidence that mind (but not minden) came with a definiteness requirement on its associate. (13) m˙eguon menden varost & m¯edent a. foldon ´ ´ ´ prt-take-pst.3sg every town-acc and every-acc the earth-sup lakozot dwell-part-acc ‘He conquered every town and every inhabitant of the land’ (Vienna C. 14) (No definiteness marking on the verb) e´g dongo le´g es mind el zagatt’a (14) a. El io ´ away tear-defo.3sg away come-3sg one buzz-part fly and all ‘And then comes a blowfly and tears it (the spider’s web) all apart’ (Bod C. 4v) b. ember, ez velagi morhat ey nappa keresi, el man this world-ly riches-acc night day-trans seek-defo.3sg away az halal, es mind el vezi o tole io ´ ´ ´ come-3.sg the death and all away take-defo.3sg he from-him ‘man pursues worldly riches night and day, but up comes death and takes them all away from him’ (Bod C. 4v) c. Es hasonlatos a test a kouer lohoz, . . . fel zokuen ´ ´ And similar the body the fat horse-all . . . up jump-part rugwan, mind az foldhoz veri ´ ´ kick-part all the earth-all cast-defo.3sg ‘The body is like a fat horse, . . . prancing, kicking, it throws (its riders) all to the ground’ (Bod C. 5v)



Ágnes Bende-Farkas

Number marking: Mind was also not inflected for number: its form was invariably mind, irrespective of the number marking on its associate (or the interpretation of its associate). This is shown in the following Modern Hungarian examples: (15) a. —Na és a leves? —Mind megettük. well and the soup? all prt-eat-pst-1pl ‘—What about the soup? —We’ve eaten it all’ b. —Na és a vendégek? —Mind-(∗-ek) eljöttek well and the guests? all(-pl) prt-come-pst.3pl ‘—What about the guests? —They have all come’ We take these data to indicate that mind (even without an overt associate) was not a free-standing DP or stranded determiner (the classic reference for a stranding analysis of floating quantifiers is Sportiche (1988)). A similar piece of evidence is that mind could not receive the partitive suffix -ik that indicated that its domain was familiar from discourse. (-ik was originally a possesive suffix that indicated group membership, in a manner similar to English of in expressions like every one of them.) Examples like (16) are relatively frequent in the codices, but there are no instances of inflected mind (???mind-ik ‘all of them’). & o egm¯edenicrè (16) viźicuala azocat o hozia ´ ´ take-3.pl-bepst that-pl.acc he all-3.sg and he one-every-poss.3pl-sub kèzet vètuen me´gu´gaz´ta vala azocat hand-poss.3sg-acc cast-part prt-heal be-pst that-pl-acc ‘they took them (those who were ailing) to Him, and He, laying His hand on each and every one of them, healed them’ (Munich C. 58vb) Adverb placement A piece of indirect evidence, from Modern Hungarian, is that today’s preverbal mind occurs between speaker-oriented and agent-oriented adverbs.8 A similar test is applied to English all in Brisson (1998), where all is ranked among so-called speaker adverbs, above agent-oriented adverbs and adverbs of manner within the syntactic and scopal hierarchy of Ernst (2002). It will be seen shortly that the test produces slightly different results for Hungarian mind, although the conclusion is consonant with Brisson’s findings. Arguing on the basis of Modern Hungarian data is justified, we think, because Modern Hungarian mind does not differ significantly from its Old Hungarian counterpart. (The only relevant difference is that Modern Hungarian mind does not modify temporal, spatial, or scalar expressions, whereas Old Hungarian mind quite frequently did so, as shown in examples (23) and (24) below.) Mind and speaker-oriented adverbs: 8 It will be in fact suggested that mind is a speaker adverb which, for reasons of scope, has to follow other speaker adverbs.

From A-quantification to D-quantification



(17) a. ???Mind valószínűleg elmentek All probably prt-go-pst-3pl Intended: ‘They have all probably left’ b. Valószínűleg mind elmentek Probably all prt-go-pst-3pl ‘They have all probably left’ In Brisson (1998) all is analyzed as a speaker adverb, because it it can occur either before or after adverbs from this class. (The sentences in (18) are Brisson’s examples (53) and (54) on page 200. Brisson’s other examples show that all does not follow agentoriented adverbs and adverbs of manner.) (18) a. The boys probably all went home b. The boys all probably went home. Hungarian mind cannot precede a speaker adverb, as shown in (17). The following examples will show that mind precedes agent-oriented adverbs and adverbs of manner. The question is whether the positioning of mind among preverbal adverbs is in fact indicative of its being one. If mind is taken to be an adverb (for the sake of the argument), the next question is how it can be seen to fit into Ernst’s hierarchy. Anticipating the summary of all types of evidence in this excursus, we can say that mind can in fact be analyzed as an adverb. As regards its position in the Ernst hierarchy, the pattern in (17) may cast doubt on its being a speaker adverb. Nevertheless Brisson’s arguments invoking the pragmatic functions of all (and all is analyzed by her as a speaker adverb) are valid for Hungarian as well: When using mind the speaker expresses a judgment concerning the status of a particular collection, emphasizing that the main predicate of the sentence holds for the entirety of that collection. The fact that mind cannot precede speaker adverbs like valószínűleg ‘probably’ or szerencsére ‘fortunately’ may be derived from scopal mechanisms: For (17a) to be acceptable the modal should be able to scope under mind; this is possible only if mind is inherently distributive, or if it can combine with a covert distributivity operator in situ. It will be seen in part 3.2 that mind was not inherently distributive; Modern Hungarian mind is also not inherently distributive, so it is not guaranteed to outscope a modal like valószínűleg. Also, apparently, covert distributivity operators cannot be inserted among preverbal speaker adverbs in Hungarian. In a word, a pattern like (17) does not preclude mind’s being a speaker adverb; the right explanation for (17) may have to do with the mechanisms of scope assignment in Hungarian. Returning to the placement of mind among preverbal adverbs, it can be seen that it precedes agent-oriented adverbs: (19) a. Mind szívesen tanultak all willingly learn-pst-3pl ‘They all studied willingly’



Ágnes Bende-Farkas b. ???Szívesen mind tanultak willingly all learn-pst-3pl Intended: the same as above

Mind can appear to precede manner adverbs, which are lowest on the scale proposed in Ernst (2002). In this case we take the adverbs to be generated higher than their preferred position, as indicated by their intonation, by the fact that the verb precedes the verbal prefix, and by a change in interpretation: (20) a. Mind lassan mentek el all slowly go-pst-3pl prt ‘They all left slowly’ b. Lassan mind elmentek slowly all prt-go-pst-3pl ‘Eventually they all left’ In (20a) the manner adverb lassan ‘slowly’ indicates the manner in which every member of some group left. The preferred syntactic position of such adverbs is the preverbal focus position, as indicated by the verb–prefix order in (20a). Sentence (20b), where the adverb is not in the focus position, is only grammatical if the adverb is not interpreted as a manner adverb. In (20b) lassan is presumably construed as a kind of speaker adverb, providing a comprehensive judgment or evaluation of the events of people leaving. Modern Hungarian mind is thus shown to fit into the hierarchy of preverbal adverbial positions. From this we cannot however conclude without further tests that mind was indeed an adverb: Hungarian quantificational DPs behave in a similar manner, precisely because the preverbal quantifier position (the dedicated landing site for quantificational expressions) precedes agent-oriented adverbs and adverbs of manner. ((21b) below, similarly to (20b), is only rescued because lassan can be reinterpreted as a speaker adverb.) (21) a. ???Szívesen mindenki tanult willingly everyone learn-pst Intended: ‘Everyone studied willingly’ b. #Lassan mindenki elment slowly everyone prt-go-pst ‘Eventually everyone left’ Not: ‘Everyone left slowly’ There is one difference regarding the preverbal placement of mind and minden relative to adverbs: minden-DPs can precede speaker adverbs, whereas mind cannot:9 9 (b) improves if the adverb following mind has the intonation indicating that it it is some kind of afterthought. The point however is that if the adverb is not an afterthought mind cannot precede it, and, mutatis mutandis, cannot outscope it.

From A-quantification to D-quantification



(22) a. (Valószínűleg) mindenki (valószínűleg) elment (probably) everyone (probably) prt-go-pst According to surface order: ‘Probably everyone left’ (Scope: Modal > ∀) Or: ‘For every x it is probable that x left’ (Scope: ∀ > modal) b. ???Mind valószínűleg elmentek all probably prt-go-pst-3pl Intended: ‘For all of them it is probable that he/they left’ c. Valószínűleg mind elmentek probably all prt-go-pst-3pl ‘It is probable that all of them have left’ In (22a) we see the Hungarian quantificational DP mindenki ‘everyone’ behaving in the same manner as English all in (18): it can freely precede or follow the speaker adverb valószínűleg ‘probably’. Obviously, this does not make mindenki a speaker adverb. All that can be concluded from (22a) is that the landing site of quantifier movement in Hungarian is within the area where speaker adverbs are generated in situ. (22a) also indicates that the quantifier can land in the very position which is scopally appropriate for it. Mind, on the other hand, does not have this option. The rigidity of mind as regards placement and scope (compared to the relative flexibility of minden-DPs) indicates that it is a better fit among preverbal adverbs than minden-DPs.10 Taken together, morphological data (no affixation) and placement relative to adverbs indicate that mind was neither a stranded determiner nor a quantificational DP. The adverbial status of Old Hungarian mind is supported by sentences where it associates with spatial or temporal expressions. In such examples (quite abundant in the codices) mind can be paraphrased as all the way, all the time until, ever since, etc. That is, in these uses mind is a synonym or near-synonym of egész(-en), teljes-(en) ‘complete(ly), full(y)’. (23a) contains mind in association with a spatial expression; in (23b) we have mind associating with a temporal expression. (23) a. az ev kyaltassok mynd menyorzagiglan fel hallyk vala. the she cry-poss.3pl all heaven-term up hear-pass.3sg be-pst ‘their cries could be heard all the way to Heaven’ (Margaret Legend 41v)

10 Data throughout this chapter indicate that ‘floating’ mind is not a stranded determiner or quantifying DP. The simplest hypothesis, then, is to take mind to be generated in situ. Data in this section also indicate that mind takes scope in situ. Being generated in situ and taking scope in situ are the hallmarks of adverbs. If we accept this conclusion, the remaining question is whether mind belongs to the Ernst hierarchy, as claimed in Brisson () for all, or whether it is an adverb of quantification similar to English always.



Ágnes Bende-Farkas b. vyselven mynd az tellyes napot nagy aytatos syralmas spend-part all the complete day-acc big pious tearful jmadsagban prayer-ine ‘spending the entire day in pious, tearful prayers’ (Margaret Legend 7r)

The sentences in (24) describe actions, and mind emphasizes that these actions were indeed carried out in full. (24) a. Idumea kiralanac tètèmit meg e´gètte mend Idumea king-poss.3sg-dat bone-poss.3sg-acc prt burned all hamuiglan ash-term ‘He burned the bones of the king of Idumea all (the way) to ashes’ (Vienna C. 216) b. & a. t¯plom mend foldiglèn letorèttètet ´ ´ and the temple all earth-term down-break-pass-pst.3sg ‘the temple was demolished all (the way) to earth’ (Vienna C. 261) ‘the temple was completely demolished’ So, in examples (23a) and (24) mind is clearly a modifier, analyzable as a specifier within a scalar expression. Without further discussion we will assume that Old Hungarian mind was not ambiguous. (Bende-Farkas (2013) contains a proposal for a uniform semantic analysis of Old Hungarian mind.) A consequence of the nonambiguity assumption for mind is that its categorial status must have been unambiguous as well. Since in (23a) and (24) mind is neither a determiner nor a DP, the only remaining option is to treat it uniformly as an adverb. To conclude discussion on the categorial status of mind, we present another type of data indicating that prenominal mind was not a determiner. Mind could and still can associate with (definite) DPs containing non-quantificational determiners such as numerals, or sok ‘much, many’, or kevés ‘little, few’. az kep elebe fel (25) a. mind az harom g´ ertyat all the three candle-acc the picture in-front-of up ragaztottauona: stick-perf.3sg-be-cond ‘He stuck all three candles in front of the picture’ (Kazinczy C. 28r) b. Azt a sok könyvet Palika mind elolvasta? that-acc the much book-acc Paulie all prt-read-pst-3sg ‘Those many books, has Paulie read them all?’ Modern Hungarian minden ‘every’ is (in principle) compatible with numerals, but it is still not interchangeable with mind. ((26a) is fashioned after examples from Szabolcsi 1981, 1983, 1994b.)

From A-quantification to D-quantification



(26) a. Minden két katona kapott egy puskát every two soldier got one gun-acc ‘Every couple of soldiers received a gun’ b. Mind a három katona kapott egy puskát all the three soldier got one gun-acc ‘All three soldiers received a gun’ In (26a) we have quantification over pairs of soldiers: in a group of soldiers, of unknown cardinality, every pair received a gun. In (26b), on the other hand, the numeral indicates the cardinality of the group: there were three soldiers altogether, and, according to the preferred interpretation, each of them received a gun. Therefore (26b) is taken to indicate that prenominal mind was not a determiner; instead, it associated with the entire DP. ... Integration with(in) the DP In this part we review data and arguments concerning the placement of the determiner minden and that of prenominal mind. A key factor will be placement relative to the nominative pronominal possessor. In Old and Modern Hungarian alike, mind can be adjacent to its ‘associate’ DP, so the question arises what position it occupies in this configuration, and how that position relates to the position of minden. In turn, one may ask of minden what position it occupied within the DP, whether, for instance it occupied the same position as other determiners, such as sok ‘much’. Perhaps a more important question as regards minden is how it became part of the DP. We assume that it started its life as a replica of mind, that is, as free-standing adverb, and that it ‘drifted’ past the leftmost edge of the DP. By the time the Munich and Vienna Codices were written (around the middle of the 15th century) minden was definitely integrated into the DP, at least in the dialect shared by the authors (scribes) of these codices. Integration into the DP (or being separate from it, as we suspect in the case of mind), has been diagnosed by checking the placement of mind or minden relative to the definite article, and relative to the pronominal possessor bearing nominative case. In Modern Hungarian the nominative possesor is part of the D-layer which forms the boundary of NP and the DP projection (Szabolcsi 1981, 1984, 1994b; Bartos 2002). If an expression follows the nominative possessor it can be taken to be placed within the NP. Assuming that in the 15th–16th centuries the DP-layer (Spec of DP, the D head, etc.) was projected, and that it was not radically different from what we find in Modern Hungarian, placement relative to the nominative possessor can be revealing. In Modern Hungarian mind precedes and minden follows the nominative pronominal possessor. (27) a. ?Mind az ő könyvei all the he book-poss.pl.3sg ‘All his books’



Ágnes Bende-Farkas b. Az ő minden könyve the he every book-poss.3sg ‘Every book of his’

A complicating factor in diagnosing DP structure in Old Hungarian manuscripts is that the possessor could remain covert, in the sense that possession can be marked solely by means of suffixation on the noun, as in Modern Hungarian (28) and Old Hungarian (29). This results in the paucity of readily usable data and makes straightforward diagnosing difficult. (28) Minden cikk-é-t elfogadták every paper-poss.3sg-acc prt-accept-pst-3pl ‘Every paper of his has been accepted’ (29) az ifv micoron mindon dolgait megh ´ the youth when every thing-poss.pl.3sg prt mondottauona say-pst.3sg-be-cond ‘when the youth had told him everything’

(Kazinczy C. 26r)

According to the data, DP-adjacent mind was not interchangeable with minden, because (i) Mind consistently preceded the definite article (when the definite article was present at all), whereas we have no data with minden co-occurring with the definite article. (ii) mind invariably preceded the nominative (pronominal) possessor, whereas minden followed it in the majority of cases. (30a) shows mind preceding the nominative possessor o ‘he’, and (30b), (30c) shows mind preceding the definite ´ article az.11 (30) a. mind o ereyet ez velagi heusagra kolti ´ ´ all he force-poss.3sg-acc this worldly vanity-sub spend-pres.3sg vala be-pst ‘He dissipated all his powers/force on worldly vanities’ (Bod C. 3v) b. Tehat mind az zentok e´getombe mondanak: Ez az zyz: ´ ´ therefore all the saint-pl together say-3pl this the virgin ‘Thus all the saints said together: This is the virgin’ (Kazinczy C. 9v)

11 Example (c) is relevant, because the associate DP has the same syntactic structure as the DP in the earlier example (), where the absence of the definite article was a complicating factor. In example (c) on the other hand, the DP az tartomanba lakozo emborok ‘the inhabitants of the province’ (lit. ‘the people ´ dwelling in the province’) is clearly definite, either in´virtue of the definiteness of the head, or in virtue of the definiteness of the locative expression.

From A-quantification to D-quantification



c. Ezben ke´gek hyre lon: mind az tartomanba ´ this-ine conj news-poss.3sg be-pst.3sg all the province-ine lakozo emboroknek: ´ ´ dwell-part man-pl-dat ‘Meanwhile all the inhabitants of the province had heard the news’ (Kazinczy C. 38r) I(erusa)lembol (31) Ioachim ked(ig) a nag p˙uspoc iouo ´ ´ ´ ´ Ioachim conj the great bishop come-pst.3sg Jerusalem-ela Betuliaba om¯eden papiual ozuo ´ ´ ´ Betulia-ill he-every priest-poss.pl.3sg-ins together ‘And Joachim the great bishop came from Jerusalem to Betulia, together with every priest of his’ (Vienna C. 44) (31) shows minden preceded by the nominative possessor o (‘he’). It appears that ´ minden occupied a DP-internal determiner position already in the 15th century,12 whereas mind was in a peripheral position. There is one type of exception where the relative position of Old Hungarian minden and the nominative possessor is concerned. The puzzle is the following: Whereas minden-DPs from the 15th century either lack nominative possessors altogether (the Jókai Codex) or else minden follows the possessor pronoun (the Vienna and Munich Codices), several 16th-century codices (among others, the Döbrentei (1508), Érdy (ca. 1524–27), Horvát (1522), Kazinczy (1526–41) or Székelyudvarhely (1526–28) Codices) appear at first sight to be throwbacks, in that they contain DPs where minden precedes the nominative possessor. That is, possessives with minden are like Modern Hungarian in the earlier codices, while in several later codices there is a small but clearly definable set of examples exhibiting a different structure. The term ‘throwback’ indicates that we take minden to have drifted into the DP from a position very much like the position occupied by mind. An example of minden preceding the nominative possessor is (33), where minden precedes the possessors mi ‘we’ and ew ‘he’, respectively. Before presenting the 16th century data we show examples from the Vienna and Munich Codices, where the nominative possessor is regularly made explicit, and where it immediately precedes minden: (32) a. Azert meglatom t˙v reiatoc t˙v menden that-fin prt-see-1sg you-pl sub-2pl you every hamissagtokat dishonesty-poss.2pl-acc ‘Thus I will uncover/visit on you all your dishonesty’

(Vienna C. 218)

12 For the purposes of this chapter it is irrelevant whether minden was in a projection reserved for (strong) quantifiers, or whether we take all determiners and numerals to have shared the same slot. What is relevant is that minden occupied a position that was definitely not on the leftmost edge of the DP.



Ágnes Bende-Farkas b. O m¯edèn muuèlkedètekèt azert teźec hog lattassanac ´ they every deed-poss.3pl-acc that-fin do-3.pl so-that see-pass-3pl èmberectol ´ people-abl ‘Every action of theirs is (only) carried out to make an impression on people’ (Munich C. 28va) ‘All their deeds are accomplished with the sole purpose of making an impression’

The sentences in (33) are from the 16th century, and show the possessor pronoun following minden: (33) a. elottunk lezon mindoon mi bineinknek emlekozety: ´ ´ ´ ´´ before-1pl be-fut.3sg every we sin-poss.1pl-dat memory-poss ‘(At the Last Judgment) we shall be confronted with the memories of all our sins’ (Kazinczy Codex 1526–41; 2r) b. boczassad megh myndden ew byneyt forgive-imp.2sg prt every he sin-poss.pl.3sg-acc ‘forgive every sin of his’ (Érdy C. ca. 1526; 153b) Our contention is that the order minden > pronoun pertains to an earlier stage of the language, even though the codices where it has been preserved were written (copied) later than the Munich and Vienna codices, in which the order pronoun > minden is prevalent. If data like (33a) or (33b) convey the dialect of their scribes, i.e. the person directly responsible for their creation, it can be concluded that in certain dialects such structures were indeed productive in the 16th century. Since very little is known about dialectal variation in Old Hungarian, and lay texts reflecting current language use are just as scarce, we cannot exclude the possibility that the order minden > pronoun did count as archaic for a large percentage of speakers in the 16th century, and sentences like (33a) or (33b) can be regarded as archaisms passed on with little change from one codex to another. Two observations can be brought to bear on this: First, DPs of the type (33a) are rare, even in these codices. In these texts the string ‘minden pronoun noun-possessive suffix’ alternates with the string ‘minden noun-possessive suffix’. The latter is by far the more frequent. Second, DPs where minden precedes the pronoun are confined to a special kind of context, which involves interaction or communication between God (or individuals in close proximity to Him: saints, angels, or the Virgin Mary) and men (typically, the addressee of the text). Conversely, structures minden > pronoun do not occur in episodic sentences with a more mundane content.13 Typically, DPs where 13 The following episodic sentences contain no pronominal possessors. They are from the same codices as (a) and (b), and describe ‘ordinary’, mundane events:

From A-quantification to D-quantification



minden precedes the possessor pronoun occur in prayers, homilies, exhortations. On account of the special status of such texts in monastic life it can be assumed that they were copied faithfully from one document to another, with minimal change. Hence, we think, the order minden > pronoun does in fact belong to a stage in Old Hungarian that preceded the actual time when the codices were produced. Returning to the question of providing a syntactic analysis for DP initial mind, the uniform categorization of mind as an adverb in the previous part commits us to an adjunction analysis, illustrated in (34) below. (The tree is a representation of the DP mend w foia nec ‘for all his kin’ from example (5).) (34)

DP

AdvP

DP

mend

w foianec

The adjunction analysis of prenominal mind implies in turn that we do not take it to project a quantifier phrase in the manner proposed for all quantifiers in Matthewson (2001). According to Matthewson, a quantifier head Q projects a quantifier phrase QP, which takes the DP as its complement. The DP in turn has to be definite. (35) shows the analysis of the same DP mend w foia nec ‘for all his kin’, from (5). (35)

QP

Q

DP[+DEF]

mend

w foia nec

Matthewson’s analysis explains the placement of mind relative to the definite article and possessive structures. It also motivates the definiteness requirement on the associate of mind: the definite article/possessive pronoun serves as a type shifter and restricts the domain of the quantifier. (i) a. az ifv micoron mindon dolgait megh mondottauona the youth when every ´ thing-poss.pl.3sg prt say-perf.3sg-be-cond ‘when the youth had told him everything’ (Kazinczy C. r) b. myndden taghyaat el zaggattataa every limb-poss.3sg-acc prt tear-caus-pst.3sg ‘He tore off every one of his limbs’ (Érdy C. b)



Ágnes Bende-Farkas

Apart from the wealth of evidence in favor of the adverbial status of mind, the problem with an analysis in this vein is that it cannot be applied to Hungarian determiners in general: mind is the only expression in Hungarian that behaves in this way. Hungarian determiners do not require their DP to be definite, and they follow both the nominative possessor and the definite article (when there is one). The reader may ask whether an analysis in Matthewson’s vein could be applicable to those Old Hungarian cases of minden where it did precede the possessive pronoun, as in (33). The problem is that minden did not in general impose a definiteness requirement, not even in those texts where it sometimes precedes the possessor pronoun. The assumption that Old Hungarian minden in these cases projected a QP with the DP as its complement would face at least two problems: (i) It would have to explain the lack of a definiteness restriction on the DP. (ii) It would also have to explain why such QPs gave way to DPs which contained minden. A purely syntactic problem with (35) is that such an analysis does not take floating mind and the possibility of right-adjacency into account. Consequently, it is not suited to handle Hungarian examples like (36). This sentence is ambiguous, with respect to the associate of mind (the object DP or the—covert—subject), and with respect to its attachment site as well. (36) Megettük a levest mind prt-eat-pst.1pl the soup-acc all ‘We have eaten all the soup’, or ‘All of us have eaten the soup’ Mind is therefore taken to be inserted into syntactic representations in a uniform manner, via adjunction. The problem presented by (36) is solved if right-adjunction is allowed. Since at least two adjunction sites are available, this also accounts for the ambiguity of the sentence. To summarize the findings of this subsection, we can say that mind was an adverb, and it can be said to have been generated in situ. As an adverb, it was not inflected and could not fill an argument slot. Minden, on the other hand, could be a determiner or a full DP, and it could move to its scope position. Minden-DPs could naturally be arguments of the verb, and were inflected for case. They could also receive the partitive suffix -ik. Mind could be adjoined to the VP or one of its extended projections, or it could be adjoined to the DP. Adjunction to the DP is indicated by the fact that mind preceded the definite article or the nominative pronominal possessor. Where minden as a quantifying DP is concerned, the question of its landing site within the sentence will be deferred until section 4.5. Within the DP, we take minden to be a determiner, belonging to a DP-internal layer. This is supported by the fact that in two 15th-century codices, the Vienna and Munich Codices, minden invariably follows the nominative pronominal possessor. A small number of 16th-century data, where minden precedes

From A-quantification to D-quantification



the possessor pronoun, can indicate that the integration of minden into the DP was not uniform, or it did not proceed with uniform speed within the linguistic community. .. Semantics ... ±Familiarity and contextual domain restriction Earlier in this chapter (in section 4.2.1) mind was characterized as an operator that could operate either on the members of a set (collection) or on the entire collection itself. A handful of examples are indicative that minden too could originally have meant ‘entire(ly), complete(ly)’. These will be presented later. It appears that although mind and minden were originally closely related semantically, a number of structure-driven changes have led to a state of affairs where minden is very different from mind as a quantifier, viz. as regards its combinatorial and scope possibilities, and as regards its discourse status. So, what we have is a difference in interpretation caused by morphological and syntactic differences. As a maximality operator, mind operated on collections that were familiar from discourse, or on the denotations of expressions that were linguistically marked as definite, so that their denotations would also be circumscribed or delimited. In other words, mind was a maximality operator that expected its domain to be quite heavily restricted. As a consequence, mind was not (and still is not) used for arbitrary generalizations; such propositions were (and are) conveyed with minden. Similarly, in Old Hungarian a meaning like ‘always’, ‘at all conceivable times’ was conveyed with a derivative of minden (minden-kor ‘every-when’).14 ... Distributivity Apart from fewer constraints on contextual domain restriction, it could be hypothesized, that the distinctive trait of minden has been distributivity. It is indeed true that today’s minden is distributive and that Old Hungarian mind was not inherently distributive. Nevertheless, it does not follow that Old Hungarian minden was also inherently distibutive. With hindsight, the distributivity of today’s minden could be attributed to the second -n suffix. Today’s -n may in principle be analyzed as a distributivity operator, but more caution is required with Old Hungarian minden. Some stray examples from Old Hungarian show that early on minden need not have been different from mind where distributivity was concerned. Example (37) shows that föld ‘earth’, ‘land’ could combine with minden, yielding sentences that involve either the coercion of föld into a count noun, or else minden was to be understood as ‘entire’, ‘all the’. Both explanations appear viable, maybe simply because codices resorted quite frequently to metonymy or other figures of speech.

14

Today’s mind-ig ‘always’ lit. ‘all-term’ is a recent creation. Codices used mindenkoron or mindenkor.



Ágnes Bende-Farkas

zongesek (37) Mynden foldreh ky meneh az o ´ ´ ´ every earth-sub out go-pst.3sg the they voice-poss.3pl . . . zerzed oketh feyeldelmol mynden foldon ... ´ ´ ´ ´ . . . appoint-past.2sg them prince-ess every earth-sup ‘Their voice filled all the earth/every land . . . you appointed them as rulers of all the earth/every land’ (Könyvecse 27 r) In example (38) minden combines with életünk, ‘our life’, ‘our lives’. If minden is to be understood distributively in this sentence, the DP mynden eletewnk lit. ‘every life of ours’ is to be construed as ‘each life from among [the lives of] a group of persons’. A simpler alternative is that at the end of the Old Hungarian period minden could still occasionally be construed as ‘entire, complete’. (38) Ez zamos zent napokban myndden eletewnket meg this numerous holy days-ine every life-poss.1pl-acc prt yobbohok improve-sbjv.1pl ‘During these many feast days we should improve our entire life’ (Érdy C. 4a) A similarly telling example is (39), where a DP containing minden is additionally marked with the temporal (pluractional) suffix -keed (roughly meaning ‘per ’, or ‘ly’). An example like (39) does not necessarily imply that minden was inherently cumulative; it may merely imply that initially the second -n suffix was not uniformly and generally reanalyzed as a distributivity marker. elót (39) zollywnk arrol ky mynden naponkeed zemewnk speak-sbjv.1pl that-del which every day-sup-dist eye-poss.1pl before forog revolve-3sg ‘Let us speak about that which is before our eyes every day’ (Érdy C. 20a) (Lit.: every daily) Examples like (37), (38), or (39) show that minden could combine with non-count nouns in a manner that is similar to mind. There is, however, an important distinction between mind and minden, which could be said to be valid in Old Hungarian already. This is the following: no examples have been found in which the Nuclear Scope of minden contains a collective expression (contributing to a collective or cumulative reading for minden), or in which minden binds a reciprocal pronoun. Several examples have been attested, on the other hand, where mind is adjoined to a phrase containing a collective predicate or a predicate modifier indicating a collective reading (as in examples (40) or (41)). Mind was also perfectly compatible with reciprocal pronouns, as shown in (42).15 15 We think we can conclude from these facts that Old Hungarian minden was indeed incompatible with expressions of collectivity or reciprocality. The reason is the relative frequency of such expressions, and their regular association with mind.

From A-quantification to D-quantification



(40) a. Tehat mind az zentok e´getombe mondanak: Ez az zyz ´ ´ thus all the saint-pl together say-3pl this the virgin ‘Thus all the saints said together: This is the virgin’ (Kazinczy C. 9v) b. Az kouetkezo ´ ´ the following allok: stand-part-pl ‘The following admired it’

nap mind az nep fel day all the people up mind Codallyak uala: all admire-3pl be-pst day, all the people having

goluen: . . . az kornol ´ ´ ´ gather-part: . . . the around

assembled, all those around it (Kazinczy C. 17r )

(41) Tehat ime az hagot napra es helre mind ozue ´ thus lo the leave-part day-sub and place-sub all together golenek: ´ gather-pst-3pl ‘Thus they all assembled on the appointed day, at the appointed place’ (Kazinczy C. 61r) e¨g maasnak es (42) kyk mind eleygben yonek ´ who-pl all before-poss.3pl-ine come-3pl one other-dat and wg tiztolyk e¨g maasth ´ that.way respect-3pl one other-acc ‘who all come forward to meet each other, and thus honour each other’ (Sándor C. 5v) In addition, mind frequently co-occured with distributivity markers. az ordog (43) a. Heten vadnak, Mel’eket, az o At’ok ´ ´ ´ seven-adv are, which-pl-acc the she father-poss-3pl the devil mynd e´genkét kazdagon el hazasyta, all oneadv-dist richly away marries ‘They (the daughters of cupidity) are seven in number, all of whom their father the devil marries off generously, one by one’ (Székelyudvarhely C. 95r–v) b. Igon meltosagossok: merth istennek mynd feyenkeed ´ very distinguished-pl because god-dat all head-adv-dist leany es fyay daughter-poss.pl.3sg and son-poss.pl.3sg ‘They are venerable, since each and every one of them is God’s daughter or son’ (Sándor C. 1v) If mind had been a distributive universal, examples like (43) would not have been possible. In such examples mind can be taken to provide the domain of the dis-



Ágnes Bende-Farkas

tributive operator via its association with a suitable expression, similarly to its interaction with distributive ki ‘who’ or ki-ki ‘who-who’, which will be discussed in part 4.4.16 If mind is taken to be a floating maximality operator, and it is not taken to be inherently distributive, it need not be assumed that it was a restricted quantifier that partitioned sentence material into Restrictor and Nuclear Scope. (44a) is a provisional definition, where mind is analyzed as a predicate (VP-)modifier, and says that the predicate P (supplied by the maximal projection mind is attached to) holds of the maximal individual ∪X constructed from the collection X. X is contributed by mind’s associate. If (44a) resembles a quantifier at all, it resembles definites and proper names treated as Generalized Quantifiers in Montague Grammar. (It can be seen that the main difference between the λ-terms in (44) is type assignment: mind is of the modifier type e, t, e, t, whereas the man and John are Generalized Quantifiers of type e, t, t.) (44) a. mind: λP.λX.[P(∪(X))] b. the man: λP.[P(ι(ιman ))] c. John: λP.[P(j)] In (45) (44a) is applied to example (3) from the Funeral Sermon and Prayer, ignoring mood and modality. The variable y stands for the (covert) subject, and z stands for the individual whose sins are to be forgiven. (45) bulsassa mend w bunet (‘may He forgive all his sins’): forgive (y, ∪X); X = λx.[sin (x, z)] Minden, on the other hand, needs to be taken as a restricted quantifier whose Restrictor and Nuclear Scope are supplied by the rules of syntax. For today’s determiner-minden, the Restrictor is the Common Noun Phrase, and the Nuclear Scope is a ‘chunk’ that contains at least the verb. (46) a. minden, distributive: λP.λQ.∀x.[P(x) → Q(x)] b. minden, non-distributive: λP.λQ.λX.[X = ∪x; P(X); Q(X)] (Alternatively: λP.λQ.[Q(∪P)]) The lambda-terms in (46) encode the Old Hungarian ambiguity of minden, possibly also reflecting a variation in semantic competence. (46a) is the textbook translation of minden: it is a distributive restricted quantifier, the Common Noun Phrase (the slot for P) contributing the Restrictor and a syntactic chunk containing the verb (the slot for Q) contributing the Nuclear Scope. (46b) provides the analysis of non-distributive 16 Examples like (), containing mind and a distributivity marker, are quite frequent. Where minden is concerned, we have so far only found a handful of examples in one document, the Érdy Codex, of its coexistence with a distributivity marker. Sentence () is one of these.

From A-quantification to D-quantification



minden, encountered in examples like (38). The formula says that the predicate Q is true of the largest individual with property P. (46b) resembles a maximality operator, but it involves a rudimentary Restrictor–Nuclear Scope structure: Unlike mind in (44a) the Common Noun Phrase (inserted into the slot reserved by P) contributes a property and not a free variable over pluralities. In other words, even though Old Hungarian minden was not inherently distributive, it did introduce a tripartite structure into the logical representation of the sentence. NB neither version of (46b) accounts for the lack of collective or reciprocal markers in the Nuclear Scope of minden; a more adequate semantic analysis of Old Hungarian minden remains a task for further research. ... Variable binding A prediction of our analysis of mind is that it did not bind variables, at least not in the usual way one would expect from logic textbooks. The hypothesis is that the relationship between mind, its associate (and further expressions when necessary, e.g. reciprocal pronouns) resembled the resolution of discourse anaphora.17 Minden, on the other hand, was in a position to bind individual variables. The most telling example is (49), where the possessive suffix in the oblique phrase ew lelkeben ‘in his soul’ is bound by the subject quantifier. Examples (47) and (48), where the Common Noun Phrase contains a bound variable, are also relevant, since it could be argued that minden binds that variable in the same manner as Modern Hungarian minden. (That is, we propose that there is a logical difference between the Old Hungarian equivalents of all his deeds and every deed of his. Also, we assume that Old Hungarian every deed of his was logically equivalent to its Modern Hungarian counterpart.) (47) mi atyank bodog fferench menden miuelkedetiben: we father-poss.1pl blessed Francis every deed-poss.pl.3sg-ine istenhez volt hassonlatos: god-all was similar ‘In all his deeds our father the Blessed Francis was like God’ (Jókai C. 1) (48) menden yozagat zegeneknec eloztuan: every asset-poss.3sg-acc poor-pl-dat away-divide-part ‘having distributed every one of his assets among the poor’

(Jókai C. 8)

(49) menden test ne gyczewlkewgyek ew lelkeben every body not glorify(-refl-)sbjv.3sg he soul-poss.3sg-ine ‘Nobody should glorify his soul’ (Jókai C. 128) ‘For everybody it holds that he is not to praise his own soul’ 17 The reader may notice that binding mind’s associate with a λ operator in (a) is not a correct move if the relation between mind and its associate is to resemble discourse anaphora. For a fully appropriate analysis, which we cannot attempt here, one would need to resort to a dynamic semantics for plurals (Van den Berg , Nouwen ).



Ágnes Bende-Farkas

In the examples above minden behaves like today’s determiner, in that it can be said to induce a Restrictor–Nuclear Scope division of sentence material, it is distributive and can safely be assumed to bind an individual variable. ((49) is very different from Modern Hungarian, for the sole reason that in Modern Hungarian universals directly scoping over negation are conveyed with senki ‘no-one’.) The following examples (quite infrequent in the codices) exhibit some rather unusual behviour with respect to binding or the Restrictor–Scope division. (50) contains a superfluous-looking relative pronoun (akynek ‘to whom’), which could stand for a variable bound by the quantifier. The main problem is that relative pronouns in the Nuclear Scope of a quantifier do not like to be bound by that quantifier. Another problem is that akynek ‘to whom’ intervenes between the head azt ‘that’ and the other relative pronoun my ‘what’, the pronoun that the head is in fact co-indexed with. So, the relative clause looks in fact like a correlative, with two relative pronouns, akynek ‘to whom’, and my ‘what’; the first of these is indexed to the quantifier, and the second to the pronominal head azt.18 (50) mendeni nek meg ada aztj akyi nek myj evuei vala everyonei -dat prt gave that-accj whoi -dat whatj hisi be-pst ‘She gave everyone his due’ (Cornides C. 178r) ‘Lit. She gave everyonei thatj to whomi whichj was hisi (due)’ Assuming that (50) is in the form intended by the author we also have to assume that there is a way for the quantifier to access the index i on the relative pronoun aky. At this stage we can see two ways this could be accomplished, and both involve some discourse mechanism. In (51a) (written in Modern Hungarian for ease of exposition) the direct object of the sentence is the bare pronoun azt ‘that’. The correlative follows as an independent clause, providing further information. In (51b) the order of clauses from (50) is reversed: the correlative sets the stage, introducing a set of pairs x, y, such that x has y as his due. In this setting the quantifier can be regarded as an afterthought. Note that in (51b) the quantifier cannot bind the relative kinek directly: kinek is part of an ordered pair, and is embedded under a maximality operator, whereas minden requires access to atoms. A prerequisite of the semantic analysis of (51b) is quantification over the atoms of the collection introduced by kinek, and the introduction of some kind of dependency such that every member of the kinek-set is awarded the corresponding member of the mit-set. That is, within the scope of mindenkinek the dependency between the members of the two collections needs to be preserved in some manner. 18 If the relative clause is indeed a correlative—which still needs to be ascertained—an additional problem is caused by the operator status of the two pronouns, and their interaction with the embedding quantifier.

From A-quantification to D-quantification



(azt). Kik nek mij az övék (51) a. Mindenkii nek megadta everyonei -dat prt-give-pst.3sg (that-acc) whok -dat whatj the hisk volt (megadta nekik ). be-pst.3sg (prt-give-pst.3sg dat.3sg) ‘She gave it to everyone. Whoever had something as his due, she gave it to him.’ mij az övéi volt, (mindenkij nek) megadtak b. Kineki who-dat what the his be-pst.3sg (everyone-dat) prt-give-pst.3sg ‘Whoever had something as his due, she gave that to him/to everyone’ A semantic analysis of the binding mechanisms needed for (50) and (51b) is well beyond the scope of this chapter. The original example (50) has been presented in order to show that in the early stages of tripartite determiner quantification in Old Hungarian such hybrid cases were possible, where loose, discourse-like relations between discourse referents interfered with the binding requirements of quantifiers. Example (52) is similar to (50), in that it contains a relative clause and a tripartite quantificational structure. The source of the problem is the presence of mind in the relative clause. If mind were absent the relative clause could be taken to belong to the Restrictor of minden (‘Everyone who believes in Him will be redeemed’). The relative clause in (52) is instead taken to be but loosely linked to the tripartite structure, as additional, ulterior domain restriction. (52) mynddennek Idwesseeg Adatyk kyk mynd ew benne every-dat salvation give-pass-3sg who-pl all he ine-3sg hyznek believe-3pl ‘Everyone will be granted salvation, all those (will be redeemed) who believe in Him’ (Érdy C. 38b–39a) To conclude this subsection we repeat an earlier example to reinforce the argument that in Old Hungarian we may find the coexistence and occasional mingling of tripartite structures with a syntactically and logically looser form of structuring information. Example (13), repeated here as (53), contains minden in what might be termed as an appositive relation with its Common Noun Phrase. This is indicated by the double accusative marking, both on m¯edent ‘everyone’ and on the nominalized nonfinite expression a foldon lakozot ‘the ones inhabiting the land’. Such double case marking ´ ´ is virtually nonexistent in Modern Hungarian, except for appositives; double case marking was also (for all practical purposes) confined to cases like (53) in Old Hungarian codices.



Ágnes Bende-Farkas

(53) m˙eguon menden varost & m¯edent a. foldon ´ ´ ´ prt-take-pst.3sg every town-acc and every-acc the earth-sup lakozot dwell-part-acc ‘He conquered every town and every inhabitant of the land’ (Vienna C. 14) Examples like (53) are not infrequent in the codices and are also found with other determiners besides minden. These too invariably occur with nominalized non-finite expressions, and not with lexical nouns like város ‘town’. That is, the reason for such loose structures can lie in the Old Hungarian syntax of nonfinite structures. ... The semantics of mind and minden: An interim summary This subsection has shown mind and minden to differ as regards contextual requirements, distributivity and variable binding. Mind placed a definiteness requirement on its associate— definiteness could mean either discourse familiarity or unique (maximal) reference. As a consequence, mind was not used to express generalizations that involved the lifting of contextual restrictions. Mind was not inherently distributive, and has remained so in Modern Hungarian. Minden was not inherently distributive, either, although today’s Modern Hungarian minden is distributive.19 Mind frequently co-occurred with distributivity operators in Old Hungarian; minden did so but sporadically. Even though minden was not inherently distributive there are no attested examples of minden co-occurring with collective or reciprocal expressions in its Nuclear Scope. Mind on the other hand could and did co-occur with such expressions. The combined effect of the definiteness restriction imposed by mind and its lack of distributivity is that the preferred analysis of mind is to take it to be a maximality operator (resembling the definite article or relative pronouns in free relatives), and not as a restricted quantifier. This is reflected in the proposed translation of mind into second-order logic (in (44a)). The main difference between mind and non-distributive minden is that the latter (but not the former) involved a division into Restrictor and Nuclear Scope (this is reflected in our first translation for non-distributive minden, in (46b)). This also meant a difference in the contribution of the relevant nominal expression: The associate of mind is said to contribute a free plural variable, whereas to minden the Common Noun Phrase contributed a property. 19 In fact, Modern Hungarian minden is still compatible with mass nouns. It does not occur with abstract nouns, unless these can be reinterpreted as count nouns: (i) a. Kibányásztak minden aranyat. out-mine-pst-3pl every gold-acc ‘All (the) gold has been excavated’ b. Minden igazságot megkérdőjeleztek every truth-acc prt-question-pst-3pl ‘Every true statement has been disputed’ Lit.: ‘Every truth has been disputed’

From A-quantification to D-quantification



Where variable binding is concerned, it has been shown that minden could and did bind variables in its Nuclear Scope, while mind was said to enter into an anaphoric relationship with its associate, which did not involve variable binding in the usual sense of this term. In other words, the landscape of quantification in Old Hungarian included D-quantification, with the logically standard way of binding individual variables, and one species of A-quantification, which relied on a form of anaphora resolution. At the end of this section a few examples show minden in non-standard configurations; we took these examples to belong to a stage when loose, discoursebased relations between operators and operands interfered with the more rigourous relations imposed by D-quantifiers. In sections 4.4 and 4.5 we will discuss another mode of A-quantification, the longdistance binding of indeterminate pronouns by propositional operators (Kratzer and Shimoyama 2002). Whereas the A-quantifier mind has remained in the inventory of Modern Hungarian, quantification over indeterminate pronouns has not.20 Its presence in Old Hungarian or Proto-Hungarian can be deduced from a handful of examples in Old Hungarian codices and from the overall behavior of languages in the Uralic family.

. Free pronouns as distributive operators In this section we discuss a consequence of the emergence of universal Dquantification. This is the reanalysis of certain indeterminate pronouns in sentence initial position as distributive operators. In Old and Modern Hungarian alike, reduplicated ki-ki lit. ‘who-who’ is a distributivity operator (cf. Farkas (1997) on Modern Hungarian ki-ki): (54) Kiki a maga istenét üvölté who.who the own god-poss.3sg-acc yell-pst-3sg ‘Each was crying to his own god’ (Mihály Babits, Jónás könyve [The Book of Jonas], 1938) In Old Hungarian the same meaning could be expressed with ki + mind: (55) nagy eremmel ky mynd hazahoz meegyen great joy-ins who all house-poss.3sg-all go-3sg ‘Greatly rejoicing each of them went home’

(Érdy C. 170a)

It is proposed here that ki-ki and ki mind have their origin in the indeterminate pronoun status of ki. These complexes can be regarded as remainders from a period when such pronouns were bound by long-distance operators (cf. Kratzer and Shimoyama 2002). The emergence of D-quantification confined such pronouns to relative 20

Unless we include binding by interrogative or relative operators.



Ágnes Bende-Farkas

and interrogative clauses. In other syntactic contexts they became superfluous, unless they could be reinterpreted. By means of reduplication or association with mind the pronoun ki ‘who’ became a distributivity operator. Other pronouns such as mi ‘what’ or hol ‘where’ do not reduplicate, nor do they associate with mind.21 .. Indeterminate pronouns in Old Hungarian The discussion of ki-ki and ki mind rests on the assumption that Proto-Hungarian had indeterminate pronouns that acquired quantificational or interrogative force via long distance binding (Kratzer and Shimoyama 2002, Kratzer 2005). Old Hungarian, at the time when the Jókai Codex was written, had a determiner system; bare pronouns were typically confined to interrogative or relative clauses. In fact, this was also a time when composite relative pronouns were formed, including ha ki (lit. ‘if who’) or az ki (lit. ‘that who’). The presence of determiners and determiner quantification implied that (i) variable binding became local, and (ii) ‘stray’ pronouns were compounded with determiners, yielding quantifying DPs such as valaki ‘someone’, mindenhol ‘everywhere’, and so on. Occasionally, however, one comes across pronouns in the codices that are neither interrogative nor relative. These are taken as remnants of an earlier mode of quantification, viz. the long distance binding of alternatives. (Cf. Kratzer and Shimoyama (2002); see also Watanabe (2009) for an application to Early Middle English.) Typically such pronouns occur in the scope of negation or in the antecedent of a conditional: (56) tevzet ievttem bochatny fevldre. es myt akarok fire-acc come-pst-1sg release-inf earth-sub and what-acc want-1sg egyebet. hanem chak hog eegyen. else-acc if-not only that burn-sbjv.3sg ‘I’ve come to release fire onto earth, and what (else) do I want but for it to burn’ (Cornides 65 r–v) ‘I’ve come to release fire on earth, and I want nothing else but for it to burn’ lelkew embernek kedeeg nagyob erdemót zerez vele (57) Az yo the good natured man-dat conj bigger merit-acc acquire inst.3sg chak ky neky ne engheggyen just who dat-3sg not yield-sbjv.3sg ‘(The devil’s temptations) (only) increase the merits of good souls; it is just that one should not yield to them’ (Érdy C. 82b) ‘(The devil’s temptations) only serve to multiply the merits of good souls; the key is that one is not to yield to them’ 21 In other languages, such as Salish, reduplication also expresses quantification. As discussed in Jelinek and Demers (), reduplication within verbs or nouns in Salish yields a weak quantifier. Reduplication of pronouns and numerals in Hungarian results in a strong quantifier, and so does the reduplication of adverbial classifiers in Chinese, cf. Cheng () and the references cited there.

From A-quantification to D-quantification



(58) Ha ky kerdenee honnan volt az. Azzonywnk marianak if who ask-cond.3sg where-from was that lady-poss.1pl Mary-dat hogy semy terheet nehesseegeet nem zenwette legyen that none burden-acc difficulty-acc not suffered be-sbjv.3sg Reea felelnek doctorok mondwan . . . sub-3sg reply-3pl doctors say-part ‘Should someone ask how come that Our Lady Mary had no difficulty (in giving birth) learned men reply saying . . . ’ (Érdy C. 44a) (59) Ha kedeeg my kewessee annal nagyobot zolt if conj what little-trans that-ade bigger-acc speak-pst.3sg volna. hyzóm hogy mind ez vylaag sem be-cond believe-1sg that all this world neither foghatta volna meg catch-possib-perf.3sg be-cond prt ‘And if he (St John) had spoken somewhat louder/any louder I believe that not even the whole wide world could have grasped it’ (Érdy C. 54a) The pronoun in (56) is taken to be in the scope of an elaborate discontinuous connective roughly paraphraseable as ‘nothing else . . . but . . . ’. The pronoun ky in (57) can be a free variable with a universal construal that scopes over negation. Pronouns in the antecedent of a conditional, as in (58), acquire a universal interpretation from binding by a conditional. (That is, sentence (58) is a so-called donkey sentence, named after the sentence standardly used to illustrate the phenomenon: Every farmer who owns a donkey beats it.) In (59) the pronoun my ‘what’ can be interpreted in two ways: it is either bound by the conditional, yielding a free choice/polarity interpretation (any increase in loudness would have been sufficient), or it is existentially closed, indicating some hypothetical increase in loudness. Given the purported meaning of the sentence, and the presence of sem ‘neither’, ‘not even’, the more likely reading is where the conditional operator ha ‘if ’ is the binder, conferring a free choice reading to the pronoun. .. On ki-ki (mind) In addition to single indeterminate pronouns codices contain the reduplicated sentence-initial pronoun ki ‘who’. Ki-ki conveys distribution over a familiar collection. A first example is (60) below. If not reduplicated, ki was immediately followed by mind, as seen in (61) below (and in (55) earlier). Reduplicated ki-ki (but not ki mind) can be found in Modern Hungarian as well. In Modern Hungarian ki-ki is commonly analyzed as a distributivity operator on a familiar set, which needs to bind a variable within the sentence (cf. Farkas 1997). In Old Hungarian ki-ki and ki mind had a similar role: they provided a link to a previously



Ágnes Bende-Farkas

introduced collection (which could be the domain of a quantifier), and distributed over the members of that collection. (60) a. ky ky mind miwelkodethe zerenth we´gón: awagh ´ who who all deed-poss.3sg according.to take-sbjv.3sg or Iot: awagh gonozth: good-acc or evil-acc: ‘Each should partake according to his deeds, whether it be of good or evil’ (Kazinczy C. 89v) b. mindonok feel tamadnak az alkolmas allapatba: meel kinek ´ ´ every-pl up surge-3pl the appropriate state-ine which who-dat kynek nezy onnon termezettit: ´ ´ who-dat regard-3sg own nature-poss.3sg-acc ‘Everyone will be resurrected in the state appropriate to his nature’ (Kazinczy C. 96v–97r) The sentences in (60) show the typical syntactic pattern of sentences with distributive ki: the pronoun occurs in a sentence-initial position, and it is either reduplicated or immediately followed by mind. In (60b) the reduplicated pronoun follows the relative pronoun meel ‘which’, indicating that a sentence with ki-ki could be embedded. (61) Ez offertorium koron zoktanak emberek adny this offertorium when-sup would-3pl man-pl give-inf ayandokozny valamyt istennek tyztessegere. ky give.as.present-inf something-acc god-dat honor-poss.3g-sub. who zerent. mynd ev tehetsege all he capability-poss.3sg according.to ‘During the offertorium people would offer presents to honor God, each according to his means.’ (Cornides C. 188r) From (60) and (61) it can be seen that ki (re)distributes over a familiar collection. In Modern Hungarian (where only reduplicated ki-ki is acceptable) the pronoun needs to bind a variable in its clause (Farkas 1997). This requirement is hard to test for Old Hungarian; provisionally, the binding requirement can be taken to hold for this period as well. In fact, both (60) and (61) satisfy this requirement. The chief question as regards distributive ki(-ki) is how a free pronoun came to be reanalyzed as a distributive operator. For Modern Hungarian the key could be reduplication; in Old Hungarian however, reduplication was not necessary for a distributive reading, as seen from (61). A necessary (but not sufficient) condition is provided by a logical property of free variables: universal generalization. If a predicate P is true of a variable x, and there are no constraints on the choice for x, then the corresponding universal statement

From A-quantification to D-quantification



(∀x.[P(x)]) will also be true. In the case of distributive ki there is one constraint (belonging to a given collection X), but this constraint holds for each and every member of X. That is to say, there is no reason to single out one particular element from the domain X. It follows, then, that the universal-distributive construal of ki is a case of contextually restricted universal generalization. When reduplicated ki-ki became the standard mode of expressing this kind of reading, reduplication became the overt reflex of a logical mechanism. A problem for what might be called the ‘all-logical’ explanation of distributive ki is the presence of mind ‘all’ in several Old Hungarian examples. In fact, ‘single’ ki on its own does not appear as a distributor in the codices. The universal construal of single ki (or mi ‘what’) in examples (58)–(59) from 4.4.1 was due to an overt operator (the conditional or negation), whereas ‘distributive’ ki was either reduplicated or was immediately followed by mind. It might appear that mind could have served as the overt universal quantifier in this construction, binding (the trace of) ki. Ki mind could even be defined as a (short-lived) composite quantifier, corresponding roughly to English each. If Old Hungarian ki mind can indeed be taken as a composite quantifier, the division of labor between its components is not what one would expect, viz. that mind is to be the universal quantifier, and ki is to provide the alternatives to be bound. The glitch in such an explanation is that Old Hungarian mind was not inherently distributive, as the examples in section 4.3 have shown. It is doubtful, therefore, whether mind could bind individual variables in the same manner as minden could. In section 4.3.2 it was proposed that the relation between mind and its associate was a form on discourse anaphora and not a way of binding individual variables. We propose therefore that in the case of ki mind as well, mind provided an anaphoric connection to a familiar collection. Mind can be seen as a ‘pluralizer’, providing a domain for the distributive operator, ensuring that the entirety of that collection would be subjected to the operation(s) carried out in the sentence. In turn, the distributivity operator was either contributed by the pronoun, or was introduced in a construction rule, as a form of syntactic ‘glue’.

. The impact of D-quantification .. The impact of D-quantification on syntactic and semantic structures In this part some of the consequences of the availablity of determiner quantification will be explored. Where the syntactic structure of Hungarian sentences is concerned, this development had two immediate consequences: (i) the consolidation of the DP-internal NumP or QP layer, and (ii) the generation of the dedicated preverbal quantifier position. ... DP structure It could be seen in 4.2.2 that the first text containing minden, the Jókai Codex, also contains a fair number of other determiners. (Determiners in



Ágnes Bende-Farkas

the Jókai Codex include sok ‘many’, ‘much’, egynéhány ‘a few’, or néminemű ‘some’, ‘(a) certain’.) That is, a DP-internal NumP or QP layer can be assumed to have existed at that time. There is no information as to the precise order in which the individual determiners appeared there. There is no information, either, whether minden occupied a pre-existing position, or whether it created its own projection. Since the other determiners attested in the Jókai, Vienna, and Munich Codices are all weak cardinals, what can be said with certainty is that strong minden as an inhabitant of NumP/QP facilitated the quantificational construal of other inhabitants of that position. ... Sentence structure Concerning sentence structure, minden-DPs were a relevant factor in the emergence of the dedicated preverbal quantifier position of Hungarian (cf. É. Kiss 1994, É. Kiss 2002, Szabolcsi and Brody 2003), and its subsequent specialization as a host for distributive expressions. (The codices do not contain a legtöbb ‘most’. Minden and its family of derivatives were practically the only strong, quantifying DPs.) What made the creation of such a position possible is, we think, the fact that minden was (inside) a DP, and, as such, it had the option to move, like other DPs. An immediate consequence of this fact is that scope relations could be established via movement. Preverbal minden-DPs occur frequently in the codices, indicating the availability of the dedicated preverbal quantifier position, which has been defined as the landing site of Quantifier Raising or Quantifier Adjunction for Modern Hungarian. Nevertheless, it cannot be said with certainty that scope order was always directly mirrored by surface order in Old Hungarian, as it is in Modern Hungarian. Example (62) at least shows that the minden-DP could in fact be interpreted lower than its surface position. Literally, the conditional clause says ‘Every day you promise to give him enough food’, which, we think, is not the intended reading. The accurate paraphrase is ‘You promise to give him enough food every day’. That is, what the addressees are requested to do every day is to feed the wolf; they are not requested to make a promise every day.22 (62) Es sonha meg nem sert tyteket valamyben ha mynden and never prt not hurt you.pl-acc something-ine if every nappon fogattok neky adnya eleg eledelt day-sup promise-2pl dat-3.sg give-inf enough food-acc ‘And he (the wolf) will never cause you any harm if you promise to give him enough food every day’ (Jókai C. 151) According to Szabolcsi (1997) (see also Szabolcsi and Brody (2003) and subsequent references), the position traditionally labelled as the quantifier position is in fact 22

The sentence is from the legend of St. Francis and the wolf of Gubbio.

From A-quantification to D-quantification



a Distributive Phrase (proposed initially in Beghelli and Stowell (1997)). The tree in (63) shows the main features of Beghelli and Stowell’s and Szabolcsi’s proposal: Referring (or specific) expressions target RefP (the topic position), whereas distributive quantifiers target DistP, whose head D contains a covert distributivity operator . (63) RefP (=TopicP)

DistP

DP

D’

D

Δ

(NegP)

FocusP

...

Quantifiers in (Modern) Hungarian are said to move to this position in order to acquire a distributive reading. Considering the compatibility of Old Hungarian minden with non-count nouns (as seen in examples (37) and (38) in 4.3.2), movement to DistP and hence the secondary, ulterior acquisition of a distributive reading seems a reasonable and attractive option. The problem, however, is that minden-DPs with non-count nouns were not distributive—had quantifier movement been movement to DistP, such DPs would automatically have been distributive when they occurred preverbally. We claim, therefore, that quantifier movement in Old Hungarian (initially, at least) was not movement to a Distributive Phrase. The target position of that movement, say, plain QuantP, could have been later reinterpreted as DistP. At this point the reader may ask whether one can in fact reckon with a genuine quantifier position in Old Hungarian. Preverbal mind was generated in a very similar position, so the two sites can easily be collapsed into one syntactic position. So, it could be argued, quantifier movement contributed no new projections. A reply to this question involves several components. As may be recalled from subsection 4.3.1, it is indeed the case that preverbally, both minden-DPs and mind appear among speaker adverbs. One difference between mind and minden is that mind is generated in that position, as an adverb. Minden-DPs, on the other hand, are moved to the same (kind



Ágnes Bende-Farkas

of) position. Movement in turn implies that minden-DPs bind their trace and the appropriate variable inside the Nuclear Scope. There is no such requirement for mind, whose association with the appropriate expression could be non-local, and did not involve binding in any case. It may also be recalled from 4.3.1 that minden-DPs show greater flexibility than mind, in that their landing site can precede or follow speaker adverbs. Depending on where they actually land, they can take scope under or over the relevant adverb. (The relevant example is (22a), repeated here as (64).) Mind in its turn cannot precede speaker adverbs, which has been taken to indicate a fixed generation site, and frozen scope. (64) (Valószínűleg) mindenki (valószínűleg) elment (probably) everyone (probably) prt-go-pst According to surface order: ‘Probably everyone left’ Or: ‘For every x it is probable that x left’

(Scope: Modal > ∀) (Scope: ∀ > modal)

(65) a. ???Mind valószínűleg elmentek all probably prt-go-pst-3pl Intended: ‘For all of them it is probable that he/they left’ b. Valószínűleg mind elmentek probably all prt-go-pst-3pl ‘It is probable that all of them have left’ The first ingredient of an answer to the issue of there being a dedicated Quantifier position in Old Hungarian relies in effect on the adverbial status of mind, and on the presumed adverbial status of the first (hypothetical) instances of minden. Since mind, an adverbial initially closely related to minden, was generated among speaker adverbs, the same area could have been re-used as a target for the movement of DPs containing minden. A slightly different argument for a dedicated quantifier position is provided by minden-DPs directly outscoping negation, as in (66) below. ((49) in (4.3.2.3) shows the same logical structure.) (66) menden titk nem lèhètètlèn tenèked every secret not impossible dat-2sg ‘No secrets are impossible before Thee’ Lit.: ‘Every secret is not impossible before Thee’

(Vienna C. 136)

In the Jókai, Vienna and Munich Codices minden . . . nem ‘every . . . not’ was frequently used in negated existential statements instead of logically equivalent senki, semmi nem ‘no-one, nothing not’, even though the n-words themselves are attested in these texts. It appears that in these cases the minden-DP was moved to a preverbal

From A-quantification to D-quantification



position so that it could outscope negation. There are no attested cases of mind in such a logically motivated configuration.23 To conclude this brief argument on the syntactic behavior of minden-DPs, we can say that preverbal minden-DPs were different from mind, precisely because they had moved to their surface position: their exact landing site could vary according to their scope needs. These DPs can be said to have bound their trace and a variable in their Nuclear Scope. Finally, minden-DPs could move in order to express universal negative statements, the logical equivalent of negated existential statements. ... Binding:±locality A further consequence of the availability of Dquantification is the local binding of variables in the manner familiar from logic, as opposed to long-distance quasi-binding of indeterminate pronouns, or anaphora resolution. If Proto-Hungarian did indeed have indeterminate pronouns bound by long-distance operators, we can assume that the landscape of quantification in that period had the properties proposed or predicted in Kratzer and Shimoyama (2002) (see also Kratzer 2005): Quantifiers (or their closest counterparts) were propositional, operating on sets of alternative propositions. Such alternative sets were built bottom-up, and indeterminate pronouns were responsible for the introduction of non-singleton sets of alternative individuals. (67) Ha ky kerdenee honnan volt az. Azzonywnk marianak if who ask-cond.3sg where-from was that. lady-poss.1pl Mary-dat hogy semy terheet nehesseegeet nem zenwette legyen that none burden-acc difficulty-acc not suffered be-sbjv.3sg Reea felelnek doctorok mondwan . sub-3sg reply-3pl doctors say-part . . . ‘Should someone ask how come Our Lady Mary had no difficulty (in giving birth) learned men reply saying . . . ’ (Érdy C. 44a) If we apply Kratzer and Shimoyama’s analysis to an Old Hungarian example like (58), repeated here as (67), we can say that the pronoun ky ‘who’ contributes to the building of a set of propositions of the form x should ask how come Our Lady Mary had an easy and painless birth. These propositions are bound by the conditional, and the outcome is the donkey sentence ‘For everyone it holds that if he asked why the Virgin Mary had an easy and painless birth, learned men would reply in such and such a manner.’ 23 When mind had scope over negation it resembled a definite and not a universal quantifier. Such an example is the matrix clause of () (repeated here, in note ), which is definitely not a negated existential statement: (i) mind ez vylaag sem foghatta volna meg all this world neither catch-possib-perf.3sg be-cond prt ‘not even the whole wide world could have grasped it’ (Érdy C. a)



Ágnes Bende-Farkas

Such operators could ‘bind’ alternatives across syntactic islands, as noted in Kratzer and Shimoyama (2002), since the construction of alternative propositions takes place in the interpretive component. These operators were also ‘unselective’, in the sense that they were insensitive to the number of actual pronouns in their operand, creating the illusion of one operator binding several variables. A prediction of Kratzer and Shimoyama’s proposal is that scope relations were determined by operator placement. Once an operator was in place, scope relations became fixed. If one adopts, even provisionally, the hypothesis that Proto-Hungarian had longdistance quantification over alternatives, one has to accept its predictions as well, viz. the lack of island constraints, the illusion of unselective binding, and the rigidity of scope relations. If this was indeed the landscape of quantification in Proto-Hungarian then the effects of determiner quantification could have been quite radical. Determiner quantification, attested in Old Hungarian at least since the 14th century, is selective, and binding is local. Scope relations are flexible, on account of quantifier movement, which is island-sensitive. .. Conclusions This chapter has discussed the syntactic and semantic differences between Old Hungarian mind ‘all’ and minden ‘every’. Minden has been argued to warrant special attention on account of its being the most prominent from a handful of strong quantifiers (the others are derived morphologically from minden itself), and on account of its clear syntactic and semantic contrast with mind, from which it was, in its turn, derived. In a type-theoretic reconstruction minden was said to have evolved from an adverb or adjective with the lexical meaning ‘entire(ly)’, ‘ful(ly)’. Within the DP, this A-expression could be assigned the type of determiners. Minden as a DP was said to have been derived from determiner-minden, by saturating its first argument slot with the universal predicate thing. The A-quantifier mind has been analyzed as a maximality operator associating with plural definites; this association has been likened to discourse anaphora. Minden, in turn, could be a determiner or a full DP. In most known cases it resembled today’s distributive, strong quantifier, introducing a division into Restrictor and Nuclear Scope, and binding a variable in each of these slots. Mind was compatible with distributivity markers and markers of collectivity and reciprocity alike. In some examples minden was seen to combine with non-count nouns, indicating that it was not inherently distributive at the time; nevertheless, no examples have been found where its Nuclear Scope contained Old Hungarian equivalents of together or each other. A handful of examples have indicated that in Old Hungarian tripartite quantificational structures may have interacted with free relatives and correlatives, exhibiting a mélange of quantificational binding and anaphoric relations not encountered in Modern Hungarian.

From A-quantification to D-quantification



This chapter has also explored the consequences of the presence of D-quantifiers. In the case of minden this includes the reinforcement of the NumP/QuantP layer, making room for other strong determiners within the DP. Movement of mindenDPs to the preverbal field could have changed the status of (one of the) preverbal adverbial positions: whereas adverbs were generated in situ, DPs could now land in that area, and take scope there. Since minden in Old Hungarian codices was not inherently distributive, a negative finding is that the Distributivity Phrase proposed in Szabolcsi (1997) was not projected, even though quantifier movement was present in the grammar. The availability of D-quantifiers made indeterminate pronouns all but redundant. A small number of existentially and universally bound indeterminate pronouns has in fact been found in Old Hungarian manuscripts, and this has led to the hypothesis that Proto-Hungarian and early Old Hungarian quantification was characterized by the long-distance binding of indeterminate pronouns, as proposed in Kratzer and Shimoyama (2002). Today’s reduplicated pronominal distributivity operator ki-ki lit. ‘who-who’ is seen as a reinterpreted (and re-fashioned) remnant of this mode of quantification. In a thought experiment sketched toward the end of part 4.5.1 we proposed that changes brought about by D-quantification could have been quite radical, once we accept the hypothesis that the landscape of Hungarian quantifiers has changed, from propositional operators binding alternatives, to determiners and DPs that bound individual variables locally. In effect, a system which was unselective and insensitive to islands has been replaced by island-sensitive and selective mechanisms of binding. An important task for further research is to find empirical evidence in support of this hypothesis, from Old Hungarian, from archaic dialects of Modern Hungarian, and from languages of the Uralic family.

 The cyclical development of Ps in Hungarian VERONIKA HEGEDŰS

. Introduction This chapter discusses the grammaticalization process and cyclical development of adpositions in Hungarian and the changes internal to the adpositional phrase (PP).1 The changes within the PP will be shown to have an effect on the syntax of clauses as well due to the appearance of functional adpositions that do not always follow their complement (contrary to the more lexical postpositions) and can, therefore, integrate into the clause as verbal particles (also called prefixes or preverbs), leaving their complement behind. In this respect, grammaticalization may result in strictly head-final structures on the one hand, and morphologically free functional elements on the other hand. Hungarian has a rich system of adpositional elements, including postpositions, oblique case suffixes, particles, and adverbs. The same system was characteristic of the Old Hungarian period already, but some syntactic changes were in progress at that time, which resulted in a slightly different syntactic distribution within the class and the emergence of new spatial elements as well. We will give an overview of the state of the PP in the beginning of the written period and discuss the changes that characterized Old Hungarian. The majority of postpositions have developed in possessive structures. The elements can go through a grammaticalization process and become (semi-)functional elements, and sometimes they end up fully grammaticalized, that is, they lose their spatial meaning. It will be shown that as a result of grammaticalization, a functional projection on the left periphery of the Hungarian PP developed, and in the written period it comes to 1 The abbreviation PP is a cover term for the different projections used below, regardless of whether they are prepositional or postpositional on the surface. The term PP will be used to refer to the phrase whenever it is not important whether we are dealing with a PlaceP, PathP or a pP.

The cyclical development of Ps in Hungarian



be filled in more and more frequently by adpositional elements that can consequently function as verbal particles within the clause. The fact that these functional elements are no longer in a strictly head-final structure results in their moving in the clause freely, independently of their complement, and in their ability to be integrated into the clause as particles that form a complex predicate with the verb. This distinguishes them from the older spatial elements that were and are still strictly head-final with a caseless complement. The chapter is organized as follows. Section 5.2 will give a brief overview of the inventory of spatial elements in the beginning of the written period. Section 5.3 will provide an analysis of the changes that characterize the internal structure of PPs. Grammaticalization of P elements in possessive constructions and in appositive constructions will be given an account, and verbal particles will be analyzed as syntactically freer Ps as well. The syntactic freedom of functional adpositions means that the word order of PPs is not necessarily head-final and some Ps can be extracted out of the PP. Section 5.4 will discuss the P(ostposition)-cyle in Hungarian and the structural conditions on the emergence of new Ps.

. Spatial elements in Old Hungarian All different types of spatial elements that characterize Modern Hungarian had already developed by the beginning of the written period. Traditional (descriptive and historical) grammars put these items into four groups, and this section provides a brief overview of the status of these groups in the earliest Old Hungarian texts. Among spatial elements, we find postpositions, case suffixes, adverbs, and particles. These elements are all adpositional and are generated within an extended PP-structure. Their different syntactic properties are a consequence of the fact that they are inserted into the structure in different positions. .. Postpositions Synchronic descriptive and generative grammars generally sort postpositions into two groups: there are Ps that take a caseless (or nominative, according to some theories, see Marácz 1989) complement and Ps that take an oblique case-marked complement, as in (1) and (2). Members of the first group often come in triplets: ones corresponding to locative, lative, and ablative meanings. The endings corresponding to these meanings come from spatial suffixes, which became obscure endings on the now morphologically simple P element: -t and -n are reconstructed to be the Proto-Hungarian locative suffixes, -á/-é were lative suffixes and -l used to be the ablative suffix in Proto-Hungarian.2 2 What are glossed as at, to, and from at the end of the P element in () and below are the obscure ProtoHungarian spatial suffixes, which determine the locative, lative, or ablative meaning of the P, but are no longer separate morphological items.



Veronika Hegedűs

(1) a. a ház előtt the house front.at ‘in front of the house’ b. a ház elé the house front.to ‘(to) in front of the house’ c. a ház elől the house front.from ‘(from) in front of the house’ Members of the other group can be either locative or directional as well, as illustrated by the examples in (2), and in some cases, we do find the tripartite system (as in (3)), although this is less typical of the whole of the class. (2) a. a ház-on belül the house-sup inside ‘(at) inside the house’ b. a ház-on keresztül the house-sup through ‘through the house’ (3) a. a ház-zal szemben the house-ins opposite-ine ‘opposite the house’ b. a ház-zal szembe the house-ins opposite-ill ‘(to) opposite the house’ c. a ház-zal szemből the house-ins opposite-ela ‘from opposite the house’ Old Hungarian has most elements with caseless complements already grammaticalized or on the way to becoming Ps. Some examples for such seemingly grammaticalized Ps are given in (4) and (5). (4) a. te toru˙ened alat ´ you law-poss.2sg under.at ‘under your law’

(Vienna C. 14)

b. az fewuen ala the sand under.to ‘(to) under the sand’

(Jókai C. 160)

The cyclical development of Ps in Hungarian c. az dys ebnec n˙ elue alol the angry dog-dat tongue-poss under.from ‘from under the tongue of the angry dog’ (5) a. uromc scine eleut Lord-poss.1pl face-poss front.at ‘before our Lord’



(Guary C. 23)

(Funeral Sermon and Prayer)

b. nekyk elue dat-3pl before.to ‘(to) before them’

(Jókai C. 21)

orcaioc èlol c. o ´ ´ they face-poss.3pl before.from ‘from before their faces’

(Vienna C. 32)

However, some of these postpositional elements are partially still nominal at this stage: they take part in a variation that is analogous to possessive constructions. I will return to this syntactic variation in section 5.3.1.1 when discussing the grammaticalization process and its stages. There are few Ps with oblique case-marked complements in the early texts. One of them is össze (often in the form öszve) ‘together’, which takes a complement in instrumental case.3 (6) a. fyal usve son.poss.3sg-ins together ‘together with (her) son’

(Königsberg Fragment)

ozuo b. o lakozoiual ´ ´ ´ it inhabitant-poss.pl.3sg-ins together ‘with all that is therein (lit. together with its inhabitants)’

(Vienna C. 225)

Some Ps exhibit variation in the form of the complement they take: some can have both unmarked and oblique marked complements. For example, által ‘through, across’ is such a P. On the one hand it occurs in constructions with an unmarked complement as in (7), on the other hand, it can take a complement with a superessive case-marking. In the latter use, however, sometimes it is not obvious whether the superessive complement and által ‘through, across’ form a constituent or not (as in (8a)–(8b)), and sometimes they are obviously separated (as in (8c)) since által is a preverbal particle while the superessive nominal is postverbal. (7) Asisia-altal-menuen Assissi-across-go-part ‘going across Assissi’

(Jókai C. 3)

3 Instrumental and comitative case have the same form in Hungarian, hence they will both be referred to as instrumental case.



Veronika Hegedűs

(8) a. a-tartomany-on altal-menuen the-province-sup across-go-part ‘going across the province’

(Jókai C. 138)

b. Ki m¯et altal tènger-˙en who go-pst.3sg across sea-sup ‘Who has ever crossed the ocean? (lit. who went across sea)’ (Vienna C. 104) c. altal-mened ez vizen across-go-inf.2sg this water-sup ‘for you to go across this water’

(Jókai C. 18)

Thus, as the previous example illustrated, the same element által is also used as a verbal particle in Old Hungarian, and in that use it can license the presence of an accusative object argument even next to motion verbs, which would otherwise be unaccusative (see (9)). (9) ky-t nem mehet uala altal who-acc not go-possib.3sg be-pst across ‘which he cannot cross’

(Jókai C. 18)

The behavior of által is typical of those elements that became functional postpositions and verbal particles, that is, they can be separated from their complement and appear preverbally in a functional projection that hosts predicative complements of the verb. I will discuss particles and their syntactic position within the PP and within the clause below. .. Oblique/adverbial suffixes Hungarian has a large number of oblique case suffixes, many of which have become suffixal in the written period of the language. The difference between the so-called postpositions and spatial case suffixes is morpho-phonological (É. Kiss 2002, Asbury 2008, Dékány 2012): suffixes are monosyllabic and most of them show vowel harmony with the stem they attach to. Syntactically, the two groups behave largely identically in Modern Hungarian. Some elements were monosyllabic and showed vowel harmony from the beginning, thus they can be considered as suffixes by the Early Old Hungarian period. Others were showing mixed features. Early examples for the ablative suffix -tól/-től and the allative suffix -hoz/-hez/-höz in (10) and (11), respectively, illustrate that some Ps were suffixes in this period. (10) a. mend paradisum-ben uolov gimilcictul all Paradise-ine be-part fruits-abl ‘from all fruits in Paradise’ (Funeral Sermon and Prayer)

The cyclical development of Ps in Hungarian b. pucul kinzotviatwl hell torture-poss-abl ‘from torture in Hell’ c. ystentewl God-abl ‘from God’ (11) a. uromchuz lord-poss.1pl-all ‘to our Lord’ b. ystense(g)nec aniahuz god-dat mother-poss-all ‘to the mother of God’ c. istenhez God-all ‘to God’



(Funeral Sermon and Prayer)

(Jókai C. 2)

(Funeral Sermon and Prayer)

(Königsberg Fragment)

(Jókai C. 1)

The three spatial elements expressing location within something or movement into or out of something, corresponding to inessive, illative, and ablative suffixes presently, were at different stages of becoming suffixal. The locative one was already monosyllabic, but in the oldest text (the Funeral Sermon and Prayer, from around 1195) it does not show vowel harmony with its complement, only the -ben form is documented (see Zsilinszky 1991: 456). In the first codex (Jókai Codex, written after 1372 with a surviving copy from around 1448), we already find the forms with either a front or a back vowel depending on the complement noun, so the locative element can be claimed to be a suffix. (12) a. paradisum-ben Paradise-in ‘in Paradise’

(Funeral Sermon and Prayer)

b. gimils-ben fruit-in ‘in fruit’

(Funeral Sermon and Prayer)

(13) a. vezedelmes vilag-ban dangerous world-ine ‘in dangerous world’ b. nagy tiztesseg-ben big honor-ine ‘in great honor’

(Jókai C. 4)

(Jókai C. 12)



Veronika Hegedűs

The form that became the illative suffix (meaning ‘into’) is still bisyllabic in the earliest texts (as in (14)), as is the ablative ‘out of ’ in (15), therefore Zsilinszky (1991) regards them as postpositions rather than suffixes at that stage. However, their form is beginning to get reduced early on as we can see in some of the examples from Jókai Codex, which has both the longer and the shorter forms. (14) vilagbele world-into ‘into (the) world’

(Funeral Sermon and Prayer)

(15) a. ez homus vilag timnuce-beleul this false world prison.cell-poss-out.of ‘out of the prison cell of this insincere world’ (Funeral Sermon and Prayer) b. bua-beleul misery-poss.3sg-out.of ‘out of his misery’ (16) a. paris-balol Paris-out.of ‘from Paris’

(Old Hungarian Lamentations of Mary)

(Jókai C. 28)

b. ez felewl mondot lang-balol this above said flame-out.of ‘from the above mentioned flame’

(Jókai C. 43)

c. az lang-bol the flame-out.of ‘out of/from the flame’

(Jókai C. 43)

d. ez vilagbol this world-out.of ‘from this world’

(Jókai C. 62)

Oblique suffixes either developed in possessive structures or by means of another type of reanalysis, namely, with the reanalysis of several inflectional morphemes into one morpheme. The spatial suffixes grammaticalized from morphologically independent postpositions that had unmarked complements, which in turn developed in possessive constructions, as will be shown in the next section. .. Particles Particles appear preverbally in neutral Modern Hungarian sentences.4 Old Hungarian is similar in this respect with two important differences: (i) the number of particles 4 A neutral sentence does not contain focus (including wh-questions), negation, and is not imperative (cf. Kálmán a,b and subsequent literature).

The cyclical development of Ps in Hungarian



and the frequency of their use is lower, and (ii) some remnant OV constructions have different word order and thus have the particles in different positions (see also Chapter 1 on word order in the clause). Some of the oldest particles had obviously grammaticalized to the extent that their original spatial use is rare and slowly disappearing, and they are mostly used as telicizers of the verbal predicate. The particle meg is the prime example for this. It used to mean ‘(to the) back’, but by now it has essentially lost its spatial meaning and in Old Hungarian already, it was frequently used without a spatial meaning. Other elements develop into particles in the written period of the language and take over some of the spatial functions of the more grammaticalized ones. One such element is vissza ‘back’, which now appears in some of the old contexts of meg. The first particles developed before the written period, as the first texts contain fully grammaticalized particles. The oldest particles are meg(e) ‘back’, ki ‘out’, le ‘down’, el(e) ‘away’, be(l) ‘into’, fel ‘up’. According to Forgács (2004), while the exact stock of particles in Modern Hungarian is constantly disputed, it is agreed upon that they go back to— and grammaticalized from—adverbs and postpositions. D. Mátai (1991) claims that they developed from spatial adverbs. In fact they all go back to spatial postpositions with a lative (mostly goal) meaning as we will propose in section 5.3.1.3. The first use of the particle meg is already a functional use with no spatial meaning, but this does not mean that it was completely grammaticalized with no lexical meaning in Old Hungarian. It first appears in the Funeral Sermon and Prayer from the end of the 12th century (see (17)). However, we still find it used in its original goal meaning in the 15th century (see (18)). (17) hug turchucat mige zocoztia vola that throat-poss.3pl-acc prt rive-3sg be-pst ‘that it was riving their throat’ (Funeral Sermon and Prayer) (18) Es fèlèlèt v˙eu˙en almocban hog ne mennenèc meg and reply take-part dream-poss.3pl-ine that not go.cond-3pl back herodèsh˙ez Herod-all ‘And being warned of God in a dream that they should not return to Herod’ (Munich C. 9ra) The following examples illustrate the other oldest particles as well. They are taken from Jókai Codex, the earliest text long enough to include a larger number of particles. (19) a. De frater ylyes meg haragwan: sebesseguel bel-teue but brother Elias prt anger-part speed-ins into-put.3sg aytayat S el-tere door-poss.3sg-acc and away-go-pst.3sg ‘But having got angry, Brother Elias quickly closed the door and left’ (Jókai C. 16)



Veronika Hegedűs b. legottan le-hagyta magatt az agyra immediately down-let-pst-3sg self-acc the bed-sub ‘he lay down on the bed immediately’

(Jókai C. 4)

c. fel-kele az agyrol up-rise-pst.3sg the bed-ela ‘he got up from the bed’

(Jókai C. 4)

d. Es valakyket akarz ky-boczatnod ky-boczatassonak and someone-pl-acc want-2sg out-let-inf-2sg out-let-pass-sbjv-3pl ‘and whoever you want to send away, should be sent away’ (Jókai C. 24) Another particle that we find in Jókai Codex that we mentioned before is által ‘through, over, across’. Although the example repeated here has the particle postverbally because it is a negative sentence, the fact that an accusative object is licensed next to a motion verb makes it certain that we are dealing with a particle in this sentence. (20) ky-t nem mehet uala altal what-acc not go-possib.3sg be-pst across ‘which he could not cross’

(Jókai C. 18)

Later, the P element által shortens to át in its use as a particle, the change mostly taking place in the 19th century (see B. Gergely 1980). However, there is still a version of által that is strictly postpositional and takes a caseless complement, but it lost its spatial meaning and became the P marking agents or instruments in passive sentences, as in (21). (21) a. a János által írt könyvek the John by written books ‘books written by John’ b. a vihar által le-tört ágak the storm by down-broken branches ‘the branches broken off by the storm’ The change we see in Old Hungarian is that particles appear in new contexts, that is, if we compare two translations of the same text, we find more particles in the later ones. (22) & mutata nèki mend èz vilagnac orzagit and show-pst.3sg dat-3sg all this world-dat country-poss.pl-acc dičosegit & o ´ ´ and they glory-poss.pl-acc ‘and showeth him all the kingdoms of the world, and the glory of them’ (Munich C. 10ra, 1466)

The cyclical development of Ps in Hungarian



(23) es meg mwtata hewneky ez vylaghnak mynden and prt show-pst.3sg he-dat-3sg this world-dat all orzaghyt, es hw dyczeseghyt country-poss.pl-acc and they glory-poss.pl-acc ‘and showeth him all the kingdoms of the world, and the glory of them’ (Jordánszky C. 363, 1516–1519) (24) & èrèztu˙¯e okèt betlechembe ´ and send-part them Betlehem-ill ‘And having sent them to Bethlehem’ (25) es el-ereztwen hwket betlehembe and away-send-part them Betlehem-ill ‘And having sent them to Bethlehem’

(Munich C. 8vb, 1466)

(Jordánszky C. 359, 1516–1519)

We also find that the most frequent particle, meg, is beginning to be replaced by other particles in spatial contexts. I will return to this observation in section 5.3.1.3. .. Locative adverbs There are several spatial elements that have been classified as belonging to two or three different word classes: they are called postpositions when they are used with a nominal complement, spatial adverbs when on their own in the clause, and particles when they are in the preverbal position. Those elements that express a spatial meaning in the clause without a nominal complement are usually classified as adverbs. In historical texts as well, the same morphological form is often identified as an element of multiple word classes depending on its distribution (e.g. D. Mátai 2007). The “independent” use of a spatial element is classified as an adverb, for example felül ‘uploc ’ and alul ‘downloc ’ in (26). (26) a. Felwl lezen az haragos biro, alol az rettenetes pokol up will.be the angry judge down the terrible hell ‘up there will be the angry judge, and down (there) the terrible hell’ (Bod C. 10r) b. ez felewl mondot rudual this above say-part stick-ins ‘with this above mentioned stick’

(Jókai C. 31)

Pais (1959: 183) argues against this view, claiming that these items do not have meaning independent of the situation they are used in, they only become meaningful by virtue of being connected to the (lingustic) context (see also Sebestyén 2002, who agrees with this view). In the analysis in the next section, we will follow this view and claim



Veronika Hegedűs

that there is an implicit (deictic) complement next to these “independent” abverbials. Under this view, the class of adverbs is then subsumed under that of adpositions; more precisely, they have the same structure as those Ps that take an oblique complement, however, their complement can be implicit or it is a silent here/there. Spatial adverbs are sometimes also found adjoined to other locative phrases, as in (27), which is similar to the context of the development of Ps with an oblique complement. (27) fen eg-ben up sky-ine ‘up in the sky’

(Jókai C. 132)

. Structural changes in the PP Hungarian Ps underwent different types of grammaticalization processes in Old Hungarian. On the one hand, Ps developed in possessive structures, with the original possessor becoming a postposition (and often becoming suffixal). On the other hand, PPs adjoined to other PPs got reanalyzed as functional P heads in the structure. This change was also accompanied by a semantic change, first of all semantic bleaching, whereby the original lexical nominal meaning disappeared and only some spatial aspect of the meaning was left. This semantic bleaching could even reach the point where all spatial meaning is lost, which has happened with some suffixes and with the oldest verbal particle as well. The structural change corresponding to this process is a change from a (lexical) nominal head to a semi-functional P head via an intermediate category that shows both nominal and adpositional properties and a possible subsequent change from a semi-functional spatial adposition to a functional telicizing element, which we hypothesize to correspond to a higher head within the extended PP. The different projections in the PP have different linearization properties in that the lower spatial heads are head-final, while the newly developing higher heads can be head-initial as well. Furthermore, the lower P heads cannot be stranded in the clause (i.e., their complement cannot be extracted), the higher particle head, however, can separate from its complement. This section will first discuss the grammaticalization process typical of PPs and then will turn to the word order properties of the adpositional elements. .. The grammaticalization of adpositions The general structure of adpositions has received some attention in the generative syntactic literature in the past couple of decades and a consensus seems to be emerging that PPs involve projections for place and path denoting elements and for functional adpositions as well, and that Path is projected on top of Place (cf. van Riemsdijk

The cyclical development of Ps in Hungarian



1990; van Riemsdijk and Huijbregts 2007; Svenonius 2004, 2010; Koopman 2010; den Dikken 2010, among others). The syntactic analyses are also based on semantic studies that have claimed that the projection of paths is based on that of places, and the two need to be distinguished semantically (e.g. Jackendoff 1983, Zwarts 2007). Svenonius (2006, 2010) also assumes that there is a projection called AxialPartP, which hosts elements that are intermediate between nouns and adpositions on the basis of their morphosyntactic behavior. The extended structure of PPs thus looks like this: (28)

pP p

PathP Path

PlaceP Place

AxPartP AxPart

DP

Diachronically, Ps often grammaticalize from nominal elements, which results in the filling of AxPart and if the element loses its nominal properties, then in the development of a new Place or Path head (cf. Waters 2009). This is also the process that took place in the development of most Hungarian postpositions. I will discuss the development of AxialParts and Place/Path heads from possessive structures in section 5.3.1.1. Then I will turn to a further grammaticalization process that Ps can undergo, whereby they become higher heads, that is, p elements in the structure. This is how Ps that take an oblique marked complement and particles developed. ... Grammaticalization in possessive structures The oldest Ps developed in possessive structures, and this is still a productive way of grammaticalization. This is a process whereby nouns lose their nominal properties and first become AxialPart elements (cf. Svenonius 2006, Waters 2009), an intermediate category with mixed properties, and then become postpositions. The original possessive structure becomes a PP, with the possessor being the Ground complement of the new spatial element. This involves reanalysis and the appearance of a new P head. Old Hungarian exhibits some syntactic variations which indicate that residual nominal properties were still retained in the developing postpositions. One of the peculiarities of some of the Old Hungarian Ps is that they can appear in a structure that is analogous to the possessive construction. In possessives, the possessum is morphologically marked and the possessor can either be morphologically unmarked (caseless/nominative) or bear dative case. An example for this is given in (29).



Veronika Hegedűs

(29) a. Wimagguc uromc isten kegilm-e-t ez lelkiert pray-imp-1pl lord-poss.1pl god mercy-poss-acc this soul.for ‘Let us pray for our Lord’s mercy for his soul.’ (Funeral Sermon and Prayer) b. De az hews vala ysten-nek angal-a but the young.man be-pst.3sg god-dat angel-poss ‘but the young man was God’s angel’ (Jókai C. 15) Similarly to how the possessor can appear in dative case, the complement of some Ps can be in dative case, and at the same time a possessedness marker analogous to the marking of possessums may appear on the P element. (30) and (31) illustrate this variation: the examples are from the same text, and there seem to be no obvious principles behind the use of the different forms, which shows that we are dealing with free variation at a transient stage of the grammar. (30) a. zemey elewt eye-poss.3sg-pl front.at ‘in front of his eyes’ b. baratok-nak elewtt-e brothers-dat front.at-poss ‘in front of (the) brothers’ (31) a. ott allo neppek kewzt there stand-part people between.at ‘among (the) people standing there’ b. Predicatorok-nak kewzewtt-e preachers-dat between.at-poss ‘among (the) preachers’

(Jókai C. 121)

(Jókai C. 84)

(Jókai C. 59)

(Jókai C. 77)

Old Hungarian possessive constructions have a property that they no longer exhibit in standard Modern Hungarian, namely, that the possessum can also agree in number with a DP-internal plural nominal possessor, thus we find examples like (32).5 (32) az Assiriosocnac k˙em˙ec the Assirian-pl-dat spy-poss-3pl ‘the spies of Assyrians’

(Vienna C. 32)

This construction does not appear in all texts, but where it does, we find similar examples for structures with Axial Part elements as well. For example, the Jordánszky Codex has both the non-agreeing and the agreeing variant with the grammaticalizing előtt ‘before, in front of ’, as illustrated in (33).

5 In standard Modern Hungarian, the plural agreement can only appear on the possessum when it does not form a constituent with the possessor (cf. É. Kiss ).

The cyclical development of Ps in Hungarian (33) a. hw labaynak elette he foot-poss.pl.3sg-dat front.at-poss ‘before his feet’



(Jordánszky C. 552)

b. ystennek angyalynak elettek God-dat angel-poss.pl-dat front.at-poss-3pl ‘in the presence of the angels of God (lit. before God’s angels)’ (Jordánszky C. 581) Hegedűs (to appear) proposes that the variation is indicative of the fact that these adpositional elements are still partly nominal at this stage, and that they are not Place or Path heads, but Axial Parts, that is, their grammaticalization from N to P is not complete at this point. The fact that some nominal property is retained allows for the agreement on the Axial Part element and the frequency of dative marked possessors in the texts of the codices facilitates the use of a similar construction as long as the grammatical properties of the element allow it, that is, until it becomes a P head. The structure of these constructions is illustrated in (34). (34)

PlaceP AxPartP DP

AxPart

zemey baratoknak

elew-

-t -tt-e

Those elements that are suffixal never show this variation; we can be certain that they are grammaticalized P heads by this time. So the examples in (12a) and (14) repeated here as (35) and (36), respectively, have the structures given here. (35) a. paradisum-ben Paradise-in ‘in Paradise’ (Funeral Sermon and Prayer) b.

PlaceP DP

Place

paradisum

‘-ben’

(36) a. vilagbele world-into ‘into (the) world’ (Funeral Sermon and Prayer)



Veronika Hegedűs b.

PathP PlaceP DP

Place

Path ‘-bele’

vilag There is another property that supports the hypothesis that some of the postpositional elements are actually not Place or Path heads in Old Hungarian but have some nominal properties. This support comes from the distribution of reflexive anaphors and pronouns. While the possessive-like variation discussed above does not concern suffixal elements, the distribution of anaphors is more restricted than the group of not yet grammaticalized postpositions. Anaphors are mostly limited to accusative and dative contexts in the early texts and slowly spread to the oblique suffixes and postpositions, that is, to PPs. (37) a. legottan lehagyta magatt az agyra muttuatuan magatt immediately down-let self-acc the bed-sub show-part self-acc alonny sleep-inf ‘he lay down on the bed immediately, pretending to be asleep’ (Jókai C. 4) b. Es ez alazatos frater magat aloytya uala leny and this humble brother self-acc believe-3sg be-pst be-inf mendentewl kewssebnek all-abl smaller-dat ‘And this humble brother believed himself to be smaller than everybody.’ (Jókai C. 73) en mondom magamnak (38) a. mert because I say-1sg self-1sg-dat ‘Because I say to myself ’ b. kibèn mag¯anac iol kellèttèm who-ine self-1sg-dat well like-pst-1sg ‘in whom I am well pleased’

(Jókai C. 32)

(Munich C. 9vb)

With other cases we find a personal pronoun in contexts where we would expect a reflexive based on the adherence to Binding Principle A in Modern Hungarian. (39) a. mert vgy uala ewn belewle kywl because so be-pst.3sg he out.of-poss outside ‘as he was beside himself so’

(Jókai C. 10)

The cyclical development of Ps in Hungarian



b. zent ferenc mene zent Jacabot meglatny: vyuen vele Saint Francis went Saint Jacob prt-see-inf take-part ins-3sg egynehan tarsokot some company-acc ‘Saint Francis went to see saint Jacob, taking some brothers with him’ (Jókai C. 13) c. Az eleuen zent kereztfat myndenkoron ev nala vagy ev the living saint cross-acc always she dat-3sg or she mellette targya vala beside-poss.3sg keep-3sg be-pst ‘She was always keeping the saint cross at or beside herself.’ (Margaret Legend) Modern English also uses pronouns in some PPs, and sometimes there is variation in English as well: (40) a. Bill found a snake [PP near him]. b. We have a whole week [PP before us]. c. John has left his family [PP behind him]. (41) Johni saw a snake near himi / himselfi . Standard Hungarian does not allow variation in most cases, but there is some variation in the use of pronouns and anaphors dialectally. Rákosi (2010) cites the examples in (42) as the standard variant and those in (43) as available options for some speakers. (42) a. Vedd magadra ezt a pulóvert! take-imp.2sg self-2sg-sub this-acc the sweater-acc ‘Put on this sweater.’ b. Vigyél magaddal pénzt! take-imp.2sg self-2sg-ins money-acc ‘Take some money with you.’ (43) a. Vedd rád ezt a pulóvert! take-imp.2sg sub-2sg this-acc the sweater-acc ‘Put on this sweater.’ b. Vigyél veled pénzt! take-imp.2sg ins-2sg money-acc ‘Take some money with you.’ We find the same variation with morphologically independent postpositional elements (Rákosi 2010: (24)).



Veronika Hegedűs

(44) a. Láttam egy kígyót mellettem. see-pst-1sg a snake-acc beside-1sg ‘I saw a snake beside me.’ b. Láttam egy kígyót magam mellett. see-pst-1sg a snake-acc self-1sg beside ‘I saw a snake beside myself.’ Rákosi (2010) claims that the Modern Hungarian Ps showing variation have a possessive structure, which forms a binding domain; the binding configuration is non-local, hence pronouns are licensed. While we have argued that suffixal spatial elements are P heads, and therefore do not expect the variation found, it seems that dialectally, the possessive nature of the original relation between the spatial element and its complement may be transparent enough to allow for a DP-like behavior with respect to binding domains. In most contexts, however, the extensive use of pronouns in PPs disappeared and the reflexive anaphor has replaced the personal pronoun. The following example compares the same lines from different translations of the Bible, and we can see that the anaphor appears already in the 16th century. (45) a. & ne akariatoc mondanotoc t˙un¯o bènnètec ´ and not want-imp-2pl say-inf-2pl you ine-2pl ‘And think not to say within yourselves’ (Munich C. 9va, 1466) b. es ne akaryatok mondany thy-magatokban and not want-imp-2pl say-inf you-self-poss.2pl-ine ‘And think not to say within yourselves’ (Jordánszky C. 631, 1516–1519) These phenomena allow us insight into the grammaticalization process of PPs and the variant syntactic properties of elements undergoing a change from N to Axial Part to P. Those elements that we found to be almost completely grammaticalized in Old Hungarian seem to have completed the process. This does not mean, however, that there are no new elements entering the same path of changes. Possessive structures are the primary source of new Ps in Hungarian. New elements began to grammaticalize in Old and Middle Hungarian, filling the AxialPart head and having the previously developed spatial elements as suffixal Ps. One such example is helyében ‘in his/her place’, which has a transparent possessive structure, but its syntactic behavior is more restricted than that of regular possessive constructions. Helyében can appear with an unmarked complement (originally possessor) or with a dative complement, as illustrated by (46a) and (46b), respectively. However, the possessor cannot be extracted out of the phrase, although regular dative possessors can leave their possessive phrase. This is shown by the ungrammaticality of (46c), where the dative phrase is supposed to be topicalized.

The cyclical development of Ps in Hungarian



(46) a. János hely-é-ben nem tenném ezt. John place-poss-ine not do-cond-1sg this-acc b. János-nak a hely-é-ben nem tenném ezt. John-dat the place-poss-ine not do-cond-1sg this-acc ezt a hely-é-ben. c. ∗ János-nak nem tenném John-dat not do-cond-1sg this-acc the place-poss-ine ‘I would not do this if I were John (lit. in John’s place).’ We can assume that the originally nominal hely ‘place’ in helyében is already an Axial Part in Modern Hungarian with -ben (the inessive suffix) filling in the Place head, thus the structure is the following:6 (47)

PP AxPartP DP

AxPart

János

helyé

P -ben

Once the element loses even more of its nominal properties, that is, the possessedness marker is no longer interpreted as such, and, therefore, the morphological form of the complement becomes invariant (unmarked or dative), we can assume that it becomes a Place head itself. ... Grammaticalization from adjuncts Some of the elements that had a postpositional use already in Old Hungarian started to appear in a new construction. First they appeared as phrasal adjuncts to other PPs, and later the adjoined phrase was reanalyzed as a higher functional head within the extended PP and became a p element, which is less lexical, expressing general direction, orientation, or position (van Riemsdijk 1990 and follow-up). This is in accordance with the economy principles described by van Gelderen (2008) whereby adjuncts get integrated into the phrase structure and specifiers become heads. One of the oldest postpositions that appears adjoined to another phrase, namely, an instrumental marked nominal, is össze ‘together’. It also appears as a particle from quite early on. (48) a. fyal usve son.poss-ins together ‘together with (her) son’

(Königsberg Fragment)

6 In fact, it is the noun and the possessedness marker together that occupy the AxPart head; the two have probably already grammaticalized into one morphological item.



Veronika Hegedűs ozuo b. o lakozoiual ´ ´ ´ it inhabitant-poss.pl.3sg-ins together ‘with all that is therein (lit. together with its inhabitants)’

(49) ewzue gewyty-uala mend az papok together collect-3sg-be.pst all the priests ‘he calls together all the priests’

(Vienna C. 225)

(Jókai C.)

The development of Ps taking oblique case marked complements is assumed to have happened when the PP element was adjoined to the oblique marked expression. The adjoined phrase could be reanalyzed as a functional p head within the phrase. (50) a. èn varosomnac kapuin bèlol ´ I city-poss.1sg-dat gate-poss.pl-sup inside ‘lit. within the gates of my city’ (51)

(Vienna C. 7)

pP p ‘belül’

PlaceP DP

Place ‘-n’

This also explains why the development of these Ps took place later in large numbers: only those elements could be adjoined as PP that had previously grammaticalized as Place or Path heads. A special case is the development of nélkül ‘without’, which exhibits a syntactic behavior identical to those Ps that developed in possessive structures, but in fact goes back to an appositive construction. The origin of this PP is hypothesized to be an appositive construction which contained an adessive case-marked nominal and the adverbial element kűl ‘outside’, that is, it meant something like ‘at a place, outside’ (see Zsilinszky 1991: 443). A hypothetical example would be something like, (52). (52) a háznál, kűl the house-ade outside ‘at the house, outside’ In such a construction, the morphological boundary between the adessive marker and the adjoined locative element could disappear and the two could be reanalyzed as one P head. The reanalysis also resulted in a semantic change. In the oldest texts, the unit nál kül or nél kül was already used in its modern meaning, so the spatial use is not documented in this construction. However, the original adjunction structure was still preserved morphologically: the form of adessive

The cyclical development of Ps in Hungarian



suffix -nál/-nél ‘at’ often still depended on the vowel quality of the stem noun it attached to, but it soon became an invariant -nél(kül) as it grammaticalized into the P head nélkül (e.g. (54)). (53) a. naka-nalkul neck-poss.3sg-without ‘without his neck’ b. paztor-nal kyl shepherd-at outside ‘without a shepherd’ (54) minden vigasag nekol ´ all joy without ‘without any joy’

(Vienna C. 41)

(Jordánszky C. 178)

(Bod C. 12v)

The new structure, thus, developed from an appositive relation, with the adverbial kűl ‘outside’ being adjoined to a PP, as in (55), and then became a PP with the now grammaticalized nélkül ‘without’ being a P head, as in (56). The change is a reanalysis whereby the adjoined PP was reanalyzed as part of the postpositional P head. The new P head developed a new meaning as well, the original place-denoting spatial meaning was lost. (55)

PlaceP PP DP

PP P

‘kűl’

‘-nál/-nél’ (56)

PP DP

P ‘nélkül’

The reanalysis was aided by the fact that a semantic change had taken place, the original spatial meaning was lost in the grammaticalization, and therefore the two morphemes could easily be re-interpreted as one postposition with the new (nonspatial) meaning. The “independent” adverbial use of spatial elements is typical of those p elements that take a locative complement and when this complement is the deictic here/there or a contextually given locative, it may remain silent (see also Dékány and Hegedűs 2013).



Veronika Hegedűs

(57) [pP [PlaceP (itt) ] alul ] This use is never possible with Ps that take unmarked/DP complements, since those complements are not locative, and the elements are not p heads. ... Becoming particles The functional p heads that we discussed in the previous section exhibit more syntactic freedom than the strictly head-final Place and Path heads. These elements can become verbal particles in the clause and in this use they are separated from their complements and appear preverbally in neutral sentences. Some of the spatial elements that grammaticalized well before the Old Hungarian period (in Proto-Hungarian) were already particles in the beginning of the written period as we saw in section 5.2.3. Particles are assumed to be functional adpositions in the highest projection of the PP—in our analysis it is pP. Grammaticalization thus equals being inserted into the structure in a higher head. (58)

pP p

PathP Path

PlaceP Place

DP

The class of particles is smaller in Old Hungarian than in Modern Hungarian but those that had grammaticalized by the early written period display the characteristic distribution and function of particles today (cf. also Chapter 1). They are telicizing elements with often little spatial meaning left due to semantic bleaching. The semantically bleached spatial particles are replaced by new, more obviously spatial (directional) elements next to motion verbs and the most grammatical, bleached element meg ‘orig. back’ has become a general telicity marker. The most frequent meg in the earlier texts was sometimes substituted by other particles with more obvious spatial meanings in later translations of the Bible. (59) Tahat meghaga otèt az ordog ´ ´ ´ then prt-left him the devil ‘Then the devil leaveth him’ (60) Ottan el-hagya hewtet az erdeg there away-left him the devil ‘Then the devil leaveth him’

(Munich C. 10ra)

(Jordánszky C. 363)

New items become particles during the Old Hungarian and Middle Hungarian period, and their grammaticalization involves the steps of having an N+P unit becoming a simple p head. Such elements include hát-ra ‘back-onto’ > hátra ‘behinddir ’ or fél-re ‘side-onto’ > félre ‘aside’.

The cyclical development of Ps in Hungarian



(61) a. [PathP [DP hát ] -ra ] b. [PathP [DP fél ] -re ] The grammaticalization of these elements is similar to those we discussed before, but they do not have a postpositional use, since they did not develop in a possessive structure. The originally nominal hát ‘back’ and fél ‘side’ and the goal denoting Path head, which used to form a PathP, were reanalyzed as a p head during the grammaticalization. After this change, the new element can become a particle in the clause. .. Functional hierarchy and linearization Hungarian used to be an SOV—and in general a head-final—language, and some of its word order characteristics are still determined by that to some extent. Most of its adpositions developed in head-final orders and are still strictly head-final, in some cases, because the element in question became a suffix. PlaceP, PathP are always postpositional, elements that lexicalize those heads surface in a head-final order. The order can be derived from an uderlyingly head-initial order via morphological merger or by movement of the complement (e.g. into the specifier of PlaceP), which line of thinking is in accordance with Kayne’s (1994) proposal regarding base order. While the lexical PlaceP and PathP projections are head-final, we also find prepositional orders in the language. This order is allowed by elements that are base-generated higher up in the PP, under a (semi-)functional head, which we call pP, following van Riemsdijk’s (1990) analysis of German PPs. The pP layer is freer in its ordering with respect to its complement, and the complement can also be extracted from below it. This results in two things: (i) the elements in p can be verbal particles, and thus appear preverbally independent of their complement, and (ii) the phrase below pP can undergo wh-movement separately from the p. While postpositional Ps are never stranded in the language, the freer p elements do not have to be pied-piped with their complements. Old Hungarian functional Ps had this property as well. For example, one of the earliest Ps with an oblique suffix is által, and it could be separated from its complement when it was in the superessive case. The repeated examples in (62) show different degrees of separation. The example in (62a) is structurally ambiguous: it is either the case that we have a postverbal pP with a prepositional order, where p precedes its PlaceP complement; or we have two separate phrases, with által being a particle that is postverbal for some reason. (62b), however, contains a particle that obviously does not form a phrase with the superessive marked PlaceP. (62) a. Ki m¯et altal tènger-˙en who went across sea-sup ‘Who has ever crossed the ocean? (lit. who went across sea)’ (Vienna C. 104)



Veronika Hegedűs b. altal-mened ez vizen across-go-inf.2sg this water-sup ‘to go across this water’

(Jókai C. 18)

The assumption is then that the higher projection in the pP can be prepositional and it can be separated from its PlaceP/PathP complement. The DP complement of strictly postpositional Place/Path heads cannot be extracted from the phrase, they have to pied-pipe the Place/Path head. The structure of the prepositional pP is given in (63): (63)

pP p

Spec

PP

p altal

P

DP

-en

tenger

With the grammaticalization of more p elements, particles became more frequent and the formerly strictly head-final PP became a mixed headed one. Particle movement also meant that the preverbal functional position PredP that became the designated landing site of particle movement (and then later more generally of predicate movement) got filled in more frequently. Chapter 1 discusses the development of PredP as part of the functional projections on the left periphery of the Hungarian clause.

. The P cycle Syntactic changes are often found to be cyclical in that the semantic bleaching of a (functional) element may lead to the reinforcement of that element and a new element filling its original role in the structure, thereby starting the cycle again.7 Changes within the PP are cyclical as well, and Waters (2009) proposes a syntactic analysis of the changes that may happen within the PP. The P-cycle starts out with nominal heads becoming Axial Parts, that is, losing some of their nominal properties. Along with lexical meaning components, grammatical features are lost as well, for example, in the Old Hungarian Axial Parts that we have seen, only the number features of nouns remained.8 Once the element loses all its nominal features and becomes a purely spatial element that expresses a location or direction in itself, it is analyzed as a Place or Path head. 7

van Gelderen () discusses the cyclical nature of changes in various domains of syntax. Svenonius () claims that Axial Parts are an intermediate category between N and P, and it varies from language to language what nominal properties they exhibit. 8

The cyclical development of Ps in Hungarian (64)



PP AxPartP DP

P

AxPart

possessor Further grammaticalization may lead to the bleaching of the spatial meaning of the adposition as well, thus it may develop into a fully functional element. According to Roberts and Rousssou (2003), P elements completely grammaticalize when they lose their spatial meaning. Some Hungarian Ps are on the way to becoming fully grammaticalized. For example, the particle meg ‘orig. back’ is now only used for the expression of telicity, the boundedness of an event, and not a spatial relation. Its original meaning back is only used in some set expressions, for example, meg-adni (tartozást) ‘to pay back (a debt)’. At this point, we argued that the element is inserted into a functional head within the extended PP, which we called pP. As we saw in the Old Hungarian examples, its original meaning still existed in its particle use, hence it could appear with motion verbs in examples like (65). (65) Es fèlèlèt v˙eu˙en almocban hog ne mennenèc meg and reply taken dream-poss.3pl-ine that not go.cond-3pl back herodèsh˙ez Herod.to ‘And being warned of God in a dream that they should not return to Herod’ (Munich C. 9ra) The other particles—while they may have non-spatial uses in some contexts—can all be used to refer to a spatial relation even in Modern Hungarian. The spatial use of meg was taken over by other particles, mostly by vissza ‘back’. Another such grammaticalized element is the suffix -nak/-nek, which was originally a P suffix with a goal meaning, but has grammaticalized as the dative suffix and secondary predicate marker with consider-type verbs. Its spatial use is quite restricted even with motion verbs and it is most often reinforced with a spatial particle (which is a duplicate of the suffix morphologically). (66) A biciklis neki-ment a kerítés-nek. the cyclist to-went the fence-dat ‘The cyclist rode into the fence.’ When the spatial meaning of the original adposition—or at least a part of that meaning—is lost, new items can be inserted into the structure in order to reinforce the spatial relation and these newly grammaticalizing elements may fill the Place or



Veronika Hegedűs

Path heads in the structure. What it means is that new AxialParts can in principle arise any time from nouns with the right meaning. Waters (2009) distinguishes between three stages in the cycle and she assumes a structure essentially similar to what we hypothesize for Hungarian. Axial Part plays a key role in the grammaticalization process, and in the first step of the cycle both AxialPart and P are filled in. In the second step the two heads are reanalyzed as one P head, they express the spatial relation together as one, but variation with the previous stage is possible. Then in the third stage there is no variation, only the structure where the Axial Part head is empty and the P head is filled with the reanalyzed spatial element. After this, the empty Axial Part head may get filled again with an element that enters the grammaticalization process, and thus the cycle may be renewed. According to this distinction of the stages, those Ps that showed syntactic variation were in the second stage of the cycle in the Old Hungarian period, and those that did not show any variation were in the third stage. The newly grammaticalizing elements, like helyében, are also assumed to be in the second stage now. The grammaticalization of this latter element was dependent on the full grammaticalization of the suffix -ben, the fact that the inessive suffix fills the Place head itself, and thereby the Axial Part head became available again for the new element hely. Another way to get into the cycle is from above, that is, from being an adjunct to becoming a functional head in the construction. This is what we saw in the development of Ps in appositive constructions.

. Conclusions This chapter has shown that the Old Hungarian period was a time of substantial changes within the Hungarian PP. Some elements were becoming postpositional heads, others were losing their morphological independence and becoming suffixes, and still others were becoming functional adpositions at this time. Changes within the PP are cyclical: The cycle often begins with a nominal element losing its nominal features and being reanalyzed as a P head. The change goes through an intermediate step, when the element may still have some nominal features but is no longer a regular noun. We analyzed this step as a grammaticalization into an AxialPart head in the extended PP, and showed that several Old Hungarian elements are at this intermediate step in their grammaticalization process. Another source of P elements is the reanalysis of a PP adjoined to another PP as a functional P element that takes the other PP as its complement. It was argued here that Old Hungarian had some elements that were on the way to being reanalyzed as such functional adpositions. These elements are the ones that take oblique case marked PPs as their complements and often share the property that they can precede the

The cyclical development of Ps in Hungarian



complement or can be separated from the complement. That is, the functional pP does not have to be head-final, and this results in the fact that the functional p elements are more loosely connected to the complement and can have their own syntactic function in the clause. The functional p elements are often preverbal particles in the clause, and we see a growing number of such elements and a growing frequency of particle use throughout the written period.

 From non-finite to finite subordination: The history of embedded clauses1 J Ú L IA B Á C S K A I -AT K Á R I A N D É VA D É K Á N Y

. Introduction In this chapter we survey the principal changes that took place in the history of Hungarian embedded clauses. We will argue that finite subordination took over nonfinite embedding alongside with the development of a functional left-periphery, that is, the CP-domain of finite embedded clauses. The leading hypothesis of this book is that between the Proto-Hungarian and Old Hungarian periods, an SOV to SVO change took place. Chapter 1 presents evidence for remnants of a head-final CP, TP, and VP in Old Hungarian. Typologically, SOV languages prefer non-finite embedding (Koptjevskaja Tamm 1994), while finite subordination is typical of SVO languages. The hypothesized SOV to SVO change thus predicts that the role of non-finite subordination decreased from Proto-Hungarian to Old Hungarian, while finite embedded clauses gained more and more importance in the language. There are no written records from the Proto-Hungarian era, and finite subordination is already present in the first written records of the language (these date back to the 12th century). This means that we cannot track the beginnings of the rise of finite subordination. In order to reconstruct Proto-Hungarian syntax, Chapter 1 employed the method of ∫-curve reconstruction of ancient languages. This method holds that new constructions in language first spread slowly, then gain momentum, and the process of spreading slows down in the end. The spread of new constructions

1 Our names are in alphabetical order. The section on non-finites is based on Éva Dékány’s work, while the section on finite subordination is based on Júlia Bácskai-Atkári’s work.

From non-finite to finite subordination



thus corresponds to an ∫-curve. Conversely, old constructions start losing ground slowly, then they decline rapidly, and their replacement slows down in the end (Kroch 1989; Croft 2000). The ousting of old constructions thus corresponds to a reverse ∫-curve. Chapter 1, section 2.1.1 argued that Old Hungarian still exhibited some rapidly vanishing non-finite constructions, which represented the last phase of reverse ∫-curves. Extending these curves backward, we arrive at the hypothesis that it was non-finite subordination that prevailed in Proto-Hungarian (cf. also É. Kiss 2013). In this chapter we complement the analysis in Chapter 2. We explore the status of non-finite and finite subordinate clauses in the written records of Old Hungarian and examine the ∫-curve of finite claues and the reverse ∫-curve of non-finites in the period between Old Hungarian and Modern Hungarian, showing that the rise of finite subordination and the fall of non-finites had not finished by Old Hungarian. Instead, both processes have continued until the present day. Thus, comparing Old Hungarian to Modern Hungarian, we find that the former has more types of non-finites, and non-finites in general are used more frequently in the language. At the same time, finite subordination gradually gains ground, and it is extended to more and more environments. The chapter is organized as follows. First, we give an overview of the distinction between non-finite and finite embedding. In section 6.3, the history of Hungarian nonfinite clauses will be examined in detail. The most interesting aspect of Old Hungarian non-finite subordination is the presence of agreement on many types of non-finite verbs, and the way the embedded subject is encoded, so these topics will receive special attention. Points of theoretical interest in this section include the gradience of nonfiniteness, the presence of overt nominative subjects without overt agreement on the non-finite verb, the non-complementary distribution between overt lexical subjects and PRO subjects of infinitives, and the existence of anti-agreement with infinitives (a phenomenon not attested in other languages, as far as we know). Finally, section 6.4 will be devoted to the changes affecting Hungarian finite clauses and to the evolution of a functional CP domain. The theoretical point of interest in this domain, and so the focus of our attention, is that the C layer is already present in Old Hungarian, but it is undergoing changes and that these changes are not unique to Hungarian but can be observed in several other languages as well. In other words, the changes to be described here follow from general principles of economy and can be linked to cyclic changes (such as the relative cycle) that contribute to the evolution of functional left peripheries in general. In particular, it can be observed in the CP-domain that the need for overtly marking finite subordination arises, which manifests in the appearance of new grammaticalized left-peripheral heads, the presence of overtly filled multiple C-layers, and a rich interaction of left-peripheral elements. The results are summarized in section 6.5.



Júlia Bácskai-Atkári and Éva Dékány

. On the definition of finite and non-finite clauses On the basis of their verbal morphology, subordinate clauses fall into two natural classes: finite and non-finite clauses. As far as the definition of finiteness is concerned, there are various approaches both in generative and non-generative grammars (see for instance Cowper 2002). Two major properties seem to be of paramount importance: finite clauses contain a tensed verb, and this tensed (finite) verb has a subject. Although the relation between subjecthood and tense cannot be viewed as one characterized by mutual entailment (see George and Kornfilt 1981), most generative analyses of finiteness still build on the relation of these two. The Inflectional Phrase (IP)—and, to a lesser extent, the Tense Phrase (TP)—is responsible not only for introducing the inflection head into the structure but also for enabling agreement between the subject and the finite verb, as well as for assigning nominative Case to it (see Chomsky 1995, 1998, 2001; Hornstein 1990, 1995). More importantly, finiteness is also related to the left periphery of the subordinate clause, that is, to the CP-domain: finite clauses are full CPs and finiteness is encoded in the C head (see Kayne 1994 and also Pesetsky and Torrego 2001). It is also infamously difficult to give a unified characterization of non-finite clauses (see Vincent 1998; Adger 2007; Ledgeway 2007 among many others), especially because while a clause can be finite only in one way, it can be non-finite in several ways (Adger 2007). There are three properties that characterize all Old Hungarian non-finites, so we define the class with the sum of these properties. Firstly, all Old Hungarian non-finites are extended verbal projections that preserve the argument structure of the base verb. Secondly, Old Hungarian non-finites either cannot head an independent main clause, or if they can, “they cannot have an independent tense interpretation, but they can only receive a modal interpretation” (Bianchi 2003). Finally, Old Hungarian non-finites don’t bear temporal, mood, and aspect affixes, and while some of them do agree with their subject, none of them distinguish the definite and indefinite conjugation like finite clauses do. Thus in this sense their agreement paradigm can be said to be defective. Non-finites typically don’t introduce a subject with independent reference (their unpronounced subject is co-referent with a DP in the matrix clause), or if they do so, that subject bears a case other than nominative. This property, however, holds only for a subset of Old Hungarian non-finites: there are several types that do introduce a referentially independent nominative subject (see section 6.3.1).

. The changes affecting non-finite subordinate clauses .. Non-finite clauses in Old Hungarian Old Hungarian had a rich system of non-finite clauses: infinitives, adjectival participles, gerunds, and adverbial participles.

From non-finite to finite subordination



... Infinitives Infinitives, marked by the suffix -ni, had either a covert controlled subject (1a) or an overt, referentially independent dative subject (1b). Control was obligatory when the matrix clause contained a potential controller DP. Referentially independent subjects were licensed only with monadic predicates whose sole argument was the infinitive, and so lacked a potential controller DP in the matrix clause (epistemic, non-directed deontic, and nominal predicates). (1) a. erezted a te angyalodat meg yzen-ny az send-pst-2sg the you angel-poss.2sg-acc prt announce-inf the isteny zyletest godly birth-acc ‘you sent your angel to announce God’s birth’ (Gömöry C. 120r) b. Hewsag [ nekthek wylaagh elewth fel kel-n-ethek: ] vanity dat-poss.2pl world in.front.of up get-inf-2pl ‘it is vanity for you to stand up in front of the world’ (Festetics C. 85) Thus PRO and lexical subjects were not in complementary distribution in Old Hungarian; either could appear in the subject position of infinitives. See also Miller (2002), Szabolcsi (2009), and Sundaresan and Mcfadden (2009) on the lack of complementary distribution between PRO and lexical subjects. ... Adjectival participles Old Hungarian had several different kinds of adjectival participles: one type employed the non-finite ending -ó/ő, and three types employed the non-finite ending -t. The -ó/ő participle, also (erroneously) known as ‘continuous’ participle, had an unpronounced, referentially controlled subject. Its base verb could be either unergative, unaccusative, or transitive.2 (2) a. az [ eci rezket-ew ] papnaki the shudder-part priest-dat ‘to the shuddering priest’ [vP eci hizolkod-o ] ordog b. Az rezogsegi ´ ´ ´ ´ coax-part devil the drunkenness ‘drunkenness is a coaxing devil’

(Jókai C. 156)

(Guary C. 7)

zegenth feel emel-ew ] c. Kyi . . . [ eci ganeebool who . . . manure-from poor-acc up raise-part ‘who raises the poor from manure’ (Festetics C. 108) The -ó/ő participial ending could be preceded by the verbal suffix -and/end. Descriptively oriented historical grammars call -and/end the future tense suffix (E. Abaffy 1991: 111). Its use on its own is illustrated in (3a), while (3b) shows how it combines with the -ó/ő participial ending. 2 In adjectival and adverbial participles, we mark the position of the gap with ec for ‘empty category’.



Júlia Bácskai-Atkári and Éva Dékány

(3) a. ha le-es-uen ymad-and-az engemeth if down-fall-part worship-mod-2sg I-acc ‘if falling down (on your knees) you worship me’

(Könyvecse 20r)

b. oluasta uala ho´g ezuilagnak megh ualtoia zyztul read-prf-3sg be.pst that this-world-dat prt savior-poss virgin-abl uona zilet-end-o ´ be-cond born-mod-part ‘she has read that the world’s savior was going to be born from a virgin’ (Kazinczy C. 20r) There is no agreement in the literature about the status of suffixes and auxiliaries with a future time reference: in some analyses they fall under the category Tense, while in others they fall under the category Modality (see van de Vate 2011: ch. 6 for a recent overview). Descriptively oriented grammars (e.g., E. Abaffy 1992, Sárosi 2003) observed that the general future was expressed by the present tense in Old Hungarian, and -and/end was restricted to the uncertain, conditional future, and so it occurred only in embedded clauses. On the basis of this fact É. Kiss (2005b) argues that the Old Hungarian -and/end suffix expressed Modality. We propose the following new arguments for É. Kiss’ analysis of -and/end as a Modal suffix. Firstly, non-finite forms in Old Hungarian are never formed from tensed verbs. If -and/end were a Tense head, we would expect it not to be followed by a participial suffix, contrary to fact. Modal suffixes, on the other hand, may co-occur with a participial ending, as in the case of the -hat/het ability/permission modal affix below. (4) az hoold mynth a čillagok:, altal lat-hat-o is lezon ´ the Moon like the star-pl through see-possib-part too be.will ‘the Moon, like stars, will be transparent’ (Sándor C. 4r) Secondly, -andó/endő may express general necessity and possibility, without any temporal orientation. As necessity and possibility are modal categories, -and/end is better described as a Modal head rather than as an instance of Tense. (5) az vy bor v˙y tomlocbè èrèźt-ènd-o ´ ´ ´ the new wine new leather.bottle-pl-ill pour-mod-part ‘new wine is to be put into new bottles’ (Munich C. 60rb) dolgok az wr istennek (6) kyk mynd lehet-end-ew possible-mod-part thing-pl the lord God-dat what-pl all ‘these are all possible for God our Lord’ (Érdy C. 510) Thirdly, if -and/end was a tense morpheme specified for future, we would not expect it to have an anterior reading. However, such anterior readings are attested (even if they are rare). These arguments support the Mod analysis of -and/end over the T analysis.

From non-finite to finite subordination



(7) ragusyabalo zarmaz-and-o Gerlandus nevew hews Ragusa-ela originate-mod-part Gerlandus named young.man ‘a young man called Gerlandus, who was from Ragusia’ (Jókai C. 162) The so-called ‘past’ participle was marked by the non-finite ending -t and obligatorily had an empty category in the position of the internal argument (the subject of unaccusatives and the object of transitives). As unergatives have no internal argument, this participle could not be formed from unergative verbs. The external argument of this participle could only be expressed as an ablative-marked DP3 (a by-phrase, cf. (8b)). As this participle could also express co-temporaneity, we will refer to it as “-t adjectival participle with a coreferential internal argument” rather than past participle. mul-t ] vetkedetti meg boczatyak (8) a. menden [ eci el away past-part sin-poss.2sg-acc prt forgive-3pl every ‘all your past sins are forgiven’ (Jókai C. 149) dolgokroli b. Meg emlekezik az [ isten tol eci meg tilt-ott ] ´ prt forbid-part thing-pl-del prt remember-3sg the god abl ‘remembers about the things forbidden by God’ (Bod C. 11r-11v) Another kind of adjectival participle ending in -t was also formed from unaccusative and transitive verbs, but its empty category (co-indexed with the modified noun) was in the position of the internal argument’s possessor. We will label this kind of non-finite as “-t adjectival participle with a coreferential possessor.” meg (9) a. & ot vala egy [[ (possessor)eci kez-e ] hand-poss.3sg prt and there was.3sg a aź-ot ] èmberi wither-part man ‘and there was a man there which had a withered hand’ (Munich C. 38ra) zeg-ot ] felesegeneci b. [[ (possessor)eci hit-e ] ´ faith-poss.3sg transgress-part wife-poss.3sg-dat ‘to his wife, who has transgressed her faith’ (Nádor C. 278v) The possessed internal argument (keze in (9a), hite in (9b)) bore the morphologically unmarked nominative case. When the base verb was unaccusative, as in (9a), the verb’s internal argument was also the subject of the participial clause. Such non-finites thus 3 In Modern Hungarian the external argument must bear the postposition által ‘by’ rather than the ablative case: () az Isten által meg-tilt-ott dolgokról the God by prt-forbid-part thing-pl-del ‘about the things forbidden by God’



Júlia Bácskai-Atkári and Éva Dékány

have an overt nominative subject (but note the lack of agreement on the participial verb). When the base verb was transitive, as in (9b), the verb’s internal argument was the object of the non-finite clause. In this case the subject of the participial clause could not be expressed overtly. Observe the similar use of the -im participle in Mansi (10), and the -m participle in Khanty (11), the Ob-Ugric languages (the closest relatives of Hungarian): (10) pukńit jakt-im e¯lmχ¯olas navel-poss.3sg cut-part person.nom ‘person whose navel has been cut’

(Riese 2001: 69)

purәś ike:-t (11) ńa:wre:mlal wo:s-na manәm child-pl-poss.3sg city-loc go-pastpart old man-pl ‘the old men whose children went to the city’ (Nikolaeva 1999: 77) The third kind of adjectival participle ending in -t employed an empty category in the place of the object and was formed only from transitive verbs. This participle had an overt nominative subject and the participial verb showed obligatory agreement with the subject. Typically, this non-finite comprised two overt elements: the agreeing participial verb and one other constituent. The latter was typically the subject (12), but in a few instances it could also be a different element (13), or it could be missing entirely (14). We will refer to this non-finite as “-t adjectival participle with a coreferent object.” bodogsagnaki (12) erdemlyók az [ ew eci megh yger-tte ] prt promise-part happiness-dat deserve-1pl the he dychósegeet glory-poss-acc ‘we deserve the glory of the happiness he promised’ (Érdy C. 96) Coronati (13) es ue´ged az [ eci nekod zorz-ott-em ] ´ ´ ´ dat-2sg procure-part-1sg crown-acc and take.imp-2sg the ‘take the crown that I procured for you’ (Kazinczy C. 17v) [ eci zy-lott-em ] fyamnak (14) az aldot zereto ´ ´ birth.give-part-1sg son-poss.1sg-dat the bless-part love-part zent vereuel holy blood-poss-with ‘with the holy blood of my blessed, beloved son that I bore’ (Nagyszombat C. 148) Compare with the similar non-finite form in Eastern Khanty: (15) [ (mä) tini-m-äm ] loγ I sell-pastpart-1sg horse ‘the horse I sold’

(Nikolaeva 1999: 79)

From non-finite to finite subordination



... The gerund The gerund of Old Hungarian also employed -t as a non-finite ending. As is characteristic of gerunds, this non-finite clause had both verbal and nominal properties. It preserved the argument structure of the base verb (either transitive, unergative, or unaccusative), its object was marked with accusative case, and it could be modified by adverbs and negation (16). (16) vetkeztem [ en erzekensegym-et iora nem sin-pst-1sg I sensibility-pl-poss.1sg-acc good-sub not byr-t-om-ba. ] hold-gerund-poss.1sg-ine ‘I have sinned in not using my sensibilities for good’ (Virginia C. 2v) The non-finite ending -t took the extended verbal projection as its complement and nominalized it: [NomP -t [clause ]]. NomP was then embedded under nominal functional projections, and the nominalized clause distributed in the clause as a noun. Owing to the presence of nominal functional projections, the nominalized clause took the possessive suffixes and the case marking of garden variety nouns (case marking reflected the grammatical role that the gerund fulfilled in the sentence, e.g., accusative, inessive, etc). Compare the possessive agreement followed by the accusative marker on -t gerunds (17) and on ordinary nouns (18). (17) haromźèr tagačme´g [ ègemèt esmèr-t-ed-èt ] three.times deny-2sg-prt I-acc know-gerund-poss.2sg-acc ‘thou shalt deny (your) knowing me thrice’ (Munich C. 81 va) (18) a. lelk-et soul-acc ‘soul’ b. lelk-ed-et soul-poss.2sg-acc ‘your soul’

(Bod C. 5r)

(Bod C. 6r)

The gerund, however, differed from ordinary nouns in that it had to be formally possessed (i.e., it had to bear possessive morphology): unlike garden variety nouns, (17) has no non-possessed variant. This non-finite form has a close parallel in Mansi, where the -ke gerundival ending is obligatorily followed by possessive agreement. The Mansi gerund is different, however, in that it always functions as a temporal adverbial and the case marking is invariantly the -t locative ending. (When used as a temporal modifier, the Old Hungarian gerund, too, bore inessive case. However, depending on the nominal function it fulfilled, the Old Hungarian gerund could also bear nominative, accusative, dative, and superessive case.)



Júlia Bácskai-Atkári and Éva Dékány

(19) màn u¯ sn jal-ke-w-t we city-lat go-gerund-poss.1pl-loc ‘when we go to the city’

(Riese 2001: 70)

The gerund’s subject could not receive case in the verbal part of the gerund, so it moved up to the nominal layers of the gerund, into the position of the possessor. Here it could be case-licensed as a possessor. The presence of the possessor explains the obligatory possessive marking on gerunds.4 (20) megakaria vol-t]-om-ban n˙ omoreitani èni [ ti ièl˙en prt-want-3sg cripple-inf I present be-gerund-poss.1sg-ine ‘will he force her (the queen) in my presence?’ (Vienna Codex 64) ièlènseg te-t]-˙e-t (21) hallottac onèkii [ ti è ´ this phenomenon do-gerund-poss.3sg-acc hear-pst-3pl he-dat ‘they heard of his doing this deed’ (Münich Codex 98 vb) If the possessor was coreferent with a matrix argument, as in (17), it underwent regular pro-drop. Its reference could be recovered from the possessive agreement on the gerund. ... Adverbial participles Old Hungarian also had four types of adverbial participles: -ván/vén, -va/ve, -val/vel, and -t participles. Adverbial participles ending in -va/ve and -ván/vén could have either an overt, referentially independent subject or an unpronounced subject coindexed with the matrix subject or object. These non-finites could be formed from unergative, unaccusative, as well as transitive verbs. (22) -ván/vén participles a· kenèr-èt a. [ Es azoc e-u˙en ] ve-ue ic and those eat-part take-pst.3sg Jesus the bread-acc ‘and as they did eat, Jesus took bread’ (Munich C. 50vb) b. kýlencz honap el mwl-waan nine month away pass-part ‘nine months having passed’

(Festetics C. 147)

c. ig [ eci meg-kotoz-u˙en a kotèlèckèl ] èl-hag-ac o-tèti ´ ´ ´ ´ this.way prt-tie-part the rope-pl-ins away-leave-3pl he-acc ‘they left him bound by ropes this way’ (Vienna C. 21) (23) -va/ve participles a. & [ m˙u alu-u¯ac ] èl vroztac otèt ´ and we sleep-part.1pl away steal-pst-3pl him ‘and they (=his disciples) stole him away while we slept’ (Munich C. 35 vb) 4 Possessors in Old Hungarian and Modern Hungarian are either morphologically unmarked or dative marked, the choice is optional. See Chapter  in this book.

From non-finite to finite subordination b. [ hal-ua ] lelic vala mellette dead-part find-3pl be-pst next.to-3sg ‘found him dead next to her’ c. kezet tew-ue ylyesnek feyere hand-acc put-part Elijah-dat head-poss-sub ‘putting his hand of Elijah’s head’



(Guary C. 103)

(Jókai C. 23)

Participles with -va/ve could optionally agree with their subject. We will take up this issue in more detail in section 6.3.3.2. Participles in -val/vel can be found in many codices, but very little is known of this type of non-finite clause. (24) meg vilagoseytateek istennek malaztyaual [ el hagy-ual prt enlighten-pass-pst.3sg God-dat grace-poss-with prt leave-part az eretneksegnek setetsegeet ] the herecy-dat darkness-poss-acc ‘he was enlightened by God’s grace, leaving the darkness of herecy’ (Domonkos C. 39v) Finally, adverbial participles in -t could have an overt, referentially independent subject with nominative case (25a), or a covert subject co-referent with the subject (25b), object (25c), or dative-marked DP (25d) of the matrix clause. These nonfinites could be formed from unergative, unaccusative, as well as transitive verbs, and obligatorily agreed with their subject. (25) a. [ azoc èuèz-ett-ec kedig ] o èlaluec ´ those row-part-3pl conj he prt-sleep-3sg ‘as they sailed he fell asleep’ (Munich C. 63vb) [ eci iar-att-oc ] gabona keźdenc b. o ˙taneitua˙nii ´ walk-part-3pl begin-pst-3pl corn he disciple-poss.pl fot źaggat-ni-oc ´ ear-acc pick-inf-3pl ‘his disciples began, as they went, to pluck the ears of corn’(Munich C. 37vb) [ eci èl-mèn-ètt-ec ] c. Es latac azokati away-go-part-3pl and see-pst-3pl those-acc ‘and the people saw them departing’

(Munich C. 41va)

[ hog io d. Es nemel´l´ècnèci [ eci a· tèmplomrol bèź˙ell-ètt-ec the temple-del speak-part-3pl that good and some-pl-dat kouèckel & aiandokockal e˙kèsitèt˙et volna ]] m¯oda ´ stone-pl-with and gift-pl-with adorn-part be-cond.3sg say-pst.3sg ‘and to some that spake of the temple, how it was adorned with goodly stones and gifts, he said’ (Munich C. 79vb)



Júlia Bácskai-Atkári and Éva Dékány

... Interim summary in Old Hungarian.

Table 6.1 below summarizes the system of non-finite forms

Table .. The system of Old Hungarian non-finite forms

infinitive adjectival participles -ó/ő -t, internal arg. gap -t, object gap -t, possessor gap adverbial participles -t -ván/vén -va/ve -val/vel gerund

independent subject

case of subject

agreement

yes, when no controller

dative

yes, optional

no no, by-phrase yes yes (w/ unaccusative V)

N/A N/A nominative nominative

no no yes, obligatory no

yes yes yes N/A yes

nominative nominative nominative N/A nominative/genitive

obligatory no yes, optional no yes, nominal

As the table shows, the often assumed correlation between nominative case and overt inflectional morphology (agreement) is not a universal property of language: -va/ve adverbial participles agreed only infrequently and optionally, yet they could have an overt nominative subject, and -ván/vén adverbial participles did not agree, but their subject was nominative if it was overt.5 Adjectival participles with a possessor gap had an overt nominative subject when the base verb was unaccusative (that is, when the internal argument also served as the subject), but these participles never agreed with their subject. That cross-linguistically there is no correlation between nominative subjects and finiteness or overt inflection in non-finite clauses is also evident in other languages. In the Turkic language Karachay-Balkar the subject of -yan participles is nominative but the participle is uninflected (26). The same pattern can be observed in Northern (Kazym) Khanty with the -әm past participle, too (27). (See Wu 2011 for further examples and Sundaresan and McFadden to appear for a recent treatment of the independence of finiteness and nominative case.) (26) oquwču al-yan kitap student buy-part book ‘the book that the student bought’

(Comrie 1998: 79–80)

5 It must be noted, though, that some researchers take the -n ending to be a third person agreement, cf. Károly , Nádasdi .)

From non-finite to finite subordination



(27) [ năŋ ewt-әm ] jOš-em χ u˘ w jăm-a ănt j˘ı-2. you cut-part hand-poss.1sg long good-lat neg become-3sg ‘my hand, which you have cut, will not heal for a long time’ (Csepregi 2012: 68) The rich inventory of non-finite forms, still in place in late Old Hungarian, has undergone dramatic changes throughout the Middle Hungarian period, and by the emergence of Modern Hungarian it has become rather impoverished compared to its previous standing. The changes were of two kinds. On the one hand, non-finite clauses gradually lost ground: some types of non-finites completely died out, and the productivity and distribution of others became narrower than before. On the other hand, those non-finites that did remain in the language came to be more typically nonfinite: some lost the ability to license referentially independent first and second person subjects or pronominal subjects or overt subjects, and others lost the ability to agree with their subject. In the next two sections we are going to discuss these processes in detail. .. Non-finite clauses losing ground From the Old Hungarian period on, finite subordination became more and more prominent, and non-finite clauses were gradually driven into the background (cf. also Gugán’s 2002 comparative study of three Old Hungarian, four Middle Hungarian, and three Modern Hungarian translations of chapters 26 and 27 from the Gospel according to Matthew). This process affected different types of non-finites to a different degree. ... Non-finites disappearing from the language The supplantation of non-finite forms had the strongest effect on -t adverbial participles and -val/vel adverbial participles. These non-finites have completely disappeared from the standard language. Adverbial participles in -t could have either a referentially independent subject (28) or a phonologically empty subject co-referent with a matrix DP (29a); both subtypes fell out of use after the era of the codices. In contemporary Hungarian, speakers would use -va/ve adverbial participles, finite subordination, or in some cases an infinitival clause instead. (28) [ Nègèd èztèndobèn Phtolomeus & Cleopatra orzagl-a˙t˙t-oc ] ´ fourth year-ine Phtolomeus and Cleopatra reign-part-3pl hozac èl Dositheus . . . & Ptolemeus o fia ´ Dosiheus and Phtolomeus he son-poss.3sg bring-pst-3pl away Fvrim e¯pl-at I¯rlm-bè Phurim letter-poss-acc Jerusalem-ill ‘In the fourth year of the reign of Ptolemeus and Cleopatra, Dositheus, . . . and Ptolemeus his son, brought this epistle of Phurim to Jerusalem.’ (Vienna C. 73)



Júlia Bácskai-Atkári and Éva Dékány

[ eci vačoral-att-a ] veue (29) a. Azockal kedig ic i a· kenèrèt those-ins conj Jesus dine-part-3sg take-pst.3sg the bread-acc & megalda and prt-bless-pst.3sg ‘And as he was eating with them, Jesus took bread, and blessed it.’ (Munich C. 32va) ¯ l˙elec [ eci all-att-a ] b. mg a· uèhm˙eti stand-part-3sg prt find-pst-3pl the donkey-acc ‘they found the colt standing’ (Munich C. 78rb) c. Mèntol vtolbzèr ke a· tizenegnc i [ eci egembè ul-ètt-ec ] ´ together sit-part-3pl all-abl last the eleven-dat c nèk-ic i ièlenec o ´ appear-pst.3sg they dat-3pl Jesus ‘Afterward he appeared unto the eleven as they sat.’ (Munich C. 53va) The Székely dialect has two adverbs that represent a small fossil of this type of non-finite, though (Károly 1956: 214): the lexicalized forms álmotta ‘sleeping’ and émëtte ‘awake’ still show the verb+t+agreement morphological make-up that characterized the productive -t adverbial participles, and their meaning is compositional. The -val/vel adverbial participle, shown in (30), also fell out of use in the standard language (though it might have been restricted to certain dialects already in Old Hungarian); it was replaced by the -va/ve adverbial participle. However, some dialects (including the most archaic Csángó dialect) have retained the -val/vel ending, too (31). (30) iarunk kelwnk . . . embereketes [ ver-uen vagdal-ual es go-1pl go.about-1pl person-pl-acc-too beat-part hew-part and meg wldwk-uen ] prt kill-part ‘we go on the loose, beating, cutting up, and killing people’ (Virginia C. 25r) (31) A legnagyobb testvérem meg van hal-val. the biggest sibling-poss.1sg prt be.3sg die-part ‘my eldest sibling is dead’ (Ivácsony 2002–2003: 44) ... Non-finites losing productivity While -t adverbial participles were completely lost and -val/vel became (or remained) dialectal, other non-finites remained in the language with crippled productivity, that is, they can be formed from a narrower class of verbs than before. The most spectacular example of this is the -t gerund, which was completely productive in Old Hungarian, and is almost completely unproductive in Modern Hungarian.

From non-finite to finite subordination (32) a. Ne zegyenletek [ alamyznaert ment-ett-ek-et ] not be.ashamed-2pl alms-final go-part-2pl-acc ‘don’t be ashamed of asking for alms’



(Jókai C. 81–82)

b. mert vetkeztem [ hytemnek tyzenket agazatyat because sin-pst-1sg faith-poss.1sg-dat twelve branch-poss-acc nem tart-at-om-ba ] es [ ellen-e vett-et-em-be ] not adhere.to-part-1sg-ill and against-3sg sin-part-1sg-ill ‘because I sinned in not adhering to the twelve branches of my faith, and in transgressing it’ (Virginia C. 7r) Only very few -t gerunds have remained that can still take possessive suffixes other than third person. (33) a. jár-t-om-ban, jár-t-od-ban, walk-gerund-poss.1sg-ine walk-gerund-poss.2sg-ine jár-t-á-ban walk-gerund-poss.3sg-ine ‘in my/your/his going about’ b. jár-t-unk-ban, jár-t-otok-ban, walk-gerund-poss.1pl-ine walk-gerund-poss.2pl-ine jár-t-uk-ban walk-gerund-poss.3pl-ine ‘in our/your/their going about’ (34) a. hol-t-om-ig, hol-t-od-ig, doom-gerund-poss.1sg-ter doom-gerund-poss.2sg-ter hol-t-á-ig doom-gerund-poss.3sg-ter ‘until my/your/his doom’ b. hol-t-unk-ig, hol-t-otok-ig, doom-gerund-poss.1pl-ter doom-gerund-poss.2pl-ter hol-t-uk-ig doom-gerund-poss.3pl-ter ‘until our/your/their doom’ (35) a. nincs nyug-t-om, nyug-t-od, nyug-t-a not.be rest-gerund-poss.1sg rest-gerund-poss.2sg rest-gerund-poss.3sg ‘I am restless, you are restless, he is restless’ b. nincs nyug-t-unk, nyug-t-otok, not.be rest-gerund-poss.1pl rest-gerund-poss.2pl nyug-t-uk rest-gerund-poss.3pl ‘we are restless, you are restless, they are restless’



Júlia Bácskai-Atkári and Éva Dékány

The rest of the remaining -t gerunds are lexicalized forms. They have become lexicalized in the third person singular form, and cannot be inflected for other combinations of person and number. These serve mostly as adverbs (36), and to a lesser extent also as nouns (37) or postpositions (38) (see Radics 1992 for a more extensive list). (36) Advs lexicalized from -t gerunds a. valami lát-t-á-ra/lát-t-á-n something see-gerund-poss-sub/see-gerund-poss-sup ‘upon seeing something’ b. valami hall-at-á-ra something hear-gerund-poss-sub ‘upon hearing something’ c. jár-t-á-nyi erő walk-gerund-poss-ful strength ‘strengh enough to walk’ (37) Ns lexicalized from -t gerunds a. nap-kel-t-e sun-rise-gerund-poss ‘sunrise’ b. valaki vesz-t-e somebody lose-gerund-poss ‘somebody’s doom’ (38) Ps lexicalized from -t gerunds a. men-t-é-n go-gerund-poss-sup ‘along’ b. múl-t-á-n past-gerund-poss-sup ‘after’ Among the adverbial participles of early Old Hungarian, -ván/vén participles were by far the most frequent. (39) zent fferencz [ fel-kel-uen ymadsagtol ] tarsanak Saint Francis up-get-part prayer-abl fellow-poss-dat eleybe mene front-poss-ill go-pst.3sg ‘Standing up from prayer, Saint Francis went to greet his fellow.’ (Jókai C. 134) They were, however, gradually ousted by -va/ve adverbial participles; by the 19th century, already -va/ve dominates in the written language (Horváth 1991). For some speakers of colloquial Modern Hungarian, -ván/vén adverbial participles sound

From non-finite to finite subordination



archaic, while the rest find them stylistically heavily marked and prefer -va/ve instead (Bartos 2009).6 The loss of productivity also affected -t adjectival participles with a coreferent possessor. It is not entirely certain how productive these participles were in Old Hungarian; only a handful of data are found in the codices. vez-ót ] yffywi (40) a. Halwan ezeket az [ eci ez-e mind-poss lose-part boy hear-part these-acc the ‘when the boy who lost his mind heard these’ (Érdy C. 199) èl-ragad-ot ] nosten mèduei b. [ eci kolk-è-y ´ ´ cub-poss.3sg-pl away-take-part female bear ‘a bear that is bereaved of her whelps’

(Vienna C. 199)

Given that the development of non-finites in Hungarian is characterized by a reverse ∫-curve, this participle must have been entirely productive at some point. Its productivity, however, has dropped close to nil in Modern Hungarian, and the generally accepted examples have a lexicalized flavor (see Nádasdi 2010 for a recent study eliciting native speaker judgments of this construction, and argumentation that at least some examples are constructed in the syntax rather than being stored in the mental lexicon). Whether the Old Hungarian period already saw the decline of this non-finite or this happened only later is not certain. But even if this participle started losing ground already in Old Hungarian, it was still more productive than it is today. In Old Hungarian the base verb of the participle could be either unaccusative or transitive (see (9b) and (40b)), while the verb in Modern Hungarian must be unaccusative (Nádasdi 2010). Furthermore, not all unaccusative verbs are acceptable either.7 In addition, in Modern Hungarian this participle must express a part–whole relationship (Nádasdi 2010), but this restriction was not operative in Old Hungarian (in (40b), for instance, the participle expresses a kinship relationship). ... Non-finites assuming a narrower external distribution So far we have seen that certain non-finites have disappeared from the language and that others have become less productive. We are now going to see that yet others remained fully productive, but in Old Hungarian they had a wider distribution than today. Infinitives, for instance, could accompany more predicates in Old Hungarian than in Modern Hungarian. In (41a) the noun meltosagh ‘honor, dignity’ takes an infinitival complement. While infinitives as complements to nouns are still possible (e.g., hiúság ‘vanity’ + infinitive), the contemporary méltóság cannot take such a complement

6

Furthermore, even for those who accept them, these participles cannot have a state adverbial reading. The participle-internal nominative DP (cf. eze in (a) and kolkèy in (b)) bears the thematic role Patient or Theme in both Old Hungarian and Modern Hungarian. ´ 7



Júlia Bácskai-Atkári and Éva Dékány

any more. In (41b) the verb ysmer ‘know’ is modified by an adjunct infinitive whose subject is controlled by the matrix object ewtet’ ‘him’. Although adjunct infinitives with object control are still possible in Hungarian, the verb ysmer cannot appear in this structure any more. In (41c) the verb tehet ‘can, able to’ takes an infinitival complement whose subject is controlled by the matrix subject. While complement infinitives with subject control are perfectly grammatical in contemporary Hungarian, too, the verb tehet is not used in this way any longer. Thus while infinitives have not lost their productivity, they can serve as adjuncts or complements to fewer predicates than before. (41) a. apostoloknak na´gh meltosagah: [ lat-ny az cristust testy disciple-pl-dat big honor-poss see-inf the Christ-acc bodily zemekkel: ] eye-pl-ins ‘it is a great honor for the disciples to see Christ with their eyes’ (Könyvecse 24r) b. hogý ewtet’ [ mý eertwenk esedez-ný ] ysmeryek that he-acc we final-1pl beg-inf know-sbjv-3pl ‘that he be known to beg (the Lord) for us’ (Festetics C. 331) c. Ees meegýs [ fel kel-n-em ] nem tehettem and yet up get-inf-1sg not can.do-possib-pst-1sg ‘and yet, I could not get up’ (Festetics C. 403) Adverbial particples with -ván/vén and -ó/ő adjectival participles have also assumed a narrower distribution. In Old Hungarian they could appear in predicative position, serving as the complement of the copula. (42) Vala kedig pèter [ al-uan ] was.3sg conj Peter stand-part ‘and Simon Peter stood’ (43) valanac [ e-uo-c & i-uo-c ] ´ be.pst-3pl eat-part-pl and drink-part-pl ‘they were eating and drinking’

(Munich C. 104va)

(Munich C. 30va)

This use completely disappeared: -ó/ő adjectival participles can only be used as nominal modifiers, and speakers who do accept -ván/vén adverbial participles in an adverbial, verb-modifying use reject them in a predicative position. Not only did -ó/ő adjectival participles lose their predicative use, but their temporal interpretation has also become narrower. As already mentioned in section 6.2, nonfinite clauses do not have an independent tense interpretation; when embedded under a matrix clause, they anchor the event time with respect to the time of the matrix event. In Old Hungarian, the event described in the non-finite clause typically precedes or is

From non-finite to finite subordination



co-temporaneous with the event in the matrix clause. In a few rare instances, however, the event described by a predicatively used -ó/ő adjectival participle is posterior to the event of the matrix clause. (44) a. ysa mend [ ozchuz ıar-ov ] vogmuc indeed all that-all go-part be-1pl ‘indeed, we are all approaching that’ (Funeral Sermon and Prayer) b. ki vala [ otèt èl arol-o ] ´ who was he-acc prt-betray-part ‘which should betray him’

(Munich C. 98va)

Both the predicative use and the posterior interpretation of -ó/ő adjectival participles were lost over time, however, and the examples in (44) would be ungrammatical in contemporary Hungarian. Their meaning would be expressed by a finite clause or by the combination of the future auxiliary fog and an infinitive. (45) a. majd mind ahhoz járul-unk one.day all that.all go-1pl ‘we will all go there’ b. mind ahhoz fog-unk járul-ni all that.all will-1pl go-inf ‘we will all go there’ In Old Hungarian and Middle Hungarian, adverbial participles ending in -va/ve (46) and -ván/vén (47) could also be coordinated with finite main clauses (Velcsov 1957; 1981; Horváth 1991; Varga 2012). This is not possible in Modern Hungarian any more. (46) Es [ bè-m¯e-u˙e-iec Moabitidisnèc videk˙ebè ] & lakoznakuala and in-go-part-3pl Moab-dat land-poss-ill and live-3pl-be.pst oth there ‘And having come into the country of Moab, they continued there’ (Vienna C. 1) (47) a. Es elewe hyw-a az O zolgalo leanyat ´ and forward call-pst.3sg the she servant maid-poss.3sg-acc zal-wan az o hazaba, ] es le vét-éé Abráth, es [ le ´ Abrat-acc and down go-part the she house-ill and off take-pst.3sg on magarol az zoor yngeth ´ ´´ own self-del the haircloth shirt-acc ‘and she called her maid, Abra, and went down into her house, and she pulled off the sackcloth which she had on’ (Székelyudvarhely C. 31v)



Júlia Bácskai-Atkári and Éva Dékány b. [ az hytetlenek azon sem eeleghód-ween ] de meeg az the incredulous-pl that-sup neither satisfy-part but even the testeet ees egy oonos edeenben rekezt-ek body-poss.3sg-acc too a tin pot-ine close-pst.3pl ‘the incredulous not being satisfied with that either, but they even hid the body in a tin pot’ (Érdy C. 336)

In some cases these adverbial participles could function as the sole predicate of a main clause, without being coordinated with another finite main clause. Such root participles are not grammatical in Modern Hungarian. (48) eztendónek alatta hal-a meg wra. year-dat under-poss.3sg die-pst.3sg prt husband-poss.3sg Akar-wan esmeg hazassagra adny myeert hogy yffyw es want-part again marriage-sub give-inf because that young and kazdagh vala; Ew kedeeg valazt-a mas eeletet rich be.pst.3sg she conj choose-pst.3sg different life-acc maganak self-dat ‘Her husband died within a year. They wanted her to marry again because she was young and rich. But she chose a different life for herself.’ (Érdy C. 371a) These data present a conundrum to standard assumptions about non-finite clauses. One of the definiting properties of non-finites is that they cannot head an independent main clause (or if they can, they can only have a modal interpretation, but the examples under consideration don’t have such an interpretation).8 Furthermore, as only similar categories can be coordinated, finite main clauses are expected to be coordinated with other finite main clauses but not non-finites (which are complement or adjunct clauses). Some descriptively oriented historical grammars suggest that in these examples the non-finite form “gets a role similar to finite verbs” (Károly 1956: 205) or it “comes near a finite form” (A. Jászó 1992: 449). We argue in line with Velcsov (1957; 1981) and Horváth (2003) that in the relevant examples the verb is not just similar to a finite predicate, but it actually is a finite form. Specifically, Old Hungarian speakers optionally reanalyzed -ván/vén and -va/ve participles as finite forms, and this naturally allowed their use as the sole predicate of a finite clause and their coordination with finite clauses. This reanalysis was optional and did not replace the previously existing non-finite structure.

8 Note, however, that adjectival participles in Khanty head finite clauses (A. Jászó , , ), and the same was possible for gerunds in Old and Early Modern Romanian (Albiou and Hill ).

From non-finite to finite subordination



Velcsov (1957; 1981) and Horváth (2003: 432) suggest that the finite use of -ván/vén was due to the analogical effect of a small group of verbs with an exceptional finite inflectional pradigm. It is typical of Hungarian throughout its history that the third person singular suffix is zero in the indefinite agreement paradigm, present tense, indicative mood. (49) a. ad-∅ give-3sg ‘he gives’ b. Fel-kèl-∅ up-get-3sg ‘he gets up’

(Bod C 3r)

(Vienna C. 161)

A small group of verbs: lesz ‘be.fut/become’, tesz ‘do, take, put’, vesz ‘take (away)’, eszik ‘eat’, iszik ‘drink’, hisz ‘believe’, visz ‘carry’, however, exceptionally took an -n ending in this case (E. Abaffy 1991; 1992).9 (50) a. tez-en do-3sg ‘he does’

(Marosvásárhely Lines)

b. vez-en take-3sg ‘he takes’

(Könyvecse 14v)

In Old Hungarian, from the earliest remaining texts (e.g., Funeral Sermon and Prayer ca. 1195, Königsberg Fragment ca. 1350) on, this group of verbs also regularly took the -n ending in third person singular, simple past, indicative mood (52). Compare the regular verbs in (51) with a zero ending in this cell of the paradigm: (51) a. ad-a-∅ give-pst-3sg ‘he gave’ b. fel-kèl-è-∅ up-get-pst-3sg ‘he got up’ (52) a. tew-n do.pst-3sg ‘he did’

(Bod C. 10r)

(Vienna C. 5)

(Jókai C. 3)

9 The -n suffixed forms appear in the early texts, too (e.g., Funeral Sermon and Prayer ca. , Königsberg Fragment ca. , Marosvásárhely lines ca. ), but some forms without -n are ambiguous between having a second person singular and third person singular subject, so in Early Old Hungarian the use of -n in the third person may not have been obligatory (E. Abaffy : ).



Júlia Bácskai-Atkári and Éva Dékány b. ve-n take.pst-3sg ‘he took’

(Jókai C. 1)

Regular verbs take an -n ending in the third person singular only in the imperative/subjunctive mood: (53) a. ad-y-on give-sbjv-3sg ‘may he give’ b. fel-kèll-en up-get.sbjv-3sg ‘may he get up’

(Bod C. 14r)

(Vienna C. 221)

Velcsov (1957; 1981) and Horváth (2003) suggest that the reanalysis of -ván/vén was triggered by the analogical effect of the third person singular -n ending of tesz, vesz, etc. We agree that the paradigm of tesz ‘do, take, put’, vesz ‘take (away)’, etc. put the process of reanalysis into motion, but add that the fact that regular verbs also take the -n ending in the (finite) imperative/subjunctive mood probably contributed to the reanalysis. Furthermore, it must have been crucial for the reanalysis that tesz ‘do, take, put’ vesz ‘take (away)’, etc. were (and still are) frequently used verbs in the language (note that these verbs have exceptional past tenses in many European fusional languages, too, and the survival of exceptional forms depends on the frequency of use). In Old Hungarian, the verbs tesz ‘do, take, put’ and vesz ‘take (away)’ used to be even more frequent than today, because these verbs were used as light verbs in ‘light verb + noun’ complex predicates with a wider range of nouns than today. For instance, the complex predicates with tesz ‘do’ in (54) are still used in Modern Hungarian, but the ones in (55) have already become obsolete. (54) a. bewnt tewtel sin-acc do.pst-2sg ‘you have sinned’ b. ue´g uaćorat totte uala ´ last supper-acc do.prf.3sg be-pst ‘was having the last supper’ (55) a. tewtell sok kart do.pst-2sg lot damage-acc ‘you did a lot of harm’ b. tot čudakat ´ do.pst.3sg miracle-pl-acc ‘he worked miracles’

(Jókai C. 32)

(Kazinczy C. 5r)

(Jókai C. 148)

(Vienna C. 208)

From non-finite to finite subordination



Over time, the paradigm of tesz ‘do, take, put’, vesz ‘take (away)’, etc. has changed, and in the standard language they don’t take the -n 3sg ending any more either in the present or the past tense. Instead, they employ the regular paradigm, with a zero agreement in both the present tense (tesz-∅ ‘does’, vesz-∅ ‘takes’, etc.) and the ordinary -t past tense (tett-∅ ‘did, took’, vett-∅ ‘took (away)’, etc). (The -n ending has remained in some dialects, though.) We hypothesize that the loss of the -n inflection from the paradigm of tesz ‘do, take, put’, vesz ‘take (away)’, etc. has contributed to the disappearance of optional reanalysis. Adverbial particples with -ván/vén are unambiguously non-finite in present day Hungarian. The exceptional finite paradigm with -n, however, cannot have had a direct effect on the finitization of -va/ve participles, as these don’t end in -n. We hypothesize that first -ván/vén forms were reanalyzed, and later this had an effect on the phonologically similar -va/ve forms. Furthermore, the third person singular past tense form of some verbs, including teremt ‘create’, vet ‘case’, hall ‘hear’, hív ‘call’, and iszik ‘drink’, also ended in a -va/ve segment (56), and this may also have contributed to the reanalysis. The renalysis must have taken place in Proto-Hungarian, as it was already in place in the Old Hungarian period. (56) a. teremt-eve . . . adamut create-pst.3sg . . . Adam-acc ‘he created Adam’

(Funeral Sermon and Prayer 2)

b. Hadl-aua choltat hear-pst.3sg death-acc ‘he heard of his death’

(Funeral Sermon and Prayer 7)

c. vet-eve wt ez munkas vilagbele cast-pst.3sg him this toilsome world-into ‘he cast him into this toilsome world’ (Funeral Sermon and Prayer 12) d. elo hiv-a ´ forth call-pst.3sg ‘he called him forth’ . . . meg iu-a e. merget poison-acc . . . prt drink-pst.3sg ‘he drank it (the poison)’

(Döbrentei C. 53v)

(Debrecen C. 73)

When the finiteness of a clause is changed, it is typically the case that finite forms get reanalyzed as non-finites. However, the change sometimes goes in the opposite direction, with non-finites being reanalyzed as finite forms (see Ledgeway 2007 and Miller 2002: ch. 4 for specific case studies in Old Neapolitan and Welsh; Evans 2007 for a cross-linguistic overview; and Chapter 2 of this volume, which claims that the Modern Hungarian -t past tense suffix came about via reanalysis of a perfective marker in Old Hungarian). The reanalysis of adverbial participles in Old Hungarian and Middle Hungarian is an example of the latter, less typical change.



Júlia Bácskai-Atkári and Éva Dékány

.. Non-finites becoming more dependent on the main clause Non-finite clauses are generally taken to have a defective C domain, lacking temporal, spatial, and speech-event information, or to entirely lack the CP (in some cases even the IP) domain (Bianchi 2003; Sigurðsson 2004; Adger 2007; Giorgi 2010; Sundaresan 2010, among many others). The more a clause is lacking such information, the more prototypically non-finite and the more dependent on the main clause it is. That is, nonfiniteness is a scalar or gradient phenomenon (see Givón 1990; Vincent 1998; Bisang 2007; Ledgeway 2007). Further properties that non-finites often exhibit and can be taken to indicate dependence on the main clause are the lack of agreement with the subject, and the lack of a referentially independent subject. Comparing Modern Hungarian to Old Hungarian, we find that not only did non-finites lose ground in the grammar, but those that did remain in the language tend to have developed a greater degree of dependence on the main clause, too. In other words, they have shifted towards being more prototypically non-finite. ... Losing the referentially independent subject Non-finite clauses often require their subject to be an empty category whose reference is determined by a DP in the matrix clause. These non-finites are thus dependent on the main clause for the identification of their subject. Old Hungarian -ó/ő adjectival participles and -t adjectival participles with a coreferent internal argument belong to this group. Other non-finites are able to license a potentially overt, referentially independent subject, thus they show a greater degree of independence from the main clause. Old Hungarian infinitives,10 -va/ve, -ván/vén, and -t adverbial participles as well as -t gerunds could license a subject without any restrictions. The -t adjectival participle with a coreferent object could certainly license a singular subject, and it possibly licensed a plural subject, too, but the latter are not attested in the remaining linguistic records. Of the Old Hungarian non-finites that could license an independent subject, three have become limited with respect to what sort of subject they may introduce. Adverbial participles with -va/ve used to place no restriction on their subject’s overtness, and overt subjects could be of any person or number. See (57a) for a first person plural subject, (57b) for a second person singular subject, (57c) for a third person plural (lexical DP) subject, and (57d) for a covert coreferent subject. (57) a. & [ m˙u alu-u¯a-c ] èl vroztac otèt ´ and we sleep-part-1pl away take-pst-3pl he-acc ‘and they stole him away while we slept’ (Munich C. 35 vb) b. [ te kezedet meg nit-ua-d: ] mendennek be you hand-poss.2sg-acc prt open-part-2sg all-pl prt tell’esednek fill-3pl ‘You having opened your hand, all are filled (with goodness).’ (Apor C. 68) 10

Only if there was no potential controller in the matrix.

From non-finite to finite subordination



c. Azert [ azoc egbè golèkez-uei-ec ] monda azocnac ´ therefore those together gather-part-pl say-pst.3sg those-dat pilatus Pilate ‘and when they gathered together, Pilate said to them’ (Munich C. 34 rb) d. az tanoytwanyok [ Nagy syr-wa ] fwtanak hozyam the disciple-pl big cry-part run-pst-3pl all-1sg ‘the disciples were running to me, crying very much’ (Apor C. 168) In Modern Hungarian these non-finites can only have a covert subject (58). (58)

∗ Meg-szület-ve

a gyereke, Jóska új életet kezdett. prt-be.born-part the child-poss.3sg Jóska new life-acc start-pst.3sg ‘His child having been borne, Jóska started a new life.’ (Komlósy 1992: 465)

Furthermore, it is highly preferred (and for some speakers, it is obligatory) for the covert subject to be co-referent with a DP in the matrix clause (Komlósy 1992; Sárik 1998; Tóth 2000a; É. Kiss 2002: ch. 9; Kenesei 2005). (These non-finites cannot agree with their subject any more either, see 6.3.3.2). (59) a. ??Beesteled-ve betértünk egy fogadóba. evening.fall-part in-go-pst-1pl an inn-ill ‘The shadows of the evening having fallen, we called at an inn.’ (Komlósy 1992: 466) haza.) A szobába be-lép-ve, a b. ∗ (Péter korán érkezett Peter early arrive-pst.3sg home the room-ill in-step-part the kutyája mindjárt elébe szaladt. dog-poss.3sg immediately to.in.front.of run-pst.3sg ‘Peter came home early. Him having entered the room, his dog immediately ran to greet him.’ (Komlósy 1992: 466) Adverbial participles with -ván/vén could have any kind of overt subject. Cf. (60a) for a first person singular subject, (60b) for a plural lexical DP subject, and (60c) for a covert coreferent subject. (60) a. Èn ked· kèrès-u˙en èn tanalčosimtol mik˙eppèn èz I conj search-part my counsellor-poss.pl-1sg-abl how this bè· tèllèsedhètnec pass-possib-cond-3sg prt ‘when I asked my counsellors how this might be brought to pass’ (Vienna C. 75)



Júlia Bácskai-Atkári and Éva Dékány ic [ aitoc bè-t˙e-u¯e ] & [ ablakoc b. Iouo ´ ´ come-pst.3sg Jesus door-pl prt-close-part and window-pl bè-rèkèźt-u¯en ] prt-close-part ‘then came Jesus, the doors and windows being shut’ (Munich C. 107 rb) c. Azert mennyetekel [ byz-uan ] therefore go.imp-2pl-away trust-part ‘therefore go away, having faith’

(Jókai C. 82)

Over time, these participles have lost the ability to co-occur with overt pronominal subjects (É. Kiss 2002: ch. 9.5). According to Nádasdi (2013), even overt lexical DP subjects are restricted to the third person singular, while Márkus (2009) remarks that some speakers reject overt independent subjects altogether, and require a covert coreferent subject instead. Finally, -t adjectival participles with a coreferent object are attested only with singular subjects in Old Hungarian, but the person of the subject is unrestricted: it can be either first (61a), second (61b), or third person (61c). [ zyl-ott-em ] fyamnak zent (61) a. az aldot zereto ´ ´ the bless-part love-part give.birth-part-1sg son-poss-1sg-dat holy vereuel blood-poss-with ‘with the blood of my blessed, beloved son that I bore’ (Nagyszombat C. 148) b. Es adom the neked es te nemednek. te and give-1sg you dat-2sg and you kind-poss.2sg-dat you ez [ te lak-t-ad ] feldet vtannad after-poss.2sg this you live-part-2sg land-acc ‘And I will give unto thee, and to thy seed after thee, the land wherein thou art living.’ (Jordánszky C. 7b) c. w´g mond rola wrwnk Iesus [ zent Mathe this.way say.3sg del-3sg lord-poss.1pl Jesus saint Matthew kenyweenek heeted reezeeben. yr-t-a ] write-part-3sg book-poss-dat seventh chapter-poss-ine ‘this is what Jesus says in chapter seven of Saint Matthew’ (Érdy C. 131) The overt subject in Modern Hungarian, however, can only be third person (singular or plural); (61a) and (61b) are ungrammatical to contemporary speakers. How could such a restriction on subjects arise? Baker (2008) argues that the following Person Licensing Condition is operative in grammar:

From non-finite to finite subordination



(62) All matrix clauses and certain embedded clauses have two special null arguments generated within the CP projection, one designated S (for speaker) and the other A (for addressee). (Baker 2008: 125) Overt first and second person pronouns in the clause get their reference from these null arguments via operator-variable agreement. We suggest that the loss of first and second person subjects with -t participles with a coreferent object can be traced back to changes in the C domain of these non-finites. Specifically, in Old Hungarian Baker’s S and A arguments could be readily introduced into the participle’s C domain, and these arguments could operator-variable agree with a first or second person subject of the participle. As this non-finite has shifted towards being more typically nonfinite, however, the introduction of the S and A arguments has become impaired: either a functional head in the C domain became defective, or the left periphery of the participle was truncated and the relevant positions were not projected any more. So in the absence of an S or A, the participle’s subject could only be third person. To summarize, several types of non-finites became constrained with respect to the kinds of subjects they can take. The -t adjectival participle with a coreferent object lost the ability to license first and second person subjects, the -ván/vén adverbial participle lost the ability to license pronominal subjects (for some speakers, all overt subjects), while the -va/ve adverbial participle lost the ability to license overt subjects. These non-finites have thus shifted towards being more dependent on the main clause. ... Losing the ability to agree It is a typical, though certainly not an obligatory, property of non-finite clauses that they do not show agreement with their subject; this is another property that can be taken to reflect the dependence of non-finites on the main clause. In Old Hungarian several types of participial verbs could agree with their subject, but some of them lost the ability to agree either completely or partially. This, too, shows that non-finites that did remain part of the language had a tendency to become more dependent on the main clause, to become more typically non-finite. The -t adjectival participle with a coreferent object and the -t adverbial participle agreed obligatorily: see (63) and (64) respectively.11 (63) Az [ teen magadnak walazt-ott-ad ] warasodban the your self-poss.2sg-dat choose-part-2sg city-poss.2sg-ine ‘in the city that you chose for yourself ’ (Thewrewk C. 2v) (64) [ t˙u a· varosba bè-men-ett-èc ] èlotocbè kel t˙unèc-tec ´ ´ you the city-ill in-go-part-2pl in.front-poss.2pl-ill come you-dat-2pl eg neminèmo èmber ´ man a certain ‘Behold, when ye are entered into the city, there shall a man meet you.’ (Munich C. 80vb) 11 The -t gerund also obligatorily bore agreement, but this is possessive agreement rather than ordinary subject-predicate agreement (recall that these non-finites are obligatorily possessed).



Júlia Bácskai-Atkári and Éva Dékány

As we have already seen, the -t adverbial participle has disappeared from the language. The -t adjectival participle with a coreferent object is still in use with limited productivity, and it still obligatorily agrees with its subject in Modern Hungarian, too. The other agreeing non-finites of Old Hungarian were infinitives and -va/ve adverbial participles. Already in Old Hungarian, these bore agreement optionally rather than obligatorily, and their ability to agree with their subject has further decreased since the Old Hungarian period. As for -va/ve adverbial participles, the lack of agreement was statistically far more frequent even in Old Hungarian. Agreeing -va/ve participles are attested in significant numbers only in Matthew in the Munich Codex (35 examples) and in the first half of the Vienna Codex (31 examples) (Károly 1956).12 There are only a handful of examples in all the other forty-some codices taken together. (65) illustrates the full paradigm.13 (65) a. sem èlèuèn˙en s˙em hal-ua-m nem tauoztatom èl neither alive neither die-part-1sg not leave-caus-1sg away ‘should I not escape it (the hand of the Almighty), neither alive, nor dead’ (Vienna C. 91) ¯ b. mel` fold t˙egedèt mg-hal-ua-d fogad¯ad ´ which earth you-acc prt-die-part-2sg accept-2sg ‘the land that you are buried in when you die’

(Vienna C. 2)

c. hogh ymar tarthassam otet meg-hal-va-ia ´ that now hold-poss-imp-1sg he-acc prt-die-part-3sg ‘so that I can hold his dead body’ (Winkler C. 116v) d. hog n¯e meg-hal-u¯a-c m˙v vèzèdelm-¯oc-bèn ´ that not prt-die-part-1pl our peril-poss.1pl-ine ‘than to die amidst this peril’

(Vienna C. 14)

e. gonozoc val-ua-toc evil-pl be-part-2pl ‘ye, being evil’ (Munich C. 18 vb) es a˙nn˙ a m˙eg-hal-ua-ioc f. o a˙t˙ta ´ she father-poss.3sg and mother-poss.3sg prt-die-part-3pl ‘her father and mother having died’ (Vienna C. 51)

12 These parts are generally assumed to be written by the same person (Szily ), so agreeing -va/ve may have been dialectal. 13 (c) is a unique piece of data in the sense that there are no other examples in all the codices in which a -va/ve participle bears the sg agreement -ja/je. According to one theory, the sg agreement suffix of -va/ve participles was actually -n (see Károly ). In this approach -ván/vén participles do not constitute a separate type of non-finite clause, they are in fact agreeing -va/ve participles (-vá+n/vé+n, the lengthening of a and e to á and é is a regular phonological process). This analysis has both advantages and disadvantages, but discussing them here would take us too far afield.

From non-finite to finite subordination



Over time these participles have completely lost their ability to agree with their subject; in Modern Hungarian they can only be uninflected. Agreeing infinitives were common in Old Hungarian. In some codices infinitives with agreement are less frequent than infinitives without agreement (Jókai Codex: 121 with and 305 without, Vienna Codex: 150 with and 262 without, cf. Károly 1956), while in others it is the other way around (Guary Codex: 79 with and 59 without, Könyvecse: 20 with and 6 without, cf. Dékány 2012b). In Old Hungarian the infinitive’s ability to agree did not correlate with any other syntactic properties of the non-finite clause. The matrix subject could control the subject of both object and adjunct infinitives (66), the matrix object could control the subject of adjunct infinitives (67), and the matrix dative could control the subject of either subject or object infinitives (68). Agreement was possible but not obligatory in all of these configurations. In the interest of space, only agreeing infinitives are shown below. (66) subject control a. ne akariatoc fel-n-etèc not want-imp-2pl fear-inf-2pl ‘don’t want to be afraid’ (Munich C. 42ra)

object inf.

b. Mert nem iottem hy-n-om igazakot ´ because not come-pst-1sg call-inf-1sg good.and.true-pl-acc ‘I have not come to call the good and true’ (Döbrentei C. 205v) adjunct inf. (67) object control a. èn èrèźtettèlec t˙utoket arat-n-otok ´ I send-pst-1sg you-pl-acc harvest-inf-2pl ‘I sent you to reap’ (Munich C. 88rb)

adjunct inf.

(68) dative control a. legyen alkolmas ennekem zol-n-om tynektek let.be appropriate I-dat-1sg say-inf-1sg you-dat-2pl ‘let it be appropriate for me to speak to you’ (Jordánszky C. 712) subject inf. b. hagyad en nekem be tellyeseyt-en-em azt. ammy-re let-2sg I dat-1sg in fulfil-inf-1sg that-acc what-sub ievttem. come-pst-1sg ‘let me fulfill what I have come for’ (Cornides C. 113v) object inf. Control infinitives in Old Hungarian could also optionally anti-agree with their subject; that is, show 3sg agreement with a non-3sg subject (Dékány 2012b). (69) Ne akaryatok feel-ny-e not want-imp-2pl fear-inf-3sg ‘do not want to be afraid’

(Jordánszky C. 55)



Júlia Bácskai-Atkári and Éva Dékány

Compare with the same sentence with regular agreement and without agreement on the infinitive: (70) a. ne akariatoc fel-n-etèc not want-imp-2pl fear-inf-3pl ‘do not want to be afraid’

(Munich C. 42ra)

b. Ne akaryatok ty ffel-ny not want-imp-2pl you fear-inf ‘do not want to be afraid’

(Jordánszky C. 450)

This might be a sign that infinitival agreement started to become less strong in this period. While infinitives have retained their ability to agree with their subject to date, an important restriction came into force: their ability to agree now correlates with the controller’s case marking. Specifically, only dative control allows agreeing infinitives (Tóth 2000b; 2011).14 Furthermore, anti-agreeing infinitives are only found with third person plural subjects in Modern Hungarian. Thus, the across-the-board optionality of agreement that characterized Old Hungarian has been lost, and infinitival agreement has become subject to strict constraints. To summarize, those non-finites that obligatorily agreed with their subject in Old Hungarian still do today (unless the non-finite in question has been lost from the language itself), while those non-finites that optionally agreed with their subject in Old Hungarian have lost this ability completely or partially. .. Linguistic fossils from the head-final period It is the hypothesis of this book that Proto-Hungarian was a head final SOV language, and this word order had changed to Topic Focus V X∗ before the Old Hungarian period, i.e., before the emergence of written documents (see also É. Kiss 2013). Lightfoot (1991) argues that syntactic innovations always begin in finite matrix clauses. The change affects non-finite subordinate clauses only later on, thus these retain the original order for a longer period. In Old Hungarian the dominant word order is already Topic Focus V X∗ (or SVO) in both finite matrix and non-finite subordinate clauses. The latter, however, do indeed preserve some SOV-related features that are not attested in finite matrix clauses. ... Head-final non-finite clauses The most obvious remnant from the SOV period is the existence of strongly or strictly head-final non-finite clasues. The -t adjectival participle with a coreferent possessor, for instance, is always head-final. meg aź-ot ] èmberi (71) & ot vala [ eci egy kez-e a hand-poss.3sg prt wither-part man and there was ‘and there was a man there which had a withered hand’ (Munich C. 38ra) 14 É. Kiss (; b), on the other hand, argues that control infinitives don’t agree at all, inflected infinitives are possible only when the infinitive’s subject is not controlled but has independent reference.

From non-finite to finite subordination



This also holds for -t adjectival participles with a co-referent object. (72) mynden [ Isten eci at-t-a ] ember el yokati every God give-part-3sg good-pl-acc man prt feledne forget-cond.3sg ‘and man would forget all the good given by God’ (Érdy C. 129) Old Hungarian -t gerunds are also strictly head-final: the object, negation, and adverbs all precede the nominalized verb. (73) vetkeztem az vr istenek elene. [ en sin-pst-1sg the lord God-dat against-poss.3sg I erzekensegymet iora nem byr-t-om-ba ] sensitivity-poss.pl-1sg-acc good-sub not have-part-1sg-ine ‘I have sinned against God, in not using my sensitivities for good’ (Virginia C. 2r) Adjectival participles ending in -ó/ő also have a strong tendency to be head-final, though a few counter-examples do exist (75). an´galoci (74) az [ eci haborusagot zerz-o ] › unrest-acc make-part angel-pl the ‘the angels who brought war’ (75) vag . . . [ meg-bočat-o gonossagokat ] be.2sg prt-forgive-part evil-pl-acc ‘forgivest thou the evil’

(Guary C. 127)

(Vienna C. 244)

There is a clear tendency for head-final structures with -t adverbial participles, too. Károly (1956) has determined that in the Jókai Codex, Vienna Codex, and Munich Codex, there are altogether 83 -t adverbial participles (1, 17, and 65 respectively). We have checked these participles and have found that 17 have an overt object (16 cases with a DP object and 1 case with a clausal object). Of these, only 4 are VO, and 13 are OV. We suggest that the few counter-examples to the head-final character of -ó/ő adverbial participles and -t adverbial participles involve clause-internal right dislocation. Chapter 2, section 2.3.2 argues that right dislocation played a crucial role in the SOV to SVO reanalysis of Proto-Hungarian. The participles mentioned above have retained their strongly or strictly head-final character up to the present day (except for the -t adverbial participle, which has been lost), preserving the old SOV order in a fossilized form. Adverbial participles with -val/vel, -va/ve, and -ván/vén had lost their previous head-final nature by the Old Hungarian period, and examples in which one of the verb’s arguments or an adverb follows the participial verb were not rare.



Júlia Bácskai-Atkári and Éva Dékány

... Preverbal unmarked objects This book hypothesizes that Proto-Hungarian was an SOV language, and the preverbal object was morphologically unmarked. Over time it became possible for the object to bear overt case-marking, and this allowed it to appear not only in the immediately preverbal position but elsewhere as well. Without the overt object marking it would not have been possible for the word order to become more relaxed, as the subject and object could not have been distinguished either on the basis of their position or their morphology. Non-finite clauses in Old Hungarian still feature morphologically unmarked objects from time to time, and such objects are always found in the immediately preverbal postion. In other words, non-finites are still able to show the previous SOV order with an unmarked object in a limited way (see also section 2.2.1.1). This is illustrated below for infinitives (76a), -ó/ő adjectival participles (76b), -t adjectival participles with a coreferent possessor (76c), -va/ve adverbial participles (76d), -ván/vén adverbial participles (76e), and -t adverbial participles (76f). (Obviously, -t adjectival participles with a coreferent internal argument or object don’t have overt objects.)15 (76) a. mykoron ez soror megyen vala [ az ora meg lat-ny ] when this sister go-3sg be.pst the clock-∅ prt see-inf ‘when this sister was going to check the clock’ (Margaret Legend 7v) tou-o-k ] b. [ o igeie ´ ´ his word-poss.3sg-∅ do-part-pl ‘those fulfilling his words’

(Apor C. 66)

c. agyad meg ymmar [ bewn-e zan-t-nak ] give.imp-2sg prt now sin-poss.3sg-∅ grieve-part-dat ‘give it to the one who is grieving his sins’ (Jókai C. 158) d. Te kedig [ alamisna t˙e-u˙e-d ] you conj alms-∅ do-part-2sg ‘and ye doing your alms’ e. [ az aitoc meg-˙nit-uan ] ki zalada the door-pl-∅ prt-open-part out run-pst.3sg ‘and opening the doors, he ran out’

(Munich C. 12ra)

(Vienna C. 171)

f. Az paraztrol ky zent fferenczet lewlteuala [ egyhaz the peasant-del who Saint Francis-acc find-prf-3sg-be.pst church-∅ sepr-ett-e ] sweep-part-3sg ‘about the peasant who found Saint Francis sweeping the church’ (Jókai C. 97) 15 (c) is an elliptical structure, where the noun ‘man/person’ modified by the participle has been elided, and the dative case marker of this noun leans on the participle for phonological support. The full structure of this example is as in (i), with the elided noun marked by ∅.

() bewn-e zant-∅(person) -nak

From non-finite to finite subordination



This fossil from the Proto-Hungarian period is already unattested in Old Hungarian finite clauses; it is featured only in a small part of the data in Old Hungarian nonfinites. Károly (1954; 1956) has found that in the Jókai Codex, for instance, out of 240 -va/ve and -ván/vén adverbial participles only 35 have unmarked objects. Unmarked objects have not survived into standard Modern Hungarian. In the most archaic Csángó dialect, however, preverbal unmarked objects are still possible (Hoppa 2012: 72).16 .. Interim summary Table 6.2 summarizes how non-finite clauses have lost ground from the Old Hungarian period to Modern Hungarian. Basically only -t adjectival participles with a coreferent internal argument (a.k.a. adjectival past participles, cf. English the fallen leaves, the reserved tables) have not suffered any loss, all the other non-finites underwent some change or another that resulted in a narrower distribution and/or greater dependence on the main clause. The gradual extension of finite subordination at the expense of non-finite subordination can also be documented in a number of other Indo-European, Uralic, and Tungusic languages. Ancient Greek, for instance, frequently employed infinitives. The category of infinitives, however, has been completely lost from the language, and Modern Greek makes use of finite subjunctives with a controlled PRO instead (Terzi 1992; 1997; Sampanis 2011). In Russia, the Tungusic and Uralic minority languages with an SOV word order are currently undergoing a shift towards finite relative subordination under contact with and cultural pressure from Russian. The Tungusic language Ewenki began to use finite relatives headed by a relative pronoun instead of non-finite relatives using the gap strategy (Comrie 1998). The same process is in effect in the Uralic language Khanty, one of Hungarian’s sister languages. In Khanty, the shift has three stages (Csepregi 2012). In the first stage, prenominal non-finite relatives using the gap strategy are replaced by postnominal non-finite relatives (still using the gap strategy), and the participial agreement is dropped. In the second stage a proto-relative pronoun is included in the postnominal non-finite relative, and the nonfinite form is used more as a predicate rather than an adnominal modifier. Finally, in the third stage the non-finite form is replaced by a finite one, and a proto-relative pronoun is near-obligatory in the clause. Thus, the shift from non-finite to finite complementation goes hand in hand with the formation of a left periphery, where the relative element is housed. (The relative cycle in Old Hungarian will be detailed in section 5.4.) Note, however, that language change may also proceed in the other direction: in Amharic (SOV), for instance, it is non-finite subordination that is gaining ground at the expense of finite embedding (Koptjevskaja Tamm 1994). 16 In the database of Hoppa, however, there are unmarked objects in postverbal position in matrix clauses as well.

Table .. Changes in the use of non-finite clauses from Old Hungarian to Modern Hungarian (the comments refer to the current state)

infinitive adjectival participles -ó/ő

disappeared entirely

narrower external distribtuion

narrower class of base verbs

independent subject loss

agreement loss

no

yes, with fewer predicates

no

no

yes, some

no

no

N/A

N/A

no no

N/A some loss, no 1st or 2nd person N/A

N/A no

N/A N/A

no

N/A yes (some: only pronominal, or only 3sg or only coreferent) yes (only covert subject)

yes, total

N/A yes

N/A N/A

N/A N/A

-t, internal arg. gap -t, object gap

no no

yes, no predicative/posterior adnominal use no no

-t, possessor gap

no

no

yes, no transitive and only some unacc.

yes for some

N/A yes, no predicative use, coordination w. finites or use as finite yes, no coordination with finites N/A no

N/A no

adverbial participles -t -ván/vén

-va/ve -val/vel gerund

no remained dialectal no

N/A

From non-finite to finite subordination



In the first Old Hungarian texts, finite subordination is already in place, so we cannot track the process by which embedded finite clauses began to emerge. There are two possible ways in which this could have happened. First, it is possible that the finite C layer emerged as head-final, in keeping with the general head-final character of Proto-Hungarian, and it was later re-analyzed as a head-first layer. Support for this position comes from the fact that while Old Hungarian yes/no questions normally feature the interrogative particle -e attached to the verb, sporadically it still occurs in a clause-final position (then it is always accompanied by a clause-initial negative interrogative discourse particle (mi)nemde, see Chapter 2). mondatic marianac e˙ (77) nemde o ańńa ´ qprt he mother-poss.3sg say-pass.3sg Mary-dat q ‘Is his mother called Mary?’

(Munich C. 20 va)

In a few cases the interrogative particle -e appears both in a clause-final and a verbadjacent position. (77) and (78) show that finite CPs are sporadically still head-final in Old Hungarian. (78) Mínemde elfeledhetí-e az ańa v´ kis qprt prt-forget-possib-3sg-q the mother she small g´ ermoket-e ´ child-poss.3sg-acc-q ‘Can the mother forget her small child?’

(Nádor C. 26r)

The intermediate stage represented by (78) is attested in contemporary finite object clauses in Udmurt (Finno-Ugric) as well, with the native šuisa ‘that’ occupying a ˇ clause-final position, and the Russian loanword čto ‘that’ appearing clause-initially (cf. Tánczos 2013 for a detailed discussion of when such complementizer doubling occurs). Udmurt features a proliferation of non-finite complementation and is currently undergoing an SOV to SVO change that also took place between Proto-Hungarian and Old Hungarian, so the processes that we can observe in this language may show us what might have happened in the period of Hungarian that preceded the era of linguistic records. (79) Mon malpaśko, čto ton bertod šuisa. ˇ I think-1sg that you get.home-fut-1sg that ‘I think that you get home.’

(Tánczos 2013)

The other logically possible course of development is that the finite C layer came to be introduced as head-initial in the first place, with a head-initial finite CP dominating a head-final TP and VP in Proto-Hungarian. (Chapter 5 analyzes the development of PPs in this vein: it is argued that the newly introduced pP layer starts its life as a head-initial projection dominating a head-final PP.) Neither option would violate the Final Over Final Constraint (Biberauer et al. 2007; 2008a,b; 2009),



Júlia Bácskai-Atkári and Éva Dékány

which predicts that diachronic changes from head-final to head-initial proceed top down.17 In the next section we are going to examine how finite clauses developed a finegrained left periphery in Old Hungarian and how they spread in the language.

. The development of finite subordinate clauses This section aims at providing an overview of how finite subordinate clauses developed in Old and Middle Hungarian and how the system of finite subordinate clauses became enriched as the importance of non-finite subordination diminished. The processes will be linked to structural changes affecting the CP-domain of embedded clauses, and we will show that these changes follow from general economy principles and hence are present in several other languages as well. In this way, the findings concerning the development of Hungarian subordination are crucially important in cross-linguistic terms as well. .. Finiteness and the CP-domain As was mentioned in the introduction, finiteness is also related to the left periphery of the subordinate clause, that is, to the CP-domain: finite clauses are full CPs and finiteness is encoded in the C head (see Kayne 1994 and also Pesetsky and Torrego 2001)—in a cartographic approach such as that of Rizzi (1997), the lowest CP is headed by the C responsible for Finiteness, while the highest one is responsible for Force. For the present investigation, what is crucial is that the marking of finiteness, as well as the marking of diverse kinds of finite clauses, is related to the CP-domain of the subordinate clause. The increasing importance of finite subordination over non-finite structures brought about changes in the CP-domain too, which fall into two major categories. On the one hand, if there was a CP-layer in Proto-Hungarian (an SOV language), then it was presumably head-final, as shown by the clause-final position of the interrogative marker -e in main clause questions (on this and for further arguments, see Chapter 1 of this volume); the evolution of a functional leftperiphery of finite subordinate clauses brought about a change from head-final to head-initial CPs. On the other hand, there are changes that can be observed within a head-initial CP domain, and these changes have two main aspects. First, one type of change involved the grammaticalization of various elements in the CP-domain, that is, elements that had previously appeared in the CP-domain only as a result of movement from within the clause now became C heads. Second, the evolution of a rich system of various 17 Note, however, that at the same time Old Hungarian also appears to present a counterexample to the Final Over Final Constraint. Old Hungarian was not an OV language any more, yet auxiliaries strictly followed the main verb, which means that VP was not head final, but TP/AspP still was.

From non-finite to finite subordination



C elements in diverse positions also enabled the combinations thereof and, until the point when the grammaticalization of all C elements into the highest node was completed, several complementizer combinations are attested. The combinations partly involved the marking of new functions, or the combination of existing functions, but were partly the result of reinforced marking of finite subordination. .. The diachronic system of finite clauses in Hungarian As far as the structure of the left periphery is concerned, we basically adopt Rizzi’s analysis, which claims that the CP is iterable, such that there are two CP projections,18 between which the optional Topic and Focus, when present, are situated (topics are iterable), if there are any (see Rizzi 1997: 297; 2004: 237–238): (80) [CP [TopP∗ [FocP [TopP∗ [CP]]]]] In what follows we will mainly be concerned with the C heads and the intermediate topic and focus projections will not be of much interest, especially because in Hungarian topics and focus normally occur below the CP-domain (see É. Kiss 2002).19 Apart from the C heads themselves, operators may also occur in the CPs, that is, operators may move to the specifier of a CP (see Chomsky 1977: 87; Kennedy and Merchant 2000: 89–90); in Hungarian, this is the lower [Spec,CP] position (see Kántor 2008). Though typically there is only one overt C head in the structure, some languages may allow both C heads to be filled at the same time. Consider the following example from Welsh (from Roberts 2005: 122): (81) Dywedais, i mai ‘r dynion fel arfer a werthith y ci. say I that the men as usual that sell the dog ‘I said that it’s the men who usually sell the dog.’ 18 In Rizzi (, ) the two CP projections are distinguished as ForceP (the higher CP) and FinP (the lower CP); that is, the higher CP is responsible for clause-typing and the lower one for defining finiteness. Though it seems to be true that Force-marking subordinators eventually become higher C heads, the clear-cut distinction is problematic for several reasons, and since Force-marking complementizers are unambiguously associated with finiteness (that is, they are all finite, at least as far as Hungarian is concerned), the distinction is not important here for our purposes. Hence the two CP projections will rather be referred to as higher and lower CP, which is not incompatible with the system outlined by Rizzi (, ) but does not exclude the possibility of a more refined theoretical analysis either. 19 Note that this is not incompatible with the system outlined by Rizzi (, ) since the main argument there is that such projections (TopP, FocP) may appear between the two CPs but it does not exclude the possibility of topics and foci appearing elsewhere in the structure. In Hungarian, for instance, these elements appear lower than the CP-domain and their positions are as follows (Brody a, b, ; É. Kiss , c; Kántor ; Bácskai-Atkári and Kántor ): (i) [CP [CP [TopP∗ [FocP]]]]

This shows that normally there are no elements intervening in between the two CPs; however, in certain cases topics may optionally occur there, as will be shown later (Bácskai-Atkári and Kántor ), hence: (ii) [CP [TopP∗ [CP]]] Furthermore, as will be shown in section ..., historically certain polarity markers (heading a PolP) could also appear as intervening elements.



Júlia Bácskai-Atkári and Éva Dékány

There are four major complementizers that have to be considered in the history of the Hungarian language: hogy ‘that’, ha ‘if ’, mint ‘as/than’ and mert ‘because’. Though in Modern Hungarian they are all complementizers located in the higher C node, historically they all derive from operators that moved to the specifier of the lower CP (cf. Juhász 1991a: 479–481; 1992: 781, 783–785, 801; Haader 1991: 729–737; 1995: 510–677). The functional split from these original operator functions did not take place at the same time, which also has a bearing on whether they still have their etymologically related operator counterparts in Modern Hungarian. The differences are summarized in Table 6.3:

Table .. The major complementizers and the related operators in Hungarian Complementizer

Original operator

Time of split

Present-day related operator

ha ‘if ’

ha ‘when’

before Old Hungarian − Early Old Hungarian

−

hogy ‘that’

hogy ‘how’

before Old Hungarian − Old Hungarian

hogyan ‘how-Int’, ahogy ‘how-Rel’

mint ‘than/as’

mint ‘how’

Old and Middle Hungarian

miképpen ‘how’, miként ‘how’, amint ‘how-Rel’

mert ‘because’

mert ‘why’

Old and Middle Hungarian

miért ‘why-Int’, amiért ‘why-Rel’

One major development in terms of the CP-domain was hence the grammaticalization of operators into C heads. Another aspect of the CP-layer being reinforced was the appearance of complementizer combinations. Consider Table 6.4. As can be seen, the system is symmetrical: if a given combination existed in the order XY, then it also existed in the YX order, such that the original meaning of the

Table .. Complementizer combinations in Hungarian

ha hogy mert mint

ha

hogy

mert

mint

– hogyha – mintha

hahogy – merthogy minthogy

– hogymert – –

hamint hogymint – –

From non-finite to finite subordination



Table .. The overview of complementizer grammaticalization in Hungarian Original (extinct) order

Grammaticalized (surviving) combination

hahogy ‘if that’ hogymint ‘that than’ hogymert ‘that because’ hamint ‘if as’ hogynemmint ‘that not than’ hogysemmint ‘that neither than’ hogyhamint ‘that if as’

hogyha ‘that if ’ minthogy ‘than that’ merthogy ‘because that’ mintha ‘as if ’ – mintsemhogy ‘than neither that’ minthogyha ‘as that if ’

two was the same. While in each pair both members are attested in Old and Middle Hungarian, it is invariably only one member that survives into Modern Hungarian; these are, as highlighted in Table 6.4, hogyha ‘that if ’, merthogy ‘because that’, mintha ‘as if ’ and minthogy ‘as that’. Apart from the basic C + C combinations given in Table 6.4, there are also combinations involving negative-like elements and ones that can morphologically be decomposed into more than two C heads. These additional combinations also tend to appear in symmetrical configurations and if so, then it is again only one of the orders that survives (in Standard Hungarian). As will be shown later on in this chapter, the surviving order is never the original one but the one derived from that. This is summarized in Table 6.5. The appearance of the combinations in the left-hand column of Table 6.5 is due to the reinforcement of the CP-domain by filling both C positions with overt elements, while the evolution of the right-hand column combinations is the result of all C elements being reanalyzed as Force-marking C heads. .. The evolution of complementizers The evolution of complementizers from the original operators involved two successive steps of reanalysis. First, one type of reanalysis was responsible for the reinterpretation of operators into (lower) C heads. This is in line with the mechanism of the relative cycle, where an operator—an original pronoun—is reanalyzed as a complementizer head, cf. Roberts and Roussou (2003), van Gelderen (2009). This is also attested for English that, and is hence far from being language-specific. Second, a further step of reanalysis caused elements to be reanalyzed from lower C heads to higher C heads, which is again attested in the case of English that (see van Gelderen 2009).



Júlia Bácskai-Atkári and Éva Dékány

The two processes are summarized in (82): (82)

CP

CP

C’

C’

C

C

CP X

X

C’ C X

CP

…

C’ C

…

X

As can be seen in the left-hand side diagram, an element X (an operator) that is located in the lower [Spec,CP] position is reanalyzed as the head of that CP (hence as a complementizer). The second step is shown in the right-hand side diagram: the element X (a complementizer) is reanalyzed as a higher C head (hence still a complementizer). Both steps are motivated by economy and are hence required by general principles governing linguistic processes. The relevant requirements on economy are summarized in terms of the Head Preference Principle (HPP) and the Late Merge Principle (LMP) by van Gelderen (2004), both going back to the idea that Merge is preferred over movement (see Chomsky 1995). The HPP states that it is preferable to be a head than a phrase, i.e., base-generation is preferred over movement—hence the reanalysis from operator to complementizer. The LMP states that it is more economical to be base-generated in a higher position than to be moved to that position—hence the reinterpretation of the original lower C as a higher one. The reason behind this latter step is simply that it is the higher C head that is responsible for defining the Force of the clause and the fact that certain overt lower C heads become associated with carrying Force implies that these elements also start moving up to the higher C head. This again leads to a choice between movement and base-generation at a higher point in the structure—and just as in the case of the HPP, the latter configuration is preferred. As has been mentioned earlier, the functional split between the original operators and the new complementizer functions took place at different times (see Table 6.3 in section 6.4.2). That is, while for hogy ‘that’ and ha ‘if ’ it happened before the Old Hungarian period and partly in Early Old Hungarian, for mint ‘than/as’ and mert ‘because’ it took place in Old and Middle Hungarian. This led to a difference in their typical positions in Old and Middle Hungarian: ha was invariably an upper C head, while hogy was typically an upper C head but could also be base-generated in the lower C position. By contrast, mint and mert were either lower C heads or were still located in the lower [Spec,CP] position.

From non-finite to finite subordination



The positional differences will be important especially in terms of combinations; for the time being, let us focus on the evolution of the individual C heads. There are two fundamental ways in which they contributed to the shift from non-finite to finite subordination. On the one hand, the general finite subordinator hogy ‘that’ was extended in its functions and came to be preferred over non-finite structures. On the other hand, the appearance and the strengthening of specific complementizers also meant that finite subordinate clauses could be used for several functions. ... The evolution of hogy ‘that’ The complementizer hogy ‘that’ is etymologically related to the operator hogy ‘how’ and the split between the two can be dated back to the period prior to Old Hungarian. Hence, even the early texts display the complementizer function in the vast majority of the cases, though there are still some examples for the original operator function: (83) furiscte musia etetý ýmletí ug hug ana bathe-3sg wash-3sg feed-3sg breastfeed-3sg so how mother sciluttet child-poss.3sg-acc ‘she bathes, washes, feeds and breastfeeds him as a mother does her child’ (Königsberg Fragment) Since hogy was grammaticalized relatively early as a (lower) C head, it appears as a higher C head already in Old Hungarian and was typically base-generated in this position and only rarely as a lower C head—in the latter case, it preferably moved up. The importance of this, as well as the arguments in favor of this stance, will be discussed in section 6.4.4 in detail. Most functions of hogy are attested in both Old/Middle and Modern Hungarian. First, hogy introduces finite declarative subclauses (that-clauses), as shown by the following example from Old Hungarian (note that in this function hogy alternates with the zero): (84) hallotta vala hog vr tèk˙ent˙ett˙e volna hear-perf-3sg be-pst that Lord see-perf-3sg be-cond on˙ep˙et ´ he-people-poss.3sg-acc ‘she had heard in the country of Moab that the Lord had visited his’ (Vienna C. 1) Second, hogy introduces embedded imperatives (again, it may alternate with zero): (85) a. & kèzdec k˙erni hog e˙ltauoznec o ´ and begin-pst-3pl ask-inf that off-depart-cond-3sg they videkecbol ´ coast-poss.2pl-ela ‘And they began to pray him to depart out of their coasts.’ (Munich C. 40ra)



Júlia Bácskai-Atkári and Éva Dékány b. dawyd keńyerewk Istennek hogy az ew ellensegyt meg David beg3sg God-dat that the he enemyposs.pl.3sg-acc prt roncza destroysbjv-3sg ‘David begs God to destroy his enemies’ (Apor C. 2)

Third, hogy appears optionally (that is, alternating with the zero) in embedded whquestions together with the wh-pronoun itself, resulting in the sequence hogy + interrogative pronoun: kozottoc nagob volna (86) vetokodtec vala hog ki o ´ ´ ´ ´ ´ ´ dispute-perf-3pl be-pst that who they among-3pl greater be-cond.3sg ‘they had disputed among themselves, who should be the greatest.’ (Munich C. 45rb) Fourth, hogy introduces purpose clauses from Old Hungarian onwards: (87) a. Meńńètec a rokon falucba & varosocba hog ot es go-imp-2pl the nearby village-pl-ill and town-pl-ill that there also p¯ dical’l’ac preach-sbjv-1sg ‘go into the next towns, that I may preach there also’ (Munich C. 37ra) vala emberi b. ada az kouetnek e´g kouet, ki ´ ´ give-pst-3sg the ambassador-dat a stone-acc which be-pst human zemnek hasonlatossagara, ho vinneek o ´ ´ eye-dat similarity-posssub that take-cond-3pl they vroknak lord-poss.3pl-dat ‘he gave the ambassador a stone, which was similar to a human eye, so that they take it to their lord’ (Bod C. 4r) In this case hogy is responsible for encoding that the subclause expresses purpose and hence cannot be replaced by the zero. Fifth, hogy is also responsible for introducing clauses with a consecutive meaning; in these cases the subclause is attached to a degree expression (DegP—olyan ‘so’ or úgy ‘so’) in the matrix clause: (88) a. Es oz gimilsnec wl keseruv uola vize hug and the fruit-dat so bitter be-pst.3sg water-poss that turchucat mige zocoztia vola throat-poss.3pl-acc prt cut-3sg be-pst ‘and the fruit tasted so bitter that it hurt their throats’ (Funeral Sermon and Prayer)

From non-finite to finite subordination



b. ug vigaziatoc m¯eden idobèn imadkozuan hog m˙eltac ´ so watch-imp-2pl all time-ine pray-part that worthy-pl legètec èltauoztatnotoc m¯ed èzekèt be-sbjv-2pl off-leave-caus-inf-2pl all these-acc ‘watch ye therefore, and pray always, that ye may be accounted worthy to escape all these things’ (Munich C. 80va) egbè ug hog sem a hazba c. & sokan golekezenc ´ and many gather-pst-3pl together so that neither the house-ill sem az aitohoz n¯e f˙ern˙enc neither the door-all not reach-cond-3pl ‘And the multitude cometh together again, so that they could not get either into the house or to the door’ (Munich C. 37rb) It is worth mentioning that the sequence of úgy and hogy was reinterpreted into the coordinating conjunction úgyhogy ‘so that’ (cf. D. Mátai 2003: 423; Rácz 1995: 699– 702); this may be the case in example (88c) as well. The same is not true for olyan since it was typically not adjacent to the subordinate clause in the linear structure: the adjective or the noun modified by olyan appears between the two and the verb may do so too. The mechanism of this kind of reinterpretation will be addressed in section 6.4.5 in more detail. Apart from the five functions mentioned above, the subordinator hogy had one additional function historically: it introduced comparative subclauses (either ones expressing equality or ones expressing inequality), typically co-occurring with the element nem ‘not’: (89) iob hog megfog’dosu¯a algukm˙eg’ vrat èlèuènèn hog better that prt-catch-part bless-sbjv-1pl-prt Lord-acc alive that n˙e m˙eghal’l’¯oc not prt-die-sbjv-1pl ‘it is better to bless the Lord if we are captured alive than to die’ (Vienna C. 25) In this function hogy was widespread and it was only in Middle Hungarian that it came to be replaced by mint ‘than/as’. These functions of hogy mentioned so far are attested in Old Hungarian; note that hogy introduces embedded yes–no questions in Modern Hungarian (appearing together with the interrogative marker -e) but this function evolved only later. These clauses were introduced by the C head ha ‘if ’ even in the 17th century, as will be shown in section 6.4.3.2. In Modern Hungarian, the complementizer is hogy, which alternates with zero in this function. Apart from functional changes, it has to be stressed that clauses introduced by hogy became more frequent. The following chart summarizes the findings of a small corpus study carried out on three translations of the gospel of Mark: the Munich Codex



Júlia Bácskai-Atkári and Éva Dékány

(1416/1466), György Káldi’s translation (1626), and the Neovulgata translation (1997). Altogether there are 219 loci where hogy occurs in at least one of the translations as a sole complementizer (hence not as part of complementizer combinations or together with relative pronouns). The occurrences of hogy are as in Table 6.6. Table .. The increased use of hogy ‘that’

hogy ‘that’

Munich Codex (1416/1466)

Káldi’s translation (1626)

Neovulgata (1997)

115

159

172

As can be seen, the number of the occurrences—and hence the frequency—of hogy increased from Old Hungarian onwards. It is important to mention that alongside with this, the number of the zero alternates of hogy also increased in the loci under discussion; that is, in cases where the earlier texts had different constructions, hence not finite subordination, later texts may contain a zero complementizer instead of hogy, and the number of these increased too, as shown in Table 6.7. Table .. The increased use of the zero subordinator

Ø ‘that’

Munich Codex (1416/1466)

Káldi’s translation (1626)

Neovulgata (1997)

11

18

20

It is worth having a closer look at the structures the Munich Codex and Káldi’s translation use where the Modern Hungarian translation has hogy (hence there are altogether 57 such instances in the Munich Codex and 13 in Káldi’s translation). Of course, there are a number of cases where the structure is too different to allow systematic comparison; disregarding these, however, there are some typical syntactic structures that appear instead of finite subordinate clauses introduced by hogy, in line with the general marginalization of non-finite subordination (see section 6.3), as seen in Table 6.8. As can be seen, the use of mert ‘that’ in that-clauses is significant in the Munich Codex; this was possible in Old Hungarian (Haader 2003: 506) but not later (when mert could only mean ‘because’); in other words, initially there were two possible candidates (hogy and mert) for the role of a general subordinator but it was clearly hogy that eventually won. Note that this is in line with the fact that hogy became a general marker of finite subordination and hence when there is no other Force to be expressed, then the head of the subclause is either hogy or its zero alternate. We will return to the issue of the functions of hogy when discussing complementizer combinations.

From non-finite to finite subordination



Table .. Structures used instead of finite subordination Munich Codex (1416/1466)

Káldi’s translation (1626)

non-finite clauses

adverbial participles infinitives

5 16

2 10

other

coordination (és ‘and’) mert ‘that’ nominal expressions

2 29 10

2 – 6

The chart above also shows that the constructions used instead of clauses introduced by hogy are in most cases not even instances of finite subordination: they are very often non-finite clauses such as infinitives and adverbial participles (see section 6.3 for more details); in addition, there are several instances of coordination and of nominal expressions (DPs containing nouns derived from verbs via the suffix -ás/-és). The frequency of these is lower already in Middle Hungarian, in accordance with the increased significance of finite subordination in general (see section 6.3 of the present chapter and also Haader 2001). It is worth noting that the complementizer introducing simple declarative subclauses tends to appear in a number of other constructions as well in languages in general. For instance, that in English has likewise several functions: (90) a. I heard that Ralph had arrived. b. Ralph was so tired that he fell asleep in class. c. I have seen the film that you mentioned last week. d. We took the train so that we would arrive on time. As can be seen, that in English can introduce simple declarative subclauses such as the one in (90a) and it also appears in the so. . .that construction, as in (90b). Furthermore, that in English is able to introduce relative clauses on its own, as in (90c), which is not the case in Hungarian—however, as will be shown later on, Hungarian hogy could also appear in relative clauses historically, if combined with relative pronouns. Last but not least, that appears in purpose clauses in the sequence so that. Apart from comparatives, that in English crucially does not appear in two constructions: embedded imperatives and embedded wh-interrogatives (or in embedded yes– no questions either). As far as the first is concerned, English uses infinitive constructions and there is simply no overt complementizer in embedded wh-interrogatives (and embedded yes–no questions are introduced by if or whether). This is demonstrated by the following examples:



Júlia Bácskai-Atkári and Éva Dékány

(91) a. I told him to clean the windows. b. I asked him when he wanted to leave. c. I asked him if he wanted to leave. This reveals an interesting property of Hungarian hogy, namely that it is truly a marker of finite subordination and was so historically as well. This has two main aspects. First, as was mentioned above, in Hungarian finite subordinate clauses introduced by hogy appear instead of non-finite structures, which is the result of finite subordination gaining over non-finite subordination. Naturally, such changes were possible only by using a complementizer that was not incompatible with the Force of the clause; that is, hogy was already general enough to accommodate even more functions. Second, unlike that in English, hogy is not specified for [−wh] but may appear in [+wh] clauses as well, which would be incompatible with the properties of a simple declarative complementizer that is inherently marked for [−wh]. ... The evolution of ha ‘if ’ The etymologically related operator of the complementizer ha ‘if ’ meant ‘when’ and since the functional split between the two took place mostly before the Old Hungarian period this latter function is relatively infrequent in Old Hungarian as well (though it is possible even later on): (92) fele mvnybe [le] ha tekunte [ek]essen tegud e[s] ha up heaven-ill when look-pst.3sg embellished you-acc too when lata ýste[n]segnec [ne]we mia rolad ozun keppe[n] see-pst.3sg deity-dat name-poss for you-del so scola speak-pst.3sg ‘when he looked up to heaven and saw you embellished, he spoke of you that way for the name of God’ (Königsberg Fragment) The complementizer ha was early grammaticalized into a lower C head and its reanalysis as a higher C head was early too: accordingly, it was base-generated as a higher C head already in Old Hungarian. This is clearly shown by its behavior in complementizer combinations, as will be shown later on. As far as its functions are concerned, ha has introduced conditional clauses from Old Hungarian onwards, as in the Old Hungarian examples in (93): (93) a. mert ha ýsten ev num uolna my benne býnut because if God he not be-cond.3sg we he-ine sin-acc lelhetneýnc find-possib-cond-1pl ‘because if he were not God, then we would be able to find sins in him’ (Königsberg Fragment)

From non-finite to finite subordination



b. ha te io´g kèzed meggonozbeitand tegedèt med if you right hand-poss.2sg prt-offend.3sg you-acc cut-imp-2sg èl otet ´ off it-acc ‘if thy right hand offend thee, cut it off ’ (Munich C. 11va) In addition, ha was responsible for introducing embedded yes–no questions in Old Hungarian: this function was preserved in Middle Hungarian as well but then the interrogative marker -e was also present in the subclause.20 In Modern Hungarian, this type is introduced by hogy or its zero alternate, and the presence of the interrogative marker -e is obligatory (see É. Kiss 2002: 239). The change is illustrated below in (94), with a zero complementizer in the Modern Hungarian example: ¯ (94) a. m¯ogadmg nèkonc ha te vag xc ´ tell-imp-2sg-prt dat-1pl if you are Christ ‘tell us whether thou be the Christ’ b. mondd meg nekünk, ha te vagy-e Krisztus tell-imp-2sg prt dat-1pl if you are-q Christ ‘tell us whether thou be the Christ’

(Munich C. 33va)

(Káldi, Mark 26:63)

c. mondd meg nekünk, Ø te vagy-e a Krisztus tell-imp-2sg prt dat-1pl you are-q the Christ ‘tell us whether thou be the Christ’ (Neovulgata, Mark 26:63) Consider also the following set of examples, with an overt hogy in the Modern Hungarian translation: (95) a. kèrde otèt ha mit latna ´ ask-pst.3sg he-acc if what-acc see-cond.3sg ‘he asked him if he saw ought’ b. kérdé őt, ha lát-e valamit ask-pst.3sg he-acc if see.3sg-q something-acc ‘he asked him if he saw ought’

(Munich C. 44ra)

(Káldi, Mark 8:23)

c. megkérdezte tőle, hogy lát-e valamit prt-ask-pst.3sg abl-3sg that see.3sg-q something-acc ‘he asked him if he saw ought’ (Neovulgata, Mark 8:23)

20 As described in Chapter  of this volume, -e was originally a C head and since it was left-branching (CPs being head-final in the SOV setting of Proto-Hungarian), it appeared at the right edge of interrogative clauses in the phonological structure. With the change from SOV to SVO and hence from head-final to head-initial, -e in Middle Hungarian is a left-branching functional head in the left periphery (identified as the head of an IntP by Bácskai-Atkári , in line with the IntP proposed by Rizzi ).



Júlia Bácskai-Atkári and Éva Dékány

As can be seen, ha in this function was initially responsible for marking the [+wh] nature of the subclause in itself; however, later on the phonologically visible marker came to be the interrogative marker -e, which is inherently [+wh] and has been appearing in main clause questions from Old Hungarian onwards (often together with nemde ‘isn’t it’ in Old Hungarian). In this way it became unnecessary to mark the [+wh] nature of the clause by a separate [+wh] complementizer and as far as marking subordination, the general subordination marker is hogy ‘that’ or its zero alternate (see section 6.4.3.1), and hogy is underspecified for [±wh], see É. Kiss (2002: 239). Hence the change between (94b) and (94c), and between (95b) and (95c) is essentially the general extension of hogy for marking subordination. It has to be mentioned that there seem to be three patterns cross-linguistically with respect to the [±wh] nature of the complementizer in embedded yes–no questions. First, the C head may select exclusively for [+wh] or [−wh]—for instance, German ob ‘if ’ introducing embedded yes–no questions selects exclusively for [+wh]: (96) Ich weiß nicht, ob er kommt. I know-1sg not if he come-3sg ‘I don’t know if he will come.’ Since German ob has no other function, it is always unambiguously [+wh] and hence there is single encoding in German embedded yes–no questions, in that the C head responsible for clause-typing also encodes the [+wh] nature of the clause. Second, it is also possible that a given C head selects either for [+wh] or [−wh] depending on its function. This is the case for English if and for Old Hungarian ha: when introducing embedded yes–no questions, as in (97a), if is [+wh] but when introducing conditional clauses it is [−wh], as in (97b): (97) a. I don’t know if he will come. b. Ring me if he comes. Since there seems to be a clear-cut distinction between the two functions, it is worth distinguishing between two complementizers that have the same phonological form and are also etymologically related. Still, in cases like (97a) there is also single encoding as it is the complementizer if that is responsible for marking subordination and the [+wh] nature of the subclause. Finally, it is also possible that the C head responsible for clause-typing marks only subordination and does not select for [+wh] or [−wh], and this is the case with hogy in Modern Hungarian. In this case, there is obviously double encoding: that is, the element responsible for marking subordination (the C head) is distinct from the element overtly marking the [+wh] of the subordinate clause (the interrogative marker -e). Note that the same double encoding holds in embedded wh-questions

From non-finite to finite subordination



from Old Hungarian onwards since the overt marker of [+wh] has always been the wh-element itself and the subordinator has been hogy (or its zero counterpart, which is in fact earlier, see Chapter 1 of this volume). In this respect, the diachronic change from Old to Modern Hungarian embedded yes–no questions is essentially one from single encoding into double encoding and the Middle Hungarian configuration (the co-occurrence of ha and the interrogative marker -e) represents an intermediate change. As ha was gradually losing its function of marking [+wh] and this role was taken over by -e, the role of an overt complementizer was reduced to solely marking subordination and hence hogy, which was the general subordination marker, took over this role from ha. ... The evolution of mint ‘than/as’ The complementizer mint ‘than/as’ is etymologically related to a former operator meaning ‘how’; the functional split between the two took place during Old Hungarian and partially also during Middle Hungarian, hence both functions can be observed for a long time. It is important to mention that mint could alternate with the operators miként ‘how’ and miképpen ‘how’ in comparatives expressing equality (see Haader 2003: 539); however, the latter did not develop into C heads. The operator mint can be observed already in the earliest texts: esmeríuc! (98) Ez oz ýsten myntevt this the God how-he-acc know-1pl ‘this is God as we know him’

(Königsberg Fragment)

We hypothesize that mint in Old Hungarian was either an operator in the specifier of the lower CP or a lower C head: it started moving up to the higher C head position in this period but was not grammaticalized there yet. The importance of this will become clear when considering combinations: as was seen earlier, the comparative complementizer was initially hogy and hence mint (and its alternates) could appear only in a lower—specifier, then head—position. This also means that the presentday complementizer function of mint evolved during the Old and Middle Hungarian periods due to the grammaticalization of mint as a higher C head and the disappearance of hogy from comparatives. We will return to this issue later on in more detail. The complementizer mint is responsible for introducing comparative subclauses both in comparatives expressing equality and in ones expressing inequality. The difference between the two types can hence primarily be observed in the degree expression in the matrix clause: in structures expressing equality, the adjective or the functional head of the degree expression is in the positive degree (e.g., olyan magas ‘as tall’ or annyi ‘as much’), while in comparatives expressing inequality it is in the comparative degree (e.g., magasabb ‘taller’ or több ‘more’). The following examples show mint in comparatives expressing equality. Note that in Old Hungarian miként (and miképpen) can still appear in this function:



Júlia Bácskai-Atkári and Éva Dékány

(99) M˙et istèn nem vg¯a fen˙egèt m¯et èmber sem because God not as threaten.3sg as human neither g´ erièztètic haragra mik¯et èmb˙ernc. fia induce-pass-3sg wrath-sub how human-dat son-poss ‘because God does not threaten as humans do, nor does he get enraged as humans’ (Vienna C. 27) In the following example mint appears in comparatives expressing inequality: (100) Es par¯ac˘ola hog a kèmencè hètzer inkab and command-pst.3sg that the furnace seven-times rather gerièztètnec m¯et zokotvala gerièztètni heat-caus-cond-3sg than use-perf-be-pst heat-pass-inf ‘and he commanded that they should heat the furnace one seven times more than it was wont to be heated’ (Vienna C. 127) In the case of mint there are no functional changes: essentially what happened is that an original operator grammaticalized into a C head, which is in line with the expectation that finite subordinators become more diversified as finite subordination becomes more important. Note that the grammaticalization of operators into C heads is in fact very frequent in comparatives cross-linguistically; what may seem to be peculiar in Hungarian is that this change took place (at least) twice, first with hogy and later with mint. On the other hand, it seems that the change affected comparatives expressing equality and ones expressing inequality at the same time. A similar grammaticalization process is argued for by Jäger (2012) for German als ‘than’ and wie ‘as’, such that wie grammaticalized later; she also points out that wie is permitted to co-occur with als in some dialects and in others it has in fact already taken over the role of als. The following examples illustrate these various possibilities: (101) a. Ralf ist so groß wie Michael. Ralph be.3sg as tall as Michael ‘Ralph is as tall as Michael.’ b. Ralf ist größer als Michael. Ralph be.3sg taller than Michael ‘Ralph is taller than Michael.’ c. Ralf ist größer als wie Michael. Ralph be.3sg taller than as Michael ‘Ralph is taller than Michael.’ d. Ralf ist größer wie Michael. Ralph be.3sg taller as Michael ‘Ralph is taller than Michael.’

From non-finite to finite subordination



The structures given in (101a) and (101b) represent the Standard German (Modern High German) setting, where the complementizer responsible for introducing clauses expressing equality is wie and the one responsible for introducing clauses expressing inequality is als. Depending on the dialect and on the speaker, (101c) is possible (this is common in Western dialects such as Hessian): this involves the co-presence of two complementizers, in the same fashion as hogy and mint could co-occur in Old and Middle Hungarian, as will be shown in section 6.4.4.2.1. However, configurations like (101c) in some dialects led to wie taking over als in comparatives expressing inequality (this is common in Southern dialects such as Bavarian): in these dialects, wie is hence a general comparative complementizer in essentially the same way as mint is in Hungarian, as shown in (101d). ... The evolution of mert ‘because’ The operator etymologically related to mert ‘because’ had the meaning of ‘why’; the functional split between the two took place during Old and partially also Middle Hungarian. Hence the split between the two forms mert ‘because’ and miért ‘why’ that is true for Modern Hungarian was not attested for a long time and the two forms were fundamentally free variants (see Haader 2003: 542–543); that is, the form mert could have both the functions ‘because’ and ‘why’, and the same is true of the form miért. We hypothesize that mert in Old Hungarian was either an operator in the specifier of the lower CP or a lower C head: it started moving up to the higher C head only in this period and was not grammaticalized there yet. Again, the importance of this will become clear when considering combinations, which will be addressed later. The following example shows the complementizer mert appearing in the form of mert: (102) Halgassad vr¯a & irgalmazy mert irgalmas istèn listen-imp-2sg lord-poss.1sg and pity-imp-2sg because merciful God vag & irgalmazih m˙vnèkonc mert b˙unhot˙tonc te ´ ´ ´ be-2sg and pity-imp-2sg we-dat-1pl because sin-pst-1pl you èlottèd ´ before-2sg ‘Listen and have pity, Lord, for we have sinned before you.’ (Vienna C. 102) In the following example the complementizer mert appears in the form of miért: istèneknèc istènè & kiraloknac (103) Bizon˙ a t˙v istèntec ´ indeed the you God-poss.2pl god-pl-dat God-poss and king-pl-dat vra megièlentuen titkokat miért lord-poss prt-reveal-part secret-pl-acc because meg˙nithata è titkot prt-open-possib-pst.3sg this secret-acc ‘of a truth it is, that your God is a God of gods, and a Lord of kings, and a revealer of secrets, seeing thou couldest reveal this secret’ (Vienna C. 124)



Júlia Bácskai-Atkári and Éva Dékány

In Old Hungarian, mert could also introduce that-clauses, as was mentioned in section 6.4.3.1. This function disappeared before the Middle Hungarian period (cf. Haader 2003: 506) but consider the following example from Old Hungarian: miat o scent zèllete (104) Kit legottan ic ´ who-acc immediately Jesus he sacred spirit-poss.3sg for megèsmèru¯e mert ig gondolnanac bènnèc m¯oda o¯ ´ prt-recognize-part that so think-cond-3pl they ine-3pl say-pst.3sg nèkic o ´ they dat-3pl ‘and immediately when Jesus perceived in his spirit that they so reasoned within themselves, he said unto them’ (Munich C. 37rb) Apart from the disappearance of this function, there are no considerable changes in the use of mert in Old and Modern Hungarian, and hence it can be concluded that there are no significant functional changes in the case of mert either. It has to be mentioned that the grammaticalization of mert into a C head involves an important change in the [±wh] nature of this element since mert as a complementizer is [−wh] while as an operator—either interrogative or relative—it was [+wh]. This difference can be observed in other languages as well: in Italian, for instance, perché ‘because’ is a complementizer and perché ‘why’ is an interrogative (though not relative) operator. The two functions are illustrated by the following examples: (105) a. Ti ho chiesto una mano perchè stavo you.dat have-1sg ask-part a-fem hand because be-pst-1sg cadendo. fall-part ‘I asked you to give me a hand because I was falling.’ b. Ho chiesto perchè questa canzone piace alla have-1sg ask-part why this-fem song please-3sg to-the-fem gente. people ‘I asked why people liked this song.’ c. Ho chiesto perchè questa canzone piaccia this-fem song please-sbjv-3sg have-1sg ask-part why alla gente. to-the-fem people ‘I asked why people liked this song.’

From non-finite to finite subordination



In (105a), the C head perché introduces a [−wh] subclause and the verb is in its indicative form. By contrast, in (105b) and (105c) the subordinate clause is [+wh] and contains the operator perché and the verb is either indicative, as in (105b), or is in the subjunctive, as in (105c), the latter representing a more formal/elevated style. Note, however, that embedding in itself does not require the use of the subjunctive and the reason behind its availability in (105c) is due to the [+wh] nature of the clause. What is important for us here is that the clear feature distinction between a [−wh] complementizer mert and a former [+wh] operator mert means that as soon as mert is reanalyzed as a C head and hence [−wh], the [±wh] nature of the clause is set by the overt complementizer itself. Hence, in a clause introduced by mert it is not possible for other, new operators to appear even after mert has been reanalyzed as a higher C head and grammaticalization at the left periphery of clauses of reason is thus not recursive in this sense. This is different from what was attested in comparatives, where the reanalysis of former operators into C heads actually feeds the appearance of new overt operators in the long run: the same process taking place in clauses of reason bleeds the appearance of new operators. ... Interim summary The general change from non-finite to finite embedding brought about the evolution of a functional left periphery (a CP-domain) in finite subordinate clauses; apart from the strengthening of a CP-domain, this also involved the establishment of CPs as head-initial projections. Grammaticalization processes in this CP-domain involved two main aspects. First, elements initially moving to the left periphery grammaticalized from operators into C heads, and this resulted in a variety of finite complementizers expressing various functions. Second, marking finite subordination became more important, and hence hogy ‘that’ was extended to a wide range of clauses as a general subordination marker. .. The appearance and disappearance of multiple complementizers As was mentioned in the introduction, complementizers could also appear in various combinations. This involves the combinations of the four complementizers— hogy ‘that’, ha ‘if ’, mint ‘than/as’, and mert ‘because’—with each other, and combinations that involve other heads in the left periphery. First, the general mechanisms of complementizer combinations will be considered, with special attention paid to the distinction between syntactic and morphological combinations and the importance thereof. Second, we will turn to the examination of Hungarian complementizers combining with each other and with negative-like heads, with special focus on comparatives. Third, we will briefly consider the issue of multiple combinations.



Júlia Bácskai-Atkári and Éva Dékány

... Syntactic and morphological combinations As was outlined in the introduction, there are several combinations attested in various periods of Hungarian: some of these are already extinct (e.g., hahogy ‘if that’ or hogyhamint ‘that if as’), while others are still used (e.g., hogyha ‘that if ’ or minthogyha ‘as that if ’). An important question concerning complementizer combinations is, in case the two (or more) elements that are involved in the combination can function as complementizers on their own as well, what the grammatical status of the combination is, that is, whether the combination is formed during the syntactic derivation or whether the combination enters the derivation already as a complex unit. The two types to be distinguished here are syntactic and morphological combinations. In syntactic combinations the parts of the combination are base-generated as separate heads in the syntax and combination hence either means the adjacency of these separate elements at PF, or there are complex heads that are formed by adjunction during the derivation. In morphological combinations the entire complex is base-generated as a single head in the syntax, and hence the notion of combination can be applied only as far as morphology is concerned. The two types are nevertheless strongly related to each other historically: as will be shown, morphological combinations came into being by the grammaticalization of syntactic combinations. As was discussed in section 6.4.3, the individual complementizers underwent grammaticalization at different times, and hence their typical positions were also different in Old Hungarian. This also enabled both C head positions to be filled by overt elements, which can also be observed in other languages, cf. Roberts (2005) and van Gelderen (2005). In addition, it was also possible for future complementizers that were still operators to appear together with an overt higher C head. In these cases the combinations are purely syntactic and the linear PF order is the same as the basegenerated order. At the same time, as has already been mentioned, lower C heads were ultimately reanalyzed as higher C heads. This obviously meant that complementizers basegenerated in the lower C position started to move up to the higher C head, and later came to be base-generated there. When a lower C head moved up to the higher C position when the latter was already filled by another (overt) element, then the original lower C head was left-adjoined to the original higher C head, following the Linear Correspondence Axiom of Kayne (1994) and the Mirror Principle of Baker (1985, 1988). Later on, these combinations were grammaticalized, that is, they came to be base-generated as a single—morphologically complex—head in the higher C position.

From non-finite to finite subordination



The four stages described above are represented in (106): (106)

CP

CP

C’

C’

C X

CP Y

C X

C’ C

CP C’

…

C Y

Ø

C

CP

CP

C’

C’ CP

YX

C

Y

CP

YX

C’ C

…

…

C’ C

…

Ø

As can be seen, grammaticalization resulting in morphological combinations leads to the inverse order of the original one but since this is an instance of predictable syntactic derivation, the variants XY and YX (at PF) are free alternates, at least as far as their original functions are concerned. This will be shown to be the case for Hungarian complex complementizers as well. ... Combinations of two C heads The system of combinations involving two C heads was outlined in Table 6.4 and it was also pointed out in the introduction that the system is completely symmetrical in that if a given combination existed in the order XY, then the order YX is also attested; this follows from the nature of the mechanisms given in (106). Moreover, for every pair XY and YX it is true that only one member remained in the language—that is, the one that grammaticalized into a morphological combination. On the other hand, the underlying order is directly influenced by when the individual complementizers grammaticalized. As was pointed out in section 6.4.3, grammaticalization did not take place at the same time for all the four complementizers, and hence they occupied different positions in Old Hungarian and partly in Middle Hungarian as well. These typical positions are given below:



Júlia Bácskai-Atkári and Éva Dékány CP

(107)

C’ C

CP

ha mint hogy mert

C’ C

…

mint mert (hogy)

As can be seen, ha ‘if ’ is base-generated as a higher C head, while hogy ‘that’ is typically a higher and less frequently a lower C head; mint ‘than/as’ and mert ‘because’ are either lower C heads or still operators moving to the specifier position of the lower CP. This has three important consequences. First, the order of two complementizers appearing in one left periphery is predictable. Since ha is always a higher C head, it always appears as the first member in the original combinations, hence: hahogy ‘if that’ and hamint ‘if as’. Similarly, since mint and mert were lower C heads (or operators), they appear as the second members in the original combinations, hence: hogymert ‘that because’, hogymint ‘that than’ and hamint ‘if as’. Finally, hogy, typically being a higher C head, could combine with lower C heads (and operators), hence: hogymert ‘that because’ and hogymint ‘that than’, but since it could appear as a lower C head as well, it could also be combined with ha, hence: hahogy ‘if that’. Second, since lower C heads systematically moved up to the higher C position, it is also explained why all the four combinations reflecting the underlying order (hahogy, hamint, hogymert, hogymint) have their counterparts with the reverse order but the same original meaning (hogyha, mintha, merthogy, minthogy), as shown in Table 6.4. Third, apart from the fact that an original XY combination also had a YX counterpart, it is also predictable that out of the two it is always the one showing the YX order that remained in the language: since all lower C heads ultimately grammaticalized into higher C heads, there remained no complementizer to appear in the lower C position. This is in accordance with the historical data, as highlighted in Table 6.4. There is yet one more question to be addressed in connection with the status of hogy in the combinations under discussion. As was mentioned, hogy was typically a higher C head and it was base-generated as a lower C head only in the case of hahogy (more arguments for this will be presented in section 6.4.4.2.3). However, since hogy preferably moved up in general, it is expected that movement preferably took place

From non-finite to finite subordination



Table .. The appearance of complex C heads

hogyha mintha minthogy merthogy

Munich Codex (1416/1466)

Jordánszky Codex (1516–1519)

Káldi’s translation (1626)

Neovulgata translation (1997)

9 – – –

8 1 – 1

9 3 4 –

– 7 1 –

also in combinations with ha: that is, the appearance of the combination hogyha is expected to be significantly earlier than that of the other grammaticalized complex C heads. This is indeed the case, as demonstrated by the comparative study carried out on four different translations of all the four gospels (cf. Bácskai-Atkári 2012a); the translations are the Munich Codex (1416/1466) and the Jordánszky Codex (1516–1519) from the Old Hungarian period, György Káldi’s translation (1626) from the Middle Hungarian period, and the Neovulgata translation (1997) from the Modern Hungarian period. We searched for the occurrences of the complex C heads hogyha, mintha, minthogy, and merthogy; the results are summarized in Table 6.9. As can be seen, hogyha appears considerably earlier than the other three combinations, which is in line with the expectations: the Munich Codex contains examples only for this combination out of all the four, while the other three combinations appear later with some sporadic examples in the Jordánszky Codex and in greater numbers in later texts. It must be mentioned that all the four combinations exist in Modern Hungarian and hence it is accidental that some of them do not appear in the Neovulgata translation at all. On the other hand, not only hogyha but also the other three combinations date back to Old Hungarian and hence their absence or low numbers can be interpreted only in terms of the comparative analysis carried out on the given texts. What is important for the present discussion is that the early and frequent appearance of hogyha is not surprising considering that hogy preferably moved up even when combining with the complementizer ha since it was preferably located in that position anyway—from this it follows that hogyha should appear considerably earlier than the other three combinations. In what follows we will briefly review the individual combination pairs and their functions. .... The combinations of hogy ‘that’ and mint ‘than/as’ As has already been discussed, comparative subclauses were originally headed by the complementizer hogy ‘that’; mint ‘than/as’ appeared in Old Hungarian, first as an operator and later as a lower



Júlia Bácskai-Atkári and Éva Dékány

C head, resulting in the combination hogymint ‘that than’. It has to be mentioned that originally nem ‘not’ or sem ‘neither’ also appeared alongside hogy; this issue will be readdressed later and hence for the time being let us disregard the question of how negative elements appeared in the combinations of hogy and mint. The following examples illustrate the function of hogymint introducing comparative subclauses expressing inequality, as in (108a), and in ones expressing equality, as in (108b): (108) a. edesseget erze nagyoban hogymint annak elotte ´ sweetness-acc feel-pst.3sg greater that-than that-dat before-3sg ‘(s)he felt sweetness even more than before’ (Lázár C. 71r) b. mínd anne bosegos końhullatasoc mene a vízeknec ´ ´ ´ all so.much plenty crying-pl as.much the water-pl-dat sokassaghí sem volnanac en elottem kellemetosek ´ ´ multitude-poss.pl neither be-cond.3pl I before-1sg pleasant-pl Aua´g foganatosoc ho´g mint akki zonetlen a kereztfanac o ´ ´ or effective-pl that as who incessantly the rood-dat he víselí keserúseget v´ testeben bitterness-poss.3sg-acc he body-poss.3sg-ine bear-3sg ‘not even as much crying as the multitude of waters would be as pleasant and touching to me as the one who incessantly bears the bitterness of the rood in his body’ (Nagyszombat C. 40–41) By way of mint moving up to the higher C head the complex minthogy ‘than that’ was formed, which was originally also a comparative complementizer (in clauses expressing inequality): (109) my lehet ezneel chodalatosb allat mynt hog ember what be-possib.3sg this-ade more.wonderful state than that human lenne istèn be-cond.3sg God ‘what can be a more wonderful state than that God be a man’ (Horvát C. 1v) It should be mentioned that in the case of minthogy there arose an explanatory meaning as well (Haader 1995: 619), and it is this function that continues to exist in Modern Hungarian. Since the question of how minthogy developed this function would lead to questions concerning the relation between subordination and coordination and the boundary cases in between the two (cf. Kenesei 1992: 537–552), we will not venture to examine this question here in more detail. Turning back to comparatives, what is important is that the co-occurrence of hogy and mint—either with or without a negative element—made it possible for the original comparative complementizer (hogy) to be gradually replaced be a new one (mint), as will be shown later on in more detail.

From non-finite to finite subordination



Combinations of a general(/declarative) and a comparative complementizer are attested in other languages as well; for instance, German has the combinations als dass ‘than that’: (110) Es war zu schrecklich, als dass man es mit Worten it be.pst.3sg too awful than that pronoun it with words beschreiben könnte. describe-inf can.cond-3sg ‘It was too awful, more than one could describe.’ In cases like (110), the combination als dass serves to introduce an unreal comparison. The combinations that emerged in Old and Middle Hungarian are different in the sense that they involved two comparative complementizers—this is also attested in German in dialects that have als wie ‘than as’, see section 6.4.3.3. .... The combinations of hogy ‘that’ and mert ‘because’ Clauses of reason were introduced by the complementizer mert ‘because’ but hogy ‘that’ could also appear in these constructions, resulting in the combination hogymert ‘that because’. It is important to mention that, unlike in comparative subclauses (hogymint ‘that than’ and minthogy ‘than that’), hogy in clauses of reason appeared later and did not modify the meaning of the construction; exceptionally, though, hogy could appear on its own in clauses expressing reason, as in (110b) above. This is in line with the fact that hogy in Old and Middle Hungarian came to be a general marker of subordination and the appearance of hogy together with mert was also motivated because mert was still not a grammaticalized higher C head, which is ultimately the position responsible for marking subordination. Since the functional split between mert and miért ‘why’ was not completed in Old Hungarian, mert in these combinations can naturally occur both in the form mert and miért. The following examples illustrate the function of hogymert as a head of clauses of reason: in (111a) mert appears in the form mert, while in (111b) it is in the form miért. ewtett (111) a. Dehogy mert zent ferenc ygen zeretiuala but-that because saint Francis well like-3sg-be-pst him-acc tyztasagert es alazatossagaert kyt valuala purity-final and humility-poss.3sg-final who-acc have.3sg-be-pst Monda neky say-pst.3sg him-dat ‘but because Saint Francis liked him well for his purity and for his humility that he had, he said to him’ (Jókai C. 46) b. De hogy meyerth dichewlth testbe wagyok en Nen but that because redeem-part body-ill be-1sg I not syrhatok cry-possib-1sg ‘but because I am in a redeemed body, I cannot cry’ (Apor C. 158)



Júlia Bácskai-Atkári and Éva Dékány

The movement of mert to the higher C head position resulted in the complex head merthogy ‘because that’, which likewise introduces clauses of reason: (112) De azonkezbe az baratok bel yewuenek az aztalra: De but meanwhile the brother-pl in come-pst-3pl the table-sub but mer hogy bodog ferencz zerzetteuala hogy ne because that blessed Francis command-perf-3sg-be-pst that not varnak wait-sbjv-3pl ‘but meanwhile the brothers had sat down to the table because blessed Francis had ordered that they should not wait for him’ (Jókai C. 84) As can be expected, merthogy survives into Modern Hungarian as a grammaticalized C head, while hogymert disappeared. Interestingly, a similar complementizer combination is also attested in Middle English in the form of for that (van Gelderen 2005): (113) Thy wyf and thou moote hange fer atwynne, / For that bitwixe yow shal be no synne. ‘Your wife and you must hang apart, that in the night shall come no chance for you to sin.’ (Chaucer, The Canterbury Tales: Miller’s Tale) Such combinations were possible when that was still located in the lower C head but not later, i.e., when that is already a higher C head (van Gelderen 2005). On the other hand, in English the inverse order of the original C + C combination is not attested: lower C heads moving up to the higher C position did not engage in head adjunction, unlike Hungarian. .... The combinations of hogy ‘that’ and ha ‘if ’ As was seen before, conditional clauses were introduced by ha ‘if ’; however, hogy ‘that’ could appear even in these constructions—when it did, it appeared in the lower C head position, the higher one already being filled by ha, hence resulting in the combination hahogy ‘if that’. Just as with clauses of reason, the function of hogy was to mark finite subordination, and hence it preferably moved up to the higher C position, as was mentioned previously. In addition, since hogy in Old Hungarian could still function as a comparative complementizer, its combination with ha could also serve to introduce conditional comparative clauses. That the underlying (C + C) order is represented by hahogy is also demonstrated by the fact that constituents could potentially move to a position between ha and hogy at the left periphery:

From non-finite to finite subordination (114) Ha késen hogy el nyugot az nap, hamar esőt if late that off set-pst.3sg the sun soon rain-acc váry expect-imp-2sg ‘if the sun has set late, expect rain soon’



(Cisio)

As can be seen, the complementizers ha and hogy are located within a single left periphery but the adverbial késen ‘late’ can appear between the two. Note that if ha and hogy in (114) were located in two different left peripheries, then the first clause (ha késen ‘if late’) should obligatorily contain an overt copula marked for past tense. Since this is not the case, the string ha késen cannot be considered a separate clause, and hence ha and hogy are located in one and the same left periphery. This kind of construction is rare because topics and foci in Hungarian normally move below the C-domain and hence not between the two C heads (see section 6.4.1). What is important for us here is that (114) is possible only if ha and hogy are distinct C heads, which in turn means that hahogy represents an underlying order. Apart from the example in (114), the following sentence also represents hahogy introducing conditional clauses: (115) Az én jó istenem, ha hogy sok ellenség, reám the I good God-poss.1sg if that many enemy sub-1sg fegyverkezék, tölök megmente arm-pst-3sg abl-3pl prt-save-pst.3sg ‘my good God, if many enemies armed against me, saved me from them’ (Balassi: Ének 32) As was argued for earlier, hogy in these constructions preferably moved up, which resulted in the complex hogyha that is still used in Modern Hungarian too: (116) a. Es az lattatíc ennekom ho´g ha az paradíčomnac ´ and that see-pass-3sg I-dat-1sg that if the Paradise-dat g´ eńerúseges edes lakodalmaban lakoznam beautiful sweet dwelling-poss-ine dwell-cond-1sg ‘and it was shown to be as if I had been living in the beautiful and sweet Paradise’ (Nagyszombat C. 118) b. gondolya vala ewnenb¯ene ezt hogiha ew hozya think-3sg be-pst himself-ine this-acc that-if he all-3sg menne zerzetes rwhaba hogi el futhna ew go-cond.3sg monk garment-ill that off run-cond.3sg he elewle before-3sg ‘he thought that if he went up to him dressed as a monk, then he would run away’ (Példák könyve 15–16)



Júlia Bácskai-Atkári and Éva Dékány

In example (116a) hogyha is used in a conditional comparative, while (116b) shows hogyha introducing an ordinary conditional clause. It has to be mentioned that in Modern Hungarian hogyha is used only in conditional clauses, but not in conditional comparatives, which is in parallel with hogy having lost its comparative function. Again, it is worth mentioning that a similar combination existed in Middle English as well in the form of that if (van Gelderen 2005) and just as in the case of Hungarian, the role of that is purely marking subordination but it does not change the meaning defined by if : (117) Blameth nat me if that ye chese amys. ‘And blame not me if you do choose amiss.’ (Chaucer, The Canterbury Tales: Miller’s Prologue) Just as in the case of for that, the reverse order is not attested and hence with the grammaticalization of that in the higher C head the combination if that disappeared from the English language. .... The combinations of ha ‘if ’ and mint ‘as’ Conditional comparatives represent a mixed type between two basic clause types and hence they are typically represented by both complementizers that otherwise introduce these two types. In Hungarian, conditional Force has always been represented by ha ‘if ’ and comparative Force was first associated with hogy ‘that’ and later with mint ‘as’. The combination hamint ‘if as’ is the result of mint starting to co-occur with the higher C head ha. The combination is illustrated by the following example: (118) de ha mynt čak el aluttak volna lelkoketh ´ but if as only off sleep-perf-3pl be-cond soul-poss.3pl-acc istennek meg adaak God-dat prt give-pst-3pl ‘but as if they had only fallen asleep, they gave their souls to God’ (Sándor C. 14v) The combination mintha ‘as if ’ is the result of mint moving up to the higher C head: (119) lelek zent x¯p¯ınat mint ha az ferodobol ´ ´ ´ find-pst-3pl saint Christine-acc as if the bath-ela ione ky ´ come-cond.3sg out ‘and they found Saint Christine as if she had come out from the bath’ (Christina Legend 19v) As far as the function of mintha is concerned, there are no changes to be considered as it introduces conditional comparatives even in Modern Hungarian.

From non-finite to finite subordination



The mixed nature of conditional comparatives involves the formation of similar combinations in other languages too, as shown by the following examples from English and German: (120) a. She acts as if she were a princess. b. Er gibt Geld aus, als ob er Millionär wäre. he give-3sg money out than if he millionaire be.cond-3sg ‘He spends money as if he were a millionaire.’ In such combinations the complementizer expressing comparison is normally one that is otherwise used in comparatives expressing equality, hence it is as and not than in English. Interestingly, in German it is als ‘than’ and not wie ‘as’: the reason behind this is that the combination was born in a period when the complementizer used in equatives was still als (see previous discussion in this chapter and also Jäger 2012). This means that standard combinations are not necessarily affected in the same way as single complementizers are. ... Negative-like elements in comparatives Let us now turn to the role and structural properties of negative-like elements in comparatives expressing inequality. As will be shown, these syntactic heads could participate in left peripheral combinations in fundamentally the same way as C heads did and hence the resulting multiple combinations can be described similarly. First the status of nem ‘not’ and sem ‘neither’ will be considered, also discussing the differences between them, and then we will proceed to show how they appeared in comparatives in combinations such as hogynem ‘that not’ and hogysem ‘that neither’. Then an examination of combinations containing mint ‘than’ will follow, hence the combinations hogynemmint ‘that not than’ and hogysemmint ‘that neither than’ and finally we will briefly discuss the conditions licensing the combination mintsemhogy ‘than neither that’ and the predictability thereof. .... The elements nem ‘not’ and sem ‘neither’ Originally, comparative subclauses were introduced by the complementizer hogy ‘that’, and this was accompanied by a negative element—typically nem ‘not’ and less frequently sem ‘neither’—in comparatives expressing inequality. Consider the following example: (121) Mert io ho megfog’dosu¯a algukm˙eg’ vrat because better that prt-catch-part bless-sbjv-1pl-prt Lord-acc èlèuènèn hog n˙e m˙eghal’l’¯oc alive that not prt-die-sbjv-1pl ‘because it is better to bless the Lord if we are captured alive than to die’ (Vienna C. 25) As can be seen, the complementizer hogy is followed by the negative element nem but the structure does not actually express negation and hence nem cannot be the syntactic



Júlia Bácskai-Atkári and Éva Dékány

head of a true NegP. Negative-like elements of this type are in fact related to polarity: comparative subclauses have negative polarity and hence some languages may require an overt negative polarity head (such as Old Hungarian), while in other languages this is optional (e.g., in Italian, see Salvi and Vanelli 2004: 283–285). In addition, there are languages where negative polarity is shown by the fact that negative polarity items are licensed in the comparative subclause (e.g., English, cf. Seuren 1973: 532–537; on cross-linguistic differences see also Bácskai-Atkári 2011): (122) He prefers to rant about a problem rather than lift a finger to fix it. Negative polarity items, such as lift a finger in (122), can appear only in clauses that have negative polarity, and they are perfectly acceptable in comparative subclauses. This is important for us here because nem and sem are hence not Neg heads in syntax but they head a PolP responsible for the polarity of the subclause (see Homer 2011). This projection appears between the two CPs, as given in the diagram below: CP

(123)

C’ C

PolP

hogy

Pol’ Pol

CP

nem/sem

C’ C

…

Ø

Though the syntactic position of nem and sem is the same, it has to be mentioned that there is an important difference between the two: while nem (as a Pol head) is a clitic, sem is not.21 This will be important when it comes to the discussion of combinations. .... The combinations hogynem ‘than not’ and hogysem ‘than neither’ Since nem ‘not’ and sem ‘neither’ in comparatives expressing inequality appeared together with hogy ‘that’, the combinations hogynem ‘that not’ and hogysem ‘that neither’ naturally follow. The combination hogynem is illustrated in (121) above and in (124) below:

21 Note that in this respect, nem and sem as Pol head behave exactly in the opposite way as nem and sem as Neg heads, since in the latter case sem is the element that behaves like a clitic (for more details, see Chapter , section . of this volume).

From non-finite to finite subordination



(124) iob hog èlèuènèn zolgallonc Nabuhodonozor nag kiralnac better that alive serve-sbjv-1pl Nebuchadnezzar great king-dat & alazkoggonc te nèk˙ed hog n¯e meghalu¯ac m˙v and cringe-sbjv-1pl you dat-2sg that not prt-die-part-1pl we vèzèdelm¯ocbèn m˙vnmagonc m˙v zolgalat¯oknac ´ peril-poss.1pl-ine ourselves we service-poss.1pl-dat karat zènuègg˙uc damage-poss-acc suffer-sbjv.1pl ‘it is better for us to serve the great king Nebuchadnezzar alive and to cringe before you than to suffer the damages of our service dying’ (Vienna C. 14) Hence in these cases a higher C head (hogy) co-occurs with a Pol head (nem or sem). As has been mentioned, nem in these constructions is a clitic, unlike sem. This has two consequences. First, nem cliticizes onto the preceding element hogy, which results in the form honnem showing phonological assimilation. This change can be observed between the Munich Codex and the Jordánszky Codex: the loci containing hogynem in the former text show honnem in the latter (possibly combined with mint ‘than’). This is illustrated by the following pair of examples: te (125) a. iob tenèked hog eg èluèzien better you-dat-2sg that one off-perish-sbjv-3sg you tagid kozzol hog n¯e mend te tèsted ´ ´ member-poss.pl-2sg among that not all you body-poss.2sg èrèztèssec pokolba cast-sbjv-3sg hell-ill ‘it is profitable for thee that one of thy members should perish, and not that thy whole body should be cast into hell.’ (Munich C. 11rb–11va) b. yncab yllyk teneked ho˝g el vezyen egik rather fit-3sg you-dat-2sg that off perish-sbjv-3sg one tagod, honnem te tellyes tested vettesseg member-poss.2sg that.not you entire body-poss.2sg cast-sbjv-3sg pokorra hell-sub ‘it is profitable for thee that one of thy members should perish, and not that thy whole body should be cast into hell.’ (Jordánszky C. 367) There is no such assimilation to be observed in the case of sem, as it is not a clitic. The second crucial difference is that sem could undergo head movement to the higher C head filled by hogy, hence resulting in the inverse order pair of hogysem: semhogy ‘neither than’ continues to exist in Modern Hungarian too. Since nem as a clitic was



Júlia Bácskai-Atkári and Éva Dékány

attached to the preceding element, it did not move up to the higher C head and hence hogynem has no inverse order counterpart.22 .... The combinations hogynemmint ‘that not than’ and hogysemmint ‘that neither than’ As was seen before, mint ‘than’ started to appear in comparative subclauses, first as an operator and later as a lower C head. Since hogy ‘that’—and in comparatives expressing inequality, the combinations hogynem ‘that not’ and hogysem ‘that neither’—were still present in the structure, the combinations hogynemmint ‘that not than’ and hogysemmint ‘that neither than’ arose. Consider the following examples for hogynemmint: (126) a. ha naual·as lelek keuelb lezen kazdaksagot meg vtaluã: if wretched soul prouder be.mod-3sg richness-acc prt hate-part ho˙g nem m˜ıt volt 8tet bíruan that not than be.pst.3sg it-acc possess-part ‘if the wretched soul becomes prouder when despising richness than it was when possessing it’ (Birk C. 1a) b. mert mastan kózelben vagyon a” my Idwesseegwnk ˝ because now nearer-sup be.3sg the we salvation-poss.1pl honnem mynt eleeb hyttók that.not than before think-pst-1pl ‘because now our salvation is nearer than we thought before’ (Érdy C. 3b) As can be seen, hogynemmint appears without phonological assimilation in (126a), while (126b) represents the form honnemmint. Finally, (127) shows hogysemmint: (127) thewbzer wacharalth az warban az kyralne azz[ony] more.times supper-pst.3sg the castle-ine the queen lady leanywal, Borbara azzannal, hogh sem mynth warasban daughter-poss-com Barbara lady-com that neither than town-ine az wrak kezewth the lord-pl among ‘he had supper in the castle with lady Barbara, the queen’s daughter more often than in town with the gentlemen’ (Hegedűs and Papp #139) Both hogynemmint and hogysemmint can be attributed the following structure (considering the case when mint is already a lower C head but it could initially be an operator as well):

22 Note that there is in fact a word nemhogy ‘instead of; not the least; not just’ in Modern Hungarian but this has never had a comparative function and is hence completely unrelated to hogynem.

From non-finite to finite subordination



CP

(128)

C’ C

PolP

hogy

Pol’ Pol

CP

nem/sem

C’ C

…

mint

In this case there are hence three overt heads in the left periphery: two C heads and a Pol head in between. It is worth mentioning that the gradual disappearance of hogy from comparative constructions could lead to the rise of semmint ‘neither than’: this is rare but it nevertheless fits into the system of complementizer combinations. Obviously, nem in this case again does not parallel with sem, since as a clitic it could not have appeared in a structure without a preceding element that it could cliticize onto. Returning now to the appearance of mint, it was mentioned in connection with hogymint ‘that than’ (and minthogy ‘than that’) that combinations containing both the previous and the later complementizer contributed to the loss of hogy and the takeover by mint in comparatives: this would not have been possible if mint had not been able to appear in such subclauses at all. The changes affecting the complementizers and complementizer combinations in comparatives expressing inequality can be summarized as follows: (129) hogynem (→ honnem), hogysem

(semhogy)

hogynemmint, hogysemmint, semmint

(mintsemhogy, mintsem)

hogymint

(minthogy)

mint

The two main lines of change are hence related to the appearance of mint and the disappearance of the negative element. Naturally, the individual stages cannot be sharply distinguished and hence the various forms are expected to co-occur in texts for considerable time both in Old Hungarian and in (early) Middle Hungarian.



Júlia Bácskai-Atkári and Éva Dékány Table .. Elements introducing comparative subclauses

hogynem hogynemmint mint

Munich Codex (1416/1466)

Jordánszky Codex (1516–1519)

Káldi‘s translation (1626)

Neovulgata translation (1997)

34 – –

20 11 4

– – 23

– – 20

However, the change can be clearly observed in the comparative study carried out on four translations of the gospels (Bácskai-Atkári 2012b): here the number of comparative structures is approximately the same in all the texts. It has to be mentioned that differences do occur especially because the standard value of comparison (that is, to which something is compared) can be expressed not only by a subordinate clause. As far as subclauses are concerned, the numbers that were found are shown in Table 6.10. The data clearly show that while comparative subclauses expressing inequality were introduced by hogynem in the Munich Codex, the picture is more diversified already in the Jordánszky Codex: the number of the occurrences of hogynem is significantly lower, resulting in a high frequency of hogynemmint and the possibility of mint. By contrast, both Káldi’s translation and the Neovulgata translation contain only mint. It has to be mentioned that hogynemmint (and hogysemmint) was nevertheless still possible in Middle Hungarian. Furthermore, neither hogynem nor hogynemmint can be considered as a Latin reflex: in all the instances above the Latin text contains quam ‘than’; hence, when considering changes in Hungarian comparative subclauses, one is essentially examining language-internal processes. An important conclusion to be drawn here is that apart from C + C combinations, there arose C + Pol + C combinations in an analogous way. .... The combination mintsemhogy ‘than not that’ As can be expected based on the analysis presented in section 6.4.4.3.3, the inverse order pair of hogynemmint ‘that not than’ did not arise, nem ‘not’ being a clitic—however, the same does not hold for hogysemmint ‘that neither than’. This is indeed so and the derivation of mintsemhogy ‘than neither that’ can be described with the same mechanisms that we saw in connection with grammaticalized, morphologically complex complementizers. Note that mintsemhogy also continues to exist in Modern Hungarian.

From non-finite to finite subordination



The head movements resulting in mintsemhogy are given below: CP

(130)

C’ C

PolP

mintsemhogy

Pol’ Pol

CP

mintsem

C’

C

…

mint

Hence the grammaticalization of C + Pol + C heads does not differ from that of C + C combinations: the complementizer mint ‘than’ moves up to the Pol head and is left-adjoined to the head sem ‘neither’, resulting in mintsem ‘than neither’; as a second step, this complex head moves up to the higher C head and is left adjoined to hogy ‘that’ there, resulting in mintsemhogy. Naturally, mintsemhogy also grammaticalized as a higher C head. Note that if the initial structure contained only sem and mint overtly, then this resulted in the complementizer mintsem ‘than neither’. From all this it follows that the properties of the syntactic and morphological combinations of two C heads are also valid in the case of two C heads (one of which may be covert) combining with a Pol head: the way heads can move up and combinations may grammaticalize is predictable, as is the fact that only combinations representing a grammaticalized higher C survive. ... Multiple combinations As has been shown, apart from combinations involving two C heads it was also possible for a negative element to appear in complementizer combinations. In what follows we will briefly examine the question of how combinations that morphologically consist of three complementizers can be analyzed since there are only two C positions in the left periphery hence there are not enough positions for generating three distinct C heads. First we will consider the conditions on the appearance of such combinations, and then we will discuss the case of hogyhamint ‘that if as’ and minthogyha ‘as that if ’, also showing that these fit into the system described above. Since (morphologically) complex complementizers ultimately grammaticalized into higher C heads, leaving the lower C position unfilled, the question arises whether such higher C heads could co-occur with a new, overt lower C head. This is naturally



Júlia Bácskai-Atkári and Éva Dékány

possible only if the complex C head grammaticalized relatively early, otherwise there would have been no complementizer potentially appearing in the lower C head, as all complementizers were reanalyzed as higher C heads. As was argued for before, it was hogyha ‘that if ’ to grammaticalize first as a complex C head, due to the preferred upward movement of hogy ‘that’. From this it follows that if there are combinations of the type complex C head + simplex C head attested historically then these should primarily be linked to the complex complementizer hogyha. This is in fact borne out: as pointed out by D. Mátai (2003: 424), the combination hogyhamint ‘that if as’ existed in Old and Middle Hungarian; the inverse order pair minthogyha ‘as that if ’ is still possible in Modern Hungarian. Both are, as expected, complementizers introducing conditional comparatives. The structure of hogyhamint is given below: CP

(131)

C’ C

CP

hogyha

C’ C

...

mint

Just as in the case of ordinary C + C combinations, there are two distinct heads in the structure: the higher one is hogyha and the lower one is mint ‘as’. The fact that the higher one is complex in itself is a matter of morphology but not the result of syntactic derivation. In this sense hogyhamint does not differ from C + C combinations described above. The derivation of minthogyha is in turn the result of head movement: CP

(132)

C’ C

CP

minthogyha C mint

C’ ...

From non-finite to finite subordination



The complementizer minthogyha is the result of mint moving up from the lower C head to the higher one and adjoining to hogyha there—hence in exactly the same way as was seen in connection with combinations containing two simplex C heads. Naturally, minthogyha ultimately also grammaticalized as a higher C head. ... Interim summary The development of the left periphery of Hungarian finite subordinate clauses involved the appearance of various complementizer combinations, in addition to the grammaticalization of diverse complementizers, as described in section 6.4.3. This was possible because the grammaticalization of the individual C heads did not take place exactly at the same time, and hence elements that grammaticalized earlier could appear higher in the structure and in a position distinct from where other elements appeared; in this way, it was possible for distinct heads to co-occur in one left periphery. These combinations either expressed new functions or they included the general finite subordination marker hogy ‘that’—in both cases, the combinations are in line with the development of finite subordination and the increased demand for the explicit marking thereof. The mechanisms of complementizer combinations can also be extended to combinations with negativelike polarity heads, the presence of which again contributed to the evolution of a robust functional left periphery. As was seen, the grammaticalization of complementizers ultimately contributed to the loss of syntactic combinations and Modern Hungarian has combinations only that are morphological in nature—that is, combinations that are base-generated as morphologically complex units in a single syntactic head. .. Complementizers in relative clauses The last part of this chapter is devoted to the discussion of relative clauses, which constitute a major subtype of finite subordinate structures. Relative clauses tend to be introduced by a relative pronoun, which occupies an operator position in the subordinate clause but there are languages that allow relative clauses to be introduced by a finite complementizer and a phonologically zero relative pronoun, such as that in English. As far as Hungarian is concerned, it seems that relative clauses have always required the overt presence a relative pronoun but it does not exclude the possibility of an overt complementizer at the same time, which is attested in Old and Middle Hungarian. The importance of this is that the presence of overt finite subordinators was motivated by the need to mark finite subordination, in parallel with the increased importance of finite subordination at the expense of non-finite structures. As was mentioned before, relative pronouns move to the specifier position of the lower CP (Kántor 2008, but see also É. Kiss 2002: 243–244 on relative pronouns moving to a [Spec,CP] position); this position is the same as the one where present-day complementizers moved to as operators in the left periphery (see section 6.4.3):



Júlia Bácskai-Atkári and Éva Dékány CP

(133)

C’ C Ø

CP Op.

C’ C

…

Ø

Relative pronouns can be found in the earliest texts already: (134) Es uimagguc szent peter urot kinec odut hotolm and pray-sbjv-1pl saint Peter lord-acc who-dat give-pass.3sg power ovdonia es ketnie loose-inf-3sg and bind-inf-3sg ‘and let us pray to the lord Saint Peter, to whom the power was given to loose and to bind’ (Funeral Sermon and Prayer) Since relative pronouns are arguments or adjuncts in the subordinate clause, there is considerable diversity to be observed among them. It has to be stressed that in Old Hungarian relative pronouns were not phonologically different from their interrogative counterparts, hence a form ki ‘who’ could stand for both an interrogative and a relative operator. The distinctive form of relative operators started to appear during Middle Hungarian and hence in Modern Hungarian there is a clear distinction between ki ‘who-Int.’ and aki ‘who-Rel.’. Furthermore, the distinction between ki ‘who’ and mi ‘what’ was not as clear-cut as it is in Modern Hungarian: in Modern Hungarian, ki is invariably associated with a [+animate] antecedent, while mi is [–animate]; by contrast, ki in Old Hungarian could also be associated with a [–animate] antecedent. In the example below the first instance of ki (kiknek ‘who-Dat.’) is associated with a [+animate] antecedent, while the second ki with a [–animate] one: (135) Vr¯a istèn ne tekozlad èl te testam¯etomodat lord-poss.1sg God not waste-imp-2sg off you testament-poss.2sg-acc se vegèd èl te irgalmassagodat m˙v tollonc ´ ´ neither take-imp-2sg off you piety-poss.2sg-acc we abl-1pl & Isaakert te Abrachamert te zèrètodert ´ Abraham-final you lover-poss.2sg-final and Isaac-final you zolgadert & Is¯rlert te scèntedert servant-poss.2sg-final and Israel-final you saint-poss.2sg-final

From non-finite to finite subordination



kiknc. bèzellettèl fogaduan hog megsokaseitanad who-pl-dat speak-pst-2sg swear-part that prt-multiply-cond-2sg magza˙t˙tokat o ´ they offspring-poss.2pl-acc ‘my Lord, do not waste your testament or take away your piety, for the sake of Abraham and Isaac, your faithful servants, whom you promised to multiply their descendants’ (Vienna C. 129) As shown by the representation in (133), relative pronouns moved to the lower [Spec,CP] position just as present-day complementizers originally did but whilst the latter grammaticalized as C heads, this is not true for present-day relative pronouns. This can easily be explained by taking into consideration that present-day relative pronouns have features that C heads are not allowed to have in Hungarian. That is, operators grammaticalizing into C heads have to lose e.g. person and number features (if they have any), which is not the case with present-day relative pronouns. In other words, an operator can grammaticalize into a C head if it loses its original syntactic and semantic roles or does not have features that would exclude its interpretation as a C head. This can be observed in other languages as well (for English, see Comrie 1999: 88 and Brook 2011; for German, see Bayer and Brandner 2008). The loss of features is described by Hancock and Bever (2009: 305) as the result of the Late Merge Principle, that is, a word that originally had a theta-role in the clause becomes a purely “syntactic” word (hence a functional head). This is precisely what has not happened in the case of relative pronouns and therefore they have not grammaticalized into C heads. Since this follows from general syntactic principles, the behavior of Hungarian relative pronouns is far from being exceptional, just as the grammaticalization of present-day complementizers. Relative clauses were introduced by single overt relative pronouns already in Old Hungarian; unlike Modern Hungarian, however, it was possible for relative clauses to be introduced by the sequence hogy ‘that’ + relative pronoun or ha ‘if ’ + relative pronoun in Old and Middle Hungarian (cf. Galambos 1907: 14–18; cf. also Haader 1995; Dömötör 1995). Consider the following examples: (136) a. olyaat tezok raytad hog kytol felz ´ ´ such-acc do-1sg you-sup that who-abl fear-2sg ‘I will do such a thing on you that you fear’

(Sándor C. 14v)

b. ky tegod zereth az nem epedh: ha ky keserg akkor ´ who you-acc love.3sg that not long.3sg if who moan.3sg then wygad rejoice.3sg ‘those who love you, do not long: those who moan, then rejoice’ (Czech C. 51–52)



Júlia Bácskai-Atkári and Éva Dékány

As shown by (136a), constructions with hogy could also have a consecutive meaning, though typically neither hogy nor ha contribute to the meaning of the construction and hence the clauses are purely relative: (137) a. ha mit keèndetec èn at’amtol èn if what-acc ask-mod-2pl I father-poss.1sg-abl I nèuembè agga t˙unèctec name-poss.1sg-ill give-3sg you-dat-2pl ‘whatsoever ye shall ask the Father in my name, he will give it you’ (Munich C. 103ra) b. ha myn kerendytek en atyamat en if what-sup ask-mod-2pl I father-poss.1sg-acc I newembe, aggya tynektek name-poss.1sg-ill give-3sg you-dat-2pl ‘whatsoever ye shall ask the Father in my name, he will give it you’ (Jordánszky C. 685) The examples in (137) show that the combinations ha + relative operator could appear in relative clauses without any additional meaning; such combinations were possible even in Middle Hungarian but are no longer available in Modern Hungarian. Similar structures can be found in Latin as well, e.g., si quid ‘if what’ and hence in what follows we will briefly examine the distribution of the Hungarian structure and its potential relatedness to the Latin counterparts. In Hungarian the combinations in question were quite productive, which is reinforced by the results of the research carried out on the four different Bible translations (cf. Bácskai-Atkári 2012b). Table 6.11 shows the number of occurrences for hogy/ha + relative operator combinations in the four gospels. There are only a few examples for combinations with hogy, but these appear already in Old Hungarian. More importantly, ha + operator combinations can be found in large numbers in Old Hungarian texts, decreasing in the Middle Hungarian trans-

Table .. The occurrences of hogy/ha + relative operator combinations

hogy + relative operator ha + relative operator

Munich Codex (1416/1466)

Jordánszky Codex (1516–1519)

Káldi‘s translation (1626)

Neovulgata translation (1997)

1

2

–

–

14

20

8

–

From non-finite to finite subordination



lation and—as expected—there are no examples for such combinations in Modern Hungarian. The relatively high number of ha + relative operator combinations in the Old Hungarian texts—especially compared to the 8 occurrences in Káldi’s translation— shows that combinations involving hogy and ha with a relative operator were quite frequent already in Old Hungarian. Of course, this is not to say that the frequency of such combinations in the selected texts strictly mirrors their frequency in Old or Middle Hungarian. In other words, the fact that ha + relative operator combinations are less frequent in the Middle Hungarian text than in the two Old Hungarian ones does not imply that it was also less frequent in Middle Hungarian than in Old Hungarian. Still, it should be obvious that the frequency of such combinations in Old Hungarian is far from insignificant. It is also worth considering that the texts discussed here are translations, which raises the question of how far the (Latin) original could induce the appearance of the Hungarian combinations. As far as the Munich Codex is concerned, all the combinations of the form ha + relative operator correspond to a Latin si + relative operator. In the Jordánszky Codex there are 6 additional occurrences of ha + relative operator (the other 14 having the same loci as the ones in the Munich Codex) but these correspond to a single relative operator in the Latin text and not to the complex of a C head and a relative operator. This clearly shows that the structure under scrutiny was in fact very productive and it cannot be considered as a Latin reflex in Old Hungarian either. The structure of the combinations is given below: CP

(138)

C’ C hogy Op. ha

CP C’ C

…

Ø

As can be seen, the structure is different from relative clauses introduced by a single relative pronoun (that is, not by a C head + relative pronoun combination) only in the presence of an overt complementizer in the higher C position. On the other hand, this kind of structure is the same configuration as the one where a higher C head (ha and hogy) combined with future complementizers still having an operator status (that is, mint ‘than/as’ and mert ‘because’). Hence the appearance of the combinations ha/hogy + relative pronoun is in line with the diachronic system outlined so far. Note that the co-occurrence of a C head and a relative pronoun in relative clauses is not unique to (earlier periods of) Hungarian: combinations like who that were also



Júlia Bácskai-Atkári and Éva Dékány

available for instance in Middle English: as described by van Gelderen (2004: 82, 105– 106), the grammaticalization of that into a C head meant that new operators could appear in the subclause as a way of reinforcement. This is shown by the following example (van Gelderen 2004: 106, ex. 19): (139) or who that dothe it I wyll paye

(Paston Letters #346, anno 1471)

As shown by (139), in English the relative pronoun preceded the complementizer and they were in fact located in the same CP projection; this is no longer possible since in Modern English such configurations are ruled out by the Doubly Filled COMP Filter. Finally, it is worth mentioning that the same type of combination can be observed in Modern Hungarian with mint: this became possible with the grammaticalization of mint as a higher C head. Just as with hogy and ha in (138), mint in these cases can be followed by an overt operator in the lower [Spec,CP] position. These are also relative operators that are comparative at the same time (in the same way as mint was before); there are various operators of this kind and hence there are several possible combinations, such as mint amilyen ‘than how’, mint ahány ‘than how much’ or mint ahogy ‘than how’. The structure of these combinations can be represented as follows: CP

(140)

C’ CP

C mint

Op.

C’ C

…

Ø

As can be seen, mint is base-generated in the higher C head and the operator moves to the lower [Spec,CP] position. This configuration is the same as the one for hogy/ha + operator combinations and structures containing a higher C head (hogy or ha) and an operator that came to be a complementizer later (mint and mert). Note that comparative subclauses always contain an operator (the comparative operator) but this can also be phonologically null (Chomsky 1977; Kennedy and Merchant 2000).

. Conclusion The aim of this chapter was to investigate the major changes in the history of Hungarian subordinate clauses and to show that finite subordination ultimately took over non-finite structures, which was strongly intertwined with the evolution of a

From non-finite to finite subordination



functional left periphery (the CP-domain) in finite embedded clauses. As was shown, this is also in line with the general change from SOV to SVO that took place between Proto-Hungarian and Old Hungarian: while an SOV setting typologically prefers nonfinite embedding, SVO languages tend to have finite subordination instead. Hence the frequency of finite subordinate structures increased at the expense of non-finites, accompanied by the loss of specific non-finite structures and the enrichment of finite ones. The ousting of non-finite structures is evidenced by the complete loss of an adverbial participle as well as the narrowing external distribution and reduced productivity (i.e. narrower class of base verbs) of several types of non-finites. In addition, the remaining types of non-finites became more prototypically non-finite, some of them losing the ability to license a referentially independent subject or agree with the subject. In finite clauses the development of a functional left-periphery can be observed, which came to be head-initial and started to exhibit multiple layers overtly. This involved the grammaticalization of various elements as C heads and the interaction thereof; the changes in question were also shown to be in line with general economy principles that can be observed in other, unrelated languages as well.

Appendix: Corpus building from Old Hungarian codices ESZTER SIMON

 Introduction The availability of annotated language resources is becoming an increasingly important factor in more and more domains of linguistic research, since high-quality linguistic databases can provide a fertile ground for theoretical investigations. Historical corpora represent a rich source of data, but only if the relevant information is specified in a computationally interpretable and retrievable way. Digitization should not be confined to scanning old manuscripts as images but should extend to making the primary data available in digital form. After linguistic enrichment, data sources ensure the possibility to access the data in a much more sophisticated way. Computers can provide support in ensuring consistency, completeness, and reliability of the metadata. Several databases of historical texts enriched with some kind of linguistic information and metadata have recently been created for various Indo-European languages, such as the PennHelsinki Parsed Corpus of Middle English (Kroch and Taylor 2000), the Tycho Brahe Parsed Corpus of Historical Portuguese (Galves and Faria 2010), or the Welsh Prose corpus (Thomas et al. 2007). One of the major aims of our project was to produce such an annotated corpus for specific stages of the history of the Hungarian language, similar in purpose to the initiatives of the database building projects mentioned above. The collection of Old Hungarian texts and presenting them in a computationally retrievable way started without any Hungarian predecessors. Thus, building the Old Hungarian Corpus was a pioneering effort. Building databases of historical texts and developing language processing tools for the cultural heritage domains is a highly interdisciplinary endeavor, which requires close collaboration across disciplines. General corpus building attempts tend to process texts which have been already digitized or originally created in an electronic format, but this is not the case with historical documents. Building corpora from the time before electronic formats is more costly and time-consuming and needs more laborious methods. The goal of this appendix is to describe the full workflow of text processing from scanning the codices to submitting queries through an online search service. Section 2 presents the acquisition of source data and the digitization process of the original linguistic material. It discusses the heterogeneity of the Old Hungarian orthographic system and several challenges

Appendix: Corpus building from Old Hungarian codices



during Optical Character Recognition (OCR) and text encoding. Section 3 gives an overview of corpus annotation, the normalization of tokens, the morphological analysis, and the morphosyntactic disambiguation. Section 4 presents the structure of the corpus, the text processing levels, and how linguistic annotation and several metadata are represented in the corpus. Section 5 describes the corpus query tool, which facilitates the linguistic analysis of large amounts of linguistic data.  Collecting the original linguistic material A corpus is a well-organized collection of data, “collected within the boundaries of a sampling frame designed to allow the exploration of certain linguistic features (or set of features) via the data collected” (McEnery 2003). The corpus should aim for balance and representativeness within a specific sampling frame, in order to allow a particular variety of a language to be studied. However, if the object of study is a highly restricted sublanguage or a dead language, identifying the texts to be included in the corpus is straightforward. This is the case with Old Hungarian texts: when constructing the Old Hungarian Corpus, we acquired all available sources from the Old Hungarian period, creating a corpus of a fixed size (more than 2.2 million tokens). The project aimed to collect and process only the continuous texts: codices and several minor texts, thus Hungarian fragments found in foreign texts were not considered. As a result, 47 codices have been made available digitally in their original orthographic form. Eleven of them have also been normalized, and four of them have been morphologically analyzed and morphosyntactically disambiguated. Furthermore, the original and normalized versions of several minor texts have also been produced. Work in the first phase started with the acquisition of source data, part of which has already been converted into some electronic text format. Documents coming from various sources (publishing companies or historical linguists) were converted into uniform, UTF-8 encoded plain text files. Another source was the Computational Database for Historical Linguistics (Jakab and Kiss 1994, 2001, Jakab 2002). The database contains the stems of the words of a few Old Hungarian codices in modern transcription, in alphabetical order. The corresponding tokens in their original form are presented with locus markers (page and row numbers), and orthographic, etymological, phonological, morphological, and semantic information. Since the information about the order of tokens in a row is not provided, recovering of the original word order was needed. Afterwards, the documents were converted into a uniform, UTF-8 encoded simple text. Normalized word forms were reconstructed from the combination of stems in modern transcription and the corresponding morphological information. The latter was also used to supply the part-of-speech (POS) tags and the full morphological analysis for each token. After a manual proofreading and correction, we obtained three codices digitally available in their original orthographic form, normalized, and morphologically analyzed. 2.1 The digitization process A significant part of the linguistic material was only available in print. In this case, digitization was carried out by manual typesetting or by scanning followed by a conversion process from the scanned images into regular text files aided by an OCR software.



Eszter Simon

2.1.1 Optical character recognition and the orthographic system of Old Hungarian Old Hungarian texts are heterogeneous mainly because of the absence of a spelling norm. The adaptation of the Latin alphabet to Hungarian posed several problems. The main challenge was that there are Hungarian phonemes which do not exist in Latin, so new characters were needed to represent them. The orthography in the 14th–16th centuries was far from uniform. In addition, one codex could be written by more than one author, which causes even more heterogeneity in the texts. Typically, sound–letter correspondences vary a great deal even within a single text sample. One sound is often written with various characters, e.g., vyragnac uiraga [virágnak virága] ‘flower of flower’ (Old Hungarian Lamentations of Mary). In addition, one letter can stand for multiple sounds, e.g., zerzete zerent [szerzete szerint] ‘after his order’ (Jókai C. 124). Moreover, some letters can represent either vowels or consonants as well, e.g., the letter v was used to represent the sounds [v, u, u:, y, y:] for centuries. According to Kniezsa’s classification (Kniezsa 1952), Hungarian phonemes not existing in Latin are represented in three ways: 1. In the first type, scribes work without diacritics: they combine more letters to represent a sound, e.g., [t∫] → ch ∼ cz ∼ chy ∼ chi ∼ cy. 2. The second type works with diacritics: letters with diacritical marks are used for representing Hungarian sounds, e.g., [t∫] → ć ∼ l ∼ l’ (l is the so-called Hussite [t∫], see section 2). 3. The third type is a kind of mixture of the first and the second types. In this case, the scribe applies letter combinations and diacritical marks as well, e.g., [t∫] → ch ∼ chy ∼ cyh ∼ c ∼ chi ∼ ch’ ∼ cz ∼ ts ∼ ć ∼ l ∼ l’ ∼ lh ∼ lz. As can be seen, Old Hungarian texts contain a large number of special characters, so a key aspect of an OCR software was its ability to be trained. This means that the software does not work with a closed set of characters, but has a training system built in, enabling it to deal with characters different from basic Latin ones. For this purpose, we used Abbyy FineReader 9.0 Professional Edition,1 which can be trained in an interactive way and produces a fairly goodquality result. The performance of the OCR system was evaluated by counting word accuracy, which is the rate of the number of correctly recognized words and the number of all words in a document. As expected, the results show that accuracy highly depends on the orthographic system used in a codex. We chose three codices representing the three orthographic types mentioned above for evaluation. A Modern Hungarian text was also processed as a baseline. As can be seen in Table 1, the best result was produced on the first orthographic type which does not use diacritical marks: this is similar to the result on Modern Hungarian text. The large number of special characters used in codices of the second and third types decreased the performance by approximately 30. In the case of codices using characters without diacritical marks, the task of the OCR system is recognizing the basic Latin characters, so it produces a fairly good quality result. However, recognizing complex, combined characters, which can be very similar to each other, causes difficulties. This is due to the fact that the OCR system cannot

1

.

Appendix: Corpus building from Old Hungarian codices



Table . Word accuracy of the OCR system for orthographic types codex

type

Kulcsár Munich Czech –

no diacritics diacritics mixed modern

word count

correct

word acc. ()

36,321 74,657 11,478 5,121

35,258 50,790 7,910 5,068

97.07 68.03 68.91 98.97

handle the diacritics properly, as is also mentioned in reports of similar projects, e.g., Volk et al. (2010). The OCR step was completed by extensive manual proof-reading and correction to ensure good quality initial resources as input to further processing steps. 2.1.2 Text encoding Character encoding is rarely an issue for languages like English that use basic Latin characters. However, for languages which use a large number of special characters, encoding is an important issue if one wants to build a consistent corpus to be searched reliably and displayed properly. Consistency is a basic requirement so that one can ask a query on the whole corpus. One of the great advantages of corpora is that they provide not only separate examples but all instances of the searched term, so analyses based on frequency become available. This important property of corpora can be ensured only if one follows the principle of consistency and always uses the same appropriate character for representing the same letter and different characters for representing different letters. For this purpose, we use UTF-8 encoded standard Unicode characters in the entire corpus. The Unicode Standard2 is a multilingual coding system which provides a consistent encoding for most of the world’s writing systems. Recently it became an international standard, which supports the worldwide interchange, processing, and display of written texts in diverse languages. One of the great advantages of Unicode is that it properly handles various accented and multiaccented characters, since basic characters and combining diacritical marks are represented by their own codes. For example, the character ÿ, which frequently appears in Old Hungarian texts, can be easily composed from a y and a combining diaeresis. Combining diacritical marks can also be accumulated, so that most of the special Old Hungarian characters can be represented by standard Unicode characters. However, it is a hard task to ensure consistency when dealing with such an extremely diverse language material as the one at hand. There is still an Old Hungarian character which is not present in Unicode charts: this is the so-called Hussite [t∫]. It is widely used in the Hussite Bible, the orthography of which was influenced by early 15th-century Czech spelling. This orthography later spread among Hungarian scribes and had a great influence on the spelling of later 16thcentury Hungarian codices. Later it disappeared and is not used in the Modern Hungarian alphabet. It looks like a small capital L (l) and is similar to some Unicode characters. However, one of the Unicode design principles is that characters have well-defined semantics, thus if we 2

.



Eszter Simon

want to be consistent, we cannot use the characters which only look alike, but are not the same in their semantics. For this reason, we decided to follow Volf (1874) and replace this character by č, which is used if and only if the Hussite [t∫] is used in the original codex. Codices from the Old Hungarian period are hand-written texts, which already have transcribed editions. We opted for using editions as the basis of corpus compilation. However, the editions were prepared in different periods, following different scientific requirements and restricted by varying typographical possibilities. Thus, the same character is often displayed in different ways in different editions. To ensure consistency, we eliminated this kind of randomness by using the same standard Unicode character for characters with the same semantics. When constructing the texts in their original orthographic form, we kept the punctuation marks, the hyphenation (or the lack thereof), and the upper- and lower-case letters as they are in the codices. However, we did not record the colors, boldface markings, and other kinds of emphasis applied in the codices. We did not aim for strict paleographic adherence, but our goal was to build a consistent database for linguistic research purposes.  Corpus annotation In the second phase of the corpus building workflow, linguistic annotation was developed. The development of an annotation requires a number of standard computational language processing tasks:

• tokenization and sentence segmentation; • normalization of tokens; • morphological analysis and morphosyntactic disambiguation. However, processing of texts from the time before electronic formats is far from trivial. Since spelling and punctuation rules in this early period of the Hungarian language were not regularized, this step requires manual work, which can be aided by automatic pre-processing tools. 3.1 Tokenization and sentence segmentation In the case of codices which have not been normalized, but are digitized only in their original orthographic form, tokenization means that we simply separate words from each other and concatenate hyphenated word parts. In the case of normalized codices, tokenization was done manually during the normalization step. We followed the Modern Hungarian spelling rules, so some words had to be split up, while others had to be joined together. When a word in the original text belongs to different constituents (as defined by our normalization guidelines), the word is split into the relevant parts. It is marked by double equals signs at the end of the first word and at the beginning of the next word. In (1), we present a typical case: verbal particles appear postverbally in imperative sentences and are spelled apart from the verb according to the Modern Hungarian spelling rules. However, they are often spelled together in Old Hungarian texts.

Appendix: Corpus building from Old Hungarian codices (1)

de säbädicz== ==mk mikët a gonostwl de szabadít-s meg mink-et a gonosz-tól but deliver-imp prt we-acc the evil-abl ‘but deliver us from evil’



(Munich Language Record 114v)

Words which are spelled apart in the original text, but constitute one word in Modern Hungarian, are joined, e.g., the noun phrase and the adverbial suffix in (2). When the original word is broken apart by a line or page break, it is marked by double at signs, as can be seen in (3). (2)

harmal napon halottay bool felthamata harmad nap-on halott-a-i-ból fel-támad-a third day-sup dead-poss-pl-ela up-rise-pst.3sg ‘on the third day he is risen from the dead’ (Munich Language Record 114v)

(3)

n¯e zorongat’t’a egmen-@@denic o at’t’afiat ´ nem szorongat-ja ő atyjafiá-t egymindenik he brother.poss.3sg-acc not thrust-def.3sg either ‘neither shall one thrust another’

(Vienna C. 205)

Since modern punctuation rules were created only in the 17th century, we cannot split the text into sentences based on the punctuation marks used in the original texts. For this reason, sentence splitting was made manually during the normalization step. In the case of non-normalized texts, we applied a quasi-sentence splitting, i.e., the text was split into 10-token sequences. 3.2 Normalization Because of the heterogeneity of the Old Hungarian orthographic system, a normalization step is required, in which the original tokens are transcribed into their modern form. This is a common step applied in most of the projects aiming at processing historical linguistic material, e.g., McEnery and Hardie (2003). Normalization is inevitable and is obviously of critical importance: without normalization the performance of automatic annotation in later stages will suffer a dramatic decrease (Rayson et al. 2007). One of the principal criteria of the normalization step is adherence to the original text—at least at the level of the morphosyntactic representation. Thus, we aimed for preserving all words and morphemes, even those which do not exist in Modern Hungarian. In (4), the word ýsa is an adverb which is known from the Funeral Sermon and Prayer. The word means ‘certainly’ and soon disappeared from the Hungarian language. Since we wanted to preserve all words, we normalized it as ‘isa’, not as ‘bizony’, its most appropriate translation into Modern Hungarian. (4)

ýsa pur es chomuv uogmuc isa por és hamu vagyunk sure dust and ash be.1pl ‘sure, we are dust and ashes’

(Funeral Sermon and Prayer)

In (5), the word form fekette preserves a morphological construction which does not exist in Modern Hungarian. It is an adverbial participle which is used to modify the verb phrase or the



Eszter Simon

whole sentence, thus it plays a role similar to that of an adverb. Its speciality is that it agrees with the subject of the construction in number and person (cf. section 6.3.1.4). There are adverbial participles even in Modern Hungarian, but they have only one form and do not agree with the subject. Following the principle of adherence to the original text, this and similar morphological constructions are preserved in the normalization process. (5)

fèkette lata o napat ´ ő napá-t fek-ett-e lát-á see-pst.def.3sg he mother.in.law.poss.3sg-acc lie-part-3sg ‘he saw his wife’s mother laid’

(Munich C. 14rb)

The second principle of normalization is consistency, thus orthographic variants of the same lexical item must be neutralized and converted into the same normalized version, e.g., m¯eden ∼ menden ∼ minden ∼ m˙enden ∼ m¯end˙en → minden ‘all’. We always followed the Modern Hungarian spelling rules during the normalization process. Since the Old Hungarian linguistic material mostly consists of Bible translations and religious texts, there is a large number of biblical names in it. Following the principle of consistency, proper names written in several diverse forms were also normalized. For this purpose, we used a modern Bible translation,3 and all names were transcribed to the form used in this translation. 3.3 Morphological analysis and disambiguation The normalization step has two main purposes: on the one hand, it makes it possible to find all instances of a word, irrespectively of how they are originally written; on the other hand, the normalized word form is the input to the morphological analysis. Since the original word forms are converted into Modern Hungarian spelling, technology developed for the morphological analysis of Modern Hungarian can be applied to the historical texts. We used the morphological analyzer engine called Humor (High speed Unification MORphology) (Prószéky and Kis 1999), which was originally developed for Modern Hungarian and has been adapted to Old Hungarian. First, Old Hungarian morphological constructions which are now extinct have been formalized and added to the grammar of the analyzer. Second, its lexicon has been expanded by adding special old words which are not used in the modern language. Since the analyzer generates all potential morphological analyses for each token, a disambiguation step is required to select the most appropriate analysis. We used HunPos (Halácsy et al. 2007), a statistical POS tagger, which requires a large amount of manually disambiguated Old Hungarian texts as a training corpus. For this purpose, morphologically analyzed and disambiguated texts produced from the Computational Database for Historical Linguistics (see Section 2) were used. For getting a corpus as error-free as possible, we manually validated and corrected the output. Since the theoretical aim of the project was to investigate syntactic changes in the history of the Hungarian language, we do not aim at full morphophonological representation, thus we do not mirror the whole morphemic structure of tokens in the analysis. The inflectional suffixes are

3

St. Stephen Association Bible translation ().

Appendix: Corpus building from Old Hungarian codices



fully encoded, while the derivational suffixes are not. Since Hungarian is a highly inflectional language, it expresses grammatical elements in a single word form using affixes for expressing grammatical phenomena. Suffixes, often multiple ones, must be attached to the word stem in strict order, and the last one always provides information about the syntactic role of the word form in the sentence. For this reason, several syntactic phenomena can be explored even at the morphological annotation level.  The structure of the corpus The structure of the corpus, i.e., the annotation levels, are parallel with the text processing steps, which are presented in Table 2. Based on this, six levels and five tasks can be distinguished throughout the text processing workflow. A sophisticated, linguistically relevant query often refers to different levels of language information contained in the corpus. For making all pieces of information available, the corpus contains all kinds of textual data corresponding to the text processing levels. Thus, for each token, the corpus provides the following pieces of linguistic information:

• • • •

original orthographic form (3): adÿad normalized form (4): adjad lemma (6): ad morphological analysis (6): V.Sub.S2.Def

Table . Text processing levels (1) scanned codex → automatic OCR (2) raw OCR output → manual correction (3) original orthographic form → manual normalization (4) normalized form → automatic morphological analysis (5) lemmatized and morphologically analyzed form → semi-automatic disambiguation (6) disambiguated form The example is a definite subjunctive/imperative form of the Hungarian verb ad ‘give’ in 2nd person singular. The numbers in parentheses are the numbers of the text processing levels (see Table 2) from which the information comes. The basic data format is the so-called multitag format, i.e., a tab separated simple text file which contains one token in every row and additional information corresponding to text processing levels in columns, as can be seen in Table 3. Sentence boundaries are marked by empty lines.



Eszter Simon Table . The multitag format page

original

normalized

lemma

analysis

1 1 1 1

Vram engem segeýtheny syees

Uram engem segíteni siess

Úr én segít siet

N:P.PxS1 N:Pro.S1.Acc V.Inf V.Subj.S2

The morphological analyzer adapted to Old Hungarian was originally developed for Modern Hungarian, for which it is widely used in the Hungarian language technology community. However, its linguistic formalism does not fit into any international annotation schemes. Therefore we plan to convert it into one of the widely used international standard formalisms. For clarification, here we provide the linguistic glosses of the example in Table 3. (6)

Vram engem segeýtheny syees Ur-am engem segít-eni sies-s Lord-poss.1sg I.acc help-inf hurry-imp.2sg ‘my Lord, hurry to help me’

(Festetics C. 1)

Besides the linguistic annotations, the corpus is also enriched with several kinds of metadata. The primary metadata are locus markers, which provide information about the place of the token in the original document (page, line, etc.). In texts containing Bible translations, biblical markers (book, chapter, verse) are also provided in a standard way, according to the modern Bible translation of the Szent István Társulat (St. Stephen Association), which provides the possibility of finding the given part in other Bible translations. The multitag format files also contain several other metadata in the form of the following codes:

• If a title or subtitle is part of the original text, it will have the TITLE code. Otherwise, it functions as a locus marker.

• Old Hungarian codices often contain parts in another language (mostly Latin). If the foreign word is provided with some kind of Hungarian inflection, i.e., it functions as a standard part of the Old Hungarian language, it will be normalized and morphologically analyzed as usual. However, if it is not inflected, it is only a foreign word wedged between Hungarian words, it will have the LANG{latin} code and will not be normalized and analyzed. • The scribe’s corrections in the original text material are also marked by codes: supplementary addition (ADD), cancellation (STRIKE), failed, but not cancelled word (FAIL), fragmentary word (FRAG).  The corpus query tool The effective retrieval of relevant information is fundamental for linguistic research. For this purpose, we have constructed a publicly available query interface (),

Appendix: Corpus building from Old Hungarian codices



which offers the user several features that greatly facilitate the linguistic analysis of large amounts of linguistic data.4 Text files in multitag format are converted into XML files, which are then validated, thereby checking the consistency of the database. These validated XML files are the suitable input for the Emdros corpus query engine (Petersen 2004), on which we have built the query interface. A good corpus query tool has to be able to formalize sophisticated linguistically relevant queries. Such queries often refer to different levels of language information contained in the corpus. Therefore, our corpus contains all of the linguistically relevant levels of language data (see Table 2), and the query interface allows the user to refer to these levels even simultaneously. The presentation of corpus results is independent of the query, in the sense that text processing levels different from the query can also be displayed. The corpus query interface allows the user to specify a query with the help of easy-to-use buttons and pop-up menus. The query is then formalized in the query language of Emdros, which can be edited for enabling more sophisticated queries. Figure 1 shows a sample part of the result page of a query in concordance format. We submitted a search for the normalized version of the Hungarian word föld ‘ground, earth’. In corcondance format, one result is one sentence, but the context can be broadened in a 5-sentence window. Above each result, there is a marker containing the name of the document in which the sample is found, the locus marker inside the document, and the unique identifier of the token. Each result is displayed in a tabular form: the original orthographic form in the first row (in green), the normalized form in the second row (in black), and the morphological information (lemma and analysis) in the third one (in grey). The requested word is always highlighted with boldface setting. [384] Konyvecse - 27r - 1/113219 Mÿnden fo͵ldreh kÿmeneh minden földre Kimene minden föld klmegy N.Sub N:Pro VPfx.V.Ipf.S3

az

o͵

zo͵ngesek

͵

az

ő

zöngésük

͵

zöngés az ő Det N:Pro.S3 N.PxP3

[385] Konyvecse - 27r - 1/113241 zer-@@zed

o͵keth

feÿedelmo͵l

mÿnden

fo͵ldo͵n

:

szerzéd

őket

fejedelemül

minden

földön

:

fejedelem N.Ess

minden N:Pro

föld N.Sub

ők szerez V.Ipf.S2.Def N:Pro.P3.Acc [386] Festk - 3 - 1/115123 merth ew

kezeeben

wadnak

fewldnek

. mynden

wėgey

:

merth ő

kezében

vannak

földnek

minden

végei

͵

merth ő C N:Pro.S3.Nom_gen

kéz van N.PxS3.Ine V.P3

föld N.Dat_gen

minden N.Pro

vég N.PxS3.Pl

Figure  A sample part of the result page in concordance format. 4 The query interface has been built in our project web site which is available via the URL . The Old Hungarian texts in their original orthographic form and their normalized versions are also available on this web site.



Eszter Simon

There is another feature of the corpus query tool in concordance format: the user can add more queries to the one already submitted in an n-word window, where n can be specified by the user. This provides the possibility to search several morphosyntactic patterns in sentences. To illustrate the usefulness of this feature, we present a real research question which is relevant for Hungarian in a diachronic perspective. Modern Hungarian makes extensive use of the definite article, but in Old Hungarian, the definite article appears only in constructions where the referent of the noun phrase is not anchored in another way (cf. section 3.2.4). Possessor expressions represent one instance of such constructions, where the definite article is an obligatory element of the possessive construction with a dative-marked possessor in Modern Hungarian, e.g., in (7), but is absent in Old Hungarian, e.g., in (8). (7)

az ember-ek-nek a fia-i-val the man-pl-dat the son.poss-pl-ins ‘with the sons of the people’

(8)

embereknek fÿaÿual ∅ ember-ek-nek ∅ fia-i-val man-pl-dat son.poss-pl-ins ‘with the sons of the people’

(Könyvecse 4v)

How can we check such a linguistic hypothesis against a corpus, that is, how can we search for something which is absent? In this case, we can submit a query on the morphologically analyzed part of the corpus by using the morphosyntactic annotation and the feature of context addition. Here we want to find a dative-marked possessor directly followed by a possessive-marked noun. We can formalize this query in the following way: [W FOCUS w_6e ~ ‘Dat_gen\)\)$’] [W FOCUS w_6e ~ ‘^6e\(\(N.Px’] In this query, we do not allow any other elements to stand between the possessor and the possessive. This query resulted in more than 2,000 hits. We can also formalize the case when one additional element is allowed to stand between them, either a determiner or an adjective: [W FOCUS w_6e ~ ‘Dat_gen\)\)$’] .. BETWEEN 1 AND 1 [W FOCUS w_6e ~ ‘^6e\(\(N.Px’] After investigating the results, we found that most of the results contain an adjective before the noun, not an article. To reduce the number of the results to the relevant hits where a possessor is directly followed by a definite article and a possessive-marked noun, we have to use the normalized level of the corpus. Since definite and non-definite articles are not distinguished on the morphosyntactic annotation level, but the definite article has only two forms (a and az), the formalization of the word between the possessor and the possessive can be specified on the normalization level in the following way: [W FOCUS w_6e ~ ‘Dat_gen\)\)$’] [W FOCUS w_4 ~ ‘^4\(\(az?\)\)$’] [W FOCUS w_6e ~ ‘^6e\(\(N.Px’]

Appendix: Corpus building from Old Hungarian codices



Since this query resulted in only one hit, this is a good indicator of the fact that the definite article was not used in dative-marked possessor constructions in Old Hungarian. Besides the concordance format, the corpus query tool also allows the user to ask for a frequency list. This service is not only for listing all possible variants of each word in the texts, but it calculates the total amount of each original orthographic form and the normalized word form itself. As can be seen in Figure 2, the corpus contains 596 occurrences of the word föld ‘ground, earth’. (Recall that only part of the entire linguistic material has been normalized, i.e., the result can only be interpreted on the normalized subcorpus.) Query: [W FOCUS w_4 ~ ' fŏld ' ] Number of hits: 596 - Elapsed time: 25s 49 db fo͵ldo͵n

földön fo͵ldnèc

44 db

fo͵ld

42 db

fo͵ldèt

29 db

fo͵ldebo͵l

18 db

fo͵ldo͵n

18 db

földnek föld földet földjéből földön föld N.Sup

Figure  A sample part of the result page in frequency list format.  Final remarks The method of gaining empirical linguistic data from searchable historical corpora in order to describe and reconstruct diachronic changes has recently become extremely popular and can be considered as one of the mainstream linguistic trends. In view of that, our efforts made so far for developing an open access, digitized historical corpus for Hungarian are state-of-the-art efforts. Moreover, since Hungarian is the longest documented language of the Uralic language family, the searchable corpus and the theoretical results based on it can be useful not only for the study of Hungarian but also for comparative Uralic studies. Additionally, both the corpus and the theoretical findings are extremely useful for researchers who are interested in approaching language change from a perspective broader than the well-established Indo-European one. However, one can never say that a corpus is finished; it can be extended both in a horizontal and in a vertical dimension to be the source for wider and deeper theoretical investigations in the future. As for the horizontal dimension, we plan to expand the database by adding Middle Hungarian sources, mainly Bible translations. These texts can then be compared to the Old Hungarian records of similar content in a particularly efficient way to observe and track the gradual changes in the Hungarian language. Developing the database in a vertical way includes



Eszter Simon

the linguistic annotation (normalization, morphological analysis, and disambiguation) of the Old Hungarian texts which have not been normalized yet. The more support linguists can get from the historical corpus, the larger the quantity of results that can be expected. One of the prospective extensions is making the corpus bilingual. The addition of an English vocabulary to the already normalized parts of the corpus, would make the historical corpus accessible to non-Hungarian users as well in the future.

Primary Sources Apor Codex

After 1416/end of the 15th century–beginning of the 16th century. György Volf (ed.), Apor-kódex. Budapest, 1879.

Babits Mihály: Jónás könyve

1938. Mihály Babits: Jónás könyve (= The Book of Jonas). E-edition at .

Balassi Ének

Balassi Bálint énekei (= The songs of Bálint Balassi). Lőcse/ Levoča, 1693.

Birk Codex

1474. István Pusztai (ed.), Birk-kódex. Facsimile, transcription of the original record, and the corresponding Latin text. Budapest: Akadémiai Kiadó, 1960.

Bod Codex

Beginning of the 16th century. István Pusztai (ed.), Bod-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1987.

Christina Legend

Beginning of the 16th century. Lilla Vekerdy (ed.), Krisztinalegenda. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1988.

Cisio

1592. Cisio. Kolozsvár/Cluj-Napoca (= RMNy 684). In: Gedeon Borsa and Ferenc Hervay (eds.), Régi Magyarországi Nyomtatványok I. (1473–1600), Budapest: Akadémiai Kiadó, 1971.

Cornides Codex

1514–1519. András Bognár and Ferenc Levárdy (eds.), Cornideskódex. Facsimile, and critical edition. Budapest: Akadémiai Kiadó, 1967.

Czech Codex

1513. Csilla N. Abaffy (ed.), Czech-kódex. Facsimile, transcription of the original record. Budapest: Magyar Nyelvtudományi Társaság, 1990.

Debrecen Codex

1519. Edit Madas, Andrea Reményi, and Csilla Abaffy (eds.), Debreceni kódex. Facsimile, transcription of the original record. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 1997.

Döbrentei Codex

1508. Csilla Abaffy, Csilla T. Szabó, and Edit Madas (eds.), Döbrentei-kódex. Halábori Bertalan keze írásával. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 1995.



Primary Sources

Domonkos Codex

1517. Gyöngyi Komlóssy and Klára Korompay (eds.), Domonkos-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1990.

Érdy Codex

1524–1527. Unpublished transcription of the original text, received from the Sermones project, .

Festetics Codex

Before 1494. Csilla N. Abaffy (ed.), Festetics-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 1996.

Funeral Sermon and Prayer

Around 1195. Halotti beszéd és könyörgés. In: Loránd Benkő: Az Árpád-kor magyar nyelvű szövegemlékei. Budapest: Akadémiai Kiadó, 1980. 47–49. A more recent reading: Abaffy Erzsébet: Korai kis szövegemlékeink újabb olvasata. Magyar Nyelv 86 (1990) 124–127.

Gömöry Codex

1516. Lea Haader and Zsuzsanna Papp (eds.), Gömöry-kódex. Facsimile, transcription of the original record. Budapest: MTA Nyelvtudományi Intézet, 2001.

Guary Codex

Before 1495. Dénes Szabó (ed.), Guary-kódex. Budapest, 1944.

Gyulafehérvár Lines

Second half of the 13th century. In: Loránd Benkő: Az Árpádkor magyar nyelvű szövegemlékei. Budapest: Akadémiai Kiadó, 1980. 47–49. A more recent reading: Abaffy, Erzsébet: Korai kis szövegemlékeink újabb olvasata. Magyar Nyelv 86 (1990) 124–127.

Horvát Codex

1522. Lea Haader and Zsuzsanna Papp (eds.), Horvát-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1994.

Jókai Codex

After 1372/around 1448. János P. Balázs (ed.), Jókai-kódex. Transcription of the original record, the corresponding Latin text, with introduction and notes. Budapest: Akadémiai Kiadó, 1981.

Jordánszky Codex

1516–1519. Ferenc Toldy and György Volf (eds.), A Jordánszkykódex bibliafordítása. Buda, 1888.

Káldi

1626. Szent Biblia (= The Holy Bible). Translated by György Káldi. Wien, 1626.

Károli

1590. Szent Biblia (= The Holy Bible). Translated by Gáspár Károli. Vizsoly, 1590. Facsimile published by Helikon Kiadó, Budapest, 1990.

Kazinczy Codex

1526–1541. Zsuzsa Kovács (ed.), Kazinczy-kódex. Budapest: Magyar Nyelvtudományi Társaság, 2003.

Primary Sources



Königsberg Fragment and Ribbons

Beginning of the 13th century/around 1350. In: Loránd Benkő: Az Árpád-kor magyar nyelvű szövegemlékei. Budapest: Akadémiai Kiadó, 1980. 49–52.

Könyvecse

(= Booklet on the Dignity of the Apostles) 1521. István Pusztai (ed.), Könyvecse a szent apostoloknak méltóságokról. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1985.

Kulcsár Codex

1539. Lea Haader and Zsuzsanna Papp (eds.), Kulcsárkódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: MTA Nyelvtudományi Intézet, 1999.

Lázár Codex

First quarter of the 16th century. Csilla N. Abaffy (ed.), Lázár Zelma-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1992.

Lobkowicz Codex

1514. Andrea Reményi (ed.), Lobkowicz-kódex. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 1999.

Margaret Legend

1510. János P. Balázs, Adrienne Dömötör, and Katalin Pólya (eds.), Margit-legenda. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1990.

Marosvásárhely Lines

Around 1410. In: József Molnár and Gyöngyi Simon: Magyar nyelvemlékek. 63. Budapest: Tankönyvkiadó, 1976.

Miskolc Fragment

1525. Zsuzsanna Papp and Zsuzsa Kovács (eds.), Vitkovics-kódex és Miskolci töredék. Facsimile, transcription of the original record, with notes. Budapest: MTA Nyelvtudományi Intézet, 1991.

Munich Codex

1466. Antal Nyíri (ed.), Müncheni kódex. Critical edition with the corresponding Latin text. Budapest: Akadémiai Kiadó, 1971.

Munich Language Record

Sarbak Gábor: Magyar nyelvemlék a XVI. század elejéről a Bajor Nemzeti Könyvtárban. Magyar Nyelv 101 (2005) 147–160; Lea Haader: A Müncheni emlék. Magyar Nyelv 101 (2005) 161–178.

Nádor Codex

1508. István Pusztai and Edit Madas (eds.), Nádorkódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1994.



Primary Sources

Nagyszombat Codex

1512–1513. Csilla T. Szabó (ed.), Nagyszombati kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 2000.

Neovulgata

The edition of the so-called Káldi-Neovulgata Bible. Budapest: Szent Jeromos Katolikus Bibliatársulat, 1997.

Old Hungarian Lamentations of Mary

Middle of the 13th century. In: Benkő, Loránd: Az Árpád-kor Magyar nyelvű szövegemlékei. Budapest: Akadémiai Kiadó, 1980. 52–53. A more recent reading: Abaffy, Erzsébet: Korai kis szövegemlékeink újabb olvasata. Magyar Nyelv 86 (1990) 124–127.

Peer Codex

Around 1518. Andrea Kacskovics-Reményi and Beatrix Oszkó (eds.), Peer-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 2000.

Példák könyve

(= Book of Proverbs) 1510. András Bognár and Ferenc Levárdy (eds.), Példák könyve. Budapest: Akadémiai Kiadó, 1960.

Pesti NT

1536. Pesti Mizsér Gábor: Új Testamentum magyar nyelven (= The New Testament in Hungarian). Facsimile edition: Budapest: MTA, [2002].

Pressburg Codex

1520. Csilla Abaffy, Erzsébet Abaffy, and Edit Madas (eds), Pozsonyi kódex. Budapest: Magyar Nyelvtudományi Társaság, 2004.

Sándor Codex

Around 1518. István Pusztai (ed.), Sándor-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1987.

Simor Codex

Beginning of the 16th century. Lilla Vekerdy (ed.), Simor-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1988.

Székelyudvarhely Codex

1526–1528. Csilla N. Abaffy (ed.), Székelyudvarhelyi kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1993.

Thewrewk Codex

1531. Judit Balázs and Gabriella Uhl (eds.), Thewrewkkódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: MTA Nyelvtudományi Intézet, 1995.

Primary Sources



Tihanyi Codex

1530–1532. Zsuzsa Kovács (ed.), Tihanyi kódex. Budapest: Magyar Nyelvtudományi Társaság, 2007.

Vienna Codex

After 1416/around 1450. Gedeon Mészöly (ed.), Bécsi kódex. Budapest: MTA, 1916.

Virginia Codex

Beginning of the 16th century. Zsuzsa Kovács (ed.), Virginia-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1990.

Vitkovics Codex

1525. Zsuzsanna Papp, István Pusztai and Zsuzsa Kovács (eds.), Vitkovicskódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: MTA Nyelvtudományi Intézet, 1991.

Weszprémi Codex

Around 1512. István Pusztai (ed.), Weszprémi-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1988.

Winkler Codex

1506. István Pusztai (ed.), Winkler-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Akadémiai Kiadó, 1988.

References Abbott, Barbara, 2004. Definiteness and indefiniteness. In: Laurence R. Horn and Gregory Ward, eds. The handbook of pragmatics. Oxford: Blackwell Publishing, 122–149. Adger, David, 2007. Three domains of finiteness: A minimalist perspective. In: Irina A. Nikolaeva, ed. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press, 23–58. A. Jászó, Anna, 1970. Melléknévi igenevek verbum finitumként az osztják nyelv szigvai nyelvjárásában [Adjectival participles as finite verbs in Sigva Khanty]. Nyelvtudományi Közlemények 71: 123–127. A. Jászó, Anna, 1975. A szófaji átcsapás egy esete az osztjákban [A case of change of word class in Khanty]. Nyelvtudományi Közlemények 77: 155–161. A. Jászó, Anna, 1976. Megjegyzések a participiumból alakult verbum finitumok mondattanához az északi osztjákban [Notes on the syntax of finite verbs derived from participles in Northern Khanty]. Nyelvtudományi Közlemények 78: 353–358. A. Jászó, Anna, 1992. Az igenevek [Non-finite verbs]. In: Loránd Benkő, Erzsébet E. Abaffy, and Endre Rácz, eds. A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor morfematikája. Budapest: Akadémiai Kiadó, 411–454. Alberti, Gábor, 1997. Restrictions on the degree of referentiality of arguments in Hungarian sentences. Acta Linguistica Hungarica 44: 341–362. Alboiu, Gabriela and Virginia Hill, 2013. Gerunds in Early Modern Romanian. Handout of a talk delivered at the 43rd Linguistic Symposium on Romance Languages. Alexiadou, Artemis et al., 2007. Noun phrase in the generative perspective. Studies in Generative Grammar 71. Berlin–New York: Mouton de Gruyter. Asbury, Anna, 2008. The morphosyntax of case and adpositions. Ph.D. dissertation, Utrecht University: LOT. Bach, Emmon, Eloise Jelinek, Angelika Kratzer, and Barbara Hall Partee, eds. 1995. Quantification in natural languages. Studies in Linguistics and Philosophy 54. Dordrecht: Kluwer. Bácskai-Atkári, Júlia, 2011. A komparatív operátor esete a mondatbevezetővel: Szintaktikai változások a magyar hasonlító mellékmondatokban [The comparative operator’s affair with the complementizer: Syntactic changes in Hungarian comparative subclauses]. In: Katalin É. Kiss and Attila Hegedűs, eds. Nyelvelmélet és diakrónia. Budapest and Piliscsaba: Szent István Társulat, 103–119. Bácskai-Atkári, Júlia, 2012a. The relative cycle in Hungarian declaratives. Talk delivered to: 14th Diachronic Generative Syntax Conference (DiGS14), Lisbon, 4–6 July 2012. Bácskai-Atkári, Júlia, 2012b. 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. Bácskai-Atkári, Júlia, 2013. Hungarian embedded interrogatives from a diachronic perspective. Talk delivered to: 3. Sprachwissenschaftliche Tagung für Promotionsstudierende (3. STaPs), Heidelberg, 5–6 April 2013.

References



Bácskai-Atkári, Júlia, and Gergely Kántor, 2012. Deletion in Hungarian, Finnish and Estonian comparatives. Finno-Ugric Languages and Linguistics 1/1–2: 44–66. Baker, Mark, 1985. Mirror Theory and morphosyntactic explanation. Linguistic Inquiry 15: 373–415. Baker, Mark, 1988. Incorporation: a theory of grammatical function changing. Chicago: University of Chicago Press. Baker, Mark, 2008. The syntax of agreement and concord. Cambridge: Cambridge University Press. Bakró-Nagy, Marianne, 1999. A magyar határozott névelő kialakulásának szintaktikai vonatkozásai [Syntactic aspects of the development of the Hungarian definite article]. In László Büky and Tamás Forgács, eds. A nyelvtörténeti kutatások újabb eredményei I. Magyar és finnugor mondattörténet. Szeged: JATE Magyar Nyelvészeti Tanszék, 5–13. Bartos, Huba, 2000. Az inflexiós jelenségek szintaktikai háttere [The syntactic background of inflectional phenomena]. In Ferenc Kiefer, ed. Strukturális magyar nyelvtan 3. Morfológia. Budapest: Akadémiai Kiadó, 653–762. Bartos, Huba, 2001. Mutató névmási módosítók a magyar főnévi szerkezetben: egyezés vagy osztozás? [Demonstrative modifiers in the Hungarian noun phrase: agreement or sharing?]. In Marianne Bakró-Nagy, Zoltán Bánréti, and Katalin É. Kiss, eds. Újabb tanulmányok a strukturális magyar nyelvtan és a nyelvtörténet köréből. Budapest: Osiris, 19–41. Bartos, Huba, 2009. The syntax of Hungarian -vA adverbial participles: a single affix with variable merge-in locations. In Katalin É. Kiss, ed. Adverbs and adverbial adjuncts at the interfaces. Berlin and New York: Mouton de Gruyter, 75–101. Barwise, Jon, and Robin Cooper, 1981. Generalized quantifiers and natural language. Linguistics and Philosophy. 4: 159–219. Bayer, Josef, and Ellen Brandner, 2008. On wh-head-movement and the Doubly-Filled-Comp Filter. In: Charles B. Chang and Hannah J. Haynie eds. Proceedings of the 26th West Coast conference on formal linguistics. Somerville, MA: Cascadilla Proceedings Project, 87–95. Beghelli, Filippo, and Timothy Stowell, 1997. Distributivity and negation: The syntax of each and every. In: Anna Szabolcsi, ed. Ways of scope taking. Dordrecht: Kluwer, 71–107. Behaghel, Otto, 1932. Deutsche Syntax IV. Heidelberg: Carl Winters. Bende-Farkas, Ágnes, 2013. The semantics of Old Hungarian floating mind. Paper presented at the 11th International Conference on the Structure of Hungarian. Pázmány Péter Catholic University, Piliscsaba. 29–31 August 2013. Benkő, Loránd, 1980. Az Árpád-kor magyar nyelvű szövegemlékei [Hungarian text records from the Árpád Age]. Budapest: Akadémiai Kiadó. Benkő, Loránd, ed., 1964–1987. A magyar nyelv történeti-etimológiai szótára. I–IV. [The historical-etymological dictionary of Hungarian]. Budapest: Akadémiai Kiadó. Benkő, Loránd, ed., 1993. Etymologisches Wörterbuch des Ungarischen. I–II. Budapest: Akadémiai Kiadó. Bereczki, Gábor, 1983. A Volga–Káma-vidék nyelveinek areális kapcsolatai [Areal contacts of the languages of the Volga–Kama area]. In: János Balázs, ed. Areális nyelvészeti tanulmányok. Budapest: Tankönyvkiadó, 207–237. Berg, Martin van den, 1996. The internal structure of discourse. Ph.D. dissertation, Amsterdam: University of Amsterdam.



References

Bernstein, Judy B., 1997. Demonstratives and reinforcers in Romance and Germanic languages. Lingua 102: 87–113. B. Gergely, Piroska, 1980. Az által ∼ át igekötő, illetve névutó történetéhez [On the history of the particle and postposition által ∼ át ‘through, across’]. In: Samu Imre, István Szathmári, and László Szűts, eds. A magyar nyelv grammatikája. A magyar nyelvészek III. nemzetközi kongresszusának előadásai. Budapest: Akadémiai Kiadó, 351–356. Bianchi, Valentina, 2003. On finiteness and logophoric anchoring. In: Jacqueline Guéron and Liliane Tasmowski eds. Temps et point de vue/Tense and point of view. Nanterre: Université Paris X, 213–246. Biberauer, Theresa, Anders Holmberg, and Ian Roberts, 2007. Disharmonic word-order systems and the Final-over-Final-Constraint FOFC. In: Antonietta Bisetto and Francesco E. Barbieri eds. Proceedings of the XXXIII incontro di grammatica generativa. Bologna: Università di Bologna, 86–105. Biberauer, Theresa, Anders Holmberg, and Ian Roberts, 2008a. Linearising disharmonic word orders: the Final-over-Final Constraint. In: Jong Yurl Yoon and Kyoung Ae Kim, eds. The perspectives of linguistics in the 21st century. Seoul: ICLK, 301–318. Biberauer, Theresa, Anders Holmberg, and Ian Roberts, 2008b. Structure and linearization in disharmonic word orders. In: Charles B. Chang and Hannah J. Haynie, eds. Proceedings of the 26th West Coast conference on formal linguistics. Somerville, MA: Cascadilla, 96–104. Biberauer, Theresa, Anders Holmberg, and Ian Roberts, 2009. Linearization and the architecture of grammar: a view from the Final-over-Final Constraint. In: Vincenzo Moscati, and Emilio Servidio, eds. Proceedings of XXXV incontro di grammatica generativa, University of Siena CISCL Working Papers 3: 77–89. Bíró, Bernadett, and Katalin Sipőcz, 2013. Syntactic variation in Mansi ditransitive constructions. In: Grammar and context IV: New approaches to the Uralic languages, 6–8 June, Tartu, book of abstracts, 32–35. Bisang, Walter, 2007. Categories that make finiteness: Discreteness from a functional perspective and some of its repercussions. In: Irina A. Nikolaeva, ed. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press, 115–137. Brassai, Sámuel, 1860–65. A magyar mondat [The Hungarian sentence]. Akadémiai Értesítő 1: 279–399, 3: 3–128, 173–409. Brisson, Christine, 1998. Distributivity, maximality, and floating quantifiers. Ph.D. dissertation, New Brunswick, NY: Rutgers University. Brody, Michael, 1990a. Remarks on the order of elements in the Hungarian focus field. In: István Kenesei, ed. Approaches to Hungarian 3. Szeged: JATE, 95–122. Brody, Michael, 1990b. Some remarks on the focus field in Hungarian. UCL Working Papers in Linguistics 2: 201–225. Brody, Michael, 1995. Focus and Checking Theory. In: István Kenesei, ed. Approaches to Hungarian 5. Szeged: JATE, 29–44. Brook, Marisa, 2011. One of those situations where a relative pronoun becomes a complementizer: A case of grammaticalization in progress. . . again. In: Proceedings of the 2011 Canadian Linguistics Association annual conference. (accessed: 11 October 2012).

References



Campbell, Lyle, and Alice C. Harris, 2002. Syntactic reconstruction and demythologizing “Myths and the prehistory of grammars”. Journal of Linguistics 38: 599–618. Cheng, Lisa Lai-Shen, 2009. On every type of quantificational expression in Chinese. In: Anastasia Giannakidou and Monika Rathert, eds. Quantification, definiteness, and nominalization. Oxford: Oxford University Press, 53–75. Chomsky, Noam, 1977. On WH-movement. In: Peter W. Culicover et al., eds. Formal syntax. New York: Academic Press, 71–132. Chomsky, Noam, 1995. The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam, 1998. Minimalist inquiries: The framework. MIT Working Papers in Linguistics 15: 1–61. Chomsky, Noam, 2001. Derivation by phase. In: Michael Kenstowicz, ed. Ken Hale: A life in language. Cambridge, MA: MIT Press, 1–52. Collinder, Björn, 1960. Comparative grammar of the Uralic languages. Stockholm: Almquist and Wiksell. Comrie, Bernard, 1980. Inverse verb forms in Siberia: Evidence from Chukchee, Koryak and Kamchadal. Folia Linguistica 1: 61–74. Comrie, Bernard, 1998. Rethinking the typology of relative clauses. Language Design 1: 59–86. Comrie, Bernard, 1999. Relative clauses: structure and typology on the periphery of Standard English. In: Peter Collins and David A. Lee, eds. The clause in English: In honor of Rodney Huddleston. Philadelphia and Amsterdam: John Benjamins, 81–91. Cowper, Elizabeth, 2002. Finiteness. Manuscript, University of Toronto. Available at (accessed: 6 May 2013). Croft, William, 2000. Explaining language change: An evolutionary approach. London and New York: Longman. Csepregi, Márta, 2012. Participiális jelzős szerkezetek két hanti nyelvjárásban [Adjectival participles in two dialects of Khanty]. Nyelvtudományi Közlemények 108: 61–94. Dalrymple, Mary, and Irina Nikolaeva, 2011. Objects and information structure. Cambridge: Cambridge University Press. Décsy, Gyula, 1990. The Uralic protolanguage: A comprehensive reconstruction. Bloomington, Indiana: Eurolingua. Dékány, Éva, 2012a. A profile of the Hungarian DP: The interaction of lexicalization, agreement, and linearization with the functional sequence. Ph.D. dissertation, University of Tromsø. Dékány, Éva, 2012b. Nem jöttem hínia az igazakat. Az ómagyar anti-egyeztetett főnévi igenevekről [On anti-agreeing infinitives in Old Hungarian]. Nyelvtudományi Közlemények 108: 219–252. Dékány, Éva, and Veronika Hegedűs, 2013. Word order variation in Hungarian PPs. Talk presented at the 11th International Conference on the Structure of Hungarian. Pázmány Péter Catholic University, Piliscsaba, 29–31 August 2013. Dikken, Marcel den, 2010. On the functional structure of locative and directional PPs. In: Guglielmo Cinque and Luigi Rizzi, eds. Mapping spatial PPs. The cartography of syntactic structures, volume 6. Oxford and New York: Oxford University Press, 74–126. D. Mátai, Mária, 1991. Az igekötők [Particles]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana I. A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 433–441.



References

D. Mátai, Mária, 2003. Szófajtörténet [The history of word classes]. In: Jenő Kiss and Ferenc Pusztai, eds. Magyar nyelvtörténet. Budapest: Osiris Kiadó. 204–233 and 393–429. D. Mátai, Mária, 2007. A magyar szófajtörténet általános kérdései [General questions about the history of Hungarian word classes]. Budapest: Akadémiai Kiadó. Dömötör, Adrienne, 1995. Az alárendelő mondatok: A jelzői mellékmondatok [Subordinate clauses: Attributive subclauses]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/2.: A kései ómagyar kor: Mondattan. Szöveggrammatika [A historical grammar of Hungarian II/2. Syntax and text in Late Old Hungarian]. Budapest: Akadémiai Kiadó, 666–693. Dömötör, Adrienne, 2008. A főnévi névmási kijelölő jelző a középmagyar korban [The pronominal demonstrative attribute in the Middle Hungarian period]. In: László Büky, Tamás Forgács, and Balázs Sinkovics, eds. A nyelvtörténeti kutatások újabb eredményei V. Szeged: Szegedi Tudományegyetem Magyar Nyelvészeti Tanszék, 17–25. E. Abaffy, Erzsébet, 1991. Az igemód- és igeidő-rendszer [Verbal mood and tense]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana I. A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 104–121. E. Abaffy, Erzsébet, 1992. Az igemód- és igeidő-rendszer [Verbal mood and tense]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor morfematikája Budapest: Akadémiai Kiadó, 120–183. Eckardt, Regine, 2006. Meaning change in grammaticalization: An enquiry into semantic reanalysis. Oxford: Oxford University Press. Egedi, Barbara, 2013. Grammatical encoding of referentiality in the history of Hungarian. In: Anna Giacolone Ramat, Caterina Mauri, and Piera Molinelli, eds. Synchrony and diachrony: A dynamic interface. Studies in Language Companion Series 138. Amsterdam: John Benjamins, 367–390. Egedi, Barbara, 2014. A latin forrásszövegek hatása az ómagyar szintaktikai jelenségekre [The impact of Latin source texts on Old Hungarian syntactic phenomena]. In: Katalin É. Kiss, ed. Magyar történeti mondattan. Budapest: Akadémiai Kiadó. Egedi, Barbara, forthcoming. Word order change at the left periphery of the Hungarian noun phrase. To appear in Anna Maria Martins and Adriana Cardoso, eds. Word order change. Oxford: Oxford University Press. Egedi, Barbara, and Eszter Simon, 2012. Gradual expansion in the use of the definite article: Checking a theory against the Old Hungarian Corpus. Talk presented at Exploring ancient languages through corpora. University of Oslo. 14–17. June 2012. . É. Kiss, Katalin, 1987. Configurationality in Hungarian. Dordrecht: Kluwer. É. Kiss, Katalin, 1994. Sentence structure and word order. In: Ferenc Kiefer and Katalin É. Kiss, eds. The syntactic structure of Hungarian. Syntax and Semantics 27. New York: Academic Press, 1–90. É. Kiss, Katalin, 2000. The Hungarian noun phrase is like the English noun phrase. In: Gábor Alberti and István Kenesei, eds. Approaches to Hungarian 7. Szeged: JATE Press, 121–149. É. Kiss, Katalin, 2002. The syntax of Hungarian. Cambridge: Cambridge University Press.

References



É. Kiss, Katalin, 2005a. The inverse agreement constraint in Hungarian: A relic of a Uralic– Siberian Sprachbund? In: Hans Broekhuis, Norbert Corver, Jan Koster, Riny Huybregts, and Ursula Kleinhenz, eds. Organizing grammar. Linguistic studies in honor of Henk van Riemsdijk. Berlin: Mouton de Gruyter, 108–116. É. Kiss, Katalin, 2005b. Az ómagyar igeidőrendszer morfoszintaxisáról [On the morphosyntax of the Old Hungarian tense system]. Magyar Nyelv 101: 420–435. É. Kiss, Katalin, 2006a. From the grammaticalization of viewpoint aspect to the grammaticalization of situation aspect. In: Katalin É. Kiss, ed. Event structure and the left periphery. Dordrecht: Springer, 129–158. É. Kiss, Katalin, 2006b. Focussing as predication. In: Valeria Molnár and Susanne Winkler, eds. The architecture of focus. Berlin: Mouton de Gruyter, 169–193. É. Kiss, Katalin, 2006c. Jól megoldottuk? Rosszul oldottuk meg? Az összefoglaló és a kirekesztő kifejezést tartalmazó mondatok szórendjének magyarázata [Have we solved it well? Have we solved it badly? An explanation for the word order in clauses containing summarizing and exhaustive expressions]. Magyar Nyelv 102.4: 442–459. É. Kiss, Katalin, 2008a. Free word order, (non-)configurationality. and phases. Linguistic Inquiry 39: 441–475. É. Kiss, Katalin, 2008b. Deriving the properties of structural focus. In: The 18th international congress of linguists, July 21–26, 2008, Seoul, abstracts volume 2. Seoul: Korea University, 223– 224. É. Kiss, Katalin, 2009a. Negative quantifiers in Hungarian. In: Marcel den Dikken and Robert M. Vago, eds. Approaches to Hungarian 11. Papers from the 2007 New York conference. Amsterdam: John Benjamins, 93–118. É. Kiss, Katalin, 2009b. Nekem el kell menni/el kell mennem/el kell, hogy menjek/el kell menjek/el kell menni [On different ways of expressing ‘I must go’]. In: Katalin É. Kiss and Attila Hegedűs, eds. Nyelvelmélet és Dialektológia. Piliscsaba: Pázmány Péter Katolikus Egyetem, 213–230. É. Kiss, Katalin, 2010. Structural focus and exhaustivity. In: Malte Zimmermann and Caroline Féry, eds. Information structure. Theoretical, typological and experimental perspectives. Oxford: Oxford University Press, 64–88. É. Kiss, Katalin, ed., 2013a. Magyar történeti mondattan [Hungarian diachronic syntax]. Budapest: Akadémiai Kiadó. É. Kiss, Katalin, 2013b. From Proto-Hungarian SOV to Old Hungarian Top Foc V X∗ . Diachronica 30: 202–231. É. Kiss, Katalin, 2013c. Az ótörök-ősmagyar kontaktus nyomai az ómagyar igeidőrendszerben és a birtokos szerkezetben [Traces of Old Turkic and Proto-Hungarian contact in the ProtoHungarian temporal system and possessive construction]. In: Klára Agyagási, Katalin É. Kiss, and Attila Hegedűs, eds. Nyelvelmélet és Kontaktológia 2. Piliscsaba: PPKE. É. Kiss, Katalin, 2013d. The Inverse Agreement Constraint in Uralic languages. Finno-Ugric Languages and Linguistics 2: 2–21. É. Kiss, Katalin, 2014. Ways of licensing Hungarian external possessors. Acta Linguistica Hungarica 61, 45–68. Ernst, Thomas, 2002. The syntax of adjuncts. Cambridge: Cambridge University Press.



References

Etxeberria, Urtzi, 2009. Contextually restricted quantification in Basque. In: Anastasia Giannakidou and Monika Rathert, eds. Quantification, definiteness, and nominalization, Oxford: Oxford University Press, 76–107. Evans, Nicholas, 2007. Insubordination and its uses. In: Irina Nikolaeva, ed. Finiteness. Oxford: Oxford University Press, 365–431. Farkas, Donka, 1997. Dependent indefinites. In: Francis Corblin, Danièle Godard, and JeanMarie Marandin, eds. Empirical issues in formal syntax and semantics. Bern: Peter Lang Publishers, 243–67. Ferraresi, Gisella, and Maria Goldbach, eds., 2008. Principles of syntactic reconstruction. Amsterdam: John Benjamins. Filchenko, A. Y., 2011. Prosody and pragmatics of Eastern Khanty narratives. Vestnik. Tomsk State Pedagogical University Bulletin 9: 139–145. Forgács, Tamás, 2004. Grammaticalisation and preverbs. Acta Linguistica Hungarica 51: 45–84. Fraurud, Kari, 2001. Possessive with extensive use: A source of definite articles? In: Irène Baron, Michael Herslund, and Finn Sørensen, eds. Dimensions of possession. Amsterdam: John Benjamins, 243–267. Galambos, Dezső, 1907. Tanulmányok a magyar relatívum mondattanáról [Studies on the syntax of Hungarian relatives]. Budapest: Athenaeum. Galves, Charlotte, and Pablo Faria, 2010. Tycho Brahe parsed corpus of historical Portuguese. . Gelderen, Elly van, 2004. Grammaticalization as economy. Linguistik Aktuell 71. Amsterdam: John Benjamins. Gelderen, Elly van, 2005. The CP and split CP cross-linguistically. WORD 55.3: 369–403. 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: Thórhallur Eythórsson, ed. Grammatical change and linguistic theory: The Rosendal papers. Linguistik Aktuell 113. Amsterdam: John Benjamins, 245–264. Gelderen, Elly van, 2009. Renewal in the left periphery: economy and the complementizer layer. Transactions of the Philological Society 107.2: 131–195. Gelderen, Elly van, 2011. The linguistic cycle. Language change and the language faculty. Oxford: Oxford University Press. George, Leland, and Jaklin Kornfilt, 1981. Finiteness and boundedness in Turkish. In: Frank Heny, ed. Binding and filtering. London: Croom Helm, 105–127. Gerőcs, Mátyás, 2011. Aspektus és igekötő. A mondat belső időszerkezete a kései ómagyarban [The internal temporal organization of the sentence in Old Hungarian]. In: Katalin É. Kiss and Attila Hegedűs, eds. Nyelvelmélet és diakrónia. Budapest–Piliscsaba: Szent István Társulat, 153–170. Giorgi, Alessandra, 2010. Towards a syntax of indexicality. Oxford: Oxford University Press. Giusti, Giuliana, 1994. Enclitic articles and double definiteness: A comparative analysis of nominal structure in Romance and Germanic. The Linguistic Review 11: 241–255. 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

References



studies in Italian syntax: essays offered to Lorenzo Renzi. North Holland New Linguistic Series 59. York: Elsevier, 157–171. Givón, Talmy, 1990. Syntax: a functional-typological introduction. Amsterdam: John Benjamins. Greenberg, Joseph H., 1978. How does a language acquire gender markers? In: Joseph H. Greenberg, Charles A. Ferguson, and Edith Moravcsik, eds. Universals of human language. Vol. 3. 47–82. Stanford: Stanford University Press. Gugán, Katalin, 2002. Syntactic synonymy: a case study. Acta Linguistica Hungarica 49: 23–49. Gugán, Katalin, 2013. Zigzagging in language history: negation and negative concord in Hungarian. Finno-Ugric Languages and Linguistics 1.1–2: 89–97. G. Varga, Györgyi, 1992. Névmások [Pronouns]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor. Morfematika. Budapest: Akadémiai Kiadó, 455–569. Haader, Lea, 1991. Az alárendelő mondatok: Az alanyi, állítmányi, tárgyi és határozói mellékmondatok [Subordinate clauses: Subject, predicate, object and adverbial subclauses]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana I.: A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 728–741. Haader, Lea, 1995. Az alárendelő mondatok: Az alanyi, állítmányi, tárgyi és határozói mellékmondatok [Subordinate clauses: Subject, predicate, object and adverbial subclauses]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/2.: A kései ómagyar kor: Mondattan. Szöveggrammatika. Budapest: Akadémiai Kiadó, 506–665. Haader, Lea, 2001. Mikrodiakrónia és változásvizsgálat (az összetett mondatokban) [Microdiachrony and the analysis of changes (in complex sentences)]. Magyar Nyelvőr 125: 354–371. Haader, Lea, 2005. A Müncheni emlék [The Munich language record]. Magyar Nyelv 101: 161–178. Hajdu, Péter, 1975. Finno-Ugrian languages and peoples, translated by G. F. Cushing. London: André Deutsch. Halácsy, Péter, András Kornai, and Csaba Oravecz, 2007. HunPos: an open source trigram tagger. In Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics: Companion Volume Proceedings of the Demo and Poster Sessions, Prague: Association for Computational Linguistics, 209–212. Hámori, Antónia, 1957. Mind [All]. Magyar Nyelv 53: 138–47. Hancock, Roeland and Thomas G. Bever, 2009. The study of syntactic cycles as an experimental science. In: Elly van Gelderen, ed. Cyclical change. Amsterdam: John Benjamins, 303–322. Haspelmath, Martin, 1995. Diachronic sources of all and every. In: Emmon Bach, Eloise Jelinek, Angelika Kratzer, and Barbara Hall Partee, eds. Quantification in natural languages. Dordrecht: Kluwer, 363–382. Haspelmath, Martin, Matthew S. Dryer, David Gil, and Bernard Comrie, eds. 2005. The world atlas of language structures. Oxford: Oxford University Press. Hawkins, John A., 1978. Definiteness and indefiniteness: a study in reference and grammaticality prediction. London: Croom Helm. Hawkins John A., 1991. On (in)definite articles: implicatures and ungrammaticality prediction. Journal of Linguistics 27: 405–422. Hawkins, John A., 2004. Efficiency and complexity in grammars. Oxford: Oxford University Press Hegedűs, Veronika, to appear. The grammaticalization of postpositions in Old Hungarian. In: Theresa Biberauer and George Walkden, eds. Syntax over time: lexical, morphological and information-structural interactions. Oxford: Oxford University Press.



References

Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer, 1991. Grammaticalization: A conceptual framework. Chicago: University of Chicago Press. Himmelmann, Nikolaus P., 1997. Deiktikon, Artikel, Nominalphrase: zur Emergenz syntaktische Struktur. Linguistische Arbeiten 362. Tübingen: Niemeyer. Himmelmann, Nikolaus P., 2001. Articles. In: Martin Haspelmath et al., eds. Language typology and language universals. Handbücher der Sprach und Kommunikationswissenschaft. Band 20.1. Berlin and New York: Walter de Gruyter, 831–841. Homer, Vincent, 2011. Polarity and modality. Ph.D. dissertation. Los Angeles: University of California. Hoop, Helen de, 1992. Case configuration and noun phrase interpretation. Ph.D. dissertation, University of Groningen. Hoppa, Enikő, 2012. “Mük csángóul beszélünk”. A moldvai magyar nyelvjárás [‘We are speaking in Csángó.’ The Moldavian dialect of Hungarian]. Pécs: Pro Pannonia Kiadói Alapítvány. Hornstein, Norbert, 1990. As time goes by. Cambridge, MA: MIT Press. Hornstein, Norbert, 1995. Logical form: From GB to Minimalism. Oxford: Blackwell. Horvath, Júlia, 1985. Focus in the theory of grammar and the syntax of Hungarian. Dordrecht: Foris. Horváth, László, 1991. Három vázlatos szinkrón metszet határozói igeneveink történetéből [On three stages in the history of adverbial participles]. Budapest: Akadémiai Kiadó. Horváth, László, 2003. Szószerkezet-történet [The history of the internal structure of words]. In: Jenő Kiss and Ferenc Pusztai, eds. Magyar nyelvtörténet. Budapest: Osiris, 430–482. Hyman, Larry, 1975. On the change from SOV to SVO: Evidence from Niger Congo. In: Charles Li, ed. Word order and word order change. Austin: University of Texas, 113–144. I. Gallasy, Magdolna, 1992. A névelők [The articles], in Benkő, Loránd, ed. A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor. Morfematika. Budapest: Akadémiai Kiadó, 716–770. Ivácsony, Zsuzsa, 2002–2003. Déli csángó [Southern Csángó]. Transcripts of interviews with speakers of the Southern Csángó dialect. Jackendoff, Ray S., 1983. Semantics and cognition. Cambridge, MA: MIT Press. Jäger, Agnes, 2008. History of German negation. Amsterdam: John Benjamins. Jäger, Agnes, 2012. ‘How’ to become a comparison particle. Talk delivered to: 14th Diachronic Generative Syntax Conference (DiGS 14), Lisbon, 4–6 July 2012. Jakab, László, 2002. A Jókai-kódex mint nyelvi emlék szótárszerű feldolgozásban [The Jókai Codex as a language record in a dictionary format]. Számítógépes Nyelvtörténeti Adattár 10. [Computational Database for Historical Linguistics 10]. Department of Hungarian Linguistics, University of Debrecen, Debrecen. Jakab, László, and Antal Kiss, 1994. A Guary-kódex ábécérendes adattára [The alphabetical database of the Guary Codex]. Számítógépes Nyelvtörténeti Adattár 6. [Computational Database for Historical Linguistics 6.]. Department of Hungarian Linguistics, University of Debrecen, Debrecen. Jakab, László, and Antal Kiss, 2001. A Festetics-kódex ábécérendes adattára [The alphabetical database of the Festetics Codex]. Számítógépes Nyelvtörténeti Adattár 9. [Computational Database for Historical Linguistics 9.]. Department of Hungarian Linguistics, University of Debrecen, Debrecen.

References



Jelinek, Eloise, and Richard Demers, 1997. Reduplication as a Quantifier in Salish. International Journal of American Linguistics. 63.3: 302–51. Juhász, Dezső, 1991a. A kötőszók [Conjunctions]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana I.: A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 476–500. Juhász, Dezső, 1991b. A módosítószók [Modifying particles]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana. I.: A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 501–513. Juhász, Dezső, 1992. A kötőszók [Conjunctions]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/1.: A kései ómagyar kor: Morfematika. Budapest: Akadémiai Kiadó, 772–814. Kálmán, Béla, 1966. Nyelvjárásaink [Our dialects]. Budapest: Tankönyvkiadó. Kántor, Gergely, 2008. On Hungarian relative operators. The Even Yearbook 8: 1–12. Károly, Sándor, 1954. Ragtalan tárgy határozói igenév mellett [Unmarked objects of adverbial participles]. Magyar Nyelv 50: 43–50. Károly, Sándor, 1956. Igenévrendszerünk a kódexirodalom első szakaszában [The system of nonfinite forms in the early codices]. Budapest: Akadémiai Kiadó. Kayne, Richard, 1994. The Antisymmetry of syntax. Cambridge, MA: MIT Press. Kenesei, István, 1992. Az alárendelt mondatok szerkezete [The structure of subordinate clauses]. In Ferenc Kiefer, ed. Strukturális magyar nyelvtan I.: Mondattan. Budapest: Akadémiai Kiadó, 529–713. Kenesei, István, 1994. Subordinate clauses. In: Ferenc Kiefer and Katalin É. Kiss, eds. The syntactic structure of Hungarian. Syntax and Semantics 27. San Diego: Academic Press, 275–354. Kenesei, István, 2005. Non-finite clauses in derived nominals. In Christopher Piñón and Szilárd Szentgyörgyi, eds. Approaches to Hungarian 9: papers from the Düsseldorf Conference. Budapest: Akadémiai Kiadó, 159–186. Kennedy, Christopher and Jason Merchant, 2000. Attributive Comparative Deletion. Natural Language & Linguistic Theory 18: 89–146. Király, Lajos, 1991. A mai magyar nyelvjárások [Modern Hungarian dialects]. In Anna A. Jászó, ed. A magyar nyelv könyve. Budapest: Trezor, 521–558. Kiss, Jenő, 2006. Nyelvjárások, regionális nyelvváltozatok [Dialects, regional varieties]. In: Kiefer, Ferenc, ed. Magyar nyelv. Budapest: Akadémiai, 517–548. Klemm, Antal, 1928. Magyar történeti mondattan [Hungarian Diachronic Syntax]. Budapest: MTA. Klumpp, Gerson, 2009. Variation in Komi object marking. In: Andreas Dufter, Jürg Fleischer, and Guido Seiler, eds. Trends in linguistics: describing and modeling variation in grammar. Berlin: Mouton de Gruyter, 325–359. Kniezsa, István, 1952. Helyesírásunk története a könyvnyomtatás koráig [The history of our orthography until the age of the printing press]. Budapest: Akadémiai Kiadó. Komlósy, András, 1992. Régensek és vonzatok [Predicates and Arguments]. In: Ferenc Kiefer, ed. Strukturális magyar nyelvtan 1. Mondattan. Budapest: Akadémiai Kiadó, 299–529. Koopman, Hilda, 2010. Prepositions, postpositions, circumpositions, and particles: The structure of Dutch PPs. In: Guglielmo Cinque and Luigi Rizzi, eds. Mapping Spatial PPs. Oxford and New York: Oxford University Press, 26–73.



References

Koptjevskaja Tamm, Maria, 1994. Finiteness. In: R. E. Asher, ed. The encyclopedia of language and linguistics. Oxford: Pergamon Press, 1245–1248. Koster, Jan, 1994. Predicate incorporation and the word order of Dutch. In: Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini, eds. Paths towards Universal Grammar: studies in honor of Richard S. Kayne. Oxford: Oxford University Press, 255–276. Kratzer, Angelika, 2005. Indefinites and the operators they depend on: From Japanese to Salish. In: Gregory Carlson and Francis Jeffrey Pelletier, eds. Reference and quantification: The Partee effect. Palo Alto: CSLI Publications, 113–142. Kratzer, Angelika, and Junko Shimoyama, 2002. Indeterminate pronouns: The view from Japanese. In: Otsu, Yukio, ed. Proceedings of the Third Tokyo Psycholinguistics Conference. Tokyo: Hituzi Shiboyo, 1–25. Kroch, Anthony S., 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Kroch, Anthony S., and Ann Taylor, 2000. The Penn-Helsinki parsed corpus of Middle English (PPCME2). CD-ROM. Kulonen, Ulla-Maija, 1989. The Passive in Ob-Ugrian. Helsinki: Finno–Ugrian Society. Kuznecova, Ariadna, Evgenij Xelimskij, and Elena Gruskina, 1982. Ocherki po sel’kupskomu jazyku [Studies in Selkup]. Moscow: MGU. Künnap, Ago, 2004. About the non-personal definite function of the Uralic 3rd person possessive suffix. Linguistica Uralica XL: 1–4. Laury, Ritva, 1997. Demonstratives in interaction: The emergence of a definite article in Finnish. Studies in Discourse and Grammar 7. Amsterdam: John Benjamins. Ledgeway, Adam, 2007. Diachrony of finiteness: subordination in the dialects of southern Italy. In: Irina A. Nikolaeva, ed. Finiteness: Theoretical and empirical foundations. Oxford: Oxford University Press, 335–365. Lightfoot, David, 1991. How to set parameters: Arguments from language change. Cambridge, MA: MIT Press/Bradford Books. Lightfoot, David, 2002. Myths and the prehistory of grammars. Journal of Linguistics 38: 113–136. Lyons, Christopher, 1999. Definiteness. Cambridge: Cambridge University Press. Marácz, László, 1989. Asymmetries in Hungarian. Ph.D. dissertation, Groningen. Marcantonio, Angela, 1985. On the definite vs. indefinite conjugation in Hungarian: a typological and diachronic analysis. Acta Linguistica Hungarica 35: 267–298. Márkus, Andrea, 2009. Egy sokak vitatta szerkezetről [On a structure debated by many]. In: Balázs Sinkovics, ed. Lingdok 8. Szeged: Szegedi Tudományegyetem, 27–46. Matthewson, Lisa, 2001. Quantification and the nature of crosslinguistic variation. Natural Language Semantics 9: 145–189. McEnery, Tony, 2003. Corpus linguistics. In: Ruslan Mitkov, ed. The Oxford handbook of computational linguistics. New York: Oxford University Press. 448–463. McEnery, Tony, and Andrew Hardie, 2003. Lancaster newsbooks corpus. Miller, Gary D., 2002. Nonfinite structures in theory and change. Oxford: Oxford University Press. Milsark, Gary, 1974. Existential sentences in English. Ph.D. dissertation. Cambridge, MA: MIT. Milsark, Gary, 1977. Towards an explanation of certain peculiarities of the existential construction in English. Linguistic Analysis 5.3: 1–29.

References



Nádasdi, Péter, 2010. Egy a nominális elemen személyjelölt jelzői tagmondat régen és ma– esettanulmány [A case study of an adjectival clause with agreement on the nominal element, nowadays and earlier], Magyar Nyelvőr 134: 471–483. Nádasdi, Péter, 2013. Az univerzális grammatika szempontjából lehet-e a -ván, -vén-nek alanya? [Can -ván, -vén participles have a nominative subject from the viewpoint of Universal Grammar?]. In Zsuzsanna Gécseg, ed. Lingdok 12. Szeged: Szegedi Tudományegyetem Nyelvtudományi Doktori Iskola, 157–176. Németh, Sándor, 1913. A domokosi nyelvjárás [The dialect of Domokos], Magyar Nyelvőr 42: 65–71. Nikolaeva, Irina, 1999. Ostyak: languages of the world: materials 305. München: LINCOM Europa. Nikolaeva, Irina, 2001. Secondary topic as a relation in information structure. Linguistics 39: 1–49. Nikolaeva, Irina, 2003. Possessive affixes as markers of information structuring: Evidence from Uralic. In: Pirkko Suihkonen and Bernard Comrie, eds. International symposium on deictic systems and quantification in languages spoken in Europe and North and Central Asia: Collection of papers. Izhevsk: Udmurt State University; Leipzig: Max Planck Institute of Evolutionary Anthropology, 130–145. Nouwen, Rick, 2003. Plural pronominal anaphora in context: Dynamic aspects of quantification. Ph.D. dissertation. LOT: Utrecht University. Olsvay, Csaba, 2006. Negative universal quantifiers in Hungarian. Lingua 116: 245–270. Pais, Dezső, 1959. Az igekötők mivoltához és keletkezéséhez [On the nature and emergence of particles], Magyar Nyelv 55: 181–184. Pancheva, Roumyana, 2008. Head-directionality of TP in Old Church Slavonic. In: Andrei Antonenko, John F. Bailyn, and Christina Y. Bethin, eds. Formal approaches to Slavic linguistics: The Stony Brook meeting, 2007, Ann Arbor: Michigan Slavic Publications, 313–332. Pápay, József, 1906–8. Északi-osztják nyelvtanulmányok [Northern Ostyak linguistic studies]. Nyelvtudományi Közlemények 36: 345–98; 37: 52–79, 164–95, 258–75; 38: 111–50, 313–29. Partee, Barbara Hall, 1986. Noun Phrase interpretation and type-shifting principles. In: Jeroen Groenendijk, ed. Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers. Dordrecht: Foris, 361–383. Partee, Barbara Hall, and Mats Rooth, 1983. Generalized conjunction and type-shifting principles. In: Rainer Bäuerle, Christoph Schwarze, and Arnim von Stechow, eds. Meaning, use and interpretation of language. Berlin and New York: Mouton de Gruyter, 361–383. Peredy, Márta, 2009. A stochastic account for the variation in Hungarian object agreement. Manuscript. Budapest: Research Institute for Linguistics of the Hungarian Academy. Pesetsky, David, and Esther Torrego, 2001. T-to-C movement: Causes and consequences. In: Michael Kenstowicz, ed. Ken Hale: A life in language. Cambridge, MA: MIT Press, 355–426. Petersen, Ulrik, 2004. Emdros: a text database engine for analyzed or annotated text. In: COLING 2004: Proceedings of the 20th International Conference on Computational Linguistics, Geneva, 1190–1193. Piñón, Christopher, 1992. Heads in the Hungarian focus field. In: István Kenesei and Csaba Pléh, eds. Approaches to Hungarian 4. Szeged: JATE, 99–121.



References

Polo, Chiara, 2005. Latin word order in generative perspective: An explanatory proposal within the sentence domain. In: Katalin É. Kiss, ed. Universal Grammar in the reconstruction of ancient languages. Berlin: Mouton de Gruyter, 373–428. Prószéky, Gábor, and Balázs Kis, 1999. A unification-based approach to morphosyntactic parsing of agglutinative and other (highly) inflectional languages. In: Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics, College Park, MD, 261–268. Rácz, Endre, 1995. Nem mondatrészkifejtő mellékmondatok [Non-argument subordinate clauses]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana II/2.: A kései ómagyar kor: Mondattan. Szöveggrammatika. Budapest: Akadémiai Kiadó, 694–718. Radics, Katalin, 1992. Fossilized gerunds with possessive endings in Hungarian. In: István Kenesei and Csaba Pléh, eds. Approaches to Hungarian 4. Szeged: JATE, 283–300. Rákosi, György, 2010. On snakes and locative binding in Hungarian. In: Miriam Butt and Tracy Holloway King, eds. The proceedings of the LFG10 conference. Stanford: CSLI Publications, 395–415. Rayson, Paul, Dawn Archer, Alistair Baron, Jonathan Culpeper, and Nicholas Smith, 2007. Tagging the Bard: Evaluating the accuracy of a modern POS tagger on Early Modern English corpora. In: Proceedings of the Corpus Linguistics Conference (CL2007), UK. University of Birmingham. Rédei, Károly, 1978. Syrjänische Chrestomathie mit Grammatik und Glossar. Studia Uralica 1. Verband der wissenschaftlichen Gesellschaften Österreichs, Vienna. Rédei, Károly, ed., 1986–1991. Uralisches etymologisches Wörterbuch. Budapest: Akadémiai Kiadó. Riemsdijk, Henk van, 1990. Functional prepositions. In: Harm Pinkster, and Inge Genee, eds. Unity in diversity: Papers presented to Simon C. Dik on his 50th birthday. Dordrecht: Foris, 229–241. Riemsdijk, Henk van, and Riny Huijbregts, 2007. Location and locality. In: Simin Karimi, Vida Samiian, and Wendy K. Wilkins, eds. Phrasal and clausal architecture: syntactic derivation and interpretation, in honor of Joseph E. Emonds. Amsterdam: John Benajmins, 339–364. Riese, Timothy, 2001. Vogul (Mansi). München: LINCOM Publishers. Rizzi, Luigi, 1997. The fine structure of the left periphery. In: Liliane Haegeman, ed. Elements of grammar. Dordrecht: Kluwer, 281–337. 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. Amsterdam: Elsevier North-Holland, 287–296. Rizzi, Luigi, 2004. Locality in the left periphery. In: Adriana Belletti, ed. Structures and beyond: The cartography of syntactic structures, volume 3. Oxford: Oxford University Press, 223–251. Roberts, Craige, 2003. Uniqueness in definite noun phrases. Linguistics and Philosophy 26. 287–350. Roberts, Ian G., 2005. Principles and Parameters in a VSO language. Oxford: Oxford University Press. Roberts, Ian G., and Anna Rousssou, 2003. Syntactic change: A Minimalist approach to grammaticalization. Cambridge: Cambridge University Press.

References



Róna-Tas, András, 1999. Hungarians and Europe in the Early Middle Ages. Budapest: CEU Press. Róna-Tas, András, 2007. The Khazars and the Magyars. In: Peter B. Golden, Haggai BenShammai, and András Róna-Tas, eds. The World of the Khazars. Leiden and Boston: Brill, 269–278. Róna-Tas, András, Árpád Berta, and László Károly, 2011. West Old Turkic. Turkic loanwords in Hungarian. I–II. Wiesbaden: Harrassowitz. Salvi, Giampaolo, and Laura Vanelli, 2004. Nuova grammatica italiana. Bologna: il Mulino. Sampanis, Konstantinos, 2011. Parallel and “Antagonistic” complementation structures in the history of the Greek language. Wiener Linguistische Gazette 75: 112–125. Sárik, Pál, 1998. A határozói igenevek néhány problémája [Some problems surrounding adverbial participles]. Magyar Nyelv 94: 423–435. Sárosi, Zsófia, 2003. Morfématörténet. Az ősmagyar kor [The history of morphology: The ProtoHungarian period]. In: Jenő Kiss and Ferenc Pusztai, eds. Magyar nyelvtörténet. Budapest: Osiris, 129–172. Sebestyén, Árpád, 2002. A névutók állománya és rendszere a Jókai-kódexben (1372 u.) [The stock and system of postpositions in the Jókai Codex (after 1372)] Debrecen: Debreceni Egyetem Magyar Nyelvtudományi Intézet. Seuren, Pieter, 1973. The comparative. In: Ferenc Kiefer and Nicolas Ruwet, eds., Generative Grammar in Europe. Dordrecht: Reidel, 528–564. Sigurðsson, H. Ármann, 2004. The syntax of person, tense, and speech features. Italian Journal of Linguistics 16: 219–151. Simonyi, Zsigmond, 1914. A jelzők mondattana. Nyelvtörténeti tanulmány [The syntax of attributes: A diachronic study]. Budapest, MTA. Sinor, Denis, ed. 1988. The Uralic languages: Description, history and foreign influences. Leiden: Brill. Skribnik, Elena, 2001. Pragmatic structuring in Northern Mansi. In: Tönu Seidenthal, ed. Congressus Nonus Internationalis Fenno-ugristarum. Pars IV. Dissertationes sectionum: Linguistica III. Tartu: Tartu University. Sommerer, Lotte, 2011. Old English se: From demonstrative to article. A usage-based study of nominal determination and category emergence. Ph.D. dissertation, Universität Wien. Sportiche, Dominique, 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19: 425–49. Steinitz, Wolfgang, 1950. Ostjakische Grammatik und Chrestomatie mit Wörterverzeichnis. Leipzig: O. Harrassowitz. Sundaresan, Sandhya, 2010. A reductionist treatment of control and anaphora. Manuscript, University of Tromsø. Sundaresan, Sandhya, and Thomas McFadden, 2009. Subject distribution in Tamil and other languages: Selection vs. case, Journal of South Asian Linguistics 2: 5–34. Sundaresan, Sandhya, and Thomas McFadden, to appear. Nominative case is independent of finiteness and agreement. In: Papers from BCGL 5: Case at the interfaces. Surányi, Balázs, 2002. Multiple operator movements in Hungarian. Ph.D. dissertation. Utrecht: LOT. Surányi, Balázs, 2009. Verbal particles inside and outside vP. Acta Linguistica Hungarica 56: 201–249.



References

Surányi, Balázs, 2011. An interface account of identificational focus movement. In: Tibor Laczkó and Catherine Ringen, eds. Approaches to Hungarian 12. Amsterdam: John Benjamins, 163–208. Surányi, Balázs, and Orsolya Tánczos, 2011. Iterated syntax and focus in Udmurt. Talk presented at the 10th International Conference on the Structure of Hungarian, Lund, July 12, 2011. Svenonius, Peter, 2004. Spatial prepositions in English. Manuscript. University of Tromsø: . Svenonius, Peter, 2006. The emergence of axial parts. Nordlyd: Tromsø Working Papers in Linguistics 33/1 (special issue on adpositions, ed. by Peter Svenonius and Marina Pantcheva), CASTL, Tromsø, 49–77. Svenonius, Peter, 2010. Spatial P in English. In: Guglielmo Cinque and Luigi Rizzi, eds. Mapping spatial PPs: The cartography of syntactic structures, volume 6. Oxford and New York: Oxford University Press, 127–160. Szabolcsi, Anna, 1981. The possessive construction in Hungarian. Acta Linguistica Hungarica 31: 261–289. Szabolcsi, Anna, 1983. The possessor that ran away from home. The Linguistic Review 3: 89–105. Szabolcsi, Anna, 1994a. All quantifiers are not equal: The case of focus. Acta Linguistica Hungarica 42: 171–187. Szabolcsi, Anna, 1994b. The noun phrase. In: Ferenc Kiefer and Katalin É. Kiss, eds. The syntactic structure of Hungarian. Syntax and Semantics 27. San Diego: Academic Press, 179–274. Szabolcsi, Anna, 1997. Strategies for scope taking. In: Anna Szabolcsi, ed. Ways of scope taking. Dordrecht: Kluwer, 109–155. Szabolcsi, Anna, 2009. Overt nominative subjects in infinitival complements in Hungarian. In: Robert M. Vago and Marcel den Dikken, eds. Approaches to Hungarian 11. Amsterdam: John Benjamins, 251–276. Szabolcsi, Anna, 2010. Quantification. Cambridge: Cambridge University Press. Szabolcsi, Anna, and Michael Brody, 2003. Overt scope in Hungarian. Syntax 6.1: 19–51. Szily, Kálmán, 1911. Tamás és Bálint: a Biblia első magyar fordítói [Tamás and Bálint: The first Hungarian translators of the Bible]. Magyar Nyelv 7: 441–445. Tánczos, Orsolya, 2013. A mixed-headed CP layer in Udmurt from a historical perspective. Manuscript. Budapest: Research Institute for Linguistics of the Hungarian Academy of Sciences. Tánczos, Orsolya, 2013. ‘Hogy . . . hogy? kettős kötőszók az udmurtban [That . . . that? Double complementizers in Udmurt]. In: Katalin É. Kiss and Attila Hegedűs, eds. Nyelvemlélet és Kontaktológia 2. PPKE BTK: Piliscsaba, 279–293. Terzi, Arhonto, 1992. Control in finite clauses: A study in the inflectional heads of the Balkan languages. Doctoral Dissertation, City University of New York. Terzi, Arhonto, 1997. PRO and null case in finite clauses. The Linguistic Review 14: 335–360. Thomas, Peter Wynn, D. Mark Smith, and Diana Luft, 2007. Rhyddiaith Gymraeg 1350–1425, Cardiff University. Tóth, Ildikó, 2000a. -Va and -ván participles in Hungarian. In Gábor Alberti and István Kenesei, eds. Approaches to Hungarian 7: Papers from the Pécs Conference. Szeged: JATE, 239–256. Tóth, Ildikó, 2000b. Inflected infinitives in Hungarian. Doctoral dissertation, University of Tilburg.

References



Tóth, Ildikó, 2011. A ragozott főnévi igenevek a kései ómagyar korban [Inflected infinitives in Late Old Hungarian]. In: Marianne Bakró-Nagy and Forgács Tamás, eds. A nyelvtörténeti kutatások legújabb eredményei VI. Szeged: SZTE, 249–265. Ürögdi, Barbara, 2012. Operator movements in embedded clauses. Ph.D. dissertation. Budapest: Eötvös Loránd University. Varga, Mónika, 2012. A határozói igenevek lehetséges állítmányi szerepéről boszorkányperek szövegeiben [On the possible predicative use of adverbial participles in transcripts of witch trials]. Első Század Online 11: 1–26. Vate, Marleen Susanne van de, 2011. Tense, aspect, and modality in a radical creole: The case of Saamáka. Doctoral dissertation, University of Tromsø. Velcsov, Mártonné, 1957. Az -n igei személyrag eredetéhez [On the origin of the -n person inflection of verbs]. Néprajz és Nyelvtudomány 1: 105–110. Velcsov, Mártonné, 1981. A határozói igeneveknek egy sajátos szerepe nyelvemlékeinkben [A special use of adverbial participles in our linguistic records]. Magyar Nyelv 77: 308–315. Vilkuna, Maria, 1998. Word order in European Uralic. In: Anna Siewierska, ed. Constituent order in the languages of Europe. Berlin: Mouton de Gruyter, 173–234. Vincent, Nigel, 1998. On the grammar of inflected non-finite forms (with special reference to Old Neapolitan). In: Iorn Korzen and Michael Herslun, eds. Clause combining and text structure. Copenhagen: Samfundslitteretur, 135–158. Virtanen, Susanna, 2012. Variation in three-participant constructions in Eastern Mansi. Linguistica Uralica 48: 120–130. Volf, György, 1874. Nyelvemléktár I. [Repository of language records I.]. Budapest: Magyar Tudományos Akadémia Kiadó Hivatala. Volk, Martin, Torsten Marek, and Rico Sennrich, 2010. Reducing OCR errors by combining two OCR systems. In: Proceedings of the ECAI 2010 Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities (LaTeCH 2010), Lisbon. Faculty of Science, University of Lisbon. Watanabe, Akira, 2009. A parametric shift in the D-system in Early Middle English: relativization, articles, adjectival inflection, and indeterminates. In: Paola Crisma and Giuseppe Longobardi, eds. Historical syntax and linguistic theory. Oxford: Oxford University Press, 321–74. Waters, Cathleen, 2009. The preposition cycle in English. In: Elly van Gelderen, ed. Cyclical change. Amsterdam: John Benjamins, 285–300. Weinreich, Uriel, William Labov, and Marvin Herzog, 1968. Empirical foundations for a theory of language change. In: Winfred Lehmann and Yakov Malkiel, eds. Directions for historical linguistics. Austin: University of Texas Press, 95–189. Westerståhl, Dag, 1985. Determiners and context sets. In: Johan van Benthem, and Alice G. B. ter Meulen, eds. Generalized quantifiers in natural language. Dordrecht: Foris, 45–72. Winter, Yoad, 2001. Flexibility principles in Boolean semantics: Coordination, plurality and scope in natural language. Cambridge, MA: MIT Press. Wu, Tong, 2011. The syntax of prenominal relative clauses: A typological study. Linguistic Typology 15: 569–623. Zaicz, Gábor, 1998. Mordva. In: Daniel Abondolo, ed. The Uralic languages. London and New York: Routledge, 184–218.



References

Zwart, Jan-Wouter, 1993. Dutch syntax: A minimalist approach. Ph.D. dissertation. University of Groningen. Zwarts, Joost, 2007. Vectors as relative positions: A compositional semantics of modified PPs. Journal of Semantics 14: 57–86. Zsilinszky, Éva, 1991. A névutók [Postpositions]. In: Loránd Benkő, ed. A magyar nyelv történeti nyelvtana I. A korai ómagyar kor és előzményei. Budapest: Akadémiai Kiadó, 442–460.

Index adjectival participle -ó/ő adjectival participle 151, 158, 164, 165, 170, 177, 178, 180 past participle see -t adjectival participle with a coreferential internal argument present participle see -ó/ő adjectival participle -t adjectival participle with a coreferential internal argument 153, 158, 170, 178–80 -t adjectival participle with a coreferential object 154, 158, 170, 172–4, 180 -t adjectival participle with a coreferential possessor 153, 154, 158, 163, 176, 178, 180 adjunction 5, 57, 73–5, 82, 101, 102, 116, 140, 200, 206 adjunct 40–3, 74–5, 101–2, 139–40, 164, 175, 218 Adpositional Phrase 122 adverb 5, 6, 10, 45, 46, 49, 53, 55, 68–70, 84, 87–9, 92–7, 101, 102, 117, 118, 120–3, 129, 131, 132, 141, 155, 160, 162, 177, 229, 230 agent-oriented adverb 92–4 locative adverb 68, 131–2 manner adverb 94 speaker-oriented (speaker adverb) 92–5, 117–18 adverbial participle inflected adverbial participle 174 -t adverbial participle 157–60, 170, 173, 174, 177, 178, 180 -va/ve adverbial participle 156–60, 162, 163, 165, 166, 169, 170, 173, 174, 177–80 val/vel adverbial participle 156–60, 177, 180 -ván/vén adverbial participle 18, 19, 156, 158, 162, 164–71, 173, 174, 177–80 as sole predicate of a main clause 166 coordinated with a finite clause 165

adverbial suffix see oblique case agreement 4–5, 9–10, 12, 14, 19–25, 27, 30, 39–40, 47, 55, 76, 78, 80–1, 90–1, 134–5, 149–50, 152, 154–6, 158, 160, 167, 169–70, 173–6, 179–80 object–verb agreement 10, 19–23 ambiguity 58, 61, 77, 80, 88, 96, 102, 106, 143, 167 analogy 167, 168 anaphor, anaphora, anaphoric 58, 61–2, 64, 70, 107, 111, 115, 119–20, 136–8 associative-anaphoric use 64 Ancient Greek 179 -and/end suffix 151, 152 animacy hierarchy 21, 22 antecedent 112, 113, 218 anti-agreement (of infinitives) 149, 175, 176 appositive 5, 68, 79, 109, 123, 140–1, 146 article 5, 25, 56–69, 71–82, 85, 89, 97–8, 101, 102, 110, 234–5 article, identification of 58, 61–4 definite article, emergence of 5, 58–60, 67, 81 definite article, spreading 5, 65–7, 81 aspect viewpoint aspect 14, 35–7 situation aspect 14, 37 Axial Part 133, 138, 146 base-generation 31, 41, 53, 67, 80, 143, 186, 187, 192, 200–2, 217, 222 Bible translation 11, 17–18, 64, 220, 230, 232, 235 binding 20, 83, 107–15, 118–21, 136, 138 unselective binding 120 character encoding 227–8 clauses of reason 197–9, 205, 206



Index

clitic 16, 17, 59, 210–14 cliticization 16, 17, 210–14 comparative clause 89, 191, 195–7, 199, 203–17, 222 complementizer 4, 7, 11, 16, 17, 25, 26, 44, 53–5, 181–222 clause-final 11, 16, 17, 25 interrogative 17, 25 complementizer combinations 181–7, 190, 195, 197, 199–217, 221 hahogy ‘if that’ 184, 185, 200, 202, 206–8 hamint ‘if as’ 184, 185, 200, 202, 208, 209 hogyha ‘that if ’ 184, 185, 200, 202, 203, 206–8, 215–17 hogyhamint ‘that if as’ 185, 200, 215–17 hogymert ‘that because’ 184, 185, 202, 205, 206 hogymint ‘that than’ 184, 185, 202–5, 213 hogynem ‘that not’ 209–14 hogynemmint ‘that not than’ 185, 209, 212–14 hogysem ‘that neither’ 209–14 hogysemmint ‘that neither than’ 185, 209, 212–14 merthogy ‘because that’ 184, 185, 202, 203, 206 mintha ‘as if ’ 184, 185, 202, 203, 208, 209 minthogy ‘than that’ 184, 185, 200, 202–5, 213 minthogyha ‘as that if ’ 85, 200, 215–17 mintsemhogy ‘than neither that’ 185, 209, 213–15 complex tenses 14–16, 35–7 conditional clause 15, 116, 192–5, 206–9, 216, 217 conditional comparative 207–9, 216, 217 consecutive meaning 188, 220 corpus 3, 7, 57, 61, 64–6, 71, 75–7, 189, 224–36 corpus query 65, 71, 227, 231–5 Old Hungarian Corpus 7, 57, 65–6, 71, 76, 224–5 CP-domain 17, 148–50, 182–223

Csángó dialect 89, 160, 179 cycle 56, 68, 81, 144, 146, 185 definite article see article definiteness 5, 23, 25, 56–9, 64, 65, 72, 77, 81, 91, 98, 101, 102, 110 definiteness restriction 102, 110 degree expression 188, 195 deictic 61, 62, 69, 72, 132, 141 demonstrative 5, 6, 56–82, 90 distal demonstrative 58, 67, 72, 77 proximal demonstrative 68, 77 determination 59–61, 69, 74 determiner 5, 6, 57, 58, 60–7, 69, 71–4, 76–84, 87–90, 92, 95–7, 99, 102, 106, 108–10, 112, 115, 116, 120, 121, 234 determiner doubling 5, 57, 69, 72, 74, 76–8, 81, 82 determiner quantification, D-quantification 6, 84, 109, 112, 115, 120, 121 determiner quantifier 67, 79, 109, 112, 115, 120 D-head 68, 69, 72, 97 differential object–verb agreement 10, 19–23 differential object marking 20, 23 digitization 3, 66, 73, 75–7, 224, 225, 228, 235 discourse 5, 17, 19, 20, 25, 26, 54, 55, 57, 59, 62, 92, 103, 107–11, 115, 120, 181 distributivity 6, 83, 84, 93, 103–6, 110, 120, 121 distributivity operator 6, 83, 84, 93, 103, 110–13, 115, 117, 121 doubly filled projection 72, 75, 222 DP 5, 6, 38, 56, 57, 67, 68, 72–5, 78–85, 87–92, 94–104, 112, 115–21, 133–6, 138–44, 150, 151, 153, 157, 159, 163, 170–2, 177, 191 DP-cycle 5, 56, 68, 81 D-position 67, 68, 79 −e 16, 17, 181 economy, economy principle 67, 68, 72, 74, 139, 149, 182, 185, 186, 223 embedded imperative 187, 188, 191, 192

Index



embedded wh-question 43, 44, 188, 191, 194, 195 embedded yes-no question 189, 191, 193–5 English 44, 58, 83, 86, 88, 89, 92, 95, 112, 115, 137, 179, 185, 191, 192, 194, 206, 208–10, 217, 219, 222, 224, 227 Ewenki 179 exhausitivity 28, 41, 44, 46

inherently unique nouns 57, 62, 65 intonation 58, 67, 94 inverse agreement constraint 21–5 Italian 198, 210

Finnish 11, 59 focus 20, 21, 23–8, 31, 32, 34, 39–47, 50, 176 information focus 10, 41 focus movement 10, 46 focus operator 10, 41 Force 112, 182, 183, 185, 186, 190, 192, 208 functional split 58, 184, 186, 192, 195, 197, 205

larger situation use 58, 65 Late Merge Principle 68, 186, 219 law of growing constituents 32 left edge 10, 50, 97, 99, 193 left periphery 5, 7, 9, 16, 17, 31, 32, 34, 49, 50, 53–6, 73–5, 81–3, 122, 144, 148, 150, 173, 179, 182, 183, 193, 199, 202, 206, 207, 213, 215, 217, 223 Linear Correspondence Axiom 200 linguistic annotation 3, 71, 224, 228–9, 231–2, 234, 236 linguistic fossils 4, 9, 11, 23, 51, 54, 176

general (finite) subordinator 44, 187, 190 generic 37, 66 German 31, 36, 47, 48, 143, 194, 196, 197, 205, 209, 219 gerund 18, 39, 150, 155, 156, 158, 160–2, 166, 170, 173, 177, 179, 180 grammaticalization 4, 5, 7, 12, 24, 25, 32, 37, 43, 55–9, 61, 64, 67–9, 75, 122–5, 128, 129, 132–6, 138–46, 149, 182–5, 187, 192, 195–203, 205, 206, 208, 214–17, 219, 222, 223 ha ‘if ’ 113, 184–6, 189, 192–5, 199, 200, 202, 203, 206–9, 215–17, 219–22 Head Preference Principle 67, 186 head-final 2, 4, 5, 7, 9–11, 14, 16, 17, 25, 30, 54, 122, 123, 132, 142–4, 147, 148, 176, 177, 181, 182, 193 head-initial 2, 4, 5, 7, 9, 10, 17, 25, 30, 132, 143, 181, 182, 193, 199, 223 hogy ‘that’ 44, 184–97, 199, 200, 202–17, 219–22 identity marker 70 incorporation 5, 31, 37, 68, 75, 81 infinitive 149–1, 158, 163, 164, 165, 170, 174–6, 178–80, 191 inflected infinitive 175, 176

Karachay-Balkar 158 Khanty (Ostyak) 1, 2, 9, 11, 16, 17, 20, 23–5, 30, 34, 39–41, 47, 60, 154, 158, 179

Mansi (Vogul) 1, 2, 4, 9, 11, 16, 17, 20, 21, 23, 40, 47, 54, 55, 60, 154, 155 mert ‘because’ 184–6, 190, 191, 197–9, 202, 203, 205, 206, 221, 222 metadata 224, 232 Middle Hungarian period 2, 3, 5, 37, 38, 54, 57, 66, 72–7, 79, 80, 82, 138, 142, 159, 198, 203 miért ‘why’ 43, 184, 197, 205 miként ‘how’ 184, 195 miképpen ‘how’ 42, 184, 195 mint ‘than/as’ 184–6, 189, 195–7, 199, 200, 202–5, 208, 209, 211–17, 221, 222 Mirror Principle 200 Modern Greek 179 Mordvin 11, 14, 59 morphological analysis 225, 228, 230–4 morphosyntactic disambiguationv 225, 228, 230–1 movement 3, 5, 6, 10, 15, 20, 24, 28, 29, 31, 34, 38, 41–3, 46, 49, 50, 53–5, 68, 85, 95, 116–18, 120, 121, 127, 143, 144, 182, 186, 202, 206, 211, 215, 216



Index

negation 34, 46–53, 112, 113, 115, 118, 119, 155, 177, 209 negative concord 50, 51 negative element 50, 204, 209–15 negative pronouns 19, 50, 51 nemde 17, 181, 194 non-finite adverbial clauses 17 non-finite clause 4, 7, 11, 12, 14, 17–19, 25, 27, 51, 110, 148–58, 160, 163, 164, 166, 169–71, 173–6, 178–82, 187, 190–2, 199, 217, 222, 223 non-finite complementation 17–19, 181 non-finite relative clauses 17, 18 normalization 228–30, 233 noun phrase 3, 5, 30, 38, 49, 56–9, 61, 62, 65, 67, 72–6, 81–3, 106, 107, 109, 110, 229, 234 Nuclear Scope 104, 106–8, 110, 111, 118–20 oblique, oblique case 128, 136 Ob-Ugric languages 1, 11, 47, 50, 60, 154 Old Hungarian period 2, 7, 10, 30, 36, 39, 51, 54, 61, 63–5, 71, 75, 104, 122, 146, 148, 159, 169, 174, 176, 177, 186, 203, 228 operator 6, 10, 26, 27, 41, 51, 55, 835, 93, 103, 106–8, 110–15, 117, 119–21, 173, 183–7, 192, 195–200, 202, 203, 212, 217–22 maximality operator 6, 85, 103, 106–8, 110, 120 propositional operator 111, 121 Optical Character Recognition (OCR) 226–7, 231 orthography 226–7, 230 Ostyak see Khanty particle focus particle 50, 53 interrogative particle 16, 17, 73, 181 negative particle 10, 17, 27, 30, 46–55, 181 scalar particle 73 verbal particle 5, 7, 14, 28, 32, 34–9, 46–9, 122–6, 132, 142, 143, 228 phonologically zero complementizer 190, 193, 195 polarity 113, 183, 210, 217

possessive possessive construction 12, 56, 65, 74, 78, 123, 125, 128, 133, 134, 138, 234 possessive affix/suffix 59–61, 100, 107, 155, 161 possessor 5, 24–6, 41, 51, 55, 84, 85, 103, 106–8, 11, 113–15, 119–21, 173, 183–7, 192, 195–7 dative-marked possessor 5, 66, 74–5, 78, 82, 234–5 possessor expression 5, 65, 67, 78–9, 82, 234 pronominal possessor 66–7, 80, 85, 97–8, 100, 102 unmarked possessor 78 postposition 78, 123–6, 132, 133, 136, 141, 143, 144, 162 PredP 28, 29, 31–5, 37, 38, 42, 48, 49, 144 pronoun 5, 6, 10, 11, 18, 19, 23, 25, 26, 47, 50–3, 55, 60, 64, 66, 68–71, 74, 79, 80, 82, 83–5, 87, 99-104, 107, 108, 110–15, 119–21, 136–8, 150, 151, 156, 173, 179, 185, 188, 190, 191, 205, 217–19, 221, 222 indeterminate pronoun 6, 83–5, 111–13, 115, 117, 121 personal pronoun 23, 26, 60, 66, 71, 80, 99–103, 136–8 reciprocal pronoun 104, 107 reduplicated pronoun 83, 114 proper names 65, 85, 106, 230 Proto-Hungarian 2, 4–7, 9–11, 17, 23, 25, 30–2, 34, 35, 37, 39, 41, 46, 50, 53–6, 59, 60, 70, 111, 112, 119–21, 123, 142, 148, 149, 169, 176–9, 181, 182, 193, 223 Proto-Uralic 47, 60 purpose clause 188, 191 quantification 6, 83, 84, 88, 90, 95, 97, 108, 111, 112, 119–21 universal quantification 27, 55, 115, 119 quantifier 6, 9, 10, 26, 27, 29, 37, 55, 67, 79, 80, 83–111, 114–20 A-quantifier 6, 84, 87, 111, 120 D-quantifier 6, 87, 111, 121

Index floating quantifier 85, 86, 92 quantifier movement (see also Quantifier raising) 6, 95, 117, 120, 121 strong quantifier 92, 112, 120 weak quantifier 112 quantifier raising 9, 27, 29, 116 question particle 16, 73, 181 questions 17, 41–4 embedded questions 43, 44 reanalysis 30–2, 41, 52, 57, 67, 69, 82, 111, 128, 133, 140, 141, 149, 166, 168–70, 177, 185, 186, 192, 199 reconstruction 4, 11, 54 reduplication 112, 114, 115 reference 30, 57, 65, 92, 110, 150, 152, 156, 170, 173, 176 referential identification 57, 65, 67 reinforcement 5, 6, 57, 69–72, 75–7, 80, 81, 89, 121, 144, 145, 183–5, 222 relative cycle 7, 149, 179, 185, 186 relative pronouns 18, 55, 108, 110, 112, 114, 179, 190, 191, 217–22 restrictor 106–10, 120 right dislocation 17, 30–2, 37, 177 right edge 153 right extraposition 5, 9, 17, 30–2, 37, 39, 55, 177 Romance languages 67, 69 root participle 166 scope 6, 10, 26, 40, 46, 50, 52, 53, 90, 92–5, 102, 103, 108, 112, 113, 116, 118–21  -curve 4, 10, 11, 54, 149, 163



semhogy ‘neither that’ 211, 213 semmint ‘neither than’ 213 SOV 10, 11, 16, 17, 19–21, 23–5, 30–2, 37, 46, 48, 51, 143, 148, 176–9, 181, 182, 223 spatial element 123, 133, 138, 142, 144 specifier 5, 28, 32, 33, 37, 47, 52, 53, 57, 67, 68, 72–75, 78, 82, 96, 139, 143, 183, 184, 195, 197, 202, 217 specifier incorporation 75 stylistically marked 79, 80, 163 SVO 11, 15, 17, 31, 32, 148, 176, 177, 181, 223 Székely dialect 160 telicityv 9, 35–7, 142, 145 temporal auxiliary 14–16, 49 text processing 3, 71, 225, 228, 230, 231, 233 topic 10, 12, 19–21, 23–7, 30, 31, 38, 117, 176, 183 secondary topic 11, 12, 20, 21, 23, 24, 39 topicalization 75 tripartite determiner quantificationv 109 tripartite structure 6, 107, 109 tripartite quantificational structure (see also Restrictor and Nuclear Scope) 109, 120 Udmurt 14, 16, 17, 30, 40, 181 unmarked object 11–14, 24, 40, 178, 179 variation 38, 100, 125, 133–8, 146 verb modifier 28, 31, 32, 34 verb movement 15, 29, 31, 34, 41–3, 46, 49, 53 Vogul see Mansi Welsh 169, 183, 224

OX F O R D S T U D I E S I N D IAC H R O N IC A N D H I S T O R I C A L L I N G U I S T I C S general editors Adam Ledgeway and Ian Roberts, University of Cambridge advisory editors Cynthia 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; 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 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 in preparation Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden The History of Negation in Low German Anne Breitbarth Variation and Change in the Syntax of Portuguese Relative Clauses Adriana Cardoso Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou The Syntax of Old Romanian Edited by Gabriela Pană Dindelegan Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti The Historical Dialectology of Arabic: Linguistic and Sociolinguistic Approaches Edited by Clive Holes A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τ oς and its Implications for Language Change and Structure Brian D. Joseph Gender from Latin to Romance Michele Loporcaro Participles in Rigvedic Sanskrit The Syntax and Semantics of Adjectival Verb Forms John J. Lowe Quantitative Historical Linguistics Barbara McGillivray and Gard Jenset Syllable and Segment in Latin Ranjan Sen Syntactic Change and Stability Joel Wallenberg The History of Negation in the Languages of Europe and the Mediterranean Volume II: Patterns and Processes Edited by David Willis, Christopher Lucas, and Anne Breitbarth