Historical Linguistics 2003: International Conference On Historical Linguistics, Copenhagen, 11-15 August 2003 (Amsterdam Studies in the Theory and History ... IV: Current Issues in Linguistic Theory)

HISTORICAL LINGUISTICS 2003 AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KON...

Author: Erik jens Mogensen | INTERNATIONAL CONFERENCE ON HISTORICAL L

42 downloads 1232 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

HISTORICAL LINGUISTICS 2003

AMSTERDAM STUDIES IN THE THEORY AND HISTORY OF LINGUISTIC SCIENCE General Editor E.F. KONRAD KOERNER (Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin) Series IV – CURRENT ISSUES IN LINGUISTIC THEORY Advisory Editorial Board Lyle Campbell (Salt Lake City); Sheila Embleton (Toronto) Brian D. Joseph (Columbus, Ohio); John E. Joseph (Edinburgh) Manfred Krifka (Berlin); E. Wyn Roberts (Vancouver, B.C.) Joseph C. Salmons (Madison, Wis.); Hans-Jürgen Sasse (Köln)

Volume 257

Michael Fortescue, Eva Skafte Jensen, Jens Erik Mogensen and Lene Schøsler (eds.) Historical Linguistics 2003. Selected papers from the 6th International Conference on Historical Linguistics, Copenhagen, –5 August 2003

HISTORICAL LINGUISTICS 2003 SELECTED PAPERS FROM THE 16TH INTERNATIONAL CONFERENCE ON HISTORICAL LINGUISTICS, COPENHAGEN, 11–15 AUGUST 2003

Edited by

MICHAEL FORTESCUE University of Copenhagen

EVA SKAFTE JENSEN Roskilde University

JENS ERIK MOGENSEN University of Copenhagen

LENE SCHØSLER

University of Copenhagen

JOHN BENJAMINS PUBLISHING COMPANY AMSTERDAM/PHILADELPHIA

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences — Permanence of Paper for Printed Library Materials, ANSI Z39.48-984.

Library of Congress Cataloging-in-Publication Data International Conference on Historical Linguistics (6th : 2003 : Copenhagen, Denmark) Historical linguistics 2003 : selected papers from the 6th International Conference on Historical Linguistics, Copenhagen, -5 August 2003 / edited by Michael Fortescue … [et al.]. p. cm. -- (Amsterdam studies in the theory and history of linguistic science. Series IV, Current issues in linguistic theory, ISSN 0304-0763 ; v. 257) Includes bibliographical references and index. . Historical linguistics--Congresses. I. Fortescue, Michael D. II. Title. III. Series. P40.I5 2003 47'.7--dc22 200504624 ISBN 90 272 477 4 (Eur.) /  588 586 0 (US) (Hb; alk. paper) © 2005 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microﬁlm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. • P.O.Box 36224 • 020 ME Amsterdam • The Netherlands John Benjamins North America • P.O.Box 2759 • Philadelphia PA 98-059 • USA

Table of contents

Preface vii Typological reflections on loss of morphological case in Middle Low German and in the Mainland Scandinavian languages 1 John Ole Askedal Ethnoreconstruction in Kok-Papónk 21 Paul Black Raising verbs vs. auxiliaries 31 Kasper Boye On the origin of the final unstressed [i] in Brazilian and other varieties of Portuguese: New evidence in an enduring debate 47 Maria José Carvalho Socio-historical evidence for copula variability in rural Southern America 61 Gaillynn D. Clements Main stress left in Early Middle English 75 B. Elan Dresher and Aditi Lahiri Some dialectal, sociolectal and communicative aspects of word order variation and change in Late Middle English 87 Tamás Eitler Using universal principles of phonetic qualitative reduction in grammaticalization to explain the Old Spanish shift from ge to se 103 Andrés Enrique-Arias The origin of transitive auxiliary verbs in Chukotko-Kamchatkan Michael Fortescue

115



Table of contents

Grammaticalisation and Latin Michèle Fruyt

131

Paths of semantic extension: From cause to beneficiary and purpose 141 Silvia Luraghi Vanishing discourse markers: Lat. et vs. sic in Old French and Old Romanian 159 Maria M. Manoliu From ditransitive to monotransitive structure in the history of the Spanish language. Reanalysis of objects: A case of incorporation and monotransitivization 179 Rosa María Ortiz Ciscomani Reflexive intensification in Spanish: Toward a complex reflexive? 199 Johan Pedersen Modern Swedish bara: From adjective to conditional subordinator 225 Henrik Rosenkvist Nordic prefix loss and metrical stress theory with particular reference to ga- and bi- 241 Michael Schulte The origin and development of lär, a modal epistemic in Swedish Gudrun Svensson

257

The development of the Spanish verb in to an auxiliary of voice 279 Thora Vinther The development of continuous aspect Kazuha Watanabe Index 317

301

Preface

The XVIth International Conference on Historical Linguistics that took place in Copenhagen, 11–15 August 2003, gathered the largest number of participants, papers, and languages this conference has ever known. There were 370 participants from 35 countries and 5 continents. Out of a very large number of abstracts submitted and reviewed, 246 were accepted and 236 actually presented. In addition to the general programme, 15 thematic sections were organized and there were between 7 and 9 parallel sessions each day. The sections were as follows: 1. North American and Siberian languages, organized by Una Canger and Michael Fortescue 2. Internal reconstruction in Indo-European: Methods, results and problems, organized by Jens Elmegaard Rasmussen 3. Latin and Greek, organized by Gerd V. M. Haverling 4. Different models of change organized by Ole Nedergaard Thomsen 5. Grammaticalization and reanalysis, organized by Jan Terje Faarlund 6. Language contact, bilingualism and change, organized by Jens Normann Jørgensen and Tore Kristiansen 7. Aboriginal languages, organized by William McGregor 8. Deliberate linguistic changes: when, how, and how often? organized by Sally Thomason 9. Historical pragmatics, organized by Maj-Britt Mosegaard Hansen and Corinne Rossari 10. Language and media, organized by Susan C. Herring 11. Old Germanic syntax, organized by Ulrike Demske and Ans van Kemenade 12. The diachrony of writing, organized by Søren Wichmann 13. Proto-Japanese, organized by Bjarke Frellesvig and John Whitman 14. From Old to Classical French: Parameters of linguistic change, organized by France Martineau and Mireille Tremblay 15. Romance variation and change in a comparative perspective, organized by Thomas D. Cravens and Fernando Tejedo-Herrero.

 Preface

Separate publications are being prepared for papers selected from several of these sections. As usual, plenary speakers were chosen to represent important methodological orientations within Historical Linguistics. They were linked thematically to a number of the sections. The speakers were: Henning Andersen, UCLA, Los Angeles: Synchrony, diachrony, and evolution (Section on different models of change) Joan Bybee, University of New Mexico, Alberquerque: The development of the category auxiliary in English (Section on grammaticalization and reanalysis) William Croft, University of Manchester: Can an evolutionary model alter the way we do historical linguistics? (Section on different models of change) Scott De Lancey, University of Oregon: The semantic transparency of Klamath bipartite stems (Section on North American and Siberian languages) Michèle Fruyt, Université de Sorbonne, Paris IV: Grammaticalization in Latin with special regard to modality (Section on Latin and Greek) Ricardo Otheguy, CUNY Graduate Center NYC: Change and persistence in bilingual and multidialectal communities: Pronominal variation and lexical borrowing in Spanish-English bilingualism in New York (Section on language contact, bilingualism and change) Sally Thomason, University of Michigan: Can you change your language? The limits of historical determination in linguistic change (Section on deliberate linguistic changes: When, how, and how often?) Diane Vincent, Université de Laval, Québec: The history of non-standard discourse markers (Section on historical pragmatics) Prof. Jay Jasanoff, University of Harvard: Internal reconstruction and the PIE bh-cases (Section on internal reconstruction in Indo-European). The Organizing Committee consisted of the following professors and Ph.D. students from the University of Copenhagen and the University of Roskilde: Una Canger, Michael Fortescue, Frans Gregersen, Jan Heegård, Kirsten Kragh, Xavier Lepetit, Anne Jensen, Eva Skafte Jensen, Jens Erik Mogensen, Maj-Britt Mosegaard Hansen, and Ole Nedergaard Thomsen. Student assistants were: Julie Andersen, Ea Feldfos, Marta G. Kirilova, Soon Trolle Nielsen, Ruben Schachtenhaufen, and Louise Torp-Pedersen. Secretary and webmaster: Esben Lydiksen.

Preface

Three members of the Organizing Committee, Michael Fortescue, Eva Skafte Jensen, Jens Erik Mogensen, and the Conference Director, have prepared the present publication. All papers received have been reviewed and 19 papers selected – the criteria for selection were: quality, variety of methods, and variety of topics. They cover the domains of morphology, phonology, pragmatics, and syntax, and include studies on ethnolinguistics, grammaticalisation, language contact, sociolinguistics, and typology. Among the languages treated are: Brazilian, Chukchi, Korean, Danish, English, German, Greek, Japanese, KokPapónk, Latin, Newar, Old Norse, Romanian, Seneca, Spanish, and Swedish.

Lene Schøsler Conference Director



Typological reflections on loss of morphological case in Middle Low German and in the Mainland Scandinavian languages John Ole Askedal University of Oslo

.

Introduction

In what follows, my aim is to assess in a typological perspective the claim that the demise of morphological case marking in Mainland Scandinavian (henceforth: MSc) since the Middle Ages is in some way influenced or even caused by contact with Middle Low German (henceforth: MLG) during the Hanseatic period (12th–16th century). I shall first discuss the question of chronology that is essential to any attempt to establish a cause-effect relationship. Second, case marking being a special instance of morphological marking, it is necessary to deal with the morphological techniques used for marking inflexional categories, including in particular case, in MLG and MSc NPs and their historical development in a comparative perspective. Third, there is the question whether loss of case marking is more or less natural with regard to other typological characteristics of MLG and MSc, in particular verb position (cf. “Universal 41” in Greenberg 1966: 96). Finally, both MLG and MSc are languages spoken in a part of Northern Europe where there has been extensive cultural exchange and migrational activity for centuries even long before the Hanseatic period. It is therefore necessary to consider the relevance of areal factors extending beyond the Baltic and Scandinavian domain in connection with the loss of morphological case in MSc and MLG.



John Ole Askedal

. Chronology Three main hypotheses concerning a causal relationship between MLG and MSc loss of morphological case are represented in the literature: (i) case loss in MLG caused case loss in MSc; (ii) case loss in MSc occurred independently of morphosyntactic developments in MLG; and, as a modified version of (ii), (iii) case loss in MLG did not cause case loss in MSc but reinforced it. Both hypotheses (i) and (iii) presuppose a contact situation of some sort; hypothesis (ii) of course does not, but it does not preclude one either. On the basis of hypothesis (i), case loss in MSc may be explained by system-external factors; hypothesis (ii) implies a system-internal explanation for MSc case loss, and (iii) allows for an explanation of MSc case loss in terms of a combination of system-internal and system-external factors. In the recent literature, I have only been able to find one explicit ‘pure’ adherent of (i) (Jahr 1998: 42). Marold (1981: 147) can be read as a representative of (ii). Most researchers in the field appear, however, to favour some variety of (iii) (cf. e.g. Wessén 1965: 147f.; Haugen 1976: 208f.; Norde 1997a: 35–45; Norde 1997b: 403; Barðdal et al. 1997: 83). Logically, it is trivial that for a state x to be able to cause another state y, x has to precede y (but chronological precedence is not in itself a sufficient condition for a causal relationship). If x follows y, then y cannot have caused x, but, once arisen, a chronologically later state x may of course exert influence on the chronologically earlier state y, for instance by giving further impetus to developments that have y as their basis. Thus, if loss of morphological case can be shown to have occurred earlier in MLG (state x) than in MSc (state y), MLG case loss may, by way of cultural and commercial contact, have been the cause of MSc case loss (but it need not be). If, on the other hand, it can be shown that case loss in MSc (state y) precedes case loss in Low German (state x), then one may safely conclude that case loss in MLG cannot have been the cause of MSc case loss, but one is still entitled to believe that MSc case loss may later on have been reinforced by case loss in MLG. According to the empirical findings presented by Ringgaard (1986) and Marold (1981; 1984), the first indications of case neutralisations within the MSc area can be observed in Danish as early as the 12th century, at a time when MLG had a functioning case system and prior to the rise of extensive commercial contacts between MSc and the Hanseatic league with its MLG language.1 These findings may be taken to refute hypothesis (i) above, whereas both hypotheses (ii) and (iii) are compatible with them. The fact that (iii) is the most favoured hypothesis may be in part due to the overwhelming impression the extent of lexical borrowing from MLG into MSc has always made on

Typological reflections on loss of morphological case

researchers in the field. In particular, the massive MLG impact on the MSc lexicon has given rise to the view that morphological – declensional and allomorphic – differences between MLG and MSc may have caused problems for the integration of MLG loan words into MSc inflexional morphology, thereby undermining MSc case distinctions (cf. Hyldgaard-Jensen 1983: 670). However, empirical investigations by Simensen (1995: 64) and Norde (1997b: 398–402) have convincingly shown that morphological integration of nouns borrowed from MLG into Middle Norwegian and Middle Swedish, respectively, do not seem to have posed any great problems for the MSc case marking systems existing at the time; and Marold (1984: 311–318, 323) has, to my mind equally convincingly, been able to show that the morphological inventory differences between MLG and MSc noun inflexion were in all probability not sufficiently extensive to have been able to endanger the system of case inflexion in these MSc languages (cf. also Marold 1981: 142–144). Hypothesis (i) is thus far from being unequivocally corroborated by the facts. In particular, the empirical basis of the assumption concerning morphological incompatibility between MSc and MLG that has been adduced in support of hypothesis (iii) is less well-founded than is often thought. It should, however, be noted that the rejection of both hypotheses (i) and (iii), and hence the acceptance of hypothesis (ii), is compatible with assuming that case loss may have similar system-internal causes both in MSc and in MLG. Acceptance of hypothesis (ii) is of course also compatible with the view that case loss may in both languages have similar areal connections beyond the more local, geographical and cultural relationship between MLG and MSc.

. Case marking in Middle Low German and ways of marking morphological case in NPs MLG and Old MSc have the same morphological cases – nominative, accusative, dative, and genitive – and the syntactical and lexical case assignment rules are also to a very large extent the same. It is, however, important to realize that this need not imply that the rules for morphological case marking in NPs operate according to the same structural pattern. To illustrate this claim, I shall introduce a typological distinction between two different strategies for categorial marking in NPs which I propose to dub ‘distributive’ and ‘selective’ marking, respectively, and on this basis compare the state of affairs in Modern Standard German, MLG and Modern Low German, and Modern MSc.





John Ole Askedal

The traditional Indo-European way of categorial marking in NPs is by means of socalled ‘agreement’, meaning that categorially corresponding morphemes are distributed across the constituents of the NP, hence the term ‘distributive’ marking. This marking system persists in a number of modern Indo-European languages. A case in point is Modern Russian. Cf. e.g. (1): (1) a. etot thismasc.sg.nom b. etomu thismasc.sg.dat c. etogo thismasc.sg.gen

krasivyj beautifulmasc.sg.nom krasivomu beautifulmasc.sg.dat krasivogo beautifulmasc.sg.gen

dom-Ø housemasc.sg.nom domu housemasc.sg.dat doma housemasc.sg.gen

Like Russian, Modern Standard German is also a case-marking language but, unlike Russian, categorial marking in NPs does not operate according to the ‘across the board’ principle illustrated in (1). With the exception of the -s ending in article words and the allomorph -es in nouns, both of which are uniquely gen.sg., and dat.sg. -em, individual endings in modern German are not categorially distinctive the way the endings in (1) are, but combinations of endings may be so.2 Cf. the examples in (2): (2) a. die Wünsche des alten Kunden = die Wünsche der alten Kunden b. Sie unterstützten die alten Frauen. = Sie halfen den alten Frauen. c. Sie wollten den jungen Mann unterstützen. = Sie wollten dem jungen Mann helfen. d. die Wünsche des jungen Mann(e)s e. den alten Männern f. dem jungen Mann(e)

In the NPs in (2), not all individual constituents are distinctively marked, hence the term ‘selective’ marking. Instances like (2a)–(2c), where difference in one morpheme position marks the opposition, are commonly called ‘monoflexional’, and instances like (2d)–(2e), where two endings cooperate to mark the morphological category in question, are called ‘biflexional’. (2f) exemplifies the (rare) possibility of a choice between ‘mono-’ and ‘biflexionality’ in certain instances. As first noted by Werner (1979), this system represents a development away from the original Indo-European ‘distributive’ agreement system that still predominates in modern inflecting Indo-European languages like Russian. Concerning the marking of morphological case in MLG, the dissertation of Ulla Björnheden on the South Westphalian “Wolfenbütteler Psaltertext” (from

Typological reflections on loss of morphological case

the first half of the 14th century, in the edition of Rooth 1919; cf. Björnheden 1997: 15f.) provides interesting material.3 This text contains no less than 12885 NPs and 683 different verbs. This ought to be a sufficient basis for an investigation of case assignment and case distinctiveness. The main content of Björnhedens investigation may be summarized thus (cf. in particular Björnheden 1997: 174f.): (3) – –

–

–

–

–

–

37% of all NPs in the text display unequivocally identifiable, distinctive case marking. The distribution of distinctively case-marked NPs on the four cases is as follows: nominative – 21%, dative – 8%, genitive – 6%, accusative – 2%. In 20% of all NPs, unequivocal case identification is possible on the basis of distinctive case marking of single words (pronouns, certain adjective endings).4 In the remaining 17% of distinctively case-marked NPs, the distinctiveness results from a combination of words and morphemes which would not be distinctive individually.5 When in addition to morphological marking the syntactic context – verb valency and agreement phenomena – is also taken into consideration, there is a residue of a mere 5% of all NPs which do not allow for clear-cut case identification. 13% of the NPs in the text show case syncretism between the nominative and the accusative but when the syntactic context is also considered, only 3% remain where case is not clearly identifiable. In the “Wolfenbütteler Psaltertext”, case distinctiveness in non-pronominal NPs is hardly ever a matter of morphological distinctiveness of one word-form only but manifests itself as a distinctive combination of morphemes (Björnheden 1997: 104).

The last observation specifically concerning case marking in non-pronominal NPs can be illustrated by the following examples:6 (4) a. b. c. d. e. f.

din-Ømunt-Ø (masc.sg.nom.) din-Øregtheit-Ø (fem.sg.nom.) den weg-Ø (masc.Sg.acc.) dine hande (fem.pl.nom./acc.) den rechten pad-Ø (masc.sg.acc.) den scarpen worde (neut.sg.dat.)

= = = = = =

dinen munt-Ø (masc.sg.acc.) dine regtheit-Ø (fem.sg.acc.) den wege (masc.sg.dat.) dinen handen (fem.pl.dat.) den rechten pade (masc.sg.dat.) den scarpen worden (neut.pl.dat.)





John Ole Askedal

The single morphemes -Ø, -e, -en are in general not distinctive in isolation but combinations of two morphemes are; cf. (5), where the endings in (4) have been stripped of lexical support: (5) a. b. c. d. e. f.

-Ø... -Ø (masc.sg.nom.) -Ø... -Ø (fem.sg.nom.) -en... -Ø (mask.sg.acc.) -e... -e (fem.pl.nom./acc.) -en... -en... -Ø (masc.sg.acc.) -en... -en... -e (neut.sg.dat.)

= = = = = =

-en... -Ø (masc.sg.acc.) -e... -Ø (fem.sg.acc.) -en... -e (masc.sg.dat.) -en... -en (fem.pl.dat.) -en... -en... -e (mask.sg.dat.) -en... -en... -en (neut.pl.dat.)

In the MLG text in question, morphological distinctiveness of non-pronominal NPs is, as a rule, a matter of two cooperating morphemes.7 The MLG case marking system illustrated in (4)–(5) is clearly typologically akin to the Modern Standard German system in (2). The same is true of the Modern Low German system in (6), where certain remnants of case distinctions still exist (cf. Lindow & al. 1998: 144–145, 191f.) within the same overall selective marking pattern:8 (6) a. NPs with a definite determiner Sg. Masc. Fem. Nom. de dicke Kopp de oole Kann Non-nom. den dicken Kopp de oole Kann Pl. Masc. Fem. Nom. de dicken Köpp de oolen Kannen Non-nom. de dicken Köpp de oolen Kannen b. NPs without a definite determiner Sg. Masc. Fem. Nom. dicke Kopp oole Kann Non-nom. dicken Kopp oole Kann Pl. Masc. Fem. Nom. dicke Köpp oole Kannen Non-nom. dicke Köpp oole Kannen

Neut. dat oole Book dat oole Book Neut. de oolen Böker de oolen Böker Neut. ool Book ool Book Neut. oole Böker oole Böker

In this system, the opposition between a nominative and a non-nominative case, the latter deriving from the former accusative and dative, is only retained in the masc. sg. In the singular, the nominative–non-nominative opposition is marked monoflexionally, and in the plural biflexionally in the attributive determiner and the following adjective.9 Present-day Low German thus has a selective marking system of the same kind that is found in MLG (as described by Björnheden 1997) and in Modern German.10

Typological reflections on loss of morphological case

As noted by Mogensen (1999: 113), the South Westphalian text described in (3) cannot unconditionally be considered representative of the MLG dialect area as a whole. Besides, one may reasonably assume written MLG in general to have been somewhat more morphologically conservative than its presumably more “progressive” spoken varieties (cf. Lasch 1974: 3, 213), in which morphological neutralizations of various kinds were in all probability more widespread. Medieval Scandinavians were of course on the whole more likely to come into contact with these spoken varieties than with written MLG.11 Categorial marking in MSc NPs differs systematically from the Low German state of affairs. The most striking difference is that definite nouns are marked for definiteness with a grammatical ending that, e.g. in Modern Norwegian Riksmål, also marks the opposition between singular and plural, and within the singular, the opposition between common gender and neuter. In official Norwegian Bokmål and in Nynorsk (‘New-Norwegian’), there is an opposition between masculine, feminine and neuter gender in the singular and a non-neuter–neuter opposition in the plural.12 Cf. the Norwegian examples in (7),13 which are also representative of the corresponding Swedish structures:14 (7) a.

den gamle boken (Riksmål/Bokmål) theDEF.SG. COMMON GENDER oldWEAK bookSG.DEF. COMMON GENDER b. den gamle boka (Bokmål/Nynorsk) theDEF.SG. COMMON GENDER oldWEAK bookSG. DEF.FEM c. det store huset (Riksmål/Bokmål/Nynorsk) theDEF.SG.NEUT largeWEAK houseDEF.SG.NEUT d. de store husene (Riksmål/Bokmål) theDEF.PL largeWEAK housesDEF.PL e. de store husa (Bokmål) theDEF.PL largeWEAK housesDEF.PL.NEUT

In addition to the category neutralisations referred to already, one may note the neutralisation of the opposition between the masculine and the feminine in the prenominal definite article in all varieties of Norwegian. The neutralisation of all category oppositions in the weak adjective would seem to imply that definite NPs in modern Norwegian are characterised by biflexional distributive plural or – in the singular – gender marking. In (7), the ancient distributive marking strategy is easily recognisable despite these neutralisations. In Danish, where the definiteness suffix on the noun is absent in the presence of a prenominal determiner with a following adjective, one may in fact in instances like (8) assume selective monoflexional gender or plural marking of the German type:15





John Ole Askedal

(8) a.

det store hus (Danish) theDEF.SG.NEUT largeWEAK houseSG b. de store huse (Danish) theDEF.PL largeWEAK housesPL

In NPs with an indefinite determiner and an attributive adjective (9) and in definite NPs without an attributive adjective (10), Norwegian, Swedish and Danish have identical structures (Norwegian examples): (9) a.

en gammel bok (Riksmål/Bokmål) anINDEF.SG.COMMON GENDER oldSG. COMMON GENDER bookSG.INDEF b. ei gammal bok (Nynorsk) anINDEF.SG.FEM oldSG.FEM bookSG.INDEF c. et stort hus (Riksmål/Bokmål) aINDEF.SG.NEUT largeSG.NEUTR houseSG.INDEF d. store hus (Riksmål/Bokmål/Nynorsk) largeWEAK housesINDEF.PL.NEUT

(10) a.

mannen (Riksmål/Bokmål/Nynorsk) manDEF.SG.COMMON GENDER “the man” b. huset (Riksmål/Bokmål/Nynorsk) houseDEF.SG.NEUT “the house” c. husene (Riksmål/Bokmål) housesDEF.PL “the houses”

In syntagmatically complex indefinite NPs (9), the prenominal article and adjective are marked correspondingly, and in Norwegian and Swedish definite NPs, the preposed article and the noun marked for definiteness have corresponding marking. On the whole, the ancient distributive agreement system still persists in MSc complex NPs. In a historical perspective it has become categorially simpler insofar as case marking is lost and only number and gender marking remain. It may, however, also be said to have strengthened its position due to the establishment of definiteness as an inflexional category in nouns, cf. in particular the categorial correspondence between article word and noun in Norwegian (and Swedish) in (7). As noted, the corresponding definite structures in Danish are more German-like (but indefinite structures like et stort hus “a large house” show material morphological correspondence between article and ‘strong’ adjective in all three languages). With regard to category marking in non-pronominal NPs, there is thus a clear-cut difference between a German mono-/biflexionality marking system in MLG (and in Modern Low German) on the one hand, and a more traditional

Typological reflections on loss of morphological case

agreement system in MSc on the other. Whereas both MLG and medieval MSc were case marking languages, morphological case is fully neutralised in nonpronominal NPs in Modern MSc16 and only vestigially present in most Modern Low German dialects. In my view, this scenario does not support the assumption that MLG is in some way responsible for the loss of morphological case in MSc. Since the category marking rules for NPs in MSc, in particular in Swedish and Norwegian, are still in principle different from what we find in MLG (and in Modern Low German), MSc speakers cannot have copied the MLG category marking system or the case system that was part of it; if they had done so, Modern MSc would have turned out quite differently. This holds good whether one takes as one’s point of departure a system with more morphological and case marking, like the one studied by Björnheden (1997) (cf. (3)), or a spoken variety with more morphological neutralisation. In both cases, the question remains why only the case marking part of the overall Scandinavian category marking system was shattered, not its basic structural traits as a distributive category-marking system as such. There is thus no real structural evidence to the effect that MSc speakers suffered structurally damaging confusion when exposed to MLG case marking (or the lack of it).

. The borrowing hierarchy For several centuries, MLG exerted a profound and lasting influence on MSc in the lexical domain that led to partial relexification of MSc.17 Equally important is, however, the insight that MLG influence was far from being equally strong in other parts of the linguistic system of the MSc languages. A number of researchers place great emphasis on the borrowing of word formation patterns and elements in the shape of certain suffixes and prefixes. There is, however, a productivity difference between the MLG suffixal and prefixal formations in the MSc languages. Whereas certain suffixes, like -hed/ -heit/-het and -else in abstract nouns and -er/-are in agent nouns have become productive in the sense that they came to be used increasingly freely with indigenous MSc word stems (e.g. Modern Norw. valgfrihet “optionality”, tømrer “carpenter”, nytelse “enjoyment, pleasure”), the prefixes (e.g. Modern Norw. an-, be-, for-, unn-) have hardly expanded beyond the lexical entities of which they were integral parts when the borrowing took place. On the other hand, the word formation pattern ‘prefix + verb stem’, which had become extinct in Proto-Nordic (cf. Schulte 2003), was reintroduced into MSc through the loan words in question and became productive with indigenous MSc prepositions





John Ole Askedal

and adverbs which came to be used as prefixes too; cf. e.g. the MLG loanword anrope “hail” with the foreign prefix an- vs. påkalle “call upon, beseech, claim, demand” with the indigenous prepositional prefix på- with roughly the same basic meaning; cf. in this connection also lexical pairs like gå inn i “enter” vs. inngå i “be part of ”. It is, however, well known that massive lexical influence need not entail any structural syntactic influence at all (cf. e.g. Thomason & Kaufman 1988: 258 on Chinook jargon). To all appearances, the MLG syntactic influence on MSc is very modest compared with the extent of lexical influence but two more or less widespead syntactic constructions in Modern MSc are most likely of MLG origin: (i) In NPs consisting of a title and a given name, the old MSc order with the name before the title was discarded in favour of the MLG order with the title before the name. Cf. e.g. (11): (11) Old Norse: Haraldr konungr – Modern Norw.: kong Harald “King Harold”

This change was carried through in the whole MSc area. On the other hand, the new ordering was not generalised to all appositional structures where it might have been applicable; for instance, place names like Akershus slott “Akershus Castle” have not undergone a corresponding change to *Slott Akershus in accordance with the German pattern in e.g. Schloß Moritzburg, Stift Melk, etc. (ii) Of particular interest is the introduction into MSc of a MLG genitive periphrasis (12) (resulting from the syntactic reanalysis of a verbal construction with a possessive dative):18 (12) mannen si ku manMASC.SG.DEF hisPOSS-PRON. REFL.FEM.SG cowFEM.SG “the man’s cow”

Until fairly recently, this construction was restricted to the western periphery of MSc. In Danish, it is found in Jutlandic dialects, and in Norway it has traditionally been a syntactic hallmark of western and northern dialects (cf. Torp 1992). It has also won official recognition as an integral part of present-day Nynorsk syntax. At present, it is gaining ground in colloquial East Norwegian and from there, it is encroaching on, and is becoming increasingly frequent in, written Bokmål too (but not in the more traditional Riksmål variety). The general conclusion is that MLG influence on MSc is hierarchically ordered in accordance with grammatical domains: MLG has had a massive impact on the MSc lexicon and has also left its imprint on word formation in

Typological reflections on loss of morphological case

the guise of certain frequent and in part productive derivational patterns. In the syntactic domain, MLG is also responsible for the restructuring of certain NP constructions and the introduction of, in certain parts of the MSc area, a productive analytical genitive construction (12), but both these changes belong to the, relatively speaking, lower NP level. On the higher clause and sentence levels of MSc, there is no evidence of lasting MLG syntactic influence on MSc. Notably, MLG has left no imprint on MSc word order above the NP level. Cf. for instance Braunmüller (1998: 330; author’s emphasis): “Obwohl Mnd. die dominante Prestigesprache im Ostseeraum, im Bereich der Nordsee und des östlichen Atlantiks war, bestimmt das Mnd. nicht die Wortstellungsnormen in den skand. Kontaktsprachen.” This general picture of the relationship between MLG and MSc lends itself naturally to an interpretation in terms of Thomason & Kaufman’s (1988: 74– 76) ‘borrowing scale’ (13): (13)

1. Casual contact: lexical borrowing only < 2. Slightly more intense contact: slight structural borrowing < 3. More intense contact: slightly more structural borrowing < 4. Strong cultural pressure: moderate structural borrowing < 5. Very strong cultural pressure: heavy structural borrowing.

MLG influence on MSc would seem to belong in category 2, or at most in category 3 in (13).

. Typological and areal perspectives on the development of case marking The question is, then, to what extent MLG influence may be assumed to have triggered or reinforced loss of morphological case in MSc, given the overall impression that the degree of MSc structural, as opposed to lexical, borrowing from MLG was, to all appearances, only “slight”. An answer to this question should be in keeping with, first, universal correlations between case marking and other functionally related syntactic phenomena and, second, what is known about the genetic and areal stability of basic structural properties of languages. Concerning, first, the relationship between case and functionally related grammatical properties, Greenberg’s (1966: 96) wellknown “Universal 41” provides a pertinent point of departure: “If in a language the verb follows both the nominal subject and nominal object as the dominant order, the language





John Ole Askedal

almost always has a case system.” Greenberg’s “Universal 41” entails the four different combinations of main verb order and presence or absence of case marking in (14), of which the second is deemed less natural (marked by – ) than the other three (marked by + ): (14)

+

Type 1. Type 2. + Type 3. + Type 4. –

XV (‘Verb/Final’), XV (‘Verb/Final’), VX (‘Verb/non-Final’), VX (‘Verb/non-Final’),

+ case marking of NPs – case marking of NPs + case marking of NPs – case marking of NPs

The following (15) are Germanic manifestations of the types in (14):19 (15) Type 1. Proto-Germanic, MLG: XV (Verb/Final) Type 2. Modern Low German: XV (‘Verb/Final’), Type 3. Old Norse: VX (‘Verb/non-Final’), Type 4. Modern MSc: VX (‘Verb/non-Final’),

N, A, D, G Non-Oblique, Oblique N, A, D, G Non-Obl, Obl (only in personal pronouns)

This means that MSc has gone through a long-term ‘natural’ development from Proto-Nordic (as represented by the language of the early Runic inscriptions) through Ancient Nordic (Old Norse etc.) to its modern stage, going from the universally common and possibly original Indo-European type 1 in (14) to the common modern West European Type 4, with Type 3 as a transitional stage. On the other hand, the development of Low German has followed a different path from Type 1 to the, universally speaking, less natural or favoured Type 2. Considering MLG influence as the cause, or at least as an accelerating factor in the MSc loss of morphological case amounts to assuming that one case language A (MLG) should cause case loss in another partly typologically different case language B (MSc) possessing different case marking rules, in the course of a long-term development where B (MSc) evolves towards a universally more favoured type than A (MLG) with regard to the absence of morphological case marking. In a typological perspective, this is, to say the least, not the most likely of contact-linguistic scenarios. The Germanic long-term developments summarized in connection with (15) can also be seen in the light of the genetic and areal stability hierarchies proposed by Nichols (1992: 170). Cf. (16):20

Typological reflections on loss of morphological case

(16) a.

Hierarchy of genetic stability Genetically most stable Genetically least stable Alignment > Head/dependent marking > Complexity > Word order

b. Hierarchy for areal groupings Most areal Least areal Word order > Alignment > Head/dependent marking > Complexity

According to (16), word order is least genetically stable and strongly areal (cf. also Nettle 1999: 137). In a local Baltic and Scandinavian perspective, one might therefore prima facie expect MLG, being the culturally and commercially dominant language, to have influenced MSc with regard to verb order but evidently there has been no radical restructuring of MSc word order due to MLG (cf. again Braunmüller 1998: 316).21 Morphological complexity is least areal and somewhat more genetically stable than word order.22 From this one may at the very least conclude that the universal tendencies proposed by Nichols do not favour the assumption of a decisive influence in either direction as far as loss of morphological case in the Baltic and Scandinavian area is concerned.

. Concluding remarks Modern German exists in two typologically partly different varieties: first Modern Standard High German with its four cases in combination with SOV order (Type 1 in (14)) and, second, Low German with its highly reduced case system and basic SOV order (Type 2 in (14)). In Western Europe, the latter type is also represented by the neighbouring Germanic languages Dutch and Frisian (cf. e.g. Askedal 1995: 96f.). To the East, one finds the Baltic and Slavic case languages with at least statistically predominant SVO order (corresponding to Type 3 in (14)), and to the North, West and South there are MSc, English, and the western Romance languages, all of which are SVO languages with far-reaching neutralisation of morphological case oppositions (cf. Type 4 in (14)).23 The western extremity of Europe is the homeland of the modern Celtic languages which also lack morphological case distinctions and have VSO order (which can be considered another variety of Type 4 in (14)).24 In a wider typological and areal European perspective, it is thus not MSc with its common European combination of SVO and mere relics of case distinctions which is a special case, it is rather Low German, which, together with Dutch and Frisian, represents a universally non-favoured combination of ancient West Germanic





John Ole Askedal

SOV ordering and modern West European neutralisation of morphological case (Type 2 in (14)). Inasmuch as case neutralisation is the more far-reaching West European structural characteristic, this phenomenon should be explained in more general structural terms. In my contribution to the Vancouver conference (Askedal 2001; cf. also Askedal 1993), I argued that what lies behind this whole process is, first, a basic shift from a ‘non-configurational’ to a ‘configurational’ syntactic type, where the coding of syntactic relations is transferred from morphology to topology, through a process of grammaticalization of element (‘word’) order.25 Second, on the basis of the socalled ‘oblique subjects’ found in Modern Icelandic and Faroese (cf. also Faarlund 2001a), I argued that the basic syntactic shift occurs prior to loss of morphological case and hence that loss of morphological case is not a precondition for, but rather a consequence of, the basic syntactic shift. This view of the general line of development receives support from Cynthia Allen’s studies on English (cf. e.g. Allen 1995; 1996).26 From this it follows that it is the state of affairs in Low German, being a Type 2 language with SOV order and absence of functional case, not MSc, which is in need of the more ‘local’, areal explanation. In Low German, as in Dutch and West Frisian, word order is genetically stable whereas the traditional Germanic case system has been subjected to common West European levelling tendencies. In these languages, the result is a correlation of basic Verb/Final order and case loss which in a European and universal perspective is less common and less favoured than the type represented by MSc and the majority of West European languages.

Notes . According to Haugen (1976: 208f.), case marking in the dialect of Zealand had reached the stage of present-day Danish as early as ca. 1350. In Sweden and Norway, the process of case loss on the whole progressed at a somewhat slower pace. . Cf. e.g. Werner (1979), Weinrich (1993: 487), Ronneberger-Sibold (1997: 314–324). . One has, of course, to acknowledge that one single text from one part of the vast Low German area cannot be considered representative of the whole area (Mogensen 1999: 113). The point is, however, that Björnhedens material allows one to identify a structural principle that may be assumed to have more general, not regionally restricted validity when seen in the light of a pervasive structural development in German. . “Einkasusmorphem” in the terminology of Björnheden (cf. Björnheden 1997: 104–113).

Typological reflections on loss of morphological case . “Einkasuskombination” in the not quite transparent terminology of Björnheden (1997: 114–130); cf. also Mogensen (1999: 110, 113). . Cordes (1983: 229) remarks that the neutralization of the opposition between the accusative and the dative is particularly characteristic of MLG but adds that this pertains more to the declension of pronouns than to the declension of nouns. . There are on record combinations like (i) (quoted form Lasch 1974: 204, n. 2, and (ii) (from Lundemo, ms.: 34), but such instances are of marginal occurrence: (i) des naghestes dincdaghes (Hamburger Statdtrecht 1292) (ii) des gudes weges The explanation generally assumed for instances like (i)–(ii) is that they represent some sort of morphological ‘attraction’ (Lasch 1974: 204, n. 2). – Similar examples occurr in Middle High German (cf. e.g. Weddige 2003: 62) . Kopp “head”, Kann “can”, Book “book”, Disk “table”, Düör “door”, Huus “house”. . Certain modern South Westphalian dialects have retained an additional opposition between the accusative and dative. Cf. (i) (cf. Lasch 1974: 212; the paradigms below are quoted from Lindow & al. 1998: 144f. – adjectives follow the pattern in (6) and are therefore left out here): (i) NPs with a definite determiner Sg. Masc. Fem. Nom. de Disk de Düör Acc. den Disk de Düör Dat. dem Diske de Düöre Pl. Masc. Fem. Nom. de Diske de Düören Acc. de Diske de Düören Dat. den Disken den Düören

Neut. dat Huus dat Huus dem Huse Neut. de Hüser de Hüser den Hüsern

The repertoire of endings is very limited and marking is a question of mono- or biflexionality in this southwestern system too. It should be noted that the suffixes -e and -n on nouns are true case endings, just as in Modern Standard German. . I disregard the genitive, which in Modern Low German only exists as an -s ending on proper nouns. In many varieties of Modern MSc, it has a wider use but, presumably like its Low German counterpart, it represents a special agglutinative morphological structure (cf. Askedal 2003) of its own which sets it apart in the present context. With regard to the terminology proposed here, the -s genitive – in e.g. Norw. den gamle mannens hus “the old man’s house” – can be characterized as strictly monoflexionally selective. . Cf. e.g. Braunmüller (1996) for an analysis of communication between Scandinavians and MLG speakers with reference to the concept of ‘semicommunication’. . On the rather intricate Norwegian linguistic situation, in particular the three main varities Riksmål, Bokmål and Nynorsk, cf. Haugen (1966). . The weak declension of adjectives has in Norwegian uniformly the ending -e and is therefore not further specified with regard to inflexional categories.





John Ole Askedal . Like Norwegian Riksmål, Swedish has no feminine gender. In the domain of nouns, adjectives and pronouns there exist a few other rather minor inflexional differences between the two languages which are of no consequence in the present context. . On the use of the ‘Danish’ type in (8) in Norwegian Riksmål/Bokmål, cf. Lundeby (1965: in particular 304–310). . On the genitive as a very special case not relevant in this context see Note 10. . Cf. e.g., concerning Swedish, Wessén (1954: 13–27; 1965: 70), Pettersson (1996: 120– 124). . Cf. in particular Nesse (2002: 37f., 170–193) and, on the corresponding Modern Low German construction, Lindow & al. (1998: 269f.). . The classification of MLG and Modern Low German as XV (SOV) is based on the same kind of syntactic evidence that underlies the corresponding classification of Old and Modern High German, i.e. the manifest Verb/Final structure of subordinate clauses and the numerous ‘frame’ constructions (“Rahmenkonstruktionen”). – For a fuller picture of the areal relationships within Germanic see e.g. Askedal (1995). . Cf. also Nichols (1992: 171f., 181, 195, 245f., 249f.). . It should be noted, though, that the V/3 order in MSc subordinate clauses may be due to MLG influence (cf. Larsson 1931: 52, 136–148, 152–158, 169, 172f., 178, 189f.). However, this is a far cry from the Verb/Final order of Modern Low German subordinate clauses (Lindow et al. 1998: 197) (which may have been influenced by High German). . Thomason & Kaufman (1988: 132) point to the fact that language contact in a universal perspective often leads to the establishment of new morphological oppositions. . On the individual Romance languages cf. e.g. Posner (1996: 115–125). . On the individual Celtic languages cf. Ball & Fife (1993: 114, 172, 240, 311 f., 364). . Cf. also Faarlund (2003). . Cf. also Benskin (2001: 107–115) and, on the development from Old Norse to Modern Norwegian, Mørck (1994; 1998).

References Allen, Cynthia L. (1995). Case marking and reanalysis: Grammatical relations from Old to Modern English. Oxford: Oxford University Press. Allen, Cynthia L. (1996). A change in structural case marking in Early Middle English. In Höskuldur Thráinsson, Samuel David Epstein, & Steve Peter (Eds.), Studies in Comparative Germanic Syntax, Vol. II (pp. 3–20). Dordrecht, Boston and London: Kluwer.

Typological reflections on loss of morphological case

Askedal, John Ole (1993). Configurationality in language typology and diachronic syntax. In André Crochetière, Jean Claude Boulanger, & Conrad Ouellon (Eds.), Actes du XVe Congrès international des Linguistes, Québec, Université Laval, 9–14 août 1992 / Proceedings of the XVth International Congress of Linguists, Québec, Université Laval, 9–14 August 1992, Vol. IV (pp. 3–6). Sainte-Foy: Les Presses de l’Université Laval. Askedal, John Ole (1995). Geographical and Typological Description of Verbal Constructions in the Modern Germanic Languages). In John Ole Askedal & Harald Bjorvand (Eds.), Drei Studien zum Germanischen in alter und neuer Zeit (= NOWELE Supplement Vol. 13) (pp. 95–146). Odense: Odense University Press. Askedal, John Ole (Ed.). (1998). Historische germanische und deutsche Syntax: Akten des Internationalen Symposiums anläßlich des 100. Geburtstages von Ingerid Dal, Oslo, 27.9.– 1.10.1995 (= Osloer Beiträge zur Germanistik 21). Frankfurt am Main etc.: Peter Lang. Askedal, John Ole (2001). ‘Oblique subjects’, structural and lexical case marking: Some thoughts on case assignment in North Germanic and German. Faarlund (Ed.). 65–97. Askedal, John Ole (2003). Grammaticalization and the historical development of the genitive in Mainland Scandinavian. Blake & Burridge (Eds.). 21–32. Ball, Martin J., & James Fife (Eds.). (1993). The Celtic Languages. London: Routledge. Barðdal, Jóhanna, Nils Jörgensen, Gorm Larsen, & Bente Martinussen (1997). Nordiska. Våra språk förr och nu. Lund: Studentlitteratur. Benskin, Michael (2001). [Review article] A New History of Early English: Hans Frede Nielsen: A Journey Through the History of the English Language in England and America, Volume I: The Continental Backgrounds of English and its Insular Development until 1154 (North-Western European Language Evolution [NOWELE] Supplement Vol. (19) [...]. Norsk Lingvistisk Tidsskrift, 19, 93–122. Björnheden, Ulla (1997). Zum Vierkasussystem des Mittelniederdeutschen (= Göteborger Germanistische Forschungen 38). Göteborg: Acta Universitatis Gothoburgensis. Blake, Barry J., & Kate Burridge (Eds. with the assistance of Jo Taylor). (2003). Selected Papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001 (= Current Issues in Linguistic Theory; Amsterdam Studies in the Theory and History of Linguistic Science 237). Amsterdam and Philadelphia, PA: John Benjamins. Braunmüller, Kurt (1996). Forms of Language Contact in the Area of the Hanseatic League: Dialect Contact Phenomena and Semicommunication. Nordic Journal of Linguistics, 19, 141–154. Braunmüller, Kurt (1998). Wortstellungstypologische Untersuchungen zu den Kontaktsprachen der Hansezeit (Mittelniederdeutsch, Dänisch, Schwedisch). Askedal (Ed.). 315–334. Cordes, Gerhard (1983). Mittelniederdeutsche Grammatik. Cordes & Möhn (Eds.). 209– 237. Cordes, Gerhard, & Dieter Möhn (Eds.). (1983). Handbuch zur niederdeutschen Sprach- und Literaturwissenschaft. Berlin: Erich Schmidt. Faarlund, Jan Terje (2001a). The notion of oblique subject and its status in the history of Icelandic. Faarlund (Ed.). 99–135. Faarlund, Jan Terje (Ed.). (2001b). Grammatical Relations in Change (= Studies in Language Companion Series 56). Amsterdam and Philadelphia, PA: John Benjamins.





John Ole Askedal

Faarlund, Jan Terje (2003). The grammaticalization of movement: Word order change in Nordic. Blake & Burridge (Eds.). 129–142. Greenberg, Joseph H. (1966). Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements. In Joseph H. Greenberg (Ed.), Universals of Language, 2nd ed. (pp. 73–113). Cambridge, MA and London: M.I.T. Press. Haugen, Einar (1966). Language Conflict and Language Planning: The Case of Modern Norwegian. Cambridge, MA: Harvard University Press. Haugen, Einar (1976). The Scandinavian Languages: An introduction to their history. London: Faber and Faber. Hyldgaard-Jensen, Karl (1983). Mittelniederdeutsch und die skandinavischen Sprachen. Cordes & Möhn (Eds.). 666–677. Jahr, Ernst Håkon (1998). Kontakten nordisk – nedertysk i hansatida: mer enn bare lån? Norsklæraren, 1 (1998), 34–42. Larsson, Carl (1931). Ordföljdsstudier över det finita verbet i de nordiska fornspråken I. Mit einem Resumé in deutscher Sprache (= Uppsala universitets årsskrift 1931. Filosofi, språkvetenskap och historiske vetenskaper 1). Uppsala: A.-B. Lundequistska bokhandeln. Lasch, Agathe (1974). Mittelniederdeutsche Grammatik, 2nd ed. Tübingen: Max Niemeyer. Lindow, Wolfgang, Dieter Möhn, Hermann Niebaum, Dieter Stellmacher, Hans Taubken, & Jan Wirrer (1998). Niederdeutsche Grammatik. Bremen: Schuster Leer. Lundeby, Einar (1965). Overbestemt substantiv i norsk og de andre nordiske språk. Mit einer deutschen Zusammenfassung. Oslo: Universitetsforlaget. Lundemo, Frode (ms.). Zur Verwendung von Dativ und Genitiv in mittelniederdeutschen Verwaltungstexten. Halden: Høgskolen i Østfold. Marold, Edith (1981). Mischsprache oder Kontinuum? Die skandinavisch-deutschen Sprachbeziehungen im Mittelalter im Licht der neueren Forschungen zum Sprachenkontakt. In Heinz Rupp & Hans-Gerd Roloff (Eds.), Akten des VI. Internationalen Germanisten-Kongresses Basel 1980 (pp. 142–148). Bern: Peter Lang. Marold, Edith (1984). Überlegungen zur Entwicklung der Substantiv-Flexion in den skandinavischen Sprachen. In Otto Gschwantler et al. (Eds.), Linguistica et philologica. Gedenkschrift für Björn Collinder (1894–1983) (pp. 309–330). Wien: Wilhelm Braumüller. Mogensen, Jens Erik (1999). Review Article. Ulla Björnheden: Zum Vierkasussystem des Mittelniederdeutschen. [...]. Studia Neophilologica, 71, 109–114. Mørck, Endre (1994). The distribution of subject properties and the acquisition of subjecthood in the West Scandinavian languages. In Toril Swan, Endre Mørck, & Olaf Jansen Westvik (Eds.), Language Change and Language Structure. Older Germanic Languages in a Comparative Perspective (= Trends in Linguistics. Studies and Monographs 73) (pp. 159–194). Berlin and New York: Mouton de Gruyter. Mørck, Endre (1998). The development of direct and indirect object as distinct functional categories from Old Norse to Modern Norwegian. Askedal (Ed.). 365–385. Nesse, Agnete (2002). Språkkontakt mellom norsk og tysk i hansatidens Bergen (= Det Norske Videnskaps-Akademi. II. Hist.-Filos. Klasse. Skrifter og avhandlinger Nr. 2). Oslo: Novus. Nettle, Daniel (1999). Linguistic Diversity. Oxford etc.: Oxford University Press. Nichols, Johanna (1992). Linguistic Diversity in Space and Time. Chicago and London: Chicago University Press.

Typological reflections on loss of morphological case

Norde, Muriel (1997a). The history of the genitive in Swedish: A case study in degrammaticalization. Amsterdam: Faculteit der Letteren, Universiteit van Amsterdam. Norde, Muriel (1997b). Middle Low German–Middle Scandinavian language contact and morphological simplification. Multilingua, 16, 389–409. Pettersson, Gertrud (1996). Svenska språket under sjuhundra år: En historia om svenskan och dess utforskande. Lund: Studentlitteratur. Posner, Rebecca (1996). The Romance Languages. Cambridge: Cambridge University Press. Ringgaard, Kristian (1986). Flektionssystemets forenkling og middelnedertysk. Arkiv för nordisk filologi, 101, 173–183. Ronneberger-Sibold, Elke (1997). Typology and the diachronic evolution of German morphosyntax. In Jacek Fisiak (Ed.), Linguistic Reconstruction and Typology (= Trends in Linguistics. Studies and Monographs 96) (pp. 313–335). Berlin and New York: Mouton de Gruyter. Schulte, Michael (2003). Early Nordic language history and modern runology: With particular reference to reduction and prefix loss. Blake & Burridge (Eds.). 391–402. Simensen, Erik (1995). Påverknaden frå mellomnedertysk på norsk i diplom frå det 14. hundreåret. In Ernst Håkon Jahr (Ed.), Nordisk og nedertysk. Språkkontakt og språkutvikling i seinmellomalderen (pp. 55–80). Oslo: Novus. Thomason, Sarah Grey, & Terrence Kaufman (1988). Language Contact, Creolization and Genetic Linguistics. Berkeley, Los Angeles and Oxford: University of California Press. Torp, Arne (1992). Der sogenannte “Garpegenitiv” – Ursprung, Alter und Verbreitung im heutigen Norwegisch. In Lennart Elmevik & Kurt Erich Schöndorf (Eds.), Niederdeutsch in Skandinavien III (= Beihefte zur Zeitschrift für Deutsche Philologie III) (pp. 151–166). Berlin: Erich Schmidt Verlag. Weddige, Hilkert (2003). Mittelhochdeutsch. Eine Einführung. 5th ed. München: C.H. Beck. Weinrich, Harald (1993). Textgrammatik der deutschen Sprache. Unter Mitarbeit von Maria Thurmair, Eva Breindl, Eva-Maria Willkop. Mannheim, Leipzig, Vienna and Zürich: Dudenverlag. Werner, Otmar (1979). Kongruenz wird zu Diskontinuität im Deutschen. In Bela Brogyanyi (Ed.), Studies in Diachronic, Synchronic, and Typological Linguistics: Festschrift für Oswald Szemerényi on the Occasion of his 65th Birthday, Part. II (= Amsterdam Studies in the Theory and History of Linguistic Science. Series 4. Current Issues in Linguistic Theory 11) (pp. 959–988). Amsterdam: John Benjamins. Wessén, Elias (1954). Om det tyska inflytandet på svenskt språk under medeltiden. Stockholm: Svenska bokförlaget. Norstedts. Wessén, Elias (1965). Svensk språkhistoria I: Ljudlära och ordböjningslära. Stockholm, Göteborg and Uppsala: Almqvist & Wiksell.



Ethnoreconstruction in Kok-Papónk Paul Black Charles Darwin University and Waseda University

.

Introduction

Along the western coast of Cape York Peninsula, Australia, Kok-Papónk was a relatively conservative dialect of the same language as Koko-Bera. Accordingly one might hope that it could provide especially useful data for comparative reconstruction because its many cognates with Koko-Bera should make it easier to pinpoint the environments for the changes in the latter, more innovative variety than would be possible on the basis of the fewer cognates the latter shares with more distant relatives. Regrettably, however, Kok-Papónk has largely been attested only from one marginal speaker, and a study of her data suggests that she is likely to have constructed some of the Kok-Papónk forms by modifying Koko-Bera forms in accord with regular sound correspondences between the two varieties, but not necessarily in appropriate environments. For example, she once gave the KokPapónk (KP) form for “eye” as chol, which corresponds to Koko-Bera (KB) chel. The correspondence between KP o < and KB e < *o < *u is quite common, but it is not expected in this particular word, since KB chel should reflect earlier *tyili, not the *tyulu required by KP chol. Accordingly it seems that the informant generated KP chol on the basis of her familiarity with the common sound correspondence, and this is also supported by the fact that on another occasion she gave the Kok-Papónk form as chel, which is the expected reflex of *tyili in both varieties. It thus seems that the informant was generating Kok-Papónk forms in much the same way that a historical linguist might predict possible forms, if without the benefit of detailed reconstruction. Accordingly ‘ethnoreconstruction’ seems quite a suitable term for this phenomenon, and it was in fact



Paul Black

introduced by McClure and McClure (1977) to refer to “a strategy by which a speaker of one language (or dialect) attempts to speak a related language (or dialect) by systematically transforming the elements in the former”. Ethnoreconstruction can be viewed as a particular type of hypercorrection, which may or may not involve such related varieties. It is similar to what Alpher and Nash (1999) have called ‘correspondence mimicry’ in that it is based on some recognition of sound correspondences, but they have used that term to refer specifically to the remodelling of borrowed vocabulary on that basis. Both ethnoreconstruction and correspondence mimicry are interesting for what they reveal about the ability of even unschooled language speakers to apply basic principles of comparative linguistics. Unlike correspondence mimicry and some other types of hypercorrection (e.g. Labov 1966), ethnoreconstruction seems unlikely to produce lasting changes inasmuch as it involves an attempt to provide forms not normally used by the informant. Even so, in a world in which indigenous languages are often attested from people who no longer actively use them, it represents a potential problem in our comparative study of such languages. The evidence for ethnoreconstruction in Kok-Papónk tends to be indirect, based largely on forms that show common sound correspondence in inappropriate environments. This evidence is presented below after a brief introduction to Kok-Papónk and Koko-Bera and their historical development.

. The languages and their historical development Kok-Papónk was once spoken around the mouth of the Nassau River along the western coast of Cape York Peninsula. Aside from four forms in Sharp (1939: 268–273) and a few attested by Alpher (n.d.: 58–59), it is known only from a couple hundred forms and a few sentences gathered independently by Sommer (1972: 70–78, 85) and Black (1978: 78–83) from a single informant who apparently had little opportunity to speak this variety for many years, although she was a fluent speaker of the closely related Koko-Bera language (indigenously called Kok-Kapér). The latter is a better known variety traditionally spoken to the north, along Topsy Creek and in the adjacent area between present day town of Kowanyama and the coast. The Koko-Bera forms cited here are from my own data (as also in e.g. Edwards & Black 1998); I am grateful to Barry Alpher and to Bruce Sommer for kindly allowing me access to their Kok-Papónk data.

Ethnoreconstruction in Kok-Papónk

Like most of the languages of Cape York Peninsula, Koko-Bera and KokPapónk are considered to go back to a putative ‘Proto-Paman’ ancestor first reconstructed by Hale (e.g. 1964, n.d.a, n.d.b, and in Sutton 1976). Since the membership and subgrouping of the so-called Paman group is not well established, it is not clear that all of the protoforms reconstructed by Hale (listed in Sommer 1969: 62–66) and others (e.g. in Sutton 1976; Black 1980) actually go back to a single, earliest common ancestor. At the same time, however, only rarely does evidence of shared innovations suggest any phonological differences among the various protolanguages that may be represented by these protoforms (see Black 2004). The consonants reconstructed for Proto-Paman include a series of stops *p, *t, *ty, *k; a matching series of nasals *m, *n, *ny, *ng; a lateral *l, alveolar *rr, retroflex *r, and the glides *w and *y. Koko-Bera and Kok-Papónk maintain a similar inventory except that they share a split in the laminals *ty and *ny that led to a contrast between lamino-palatals (written as ch and ny) and laminodentals (written th and nh). Both varieties also show the occasional lenition of *p and *k to w and *ty and (in pronouns) *ny to y, but the environmental conditioning for this remains unclear. The vowels of Proto-Paman included *i, *a, and *u, with contrastive length in initial syllables. Through the lowering of *i and *u to e and o respectively in certain environments, Koko-Bera and Kok-Papónk developed a system of five main vowels i, e, a, o, u, with a system of contrastive stress instead of length. Word-final vowels were lost, as were some vowel contrasts in unstressed syllables, where a sixth, schwa vowel is written as v(following Alpher’s (1991) practice for Yir-Yoront) to avoid identifying it with any particular stressed vowel. (Some speakers have unstressed versions of other vowels where others have schwa, as in KB ngvlthórr or ngolthórr “black”, and such variations in spelling should not be regarded as more significant than this.) In Kok-Papónk alone an additional, unrounded central vowel oe was marginally attested before r or rr in just three words. Out of the some 225 attested Kok-Papónk (KP) forms, only ten were are not cognate with Koko-Bera (KB) forms; one is the form found in the indigenous name of each language, where both KP Kok-Papónk and KB Kok-Kapér literally mean “language-proper”. There are also striking differences in pronoun morphology, where KP -nk appears in place of KB -ntvw on first person subjects (e.g. KP ngank, KB ngántvw “I”) and KP -vngk in place of KB -vw or -tvw on third person subjects (e.g. KP púlvngk, KB púlvw “they (dual)”). Otherwise the attested differences between the two varieties are only in phonological development. With regard to consonants, Kok-Papónk retains





Paul Black

*rr before velars, where it was clearly lost in Koko-Bera: e.g. *yirrka- > KP yirrké-y, KB yiké-y “say”. While rr is attested before velars in a few Koko-Bera forms, of which min-yirrkáy “sulphur crested cockatoo” may also involve the root for “say”, some could result from borrowing and others from compounding. A more poorly understood change involves a correspondence between KP w and KB ch or y, which seem to reflect earlier *k before *i in at least KP wéngal (also attested as wérngvl), KB chéngvl “boomerang” < *kiingalV. Another consonantal difference is that Kok-Papónk has a non-homorganic cluster nk while Koko-Bera normally does not (except in one transparent compound, an uncommon plant name), but there are no cognate forms involving this difference. It is in connection with the phonological development of vowels that the problem of ethnoreconstruction becomes apparent. We’ll first consider evidence relating to the fronting of *o (< *u) to KB e and then that for the raising of *a to KB *e, changes which Kok-Papónk did not undergo. Other differences in vowels between the two languages are only marginally attested; these include two KP forms that have u, oe or o before rr or r where Koko-Bera has i or u (KP murrm or moerrm, KB mirrm “house”; KP yorrk, KB yik “large grindstone”; KP thoer, KB thur “bone”; KP móerrkvlvm, KB múkvlvm “an itch”).

. The fronting of *o (< *u) to KB e As noted earlier, Proto-Paman *u was lowered to *o in certain environments, most obviously when the following syllable also contained *u; e.g. *kuyu > KP, KB koy “fish”. In Kok-Papónk this vowel remains o, but in Koko-Bera it was fronted to e in certain environments. However, an attempt to work out the precise environment for the fronting reveals that the evidence provided by the Kok-Papónk forms sometimes disagrees with that of more distant languages, here normally represented only in terms of the resulting Proto-Paman (PP) reconstructions. The most likely explanation seems to be that these and perhaps other forms attested for Kok-Papónk are not actually authentic, but rather created through ethnoreconstruction. Firstly, it seems clear that *u > *o > KB e after the labials *p and *m, as in *pungku > KP pongk, KB pengk (or pengkvchích) “knee” (also in KP pongkvpírpvr, KB pengkvpírpvr “knee cap”); and *ngamurr > KP ngvmórr, KB ngvmérr “armpit”. There is an anomaly in the Kok-Papónk data in that it is attested as being like Koko-Bera in having peth for “ti-tree bark”, where KP *poth

Ethnoreconstruction in Kok-Papónk

would be the expected as a reflex of PP *putyu, but this is presumably due to contamination from Koko-Bera. Aside from this the Kok-Papónk attestation provides fourteen examples of the expected correspondence after p, six after m, and also three after w, where there is otherwise no straightforward evidence (aside from the fact that wo is rare or non-existent in Koko-Bera in word-initial position). However, broader comparative evidence suggests that two of the correspondences involve ethnoreconstruction in Kok-Paponk; these are KP thamól, KB thamél, “foot”, where KP *thamál is the expected reflex of PP *tyamal, and KP pamól, KB pamél, “person (ergative)”, where KP *pamál is the expected reflex of PP *pama-lu. One can wonder which of the other Kok-Papónk forms can be trusted in cases where such broader evidence is lacking. After velars there seems to be good evidence that *o < *u was not fronted to e in either Kok-Papónk or Koko-Bera: e.g. *kuuku > KP, KB kok “language”; and *kungulV > KB kongól “mosquito”. Even so there are a small number of Kok-Papónk forms with o where Koko-Bera has e after k or ng, presumably due to ethnoreconstruction in Kok-Papónk. For example, KP pvngków, KB pvngkéw “brolga” suggests earlier *pangkupV, with KB e < *o < *u even after k, but Black (1980) took such other evidence as Kurtjar ngaabh “brolga” as a basis for reconstructing PP *pangkapV, which should yield KP *pvngkáw. Other correspondences involving k include KP kungkórr thorr, KB kungkarrvthérr “stomach”; KP pa-yingkóngvyvrr, KB pa-yingkéngvyvrr “younger sister”; KP thokórvngk, KB thvkérvngk “stone”; and KP walmvkól, KB walmvkél “flat”, while one involving ng is KP ma-kvngórr, KB ma-kvngérr “black plum”. With regard to apicals, there is one Proto-Paman reconstruction that suggests that *u > *o > KB e after *rr, namely *wurrukV > KP wvrrók, KB wvrrék (cf. Kurtjar worrok) “long ago”. Aside from this, there are Kok-Papónk forms that suggest that it occurred after all apicals: five each after *rr and *l and one each after *t and *n. Again, however, it is not clear to what extent this evidence can be trusted. In one case, if KP mvrróng and KB mvrréng “cold” are cognate with Kurtjar moerr “cold”, they would be expected to reflect PP *murra with the addition of a suffix *-ngV, but in this case the expected KP reflex would be *mvrráng. With regard to laminals, KB ngvlthórr “black” < *ngultyurr (cf. Kurtjar lthoorgh “black”) provides evidence that the fronting did not occur after *ty. A number of correspondences with Kok-Papónk forms suggest that it did, but at least one is clearly due to ethnoreconstruction, namely KP pithórr, KB pithérr “dream”, since KP *pithárr is the expected reflex of PP *pityarrV. (The KokPapónk form is also attested as pithérr, presumably due to contamination from





Paul Black

Koko-Bera.) Some of the other Kok-Papónk forms are unlikely because they involve the ch reflex of *ty, which is not otherwise known to have developed before *u. Most obviously KP chol “eye” surely results from ethnoreconstruction because the expected KP (as well as KB) chel < *tyili is also attested. The other instances of the correspondence after ch include KP chornt, KB (yok) chernt “guttapercha”; KP kolchóm, KB kvlchém “dilly bag”; KP kvrrvchól, KB kvrrvchél “navel”; KP marvchól, KB marvchél “fingernail”; KP panh pinchól, KB panh pinchél “sick”; and KP pichólngk, KB pichélngk “skin”. Instances of the correspondence after th include KP mvthórr, KB mvthérr “a sore”; KP pa-kvnthórr, KB pa-kvnthérr “children”; KP pvthónt, KB pvthént “(he) ate (it)”; KP thorr, KB therr “arm”; KP wvthól, KB wvthél “cook it”; and the second element of KP chol/chel kvthóm(p), KB chel kvthém “eye”. If fronting did not actually occur after *ty then all of those Kok-Papónk forms would seem to be spurious, due to ethnoreconstruction. With regard to the laminal *ny, on the other hand, evidence from the pronoun KP nhólvngk, KB yélvw “he, she” < *nyulV suggests *o < *u was indeed fronted to KB e, but reconstruction of such pronouns is not straightforward. The fronting is also supported by KP yon, KB yen “you” and by KP yont, KB yent “yours”, which presumably also reflect a root *nyun-, but it is not clear why the *ny should yield KP *y here but *nh in the third person pronoun. The only other such correspondences after *ny is the purposive form KP minyów, KB minyéw “for meat”. Here KP *minyáw might be expected as a reflex of PP *minya-ku, but conceivably the o in the attested KP minyów could be due to analogy with such other forms as KP koyów, KB koyéw “for fish” < PP *kuyuku rather than to ethnoreconstruction. The latter set of forms seem to provide evidence for the fronting of *o to KB e after the laminal *y, but here the KB e could similarly be due to analogy with other such forms as KB mayéw, KP mayéw or mayów “for food” < PP *mayi-ku. Thus the Kok-Papónk attestation has not enabled us to clarify the environment for the fronting of *o to KB e, but instead it simply raises problems. Conceivably some of these could be due to the environmental conditioning being more complex than first thought. However, such forms as KP pamól “person (ergative)” provide good evidence for ethnoreconstruction (barring an otherwise unattested development of earlier *a into KP o, in this case in PP *pama-lu), and the remaining problems could simply result from the same phenomenon.

Ethnoreconstruction in Kok-Papónk

. The raising of *a to KB e Earlier *a is reflected as KP a but KB e in certain environments, as in PP *kutaka > KP kutáw but KB kutéw “dog”. The correspondence between KP a and KB e can be seen in about two dozen cognate sets, and from about two-thirds of these it would seem that a major environment for the change involved the vowel u in the preceding syllable; other examples would include KP pulán(t)hvrr, KB pulén(t)hvrr “two” and KP pulánvngk, KB pulénvngk “their (dual)”, both of which could contain reflexes of PP *pula, and KP yurránt, KB yurrént “yours (plural)”, which should contain reflexes of PP *nyurra. However, there are also Koko-Bera forms which show a < *a rather than e in such an environment, such as KB pukán “grass” < *pukan. In addition, two of the Kok-Papónk forms are anomalous and thus may well involve ethnoreconstruction. One is KP kunthán, KB kunthén “tail”, presumably compounds of kun < *kuna (generally referring to faeces or intestines) followed by reflexes of PP *tyu(u)n, whose expected Kok-Papónk reflex thon is also attested in the same meaning. The other is KP pa-pumáriyvrr, KB pa-pumérriyvrr “father’s sister” if these include reflexes of what has been reconstructed as a PP root *piimur even though stress does not fall on the attested forms as required by the long vowel in that reconstruction. This can make one wonder if the Kok-Papónk informant simply constructed other forms to have the pattern uCá (with C representing any consonant or cluster) where Koko-Bera has uCé. Other such cognate sets includes KP man ngurrchám, KB ngurrchém “mourning scarf ”; KP ming-kunchám, KB ming-kunchém “beef, flesh”; KP ming-kutál, KB ming-kutél “fish hawk”; KP murámuriny, KB murémuriny “smooth”; KP ngumár, KB mi-ngumér “fat, grease”; KP purrmán, KB purrmén “blood”; KP purrmápurrmvn, KB purrmépurrmen “red”; KP purrpáng, KB purrpéng “soft”; KP thunpány, KB thunpény “narrow”; and KP yok pulpál, KB yok pulpél “cotton tree”. Whether or not a first-syllable u might sometimes have been involved as a conditioning factor, some other environment is needed to account for the remaining evidence. The same correspondence can be seen in initial syllables in KP wat, KB wet “bad”; KP wátvwvt, KB wétvwvt “woman”; probably KP nhar (also attested as nhor), KB yer “this”; and perhaps KP wanth (also attested as wonth), KB wenth “crooked”. It can also be seen in non-initial syllables that are not preceded by a syllable containing u, as in KP ngarpánchvl, KB ngarpénchvl “sweat”; KP thvnánvngk, KB thvnénvngk “their”; and KP wvlkány, KB wvlkény “lie down”.





Paul Black

The raising of *a to KB e also led to a morphophonemic alternation between a and e before different suffixes in one class of Koko-Bera verbs, as in KB yiká-wvny “was saying” but yiké-m “says” and yiké-y “say (imperative)” < *yirrka-. In these cases it would seem that the critical aspect of the environment was something after the *a. Perhaps the change depended on the vowel in the following syllable, which tends to be lost or reduced to schwa in both of these varieties. Such forms as KP rrungáy, KB rrungéy “dance” and KP kuncháy, KB kunchéy “run” seem to confirm that Kok-Papónk did not undergo this raising of *a to e, but some forms with u in the preceding syllable seem to be due to ethnoreconstruction, as discussed earlier. Kok-Papónk appears to be like KokoBera in having e < *a in the verb KP yirrké-y, KB yiké-y “say” < *yirrka, but this could simply be due to contamination from Koko-Bera.

. Conclusion The Kok-Papónk evidence relating to the fronting and raising of vowels in Koko-Bera thus sometimes conflicts with the evidence from more distant languages. Since some of the problematic forms clearly seem to be due to ethnoreconstruction in Kok-Papónk, it seems reasonable to suppose that this might be behind the other problems as well. From the evidence cited in those sections it should also be apparent that the informant gave two versions of some Kok-Papónk forms, in suggests that one version is incorrect, whether due to ethnoreconstruction or to contamination from Koko-Bera. The Kok-Papónk data thus tends to illustrate the ability of even unschooled speakers of indigenous languages to recognise sound correspondences between related varieties, but regrettably it contributes little to our evidence for comparative reconstruction. This is not to say that the Kok-Papónk data has no value whatsoever. On those occasions where its forms are unlike those of Koko-Bera and yet can not be predicted on the basis of common sound correspondences, we can hope that they accurately represent the language. For example, it is clear that Kok-Papónk pronouns must have had quite different affixes from those in the corresponding Koko-Bera forms, although it is less clear that they are entirely free from contamination and ethnoreconstruction.

Ethnoreconstruction in Kok-Papónk

References Alpher, Barry (n.d.). “Fieldbook 2 PC”. Manuscript held by the Australian Institute of Aboriginal and Torres Strait Islander Studies, Canberra. Alpher, Barry. (1991). Yir-Yoront Lexicon: Sketch and Dictionary of an Australian Language. Berlin: Mouton de Gruyter. Alpher, Barry & David Nash (1999). “Lexical Replacement and Cognate Equilibrium in Australia”. Australian Journal of Linguistics, 19, 5–56. Black, Paul (1974–1978). “Fieldbooks”, Vol. 20. Manuscript held by the Australian Institute for Aboriginal and Torres Strait Islander Studies, Canberra. Black, Paul (1980). “Norman Pama Historical Phonology”. In B. Rigsby & P. Sutton (Eds.), Papers in Australian Linguistics No. 13: Contributions to Australian Linguistics (181– 239). (= Pacific Linguistics, A-59.) Canberra: Australian National University. Black, Paul (2004). “The Failure of the Evidence of Shared Innovations in Cape York Peninsula”. In Claire Bowern & Harold Koch (Eds.), Australian Languages: Classification and Comparative Method (pp. 241–267). Amsterdam: Jojn Benjamins. Edwards, Shaun Kalk & Paul Black (1998). “Kokoberrin Dictionary”. Manuscript compiled for the Kokoberrin Tribal Aboriginal Corporation, Cairns. Hale, Kenneth (n.d.a). “Attestation”. Manuscript. Hale, Kenneth (n.d.b). “Other Paman Languages”. Manuscript. Hale, Kenneth (1964). “Classification of Northern Paman Languages, Cape York Peninsula, Australia: A Research Report”. Oceanic Linguistics, 3 (2), 248–265. Labov, William (1966). “Hypercorrection by the Lower Middle Class as a Factor in Linguistic Change”. In William Bright (Ed.), Sociolinguistics (pp. 84–102). The Hague: Mouton. McClure, Erica F. & Malcolm McClure (1977). “Ethnoreconstruction”. Anthropological Linguistics, 19 (3), 104–110. Sharp, R. Lauriston (1939). “Tribes and Totemism in North-east Australia”. Oceania, 9, 254– 275, 439–461. Sommer, B. A. (1969). Kunjen Phonology: Synchronic and Diachronic (= Pacific Linguistics B-11.). Canberra: Australian National University. Sommer, Bruce (1972). “[1971-2 Field Book] #4”. Manuscript held by the Australian Institute for Aboriginal and Torres Strait Islander Studies, Canberra. Sutton, Peter (Ed.). (1976). Languages of Cape York: Papers Presented to a Symposium Organised by the Australian Institute of Aboriginal Studies. Canberra: Australian Institute of Aboriginal Studies.



Raising verbs vs. auxiliaries Kasper Boye University of Copenhagen

.

Introduction1

It is widely held that auxiliaries should be analyzed as raising main or full verbs. The analysis goes back to Ross (1969) and is currently maintained, for instance, within the framework of HPSG (cf. e.g. Borsley 1996: 140–143). The present paper argues against this analysis. It aims at showing that both synchronically and diachronically, auxiliaries should be regarded as distinct from raising verbs. Synchronically, I argue, auxiliaries are grammaticalized verbs while raising verbs are lexical full verbs occurring in a peculiar construction. They are formally distinguishable in that auxiliaries meet criteria of grammaticalization and auxiliaryhood while raising verbs do not. Diachronically, I argue, auxiliaries and raising verbs are the result of two different processes. Auxiliaries are the result of a grammaticalization (in the sense of Meillet 1921 and Kurylowicz 1965) involving, probably, reanalysis as an initial step. Raising verbs, on the other hand, are the result of a syntactic reanalysis (in the sense of Langacker 1977) not in itself a matter of grammaticalization. However, the former process may take the output of the latter as its input; auxiliaries have raising verbs as one of their diachronic sources. The distinction between auxiliaries and raising verbs is a general one. That is, when one randomly chosen auxiliary is compared with one randomly chosen raising verb, the two may be distinguished synchronically and diachronically along the lines described above. However, in my argumentation I focus on a number of Danish verbs that seem to have both an auxiliary variant and a raising-verb variant. These verbs occur in a special syntactic construction and belong to four semantic groups. In Section 2, I introduce the construction and the verbs I focus on. In Section 3, I argue that some of the verbs that occur in this construction have two



Kasper Boye

variants. In Section 4, I present synchronic evidence that the distinction between these two variants corresponds to a distinction between raising full verbs and grammaticalized auxiliaries. In Section 5, I sketch hypothetical developmental paths for each of the two variants of the verbs I focus on in particular, and for auxiliaries and raising verbs in general. Finally, in Section 6, I present my conclusion.

. The Danish at-nci-nexus constructions The construction I focus on is exemplified in (1)–(3). It involves an infinitival predication, marked by italics, and a finite verb. Following the terminology of Jespersen 1924, I shall classify the infinitival predication as an infinitival nexus. More precisely, in order to distinguish the infinitival predication in (1)–(3) from other related types of infinitival predication, I shall classify it as an at (‘to’)-nci (‘nominativus cum infinitivo’) nexus (cf. Boye 2002). (1) Han synes at være død. He seems to be dead. “He seems to be dead”. (2) Han siges/menes at være død. He say-pass/think-pass to be dead. “He is said/thought to be dead”. (3) Han ses at være død.2 He see-pass to be dead. “He is seen to be dead”.

The verbs I focus on are the verbs that are capable of occurring in construction with at-nci nexus. These verbs belong to four semantic groups. The four groups are: Appearance verbs such as synes ‘seem’ in (1) (and additionally, for instance, forekomme ‘appear’, lade til ‘appear’ (lit.: ‘let to’), lyde til ‘sound’ (lit.: ‘sound to’), se ud til ‘look, seem’ (lit.: ‘see out to’)), Utterance verbs such as sige ‘say’ in (2) (and additionally, for instance, hævde ‘claim’, rapportere ‘report’), Attitude verbs such as mene ‘think’ also in (2) (and additionally, for instance, antage ‘assume’, frygte ‘fear’) and (perhaps marginally, cf. Note 2) Perception verbs such as se ‘see’ in (3) (cf. Boye 2002 for a comprehensive list of verbs occurring in construction with at-nci nexus). When verbs from these four groups are constructed with at-nci nexus, most of them (viz. the Utterance, Attitude and Perception verbs) appear formally to be passive transitive verbs.

Raising verbs vs. auxiliaries

The Appearance verbs (including deponent synes), however, appear formally to be active intransitive verbs. Henceforth, I shall refer to verbs belonging to these four groups in short as APUA (Appearance, Perception, Utterance and Attitude) verbs.

. Predicative vs. non-predicative verbs Some APUA verbs have two variants. This is the case when they occur in construction with finite predications (cf. e.g. Urmson 1952, Benveniste 1971 and Thompson & Mulac 1991 on ‘parenthetical verbs’, Hooper 1975 on weak assertive predicates, and Boye & Harder forthc., for a detailed account), but also when they occur in construction with at-nci nexus. The two variants may be distinguished semantically as well as syntacticallydistributionally. Semantically, a distinction can be made between an inherently addressable and an inherently non-addressable variant. Consider synes ‘seem’ in (4) as an example. (4) Han synes at være død. He seems to be dead. “He seems to be dead”.

The inherently addressable variant of synes can be (but may not be) addressed by the utterance of (5). (5) Er det virkelig rigtigt? “Is that really so?”

In contrast, the inherently non-addressable variant of synes cannot be addressed by the utterance of (5). If (4) involves inherently non-addressable synes, only the at-nci nexus han at være død ‘he to be dead’ can be addressed by the utterance of (5). Inherently non-addressable synes merely serves to ‘modify’ the at-nci nexus. Syntactically-distributionally, a distinction can be made between a variant that allows of and a variant that does not allow of neg anteposition (or neg raising) and the more general Danish phenomenon of adverbial anteposition (or adverbial raising) (cf. Boye 2001 and Hansen 2000; I refer to these phenomena with the term anteposition, suggested by Hansen 2000, rather than raising in order to stress the distinction between them and the phenomenon of raising verbs. Consider the APUA verbs in (6)–(8) as examples. The negation ikke





Kasper Boye

‘not’ may be read in two ways. It may be read as non-anteposed, i.e. with wide scope. That is, it may be read in accordance with its position which links it semantically to the finite APUA verb (reading a). Alternatively, it may be read as anteposed, i.e. with a narrow scope. That is, in spite of its position, it may be read as semantically3 linked to the at-nci nexus (reading b).4 (6) ...som ikke synes at være død. ...who not seems to be dead. a. “...who [doesn’t seem] to be dead”. b. “...who seems not to be dead”. (7) ...som ikke siges/menes at være død. ...who not say-pass/think-pass to be dead. a. “...who [isn’t said/thought] to be dead”. b. “...who is said/thought not to be dead”. (8) Han lader ikke til at være død. He lets not to to be dead. a. “He [doesn’t appear] to be dead”. b. “He appears not to be dead”.

The reading of the negation in (6)–(8) as anteposed or non-anteposed is bound up with the addressability of the involved APUA verb. In construction with an anteposed adverbial, an APUA verb cannot be addressed (outside metalinguistic contexts) – it is inherently non-addressable. Thus, in the b-readings of (6)–(8), only the at-nci nexus han at være død ‘he to be dead’ can be addressed by the utterance of (5). In the a-readings of (6)–(8), on the other hand, the negation ‘profiles’ or ‘foregrounds’ the APUA verbs and makes them highly susceptible of being addressed by the utterance of (5). In sum, then, a number of APUA verbs occurring in construction with at-nci nexus have two semantically and syntactically-distributionally distinguishable variants. What I shall refer to as the predicative variant is asserted as part of a predication and does not allow of adverbial anteposition. What I shall refer to as the nonpredicative variant is not asserted as part of a predication and does allow of adverbial anteposition. Not all APUA verbs occurring with at-nci nexus have a non-predicative variant. This is the case, for instance, with the Attitude verb vide ‘know.’ With some APUA verbs, moreover, a predicative variant seems marginal. This is the case, for instance, with the Perception verb se ‘see’. (9) ...som ikke ses at være død. ...who not see-pass to be dead.

Raising verbs vs. auxiliaries

a. “...who [isn’t seen] to be dead”. b. “...who is seen not to be dead”.

It may be hard to maintain a non-predicative reading of se in (9). Accordingly, it may be hard to read the adverb ikke as anteposed as in reading b.

. Raising verbs vs. auxiliaries – a synchronic analysis I shall now present synchronic evidence that the distinction between predicative and non-predicative variants of APUA verbs occurring in construction with at-nci nexus corresponds to a distinction between raising full verbs and grammaticalized auxiliaries, respectively. . Raising verbs Let me first consider the predicative APUA variants. The analysis of these variants as full verbs follows from my argumentation later on for the analysis of the non-predicative variants as grammaticalized verbs. Being full verbs, the predicative variants of APUA verbs occurring in construction with at-nci nexus must be analyzed as raising verbs. The raising-verb analysis may be conceived of as simply covering a divergence between the ‘expression syntax’ and the ‘content syntax’ (cf. Harder 1996) of the at-nci nexus. Syntactically and morphologically, as it appears from (1)–(3), the primary argument of the infinitival predicator of the at-nci nexus is expressed as the sentence subject: it is expressed in either topic position or subject position (cf. Diderichsen 1962), and if it is expressed by a pronoun, this pronoun appears in the nominative case. Semantically, however, the at-nci nexus is a unit the constituents of which have no external semantic predicational relations on their own, a unit constituting itself the embedded primary argument of the APUA full verb. We may represent this analysis as in (10). Here, synes ‘seem’ is the predicator in a predication that functions as a main clause. The primary (and only) argument of synes is the at-nci nexus han at spille fodbold ‘he to play football’ which functions as an embedded argument clause. The brackets indicate that there are no semantic predicational relations between synes and individual elements of the at-nci nexus, only between synes and the at-nci nexus as a whole. (10) Han synes at spille fodbold. synes-RAISING VERB [han at spille fodbold]-ARGUMENT CLAUSE seem-RAISING VERB [he to play football]-ARGUMENT CLAUSE





Kasper Boye

There are at least three arguments for this semantic analysis of the at-nci nexus – and thus for the analysis of APUA full verbs as raising verbs (cf. Boye 2002). 1) The infinitival predicate (i.e. the infinitival predicator plus non-primary arguments: at være død ‘to be dead’) cannot be brought into focus in a cleft or pseudo-cleft construction independently as in (11) (cf. Rosenbaum 1967:59). (11) *Hvad han synes er at være død. “What he seems is to be dead”.

This suggests that the infinitival predicate at være død ‘to be dead’ is not a separate argument of the APUA verb. 2) The infinitival predicate cannot be omitted as in (12). (12) *Han synes. “He seems”.

This suggests that the primary argument han ‘he’ of the infinitival predicator is not a separate argument of the APUA verb. 3) The primary argument of the infinitival predicate may be incompatible with the APUA verb. In (13), for instance, the primary argument, han ‘he’, of the infinitival predicator is semantically incompatible with the APUA verb menes which can only take a predicational argument. In (14), in fact, the infinitival predicator regne ‘rain’ has no argument at all, but is constructed with the dummy det which is incompatible with all other verbs – including APUA verbs – than weather verbs. (13) Han menes at være død. He think-pass to be dead. “He is thought to be dead”. (14) Det synes at regne. It seems to rain. “It seems to rain”.

This again suggests that the primary argument of the infinitival predicator is not a separate argument of the (finite) APUA verb. . Auxiliaries Let me now consider the non-predicative APUA variants. Below, I present arguments for the analysis of these variants as grammaticalized verbs. Before I do so, however, three things should be noted. First, if the non-predicative vari-

Raising verbs vs. auxiliaries

ants of APUA verbs occurring in construction with at-nci nexus as in (1)–(3) are grammatical rather than full lexical verbs, they must be grammaticalized as auxiliaries. Like typical auxiliaries they are accompanied by infinitives and they may appear as the only finite verb in a predication. Second, if non-predicative APUA variants are grammaticalized, they are weakly grammaticalized; they are not morphologically distinguishable from their non-grammaticalized counterparts, the predicative variants. Third, some APUA verbs, as mentioned, have a non-predicative variant also when they occur in construction with a finite predication (for instance in jeg mener han er død ‘I think he is dead’). The arguments for the analysis of non-predicative APUA variants as grammaticalized apply also with respect to this construction. In this construction, however, non-predicative, grammaticalized APUA variants are not auxiliaries. They are not accompanied by infinitives. Rather, they must be analyzed as being part of grammaticalized particles or adverbials (cf. Thompson & Mulac 1991). That is, in jeg mener han er død, non-predicative mene ‘think’, if indeed grammaticalized, must be analyzed as being part of the particle or adverbial jeg mener ‘I think’ (cf. Boye & Harder forthc.). I now return to the matter at hand, i.e. the arguments for the analysis of the non-predicative variants of APUA verbs as grammaticalized items. Arguing that non-predicative variants of APUA verbs occurring in construction with atnci nexus are grammaticalized, amounts to arguing that they are auxiliaries (cf. above). We may represent the auxiliary analysis as in (15). Here, synes ‘seem’ is an auxiliary, and the at-nci nexus functions as a main clause. The brackets indicate that there are no semantic predicational relations between synes and individual elements of the at-nci nexus. (15) Han synes at spille fodbold. synes-AUX [han at spille fodbold]-MAIN CLAUSE seem-AUX [he to play football]-MAIN CLAUSE

The auxiliary analysis represented in (15) contrasts with the raising analysis represented in (10) in two respects. First, the at-nci nexus is not an embedded argument clause, but a main clause. Second, the APUA verb is not a predicator, but a ‘modifier’ or ‘operator’. My basic argument for the analysis of non-predicative APUA variants as grammaticalized verbs is as follows: the facts that distinguish non-predicative APUA variants from predicative ones, viz. the inherent non-addressability and the allowance of adverbial anteposition, are symptoms of grammaticalized verbs. Accordingly, the facts that distinguish predicative APUA variants from





Kasper Boye

non-predicative ones, viz. the inherent addressability and the disallowance of adverbial anteposition, are symptoms of full verbs. 1) As regards inherent addressability and non-addressability, these two features distinguish lexical items from grammaticalized items in general. With a lexical item, it is always possible to come up with an utterance in which the item can be addressed. With a grammatical item this is never possible (outside metalinguistic contexts). Auxiliaries, clitics, affixes and so on can normally never be addressed in discourse. It follows that inherently non-addressable APUA variants are grammaticalized. 2) As regards allowance and disallowance of adverbial anteposition, an adverbial positionally attached to a verb is, under normal circumstances, also semantically attached to this verb or the predication or predicate in which the verb is a predicator. This holds for adverbials positionally attached to all sorts of full verbs, and it holds for adverbials positionally attached to predicative APUA variants, as we have seen in Section 3. Thus, these variants must be full verbs. With non-predicative APUA variants, on the other hand, an adverbial positionally attached to the APUA verb is not semantically attached to it. Thus, non-predicative APUA variants must be grammaticalized. In other words, the reason why an adverbial can be positionally attached to a non-predicative APUA variant without being semantically attached is that the non-predicative APUA variant has lost its full-verbal status and thus its capability of being adverbially modified. The reason why an adverbial can be positionally attached to a non-predicative APUA variant at all probably resides in the fact that nonpredicative APUA variants are relatively weakly grammaticalized (cf. above) and thus still formally have a slot vacant for adverbials. The analysis of non-predicative APUA variants as auxiliaries finds additional support in four facts (cf. Boye 2001). 1) The grammaticalization of APUA verbs, in accordance with one of Hopper’s (1991) principles of grammaticalization, exhibits a functional divergence between etymologically related items. The auxiliary variants exist side by side with the full-verbal variants. 2) The grammaticalization of APUA verbs, in accordance with one of Hopper’s (1991) principles of grammaticalization, seems to involve functional specialization of a part of a bigger lexical paradigm. In Section 3. I argued that not all APUA verbs have a non-predicative variant and that with some APUA verbs a non-predicative variant seems marginal. This amounts to saying that APUA verbs do not all have a grammaticalized variant and that the grammaticalized variants are not grammaticalized to the same extent.

Raising verbs vs. auxiliaries

3) The grammaticalization of APUA verbs follows a cross-linguistically frequently found path of grammaticalization. Non-predicative, auxiliary APUA variants are evidentials (cf. Boye 2001, and cf. e.g. Östman 1981, Thompson 2002 and Boye & Harder forthc., for a related view of (grammaticalized) APUA verbs in construction with finite predications). The at-nci nexus expresses a proposition (cf. Vendler’s (1967) fact and Lyons’s (1977) third-order entity) which has a truth-value and represents a ‘possible fact in the world’, rather than a state-of-affairs (cf. Vendler’s (1967) event and Lyons’s (1977) secondorder entity) which represents a ‘situation in the world’, but has no truth value (cf. Boye 2002). The auxiliary APUA variants all ‘modify’ this proposition evidentially. They express the three general kinds of evidential meaning found by Willett (1988) to be coded cross-linguistically: the Appearance verb in (1) signals indirect, inferential evidence, the Utterance and Attitude verbs in (2) signal indirect, reported evidence, and the Perception verb in (3) signals direct, visual evidence. Cross-linguistically, grammatical evidentials often evolve from Perception, Attitude, Utterance and Appearance verbs (Willett 1988:61; Aikhenvald 2003: 21) – and occasionally from passives, as in Western Apache (de Reuse 2003). 4) The grammaticalization of APUA verbs, in accordance with one of Hopper’s (1991) principles of grammaticalization, involves the persistence of remnants of the lexical meaning of APUA verbs. The evidential meaning expressed by the auxiliary variants is implied by the full-verbal variants. For instance, to state that somebody is said or thought to be dead as in (2) clearly implies that there is reported evidence for the truth of the proposition that he is dead. To sum up the analyses, a number of APUA verbs occurring in construction with at-nci nexus have two variants: a full-verbal variant appearing as a raising verb, and a grammaticalized variant appearing as an auxiliary.

. Raising verbs vs. auxiliaries – a diachronic hypothesis Having established the synchronic distinction between raising verbs and auxiliaries, we are now in a position to consider this distinction as the result of two different diachronic processes. I shall first consider the particular development of Danish APUA raising full verbs and Danish APUA auxiliaries. On the one hand, with most APUA verbs, the full-verbal raising construction goes back to Old Danish. Here, the construction of APUA verbs with at-nci nexus was introduced as a translation of the similar Latin construction (cf. Boye 2002). There are two arguments





Kasper Boye

that the construction was actually introduced as a raising construction (cf. remarks made by the Danish grammarian Mikkelsen (1911:112)). First, the occurrence of at-nci nexus in construction with passive Attitude verbs such as menes ‘think-pass’ indicates that no other analysis is appropriate. To the best of my knowledge, passive Attitude verbs all the way back to Common Nordic have only been capable of taking predicational arguments. This suggests that the atnci nexus has always been a separate predicational unit. Second, Bolkestein (1979) acknowledges a raising analysis of the Latin nci-nexus construction (even though she argues against a similar analysis of the Latin accusativus cum infinitivo nexus construction). As mentioned, this Latin construction was the template of the Danish one. Danish APUA raising verbs, then, are rather old and seem to be the result of revolution rather than evolution. On the other hand, the APUA auxiliaries seem to be of a recent date. Of course, the fact that the APUA auxiliaries are rather weakly grammaticalized (cf. Section 4) does not necessarily imply that they are young. However, the young age of APUA auxiliaries is indicated by the fact that the phenomenon of adverbial anteposition seems to be of a recent date. In all the instances that I have found of an APUA verb occurring in construction with a negated predication in Early Modern Danish, the negation is positionally attached to this predication and not to the APUA verb. In negative finite predications functioning as subordinate clauses the negation takes a position within the finite predication. Thus, Danish APUA raising verbs seem to be considerably older than Danish APUA auxiliaries. Since the two types of verbs are not formally distinguishable (the sentences in (1)–(3), for instance, may involve either), the raising verbs seem to be obvious candidates for being the diachronic sources of the corresponding auxiliaries. Indeed, there are no other candidates. This suggests the development in (16), linking (10) to (15).5 (16) Han synes at spille fodbold. synes-RAISING VERB [han at spille fodbold]-ARGUMENT CLAUSE seem-RAISING VERB [he to play football]-ARGUMENT CLAUSE ↓ grammaticalization (after Old Danish) synes-AUX [han at spille fodbold]-MAIN CLAUSE seem-AUX [he to play football]-MAIN CLAUSE

As argued in Boye & Harder forthc., this grammaticalization process probably involves an initial proliferation of the use of the APUA raising verb with ‘unprofiled’ or ‘backgrounded’ discourse status, and a subsequent structural codification of this status as grammatical.

Raising verbs vs. auxiliaries

We can now, in general, consider the distinction between raising verbs and auxiliaries as the result of two different diachronic processes. On the one hand, the development of the Danish at-nci-nexus construction suggests that auxiliaries may develop from raising verbs through the process of grammaticalization, as sketched in (17). (17) RAISING VERB + ARGUMENT CLAUSE ↓ grammaticalization AUXILIARY + MAIN CLAUSE

This process involves many of the characteristics of weak grammaticalization: change of the original lexical content, slight decategorialization, onset of paradigmatization, etc., and all the characteristics of grammaticalization of full verbs into auxiliaries: loss of predicative function, reanalysis of the embedded argument clause into a main clause, etc. The initial part of this process, of course, may be conceived of as a special case of reanalysis: ‘auxiliarization’ involves the reanalysis of complement-taking lexical verbs into noncomplement-taking auxiliary operators, and thus the reanalysis of argumentassigning predicators into non-argument-assigning ‘modifiers’. On the other hand, raising verbs, I think, are normally the result of nongrammaticalization syntactic reanalysis. This is, of course, not the case with most Danish APUA verbs. As mentioned, the construction of most APUA verbs as raising full verbs with at-nci nexus was not the result of languageinternal evolution. Rather, it was introduced into Old Danish by translation of Latin raising constructions; its appearance must be considered a contact phenomenon. Clearly, however, even if the Latin raising verbs were not the result of language-internal evolution either, at some point raising verbs must be the result of a language-internal evolutionary process, a non-grammaticalization syntactic reanalysis. In the case of a few Danish APUA verbs, moreover, one might actually suspect such a process. The verb se ud til ‘look, seem’ (lit.: ‘see out to’) is a case in point. According to my data collection, this verb was not among the verbs appearing in construction with at-nci nexus in Old Danish. It seems, then, that the construction of this verb with at-nci nexus, exemplified in (18), has evolved later, probably as an analogy to constructions such as those in (1)–(3). (18) Jon ser ud til at være irriteret/borte. Jon sees out to to be annoyed/gone. “Jon seems to be annoyed/gone”.





Kasper Boye

There are indications that se ud til in this construction was not originally a raising verb. On the one hand, (18) clearly demonstrates the availability of the reading of se ud til as a raising verb. First, a reading must be available where there is no semantic verb-argument relation between se ud til and Jon. Otherwise, the example would be unacceptable on the grounds that Jon cannot do any ‘seeming’ if he is gone. Second, this reading is compatible with se ud til being addressable. Thus, the acceptability of (18) does not hinge on a reading of se ud til as an auxiliary. On the other hand, there is evidence for the remnants of a possible reading of se ud til as a control verb. As a consequence of the fact that raising verbs have no semantic argument relations to the individual elements of the predication they co-occur with, they are not permitted in pseudoclefts (cf. Section 4.1). However, Eide (2002: 157) claims that the Norwegian cognate of se ud til, se ut til, under one condition is “marginally permitted” in a pseudocleft. (19)

?Det Jon ser ut til, er å være irritert

(Eide 2002: 157). “What Jon seems (to be), is annoyed”.

(20) *Det Jon ser ut til, er å være borte (Eide 2002: 157). “What Jon seems (to be), is gone”.

Obviously, the condition is that the referent of the primary argument of the infinitival predicate should be visible. In (19) it is, but in (20) it is not: Jon is gone. This amounts to saying that (19) and (20) demonstrate the existence of a remnant of a semantic verb argument-relation between se ut til and the primary argument of the infinitival predicate. In other words, (19) can only be accounted for if a reading of se ut til as a control verb is recognized. This reading, however, can only be a remnant. The reading of se ut til as a raising verb (or auxiliary) must be dominant with respect to availability – otherwise (19) would be fully and not only marginally acceptable. I do not wish to go into acceptability judgements of the Danish counterpart of (19). The Danish and Norwegian languages are so closely related that the developmental scenario in (21) may be posited as a hypothesis. (21) Han ser ud til at være irriteret. se ud til-CONTROL VERB [han]-ARG.1 [at være irriteret]-ARG.2 seem-CONTROL VERB [he]-ARG.1 [to be annoyed]-ARG.2 ↓ non-grammaticalization syntactic reanalysis se ud til-RAISING VERB [han at være irriteret]-ARG. CLAUSE seem-RAISING VERB [he to be annoyed]-ARGUMENT CLAUSE

(21), in turn, may be generalized to (22).

Raising verbs vs. auxiliaries

(22) CONTROL VERB + ENTITY-ARG. + PREDICATE-ARG. ↓ non-grammaticalization syntactic reanalysis RAISING VERB + ARGUMENT CLAUSE

The hypothesis sketched in (22) finds support in remarks made by Jespersen (1924: 118) and Barron (1999). It should be noted, however, that control verbs may not be the only possible origin of raising verbs and that nongrammaticalization syntactic reanalysis is not necessarily a one-step process (cf. Barron 1999 for a discussion of the “continuum between fully control and fully raising verbs”). The hypothesis in (22) involves no diachronic process of literally raising. Rather, raising verbs are seen as the result of a syntactic reanalysis, i.e. a reanalysis preserving the expression side of linguistic items, but changing their content-side relations. More specifically, they are seen as the result of a nongrammaticalization syntactic reanalysis. In contrast to the process resulting in auxiliaries, the process resulting in raising verbs maintains the complementtaking and argument-assigning status – and thus the lexical status – of verbal predicators. However, it changes the argument structure of these verbal predicators.

. Conclusion In this paper I have argued for a distinction between raising verbs and auxiliaries. I have argued that synchronically, raising verbs and auxiliaries are formally distinguishable as full verbs and grammaticalized verbs, respectively, and that diachronically, they are distinguishable as the results of nongrammaticalization syntactic reanalysis and grammaticalization, respectively. My argumentation has been based on a Danish case study. The subject of this study is the construction of the Danish predication referred to as the at-nci nexus with verbs referred to as APUA verbs. In a synchronic analysis of these verbs, I have tried to show that in construction with at-nci nexus they occur in two variants: a full verbal variant appearing as a raising verb and a grammaticalized variant appearing as an auxiliary. I have argued that these variants are not only semantically, but also formally distinguishable in that the full verbal variant does not allow of adverbial anteposition while the grammaticalized variant does. In a diachronic sketch, I have argued that these two variants are the results of different diachronic processes. Whereas the full verbal raising variant has been introduced into Danish by translation from Latin, the





Kasper Boye

auxiliary variant is the result of a grammaticalization. However, these two diachronic processes are linked in that the auxiliary variant has the full verbal variant as its diachronic source. In my argumentation I have suggested two things. First, I have suggested that the development of the Danish at-nci nexus construction is linked to the coding of evidentiality. This is also the case, I think, with, for instance, English and German cognate constructions (cf. also Svensson (this volume) on Swedish). Both English seem in he seems to be dead and German scheinen ‘seem’ in er scheint Tot zu sein ‘he seems to be dead’ have or imply evidential meaning, and both are occasionally considered auxiliaries. Second, I have suggested that phenomena linked by Hooper (1975) to weak assertive predicates have to do with grammaticalization, weak assertive predicates being grammaticalized predicates. In particular, I have tried to explain the phenomenon of adverbial anteposition with reference to the phenomenon of grammaticalization.

Notes . I am grateful to the Danish Research Council for the Humanities for granting me a research scholarship that has made my work on this paper possible. I am indebted to Peter Harder, Eva Skafte Jensen, Johan Pedersen and Gudrun Svensson for providing me with valuable criticism and comments. . The occurrence of passive se ‘see’ in construction with at-nci nexus may seem marginal. It is, however, abundantly attested, especially in texts written in a formal style. In April 2004, a google-search on the string “ses at være” ‘is seen to be’ resulted in 2020 hits. . It has been suggested that neg anteposition (and thus presumably adverbial anteposition) is not a semantic, but a pragmatic phenomenon, an implicature (cf. e.g. Horn 1989 and Levinson 2001). As pointed out by Croft (2001: 213), however, this suggestion does not hold. Adverbial anteposition, including neg anteposition, is really a semantic phenomenon. This follows from constructions such as jeg tror ikke jeg nogensinde har elsket hende ‘I don’t think I have ever loved her’ where the presence of the adverb nogensinde ‘ever’, which is positionally as well as semantically attached to the verb elske, hinges on the meaning of the negation positionally attached to the verb tro. . With simplex finite verbs and main-clause word order it is impossible in Danish to distinguish formally between adverbials attached to the finite verb and adverbials attached to the infinitive. For this reason, (6) and (7) have simplex APUA verbs appearing in embeddedclause constructions. (8), however, has the complex APUA verb lade til ‘appear’ (lit.: ‘let to’) appearing in a main-clause construction. . One might suggest that one prerequisite for this development resides in the fact that most Danish APUA verbs occurring in construction with at-nci nexus are simplex verbs. The simplex s-passives in (2) and (3), for instance, might be assumed to be relatively more prone

Raising verbs vs. auxiliaries

to grammaticalization than the corresponding complex English passives in constructions such as he is said to be dead. However, a number of Danish APUA verbs are actually complex verbs, and some of these, for instance lade til ‘appear’, seem to have an auxiliary variant.

References Aikhenvald, A. Y. (2003). “Evidentiality in typological perspective”. In A. Y. Aikhenvald & R. M. W. Dixon (Eds.), Studies in Evidentiality (pp. 1–31). Amsterdam: John Benjamins. Barron, J. (1999). Perception, volition and reduced clausal complementation. Unpublished Ph.D.-dissertation (University of Manchester). Benveniste, É. (1971). “Subjectivity in language”. Problems in general linguistics. Transl. by M. E. Meek (pp. 223–230). Coral Gables: University of Miami Press. Bolkestein, A. M. (1979). “Subject-to-object Raising in Latin?” Lingua, 48, 15–34. Borsley, R. D. (1996). Modern Phrase Structure Grammar. Oxford: Blackwell. Boye, K. (2001). “Evidence for Evidentiality in Danish”. In H. H. Müller (Ed.), Reflections on Modality (pp. 97–110). Copenhagen: Samfundslitteratur. Boye, K. (2002). “Den danske infinitivneksus – distribution, funktion og diakroni”. Danske Studier, 2002, 17–68. Boye, K. & P. Harder (forthc.). “Complement-taking predicates: Usage and linguistic structure”. Croft, W. (2001). Radical Construction Grammar. Oxford: Oxford University Press. Diderichsen, P. (1962). Elementær Dansk Grammatik. 3rd edition. København: Gyldendal. Eide, K. M. (2002). Norwegian Modals. Doctoral dissertation (Norwegian University of Science and Technology). Hansen, E. (2000). “Anteponeret adverbial”. In C. Bache (Ed.), Ny forskning i grammatik 7 (pp. 73–86). Odense: Odense University Press. Harder, P. (1996). Functional Semantics. Berlin: Mouton de Gruyter. Hooper, J. B. (1975). “On Assertive Predicates”. In J. P. Kimball (Ed.), Syntax and Semantics 4 (pp. 91–124). New York: Academic Press. Hopper, P. J. (1991). “On Some Principles of Grammaticalization”. In E. C. Traugott & B. Heine (Eds.), Approaches to Grammaticalization, I (pp. 17–35). Amsterdam: John Benjamins. Horn, L. R. (1989). A natural history of negation. Chicago: University of Chicago Press. Jespersen, O. (1924). The Philosophy of Grammar. London: Allen & Unwin. Kurylowicz, J. (1965). “The evolution of grammatical categories”. Diogenes, 51, 55–71. Langacker, R. W. (1977). “Syntactic reanalysis”. In C. N. Li (Ed.), Mechanisms of Syntactic Change (pp. 57–139). Austin: University of Texas Press. Levinson, S. C. (2001). Presumptive Meanings. Cambridge: MIT Press. Lyons, J. (1977). Semantics. Cambridge: Cambridge University Press. Meillet, A. (1921 [1912]). “L’évolution des formes grammaticales”. Linguistique historique et linguistique générale, Vol. 1 (pp. 130-148). Paris: La société de Linguistique de Paris. Mikkelsen, K. (1911). Dansk Ordföjningslære. Copenhagen: Lehmann & Stages Forlag. Östman, J.-O. (1981). “You know”: A discourse functional view. Amsterdam: John Benjamins.





Kasper Boye

de Reuse, W. J. (2003). “Evidentiality in Western Apache (Athabaskan)”. In A. Y. Aikhenvald & R. M. W. Dixon (Eds.), Studies in Evidentiality (pp. 79–100). Amsterdam: John Benjamins. Rosenbaum, P. S. (1967). The Grammar of English Predicate Complement Constructions. Cambridge: MIT Press. Ross, J. R. (1969). “Auxiliaries as main verbs”. In W. Todd (Ed.), Studies in Philosophical Linguistics, 1 (pp. 77–102). Evanston: Great Expectations Press. Thompson, S. A. (2002). “‘Object complements’ and conversation. Towards a realistic account”. Studies in Language, 26 (1), 125–164. Thompson, S. A. & A. Mulac. (1991). “A Quantitative Perspective on the Grammaticization of Epistemic Parentheticals in English”. In E. C. Traugott & B. Heine (Eds.), Approaches to Grammaticalization, II (pp. 313–329). Amsterdam: John Benjamins. Urmson, J. O. (1952). “Parenthetical verbs”. Mind, 61, 234–246. Vendler, Z. (1967). Linguistics in Philosophy. Ithaca: Cornell University Press. Willett, T. (1988). “A Cross-Linguistic Survey of the Grammaticization of Evidentiality” Studies in Language, 12 (1), 51–97.

On the origin of the final unstressed [i] in Brazilian and other varieties of Portuguese New evidence in an enduring debate* Maria José Carvalho University of Coimbra/CELGA

.

Introduction

In standard Portuguese and some of its local variants the sound [6], written as e, is commonly found in the unstressed position, especially the final one. However, in Brazilian Portuguese, in some Portuguese-based creoles, in some dialectal varieties of Portugal itself and in most European varieties we find elevation of final unstressed e (BP cidade [si´dadži], ponte [´põtši], verde [´verdži], etc.), which is one of the most intriguing phonetic phenomena of Portuguese for the linguistic community.1 The fact that territories, geographically unconnected but colonised by speakers of European Portuguese in the 16th century, show identical versions of the final unstressed vowel leads us to suppose that -[i] was the most common pronunciation of this vowel in Portuguese at that time, or even earlier. However, the absence of convincing documentary evidence, and the preservation of non-elevated unstressed vowels in Brazilian Portuguese and in Portuguese-based creoles, has not only made it difficult to establish a chronology, but has also led to synchronic explanations for an intrinsically historical phenomenon.2 . Status quæstionis Herculano de Carvalho’s article, athough well-respected in this field, does not seem very helpful here. He begins by claiming, on the basis of 18th century



Maria José Carvalho

grammars, that in the contemporary Lisbon variety unstressed graphemes e and o were equivalent to the vocalic phones [e] and [o], even in the final syllable (Carvalho 1962b: 14), but referring to some Brazilian varieties, he poses the question: Representarão as realizações [i u] a persistência de realizações já correntes, senão gerais, no português dos séculos XVI-XVII? Ou pelo contrário não representarão [e o] das variedades brasileiras ùltimamente mencionadas a conservação da fase mais antiga, que seria efectivamente a do português do século XVI, de que só recentemente teria resultado [i u] no Brasil, em Portugal (já nas primeiras décadas do século XVIII!) e nos falares crioulos? (Carvalho 1962b: 19).3 [Might the sounds [i u] show the survival of forms already current, if not general, in 16th and 17th century Portuguese? Or conversely might not [e o] in these Brazilian varieties represent the preservation of the earliest phase, which would in fact be 16th century Portuguese, which would recently have resulted in [i u] in Brazil, in Portugal (even already in the first decades of the 18th century!) and in creoles?]

Still on the hypothetical level, though now basing his arguments on evidence from modern Portuguese varieties, he admits “a possibilidade de que estes [-e -o] finais actuais continuem ininterruptamente uma realização antiga, digamos, quinhentista [-e -o], sem que isso implique que ao lado dela não existisse uma outra com [-i -u]” (Carvalho 1962b: 20). (“the possibility that current final [-e -o] represents the unbroken survival of an ancient, perhaps sixteenthcentury sound [-e -o], without implying that another with [-i -u] did not exist alongside it”). However, his claims are inconclusive and lack empirical support. Anthony Naro, in an attempt to explain by generative phonology the ‘drift’(Naro 1971: 615, 642) seen in the data,4 recognises that the elevation of the final unstressed -e remains unexplained due to the lack of data (Naro 1971: 637). However, he starts from the assumption that this is a relatively late change in Europe: (...) final -e became -i in Portugal sometime between the last quarter of the 17th century (Barreto) and the mid 18th century (Verney). This same change also occurred in large sections of Brazil at an unknown date, and from the late occurrence of the change in Europe it is clear that the two developments were independent (Naro 1971: 641)

Clarinda de Azevedo Maia recognises that the forms with -i, found in the corpus from the original Galician-Portuguese area which she has edited, are very ancient. She adds in a note that the pronunciation [i] is

On the origin of the final unstressed [i] in Brazilian

atestada apenas pelos gramáticos e ortógrafos a partir do século XVIII, mas que certamente sempre existiu. Aliás, só assim se explica a sua sobrevivência nalguns falares de Portugal continental e insular, no português do Brasil e em quase todos os crioulos (Maia 1997: 377, continuation of the note on the previous page). [only attested by grammarians and orthographers from the 18th century, but it had certainly always existed. This is the only explanation for its survival in certain dialects of continental Portugal and the islands, in Brazilian Portuguese and in nearly all the creoles].

She raises the question of the early date of the phenomenon, and confirms Serafim da Silva Neto’s claim: “Essa distribuição geográfica actual leva-nos a crer que outrora a pronúncia de -e final era -i em todo ou quase todo o território de Portugal” (Neto 1992: 483) (“This present geographical distribution leads us to believe that formerly the pronunciation of final -e was -i in all or nearly all of Portugal”). We have deliberately emphasised the adverbs always and formerly in the two quotations above, since it seems to us that what is missing from existing studies of this subject is an empirical basis on which to set up a precise chronology for the appearance of final i, a crucial question for the history of unstressed vowels in Portuguese. The most recent allusion to the phenomenon of elevation of final unstressed -e is, as far as we know, that by Helmut Lüdtke. Referring to the origins of the different graphic -i in Old Portuguese, the author states that “not all of them originate directly from classical Latin -i (. . .), as for example outri and longi” (Lüdke 1993: 284).5 Following Lüdke’s line of reasoning, this paper aims to help clarify the place of the phenomenon in the Medieval period, based on a corpus of original juridical sources (13th–16th centuries) which we have transcribed,6 originating from a particularly important centre in medieval Portuguese culture , the monastery of Alcobaça7 (between Coimbra and Lisbon). The corpus consists of 153 original documents not only from the monastery but also from the outlying areas under its jurisdiction (coutos), dating from between 1289 and 1565.8 (Portuguese was, of course, first taken to Brazil in 1500, at the time of Pedro Álvares Cabral’s first voyage).





Maria José Carvalho

. Analysis of the corpus (13th to 16th centuries) The data collected in our corpus seem to point clearly to two types of forms of final i, connected to two types of chronology. . Forms with final etymological -i (< ¯I) Let us look at the forms found in the period between the end of the 13th century and about 1350, usually considered as corresponding to the first phase of archaic Portuguese:9 (1) dar˜emj infl inf-1sg. clit “to give me” (year 1307) (2) deu˜elhis pres-3pl. clit ‘they should. . .them’ (year 1307) (3) ffazer˜emj infl inf-1sg. clit “to make me” (year 1307) (4) lhi/lhj/lhy 3sg. clit “him” (years 1328/1330/ 1298; 1300; 1306; 1315, 2; 1326, 6; 1328; 1336, 7; 1338, 3; 1343, 2; 1346, 2; 1351, 6) (5) lhis/lhjs/lhys 3pl. clit “them” (1304, 2; 1328; 1330; 1336, 3/1329; 1340/1289, 2; 1306; 1328, 4; 1334; 1343, 2; 1351, 2) (6) mãdolhy pres-3sg. clit “I order him” (year 1343) (7) mi/mj/my 1sg. clit “me” (years 1289/1297/1315; 1343) (8) outrim pron “others” (year 1307)

On the origin of the final unstressed [i] in Brazilian

(9) puggi/pugi/pugj/pugy/puwwy simp past 1sg. “I put” (years 1306/1304; 1336/1305; 1307; 1328; 1328; 1329; 1330; 1334; 1338; 1340; 1343; 1346; 1346/1326/1299) (10) vijntj num “twenty” (year 1332).

If we examine the forms of the first person singular of the Simple Past tense ˘ ı (CP pus “i put”) during this period, we see that these are derived from posu¯ the only forms used. As for the subsystem of clitics (coming from Latin mih¯ı and ill¯ı, since the second person is never found), forms with final i consist of about 94% of the total in this period. In the same chronological period we find the following patronymics (< -¯ıs):10 Dom˜ıguiz (years 1289; 1297; 1298; 1300; 1305, 2; 1307, 2), Ewteuãiz (year 1297), Ewteueyz (year 1298, 2), Fernadiz (year 1321), Fernãdiz (years 1300; 1350, 2), Ffernãdiz (years 1306; 1343, 3; 1346; 1350, 4), Giraldiz (year 1328), Gõçalujz (years 1342, 3; 1350), Gonçalujz (year 1343), Gonçaluiz (year 1321), Gregoriz (years 1328, 2; 1338), M˜e˜ediz (years 1300; 1307), Meendiz (year 1304), Nuniz (year 1300), Rejnaldiz (year 1351), wanchyz (year 1299), wanchiz (year 1300, 2), woariz (year 1304), wuariz (year 1304), wteueiz (year 1304, 2), Uaawquiz (year 1336) and Vááquiz (year 1346).

In quantitative terms, we may say that at this period patronymic forms with final i are about 75% of the total. There is a fairly sharp reduction in the forms outrim “others” (< alter¯ı), down by 17%, and vijntj (< vigint¯ı), down by 8%.11 We shall now look at the frequency of occurrence of final i in the second half of the 14th century. The clearest features are the total disappearance of the cardinal number historically representing vigint¯ı with final i/j, the replacement of the forms of the first person of the Simple Past tense pugj (and variants) by puge, and the almost complete transformation of i in patronymics to e; here, the only remaining examples of the previous form are Ffernandjz (year 1355) and Uaasquiz (year 1362), which amount to only 9% of the total. Let us look again at clitics: (11) damowlhy pres-3sg. clit “we give him” (year 1379)





Maria José Carvalho

(12) lhi/lhj/lhy 3sg. clit “him” (years 1362; 1363, 4; 1375, 3/1366; 1372/1362, 3; 1375; 1379; 1383, 4; 1385; 1391; 1397, 3; 1399, 2) (13) lhis/ lhys 3pl. clit “them” (years 1388, 2; 1391; 1399, 2/1379, 3)

One of the conclusions we can draw here is that there are slightly fewer occurrences with final i/j/y in this kind of form in the second half of this century. The percentage falls to 70%, but it is still higher than forms with e. What is surprising in this collection of documents is that clitic pronouns with i (as well as patronymics ending with this vowel) disappear completely from the 15th century onwards, falling dramatically from 70% to 0%. In fact, the tendency towards the spread of final e in forms of this type ¯ (< ı) is seen from an early period, probably as a result of normative orthographic pressure by the central powers: the scriptorium of the Monastery of Alcobaça was apparently not immune from this process of standardisation. The following table summarises the quantitative data on forms with final i(y/j), as outlined above: Table 1. Chronology of forms with final i(< ¯ı) Forms with final i (<¯ı)

1289–1351

1352–1399

1st half of 15th cent.

Forms of the 1st person of Simple Past tense (< p˘osu¯ı) Clitics (mih¯ı and ill¯ı) Patronymic forms (-¯ıs) Form representing Latin dative alter¯ı Numeral representing vigint¯ı

100%

0%

0%

94% 75% 17% 8%

70% 9% 0% 0%

0% 0% 0% 0%

. Forms with final “non-etymological” -i (< E¯ ; ˘I) What is now required is information of a qualitative kind on the forms with final i which appear as exceptions in the 14th century but which only become frequent from the last quarter of the 15th century. In nearly all cases these forms are different in nature from those presented above, since according to their etymological origin they would have been written with an e. Thus, if the spread of forms with e to those which came from Latin ¯ı was easily accepted

On the origin of the final unstressed [i] in Brazilian

in society from the early days of the language, it seems that the inverse process of substitution also took place, but was part of a different social phenomenon. That is, from the end of the 13th century the phenomenon of the spread of final i to forms with a vowel historically deriving from E¯ and ˘I was taking place, but it was an impulse which was checked by an overriding normative tendency, since political and socio-cultural conditions tended towards a degree of stability. We can therefore see two distinct chronological phases in the following table:12 ¯ ˘I and occasionally ¯ı13 ) Table 2. First and second phases of final i (< E; First phase 1291 1300

1321 1336 1380

ordim “order” ç˜ıquj “five” Ffroylhy (personal name) ord“‘, 2 “order” xi impers clit sabersy infin-impers clit “to know”

¯ ˘I and occasionally ¯ı) Final i (< E; Second phase 1440

1443

1482

1489

1490

1491

1496

quiti “discharged” ljuri “free” firmj “valid” quity ljuri, 2 firmj cõprim15 pres. subj. “they keep” v[i]rym infl inf 3pl. “to come” obrigousy simp past 3sg.-refl clit “he committed himself ” mj 1sg clit “me” Aluariz, 2 (patronymic)

1529 1532

hutily14 “useful” djãtj “afterwards” partj “part” cor˜etj “current” quitj “discharged” pres˜etj “present” estj “this” Gomiz16 (patronymic)

In fact, in the 14th century such pressure seems to be seen roughly between the period just before the battle of Salado (1340), which was the last crusade against the Moors in the Iberian Peninsula, and the battle of Aljubarrota (1383– 1385), which set the seal on national identity. This normative graphical stability was to last into the first half of the 15th century, with the accession to the throne of the Infantes of Avis, until the crisis which led to the disaster of Alfarrobeira





Maria José Carvalho

(1449) began to have repercussions in monastic life. The 1380 document comes from the monastery’s most rural and outlying couto, so that it may be a good example of the language as used throughout practically the whole country, uninfluenced by trends towards standardisation from cultural centres. The notary who in the 1440s changed -e to -i in these forms in two quittance documents (certifying that the debtor had paid in full and was now free of any debt) did it in very precise formulaic cotexts. It is clear that the elevation of the final unstressed vowel is here influenced by assimilation with the stressed vowel, but in the same decade the same form is seen in a document on the same subject, produced nearby by another notary, the unstressed vowel here being e rather than i: (14) dou por quite e ljure, 2 “I account discharged and free” (year 1448).

This leads us to the hypothesis that the focus of innovation in the 15th century was related to parameters of a socio-cultural nature, but it is also true that there is nothing to show that the first notary did not use the same adjectives with e rather than final i in other texts, since he used ewtauell and ewtaujll “valid” indiscriminately in two different documents. Nor, on the other hand, is there anything to show that the second notary did not pronounce the final graph e as [i] or as a sound somewhere in between, since we know that the degree of awareness of the distance between the two semiotic systems, written and spoken – both now and in earlier periods of the language – may vary considerably from person to person. It is clear however that from the last quarter of the 15th century this highly stigmatised ‘popular style’ of final unstressed i, once originally restricted to forms which had Latin final ¯ı, began to spread once more, now not only to clitics and patronymics, but particularly to other forms where it had no etymological reason to exist. That the phenomenon was already seen at certain regional or social centres seems to be shown by the document of 1532 (a contract of sale). This was produced, curiously, by a notary working in the most rural couto in the extreme south of the Alcobaça area; another point worth mentioning is that all the clitic forms lhe “him” (5 in all) keep the final e, not changing it to i. We can see from this that the notary may have been poorly educated – he hyper-corrected by keeping the e when i would have been expected (since etymologically it was a long i) and changed it to i in forms where, in accordance with its etymological origin, it had been established as e. We do not have the evidence to claim that the phenomenon was common throughout the country, but it is important to recognise that this analogical spread of

On the origin of the final unstressed [i] in Brazilian

final i is found repeatedly in a document by a notary from the most remote and rustic couto of Alcobaça.17 Other phonetic-phonological phenomena we have analysed in our researches have characterised the town regionally, allowing us to imagine what could have happened in the rest of the country, away from linguistic and orthographic coercion by a monastic centre. In fact, the evidence recorded in the História do galego-português (Maia 1997: 375) leads us to believe that the spread of non-etymological final i was also an early phenomenon in the Galician-Portuguese area, and that it continued more freely in these 15th century texts than in the documents we are analysing here. This, added to the quantification of data from our corpus, lends support to the idea that, in the documents produced at the most significant monastery in this dialectal region from the last two decades of the 14th century onwards, strong pressures towards orthographic standardisation were in evidence, which would have reflected some aesthetic ideal in the world of letters, originating from the centre. From the last two decades of the 15th century, with Portugal changing direction and Lisbon opening up to the whole continent, these normative forces would have relaxed considerably, allowing for the formation of different socio-regional linguistic identities. Thus the language was ‘denationalised’,18 and was to be taken, along with the particular shades lent to it by those who ventured across the oceans, to many continents, especially America.

. Conclusions: Towards a sociolinguistic reconstruction The observations we have made therefore tend to confirm Paul Teyssier’s hypothesis, clarifying his proposed chronology: Ce [ë] n’a pu évidemment sortir directement de l’ancien [e.], réalisation primitive de -e final atone, mais il provient du [i] attesté dans la première moitié du XVIIIe siècle, l’évolution s’étant faite selon le schéma suivant: [e.] > [i] > [ë]. (Teyssier 1980: 73)19

Since our investigation has led us to reject the previously accepted theory that the elevation of Portuguese final unstressed e is an 18th century phenomenon, we would claim that a consistent interpretation of this phonetic aspect in the history of Portuguese demands consideration at three distinct levels: at the linguistic level, the sociolinguistic level and at the level of relationship between oral and written language:





Maria José Carvalho

a. Linguistic level We must emphasise that 18th century Lisbon forms, as seen in contemporary grammar books, should not have the same linguistic status as those in our corpus from the 13th and 14th centuries, where the i results from the evolution of ¯I; that is, not to recognise that there were (etymologically) ‘primary’ forms like lhi, mi, pugi, vijntj, etc., at an early stage of the language, and others certainly later, which might include the former, but particularly forms such as estj, presentj, djantj, etc. (< e¯ ; ˘i) probably deriving from analogy with final i, is to follow a misleading chronology. In fact, although this is an intra-linguistic matter, we must distinguish two different phenomena. On the other hand, evidence from the 16th century allows us to infer that the cotext most favourable to the vowel elevation in this analogical phase was that when e follows an apical-dental phoneme (/t/ and /d/). b. Sociolinguistic level A decentralised popular tendency to elevate the final vowel was competing with the e (reflecting an orthographic taste), and flourished in texts from periods of political instability, and in others of clearly rustic origin. Examples, though rather few, seem enough to indicate that from the 1440s onwards the voice of the speakers begins to show through in the writing, thus violating a certain orthographic unity which was probably based on an orthoepic conception of language. This popular tendency for the spread of final i could not have sounded good at Court, especially to King João I, the Master of Avis, and his successor, King Duarte, since it seems that between 1380 and 1439 there is not the slightest trace of i in written texts. We do not know, therefore, what would have been the phonetic value of final e in this stage of relative political stability. It must have been the excessive ‘nationalism’ of Fernão de Oliveira, author of the first Portuguese grammar (1536), and his purist vision of the language, which made him conceal this phenomenon: since he was born at the beginning of the 16th century, he did not yet have the necessary distance to reflect on a change (apparently socially stigmatised) which was happening under his very eyes. We believe, however, that it must have been the importance given by researchers to the study of the ‘standard’ Medieval variety, or perhaps the scarcity of sources of other kinds, from areas remote from centralising pressures, which has led to some difficulties in establishing the chronology of the phenomenon, and thus made it “un des points les plus importants, mais aussi les plus obscurs, de l’histoire du portugais”(Teyssier 1980: 70), which goes back “dans ses origines, à une époque bien plus ancienne qu’on ne serait porté à croire”(Carvalho 1962a: 11).

On the origin of the final unstressed [i] in Brazilian

c. Level of relationship between oral and written language This level is inevitably linked with the preceding one.20 In previous studies we have tried to show how essential it is to understand the development of Portuguese from a sociolinguistic perspective . In this present work of reexamining the available evidence, the researcher is inevitably faced with the unavoidable difference between two distinct semiotic systems. Recalling what we have been able to ascertain so far about the periods of Portuguese linguistic history within this theoretical and methodological framework, we can claim the following: the central power, represented in the field of letters by the Infantes of Avis, in view of the linguistic ferment of the end of the previous century, and fearing the spread of features leading away from the aesthetic ideal, fixed and sealed the orthography with a “homogeneous layer of sand”(Meier 1943: 499),21 in an unworthy attempt to restrain non-central forces. This apparent stability, expressive of the power of standardisation and which, never, we believe, obliterated the linguistic thesaurus of the people, was to endure for about half a century. Finally, we believe that the difficulty in tracing precise chronologies is due to lack of use, on this level, of the sources which are available. In our view, the editions of literary texts may mislead the researcher who seeks to capture the orality of earlier phases in the history of any language. In fact, we strongly believe that official documents, since they are located in time and space, contain valuable orthographic evidence (if used with caution) and are therefore the best sources to construct satisfactory chronologies for linguistic change.

Notes * The presentation of this paper was supported by a grant from the Fundação Calouste Gulbenkian. This support is gratefully acknowledged. I also thank my audience at the ICHL 2003. . This elevation seems to be a natural consequence of decrease in energy of articulation. In parts of southern Portugal, including the Algarve, and in some creoles, these vocalic phonemes were suppressed, but traces of their previous existence, and of their pronunciation as closed vowels, remain, through metaphonic phenomena, in the stressed vowels of various forms (Algarve, Cape Verde). According to Hammarström, “Des localités situées dans toutes les parties de l’Algarve et sans lien géographique entre elles présentent des emplois contenant le même i final; il est donc naturel de voir dans ces formes des survivances d’un stade que les autres parlers ont dépassé” (borrowed from Hart Jr. 1955: 408). In presentday Portuguese dialects, the best-known example of elevation of final e is in Alentejan pronunciation, although the same kind of elevation is also found in other dialect areas. Her-





Maria José Carvalho

culano de Carvalho’s comments on Mirandês are also relevant here: Carvalho 1958: 68–69; 110. For Cape Verde creole, see Carvalho 196–?: 46–47 and Carvalho 1962a: 4–5. . The creolist John Holm mentions this phenomenon in S. Tomé creole: “Portuguese /6/ often became /i/ in the Gulf of Guinea creoles, e.g. São Tomé CP nglãji ‘large’ (P grande /gr"nd6) (...)”. Cf. Holm 1988: 116. . Referring to the unstressed vowels of Cape Verde creoles, he shows the same doubts on the chronology of vocalic elevation: “Le vocalisme atone que je viens de décrire dans ses larges traits continue clairement un système portugais fondamentalement identique, ayant /i a u/ dans la finale, /i e a o u/ (/e o/ réalisés comme [e.] et [o.]) dans toutes les autres syllabes non accentuées, c’est à dire, un système qui pour la finale et la postonique est le même que nous connaissons pour le portugais du XVIIIe siècle et qui pourrait être déjà celui du XVIe”. (Carvalho 1962a: 4–5). . Paul Teyssier also says, on the subject of final e and o: “Comme il n’existait en cette position que trois phonèmes, et qu’il n’y avait par conséquent aucun moyen d’opposer [o.] a [u] et [e.] a [i], les locuteurs lusophones, pendant longtemps, n’ont pas pris conscience de cette dérive, et il faut attendre la première moitié du XVIIIe siècle pour que certains observateurs remarquent qu’on prononçait [u] et [i]” (Teyssier 1980: 75). . The author suggests analysing longi by comparison to many other adverbial forms in -i found in the Astúrias and in Italy; and for outri, he believes this form to be part of a series of analogical transformations, which he explains further on. . Part of our doctorate, presently in progress, and hence unpublished. The transcription has been based on rigorous philological criteria. . The selected documents are part of the collection Mosteiro de Alcobaça (1st and 2nd Incorporations) from the “Arquivo Nacional da Torre do Tombo” (Lisbon), the largest and most important national archive. The need to transcribe, according to linguistic criteria, and to analyse the original documents preserved in archives, is implicit in the following statement by Thomas Hart: “I have stressed the inadequacy of the materials with we must work and the difficulties of using them properly. We linguists must learn to distinguish more carefully between those things which are demonstrably true and those which, at least in the present state of our knowledge, are incapable of proof. And it has seemed to me particularly important to stress the limitations of the materials and methods at our disposal as a preliminary to pleading for the collecting of more extensive data and more precise ways of interpreting it”. Cf. Hart 1959: 272. . The selection criteria were essentially diatopic and diachronic, but obviously the corpus would never be perfect. We may recall William Labov: “But the data that are rich in so many ways are impoverished in others. Historical documents survive by chance, not by design, and the selection that is available is the product of an unpredictable series of historical accidents (. . .). Historical linguistics can then be thought of as the art of making the best use of bad data. Cf. Labov 1994: 11. . When there is more than one occurrence of the form, the number is separated from the year by a comma. . We also find the toponym Méédj (< * MENEND¯I) (year 1299).

On the origin of the final unstressed [i] in Brazilian . Forms with the grapheme e already make their appearance in the last years of the 13th century. . The chronological distribution of the documents in the corpus is as follows: 13th cent. – 7; 1st half of the 14th cent. – 29; 2nd half of the 14th cent. – 30; 1st half of the 15th cent. – 38; 2nd half of the 15th cent. – 31; 16th cent. – 18. Thus the second half of 14th century and the first half of the 15th century are the best represented periods in terms of documentation, which rules out the theory relating the absence of textual evidence to possible gaps in the chronological distribution of documents. . Examples of this last are mj (1491), Aluariz (1496) and Gomiz (1532). . This is a form showing later paragogic development. . The vernacular form is cumpram, which shows that the phenomenon of elevation of final e must have been so common in the spoken language that it extended, in probably less formal registers, into other contexts where the vowel did not exist. . Curiously, this is the only occurrence of this patronymic form showing final i: the form had e in this cotext since the archaic period. . In the section on “Problems in the interpretation of historical data”, William Labov says: “Except for very recent times, no phonetic records are available for instrumental measurements. We usually know very little about the social position of the writers, and not much more about the social structure of the community”. Cf. Labov 1994: 11. . A term based on “desgaleguização”, used for the gradual process of separation of Galician from Portuguese (first seen from the mid-14th century). . It is worth noting that the forms rresydents and segints appear in a document from 1485, leading us to suppose that at this time there was already a form of final [6], which must have been parallel to the [i]. The following statement from I. S. Révah must therefore be revised: “Les actuelles pronunciations de la langue commune (-E central ou disparition de la voyelle) doivent donc dater de la fin du XVIIIe siècle ou du commencement du XIXe ”. Cf. Révah 1959: 282. . We recall here that “La reconstrucción de la lengua del pasado en su contexto social suscita cuestiones tales como la relación texto escrito/texto oral y la covariación entre variables lingüísticas con factores lingüísticos y sociales”. See Gimeno Menéndez 1995: 176. . Metaphor used by Harri Meier, referring to Wartburg’s research, which led him to claim that the Iberian Peninsula was a linguistic unit up to 1000 AD. Cf. Meier 1943: 499.

References Carvalho, José Gonçalo Herculano de (1958). Fonologia mirandesa, Vol. I. Coimbra: Imprensa de Coimbra. Carvalho, José Gonçalo Herculano de (1962a). “Le vocalisme atone des parlers créoles du Cap Vert”. Actas do IX Congresso Internacional de Linguística Românica, Vol. III (= Boletim de Filologia 20, pt. 1 and 2, 1961), 3–12.





Maria José Carvalho

Carvalho, José Gonçalo Herculano de (1962b). “Nota sobre o vocalismo antigo português: valor dos grafemas e e o em sílaba átona”. Revista Portuguesa de Filologia 12, pt. 1. 1–23 [Separata]. Carvalho, José Gonçalo Herculano de (196?). “Sincronia e diacronia nos sistemas vocálicos do crioulo caboverdiano”. In Diego Catalán (Ed.), Miscelánea Homenaje a André Martinet “Estructuralismo e Historia”, Vol. III (pp. 1–25). Laguna: Universidade de la Laguna. [Separata]. Castro, Ivo (1991). Curso de história da língua portuguesa. Lisboa: Universidade Aberta. Drinka, Bridget (1990). “The chronology of phonological change”. In Edgar C. Polomé (Ed.), Research Guide on Language Change (pp. 107–136). Berlin, New York: Mouton de Gruyter. Hart Jr, Thomas R. (1955). “Notes on Sixteenth Century Portuguese Pronunciation”. Word, 11 (3), 404–415. Hart Jr, Thomas R. (1959). “The overseas Dialects as Sources for the History of Portuguese Pronounciation”. Actas do III Colóquio Internacional de Estudos Luso-Brasileiros (Lisboa 1957), Vol. I (pp. 261–272). Lisboa. Holm, John (1988). Pidgins and Creoles. Vol. I: Theory and Structure. Cambridge: Cambridge University Press. Gimeno Menéndez, Francisco (1995). Sociolingüística histórica (siglos X–XII). Madrid, Universidad de Alicante: Visor Libros. Labov, William (1994). Principles of Linguistic Change. Internal Factors. Oxford UK and Cambridge USA: Blackwell. Lüdke, Helmut (1993). “As origens do -i final átono do português antigo”. In Jürgen Schmidt-Radefelt (Ed.), Semiótica e Linguística portuguesa e românica. Homenagem a José Gonçalo Herculano de Carvalho (pp. 281–284). Tübingen: Gunter Narr Verlag. Maia, Clarinda de Azevedo (1997). História do galego-português. Estado linguístico da Galiza e do Noroeste de Portugal desde o século XIII ao século XVI (Com referência à situação do galego moderno). Lisboa: FCG and JNICT. Meier, Harri (1943). “A evolução do português dentro do quadro das língua iberoromânicas”. Biblos, 18 (pt. II), 497–515. Naro, Anthony J. (1971). “The history of e and o in Portuguese: a study in linguistic drift”. Language, 47 (3), 615–645. Neto, Serafim da Silva (1992). História da língua portuguesa (6th ed.). Rio de Janeiro: Presença/DinaLivro. Révah, I. S. (1958). “L’évolution de la prononciation au Portugal et au Brésil du XVIème siècle à nos jours”. Anais do Primeiro Congresso Brasileiro de Língua Falada no Teatro (1956) (pp. 387–399). Rio de Janeiro: Biblioteca Nacional & Universidade Federal da Bahia. Révah, I. S. (1959). “Comment et jusqu’à quel point les parlers brésiliens permettent-ils de reconstituer le système phonétique des parlers portugais des XVIe-XVIIe siècles?”. Actas do III Colóquio Internacional de Estudos Luso-Brasileiros (Lisboa, 1957), Vol. I (pp. 273– 291). Lisboa. Teyssier, Paul (1980). Histoire de la Langue Portugaise. Paris: Presses Universitaires de France.

Socio-historical evidence for copula variability in rural Southern America Gaillynn D. Clements University of Cambridge

.

Introduction

The purpose of this paper is to examine copula variability in Southern white vernacular English. The goals of this study are to (1) examine copula variability in Advance, North Carolina, to see at what frequency these speakers have copula absence and habitual be if at all, (2) compare my results to three previous copula variability studies by Wolfram, Feagin, and Hazen, looking for similarities in patterning whether by age, class, gender, or environment, and finally (3) if upon comparison there is a difference in findings of my study to the three previous studies mentioned in point (2), I will account for this difference with socio-historical evidence. Dialectologists who are mainly interested in white English tend to view “Southern white English, especially as spoken by the lower classes, as an archaic form of English which preserves regional features of the British Isles” (Feagin 1979: 4). Why has the South preserved archaic features? The South is more resistant to change: political, social, and linguistic because of isolation particularly between 1860 and 1940. Poverty and lack of immigration preserved archaic and non-standard features which were largely lost in the rest of the United States through the leveling of social differences and greater interest in education than the South (McDavid 1967). If we accept that non-standard white speakers have retained older forms, where does the variability of the copula come from since the paradigm of textual forms and narrative examples did not include be+verb+ing up to and including the Early Modern English period?



Gaillynn D. Clements

. History of the copula in English Old English had complex alternatives for the present indicative be form. There were two inflected forms, but each of these forms differed according to dialect (Hazen 2000: 96). The two forms from Proto-Indo-European (PIE) were *es, closer to the paradigm we see today, and *bheu. These changed into two different forms which are in Table 1. In Middle English, (see Table 2) Burrow and Truville-Petre (1996: 37) cite two paradigms from West Midland texts, with variation for be. Note there is variation in first, second, and third person singular in Middle English. Examples (1)–(5) are taken from Visser 1970 which show copula deletion dating from about 1400. Translations and glosses of (1) and (2) are by Dr. Scott Gwara at the University of South Carolina, translations of (3)–(5) are by Visser. Table 1. Old English copula roots in Present Indicative Sievers and Cook 1903: 351

West Saxon *es 1 2 3 Northumbrian *bheu 1 2 3

Singular

Plural

eom eart is

sindom sind sint

bio bist bi

bio bio bio

Table 2. Variation of be in West Midland texts Adapted from Burrow and Truville-Petre (1996: 37)

Ancrene Wisse *beon 1 2 3 Sir Gawain and the Green Knight *be, bene 1 2 3

Singular

Plural

am, beo art, bist is, bi

beo beo beo

am art is, betz

ar(n), ben ar(n), ben ar(n), ben

Copula variability in rural Southern America

(1) Wundur hwar þonne earl ellenrof ende gefere lifgesceafta Mystery where nobleman brave end may-reach lifespan “It Ø a mystery where a nobleman brave will then reach the end of his lifespan” (Beowolf 3062) (2) Hayl Marie. . . y -blissed þou ine wymmen, and Hail Mary... Perfective Prefix -blessed thou among women and y -blissed þet ouet of þine wombe Perf. Prefix blessed be-eth fruit of thy womb “Hail Mary...blessed Ø thou among women and blessed be the fruit of thy womb” (Ayenbite (Morris/Skeat, Spec. II p. 106) circa 1340) (3) he declared then, that king Eduard was neuer lawfully married, vnto the queen,. . . & so his children bastardes “Then, he declared that King Edward had never lawfully married the queen, ... and so his children Ø bastards” (St. Thomas More (Wks 1557) 6I A6 c 1513) (4) She a beauty “She Ø a beauty” (Austin, Pride and Prejudice, 333 1796–1812) (5) Awful rotters those Portuguese! “Those Portuguese Ø awful rotters” (Galsworthy, Plays, Joy 51 c 1912)

Visser points to the semantic emptiness of to be as the reason for the omission of the copula. Since he has only compiled forty-three examples over a five hundred twenty year period, the deletion of the copula in Old and Middle English seems to be a minor tendency employed at time for prosody or variable contraction of unstressed auxiliaries and vowel elision, which is conditioned by prosody (Selkirk 1984). Selkirk 1984 argues that prosody explains deletion in literary constructions, as in examples (1) and (2). Through the Early Modern English period the paradigm of written forms became standardized. Copula absence was around in scant numbers in Old, Middle, and Early Modern English. So where did the increased frequency of copula absence and other variable forms come from? Closer research done by Laura Wright on Early Modern London prison narratives (2001, 2002) from Bridewell and Bethlem, which were the first to send prisoners to the Jamestown colony, show no copula absence, but they do show instances of invariant be, as in “They be both whores“. In Early Modern English, invariant be was a minority variant like absence in Old and Middle English. But Wright found no instances of invariant be+verb+ing. She did find many other similarities between the narratives and Southern American English, which are prescriptively wrong, such as





Gaillynn D. Clements

a+verb+ing or a-prefixing seen in Appalachian English, and Southern American “He’s a-painting his house” and double marked preterites seen in English, as in in Ocracoke and Texas Southern American English. The fact that Wright found no copula absence or be+verb+ing in the prison narratives and Visser’s examples were meager, would suggest that these two particular forms of copula variability in Southern American English are not maintained features from British English.

. American innovation Labov (1969) does admit that European American Southerners in North Carolina and Georgia do show are deletion, but his anecdotal research does not account for any is deletion. He attributes the deletion of are to the fact that the speakers are r-less, therefore, reducing their percentage of are use. Feagin also mentions that her informants have more r-constriction in word final position than Wolfram’s Mississippi informants, which under the Labovian theory, would account for her informants having less are deletion. But Wolfram does not agree with Labov citing his results as evidence that there is is and are deletion, and because there is evidence of both, which was found in my study as well, Labov cannot dismiss are deletion as a result of r-lessness. Hazen does not look at r-constriction. . R-lessness as a reason for absence in Advance I looked at r-constriction in Advance to see if this could be an explanatory factor for are absence. I looked for r medially and finally in each speaker. I found that r-lessness is not a predominate feature in Advance speech and could not have caused the high frequency of copula absence. Since r-lessness was only shown in three tokens out of 165 postvocalic r tokens, or 1.8%, there is not substantial evidence to claim that these speakers are r-less, and thus copula absence is most likely not an American innovation in Advance. . Influence from AAVE One could argue, like Feagin 1997, that African-American Vernacular English, AAVE, has influenced Southern American English through contact of blacks and whites during and following slavery. Wolfram states that only Southern European American speakers appear to have patterns of are absence similar

Copula variability in rural Southern America

to AAVE. This would seem to be evidence for change through contact, but it would take prolonged and extreme contact to change an underlying grammar. In addition, in contact situations, such as that between blacks and whites in the South, the prestigious language1 normally does not change, the nonprestigious one does. Feagin’s 1997 study focuses on r-lessness, the Southern drawl, and falsetto, or pitch with tensed vocal cords, and even though these are phonological features, it does show that a nonprestigious variety can change the prestigious variety. We have also seen unconscious changes as Bright (1960) describes them when the lower class, in attempt to imitate the upper classes’ phonetic changes, brings about phonemic change, which eventually brings structural changes for the language.

. Other Southern American copula studies There are three studies I turned to that analyzed the copula in Southern American speech, Wolfram (1974), Feagin (1979), and Hazen (2000). Wolfram’s study focused on the speech of whites in Meadville, Mississippi, where he found are deletion to be extremely common in his speakers: his thirty-three speakers deleted are 64.1 % overall, and nine of his speakers had categorical are deletion. He only found 6.5% is deletion. Feagin’s study focused on the speech of whites in Anniston, Alabama, which is more metropolitan. Feagin’s results were similar to Wolfram’s: she found a higher frequency of are deletion than is. Her speakers deleted are 28.79% of the time, although none had categorical deletion of are, and they had only 4.5% is deletion. Hazen’s study of Warren County, North Carolina, focuses on white, black, and Native American speech in the county. He found white speakers deleted is/are 8% of the time (15 tokens out of 179). Table 3 compares Wolfram, Feagin, and Hazen’s studies’ demographics to the demographics of the Advance study, while Table 4 gives the Advance study’s speakers’ information.





Gaillynn D. Clements

Table 3. Comparison of copula studies’ statistics Town, State/researcher

population location

economic situation

Meadville, MS/ Wolfram (1974)

700

isolated – in a national forest

in one of the poorest counties in MS

Anniston, AL/ Feagin (1979)

60,000

main highway between Birmingham and Atlanta

in one of the richest counties in AL

Warren County, NC/ Hazen (2000)

19,9721

NC-VA border eastern half of $23,0252 the state

Advance, NC/ Clements (2002)

5,8751

western Piedmont $39,8722 10–15 miles outside of the Winston-Salem/Greensboro/ High Point metropolitan area

1 U.S. 2 U.S.

Census Bureau: 2000 census Census Bureau: median household income, 1997 model. NC state average is $35,320.

Table 4. Speaker information for the Advance, NC study Speaker

Age

Gender

Class

Born/Raised in Advnace

GL CT LR DJ SD KM WH CC RE LN KH RB LB

26 28 38 44 54 60’s 85 21 59 59 60’s 67 73

f f f f f f f m m m m m m

MMC LMC MWC LMC LMC LMC MWC UWC1 LMC LWC MWC MMC LMC

no-moved here at 92 no-moved here at 73 yes yes yes yes yes no-moved here at 34 yes yes yes yes yes

Age categories: 15–29: younger, 30–64: middle-aged, and 65+: older. Class was decided by a points system based on their highest level of education, income per annum, and residence size. 1 CC still lives with his parents, so their residence size was used, but he has a full time job, so his income was used. 2 GL moved from Virginia (right above North Carolina) when she was 9. 3 CT moved from Georgia (right below North Carolina) when she was 7. 4 CC moved from the closest city (Winston-Salem, North Carolina) to Advance when he was 3.

Copula variability in rural Southern America

. The copula in Advance The copula consists of inflected forms of be, as am, is, and are, either as the main verb or an auxiliary. Copula variability in Advance includes regularization, not focused on here, absence and habitual be. Copula absence, as seen in examples (6)–(10), was produced by all different types of speakers. (6) You Ø right “You are right” (7) See it Ø just like that “See, it is just like that” (8) It Ø like it don’t bother him “It is like it does not bother him” (9) You Ø gonna show my ignorance “You are gonna show my ignorance” (10) You Ø pretty for a little girl “You are pretty for a little girl”

I did find instances of be in utterances, see examples (11)–(14) below. (11) There he be (12) I be ready (13) And their nose be runnin’ (14) He be sittin’ there

Examples (11) and (12) are like the invariant be found in Wright’s Early Modern prison narratives and also in Bailey and Maynor (1985), but nothing like examples (13) and (14) were found in the narratives or focused on in the other three copula studies of Southern white speakers. These two examples parallel the habitual be of be+verb+ing seen in African-American English. . Results of Advance, N.C. The overall breakdown of numbers for the copula in Advance is in Table 5. . Age The three age categories are younger, ages 15–29, middle-aged, 30–64, and older, 65+. The younger speakers had 8% is/are absence and only two invariant





Gaillynn D. Clements

Table 5. Percentages of copula activity in Advance

Overall percentages and numbers for is/are:

%

N

17.8 67.5 10.5 1.7 2.5

105/589 397/589 62/589 10/589 15/589

full contraction absence invariant be be+verb+ing

be forms and two be+verb+ing or 2.5% be constructions. The six middle-aged speakers had a total of 13% is/are absence and 3.7% be use, and, the four older speaker’s absent rate equals the middle-aged speakers for absence at 13%, and they used a be construction 7.6% of the time. Feagin also found in her middle/upper class speakers that older males and females had absent copula more than teenagers. On the other hand, in the urban working class speakers she found the opposite, that teenage males had more absence than older males, and then older males had more absence than older females, which supports what Wolfram found, that teenagers showed more copula absence than the older speakers. Hazen found absence at a gradually declining rate starting with the older speakers at 18% absence, the middle-aged speakers at 12%, and the younger speakers with only 2%. This could show that copula absence is decreasing or that it is a feature of age grading. It could also show that because of mass media, more interest in education, and less isolation in the rural South, people are exposed to more mainstream standard English, and many others are leaving the area for jobs and college where their vernacular may be seen as substandard, and so they shift to a more standard variety. This could be the case for my data as well since my three younger speakers all either were in college and graduated or are still in school and have absent copula 50% less than the middle-aged and older speakers. . Gender The six females in this study show 8% copula absence and 1.9% be use while the seven males show 18% absence and 9% be use. The males have more absent tokens, which is comparable to what Feagin found, and Hazen who found males had 12% absence while females only had 6% absence. Concerning are absence, Feagin found that older men – whether upper class, urban working class, or rural working class – had higher absence rates than women or younger speakers. Wolfram’s findings were different; he found the highest percentage of absence in teenage girls with teenage boys a close sec-

Copula variability in rural Southern America

ond. In regard to is absence, Wolfram found that teenage boys had the highest rate of absence, while Feagin found males, in general, had a higher frequency of absence. In her upper class and rural working class speakers older men had more is absence, whereas in urban working class, teenage boys had the highest percent of is absence than any other group. The fact that in most cases males have a higher frequency of copula absence probably stems from males using stigmatized forms as identity markers. . Class Overall the working class speakers had a higher rate of copula absence than the middle class speakers. The working class and middle class speakers had almost equal rates of uninflected be, 4.6% for the working class and 4.5% for the middle class. Wolfram relates his categorical are absence results to social class and linguistic environment. The lower class has are absence more frequently when the preceding environment is a pronoun, and the higher classes show are absence more frequently when the following environment is gonna. I also had two older male speakers with categorical are absence: RB and LB. From what I have here, my speakers act the same as Wolfram’s. The lower middle class speaker, LB, has categorical absence after pronoun, and the middle class speaker, RB, has absence before gonna. . Linguistic environments – Absence only Hazen (2000) and Wolfram (1974), have found highly favored environments for is/are absence when the subject is a Pro, or a NP, which is a little different than what I found. I only had 5 null subject tokens of which 2 were absent, 5 tokens is not a lot in comparison with Pro, which had 40 and NP, which had 10. If I take out the null subject category, then I am left with the scale others have found, which shows that absence favors Pro and then NP. Copula absence also prefers certain following environments over others; Hazen (2000) found absence more likely, for his white speakers, when the following environments were V+ing>NP>AdjP; he found 0% absence with Loc and Gonna. But Hazen follows Fasold (1990) and compensates for low copula absence with copula contraction, where a more familiar pattern emerges: Gonna>V+ing>Loc>AdjP>NP, which is also seen in African-American Vernacular English in Labov (1969) and Cukor-Avila (2001).





Gaillynn D. Clements

. African-American Influence Lastly, since there is no substantial historical evidence of copula absence and no evidence at all of be+verb+ing in British English, and if they are not American innovations, then the only other dialect, which could have influenced Southern speakers, is African-American Vernacular English or AfricanAmerican English. Was the percentage of African-Americans in the area high enough over the years to allow and even promote borrowing of features between the AfricanAmericans and white speakers? If so, then there could have been influence on the white speakers. Because of the decline of farming and the decline in need for extra workers because of smaller families and modern conveniences, the younger speakers do not have the close contact with AAVE speakers like older speakers did, and this could be the reasoning for the younger speakers having less copula absence and be. I researched African-Americans in Davie County and Advance in particular, what I found was quite interesting. In the 1790 Federal census, which was the first conducted, Davie County/Advance had not yet been named and incorporated, so I traced settling families that moved into the Heidelberg settlement, which would later be named Advance. By the time of the 1850 census Davie County had been formed. There were 5695 free people: 2804 white males, 2809 white females, 45 African-American males, and 37 African-American females. Several of the people listed in this as coloreds census were mulattos which further supports the physical closeness between whites and blacks at the time. Forty-six of the African-Americans counted as free coloreds were actually mulattos. There were 2170 slaves in Davie County. Adding the free and enslaved blacks gives us 2252 blacks in Davie County making blacks 28.6% of the total population. By the time of the 1860 census, most families, 175 out of 260 or 67% that had slaves only listed 1–3 slaves for poll taxes. Males were the only ones listed for taxes, so I cannot be sure about the number of female slaves per household. Personal interviews with RB and LB revealed that even after slavery in 1920’s until the 50’s and 60’s, there were African-Americans who worked on farms and in houses in Advance, side by side with the white workers; this close daily contact would have maintained the white speakers’ copula absence and habitual be. By 1950, the number of African-Americans in Davie County had decreased to 2150 making blacks only 13.9% of the total population, and by 2001 African-Americans only made up 6.8% of the population. Along with the decrease in numbers of African-Americans comes decrease in contact among whites and blacks. This

Copula variability in rural Southern America

is further support for my contact argument; since contact has decreased, the number of shared features or instances of copula absence and be+verb+ing in the younger white speakers has decreased also.

. Conclusion All of this information is interesting, but what does it mean for the copula absence and variability discussion? Since many of the families in Davie County and Advance only had a few slaves making for closer relationships with the slaves, even to the point of whites having children with slave and free blacks, then this is supporting evidence that there was a high amount of contact between the whites and the slave and free blacks. This type of contact is what is needed, and it has already been shown that changes from nonprestigious to prestigious varieties can happen. Since copula absence and habitual be are not from British English varieties, not American innovations2 from r-lessness, there is only one source left. African-Americans’ speech is the only other speech community with these features that has most likely influenced the speech of the whites in Advance. With the isolation of the South and the closeness of whites and blacks in rural areas like Advance, the need for ease of communication and for purposes of self-identification with similar backgrounds of people, copula absence and habitual be +verb+ing were added to the optional variants for white speakers in Advance, but these two features are now decreasing as variants. What does this mean for the Southern American dialect and AfricanAmerican English? It supports the theory that these two are diverging not converging as many argue. For many of the middle-aged and older speakers, these copula options in white speech identify them with the African-American speakers from the same area who went through the same hard work and rough times – rurality was the common bond. The decreased amount of contact and decreased amount of shared identity has caused the younger white speakers to start eliminating copula absence and habitual be +verb+ing – rurality is no longer their common ground.

Notes . During and post-slavery times even though most rural Southern white Americans did not speak what all others would call the prestigious variety in America, it was still the variety of





Gaillynn D. Clements

the dominate race, whites. But, as the South was more isolated than the North and the racial power dynamic was slightly shifting, the need for communication overcame the need to remain linguistically separate. . There could have been another path of American innovation involved in these two forms. There is no way to know for sure. The likelihood of both coming from different paths of innovation seems less possible than the explanation of African-American influence.

References Bailey, Guy & Natalie Maynor (1985). “The present tense of be in white folk speech of the Southern United States”. English World Wide, 6 (2), 199–216. Cukor-Avila, Patricia (2001). “Co-existing grammars: The relationship between the evolution of African American and Southern White Vernacular English in the South”. Sociocultural and historical Contexts of African American English (pp. 93–128). Philadelphia: John Benjamins. Dillard, J. L. (1972). Black English: Its history and usage in the United States. New York, NY.: Random House. Fasold, Ralph W. (1981). “The relation between black and white speech in the South”. American Speech, 56, 163–189. Fasold, Ralph W. (1987). “Are Black and White Vernaculars Diverging?”. American Speech, 62 (1), 55–62. Fasold, Ralph W. (1990). The Sociolinguistics of Language. Cambridge, MA: Blackwell. Feagin, Crawford (1979). Variation and change in Alabama English: A sociolinguistic study of the white community. Washington, DC: Georgetown University Press. Feagin, Crawford (1997). “The African contribution to Southern states English”. Language Variety in the South Revisited (pp. 123–139). Tuscaloosa, AL: University of Alabama Press. Fischer, John (1958). “Social influences on the choice of a linguistic variant”. Word, 14, 47– 56. Hazen, Kirk (2000). Identity and ethnicity in the rural South: A sociolinguistic view through past and present BE. Durham, NC: Duke University Press. Labov, William (1963). “The social motivation of a sound change”. Word, 19, 273–309. Labov, William (1966). The social stratification of English in New York City. Washington, DC: Center for Applied Linguistics. Labov, William (1969). “Contraction, deletion, and inherent variability of the English copula”. Language, 45 (4), 715–762. Labov, William (1972). Language in the Inner-City: Studies in the Black English vernacular. Philadelphia, PA: University of Pennsylvania Press. Labov, William et al. (1968). A study of the nonstandard English of Negro and Puerto Rican speakers in New York City. Report in Co-operative Research Project 3288. New York, NY: Columbia University. McElhinny, Bonnie (1993). “Copula and auxiliary contraction in the speech of white Americans”. American Speech, 68 (4), 371–399.

Copula variability in rural Southern America

Montgomery, Michael et al. (1993). “The black men has wives and Sweet harts [and third person plural -s] jest like the white men: Evidence for verbal -s from written documents on 19th century African American speech”. Language Variation and Change, 5, 335–357. Rickford, John (1985). “Ethnicity as a sociolinguistic boundary”. American Speech, 60 (2), 99–125. Rickford, John (1998). “The development of African-American vernacular English: Focusing on the creole origins issue”. In Salikoko Mufwene, John Rickford, Guy Bailey & John Baugh (Eds.), African-American English (pp. 154–200). London, Eng.: Routledge. Rickford, John, Arnetha Ball, Renee Blake, Raina Jackson & Nomi Martin (1991). “Rappin on the copula coffin: Theoretical and methodological issues in the analysis of copula variation in African American Vernacular English”. Language Variation and Change, 3, 103–132. Rickford, John & Renee Blake. (1990). “Copula contraction and absence in Barbadian English, Samana English, and vernacular Black English”. Berkley Linguistics Society, 16, 257–268. Romaine, Suzanne (1981). “The status of variable rules in sociolinguistic theory”. The Journal of Linguistics, 17, 93–119. Romaine, Suzanne (1982). Socio-historical linguistics: Its status and methodology. Cambridge, Eng.: Cambridge University Press. United States Census Bureau. 10-15-01. <www.census.gov> Visser, F. Th. (1970). An historical syntax of the English language. Leiden, Netherlands: E.J. Brill. Walker, James (2000). “Rephrasing the copula: Contraction and zero in early AfricanAmerican English”. In Shana Poplack (Ed.), The English History of African-American English (pp. 34–72). Malden, MA: Blackwell Publishers. Winford, Donald (1992). “Another look at the copula in black English and Caribbean creoles”. American Speech, 67, 21–60. Wolfram, Walt (1969). A sociolinguistic description of Detroit Negro speech. Arlington, VA: Center for Applied Linguistics. Wolfram, Walt (1974). “The relationship of white Southern speech to vernacular black English”. Language, 50, 498–527. Wolfram, Walt, Kirk Hazen & Jennifer Ruff Tamburro (1997). “Isolation within isolation: A solitary century of African-American vernacular English”. Journal of Sociolinguistics, 1, 7–38. Wolfram, Walt & Natalie Schilling-Estes (1998). American English. Malden, MA: Blackwell. Wright, Laura (2002). “Five syntactic features present in AAVE/SWVE and Early Modern London English”. Paper presented at the Southeastern Conference on Linguistics. Memphis, TN., April 2002.



Main stress left in Early Middle English* B. Elan Dresher and Aditi Lahiri University of Toronto / University of Konstanz

.

Pertinacity in grammar

Lahiri (2002) has called attention to a characteristic of grammars called pertinacity. A rule or pattern may persist over time, though its realization may change. An example is the persistence of a particular metrical pattern in a language, though it may apply to new forms and no longer apply to forms that it used to apply to (e.g., the Germanic Foot, Lahiri & Dresher 1999). This type of pertinacity can be summed up as: same pattern, different output realization. We will also look at an example of the converse kind of pertinacity. It concerns persistence of output forms despite changes in the grammar. This type of change can occur under various conditions. Such change always involves a reanalysis of the output form, provoked by changes elsewhere in the system. This type of pertinacity can be summed up as: different pattern, same output realization. Since learners acquire their grammars guided by the output forms they are exposed to, we do not expect these forms – especially those that make up the ‘core’ or ‘primary’ data – to change in the course of acquisition. Reanalysis of grammar that does not involve an immediate change in output forms is thus a significant type of language change.

. Change in the English stress system Our example is the change from the Old English Germanic stress system to the Modern English Latinate stress system. This represents a radical change:



B. Elan Dresher and Aditi Lahiri

(1) Germanic: stress on the stem-initial syllable, regardless of quantity, building secondary stress from left to right. a. L L H L

b. L H L

c. L L L

(2) Latin: stress on the penult if heavy, otherwise on the antepenult, secondary stresses from right to left. a. L L H L

b. L H L

c. L L L

Did a new group of people take up English and bring with them their native Latinate prosody? We know this is not what happened. But even if we did not know this directly, we would have been tipped off by a peculiar fact. Through the changes in the grammar of stress, all native Old English words retained their output stress contours: wáter, hópefulness, begín all retain their original stress contours, though the metrical structures that underlie them have changed. This fact suggests the change was carried out by by native speakers who maintained continuity with the stress patterns of previous generations. So what brought about the change? Contrary to Halle & Keyser (1971), who place the origins of the change in the time of Chaucer, we date the important innovations to a later time, due to the influence of Latin borrowings. If this is correct, we must answer another question: How are borrowings able to effect such a big change in the system? Our hypothesis is that borrowings can be decisive when the core native vocabulary does not decide between grammars. The pertinacity of surface stress contours of native forms suggests the change was carried out by native speakers, influenced by new vocabulary that resolved ambiguities in the grammar.

. An early generative account: Halle & Keyser (1971) Halle & Keyser (1971) propose that the Romance stress rule was added to English in the time of Chaucer. This rule originally competed against the dominant Old English stress rule, and was gradually extended over the subsequent centuries. According to Halle & Keyser, the Romance stress rule subsumed two different patterns commonly attributed to separate stress rules: A. the French pattern responsible for Chaucer’s final stress in words like honóur and vertú.

Main stress left in Early Middle English

B. the Latin pattern – stress on the penultimate syllable iff heavy, otherwise on the antepenult – that came to be dominant in later English. The great advantage of this scenario, with respect to our topic, is that it accounts for the origin of the Latin stress pattern in English. The relatively few early borrowings from Latin are now reinforced by the more numerous words with the French stress pattern (since the French and Latin stress rules are united in this analysis). The later flood of Latin borrowings could thus simply provide further evidence for a pattern that had already gained a foothold in English. Unfortunately, this account appears to be wrong in a number of respects. The first problem is that, under any analysis, the French and Latin stress rules are different. (3) The French stress rule a. Stress the final vowel unless it is schwa: abbót, chanóun, degrée, honóur, vertú; b. Otherwise, stress the penultimate vowel: divíne, Egípte, exíled, govérne, servíce. (4) The Latin stress rule a. Stress a tense final vowel: chanóun, degrée, honóur, vertú; b. Otherwise, stress the penultimate syllable iff it is heavy (either having a tense vowel or closed by a consonant): Caríbdis, divíne, govérne, Neptúnus; c. Otherwise, stress the antepenult: Cappáneus, Týdeus, Zépherus.

The stress rules differ in two cases. When the final vowel is lax but not schwa, the French rule gives final stress (5), and the Latin rule would give nonfinal stress (6): (5) French: abbót, Jhesús, Judíth, Oréb, tempést (6) Latin: ábbot, Jhésus, Júdith, Óreb, témpest

In the above examples alternants with initial stress would be generated both by the Germanic and the Latin stress rules. However, under the Latin stress rule there would be no source for the forms in (5). Thus, Halle & Keyser (1971) must mark these as exceptions to the unified Romance stress rule. A second discrepancy occurs when a word has more than two syllables, where the final vowel is schwa and the penult is in a light syllable. In such cases, the French rule gives penultimate stress, but the Latin rule gives antepenultimate stress. According to Halle & Keyser (1971), Old French words all had heavy penults, thus avoiding a conflict in words borrowed from that source.





B. Elan Dresher and Aditi Lahiri

But many words borrowed into English from Latin had light penults and followed the Latin, not the French rule, undermining a unified approach to Old French and Latin stress patterns. The second problem with the Halle-Keyser account is that the French stress pattern had no lasting effect on English prosody. It should be noted that words with French stress in Chaucer could in general also be stressed according to the native English pattern, as required by the meter. Hence we find many doublets: (7) citée ∼ cítee; comfórt ∼ cómfort; divérs ∼ díverse; geáunt ∼ géant; Plató ∼ Pláto; presént ∼ présent.

With very few exceptions, the Present Day English reflexes of Romance words with French stress in Chaucer have initial stress consistent with Germanic stressing: (8) ábbot, bárren, cíty, cómfort, díscord, Égypt, fórtune, gíant, góvern, hónour, Jésus, Júdith, mércy, Pláto, présent, sérvant, sólemn, témpest, tórment (noun), týrant, vírtue.

In addition, we find PDE ascéndant (cf. ascénd, Chaucer ascendént), cánon (Chaucer chanóun), and purvéyance (cf. purvéy, Chaucer purveyáunce; the more usual Middle English form was púrvey-). There are only a few exceptions to the above generalization: degrée, divíne, and rewárd, among others, have final stress. More generally, bisyllabic Romance loans borrowed before the fifteenth century have initial stress in Present Day English (see also Lahiri & Fikkert 1999 and Svensson & Hering 2003): (9) Stem vowel is short in Present Day English talent (893), baron (1200), senate (1205), jealous (1250), palace (1290), channel (1300), gallon (1300), panel (1300), coral (1305), profit (1325), metal (1340), satin (1366), moral (1380), volume (1380), second (1391), Latin (1391). (10) Stem vowel is long in Present Day English basin (1220), moment (1240), vacant (1290), odour (1300), process (1330), paper (1374), raisin (1382), patent (1387), famous (1400).

By contrast, bisyllabic Romance loans with final stress in Present Day English tend to have been borrowed much later: (11) cement (1300) (but ME síment had initial stress until the 19th c.), canal (1449), bourgeois (1564), gazelle (1582/1700), moustache (1585), gazette

Main stress left in Early Middle English

(1605), hotel (1644), champagne (1664), ballet (1667), salon (1715), bouquet (1716), brochure (1765), beret (1850).

A third problem with the Halle-Keyser account is that the Latin stress pattern, as distinct from the Old French one, is hardly attested in Chaucer. We have argued that the Latin stress pattern is distinct from that of Old French. Thus, evidence for the introduction of a Latin-type stress rule into English in Chaucer’s time must rest on words that particularly exemplify this pattern. However, such words are quite rare in Chaucer, and tend to be Latin names. Halle & Keyser cite the following instances: (12) C˘appán˘eu˘ s, C˘aríbd˘ıs, N˘eptún˘us, S˘atúrn˘es, Týd˘eu˘ s, Zéph˘er˘us.

We conclude that there is no evidence that either the Old French or the Latin stress rule gained a foothold in English at the time of Chaucer (see also Minkova 1997; Redford 2003). We must look to a later period for the introduction of the Latin stress pattern.

. A Parametric Account Our hypothesis is that the Latin stress rule was not added all at once to the grammar of English. A stress system is the result of interacting parameters. These parameters can change independently. In the case of English, the main changes involved: A.

a change in directionality (parsing from the left vs. parsing from the right); and

B.

the position of main stress (left vs. right).

. Old English stress (Dresher & Lahiri 1991) We assume that the Old English foot type is the Germanic Foot, a resolved and expanded moraic trochee of the form ([head] dependent), where the head must consist of at least two moras and the dependent may have at most one mora. The two moras of the head do not have to come from the same syllable. The direction of parsing is left to right, and main stress is on the left.





B. Elan Dresher and Aditi Lahiri

(13) Old English stress: Sample parsings a. (x .) b. (x .) ([µµ] µ) ([µ µ] µ) H L L L L wor da we ru da

c. (x .) ([µ µµ] µ) L H L cy nin ga

In (13a), the initial heavy syllable has two moras and occupies the head of the foot; the second syllable is light (one mora), and occupies the dependent branch. In (13b), the initial syllable is light, and so the second light syllable joins it (a process called Resolution) to make up the head position of the foot. The third syllable occupies the dependent position. (13c) is similar, except resolution is with a heavy syllable. The forms in (14a–c) illustrate High Vowel Deletion, whereby a short high vowel in an open syllable is deleted when it would occupy the dependent position of a foot. Underlined vowels are deleted; observe that a high vowel is not deleted in closed syllables (14d), or when it is resolved with the initial syllable (14c, e), or when it falls outside the foot (14f). (14) High Vowel Deletion in Old English a. (x .) . b. (x .) ([µµ] µ) µ ([µµ] µ) H L L H L wor du h¯ea fu de d. (x ([µµ] H wor

.) µ) H dum

e. (x) ([µ µ]) L L lo fu

c. (x ([µ L we

.) µ]) µ L L ru du

f. (x ([µµ] H n¯i

.) µ L te

. µ) L nu

Old English lacked secondary stress in final syllables. In (15), the underlined final syllables might be expected to have a secondary stress (because they ought to be the head of a foot), but they do not. (15) Final syllables a. (x) ([µµ]) ([µµ]) H H ðer o¯´ c. (x) ([µ L æ´

µ]) L ðe

([µµ]) H ling

b. (x) (x ([µµ]) ([µµ]) H H o¯´ ðèr

.) nµ) L ne

d. (x) ([µ L æ´

(x) ([µµ]) ([µµ]) H H lìn ges

µ]) L ðe

Main stress left in Early Middle English

In the earlier period, when long vowels could occur in final syllables, lack of final stress has to be accounted for by Final Destressing (FD): (16) Final Destressing (FD) Defoot a final foot that does not carry main stress and that has no dependent branch.

Later, long vowels in unstressed final syllables were shortened. Therefore, the only word-final syllables that appeared to be heavy were those ending in a consonant. The fact that such syllables did not receive a secondary stress was subject to reanalysis in terms of Final Consonant Extrametricality (CEM): (17) Final Consonant Extrametricality (CEM) Final consonants are extrametrical.

Another rule that came to play a role in the transition from Old to Middle English was Trisyllabic Shortening (TSS): (18) Trisyllabic Shortening (TSS) A stressed long vowel is shortened when preceding two unstressed syllables.

. Middle English stress The changes sketched above had no effect on the position of main stress, and the stress system in Middle English remained essentially as in Old English. However, the various changes did have the effect of metrically ‘shortening’ words. Thus, many words that had more than one foot in Old English were reduced to a single foot in Middle English (Lahiri & Dresher 1999: 709). (19) Metrical shortening from Old to Middle English OE *he¯´ rìnges *la¯´ vèrke *c¯ı´cenes FD ([H])([H])([H]) ([H])([H]L) ([H] L)([H]) CEM ([H]) ([H]L) — ([H] L) L TSS ([L H] L) ([L H] L) ([L L] L) ME héringes láverke cícenes

*cla¯´ vere ([H] L)L — ([L L] L) clávere

Old English words already tended to be short. Moreover, many Old English suffixes were, as they still are today, ‘stress neutral’, meaning they do not participate in the stress domain. Adding the further metrical shortenings described above, native English words tended to be no longer than a single foot. There-





B. Elan Dresher and Aditi Lahiri

fore, evidence for setting the parameters of directionality and main stress was in short supply. Among the Latin words that began entering the language in great numbers in the sixteenth century were many that were relatively long. These Latin loan words were thus able to fill the gap left by the native words. Without contradicting the majority of the native words, the loan words eventually caused the resetting of the directionality and main stress parameters. . Early Latin borrowings We follow Lahiri & Fikkert (1999) in claiming that Latin words were originally borrowed as morphologically simplex (see also Minkova & Stockwell 1996). Thus, reverence was not initially derived from revere, nor austerity from austere. Often, the ‘derived’ word was borrowed earlier. This hypothesis accounts for the stress patterns of these words, and provides evidence that direction of parsing and placement of main stress had not changed before 1530. Consider words exhibiting so-called ‘Medial Laxing’. The stressed vowel in the ‘underived’ word is unstressed and laxed in the ‘derived’ word: (20) Medial Laxing (‘underived’ ∼ ‘derived’) admíre ∼ ádmirable; confíde ∼ cónfident; presíde ∼ président, présidence; reláte ∼ rélative; revére ∼ réverence.

These words are problematic in all morphophonological analyses, including those of Liberman & Prince 1977 (morphological shortening); Kiparsky 1979 (sonorant destressing); Myers 1987 (Medial Laxing); and Kager 1989 and Gussenhoven 1994 (lexical exceptions). The main point of interest here is that in Medial Laxing alternations, ‘derived’ forms have stress consistent with the Old English pattern. (21) ‘Derived’ words borrowed earlier than ‘underived’ words abstain (1380) ∼ abstinence (1300); confide (1455) ∼ confidence (1430); reside (1460) ∼ resident (adj.) (1382); finite (1493/1597) ∼ infinite (1385); potent (1500) ∼ impotent (1390): preside (1611) ∼ president (1375); revere (1661) ∼ reverence (1290).

. Changes in direction of parsing and main stress The main stress parameter did not change together with directionality. We assume the following sequence:

Main stress left in Early Middle English

(22) Approximate dates of changes in metrical structure Foot = Resolved moraic trochee throughout. 1400: Foot direction left, Main stress left (as in Old English). 1530: Foot direction right, Main stress left. 1660: Foot direction right, Main stress right.

.. Change of direction of parsing The preceding forms show that it was not sufficient to borrow Latin words to provoke a change in directionality. Following commentators such as Danielsson (1948) and Poldauf (1981), we associate this change with the introduction of words with Latin suffixes such as -abl/-ible, -ation, -ic(al), -ity, -ator, etc. In such forms, stress is computed from the right side. Compare the analyses of cómparable and résidence, borrowed when direction of parsing was still from the left, with those of sevérity and rárity, borrowed after the change in parsing direction. Notice that the change in direction is evident only in (23a) and (24a), not in (23b) and (24b). (23) Early borrowings: Foot direction left, Main stress left a. (x .) (x) b. (x .) ([µµ] µ) ([µ µ]) ([µ µ] µ) com pa ra ble (1413) re si de (1386) (24) Later borrowings: Foot direction right, Main stress left a. . (x .) b. (x .) µ ([µ µ] µ) ([µ µ] µ) se ve ri ty (1530) ra ri ty (1560)

.. The ‘Countertonic Principle’ Danielsson (1948) attributes to Walker (1791) the observation that classical words were pronounced, in the English pronunciation, with alternating secondary stresses two before the tonic (e.g., Latin àcadémia). When Englished, the tonic and countertonic change places to conform to English ‘speech habits’ (e.g., ácadèmy). Reference here is specifically to the habit of putting the main stress left. The Countertonic Principle shows that the main stress parameter remained set to left for some time after the change of directionality to right. It is worth noting that the addition of words stressed according to the Countertonic Principle would have increased the evidence for main stress left. Thus, a word like ácadèmy clearly shows two feet, of which the left has the main stress. There-





B. Elan Dresher and Aditi Lahiri

fore, it is not correct to say that English gradually moved from a ‘Germanic’ to a ‘Romance’ stress system. In this case, the same words that provoked a change of directionality to right reinforced the evidence for main stress left. .. Main stress right What exactly caused the main stress parameter to finally switch to right is not entirely clear to us. However, a likely place to look is around or before 1660. According to Danielsson (1948: 29), that year was the ‘turning point’ when French words kept final accent in English, as with suffixes like those in (25). (25) Suffixes retaining main stress -ade, -ee, -eer, -esque, -ette, -oon. (26) Words with final stressed suffixes in Present Day English parade (1656), payee (1758), cannoneer (1562), grenadier (1676), arabesque (1611), musette (1811), bassoon (1727).

Though some words like those in (26) may have entered the language before 1660, they may not have systematically retained final stress until around that date. It is plausible to suppose that final stress in words with these suffixes became more systematic after the change of main stress to right.

. Conclusion: Conservatism amid change It emerges from our analysis that both the core grammar (foot type, quantity sensitivity) and the core data (surface stress patterns) remain essentially unchanged in the course of seemingly radical changes to the English stress system. Change occurred most readily in areas of the grammar where the native vocabulary did not provide decisive cues. Only in these areas could the new loan vocabulary provide native speakers with the key evidence that led to a reanalysis of the grammar.

Note * We would like to thank Paula Fikkert, Astrid Kraehenmann, Martha McGinnis, and Frans Plank for their comments and assistance. This research was partly supported by a grant (4102003-0913) from the Social Sciences and Humanities Research Council of Canada to Elan Dresher, and by funds from the German Research Foundation (Sonderforschungsbereich 471 and the Leibniz Prize) to Aditi Lahiri.

Main stress left in Early Middle English

References Danielsson, Bror (1948). Studies on Accentuation of Polysyllabic Latin, Greek, and Romance Loan-words in English. Stockholm: Almqvist & Wiksells. Dresher, B. Elan, & Aditi Lahiri (1991). “The Germanic Foot: Metrical Coherence in Old English”. Linguistic Inquiry, 22, 251–286. Gussenhoven, Carlos (1994). “English Stress in Lexical Phonology”. In Wolfgang U. Dressler, Martin Prinzhorn, & John R. Rennison (Eds.), Phonologica 1992: Proceedings of the 7th International Phonology Meeting (pp. 87–96). Torino: Rosenberg & Sellier. Halle, Morris, & Samuel Jay Keyser (1971). English Stress: Its Form, its Growth, and its Role in Verse. New York: Harper & Row. Kager, René (1989). A Metrical Theory of Stress and Destressing in English. Dorderecht: Foris. Kiparsky, Paul (1979). “Metrical Structure Assignment is Cyclic”. Linguistic Inquiry, 10, 421– 441. Lahiri, Aditi (2002). “Pertinacity in Representation and Change”. Paper presented at the Workshop on Pertinacity, Schloss Freudental, July 10–14, 2002. Lahiri, Aditi, & B. Elan Dresher (1999). “Open Syllable Lengthening in West Germanic”. Language, 75, 678–719. Lahiri, Aditi, & Paula Fikkert (1999). “Trisyllabic Shortening in English: Past and Present”. English Language and Linguistics, 3 (2), 229–267. Liberman, Mark, & Alan Prince (1977). “On Stress and Linguistic Rhythm”. Linguistic Inquiry, 8, 249–336. Myers, Scott (1987). “Vowel Shortening in English”. Natural Language and Linguistic Theory, 5, 485–518. Minkova, Donka (1997). “Constraint Ranking in Middle English Stress-shifting”. English Language and Linguistics, 1, 135–175. Minkova, Donka, & Robert P. Stockwell (1996). “The Origins of Long-short Allomorphy in English”. In Jacek Fisiak & Marcin Krygier (Eds.), Advances in English Historical Linguistics 1996 (Trends in Linguistics. Studies and Monographs 112) (pp. 211–239). Berlin: Mouton de Gruyter. Poldauf, Ivan (1981). “The Genesis of Terminational Stress in English”. Lingua, 54, 335–359. Redford, Michael (2003). “English Stress Doubles: New Evidence from Chaucer’s Meter”. In Paula Fikkert & Haike Jacobs (Eds.), Development in Prosodic Systems (Studies in Generative Grammar 58) (pp. 159–196). Berlin: Mouton de Gruyter. Svensson, Ann-Marie, & Jürgen Hering (2003). “Germanic Prosody and French Loanwords”. Paper presented at NWAVE 32, University of Pennsylvania, Philadelphia, PA, October 9–12, 2003. Walker, John (1791). A Critical Pronouncing Dictionary and Expositor of the English Language. London: Robinson.



Some dialectal, sociolectal and communicative aspects of word order variation and change in Late Middle English* Tamás Eitler ELTE University, Budapest

.

Introduction

Anyone studying Middle English will sooner or later have to account for a certain degree of variation within a single text or within all the texts by the same author. The factors causing variation include dialect or sociolect contact, differences in genre or text type, scribal interference in the case of copying, and influence from different source languages in the case of translation. Through two case studies, the present paper examines intrapersonal syntactic variation. It will be stated that the differences in the use of the word order variants V2 and V3 in the Late Middle English period can also be ascribed to communicative factors including accommodation to the sociolectally and dialectally determined target audience. Of these two determining factors, regional input will be claimed to have been primary, whereas sociolects will be argued to be partly derivable from dialects. As the examined texts were written by authors with an East Anglian background, the East Anglian input into the late mediaeval London sociolects will be examined in greater detail. Finally, as a “by-product” of the research, it will be claimed that on the basis of word order characteristics, the authorship of the manuscript of The Equatorie of the Planetis can be assigned to Geoffrey Chaucer. It is hoped that this will demonstrate that even the so far somewhat neglected syntactic criteria can aid in linguistic reprofiling and author identification.



Tamás Eitler

. Researching word order competition In the course of its history, English changed from an inflecting V2 language with SOV base word order to an analytic V3 language in which word order is fixed as SVO. By 1400 Middle English had started to dispose of the V2 variant in the dialects underlying the standard (van Kemenade 1987: 183ff.). In the present paper only cases of main clause non-operator fronting (as defined by Haeberli 2000) with V2 or V3 order are dealt with. The Middle English period saw the competition of the two parametric variants V2 and V3 in similar syntactic contexts for an extended period of time. The word order change stemming from this variation seems to have been a gradual process at the level of language use. However, as one can find variation between V2 and V3 within the same non-copied text written by the same individual, it becomes highly questionable whether variation is simply a performance phenomenon only. According to the variationist approach, as a result of acquisition of both the older and the novel parameter settings, variation is located within the individual who can be claimed to have syntactic diglossia. It is exactly this sort of syntactic diglossia which must have existed for many Middle English speakers, and since two mutually exclusive variants cannot be tolerated without functional differentiation, the linguistic outcome was the gradual ousting of the older V2 variant in the relevant linguistic contexts. Some of the stages of this gradual replacement can be seen in the chosen texts, displaying different usage frequencies even within the same author’s oeuvre, which variation cannot be ascribed to mere chronological differences. Intrapersonal syntactic variation can be caused by many factors. The possible external factors include the influence of translation and scribal interference. Also, sociolinguistic factors such as class, age and gender can influence the nature of the given variational phenomenon. Furthermore, intrapersonal but intertextual variation can appear due to the different types of texts which may tolerate innovative, spoken or more vernacular forms to a various extent. A further dimension, register variation, seems to be dependent on the factors underlying the given interaction. One such factor can be the target audience, whose linguistic behaviour and expectations may trigger accommodation on the part of the encoder, who vies for successful communication, a component of social success (Keller 1994: 86–90). According to the Communication Accommodation Theory (Coupland et al. 1991), accommodation takes place when the interlocutors converge their speech in order to increase their linguistic similarities, which promotes the success of the social interaction. Using the audience design model it can be predicted that this audience-designed shift

Word order in Late Middle English

will appear in the use of only those linguistic forms which are available in the linguistic repertoire (hence in the internalised grammar with competing parameter settings) of the speech community to which the interlocutors belong (Bell 2001: 145). On the basis of the aforementioned possibilities, the principal question to be answered in Section 8 is whether it is the text type that influences word order (and hence it can cause intrapersonal variation) or there is another decisive factor which is immediately related to the embedding of the interaction. Variation can be the manifestation of ongoing changes in the language. The two or more mutually incompatible parametric variants are vying with each other and gradually, one of them (sometimes dubbed as conservative, outgoing) will be ousted by the other one (dubbed as innovative, novel or incoming). This gradual replacement of the conservative variant with the novel one can be displayed graphically. The trajectory of the replacement is idealtypically an S-shaped curve, which is based on the logistic function (Kroch 1989: 203– 204). The existence of the S-shaped trajectory of linguistic replacement has been tested through various computer simulations. These simulations showed the logistic function in the case of variation between variants of equal fitness (Wonnacott 2000: 30–34). However, there seem to be no attempts to locate texts along an S-curve on the basis of their word order syntax as the S-curve for the whole replacement is lacking, too. In light of this, the position of the examined texts will be identified and explained in the contemporary linguistic context and in the word order replacement trajectory alike. A recent strand of investigation is concerned with identifying the sociolects of the mediaeval City of London and the regional input into these sociolects. This interest may be due to the fact that no generally accepted account exists as to the exact regional source of the emerging standard. If one assumes that the selection of linguistic forms and structures into the standard was socially motivated, the identification of the regional population input into the social classes of London may lead to the identification of the regional linguistic input into the sociolects which could have shaped the incipient standard. Since both of the authors examined in this paper come from the East Anglian dialect area, whose role in shaping the standard has recently been given a lot of attention (Kristensson 2001: 71–77), it seems in order to identify (1) the word order syntax of the upper class sociolect, spoken by the wealthy immigrants from East Anglia; (2) which London sociolect Chaucer’s idiom represents. Once Chaucer’s indigenous idiom is identified on the basis of the two works allegedly written initially for his son, and once this idiom is contrasted with the one used in his literary works written for a different target audience with a different idiom, it





Tamás Eitler

will be possible to decide whether MS Peterhouse 75.1 can really be assigned to Chaucer on syntactic grounds. In the light of the aforementioned interrelatedness of dialectal and sociolinguistic factors in shaping the incipient standard, it will be interesting to explain additionally (1) why the systematic CP-V2 syntax of the East Anglianbased immigrants forming the upper section of London’s society was not selected as the word order of the standard; (2) how exactly the incoming V3 variant spread to East Anglia. During the research the second version of the computerised Penn-Helsinki Parsed Corpus of Middle English (PPCME) was used, which is a morphologically tagged and syntactically analysed collection of Middle English texts (Kroch & Taylor 2000). For the purpose, four texts were chosen for closer examination: two by Geoffrey Chaucer and two by John Capgrave. Geoffrey Chaucer is represented by his Treatise on the Astrolabe (hereafter Astrolabe) and tentatively by The Equatorie of the Planetis (hereafter Equatorie). For comparison, data were drawn from Haeberli (2000), who besides the Astrolabe examined three literary prose works as well: Boethius and two prose tales from the Canterbury Tales: The Tale of Melibee and The Parson’s Tale. As Haeberli provided only conflated data for Chaucer (reproduced in Table 1), the aggregated word order usage frequencies in Chaucer’s prose literary works (in Table 1) were arrived at by deducing the number of V2 and V3 clauses in the Astrolabe from Haeberli’s respective conflated figures of V2 and V3 clauses for his Chaucer corpus. John Capgrave is represented by his Sermon and his Abbreuiacion of Cronicles (hereafter Chronicle), from which data were drawn from Haeberli (2000) and from a previous investigation (Eitler 1999). As a comparison, data will be drawn from the Wycliffite Old Testament, investigated by Haeberli (2000). Both authors descended from East Anglian families. Chaucer’s family originated from Ipswich, Suffolk, whereas Capgrave came from King’s Lynn, Norfolk. Capgrave’s idiom is unquestionably East Anglian, whereas research into Chaucer’s phonology could detect East Midlands traces in his texts (Kroch & Taylor 1997: 324, Note 16). Chaucer’s Astrolabe was written in 1391; however, the extant authoritative manuscript was copied in about 1450, possibly being at one remove from the original. The Equatorie, also attributed to Chaucer, is a holograph written in 1392. Both of these texts can be classified as scientific (more closely astronomical) treatises, or using the terminology of Rissanen (2000: 119), as expository texts. Due to this typological similarity, the difference in the extent of the variation to be found in the two texts must be ascribed to other factors. Capgrave’s texts are both holographs, which means that the difference in the extent of the variation to be found in the two texts cannot

Word order in Late Middle English

Table 1. The rate of V2 and V3 in main clauses Texts

Nominal subject Pronominal subject V2 Rate of V2 V3 Rate of V3 V2 Rate of V2 V3 Rate of V3

Treatise on the Astrolabe

38

88.4%

5

11.6%

59

83.1%

12

16.9%

Equatorie of the Planetis

7

100%

0

0%

19

95%

1

5%

Chaucer’s prose works in PPCME-1 (conflated in Haeberli 2000)

85

57%

64

43%

95

50%

95

50%

Chaucer’s literary prose works in PPCME-1 (after deduction)

52

45%

63

55%

59

40%

90

60%

Sermon

4

100%

0

0%

10

90.9%

1

9.1%

Abbreuiacion of Cronicles (Haeberli 2000)

36

40%

54

60%

31

51.7%

29

48.3%

Abbreuiacion of Cronicles (Eitler 1999)

9

47.3%

10

52.7%

3

42.8%

4

57.2%

Wycliffite Old Testament (Haeberli 2000)

6

5.3%

107

94.7%

1

2.1%

46

97.9%

be due to any scribal interference. His Sermon was written in 1452 (though it had been preached two decades earlier), whereas his Chronicle was written in 1464. In Rissanen’s terminology (2000: 119), the Sermon can be classified as a religious instruction, while the Chronicle is a non-imaginative narration.

. Dialect input into London sociolects As mentioned previously, Middle English was characterised by the competition of the word order variants V2 and V3. The innovative variant, V3, was percolating from the northern to the southern dialect areas (Kroch et al. 2001). Originally, northern Middle English was CP-V2 due to previous contact with Scandinavian, whereas southern Middle English was IP-V2. Speakers of the





Tamás Eitler

northern CP-V2 language accommodated to their southern interlocutors, by positing to their language V3 word order, as they perceived that their interlocutors used V3 to a larger extent than themselves (Kroch et al. 2001). This V3 word order was acquired by the subsequent generations and began spreading southwards in a manner akin to some other northern innovations finally finding their way into the standard. The first traces of the innovative V3 grammar can be found in Richard Rolle of Hampole’s works from around 1350 (written in the Yorkshire dialect), and it seems that the (Northern and Southern) Central Midlands dialect was the one that through contact next borrowed this V3. This is important as it has been claimed that the Central Midlands dialect underlay the standard, the formation of which began centring on London in the 14th and 15th centuries. When investigating the linguistic situation of late mediaeval London, one must take into consideration the influence of consecutive waves of immigrants from different dialect areas. It can be claimed that each immigrant group brought with themselves their own respective dialect and continued using it within the city in special interactions. There they formed segregated neighbourhoods, where everyday communication in the same indigenous idiom was still possible (Keene 2000: iii). Since in different periods distinct immigrant groups settled in London from different areas, the areal and hence dialect provenance of these groups can be claimed to have tended to correlate with their socioeconomic status within the city. It appears that those who first moved to the city typically began to constitute (or become overrepresented in) the upper class, whereas the relative latecomers began to form (or become overrepresented in) the middle class. According to Samuels (1963), who relies on Ekvall’s studies on London’s population history (1956), there were three candidate idioms for the standard-defining status in 14th and 15th century London: (1) the indigenous language of London, which was a typical Southern dialect, (2) Chaucer’s dialect, and (3) a later Central Midlands dialect, very close to the language of the Wycliffite texts. The speakers of the indigenous London dialect formed the oldest layer of the population of the city and therefore presumably they could be found in all social strata. As their language was a southern one, it can be assumed that it had IP-V2 word order. An example of this indigenous London dialect is the Proclamation of Henry III from the mid-13th century. The Central Midlands Wycliffite language started to become standardlike in the late 14th century, but it finally failed. This dialect is exemplified by the Wycliffite Old Testament (1425), which, as shown in Table 1, was almost systematically V3: in main

Word order in Late Middle English

clauses the rate of V3 is nearly 100 percent. It can be assumed that the somewhat earlier Central Midlands dialect must have had a similar, though perhaps not so systematic V3 syntax, which was taken to London by more and more immigrants from the second half of the 14th century on (Samuels 1963: 90– 91). The East Anglian immigrants, however, had started to move to London in greater numbers in the late 13th century (Samuels 1963: 88–89). According to Kristensson (2001: 75), they were mostly prosperous merchants who subsequently formed (or at least were overrepresented in) the relatively upper class and were later also to be found in prominent positions in the society of the City of London. As they had arrived earlier than the aforementioned immigrants from the Central Midlands, these latter could typically only form the middle (or the relatively lower than the upper) stratum of the contemporary London society. In Section 6, the word order syntax of the East Anglians will be discussed, which was also the word order syntax of their respective upper class London sociolect.

. Findings in close-up As can be seen in Table 1, in the Astrolabe in main clauses containing a nominal subject the rate of V2 is 88.4 percent, whereas in main clauses with a pronominal subject the rate of V2 is slightly lower, 83.1 percent. This almost systematic pronominal V2, however, can be regarded as a sufficient diagnostic of the presence of a CP-V2 syntax, somewhat distorted by the intrusion of a clearly V3 syntax. The other closely examined Chaucerian text, the Equatorie, exhibits a completely systematic V2 syntax, which is clearly a CP-V2 syntax. What is important to note is that at the very end of the 14th century, besides the centrally located dialects (like the Wycliffite one) having V3 syntax, there existed a dialect that still preserved a CP-V2 syntax almost intact. In Chaucer’s literary prose texts, on the other hand, the rate of V3 is 55 and 60 percent, respectively. This dichotomy needs to be accounted for. In Capgrave’s Sermon, V2 is systematic in main clauses containing a nominal subject, whereas in main clauses with a pronominal subject the rate of V2 is 90.9 percent. These figures indicate the presence of a systematic CP-V2 syntax into which a V3 syntax has slightly intruded. According to Haeberli’s data (2000), in the Chronicle, in main clauses with a nominal subject, the rate of V2 is 40 percent, whereas in main clauses with a pronominal subject the rate of V2 is higher, 51.7 percent. According to the data based on a search on a 100clause extract only (Eitler 1999), provided here for comparison, the Chronicle





Tamás Eitler

exhibits 47.3 percent V2 in main clauses with a nominal subject, and 42.8 percent V2 with a pronominal subject. These figures indicate a replacement very much in progress. The text itself can be located somewhere around the middle phase of the S-curve trajectory in the third quarter of the 15th century, when the more centrally located dialects and even the more northern dialects were already nearing the tailing off phase of the change.

. Word order competition in East Anglia The findings for Capgrave show that in the second part of the 15th century, in sharp contrast with the more central dialects, in Norfolk and possibly in the whole East Anglian region there was still a strong CP-V2 syntax, or at least it can be claimed that the change to V3 was not yet at an advanced stage. As a reason for this relative lateness one might consider the fact that linguistically East Anglia has been a relic area due to its location on the periphery of the East Midlands: to the east there is the North Sea, and to the west it is cut off from its neighbouring areas by the Fenland. This marshy area may have been more extensive in the Middle Ages, when for that matter communication was even more restricted. Thus it can be argued that East Anglia did not have substantial communications with the neighbouring part of the East Midlands areas or with the Central Midlands. As a consequence, the allegedly Central Midlandsbased standard’s V3 order may have been spreading from London (a direction more easily navigable) once it had been adopted there as the emerging standard variant, besides percolating directly from the Central Midlands, where it had been adopted even earlier (see Section 3). Thus, what most probably happened besides dialect levelling was the process of standardisation which ousted the conservative V2. During this process linguistic forms travelled encapsulated in the legal and administrative texts issued in the Chancery and other government departments. These statutory texts were instrumental in the establishment of the innovative forms as elements of the emerging standard (Rissanen 2000: 120). Furthermore, there is an important implication of this relic area behaviour. Kroch and Taylor (1997: 324–325, Note 24) stated that apart from appearing in the northern dialect area, CP-V2 was also present in the North East Midlands. Accepting their reasoning and applying it to the East Anglian (a part of the South East Midlands) dialect area, it can be stated that the presence of the CP-V2 syntax in the late 15th century East Anglian dialect, as shown by the findings in Table 1, must originally be a result of Scandinavian language con-

Word order in Late Middle English

tact, similarly to the result of the contact with Scandinavian in the north. This claim is not surprising since a great number of secondary Scandinavian immigrants settled peacefully in East Anglia in the Late Old English period (Hansen 1984: 55). Besides the attested lexical and even grammatical traces, which suggest an intimate relationship between the speakers of the two languages, this peaceful population influx is noticeable in the toponymy of the coastal and the more inland areas alike.

. The London sociolects and the emerging standard Kristensson’s (2001) pioneering study on the sociolects of late 14th century London revealed the existence of two distinct social layers composed of immigrant groups with two distinct dialect backgrounds: the middle class (predominantly originating from the Central Midlands areas), and the upper class (from the East Anglian areas). As was shown in the third section, the Central Midlands-based immigrants forming (at least a well-defined part of) the middle class had an almost systematic V3 syntax in their sociolect. On the basis of the data drawn from Capgrave’s two holograph texts, representing the East Anglian dialect in the second part of the 15th century, it can be inferred that the East Anglian-based immigrants forming (at least a well-defined part of) the upper class in the 14th (and possibly in the 15th) century must have had a strong if not systematic CP-V2 in their respective sociolect. As the change was from V2 to V3, it can be assumed that the strong CP-V2 syntax of the 15th century must have been the reflex of at least a similarly strong, if not an even stronger, systematic CP-V2 syntax. Until recently it has been accepted that the standard, which was emerging in the 15th century via the practices of the Chancery and some other government departments, had only a single underlying source, i.e. it descended from a single ancestor. The East or Central Midlands dialect area was identified as the possible source of the incipient standard (Samuels 1963). This theory became known as the single ancestor-dialect hypothesis (Hope 2000: 49–50). Recently, however, a new line of research has revealed that there were many dialects with competing grammars from which the incipient standard selected different forms (Hope 2000: 50–51). The multisource or amalgam nature of Standard English can explain why one can find so many Northern, Southern and East Anglian features besides the Central and East Midlands ones. Kristensson (2001: 74–76) claims that among the sources of Chancery English the East Anglian dialect played an important role. Following Ekvall’s





Tamás Eitler

claims on the population history of mediaeval London, Kristensson argues that in the first half of the 14th century there was an influx of East Anglian immigrants, who, despite their small number, exerted great influence on the shaping of the standard. These immigrants, who were typically prosperous merchants, started to constitute the upper class of the London society and import their respective East Anglian forms into their upper class sociolect. Kristensson cites an example of phonological change influencing spelling conventions during which the upper class variant (based on the East Anglian variant) was spreading. It can be seen, however, that as regards the word order syntax in the context of non-operator fronting in main clauses, the innovation came from another sociolect of London. The innovative variant, V3, was typical of the sociolect of the immigrants who hailed from the Central Midlands and came to form the middle class. Apparently, in this case a middle class variant was spreading, which needs to be somehow reconciled with the fact that as regards some phonological changes and spelling conventions, the (relatively more) upper class variant was the successful one. The problem that the multisource nature of the standard is the result not only of a dialect-based, but also of a (dialect-fed) sociolect-based grammar competition, will be addressed in future research.

. Reprofiling Geoffrey Chaucer According to the popular profile, Chaucer was a southern, conservative London speaker. The data from the present research show, however, that in one of his competing grammars, Chaucer had a systematic V2 syntax indicative of the CP-V2 syntax and not of the typically southern IP-V2 one. In the following the question will be posed whether his CP-V2 syntax can be connected somehow to the sociolect of the London upper class, in which CP-V2 was the typical word order pattern (see Section 6). It is often overlooked that Chaucer was a member of the influential mercantile elite, descending from an Ipswich-based East Anglian family of wealthy vintners and brought up in Upper Thames Street, in the Vintry Ward of the City of London. As the V2 syntax found in Capgrave’s texts is indicative of the language used in the second part of the 15th century, it can be assumed that in the late 14th century immigrants from the same East Anglian dialect area must have had an even more systematic CP-V2 (see Section 6). As this systematic CP-V2 syntax can also be found in the texts of Chaucer and since his family background is similar to that of the wealthy

Word order in Late Middle English

East Anglian immigrants, it can be asserted that as far as syntax goes, Chaucer’s idiom belongs to the (relatively) upper class sociolect of the London of his time. The picture, however, becomes more complex if one looks at Haeberli’s findings. The first version of PPCME did not include the Equatorie and thus Haeberli’s (2000) conflated findings on the word order syntax of Chaucer were based on the PPCME-1 version of Boethius, the Astrolabe and two tales from the Canterbury Tales. It can be claimed that Chaucer had syntactic diglossia with two competing parameter settings, a systematic CP-V2 and a V3, and his nonliterary prose works have a higher rate of V2 (Astrolabe), whereas his literary prose texts have a higher rate of V3. However, as shown in Section 9 and in Eitler (2004: 55–56), the indigenous word order parameter setting of Chaucer must have been according to the CP-V2 syntax. It will also be claimed that the use of word order variants was conditioned by text type (where the further factor of influence of translation from different source languages may also have played a role) and/or by accommodation to the target audience.

. Variation and accommodation in John Capgrave As the two works of Capgrave (1) represent two distinct text types, (2) are both holographs, hence free of scribal interference, (3) were both translated from the same source language, they can be analysed as to whether this text type or genre difference may cause any differences in the word order patterns. As can be seen in the findings in Table 1, the usage frequencies are different in the two texts. Among the possible genre-related factors causing this divergent behaviour one can mention two likely candidates. One of these can be the medium’s role, i.e. whether the text is the product of a primarily oral or written process. The second factor can be style: whether the text has a formal or informal style. It could be expected that since sermons are as a rule spoken and hence also may tend to be somewhat more informal (than most written texts), innovative variants can be more numerous in them. As in our case the innovative variant is V3, the rate of V3 could be expected to be higher. Similarly, it could be expected that since chronicles are as a rule written and hence also may tend to be somewhat more formal, conservative variants are more numerous in them. As in our case the conservative variant is the V2 order, the rate of V2 could be expected to be higher. While it can be seen that the rate of V2 is high in the Chronicle, it is higher, even systematic in the Sermon. This means that the data do not meet the aforementioned expectations, and thus one has to consider two other possible





Tamás Eitler

factors underlying the divergent syntactic behaviour of the two texts. Whereas the first of these factors is the possible chronological difference, the second one could be dubbed as the language of the target audience of the text. Since between the composition of the Sermon and the Chronicle there is only a 12-year time gap, which presumably could not have witnessed a drop of 60 percent (with nominal subjects) and 38 percent (with pronominal subjects) in usage frequency, one has to resort to the possible influence of the second factor. Here one clearly deals with the communicative aspect mentioned in Section 2. In a communication situation the interaction between the interlocutors can be more successful if they accommodate to each other linguistically. In the case of persuasive discourse, the need to convince and influence the target audience can be an even stronger driving force resulting in the encoder’s approximation to the listener’s language, which idiom naturally has to be part of the encoder’s competence or linguistic repertoire, too. Reassessing the situation in compliance with the above, it can be argued that as a rule sermons are preached to a local community. In Capgrave’s case it was a religious community in Cambridge that the Sermon was intended for. As the target audience is a local one, one can expect more locally used and accepted forms to occur in the text and so the rate of the V2 variant to be higher. In a chronicle, however, the eminent aim of the encoder is mainly to inform the readers that constitute a wider audience possibly recruited from all over the country. It can be argued that in this case the encoder is not compelled to use any linguistic forms preferred in a specific dialect area. Instead, he can opt for more widespread forms that are understandable in larger areas: besides the emerging standard itself, a candidate type of a language having these forms can be the colourless regional standard. As by the time of the composition of the Chronicle V3 had gained acceptance in the standard and in many dialect areas, it can be estimated that the colourless regional standard possibly used in the text and understandable in inter alia wider East Anglian and Central Midlands areas must have had a lower rate of the conservative variant V2. It is exactly this lower rate that is found in the Chronicle. Further investigation is necessary, though. In summary, it can be claimed that the need of accommodation to the target audience can overwrite or at least influence the effect of the text type. However, more evidence is still needed to confirm this claim. The possible intersecting influence of genre and the regionally (hence dialectally) identifiable target audience can also be detected in Chaucer, which issue is briefly taken up in the next section (for more detail, cf. Eitler 2004: 58).

Word order in Late Middle English

. Geoffrey Chaucer’s audiences As regards Chaucer’s syntactic diglossia, similarly to what can be found in Capgrave, accommodation to different target audiences can account for the distinct word order usage frequencies in the various works (mentioned in Section 7). Chaucer intended the Astrolabe for the use of his son Lowys. If the Equatorie is a natural complement to the Astrolabe, as it is believed to be (counting as a further piece of evidence for Chaucer’s authorship, cf. Price ed. 1955; Rand Schmidt 1993), it can be claimed that it too was written for the use of his son. As intrafamily language use is often associated with the vernacular, one might consider the possibility that Chaucer used his indigenous idiom in the aforementioned two texts (Eitler 2004: 55–56). Given the findings for these texts, it can be argued that his indigenous syntax was of the CP-V2 type. By contrast, the literary prose texts targeted another audience: their function was to entertain the highest stratum in England (whose representatives were mostly to be found at Court) and/or a nationwide audience. It can be claimed that in order to ensure the reception of his works, now Chaucer accommodated to the language spoken by these audiences. Since he seems to have substituted his systematic CP-V2 syntax with a mixture of V3 and CP-V2 syntax, it can be assumed that the syntax of the uppermost social layer (which, however, was not identical with the upper layer of the City of London) and/or that of the nationwide audience (as presumed or perceived by Chaucer) may have been closer to a V3 syntax, the dialect source of which was identified in Section 3. This issue can only be clarified through further research. A philological excursus, however, is in order. Ever since the discovery of MS Peterhouse 75.1, its authorship has been a much debated issue. Interestingly enough, in the debate the syntactic character and the word order pattern in particular were not used as arguments. The present findings show that the Astrolabe has a rate of 83.1–88.4 percent of V2. Besides, one can recall at this point that the manuscript of the Astrolabe is at least at one remove from the original manuscript. As the tendency in the Middle English period was the gradual weakening of V2, the possible scribal interference (appearing in the copying process between the original (around 1391) and the copied manuscript (around 1450)) must have included the replacement of V2 clauses with V3 ones. Thus one can assume that the word order in the original manuscript of the Astrolabe (i.e. in Chaucer’s idiom) must have been a more systematic V2. As the findings show that in the Equatorie the rate of V2 is almost 100 percent in main clauses, and as this manuscript is claimed to be a holograph from 1392, it can be argued that Chaucer cannot be excluded from the group of pos-



 Tamás Eitler

sible authors of MS Peterhouse 75.1. In short, the word order character of the Equatorie lends further support to Chaucer’s authorship of the text. It must be seen, however, that syntactic character in itself is not conclusive evidence for authorship.

. Conclusion Middle English word order variation displays various patterns of complexity. A systematic focusing on the links between dialectal, sociolectal and audiencerelated communicative factors, however, may result in a somewhat better understanding of this complexity. In line with these considerations, the present paper offered tentative solutions to problems surfacing in intrapersonal and intratextual variation, as evidenced in the various works of Geoffrey Chaucer and John Capgrave. The use of the two competing word order variants was claimed to have been influenced by accommodation to the sociolectally and dialectally determined target audience. In addition, it was pointed out that the linguistic impact of the target audience was primarily dialect-fed. Clearly, there is much left to be done, and in order to arrive at anything more certain than the present state of affairs, the future accounts of word order variation and change in Middle English should definitely be based on investigations into the interaction between sociolects, social networks, contemporary audiences and accommodation phenomena.

Note * This paper reports on part of my research conducted at the Research Centre for Variation and Change in English, hosted by the English Department of the University of Helsinki. The funding was provided by MÖB and CIMO (UA1 056). I am grateful to the researchers and research assistants there for making my stay both utterly useful and enjoyable. I am especially obliged to Kirsti Peitsara, Leena Kahlas-Tarkka and Matti Rissanen for their guidance.

References Bell, Allan (2001). “Back in Style: Reworking Audience Design”. In Penelope Eckert & John R. Rickford (Eds.), Style and Sociolinguistic Variation (pp. 139–169). Cambridge: Cambridge University Press.

Word order in Late Middle English 

Coupland, Justine, Nikolas Coupland, & Howard Giles (1991). “Accommodation Theory. Communication, Context and Consequences”. In Justine Coupland, Nikolas Coupland, & Howard Giles (Eds.), Contexts of Accommodation (pp. 1–68). Cambridge: Cambridge University Press; Paris: Éditions de la maison des sciences de l’homme. Eitler, Tamás (1999). “Syntactic Tendencies in Late Old English and Middle English: Constituent Order Restructuring as an Example of Grammar Competition and Dialectally Diffusive Parameter Resetting”. Unpublished manuscript. Budapest: ELTE SEAS. Eitler, Tamás (2004). “Late Middle English Syntactic Variation and Change: Towards an Integrated Model of Accommodation in Social Networks.” In László Varga (Ed.), The EVEN Yearbook 6: ELTE SEAS Working Papers in Linguistics (pp. 49–59). Budapest: ELTE SEAS. Ekvall, Eilert (1956). Studies on the Population of Medieval London (= Kungl. Vitterhets Historie och Antikvitets Akademiens Handlingar, Filologisk filosofiska Serien 2.). Stockholm: Almqvist and Wiksell. Haeberli, Eric (2000). “Observations on the Loss of Verb Second in the History of English”. Unpublished manuscript. University of Reading. Hope, Jonathan (2000). “Rats, Bats, Sparrows and Dogs: Biology, Linguistics and the Nature of Standard English”. Wright, 2000, 49–56. Hansen, Bente Hyldegaard (1984). “The Historical Implications of the Scandinavian Linguistic Element in English: A Theoretical Evaluation”. Nowele, 4, 53–95. Keene, Derek (2000). “Metropolitan Values: Migration, Mobility and Cultural Norms, London 1100–1700”. Wright, 2000, 93–114. Keller, Rudi (1994). On Language Change: The Invisible Hand in Language. Transl. Brigitte Nerlich. London and New York: Routledge. Kemenade, Ans van (1987). Syntactic Case and Morphological Case in the History of English. Dordrecht: ICG Publishing. Kristensson, Gillis (2001). “Sociolects in Fourteenth-Century London”. In Jacek Fisiak & Peter Trudgill (Eds.), East Anglian English (pp. 71–77). Cambridge: D. S. Brewer. Kroch, Anthony (1989). “Reflexes of Grammar in Patterns of Language Change”. Language Variation and Change, 1, 199–244. Kroch, Anthony, & Ann Taylor (1997). “Verb Movement in Old and Middle English: Dialect Variation and Language Contact”. In Ans van Kemenade & Nigel Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 297–325). Cambridge: Cambridge University Press. Kroch, Anthony, & Ann Taylor (2000). The Penn-Helsinki Parsed Corpus of Middle English-2. Philadelphia: Department of Linguistics, University of Pennsylvania. Kroch, Anthony, Ann Taylor, & Donald Ringe (2001). “The Middle English Verb-Second Constraint: A Case Study in Language Contact and Language Change”. In Susan Hering, Pieter van Reenen, & Lene Schøsler (Eds.), Textual Parameters in Older Languages (pp. 353–391). Philadelphia: John Benjamins. Price, Derek (Ed.). (1955). The Equatorie of the Planetis. With a linguistic analysis by R. M Wilson. Cambridge: Cambridge University Press. Rand Schmidt, Kari Anne (1993). The Authorship of The Equatorie of the Planetis. Cambridge: Boydell and Brewer.

 Tamás Eitler

Rissanen, Matti (2000). “Standardisation and the Language of Early Statutes”. Wright, 2000, 117–130. Samuels, M. L. (1963). “Some Applications of Middle English Dialectology”. English Studies, 44, 81–94. Wonnacott, Elizabeth (2000). “Investigating the Trajectory of Language Change”. Unpublished manuscript. University of Edinburgh. Wright, Laura (Ed.). (2000). The Development of Standard English, 1300–1800: Theories, Descriptions, Conflicts. Cambridge: Cambridge University Press.

Using universal principles of phonetic qualitative reduction in grammaticalization to explain the Old Spanish shift from ge to se* Andrés Enrique-Arias University of the Balearic Islands

.

Introduction

One of the unsolved problems of Spanish historical linguistics involves the shift of the third person dative clitic pronoun in combination with a third person accusative pronoun from gelo to selo towards the end of the Middle Ages. Old Spanish gelo had originated as a result of the evolution of the Latin combination ˘ılli-˘ıllu(m) as a single word. It is generally accepted that the initial vowel of the second pronoun was preserved; thus, *(˘ı)lli-˘ıllu would have yielded [λjeλo], but was further reduced to [λjelo], a hypothesis supported by the existence of Italian glielo.1 Finally, the combination of the initial palatal with the glide resulted in a palatal fricative: [λjelo] > [Šelo]. However, the ensuing shift from Old Spanish ge to se, which resulted in the present ambiguity as to reflexivity/non-reflexivity of the Spanish pronoun se when immediately followed by a third person direct object pronoun, has not been explained satisfactorily.

. Confusion of sibilants The traditional explanation for the change ge > se is phonological in nature and involves the different interchanges of sibilants that take place throughout Old Spanish (Menéndez Pidal 1940: 254; Lapesa 1981: 397). These are sporadic

 Andrés Enrique-Arias

changes (which in some cases go as far back as the origins of the documented history of the language) that typically have lexeme specific explanations, such as dissimilation, assimilation, or influence from semantically related lexical items. Alonso (1947) studied these changes in great detail and identified the six most common types: [s]>[ts] (qui sapit > quiçab ‘perhaps’), [s]>[w] (passer > paxaro ‘bird’), [s]>[tw] (*soriceu > chorizo ‘sausage’), [ts]>[tw] (cistella > chistera ‘top hat’), [z]>[Š] (tonsoria > tigera ‘scissors’), [z]>[dz] (*cinisia > ceniza ‘ash’). In Alonso’s analysis there are three common underlying trends: (i) the changes involve apicals becoming palatals; (ii) there are no changes involving voice, that is, voiceless becoming voiced or vice-versa; and (iii) the changes represent a reinforcement of articulation. It is impossible, then, to integrate the replacement of gelo by selo within the general direction of these purely phonetic replacements: this change is a devoicing, that is, it is a reduction rather than a reinforcement of articulation, and it involves a palatal becoming an apical. Therefore, it contradicts the general changes in every respect. Changes in the same direction as gelo > selo ([Š]>[z] as in relisión, colesio, mesor) are relatively rare, none of them has survived, and take place later in time, which makes it hard to relate them to the change into selo. Some studies that advocate a phonological explanation (e.g. Penny 2002: 136) appeal to another process: the elimination of the distinction of voice undergone by the Old Spanish pair /Š/ – /w/. This proposal assumes that the voiced palatal fricative became voiceless and then later was replaced by /s/. But the exact chronology of this change is highly controversial. Traditionally, it has been established that these replacements originate in northern dialects and spread to central and southern Spain, the changes becoming frequent during the second half of the 16th century (Lapesa 1981: 370–371). Since the substitution of selo for gelo began possibly as early as the fourteenth century (cf. Menéndez Pidal 1940: 254), this change could hardly have had any relation with the general devoicing of sibilants that took place much later. However, recent scholarship contends that the loss of voice distinction in the sibilants originates in the south and takes place much earlier than previously thought (cf. Frago 1985), in which case it is possible that the first instances of dative se happened in areas where ge was pronounced voiceless. At any rate, in the absence of any attestation of [welo] (spelled xelo), linking the change gelo > selo to the devoicing of sibilants is, at this point, speculative.

Old Spanish ge > se due to qualitative reduction 

. Analogy with reflexive se Since a purely phonetic account is less than satisfactory, most authors that have supported an explanation based on the interchanges of sibilants have appealed to a second factor: the analogical influence of the reflexive construction (Menéndez Pidal 1940: 254; Lapesa 1981: 397 among others). The basic argument is that ge lo and se lo alternated in similar syntactic contexts: dióse-lo (gave-3sgREF-3sgACC) ‘he gave it to himself ’, dió-ge-lo (gave-3sgDAT3sgACC) ‘he gave it to him/her.’ But again, it is not clear how this fact alone could have contributed to promoting the change in the direction it happened. In Spanish, there are areas within the pronominal system where speakers can infer a reflexive meaning with non-reflexive pronouns. For instance, in utterances like Juan lo quiere todo para él, literally ‘John wants everything for him’ (where él ‘him’ is a non-reflexive third person pronoun), the reflexive interpretation ‘John wants everything for himself ’ is perfectly possible (the same situation is found in 16th century Spanish as illustrated by the examples collected by Keniston 1937: 55). But the opposite is not possible: because the reflexive pronoun sí is more restricted semantically, it cannot have non-reflexive uses replacing él (‘him’) or ella (‘her’). This distribution suggests that on purely semantic grounds, then, if there were a merger of functions one would expect dative expression to replace reflexive, not the other way around.2 Another challenge for the assumption that the merger of gelo and selo was due to analogy is the relative frequency of the constructions involved: the sequence gelo was far more frequent than selo at the time of the change. Table 1 exhibits frequency counts of 3rd person dative and reflexive pronouns in combination with lo, la, los, las in several studies. For this study I have coded all cases of the sequence gelo and selo in 40 texts from the Madison corpus (see Appendix) covering the mid 14th to early 15th centuries. The figures in Klein (1972) are late 15th century and early 16th century texts, and Keniston (1937) studies 16th century texts exclusively. In all cases the combination dative + lo is overwhelmingly more frequent. Table 1. Relative frequency of dative +lo and reflexive +lo in several studies

Madison corpus (14th) Madison corpus (early 15th) Klein (1972) Keniston (1937)

dative + lo

reflexive + lo

598 96% 696 92% 692 92% 118 86%

26 4% 58 8% 59 8% 19 14%

 Andrés Enrique-Arias

Furthermore, Klein (1972: 135–36) tested the effect of analogy in her data and found that with third person verb forms, which is the only context shared by both reflexive and dative pronouns, occurrence of innovative dative se was actually lower than with other verb forms.3

. Substitution by reflexive se Finally, a number of studies have proposed that ge was replaced by the reflexive pronoun se that already existed in the language rather than a phonetically conditioned variant of ge (Klein 1972; Schmidely 1978; Hewson 1981). Klein (1972) argues that reflexive se, thanks to the extension of its reference to include impersonal agents as in the “impersonal passive” construction, had become suitable for use in place of ge. This led to a transitional period in which both ge and dative se are used, not as the result of a mere loss of distinction, but with distributions that are consistent with a semantically based selectivity in the particular choice of pronoun. Eventually, because of the greater semantic complexity of the construction with ge, speakers decided in favor of its abandonment. However, Klein’s account is challenged by the existence of hypercorrect reflexive uses of ge (see examples (5)–(6) discussed below). These examples are evidence that indeed there was a loss of distinction between ge and se, and thus weaken the plausibility of a substitution in terms such as she advocates. In addition, because Klein leaves out early examples of dative se and uses only later texts where the change was already widespread, her study is limited to defining aspects related to the spread of the change. Her study, thus, is not well suited to identifying the cause for the first instances of dative se.4

. Phonological concomitants of grammaticalization Recent work on the origin and evolution of grammatical markers (cf. among others Hopper & Traugott 1993: 145; and Bybee, Perkins, & Pagliuca 1994: 107) has defined two main phonological concomitants of morphologization: erosion and qualitative reduction. Erosion tends to make morphemes less marked by deleting segments, inducing assimilations, reducing stress, eliminating features (such as voiced or nasal) and reducing segments to non-marked coronal position. A direct consequence of erosion on grammatical material is qualitative reduction, that is, the remaining phonological segments in the form tend to

Old Spanish ge > se due to qualitative reduction 

belong to a restricted subset of the whole inventory of phonemes present in the language. These segments usually reflect universally unmarked sounds such as apical consonants [n], [t] and [s]. The reason is that inflections are functional and frequent and should therefore, in principle, be composed of phonemes that are easy to produce. Apicals are preferred in these environments because the apex of the tongue is the most flexible, sensitive and the easiest to control of all the active articulators, and it has access to most passive receptors. Function morphemes are more vulnerable to erosion because typically they are redundant, unstressed, and occupy non-prominent positions and therefore are pronounced at a faster rate. Diachronically speaking, the tendency for favoring unmarked or relatively easier phonemes in inflectional morphology increases over time. Erosion on functional affixes promotes reductions that are idiosyncratic in that they cannot be explained under regular sound changes, borrowing or analogy. Examples of such irregular changes have been studied in ´ detail in woks like Manczak (1980), Shield (1980), and Picard (1997).

. The change ge > se due to qualitative reduction The replacement of ge with se has all the characteristics that define qualitative reduction in forms undergoing grammaticalization. Towards the end of the Middle Ages, the spread of clitic doubling brought about an increase in the frequency of use of redundant object pronouns. As a consequence, the already semantically and phonologically weak unstressed pronouns became increasingly available for affixation to adjacent verb forms eventually becoming verbal object agreement markers (Enrique-Arias 2003).5 The implication for the historical development of morphology is that, as grammatical markers evolve over time and their textual frequency increases, they should exhibit an increasing tendency for favoring unmarked phonemes. The evolution of Spanish object agreement markers exhibits two changes during the 15th century that go in this direction: the replacement of ge by se and the change from unstressed vos > os. Table 2 exhibits the relative textual frequency of the Spanish phonemes that appear as initial segments in the medieval Spanish pronoun system (adapted from Navarro Tomás 1968: 114). It can be observed that these two changes eliminated precisely those phonemes that were least frequent (i.e. more marked). The change from vos to os, which has been unanimously analyzed as a case of erosion (García et al. 1990; Eberenz 2000), is in many ways parallel to the change from ge to se. First, the chronology is rather similar: os appears sporad-

 Andrés Enrique-Arias

Table 2. Phoneme frequency of initial segments in Old Spanish clitic pronouns Phoneme

Frequency

/s/ [se] /l/ [lo], [la], [los], [las], [le], [les] /n/ [nos] /t/ [te] /m/ [me] /β/ [βos] // [e]

7.59% 7.41% 7.41% 2.80% 2.41% 1.69% 0.65%

ically in Old Spanish, always in contexts that are more vulnerable to erosion, such as enclitic, postonic position, and spreads quickly to all environments in the last decades of the 15th century to replace vos completely by mid 16th century. Second, both changes represent reductions that eliminate a marked initial segment: in vos > os through the deletion of the initial segment, and in ge > se via the elimination of the voice feature and reduction to the unmarked coronal position. And, third, in both cases the change cannot be explained as a regular phonetic replacement. The change ge > se, if due to qualitative reduction would have started in postonic, enclitic position, where sounds are more vulnerable to erosion. In fast speech, in order to facilitate production, speakers would have moved forward the point of articulation to an apical position. Such a move would have resulted in [z], which is the apical sound that is closest to the voiced palatal sibilant /Š/, but in Old Spanish /z/ could not occur immediately after a morpheme boundary (that is, dágelo, ‘give it to him’, could not be pronounced *[dázelo]). Therefore, the only possible outcome for an apical pronunciation of this segment was [s], which could occur unrestricted in this environment.

. Textual evidence We turn now to review textual evidence from the time of the change. If dative se emerged as a fast speech variant of ge, we should expect it to develop in prosodic environments that are more vulnerable to erosion, such as enclitic, postonic position, and this is indeed the case. All the early examples reported in the literature are enclitic. Frago (1990) reports four examples in the Actas Capitulares de Morón, written between 1402 and 1426. The same hand that uses ge in over two dozen cases uses se on four occasions for the same kind of structure. The oldest example comes from a 1414 document.

Old Spanish ge > se due to qualitative reduction 

(1) ochoçientos e nouenta e vn maravedis [...] et él dando selos, dieron lo por quito eight-hundred and ninety one maravedis, and he giving-3sgDAT-3plACC, gave him for cleared “Eight-hundred and ninety one maravedís, and by paying them he was cleared of all charges.”

There are three more occurrences; two are from 1419 and one from 1422. All of them are in enclitic position. Likewise, the earliest examples reported in Eberenz (2000: 217) are all enclitic. The M codex of Ayala’s Libro de cetrería, composed in 1386 and copied in the mid 15th century, exhibits three examples of se next to 73 cases of ge. Once again, all three examples are enclitic: (2) [...] cosele el ala e cierasselo sew-him the wing and close-3sgDAT-3sgACC “Sew the wing and close it.” (3) aquella sangre [...] dasela al falcón [...] dasela asy that blood [...] give-3sgDAT-3sgACC to-the falcon [...] give-3sgDAT-3sgACC like-this “Give that blood to the falcon, give it like this.”

In Alfonso de Cartagena’s Discurso (composed in 1434) there is only one case of the structure studied, also in enclitic position, which six of the extant versions render as se: (4) E ellos fueron al rey et otorgoselo graçiosamente And they went to-the king and granted-3sgDAT-3sgACC graciously “And they went to the king, and he granted it to them graciously.”

The numerical preponderance of enclisis in the early examples is particularly significant because at this moment in the history of the language (early 15th century) proclisis is textually more frequent for this structure. In the texts studied from the HSMS corpus we have 64% proclitic pronouns (461/723). A second argument in favor of the fast-speech variant explanation is the presence of hypercorrect forms, which are characteristic of linguistic variants that originate in fast or colloquial speech. When speakers attempt to restore the prestige structure, they may do it in the “wrong” place, due to linguistic insecurity. For instance, there is a passage in the Libro de las Tafurerías (written ca. 1332-34) where some of the versions exhibit hypercorrect ge. The Madison corpus (O’Neill 1999) contains 17 different versions of this text transcribed from various manuscripts copied throughout the 15th century. Three of them

 Andrés Enrique-Arias

are incomplete and do not contain the passage in question. Of the remaining 14 manuscripts, eleven exhibit this passage with reciprocal se coreferential with the noun phrase unos a otros, which is the traditional usage. (5) que se lo-demanden vnos a otros por sus fueros [Ms. D, fol. 7 v] that 3sgREF-3sgACC-sue ones to other by their laws “Let them sue each other according to their laws.”

However, manuscripts A, I and N exhibit ge (cf. Ms. I, fol. 138r que ge lo demanden vnos a otros). In texts from periods in which ge and se kept their distinctive functions the noun phrase unos a otros is always co-referential with se; thus the use of ge in this structure is hypercorrect.6 Another example of hypercorrection can be observed in Corbacho: (6) algunos [...] non dan pena luego que ge la mereçen [CBO fol. 106 v] some [...] not give punishment even that 3sgREF-3sgACC-they-deserve “Some do not punish [others] even if they deserve it.”

A final argument in favor of the fast-speech variant hypothesis is that the earliest cases of dative se, as we have seen, appear in administrative documents (such as the four examples from the Actas de Morón written between 1414 and 1422 recorded in Frago 1990) while at the same time, examples of dative se are extremely rare in literary texts. The reason is that administrative documents, especially in rural areas, are often written, without any intention of elegance, by semi-learned persons for purely utilitarian purposes. As a result, these texts often exhibit features that render information on the vernacular of the scribe. The higher occurrence of dative se in non-literary texts is relevant because in all its effects fast speech works in the same way as casual speech (Zwicky 1972), and thus it can be assumed that the higher frequency of dative se in less formal registers reflects the fact that this form is characteristic of fast speech. Likewise, Eberenz (2000: 219) observes that in his 15th century corpus, examples of dative se are sporadic in literary texts, but very frequent in the records from the Inquisition of Ciudad Real, where occurrences of dative se account for a third of all cases of dative + lo. This means that by searching further in non-literary documents from the late 14th and early 15th centuries we will be able to get a clearer picture of the way this change originated.

Old Spanish ge > se due to qualitative reduction

. Summary and conclusions The increase in textual frequency and loss of phonetic prominence brought about by the progressive grammaticalization of object clitic pronouns to become affixal object agreement markers sets up the typical scenario for erosion and the replacement of marked phonological segments with unmarked sounds. Spanish dative pronoun se then originated as a fast speech variant of ge in connection with a universal tendency towards phonetic qualitative reduction as a result of morphologization. This explanation is supported by three facts: a) the earliest examples of dative se appear in enclitic position, which is an environment more vulnerable to erosion; b) during the period of the spread of dative se there are occurrences of hypercorrect reflexive ge; c) dative se appears first and is more common in less formal registers. Since, as we have seen, the repertoire of phonological segments that can appear in function affixes is somewhat limited, it is fairly common, and indeed only to be expected, that a change in this context results in the creation of a grammatical marker that is homonymous with a preexisting one. It is quite common cross-linguistically that the same suffix appears in more than one system (e.g. English -s marks plural and possessor in nouns, and third person singular present indicative in verbs). In Spanish, after changing from ge into se, the dative pronoun became homonymous not only with reflexive, reciprocal and impersonal se, but also with the past subjunctive suffix (i.e. canta-se). Homonymy with reflexive se, then, can be regarded as a mere accident attributable to the limited inventory of sounds that are typically present in function affixes; appealing to analogy simply because there is a merger of forms does not contribute much to explaining this change. The qualitative reduction explanation defended here has a number of advantages over previous ones. It provides a strong yet simple motivation for the change ge > se that is widely attested cross-linguistically. In contrast, the interchanges of sibilants are a lexeme-specific phenomenon with no unitary functional principle that can be applied to the problem at issue. Also, the qualitative reduction explanation works better than Klein’s (1972) proposal because speakers cannot normally replace morphological material in highly grammaticalized constructions. In order to economize, speakers reduce their utterances phonetically, not morphosyntactically (Lütdke 1985). The explanation defended here is not necessarily in conflict with traditional explanations. Once the first apical pronunciations of ge took place, the spread of the change was most likely encouraged by factors proposed in previous proposals, such as the existence of the parallel reflexive structure and the





Andrés Enrique-Arias

fact that, as Klein suggests, a dative use of se in this structure is semantically and pragmatically viable. The diachronic facts, the textual evidence from the time of the change, and the cross-linguistic data together support a position in which the replacement of ge with se correlates with the selection of universally unmarked phonological segments in forms undergoing morphologization.

Notes * My thanks to Concepción Company, José Ignacio Hualde, Flora Klein-Andreu, Robert Orr, Rémy Videraz, and ICHL participants for suggestions and remarks on this paper. . The change from [λjeλo] to [λjelo] has received different explanations. Menéndez Pidal (1940: 253) advocates an analogical reduction of the second [λ] influenced by other unstressed pronouns that begin with an [l] sound; Alvar and Pottier (1983: 30–31) invoke dissimilation of the palatals, and Frago (1990: 218) suggests that Latin -ll- in the second pronoun never palatalized due to its unstressed nature. . This is not to say that a change where reflexive expression replaces non-reflexive is not possible, but those that advocate an explanation based on analogy with reflexive se have not explained what motivated the merger in the direction it did. . Eberenz (2000: 216, 221) supports the role of analogy in this change by pointing out that the construction with se was ambiguous in certain cases, allowing a dative and a reflexive interpretation. However, the examples that he provides simply show that there is ambiguity after the change; obviously, while ge and se kept their distinctive functions there could not be ambiguity. . Additionally, the accounts that claim the incompatibility of combinations of le (or its variant ge) and lo are challenged by the fact that these combinations survived in Spanish for many centuries, and are now widespread in Romance varieties such as Iberian Portuguese, Leonese, Aragonese, French, Italian and Sardinian (Brakel: 1979). . As far as I know there are no quantitative studies that measure the textual frequency of Spanish clitic pronouns diachronically. However we may infer this development from the studies that show a steady increase in the frequency of clitic doubling throughout the Middle Ages (cf. Rini 1991). . A search of the sequence unos a otros in corpusdelespanol.org yields 67 13th century examples that are coreferential with se, and none with ge, le, or les.

References Alonso, Amado (1947). “Trueques de sibilantes en antiguo español”. Nueva revista de Filología Hispánica, 1 (1), 1–12. Alvar, Manuel, & Bernard Pottier (1983). Morfología histórica del español. Madrid: Gredos.

Old Spanish ge > se due to qualitative reduction

Brakel, Arthur (1979). “The provenience and present status of Spanish selo”. Linguistics, 17, 659–670. Bybee, Joan L., Revere D. Perkins, & William Pagliuca (1994). The Evolution of Grammar. Chicago and London: The University of Chicago Press. Eberenz, Rolf (2000). El español en el otoño de la Edad Media. Madrid: Gredos. Enrique Arias, Andrés (2003). “From clitics to inflections: Diachronic and typological evidence for affixal object agreement marking in Spanish”. In Bernard Fradin et al. (Eds.), Les unites morphologiques. Forum de Morphologie (3e rencontres). Actes du colloque de Vielleneuve d’Ascq (pp. 67–75). Lille: Université. Frago Gracia, Juan (1985). “Valor histórico de las alternancias grafémicas en los fonemas del orden velar”. Revista de Filología Española, 65, 273–304. Frago Gracia, Juan (1990). “El cambio de ge lo a se lo en testimonios andaluces”. Anuario de Lingüística Hispánica, 6, 217–224. García, Érica, Robert de Jonge, Dorine Nieuwenhuijsen, & C. Lechner (1990). “(V)os(otros): ¿dos y el mismo cambio?”. Nueva Revista de Filología Española, 38, 63–132. Hewson, John (1981). “More on Spanish selo”. Linguistics, 19, 439–447. Hopper, Paul J., & Elizabeth C. Traugott (1993). Grammaticalization. Cambridge: University Press. Keniston, Hayward (1937). The Syntax of Castilian Prose. The Sixteenth Century. Chicago: The University of Chicago Press. Klein, Flora (1972). The Role of Meaning in Grammatical Change: The Shift from Old Spanish ge to se. Columbia University Doctoral Dissertation. Lapesa, Rafael (1981). Historia de la lengua española. Madrid: Gredos. Lütdke, Helmut (1985). “Diachronic irreversibility in word-formation and semantics”. In Jacek Fisiak (Ed.), Historical semantics, historical word-formation (pp. 355–366). Berlin: Mouton de Gruyter. Menéndez Pidal, Ramón (1940 [1999 this edition]). Manual de gramática histórica española. Madrid: Espasa-Calpe. Mañczak, Witold (1980). “Irregular Sound Change due to Frequency in German”. In Jacek Fisiak (Ed.), Historical Morphology (pp. 309–319). The Hague: Mouton. Navarro, Tomás (1968). Studies in Spanish Phonology. Coral Gables: University of Miami Press. O’Neill, John (Ed.). (1999). Electronic texts and concordances of the Madison Corpus of early Spanish manuscripts and printings. Madison: Hispanic Seminary of Medieval Studies. Penny, Ralph (2002). A History of the Spanish Language. Cambridge: Cambridge University Press. Picard, Marc (1997). “On the Effects of Frequency-induced Phonological Change”. Diachronica, XIV (1), 109–118. Rini, Joel (1991). “The Redundant Indirect Object Constructions in Spanish: A New Perspective”. Romance Philology, 45, 269–286. Schmidely, Jack (1978). “De ‘ge lo’ a ‘se lo”’. Cahiers de linguistique hispanique médiévale, 4, 63–70. Shield, Kenneth (1980). “Fast Speech and the Origin of the Standard English Verbal Suffix -s”. Journal of English Linguistics, 14, 24–35.





Andrés Enrique-Arias

Zwicky, Arnold (1972). “On Casual Speech”. Papers from the 8th Regional Meeting of the Chicago Linguistic Society (pp. 607–615).

Appendix: Sources of texts used in the study All the texts used in the study are taken from O’Neill (1999) except ALB and MAC, taken from my own transcriptions. For each text the following information is listed bellow: abbreviation, title (reproduced exactly as given in O’Neill 1999), number of occurrences of dative ge + accusative (gelo, gela, gelos, gelas), and number of occurrences of reflexive or reciprocal se + accusative (selo, sela, selos, selas). Next to the texts copied in the 15th century there is a third figure indicating number of occurrences of selo, sela, selos, selas where se has a dative interpretation. There are no examples of this structure in the texts copied in the 14th century. Texts copied in the 14th century: BSM, Bendita Santa Marta, 1, 0; CCM, Cuento del emperador Carlos Maynes, 21, 0; EMZ, De una santa emperatriz que hubo en Roma, 10, 0; EOR, El emperador Otas de Roma, 35, 4; ERG, Estoria del rey Gujllelme, 23, 0; GRF, Gerardus falconarius, 61, 0; HCP, Historia del caballero Plácidas, 9, 0; ICR, Crónica de San Isidoro, 0, 0; IHG, Historia de los godos de San Isidoro, 0, 0; LBG, Libro de buen amor-G, 21, 2; LBT, Libro de buen amor-S, 14, 0; LCB, Libro de los caballos, 195, 3; LEO, Leomarte: Sumas de la historia troyana, 80, 6; MDR, Mocedades de Rodrigo, 16, 1; OA1, Ordenamiento de Alcalá, 47, 2; PAL, Libro de Palladio, 6, 6; SMM, Santa María Madalena, 2, 0; TFI, Libro de las tahurerías, 8, 0; TRE, Libro de las tres creencias, 9, 0; TRO, Tres Reyes d’Orient, 4, 0; VSC, Vida de Santa Catalina, 3, 0; VSM, Vida de Santa María Egipciaca, 6, 0. Total: ge + accusative 571 (96%), se + accusative 24 (4%) Texts copied in the 15th Century: ABC, Exemplario por ABC, 62, 4, 1; ALB, Biblia de Alba (prólogo), 3, 0, 0; ATA, Atalaya de las corónicas, 174, 4, 1; CBO, Arcipreste de Talavera: Corbacho, 50, 18, 1; CIS, Arte cisoria, 12, 1, 0; CRN, Crónica de 1344, 25, 2, 0; DON, Libro de las donas, 31, 5, 0; LBS, Libro de buen amor-S, 27, 2, 0; LC1, Libro del conocimiento de todos los reinos, 0, 0, 0; LCA, Libro de la caça de las aues, 90, 3, 0; LTR, Letras de Hernando del Pulgar, 3, 0, 0; MAC, BNM ms. 1518, 5, 0, 0; OFI, De officiis de Cicerón, 11, 1, 0; PER, Mostrador e enseñador de los turbados (perplejos), 58, 5, 0; TAM, Historia del gran Tamerlán, 60, 0, 0; TFM, Libro de las tahurerías, 10, 1, 0; TRS, Cuento de Tristán de Leonis, 57, 1, 1; VMZ, Los nueve libros de Valerio Máximo, 45, 13, 0 Total: ge + accusative 723 (92%), se + accusative 60 (8%)

The origin of transitive auxiliary verbs in Chukotko-Kamchatkan Michael Fortescue University of Copenhagen

.

Copular/auxiliary polysemy in Chukchi

In Fortescue (2003) I presented the cumulative burden of proof for the unity of the controversial Chukotko-Kamchatkan (CK) family. The present paper represents additional proof of the genetic relationship between the two branches of the family, Chukotian and Itelmen. This additional proof takes as point of departure the extensive use of analytic verbal constructions employing auxiliary verbs made by all Chukotian languages (Chukchi, Koryak, Alutor, & Kerek). The auxiliaries involved are typologically unusual in so far as they include transitive (‘have’) as well as intransitive (‘be’) auxiliaries applicable to the same bases.1 The former are moreover usually ditransitive – ‘have X as Y’ – rather than the more usual transitive ‘have X’ type. Historically, they appear to be of the copular type widely known from northern and central Asia, allowing the attachment of TAM categories to otherwise fixed, uninflected forms of lexical verbs. Unlike in adjacent language families, however, the auxiliary verbal stems do not themselves introduce the TAM meaning – they must be inflected for these. In some constructions they are obligatory (e.g. in certain types of negative sentence), while in others there may be a choice depending on semantico-pragmatic factors (e.g. with verbs of mental/emotional activity or state). The semantic idiosyncracies of these verbal constructions – especially the transitive ones – and their common distributional constraints can be shown to be shared in considerable detail by Itelmen. The findings of grammaticalization theory will be relevant here, since the direction of change can be shown to have been from common – or at least overlapping – lexical sources to parallel copular and then auxiliary functions.



Michael Fortescue

In the Chukotian branch of the family there are six basic stems which are used in auxiliary constructions. Dunn presents them for Chukchi as in Table 1 (Dunn 1999: 311–312); I have simply added glosses for the non-auxiliary (copular) meanings: Table 1. Chukchi auxiliary stems Intransitive: it- (stative, deverbal verb bases) ‘be (identity)’ wa- (-twa-) (stative, deadjectival verb bases) ‘be (locational or existential)’ n‘el- (inchoative) ‘become’ Transitive: r6t- (-nt-) (non-mental predicates) ‘have (as)’ l6]- (-l>-) (non-resultative mental predicates) ‘consider (as)’ r6tc(6)- (-tc-) (resultative mental predicates) ‘treat in such a way’

In the following sentences some examples (from Dunn) are given of their use as lexical verbs/copulas: (1) mej]6-wil-u n-it-qin big-price-equ imperf-be-3sg “That was a lot of money.” (2) tej]et n6-twa-qen food imperf-be-3sg “There was food.” (3) Ta>6n n‘el->‘i brigadir-o Tagyn become-3sg.aor brigade.leader-equ “Tagyn became a brigade-leader.” (4) 6ntuulp6re-nu l6>-nin son.in.law-equ consider.as-3sg/3sg.aor “He had/took him as son-in-law.” (5) erm-u m6-tc6->6t leader-equ 1sg.opt-make.into-2sg “I’ll make you leader.” (6) ]utku t6-re-nt6-]-6n reluur‘6-qej here 1sg.fut-have-fut-3sg chewing.tobacco-little “Here I’ll have some chewing tobacco.”2

Sentences (7) and (8) below from Skorik (1977) illustrate their use as auxiliaries, with respectively intransitive it- and transitive r6t-:3

Auxiliary verbs in Chukotko-Kamchatkan

(7) >6m-nin epeqej >6lo-ta it6-rk6n cinit akka->t6 me-poss grandmother miss-instr it-3sg.pres her.own son-all >6m-6m >6lo-ta t6-n‘el-6k 6tl6>-et6 me-emph miss-instr 1sg-n‘el-aor father-all “My grandmother misses her son, and I have begun to miss my father.” (8) waj6]qen iwinil‘6n amq6n-‘6co em-ret-e that hunter every-day only-bring.home-instr r6t6-tku-l‘6n 6nnan-mem6l-a r6t-freq-part one-seal-instr “That hunter brings home every day just one seal.”

The lexical stem (the main verb) is, as here, typically in the instrumental or some other nominal case form – or takes the uninflected negative circumfix as in (9): (9) >6m-ninet ekke-t i>6r e-tejkew-ke r-in-]6t me-poss.pl son-pl today neg-fight-neg fut-it-3pl “My sons will not fight today.”

As an example of the possible choice between an analytic and an ordinary (synthetic) verbal expression compare 10a (focus on the predicate) and 10b (with focus on the locative frame expression): ]inqe>ti i>6r tip‘ej]e-te n-it-qinet boy.pl today sing-instr imperf-it-3pl “The boys are singing today.” b. ]utin-qel>i-k-6m ]inqe>-ti n6-tip‘ej]e-qinet this-room-loc-emph boy-pl imperf-sing-3pl “The boys are singing in this room.”

(10) a.

. Copulas and auxiliaries in Itelmen There is a very similar distinction between analytical and synthetic verbal constructions also in the Itelmen branch of the family, involving by and large the same set of auxiliary verbs, although conflations have obscured the picture somewhat and the transitive ‘have’ type auxiliary corresponding to r6t- of Chukchi does not appear at first sight to have any independent lexical source there at all. It would therefore be quite a coincidence if the two were not closely related – unless of course Itelmen somehow borrowed the entire system (involving both lexical and auxiliary functions) from neighbouring Chukotian





Michael Fortescue

languages, which is what Volodin’s convergence theory actually suggests (cf. Georg & Volodin 1999: 224ff.). I shall return to this point later. Of the six Chukchi auxiliaries treated in Section 1 at least four and possibly five have cognates in western Itelmen (the sole surviving member of this branch of the family), most of them used in similar analytic constructional contexts, although some of them are only found in one or two fossilized phrases today. They are in general much less widely used than in Chukchi despite Itelmen being more analytic in other respects. The only ones with any degree of productivity are eS(e)- (cf. Chukchi l6]-), tel(e)- (similar in sense to Chukchi r6tc6-), and le-, although Volodin does not consider the latter as an auxiliary as such. I in turn do not regard his ite- ‘force, order to’ as an auxiliary today since it occurs as an auxiliary in his listing only with negative particle qa‘m in the lexicalized sense ‘forbid’. The Itelmen auxiliaries are presented in Table 2 (in their infinitive shape) along with typical – or sole – collocations (from Volodin 1976): Table 2. Itelmen auxiliaries Intransitive: it-: 6]qan it- ‘squirm, act in inexplicable way’4 le-: tmalk le- ‘approach’ (cf. tmalk ‘nearby’) Transitive: ent(e)-: 6]qan ent- ‘harm’ eS(e)- (stem iS-): anzan eS(e)- ‘accuse’; xiq eS(e)- ‘know’, SmaSu eS(e)- ‘believe’ tel(e)- (stem til-): xiq tel(e)- ‘recognize’

Analytic constructions are required in negative constructions in much the same way in Itelmen as in Chukchi, only the particles involved and the distribution is somewhat different, owing to the differently organized TAM systems: an auxiliary construction is required in Itelmen in the present and past but not in the future (in the latter case a synthetic verb in the imperative/optative is required, just like in Chukchi). A single example will suffice: (11) qa‘m cawa-kaq t-tile-qzu-kicen not meet-adv 1sg-til-cont-3sg.aor “I did not meet him.”

Most of these auxiliary stems have lexical (copular) uses, as in: massu-k’a le‘become a bear’, isxe-k’i eS(e)- ‘recognize as (one’s) father’, matuxa-k’a t-til(l)axen ‘I shall turn (you) into a female bear’ (with tel(e)- and the ‘transformative’ case ending on the noun), and nisilan s-in ‘he is a hunter’ (with existential/locational it- absorbed by present tense -s-, and the absolutive form of

Auxiliary verbs in Chukotko-Kamchatkan 

the noun). The instances where there is some choice of use between analytic and synthetic constructions with the same verbal (or other) base are more restricted than in Chukchi – Volodin mentions (1976: 217) just a few cases where le- ‘become’ can be used in both, apparently with no difference of meaning (e.g. txlo] le- alongside txlote- ‘get old (of woman)’). This – plus the predictable and productive meaning of this stem with bases – is the principal reason he prefers not to see le- as an auxiliary like the others. However, it would not be surprising if the distinction here is (or was once) similar to that between analytic and corresponding synthetic constructions in Chukchi – i.e. one of foregrounding/potential contrast. The adverbial forms of bases in all the analytic constructions illustrated above are quite varied (as in the Chukotian branch) and include case-marked ones. Instances with eS(e)- such as SmaSu eS(e)- ‘believe’ above (going with Chukchi l6mal-o l6]- ‘believe’) are particularly interesting since they appear to contain the original Chukotko-Kamchatkan ‘equative’ (or ‘attributive’) case marker -(n)u widely used in the Chukotian branch in these constructions but otherwise replaced by -k’i/-k’a in Itelmen (the so-called ‘transformative’ case of similar function, deriving historically from an old allative as in Chukchi 6nk6ri ‘to there’). Although one might argue, as Volodin doubtless would, that these represent loans from Chukotian, the only other instance he gives, namely Senwej-u eSe- ‘laugh at’, is more difficult to relate to its Chukotian equivalent and unlikely to be a loan (cf. Chukchi ten]-u l6]- of this meaning).

. Proto-Chukotko-Kamchatkan reconstructions I turn now to the hypothetical proto-language out of which both branches of the family developed. Four (and perhaps five) of the six copular stems discussed above can be reconstructed for this PCK stage (Proto-ChuktokoKamchatkan) while the others can only with certainty be reconstructed for PC (Proto-Chukotian, i.e. also attested in Koryak), as in Table 3: Table 3. PC and PCK copula/auxiliary stems Intrans. PCK it- ‘be (or do or happen)’ PCK t6va- ‘be or live’ (or ‘sit’, as in I Sa-) PC n6ræl- ‘become’

Trans. PCK ð6nt6- ‘have (as)’ PCK l6]- ‘consider (as)’ PC ð6cc6- ‘make or turn into’

 Michael Fortescue

Itelmen cognate Sa- under PCK t6va- is not used as an auxiliary at all.5 More specific collocational correspondences between Chukchi and Itelmen can be adduced, such as Chukchi l6>i l6]- ‘know’ (with auxiliary *l6]- and adverbial base l6>i ‘known’) and Itelmen xiq eS(e)- ‘know’; also l6>i r6tc6- ‘recognize, get to know’ (with *ð6cc6- and the same base) has its parallel in Itelmen xiq tel(e)- ’recognize, understand’ above (with productive transitive auxiliary tel(e)-). In fact even the adverbial stem is cognate here, since xi- is from *l6>i (an additional adverbial ending -q is added to it in Itelmen). There are problems concerning the sound correspondences between certain of the cognates behind the entries above – not surprising given the great frequency and phonological brevity of these items – but by and large they are regular.6 Auxiliaries, it should be borne in mind, are also found in somewhat different distribution from corresponding full verbs (e.g. more often in tensed than infinite forms) and may easily undergo syncope and generalization of allomorphs. All that needs be said about *l6]- is that it may be related to PCK *ð6l]6- ‘count’ (and ð6l>6 ‘finger’), and Itelmen eS- (which has also largely taken over the function of *ð6nt6-) can be compared with Sne- ‘count, read’ from the latter. In Kerek the direct reflex of *ð6l]6- is illi- ‘count’, alternating with l6]- in the auxiliary construction. *n6ræl-, which has no obvious cognate in Itelmen, requires further comment, since the Itelmen functional equivalent, le- ’become’ (also used as an auxiliary verb) does in fact have a Chukotian cognate, namely Alutor li- ‘become, turn to’, which may be the intransitive equivalent of PC ð6(n)li- ‘turn inside out’, but at all events is used in auxiliary constructions such as alva] li- ‘become different, change’ (alva] is an adverbial form ‘different(ly)’). It would appear then that li- ‘become’ is common Chukotko-Kamchatkan, while n6ræl- (at least as an auxiliary) is an innovation in Chukotian that replaced li-, except in Alutor, which has both. The relationship of Itelmen tel(e)- to *ð6cc6- is also uncertain – it may in fact be unrelated.7 According to Volodin it contains a causative prefix t- plus le- ‘become’ above (Volodin 1976: 156). It could represent a newer formation in Itelmen (although Kerek illa- in alwa] illa- ‘remove’, with alwa] ‘elsewhere’, appears to be cognate), imminent conflation with it- and eS- (all with /S/ ∼ /s/ ∼ /Ø/ before a following consonant) having prompted the development. It is evidently the same as the full transitive ‘-s’ conjugation verb tel(e)- ‘cause to become’, although as an auxiliary it belongs to the intransitive ‘-kas’ conjugation. So it could have replaced the original transitive auxiliary *ð6cc6-, which in turn could come from a full verb meaning ‘make or let’ – note the use of r6tc6- in Chukchi with suffix -j>ut ‘in order to’ on the preceding subordinate verb: -j>ut r6tc6- ‘order to, make’. However, it could also come from causativizer

Auxiliary verbs in Chukotko-Kamchatkan

*ð6n- plus li- ‘become’, like Itelmen tel(e)- but with the /l/ ∼ /c/ alternation typical of Chukotian (but not Itelmen), i.e. ultimately be cognate. This possibility is bolstered by the fact that the causativizing prefix t- in Itelmen is not at all productive, thus suggesting an old form. The original system of auxiliaries in the common ancestral language seems, at all events, to have been built up upon three intransitive copular verbs, respectively a ‘be (equational), happen’, a ‘be (locational, existential)’, and a ‘become’ item, plus three corresponding transitive verbs meaning respectively ‘consider (to be)’, ‘have (as)’, and ‘cause to become’. It is clear that much of the productivity of the original system has been lost or reorganized, especially in Itelmen, with various conflations confusing the picture, but the basic dimensions and correlations are so evident that one can hardly be in doubt that it really does represent a common heritage. I would claim that the system as such could not have been borrowed lock stock and barrel from Chukotian, since the parallel concerns not just the auxiliaries themselves but also their more basic lexical uses, from which – according to all that is known about grammaticalization chains in this area – the auxiliary uses must have arisen. So Itelmen would have had to ‘borrow’ the same items two times – first as lexical items, then later as auxiliaries.

. The missing ‘have’ equivalent in Itelmen What I shall focus on now is the most egregious difference between the auxiliary systems of contemporary Chukotian languages and Itelmen, namely the fact that one of the three intransitive auxiliaries of the former, t6va-, does not match any auxiliary in Itelmen. On the other hand, Itelmen S- from *it- appears to have merged the equational and locational senses of PCK it- and t6va- respectively (*/t/ regularly goes to /S/ in Itelmen in both of these), a development also occurring with neighbouring Alutor. This will be seen to tie in with an innovative aspect of Itelmen verbal inflection that in the past has been hard (if not impossible) to correlate with anything within Chukotian, namely its defective ‘possessive’ and (related) ‘indirect object’ paradigms, in which indication of a pronominal possessor or indirect object is incorporated into the verbal inflection (semantically rather like ‘ethic dative’ constructions in German and Russian). Itelmen moreover displays a basic ‘have’ verb that seems to be totally lacking in the Chukotian branch. The key here lies in the early conflations that must have taken place in Itelmen between the two items above, namely *t6va- and *it-, both centrally involved in the expression of possession in the



 Michael Fortescue

CK family, and between these and *ð6nð6- ‘hold’ (originally perhaps ‘grasp’). This resulted in a new lexical stem ci- (infinitive ce-kas) ‘have’ (not itself an auxiliary) which displays its own complex suppletive paradigm.8 In Table 4 is given the present tense form of this verb (from Volodin 1976:257) – the construction includes the partially redundant locative case pronoun possessor as given (the -z-/-s- is the present tense affix): Table 4. Itelmen possessive paradigm Subject of possession 1s k6mmank 2s knank 3s 6nnank 1pl mizwank 2pl tizwank 3pl tXi‘ink

Object of possession sing. ci-z-wen ci-z-wen ci-s-kinen ci-z-wen ci-s-kisxen ci-s-kip6nen

plural ci-z-we‘n ci-z-we‘n ci-s-kine‘n ci-z-we‘n ci-s-kisxe‘n ci-s-kip6ne‘n

A comparison with the basic intransitive and transitive aorist paradigms shows that there is a difference between on the one hand the 3rd person (singular and plural) and 2pl subject forms – which have recognizable transitive endings – and on the other hand the rest of the forms, where the endings -wen and -we‘n look like the corresponding intransitive 3rd person subject forms.9 This latter construction is termed ‘impersonal’ by Bogoras, corresponding to the ‘there is to me/you’ type existential possession construction common to many languages (including Russian). The other, ‘personal’ forms align on the other hand with transitive lexical ‘X have Y’ type constructions, as in English, with the transitive object marked on the verb. In this case the locative pronouns (‘to him/them’, etc.) are quite redundant and must have spread to the whole paradigm from the ‘impersonal’ items. Everything points then in the direction of the conflation of two different constructions. But what evidence points specifically towards *t6va- and *ð6nð6- as the historical ingredients? First let us consider function: *t6va- is, as we have seen, intransitive as a lexical verb, whereas *ð6nð6- ‘hold’ is transitive (or rather ditransitive). Heine (1997: 231) points out that the reason English ‘have’ happens to be transitive is that its historical source lies in a ‘seizing’ verb corresponding to an underlying cognitive ‘Action’ schema, and in a similar manner one can argue that the intransitivelike inflections of the ci- paradigm reflect an intransitive existential verb like t6va-, while the transitive-like ones reflect a transitive ‘have’ verb like ð6nð6-.10 The fact that *t6va- already has a reasonably regular cognate in Itelmen, namely Sa- ‘sit’ (as mentioned above) need not be a problem – other items among the

Auxiliary verbs in Chukotko-Kamchatkan 

six original ones have somewhat different reflexes in respectively full lexical and auxiliary usage. It must be admitted as regards the form that there are idiosyncracies here that do not add up to strict ‘regularity’; nevertheless there are indices and partial parallels that point towards the result of the hypothetical conflation precisely in a verb of the form ci-. The initial consonant seems to reflect that of *ð6nð6-, with /c/ from */ð/ (or */t/) being an attested – but sporadic – development in Itelmen when there is another alveolar later in the stem, as in kcec ‘forehead’ (< *k6ððel) and c’sal ‘fox’ (< *ðaðjola). Confusingly, however, certain alternative forms of the stem, c>e-/cwe-, are cited in both newer and older ´ sources. Thus corresponding to modern ci- Radlinski had an earlier ‘impersonal’ form c>izin and Bogoras chižin. Dürr et al. have modern 1/2 > 3 present form cvezin but 3s cizvin (with ‘v’ for my ‘w’). The />/ in such forms seems to represent the result of metathesis from *ciz>in (->in/>en being the usual 3s marker and -z- the present tense affix). Note that Itelmen S- ‘be’ from *it- also varies in an irregular way with /s/ (thus the /t/ does not go to expected /s/ before the -kas infinitive ending, but it does – unexpectedly! – in 3s form s-in ‘he is’), so one should not expect too much regularity with high frequency items like this – especially when conflation is involved. In general, there is much syncope and irregular metathesis in Itelmen compared to the phonologically more conservative Chukotian languages (see Fortescue 2003). The other ingredient in this conflation (perhaps newer) that has complicated the picture is precisely it- ‘be’. This has played a plausible role in the ‘disappearance’ of t6va- as an auxiliary in Itelmen. The facts here are reasonably clear (and as such have been adequately described before). In the past and future tenses ci- can be replaced in the paradigm given above by S-, the Itelmen reflex of lexical copular *it-. Thus 6nnan-k S-qzu-kinen ‘he had it’ (Bogoras 1922: 767, also has auxiliary le- ‘become’ in this possessive construction, which he glossed as ‘become as a – for’). That this is not possible in the present tense could well reflect the original semantic trait distinguishing t6va- (one of the sources of ci-) from it-, before the two may have fallen together in auxiliary function, namely that of describing a more enduring state, which is clearly more suitable to the present than to the past or future in this construction (although both lexical copulas can be used independently in any tense). In fact Dürr et al. have ci- also in the meaning ‘occur, be located’ (like t6va-). By contrast, it- seems much closer to a general proverb (‘do, happen’ as well as equative ‘be’). The net result, then, was the appearance of a ‘new’ verbal possessive construction of the type seen in the following:

 Michael Fortescue

(12) 6nnan-k ci-s-kinen tXt6m he.loc have-pres-3sg/3sg boat “He has a boat.”

As the person inflection shows, the construction was originally transitive for 3rd person subject, but has now been brought in line with the intransitive non-3rd person forms – hence the -kas conjugation marker -ki- (typical of intransitive verbs). Observe the remarkable parallel with the Chukchi equivalent: (13) 6nan r6t-6rk6nin ‘6tt‘6t he.ERG have-pres.3sg/3sg boat “He has (it as) a boat.”

All the morphemes here are probably directly cognate (even the -kas conjungation marker -ki- following the present tense affix -s- may be from reanalysis of such forms). Moreover, the opposition of oblique (Itelmen locative, Chukchi ergative – corresponding to the locative case on most nouns) and absolutive case is the same. This is remarkable since Chukchi is an ergative but Itelmen a nominative-accusative language; in fact it presents a further clue as to how ergativity arose in the Chukotian branch (perhaps under Eskimo influence – see Fortescue 1997). Later this new, composite inflectional type was extended in Itelmen to other verbs (initially of an existential, locational type), then on to form the completely general ‘indirect object’ paradigm. This may have been facilitated by the behaviour of ditransitive stem *j6l- ‘give’ (present in all CK languages), the inflectional object of which may refer to either the direct or indirect object. In fact many ‘ambitransitive’ verbs (and transitive ones with incorporated objects) can have a transitive applicative usage (a kind of ‘Ø – applicative’ derivation) in the Chukotian languages, thus Chukchi uwi- can be both ‘cook (s.th.)’ and ‘cook for (s.o.)’ – the latter is the equivalent of the Itelmen indirect object paradigm. Both t6va- and it- have been further grammaticalized in the Chukotian languages in a derivational/synthetic direction, respectively as a resultative affix and as an affixal copula (in nominal predicates and participial verbal mood paradigms with certain non-3rd person subjects). This has no parallel in generally less synthetic Itelmen.

. ‘Have’ vs. ‘be’ constructions in Chukotko-Kamchatkan Chukotko-Kamchatkan languages in general show a remarkable variety of possessive constructions – in fact most of Heine’s types (reflecting his Location,

Auxiliary verbs in Chukotko-Kamchatkan 

Action, Companion, Goal and Genitive schemas) are represented in all of these languages, side by side, with some unexplored semantic differences (Heine 1997: 45ff.). There is also a derivational possession construction with affix -l‘6n ‘one having a-’, of the sort Heine notes in passing (1997: 71). Heine does not, however, consider ‘mixed’ constructions separately, such as the ci- one in Itelmen, which corresponds partly to his ‘Location schema’ (‘X exists for me’) and partly – at least historically – to his ‘Action schema’ (e.g. ‘X holds Y’) and seems to have been motivated largely by phonological syncretisms. He does discuss ‘hybrid’ constructions (1997: 111), but these are all cases of historical shift from one schema to another and not of the blending of preexisting constructions corresponding to two different schemas. One cannot simply reconstruct the evolution of a ‘have’ construction from a ‘be’ one here (in the manner of Hagège 1993: 63f.), since both the ‘have’ and the ‘be’ constructions concerned can be assigned to the proto-language. The impersonal ‘there is’ sense of ci- could well be later – following the well-substantiated grammaticalization scale Action > Location > Possession > ‘Nuclear’ Existence (Heine 1997: 207), as is the case with *t6va- itself within Chukotian (assuming that the postural ‘sit’ sense is indeed original). Nevertheless, the ‘impersonal’ form of ci-, ciz(w)en, is clearly intransitive in origin (the 3s aorist form reconstructed for the proto-language is *->6ræn, regularly > I ->en/-wen), different from the ‘personal’ forms which directly reflect the transitive Action schema. There remain to be said a few words about the semantics of Chukotian *ð6nt6- ‘have as’ as opposed to Itelmen ci- ‘have’. Recall the typical Chukchi ditransitive usage of r6t-, as in the following entry in Bogoras’ 1937 dictionary: (14) r6jul‘-u t6-nt6-rk6ni->6t watchman-equ 1sg-have.as-2sg.pres “I have you as watchmen (over my herd).”

This can be alternatively glossed in English as ‘you are (a) watchman for me’, and it is indeed a similar copular construction (with it-/S-) that forms the nearest equivalent in Itelmen to this kind of sentence rather than ci-, as in (15) from Volodin (1976: 261).11 (15) isx t-s-kinen 6nnan-ke father 1sg-be.pres-3sg.poss he-dat “I am a father to him.”

The question then arises as to whether the rather unusual ditransitive possession (‘have as’) meaning of *ð6nt6- in the Chukotian branch of the family might not be secondary, due to areal – or specifically Eskimo – influence. There is a

 Michael Fortescue

ditransitive possession morpheme that is extremely common in Eskimo languages, but it is manifested there by a bound affix (-k6- ‘have as, consider as’) rather than by an analytic construction (thus CAY aana-k-aa ‘he has her as his mother/she is his mother’). More significant than this alone is the fact that this morpheme is also used with mental/emotional verb stems used transitively, as in kum-k-aa ‘he feels affectionate towards’, parallel with Chukchi te]-u l6]- ‘like’ with auxiliary l6]-/-l>- (which has - perhaps coincidentally – a remarkable formal parallel in Eskimo affix -l6> ‘provided with’, -l>u- ‘have’). In both languages the construction is obligatory when transitive (there are synthetic equivalents – with specific affixes – in both languages in the corresponding intransitive construction). This obligatory construction, it will be recalled, is common to both branches of Chukotko-Kamchatkan, so Itelmen may simply have lost the original ditransitive use of *ð6nt6- (replaced by ci-).12 We may in fact rather be seeing the parallel result of similar grammaticalization processes in the Eskimo-Aleut and Chukotko-Kamchatkan families. It is typologically likely, at least, that these bound affixes in Eskimo derive prehistorically from auxiliaries and – before that – lexical stems, as we still see in Chukotko-Kamchatkan. We lack, however, coverage of such phenomenon across the world’s languages to say much more on the matter. All we can state at the present is that this is indeed an areal phenomenon, one in which also neighbouring accusative-nominative Yukagir may be involved, in so far as it displays both a ‘have as’ verb like Chukchi (Kolyma o¯ži- a causative form of o¯- ‘be’) and ‘use as’ and ‘provide with’ ditransitive derivational affixes like Eskimo (so the phenomenon can not be attributed to the ergative morphology of Eskimo). It appears at all events that the typical Location schema possessive construction in CK is newer than the Action schema construction involving *ð6nt6since the former is limited to physical possession whereas the latter has developed a more abstract possessive sense (further to the right on Heine’s diachronic scale), parallel with the case of the two possessive constructions in Slavic languages (Heine 1997: 232). Note, incidentally, that the principal Eskimo affix of possession *-]qar- also probably reflects an Action schema origin since it appears to contain an affix meaning ‘get, obtain’ (see Fortescue et al. 1994: 419). There is reason to believe, then, that possession constructions based on the Action schema are very old in the northern Siberian/Arctic regions.

Auxiliary verbs in Chukotko-Kamchatkan 

. Summary To sum up, the diachronic picture of the development of auxiliary verb constructions in the CK family as a whole, both of the ‘be’ and the less common ‘have as’ type, is one of idiosyncratic – though in part parallel – shifts and conflations within an interrelated network that presupposes a common starting point. This cannot simply be explained by wholesale borrowing from Chukotian into Itelmen since the morphemes involved display polysemies reflecting both earlier and later points on the same grammaticalization pathways (in respectively lexical/copular and auxiliary functions). The situation is similar to that found in the areas of nominal inflection and of (synthetic) verbal TAM paradigms, where the various nominal declensions and verbal paradigms of the contemporary languages can be seen to have common ingredient sub-patterns, but where they first developed – in somewhat divergent ways – into fully integrated systems in the separate branches of the family, so that a point at which the entire system could have been borrowed one way or the other never existed (cf. Fortescue 2003 as regards nominal declensions and Fortescue 1997 as regards transitive verbal paradigms). Those developments that cognate items underwent along plausible grammaticalization pathways must be assigned to the common proto-language itself, in so far as these items were, at all reconstructable stages, already linked into the same broader system of contrasts. In the case of the auxiliary constructions this paper is concerned with, the system takes the form of a semantically unified set of auxiliary constructions built on two times three full lexical verbs. These had already developed into copulas in their own right at an earlier stage within the ancestral language, although traces of the entire process remain in the polysemies and formal residues left in the contemporary languages.

Notes . The only languages with comparable transitive auxiliaries that I am aware of are Korean (Martin 1992: 227) and possibly Haida (but see Fortescue 1998: 88, Note 6, on the uncertain status of the ‘directive causative’ morphemes in that language, which may rather be affixes already). These Chukotko-Kamchatkan auxiliaries are transitivizing and thus different from the transitive auxiliary-like verbs of languages such as Hindi which are attached to already transitive verbs in a benefactive/applicative sense (thus de-na ‘give’ in swna de-na ‘to relate to/for’, compounded with swna- ‘relate’). . This sentence (from Dunn 1999: 322) may have an implicit ‘as (in capacity of)’ argument, but from Dunn’s explanation of the context (one of stopping at certain points along a jour-

 Michael Fortescue

ney to rest and take some tobacco) it seems best to gloss it here as simple transitive ‘take, use’. Compare the more usual ditransitive usage in ekke-nu t6-nt6->‘en Rinto (son-equ 1sgr6t-3sg.aor) “I have Rinto as son” (Bogoras 1922: 765). Observe the use of the aorist here (for present relevance of a resultative state), suggestive of an earlier active ‘acquire’ sense (similarly, l6]- in the aorist can have an ‘acquire as -’ sense). Skorik (1977: 243) denies that r6t- ever has an independent lexical use, but this is surely an exaggeration. . The auxiliary use of wa- (-twa- non-initially) is somewhat more restricted, but it is common enough with negative adjectives. Compare also mi]q6ri wa-l‘6n (how (t)wa- part) “like what, what kind” (asking about a long-term quality). . The first word = ‘somehow, something’. Bogoras (1922: 768) also has S(a)me-l m-(i)S-in “I will kill him” (apparently with ‘6th infinitive’/instr. case suffix S- on the verbal stem), parallel to Chukchi ‘restrictive’ equivalent am-t6m-a m6-nt6->‘en (only-kill-instr 1sg.optr6t-3sg)). The Itelmen ‘6th infinitive’ usually functions as object of modal verbs of inception and inability, but Georg and Volodin (1999: 184) give one example of its use in an analytic construction with an auxiliary, namely t6l- (= tel(e)-): qec’eS q-t6l-x tx]ate-S (enough optt6l-2sg/3sg hide-INF6) ”Do not hide it!”. Note that Volodin 1976 has it- ’be, happen’ in analytic expressions, though its usual shape is S-. . But it does enter into 6n-Sa-S(e)- ‘put, set’ from *ð6nt6vat- (Chukchi r6twat-) of that meaning, a causative derivation of t6va-. Note also the parallel meaning in Chukchi wak‘o‘sit down’, wak‘otwa- (with resultative affix -twa-) ‘be seated’ < *vak(6r)o-, which may be related (the affix certainly is). Actually, the Itelmen positional meaning, ‘sit’, may reflect the earlier lexical source of the locational/existential copula, a known grammaticalization route (cf. Heine & Kuteva 2002: 278). . Thus */t/ usually goes to /s/ before a stop or to /S/ before a sonorant, */]/ is often dropped, and */w / goes to /x/ or voiceless /f/ in Itelmen. */r/ goes to glottal stop /‘/ in Chukchi but is usually dropped (at least initially) in Itelmen; */v/ goes to /w/ in Chukchi and is usually dropped in Itelmen; */ð/ goes to /r/ in Chukchi (Koryak and Kerek /j/, Alutor /t/) but goes to /z/ (falling together with the result of */j/) in Itelmen when it is not dropped (especially initially). /c/ alternates with /l/ in Chukotian. Productive ‘root retraction’ vowel harmony is by and large retained in Chukchi, but lost in Itelmen (with a residue), where there is also considerable syncope and hence rich consonant clusters. . Bogoras (1922: 768) distinguishes it from iss6- ’make (into)’ as in 1s/2s t-iss6-hin, apparently going with Chukchi r6tc6-, but this may rather reflect eS(e)- under l6]- plus present tense marker -s-. . Compare the most usual way of expressing physical possession in Chukchi, namely with (t)wa- and the locative or possessive form of the possessor: >6m-6k wa-rk6t ]ireq k(6)mi]6t (I-loc be-3pl.pres two children) or, with possessive ‘mine’: >6m-nin wa-rk6t ]ireq k(6)mi]6t (I-poss be-3pl.pres two children) “I have two children/sons”. There is also a synthetic construction with verbalization of the comitative case as in >6m >e-kmi]-i>6m (I COMchild-1sg) “I have children” (emphasizing the subject’s state of possession rather than its precise object?). . Thus compare 3s intransitive aorist ->en/-wen, 3pl ->e‘n/-we‘n (the forms with /w/ – actually a labio-velar – are newer), and, on the other hand, such transitive aorist forms as 3s/3s

Auxiliary verbs in Chukotko-Kamchatkan 

-nen, 3s/3pl -ne‘n, and 2pl/3s -sx, 2pl/3pl -sxi‘n. The ‘intransitive’ inflections are precisely the ones that take a preceding -ki-, which characterizes -kas conjugation verbs (basically intransitive, but extended to certain transitive stems, including – significantly – ci- itself ). All of these -kas conjugation endings display irregular contractions/metathesis. There has been considerable readjustment and simplification of the verbal paradigms in general. . The PC stem *ð6nð6- ‘hold’ may well be related to the transitive auxiliary *ð6nt6- above, namely if it is the source of causativizing prefix *ð6n- and ð6nt6- further contains it- ‘be’ as Volodin has suggested. This would explain the idiosyncratic morphophonemic behaviour of Chukchi auxiliary r6t- (or its cognates) in all Chukotian, which appears as -nt6- medially, an idiosyncratic development found only with this lexeme and those containing the (productive) causative prefix r6-/-n – (Itelmen 6n-) from *ð6n- – the medial correlate of Chukchi r6nr6- ‘hold’ is -nr6-, also losing the first consonant. That a verb of this meaning should also have resulted in a causative prefix is not unlikely – compare English ‘have s.o. do s.th.’. See also Heine and Kuteva (2002: 286) for the known source of many causativizers in lexical verbs of ‘taking’, of which this may be a further instance, i.e. ‘take’ > causativizer. The causative affix in CK languages is actually a circumfix, by the way, the earliest (valencychanging) part of it probably lying in the associated suffix (*-æv- or *-æt-). . Note that the 1s/3s suffix -kinen of the indirect object paradigm is not found in the conjugation of ci- given above (where the ‘impersonal’ possessive construction rather is used with 1st person subject). Bogoras (1922: 766) has tckinin (for tskinen) p’e-k’i ‘I am (as) a son to him’, with translative/equative case -k’i on the complement rather than the absolutive. . The transitive ‘possessive/indirect object’ paradigm of Itelmen makes up for this since it- ‘be’ and le- ‘become’ can also be used with these transitive inflections in the senses ‘be for’ and ‘become for’, as illustrated with it- (S-) above.

References Bogoras, Waldemar (1922). “Chukchee”. In Franz Boas (Ed.), Handbook of American Indian Languages, Part 2 (pp. 631–903). Washington: Bureau of American Ethnology (Smithsonian Institute, Bulletin 40). Bogoras, Waldemar (1937). Luoravetlansko-russkij slovar’. Moskva/Leningrad: Gosudarstvennoje uˇcebno-pedagogiˇceskoje izdatel’stvo. Dunn, Michael (1999). A Grammar of Chukchi. Canberra: Australian National University (Ph.D. thesis). Dürr, Michael, Erich Kasten & Klavdiya Khaloimova (2001). Itelmen Language and Culture (CD-ROM). München: Waxmann. Fortescue, Michael (1997). “Eskimo Influence on the Formation of the Chukotkan Ergative Clause”. Studies in Language, 21 (2), 369–409. Fortescue, Michael (1998). Language Relations across Bering Strait: Reappraising the Archaeological and Linguistic Evidence. London: Cassell Academic. Fortescue, Michael (2003). “Diachronic typology and the genealogical unity of ChukotkoKamchatkan”. Linguistic Typology, 7, 51–88.

 Michael Fortescue

Fortescue, Michael, Lawrence Kaplan, & Steven Jacobson (1994). Comparative Eskimo Dictionary with Aleut Cognates. Fairbanks: Alaska Native Language Center Press. Georg, Stefan, & Alexander P. Volodin (1999). Die itelmenische Sprache. Wiesbaden: Harrassowitz. Hagège, Claude (1993). The Language Builder. Amsterdam/Philadelphia: John Benjamins. Heine, Bernd (1997). Possession. Cambridge: Cambridge University Press. Heine, Bernd, & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Martin, Samuel (1992). A Reference Grammar of Korean. Rutland, Vermont, and Tokyo: Charles Tuttle. ´ Radlinski, I. (1891–1892). SSowniki narzeczny ludów Kamczackich (compiler B. Dybowski). Kraków: Rozprawy Akademii Umiej˛etno´sci, wydzial – filologiczny, 16–18. Skorik, Petr J (1977). Grammatika ˇcukotskogo jazyka, T. 2. Moskva/Leningrad: Izdatel’stvo Akademii Nauk SSSR. Volodin, Alexander P (1976). Itel’menskij jazyk. Leningrad: Nauka.

Grammaticalisation and Latin Michèle Fruyt University of Paris IV

The term grammaticalisation was created by Antoine Meillet in two articles published in 1912 and 1916 and which appeared again in his better-known work of 1921, Linguistique historique et linguistique générale. A. Meillet was the great French comparatist of the time. He very often used Latin in his examples, and all today’s French Latin scholars are necessarily, in one way or another, his pupils. This link explains my choice of subject: I would like to show the very important and special role of Latin in illustrating the mechanisms and results of grammaticalisation. This is due in large part to the language’s longevity – over eight hundred years between Plautus at the end of the third century B.C. and Gregory of Tours in the sixth century A.D., not to mention the documents that preceded and followed those two writers. What is more, we can study the origins of Latin through comparative linguistics, and we can follow developments that have their origin in Latin through into the Romance languages and even into modern Creole languages. Indeed, each of the Romance languages would deserve to be called “Latin” as there is no break in the diachronic continuity. This extremely long “lifespan” of Latin allows us vast scope for diachronic research, providing us with examples of repetition, resurgence, cyclical and spiral development, and successive re-grammaticalisations at different periods. A good example of these is to be found in the French causative periphrasis “faire + infinitive” (je fais construire une maison), which has its origins in late Latin “facere + infinitive”. This Late Latin causative periphrasis is simply reproducing a structure (with a verbal noun, ancestor of the infinitive, plus facio) which already existed in pre-literary Latin and is illustrated by the Latin causative verbs in . . .e-faci¯o, like cale-faci¯o “to warm something”, tepe-faci¯o “to make lukewarm”, etc. Since this structure is also present in Sanskrit, it is

 Michèle Fruyt

probably itself simply a Latin resurgence of an Indo-European structure (Fruyt 1998: 2, 2001). Grammaticalisation has been studied in a large number of languages, but for many of these we do not have sufficient historical evidence, so that we cannot rely on a solid comparative linguistic base. Yet grammaticalisation is a diachronic subject par excellence. The longer the chronology at our disposal and the more the attested examples we have, the better the chance we have of reaching a coherent understanding of the different phenomena. In this context, Latin is of such importance in its many manifestations of the different aspects of grammaticalisation, that we can reasonably say that there can be no general, and much less a universal theory on the subject that does not take it into account. It was our research into word formation in Latin that led us to study grammaticalisation, and in particular our research into word formation by the process of agglutination. Agglutination is extremely common in Latin, and it often co-exists with and facilitates grammaticalisation. Similarly, our studies on motivation, de-motivation, re-motivation, and re-interpretation or re-analysis also led naturally to grammaticalisation. It was thus through its concrete manifestations, through words like licet – which we will return to later – that we became aware of the wide variety of examples and the continual creativity of Latin in this area. But to talk of grammaticalisation presupposes that we have already defined the meaning of grammatical. In this, we will follow Touratier (1994) in distinguishing between two types of morphemes: lexical morphemes and grammatical morphemes – where the term morpheme can be defined in the traditional way in general linguistics as a “minimal unit of meaning” (in French, unité minimale de signification) by opposition to French unité morphologique, which does not denote a morpheme, but only a sequence perceived as a morphological unit, which is not a significant unit. Lexical morphemes include all nouns, adjectives and verbs, as well as some adverbs. In all these cases the morpheme is effectively a lexeme in its own right. But a lexical morpheme can also be a bound morpheme, for example a synchronic root (= French radical latin) such as Latin p¯o- (“to drink”), in p¯o- t¯a-re “drink” and p¯o-culum “recipient used for drinking”, or can- (“to sing”) in can-ere. Grammatical morphemes may be of two types. Firstly, lexemes such as: personal pronouns, endophoric and deictic pronouns, relators such as prepositions, conjunctions, the so-called “free syntactic negations”. Secondly, they may also be bound morphemes, such as the negative morpheme ne- in ne-f¯as and ne-sci¯o, and the n- of n-¯ullus. Certain of these so-called grammatical mor-

Grammaticalisation and Latin

phemes also contain lexical elements – they also have a “meaning” beyond their simple grammatical function. Prepositions and subordinating conjunctions, for example, are relators, but they also have specific semantic features of their own. Their “signifié” (to use Saussure’s term) is a combination of a semantic value and a relating function. There is thus effectively a continuum between lexical morphemes on the one hand and grammatical morphemes on the other (Fruyt 1998-1). Grammaticalisation makes a given morpheme pass from being “lexical” to being “grammatical”. Or it can increase the degree of “grammaticality” of an existing grammatical morpheme and make something already grammatical even more grammatical (Lehmann 1985; Fruyt 1998-1). This is the case when the collective suffixe *-eh2- is downgraded into the Latin inflectional ending -a (Lat. templ-a, mari-a) of the neuter plural. Another example of the latter process is shown by licet, a modal verb that becomes a subordinating conjunction – as we will see later –, as well as by “support-verbs” (to use the terminology of Gaston Gross – in French: verbes supports) such as facere, habere, which become true auxiliary verbs. Grammaticalisation can also go further and reduce a morpheme to a simple phonetic sequence which has no feature of a “signifié” (Fruyt 1998-1). A well-known example of grammaticalisation is given by the Latin mens, mentis (“the mind”), frozen in the instrumental-ablative with an accompanying feminine adjective in . . . a¯ plus mente. This process gave rise to a category of adverbs of manner in Italian, French, etc. such as Fr. lente-ment, juste-ment, adroite-ment, where -ment has become a suffix. It is now a bound morpheme whose function has become morpho-syntactic, creating an adverb from an adjective, and whose only semantic content is that of the original adjective. The morpheme -ment itself has lost all semantic features, and remains just a morpho-syntactic function. The transformation of -ment was slow and progressive. It started in Latin, but was only completed in the Romance languages, where -ment appears in its totally de-lexicalised form. In attested Latin, however, mente continues to be accompanied by an adjective compatible with a state of mind, a psychological disposition, or a trait of character, although we find an increasing number of ambiguous cases. Before looking at some areas of particular interest in more detail, it must not be forgotten that grammaticalisation occurs in Latin in very many instances and in very many linguistic areas. Although we cannot pretend to be exhaustive, the major areas where it is an important diachronic force can be mentioned as: negation, prepositions from frozen nouns or from frozen participles, subordinating conjunctions, verbal periphrasis and the development



 Michèle Fruyt

of auxiliaries, formation of adverbs of all kinds, quantifiers, the second term of a nominal compound that is de-motivated, lexemes expressing modality, constitution of the verbal paradigm, suppletism in the nominal paradigm, suffixes or particles becoming inflectional endings, incorporation of particles into pronouns and change of position for the inflexional ending, etc. This list could go on, and each point could be the theme of a detailed study, as, for example, subordinating conjunctions were the theme of Meillet ’s seminar in 1915–1916. We shall develop here just one area from the above list: negation. The Latin negation, which was inherited from Indo-European *ne with a short e, corresponding to the Sanskrit na, has undergone several processes of grammaticalisation. Phonetically weak, this monosyllable with a short vowel was strengthened by a lexeme which originally had a meaning of its own, the numeral “one”: *ne oinom “not (even) one”, to produce the usual negation n¯on. Not only has the meaning of *oinom (“one”) disappeared, but even the word-boundary, which had become a morpheme boundary, has disappeared. All that remains of *oinom are the two phonemes: long o and n, which do not themselves constitute a morpheme, the morpheme being the lexeme n¯on. The sequence /¯on/ is not even perceived by the speaker as a morphological element. The word has become totally de-motivated, and the speaker is quite unaware of its origin. We have witnessed here the total reduction of *oinom, since in the process it has lost all its features of a “signifié” and it is only represented as a “signifiant” by two phonemes. These have been incorporated into the morpheme *ne without any change of “signifié”; *ne now only incorporates a variant of its original form. So grammaticalisation here only affects the second element – *oinom – which has been de-lexicalised. It has lost its lexical meaning without acquiring a grammatical meaning or function. It has undergone a downgrading in the linguistic hierarchy, since it has gone from being a lexeme and a morpheme (an independent minimal significant unit) to a simple phonetic sequence, which is neither a free lexeme nor a morpheme and not even a mophological unit. But *ne gives us further examples of grammaticalisation, in nihil (“nothing”) for example. This is formed from *ne h¯ılum with the original meaning “not even a minimal unit”, using the logical means – already seen with n¯on – of denying the existence of even the smallest unit possible. H¯ılum denotes, in fact, the tiny part in a bean where the sprout comes out. There was a resurgence of this phenomenon in Old French with the negation ne . . . pas, ne. . . mie, ne. . . goutte.1 Another parallel phenomenon occurs with English nought > not. The lexical value of h¯ılum has disappeared completely in nihil (leaving just a general sense of “inanimate entity”), as the word-boundary and then the

Grammaticalisation and Latin

morpheme-boundary have disappeared. Nihil functions as a single morpheme, historically un-analysable by the speaker, and which cannot even be divided by him into smaller morphological units. In n-¯ullus (“none”) – from *ne oinolos and by opposition to u¯ llus – the word-boundary has been transformed into a morpheme-boundary. The initial n- remains a negative morpheme. There has been no loss of “signifié”, but there has been a downgrading on the scale of linguistic units, since the initial n- is now a bound morpheme. From a synchronic point of view, it has become a prefix, and is no longer a lexeme. In the cases of ne-f¯as (“it is not allowed to”), by opposition to f¯as, and of ne-sci¯o (“I don’t know”) by opposition to sci¯o, the process of grammaticalisation has been limited to a change of linguistic status for *ne-, which has become a prefix and has lost its status as a lexeme. However, its status as a morpheme and the nature of that morpheme remain unchanged. But negation in Latin has still more to show us. The second Latin negation, *n¯e with a long e (used with the volitive subjunctive), shows us another type of grammaticalisation in its use with verbs of fearing of the type: timeo ne veniat (“I fear that he will come”), as opposed to timeo ne non veniat (“I fear that he will not come”). Since the n¯e that introduces the subordinate clause is no longer a negative morpheme here, in the second example it is necessary to add a n¯on in order to make the subordinate clause negative. In this construction, n¯e has undergone a complex process of grammaticalisation which requires us to start from a sentence with two main clauses: timeo and ne veniat! (“I am afraid” and “May he not come!”). The verb of fearing is thus cataphoric – it announces the following clause. The second clause expresses a negative wish in a main clause using a subjunctive of wishing accompanied by n¯e. A re-analysis and a re-segmentation of this word sequence has led to the creation of a complex sentence comprising a main clause followed by a dependent clause. The intonation and syntactic structure change, and the negative adverb of the original becomes a subordinating conjunction, loosing its negative function in the process. N¯e is no longer a negative morpheme, since the speaker no longer has a choice – he or she is obliged to use n¯e in order to introduce any subordinate clause following time¯o. In those constructions involving verbs of fearing, n¯e has thus become more grammaticalised by desemantisation and by functional change. It has changed from an adverb in a main clause to a subordinating conjunction introducing a dependent clause, from a morpheme denoting a certain kind of negation to a morpheme denoting subordination in a context which is lexically conditioned.2



 Michèle Fruyt

Conclusions Finally, we can return to our initial point about the “unavoidability” of Latin in building any general historical linguistic theory. For example, some crosslinguistic studies looking for predictable and possibly universal paths of grammaticalisation in modality have suggested that there is a general evolution from deontic to epistemic modality by loss of semantic features. In the Latin lexicon, d¯ebe¯o is the only verb that could possibly exemplify this type of evolution, but even it is not a good example, since its epistemic and alethic values are only rarely documented. The semantic developments found in the other Latin modal lexemes are different from the one suggested in the cross-linguistic studies (Bybee, etc. 1994). Possum and potest evolve from ability towards possibility in the epistemic modality. Licet is linked to possibility in epistemic modality then to permission in deontic modality. And all Latin modal verbs are polysemic and stay so in the development of the Latin language. Their evolution is not a linear one. We have not listed in our introduction every case of grammaticalisation documented in Latin and, anyway, it would probably be impossible: they are too numerous and some of them remain enigmatic, with obscure origins. Nevertheless, their number, attestations and longevity make it clear that a typological study of grammaticalisation that ignored Latin would at best deprive itself of a very rich diachronic data source, and, at worst, could produce ill-founded conclusions. Indeed, among all the world’s known languages, Latin is quite exceptional in its role as a source for both the diachronic and synchronic studies. We have traces of it over a remarkably long period of time. The earliest Latin inscriptions that we know are dated from the 7th century B.C. and the first Latin texts from the end of the 3rd century B.C., not to mention texts written earlier, but transmitted by later authors (mainly religious and legal or political formulas, prayers and treaties). At the other end of the diachronic scale, it is usually believed that Latin ceased to exist when the Romance languages appeared. E. Löfstedt (1959: 1) remarked that there is no sudden transition in language and it is difficult “to determine when Late Latin begins and ends”: whether we are to make Late Latin start with such and such author is “a question of terminology rather than of substance”. We can say the same thing for the end of Late Latin and the beginning of the Romance languages. It is only a problem of terminology,3 as the diachronic evolution of languages as well as the synchronic diatopic variations are continuous, i.e. effectively displayed along a continuum.4

Grammaticalisation and Latin 

The various current denominations of the Romance languages as Italian, Spanish, French, etc. are essentially related to the political units established in the associated linguistic areas. The socio-political situation did indeed subsequently influence these languages, as the political boundaries often became linguistic ones. But from a linguistic point of view, from the beginning of Latin (whenever such a “beginning” was) to the contemporary state of a particular Romance language, we observe the development of one and the same language through many synchronies. From this point of view, there are as many “Latins” today as there are Romance languages. This gives us an incomparable and an immensely rich field for linguistic studies, particularly for those on grammaticalisation. This method is precisely that adopted by L. Schoesler5 when, breaking through the classifications artificially created by the denominations of languages, she does not treat Latin and French as two separate languages, but as different synchronies inside the same language.

Notes . But as the French negation is becoming pas nowadays in colloquial speech, this pas is being effectively upgraded again into a full morpheme instead of a part of a morpheme. . For this interpretation of a paratactic origin for the construction of the verbs of fearing in Latin and the previous bibliography, see Fleck 2002. This interpretation is limited to verbs of fearing and does not apply to the other uses of subordinating n¯e. It is particularly useful for the analysis of timeo ut ueniat (which is the equivalent of timeo ne non ueniat ) as timeo “I am afraid” and ut ueniat “May he come!”. . Answering the question “When does Spanish start?”, R. Wright (Forthcoming) suggested the date when Latin was introduced into the peninsula by the Romans. He thus incorporates Latin into what he calls “Spanish”. One could, conversely, incorporate Spanish into what we call “Latin”, as Spanish represents one of the direct evolutions of Latin. . As has been shown by F. de Saussure. . Schøsler (2001) studies the expression of the category of definiteness (and the rise of the definite article) from Latin to Modern French, showing the evolutions along the successive synchronies of the same language (Classical Latin, Late Latin, Old French, Early Modern French, Modern French).

 Michèle Fruyt

References Bybee J. L., Perkins R., & Pagliuca W. (1994). The Evolution of Grammar. Tense, Aspect and Modality in the Languages of the World. Chicago-London: The University of Chicago Press. Ernout, Alfred, & Meillet, Antoine (1967). Dictionnaire étymologique de la langue latine, 4th ed. Paris: Klincksieck. Fleck, Frédérique (2002). La négation en latin archaïque: étude syntaxique. Le subordonnant quin dans l’œuvre de Plaute. DEA (Diplôme d’études approfondies) in Latin linguistics, presented at the University of Paris IV-Sorbonne, directed by M. Fruyt. Fruyt, Michèle (1990). “La formation des mots par agglutination en latin”. Bulletin de la Société de Linguistique de Paris, 85, 173–209. Fruyt, Michèle (1998-1). “La grammaticalisation en latin”. In B. Garcia-Hernandez (Ed.), Estudios de Lingüistica Latina, Actas del IX Coloquio Internacional de Lingüistica Latina (pp. 877–890). Madrid, Ediciones clasicas (9th International Colloquium of Latin Linguistics, Madrid, Autonoma, April 14–18th 1997). Fruyt, Michèle (1998-2). “Le renouvellement dans l’évolution linguistique: quelques faits latins”. In B. Bureau & Ch. Nicolas (Eds.), Moussyllanea, Mélanges de linguistique et de littérature anciennes offerts à Claude Moussy (pp. 77–87). Peeters: Louvain-Paris. Fruyt, Michèle (2001). “Réflexions sur la notion de ’mot’ en latin: les verbes du type calefacio”. In C. Moussy (Ed.), De lingua latina novae quaestiones (Proceedings of the 10th International Colloquium on Latin Linguistics, Paris-Sèvres, April 19th–23rd 1999) (pp. 81–94). Peeters: Paris-Louvain. Fruyt, Michèle (1997–1998). “Les frontières du lexique”. Voces, Vol. 8–9 (published in 2000) (pp. 265–292). (Ediciones Universidad, Salamanca) (= Proceedings of the colloquium “Encuentro Internacional sobre Lexico Latino” Salamanca, Universidad, November 18th–20th 1999). Fruyt, Michèle (forthcoming-1). “La corrélation en latin: définition et description”. La corrélation dans les langues anciennes”, Université Michel de Montaigne, Bordeaux III, September 26th–27th 2002. Presses universitaires de l’Université de Saint-Etienne. Fruyt, Michèle (forthcoming-2). “La corrélation en latin: son rôle dans la subordination et l’endophore”. In C.Bodelot (Ed.), Anaphore, cataphore et corrélation dans la phrase complexe du latin. Université Blaise Pascal, Clermont-Ferrand II, January 7th 2003. Presses universitaires de l’Université Blaise Pascal, Clermont-Ferrand. Fruyt, Michèle & Orlandini, Anna (forthcoming). “L’expression de la modalité et les verbes perfecto-présents en latin”. In André Rousseau (Ed.), La modalité dans tous ses états, thematic volume: Revue belge de philologie et d’histoire (Section: Langues et littératures modernes). Hagège, Claude (1993). The Language Builder, An Essay on the human Signature in Linguistic Morphogenesis. Amsterdam: John Benjamins. Lehmann, Christian (1985). “Grammaticalization: Synchronic Variation and Diachronic Change”. Lingua e stile, 20, 303–318. Lehmann, Christian (1991). “Grammaticalization an Related Changes in Contemporary German”. In E. Traugott & B. Heine (Eds.), Approaches to Grammaticalisation (pp. 493– 535). Amsterdam: John Benjamins.

Grammaticalisation and Latin 

Leumann-Hofmann-Szantyr, Lateinische Grammatik I, Leumann, Manu, Lateinische Lautund Formen-Lehre (1977); II, Hofmann, J. B., Lateinische Syntax und Stilistik (1963). Munich: Beck. Löfstedt, Einar (1959). Late Latin. Oslo: H. Aschehoug & Co., Instituttet for Sammenlignende kulturforskning. Meillet, Antoine (1912). “L’évolution des formes grammaticales”. Rivista di scienza, Vol. XII, no. XXVI, 6 (=Lingu. hist. et lingu. gén.) 1921. 130–148. Meillet, Antoine (1916). “Le renouvellement des conjonctions”. Annuaire de l’Ecole Pratique des Hautes Etudes 1915–1916, section historique et philologique, Paris (= Lingu. hist. et lingu. gén.) 1921. 159–174. Meillet, Antoine (1921). Linguistique historique et linguistique générale, Paris: Champion. Schøsler, Lene (2001). “Reanalysing Structure. The Modern French definite Article, its Predecessors and Development”. In Acta Linguistica Hafniensia, 33, 91–108. Copenhagen (The Linguistic Circle of Copenhagen). Touratier, Christian, Syntaxe latine. Louvain-Paris: Peeters. Traugott, E., & Heine, B. (1991). “Introduction”. In E. Traugott & B. Heine (Eds.), Approaches to Grammaticalisation (pp. 1–14). Amsterdam: Benjamins. Wright, Roger (forthcoming). “La representacion de las vocales en las Actas del Concilio de Cordoba de 839”. In Carman Arias-Abellan (Ed.), Proceedings of the 7th International Colloquium on Vulgar and Late Latin (VII Coloquio Internacional de Latin Vulgar y Tardio), Sevilla, September 2nd–6th 2003.

Paths of semantic extension From cause to beneficiary and purpose Silvia Luraghi University of Pavia

.

Introduction1

The present paper is devoted to the relation among three semantic roles – cause, beneficiary, and purpose – in Greek. By analyzing in particular the semantic development of the Ancient Greek preposition diá (Modern Greek ja) I hope to shed some light on possible patterns of syncretism among the three roles. In the last part of the paper I compare the Greek data with data from Classical and Vulgar Latin, which show that the same syncretism may develop from a different starting point. A relevant factor in allowing syncretims of the three roles is the type of metaphor that connects the domain of space (source domain) with the domain of causation (target domain). Semantic features of the NPs involved in the expressions of the three roles play a crucial role in establishing or ruling out possible syncretism: in particular, comparison of Greek and Latin shows that syncretism of cause and beneficiary only holds in cases where animacy can disambiguate the semantic roles.

. The causal approach . Antecedent and subsequent roles Based on the assumption that causal structure defines events, Croft (1991) put forward a model of event structure that can be defined as the ‘causal approach’. The prototypical event type, denoted by a prototypical transitive verb, involves a three part causal chain, with an initiator (the agent), and an endpoint (the

 Silvia Luraghi

patient). Between initiator and endpoint, the event involves a transmission of force. In other words, the semantic roles agent and patient carry with them a causal perspective on event structure, as do a number of other semantic roles. To make his approach clearer, Croft gives definitions of the semantic roles in terms of configurations in causal chains, and classifies them into antecedent and subsequent roles, according to whether they precede or follow the object in the transmission of force in the causal chain. According to this classification, roles such as instrument and cause are antecedent, whereas roles such as beneficiary (or benefactive) and purpose are subsequent. In support of this distinction, Croft points out that, based on a sample of genetically unrelated languages, instances of case syncretisms involving roles within either the antecedent or the subsequent group are frequent, while syncretisms across the two groups are virtually inexistent. Among the few exceptions found, Croft mentions syncretism of cause and purpose, as instantiated, for example, by English for. Indeed syncretism of cause and purpose is widespread, at least in the Indo-European languages, and it is often accompanied by syncretism of these two roles with beneficiary (as, again, in the case of English for). In order to explain this exception, Croft refers to another semantic role, reason, as constituting a possible link between purpose and cause. Much in the same way as purpose, reason involves intentionality, but, similar to cause, it can be conceived as preceding the transmission of force, being the motivation for an agent to act. As for the direction of semantic extension, Croft (1991: 293) writes that possibly “subsequent forms can spread to the cause by means of expressions of reason. ... The hypothesis is that normal subsequent expressions spread to purpose, then to reason (which is nondirectional), and thence to true cause”. In Luraghi (2001) I have shown that this is not necessarily the only possible direction: the extension from cause to purpose is also attested. As I will argue below, the fact that reason is non-directional creates an area of possible overlap for antecedent and subsequent roles, which, as we will see, also includes beneficiary. Note that Croft does not include reason in the causal approach, because, as he correctly remarks, reason “...is a category of intention, not of causation” (1991: 293). As remarked above, a further feature of reason is that it lacks directionality. Croft does not inquire further into the possible group of semantic roles that share this latter feature. To my mind, however, this is worth doing, especially in the case of spatial semantic roles, as I will argue in the next section.

From cause to beneficiary and purpose 

. Space and causation Assuming that the domain of space serves as a source domain for the domain of causation (target domain, see Croft 1991: 194 fll.), when we map the former onto the latter markers of spatial semantic roles as source or origin (ablative markers) will provide the source for the expression of antecedent semantic roles, while markers of direction (allative markers) will serve as source for the expression of subsequent roles. This is what Croft calls ‘the object-location metaphor’. The object-location metaphor is based on a complex schema; various parts of it have been described by other authors as distinct metaphors. In particular, the extension of allative forms to purpose (subsequent) is described as a metaphor by which “purposes are destinations”, and the extension of ablative forms to cause (antecedent) is captured by the metaphor “causes are origins”, according to Lakoff and Johnson (1980). The position of location, and the possible extension of locative forms, in Croft’s model is not as clear. Croft treats location as a subsequent role; when he finds instances of syncretism of location and antecedent roles, he holds that “location is mapped onto the verbal segment as a whole (and thus all the roles included in it) instead of onto just the object” (1991: 196). This sounds rather like an ad hoc solution: perhaps the tendency of locative markers to provide a source for markers of both antecedent and subsequent roles should better be regarded in the fact that location does not belong into a movement chain, as do source and direction. In other words, location lacks directionality, much in the same way as reason does on the abstract plane. This is an important feature of location, which should not be overlooked, because it can explain why certain spatial expressions provide a source for both purpose and cause, while others do not. . Purpose and beneficiary According to the data in Croft (1991), syncretism of purpose and beneficiary is frequent across languages. Since both purpose and beneficiary are subsequent roles, this type of syncretism is in accordance with the predictions of the causal approach. In the Indo-European languages, the meaning of many prepositions seems to switch from purpose to beneficiary simply based on animacy, as shown in (1) a. I did this for fun b. I did this for you.

 Silvia Luraghi

As I will argue further in §5, beneficiary seems more complicated than purpose, because it involves the occurrence of another human participant beside the agent (see also Luraghi, forthcoming b): however, the two roles may be in complementary distribution. Apparently, human beings are not normally purposes for the action of other human beings: rather, it seems that the benefit of another human being is a possible purpose. This fact has the important consequence that there is no need to distinguish between purpose and beneficiary with animate NPs, simply because purposes are not normally animate. As we will see below, however, things are not so straightforward for cause: since causes can be both inanimate and animate, apparently, markers of cause can extend to beneficiary only if animate causes are kept formally distinct. In other words, cause with animate NPs cannot normally be expressed by the same form that also denotes beneficiary: so syncretism of the roles is only partial. Some authors do not distinguish beneficiary from recipient. Indeed typical markers of recipient are often also markers of beneficiary: in the IndoEuropean languages that have case systems, the dative case can express either role. The contrary, however, is not often the case: following the semantic developments of spatial prepositions, we can see that those that extend from space to beneficiary normally do not extend to recipient (see Luraghi 2003: 40–42). If we analyze the matter a little more closely, we see that in the Indo-European languages in which beneficiary and recipient are expressed by the same marker, this marker also denotes, or used to denote, direction: in other words, beneficiary and recipient normally merge where they are both consistently treated as subsequent roles. However, as I have argued in Luraghi (forthcoming b), a beneficiary can be conceptualized as a reason for an agent to act, thus opening a path to syncretism with antecedent roles. This path, apparently, is not easily taken by the recipient.

. Types of cause, purpose, and beneficiary expressions in Ancient Greek For the purposes of the present investigation, it is important to remark that all means of encoding purpose in Classical Greek can be understood as encoding beneficiary, based on the feature of animacy. The contrary is not the case: the preposition hupér, ‘over’, with the genitive,2 can denote beneficiary, but not purpose. Note that hupér most often denotes the behalf type of beneficiary; its occurrences with inanimate nouns is virtually limited to concrete spatial

From cause to beneficiary and purpose 

expressions. Because it does not display syncretisms of the roles discussed here, I will leave this preposition out of account. As we will see below, the possible occurrence of animate nouns with a specific preposition is crucial in determining the pattern of syncretism: apparently, all three semantic roles under discussion can co-occur, but only when cause is limited to inanimate referents. Syncretism of beneficiary and cause with animate nouns does not seem possible based on the evidence in the following sections. . Purpose and beneficiary as direction In Classical Greek, both purpose and beneficiary can be expressed by the preposition eis, ‘to’, ‘toward’. When used in its original spatial meaning, this preposition denotes direction; occasionally it can also extend to addressee and recipient (see Luraghi, 2003 §3.3). Further syncretism of purpose, beneficiary, and recipient is attested by the dative case. Although it usually does not denote direction, the dative is clearly connected with subsequent roles in all Indo-European languages, including Greek. . Purpose and beneficiary as location Various prepositions whose primary spatial function is to express some sort of static location can denote purpose and beneficiary in Ancient Greek. These prepositions do not denote directionality on the spatial plane, so when mapped onto the abstract plane of causation the metaphor that connects the source to the target cannot be the same as in the case of allative markers, and the nature of purpose and beneficiary as subsequent roles does not seem to play a role in providing some sort of similarity with the source domain. An example is the use of perí, ‘around’, ‘about’, with the genitive: (2) perì ptóliós te makh´¯esetai ¯edè gunaikôn about city:gen.f and fight:fut.mid.3sg ptc woman:gen.pl.f “he will fight for (our) city and for (our) women” (Il. 18.265).

Perí also extends to Topic in Greek. Most likely this is the role that provides the link with purpose and beneficiary: in (2), the city and the women are what the fight is about, they are the ‘topic’ of the fight. Note that the notion of reason is also relevant, as for all prepositions in this section. In (3) and (4) I give examples of epí, ‘on’, with the dative:

 Silvia Luraghi

(3) gastéres haíd’ aigôn ... tàs epì belly:nom.pl.f dem.nom.pl.f goat:gen.pl dem.acc.pl.f on dórp¯oi kathémetha supper:dat set:aor.mid.1pl “those goats’ paunches ... which we set there for supper” (Od. 18.44–45); (4) ei mèn nûn epì áll¯oi aethleúoimen if ptc now on indef.dat.m contend:opt.prs.1pl “if for some other’s honor we were now holding contests” (Il. 23.274).

Again, the semantic extension does not rely on directionality. The landmark is conceived as a physical support for the trajector; when shifted to the abstract plane, it provides a basis, hence a reason, for a certain action. The notions of exchange and substitution also provide a path for extension to beneficiary and purpose: this can be seen in the case of the preposition pró, ‘before’. In Homer, pró denotes beneficiary or purpose with verbs of fighting and animate or inanimate NPs in examples similar to perí (2); examples from later Greek include occurrences such as (5) and (6): (5) m¯edè prò spodoû ménontas kinduneúein neg before dust:gen stay:part.prs.acc.pl risk:inf.prs “(there was no need) to stay and risk their lives for the dust (of the desert)” (Hdt. 4.11.2); (6) ou gár ti ... moúnous prò humé¯on deî neg ptc indef.n/a alone:acc.pl before 2pl.gen need:prs.3sg apolésthai perish:inf.aor.mid “for it is not right that (we) should perish alone for your sakes” (Hdt. 7.172.2).

. Cause as source/origin In Ancient Greek, cause can be expressed by prepositions that denote source or origin, such as ek, ‘out of ’, and apó, ‘from’. This type of syncretism is predicted by the position of cause in the causal chain, and is in accordance with its definition as an antecedent role. Note that source prepositions can denote cause only with inanimate nouns: animate nouns with ablative prepositions are taken to denote source, origin, or agent. In Homeric Greek, lexical constraints were even greater: only ek could occur in cause expression and limited to abstract nouns (see Luraghi 2003: § 3.2).

From cause to beneficiary and purpose 

. Cause as location Several of the prepositions seen in §3.2 can also denote cause. It is the case of perí, as shown in (7) autoîsi toîsi Sparti´¯et¯eisi ... éris dem.dat.pl.m art.dat.pl.m Spartans:dat.pl.m feud:nom.f eoûsa pròs Argeíous perì kho´¯rou be:part.prs.nom.f toward Argive:acc.pl about country:gen kaleoménou Thuré¯es call:part.prs.m/p.gen Thyrea:gen “the Spartans themselves had a feud on hand with the Argives, on account of the country called Thyrea” (Hdt. 1.82.1).

Again, this type of cause should better be reagarded as reason, since it involves an agent’s intentionality. This is not always the case: sometimes, the fact that a PP can be taken as denoting cause, rather than purpose or beneficiary, depends on its occurrence with uncontrolled states of affairs (a purpose requires the agent’s intentionality, a cause does not), as in (8) epì soì mála pollà páthon on 2sg.dat very many:n/a.pl suffer:aor.1sg “because of you I have suffered much” (Il. 9.492).

Limited to Homeric Greek, pró, ‘before’, can also denote cause, in expressions such as prò phóboio, ‘out of fear’. Pró occurs with nouns that denote emotions: rather than by the notion of exchange, the semantic extension to cause seems to be motivated by the concrete spatial meaning of the preposition. An emotion is conceived as something that manifests itself in front of an experiencer, thus prompting his/her reaction (see Luraghi, forthcoming b, on a similar development in Latin). The preposition that most frequently denotes cause in Ancient Greek is diá with the accusative. Its spatial meaning, ‘about’, is limited to Homeric Greek (later this type of PP remains limited to cause); it denotes non-directional movement of a trajector inside a bounded landmark (multiple path):3 (9) autàr ho bê dià dôma ... óphr’ ptc dem.nom walk:aor.3sg about hall:n/a until híket’ Ar´¯et¯en reach:aor.mid.3sg A.:acc “but he went about in the hall until he came to Arete” (Od. 7.139–141).

 Silvia Luraghi

Diá denotes cause with all types of noun, including animate and concrete inanimate nouns. It must be stressed that this is the only preposition that can denote cause with human referents without further conditions (as e.g. that the state of affairs is uncontrolled, as in the case of epí, illustrated above): (10) gêmen heòn dià kállos marry:aor.3sg poss.3sg.n/a for beauty:n/a “(he) married (her) because of her beauty” (Od. 11.282); (11) deísantes ôn hoi Lampsak¯enoì Kroîson fear:part.aor.nom.pl ptc art.nom.pl from.L.:nom.pl K.:acc lúsantes metêkan Miltiádea. hoûtos mèn d`¯e free:part.aor.nom.pl release:aor.3pl M:acc dem.nom ptc ptc dià Kroîson ekphéugei for K.:acc escape:prs.3sg “in fear of Croesus the inhabitants of Lampsacus freed Miltiades and let him go. So Miltiades was saved because of Croesus” (Hdt. 6.37.2–38.1); (12) Thebaîoi mén nun kaì hósoi dià toútous Theban:nom.pl ptc ptc and rel.nom.pl through dem.acc.pl oí¯on apékhontai sheep:gen.pl be.away:inf.prs.m/p.3pl “the Thebans and those who by the Theban example will not touch sheep” (Hdt. 2.42.3). (13) dià t`¯en kuné¯en pheúgein es tà hélea for art.acc.f helmet:acc.f flee:inf.prs to art.n/a.pl marsh:n/a.pl “to be driven away into the marshes by reason of the matter of the helmet” (Hdt. 2.152.2).

Another preposition that can frequently denote cause is hupó with the genitive (or dative limited to Homeric Greek). Its spatial meaning is ‘under’, and the extension to cause is based on a metaphor by which the idea of physical superiority is taken as denoting control over an event (see Lakoff & Johnson 1980: 15). Examples are: (14) hupò pl´¯etheos oudeìs àn eípoi arithmón under multitude:gen indef.nom ptc say:opt.aor.3sg number:acc “no one could tell the number, so many they were” (Hdt. 7.187.1); (15) hoi ánthr¯opoi hupò toû kaúmatos mélanes art.nom man:nom under art.gen heat:gen black:nom.pl eóntes be:part.prs.nom.pl “the men (of the country) are black by reason of the heat” (Hdt. 2.22.3).

From cause to beneficiary and purpose 

Hupó can denote cause with abstract and concrete nouns, but not with animate nouns, because in such case it usually denotes agent and occurs with passive verbs. . Cause, reason, and purpose Already in Classical Greek, diá with the accusative is found in some occurrences where its meaning shifts from cause to purpose. The connecting function of reason is clearly visible: (16) h¯egoúmenoi autôn dià t`¯en sphetéran lead:part.prs.m/p.nom.pl dem.gen.pl for art.acc.f poss.3pl.acc.f dóxan glory:acc.f “using their supremacy over them to promote their own glory” (Th. 2.89.4). (17) éti h¯os tò télos: toûto d’ estì tò ptc so art.n/a end:n/a dem.n/a ptc be:3sg.prs art.n/a hoû héneka, hoîon toû peripateîn h¯e rel.gen.n for such:n/a art.gen.n walk:inf.prs art.nom.f hugíeia. dià tí gàr peripateî? phamén. hína health:nom.f for int.n/a ptc walk:inf.prs say:prs.1pl in.order.to hugiaín¯ei. kaì eipóntes hoút¯os be.healthy:subj.prs.3sg and say:part.prs.nom.pl thus oiómetha apoded¯okénai tò aítion think:prs.m/p.1pl give:inf.prf art.n/a cause:n/a “the same as end; i.e. the final cause; e.g., as the end of walking is health. For why does one walk? ‘To be healthy,’ we say, and by saying this we consider that we have supplied the cause.” (Arist. Metaph. 1013a 34–35).

In (16) dià t`¯en sphetéran dóxan can be said to denote the reason for the understood agent to act, but it certainly refers to something that follows the action, rather than precede it, so it certainly cannot denote cause (compare (16) with (11), where the diá phrase denotes a reason that precedes the event, and can be understood as a cause). Example (17) is of particular interest: the expression dià tí, ‘why’, normally occurs when the answer contains a cause expression. Here the answer is a purpose clause introduced by hína, ‘in order to’: through the ambiguity of diá when denoting reason, Aristotle captures the link between a purpose and a cause.

 Silvia Luraghi

. Byzantine Greek In Byzantine Greek the extension of diá to subsequent roles continues: besides purpose expressions, as in (18) and (19), we now also find beneficiary expressions, including behalf, as in (21): (18) hoi dè Argonaûtai ... anépleusan dià tò krúseon art.nom.pl ptc A.:nom.pl sail:aor.3pl for art.n/a golden:n/a déras fleece:n/a “the Argonauts sailed for (i.e. to find) the golden fleece” (Malalas Chron. 79.10–11); (19) bállousi pûr eis tòn b¯omòn dià t¯`en put:prs.3pl fire:n/a to art.acc altar:acc for art.acc.f thusían sacrifice:acc.f “they put fire on the altar for the sacrifice” (Malalas Chron. 112.2); (20) dià toûto épempsá soi skûtos kaì sphaîran kaì for dem.n/a sent:aor.1sg 2sg.dat thong:n/a and globe:acc.f and khrusíon, hína ár¯eis hóti potè boúlei gold:acc for take:subj.prs.2sg indef.n/a.sg ptc want:prs.m/p.2sg “for this reason I sent you a thing, a globe, and gold, in order for you to take whatever you may want” (Historia Alexandri Magni 1.36); (21) grápsas Rh¯omaíois dià Sérgión tina write:part.aor.nom.sg Roman:dat.pl for Sergius:acc indef.acc diákonon diaconus:acc “having written to the Romans on behalf of a diaconus Sergius” (Malalas Chron. 466.20); (22) pántes, gár ph¯esin, leitourgikà pneúmatá eisin all:nom.pl ptc say:prs.3sg minister:n/a.pl soul:n/a.pl be:prs.3pl eis diakonían apostellómena dià toùs to attendance:acc.f send:part.prs.m/p.n/a.pl for art.acc.pl méllontas kl¯eronomeîn s¯oterían be.about:part.prs.acc.pl acquire:inf.prs salvation:acc.f “all, it is said, are ministering souls which are sent as attendants for those who are about to receive salvation” (Georgius Monachus 110.19);

From cause to beneficiary and purpose

(23) dià t`¯en kh´¯eran kaì tà orphanà ho for art.acc.f widow:acc.f and art.n/a.pl orphans:n/a.pl art.nom theòs epoí¯ese tò prâgma god:nom do:aor.3sg art.n/a deed:n/a “God did this for the sake of the widow and the orphans” (Apophth. 252c, from Jannaris, 1897: 375).

In examples (18)–(20) the meaning extension is similar to the extension sporadically found in Classical Greek, and illustrated above in examples (16) and (17): in (20) in particular we again find a purpose clause with hína, similar to (17). The real shift happens with the occurrence of animate NPs: in this case, the meaning of diá in Byzantine Greek no longer continues any of the possible meanings the preposition had in earlier language stages. If we compare the occurrences in (21)–(23) with (11) and (12) we see that the two meanings seem incompatible: in (11) and (12) two actions are said to be performed because of some human referents who, however, do not benefit from the actions. Indeed cause expressions such as those in (11) and (12) no longer occur in Byzantine Greek: apparently, when the preposition extends to beneficiary, its causal meaning remains, limited to contexts where it cannot cause ambiguity. If the landmark is inanimate, possible ambiguity between cause and purpose is acceptable because of the overlap of the two roles in the area of reason: as Radden (1989: 562) puts it, “the objects we aspire for are usually also the cause for our aspiration”. However, this does not work for beneficiary and cause: as I remarked in §2.3, a beneficiary is something more complex than a human purpose. In (1a) fun denotes a purpose and as such it can be taken to refer to the object of intention or to its motivation, but you in (1b) is not the object of intention: rather, the object of intention is something such as your benefit. So beneficiary introduces a further level, and can hardly be interpreted as a cause.

. Modern Greek In Modern Greek most of the Ancient Greek prepositions surveyed in §3 have disappeared. Among those which still continue, we find s(e), derived from eis, which can denote recipient and direction, and has substituted the dative case, and apó, ‘from’, which, as in Ancient Greek, denotes source, and also partly substitutes the genitive case. In the case of these prepositions there is no syncretism of subsequent and antecedent roles; their meaning has become increasingly abstract due to the reduction of the case system.4 In particular, note that apó





Silvia Luraghi

can still denote cause in Modern Greek, but it does not denote purpose or beneficiary, i.e. it has not taken part in the merger of these semantic roles, discussed below. The preposition diá has undergone a phonological change and corresponds to Modern Greek ja. It now denotes cause, as in (24), purpose, as in (25), and beneficiary, as in (26). Cause expressions with ja are basically limited to inanimate nouns, and frequently contain abstract nouns: (24) siγnómi ja tin kaθistérisi I.regret for the delay “I am sorry for the delay”; (25) piγe sti vrísi ja neró s/he.went to.the tap for water “s/he went to the tap for water”; (26) to ’kana ja séna it I.did for you “I did it for you” (from Holton et al. 1997: 385–386).

It is interesting to observe how Ancient Greek diá in cause expressions can be translated into Modern Greek. With inanimate nouns, já is possible, especially with abstract nouns, as shown in (27), to be compared with (10):5 (27) tin pandreftike ja tin omorfia tis her he.married for the beauty hers “he married her for her beauty”.

Concrete nouns are less flexible: if a purpose interpretation is possible, it is preferred, and cause must be expressed differently. Compare (28a, b) with (13): (28) a.

épese sto élos eksetías tu kránus tu he.fell in.the marsh because of.the helmet his “he fell in the marsh because of his helmet” b. épese sto élos ja to krános tu he.fell in.the marsh for the helmet his “he fell in the marsh in order to find his helmet”.

Animate nouns trigger the beneficiary interpretation; again, cause must be expressed through another preposition, as shown by comparison of (29a, b) with (11): (29) a.

sóθike xári s-ton Kríso he.was.saved because to.the Cresus “he was rescued because of Cresus”

From cause to beneficiary and purpose

b. sóθike ja ton Kríso “he was saved in the interest of Cresus”.

. Discussion The semantic development of the preposition diá, and its relation with other possible means of encoding cause, beneficiary, and purpose in the history of Greek give rise to some insightful remarks: a. the direction of semantic spread does not necessarily procede from purpose to cause, as predicted by Croft, but is can also go contrarywise; b. some lexical features are crucial to the development of possible syncretism: in the first place, animate nouns seem to play a special role, because they introduce a third semantic role, beneficiary, which cannot easily merge with cause; c. intentionality also plays a special role in bringing about possible polysemy: cause can co-occur with both uncontrolled and controlled states of affairs, while purpose can only co-occur with the latter. With controlled states of affairs there is an area of overlap of the two roles, which is the area covered by reason. One may further wonder which spatial bases can be found for syncretism of antecedent and subsequent roles in the causal chain. As I have remarked above, prepositions that allow this type of syncretism are prepositions that do not have a directional component in their basic spatial meaning. Indeed direction and source preposition, when their meaning is shifted to the abstract plane of causation, apparently follow the prediction in the causal approach and only extend to either antecedent or subsequent roles. We have seen that many prepositions displayed possible syncretism of cause and purpose already in Ancient Greek: but it must be stressed that these are all prepositions whose basic spatial meaning is location. Far from being a peculiarity of Greek, this same situation can also be seen in other languages. In Luraghi (forthcoming b) I have shown that the same patterns of syncretism held in Latin and led to developments in the Romance languages that are partly similar to the Greek development. Because the Latin evidence displays some similarities, but also some interesting differences with respect to Greek, I will briefly survey it in the next section.



 Silvia Luraghi

. Cause, purpose, and beneficiary in Latin In Latin several prepositions can denote cause; the most interesting for the purposes of this paper are propter and pro. Similar to Greek diá, the former can denote cause with all types of referent, including animate nouns, and does not display syncretism with other roles until Vulgar Latin. Pro, instead, denotes exchange as its basic meaning, and it can occur in cause expressions with inanimate nouns.6 When occurring with animate nouns, it either preserves the meaning of exchange, or denotes beneficiary. So there is a different syncretism of cause with other semantic roles, depending on animacy: Greek

dative and allative prepositions non-directional prepositions beneficiary [+a]/purpose [–a]

diá (non-directional) [+/–a] cause

Latin

dative (subsequent) beneficiary [+a]/ purpose [–a] ad (subsequent) purpose [–a]

pro (non-directional) beneficiary [+a]/ cause [–a]

propter (non-directional) cause [+/–a]

In Latin, purpose is originally only expressed through markers of subsequent roles, while syncretism concerns beneficiary and cause, already in the earliest texts. As noted in §2.3, this syncretism is only partial: crucially, the preposition which displays it cannot be used if the cause is animate. In this case, another preposition occurs, which unambiguously denotes cause. Ambiguity does not concern prepositions that can denote beneficiary and purpose, not because these two roles are more closely related than beneficiary and cause, but simply because the distribution of the two roles based on animacy is so to speak automatic: as remarked in §2.3, under normal circumstances purposes are inanimate. Note further that in Latin, too, cause could also be expressed by source prepositions, albeit to a limited extent. In Vulgar Latin, the type of syncretism that we have seen in Byzantine Greek also took place: pro and propter occur in purpose, beneficiary, and cause expressions. At this stage, animate nouns tend to be interpreted as denoting beneficiary even when occurring with propter. A further similarity between the Latin and the Greek developments is that allative markers that can denote purpose and beneficiary do not extend to cause, and ablative markers that denote cause do not extend to purpose or

From cause to beneficiary and purpose

beneficiary: in other words, possible syncretism among antecedent and subsequent roles seems to be constrained by the original spatial meaning of specific coding devices involved.

. Summary and conclusions In the present paper I have surveyed various ways of encoding cause, purpose, and beneficiary throughout the history of Greek, and have compared the Greek data with similar data from Latin. I have shown that the three roles can be understood through a spatial metaphor based on directionality, in which case ablative forms extend to cause and allative forms extend to purpose and beneficiary. When this happens, the three roles do not seem to merge with each other in subsequent semantic developments. The three roles can also be understood as metaphorically representing some sort of location. In this latter case, all three role are encoded through locative forms, which do not imply directionality on the spatial plane. Lack of spatial directionality matches lack of causal directionality on the abstract plane, and proves particularly fit to encode the reason role. Reason provides an area in which cause can overlap with purpose when it occurs in controlled states of affairs (i.e. when it motivates the action of an agent). Thus possible syncretism of cause (antecedent) and purpose (subsequent) appears to crucially depend on the type of spatial metaphor that underlies the extension of specific forms form the source domain (space) to the target domain. An important difference between possible syncretism of beneficiary with either purpose or cause is constituted by the treatment of animate NPs. While purpose and beneficiary are possibly complementary roles with respect to animacy (purpose being assigned to inanimate NPs and beneficiary to animate ones), cause does not have this peculiarity: causes can be inanimate as well as animate. Consequently, syncretism of cause and beneficiary appears to be limited to patterns in which the same form that can denote cause with inanimate nouns can denote beneficiary with animate ones; another form must be used to express cause with animate nouns. Note that easier merger of purpose and beneficiary does not imply that this syncretism necessarily precedes syncretism of either role with cause. In this respect, Greek and Latin display different patterns: while in Greek the two subsequent roles indeed display syncretism already at an early stage, and only later merge with cause, in Latin the earliest syncretism concerns cause and beneficiary.7



 Silvia Luraghi

Notes . I thank Bill Croft and Kiki Nikiforidou for helpful comments on an earlier draft. . Most of the Greek prepositions discussed in this paper can take more than one case. It must be stressed that whatever is said about a certain preposition with a certain case only holds limited to the case in question: prepositions have other meanings connected with case variation, that I am not going to discuss here. . Note that the meaning of diá with the accusative which provides the basis for the extension to cause is non-directional: diá with the accusative does not denote a straight trajectory which crosses a landmark from one side to the other, but rather, as shown in the example, a multi-directional trajectory that remains inside the landmark (as do two-way prepositions in German when taking the dative; see Luraghi forthcoming a and 2003: §3.9). . On the semantic development of eis and apó see also Bortone (2000). . I thank Stavros Skopeteas for helping me with the Modern Greek translations of the Ancient Greek examples. . Pro also denotes purpose, but apparently not in Early Latin: in other words, extension form exchange to cause seems to precede extension to purpose. . This is only partially true of course: the dative case in Latin displays syncretism of beneficiary and purpose at an early stage. Note however that the dative is outside the scope of the present discussion, because it remains limited to subsequent roles throughout its history. In this paper I am concerned only with forms that display syncretism of subsequent and antecedent roles.

Abbreviations acc aor art dat dem f fut gen impf indef inf int m mid m/p

accusative aorist article dative demonstrative feminine future genitive imperfect indefinite infinitive interrogative masculine middle medio-passive

n n/a neg nom opt p part pf pl poss prs ptc rel sg subj

neuter nominative/accusative neuter negation nominative optative passive participle perfect plural possessive present particle relative singular subjunctive

Note: singular number is not indicated for nominal categories (except for personal and possessive pronouns), while active diathesis and indicative mood are not indicated for verbal

From cause to beneficiary and purpose

categories; feminine gender is always indicated as f; neuter and masculine nominative and accusative have distinct glosses (nom and acc for masculine, n/a for neuter; the gloss m is only used in case of pronouns that have identical forms for masculine and feminine); for dative and genitive forms of masculine and neuter nouns gender is indicated only where it can help understand phrase structure.

References Bortone, Pietro (2000). Aspects of the history of Greek prepositions. Hilary Term: University of Oxford. Croft, William (1991). Syntactic Categories and Grammatical Relations, Chicago: Chicago UP. Holton, David et al. (1997). Greek. A Comprehensive Grammar of the Modern Language. London – New York: Routledge. Jannaris, A. N. (1897). An Historical Greek Grammar. London: Macmillan and Co. Lakoff, George, & Mark Johnson (1980). Metaphors we live by. Chicago: Chicago UP Luraghi, Silvia (2001). Syncretism and the classification of semantic roles. Sprachtypologie und Universalienforschung, 54 (1), 35–51. Luraghi, Silvia (2003). On the meaning of prepositions and cases. Amsterdam/Philadelphia: John Benjamins. Luraghi, Silvia (forthcoming a). The local meaning of diá with the accusative in Homer. In Hubert Cuyckens & Walter de Mulder (Eds.), Variation and Change in Adpositions of Movement. Amsterdam/Philadelphia: John Benjamins. Luraghi, Silvia (forthcoming b). Cause expressions in Latin. In Proceedings of the 12th Colloquium on Latin Linguistics (Bologna, 9–14 June 2003). Radden, Günter (1989). Figurative use of prepositions. In R. Dirven & R. A. Geiger (Eds.), A User’s Grammar of English (pp. 551–575). Frankfurt am Mein: Peter Lang.



Vanishing discourse markers Lat. et vs. sic in Old French and Old Romanian Maria M. Manoliu University of California, Davis

.

Introduction

The evolution of an adverb or a conjunction into a discourse marker has been widely studied providing convincing evidence for Traugott’s hypothesis (1991) that the development of adverbs and/or conjunctions into discourse markers meets the universal characteristics of grammaticalization. The evolution of Lat. et and sic in French and Romanian adds a new dimension to this evolution by showing how a discourse marker of text-coherence loses its capacity for signaling topic continuity and ends up by being a filler, with little chance of survival. Consequently various synonymous items fulfilling more appropriately its pragmatic and discourse functions replace it altogether. We shall first examine briefly the distribution of et and sic in Latin, then we shall present the functions of their offspring in Old Romanian, after bringing into the picture a few missing links between Latin and Romanian that are attested in Old French texts.

. Lat. et vs sic . Lat. et as a discourse marker As a conjunction, Lat. et could connect utterances with either coreferential or non-coreferential subjects. Compare (1) and (2).

 Maria M. Manoliu

(1) Pueri, plaudite et mihi ob lactum cantharo mulsum date (Plautus, Asinaria, v. 906). “Slaves, applaud and give me, for this successful deed, a cup of wine with honey”.

When connecting sentences with non-coreferential subjects, it may fulfill two functions: i.

When the connected terms are logical contraries (expressing opposite referents or events), the copulative conjunction takes a contextual adversative value (see (2)). (2) Lat. si tuus servus nullus fuerit et omnes alieni ac mercenarii (Cicero, Caecin. 58) “if there would be none of your servants but all foreigners and mercenaries”.

In a pragmatic framework, the adversative value is an effet de sens (side effect) brought in by the conversational implicature that presents the second term as being an opposite to the first term; in brief, it implies that ‘b is non-a’, usually asserting ‘a and b’. In (2) it explicitly asserts ‘non-a and b’.1 ii. Within larger contexts, et can function as a discourse marker connecting utterances that belong to the same narrative unit, with a super-ordinate topic (see (3)). (3) Megadorus: Nam meo quidem animo si idem faciant ceteri, Opluentiores pauperiorum filias Vt indotas ducant uxores domum, Et multo fiat ciuitas concordior, Et inuidia nos minore utamur quam utimur, Et illac malam rem metuant quam metuunt magis, Et nos minore sumptu simus quam sumus. (Plautus, Aulularia, v. 480–486) “Megadorus: Since I think that if everybody else would do as I did, If the rich people would marry without dowry the poor citizens’ daughters, Then (lit. And) there would be more agreement in the city, And we would be less envious, And they [=our women] would fear our severity more than they do now, And, as for us, we would have to spend less than we do now.”

Vanishing discourse markers

In Kroon’s opinion (1995: 106–107), in such cases et has a ‘presentational function’ because ‘the first clause of such units functions as a preliminary with regard to the following one(s), which by means of et is (are) explicitly introduced as a new, important stage in the development of the story’. iii. The focalizing et . . . et. The repetition of et before two terms expresses the idea that they make a unitary compound, despite a previous negative expectation: (4) Alcvmena: Equidem ecastor uigilo et uigilans id quod factum est fabulor; Nam dudum ante lucem et istunc et te uidi (Plautus, Amphitrvo, 2, 2, 68) “Alcimena: I am very much awake and, I swear it, being awake I am telling you what happened. Since, just before daylight, I saw both (lit. and) this one and you.”

. Lat. sic as adverb and/or conjunction In our opinion, thanks to its anaphoric value of “in this way, so, yes”, Lat. sic must have been used in spoken registers, first of all, for intensifying the value of et as a marker of confirming the addition of another argument of the predicate, which enabled it to invalidate a contrary expectation. In fact, it has always been considered that a copulative sic ‘thus, so’ could be found already in Latin texts. It is rather interesting that in the same Vulgar Latin text, namely Peregrinatio Aetheriae, sic can co-occur with et or stand alone fulfilling the same function. In (5) sic follows et for the purpose of expressing the equal treatment of both categories (the catechumens and the faithful); or to put it in pragmatic terms, sic (as a so called “intensifier” or “reinforcing device”) served the purpose of denying any expectation that only one category received the benediction.2 (5) Lat. benedicuntur cathecumini et sic fideles, et inde denuo [. . . ] itur ad Anastase (Aeth. 43,7) “both the catechumens and [lit. in the same way] the faithful [people] were blessed, and then they went to Anastasis”.

In (6) sic alone occurs in the same context: (6) benedicuntur cathecumini sic fideles (Aeth. 43,6) “both the catechumens and the faithful [people] were blessed.”

Once sic replaced et in contexts where it expressed the denial of an expectation as in (2), it could also be used in larger narrative units to emphasize the fact that they make a unitary compound, despite a negative expectation.



 Maria M. Manoliu

. Old French et versus si . French et We shall now turn our attention to the distribution of et in Old French, because there are interesting similarities of functions with O. Rom. e. Given the fact that the oldest Romanian texts that have been preserved were written as late as the 16th century, it is very likely that the distribution of Old French e(t) can provide some evidence for the missing links between the eastern variants of Vulgar Latin and 16th-century Romanian. In Old French e(t) could have all the functions it fulfills nowadays. Moreover it continued to fulfill several of the functions of its Latin etymon that later on were lost or became less frequent. For example, O. Fr. e(t) could function as a discourse or a pragmatic marker, denying various expectations. i.

Unexpected inclusion in the same category (to which the same predicate applies) was mostly expressed by the repetition of et before each term: (7) Il en apelet e ses dux e ses cuntes (Roland, 15 in Moignet 1973: 330) “He called both [lit. and] his dukes and his counts”.

Utterance (7) denies the expectation that ‘he called only one category’ (for instance, ses ducs ‘his dukes’) by asserting the fact that he called both his dukes and his counts; if no contrary expectation had been invalidated, it would have been normal to say: (7) a.

Il en apelet ses dux e ses cuntes. “He called his dukes and his counts”.

In contemporary French et can also be repeated with a similar value. According to Grevisse (1991: 1557), et and ou may be repeated before each term for energizing/intensifying or foregrounding the expression. In our opinion, what Grevisse defines as ‘l’énérgie ou le relief de l’expression’ corresponds to the pragmatically defined function of a conventional implicature consisting in the ‘denial of an expectation’ In this case the ‘energy of the expression’ is the effet de sens of the unexpected addition of an entity (expressed by either a NP or a determiner). ii. As a discourse marker of topic continuity within the same narrative unit, et could occur in sentence-initial position more often in Old French than nowadays, either preceding puis ‘then’ or not (see (8)).3

Vanishing discourse markers 

(8) E Deu l’en rendi gueredon, Quant il el ventre del peisson Le salva e gari de mort Et puis le mena a bon port (Best. 3133, in Reenen & Schøsler 2000: 87) “And God rewarded him When he in the belly of the fish Saved him and protected him from death, And then brought him to a good harbor.”

Buridant (2000: 552–555) gives a comprehensive inventory of the discursive functions of inter-sentential et. We shall recall here just one of them which provides ample evidence for the discourse function of et at the beginning of sentences, paralleling the distribution of e in Old Romanian: En plus de l’identité catégorielle entre deux propositions, et pleinement coordonant asserte leur solidarité discursive en posant que l’énonciation de l’un, à un moment donné et dans un certain univers de croyance, n’a de portée qu’en connexion avec l’énonciation de l’autre, en même temps que l’identité locutoire des éléments qu’il associe. “Besides the categorial identity of two clauses, a fully coordinating et asserts their discourse solidarity expressing the fact that the enunciation of one, at a certain moment and in a certain universe of beliefs, has truthvalue only in connection with the enunciation of the other, and in the same time the locutionary identity of the entities it associates.” (see example (9) below). (9) Après si sanlerent un jour tout li baron u palais l’empereur, et demanderent leur convenanches a l’empereur; et respondit qu’il les tenroit bien, mes il voloit anchois estre coronés. Et puis si atirerent et prisent un jour de li coroner; et a chel jour il fu coronés hautement come empereres par le volenté de sen pere qui li otria boinement. Et quant il fu coronés, li baron requisent leur paiement; et il dist qu’il parleroit molt volontiers chou qu’il porroit, et paia adonques bien.cm. mars. (Conqueste C, LVI). “Then all the lords got together one day at the emperor’s palace and asked him to keep his promises to them; and he answered that he would keep them, but first he wished to be crowned. Then they chose a day for doing that, and that day he was crowned solemnly as emperor by his father’s will, who agreed with no reserve. And when he was crowned, the lords asked for their pay; and he said that he would willingly pay what he could, and right away he gave them a hundred thousand marks.”

 Maria M. Manoliu

iii. Like its Latin counterpart, et could also have a contextual adversative value triggered by a context in which the coordinating sentences express opposite events. (10) Fr. Mes comme il plus le demanderent, et meins en But as they more him asked and less of.it apristrent (Artu, 37.7) learned:they. “But the more they asked him, the less they learned of it”.

. French si The hypothesis concerning the discourse function of Lat. sic as an intensifier implying “equality”, by inclusion into a superordinate group can explain the function of Old French si as a topic continuity marker. According to Thomasset and Ueltschi (1993: 118), ‘si fait de la première proposition un posé préalable, << un thème >> qui rend possible l’énonciation de la deuxième proposition.’ Unlike et, si always has an anaphoric function. It signals the temporal succession of events or the logical consequence similar to the use of alors in contemporary French (see Moignet 1973: 287). A detailed description of the discourse functions of O. Fr. se/si as a marker of topical continuity/discontinuity may be also found in Buridant (2000: 505–515). Here are a few examples of the uses of si as a discourse marker from Reenen & Schøsler (2000): i.

Sentence-initial si: when the coreferential subject of the sentence introduced by si is not expressed (zero anaphor). In this case, si has no corresponding translation in English:4 (11) Quant il eut che fait, si prist deux grandesmes pierres, si leur fist lier as cous et puis si les fist geter en le mer. Après si se fist coroner tot a force a empereur (Clari: XXI, 18–21). “When he had done this, he took two very big rocks, had them tied around their necks and then had them thrown into the sea. After that he had himself crowned emperor by force.” (Reenen & Schøsler 2000: 86).

ii. In (12) si carries the implicature of denying an expected entailment, traditionally labeled as an adversative value, which recalls one of the contextual values of Latin et:

Vanishing discourse markers 

(12) Et s’ or en deviez crever, Si diroie je tote And if now of.it must.you perish [si] would.say I every voie [. . . ] (Perc., 5498–5500) way [. . . ] “And if you were to perish now, I would say however . . . ”

iii. As in contemporary French, si may have the function of a pragmatic marker contradicting a previous negative statement: (13) – Por quoi le demandez vos? fet li roi. – Por ce, fet messier Gauvains, que ge ne cuit pas que vos le sachiez. – Si sai bien, fet li rois, mes vos ne le savez pas. (Artu, 29,30, in Moignet 1979: 288) – “Why are you asking this? the king asks. – Because, Sir G. says, I do not think you know it. – But I know it very well, the king says, but you do not know it.”

The spread of et and the very restricted use of si (only as an adverbial intensifier and as a pragmatic marker) in French are probably due to two factors: a. The competition with synonymous items such as alors carrying a conventional implicature of ‘temporal sequencing’ or ainsi, car, alors for expressing the relation of ‘consequence’. b. The homonymy with si (O. Fr. se/si), derived from Lat. conditional conjunction si. Since the evolution of its synonymous replacements as well as the contemporary functions of French si are well known thanks to a variety of studies (Moignet 1973; Marchello-Nizia 1995; Fleischman 1992; Hansen 1998; Reenen & Schøsler 2000; Buridant 2000: 505–515, to mention only a few), we shall now turn our attention to the evolution of its Romanian counterpart, ¸si.

. Old Romanian e versus ¸si Like French, Old Romanian had two copulative conjunctions: (i) the panRomance e originating in the Lat. conjunction et ‘and’; and (ii) ¸si, which originates in Lat. sic. As will be shown in what follows, in old Romanian texts, the two copulative conjunctions were not interchangeable but had a different distribution.

 Maria M. Manoliu

. Old Romanian e Most of the time, O. Rom. e occurs in sentence-initial position when the conjoined sentences do not have coreferential subjects. (i) The inter-sentential e (after stop) is used mainly as a presentational marker of discourse coherence in spite of a topic switch. The text-coherence is also marked by the fact that the subject of the sentence introduced by e has already been identified by the previous co-text. In (14) there is an explicit change in subject from the people defending their goddess (Artemis) to Alexander. The latest has been identified as the direct object of the previous sentence. (14) Era b˘aseareca turrburat˘a ¸si mai mul¸ti nu-¸s ¸stiia derep ce au veritu. Din gloat˘a aleaser˘a Alexandru ¸si-1 scoaser˘a elu iudeiloru. E Alexandru m˘ah˘ai cu mînra ¸si venre se rr˘aspundz˘a gloateei. În¸teleaser˘a c˘a iudeianinu easte. . . (CV: 239). “In the church, the crowd was in uproar and most of them did not even know why they had come. From the crowd [they] chose Alexander and took him in front of the Judeans. And Alexander motioned with his hand [for the crowd to be silent] and tried to answer the crowd. They understood that he was a Judean.”

(ii) When inter-clausal, after a pause (marked by a coma) e has also a presentational function activating a conversational implicature of contrast between the conjoined constituents ‘a and b’, namely that ‘b = non a’(see (15)): (15) Gr˘aescu voao, c˘a de¸stinse acesta dereptat˘u în casa lui, mai vârtos˘u de alalt˘u, c˘a to¸ti cine se înal¸ta˘ , pleac˘a-se, e cine se va plecà, de sine în˘al¸tà-se-va (CÎ: 17). “I am telling you, that this one [= the tax collector], and not the other [Pharisee] was the righteous one, when he went home, for everyone who makes himself great, will be humbled, and everyone who humbles himself, will be made great”.

In (16) e introduces a concessive clause after the main (chain of) clause(s). In other words, e activates a conventional implicature of contrast denying an expected entailment: (16) Nu numai aceasta nevoie preimea¸ste a noastr˘a parrte întru oblicire a venri, ce ¸si a mariei dumnedzeaie cas˘a, a Arrtemideei, întru nemic˘a o pomenea¸ste ¸si va se sparrg˘a m˘ariia ei, e ea toat˘a lumea o cinstea¸ste’. (CV:237); “But not only great misfortune will come upon our business, but there is also danger that the temple of the great goddess Artemis will come to

Vanishing discourse markers 

mean nothing and her greatness will be destroyed, and (= though) she is worshipped by the whole world.”

As mentioned before, in a pragmatic framework, such a contrast is a side effect brought in by the conversational implicature that ‘b is non-a’, ‘a → non b’, asserting ‘a and b’, where ‘a’ stands for ‘she is worshipped by the whole world’ and ‘b’ for the sequence of unfortunate happenings (‘Artemis will come to mean nothing and her greatness will be destroyed’). As will be shown below, this conversational implicature will attract the use of a marker carrying a synonymous conventional implicature (see 5.2 below). The proof that e had already become a mere marker of text coherence (or even only a filler) is provided by contexts where it occurs after a stop and is followed by the adversative conjunction îns˘a ‘but’, like in (17): (17) S¸ i acmu eu legatu-s˘u cu Duhul de viiu întru Ierusalim ¸si nu ¸stiu ce mi se va tâmpla într-însu. E îns˘a cum Duhul Sfântu prin toate cet˘a¸tile m˘arrturisea¸ste ¸si gr˘aia¸ste c˘a leg˘aturi ¸si scârbi menre a¸steapt˘a-me. (CV, 247–248). “And now I am bound [by oath] with the Holy Spirit to come to Jerusalem and I do not know what will happen there. But [lit. and but] as the Holy Spirit warns and says in every town that troubles and sufferings await me.”

The last stages of grammaticalization, when e becomes a mere filler rather than a discourse marker, are attested in contemporary Istro-Romanian, a rather conservative Southern-Danubian dialect, where only a sentence-initial e (after a full stop) is still in use:5 ˇ ¸sti¸t. Iå åre [. . . ]. Fil’u dvåiset ¸si doi åñ va (18) Bíserca, sora jivé-n Lovrån. Cå ave˛-n novémber. E ¸si om åre din Her¸tegóvina. . . (Túrcovit, Maria, 68 ani, in Sârbu and Fr˘at¸il˘a, 1998: 102). “Biserca, [my] sister lives in Lovran. Cause [you] know. She is . . . [Her] son will be twenty-two years in November. And [she] also has a husband from Herzegovina. . .”

As in Old Romanian, Istro-Romanian e may be synonymous with the DacoRomanian iar, a weak adversative conjunction: (19) Catrída-i óbiˇcna po nå¸su ˇce cuvintåm catrida, cu re¸t.. hârvå¸tki. E scånd a fost opilíta de ur cus de búcva, pac se ras¸tipe˛it-o ¸s-atúnˇc s-å fost tu¸si ˇcå flanca ¸si pus-av trei piˇcóre. . . (Ióje lu Panoc, 84 years, in Sârbu and Fr˘at¸il˘a, 1998: 102) Daco-rom: Banca, deobicei, este pe graiul ce-l vorbim ‘catrida’, cum spune¸ti . . . în croat˘a. Iar când a fost cioplit˘a dintr’o bucat˘a de stejar, atunci s’a despi-

 Maria M. Manoliu

cat [o bucat˘a de stejar] ¸si-atunci acea blan˘a s’a cioplit ¸si s-au pus trei picioare (translation by Sârbu and Fr˘at¸il˘a, 1998: 243) “The bench, we usually call ‘catrida’, as you say . . . in Croatian. But when it was being cut from a piece of beech, immediately it split and then that piece of beech [that was left] was carved/ cut even and three legs were added [to it] . . .”.

. Si in Old Romanian .. Si ¸ as a discourse marker The conventional implicature carried by sic (as marker of “equality”) in colloquial Latin can also account for the fact that, in Old Romanian, unlike e, the conjunction ¸si is the preferred choice when the subjects are coreferential (see (20)). (20) Din gloat˘a aleaser˘a Alexandru ¸si-1 scoaser˘a elu iudeiloru. E Alexandru m˘ah˘ai cu mînra ¸si venre se rr˘aspundz˘a gloateei. (CV: 239) “From the crowd [they] chose Alexander and took him in front of the Judeans. And Alexander motioned with his hand for [the crowd to be silent] and tried to answer the crowd.”

In sermons of the same period, the sentence-initial e is retained mainly in quotations from the Bible, whereas ¸si replaces e as the preferred marker of topic continuity (see (21)). (21) Înv˘a¸ta˘ pre noi cu bun˘a în¸telep¸tie calea sp˘aseniei. S¸ i s˘a ne întoarc˘a pre noi cu bun˘a în¸telep¸tie pe calea sp˘aseniei. S¸ i s˘a ne întoarc˘a den întunearec˘u c˘atr˘a lumin˘a, acesta amu trudele ¸si boalele noastre r˘abd˘a; s˘a se munceasc˘a ¸si s˘a moar˘a derept˘u p˘acatele noastre vru. S¸ i ranele lui noi to¸ti ne-am˘u vindecat˘u. (CÎ: 2) “He taught us with good wisdom the way of repentance. And to bring us back with good wisdom to the road of repentance. And, to bring us back from darkness to light, he endured our pains and illnesses; he wished to suffer and to die for our sins,. And [thanks to] his wounds we are all healed.”

As shown by Tables 1 and 2, in the 16th– and 17th-cs.–texts, ¸si has a high percentage of the representative function (inter-clausal and inner-clausal) whereas e functions mostly as a presentative discourse marker (in inter-sentencial position, after a full stop, a colon or a coma).

Vanishing discourse markers 

Table 1. The distribution of e in 16th–17th centuries Text

Total

Inter-sentential

After a pause(,)

Inter-clausal

CV 15th–16th

23 100%

16 70%

2 9%

5 21%

Coresi 1581 (2–33)

24 100%

8 33.4%

14 58.3%

2 8.3%

CS 1583–1614 (237–256 and 280–289)

22 100%

12 55%

8 36%

2 9%;

Table 2. The distribution of ¸si in 16th–17th centuries Text

Total

Inter-sentential

After a pause(,)

Inter-clausal

Inner-clausal

CV 15th–16th

245 100

58 24%

11 4%

136 56%

40 16%

Coresi 1581 (2–33)

709 100%

28 4%

73 10%

338 48%

270 38%

Înv. 508–538 1642

286 100%

41 14%

13 5%

135 47%;

97 34%

.. Rom. s¸ i as a pragmatic marker Though much more frequently than French et, Romanian ¸si can also function as a pragmatic marker. When strongly stressed, adverbial ¸si (< Lat. sic) occurs as an expression denying an expected non-inclusion in a superordinate unit in both Old and Contemporary Romanian. As such, ¸si has two types of occurrences: (a) the repetitive ¸si: when it occurs before each term of the unitary entity, and (b) the intensive ¸si: when it occurs only once before the constituent referring to an unexpected additional entity (object, quality, event). The repetition of ¸si before each constituent is frequently used in religious texts as a means of persuasion (see also Lat. et in (4) above) or Fr. et in (7) above): (22) M˘acar˘a preo¸ti, m˘acar˘a mireani, înc˘a ¸si arhierei, ori-fiecarele, tot li se cade s˘a fie deplin singuri ca Hristos, smeri¸ti, milostivi, iubitori de Dumnezeu, milosârzi, ndelungu r˘abd˘atori, mult milostivi ¸si de alte de toate bun˘at˘a¸tile [. . . ] iar˘a nu v˘anz˘atori [. . . ] ¸si nemilosârzi ¸si nemilostivi ¸si n˘aprasnici ¸si m˘anio¸si ca mine, ocaianicul de mine (Inv.: 522)

 Maria M. Manoliu

“Even if [we are] priests, even lay people, even more so [if we are] bishops, no matter who [we are], we all have to be ourselves like Christ, humble, generous, loving God, merciful, longsuffering, very generous and [have] all the good qualities [. . .] and not traitors [..] and mean and greedy and violent and angry, like poor me.”

Depending upon the type of syntactic constituent it determines, the intensive ¸si has the following contextual meanings attested since the 16th c.: (i) Before NPs or adjectives ¸si means ‘also’: (23) Dzise: ‘Fiindu acolo, cade-mi-se ¸sí Rrâmul a vedea’. (CV: 235) “He said: ‘Once I am there, I should also see Rome’.”

(ii) Before verbs it may have two values: (a) ‘even’ (see (24)) or (b) ‘right away’ (see (25)).6 (24) E s’ai ¸sí luat˘u, pentru ce te lauzi c˘a n’ai luat˘u? (CÎ: 16) “And even if you took [it], why do you claim that you did not take [it]?” (25) domnii au ¸sí trecut Dun˘area cu o¸stile lor f˘ar˘a de ¸stirea veziriului (Ureche: 37) “the kings crossed the Danube right away with their troops without the vizier’s knowledge”.

. Factors favoring the loss of e As Table 3 shows, the 16th and 17th cs. – texts display an increasing difference between the frequency of e and ¸si. In a later text, namely Melchisedec’s Înv˘a¸ta˘ turi preste toate zilele (1642), e is no longer attested. The factors favoring the disappearance of e and the spread of ¸si are mainly due to similar pragmatic or discourse functions. . The competition between e and s¸ i O. Rom. e occurs as a copulative conjunction connecting mainly clauses with different subjects and/or contrary predicates. Such contexts that could activate a conversational implicature that ‘b = non-a’, favored its reinterpretation as a weak adversative conjunction, carrying the corresponding conventional implicature. As such it became even less appropriate for fulfilling the function of a representative marker expressing close connection between events. Its competitor, ¸si, was able to express precisely the equality between the conjoined

Vanishing discourse markers

Table 3. e versus ¸si in 15th–17th centuries Texts

Conjunctions Total per Inter-sentential: Inter-clausal Inner-clausal conjunction presentative

CV E 15th–16th cs. (229–279) S¸ I

23 9% 245 91 % Total conjunc. 268

18 21% 69 79% 87

5 4% 136 96% 141

0

Coresi, CÎ (2–33) 1581

E

22 18% 101 82% 123

2 0.5% 338 99.5% 340

0

12 5% 235 95% 247

10 3% 333 97% 343

0

24 3.4% S¸ I 709 96.6% Total conjunc. 733

CS, E (237–256 and 280–289) S¸ i 1583–1614 Total conjunc.

22 3% 731 97% 753

40 100% 40

270 100% 270

163 100% 163

members and, unlike O. Fr. si(e), did not have a homophonous hypothetical counterpart originating in Lat. si ‘if ’, which resulted in O. Rom. se (see (26)). (26) se întreb˘ari sâmtu (CV: 229. 1505–1525) if questions are “if there are any questions”.

Moreover, the hypothetical se was early replaced by deaca (Cont. Rom. dac˘a) ‘if ’: (27) deaca ne grijim˘u noi de ale trupurilor˘u noastre, datori sântem˘u mai vârtos˘u s˘a av˘am˘u ¸si s˘a grijim˘u de a sufletului (CÎ: 3). “if we take care of our body, even more so we have to take care of our soul”.

When introducing the subjunctive-clause, se was replaced by s˘a, which became the marker of modality rather than a conjunction. Comp. s˘a in (27) above with se in (28) below. (28) to¸ti ceia ce viia întru Asiia se audz˘a cuvântul Domnului all hose who came to Asia that hear:subj word Lord Isus (CV:232) Jesus “all those who came to Asia to hear the word of Lord Jesus”.



 Maria M. Manoliu

. e versus iar˘a The copulative/adversative e was also in competition with another conjunction, namely iar˘a. Of Greek origin, iar(˘a) (Gr. έ/άρα) also had an interesting evolution.7 Basically it expressed an additional event, either identical (‘again’) or not (‘also’), with the aim of invalidating an expectation of non-addition (see (29)). (29) Zise amu Hristos˘u întru sfânta evanghelie: ‘Nu venii s˘a osândesc lumea, ce s˘a sp˘asesc˘u lumea; nu venii a chemà derep¸tii, ce p˘ac˘ato¸sii la poc˘ain¸ta˘ !’. S¸ i: ‘Nu trebuia¸ste s˘an˘ato¸silor˘u vraci, ce bolnavilor˘u!’. S¸ i iar˘a: ‘Poc˘ai¸ti-v˘a, c˘a se aproprie împ˘ar˘a¸tiia ceriului!’ S¸ i iar˘a:. . . (Coresi, CÎ: 5) “So Christ said in the Holy Gospel: ‘I did not come to condemn the world but to redeem the world. I did not come to call the righteous, but sinners to repentance.’ And: ‘the healthy do not need a healer, but the sick!’ And again/ also [he says]: ‘Repent, because the kingdom of God is at hand!’ And again/ also:. . . ”

In Coresi’s Carte cu înv˘a¸ta˘ tur˘a, iar˘a may still have a repetitive value, mostly when following a sentence-initial ¸si: see S¸ i iar˘a in (29) above. But, as (30) shows, iar˘a could often be preceded by the conjunctions înc˘a ‘once again’, which shows that iar˘a alone could not fully express ‘the repetition of the same event’: (30) S¸ i înc˘a iar˘a gr˘aia¸ste David˘u: [. . . ] (Coresi, CÎ: 4) And also again speaks David: [. . . ] “And David also (again) says: [. . . ]”

In order to reinforce its function of expressing the similarity (if not identity) of additional events, except their time span, iar˘a combined with the reflexive ¸si (
As shown by (32), iar˘a usually takes the value of a weak adversative conjunction, in contexts referring to opposite events (carrying the conventional implicature: ‘b = non-a’).

Vanishing discourse markers 

Table 4. e versus iar in 15th–17th centuries Texts

Total per conjunction

Inter-sentential

Inter-clausal

Inner-clausal

CV 15th–16th cs. (229–279)

E

23 77% 7 23% 30 100%

18 82% 4 18% 22 73%

2 40% 3 60% 5 16%

3 100% 0

24 22% 84 78% 108 100%

22 22% 76 78% 98 91%

2 20% 8 80% 10 9%

0

22 45% 27 55% 49 100%

12 50% 12 50% 22 100%

10 40% 15 60% 25 100%

0

IAR Total conj.

Coresi, CÎ 1581 (2–33)

E IAR Total conj.

CS 1583–1614 (237–256 and 280–289)

E IAR Total conj.

3 1%

0 0

0 0

(32) ca s˘a învea¸te pre to¸ti trufiia s˘a urasc˘a, iar˘a so that teach: subj acc all haughtiness that hate: subj and smeareniia s˘a iubeasc˘a (CÎ: 14) humility that love: subj “so that he teaches everybody to hate haughtiness, and to love humility”.

It was precisely in this kind of contexts, where inter-sentential e had a presentative value, that iar˘a could easily replace the adversative e (see Table 4).

. Conclusions By comparing Old French and Old Romanian, we have been able not only to reconstruct some unattested links between Latin and Old Romanian, but also to show how different factors, among them phonetic change, language contact and pragmatic functions, can explain the divergent evolution of Lat. et and sic in these two languages. (i) In Old French, si was a discourse marker of topic continuity. Unlike et, si carried the conventional implicature of ‘succession (temporal or logical)’.

 Maria M. Manoliu

In modern French, si is no longer a discourse marker of continuity but has retained its pragmatic function carrying the implicature of denying a negative expectation expressed in the previous utterance. Its restricted role must have been the result of the competition of the homophonous si ‘if ’ and the spread of its synonymous competitors: alors ‘then’, puis ‘then, after’, ensuite ‘then, after that’ for expressing temporal succession or alors ‘then’, or, and car ‘since’, for signaling a logical relation of consequence. (ii) In a sequence of sentences, Old Rom. e (< Lat. et ‘and’) developed its presentative function of signaling text-coherence. It also occurs as a copulative conjunction connecting mainly clauses with different subjects and/or contrary predicates. Such contexts that could activate the conversational implicature that ‘b = non-a’, favored its reinterpretation as a weak adversative conjunction, carrying the corresponding conventional implicature. As such it became even less appropriate for fulfilling the function of a representative marker expressing close connection between events. Consequently, it was replaced by a variety of items that express each of its etymological functions: representative connective, topicality marker (Rom. conjunction ¸si), marker signaling the addition/ inclusion of an unexpected referent (Rom. adverbial ¸si), contrast marker (Rom. iar). The value of ‘equality’ carried by the adverbial sic developed into equality by ‘inclusion in the same superordinate unit’. This feature became the invariant semantic feature underlying the present functions of Rom. ¸si.

Notes . For the concepts of ‘conversational and conventional implicature’ see Levinson (1983, 2000) and, more recently, the survey of various positions concerning Grice’s ‘Principle of Cooperation’ in Jaszczolt (2002: 207–223). The distinction between pragmatic markers and discourse markers may be formulated as follows: pragmatic markers signal the speaker’s attitude toward the propositional content; discourse markers signal the organization of discourse units (text coherence, topic switch, turn switch, etc.). . As Robert Martin (1983) puts it, ‘expectation is a way of representing a possible world which has the best chances of realization according to the speaker’s universe of beliefs’. See also Manoliu (1994: 183). An updated view on the predicate logic and the propositional semantics may be found in Jaszczolt (2002), especially Chapter 6. . Although everybody seems to agree that, in contemporary spoken registers, puis ‘after, then’ has become a preferred marker of discourse continuity, a comparable example where et is repeated as a presentative marker may be found in Ball (2000: 160). It reproduces the

Vanishing discourse markers

speech of a ten-year-old girl describing how her friends tried to dry their clothes during a skiing holiday: Alors on a suspendu une ficelle. Et on étendait les anoraks et tout ça parce que elles les lavaient pas mais après le ski elles les . . . on les . . . on les suspendaient, elles . . . elles les étendaient, c’était la même chose dans la chambre à trois. Et puis au bout d’un moment elles mettaient un anorak deux trois, euh, des culottes, des chandails, des pulls, de tout; et puis à euh, à peu près à la vingtième chose, elle met la vingtième chose crac! (CR190). “Then we hang up a cord. And we hung on the anoraks and that not because they washed them but [because] after skiing they them. . . they hung them, they . . . they spread them, it was the same thing in our room for three people. And then after a short while they hung an anorak two three, hmm, pants, sweaters, pullovers, everything; and then when, hmm. . ., about the twentieth thing, she hangs the twentieth thing crack!” . Similar uses of si may be found in Northern Italian dialects (see Rohlfs 1969: 165–166). . My thanks to Martin Maiden who pointed out to me that e is still used in some SouthernDanubian Romanian dialects. . Examples of the fact that the value of ‘unexpected immediacy’ is already attested in Vulgar Latin (in contexts in which sic accompanies statim) may be found in Iliescu (1991). . Cf. Ar. iar, Mgl. ar˘a, ‘and, but, again’, similar to Eng. eir ‘, Pv. (e)ra, eras ‘also, and’, which are also related to a Greek form έρα (see Pu¸scariu, 1975, sv. 756). . For the changes undergone by the Latin reflexive pronoun sibi when becoming a marker of identity in Romanian see Manoliu (2003).

Sources Aeth: Silviae uel potius Aetheriae Peregrinatio ad loca sancta. Edited by W. Heraeus, 2nd ed. Heidelberg, 1921. Artu: La mort le Roi Artu, roman du 13e siècle. Edited by J. Frappier, Genève-Paris: Droz & Minard, 1964. CÎ: Diaconul Coresi, Carte cu înv˘a¸ta˘tur˘a (1581). Publicat˘a de Sextil Pu¸scariu s¸ i Alexie Procopovici, 1. Textul, Bucharest: Socec and Co., 1914. Creang˘a, PAP: Ion Creang˘a, Pove¸sti, Amintiri, Povestiri. Bucharest: Editura pentru literatur˘a, 1965. CS: Codex Sturdzanus (1583–1614). Studiu filologic, studiu lingvistic, edi¸tie de text s¸ i indice de cuvinte de Gheorghe Chivu. Bucharest: Editura Academiei Române, 1993. CV: Codicele Vorone¸tean (15th c.–16th c.). Edi¸tie critic˘a. Studiu filologic s¸ i studiu lingvistic de Mariana Costinescu. Bucharest: Minerva, 1981. Înv: Înv˘a¸ta˘turi preste toate zilele (1642), 1–2. Edited by W. van Eeden. Amsterdam: Rodopi, 1985. Plautus: Plaute, 1. Amphitrion, Asinaria, Aulularia. Texte établi et traduit par Alfred Ernout. Paris: Les Belles Lettres, 1932.



 Maria M. Manoliu

Roland: La Chanson de Roland. (1080). Publiée d’après le manuscrit d’Oxford (XIIe siècle.) et traduite par Joseph Bédier de l’Académie Française. Édition définitive. Paris: Édition d’art H. Piazza, 1947. Sârbu, Richard and Vasile Fr˘at¸il˘a, 1998. Dialectul istroromân. Texte ¸si glosar. Timi¸soara: Amarcord. Ureche: Grigore Ureche, Letopise¸tul T˘ ¸ arii Moldovei (16th c). Texte stabilite, studiu introductiv s¸i glosar de Liviu Onu. Bucharest: Editura Stiin¸ ¸ tific˘a, 1967.

References Ball, Rodney (1940/2000). Colloquial French Grammar. A Practical Guide (Blackwell Reference Grammars). Oxford: Blackwell. Buridant, Claude (2000). Grammaire nouvelle de l’ancien français. Saint-Juste-la Pendue: Sedes. Ferro, Teresa (1999). I caratteri ‘adanubiani’ del latino di Iordanes. In Fusco, Fabiana, Vincenzo Orioles, & Alice Parmeggiani (Eds.), Processi di Convergenza e differenziazione nelle lingue dell’Europa Medievale e Moderna/ Processes of Convergence and Differentiation in the Languages of Medieval and Modern Europe. Atti del Convegno Internazionale Udine, 9–11 dicembre 1999 (pp. 48–56). Udine: Forum. Fleischman, Suzanne (1992). Discourse and Diachrony: The Rise and Fall of Old French SI. In Gerristen, Marinel & Dieter Stein (Eds.), Internal and External Factors in Syntactic Change (pp. 433–473). Berlin: Mouton de Gruyter. Fraser, Bruce (1990). An approach to discourse markers. Journal of Pragmatics, 14, 383–395. Grevisse, Maurice (19861 /1991). Le bon usage. Grammaire française. 12e éd. Refondue par André Goose. Paris/Louvain-la-Neuve: Duculot. Iliescu, Maria (1991). Lat. sic – Roum. ¸si. Linguistica, 31 (1), 121–129. Paulo Tekavˇciˇc Sexagenario In Honorem Oblata. Liubliana. Iordan, Iorgu (1975). Stilistica limbii române. Edi¸tie definitiv˘a. Bucure¸sti: Editura Stiin¸ ¸ tific˘a. Jaszczolt, Katarzyna M. (2002). Semantics and Pragmatics. Meaning in Language and Discourse. London: Pearson Education/Longman. Kroon, Caroline Henrica Maria (1995). Discourse particles in Latin: a study of NAM, ENIM, AUTEM, VERO, and AT (Amsterdam studies in classical philology, 4). Amsterdam: J.C. Gieben. Levinson, Stephen C. (1983). Pragmatics. Cambridge: Cambridge University Press. Manoliu-Manea, Maria (1994). Discourse and Pragmatic Constraints on Grammatical Choices. A Grammar of Surprises (North Holland Linguistic Series 57). Amsterdam/ Lausanne/ New York/ Oxford/ Shannon/ Tokyo: Elsevier. Manoliu-Manea, Maria (2003). Conventional Implicature and Language Change: The cyclic evolution of the emphatic pronouns in Romanian.” In R. Blake (Ed.), Historical Linguistics 2001). Selected papers from the 15th International Conference on Historical Linguistics (pp. 301–318). Amsterdam: John Benjamins. Marchello-Nizia, Christiane (1985). Dire le vrai: L’adverbe ‘si’ en français médiéval (Publications romanes et française, 168). Genève: Droz.

Vanishing discourse markers 

Martin, Robert (1983). Pour une logique du sens. Paris: P.U.F. Moignet, Gérard (1979). Grammaire de l’ancien français. Paris: Klincksieck. Pu¸scariu, Sextil (1905)19752. Etymologisches Wörterbuch der rumänischen Sprache: Lateinisches Element mit Berücksichtigung aller romanischen Sprachen. Heidelberg: C. Winter. Reenen, Pieter van, & Lene Schøsler. (1992). Ancien et moyen français. Si thématique. Analyse exhaustive d’une série de textes. Vox Rom, 51, 101–129. Reenen, Pieter van, & Lene Schøsler (2000). The Pragmatic Function of the Old French Particles AINZ, AORES, DONC, LORS, OR, PUIS, and SI. In Susan C. Herring, Pieter van Reenen, & Lene Schøsler (Eds.), Textual Parameters in Older Languages (Current Issues in Linguistic Theory) (pp. 59–105). Amsterdam: John Benjamins. Rossari, Corinne (2000). Connecteurs et relations de discours: des liens entre cognition et signification (Langage-Cognition-Interaction). Nancy: Presses Universitaires de Nancy. Thomasset, Claude & Karin Ueltschi, 1993. Pour lire l’ancien français. Paris: Nathan. Traugott, Elizabeth Closs, & Ekkehard König (1991). The semantics-pragmatics of grammaticalization revisited. In Traugott, Elizabeth Closs, & Bernard Heine (Eds.), Approaches to Grammaticalization, 1. Focus on Theoretical and Methodological Issues (pp. 189–218). Amsterdam/Philadelphia: John Benjamins. Wilmet, Marc (1997). Grammaire critique du français (Hachette Supérieur). Paris: Duculot.

From ditransitive to monotransitive structure in the history of the Spanish language. Reanalysis of objects A case of incorporation and monotransitivization Rosa María Ortiz Ciscomani University of Sonora

.

The phenomenon1

The number of studies about the ditransitive construction, despite its typological importance, is not great. Consequently, diachronic knowledge of this kind of structure in the world’s languages is scarce. My concern in this paper is with ditransitive Spanish structures, that is, constructions whose verbs take two objects, a DO and an IO, with the preposition a, and sometimes the dative clitic le;2 this is a subject about which there is little previous work.3 I am dealing with ditransitivity as an scalar category in the sense of Hopper and Thompson (1980), as a continuum with zones associated to the object noun phrases features, being crucial the one which I call ‘asymmetry’ or lexical contrast: DO-inanimate, inactive/IO-animate, active. The asymmetry is characteristic of the strong or prototypical structure; the ‘symmetry’, or identical lexical features, is characteristic of the reanalysed or weak ditransitive structure or monotransitive structure. The analysis shows the weak ditransitive structure as a grammaticalised monovalent structure which has been diachronically stable but which has increased in nineteenth and twentieth centuries. Structures like those in (1) can be found throughout the history of Spanish. (1a) with the verb dar ‘to give’ is an example of typical ditransitive structure; (1b) and (1c) with the same verb, are less typical:

 Rosa María Ortiz Ciscomani

(1) a. E dio el vaso al rey que gelo diese el (GEII, 87b.6) and gave the glass-do to the king-io that him io-it od had given he “He gave the glass to the king who had given it to him” b. Y desque nos vimos salvos de aquellas refriegas, And since ourselves saw 1pl safe from those brawls, “And since we found ourselves safe from those brawls, dimos muchas gracias a Dios por ello (HistVerd, 6) gave 1pl much thanks-do to God-io for it we thanked God profusely for that” c. Y no me refiero aquí al hecho de que el tráfico continuó cuando And neg 1sg refer1sg here to the fact that traffic continued when “And here I’m not referring to the fact that the traffic continued when ya estaba proscrita; lo que dio lugar a mayores it had already was proscribed; which gave place-do to greater atrocidades (Noticias, 227) atrocities-io already been banned; which gave rise to greater atrocities”

The sentences in (1), as well as others I will deal with in this paper, show the same verb (in italics), the verb dar ‘to give’ – dio [past 3S (a)]: dimos [past 1PL (b)]; dio [past 3S (c)] –, and two noun phrases, DO (in bold) and IO (bold and italics). In contrast to the strong ditransitivity shown by the sentence in (a) with a concrete DO, el vaso, ‘the glass’, and a human IO al rey, ‘to the king’, those in (b) and (c) are weaker ditransitives: they are examples of sentences that have undergone semantic and syntactic changes, not always to the same degree, so that some of them now function as fixed phrases. In this paper I intend to outline the properties of the strong or prototypical Spanish ditransitive, and to give an account of the weak or reanalysed Spanish ditransitive;4 assuming that the ditransitive is a construction, in the sense of Goldberg (1995), and that constructional meanings exist independently of particular verb meanings. Likewise I intend to show that weak Spanish ditransitives are characterized mainly by the loss of the object’s asymmetry characteristic of the prototypical construction, that is, the loss of the lexical semantic contrast which conveys a clear default distinction between the DO participant, an inanimate inactive entity, and the IO participant –a person virtually active–;5 that they involve semantic change, that is, lexical meaning impoverishment or bleaching of both the verb and the DO noun, and that they also involve grammatical change, that is, the DO noun has more syntactic constraints, in modification as well as in word order, and the DO incorporates into the verb. These behaviour features have an impact on the construction itself,

Ditransitive to monotransitive structure in Spanish

which goes from a ditransitive structure with three arguments or slots [(S)-VDO-IO] to a monotransitive structure with only two slots [(S)-vo-O], in which both DO and IO are reanalysed, now the verb and the DO being a single functional and semantic complex unit and the IO remaining as the single object. The overall account depicts the weak ditransitive as reanalysed construction which becomes grammaticalized (Heine et al. 1991) as its frequency of use increases (Haiman 1994; Haspelmath 1999; Hopper & Thompson 1985; Keller 1994; Lehmann 1982). These conclusions are supported by a large corpus which takes into account twenty-two texts, from eight centuries of the history of the Spanish language, from the 13th to the 20th century,6 a long period not covered by analyses for other languages with respect to the structures which I focus on in this paper. The data I deal with in this presentation belong to the 13th, 16th, 19th and 20th centuries, which are relevant periods for the point at issue [cf. corpus (a) in the Bibliography]. The paper is organized as follows: in Section 2, I present the aim of this presentation; in Section 3, I analyse the Spanish low ditransitive construction, and I show the behaviour which proves the reanalysis of the ditransitive as well as the monotransitive construction, and their grammaticalization; and finally in Section 4, I present my findings as a summary.

. The aim The aim is twofold: First, I offer an analysis of the weak ditransitives sentences as those in (1b) and (1c) as a type of incorporation,7 with diachronic stability and some degree of compositionality which results in monotransitivization. I am addressing only those Spanish ditransitive constructions with both objects, direct (DO) and indirect (IO), expressed by full nominal phrases, that move away from typical ditransitivity. These structures have a DO with a zero mark and an IO marked with a, that is, the syntactic pattern is: (Subject)Verb-DO-IO (1a). This is the prototypical pattern too, but in contrast to the prototypical or strong ditransitive structures, the weak ditransitive structures combine symmetrical8 noun phrases, that is, nouns with similar lexical semantic features, without clear semantic distinction between the DO and the IO and their background; this symmetry, conveys some semantic complexity to the construction that rejects simple analysis. The overall analysis involves some particular changes:



 Rosa María Ortiz Ciscomani

i.

from concreteness [sentences in (1a)] to abstractness [sentences in (1b) and (1c)] ii. decategorization of DO nouns which lose syntactic and conceptual or semantic freedom iii. reanalysis, both of the IO as DO, and of the DO so that DO becomes incorporated into the verb, shaping a composite predicate, a verbo-nominal structure Secondly, I intend to present the features that characterize these low ditransitive constructions, their diachronic stability and, at the same time, the increase they present through time, especially in the 19th and 20th centuries. I also intend to show that they involve a process which has the effect of redefining constituent boundaries, a kind of reanalysis of the construction, via abstractness, fixation and routinization of forms.9 As we will see later on, constructions with weak ditransitivity involve rigidification of word order, the morphosyntactic behaviour of verb and DO nouns is constrained, and they are semantically impoverished or bleached (Traugott & Heine 1991: 5; Haspelmath 1998: 52, 78); these are characteristic features of grammaticalization.

. The analysis The analysis has two main angles: One concerns objects; the other concerns verbs. . Objects: The evidence Objects in low ditransitivity constructions show some properties, both lexical and morphosyntactic. In Table 1 we can see the properties involved, and their coding both in the prototypical structure (middle column), and the weak ditransitive structure (right column). Two main lexical changes are involved: one has to do with the object’s asymmetry, the other with DO abstractness. Besides lexical constraints, a construction with low ditransitivity has some syntactic constraints. .. Object’s asymmetry: From asymmetrical objects to symmetrical objects One of the main properties of a typical ditransitive construction is that it combines objects which are chiefly asymmetric. In fact in a prototypical Span-

Ditransitive to monotransitive structure in Spanish 

Table 1. Changes from Prototypical to weak ditransitive Properties

Prototypical ditransitive

Weak ditransitive-monotransitive

Object’s asymmetry

+ OD inanimate/IO human

DO noun DO Modification DO Expansion DO Number DO Quantification DO Referentiality Order

Concrete + + Flexible + + Free

– DO inanimate/IO flexible (animate/inanimate) Abstract Restricted – Frozen – Indefinite –non referential DO Rigid or fixed

ish ditransitive, a transference verb like give co-occurs with two objects, an animate-human IO + an inanimate concrete DO (2): (2) E dixo luego donna Juno: “Amigas, dad a mí esta manzana” (GEII, 91b.47) And said later lady Juno: “Friends, give to me-io that apple-do” “And Lady Juno later said: “Friends, give me that apple””

In contrast, in constructions with low or weak ditransitivity, the IO is lexically flexible, that is, its place may be filled by an inanimate noun. This feature makes an IO to resemble a DO, approaches the IO to the DO, and dissolves the asymmetry of prototypical or strong ditransitivity. See IO a las espadas ‘to the swords’ in (3), inanimate, as is DO mano ‘hand’, versus asymmetrical concurrent objects in prototypical ditransitives as in the sentence (2) above. (3) e metieron mano a las espadas e dieronse grandes golpes con ellas and get hand-do to the swords-io and gave themselves big strokes “and they took up the swords and gave each other great blows with them sobre los escudos (GEII, 153a.40) on the shields on their shields”

.. DO: From concrete to abstract The prototypical construction’s DO is a concrete noun as in (4): (4) Bien dicen que en Zaragoza y en el ejército del Centro se dieron los Well say that in Zaragoza and in the army of the Center 3refl give the “It can well be said that in Zaragoza and in the Central Army, ranks were

 Rosa María Ortiz Ciscomani

grados como quien echa almorzadas de trigo a las gallinas (Gerona, 5) ranks as who throw double handful of wheat to the chicken given out as one would throw brimming handfuls of wheat to the chickens”

In contrast to the prototypical ditransitive, in the weak ditransitive Spanish construction the DO is usually not a concrete entity but an abstract noun. Compare the abstract DO un vistazo ‘a glance’ in the weak ditransitive in (5) with the DO noun almorzadas de trigo ‘double handfuls of wheat’ in the prototypical ditransitive in (4), both with the same verb echar ‘to throw’: (5) Puedo echarlei un vistazo a tus cuadrosi ? Le pregunté (Tempestad, 28) Can (I) give-3sdat a glance-do to your pictures-io? 3sdat asked (I), “Can I take a look at your pictures? I asked him”

.. DO Modification and expansion: From unrestricted to restricted The DO in prototypical ditransitives permits modification and expansion, as we can see in (6) (the DO noun phrase in bold): (6) el ministro a su vez comunicó las intenciones del emperador de los the minister at his turn told the intentions of the emperor of the French-do “The minister in turn told the Spanish premier Calderón Collantes the franceses al premier español Calderón Collantes, quien a su vez to the premier Spanish Calderón Collantes and Calderón Collantes in turn intentions of the French emperor, who at his turn le escribió a su embajador en Londres (Noticias, 90) 3Sgdat wrote to his ambassador in London-oi wrote to his ambassador in London”

In contrast, the concurrent DO in a weak ditransitive structure is generally a ‘bare’ noun phrase, that is, a noun phrase without determiners nor expansion. See mano ‘hand’ in (7a) and largas ‘delay’ in (7b): eché mano a la navajai (Cachito, 113) yo lei I it-dat put hand-do to the switchable knife-io “I took up the switchblade” b. Así que consciente de que eso era autosentenciarme a So that conscious of that that was to autocondemn-1sdo to “And so, conscious that that was the same as condemning myself to

(7) a.

Ditransitive to monotransitive structure in Spanish 

dos años de galeras, lei daba largas a mi editor i (Cachito, 113) two years of galleys 3sdat give delay-do to my editor-io two years in the galleys, I kept putting things off to my editor”

.. DO number: From flexible to frozen In a prototypical ditransitive construction, the DO is flexible concerning number, that is, it can appear in the singular or in the plural as in (8), unos cuantos consejos, with the verb dar ‘to give’ with its normal, inherent, ‘transference’ meaning: (8) Lamparilla se despidió de su comadre, dio unos cuantos consejos Lamparilla said goodbye to his ‘comadre’, gave a few advices-do “Lamparilla said goodbye to his ‘comadre’, gave a few words of advice a Juan y ...emprendió a galope el viaje para la capital (Bandidos, 219) to Juan-io and . . . began at a gallop the journey to the capital to Juan and, at a gallop, began his journey to the capital”

The DO noun in a low ditransitivity construction, in contrast, is a frozen form with regard to number; it appears only in the singular, or only in the plural; that is, it does not have inflectional flexibility, and it loses its independence from the verb. In a weak Spanish ditransitive the DO noun usually occurs in the singular, such as lugar in (9), in which the same verb dar ‘to give’ (cf. examples (8) and (2)) combined with a singular, abstract DO noun loses its ‘transfer’ sense: (9) Y no me refiero aquí al hecho de que el tráfico continuó And neg 1sg refer1sg here to the fact that the traffic continued “Here I’m not talking about the fact that the traffic continued even cuando ya estaba proscrita; lo que dio lugar a mayores when already was proscribed; which gave place-do to greater when it had already been banned; which gave rise to greater atrocidades (Noticias, 227) atrocities-io atrocities”

.. DO quantification: From flexible to inflexible In contrast to the prototypical ditransitive structure which admits quantifiers, like dos pesos ‘two pesos’ in (10): nomás tei doy (10) A tii dos pesos carajo (Suerte, 121) To you-io only 2soi give two pesos-do damn “To you I only give two pesos damn it”

 Rosa María Ortiz Ciscomani

in a weak Spanish ditransitive structure a DO usually appears without quantifiers; this type of construction shows some syntactic heterogeneity in this respect (compare mano (7a) versus una ojeada in (11), but in both cases we have semantically indefinite noun phrases. The quantification inflexibility of weak ditransitive structure goes parallel to a process of increasing cohesion and abstraction. (11) Y basta echarlei una ojeada And suffice give3sdat a quick glimpse-do “And it suffices to take a quick glimpse a la historia de la esclavitudi to the history of the slavery-io, at the history of slavery para enterarse, con horror, que las tribus de Africa Occidental se pasaban for be realize, with horror, that the tribes of Africa western passed in order to realize, with horror, that the tribes of western Africa spent la vida luchando entre sí para hacerse prisioneros de guerra (Noticias, 226) the life fighting among themselves for be made prisoners of war their lives fighting among themselves to take prisoners of war”

In (12) the DO noun appears with a quantified adjective phrase. This modifier diminishes the cohesion with the verb and the construction moves away from being a weak ditransitive construction such as the ones exemplified in (7a) and (7b) above. (12) E daré muy grand consejo a todo su fecho (GRII, 64b.19) And (I) will give very great advice-od to all her deed-oi “And I will give very great advice to all her deeds”

.. Referentiality: From referential to non-referential DO As we have seen, in the prototypical Spanish construction the DO is an inanimate referential noun, modified and/or expanded as in sentences (2), (4), (6), and (8). In a weak ditransitive, as we have shown above, a DO noun does not usually accept either modification nor expansion which would attribute some determinacy to it. In a general sense, these nouns are non-countable nouns of scarce or no referentiality (Mithun 1986: 34). When a DO appears with a modifier or expansion, the verb regains some independence. Compare (13a), with the verb

Ditransitive to monotransitive structure in Spanish 

facer ‘to make’ with a bare DO noun, to (13b), with the same verb but with a DO noun modified by possessive sus: (13) a.

fazed serviçio a vuestro padre (GEII, 79a.16) Make service-od to your father-oi “Serve your father” b. Pues que ovo fecha su avenencia el rey Minos con los After that have made his agreeing the king Minos with the (people) “After having made an agreement with the people of Athens, King de Athenas, fizo sus sacrificios a Júpiter por quanta from Athenas, made his sacrifices-od to Jupiter-oi for all Minos made his sacrifices to Jupiter in thanks for all bien andança le avie dada (GE, 415b.16) good fortune 3Sdat had given the good fortune he had been given”

.. Order: From free to fixed order In addition to the former features, the prototypical and the weak ditransitive construction contrast in respect to order: the typical ditransitive structures are free structures, that is, they have highly flexible constituent order, as opposed to the weak ditransitive structures, whose order is rigid, frozen: if a constituent moves, the structure becomes agrammatical [*] (14). (14) yo lei eché mano a la navajai (Cachito, 113) [*eché a la navaja mano] I it-dat put hand-do to the switchable knife-io “I took up the switchblade lo que dio lugar a mayores atrocidades (Noticias, 227) [*dio a mayores atrocidades lugar] which gave place-do to greater atrocities-io been banned; which gave rise to greater atrocities”

In the typical ditransitives the objects may appear in the order DO-IO (15a) or IO-DO (15b), adjacent (15a) or not adjacent to each other (15b): (15) a. E en pos esto fincando yo muy espantada, demande consejo a And in pursuit of this I very frightened, (I) asked advice-do to “And in pursuit of this I, being very frightened, sought advice from las buenas du[ennas] ancianas e a los viejos mayores de edad (GEII, 120b.12) the benevolent ladies old and to the old men elder in age-io the benevolent old ladies and the older men”

 Rosa María Ortiz Ciscomani

daba joyas, a otros b. y era tan bueno que a todosi nosi and was so good that to all-io us-dat gave jewels-do, to others “And he was so good that he gave us all jewels, and to others mantas e indias hermosas (Hist Verd, 282) blankets and indians beautiful blankets and beautiful Indian girls”

In sum, in low ditransitivity constructions, the objects tend to appear in a fixed order V-DO-IO, with the DO next to the verb; that is, the DO follows the verb and stands before the IO. This DO is an inflexible form with respect to modification, expansion, and quantification; that is, it is a frozen form. In other words, there is syntactic rigidization. Now we have a V and a DO unit so semantically and syntactically close, that we have a DO incorporated into the verb, [V-DO] (Mithun 1984: 850). This syntactic proximity between the DO ant the verb is iconic; that is, its syntactic cohesion mirrors its semantic cohesion (Kirsner 1985). Certainly in sentences like those in (7a) and (11), it can be seen that the DOs are semantically impoverished lexical items; they cannot be understood alone, but only in constructions with those verbs. Cohesiveness, nevertheless, is a matter of degree [cf. (7a) eché mano versus (11) echarle una ojeada and (7b) daba largas versus (12) daré muy grand consejo]. The above mentioned DO nouns’ features are consistent with well-known properties of grammaticalization; including at least, attrition or gradual loss of semantic substance, coalescence or collapsing together of adjacent forms, fixation of linear order (Lehmann 1982: 307) and de-categorialization, that is, loss of morphological markers and syntactic privileges characteristic of the full category (DO noun in this case) (Hopper 1991: 21). . Verbs With respect to the verbs, a restricted series of them appear in weak constructions. The verbs which recur during the eight centuries studied, are: dar ‘to give’, hacer ‘to make’, poner ‘to put’, and echar ‘to throw’; that is, they are verbs of general active meaning, common verbs with a general meaning [Quirk et al., 1985: 750) or ‘light verbs’ [Jespersen 1984 (1921): 138–139]. In the weak ditransitive construction the verbs, also present in prototypical constructions, combine with abstract DO nouns of the kind mentioned in the previous section. Both the verb and the DO noun, in the weak ditransitive show semantic bleaching, loss of the verbal independent lexical meaning, and

Ditransitive to monotransitive structure in Spanish 

loss of the objects referential meaning, so that they both may be understood as a combination V-DO + IO. In (17) we have sentences with each of the mentioned recurrent verbs, dar ‘to give’, hacer ‘to make’, poner ‘to put’, and echar ‘to throw’, combined with a DO abstract noun plus an IO which is sometimes an animate and sometimes an inanimate noun. (16) a.

Porque si arremetían o daban alcance a los escuadrones que Because if assailed3pl or gave3pl reach-od to the squadrons which “Because if they attacked or caught up with the squadrons that were con nosotros peleaban, luego se les arrojaban en el agua (Hist Verd, 506) with us struggled-oi, after themselves to them threw in the water fighting against us, then they threw themselves onto them in the water” b. Separó con la mano al que hacía frente a Pancha Put aside with the hand the one who made front-od to Pancha-oi “He pushed aside with his hand the one who was facing up to Pancha y continuó bailando y taconeando (Bandidos, 92) and continued dancing and tap dancing and continued dancing and tapping his feet” c. yo porne fuego a las sus reales casas, e fare que I will put fire-od to the their royal homes-oi, and (I) will make that “And or I will set fire to their royal homes, and I will make Thereo, Thereo, el fazedor de tanta falsedat, arda en ellas (GEI, 256a.37) Thereo, the maker of so much untruth, burn in them the inventor of so many falsehoods, burn in them” d. “¿Dónde has estado todo el día?” – preguntó Siseta echando Where have (you) been all the day? – asked Siseta laying mano hand-od “Where have you been all day?” – asked Siseta, laying her hand a Badoret y deteniéndole (Gerona, 33) to Badoret-oi and stopping-3Sgdat=acc on Badoret and stopping him”

Other verbs have occurred in these constructions when combined with abstract, bare, non referential DO nouns. Table 2 registers the thirteen nonrecurrent verbs and the century in which they were documented. Table 2 shows that, in comparison to the previous centuries, the number of verbs in this kind of construction increases in the 20th century.

 Rosa María Ortiz Ciscomani

Table 2. Non-recurrent verbs in low ditransitive constructions Verb

13th

meter ‘to put in’ parar ‘to prepare’ pedir ‘to ask’ tomar ‘to take’ ajustar ‘to adjust’ llevar ‘to carry’ prestar ‘to lend’ pasar ‘to pass’ declarar ‘to declare’ dejar ‘to permit’ romper ‘to break’ tirar ‘to throw’ ver ‘to see’ Total

3

16th

19th

20th

1

2

8

In contrast to the low frequency of verb types admitted in these constructions, the frequency of nominal types in these constructions is high and has increased during the last two centuries. See Table 3, with constructions of the verb dar ‘to give’ and DO nouns documented in the 13th, 16th, 19th and 20th centuries. The data analysed show that diachronically these kinds of structure have been present throughout the history of Spanish since the 13th century. Most of the nouns in these constructions have a short life cycle, but the construction pattern or model is diachronically stable and still productive as Schøsler (2003: 410–411) points out about support verb constructions in French; variation in frequency of tokens depends on the thematic nature of texts (Schøsler 2003: 408). In Table 4 we can see the increase of the constructions with a verb combined with an abstract DO noun and an inanimate IO, that is, constructions with lexically symmetric object nouns, across the time. Some verbs in the low ditransitive constructions (as verb to give in Table) 3 are the same verbs present in typical ditransitive constructions; but in the low ditransitive construction both the verb and the abstract DO itself have lost their normal inherent meaning not always at the same degree, so that the degree of semantic cohesiveness differs as well. In this respect, constructions can be seen as located in a gradient continuum of semantic compositionality. Cohesiveness results in expressions whose meaning is not predictable from the meanings of its elements.

Ditransitive to monotransitive structure in Spanish

Table 3. DO Nouns combined with the verb dar ‘to give’ DO noun

13th

16 th

19th

20th

gracias ‘thanks’ batalla ‘battle’ conseio ‘advise’ respuesta ‘answer’ ayuda ‘help’ fuego ‘fire’ beneficios ‘benefit’ sepulturas ‘burial’ muerte ‘death’ nombre ‘name’ lugar ‘place’ la vuelta ‘turn’ fin ‘end’ vista ‘sight’ guerra ‘war’ garrote ‘stick’ alcance ‘reach’ caza ‘hunt’ causa ‘cause’ oreja/oídos ‘ears’ cuenta ‘account’ voces ‘voices’ orden ‘order’ ocasión ‘occasion’ crédito ‘credit’ parte ‘part’ gusto ‘delight’ salida ‘exit’ término ‘finish’ comienzo ‘beginning’ estocadas ‘thrust’ un vistazo ‘a glance’ entrada ‘entrance’ lectura ‘reading’ vuelo ‘flight’ pábulo ‘food’ fuerza ‘power’ paso’step’ rienda suelta ‘free rein’ inicio ‘start’ tiempo ‘time’ puerta ‘door’ pie ‘foot’ mano libre ‘free hand’ cabida ‘room’ largas ‘delay’ TOTAL10

1 1 10 1 1 1 1 1 1 2

45

8

3

1 1 1 12 1 1 4 2 1

1 2 1

2 3 3 3

1 4

1 2 2 3 3 5 1 1 1 1 2 1 1 1 1 1 1 2

11/20

11/70

21/44

2 1 1 1 1 1 1 1 1 15/20



 Rosa María Ortiz Ciscomani

Table 4. Constructions with lexically symmetric objects: V-abstract DO + inanimate IO 13th 16th 19th 20th

13% 4% 18% 23%

(40/316) (21/567) (67/381) (66/282)

They involve an abstraction process, or a metaphor process in principle context-dependent, contingent, and pragmatic. It is the context, purpose or point of view, which determines such behaviour. Some contexts are more general, frequent, obvious, and thus, “stable” (Givón 1984: 18); some of these constructions are preferred sequences of V-DO noun + IO, with such a high frequency of tokens that they become formulae, a routinized construction, identical in form to typical constructions but whose whole meaning cannot be interpreted or obtained by working out the properties of their components. In other words, when a construction increases in frequency of use, it acquires some idiomatic value; high cohesion concurrent with high frequency of use makes idioms arise (17). As consequence of habituation through frequency of use or routinization there is some loss of pragmatic salience of the DO noun. In (17) we have two of these routinized sentences, the first with a recurrent verb dar ‘give’, and the second with a non-recurrent verb parar ‘to prepare’: (17) a.

Así que consciente de que eso era autosentenciarme a dos años So that conscious of that that was to autocondemn-1sdo to two years “And so, conscious that that was the same as condemning myself to two years a mi editor i (Cachito, 113) de galeras, lei daba largas of galleys 3sdat give delay-do to my editor-io in the galleys, I kept putting things off to my editor” b. e parat mientes a mí e como veriedes que yo fago, And prepare minds-do to me-io and as (you) will see that I do, “And prepare your minds as you see I do, fazet vos assi (GEI, 314a.6) do you alike and do as I do”

Thus in these constructions both the verb and the DO have lost their lexical freedom, resulting in a reorganization of constituent boundaries. They have undergone a kind of reanalysis like the so called constituent-internal reanalysis (Heine and Reh 1984: 110), i.e. it is one specific form of the more general pro-

Ditransitive to monotransitive structure in Spanish 

Figure 1. From ditransitive to monotransitive construction

cess of reanalysis which has the effect of redefining constituent boundaries in a construction. The DO is now incorporated into the verb; it functions as a so-called composite predicate, a complex verb or verbo-nominal phrase. The semantic and syntactic cohesion V-DO causes the DO to be no longer a DO; so the construction with three slots, V-DO-IO, is now reanalysed as a two slot construction, V-Noun + Noun. The DO slot is now filled by the previous IO Noun which is reanalysed as the first object, gaining topicality and relevance. Grammatical relations have changed also, the ditransitive construction is now a monotransitive construction (Figure 1).

. Summary I have attempted to show that low ditransitive constructions exist in the history of Spanish and to describe the form and frequency of this kind of ditransitive structure, which has been relatively neglected in cross-linguistic studies as well. I have shown that as opposed to typical ditransitive constructions, with relatively free word order, diverse verb types and chiefly asymmetric objects, the constructions I have accounted for are characterized by rigidification of word order, and by some lexical, semantic and morphosyntactic constraints, mainly on the DO noun. The DOs in these constructions are abstract, non-referential, indefinite nouns, syntactically restricted with regards to modifiers and expansion, as well as in their number and quantifiers, combined with verbs of general meaning, or ‘light’ verbs, and with IOs which are lexically flexible but with increasing inanimacy. I have shown that the lexical meanings of both verb and DO are bleached. Furthermore, I have shown that they have lost their semantic and syntactic freedom, to different degrees, thus making their cohesiveness a matter of degree and giving rise to a continuum. Finally, the features of fixed order, semantic bleaching and syntactic restrictions on both the verb and the DO, have syntactic effects: a reanalysis

 Rosa María Ortiz Ciscomani

of constituent boundaries through object incorporation, that results in decategorialization of both the verb and the DO, and a change in the nature of the construction from ditransitive to monotransitive. I conclude that these changes constitute instances of grammaticalization and idiomatization, driven by frequency of use. Grammar shapes use, discourse, and use, in turn, shapes grammar (Lichtenberk 1991: 76). These kinds of structure are found throughout all periods of the history of the Spanish language, and have shown some increase in frequency of both types and tokens.

Notes . I am indebted to Concepción Company for her careful and insightful comments on this paper. I am also very grateful to Lene Schøsler for her valuable suggestions, and to the anonymous referees for helpful comments that profit this paper. Of course, responsability for remaining faults is mine. This research was sponsored by University of Sonora and SEP- PROMEP. . Company (2001) considers this dative clitic as an agreement object marking. . Spanish ditransitivity is a realm only recently studied systematically in either a synchronic or diachronic perspective (Ortiz Ciscomani 2002, 2003). Recent works, interested mainly in the dative, take this kind of structure into account. Company (2002a, 2002b, and 2003) provides information, both diachronic and typological, about the dative in these structures as an index of the status of Spanish as a primary object language. Delbecque and Lamiroy (1996) make reference to the dative of these structures in a typological treatment. In synchronic perspective see Cano (1981) whose study refers to the Spanish ditransitive as a transitive structure, and Gutiérrez Ordóñez (1999), who is interested mainly in the dative. . ‘Strong’ and ‘weak’ ditransitivity are the equivalent of “high” and “low” transitivity in Hopper and Thompson (1980). The prototypical ditransitive is conceptualised as a causal chain (Croft 1991) with individuals acting on individuals, with some notion of transmission of force which imposes a ordering of participants in the causal event, so that subject acts upon an inanimate referential DO to move it or to ‘transfer’ it to an animate and referential IO. . This is one of the properties of individuation, a component of scalar transitivity (Hopper & Thompson 1980; Timberlake 1977). . Traugott (1999: 239) affirms that “only brief glimpses at a large scale historical perspective over the thirteen hundred years of the textual history of English have been possible”, related to the kind of structures I analyse here. Recently, Schøsler (2003) deals with support verb constructions in French, based on diachronic corpora covering the Old French and Middle French, as an ongoing process of grammaticalization. . Incorporation is used here in the sense of Mithun (1986), a process by which “a N stem is compounded with V stem to yield a larger, derived V stem”, with a unitary concept, and with constituents with different degree of cohesion (p. 847–849); see also Mithun (1984).

Ditransitive to monotransitive structure in Spanish  . Opposed to asymmetrical or clearly distinct or differentiated participants, symmetrical refers to absence of that clear distinction. . The term syntactic reanalysis refers to “syntactic/semantic reformulation”, that is, a change in the structure of an expression that does not involve any immediate or intrinsic modification of its surface manifestation (Langacker 1977: 59, 79). Heine and Reh (1984: 95ff.) propose to separate reanalysis from grammaticalization, essentially because of the unidirectionality principle, which has been assumed as an inherent property of the latter but not necessarily of the former. Haspelmath (1998: 318) also considers them as two distinct kinds of syntactic change. In this work I keep them separate but I am not going to discuss the theoretical point. . The number in bold corresponds to total of noun types; the other to total of tokens.

Bibliography a) Corpus GEI = Alfonso X El Sabio, General estoria, Second part, I, edition by Antonio G. Solalinde, L Loyd A. Kasten and Víctor R. B. Oelschläger, Consejo Superior de Investigaciones Científicas, Instituto “Miguel de Cervantes”, Madrid, MCMLVII. GEII = Alfonso X El Sabio, General estoria, Second part, II, edition by Antonio G. Solalinde, LLoyd A. Kasten and Víctor R. B. Oelschläger, Consejo Superior de Investigaciones Científicas, Instituto “Miguel de Cervantes”, Madrid, MCMLXI. Hist Verd = Bernal Díaz del Castillo, Historia verdadera de la Conquista de la Nueva España, Edition, index and prologue by Carmelo Sáenz de Santa María, Alianza, Madrid, 1991. Bandidos = Manuel Payno, Los bandidos de Río Frío, Porrúa, México, 1997. Gerona = Benito Pérez Galdós, Gerona, in Episodios nacionales, Gerona. Cádiz, Porrúa, México, 1960, 77–107. Pepita= Juan Valera, Pepita Jiménez, REI, México, 1990. Obras IV = Manuel Gutiérrez Nájera, Obras IV/Crónicas y artículos sobre teatro, II (1881– 1882), Ed. Yolanda Bache Cortés and Ana Elena Díaz Alejo, UNAM, México, 1984. Obras VII = Manuel Gutiérrez Nájera, Obras VII/Crónicas y artículos sobre teatro, V (1890– 1892), Ed. Ana Elena Díaz Alejo and Elvira López Aparicio, UNAM, México, 1990. Suerte = Luis Miguel Aguilar, Suerte con las mujeres, Cal y Arena, México, 1992. Noticias = Fernando del Paso, Noticias del Imperio, 2◦ edition. ninth reprint, Diana, México, 1997. Noche = Juan Villoro, Noche navegable, second edition, 1st. reprint, Joaquín Mortiz, México, 1995. Tempestad = Juan Manuel de Prada, La tempestad, Planeta, México, 1998. Maestro = Arturo Pérez Reverte, El maestro de esgrima, Alfaguara, Madrid, 1998. Cachito = Arturo Pérez Reverte, Cachito (Un asunto de honor), Alfaguara, Madrid, 1996.

 Rosa María Ortiz Ciscomani

b) References Cano Aguilar, R. (1981). Estructuras transitivas en el español actual. Madrid: Gredos. Company, Concepción (2002a). “Reanálisis en cadena y gramaticalización. Dativos problemáticos en la historia del español”. Verba, 29, 31–69. — — — (2002b). “Grammaticalization and Category Weakness”. In G. Diewald & I. Wischer (Eds.), New Reflections on grammaticalization (pp. 201–216). Amsterdam: John Benjamins. — — — (2001). “Multiple Dative Marking Grammaticalization. Spanish as a Special Kind of Primary Object Language”. Studies in Language, 25 (1), 1–47. — — — (2003). “Transitivity and grammaticalization of objects. The diachrony struggle of direct object and indirect object in Spanish”. In G. Fiorentino (Ed.), Transitivity in Romance Languages. Semantics and Pragmatics (pp. 217–260). Berlin-New York: Mouton de Gruyter. Croft, William (1991). Syntactic Categories and Grammatical Relations. Chicago and London: The University of Chicago Press. Delbecque, N., & B. Lamiroy (1996). “Towards a typology of the Spanish dative”. In W. Van Belle & W. Van Langendonck (Eds.), The dative, Vol. 1: Descriptive studies (pp. 73–117). Amsterdam: John Benjamins. Givón, Talmy (1984). Syntax. A Functional-typological Introduction. Vol I. Amsterdam: John Benjamins. Goldberg, Adele E. (1995). A Construction Grammar Approach to Argument Structure. Chicago and London: The University of Chicago Press. Haiman, John (1994). “Ritualization and the Development of Language”. In William Pagliuca (Ed.), Perspectives on Grammaticalization (pp. 3–28). Amsterdam/Philadelphia: John Benjamins. Haspelmath, Martin (1998). “Does grammaticalization need reanalysis?” Language, 22 (2), 315–351. — — — (1999). “Why is grammaticalization irreversible?” Linguistics, 37 (6), 1043–1068. Heine, Bernard, & M. Reh (1984). Grammaticalization and Reanalysis in African Languages. Hamburg: Buske. Heine, Bernard Claudi, U., & Hünnemeyer, F. (1991). “From cognition to grammar. Evidence from African languages”. In Elizabeth C. Traugott & B. Heine (Eds.), Approaches to Grammaticalization. Vol. I (pp. 149–187). Amsterdam/Philadelphia: John Benjamins. Hopper, Paul (1991). “On some principles of grammaticization”. In Elizabeth C. Traugott & B. Heine (Eds.), Approaches to Grammaticalization. Vol. I (pp. 17–35). Amsterdam/Philadelphia: John Benjamins. Hopper, Paul, & S. Thompson. (1980). “Transitivity in Grammar and Discourse”. Language, 56 (2), 703–753. — — — (1985). “The iconicity of the universal categories ‘noun’ and ‘verb”’. In John Haiman (Ed.), Iconicity in Syntax (pp. 151–183). Amsterdam: John Benjamins. Jespersen, Otto (1984). Analytic Syntax. With a Introduction by James D. McCawley, Chicago: University of Chicago Press. Keller, Rudi (1994). Language change: the Invisible Hand in Language. London: Routledge.

Ditransitive to monotransitive structure in Spanish 

Kirsner, Robert (1985). “Iconicity and grammatical meaning”. In John Haiman (Ed.), Iconicity in Syntax (pp. 249–267). Amsterdam: John Benjamins. Langacker. R. (1977). “Syntactic Reanalysis”. In Charles Li (Ed.), Mechanisms of Syntactic Change (pp. 57–139). Austin & London: University of Texas. Lehmann, Christian (1982). Thoughts on Grammaticalization. A Programmatic Sketch, Vol. I. University of Cologne: Arbeiten des Kölner Universaliensprojekts 48. Lichtenberck, Frantisek (1991). “On the gradualness of grammaticalization”. In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to grammaticalization. Vol. I. (pp. 37– 80). Amsterdam/Philadelphia: John Benjamins. Mithun, Marianne (1984). “The evolution of noun incorporation”. Language, 60 (4), 847– 893. Mithun, Marianne (1986). “On the nature of noun incorporation”. Language, 62 (1), 32–37. Gutiérrez Ordóñez, Salvador (1999). “Los dativos”. Gramática descriptiva de la lengua española. Vol. 2. dir. by Ignacio Bosque & Violeta Demonte. 1855–1930. Madrid: Espasa Calpe. Ortiz Ciscomani, Rosa María (2002). “La bitransitividad en el español. Estructura y evolución”. Ph. D. Dissertation unpublished. México: UNAM. — — — (2003). “Non diachrony of ditransitivity in Spanish”. In G. Fiorentino (ed.), Transitivity in Romance Languages. Semantics and Pragmatics (pp. 261–298). BerlinNew York: Mouton de Gruyter. Quirk, Randolph, S. Greenbaum, G. Leech, & J. Svartvik (1985). A comprehensive Grammar of the English Language. London and New York: Longman. Schøsler, Lene (2003). “Grammaticalisation of Valency Patterns? – An Investigation into Valency Patterns and Support Verb Constructions, based on Diachronic Corpora”. Forum for Modern Language Studies, 39 (4), 400–413. Timberlake, Alan (1977). “Reanalysis and actualization in syntactic change”. In Charles Li (Ed.), Mechanisms of Syntactic Change (pp. 141–177). Austin: Texas University Press. Traugott, Elizabeth (1999). “A Historical Overview of Complex Predicate Types”. In Laurel J. Brinton & M. Akimoto (Eds.), Collocational and Idiomatic Aspects of Composite Predicates in the History of English (pp. 239–260). Amsterdam/Philadelphia: John Benjamins. Traugott, Elizabeth C., & Bernd Heine (1991). “Introduction”. In Elizabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization. Vol. I (pp. 1–20). Amsterdam/Philadelphia: John Benjamins.

Reflexive intensification in Spanish Toward a complex reflexive?* Johan Pedersen University of Copenhagen

.

Introduction

This study aims to analyze the Spanish reflexive / intensifier distinction in a diachronic-typological perspective. Many languages have reflexive markers that can be combined with optional intensifiers. This is exemplified in the following Spanish construction: (1) Pedro confía en sí (mismo) Pedro relies on refl self “Pedro relies on himself ”

This also applies to e.g. German, which is typologically similar to Spanish in this respect (Peregrín Otero 1999: 1456): (2) Der Kanzler ist stolz auf sich (selbst) The chancellor is proud of refl self “The chancellor is proud of himself ” (König & Siemund 2000: 50)

It is generally assumed, thus, that in (1) and (2) the linguistic marker of reflexive meaning (sí/sich) has to be distinguished grammatically from the marker of intensification (mismo/selbst).1 See e.g. König 2001, König & Siemund 2000, Otero 1999, Baker 1995, Kemmer 1995. In this paper, however, it will be argued, on the basis of empirical evidence, that the Spanish reflexive – intensifier construction sí (mism-) is in a process of being reanalyzed as a complex reflexive sí mism-. Let me first point out what might be considered “the general view” on reflexive intensification in Spanish linguistics, as recently stated by Carlos Pere-

 Johan Pedersen

grín Otero in the extensive and detailed Spanish grammar: Gramática Descriptiva del Español (Otero 1999: 1449, 1509): ‘The modifier mism- is always possible, but it is never necessary, . . ., so we can deduce that it plays no role in the expression of reflexivity . . .. The distinction between morphologically simple and complex anaphors does not seem to have any consequences for Spanish that are worth mentioning’. (my translation from Spanish)

Basically he claims, maintaining the traditional point of view, that mism- has nothing to do with the expression of reflexivity in Spanish. See e.g. Alarcos Llorach 1995: 127; Real Academia Española 1981: 211; Franch y Blecua 1975: 677f.; Bello & Cuervo 1954: 246. Defenders of this position are forced to face the following question: Is mism- always optional? In Spanish reflexivity is basically expressed in two different constructions: the prototypical argument structure se-construction, in which se is a clitic reflexive, sometimes doubled by an unbound complex form in the prepositional case (a sí mism-), as in (3); and a non-clitic sí-construction, sometimes intensified by mism-, in which sí is an unbound reflexive in the prepositional case, as in (1), repeated for convenience: (3) Fernando se lava (a sí mismo) Fernando refl washes obj mark refl self “Fernando washes himself ” (1) Pedro confía en sí (mismo) Pedro relies on refl self “Pedro relies on himself ”

It is well known that in the prototypical argument structure construction, see (3), in which the clitic se is required, mism- is certainly not optional in Modern Spanish. You can’t say: *se lava a sí; it has to be: se lava a sí mism-. If we look at the non-clitic constructions, in many contexts the complex form, including mismo, is required for most speakers (Otero 1999), as exemplified in (4). This is not the case in similar German constructions of intensification, in which the use of selbst, on the contrary, is restricted to the most extreme cases of non-other directed predicates (König & Siemund 2000: 63f.; see also Otero 1999: 1456 and Baker 1995: 98f.). This is exemplified in (5): (4) Pedro está orgulloso de sí mismo Pedro is proud of refl selv “Pedro is proud of himself ”

Reflexive intensification in Spanish 

(5) Paul solperte über sich selbst Paul stumbled over refl self “Paul stumbled over himself ”

(König & Siemund 2000: Note 18)

On the other hand, Danish, as well as other languages, has a complex reflexive with an incorporated (obligatory) element of intensification. In Danish you would always say: (6) Han sad og tænkte over det for sig selv He sat and thought about it for refl self “He was thinking about it on his own” (han sad og tænkte over det for *sig)

Consequently, for the speakers concerned, Spanish mism- seems to some extent, just as the Danish selv in (6), to be relevant for the construction of reflexivity in certain contexts. If it is true that mism- is in a process of becoming an obligatory part of the prepositional construction of reflexivity in Spanish, we may expect that the use of complex constructions containing mism- has increased markedly as related to the use of simple constructions without mism-. In order to approve, or discard, the idea of a complex reflexive being under development in Spanish, I have carried out a quantitative diachronic study of the use of mism- in reflexive constructions. Thus, the principal empirical question I seek to answer in this paper is: Is there a clear indication of historical change in the use of mism- in reflexive constructions? The paper is structured as follows: In Section 2 some major contributions to the analysis of reflexive intensification in the literature will be sketched out. The empirical method will be outlined in Section 3. Subsequently, results from the empirical study of the diachrony of sí mism- will be presented in Sections 4 and 5. In Section 6 the data will be analyzed and interpreted, basically by examining the plausibility of two distinct hypotheses. Finally, the results will be summarized and a conclusion will be drawn in Section 7.

. The construction of reflexivity and intensification In many European languages, including Spanish, the reflexive and the intensifier may be distinguished as regards form as well as syntactic distribution (e.g. German: sich/selbst, Italian: sè/stesso, Spanish: sí/mism-). In these languages the reflexive pronoun and the intensifier are combinable: sich selbst, sè stesso, sí mism- etc., see (7)–(10):

 Johan Pedersen

= Intensifier (7) Der Kanzler selbst wird anwesend sein The Chancellor self will present be “The Chancellor himself will be present” (König & Siemund 2000: 50) = Refl. (+intensifier) (8) Der Kanzler ist stolz auf sich (selbst) The Chancellor is proud of refl self “The Chancellor is proud of himself ” (König & Siemund 2000: 50) (9) Él mismo dijo que no lo iba a hacer. He self said that not it would do “He himself said that he wasn’t going to do it” (10) Pedro está orgulloso de sí (mismo) Pedro is proud of refl self “He is proud of himself ”

= Intensifier

= Refl. (+ intensifier)

Correspondingly, the term emphatic reflexives (emphatics), which is used in part of the literature (e.g. Kemmer 1995; Quirk et al. 1985), indicates that intensifiers are supposed to be understood merely as providing a special emphatic use of reflexives. Nevertheless, in theoretical terms, in particular when semantic aspects are to be taken into account, various scholars have argued that it is in some cases highly problematic to draw a clear distinction between reflexives and reflexive intensifiers (e.g. König & Siemund 2000, Baker 1995, Zribi-Hertz 1995). On the other hand, some recent theoretical studies of reflexivity distinguish between at least two types of reflexives: 1) morphologically simple (simplex or SE-) anaphors (Dutch zich, Danish sig, Italian sè, etc.) and 2) morphologically complex (complex or self -) anaphors (e.g. English himself, Dutch zichzelf, Danish sig selv) (see e.g. Burzio 1996, 1991; Reinhard & Reuland 1993; Pica 1991; Everaert 1986). In the latter type, the simple reflexive is typically obligatorily combined with an element of intensification, constituting together a complex reflexive.2 Basically there are: A) Simple reflexives (11) Hans zag de hond naast zich / *zichzelf Hans saw the dog next simp refl / self-refl “Hans saw the dog next to him”

Dutch (Everaert 1986: 2)

B) Complex reflexives (12) Han er stolt af sig selv / *sig He is proud of self-refl / *simp-refl “He is proud of himself ”

C) Simple reflexives + optional intensifier

Danish

Reflexive intensification in Spanish 

(13) Der Kanzler ist stolz auf sich (selbst) The chancellor is proud of refl self “The chancellor is proud of himself ”

German

It is rather unclear how intensifiers have developed historically, but the typical case seems to be that reflexive anaphors develop from intensifiers (see e.g. König & Siemund 2000; Schladt 2000; Baker 1995; Zribi-Hertz 1995). However, this typological pattern is not necessarily applicable to Modern Spanish. It is primarily based on the fact that in a number of languages (e.g. English, but not Spanish) the same expression is used both as an intensifier and as a reflexive anaphor. Moreover, in the Spanish case we have to take into account the fact that the prototypical reflexive construction has undergone a process of cliticization, as will be shown in Section 2.2. . The meaning of reflexivity and intensification The intensifier identifies an entity that is salient in the discourse in contrast to other potential referents, which have just been mentioned, or could have been mentioned (Kemmer 1995: 66). König & Siemund claim, in line with Kemmer, that intensifiers provide potential alternatives to the accentuated NP-referent, to which they are typically collocated. The use of the intensifier implies that the status of the NP-referent has to be evaluated by inferring an actual prominence structure in discourse (König & Siemund 2000: 45–46, 1996a, b). See also Baker (1995: 80ff.) and König (1991: 87ff.). It seems clear that the construction of intensification is closely related to the construction of reflexive meaning, both diachronically and synchronically. Both constructions imply, in many cases, that something is contrary to the expectations created in discourse (Kemmer 1995). The reflexive marker typically encodes a coreference that is contrary to the expectation created in the semantic valency structure of the verb, or in context (The old horse heaved himself out of the mud, Kemmer 1995: 56). The intensifier, in turn, often appears emphasizing, unexpectedly, the prominent role in the clause, or in the discourse, of some referent as opposed to more natural candidates (I wanted Marie herself to tell me, Kemmer 1995: 56). This common semantic structure of reflexivization and intensification highlights the conceptual closeness of the categories intensifier and reflexive marker. The term “unexpectedness”, however, seems to be problematic, since the intensifier is not basically used, according to Baker, to indicate unexpectedness but to mark discourse prominence (Baker 1995: 95, Note 33). See also the discussion in Zribi-Hertz 1995 and König & Siemund 2000.

 Johan Pedersen

. The clitic argument structure construction in Spanish In modern Spanish the unbound reflexive, as direct or indirect object (e.g. a sí mism-), has to be doubled by a clitic reflexive (se). In other words, the use of the clitic is obligatory in these constructions (for more details, see Torrego 1995):3 (14) Se ha tomado el pulso (a sí mismo) refl has taken the pulse obj mark refl self “He has examined his own pulse”

On the other hand, this use of the unbound reflexive is only possible in some verbal constructions, and it is never obligatory (in brackets in (14)). Therefore, we may hypothesize in the first place that in those reflexive constructions in which the unbound form may occur as a redundant object marker, the whole expression ‘a+refl+mism-’, not only mism-, has to be interpreted as an optional marker of reflexive intensification. Notice, in addition, that the unbound form in these argument structure constructions of Modern Spanish only very rarely occurs without mism-, see (15), as in other prepositional case constructions, e.g. para sí (mism-). A sí without mism- occurs only in adverbial prepositional expressions like frente a sí, in which there is no reflexive object to be doubled, cf. (16): (15) Se ha tomado el pulso *a sí (16) Clinton tiene frente a sí 4 años para innovar... Clinton has front of refl 4 years to innovate... “Clinton has in front of him 4 years to innovate...”

(CREA, 1996)

It seems therefore plausible to further hypothesize that in Modern Spanish the use of mism-, as a compulsory part of the optional intensifier in the clitic construction (se. . . (a sí mism-)), somehow is related to the selfsame cliticization of the construction. In fact, in Old Spanish non-clitic constructions, the reflexive object was normally constructed without mism- (17), though the use of a sí mism- was also frequent (18): (17) E obligaron a sí e a sus bienes,... and mortgaged-3pl obj mark refl and obj mark their belongings “and they mortgaged themselves and their belongings,...” (CORDE, from 1308)

Reflexive intensification in Spanish 

(18) y que el que ama a sí mismo,... and that det that loves obj mark refl self “..., and that those who love themselves” (CORDE, from 1537)

It follows that the increasing use of mism- in Old Spanish cannot only be due to the cliticization of the construction at that time. It might also be related to changes in the use of the prepositional object marker a. In Old Spanish one of the basic functions of the a-marker was to focalize human referents considered particularly important (Detges in press). According to Detges, referring to Meier (1945: 247) and Delille (1970: 34), the use of the equivalents of the Spanish reflexive emphatic mism- in Medieval Portuguese and Old Catalan was normally an impediment for the use of the prepositional a-marker. For instance, when preceeding si mateix in Catalan the preposition a was never used. Detges interprets this pattern as indicating that the focalizing use of the prepositional marker was considered unnecessary when other focalizing elements were used. Thus, we may similarly hypothesize that the loss of the focalizing weight of the Spanish a-marker in Old Spanish motivated an increased use of the emphatic element mism-, which, as we have seen, has become a compulsory part of the intensifying reflexive expression a sí mism- in the argument structure construction in Modern Spanish. This interpretation is actually highly compatible with, and supports, the hypothesis stated earlier in this section, that the increased use of mism- in clitic constructions (se. . . a sí mism-) is related to the cliticization of the construction. This is so, because the focalizing weight of the a-marker was in fact weakened by the emergence of the clitic duplication itself (see the discussion in Detges in press). To summarize: In Modern Spanish, the complexity (with mism-) of the reflexive unbound form is compulsory in the clitic argument structure construction, whereas its use is optional: se. . . (a sí mism-). Correspondingly, mismseems to have achieved a new status in this construction in Modern Spanish: from being mism- an optional part of the reflexive marker in Old Spanish, the whole expression a sí mism- has in Modern Spanish an intensifying function in the clitic construction. In Section 6, data that support this interpretation will be presented. Another interesting perspective of this analysis is that both similar pragmatic factors and the syntactic pattern itself (the required complexity of the unbound reflexive form sí mism-) might have been motivating factors for an increased use of the complex form sí mism- in non-prototypical prepositional constructions (see next Section). On the basis of extensive empirical evidence, this question will be analyzed in detail in Section 6.

 Johan Pedersen

. The Spanish non-clitic reflexive In the non-clitic construction the intensifier (mism-) is, according to Spanish convention, optional: (19) Pedro está orgulloso de sí (mismo) Pedro is proud of refl self “Pedro is proud of himself“

However, many Spanish speakers prefer the complex reflexive form in many contexts. Carlos Otero (1999: 1500) mentions in passing that the complex form (+mism-) is required for some speakers in specific construction types, such as the one in (20): (20) Sólo Ana habla siempre de sí misma Only Ana talks always about refl self “Only Ana is always talking about herself ”

When the same speakers choose a non-reflexive anaphoric construction in the same context the complex form is not required: (21) Sólo Ana habla siempre de ella (misma) Only Ana talks always about pers pron-3sg self

This is surprising since the pronominal reference in (21) actually is ambiguous in the simple anaphoric construction without misma (the pronominal anaphor refers either to Ana or to someone else), while it would not be so in a simple reflexive construction in (20). In other words, misma would resolve the ambiguous interpretation in (21), whereas there is no ambiguity to resolve in (20). Such “test” constructions suggest that mism- is not merely an intensifier, but is rather closely related to the construction of reflexive meaning.4 In the subsequent sections I will present substantial empirical evidence that shows that an increased use of the complex reflexive should not be ascribed simply to an increased, pragmatically determined, use of the intensifier. I will argue that such a change has to be seen as related to the development of a complex reflexive.

. The empirical method: varying token ratio In the empirical part of this study I have measured the use of the complex form sí mism- as related to the use of the simple form sí; that is, the token ratio (TR) of the complex form.5 The token ratio has been calculated on the basis

Reflexive intensification in Spanish 

of searches in the diachronic corpus CORDE (C1) and the synchronic corpus CREA (C2) for uses of sí and sí mism-.6 The difference between the token ratio of sí mism- in two corpora C1 and C2 is calculated as a differential factor DTR (difference of token ratio): DTR(C1, C2) = TR(C2) / TR(C1). The chi square test (χ2 ) has been used to ensure that observed variations really are significant, and that they should not be interpreted simply as incidences of statistical coincidences. The advantage of having the token ratio, instead of absolute values, as the basic measure of frequency and historical change, is the possibility of comparing non-homogeneous corpora of different sizes as a reliable source for the extraction of data. Dialectal and text-stylistic variation may subsequently be taken into account by analyzing usage in differentiated corpora. The problem is that notable tendencies observed in the data may be due to over-representation of certain text types, or dialectal variants, in which the simple or the complex form is more or less frequent than elsewhere. Another issue of corpus linguistic methodology that has to be taken into account is that, for practical reasons, this study has only included the third person forms of the reflexive construction.7 That is, I have only searched for occurrences of sí mism-, while mí mism- (first person) and ti mism- (second person) have been excluded from the study.8 The data has been ajusted for formal variations, as the early period of Old Spanish presents different formal variants, in particular meísmo, misme and me(e)smo. Searches in CORDE confirm that mesmo was used almost exclusively from 1400–1700, as stated by Corominas (1954: 57). From the beginning of the 17th century mismo is the dominant form, and from the 18th century mesmo is only sporadically used in rural areas. The other formal variants are practically non-existent in CORDE.

. General data on the historical change In Table 1 the general data on the historical change are presented for the most common prepositions used in this construction type. In the last table row the values are calculated on the basis of added occurrences of ‘preposition + sí’ and ‘preposition + sí mism-’ respectively, in order to provide a measure of the general tendency. The χ2 -test confirms that there are actually significant historical changes going on for all the prepositional constructions. The data show that the token ratio (TR) has increased dramatically for all the prepositional constructions, though the change in using a sí mism- in the clitic construction is particularly striking. Also sobre sí mism- and de sí mism- have turned

 Johan Pedersen

Table 1. Token ratio of sí mism- in an overall historical perspective Corde Sí

Sí mism-

A = obj. marker

2281

3388

Para = to/for

2522

Por = by De = of

TR

χ2

Crea

TR

χ2

DTR

Sí

Sí mism-

1,49 7,8E-l97

173

4643

26,84 2,5E-231 18,07

261

0,10 2,41E-22

1166

450

0,39

2,77E-37 3,73

6061

2264

0,37 2,79E-42

2465

2184

0,89

2,75E-74 2,37

7009

2951

0,42 3,9E-107

1754

2999

1,71

2,7E-222 4,06

2218

En = in

6230

1912

0,31 2,88E-98

2552

1,15

2,5E-166 3,75

Sobre = about

967

444

0,46 0,000768 164

727

4,43

4,11E-05 9,65

Con = with

9930

690

0,07 2,83E-52

2534

873

0,34

6,1E-159 4,96

Added occ.

35000 11910

0,34 0

10474 14428

1,38

0

4,05

out to be predominant in the non-clitic constructions, which is reflected in the calculation of TR to be nearly 4,5 and 1,7 respectively in the Modern Spanish corpus (CREA). From the added occurrences the general value of TR in CREA has been calculated to be 1,38, which indicates that the complex form sí mismis substantially more common than the simple form sí in the corpus of Modern Spanish. This is, on the contrary, definitely not so in the diachronic corpus (CORDE), for which TR is calculated to be 0,34. Finally, we should also pay attention to the differential factor DTR. A sí mism- has a DTR of more than 18, and as for de sí mism-, for example, the TR has been multiplied by more than four in the modern corpus, which is also the average factor for the added occurrences. Observed changes may be due to overrepresentation in a corpus of certain text types for which the simple, or the complex form, is more frequently used.9 This is why calculations of the token ratio in general terms have to be supplemented with calculations based on differentiated corpora. Space constraints make it impossible to report on the detailed data from this part of the study. In general, I found that the use of sí mism- has increased markedly in all text types, though some variation can be observed, as is only to be expected.

Reflexive intensification in Spanish  1,6 1,4 1,2

TR

1 0,8 0,6 0,4 0,2 0 0– 1300

1300– 1400– 1500– 1600– 1700– 1800– 1900– 1975– 1400 1500 1600 1700 1800 1900 1975

Figure 1. The development of sí mism- in terms of token ratio

. More specific data on the historical change In this section, I will show in more detail how the use of the complex construction ‘preposition + sí mism-’ has developed. As illustrated in Figure 1, it has been undergoing constant change up to present time, though the change from the fifteenth to the sixteenth century is particularly remarkable. The corpus (CORDE) has very few occurrences in the period 0-1400, in particular of the complex form; in some cases (with specific prepositions) only a few, or even no occurrences at all. Therefore, we have to be particularly careful when we interpret the data from this period. This is why the discussions and interpretations of the data presented in the subsequent sections are based exclusively on data from 1400-.

. Interpreting the data The data show a continuous, not abrupt, change toward an increased use of the complex form. This study supports, therefore, the claim that language change is a gradual process (see the discussion in Croft 2000: 49ff. and references cited there). Gradualness is difficult to account for in child-based generative theories of language change, although some important attempts have been made (e.g. Lightfoot 1991). Among non-generative approaches a considerable number of frameworks are not based on child-acquisition, but rather on adult usage of

 Johan Pedersen

language (see e.g. Milroy 1992; Bybee et al. 1994; Keller 1994; Croft 2000). In such frameworks, often referred to as usage-based, the gradualness of linguistic change is certainly expected and handled straightforwardly. Moreover, changes in use are supposed in these frameworks to have frequency effects on grammar (Bybee et al. 1994; Bybee & Hopper 2001). In that respect an increased use of the intensifier mism- would be expected to alter the status of this item in grammar. In this section we will be concerned with how to interpret the data by evaluating basically two types of usage-based analysis (cf. the discussion in Newmeyer 2003: 694; Croft 2000: Chapter 5; Labov 1994: Chapter 20): 1. A paradigmatic interpretation (Section 6.1) according to which the selection and the usefulness of an item is in focus as a decisive parameter for change. A paradigmatic interpretation of the data presented in this paper would be of the following type: The optional use of mism- as an intensifier may have increased gradually during the last 700 years for discoursefunctional reasons. 2. A syntagmatic interpretation (Section 6.2), form-function reanalysis (Croft 2000: 120). Recall that in processes of grammaticalization paradigmatic changes affect only one element in the grammaticalizing construction, whereas syntagmatic changes affect more than one element (cf. Lehmann 1985; Haiman 1985). In a syntagmatic perspective we will, in order to offer an explanation of the data, consider the possibility that mism- has been in a process of becoming gradually reanalyzed as a compulsory part of the reflexive marker. . A paradigmatic interpretation First, we may hypothesize that the increase in token ratio of sí mism- reflects a general rise in the selection of mism- as a useful device for intensification. Such an interpretation of the data would predict that a similar phenomenon should be observable in constructions in which mism- is generally assumed to function as an intensifier. Another aspect of this hypothesis is that if the increased use of the complex form exclusively is related to the selection of a useful element, this inherently implies an increased speaker involvement. Therefore, if the hypothesis of usefulness is correct, it is predicted that the observed change should be notably distinct in spoken language, when compared to written language, since the speaker involvement in spoken discourse may be assumed to be relatively higher.

Reflexive intensification in Spanish

.. Distributional data The first prediction is testable by examining how the token ratio of the complex form él/ella mism- [he/she self ]10 has developed in the construction type: él (mism-) lo ha visto [he (self) it has seen] in which it is generally accepted that mism-, and equivalents in typologically similar languages likewise, functions as an intensifier (see e.g. König & Siemund 2000). By measuring the historical development of the expression él/ella (mism-), we will get an indication of how the use of mism- as an intensifier of (pro)nominal phrases has changed, if it has. Thereby, we will be able to evaluate whether or not the altered use of the reflexive construction sí (mism-) could plausibly be ascribed to the general usefulness of mism- as an intensifier in communication. The results from this part of the study are represented in Figure 2. We can observe certain similarities by comparing the development of él/ella (mism-) with the patterns of change of the reflexive construction. In particular, the shift in token ratio from the fifteenth to the sixteenth century reminds us of the course of development of the reflexive construction. However, the change is much less drastic and in fact completely neutralized in the second part of the period, or in terms of the token ratio: the token ratio has been almost constant, and maybe even with a decreasing tendency, for the last four hundred years. It should also be noticed that the simple construction (él/ella), without intensification, is in Modern Spanish nearly 50 times more frequent than the intensified construction (the token ratio of él/ella mism- is 0,02). This tells us that the use of the basic intensifier-construction él/ella mism- has always been, and still is, relatively very limited. On the other hand, the use of the reflexive intensifier construction (the complex variant) is much more extended; see Table 1. The 0,03 0,025

TR

0,02 0,015 0,01 0,005 0 0– 1300

1300– 1400– 1500– 1400 1500 1600

1600– 1700– 1800– 1900– 1700 1800 1900 1975

Figure 2. The development of él/ella mism- in terms of token ratio

1975–



 Johan Pedersen

point I want to make here is that, according to the data presented in Figure 2, the dramatically increased use of the complex reflexive form cannot convincingly be ascribed to a general, pragmatically determined, usefulness of mismas an intensifier of (pro)nominal phrases. At least, there is no indication that such a general tendency, driven by pragmatic principles of usefulness, should be the principal explanatory factor. .. Increased speaker involvement A paradigmatic interpretation of the data is, as we have seen, centered in the usefulness of the intensifier (mism-) as an optional device for foregrounding (pro)nominal phrases in discourse, cf. 2.1. Correspondingly, the use of the complex reflexive form involves, from a paradigmatic point of view, an optional promotion of the subject in discourse, and indirectly an increased speaker involvement. In Detges (in press), what he calls contrastive autothematization in spoken language is a related phenomenon, based on universal pragmatic and discourse principles (Detges in press: 9; see also Detges manuscript). The speaker marks that his contribution contrasts with a previous utterance, or with what the interlocutor presumably believes. This is shown in the following sequence, provided by Detges: (22) H1: [. . .] ¿Es imposible el entenderse entre dos generaciones diferentes, Adolfo Marsillach? H2: Yo [I] es que creo [I think] que no hace falta. Eh. . . Más bien creo que hace falta que no se entiendan. Yo [I] creo que está bien eso, porque la...hay un proceso...eh...competitivo, por el cual las generaciones nuevas necesitan afirmarse, y esa afirmación yo creo [I think] que se produce a través de un inmediato enfrentamiento con el enemigo más próximo, que evidentemente es el familiar que tienen delante. Pero a mí [to me] eso, insisto que no me parece [it seems] mal. (Corp. Oral 92, cit. Detges in press: 9).

The sequence is taken from a television program in which the participants discuss the conflicts between the generations. The speaker is expressing various acts of contrastive auto-thematization in order to indicate that his view diverges from the general opinion. The interlocutor may, according to Detges, expect the content of this kind of utterances to be relevant in the Gricean sense since it is assumed to contain new information. This is a strong motive for using contrastive auto-thematization as a rhetorical strategy; also in contexts with no contrasts of opinion at all, in which such a technique is not objectively justified. In the long run the abusive use of auto-thematization will lead to a

Reflexive intensification in Spanish 

loss of the original contrastive value in constructions that were meant as a way to indicate a high level of communicative relevance. In spoken discourse, rhetorical strategies of auto-thematization may lead to the use of “heavy” focalizing constructions. Along with the emergence of the clitic reflexive marker in the sixteenth century, this may possibly have favored the use of mism- in the reflexive argument structure construction. More specifically, it could be argued that the loss of the original focalizing weight of the Spanish a-marker must have been a strong motivation for using another emphatic element, namely mism- (See also Section 2.2). Furthermore, we have seen that the optional intensifier mism- offers itself as a candidate for contrastive discourse promotion in the non-clitic construction as well, a communicative function that seems to be related to the auto-thematization, advocated by Detges (in press). See Section 2.1. In short, it might be hypothesized that discourse rhetorical strategies of this kind, spreading from first and second person to third person, will appear to be the principal explanatory force behind the increasing use of the complex form sí mism-. Given the validity of this hypothesis we may assume that the observed change, ceteris paribus, should be particularly extended in spoken language, as compared to written language. This assumption is based on the belief that speaker involvement, and the effect of intended pragmatic strategies, must tend to be particularly high in spoken discourse. In Table 2 the data show, surprisingly, that there is not such an overall difference: Table 2. Token ratio of sí (mism-) in spoken/written Modern Spanish Sí A = obj. marker Para = to/for Por = by De = of En = in Sobre = about Con = with

Spoken language Sí mismTR

Sí

Written language Sí mism-

TR

3 16

102 11

34,00 0,69

170 1006

4541 428

26,71 0,43

149 51 271 1 79

80 49 105 8 16

0,54 0,96 0,39 8,00 0,20

2316 1703 1947 163 2455

2104 2950 2447 719 857

0,91 1,73 1,26 4,41 0,35

The prediction for the token ratio is almost fulfilled for the prototypical reflexive argument structure construction, principally a sí mism-, but also to some degree for para sí mism-.11 In the more peripheral prepositional constructions the picture is markedly different. Only sobre sí mism- shows a relatively higher token ratio in spoken language. This specific case can probably be seen

 Johan Pedersen

as a natural consequence of the type of meaning content that is usually involved in this construction.12 Moreover, the token ratio of sobre sí mism- in spoken language is based on very few occurrences (9) in CREA. Regarding the rest of the prepositions, the token ratio for the complex form is strikingly lower in spoken language than in written language. Let me recall that this pattern would be markedly unexpected if rhetorical strategies, as described above, were the primary motive for the extended use of the complex construction. The relatively sparse occurrences of the complex reflexive form in spoken language could alternatively be accounted for by applying simple principles of syntagmatic economy: speakers will tend to use shorter and/or simpler forms for more frequent contexts of use (Croft 2000: 142; Haiman 1983; Bybee 1985). From such principles we would, ceteris paribus, expect the simplest form, i.e. sí, to be the preferred one in spoken discourse, but not necessarily in written language. It might be objected that since this study only concerns third person forms, and consequently only includes occurrences of sí and sí mism-, while the first and second person forms have been excluded, the outcome of written language occurrences has been biased to some extent. The reason for this bias in the data is that spoken language to a larger extent than written language involves first and second person forms. However, the third person forms of the unbound reflexive pronoun are certainly used in spoken language as well, and given the size of the modern corpus CREA (approx. 120 million words), enough material is provided for calculating a reliable token ratio for sí mism- in spoken language. In fact, occurrences of first and second person complex forms, e.g. para mí/ti mism- (to/for my/you self ) are surprisingly rare, even in spoken language, as compared to third person forms. It is therefore highly unlikely that the exclusion of first/second person forms should be responsible for the outcome in Table 2. See also Note 8. Another serious problem for the paradigmatic interpretation is that the hypothesis of increased speaker involvement, or auto-thematization, rests on universal discourse and pragmatic principles (Detges in press: 9). It follows that if this interpretation were correct, we would expect to see the same tendency, at least to some extent, in the German reflexive intensifier-construction. However, there is no evidence that a similar tendency should have been observed in German (see e.g. König & Siemund 2000: 63f.; Baker 1995: 98f.; Hole 2002: 133ff.; Eckardt 2002a, 2002b; Siemund 2000). A convincing analysis of the development of the Spanish reflexive should preferably account for the fact that the development in Spanish differs from the development in German since the German reflexive-intensifier distinction

Reflexive intensification in Spanish 

is typologically very similar to the one we find in the Spanish non-clitic construction. The difference between the two languages must be due to language specific factors. Generalization patterns in Spanish, for instance, may have an influence on the Spanish construction that is stronger than the impact of some universal-typological pragmatic principles. In this perspective, the development of the construction in Spanish may be analyzed by drawing on specific conditions for Spanish, which have no parallel in German, such as the prototypical clitic reflexive construction and its close interaction with the non-clitic construction. See Section 6.2. All together this leads us to conclude that the data show only little evidence for the claim that the increased use of the complex reflexive form in non-clitic constructions should be interpreted in terms of pragmatically determined usefulness. The pragmatic impact on the development should therefore not be exaggerated. This conclusion applies not only for the quantitative data for the prototypical intensifier-construction (6.1.1), but also for the comparison of written and spoken language (6.1.2) and the typological prediction concerning German. The influence from the clitic argument structure construction, in which the complexity of the unbound form has become compulsory, seems to be relevant for the evolution of the more peripheral non-clitic construction. . A syntagmatic explanation: Reanalysis I am certainly not claiming that the observed change in usage may not involve an increased use of reflexive intensification due to pragmatic factors. However, the analysis in Section 6.1 suggests that pragmatic motivations for the increased use of the complex construction cannot per se explain the data. We will now examine the possibility that the observed change in the use of mism- involves not merely an entrenchment of the optional reflexive intensification, but a true reanalysis of the construction (sí + intensifier mism- being reinterpreted as a complex reflexive sí mism-) in the mental grammar of the user. Notice that in French the negative pas was originally part of an intensifier construction ne (pas), much like Modern English not a bit. Later it was reanalyzed as a part of a complex negative ne pas (see e.g. Croft 2000: 130f.; Labov 1994: 596f.; Jespersen 1917: 7f.). Similar changes are attested in many languages (Campbell 1998: 240). In this Section it will be hypothesized that the intensifier construction sí (mism-) is in a process of being reanalyzed in a similar way. Recall that concerning the clitic argument structure construction, in which the clitic se (= reflexive) has become the principal co-referential device, we have hypothesized that a sí (mism-) has been reanalyzed as a reflexive intensifier a sí

 Johan Pedersen

mism, see Section 2.2. Here the expectation is that an increasing token ratio of the clitic construction se. . . a sí. . . is correlated with an increasing token ratio of the complex form a sí mism-.13 For both the clitic and non-clitic construction the hypothesis of reanalysis implies that in the mental grammar of the user, mism- is being conceived as obligatory in the complex form sí mism-. The question is now: what would be a plausible motivation for reanalyzing sí (mism-) in the non-clitic construction as a complex reflexive sí mismif it cannot be exclusively a matter of pragmatic usefulness (cf. the analysis and discussion in Section 6.1)? Tentatively we may argue that the preference of the complex form in the prototypical clitic construction (se. . . a sí mism-) could simply have spread to non-clitic construction types by means of syntactic generalization. If this is true, we must expect the increasing token ratio of the complex form a sí mism- to be correlated with a general increase in the token ratio for the construction type ‘preposition + sí mism-’. In sum, there are two expectations: Expectation A: the increasing token ratio of the clitic construction (se) is correlated with an increasing token ratio of the complex form a sí mism-. Expectation B: the increasing token ratio of the complex form a sí mism- is correlated with a general increase in the token ratio for the construction type ‘preposition + sí mism-’. In Figure 3 the token ratio for the development of the clitic construction, contrasted with the token ratio for a sí mism-, is represented graphically,14 cf. expectation A. In order to focus on the early development of the clitic construction, calculations of the inverse token ratio: a sí / se...a sí (instead of se...a sí / a sí) and a sí / a sí mism- (instead of a sí mism- / a sí) are preferable. The data show that the developments of the clitic construction and the complex a sí mism-construction, in terms of token ratio, are strongly correlated confirming the expectation A. As regarding expectation B, when comparing the development of a sí mism- with the prepositional construction in general, I found that the observed changes in the use of the various prepositional non-clitic constructions follow quite systematically the development of the a sí mism-construction. The result of this comparative analysis, presented in Figure 4, confirms B. To summarize the argument: In the syntagmatic perspective the observed change has been interpreted as a manifestation in usage of a reanalyzed reflexive expression. Due to the cliticisation of se, the reflexive expression a sí (mism-) has been reanalyzed as a complex intensifier a sí mism- in the argument structure construction. This syntactic pattern is generalized by spreading to non-

Reflexive intensification in Spanish 

TR, inverse values

35 (Se)...a si... A si (mism-)

30 25 20 15 10 5 0

0– 1300– 1400– 1500– 1600– 1700– 1800– 1900– 1975– 1300 1400 1500 1600 1700 1800 1900 1975

TR, inverse values

Figure 3. The development of the complex a sí mism- as related to the clitic construction 30 25 20 15 10 5 0

Prep. + si mismA si mism-

0– 1300– 1400– 1500– 1600– 1700– 1800– 1900– 1975– 1300 1400 1500 1600 1700 1800 1900 1975

Figure 4. The development of the unbound reflexive

clitic, non-prototypical construction types, motivating the reanalysis of ‘preposition + sí + optional mism’ as a complex reflexive expression: ‘preposition + sí mism-‘. The entire chain of changes can be schematized as follows: Change 1: Cliticization of se ⇒ Change 2: reanalysis: a sí (mism-) → se (a sí mism-) in clitic constructions ⇒ Change 3: Syntactic generalization: sí (mism-) → sí mism- ⇒ Change 4: reanalysis: prep. sí (mism-) → prep. sí mism- in non-clitic constructions The syntagmatic interpretation is supported by general observations on language change made by William Labov (1994). He observed that in actual speech one variant is rarely chosen over another for functional reasons, that is, as a re-

 Johan Pedersen

sult of an intentional individual choice. Instead he argues that the choice is much more mechanical, and syntagmatic in nature, forming part of what he calls systemic readjustment, without regard to communicative needs. As one example he talks about a tendency to preserve parallel structures in successive sentences, a mechanism which is similar to the syntactic generalization proposed as change 3 (ibid.: 550). Labov does certainly not say that functional arguments are never valid. Rather, he claims that the pressure from specific communicative needs is relatively weak, and can be overridden by a variety of other factors. See also the discussion in Newmeyer 2003. This analysis relies tacitly on the assumption that a syntactic generalization (change 3) may trigger the loss of a semantic distinction and a pragmatic option (change 4). From a mainstream functional (usage based) point of view the analysis might therefore appear problematic, though it certainly raises some interesting questions: Is a clear cut semantic distinction between reflexivization and intensification in Modern Spanish maintainable? Is it possible for syntactic changes to trigger semantic changes? Carl L. Baker (1995) has shown that the distinction between locally binding properties, i.e. reflexivization, and properties of discourse prominence, i.e. intensification, tends to be blurred to the user. If this is so, the clear cut semantic distinction between reflexivization and intensification is not maintainable, and a reanalysis of reflexive intensification, sí (mism-), as a complex reflexive, sí mism-, is semantically and pragmatically highly motivated. There seems, thus, in the syntagmatic perspective to be a semantic-pragmatic motivation as well as a syntactic motivation for the reflexive marker + intensifier to be reanalyzed. Furthermore, the process of reanalysis will naturally tend to eliminate the explicitly coded meaning of intensification; cf. the understanding of reanalysis adopted in this paper: Form-function reanalysis is syntagmatic: it arises from the (re)mapping of form-function relations of combinations of syntactic units and semantic components. The process may nevertheless have an apparently paradigmatic result, for example, a change of meaning of a syntactic unit. (Croft 2000: 120)

I am not suggesting that semantic changes in general are triggered by syntactic changes, but simply that in this specific case a syntactic generalization seems to have had a significant impact on the process of change. Given the validity of Baker’s argumentation, the question whether a similar semantic-pragmatic motivation for reanalysis might have had an impact on the German reflexive as well becomes relevant. There is, apparently, no evidence, as I have pointed out, that a similar process of change is taking place in German. This suggests that

Reflexive intensification in Spanish 

the hypothesized reanalysis of the Spanish non-clitic reflexive may be strongly influenced by the syntactic change in the clitic construction, as proposed in this paper, since the clitic construction is specific for Spanish as opposed to German. On the other hand, this study should be replicated for other Romance languages with clitic reflexives, for instance Italian, to see if the same patterns can be observed. If this is the case, it will provide further support for the analysis offered in this paper.

. Conclusion The data show a continuous increase in the use of mism- in Spanish reflexive constructions during the last 700 years. The change from the fifteenth to the sixteenth century has been shown to be particularly striking. Assuming that the approach must be usage based, two different interpretations of the data have been discussed: A) The paradigmatic interpretation, leading to the hypothesis: The complex construction sí mism- is becoming more entrenched in the grammar as a useful device for intensification. B) The syntagmatic interpretation (of reanalysis), leading to the hypothesis: Mism- is becoming part of the reflexive marker. B in particular has shown to fit well with empirical data on the development of the reflexive construction, though A-phenomena, to some extent, may have been motivational factors for reanalysis (B); cf. Section 2.2 and 6.1. On the other hand, I have argued that the hypothesized reanalysis of the non-clitic construction seems to be influenced by the syntactic structure of the prototypical reflexive argument structure construction (the clitic construction). If this interpretation of the data is correct, a complex reflexive marker, sí mism-, is in a process of being grammaticalized in Spanish, which has not hitherto been realized in Spanish linguistics.

Notes * I owe thanks to the following persons for valuable comments and criticisms of previous drafts and presentations of this paper: Ulrich Detges, Lene Schøsler, Henning Andersen, Daniel Jacob, Jens Nørgård-Sørensen, Thora Vinther, Kasper Boye, Susana Silvia Fernández,

 Johan Pedersen

Ronald Langacker, Nicole Delbeque, Margaret E. Winters, Concepción Company Company, an anonymous reviewer and to the audience at the ICHL. . Henceforth I will refer to all possible variants of this expression by using the form mism-, e.g. para sí mism-. It should also be noted that, for reasons stated below, I have restricted the empirical part of this study to include only third person forms of the reflexive. . Notice, however, that English does not distinguish morphologically between reflexive and intensifier, e.g. he washed himself (reflexive) and Peter himself . . . (intensifier). . It is a general rule that unbound personal pronouns with object function have to be doubled by a clitic pronoun. . Note that sólo in (20) and (21) creates contextual conditions that make the realization of the mism-variant more compelling. Nevertheless, this is so in both constructions, and the point here is that if we compare the preference of the speakers, with respect to the use of mism-, in these two construction types, for some speakers the reflexive construction seems to be particularly well suited for the use of the complex form. . In some cases substantial changes of very small values of TR may be poorly represented by the use of graphs or columns. In such cases the inverse token ratio, that is: ‘preposition + sí / ‘preposition + sí mism-’, will be used in order to improve the visualization of the change, since the inverse formula will convert small values into corresponding high values. . Thus, needless to say, conclusions may be drawn only for the usage in these specific corpora. Nevertheless, if we consider the size and the composition of the corpora CORDE and CREA, available in detail on the website, it becomes clear that they have been composed with the objective of being representative of Old Spanish and Modern Spanish respectively. Each corpus contains approximately 120 million words. In CORDE the texts are from the beginning of the Spanish language to 1974. CREA contains Modern Spanish texts from 1975–. . Methodological shortcomings imply that occurrences of fixed expressions, such as de por sí, are registered in data as instances of the simple form. As they do not allow for optional intensification (*de por sí mism-) they are irrelevant for this study and should in ideal conditions have been left out. . Notice, however, that in languages that have reflexive pronouns in all three persons, as in Spanish, the use of reflexive constructions in the third person is much more frequent than in the first or second (Newmeyer 2003: 695). . This may also be true for the distinction of written vs. spoken language; see below. . Reflecting the use of the complex form as related to the use of the simple form. . It should be taken into account, however, that the process of change favoring the use of the complex form a sí mism- is already completed at the beginning of the time period covered by the corpus CREA (1975), on which the comparison in Table 2 is exclusively based. Thus the data in Table 2, concerning a sí mism-, cannot be used directly to argue for the correctness of the hypothesis of usefulness. Since we do not have diachronic data of spoken Spanish, we will have to rely on the argument presented in the first paragraph of this Section: the loss of the focalizing weight of the Spanish a-marker could have favored the use of mismin the reflexive argument structure construction.

Reflexive intensification in Spanish  . The predicate frequently involves a kind of subjective reflection that makes the use of the mism- variant more compelling, as in the following example: La función del siquiatra no es la de ”normalizar”, sino tratar de dar al paciente una posible lucidez sobre sí mismo, una visión más clara sobre el mundo que le rodea [The role of the psychiatrist is not to “normalize” but to try to give the patient a possible lucidity upon himself, a clearer vision of the world that surrounds him] (CREA). . The token ratio of the clitic construction se . . . así . . . reflects the clitic use as related to the non-clitic use, and the token ratio of the complex form a sí mism-, as defined in Section 3, the use of the complex form a sí mism- as related to the simple form a sí. . The calculation is based on registered occurrences of the word se within a distance of five words from the expression a sí.

References Alarcos Llorach, Emilio (1995). Gramática de la Lengua Española. Real Academia Española. Colección Nebrija y Bello. Madrid: Espasa Calpe. Alcina Franch, Juan & Juan M. Blecua (1975). Gramática Española. Barcelona: Ariel. Baker, Carl L. (1995). “Contrast, discourse prominence, and intensification, with special reference to locally free reflexives in British English”. Language, 71 (2), 63–101. Bello, Andrés & Rufino José Cuervo (1954). Gramática de la lengua Castellana. Prepared by Niceto Alcala-Zamora y Torres. Buenos Aires: Sopena Argentina. Bosque, Ignacio & Violeta Demonte (Eds.) (1999). Gramática Descriptiva de la Lengua Española. (GDE). Madrid: Espasa. Burzio, Luigi (1991). “The morphological basis of anaphora”. Journal of Linguistics, 27, 81– 105. Burzio, Luigi (1996). “The role of the antecedent in anaphoric relations”. In Robert Freidin (Ed.), Current Issues in Comparative Grammar (pp. 1–45). Dordrecht: Kluwer. Bybee, Joan (1985). Morphology: A study into the relation between meaning and form. Amsterdam: John Benjamins. Bybee, Joan, Revere D. Perkins & William Pagliuca (1994). The evolution of grammar: Tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press. Bybee, Joan & Paul Hopper (Eds.) (2001). Frequency and the emergence of linguistic structure. Amsterdam: John Benjamins. Campbell, Lyle (1998). Historical Linguistics. Edinburgh: Edinburgh University Press. CORDE. Internet-based database: http://www.rae.es/. The Royal Spanish Academy (RAE). Corominas, Joan (1954–57). Diccionario crítico etimológico de la lengua castellana. Madrid: Gredos. CREA. Internet-based database: http://www.rae.es/. The Royal Spanish Academy (RAE). Croft, William (2000). Explaining Language Change. An Evolutionary Approach. London: Longman. Delille, Karl Heinz (1970). Die geschichtliche Entwichlung des präpositionalen Akkusativs im Por tugiesischen. Bonn: Romanisches Seminar.

 Johan Pedersen

Detges, Ulrich (In press). “La gramaticalización de los acusativos preposicionales en las lenguas iberorrománicas. Una hipótesis pragmática”. In Valeriano Bellosta von Kolbe & Gabriele Knauer (Eds.), Variación sintáctica en español: un reto para las teorías de la sintaxis. Tübingen: Niemeyer. Detges, Ulrich (Manuscript). “Du sujet parlant au sujet grammatical. L’obligatorisation des pronoms sujets en ancien francais dans une perspective pragmatique”. Eckardt, Regine (2002a). “Reanalyzing selbst”. Natural Language Semantics, 9 (4), 371–412. Eckardt, Regine (2002b). “Semantic change in grammaticalization”. In Graham Katz et al. (Eds.), 53–67. Everaert, Martin (1986). The syntax of Reflexivization. Dordrecht: Foris. Fernández Ramírez, Salvador (1987). Gramática Española, 3.2. El pronombre (1951). Vol. Preparado por José Polo (2. edición.). Madrid: Arco/Libros. Frajzyngier, Zygmunt & Traci S. Curl (Eds.) (2000). Reflexives. Forms and functions. Amsterdam/Philadelphia: John Benjamins. Geniusiene, Emma (1987). The typology of reflexives. Berlin: Mouton de Gruyter. Haiman, John (1983). “Iconic and economic motivation”. Language, 59, 781–819. Haiman, John (1985). Natural syntax. Cambridge: Cambridge University Press. Hole, Daniel (2002). “Agentive Selbst in German”. In G. Katz et al. (Eds.), 133–150. Huang, Yan (2000). Anaphora. A Cross-linguistic Approach. Oxford: Oxford University Press. Jespersen, Otto (1917). Negation in English and other languages. Copenhagen: A. F. Høst. Katz, Graham, Sabine Reinhard & Philip Reuter (Eds.) (2002). Sinn & Bedeutung VI, Proceedings of the Sixth Annua Meeting of the Gesellschaft für Semantik. University of Osnabrück. Keller, Rudi (1994). On language change: The invisible hand in language. London: Routledge. Kemmer, Susanne (1995). “Emphatic and reflexive – self: Expectations, viewpoint, and subjectivity”. In Dieter Stein & Susan Wright (Eds.), Subjectivity and subjectification (pp. 55–82). Cambridge: Cambridge University Press. König, Ekkehard (1991). The meaning of Focus Particles: A comparative perspective. London: Routledge. König, Ekkehard & Peter Siemund (1996a). “Emphatische reflexiva und fokusstruktur”. In Inger Rosengren (Ed.), Sprache und Pragmatik 40 (pp. 1–42). Lund: Lunds Universitet. König, Ekkehard & Peter Siemund (1996b). “Selbst-Reflektionen”. In Gisela Harras (Ed.), Wenn die Semantik Arbeitet. Festschrift für Klaus Baumgärtner (pp. 277–302). Tübingen: Niemeyer. König, Ekkehard & Peter Siemund (2000). “Intensifiers and reflexives: A typological perspective”. In Frajzyngier & Curl, 41–74. König, Ekkehard (2001). “Intensifiers and Reflexive Pronouns”. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible, Language Typology and Language Universals: An international Handbook (pp. 747–760). Berlin: Walter de Gruyter. Labov, William (1994). Principles of linguistic change, Vol. 1: Internal factors. Oxford: Blackwell. Lehmann, Christian (1985). “Grammaticalization: Synchronic variation and diachronic change”. Lingua e Stile, 20, 303–18.

Reflexive intensification in Spanish 

Lightfoot, David W. (1991). How to set parameters: Arguments from language change. Cambridge: Cambridge University Press. Meier, Harri (1945). “O problema do acusativo preposicional no catalão”. Boletim de Filologia, VIII, 327–60. Milroy, James (1992). Linguistic variation and change. Oxford: Blackwell. Newmeyer, Frederic J. (2003). “Grammar is grammar and usage is usage”. Language, 79 (4), 682–707. Peregrín Otero, Carlos (1999). “Pronombres reflexivos y recíprocos”. In Bosque & Demonte, 1427–1517. Pica, Pierre (1991). “On the interaction between Antecedent-Government and Binding: The case of Long-distance Reflexivization”. In Jan Koster & Eric Reulan (Eds.), Long-distance Anaphora (pp. 119–36). Cambridge: CUP. Quirk, Randolf, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik (1985). A comprehensive Grammar of the English Language. London: Longman. Real Academia Española (1981). Esbozo de una Nueva Gramática de la Lengua Española. Madrid: Espasa-Calpe, S.A. Reinhard, Tanya & Eric Reuland (1993). “Reflexivity”. Linguistic Inquiry, 24 (4), 657–720. Schladt, Mathias (2000). “The typology and grammaticalization of reflexives”. In Frajzyngier & Curl, 103–124. Siemund, Peter (2000). Intensifiers. A Comparison of German and English. London: Routledge. Torrego, Esther (1995). “From Argumental to Non-Argumental Pronouns: Spanish Doubled Reflexives”. Probus, 7, 221–241. Van Hoek, Karen (1997). Anaphora and conceptual structure. The University of Chicago Press. Zribi-Hertz, Anna (1995). “Emphatic or Reflexive”? On the Endophoric Character of French lui-même and Similar Complex Pronouns”. Journal of Linguistics, 31, 333–374.

Modern Swedish bara From adjective to conditional subordinator Henrik Rosenkvist Lund University

.

Introduction1

The Modern Swedish conditional subordinator bara conveys two distinct submeanings; by using bara (‘if only’ or ‘if just’) instead of the default om (‘if ’) the speaker underlines that what is expressed in the apodosis is desired, and that the condition in the protasis is explicitly sufficient (‘no more than X is needed for Y to occur’, or ‘if just X, then Y’): (1) Bara det slutar regna så kan vi resa tältet. only it ceases to-rain then can we raise tent-the “If just the rain stops we may pitch the tent”

These semantic properties of conditional bara make the use of it restricted (Wijk-Andersson 1991: 179) – it can not, as a rule, be used in atypical conditional clauses (like for instance speech act conditionals, cf. Sweetser 1990). In this paper the evolution of conditional bara is investigated, from both a semantic and a syntactic perspective, and it is argued that both the path of change as well as the origin of bara have contributed to the semantic peculiarities of conditional bara. Furthermore, I claim that it is necessary to utilize both syntactic and semantic modes of explanation in order to understand the entire development. In Section 2, the background of bara is presented. The assumed development can then be found in Sections 3 and 4. The topic of Section 5 is a handful of other grams, the evolution of which may contribute to the understanding of the changes of bara: MSw allena, allenast, blott, enbart and endast. The possi-

 Henrik Rosenkvist

ble theoretical implications of the assumed changes are discussed in section 6, from the perspective of grammaticalization theory.

. The etymological origin of bara Bara was originally, according to SAOB2 (B 292), an inflected variant of the EOSw3 adjective bar, which had the meanings ‘bare, naked, uncovered’; the affix -a is frequent in the nominal paradigms of inflection in OSw. The root is probably Indo-European (Hellquist 1980: 51; Pfeifer 1989: 122), and it consequently appears, with approximately the same basic meaning, in a number of languages that are closely related to Swedish.4 SAOB (B 281–295) accounts for three lexical instances of bara: as an inflected form of the adjective bar, as an uninflectable adjective and as an adverb/ subordinator. A number of various meanings are presented in SAOB; in the following section I suggest that these can be attributed to two different paths of meaning development. One of these, the meaning of ‘obvious’ and the related meanings, was a semantic cul-de-sac (see below).

. From adjective to adverb In the beginning of the 16th century, the adjective bar could mean either ‘naked, uncovered’ (2a) or ‘obvious, manifest’ (2b): (2) a.

oc the hwggo hiälmen aff honom oc bleff sedan baar oc owänkter (Troja, 1529) and they hacked helmet-the off him and became then bare and unprotected “and they hacked off his helmet and he then became bare and unprotected” b. Heluetet är baart för honom [...] (SAOB, 1541) hell-the is obvious for him [...] “Hell is manifest for him”

In the end of the 16th century, an adverbial bara appears, with the meaning ‘completely, fully’, and some decades later we find examples of an adjectival bar with a very similar meaning. It seems likely that this meaning (‘completely’) emanates from the earlier meaning ‘obvious, manifest’ (2b). Another meaning that probably is related to this small semantic field is a meaning related

Modern Swedish bara 

to money (‘cash’), which we find as early as 1540 (SAOB: B 286). SAOB mentions the German constructions bares geld, bar bezahlen as a possible source of influence for the emergence of this meaning. In contemporary Swedish, none of these meanings are in use. The related meanings ‘obvious’, ‘completely’ and ‘cash’ consequently seem to have been dead ends, and they did not, I assume, take part in the rise of conditional bara. In the early 17th century, however, we find a new meaning of the adjectival bar: ‘only, merely, no more than’ (3a). At about the same time (the first example in SAOB is from 1620), an adverbial bara appears, with the same meaning (3b): (3) a.

[...] Gibraltar, ther man icke [...] tullar för bara passagen. (SAOB, 1640). [...] Gibraltar where you not [...] pay for only passage-the “Gibraltar, where you do not [...] pay duty for merely the passage” b. En manade den andre ut til at fäckta bara för skryyt skull. (Columbus, 1675) one urged the other outVPL5 for to fence only for boasts sake “one urged the other to fence only for the sake of boasting”

It is plausible that this meaning (‘only’) has emerged directly from the basic meaning of bar (‘naked, uncovered’), I think, since these meanings in a sense are overlapping; a naked sword is nothing but a sword, or only a sword. It is hence not surprising to find examples that may be interpreted in either of these meanings: (4) Så stule de bort hans swälzbalia, at han danza mädh bara värian. (Horn, 1657) so stole they awayVPL his scabbard that he danced with only rapier-the “then they stole his scabbard, so that he danced with only/the bare rapier”

It also seems clear that an adjective meaning ‘only’ easily may be analyzed as an adverb – the abstract meaning no longer requires a nominal complement, as an element meaning ‘only’ also can modify events. So, in the end of the 17th century, bara carries three specific meanings: ‘naked’, ‘obvious’ (which I from now will ignore, for reasons mentioned above) and the exclusive (i.e. non-inclusive) meaning ‘only’, and it may function as an adjective or as an VP-adverb, but not yet as a sentence adverb or subordinator.

 Henrik Rosenkvist

. From adverb to conditional subordinator In the end of the 17th century, or in the first decade of the 18th century, bara appears as a sentence adverb; Wijk-Andersson (1991: 69f.) assumes that bara developed from adverb to sentence adverb during the 17th century, but only presents two examples of this new usage (1991: 57f.). In both of them the form of the gram is barast, and the source is the poet Lucidor (ca. 1670, 1672). Apart from these examples, a few instances of a sentence adverbial bara can be found around the shift 1600-1700. The example below is from Stålhammar (1704): (5) Nilss [...] låtter helsa sin hustru och bara ber att hon måtte skrifva honom till. Nils [...] lets greet his wife and only asks that she may write him to “Nils [...] sends regards to his wife and only asks that she may write to him”

The change from adverb to sentence adverb entails increase of syntactic scope, a type of change described by Tabor & Traugott (1998: 256; cf. also Roberts & Roussou 1999; Abraham 2001; Traugott & Dasher 2002: 152ff.), who investigate the emergence of Present Day English anyway and a few other similar grams. It is thus well known that adverbs may change into sentence adverbs, through an expansion of the syntactic scope. As for bara, the structural possibilities can be illustrated by the following examples, in which the underlined segments represent the scope of the corresponding English adverb only (cf. Wijk-Andersson 1991: 120; Abraham 2001: 40f.): (6) a. b. c. d. e.

I can see only Calvin (and noone else). I can only see Calvin (and noone else). I can only see Calvin (I can not hear him). I can only see Calvin (I can not see Hobbes). I can only see Calvin (but Mary can see Hobbes)

No single part of the clause in (6e) is stressed, an indication that it really is the entire clause that forms the focus of only, and thus only here must be classified as a sentence adverb (cf. Trask 1996: 251). In the beginning of the 18th century, bara could be used in precisely the same manner as only in (6), which implies that bara at this time also was (and still is) a sentence adverb (SAG6 4: 84; WijkAndersson 1991: 169). The next step in the process takes place in the first decades of the 18th century, when the sentence adverb bara appears in a new meaning:

Modern Swedish bara 

(7) a.

Lät bara min Sissa Cammar-Piga komma fram. (Argus, 1732) let only my Sissa chamber-maid come forwardVPL “just let my chambermaid Sissa approach” b. Bara lyd mig åt (SAOB, 1741) only obey me atVPL “just obey me” c. Gå dumma Höna – bara gå. (SAOB, 1793) go stupid hen – just go “leave stupid hen – just leave”

Here we find that bara has acquired a new, not strictly exclusive meaning, which may be called desiderative (cf. Trask 1996: 80), since it underlines a wish or desire. Wijk-Andersson (1991: 82) considers, according to her Table 3.4.3.I, this use of bara to be modal, and she asserts that this is a new meaning of bara that is developed in the 18th century. The desiderative sentence adverb bara probably emerged when bara (in the meaning ‘only’) was used as a sentence adverb in clauses expressing desires (imperatives, desiderative exclamations etc.), whereupon it was reinterpreted by speakers who assumed that the desiderativity in these clauses was dependent on bara, a typical case of pragmatic inferencing (Hopper & Traugott 1993: 63ff.). The final stage in the development, the change into a conditional subordinator, seems to take place at about the same time, or shortly afterwards. The very earliest example of a conditional bara (8a) can be found in a protocol from the Swedish House of Knights, in an utterance of count Lewenhaupt. Most early examples of conditional bara, such as (8b–8d), actually appear in spoken contexts, it seems. (8b) is a comic comment in the colloquial magazine Argus and (8c) is spoken line in a comedy (in the context, which otherwise is ridicously highbrow, it is clearly informal). Also (8d) is written as direct speech. In her samples from Argus, Wijk-Andersson (1991: 77) finds eight instances of bara, of which four come from a chapter about the folly of servants (like 8b). She points out (1991: 78, 83) that in the 18th century bara did indeed belong to contexts of ordinary life; it is at this time unceremonious and down-to-earth, a statement confirmed by SAOB (B 3476). (8) a.

Vij skrida til vahlet, bara vij intet taga någon utom Riddarhuset. (SAOB, 1726) we go to election-the only we not take anyone outside House-ofKnights-the “we proceed to elections, if we just do not take anyone outside the house of lords”

 Henrik Rosenkvist

b. Gerna, bara hon intet will wara olåtig: Hon skal kunna gladly, only she not will be unpractical she shall be-able-to giöra god Ragout utan Champignon [...] (Argus, 1732) make good ragout without mushroom [...] “with delight, if she just is not unpractical. She must be able to make nice ragout without mushrooms [...]” c. det bästa är at han bär hiertat tämeligen lågt i böxorne, the best is that he carries heart-the quite low in trousers-the så at bara jag ser på so that only I look at mitt steke-spett här, så kusar han som en hund. (Gyllenborg, 1737) my skewer here then crouches he like a dog “the best thing is that he carries his heart quite low in his trousers, so that if I just glance at my skewer here, then he crouches like a dog” d. [...] bara jag får se er så glömmer jag bort altihop. (SAOB, 1786) [...] only I may see you then forget I awayVPL everything “if I just get to see you, then I forget everything”

In (8c, d), bara appears with the properties of a typical conditional subordinator – it is clearly the initial word in the protasis, and the apodosis is headed by så (then). An explanation for this development may be that as a desiderative sentence adverb, bara typically appeared in non-assertive or hypothetical clauses; one can not wish a fact. Such clauses often occur juxtaposed with the predicted consequence of the fulfilled desire – in (7a) the chamber maid is for instance expected to reveal to latest gossip about another noble lady. Given that bara appeared in hypothetical clauses that sometimes were followed by a clause where the desired result was presented, the two basic semantic features of a conditional sentence were present: hypotheticality and causality (cf. Comrie 1986). Hypotheticality is a necessary feature of desiderative clauses, and causality is often inferred from a temporal sequence (post hoc ergo propter hoc). From a more general perspective, we know that adverbs often are recruited as subordinators (like after, before and since), and we know that modal expressions may serve as sources for conditionals (Hopper & Traugott 1993: 179). Indeed, the Mandarin yào can, according to Hopper & Traugott, convey several related meanings: ‘wish, be necessary, if ’. The semantic resemblance to desiderative bara is obvious. These circumstances all suggest that it must have been a relatively small step for the language user to start using the desiderative sentence adverb bara also as a conditional subordinator.

Modern Swedish bara

The chronology of the development of conditional bara is a bit unsatisfactory, from a language historian’s point of view. When working with texts from such a late period, one would expect to find clear examples of all the relevant stages, neatly following each other and eventually leading to the goal, the conditional subordinator. In the case of bara, the really crucial stages and the goal nonetheless appear simultanously. But I still propose that there is a successive development, and I think that the reason why it is not observable is that bara initially was used only in colloquial speech, and hence belonged to a style that was not written in any extent until it was too late – the change had already taken place. Subordinating bara has furthermore never been frequent, as it has been competing with other conditional subordinators such as the default om and its more formal counterparts allena and blott. The scarcity of conditional bara makes the change even harder to observe.

. Some other relevant grams The topic of Wijk-Andersson’s dissertation (1991) is bara and its equivalents blott, enbart, endast, uteslutande and allena/allenast. Here, allenast and blott are our main interest, since they have been regularly used as conditional subordinators in Swedish.7 Also endast was used as a conditional subordinator for a short period, and it will be studied in comparison with enbart, which did not evolve into a conditional. Allena can be found in EOSw, and as English alone it was originally a compound of the parts all and ena (‘one’), meaning ‘solo’, like the German allein. In EOSw (Söderwall I:27), allena could, either as an adjective or an adverb, also mean ‘only’. Allenast is according to SAOB (A 995) an inflected form of allena. In OSw, it had the meaning ‘only’ (Söderwall I:27). Interestingly, allenast, but not allena, developed into a conditional subordinator: (9) Om någon [...] wil i Fastan gifta sig, bör sådant icke wägras, allenast thet if anyone [...] want in Lent marry refl should such not denied-be, only it skeer i stillheet, och utan alt Brudebång. (SAOB, 1686) happens in silence and without all bridenoise. “if anyone [...] wants to marry during Lent, then that should not be denied, if it just occurs in silence and without any marital noise”

The pair allena/allenast, which have precisely the same etymological background, thus offers an interesting asymmetry.



 Henrik Rosenkvist

Likewise, the pair enbart and endast developed differently, although they have very similar backgrounds. Enbart is, according to SAOB (E 583) the adverbial form of the adjective enbar. The adjective is, in its turn, originally a compound of the numeral en (‘one’) and the adjective bar, which of course also was the origin of bara. Like the other grams in this group, the adjective enbar developed an exclusive meaning ‘only’, and an adverbial, in the form enbart, was used with the same meaning in the end of the 19th century. Endast is also the offspring of an adjective, ende (SAOB, E 588), in the same pattern as barast and allenast. The original meaning of ende is probably ‘alone’, but as one of its submeanings ende meant ‘only’. In contrast to enbart, endast developed further. In the end of the 18th century (the example below is from SAOB, dated 1781), we find endast in use as a conditional subordinator: (10) Vi hafve intet så ondt om Patrioter endast de hinna med blifva kända. we have not so poorly of VPL patriots only they manage with VPL become known “we have no shortage of patriots, if they just get time to become known”

The adjective blott was borrowed from Low German (bloß) in the Middle Ages (Hellquist 1980: 81). In OSw it had meanings like ‘naked’ or ‘defenseless’ (Söderwall I:127), but in EMSw it could mean ‘only, mere’. Later, an adverbial blott shows up, also in the meaning ‘only’, and in the early 18th century, we find a conditional combination blott att. About half a century later, blott can be used as a bare conditional subordinator: (11) Han är nögd, blott han får detta (SAOB, 1773) he is satisfied only he receives this “he is satisfied, if he just receives this”

So, these quite similar grams (bara, blott, allena, allenast, endast and enbart) all acquired exclusive meaning, but only four of them developed further into conditional subordinators. A closer study reveals what appears to be the key factor: bara, blott, allenast and endast could actually also be used as desiderative sentence adverbs. Hence we find a very strong correlation between desiderative and conditional meaning. Although I have not been able to establish an undisputable chronological development for all of these grams at this stage, I find it plausible that there is a path of meaning change from exclusive to desiderative and further on to conditional meaning.8 I also think that the reason why desiderativity is crucial is that desiderative clauses may be the only context in which grams of this type

Modern Swedish bara 

Table 1. Correlation between desiderative and conditional meaning gram

exclusive

desiderative

conditional

bara blott

yes yes

yes yes

yes yes

allena allenast

yes yes

no yes

no yes

enbart endast

yes yes

no yes

no yes

may acquire the semantic feature of hypotheticality. Hypotheticality, in turn, is an essential meaning component of conditional sentences (Comrie 1986).

. The changes of bara: a case of grammaticalization? The development of bara consists of semantic as well as syntactic changes. In this section, both of these aspects are discussed from the perspective of grammaticalization theory. I also comment upon the mechanisms of change that are involved in the evolution of bara. . The semantic aspects of the proposed path of change Some of the most prominent semantic components of grammaticalization theory are semantic retention and source determination (Bybee et al. 1994: Chapter 1), abstraction (Heine et al. 1991: 48) and subjectification (Traugott 1980, 1982, 1989, 1995; Traugott & Dasher 2002: 89ff.). Here I investigate whether the changes of bara fit into these models of change. As was mentioned in the introduction, conditional bara signals explicit suffiency (‘if just X, then Y’). The origin of this meaning component is, I think, the basic meaning of bara (‘uncovered’, ‘bare’), which, arguably, contains a sense of negation. SAG (IV: 161ff.) claims that a streak of negativity also is present in the exclusive as well as the desiderative meanings, which both can be paraphrased as ‘no more than’. It is probable that this feature is present also in the conditional subordinator, resulting in the explicit sufficiency of MSw bara – if so, this is an instance of both semantic retention and source determination. Bara also follows Heine et al.’s proposed path of change (1991: 48); as an adjective bara modifies Objects, as an adverb it modifies Activities, and as a

 Henrik Rosenkvist

subordinator I assume it must be categorized as an item that belongs to the “most fuzzy” (Heine et al. 1991: 49) of these categories, i.e. Quality. Turning to subjectification, it again seems that the changes of bara proceed as could be expected (from this particular perspective). In Traugott’s original version of the concept (1980: 51f., 1982, 1989),9 subjectification was seen as a shift of functional-semantic meaning type (from propositional to textual to interpersonal meaning). Adjectival bara does indeed belong to the propositional level, expressing a lexical meaning. The adverbial bara falls into the textual level, however, and as the adverbial bara acquires the speaker oriented meaning of desiderativity, it also changes functional-semantic status, from textual to interpersonal level of meaning. All in all, the semantic aspects of the changes of bara indicate that we are dealing with an uncontroversial example of grammaticalization. . The syntactic aspects of the proposed path of change The predominant view on structural change within grammaticalization theory is, at least since Givón (1979: 209), that grammaticalization involves a decrease of syntactic scope: “The structural scope of a sign decreases with increasing grammaticalization” (Lehmann 1995: 143). A change of the type described here can then not be considered to be a case of grammaticalization, as both adverbs and subordinators have wider scope than adjectives. But it has been suggested that also increase of structural scope may be a possible property of grammaticalization; Tabor and Traugott (1998) and Roberts and Roussou (1999) both propose (although working within quite different linguistic frameworks) that a scope increasing change indeed can be classified as grammaticalization – Roberts and Roussou actually claim that this is the defining feature of grammaticalization, from the syntactic point of view. These conflicting assumptions concerning scope increase/decrease may be caused by an oversight; in grammaticalization theory, much attention has been paid to the semantic aspects of the investigated changes (cf. Heine et al. 1991; Bybee et al. 1994), as grammaticalization has been seen as loss of lexical meaning intertwined with acquisition of grammatical meaning. However, not only items with narrow scope, such as affixes and clitics, express grammatical (i.e. non-lexical) meaning, but so do items with clausal scope: modal verbs, subordinators, coordinators and sentence adverbs (cf. Abraham 2001). There should thus be nothing remarkable about a grammaticalization process in which the gram, like bara, goes through an increase of syntactic scope during the semantic changes.

Modern Swedish bara 

. The mechanisms of change My intention has been to describe both the syntactic and the semantic aspects of the evolution of conditional bara. In order to reach a deeper understanding of the triggering factors and of the active mechanisms of change, a theory which encompasses both of these aspects and which does not exclude structural scope increase is necessary.10 In the pursuit of such a theory, it is reasonable, I think, to depart from the observation that the changes can be seen as interacting structural reanalyses and semantic reinterpretations. Furthermore, the reanalyses do not seem to alter the meaning of bara, and the reinterpretations, conversely, do not affect the structural status of bara. I propose that the crucial factors in these changes are (structural) ambiguity and (semantic) vagueness.11 As an adjective, bara is vague; it can mean ‘uncovered’ and ‘only’, and a number of intermediate meanings. In many cases, it is not easy to specify the exact difference between these meanings. The initial reanalysis (adjective > adverb) dissolved the vagueness, since bara as an adverb solely means ‘only’. However, as an adverb bara could either be analyzed as a regular (VP-)adverb or as a sentence adverb – the adverbial bara (meaning ‘only’) is thus structurally ambiguous. The following change, when desiderative meaning rubbed off on bara through pragmatic inferencing, in turn resulted in an unambiguous sentence adverb, since desiderativity clearly is a sentential type of meaning. But also this stage of development led on to further change. Bara could, as a desiderative sentence adverb, be interpreted as a conditional gram in the proper contexts – again this is a question of vagueness, I think, since the semantic difference between a conditional clausal relation and a sequential/temporal clausal relation in many cases is hard to capture accurately. The final reanalysis (sentence adverb > subordinator ) resulted in an unambiguous, non-vague conditional subordinator.12 So, I suggest that the evolution of bara can be seen as interacting semantic reinterpretations, facilitated by structural ambiguity, and structural reanalyses, possibly triggered by vagueness. This perspective of grammaticalization also have implications for the formal/functional debate (cf. Newmeyer 1998; Haspelmath 1998) concerning gradual and non-gradual change. Most work within grammaticalization theory has focussed on semantic changes, and if one of the key factors in semantic change is vagueness, as suggested here, then it is natural to assume a principle of gradualness, I think. Vague meanings can not be easily delineated or categorized – they are rather points on a scale than distinct alternatives of in-

 Henrik Rosenkvist

Table 2. The evolution of bara seen as interacting reinterpretations and reanalyses

terpretation. On the other hand, a structural analysis of a sentence is rarely vague. Consider the sentence below: (12) Calvin gave the farmer his money.

Example (12) offers two analytical options; either his refers to Calvin or to the farmer. There are thus two distinctly different structural analyses available (with two separate meanings, in this case). The alternatives are non-gradual – there are no intermediate analyses – and the sentence is thus structurally ambiguous. To me, it seems likely that the syntactic domain of language is organized in discrete categories, while the semantic domain is “fluid, gradient and variable” (Bybee 2003). Accordingly, syntactic change will generally be abrupt, whereas semantic change will be gradual. In the model sketched here, non-gradual syntactic changes (reanalyses) accompany gradual semantic changes (reinterpretations), and there is no opposition between gradual and non-gradual change – the semantic and syntactic changes interact and enable further change throughout the evolution of bara. It is plausible, I think, that similar forms of semantic/syntactic interaction have been active in other languages and in other cases of grammaticalization. A perspective of grammaticalization which includes both of these aspects of language is a prerequisite for future work along these lines.

Notes . The content of this paper has been presented at the Danish grammaticalization Network and at ICHL 16 in Copenhagen. I wish to express my gratitude for valuable comments to

Modern Swedish bara 

Henning Andersen, Cecilia Falk, Lars Heltoft, Muriel Norde, Jens Nørgård-Sørensen and Eva Skafte Jensen. . The lexicon of the Swedish Academy. . Swedish is traditionally divided into the following chronological periods: Runic Swedish 800–1225, Early Old Swedish 1225–1375, Late Old Swedish 1375–1526, Early Modern Swedish 1526–1732, Late Modern Swedish 1732–. . Another possible explanation for the adverbial bara might be that it has been developed from the adjective bar through derivation. In the old Nordic languages, a derivating suffix -a could be employed when adjectives were turned into adverbs (Hanssen et al. 1975:86). Both ill-a and gärn-a are examples of such derivation (cf. Wijk-Andersson 1991:52f.). However, it seems as this kind of suffixation was rather rare, and it is very unlikely that it was in use at the time when adverbial bara appears (SAOB’s first examples of adverbial bara are from the end of the 16th century). Therefore the aforementioned hypothesis must be preferred. . VPL indicates that ut (‘out’) here is used as a verbal particle. . The Swedish Academy Grammar. . Allena is not used as a conditional in contemporary Swedish, but conditional blott may still appear in formal, poetic or archaic contexts. . Eckardt (2003; cf. also Abraham 2001) has shown that also German bloß can be used both as an exclusive and as a desiderative adverb. . In later work (Traugott 1995; Traugott & Dasher 2002:89ff.), Traugott has abandoned the theoretical foundation of subjectification, i.e. the hypothesis presented by Halliday & Hasan (1976), but the actual concept of subjectification remains more or less unaltered. . One such theory is Hopper & Traugott’s (1993:88) model of grammaticalization, in which they consider grammaticalization to be a succession of changes that either proceed by reanalysis/metonymy or by analogy/metaphor. However, I do not find this model to be applicable in this particular case. . Tuggy (1993) discusses vagueness/polysemy as well as ambiguity/homonymy; here I transfer his view of ambiguity to the structural domain of language. . However, since all meanings and functions of bara still are in use in MSw, there are of course constructions in which bara still is either ambiguous or vague. But in some cases bara is an indisputable conditional subordinator.

References Abraham, Werner (2001). “How far does semantic bleaching go: About grammaticalization that does not terminate in functional categories”. In Jan Terje Faarlund (Ed.), Grammatical relations and change (pp. 15–64). Amsterdam/Philadelphia: John Benjamins. Bybee, Joan, Revere Perkins, & William Pagliuca (1994). The evolution of grammar. Chicago/London: The University of Chicago Press.

 Henrik Rosenkvist

Bybee, Joan. (2003). “The development of the category of auxiliary in English”. Talk presented at ICHL 16, Copenhagen, August 2003. Comrie, Bernard (1986). “Conditionals: A typology”. In Elizabeth Closs Traugott, Alice ter Meulen, Judy Snitzer Reilly, & Charles A. Ferguson (Eds.), On conditionals (pp. 77–102). Cambridge: CUP. Declerck, Renaat, & Susan Reed (2001). Conditionals. A comprehensive empirical analysis. Berlin/New York: Mouton de Gruyter. Eckardt, Regine (2003). “A two-way pathway of change”. Talk presented at ICHL 16, Copenhagen, August 2003. Givón, Talmy (1979). On understanding grammar. New York/San Francisco/London: Academic Press. Halliday, M. A. K., & Ruqaia Hasan (1976). Cohesion in English. London: Routledge. Hanssen, Eskil, Else Mundal, & Kåre Skadberg (1975). Norrøn grammatik. Oslo/Bergen/Tromsø: Universitetsforlaget. Haspelmath, Martin (1998). “Does grammaticalization need reanalysis?”. Studies in Language, 22 (2), 315–351. Heine, Bernd, Ulrike Claudi, & Friederike Hünnemeyer (1991). Grammaticalization. A conceptual framework. Chicago/London: The U. of Chicago Press. Hellquist, Elof (1980). Svensk etymologisk ordbok. Lund: Liber. Hopper, Paul, & Elizabeth Traugott (1993). Grammaticalization. Cambridge: CUP. van Kemenade, Ans (1999). “Functional categories, morphosyntactic change, grammaticalization”. Linguistics, 37, 997–1010. Lehmann, Christian (1995). Thoughts on grammaticalization. München/Newcastle: Lincom Europa. Newmeyer, Frederick (1998). Language Form and Language Function. Cambridge, MA: MIT Press. Pfeifer, Wolfgang (1989). Etymologisches Wörterbuch des Deutschen, I–III. Berlin: AkademieVerlag. Roberts, Ian & Anna Roussou (1999). “A formal approach to ‘grammaticalization’ ”. Linguistics, 37, 1011–1042. SAG (Svenska Akademiens Grammatik), I–IV (1999). Ulf Teleman, Staffan Hellberg & Erik Andersson. Stockholm: Svenska Akademien. SAOB (Svenska Akademiens Ordbok). 1873–. Lund: Svenska Akademien. Sweetser, Eve (1988). “Grammaticalization and semantic bleaching”. Berkeley Linguistic Society, 14, 389–405. Sweetser, Eve (1990). From etymology to pragmatics. Cambridge: CUP. Söderwall, Knut F. (1884–1973). Ordbok över svenska medeltidsspråket, I–IV. Lund: Berlingska Boktryckeriet. Tabor, Whitney, & Elizabeth Traugott (1998). “Structural scope expansion and grammaticalization”. In Anna Ramat & Paul Hopper (Eds.), The limits of grammaticalization (pp. 229–272). Amsterdam/Philadelphia: John Benjamins. Trask, Robert (1996). A dictionary of grammatical terms in linguistics. New York: Routledge. Traugott, Elizabeth (1980). “Meaning-change in the development of grammatical markers.” Language Sciences, 2, 44–61.

Modern Swedish bara 

Traugott, Elizabeth (1982). “From propositional to textual and expressive meanings: Some semantic-pragmatic aspects of grammaticalization”. In W. Lehmann & Y. Malkel (Eds.), Perspectives on Historical Linguistics (pp. 245–271). Amsterdam: John Benjamins. Traugott, Elizabeth (1989). “On the rise of epistemic meanings in English: An example of subjectification in semantic change”. Language, 65, 289–307. Traugott, Elizabeth (1995). “Subjectification in grammaticalization”. In D. Stein & S. Wright (Eds.), Subjectivity and subjectivisation (pp. 31–54). Cambridge: CUP. Traugott, Elizabeth, & Richard Dasher. (2002). Regularity in semantic change. Cambridge: CUP. Tuggy, David (1993). “Ambiguity, polysemy and vagueness”. Cognitive Linguistics, 4 (3), 273– 290. Wijk-Andersson, Elsie (1991). Bara i fokus. Uppsala University: The Dep. of Scandinavian languages.

Nordic prefix loss and metrical stress theory with particular reference to *ga- and *biMichael Schulte* Agder University College

.

Introduction: aim and focus

A peculiar trait of Nordic is its almost total absence of cohering prefixes such as Gmc. */ga/ and */bi/.1 Although the evolutionary trend of Old Germanic was to expand the productive pattern of unstressed preverb plus verb, this type of word formation is poorly attested in North Germanic, before being partly reintroduced from Middle Low German in the late Middle Ages. It is not entirely clear whether this lack is generally due to genuine loss, but it may safely be assumed on the basis of different reflexes in the Scandinavian languages that deletion processes were involved; see e.g. the seminal paper of Wadstein (1895) for the reflexes of Gmc. */ga-/ in Old Norse (e.g. ON greiða ‘manage’ in relation to Gothic garaidjan). Judging from the morphophonological, lexical-semantic and syntactic legacy of preverbs, they must once have been present in Nordic (see Christiansen 1960, also Schulte 2003). Though productive in all the other Germanic languages, these preverbal structures were obviously disfavoured in Nordic and came to a decline. One might thus ponder whether tawido and worahto in the Scandinavian runic inscriptions of Gallehus (KJ 43) and Tune (KJ 72) reflect earlier Gmc */ga-tawiðf:n/ and */ga-wurhtf:n/; cf. Runic OE gewarahtæ on the Mortain Casket and Runic WGmc. gasokun on the beltbuckle of Pforzen (cf. Sections 2 and 3 below). The aim of this paper is to shed light on the motivating factors of Nordic prefix loss and the overall disinclination to preverbation, basically from a metrical perspective. As the runic data signal two distinct phases of prefix loss, a binary weight distinction is called for at the outset (Section 2). After investigating the meagre and partly ambiguous runic evidence in Section 3, I shall

 Michael Schulte

draw a distinction between regular phonetic reduction on the one hand and rhythmic-metrically based deletion processes on the other (Section 4). To this end, ‘Prosodic Repair Strategies’ in the sense of Armin Mester (1994) will be discussed for different languages such as Latin, Modern Norwegian, Modern German and – last but not least – Ancient Nordic; the relevant data shall be re-evaluated with recourse to foot structure (Section 5). The basic claim to be made is that the deletion of an unmetrified monosyllabic domain at the left edge of the word functions as a central repair in Nordic (Section 6). This technique of ‘prefix-stripping’ may be compared to other rhythmically conditioned repair strategies involving stress shifts and lengthening processes, thus providing several means of integrating light prefixes into the foot parse, or alternatively disintegrating them.

. The binary weight distinction In the absence of any indication to the contrary, I assume that cohering prefixes of the /ga/-and /bi/-type had already vanished at the time of the earliest runic inscriptions, i.e. around 150/200 AD (see Section 3). By contrast, heavy preverbs such as */unθ-/ were resistant to deletion until the syncope period of the 500s. A heavy preverb is retained in the Reistad form (KJ 74) unnam which is generally regarded as the continuation of AN */unθ-nam/ (or */unð-nam/), 1 sing. pret. ind. of */unθ-neman/, semantically ‘undertake’ vel sim. (e.g. Grønvik 1996: 114–121; cf. also Bammesberger 1996). The Reistad inscription is roughly datable to the 400s which is the final phase of Early Runic.2 If trimoraic or super-heavy syllables in pre-tonic position are excluded, the basic dichotomy is as follows (see (1)). (1) Binary weight distinction: unstressed prefixes i. light, i.e. monomoraic type (e.g. */ga/) ii. heavy, i.e. bimoraic type (e.g. */unθ/)

As mentioned, this distinction is of vital importance to the chronology of prefix loss in Nordic, since the light type (e.g. */ga/) appears to be subject to deletion at an earlier stage than the heavy type (e.g. */unθ/). As shown in Section 3, this assumption is directly corroborated by the older runic inscriptions of Scandinavia.

Nordic prefix loss and metrical stress theory 

. Early runic evidence As a matter of fact, the older runic inscriptions of Scandinavia exhibit no secure instances of */ga-/ nor of any other cohering prefix displaying a CVstructure (e.g. */bi-/). The short list of potential candidates is centred on the inscription on the Kragehul spearshaft (KJ 27) which partly defies a definite reading and interpretation (cf. Marstrander 1952: 31; Makaev 1996: 87, 105). To begin with Ottar Grønvik (1996: 51–55) who interprets the Kragehul sequence gaë gaë gaë (consisting of three bind-runes ga) as /ga-ga(n)ga/ ‘follower’ with retained *ga- in a sociative function. Similarly, Elmar Seebold (1995) ë suggested that gaë gaë gaginu gaë contains a past participle formation */ga-ganga/ (cf. OE gegang ‘chance, event’), followed by a theoretical ga-extension to AN */gin:u-/, which he rendered as ‘magisches Schicksal’.3 Although these two interpretations cannot be dismissed solely on graphic grounds, the sequence ë gaë gaë gaginu gaë is hardly readable as a meaningful unit of words.4 Rather, it represents an ideographic expression or a formula, and main stream research has agreed that the repeated ligature or bind-rune for g + a (or a + g, as Antonsen 1975: 36 would have it) most likely is some sort of votive or protective charm (e.g. Krause 1966: 66; Makaev 1996: 56). As Tineke Looijenga (2003: 114) puts it, “texts [such as the Kragehul inscription; M.S.] do not create the impression of being just everyday message, but seem instead to have some supernatural connotation”. Consequently, both Krause (1971: 152–153) and Makaev (1996: 38) stress that any linguistic interpretation of the triple bind gaë is bound to be doubtful. ë Turning to the alleged sequence ginugaë helija in the Kragehul inscription (runes 35–46), it may first be noticed that there is a lacuna between the bindë rune heë and the following sequence lija; hence the transliteration gaë he|lija. ë Following Theodor von Grienberger’s reading of a term gaë helija (von Grienberger 1907), Adolf Noreen (1923: 381) identified this sequence as an adjectival ija-formation of a strong verb to be equated with OHG, OS gihellan, gehellen vb. ‘sounding jointly, resounding loudly’ (“sehr helltönend”) with sociative *ga-; the preceding part was analyzed as an intensifier as paralleled by ginoronoR and ginA-runAR of the Björketorp and Stentoften inscriptions respectively; cf. ON ginnregin ‘the mighty powers’ and ginnheilagr ‘extremely holy’, also Mod.Norw. ginsk ‘extraordinary, lively’ and gjøngod ‘extremely good’.5 But ë the obvious weakness of this proposal lies in the root vocalism of gaë helija = /gahel:ija/ which is rather unexpected. As Krause (1966: 66) points out, the raising (or ‘vertical i-umlaut’) [ÁeC(C)i] > [ÁiC(C)i] should have occurred at that

 Michael Schulte

stage of Runic; cf. KJ 64 Barmen þirbijaR from Gmc. */θerbijaz/; for the stem */θerb-/, cf. ON þiarfr ‘unleavened’, also Mod.G derb ‘rough, sturdy’.6 Summing up, there is no proof of */ga-/ nor any other light prefix in Early ë haë Runic. Wolfgang Krause (1971: 152) therefore comments on the form mu of the Kragehul inscription that it might stem from */ga-mu:ha-/ ‘follower’ with lost *ga-, a proposal already made by Theodor von Grienberger (1907). In addition, the final part of the sequence niuhAborumR in the Stentoften inscription was once equated with OHG gib¯uro, gib¯ur < WGmc. */ga-bu:ra-/ masc., lit. ‘the one dwelling at the same farmstead’, hence ‘farmer’ (Lindroth 1918; Krause 1966: 211; 1971: 164f.). Although the o-spelling of -borumR does not necessarily run counter to this interpretation (cf. ginoronoR for */gin:uru:nf:R/ in the same inscription), a corresponding ga-formation seems to be entirely absent in North Germanic and the semantic shift toward ‘farmer’ (G Bauer) is likely to be a single-language development within West Germanic, not being paralleled by Nordic (e.g. Kluge 1995: 86; Paul 2002: 138). Moreover, the segmentation niuhA#borumR is a matter of dispute, and other interpretations seem equally viable, e.g. niu#hAborumR (for research report see Schulte 1998: 121–122). Thus, we are solely left with negative evidence regarding light prefixes in Early Runic, in particular */ga-/, and Runic WGmc. gasokun on the belt-buckle of Pforzen has no counterpart in Nordic.7

. Rhythmic-metrical deletion versus phonetic reduction Prefixes’ lack of stress alone cannot be a sufficient motivation for losing them entirely. Under this premise, it may be argued that Nordic prefix loss exceeds regular phonetic reduction, or metaphorically speaking, the ‘erosion’ of a-tonic phonetic material. At least two arguments in support of this view shall be considered here. First, the Nordic process of prefix-stripping exceeds the purely phonetic reduction of pre-tonic syllables, as evidenced for instance by Old English runic inscriptions. In general, Old English exhibits many instances of unstressed bi-, gi-, being susceptible to weakening: bi-, gi- > be-, ge- (e.g. Campbell 1959: §§265, 368–370; cf. also Lutz 1997). The weakened variant ge-, for instance, shows up in 8th- and 9th-century inscriptions (see (2)).

Nordic prefix loss and metrical stress theory 

(2) The graphic variants versus in early OE runic inscriptions

Falstone Hogback geb[id]æd Lancaster Cross gibidæþ Great Urswick Stone gebidæs Franks Casket Front side giswom, Left side gibroþær Mortain Casket gewarahtæ Ruthwell Cross NW giwundad, gistoddu[.] Thornhill Stone III gebiddaþ Harford Farm Brooch gibœtæ

This weakening process probably started around 750 AD, the earliest possible date for the Thornhill Stone III where gebiddaþ occurs (for the dating of the above-mentioned runic inscriptions, see Page 1999: passim). The graphematic representation definitely signals lowering or even centralization, whereas either involves an unreduced variant or a conservative spelling with a reduced value. Most important, this gradual weakening of pre-tonic vowels is not traceable in the older runic inscriptions of Scandinavia, whereas weakening processes of post-tonic syllables are observed in the syncope period (see Schulte 1998: 87–158). Second, the disappearance of unstressed prefixes in Nordic goes beyond regular right-hand reduction. In particular, the loss of heavy preverbs such as */unθ-/ (as still attested in KJ 74 Reistad un-nam) exceeds the reduction of post-tonic syllables at least by one mora (consider (3a) and (3b)). Thus, reduction under (3a) involves the deletion of two moras, whereas reduction under (3b) only affects one mora, hence deletion by one mora. (Note that final -z in AN */ferðunz/ may be regarded as extrametrical.) (3) Left-hand versus right-hand reduction in Nordic a. AN *[unθ.Ánam] > ON nam ‘took, undertook’ (1st and 3rd sing. pret.) σµµ σµµ (x) (X) b. AN *[Áfer.ðunz] > ON fio˛rðu ‘fjords’ (acc.plur.) σµµ σµµ (X) (x)

The data mentioned under (2) and (3) strongly indicate that apart from the general condition of unaccentedness, an additional factor is required to account for deletion processes at the left edge of the word. Provided that the foot construction plays a prominent role, Nordic prefix loss may be governed by rhythmic-metrical factors. Accordingly, I shall explain prefix-stripping in terms of a ‘Prosodic Repair Strategy’ in the following sections.

 Michael Schulte

. Prosodic Repair Strategies The basic claim to be made here is that the early abandonment of monomoraic unstressed prefixes in pre-syncope Nordic is rhythmic-metrically conditioned. To minimize irregularities in performing well-formed metrical structures, the notion of ‘Prosodic Repair Strategies’ has been introduced on a languagespecific basis (for the quantitative trochee in Latin, see Mester 1994). In a seminal paper on “Quantitative consequences of rhythmic organization”, Alan Prince (1990) argued that optimizing the foot parse is a prime motivation of metrical restructuring. Transferring this approach to Early Nordic, the exemption of pre-tonic material from the foot construction results in light syllables being ‘unfooted’ or ‘stranded’. Such unmetrified material may be subject to outright deletion (see Section 6). To begin with the prosodic effects of Latin prefixation, as analyzed by Armin Mester (1994) in the wake of Max Niedermann (1908). . Latin In Latin, optimal footing may be regarded as a prime motivation of metrical restructuring and the ideal output structure is a canonical bimoraic trochee. Thus, a series of morphophonological processes are triggered by the imposition of the canonical bimoraic foot. Analyzing morphophonological structure in Latin, Mester (1994: 15) focuses on what he calls ‘anti-trapping effects’ to integrate metrically isolated, or ‘trapped’ syllables into the foot structure (see (4)). (4) Basic Repair Strategies in Latin i. lengthening processes (affecting either the syllable in question or another domain) ii. outright deletion of unfooted syllables iii. shortening processes of a following heavy syllable (in particular Iambic shortening)

Of particular interest are initial trapping phenomena in connection with prefixation. The repair of unfooted material in Latin is bound up with the bimoraic condition, otherwise known as Prokosch’s Law (cf. Vennemann 1988: 30), i.e. the realization of stress requires two moras. The point is that preverbs, when integrated into the foot construction, modify the entire rhythmic pattern. Consider, for instance, Latin verb roots with short thematic -˘ı-, alternating with long -¯ı- in primary verbs due to prefixation (see (5)).

Nordic prefix loss and metrical stress theory 

(5) Rhythmic alternation of theme vowels: Latin [-˘ı-] ∼ [-¯ı-] a. -˘ı[Ápari-mus] ‘we bring forth’ [Ásapi-mus] ‘we taste’ óìóì F

b. -¯ı[repeÁri˜-mus] ‘we find’ [resiÁpi˜-mus] ‘we taste of ’ óìóìóìì F1 F2

c. -˘ı[de˜Ásipi-mus] ‘we are out of our mind’ óììóìóì F1 F2

The long variant of the theme vowel -¯ı- occurs when the light preverb r˘eis attached, hence repeÁr¯ımus, resiÁp¯ımus. The lengthening strategy enables the penult of these forms to bear a bimoraic foot, thus providing the dockside for the Latin stress rule. However, if the preverb itself is bimoraic as Latin d¯e-, the short theme vowel -˘ı- shows up again, hence the alternation -˘ı- ∼ -¯ı- coupled with stress shift: repeÁr¯ımus, resiÁp¯ımus versus d¯eÁsipimus. Already Niedermann (1908) argued that this alternation was not etymologically based, but rather governed by prosodic factors – the only difference being the quantity of the prefix vowel. By way of conclusion, quantitative adjustments result in persistent footing, where [resi][Ápi˜] footwise contrasts with [de˜][Ásipi]. The avoidance of unfooted material turns out to be a basic mechanism in the phonology of Latin. . Modern Norwegian Modern Norwegian provides a particular case of stress shift in conjunction with the lengthening strategy. In general, the light cohering preverb be-, which is of German origin, receives no stress in Norwegian; cf. Mod.Norw. be-tone [b6.1 tu˜.n6] ‘accentuate’ and be-kjenne [b6.1 ç7n.n6] ‘confess’ (cf. Kristoffersen 2000: 180). In connection with a subsequent main-stressed stem syllable, therefore, be- cannot constitute the strong head of a stress foot. The absence of stress is signalled by the achromatic schwa value of [b6] which would be disallowed in any stressed surface syllable of Norwegian (e.g. Kristoffersen 2000: 184). However, one notable exception to this rule is provided by /be/-prefixation to the disyllabic stem arbeid-, resulting in the exceptional stress pattern of Mod.Norw. béarbèide as opposed to Mod.G beárbeiten (see (6)).

 Michael Schulte

(6) Stress Retraction Rule: exceptional prefix stress on beMod.Norw. béarbèide [1 be˜.an.b7i.d6] ‘work on something’

This form undergoes the Stress Retraction Rule, moving stress to the initial syllable. Arguably, the preverb forms an alliance with the second syllable [-an-] to be footed. Here, the foot parse is regulated by creating two canonical trochees. To render the initial syllable stressable, a strengthened variant [be˜] shows up, with primary stress being realized as long /e˜/. This repair strategy of initial vowel lengthening gives prominence to the originally light preverb, thus fulfilling the bimoraic condition by the weight-to-stress principle (cf. Section 5.1 for Latin). To conclude, natural parsing motivates a stress shift in Mod.Norw. bearbeide. When an unmetrified disyllabic domain is cyclically created at the left edge of the word, a stress foot may be built upon it to attract main stress (cf. also Kristoffersen 2000: 181). Despite its marginal status in the language system, Mod.Norw. bearbeide clearly illustrates the repair of an unfooted monomoraic syllable. Further Norwegian data conveying the notion of Prosodic Repair Strategies are discussed in Schulte (2003). . Modern Standard German Modern German provides cases of stress-dependent morphology. A number of deletion processes and stress shifts are due to the creation of a regular trochaic pattern (e.g. Wiese 1996: 306–310). In particular, a sequence of two unstressed syllables (e.g. word-initial [C6C(C)6]) is ruled out at some level of the German grammar. To meet this prosodic requirement, the preverbs ge- and beare partly dropped. Synchronically, this deletion rule implies that prefixal ge-, be- are only attached if main stress falls immediately on the following syllable (Kiparsky 1966: 70–75; cf. also Kaltenbacher 1999: 450). For this reason, prefixal ge- is partly missing in past participle formations in Modern Standard German (see (7)). (7) Rhythmic deletion of ge- in Modern Standard German Past participle formations: i. with ge-: gerédet, geárbeitet, gehéiratet ii. without ge-: befóhlen, versúcht, entfállen, servíert, diskutíert

Historically, a systematic morphological restructuring has occurred (see Paul 1917: §§201–203). In early New High German, some participles still lacked geon a regular morphological basis in simplex verbs, e.g. bracht, funden, kommen, troffen versus Mod.G gebracht ‘brought’, gefunden ‘found’, gekommen ‘come’,

Nordic prefix loss and metrical stress theory 

getroffen ‘met’. Diametrically opposed, verbs in -ieren formed a past participle with ge- in former times, but deleted it for prosodic reasons, e.g. geregíert, gedesertíert → regíert, desertíert (belonging to the infinitives regieren ‘rule’ and desertieren ‘desert’ respectively). According to Kiparsky (1966: 75), reflexes of preverbal be- are evidenced by transitivization. Valency change is evident in the case of Mod.G applaudieren ‘applaud’ which governs both the dative and the accusative case (see (8)). In this light, transitivity is the direct reflex of dropped be- in the original prefix verb *be-applaudieren. (8) Valency change in Mod.G applaudieren SIMPLEX: Er applaudierte dem Sänger [dat]. Be-TRANS: Er applaudierte den Sänger [acc]. “He applauded the singer.”

On the whole, the selection of ge- and be- in German is a clear case of affixation determined by stress characteristics. Occasional formations like Mod.G bediskutieren vb. ‘discuss at some length’, or Getelefoniere n. ‘act of telephoning (without any purpose)’, which Susan Olsen (1991) characterizes as marginally acceptable, show that phonology and morphology neatly interact, with ge- and be- being productive prefixes in German (cf. Wiese 1992: 132–133). The adequate generalization behind this selection process seems to be that prefixing is usually blocked when it does not meet the rhythmic-prosodic requirements of Modern Standard German.

. Nordic prefix loss as a Prosodic Repair Strategy The fundamental contrast with regard to prefixation between North Germanic on the one hand and West Germanic and Gothic on the other hand calls for an explication. Reverting to the basic claim, prefixes’ lack of stress is a conditio sine qua non, but no sufficient cause of their abandonment (cf. Section 4). Nordic prefix loss clearly exceeds regular phonetic reduction of suffixes and endings. To account for these discrepancies, one viable approach invokes prosodic factors with recourse to the foot parse (e.g. Hayes 1995). In view of optimal parsing, light pre-tonic syllables at the left edge of the word represent an unmetrified domain susceptible to deletion processes. Unlike Latin and Modern Norwegian, Early Nordic with its fixed root accent cannot make use of the Lengthening Strategy in conjunction with stress shift, because it is a language with phonemic vowel length distinction which

 Michael Schulte

avoids using duration as a correlate for stress. Historically, it has not yet levelled out the monomoraic stress feet of Germanic by processes such as reduction, syncope and the Nordic quantity shift (Riad 1992; cf. also Schulte 2004b). In later phases of North Germanic, these processes will transform all remaining undersized non-canonical stress-feet (e.g. Lahiri et al. 1999: 356– 367; Schulte 2003). As a corollary, any lengthening strategy to cope with stress is excluded for Nordic, whereas prosodic deletion functions as a central Nordic Repair Strategy. Ancient Nordic displays a system of quantity-sensitive trochees, and the bimoraic condition is not fully implemented as opposed to Modern Scandinavian (cf. Section 5.2 for Modern Norwegian). For instance, the light stem syllable of AN */staði:z/ nom.plur. ‘locations’ will automatically be footed, although it constitutes an undersized, non-canonical stress-foot (see Riad 1992; Schulte 2004b). Thus, due to the fixation of main stress we are left with a monomoraic stress foot (see (9). (9) Foot-parse: Light-heavy sequences with initial stress [Ásta.ði˜z] > ON staðir (nom.plur. of staðr ‘location’) σµ σµµ (X) (x)

Diametrically opposed, light cohering prefixes are not integrated into the foot construction. The analysis of runic data, as presented in Section 3, lends support to this claim, since no clear ga-formation is met in the entire corpus of the older runic inscriptions. It may be inferred that late PrGmc. formations such as */ga-taujan-/, */ga-wurkijan-/ or */ga-felhan-/, all well attested in the shape of Gothic gataujan, gawaurkjan, gafilhan, would metri causa be abandoned or blocked in North Germanic (see (10)). (10) Deletion of unstressed light preverbs Foot-parse: Light-heavy sequences with non-initial stress + [ga.Átau.jan] → AN *taujan (cf. KJ 43 Gallehus tawido) σµ σµµ σµµ (.) (X ) (x) [ga.Áfel.han] → AN *felhan σµ σµµ σµµ (.) (X ) (x)

+

(cf. KJ 96 Stentoften felAhekA, KJ 97 Björketorp fAlAhAk)

With regard to word formation, blocking of prefixation was part of this metrically-based restructuring process, since new prefix formations developed in terms of single-language developments in the separate Germanic languages,

Nordic prefix loss and metrical stress theory

and hence were not common Germanic. Once the new metrical pattern was established, any new preverbal structure must have been blocked. This additional process must be included because it is not always clear whether genuine loss is involved. KJ 43 Gallehus tawido, for example, may represent the original simplex verb or a prefix formation with a deleted light preverb.

. Conclusion Since Neogrammarian times, linguists of Scandinavian studies have considered the specific conditions and motivating factors of Nordic prefix loss, for its abruptness is unparalleled by any other Germanic language. The data are fairly clear, but the processes involved are close to mystifying. As argued, stress properties and phonetic reduction alone cannot account for the abrupt decline of verbal prefixes in Nordic. Invoking a metrical approach, optimization of foot structure by means of ‘prefix stripping’ clearly improves well-formedness. This claim is substantiated by the fact that the two phases of prefix loss, which can be discerned in the older runic inscriptions, neatly tie in with the binary weight distinction (see Section 2 above). External evidence for a Prosodic Repair Strategy is provided by various languages which regulate the foot pattern by deleting unmetrified syllables (e.g. Hayes 1995: 149). In addition, the phonologies of Modern Norwegian, New High German, Latin avoid unfooted material in a number of ways. Initially ‘trapped’ syllables can be mended by restructuring the algorithm: Prosodic Repair Strategies may involve lengthening, shortening and deletion (e.g. Prince 1990; Mester 1994; Hayes 1995). As argued, different stress rules in different languages motivate different repairs, since stress is realized on a languagespecific basis. While it is true that metrical phonology sheds new light on the problem, it must be admitted at the same time that it lacks explanatory force, unless integrated into a broader framework. The material quoted in Section 5 is able to demonstrate the close interaction between phonological and morphological factors with regard to ge- and be-prefixation in Modern German. In a typological perspective, the loss of preverbs within North Germanic is tied in with major changes on the morphological and syntactic levels of the language system (cf. Christiansen 1960). In a stimulating paper, Winfred Lehmann (1978) assumed strong OV influence in the early North Germanic period to account for the general lack of verbal prefixes – this typological argument based on the absence of prefixes in OV languages needs to be further explored. The



 Michael Schulte

interim conclusion has to be that metrically undersized pre-tonic syllables in Early Nordic, in particular */ga/ and */bi/, are exposed to processes of aphaeresis to regulate the foot parse. Thus, highlighting single aspects of prefix loss must be the first step in an integral analysis of the phenomenon where metrical phonology constitutes one basic component.

Notes * I am indebted to Dr. Þórhallur Eyþórsson (Univ. of Manchester), Prof. Gary Holland (Univ. of California at Berkeley), Prof. Gjert Kristoffersen (Univ. of Bergen), Prof. Kurt Braunmüller (Univ. of Hamburg), Prof. Elke Ronneberger-Sibold and Dr. Gaby Waxenberger (both Univ. of Eichstätt) for valuable comments and stimulating discussion of the material presented here. . For a definition of cohering versus non-cohering prefixes, see Kristoffersen (2000: 41–46, 182–184). . For an overview of different proposals concerning the date of Reistad, see in particular Bammesberger (1996: 117, Note 2) and Grønvik (1996: 114). The runologist James Knirk (p.c.) leaves this issue rather unanswered, stating that he personally “could countenance any date from about 350 [. . .] up to about 550”. . Both Seebold’s and Grønvik’s above-mentioned proposals are inspired by Heiner Eichë ë is taken as [gæÁgo]gæ], 3rd ner’s interpretation of the Undley bracteate where gægo gæ pres. subj. of OE gegangan, gegongen ‘to go’ (Eichner 1990: 316). But despite the striking resemblance and the closeness in space and time, sound methodology requires us to put forward two different solutions for the two forms on the Kragehul spearshaft and the Undley bracteate. For further linguistic criticism, cf. MacLeod & Mees (2001: 23, 27). . As nasals are regularly omitted before homorganic obstruents in runic practice, the spelling gaë gaë gaë for /ga-ganga/ would be perfectly possible (cf. in principal Makaev 1996: 52– 54). Thus, contrary to the claim made by MacLeod & Mees (2001: 24), the ]-rune is not obligatory to denote the velar nasal []] of the phonemic sequence /ng/. Inconsistent spelling practices are confirmed by KJ 74 Reistad iuþingaR with the n-rune as a substitute for <]> (see Schulte 2004a: 44 with references). . These provide direct parallels of ginu-, since varying representations of thematic *-u- as <-o-> and <-A-> in gino-/ginA- are explicable in terms of progressive vowel weakening and centralization processes (see Schulte 1998: 123–129). ë (which is thrice attested as irilaR) in supë aR . MacLeod & Mees (2001: 27) mention eril ë port of Noreen’s interpretation of gaë helija. But the runic term erilaR/irilaR receives an ë exceptional status in the lexicon (with no direct continuation in Old Norse), and gaë helija = /gahel:ija/ is prosodically comparable to KJ 64 Barmen þirbijaR rather than erilaR, since the initial syllable is heavy. . For analysis and interpretation of the runic form gasokun of the Pforzen inscription, see e.g. Nedoma (1999) and Grønvik (2003).

Nordic prefix loss and metrical stress theory 

References Antonsen, Elmer H. (1975). A Concise Grammar of the Older Runic Inscriptions. (= Sprachstrukturen; Reihe A, Historische Sprachstrukturen, 3.) Tübingen: Niemeyer. Antonsen, Elmer H. (1996). “Die Runeninschrift auf dem Stein von Reistad”. Historische Sprachforschung, 109, 117–126. Campbell, Alastair (1959). Old English Grammar. Oxford: Clarendon Press. Christiansen, Hallfrid (1960). “De germanske uaksentuerte prefikser i nordisk”. Norsk Tidsskrift for Sprogvidenskap, 19, 340–382. Eichner, Heiner (1990). “Die Ausprägung der linguistischen Physiognomie des Englischen anno 400 bis anno 600 n. Chr.” In A. Bammesberger & A. Wollmann (Eds.), Britain 400–600: Language and History (pp. 307–333). (= Anglistische Forschungen, 217.) Heidelberg: C. Winter. Grienberger, Theodor von (1907). “Neue Beiträge zur Runenlehre (Zweite Folge)”. Zeitschrift für deutsche Philologie, 39, 50–100. Grønvik, Ottar (1996). Fra Vimose til Ødemotland. Nye studier over runeinnskrifter fra førkristen tid i Norden. Oslo: Universitetsforlaget. Grønvik, Ottar (2003). “Die Inschrift von Pforzen”. In W. Heizmann & A. van Nahl (Eds.), Runica – Germanica – Mediaevalia (pp. 174–185). (= Reallexikon der germanischen Altertumskunde; Ergänzungsbd. 37.) Berlin & New York: Walter de Gruyter. Haugen, Einar (1982). Scandinavian Language Structures. A comparative historical survey. (= Sprachstrukturen; Reihe A, Historische Sprachstrukturen, 5.) Tübingen: Niemeyer. Hayes, Bruce (1995). Metrical Stress Theory. Principles and case studies. Chicago & London: University of Chicago Press. Kaltenbacher, Erika (1999). “Prosodische Aspekte der Verbpräfigierung”. Linguistische Berichte, 180, 447–466. Kiparsky, Paul 1966). “Über den deutschen Akzent”. Studia Grammatica VII. Untersuchungen über Akzent und Intonation im Deutschen (pp. 69–98). Berlin: AkademieVerlag. Kluge, Friedrich (1995). Etymologisches Wörterbuch der deutschen Sprache. 23rd ed. Berlin & New York: Walter de Gruyter. Krause, Wolfgang (1966). Die Runeninschriften im älteren Fuþark. (= Abhandlungen der Akademie der Wissenschaften in Göttingen; Phil.-hist. Klasse; 3. Folge, Nr. 65.) Göttingen: Vandenhoeck & Ruprecht. Krause, Wolfgang (1971). Die Sprache der urnordischen Runeninschriften. (= Germanische Bibliothek, 3. Reihe.) Heidelberg: Winter. KJ plus number of inscription: see Krause 1966. Kristoffersen, Gjert (2000). The Phonology of Norwegian. (= The Phonology of the World’s Languages.) Oxford: University Press. Lahiri, Aditi, Tomas Riad, & Haike Jacobs (1999). “Diachronic Prosody”. In H. van der Hulst (Ed.), Word Prosodic Systems in the Languages of Europe (pp. 335–422). (= Empirical Approaches to Language Typology; 20. EUROTYP, 4.) Berlin & New York: Mouton de Gruyter.

 Michael Schulte

Lehmann, Winfred P. (1978). “A Shift in the Syntactic Type of Early North Germanic and its Phonological Effects”. In J. Weinstock (Ed.), Third International Conference of Nordic and General Linguistics (pp. 110–115). (= The Nordic Languages and Modern Linguistics, 3.) Austin, Texas: University of Texas. Lindroth, Hjalmar (1918). “Till den urnordiska inskriften på Stentoftens-stenen”. Studier tillägnade Esaias Tegnér, den 13 januar 1918 (pp. 167–172). Lund: Gleerup. Looijenga, Tineke (2003). Texts & Contexts of the Oldest Runic Inscriptions. (= The Northern World, 4.) Leiden & Boston: Brill. Lutz, Angelika (1997). “Sound change, word formation and the lexicon: The history of the English prefix verbs”. English Studies, 78 (3), 258–290. MacLeod, Mindy, & Bernard Mees (2001). “The Triple Binds of Kragehul and Undley”. NOWELE, 38, 17–36. Makaev, È. A. (1996). The Language of the Oldest Runic Inscriptions. A linguistic and historical-philological analysis. Transl. from the Russian by J. Meredig. (= Kungl. Vitterhets historie och Antikvitets Akademiens Handlingar, Filologisk-filosofiska serien; 21.) Stockholm: Gotab AB. Marstrander, C. J. S. (1952). “De nordiske runeinnskrifter i eldre alfabet. Skrift og språk i folkevandringstiden. I. Danske og svenske innskrifter”. Viking, 16, 1–277. Mester, Armin (1994). “The Quantitative Trochee in Latin”. Natural Language and Linguistic Theory, 12, 1–61. Nedoma, Robert (1999). “Die Runeninschrift auf der Gürtelschnalle von Pforzen”. In A. Bammesberger (Ed.), Pforzen und Bergakker. Neue Untersuchungen zu Runeninschriften (pp. 98–109). (= Historische Sprachforschung; Egänzungsheft, 41.) Göttingen: Vandenhoeck & Ruprecht. Niedermann, Max (1908). “Une loi rythmique proéthnique en latin”. Mélanges de Linguistique offerts à M. Ferdinand de Saussure (pp. 43–57). Paris: Société de Linguistique de Paris. Noreen, Adolf (1923). Altnordische Grammatik I. Altisländische und altnorwegische Grammatik. (= Sammlung kurzer Grammatiken germanischer Dialekte; A. Hauptreihe, Nr. 4.) 4th ed. Halle/Saale: Niemeyer. Olsen, Susan (1991). “Ge-Präfigierungen im heutigen Deutsch”. Beiträge zur Geschichte und Sprache der deutschen Literatur, 113, 333–366. Page, Ray (1999). An Introduction to English Runes. 2nd ed. Woodbridge: Boydell Press. Paul, Hermann (1917). Deutsche Grammatik, Vol. 2, 3. Halle: Niemeyer. Paul, Hermann (2002). Deutsches Wörterbuch. Bedeutungsgeschichte und Aufbau unseres Wortschatzes. 10th ed. Tübingen: Niemeyer. Prince, Alan (1990). “Quantitative Consequences of Rhythmic Organization”. In M. Ziolkowski et al. (Eds.), Papers from the 26th Regional Meeting of the Chicago Linguistic Society (CLS 26), Vol. II (pp. 355–398). Riad, Tomas (1992). Structures in Germanic Prosody. A diachronic study with special reference to the Nordic languages. Stockholm: Dept. of Scandinavian Languages. Schulte, Michael (1998). Grundfragen der Umlautphonemisierung. (= Reallexikon der germanischen Altertumskunde; Ergänzungsbd. 17.) Berlin & New York: Walter de Gruyter. Schulte, Michael (2003). “Metrical Phonology and Nordic Prefix Loss. A re-examination of the basic data”. Norsk lingvistisk tidsskrift, 21 (2), 163–194.

Nordic prefix loss and metrical stress theory 

Schulte, Michael (2004a). “Fonologisk baserte skriftendringer? Overgangen fra den eldre til den yngre fuþarken”. Maal og Minne, 2004 (1), 41–55. Schulte, Michael (2004b). “The Germanic Foot in Ancient Nordic. Resolution and related matters revisited”. NOWELE, 45 (2), 3–24. Seebold, Elmar (1995). “Völker und Sprachen in Dänemark zur Zeit der germanischen Wanderungen”. In E. Marold & C. Zimmermann (Eds.), Nordwestgermanisch (pp. 155– 186). (= Reallexikon der germanischen Altertumskunde, Ergänzungsbd. 13.) Berlin & New York: Walter de Gruyter. Vennemann, Theo (1988). Preference Laws for Syllable Structure and the Explanation of Sound Change. Berlin & New York, Amsterdam: Mouton de Gruyter. Wadstein, Elis (1895). “Nordische Bildungen mit dem Präfix ga-”. Indogermanische Forschungen, 5, 1–33. Wiese, Richard (1992). “Über die Bedeutung der Phonologie für die Morphologie”. Deutsch als Fremdsprache, 29, 131–135. Wiese, Richard (1996). The Phonology of German. (= The Phonology of the World’s Languages.) Oxford: University Press.

The origin and development of lär, a modal epistemic in Swedish Gudrun Svensson Lund University

.

Introduction

The Swedish modal epistemic lär is syntactically and morphologically constrained, being used in one single form, viz. the present tense. On the other hand, its meaning is more extended as it expresses a modality corresponding to subjunctive mood in other languages, based on conclusions from different circumstantial evidence. In this paper I will discuss the possible origin and development of the modal epistemic lär from various points of view. In Section 2, I describe the use of lär in Contemporary Swedish and discuss whether lär can be classified as a grammaticalized item. In Section 3, I present two proposals regarding the origin of epistemic lär, viz. whether epistemic lär originated from lära ‘learn, teach’, or from låta ‘allow, let, permit, sound, appear’. In this section I also give an overview of previous research. In Section 4, I compare and discuss the two proposals above, i.e. that lär is derived from lata/låta ‘allow, let, permit, seem, sound, appear’ or from lära ‘teach, learn’. In 4.1, I discuss phonetic and morphologic aspects and in 4.2 semantic aspects. I also present my conclusions as to why I find it likely for epistemic lär to be derived from låta ‘allow, let, permit, sound, appear’. In Section 5, I present three different Old Swedish contexts with låta which could be possible instances for reanalysis. In 5.1, I discuss the standard assumption of earlier research that lär is derived from låta ‘allow, let, permit, sound, appear’. Finding certain drawbacks with this hypothesis, I present in 5.2 other possible sources for epistemic lär that previous research has not taken into ac-

 Gudrun Svensson

count – conditional clauses in law texts, and complex sentences with låta in the matrix clause containing a deontic-epistemic implication. In Section 6, I discuss the proposals in 5 in terms of theta-roles, looking primarily at the first recorded instances of epistemic lär. On the basis of these comparisons I come to the conclusion that complex sentences with deontic epistemic implications and with present indicative singular of låta, later/läter /lätir, in the matrix clause are the most probable context for reanalysis, and hence my proposal is that epistemic lär originated in this context. Section 7 describes the development of lär from the first recorded instances up to Modern Swedish. In this paper I use the following terms for the historical periods of Swedish: Old Swedish spans the time from 1225 to 1521 and is divided into Early Old Swedish from 1225 to 1375 and Late Old Swedish from 1375 to 1521. Modern Swedish spans the time from 1521 up to modern time and is divided into Early Modern Swedish up to 1734 and Late Modern Swedish after that. The span over the latest 50 years is called Contemporary Swedish.

. Lär as a grammaticalized item in Contemporary Swedish In Nationalencyklopedin, Ordbok, the Swedish National Encyclopaedia Dictionary, (1996: 328) lär is described as an auxiliary with the inherent meaning of hearsay and also to express a supposition with a future development that probably is to occur.1 In the large Swedish Academy Dictionary, SAOB (1942, Volume 16, column 1684), a more detailed description is given, highlighting four senses of epistemic lär: – – – –

to indicate the future with some constraints to indicate a supposition or to submit an utterance with some moderation to indicate something obligatory (only in some dialects) to express what one believes from a statement or from a more or less confirmed rumour

The descriptions above may seem rather distinct and well defined, but trying to describe the use and interpretation of lär can in reality be somewhat more problematic. To explain this statement of mine I will give two examples. Firstly, I present two sentences:

Swedish lär, a modal epistemic 

(1) Anderssons köper en bil Anderssons buypres.ind a car “The Anderssons are buying a car” (2) Anderssons lär köpa en bil Anderssons lärpres.ind buy a car “The Anderssons will apparently buy a car”

The interpretation of the first sentence does not cause any problems whatsoever, but the second one gives rise to a sense of uncertainty considering the time of the propositional event, the standpoint of speaker contra the subject of the sentence, reliability, probability and circumstantial evidence. Secondly, when translating the sentence in (2) above (Anderssons lär köpa en bil) one has to use paraphrases to catch the sense of epistemic lär, which is demonstrated in (3), (4), (5) and (6) below. In the translated sentences from Danish, German, English, and Spanish below I have marked epistemic expressions corresponding to the subjunctive mood in other languages with underlining, and circumstantial evidence with bold letters. (3) Anderssons køber angivelig en bil Anderssons buy apparently a car “The Anderssons will supposedly buy a car” (4) Die Anderssons sollen sich angeblich ein Auto kaufen The Anderssons will themselves supposedly a car buy “The Anderssons will supposedly buy a car” (5) The Anderssons will supposedly buy a car (6) Parece / dicen que compraren los Anderssons un coche Seems /saypl that buysubj the Anderssson a car “It seems/they say that the Anderssons will buy a car”

These translations demonstrate the complexity and diffusion of lär ’s inherent semantics. In Spanish, compraren, a modal expression is necessary to define the meaning correctly. Phrases like dicen, they say, and they are supposed to catch the inherent sense of circumstantial evidence. The Danish adverb angivelig expresses approximately the same semantic contents. The gist of this section is that lär is an auxiliary that cannot be translated into other languages in a 1:1 word relation, it cannot even be easily described by native Swedes due to its semantic diffuseness. At the same time lär is syntactically and morphologically strictly constrained. Lär could earlier be tensed as well as inflected in various forms, but

 Gudrun Svensson

is nowadays only used in the present indicative. Syntactically lär is strictly constrained, and in a series of verbs it is always the first one. It has no compounds, no derivations and no related words. Like other epistemic modals it is declutched from the propositional contents of the sentence and the theta-roles of the sentence are not distributed by lär. All these properties of lär indicate that it is an item with a profound history of grammaticalization process behind. I will return to the grammaticalization of lär in the various chapters below, but first I will discuss the questions where it can be derived from and what is the time span for its development.

. Previous research and proposals on the origin of lär Lär is somewhat of a solitary in the linguistic world. There is, as far as I know, no equivalent in other languages, not even in the closest Scandinavian languages. Thus, it must be of internal Swedish origin. Probably, it is of limited age as there is no instance of the word attested until Late Old Swedish. In his Dictionary of Old Swedish, Ordbok öfver svenska medeltids-språket, Söderwall (1918: 804) refers to only a single record, from 1495: (7) ler iak ekke komma til tith bröllop lär I not come to your wedding “Probably I will not come to your wedding” (BSH 4: 195)

Söderwall interprets the word lär, here spelt ler, as being semantically identical to another epistemic modal auxiliary of this time, torde. Thus, according to Söderwall, lär by this time already seems to bear the same properties as today’s lär with an exception for the sense of circumstantial evidence. One reason for the absence of lär in written records prior to 1495 may be the fact that the preserved medieval texts only reflect the written language of a small learned class (Widmark 2001: 196; Johnson 2003: 265) and thus are only partially representative of Old Swedish, oral and written. Presumably, lär has developed in the oral language, a development that happens to be unattested in the extant written sources. Another reason for lär not appearing until 1495 could be that lär is a loan word from this time. However, this is unlikely due to its position as a solitary in the linguistic world. Over time there has been some discussion regarding the origin of lär, but the research is scarce and in dictionaries quite different explanations are found regarding the derivation and development of this modal epistemic. There are

Swedish lär, a modal epistemic 

two main proposals for the origin of lär, the first one claiming that it is developed from the word låta ‘allow, let, permit, sound, appear’, and the second one that it is developed from the word lära ‘learn, teach’. The different hypotheses regarding the origin of the epistemic modal are all semantically based and highlight a future component, the presupposed event or the sense of circumstantial evidence. In his dissertation (1919: 70f.) on modal auxiliaries Harald Björkstam came to the conclusion that lär was derived from lära ‘learn/teach’. Under the influence of the German word lernen, and especially in compounds like kennen lernen, it developed into an auxiliary containing the features ‘knowledge’ and ‘future’. His conclusions were supported by Elof Hellquist in his etymological dictionary (1925, 3rd edition 1957: 609). Another standpoint was taken by Elias Wessén (1941/1962: 7, 147, 211f., 221), who proposed lär to be derived from lata/låta in Old Swedish. The present indicative form latir developed into läter due to umlaut and later became lär. This conclusion has been supported by other linguists and is for instance supported by SAOB (1942, Vol. 16, column 1684).

. Comparisons between the possible origins from lära or låta In this section I will present and discuss what previous research and linguistic dictionaries say about the development of lär. In 4.1 I will discuss phonological matters and in 4.2 semantic points of view. I will scrutinize the different hypotheses and discuss them on the basis of linguistic development rules and from the point of view of grammaticalization theory. I will also present evidence from Old Swedish which supports my conclusion that the hypothesis that lär is derived from lära ‘learn and teach’, is erroneous. . Phonological discussions regarding the origin of epistemic lär As mentioned above, Harald Björkstam (1919) asserted that lär originates from lära ‘teach, learn’. He based his hypothesis first and foremost on the phonetic similarity (1919: 71f.). According to Björkstam, the two words are identical and the present indicative singular form of lära, viz. lärer, alternating with the short form lär was used also for epistemic lär. However, this way of proving relationship is a weak one when discussing linguistic kinship since historical sound developments may very well conceal

 Gudrun Svensson

derivational stems. As a matter of fact, lära is not the only potential origin of lär. Phonologically, in addition to being derived from lära ‘teach, learn’, the epistemic lär can also be derived from the verb lata/låta even if the two verbs nowadays differ from each other. In Old Swedish texts, the present form latir is often spelled lätir due to umlaut; attested spelling forms are later, latær, latir, lætar, lätir, lætær, lethær, läter. The development of lär from läter is harder to observe. According to Wessén (1941/1962: 15), the development of lär from läter presupposes a stage where t is weakened to dh, i.e. läter > *lädher > lär, when the verb is unstressed. The form lädher is, however, unattested, although there are many examples of other forms of lata/låta being spelled with d/dh, viz. lader, ladhe, ladæ, lædz, lædhz. Admittedly, since Old Swedish spelling varies from source to source, it is not obvious how to interpret the lack of the intermediate form. However, contrary to Wessén’s claim, there is a single example attested where the present singular of lata is spelled lær, namely in The Law of Uppland, Upplandslagen, from 1296: (8) Nu liggær man a skoghi ællr skipi. ællr annæer staþ. Now lies man in forest or ship. or other place. Lær sik þær til myrþe ælr til renæ. Lets himself there to murder or to run. “Now a man is lying in the forest or on a ship. Or other place. Abandoning himself to murder or loafing about.”

The expression lata sik til means ‘to devote/abandon oneself to’. In the sentence above the word lær is a short form for læter and cannot be the auxiliary lär. This single example shows, however, that there existed a short form of läter that could have been the basis for the auxiliary verb lär. This example thus also proves that there is a possible phonological development from lata/låta to lär. As a matter of fact, Björkstam (1919: 70) refers to this instance but emphasizes that it is not especially noteworthy as it has no relation to epistemic lär and is only an isolated example in a dialect. Certainly, he is right in some way but on the other hand this example shows that there existed a short form of läter already in Early Old Swedish. In The Later Västgöta Law, Yngre Västgötalagen, around 1350, I have found a text saying:

Swedish lär, a modal epistemic 

(9) Lær maþer manni þrel sin. svari hin. . . . Lets mannom mandat thrall his answerpres.subj him. . . “If a man lets his thrall to another man, he answers for him”

This lær could be based on läter but is probably based on lejer/leger ‘to let, to hire out’. Even if it is based on lejer it shows that disyllabic words with ä/e can be shortened to lær. In the texts from Late Old Swedish, around 1500, there existed two forms of epistemic lär in the present indicative singular, viz. lär see (7) above, and lärer (10) from Sturekrönikan, around 1500: (10) Han lärer wäl myndre springa He lärpres.ind.sing. probably less run tha wy honum med riis wiyliom twinga when we him with birch want to-force “He will probably run less when we subjugate him with birch”

The two forms in the present indicative, lär and lärer, were seen as proof of the relation between epistemic lär and lära ‘teach, learn’, present indicative lärer, by Björkstam (1919: 71f.), while Hellquist (1957: 609) found it strange that the form lär in that case appeared already in Late Old Swedish, assuming that lär in some instances was the unstressed form. In SAOB (1942: Vol. 16, column 1684) it is stated (my translation): ‘The form lärer is a written extension of the same sort as bärer for bär’.2 According to Hellquist the present indicative of lära ‘teach, learn’ was always lärer up to the end of the 17th century, while the same form for epistemic lär was alternating lär or lärer (1902: 144). From 1700 onwards only the form lär was used in the present indicative for both the verbs, epistemic lär and lära ‘learn, teach’ (Noréen 1913: 136). . The origin of the epistemic lär: Semantic considerations Harald Björkstam (1919: 90f.) asserts that epistemic lär and lära ‘teach, learn’, have some semantic properties in common as both words have an inherent meaning of ‘getting to know’. This common meaning of ‘getting to know’ is, according to Björkstam, a result of influence from the German expression kennen lernen with the same implication. According to Givón (1979: 209) grammaticalization starts in discourse: “Loose paratactic ‘pragmatic’ discourse structures develop – over time – into tight ‘grammaticalized’ syntactic structures”. Björkstam also starts in discourse (1919: 90) when he explains the development of lära into epistemic lär. According to him there is a shift in meaning when using the compound lära

 Gudrun Svensson

sanna ‘confirm the truth’, in expressions like lära sanna mina ord ‘confirm the truth of my words’. The word lära means both ‘teach’ and ‘learn’ and with the ‘learn’ meaning there is a shift over time towards future meaning (Björkstam 1919: 90f.). This is a case of what, according to later grammaticalization theories, would be called pragmatic inferencing leading to reanalysis (Hopper & Traugott 2002: 63f.). From this inference lär becomes a future auxiliary and supersedes another Old Swedish auxiliary expressing future, mon, at the end of the Old Swedish period (Björkstam 1919: 88). Thus the future component is dominating in lär in the texts from the next centuries, according to Björkstam (1919: 94f.). Certainly, Björkstam’s discussion of inferences in a local context is in accordance with modern grammaticalization theories, but with respect to other items of these theories his proposal leads to some problems. Consider e.g. the following quote from Hopper & Traugott (2002: 75): “for inferences to play a significant role in grammaticalization they must be frequently occurring.” In this respect lära ‘learn, teach’, is a poor candidate as the basis for epistemic lär. It does not seem to appear very frequently in Old Swedish texts. Out of roughly 500, 000 words that I have checked from laws, religious texts, stories, and historical poems the various forms of the verb lära ‘learn, teach’, appeared in the frequency of about only 0.02% while the other candidate, the verb lata/låta ‘allow, let, permit, sound, appear’, appeared in about 0.2%. Thus, in this respect, the probability of lära ‘learn, teach’, being grammaticalized seems to be very low. Another problem for Björkstam’s hypothesis, in the light of the grammaticalization theory of Hopper & Traugott (2002: 2f.), is that the source conception for grammaticalization mostly has general meaning. The meaning of the content verb lära ‘learn, teach’, is, however, very specific and restricted with just a few inherent semantic components. The opposite can be said about lata/låta ‘allow, let, permit, sound, appear’, which has many implications and can function as a strict semantic verb, transitive or intransitive, as well as an auxiliary. In Söderwall’s Dictionary of Old Swedish (1918: 803) the word lära ‘learn, teach’, takes up about one column, while lata/låta ‘allow, let, permit, sound, appear’, takes up about 6 times more (ibid.: 738f.). Still another problem with Björkstam’s hypothesis is the fact that epistemic modal verbs tend to be derived from other auxiliaries; see Bybee, Perkins & Paglicua (1994: 176f.) and Heine & Kuteva (2002: 116f.). They typically are not directly derived from content verbs. Deontic auxiliaries with the implication of capability/necessity/obligation are often the basis for developing into expressions for possibility/consent or probability. In respect to this, I find lära ‘teach,

Swedish lär, a modal epistemic 

learn’, unlikely to be the origin of epistemic lär. The other candidate, lata/låta ‘allow, let, permit, sound, appear’, had already in Old Swedish a function that is very close to modal auxiliaries. This use can be found as early as in Runic Swedish, around 800–1225 (Wessén, 1956/1992: 343f.): (11) lit raisa stain let raise stone “had the stone erected”

In this example lata ‘let’, has the meaning ‘induce’ and it expresses a situation where the subject of lata is the originator of the event (stone raising) but not the executor of this event. In 5.2 I will present sentences from Old Swedish where lata/låta, ‘let’ could have been interpreted as having a deontic meaning. I will propose that in special local contexts sentences of this kind can have been inferred from the context, reanalysed and thus the base for grammaticalization of lata/låta into epistemic lär.

. Regarding the derivation of epistemic lär from lata/låta In this section I will present some possible models for the derivation of epistemic lär from lata/låta ‘allow, let, sound’. In 5.1 I will discuss the proposal in SAOB (1942, Volume 16: column 1684) that epistemic lär probably has its origin in the meaning ‘look, seem, appear’, a meaning that cannot be found in låta of today. In 5.2 I will discuss the possibility that epistemic lär could have originated from a specific use of låta in complex conditional sentences, and hypothesize that epistemic lär through time developed out of lata/låta ‘let’, in special contexts where lata/låta, ‘let’ could be interpreted as having a deontic/epistemic meaning. . The suggested derivation from låta/lata ‘seem, appear’ According to SAOB (1942, Volume 16: column 1684), lær has derived from lata/låta with the meaning ‘look like, seem, appear’ (SAOB 1942, Volume 16: column 1486). According to SAOB this meaning of låta is a loan from Middle Low German and SAOB also refers to the occurrence of this meaning of låta in Danish, German, and Norwegian dialects. In Söderwall’s Old Swedish Dictionary there are only two recorded occurrences of lata with this meaning.

 Gudrun Svensson

However, I have found some more examples in Old Swedish texts which can illustrate this meaning. Here are some examples from Erikskrönikan, around 1350: (12) then tiid han saa den ofridh that time he saw the dissension ok han förstod at illa loth and he realized that bad seemed tha loot han trädhan gaa til foot then let he trees-the go by foot “At that time he saw the war and realized that it seemed to be bad, then he let the trees go by foot” (13) nw prövar iak at illa lather now experience I that bad seems (H)ärtogans mark affstark duke-the’s power too-strong ok sende tha bud til danmark and sendsubj then message to Denmark “Now I realize that it seems to be bad because the duke’s power is too strong and he will then send a message to Denmark”

The Swedish word låtsa(s) ‘pretend’, has also developed from Old Swedish lata/låta. The two meanings, ‘pretend’ and ‘it seems’, are quite close from the interlocutor’s point of view. Here is an example from Ivan Lejonriddaren, around 1400–1425: (14) The frugha loot som hon ville thet ey höra The lady let as she wanted that not hear “The lady pretended/looked as if she did not want to hear it”

A similar example is found in Sturekrönikan, around 1490: (15) Han lädz wara oredh med sin här He lets be unafraid with his army. “He seems /pretends not to be afraid with his army”

The examples above could support a reanalysis of läter into auxiliary lär as far as the subjunctive component is concerned and also with regard to the meaning of circumstantial evidence, but it does not explain the suggested component of future, which, according to SAOB (1942, Volume 16: column 1685), Hellquist

Swedish lär, a modal epistemic 

(1902: 192) and Björkstam (1919: 75f.), was a very dominating factor already in the first records of epistemic lär. As mentioned above in 4.2, epistemic auxiliaries are frequently developed from other auxiliaries, mostly from deontic ones. Taking this into consideration, it is quite possible that epistemic lär can have developed from lata with the meaning ‘it seems, looks, appears’ as this is an epistemic expression already in itself. . Possible development of epistemic lär from complex sentences with deontic and epistemic implication In many Early Old Swedish laws, we can attest sentences beginning with the word latir/later/læter/latær: (16) Latær maþær hæst i tjuþær. . . Lets man horse in tether. . . “If a man leaves a horse in tether. . .” (Law of Västergötland, around 1220) (17) Latær cona lægra sic. . . Lets woman mate herself. . . “If a woman lets herself be mated...” (Law of Västmanland, around 1300–1350) (18) Later man eng sinæ wardlössu liggiæ. . . Lets man meadow his careless lie. . . “If a man leaves his meadow untended. . .” (Law of Södermanland 1327)

Conditional subordinate clauses like (16)–(18) above contain hypothetical propositions. In the Early Old Swedish laws the succeeding matrix clause contains the sanction, which is expressed in a noun phrase and/or by the verb in the indicative or subjunctive mood. This combination thus indicates an epistemic possibility and the sentences in question refer to a possible case that would lead to a possible sanction. The matrix clauses express a deontic meaning, e.g. what must/should happen if the hypothetic case is fulfilled. Thus the sequence also contains a deontic implication which is a possible but not necessary prerequisite for grammaticalization into epistemicity according to Bybee, Perkins & Paglicua (1994: 176f.) and Heine & Kuteva (2002: 116f.). However, I will not claim that these kinds of conditionals were the basis for reanalysis of lata as the auxiliary lär because the deontic meaning is not

 Gudrun Svensson

implicated in the verb læter in the subordinate clause but in the succeeding matrix clause. On the other hand, this kind of context can have contributed to creating a hypothetical implication in læter and thus urged on the development into epistemic lär. There are other cases in Old Swedish where the present indicative singular of lata/låta, i.e. later/läter, can have a deontic implication. The verb lata/låta has a function close to an auxiliary in Old Swedish (Wessén 1956/1992: 203f.) and in these cases it bears a deontic implication. This deontic implication is visible in the following instance from Stockholms stads tänkeböcker 1474 (Hildebrand 1917: 18). (19) Hwat rettin och laghin giffua, thermedh lather jag mik atnögia What court-the and law-the give, therewith let I myself content “With what the court and law give, I let myself be content”

In the four sentences below the word later/läter implicates what will happen under certain circumstances. Most of the sentences are introduced by subordinate clauses containing basic conditions for the actions in the main clause where lata/låta is the most salient verb. These conditions are not yet fulfilled and so the actions are placed in the future. If the conditions are fulfilled the actions will necessarily take place, according to the speaker. As the verb later/läter is in the indicative there is more certainty than if it had been in the subjunctive mood, late/lati/läte/läti. (20) nar erik mz almogen komber ok thesse tidhende fornimber when erik with peasantry-the come and these tidings hears tha later han i same stund os oc war stad brenna i grund then lets he at same time us and our town burn to ground “when Eric comes with the peasantry and gets to know this news then at once he will burn us and our town to the ground” (Karlskrönikan, around 1450) (21) tolikä här myklä större än iach är och later idher allä dräpä such here much bigger than I am and lets you all slaughter och fang och aldrig skal een undan ganga and catch and never shall one away go “such one is much bigger than me and will have you all killed and caught and you can never escape” (Fredrik av Normandie, 1308 ) (22) ffaar thin fadher thetta wita tha later han oss oc alle gets your father this know then lets he us and all

Swedish lär, a modal epistemic 

ware slaekt fordaerfwa our kinships spoil “If your father gets to know this then he will have us killed with all our relatives” (Troens tröst, around 1420 ) (23) Ok vm han kombir in ij gasa husit. ey lätir han sik nöghia at And if he comes in into goose house. not lets he himself satisfy that han ätir ena gass he eats one goose “And if he comes into the goose house he will not be satisfied by eating only one goose” (Birgittas uppenbarelser, around, 1380)

The sentences above express events that will take place when/if the conditions are fulfilled, a deontic implication. But still there is a difference in the certainty sense of later/läter under the influence of the conditional prerequisites of the subordinate clause in these local contexts, viz. the conditions given in the subordinate clauses (22) and (23) are a bit more uncertain than (20) and (21). The conditions for the presupposed occurrence inserted by later/läter in the matrix clause are thus also more uncertain in (22) and (23). Later/läter is hereby inferred as being less certain and more of a possibility/probability; the implication can be regarded as moving over from deontic to epistemic. From sentence (19) to (23) there is a gradual shift in later/läter from a deontic to an epistemic implication. I propose that sentences like (22) and (23) could be a basis for a reanalysis, and so through time the meaning of lata /låta has shifted from an almost auxiliary use to an epistemic modal auxiliary. The fact that the verb has the present indicative form later/läter can, on the other hand, be an indication that there is more of reality than of possibility than in a sentence where the verb form is in the subjunctive. However, according to Wessén (1956/1992: 142f.) there was a considerable decrease in the use of the subjunctive already in Early Old Swedish, especially in speech, so it is quite possible to assert that the sentences (19) to (23) have epistemic modality even though the verb later/läter is in the indicative. I also regard it as likely that it was precisely the verb form later/läter and not a subjunctive form that is the origin of the final form lär (see 4.1 above). König & Traugott (1988: 110f.) discuss pragmatic inference and demonstrate how conversational implicature based on cooperative interaction can be conventionalized. When looking through the Old Swedish texts in order to find clues to the development of lär I have noted that examples like (19) to (23)

 Gudrun Svensson

are mostly found in quotations of direct speech, i.e. the closest approximation to oral Old Swedish. The development of lär is unattested in written records but instances of later/läter ‘let’, in quotations might indicate that there was an ongoing development in oral Old Swedish with the result that a deontic implicature through time can have changed into an epistemic implicature that later became conventionalized. The development could thus be illustrated in the following way: läter as an auxiliary > läter as a deontic auxiliary > läter as an epistemic auxiliary > epistemic lär

Traugott & Dasher assert (2002: 124) that concerning the English auxiliary may there is a developing implication due to the context from the sense of permission to implication of demand and deonticity and finally to epistemicity. This description can be applied to the Swedish lata/låta ‘permit, let’, which in sentences like (19) to (23) had reached the development cline of deontic implication. The hypothesis that epistemic lär originates from a deontic meaning can also be supported by the fact that in many dialects of Modern Swedish there is a modal auxiliary lär that has a deontic implication (Sundqvist 1955: 307; SAOB 1942: column 1682) which means that in these dialects lär is not an epistemic modal. (24) Jag lär väl ge hästarna nattfoder, sade han I lärpres.ind surely give horses-the night-forage, said he “I suppose I have to give the horses night forage, he said” (SAOB 1942, Volume 16: column 1682)

These occurrences also indicate the relations between deontic and epistemic implications in lär. As mentioned above in Section 1, the epistemic lär of today bears implications of possibility and a conclusion from circumstantial evidence. These implications are also strong in sentences (20) to (23). Remarkably, it is quite possible to replace later/läter in sentences (20) to (23) with epistemic lär in Modern Swedish and get the same implications.

. Discussions regarding sentence structures and theta-roles Since epistemic lär seems to have developed within a relatively short time span (1200–1500) it is not likely to have wandered too far away from its origin. This

Swedish lär, a modal epistemic 

is an argument against Björkstam’s (1919) attempt to derive epistemic lär from lära ‘teach, learn’, since there is no indication in the source constructions that this verb occurs without the full set of theta-roles, whereas epistemic lär, like other auxiliaries, does not assign any theta-roles. Consider the following example from Björkstam (1919: 93), which is said to illustrate a shift in meaning towards epistemic lär. (25) Om i icke lären sanna mina ord, . . . If you not learn confirm my words, . . . “If you do not learn to confirm my words, . . .” (Judas: 58)

The finite verb lären ‘learn’, obviously assigns theta-roles to its arguments here, thus from a grammatical point of view there is no indication of a development towards auxiliary use. In all proposed source constructions with the semantic verb lära, ‘learn, teach’, epistemic lär is thus said to be derived from a content verb with the full array of theta-roles which radically differs from the target construction with epistemic lär. From the point of view of theta-role structure, my hypothesis that epistemic lär is derived from lata used in complex sentences with deontic or epistemic implications seems more likely to be on the right track. Consider example (22) again, here repeated as (26): (26) ffaar thin fadher thetta wita tha later han oss oc alle gets your father this know then lets he us and all ware slaekt fordaerfwa our kinships destroy “If your father gets to know this he will have us killed with all relatives” (Troens tröst, around 1420)

In this example han, ‘he’ is the subject of latær, ‘lets’ and is assigned the thetarole AGENT. However, the AGENT is not the PERFORMER of the actions. This specific phenomenon in the theta-role assignment of låta, ‘let, induce’ has been discussed by Lundin (2003: 124f.). According to Lundin the word låta is an AGENT-splitter in special contexts (designated låtaIN2 ), i.e. in examples like: (27) Kalle lät öppna dörren Kalle let open door-the “Kalle had the door opened”

 Gudrun Svensson

“The protypical AGENT is ‘doer’ as well as ‘initiator’ ”, writes Lundin (2003: 134) and continues: “When combined with låtaIN2 the AGENT-concept is split up, with the result that the subject functions only as INITIATOR (=Kalle), whereas the one performing the act, the DOER (= Kalle), is never expressed [...] låta IN2 can only conflate with an agentive verb” (Lundin 2003: 129). In (26) above this means that the DOER who does what is expressed in the non-finite semantic verb, fordærfwa ‘destroy’ is not overt. Example (10) here repeated as (28) is from Late Old Swedish, Sturekrönikan, around 1500, illustrating one of the first recorded instances of epistemic lär in Swedish. (28) Han lärer wäl myndre springa He lärpres.ind.sg. probably less run tha wy honum med riis wiyliom twinga when we him with birch want to-force “He will probably run less when we subjugate him with birch”

In this sentence Han ‘he’, is the subject of the semantic non-finite springa ‘run’. The theta-role assigned to the subject by springa, ‘run’ is AGENT. The finite verb is the epistemic lär that is subject autonomous and thus does not assign theta-roles. Epistemic lär implicates ‘possibility’. Someone, presumably the speaker, must impact possibility to the event described. However, the speaker is not overtly expressed. I claim that in a target construction, here epistemic lär, there are probably some remnant traces of the source construction. This is not the case regarding the suggested origin of lär (25) from lära ‘teach, learn’, but quite possible in the examples which show reconstruction of låta on its way to be an epistemic modal, viz. (12) to (15) or (19) to (23). Hence, I find it likely that epistemic lär originated from lata/låta, ‘allow, let, permit, sound, appear’ or from local contexts with läter in complex sentence structures in combination with possible deontic and eventually epistemic implications.

. The development of lär in Modern Swedish From its first appearance around 1490 the auxiliary lär appears more or less frequently in the texts. Implications of future and probability/possibility were dominating from the beginning but could vary according to the authors and over time (Björkstam 1919: 88) and it can sometimes be difficult to determine how to interpret the implications of lär. For example in record (29) below the

Swedish lär, a modal epistemic 

expression iag wet ‘I know’, implies something that is for certain but on the other side the present subjunctive form, läre¸implicates uncertainty. In (31) the future implication refers to uncertainty but the flexion of lär is in the present indicative and thus indicates more certainty: (29) Och iag wet, thet läre så gå dig And I know, it lärepres.subj. so go you “And I know it will be like that with you” (Agneta Horn, Beskrivning över min vandringstid, 1647) (30) Så hoppas jag och att J lär bruka den behändighet som kan So hope I too that you lärpres.ind use the facility that can beqwema henne att taga dem emot bring her to take them up/to “So I also hope that you will use the facility to make her accept them” (Johan Runius, Prosastycken på svenska, 1697)

Other examples display a stronger implication of epistemic modality: (31) så lär väl Gudh hielpa dem so lärpres.ind.sing probably God help them “so God may help them” (Jon Stålhammar, Brev, 1700–1708) (32) vij lära intet få sätterijfriheet därpå we lärpres.ind.pl not get press-freedom there-after “we will probably not get freedom of the press after that (Jon Stålhammar, Brev, 1700–1708)

From around 1680 lär can be used without the implication of future (Björkstam 1919: 89). After that lär develops a stronger implication of conclusion from circumstantial evidence but the time when this becomes possible is difficult to determine exactly (Björkstam 1919: 85f., Hellquist 1902: 192). The Swedish king Charles XII (1697–1718) frequently uses lär in his letters. He uses the verb as an epistemic auxiliary with (33) or without (34) the implication of conclusion from circumstantial evidence or just as a future auxiliary (35). (33) Ty Zaren lärer hvarken tillåta honom att hålla sin Parol Because tsar-the lärpres.ind neither allow him to keep his word “Because the tsar will probably not allow him to keep his promise” (Karl XII 1710, Carlsson, 1893: 105)

 Gudrun Svensson

(34) hvad som kunde lända hertigens intresse till förmån, that which could redound duke-the’s interest to advantage så lärer iagh intet underlåta all flit dhertill at använda so lärpres.ind I not neglect all diligence thereto to use “for whatever is advantageous for the duke’s interest I will not fail to use diligently my assiduity” (Karl XII 1715, Carlsson, 1893: 9) (35) Iagh håppas, at han lärer tilltaga uti beständihg välmåga I hope that he lärpres.ind grow in steady good-health “I hope he will have increasingly good health” (Karl XII 1715, Carlsson 1893: 9)

From the first records of epistemic lär in the sources around 1500 through the next 300 years the modal is found inflected for tense and mood and person (see examples (28) to (35) above), but by time the flexions disappeared and in contemporary Swedish it is only used in the present indicative form lär.

. Summary and conclusions The modal epistemic lär seems to undergo a profound development of grammaticalization. From appearing as an already grammaticalized item in texts around 1490 it has continued the grammaticalization cline to become a very constrained grammatical form in contemporary Swedish, at the same time as the semantic implications seem to have widened and become more elusive. In this paper I have argued that lär developed from the verb lata/låta during the Old Swedish period and that it underwent further grammaticalization during Early and Late Modern Swedish. There seem to be two significant contexts in which the grammaticalization cline of lär might have started: either in cases where lata/låta in the present indicative singular form, later/läter/lätir, expressed the meaning ‘look like, seem, sound, appear’ or in cases where lata/låta in the present indicative singular form, later/läter/lätir, was used in complex sentences with deontic/epistemic implications. What will be the next step? Maybe epistemic lär will follow the fate of other Swedish epistemics like kan ske ‘may happen’ and månne ‘may be’, and end up as an adverb.

Swedish lär, a modal epistemic 

Notes . Original text: enligt vad det påstås. BET.NYANS: för att uttrycka förmodan om framtida utveckling kommer sannolikt att. . Original text: Formen lärer är en skriftspråklig utvidgning av samma slag som bärer för bär.

References Björkstam, Harald (1919). De modala hjälpverben i svenskan I. Lund: Håkan Ohlssons tryckeri. Bybee, Joan L., Revere Perkins, & William Pagliuca (1994). The Evolution of Grammar. Chicago: University of Chicago Press. Bybee, Joan L., & William Pagliuca (1987). “The revolution of future meaning”. In Anna Giacolone Ramat, Onofrio Carruba, & Giuliano Bernini (Eds.), Papers from the 7th International Conference on Historical Linguistics (pp. 108–122). Amsterdam/Philadelphia: John Benjamins Publishing Company. Givón, Talmy (1979). On understanding grammar. New York: Academic Press. Heine, Bernd, & Tania Kuteva (2002). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hellquist, Elof (1925, 1957, 3rd edition). Svensk etymologisk ordbok. Lund: Gleerups. Hellquist, Elof (1902). Studier i 1600-talets svenska. Uppsala: Akademiska bokhandeln. Hopper, Paul J., & Elisabeth Closs Traugott (1993, reprinted 2002). Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul J. (1991). “On Some Principles of Grammaticalization”. In Elisabeth Closs Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. I (pp. 17–35). Amsterdam/Philadelphia: John Benjamins Publishing Company. Johnsson, Rakel (2003). Skrivaren och språket. Göteborg: Kompendiet Aidla Trading AB. König, Ekkehard, & Elisabeth Closs Traugott (1988). “Pragmatic strengthening and semantic change: The conventionalising of conversational implicature”. In W. Hüllan & R. Schulze (Eds.), Understanding the Lexicon. Meaning, Sense and World Knowledge in Lexical Semantics (pp. 110–124). Tübingen: Max Niemeyer Verlag. Lundin, Katarina (2003). Small Clauses in Swedish. Towards a Unified Account. Lund: University Press. Nationalencyclopedin, Ordbok (1996). Höganäs: Bra böcker. Noréen, Erik, & Sven Benson (1932/1954). Fornsvensk läsebok. Lund: Gleerup. Noréen, Adolf (1913). Vårt språk. Nusvensk grammatik IV. Lund: Gleerup. SAOB (1942). Volume 16. Lund: A.B. PH Lundstedts universitetsbokhandel. Sundqvist, Anders (1955). Studier i svensk moduslära. Lund: Gleerup. Söderwall, K. F. (1884–1918). Ordbok öfver svenska medeltids-språket. Lund: Gleerup. Söderwall, K. F. (1925–1973). Ordbok över svenska medeltids-språket. Supplement utarbetat av K.F. Söderwall, W. Åkerlund, K.G. Ljunggren & E. Wessén, Lund. Teleman, Ulf, Staffan Hellberg, & Erik Andersson (1999). Svenska Akademiens Grammatik. Uddevalla: Mediaprint.

 Gudrun Svensson

Traugott, Elisabeth Closs (1980). “Meaning-change in the Development of Grammatical Markers”. Language Sciences, 2, 44–61. Traugott, Elisabeth Closs, & Richard B. Dasher (2002). Regularity in Semantic Change. Cambridge: Cambridge University Press. Wessén, Elias (1941, 6th edition 1962). Svensk språkhistoria I. Ljudlära och ordböjningslära. Stockholm: Almquist och Wiksell. Wessén, Elias (1956, 2nd edition 1992). Svensk språkhistoria III. Grundlinjer till en historisk syntax. Stockholm: Almquist & Wiksell. Wessén, Elias (1966). Våra ord, kortfattad etymologisk ordlista. Stockholm: A.B. Ph. Norstedt & söner. Widmark, Gun (2001). Det land som blev vårt. Uppsala: Kungl. Gustav Adolfs Akademien för svensk folkkultur.

Swedish lär, a modal epistemic 

Appendix: Excerpted literature

Excerpted literature, non-computerized: Carlsson, Ernst. 1893. Konung Karl XII:s egenhändiga bref. Stockholm: Norstedt & söner. Handlingar rörande Sveriges historia. 1864. band 1–3, Gustav Wasas brev 1521–1526. Stockholm: Nordstedt & söner. Hildebrand, Emil. 1917. Stockholms Stads tänkeböcker 1474–1483 samt Burspråk. Stockholm: Ivar Hæggströms boktryckeri. Wallmark P. A. 1832. Carl II:s brefvexling med sin syster Prinsessan Ulrika Eleonora från år1698 till 1709. Stockholm: Ecksteinska tryckeriet. Computer excerpted literature (chronologically ordered): Sources: Fornsvenska textbanken, Lund, www.nordlund.lu.se; Språkbanken, Göteborg, www.svenska.gu.se Early Old Swedish, 1220–1375 Äldre Västgötalagen Upplandslagen Bjärköarätten Södermanlandslagen, Smålandslagen Konungastyrelsen Äldre och Yngre Västmannalagen Legendariet (Bur) Legendariet (Bil) Gutalagen Magnus Erikssons landslag Birgittaautograferna Yngre Västgötalagen Flores och Banzeflor Ivan Lejonriddaren Fredrik av Normandie Erikskrönikan

Late Old Swedish, 1375–1521 Läkeböckerna Kristoffers landslag Namnlös och Valentin Karl Magnus Sju vise mästare Prosaiska krönikan Didrik av Bern Barlam och Josafat Historia Trojana Ängelns betraktelser Helga Manna leverne Birgittas uppenbarelser Sankt Bernhard Karlskrönikan Sturekrönikan Ordspråk Moseböckerna

Early Modern Swedish, 1521–1734 Olaus Petris krönika Peder Swarts krönika Per Brahes krönika Carl Carlsson Gyllenhielms anteckningar Agneta Horn: Beskrivning över min vandringstid Urban Hiärne: Stratonice Samuel Columbus: Roo-mål eller Mål-roo Haqvin Spegel: Dagbok Jon Stålhammar: Brev Johan Runius: Prosastycken på svenska 1734 års lag. Missgärninga och giftermålsbalken Den svenska Argus

The development of the Spanish verb ir into an auxiliary of voice* Thora Vinther University of Copenhagen

.

Introduction

. Passive constructions in Spanish? The Spanish passive construction ser + past participle (pp) is an example of linguistic development towards analytic structure, as soy amado (< sum amatum) has replaced the synthetic passive form of Latin amor → soy amado. The construction shows a lower degree of grammaticalisation as compared to the perfective periphrasis haber + pp, as its participle obligatorily concords with the gender and number of the Subject: las niñas son admiradas por sus abuelos (“the girls are admired by their grandparents”), while concord has disappeared from the pp when it realises the (non-passive) perfect: las señoras han admirado las flores (“the ladies have admired the flowers”). Although the passive is described as such in most grammar textbooks, linguists do not always agree in considering this construction a passive periphrasis in Spanish. The clearest opponent is Alarcos Llorach (1970, 1994), who sees no difference between ser + an adjective and ser + a pp, considering them both predicative complements: “sintácticamente, no hay ninguna diferencia respecto de las oraciones atributivas” (“syntactically there is no difference with respect to the predicative sentences” (1994: 303). As regards the occurrence of agentive phrases he considers the agentive merely as a semantic feature that has no influence on the syntactic structure. Writing about the agentive and causal sense of the two por phrases in El discurso es elogiado por el público (“the speech is praised by the audience”) and El discurso es elogiado por su claridad (“the speech is praised for its clarity”) he states: “Agente y causa son sentidos no distinguidos por la estructura y las relaciones sintácticas” (“Agent and cause are

 Thora Vinther

meanings that are not distinguished by the structure and the syntactic relations”) (p. 304). He thus fails to see the syntactic difference in that the causal phrase can appear first in the sentence: Por su claridad, el discurso es elogiado, a position in which the agentive phrase cannot occur. If Por el público, el discurso es elogiado is spoken with neutral intonation, the agentive sense disappears, and por el público acquires the meaning of ‘because of the audience’ or ‘for the sake of the audience’. The passive can also take the auxiliary estar, with the nuance of ‘reached state’ (‘Zustandspassiv’), and also a number of other ‘semi-auxiliary’ verbs, the inventory of which varies in the reference literature, though it normally includes andar, ir, quedar, resultar, salir and venir. In his study of Romance voice auxiliaries, Green (1982) described these constructions, together with a larger list which he formally defined as consisting of “a pp conjugated with some verb other than ser, estar, or haber” (p. 103), as semi-auxiliary constructions expressing the passive voice. His conclusion was that these verbs “are substantially comparable with estar-passives and share some important characteristics – chief among them concord patterns – with ser-passives” (p. 115). In other words, according to Green, if these semi-auxiliary constructions are not ‘passives’, all copula + pp combinations must be considered as predicative constructions, and consequently Spanish would have no analytic passive. “In my view that would be manifestly absurd” (Green 1982: 115). . Ir + past participle In this paper I shall concentrate on the construction formed by ir + pp and consider the finite verb’s degree of auxiliarity. Though such a limitation of scope may seem very narrow, it is essential to adhere to one construction only, in an attempt to avoid the danger of confusing the properties of the different verbs that are construed with a pp. I shall study the occurrence of the construction in older texts and consider some of Green’s (1982) and Lehmann’s (1985) criteria of grammaticalisation with respect to both grammatical and semantic features. Finally I shall look at the construction in modern Spanish, and try to show that it has undergone a semantic change. The majority of examples have been extracted from the two databases of the Real Academia Española: CORDE (texts before 1975) and CREA (texts from 1975 on). Matthies (1933) has also been a good source of examples.

The development of ir into an auxiliary of voice 

. Diachronic analysis In the course of the history of the Spanish language the ir + pp construction appears both as a main verb of movement modified by the pp as a predicative adjective and as a parallel construction that can be interpreted more or less as a periphrasis. According to Reichenkron (1933), quoted in Matthies (1933), the only movement verb in Latin which could be construed with a pp was venire. In the Poema de Mio Cid, Menéndez Pidal found only pp periphrases with the verb andar, but Yllera (1979: 295–296) considered the resultative meaning of the following example as the point of departure of a periphrasis that “se desarrolla tímidamente” (“develops timidly”) in the following centuries: (1) Como yo fio por Dios y en todos los sos santos, desta arrancada nos iremos pagados1 (2447–2448). “As I rely on God and all his saints, we shall go away satisfied from this victory.”

Though the main sense is that of going away, Yllera (1979: 295) adds: “pero, a la par, irse supone el resultado de la acción anterior” (“but at the same time, irse entails the result of the preceding action”). The question is if this meaning is inherent in the construction, or whether it is simply inferred from the situation described. Arguments against the interpretation of the verb as an auxiliary and in favour of its interpretation as a main verb of movement could be the fact that ir is construed as a reflexive, the use of the local preposition phrase introduced by de indicating the origin of the movement, and the contrast with the use of the same expression, pagado, used a few lines earlier in the text, but with the verb ser: (2) Sé que de lidiar bien sodes pagados, (2444). “I know that you are well satisfied with fighting”

Prepositional phrases with a indicating goal also support the sense of movement of the verb. Many early examples, here from the 13th century, carry the sense of a main verb of movement: (3) [en mal dia nasçio la Su alma] que de aquella va perdido para syenpre al logar donde non puede tornar (CORDE, 1293, Anónimo: Castigos y Documentos de Sancho IV) “[. . .] because of his soul he goes lost for ever to the place from where he cannot return”

 Thora Vinther

(4) Et en aquella manera fueron se todos los moros. como desbaratados & dexaron las tiendas con todas sus cosas. (CORDE, 1293. Anónimo: Gran Conquista de Ultramar) “And in this way all the Moors went away, as they were dispersed and left the tents and all their things”

But more periphrasis-like examples also occur in this period: (5) çiertamente la su vida es en poder de sus enemigos; otrosy en su mesura va vendido. (CORDE, 1293. Anónimo: Castigos e documentos para bien vivir ordenados por el rey Sancho IV) “his life certainly is in the power of his enemies and he also ‘goes’ left to their discretion” (6) [. . . este libro porque] todas las cosas que en él sson van ordenadas por cuento de siete (CORDE, 1252–1270. Alfonso X: Setenario) “[. . .] all the things that are in it ‘go’ arranged by seven”

. Grammaticalisation criteria: Grammar As stated by Lehmann (1985: 307), a feature of importance when considering the degree of grammaticalisation of an element is its degree of ‘obligatorification’ (or, in the terminology of Hopper (1991: 22), ‘specialization’), that is, the degree to which the choice among members of a paradigm is constrained by grammatical rules. We can thus see an indication of grammaticalisation in the ir + pp construction from the fact that it is rarely found in the perfective preterite, perfect and pluperfect tenses. The fact that the Spanish preterite forms of the verbs ir and ser coincide (fui, fuiste, fue, fuimos, fuisteis, fueron) might lead to difficulties in tracing the verbal form to the right infinitive. Semantics and the context, however, make it clear that most examples found with the preterite are instances either of the verb ser or of ir as a main verb of movement. In Green’s (1982) modern corpus, 22 cases of the verb ir were drawn from about 200,000 words, and none of the 22 appeared in any of the perfective tenses. However, a small number of counterexamples have been found. Two occur with the preterite, the first of which has been taken from a presentation of the ir + pp periphrasis: (7) Cuando murió su madre, fue vestida de negro durante un año (Fente, Rafael, Jesús Fernández & Lope Feijóo: Perífrasis verbales.

The development of ir into an auxiliary of voice 

Madrid: edi-6 (1983: 40)) “When her mother died, she ‘went’ dressed in black for a year” (8) Cayetano se negó a compartir el espectáculo (. . . ) Nadie, pues, supo si la novia fue vestida o desvestida de blanco, si el novio llegó en angarillas o apoyado en bastón de plata (1988. Sánchez, Héctor: El héroe de la familia) “Cayetano refused to share the show (. . .) Nobody, thus, got to know whether the bride ‘went’ dressed or undressed in white, whether the groom arrived on a handbarrow or leaning on a silver cane”

And an example of the use of perfect has also been found: (9) ha ido y va aun muy frecuentemente acompañada de la inmoralidad (CREA, El Imparcial, 2/2/1822) “ ‘has gone’ and still ‘goes’ very often accompanied by immorality”

Green did not register examples of the conditional either, but this fact may be due to the size of his corpus, as both the form irá and iría are documented elsewhere. I have quoted one example from the Poema de Mio Cid above (example (1): iremos pagados); and later examples are quoted by Matthies (1933): (10) Serás dello más cierto, irás más segurado (1330–1343, Ruiz, Juan: Libro de Buen Amor. Madrid: Biblioteca Calleja (Segunda serie), 1926. Apud Matthies, 1933: 36). “You will be more sure of it, you will go/ feel more certain” (11) Una duda funesta iría siempre unida a mis transportes amorosos (1889, Palacio Valdés, Armando: La hermana San Sulpicio. Paris: Nelson. Apud Matthies, 1933: 56). “A dismal doubt would always ‘go’ together with my ecstasies of love”

Since Fente, Fernández & Feijóo (1983: 40) mention the impossibility of employing the imperative in the accumulative meaning (discussed in Section 5 below), it is relevant to inquire into the occurrence of the imperative in connection with the use treated here. A search in the databases for examples of the imperative forms ve, yd, id + pp gave no results, and therefore it can be assumed that the imperative too is very rarely constructed with a pp. The limitation in the optionality of morphological forms can be taken as a signal of grammaticalisation. As Hopper (1991) explains, the development from lexical word to grammatical element goes through a process of narrowing of the options of a construction. As long as ir has the meaning of ‘displacement’, the verb can appear in all forms and tenses, but as it acquires the function of an auxiliary, some of these options disappear, and it is used only – in this

 Thora Vinther

function – in a limited range of tenses, i.e. the ones that are non perfective. Nevertheless, the few cases of fue + pp show that the grammaticalisation of the ir + pp construction is not quite as advanced as that of ir + a + infinitive, which does not appear at all with the auxiliary in the perfect or perfective tenses (Vinther in press). Another issue within grammaticalisation is the possibility (or impossibility) of separating the pp from the finite verb by one or more words. This is Lehmann’s (1985) criterion of syntagmatic bondedness or, in the diachronic view, the principle of coalescence: the less possibility there is of introducing linguistic elements between two parts of a syntagm, the greater coherence there is between them, and the greater the tendency towards fusion. The argument is that the construction with the sense of movement would allow many words between ir and the pp, while the pp should appear immediately after the form of ir if the use is periphrastic. This may be the case, though it is not easy to demonstrate, as both uses most often appear without intervening words. If something is inserted, it is generally a Subject or an adverb that modifies the pp. And this possibility seems to be current already in the 13th century. Note that two of the three 13th-century examples of the ir + pp construction conceived as verb of movement quoted above (examples (4) and (5)) appear with some words inserted between the two verbs, which minimises the coherence of the construction. The same is the case with the following example from the 16th century: (12) Tú vas ahora por Tracia desterrado, hinchando cielo y tierra con tu queja (CORDE, 1580. Herrera, Fernando de: Comentarios a Garcilaso) “You go now as an exile through Thrace, filling the sky and the earth with your complaints”

But intervening words also appear in this construction whatever the meaning of the verb; for instance, adverbs such as assí, ya, más, tan, del todo, también, aquí, siempre, adelante (“so, now, more, so, completely, also, here, always, further on”). I have also found longer expressions, such as muy particularmente (“very particularly”), de inmortales glorias (“by eternal glories”), en esta memoria (“in this report”), en tres lugares (“in three places”) in the 15th and 16th centuries, and en el mismo librillo2 (“in the same little book”) in the 17th. In the 19th and 20th centuries it seems more usual for adverbs to be inserted. Not only do we find anteriormente, debidamente, aún muy frecuentemente, general-, lógica-, necesaria-, normal- and regularmente (“earlier, duly, very frequently, generally, logically, necessarily, normally, regularly”), as well as the normal short ones bien, muy, muy bien, siempre, casi siempre (“well, very,

The development of ir into an auxiliary of voice 

very well, always, nearly always”) but longer expressions such as en ocasiones, siempre o casi siempre3 (“occasionally, always or nearly always”) also occur. On the basis of his modern corpus, Green (1982) found that a short Subject-NP or single-word adverbial element is inserted in 12.3% of all his verbs, a percentage that supports the idea of interpreting these verbs as passive auxiliaries. As to the specific verb ir, even though intervening elements are not very frequent, the corpora consulted present, as I have shown, instances of inserted words both in old and new texts, and not necessarily short or single words. My data thus run counter Green’s conclusion, and also go against the idea of coalescence. But here we might consider another parameter: it is possible that the weight – or obstructiveness – of the inserted elements is related, not as Green saw it, to their length in syllables or number of words, but to their information load.4 Though my material may not be conclusive, there seems to be a development towards a pattern in modern Spanish that admits two types of insertion between ir and the pp: 1) adverbials of intensity and 2) Subjects with a low information load. The first category consists of examples such as como, bien, muy, debidamente, (“as, well, very, duly”), and also a large group of temporal adverbials, the semantic function of which is not ‘time location’ (i.e., not a factually informative answer to when something happened), but rather an intensification of the fact that it did happen: normalmente, siempre, aún muy frecuentemente, generalmente, lógicamente, (“normally, always, very often, generally, logically”), etc. Consequently, such intervening adverbials can be interpreted as being predominantly epistemic. The second category includes the Subjects that have already been introduced in the discourse, which have the function of ensuring topic continuity. They work as ‘reminders’ of what we are talking about, and are placed somewhere in the sentence without much information weight. It follows from this conjecture that these Subjects have definite determiners, referring to a former mention of the entity: la una, este ejército, este punto, toda la confesión, todo él5 (“one of them, this army, this full stop, the whole confession, all of it”). Of special interest is the following example from the 20th century where we can observe that the indefinite un hombre is used as a synonym to uno, the impersonal pronoun, “that is often an oblique or modest way of saying ‘I’ ” (Butt & Benjamin 2000: 397) which thus carries much less weight than if it actually introduced ‘a man’ in the discourse: (13) [Pepito comentaba despectivamente que] con locas así, va un hombre siempre vendido y dando la cara (CREA, 1953. Casas, Borita: Antoñita la fantástica y Titerris. Madrid: Gilsa).

 Thora Vinther

“[. . .] with such crazy women a man always ‘goes’ in danger and exposed to dangerous things”

The conclusion to these reflections about the material fusion of ir as an auxiliary and the pp is that there is a tendency to place the pp immediately after the finite verb, which suggests syntagmatic bondedness, and that the words that in fact appear in between the two tend to be used in a special, epistemic sense, or with a low information load. Another of Green’s (1982) grammaticalisation criteria is the animate/inanimate character of the Subject. From a logical point of view, if a verb like ir, with the “lexical meaning of an intentional act or process for which ‘being alive’ is a logical prerequisite” (p. 110), occurs with an inanimate Subject, we can speak of a deviant or metaphorical use. Green applies this criterion together with a desemanticisation criterion, seeing the possibility of the verb being combined with an inanimate Subject as a degree of polysemy. His point is that as long as the construction is able to take both an animate and an inanimate Subject, the verb will continue to act as a semi-auxiliary. If, on the other hand, the verb loses the possibility of combining with the original animate Subject, the meaning will be atrophied, and the verb will develop into a true auxiliary. This corresponds with Lehmann’s (1985) criterion of attrition, according to which a sign gradually loses semantic substance. As Green concludes, none of the motion verbs has as yet developed so far that it can take only inanimate Subjects; his suggestion, in fact, is that they may have stabilised as semi-auxiliaries (personal communication). But as I try to show further on, ir + pp with an inanimate Subject seems to be developing a special use that exploits only part of the ‘displacement’ content of the verb. . Grammaticalisation criteria: Semantics A clear change in the semantics of ir when it appears together with a pp indicates a development towards grammaticalisation: in earlier times it expressed movement, which was made especially clear in those cases where the construction was accompanied by the above-mentioned reflexive pronouns or by prepositional phrases with a and de denoting the target or the point of departure of the movement. Over the centuries this relationship has gradually changed, so that today, only cases with explicit indications of displacement are interpreted as main verbs, while all other examples may be seen as constructions that accept more varied contexts in a sense of ‘being’ rather than ‘going’.

The development of ir into an auxiliary of voice 

It is in the area of semantics that the greatest differences between 20th century and older uses have been found. The ir + pp construction, which appears to be somewhat grammaticalised at an early stage, has developed a specialised meaning in modern Spanish, rendering semantic interpretations which are rarely found in older texts. Of interest here is the semantic relationship between the root of the pp and the Subject. Matthies (1933) made this the main issue of his investigation, but only to pursue the idea that ir had not lost its sense of movement completely, but retained it in a sense of ‘development’ or ‘succession’, which is often supported by the meaning of the Subject. In spite of many shrewd explanations, his account is not always completely convincing. It seems likely that early use of the ir + pp construction expressed some kind of movement, even though this has not always meant that the Subject was active. The construction exploited some semantic element of movement in the meaning of the entity that acted as Subject. This use appears especially within the semantic domain of ‘written communication’, where words such as carta (letter), letra (character), escritura, (text), etc. are presented as entities that move towards a receiver when the verb is ir + pp, and merely things that ‘have been written’ when the verb is estar + pp. It appears very often in the 15th century, in letters and legal documents, thereby assuming the character of formulae. The following example, which uses both auxiliaries, is particularly illustrative: (14) fiz escrevir esta carta de enplazamiento, que va escripta en pliego e medio de papel e cosida con filo prieto, e va escripto ençima del primero rrenglón a do dize [“e a vos, Gil Gonçález de Avila, su curador”, non le enpezca, e por ende fiz aquí este mío signo en testimonio de verdat. Gonçalo Gonçález.] Et en las espaldas de la dicha carta estava escripto esto que se sygue: En la Puente del Congosto (CORDE, 1414. Anónimo: Proceso judicial entablado por la ciudad y tierra de Avila) “I ordered this letter of summons which ‘goes’ written on one and a half sheets of paper and sewn with dark thread, and it ‘goes’ written over the first line where it says: [. . .]. And on the back of the letter was written the following: On the Bridge of Congosto”

It should be noted that the two uses of va + the pp of escribir form part of the quoted text ‘within’ the letter, and can therefore be interpreted as the text that moves from the sender to the receiver, while estava escripto is used as a comment about the letter, issued ‘outside’ the actual text and in the past. Some sense of movement is also found in the following example:

 Thora Vinther

(15) [Et, por quanto de suso faze mençión que] va esta carta sellada con mi sello, (CORDE, 1409. Anónimo: Juan Rodríguez, deán de la iglesia de Avila, a causa de sus múltiples ocupaciones, nombra... ) “[. . .] this letter ‘goes’ stamped with my seal”

In the following example, the two instances of the ir + pp expression have acquired such a formulaic character that the meaning of va is difficult to distinguish from está ‘has been’: (16) [en nombre de los dichos pueblos,] lo fiz escrevir, que va escripto en diez planas de papel deste quaderno, con ésta en que va mi signo, et en fyn de cada plana va señalada de la rrúbrica de mi nonbre. (CORDE, 1453. Anónimo: Traslado de un privilegio de Juan II) “[. . .] I had it written, it ‘goes’ written on ten pages of paper of this block, including this one which contains my signature, and on the bottom of each page it ‘goes’ marked with the flourish of my name”

The verb encorporar6 also constitutes a fixed formula in the ir + pp construction in letters, in the sense of ‘is (or ‘goes’) included in the text’: (17) Por ende, ordenamos & mandamos que la dicha ley, que de suso va encorporada, se guarde & cumpla (CORDE, 1495. Anónimo: Ordenanças & premáticas hechas por sus altezas) “Consequently, we order and command that the law mentioned, which ‘goes’ included above, be kept and obeyed” (18) yo, [Gómez Gutiérrez, escrivano público susodicho, en presençia de los testigos yuso escriptos] leý la dicha sentençia que de suso va encorporada de verbo ad verbo (CORDE, 1457. Anónimo: Recolección de castañas) “I [. . .] read the verdict mentioned which ‘goes’ included above word for word”

. Modern use In modern usage, some expressions might perhaps be considered to have a sense of movement, although it is not very prominent. (19) El acontecimiento... está subvencionado en esta ocasión por la Generalitat valenciana. Va acompañado por ‘Seis infancias quebradas’, una exposición en las instalaciones portuarias de fotografías realizadas en Albania por Jons Hilgendag. (CREA, El País 9/6/2000, p. 61)

The development of ir into an auxiliary of voice 

“The event . . . is subsidised on this occasion by the Generalitat of Valencia. It ‘goes’ accompanied by ‘Six broken childhoods’, an exhibition in the harbour of J.H.’s photos from Albania.”

Especially because the construction with ir appears in contrast to está subvencionado, the reader is invited to make a distinction, and to interpret va acompañado as information about a travelling exhibition. In the case of the semantic area of ‘communication’, some sense of movement has been preserved, but not to the same degree as in older texts. The following examples, recorded by Matthies (1933), show the ‘communication’ use of ir as opposed to estar in a ‘non-communication’ context, with diálogos as a more ‘mobile’ object than balcones y ventanas which is more ‘stable’: (20) Estos diálogos cortos iban exornados con una infinidad de miradas furtivas del marido (Cuaderno de Lectura. Madrid: Centro de Estudios Históricos. 1927. Apud Matthies, 1933: 49) “these short dialogues ‘went’ decorated with secret looks from the husband” (21) Todos los balcones y ventanas estaban adornados [con tiestos que rebosaban de flores.] (1889. Palacio Valdés: La hermana San Sulpicio. Paris: Nelson. Apud Matthies, 1933: 49). “all the balconies and windows were decorated [. . .]” (= somebody had decorated them)

But unlike the old texts, the examples of ir firmado in the modern texts do not have a formulaic character and do not stem from letter texts (though example (24) refers to a letter). The use is different, having no trace of movement, but rather a sense of ‘appearing’ in a specific way: (22) Incluso alguna de las letras van firmadas [sic] por su puño y letra. (CREA, ABC Electrónico, 17/10/1997) “Moreover, some of the song texts ‘go’ signed personally” (23) los jarros de Haro tienen [la base ancha, dos asas] y van decorados con [cobalto sucio y algo de manganeso formando siluetas de pájaros y alguna leyenda.] A veces van firmados. (CREA, 1997. Seseña, Natacha: Cacharrería popular. La alfarería de basto en España.) “the jugs from Haro have [. . .] and ‘go’ decorated with [. . .] Sometimes they ‘go’ signed.” (24) la carta,[fechada en Madrid el 17 de julio de 1789], va firmada por el picador de toros Manuel Jiménez (CREA, 1992. Tapia Bolívar, Daniel: Historia del toreo (I) De Romero a “Manolete”) “the letter [. . .] ‘goes’ signed by the bull-fighter M.J.”

 Thora Vinther

The instances of ir + incorporado (cf. examples (17) and (18)) that I have found in the modern database also differ from the old use: they appear in a somewhat technical context that has nothing to do with written communication discourse, and indicate not that the incorporation has taken place, but that the object is actually present in the location mentioned: (25) En el soporte de la maneta va incorporado el regulador de tensión del cable (CREA, 1992. Bueno, Pablo: El libro del mountain bike) “In the bracket of the handle is included the tension regulator of the cable”

The search for va escrito also rendered different examples as compared to the ones cited above (examples (14) and (16)). Either the form of ir can be interpreted as a main verb of movement: (26) [Don Felipe decidió que el canónigo Peña regresara a Sevilla para dar cuenta de todo,] que iba escrito y escondido en la albarda de una mula (CREA, 1994, Ortiz-Armengol, Pedro: Aviraneta o la intriga) “[. . .] which ‘went’ written and hidden in a mule’s saddle bag”

or the ir + pp combination expresses the meaning of ‘appearing in a specific way’: (27) [Otras veces, discretamente, han tapado el maillot de ciertos corredores,] si en ese maillot va escrito el nombre de alguna bebida [distinta a la que en ese caso concreto patrocina el Tour.] (CREA, 1994, García Sánchez, Javier: El Alpe d’Huez) “[. . .] if the name of some drink ‘goes’ written on the shirt”

In the following two examples ir + pp has this special meaning, and seems to be synonymous with the first verb in the first example, aparezca: (28) Se prohíbe que la letra E aparezca pintada en la placa de matrícula, que vaya precedida o seguida de otras letras y que como fondo de las placas o superficies en que aquella letra vaya escrita se pinten los colores de la bandera nacional o cualesquiera otros que el blanco. (CREA, 1985. Anónimo: Código de la Circulación) “It is forbidden for the letter E to appear painted on the number plate or to be followed or preceded by other characters and for the colours of the surface where it appears written to be other than white” (29) [Mariano Baso extiende ante mis ojos un extraño libro.] No va cosido sino doblado, no va escrito sino dibujado, [pero libro es, ¿a quién le cabe la menor duda?] (CREA, 1994. Boullosa, Carmen: Duerme) “It is (‘goes’) not sewn but folded, it doesn’t ‘go’ written but drawn [. . .]”

The development of ir into an auxiliary of voice 

Naturally, the verb estar could have been used instead of ir in these examples, but the sense and the effect would have been different. Apart from the stylistic intensity, which Yllera (1979) also mentions,7 ir gives the pp a different voice: it is neither passive nor active, but could perhaps be considered middle voice. Spanish has no morphologically distinct passive nor middle voice, a fact that, as mentioned in the beginning of this article, has led some linguists to deny the existence of a passive in Spanish. However, while the passive voice is widely accepted as a category expressed either by an auxiliary + the pp or by a reflexive construction with an inanimate Subject (se venden casas “houses are sold (by somebody)”), the middle voice is less accepted as a grammatical category, although it is equally widely exemplified by reflexive constructions, both with inanimate and animate Subjects: la llanura se extiende hacia el horizonte (“the plain extends (itself) towards the horizon”), Juan se levanta (“Juan gets up”). The middle voice expresses an intransitive activity which takes place within the domain of the Subject. According to Kemmer (1993) the semantic properties of the middle voice are not only the ‘subject affectedness’ described by Lyons (1969: 373): “the ‘action’ or ‘state’ affects the subject of the verb or his interests’, but also what she described as ‘the relative elaboration of events’, that is, ‘the degree to which the participants and the component subevents in a particular verbal event are distinguished” (Kemmer 1993: 121). In Spanish this variation, which reflects different conceptualisations of a verbal event as an undifferentiated whole or as an event composed of different parts and structures, is usually expressed by the variant middle se as opposed to active reflexive: se asusta (“he is frightened”) vs. se asusta a sí mismo (“he frightens himself ”), where the first is a relatively non-elaborated event, while the latter can be decomposed as an agent and a patient and what happens to him. I suggest that no va cosido (“it does not ‘go’ sewn”) in example (29) differs from no está cosido in somewhat the same way: va cosido means ‘shows’ or ‘makes its appearance’ with a specific characteristic, without any reference to any external agent and can thus be described as middle voice, while está cosido expresses the result of the activity of some external agent, and is interpreted as passive voice. As mentioned above, the por phrase is the typical agentive of the ser + pp passive, and it does not appear with the normal middle reflexives of the type la ropa se seca (“the clothes dry”). In examples (22) and (24) the impression might be that we have to do with agentives, but a closer look shows that the bearer of the passive meaning is the pp, which at all times (except in the tenses constructed with the auxiliary haber) can be modified by an agentive, but that this modification exploits the passivity of the pp without necessarily bringing

 Thora Vinther

about a passive interpretation of the construction as a whole. This applies both in predicative constructions as María llegó al baile peinada por un peluquero francés (“M. arrived at the dance with her hair done by a French stylist”) and in Object predicative constructions such as Tengo un cuadro pintado por Picasso (“I have a picture painted by P.”). An example from the database: (30) Hoy la perfumería española, [con la firma Antoni Puig, SA en primera línea] ha consolidado brillantemente el difícil camino iniciado por el señor Rodés. (CREA, La Vanguardia, 25/01/1994) “Today perfumery in Spain [. . .] has brilliantly consolidated the difficult journey initiated by Mr Rodés.”

The por phrase in example (22) is still a little puzzling, as it differs from the normal use of the expression for ‘(written) personally’, which is de su puño y letra, and usually not por . . .8 My point is that the change of meaning that has emerged with the ir + pp construction has deprived this verb from the sense of ‘displacement’, but has kept the sense of movement as activity. In the examples cited the main content of the verbal phrase is that the Subject has this set of characteristics in a particularly intense manner, as they are actively being shown off. If my interpretation of the examples quoted is valid, we might conclude that ir + pp in modern Spanish is a periphrasis of voice, not of the passive, which is expressed by ser or estar, but of the middle voice.

. The diachronic development of the middle meaning It is not easy to say when the sense of movement disappeared – or at least lost prominence – in the ir + pp construction. But the development has certainly been very gradual, and I have actually found a very small number of examples in older texts which could be interpreted as middle use of ir + pp: (31) quando va el sol muy alongado entonçes non ha la su agua del el ofiçio que las otras aguas han./ Et quando va el çielo desatenprado en medio delos grandes feruores del sol entonçe salle el njlo & trae la mucha agua. (CORDE, 1272–1284. Alfonso X: General estoria V) “when the sun is very far removed, its water does not have the same job as other waters; and when the sky is unsettled in the heat of the sun, the Nile comes out and brings a lot of water”

The development of ir into an auxiliary of voice 

Both examples – ‘the sun far removed’ and ‘the sky unsettled’ – are descriptions of a specific transient condition of the sun and the sky, and in this use, the grammatical voice could be interpreted as ‘middle’: ‘the sun has receded’, ‘the sky has turned violent’. And from a text three centuries later: (32) [en los cabellos llevava atadas dos borlas de plumas ricas que se llaman quetzalli;] ivan atadas en los cabellos [del medio de la cabeça recogidos como una gavilla de todo lo alto de la cabeça] (CORDE, 1576–1577, Sahagún, Fray Bernardino de: Historia general de las cosas de Nueva España) “[. . .] they appeared bound to the hair [. . .]”

But only in modern texts do examples with a middle interpretation abound.

. The accumulative meaning Before concluding I want to mention another meaning of the ir + pp construction which has not been included in this paper: the one we could call ‘accumulative’. It is not referred to either by Green (1982) or Matthies (1933). But in Gómez Torrego (1988) it is the only meaning suggested for the periphrastic construction: (33) Ya van matriculados quinientos alumnos (Gómez Torrego (1988: 190) “by now, 500 students have registered”

According to Lorenzo (1971: 116) the use of ir as an auxiliary in this construction entails that the passive action indicated by the pp has taken place for some time (the time it takes to carry out the action on the given number of items) and is still going on: (34) Van rescatados veinte cadáveres (Pueblo, 17.2.70. Apud Lorenzo (1971: 116). “Twenty corpses have been released so far”

In this particular sense, the construction seems to be relatively new. The earliest example I have found of the accumulative meaning is from the 18th century: (35) Encargué que me trajesen ahí la razón de lo que va vendido (CORDE, 1792. Fernández de Moratín, Leandro: La comedia nueva) “I ordered them to bring me the report of how much has been sold so far”

The ir + pp construction with accumulative meaning, which is much less documented than the ir + pp construction studied above, seems to follow a grammaticalisation process similar to the one described.

 Thora Vinther

With respect to specialisation, besides certain restrictions on verbal tenses, which are the same as the ones described for the middle use of ir + pp, there seems to be a restriction on the nature of the Subject. As shown in the examples, the support of an indication of measure or quantity is necessary for the accumulative meaning to be clear and, as a consequence, the verb will very often appear in the plural. Fente, Fernández & Feijóo (1983: 40) state that the verb ir in the accumulative construction has a limited use: “El verbo ir, en estos casos, se utiliza generalmente en tercera persona del plural y no se emplea en los tiempos compuestos, indefinido e imperativo”. (“The verb ir in these cases is generally used in the third person plural and not in the compound tenses, the preterite or the imperative”). However, the following example shows that the Subject’s restriction to the plural is only a tendency, as the accumulative interpretation here is supported by an indication of quantity that appears in the singular: (36) No iría mediada aún la lectura cuando doña Luz. . . (1925, Valera, Juan: Doña Luz. Obras escogidas II. Madrid: Biblioteca Nueva. Apud Matthies, 1933: 48). “The reading would not have reached the middle when Doña Luz. . .”

As regards the principle of coalescence, the few examples I have been able to find, all have the form of ir and the pp in close sequence, without any intervening words, thus supporting the idea that the two parts of the construction have a high degree of coherence. As regards semantics, the accumulative meaning seems to be based on the sense of movement of ir, preserved in the idea of a number of entities undergoing, in succession, the action expressed by the pp. A study of a larger number of examples of this special ir + pp construction will show to what degree these tendencies towards specialisation, coherence and semantic bleaching can be interpreted as indicators of grammaticalisation of the verb ir.

. Conclusion With the help of several examples I have shown that the ir + pp construction has some characteristics that push it towards grammaticalisation, most importantly the fact that it rarely appears in other than non-perfective verbal tenses. The construction has a preference for occurring without any intervening words between its finite verb and participle form, though a few words, albeit with

The development of ir into an auxiliary of voice 

downgraded information load, are not excluded. Furthermore, the finite verb seems, in modern Spanish, to have undergone a process of desemanticisation, losing most of its sense of movement. Following Meillet (1912) I have tried to establish whether the word ir in modern Spanish is lexical or grammatical, and I suggest that when it is followed by a pp it seems to have lost some of its grammatical possibilities and lexical characteristics and, to some degree, to act as a grammatical indication of middle voice . On the other hand, in a small number of instances, when the construction appears together with an indication of quantity, grammaticalisation leads to an accumulative interpretation of the construction.

Notes * I wish to express my gratitude to Roger Wright and Lene Schøsler, who read an earlier version of this paper and made very useful comments and suggestions. I also wish to thank the organisers of the Netværk for Grammatikalisering¸ supported by the National Research Council for the Humanities, for providing an inspiring forum for the discussion of topics within grammaticalisation. I am also grateful to Inger Mees for her insightful revision of my English. . Following Yllera I interpret nos as a reflexive. It is, however, also possible to interpret this word as a Subject pronoun, as in Michael’s 1976 edition of the Poema de Mio Cid. This interpretation is indicated by adding an accent on the o: nós. . See Appendix 1 for references. . See Appendix 2 for references. . I am indebted to Eva Skafte Jensen for this suggestion. . See Appendix 3 for references . The CORDE database (< 1975) registers 20 examples of the construction está encorporado/a, and 57 examples of va encorporado/a. . “las perífrasis con verbos de movimiento [. . .] presentan en general un matiz intensivo que les viene dado por su contenido peculiar de verbos de movimiento” (1979: 295) “periphrases with verbs of movement . . . generally present a nuance of intensity due to their specific content as verbs of movement” . A search in the CREA database yielded 88 cases of de su puño y letra, and only 3 cases of por su puño y letra.

References Alarcos Llorach, Emilio (1970). Estudios de gramática funcional del español Madrid: Gredos.

 Thora Vinther

Alarcos Llorach, Emilio (1994). Gramática de la lengua española. Madrid: Espasa Calpe. Butt, John & Carmen Benjamín (2000). A New Reference Grammar of Modern Spanish. London: Arnold. CORDE: Corpus Histórico de la Real Academia Española (<1975): http://www.rae.es/ CREA: Corpus de la Real Academia Española (> 1975) http://www.rae.es/ Fente, Rafael, Jesús Fernández & Lope Feijóo (1983). Perífrasis verbales, edi-6. Madrid. Gómez Torrego, Leonardo (1988). Perífrasis verbales: Sintaxis, semántica y estilística. Madrid: Arcolibros. Green, John (1982). “The status of the Romance auxiliaries of voice”. In Nigel Vincent & Martin Harris (Eds.), Studies in the Romance Verb (pp. 97–138). London: Croom Helm. Hopper, Poul (1991). “On some principles of grammaticization”. In Elizabeth C. Traugott & Bernd Heine (Eds.), Approaches to Grammaticalization, Vol. I (pp. 17–35). Amsterdam & Philadelphia: John Benjamins. Kemmer, Suzanne (1993). The Middle Voice. Amsterdam & Philadelphia: John Benjamins. Lehmann, Christian (1985). “Grammaticalization: synchronic variation and diachronic change”. Lingua e stile, XX (3), 303–318. Lorenzo, Emilio (1971). El español de hoy, lengua en ebullición. Madrid: Gredos. Lyons, John (1969). Introduction to Theoretical Linguistics. Cambridge: Cambridge University Press. Matthies, Werner (1933). “Die aus den intransitiven Verben der Bewegung und dem Partizip des Perfekts gebildeten Umschreibungen im Spanishen”. Berliner Beiträge zur Romanischen Philologie, III (3), 1–66. Jena & Leipzig: Verlag von Wilhelm Gronau. Meillet, André (1912). “L’évolution des formes grammaticales”, reprinted in 1975. Linguistique historique et linguistique générale (pp. 130–148). Paris: Librairie Honoré Champion. Menéndez Pidal, Ramón (1944). Cantar de Mio Cid, I – III: Texto, Gramática y Vocabulario. Madrid: Espasa Calpe. Michael, Ian (Ed.) (1978). Poema de Mio Cid. Madrid: Castalia. Reichenkron, Günter (1933). “Passivum, Medium und Reflexivum in den romanischen Sprachen”. Berliner Beiträge zur Romanischen Philologie, III (1), 1–69. Jena & Leipzig: Verlag von Wilhelm Gronau. Vinther, Thora (In press). “Den spanske konstruktion ir + a + infinitiv i grammatikaliseringsperspektiv”. In Lars Heltoft, Jens Nørgård-Sørensen & Lene Schøsler: Grammatikalisering og struktur. Copenhagen: Museum Tusculanum. Yllera, Alicia (1979). Sintaxis histórica del verbo español: Las perífrasis medievales. Zaragoza: Universidad de Zaragoza.

The development of ir into an auxiliary of voice 

Appendix 1 Virgen María, los mis lloros e gemidos; non vayan assí perdidos, pues son de triste memoria, (CORDE, 1414–1435. Manuel de Lando, Ferrand: Poesías) “Let not, Virgin Mary, my weeping and complaining be thus lost, for they are of sad remembrance” pues mi alegría del todo va ya perdida (CORDE, 1514. Fernández, Lucas: Farsa o quasi comedia... vna donzella y vn pastor y vn caual) ”for all my happiness is now lost” pues en tu compañía el que acierta va más perdido, el que te halla es peor librado, (CORDE, 1539. Guevara, Fray Antonio de: Menosprecio de corte y alabanza de aldea) “for in your company the one who succeeds is more lost, the one who finds you is more unlucky” Y las materias que universalmente se trataban, van tan acompañadas de casos particulares, (CORDE, 1538–1589. Granada, Fray Luis de: Epistolario) “and the topics that were universally treated are so much accompanied by special cases” Ténganse por dicho que van del todo perdidos los que en este mundo son de. . . , (CORDE, 1521–1543. Guevara, Fray Antonio de: Epístolas familiares, libro primero) “Let it be said that those are completely lost who in this world come from. . .” dicen: he lástima, que van también perdidas, (CORDE, 1580. Herrera, Fernando de: Comentarios a Garcilaso) “they say: I feel pity, for they are also lost” avn no se tiene entera notiçia, por lo qual no van aqui escritos, (CORDE, 1573–1581. Aguado, Fray Pedro de: Historia de Santa Marta y Nuevo Reino de Granada) “there is not yet complete information, for which reason they are not written here” avnque por la mayor parte las venas van siempre acompañadas con arterias, (CORDE, 1551. Montaña de Monserrate, Bernardino de: Anothomía) “although for the most part the veins are always connected to arteries” como ira declarado en el memorial que va adelante firmado de mi nombre, (CORDE, 1568–1570. Anónimo: Visita de los valles de Sonqo en los yunka de coca de la Paz) “as it will be declared in the memorial which is signed further on with my name” y en esta escritura va muy particularmente declarado (CORDE, 1519–1526. Cortés, Hernán: Cartas de relación) “and in this writing it is very particularly declared” y en cuanto abraça el mar y cerca el cielo, va de inmortales glorias rodeada, (CORDE, 1567–1597. Herrera, Fernando de: Poesía) “and when it embraces the sea and surrounds the sky, it ‘goes’ encompassed with immortal glories”

 Thora Vinther

va en esta memoria firmada de los dichos procuradores (CORDE, 1519–1526. Cortés, Hernán: Cartas de relación) “it ‘goes’ signed in this memorial by the mentioned attorneys” e va en tres lugares escrito dos vezes (CORDE, 1369. Anónimo: Carta de venta) “and it ‘goes’ written in three places twice” La libranza irá en el mesmo librillo firmada (CORDE, 1605. Cervantes Saavedra, Miguel de: El ingenioso hidalgo don Quijote de la Mancha) “The bill of exchange ‘will go’ written in the same little book”

Appendix 2 Acabo de declarar fuera del texto y en lo que va anteriormente escrito de este Apéndice, (CORDE, 1895. López Bago, Eduardo: El separatista) “I have just declared outside the text and in what ‘goes’ written above in this Appendix” que irá debidamente cumplimentado, firmado y sellado (CREA, 1999). Efímero 99206016 Páginas WEB) “It shall ‘go’ duly fulfilled, signed and stamped” ha ido y va aun muy frecuentemente acompañada de la inmoralidad (CREA, El Imparcial, 2/2/1822) “’has gone’ and still ‘goes’ very often accompanied by immorality” En el grupo nm la articulación de la primera consonante, en la conversación ordinaria, va generalmente cubierta por la de la m. (CREA, 1918–1932. Navarro Tomás, Tomás: Manual de pronunciación española) “In the nm group the articulation of the first consonant in ordinary conversation generally ‘goes’ covered by the articulation of the m” El avance hacia la unión política va lógicamente acompañado de un refuerzo de la democracia. (CREA, 1996. García de Enterría, Eduardo: Tizzano, Un Código de la Unión Europea) “Advance towards political union ‘goes’ logically accompanied by a reinforcement of democracy” Pero no advierten que lejos de inferirse aquella cualidad de esta operacion, en ella va necesariamente envuelta, aunque no expresada, la naturaleza negativa del substraendo respecto del minuendo. (CORDE, 1861–1865. Rey y Heredia, José María: Teoría transcendental de las cantidades imaginarias) “But they are not aware that far from being inferred from this operation, in this quality ‘goes’ necessarily involved, though not expressed, the negative nature of the substrahend with respect to the minuend.”

The development of ir into an auxiliary of voice 

Pero en esta etapa la creatividad no es libre. Toda hipótesis va normalmente encuadrada en el marco de una teoría (CREA, 1998. Morcillo Rubio et al.: Química) “But in this stage creativity is not free. Every hypothesis normally ‘goes’ framed within a theory” los hechos mentales van regularmente acompañados de ciertos hechos físicos (CREA, 1998. VV.AA.: Filosofía. 1o bachillerato) “mental acts regularly ‘go’ accompanied by certain physical acts” Y tampoco va bien servido en lo que se refiere a maletero (CREA, 1994 Anónimo: El Mundo, 22/11/1994: Prueba del coche LEXUS GS 300) “neither does it [the Lexus car] ‘go’ well provided as regards the boot” Intolerancia: “En mi caso va muy unida a la tolerancia. (CREA, 1995). La Vanguardia, 02/11/1995) ”Intolerance: ‘In my case it ‘goes’ much connected with tolerance” La Catrina es una señora muy elegante... siempre va muy bien arreglada. (CREA, 2002. Vargas, Chavela Y si quieres saber de mi pasado) “La Catrina is a very elegant lady. . . she always ‘goes’ very elegantly dressed” Para K. L. KUNZ la actividad analítica va siempre precedida de una comprensión hermenéutica. (CORDE, 1973. Ollero Tassara, Andrés: Derecho y sociedad. Dos reflexiones en torno a la Filosofía) “For K. L. Kunz analytic activity always ‘goes’ preceded by a hermeneutic understanding” la diabetes en esta edad va casi siempre unida a estados más o menos intensos (CORDE, 1919–1936. Marañón, Gregorio: Climaterio de la mujer y el hombre) “At this age diabetes nearly always ‘goes’ connected with more or less intense conditions” Esta sensación de cansancio va en ocasiones acompañada de nerviosismo (CREA, 1995. Gómez Montoya, Adolfo María: La cirugía estética) “This feeling of fatigue at times ‘goes’ accompanied by nervousness” van siempre o casi siempre acompañadas de otra fruta (CREA, 1992. Domingo, Xavier: El sabor de España) “they ‘go’ always or nearly always accompanied with other fruit”

 Thora Vinther

Appendix 3 y entonces dezimos que va la vna perdida assi como dixo don jorge (CORDE, 1481– 1496. Encina, Juan del: Cancionero) “and then we say that one ‘goes’ lost as don Jorge said ” viendo que va este ejército perdido, la lástima me mueve (CORDE, 1589–1594. Ercilla, Alonso de: La Araucana) “seeing that this army ‘goes’ defeated, compassion moves me” Por tanto será V. M. servido de que vaya este punto desterrado (CORDE, 1631. Robles, Juan de: El culto sevillano) “Therefore your honour will be pleased that this full stop ‘goes’ taken away” unas llaves y una espada, y a este modo va toda la confesión escrita por imágines (CORDE, 1590. Acosta, José de: Historia natural y moral de las Indias) “keys and a sword, and this way the whole confession ‘goes’ written by pictures” verá que va todo él enderezado a formar un hombre (CORDE, 1550. Arce de Otárola, Juan de: Coloquios de Palatino y Pinciano) “will see that it ‘goes’ all directed to form a man”

The development of continuous aspect Kazuha Watanabe Cornell University

.

Introduction

The universality of the development of aspect markers has been the subject of various investigations. These usually divide aspect markers into two categories, namely perfective and imperfective. However, this binary classification cannot account for the grammaticalization patterns in Japanese, Korean, Newar, and Parji which have a single marker (referred to as ‘continuous’1 in this paper) that signifies both progressive and resultative/perfect. In this paper, I investigate the characteristics of the continuous markers of these languages, and identify the mechanism through which these markers have developed.

. Definitions of aspect The aspectual terms discussed in this paper are 1) progressive, 2) resultative, 3) perfect, 4) continuous, and 5) continuative. The definitions of the terms differ slightly from conventional usage. First, progressive is used only when indicating an on-going action, such as ‘I am walking in the park’. However, it excludes the preliminary stages of an action, such as ‘the train is arriving’, which indicates that the arrival of the train has not happened. It also excludes the future reading, such as ‘I am going to school tomorrow’. The resultative aspect indicates the result of an event which remains in the reference time, such as ‘he has gone to Paris’; i.e. the action of going to Paris has taken place and the result of the action (i.e. that he is in Paris) pertains at the reference time. The perfect aspect indicates a situation, which occurs prior to the reference time, and is relevant to the situation at reference time, such as ‘he has sent the letter

 Kazuha Watanabe

(so that I do not need to do it anymore)’. Note that the relation between the action prior to the reference time and the state at the reference time indicated by the perfect aspect can be indirect, while the action and the state signaled by the resultative must be directly related. The continuous aspect signifies both progressive and resultative, whereas continuative indicates progressive, resultative and perfect. Both continuous and continuative may indicate other aspectual meanings such as habitual or iterative. However, the integral part of the definitions is that they indicate both progressive and resultative/perfect.

. The development of aspect Bybee et al. (1994) examine the development of aspect markers in 95 languages selected from various language families in order to identify the directionality of the grammaticalization. They first classify aspect markers into two categories, namely 1) completive, perfect, resultative and perfective, and 2) progressive, iterative, habitual and imperfective. They conclude that each group takes a distinctive path of semantic shifts, and that the path of the semantic changes is based on the meanings of the lexical items that the aspect markers originate from. They provide the following figures to illustrate each path of change.2 However, this binary classification of aspect markers does not capture the history of the aspect systems of languages such as Japanese, Korean, Newar, and Parji. The commonality among these languages is that they all possess a single aspect maker which expresses both progressive and perfect.3 That is, the development of these markers cannot be explained by any models of semantic change where the paths leading to perfective and those of imperfective never intersect. locative expressions

progressive

continuative

imperfective

Figure 1. The path of change leading to imperfective (from Bybee et al. 1994: 139) ‘be’ / ‘have’

resultative

‘come’ ‘finish’ / ‘be away’

perfect

perfective/past

completive

Figure 2. The paths of changes leading to perfective (from Bybee et al. 1994: 105)

The development of continuous aspect 

Although Bybee et al. also discuss various mechanisms for the semantic shifts, such as metaphorical extension, inference (conversational implicature), or generalization, they do not thoroughly explore the syntactic motivations for the change. They briefly state that the choice of paths of change, leading to perfective, depends on the type of predicating verb (whether active or stative) the aspect markers occur with (Bybee et al. ibid.: 51). However, I will focus on the correlation between aspect markers and verb types, i.e. the semantic property of each aspect marker as determined by the semantic types of co-occurring verbs. It is essential to understand this interaction in order to determine the function of the aspect marker properly as well as to identify the paths of semantic change involved.

. Verb types and aspect markers The relation between aspect markers and semantic types of verbs has been discussed in various investigations. For instance, Dowty (1977) notices the ‘imperfective paradox’, where English be+-ing, when used with achievement verbs, indicates the preliminary stage of the event,4 rather than an on-going action. Smith (1991, 1997) proposes five situation types, each of which is assigned to a verb constellation, rather than a verb itself, and provides analyses of the relation between the semantics of verbs and the aspect markers, including the English be+-ing form. Smith’s situation types are based on Vendler’s categorization, which uses telicity, duration and stativity as the features of each semantic type. The following chart illustrates these five categories and their semantic properties:

state activity accomplishment achievement semelfactive

state + – – – –

durative + + + – –

telic N/A – + + –

Figure 3. Smith’s verb types

She states that ‘[the] basic use of the English progressive focuses on the internal stages of durative, non-stative situations’ (Smith 1997: 171), providing the examples using activity and accomplishment situations as follows (ibid.: 171):

 Kazuha Watanabe

(1) Mary was walking in the park.

(activity)

(2) Sam was eating an apple.

(accomplishment)

On the other hand, the progressive form cannot co-occur with a stative situation as in (3) (Smith 1997:171). (3) *He is knowing the answer.

(stative)

She also notices that the progressive forms of semelfactive verbs indicate multiple actions, while some accomplishment and achievement verbs indicate a preliminary stage of an event as follows (Smith ibid.: 172): (4) Jane was knocking at the door.

(semelfactive)

(5) They were building a house.

(accomplishment)

She observes that English be+-ing expresses various aspectual meanings depending on the semantics of the other components of the sentence. However, it is problematic to assign the semantic types to the verb constellation instead of the verb itself, since some verbs are categorized into two different types depending on their arguments. For instance, the verb ‘eat’ can be categorized either as activity (as in ‘Alice is eating dinner’) or accomplishment (as in ‘Alice is eating an orange’). On the other hand, Olsen’s (1997) verb types, which allow the semantic features to be underspecified, enable us to categorize verbs into a single type. Figure 4 illustrates her proposal:5 dynamic state activity accomplishment achievement semelfactive

+ + + +

durative + + +

telic

+ +

Figure 4. Olsen’s verb types

Watanabe (2000: 5–8) discusses in detail the advantages of Olsen’s analysis. Watanabe’s major claim is that Olsen’s approach enables us to distinguish the semantic property of a verb from that of a whole sentence, which is influenced by both linguistic and extra-linguistic factors. Furthermore, Olsen’s underspecified classification allows for the verb types to constrain the distribution of grammatical aspect. In other words, it provides an explanation for

The development of continuous aspect 

the sentences with ambiguous aspectual meanings, since the sentences with aspectual ambiguity contain verbs with non-marked features. Therefore, all the verb types in this paper are based on Figure 4, which allows us to identify the aspectual meaning of a sentence in any language by determining the semantics of the verb itself.

. Continuous aspect Some researchers have compared the continuous aspect in various languages. For instance, Shirai compares Japanese, Korean, Chinese, and English, and claims that all of these aspect markers signify both progressive and resultative. While his description of Japanese -te iru is accurate, it is not certain if the examples he quotes for the other languages support his argument. For instance, he claims that Chinese -zhe, which is described as ‘imperfective focusing on the resultant state of an event’ (Sun 1998: 155), signifies the progressive and resultative aspect, citing the following example (Shirai 1998: 668 after Smith 1991: 359): (6) Quiang chang guo-zhe ji zhang huar wall on hang-asp several class picture ‘Several pictures are hanging on the wall.’ (7) Women bici shen ai-zhe, bici yilai-zhe we each other deep love-asp each other rely-asp ‘We deeply love each other, and relied on each other.’

He concludes, based on these two examples, that -zhe now signifies the progressive aspect while its original meaning is resultative. However, there is a possibility that both (6) and (7) are examples of the resultative aspect, even though their English translation uses the progressive form. That is to say, -zhe simply indicates the resultative aspect, instead of signifying both resultative and progressive. Li and Thompson (1981: 220–221) observe that the verbs ‘hang’ and ‘put on/wear’ in Chinese are active verbs. They point out that the verb ‘hang’ or ‘put on’ can indicate ‘a state associated with the activity’, i.e. resultative, when used with -zhe. On the other hand, they also state that -zhe cannot be used with active verbs such as ‘write’ in order to express progressive aspect (i.e. write + -zhe means ‘is written’, rather than ‘is writing’). That is, ‘hang’ or ‘put on’ plus -zhe means resultative aspect, although the English translation uses the progressive form.

 Kazuha Watanabe

Shirai’s English examples have a similar problem. He claims that sentences such as ‘he is lying in the bed’ or ‘the picture was hanging on the wall’ are resultative. However, the verbs ‘lie’ or ‘hang’ are stative verbs, whose [+dynamic]/[+telic] counterparts are ‘lie down’ and ‘hang up’. That is, these English sentences are examples of imperfective use of be+-ing, not resultative. Chafe (1980) examines Seneca (and Onondaga) and briefly mentions the possibility that the Iroquoian aspect system and that of Japanese share some similarity. He points out that -te iru in Japanese as in tabe-te iru, which indicates ‘(one) is eating’ or ‘(one) has eaten’, shows the same aspectual ambiguity as the Seneca counterpart ho:kõh, which means ‘he is eating’ or ‘he has eaten’, does. Ebert (1995) examines 16 languages to search for a connection between perfect and progressive. She concludes that there is no significant relation between the two, claiming that very small numbers of languages have an aspect marker that results in the ambiguity between perfect and progressive for all types of verbs. However, since the semantic requirement of perfect differs from that of progressive, it is impossible to see the ambiguity for every verb. In other words, the meanings of some verbs are not compatible with the perfect reading, whereas other verbs do not go with the progressive reading. The comparison that I make in this paper attempts to provide a system for identifying the aspectual meaning of a sentence, while taking advantage of the typological approach to the puzzling fact that a single aspect marker signifies both perfect/resultative and progressive. I also mention the diachronic aspect of the topic, which none of these previous works mention.

. Japanese The Japanese aspect marker -te iru has received much attention from researchers, because of its semantic complexity. For instance, Kindaichi (1950) notices the ambiguity of -te iru, and classifies Japanese verbs into four categories: stative, durative, instantaneous, and the ‘fourth’ kind. He states that -te iru denotes resultative aspect when used with instantaneous verbs, progressive with durative verbs. In addition, while the stative verbs never co-occur with -te iru, the fourth kind of verb must always occur with -te iru. Ogihara (1998) observes that -te iru indicates either an on-going or resultative state when used with durative verbs, while it signifies only a resultative state with instantaneous verbs. Shirai (1998) claims that -te iru indicates progressive when used with ‘a dynamic durative situation, i.e. activity and accomplishment’,6 while it signi-

The development of continuous aspect 

fies the resultative aspect when used with a punctual change of state verbs, i.e. achievement verbs’ (p. 663). Let us evaluate the proposals that previous research presents. While we can agree that the progressive reading requires a [+durative] [+dynamic] situation, the claim that [–durative] is the trigger to induce the resultative reading is problematic. The following examples demonstrate that [+telic], rather than [–durative], brings about the resultative reading. (8) Ken-ga doa-o nokku shi-te iru. Ken-nom door-acc knock-te iru ‘Ken is knocking on the door’ (semelfactive verb) (9) Ken-ha mou kaijou-ni ki-te iru. Ken-top already venue-loc come-te iru ‘Ken has already come to the venue.’ (accomplishment verb)

Example (8) contains the semelfactive verb ‘knock’, which is [–durative] (acc. Smith) or unmarked for duration i.e. used as [–durative] in this context (acc. Olsen). That is, whichever framework we adopt, a [–durative] situation does not necessarily give rise to resultative aspect. On the other hand, example (9), which contains an accomplishment verb ‘come’ brings about a resultative reading, even though the verb is [+durative]. Note, however, that while the verb ‘knock’ here is used as [–telic] ‘come’ is a [+telic] verb. Therefore, telicity, rather than duration, is the key for determining the distinction between resultative and progressive. This syntactic feature of -te iru helps us to determine its paths of historical development. The construction -te iru consists of the conjunctive particle -te plus the verb iru, ‘exist’. In Classical Japanese, iru, whose phonological form was wiru, signified ‘sit’ or ‘stay’. Watanabe (2002) examines the construction V1 +wiru, which appears to have two functions. In example (10), narabi ‘line-up’ plus wi ‘sit’ (wi is the conjunctive form of wiru) indicates a sort of compound verb ‘sit in line’. (10) nihodori-no futari narabi-wi katarahi. . .. grebe-nom two line up-wiru talk ‘Two grebes are sitting in line, and talking (to each other). . . .’7

However, wiru in (11) cannot mean ‘sit’, since the meaning of the co-occurring verb contradicts that of wiru. (11) fusi-wi nageke-do aki tara-nu. . . lie down-wiru weep-though satisfy enough-not ‘I was laid on my face, weeping, but cannot weep enough. . . .’8

 Kazuha Watanabe

In this example, V1 +wiru seems to indicate resultative meaning; that is, fusi-wi indicates ‘I have laid down (so that I am lying now)’. However, Yanagida (1990), among others, claims that V1 +wiru signifies the progressive aspect in Classical Japanese, without providing any evidence to support the claim. Watanabe examines all the occurrences of V1 +wiru in Manyooshu and finds that all the V1 are achievement verbs (i.e. [+telic]), used as [–durative], where the progressive reading is impossible. That is, this construction was a resultative marker. In Genji Monogatari,9 wiru co-occurs with both [–telic] and [+telic] situations; i.e. some of these examples must indicate the progressive aspect. In addition, there are examples which use the conjunctive particle -te between the verb and wiru (i.e. V1 -te+wiru). The use of the conjunctive particle for combining two verbs appears to be an independent syntactic change from the aspectual development. In fact, there seems to be no semantic difference between V1 +wiru and V1 -te+wiru. Yanagida (1990) points out that while there is no example of V1 -te+wiru in Manyooshuu, there are 49 in Genji Monogatari (266 V1 +wiru), and 79 in Shiki Toogen Shoo10 (no example of V1 + wiru). In this analysis, V1 +wiru and V1 -te+wiru are treated as having the same aspectual meaning. The difference between the semantics of V1 s in Manyooshuu and those in Genji Monogatari suggests that while the construction signified resultative aspect in Manyooshuu, it signaled both resultative and progressive (i.e. continuous) in Genji Monogatari. The following figure schematize this change. To summarize, in Modern Japanese, -te iru brings about the resultative reading when used with [+telic] situations, while it yields to the progressive reading when used with [+durative] situation. In addition, -te iru was originally a resultative marker, which could occur only with [+telic] situation, but expanded its syntactic range to both [–telic] and [+telic], which results in it having both progressive and resultative meanings.

Manyooshuu resultative

Genji Monogatari continuous (resultative +progressive)

Figure 5. The change in the semantics of V1 -te+wiru

The development of continuous aspect 

. Other languages . Newar11 Kiryu (1999) points out that the verb cwan-e, which means ‘stay’, ‘sit’, or ‘remain’ in Newar, is used to express both progressive and resultative aspects when it is used with a gerundive form of a verb, providing the following examples: (12) R¯am-ã saphu: bwan¯a cwan-a Ram-erg book-abs read-ger stay-pd ‘Ram is reading a book.’ (13) Thawa kh¯a sin¯a cwan-a This hen die-ger stay-pd ‘The hen is dead’

This construction in Newar shares two major characteristics with the other continuous aspect markers. First, Kiryu claims that cwan-e obtains a progressive reading when it is used with an ‘activity’12 verb, whereas it gives a resultative reading when used with ‘change of state’ verbs. That is to say, [+durative] situations (‘read’ in the example above) give rise to a progressive reading, while [+telic] situations (‘die’ in the example above) bring about a resultative reading. Second, transitive sentences yield a progressive reading, while the intransitive counterparts bring about a resultative reading: (14) B¯abur¯ajã jhy¯a cae-k-¯a cwan-a Baburaja-erg window open-caus-ger stay-pd ‘Baburaja is opening the window’ (15) Jhy¯a: c¯ae-k-¯a cwan-a window open-caus-ger stay-pd ‘The window was being opened’

Kiryu does not discuss the historical development of this construction. However, he points out that the verb ta-ye ‘keep’ is more productive for forming a resultative reading. (16) Jhy¯a: c¯ae-k-¯a tal-a window open-caus-ger keep-pd ‘The window has been opened’

Therefore, it is possible that cwan-a, which originally signified the progressive aspect, is currently expanding its meaning to the resultative aspect.

 Kazuha Watanabe

. Parji Parji is a central Dravidian language spoken in Bastar, India (Burrow & Bhattacharya 1953; Lyovin 1997). Steever (1984, 1993) observes that the verb ma (glossed here as ‘be’) plus the perfect form of a verb denotes both progressive and perfect as follows (Steever 1984: 632–633). (17) ok manjen and kiter m˜edar one man-acc kill-pp-3p be-fut-3p ‘They have killed a man’ (perfect) (18) ayal ok p¯apin oloto m˜eda. . . . woman one child-acc embrace-pp-3n be-fut-3n ‘A woman is embracing a child. . . .’ (progressive) (19) d u˜ meramti pakato m˜eda tiger jungle-loc lurk-pp 3n be-fut-3n ‘The tiger is lurking in the jungle.’ (progressive)

Steever claims that verbs that are ‘inherently imperfective’13 yield a progressive interpretation whereas verbs that are ‘inherently perfective’14 give rise to one of perfect aspect. We can reinterpret this observation using our verb types; the ‘inherently imperfective’ verbs, i.e. the verbs unmarked for telicity, are interpreted as progressive aspect, while ‘inherently perfective’ verbs, i.e. [+telic] verbs, are interpreted as perfect aspect. He does not provide any examples showing ambiguity between progressive and perfect. However, if our prediction is accurate, it will be possible to find such examples. Steever also examines the present perfect forms in other sister languages, such as Pengo, K¯ui, K˜ur ux and old Tamil. However, all the perfect forms in these languages he provides are of telic verbs, such as ‘break’ (¯en es-k-an ra‘-ckan ‘I had broken it’ in K˜ur ux) or ‘die’ (oz int-an-an-¯e ‘he has died’ in Old Tamil). Burrow and Bhattacharya (1970) also describe the perfect forms of Pengo, all of which involve telic verbs. Although we cannot conclude with certainty that none of the perfect forms in these languages can signify the progressive aspect, the fact is that the accomplishment verbs (i.e. [+telic] and [+durative]) such as ‘come’ or ‘enter’, as in v¯atini] ‘they have come’ or h¯otanader ‘you have entered’ (Burrow et al. 1970: 68) seem to produce only the resultative and perfect aspect. That is, the perfect plus the verb ‘be’ in Dravidian languages originally indicated the resultative aspect. The analogous construction in Parji subsequently gained the progressive meaning, which is identical to the development of Japanese -te iru.

The development of continuous aspect

. Korean In Korean, while -ô iss usually indicates perfect aspect, -ko iss signifies progressive. Lee examines the aspect marker -ko iss in Korean, which is usually defined as the progressive marker, as follows: (20) Chelswu-ka cip-ul cis-ko iss-ta Chelswu-nom house-acc build-prog-dec ‘Chelswu is building a house.’

She then provides two different contexts where the verb ‘open’ is used. While ko iss gives rise to progressive aspect in example (21), it brings in the resultative reading in (22) (Lee 2001: 525). (21) Chelswu-ka changmwun-ulu yel-ko iss-ta Chelswu-nom window-acc open-prog-dec ‘Chelswu is opening the window’ (22) Chelswu-ey pang-un com taptaphayse Chelswu-nun Chelswu-poss room-top a little closed-in-conj Chelswu-top pang-ey iss-ul ttay hangsang changmwun-ul yel-ko iss-ta room-loc be-when always window-acc open-prog-dec ‘Chelswu’s room is a little closed-in, so when he is in his room, he always has the window open (on him).’

She also provides another example of -ko iss resulting in a resultative reading with the help of pelsse ‘already’. (23) Chelswu-ka cake han kay-lul pelsse ta mek-ko iss-ta Chelswu-nom cake one-acc already all eat-prog-dec ‘Chelswu has already eaten the whole cake.’

Wako, Sato and Horie (2002) also point out the perfect function of -ko iss. They notice that -ko iss indicates the perfect aspect in newspaper articles when used with reporting verbs, such as ‘note’, ‘notice’ or ‘state’, as follows: (24) Camocen-i-la-nun chelcen-ul mantul-e ssu-n cop-quot-adn metal money-acc make-conj use-adpast kes-i cwuhwa-uy hyosi-la-ko (. . . ) ‘Haytongyeksa’-nun noml-nom currency-gen beginning-quot-conj -top cenha-ko iss-ta. report-KO ISS-decl ‘Haytongyeksa (. . . ) has reported that the creation and use of metal money called camocen attested the first use of currency. . . .’



 Kazuha Watanabe

The perfect reading that Wako et al. mention suggests that the progressive marker -ko iss must have taken on a function that is very common for the present tense markers in various languages, which can be called ‘reporting present’, that is, the present tense marker used for describing past events in specific circumstances (such as academic writings or newspapers). This new function eventually led to the reinterpretation that -ko iss signifies the perfect. In addition, as Lee (ibid.) points out, the use of progressive marker (which requires [+durative] situations) with achievement verbs (which can be [+durative] or [–durative] depending on the context), may cause the co-occurrence of the progressive marker and a [–durative] situation where the progressive reading is not possible. This mismatch must have led to the resultative reading, because of the semantic similarity between progressive and resultative; i.e. both of the aspects indicate a static or open-ended situation.15 The Korean -ko iss provide us with an important piece of evidence for determining the paths of two different developments. While Japanese -te iru developed from resultative to continuous, -ko iss changed from progressive to continuous. However, both of the developments share something significant. That is, while both resultative and progressive require more specific syntactic features (i.e. co-occur with a small range of verbs), continuous aspect requires less specific syntactic features (i.e. co-occurs with a wide range of verbs).

. Conclusion The comparison of four languages illustrates a synchronic system of aspect that possesses only one marker, the ‘continuous’, for signifying both progressive and perfect/resultative. That is, the semantic features of co-occurring verbs, especially durativity and telicity, determine which reading is brought about. To be specific, a [+durative] situation leads to a progressive reading, while a [+telic] situation gives rise to a resultative reading. Therefore, a situation which is both [+durative] and [+telic] results in aspectual ambiguity. In addition, while the development of some of these markers remains unknown, we can conclude that there are two possible paths to the birth of the continuous marker. The fact that both a progressive and a resultative marker can develop into a continuous aspect enable us to combine the two separate schemas that Bybee et al. propose into one unified picture as follows below. One path is from resultative aspect, which requires [+telic] situations, to continuous, which accepts both [+telic] and [–telic] situations. The other path is from progressive, which requires [+durative] situations, to continu-

The development of continuous aspect locative expressions

progressive

‘be’ ‘have’

resultative

‘come’ ‘finish’ ‘be away’

continuous

continuative

perfect

imperfective

perfective

completive

Figure 6. The development of continuous aspect

ous, which allows [–durative] situations. That is, both paths involve a change whereby an aspect marker that occurs only with a limited semantic range of verbs expands its syntactic freedom to occur with a wide range of verbs. I believe that this general mechanism of change in aspectual meanings may apply to the development of other aspect markers.

Notes . The term ‘continuous’ has been used for various purposes in previous works on aspect. For instance, Bybee et al. (1994: 127) define ‘continuous’ as progressive/stative, indicating ongoing situations, and ‘continuative’ as ‘keeping on doing’ something. However, I shall use this term strictly for aspect markers which indicate both progressive and resultative/perfect. . Figures only include the major aspectual categories. Some of the categories they provide are omitted, since they are irrelevant to the discussion of this investigation. . Some of the markers seem to express resultative instead of perfect. . This term is coined by Smith (1991). . She also includes ‘stage-level’ state, which is omitted here. See Watanabe (2000) for the argument. . Shirai adopts the verb categories that Smith (1991) proposes. . From Manyooshuu, the oldest Japanese poetry anthology, which contains about 4500 poems in 20 volumes. The poems are dated ca. 630–760 AD. This poem is number 794 in Vol. 5, dated 750-753. . Poem number 204 in Vol. 2. . ‘The Tale of Genji’, which consists of 54 volumes. Dated between 1001–1021. . By Toogen Zuisen. Dated 1455–1460. . The language is also called Newari . He does not define what types of verbs are called ‘activity’ verbs. . He defines these as referring to verbs like English ‘run’ and ‘break’.



 Kazuha Watanabe . Cited from Lee (2001: 521). Although this example does not strictly indicate the progressive aspect (preliminary stage), I have adopted it as the standard progressive example, since it is similar in usage to the English progressive. . Smith also points out that both progressive and resultative indicate stable situations that do not include the end point of the situation (1997: 84)

References Burrow, Thomas, & S. Bhattacharya (1953). The Parji Language: A Dravidian Language of Bastar. Stephen Austin: Hertford. Bybee, J., R. Parkins, & W. Pagliuca (1994). The Evolution of Grammar: Tense, aspect and Modality in the Languages of the World. Chicago: The University of Chicago Press. Chafe, Wallace, L. (1980). Consequential Verbs in the Northern Iroquoian Languages and Elsewhere. In K. Klar, M. Langdon, & S. Silver (Eds.) Trends in Linguistics: Studies and Monographs 16; American Indian and Indoeuropean Studies – papers in Honor of Madison S. Beeler (pp. 43–49). The Hague: Mouton. Dowty, David (1977). Toward a Semantic Analysis of Verb aspect and the English ‘Imperfective Progressive’. Linguistics and Philosophy, 1, 45–77. Ebert, Karen H. (1995). Ambiguous Perfect-Progressive Forms across Languages. In P. M. Bertinetto, V. Bianchi, Ö Dahl, & M. Squartini (Eds.) Temporal Reference, aspect and Actionality, Vol. 2: Typological Perspectives (pp. 185–203). Torino: Rosenberg & Sellier. Kindaichi, Haruhiko (1950). Kokugo dooshi-no Ichi Bunrui. (A Classification of Japanese verbs). Gengo Kenkyuu, 15, 48–63. Kiryu, Kazuyuki (1999). Conceptualization and aspect in Some Asian Languages. In W. Abraham & L.Kulikov (Eds.) Tense-Aspect, Transitivity and Causativity (pp. 43–62). Amsterdam: John Benjamins. Lee, EunHee (2001). The Meaning of the progressive -ko iss in Korean. In S. Kuno, I. Lee, J. Whitman, J. Maling, Y. Kang, & Y. Kim (Eds.) Harvard Studies in Korean Linguistics, IX, 521–530. Dordrecht: Kluwer Academic. Li, Charles N., & Sandra A. Thompson (1981). Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Lyovin, Anatole V. (1997). An Introduction to the Languages of the World. New York: Oxford University Press. Mithun, Marianne (1999). The Languages of Native North America. Cambridge: Cambridge University Press. Ogihara, Toshiyuki (1998). The Ambiguity of the -te iru Form in Japanese. Journal of East Asian Linguistics, 7, 87–120. Olsen, Mari Broman (1997). A Semantic and Pragmatic Model of Lexical and Grammatical Aspect. New York: Garland Publishing. Shirai, Yasuhiro (1998). Where the progressive and the resultative meet – Imperfective aspect in Japanese, Chinese, Korean and English. Studies in Language, 22 (3), 661–692. Smith, Carlota, S. (1991 (second edition 1997)). The Parameter of Aspect. Dordrecht: Kluwer Academic Publisher.

The development of continuous aspect

Steever, Sanford, B. (1984). The Evolution of the Present Perfect in Pengo. Journal of the American Oriental Society, 104 (4), 621–648. — — — (1993). Analysis to Synthesis: The Development of Complex Verb Morphology in the Dravidian Languages. New York: Oxford University. Sun, Chaofen (1998). Aspectual Categories that Overlap: A Historical and Dialectal Perspective of the Chinese -zhe. Journal of East Asian Linguistics, 7, 153–174. Vendler, Zeno (1967). Linguistics in Philosophy. Ithaca: Cornell University Press. Wako, Masakazu, Shigeru Sato, & Koaru Horie (2002). From progressive to Perfect: A CorpusBased Study of the “Perfect” Meaning of the Korean progressive form -ko iss-. Paper presented at the 12th Japanese/Korean Linguistics Conference. Watanabe, Kazuha (2000). The grammaticalization of aspect in Japanese. Unpublished M.A. thesis. University of California, Davis. — — — (2001). The Origin of progressive and resultative Markers in Japanese. Unpublished manuscript. Yanagida, Seiji (1990). Kindaigo no Shinkoutai, Kizentai Hyougen. (The expression of progressive and resultative in the modern language.) In Kindaigo Gakkai, Kindaigo Kenkyuu, 8, 3–27.



Index

A accommodation 87 accumulative meaning 283, 293, 294 adverbial raising 33 African-American Vernacular English 64, 69, 70 agglutination 132 analogy 105–107, 111 anaphors 200, 202, 203 aspect 301–303, 305–309, 311–314 asymmetry 179 auxiliary 31, 37–39, 45, 115–118, 120, 121, 126, 127, 280, 281, 283, 284, 286, 291, 293

clitic 200, 204–208, 213, 215–217, 219 clitic pronouns 108, 111 conditional 225, 227–233, 235, 237 configurational syntactic type 14 conservative dialect 21 contamination 24, 25, 28 continuative 302, 312 continuous 301, 302, 305, 309, 312 copula 61–71, 116, 117, 119, 121, 123, 124, 127, 128 countertonic principle 83

B bara 225–237 biflexional 4 binary weight distinction 242 binding 218 Björketorp 243 borrowings 76, 77, 82, 83 Brazilian Portuguese 47

D deletion processes 241 deontic 258, 264, 265, 267–272, 274 desiderative 229, 230, 232, 233, 235, 237 devoicing 104 diachronic analysis 281 dialect 62, 70, 71 distributive 3 ditransitive 179, 181, 183

C Capgrave 90 case loss 3 case marking 1 causal approach 153 change 103–108, 110–112 Chaucer 76, 78, 79, 87 chronologies 57 Chukchi 115–120, 124–126, 128, 129 Chukotko-Kamchatkan 115, 117, 119–121, 123–127, 129

E e 162, 163, 165–175 Early Middle English 75 Early Old Swedish 258, 262, 267, 269 emphatic 202, 205, 213 epistemic 258–265, 267–274 erosion 106–108, 111 et 159–165, 169, 173, 174 ethnoreconstruction 21, 22, 24–28 evidentiality 44 expectation 161, 162, 172, 174

 Index

F fast speech 108, 110, 111 French stress 77, 78 frequency 105, 107, 108, 111 G Germanic foot 75, 79 gram 228, 234, 235 grammatical morphemes 132, 133 grammaticalization 31, 38–41, 43–45, 106, 107, 111, 115, 121, 125–128, 130–137, 181, 210, 226, 233–237, 260, 261, 263–265, 267, 274, 279, 280, 282–284, 286, 293–295, 301 H habitual 61, 67, 70, 71 hypercorrection 110 hypothetical clauses 230 I iambic shortening 246 iar(˘a) 167, 172, 173, 174 implicature 160, 162, 164–168, 170, 172–174 infinitival nexus 32 intensifier 199, 201, 203, 204, 206, 210–216, 218 Itelmen 115, 117–129 K Kok-Papónk 21–28 Koko-Bera 21–25, 27, 28 Kragehul spearshaft 243 L lär 257–274 lära 257, 261–264, 271, 272 läter 258, 261–263, 266, 268–270, 272, 274 lata 257, 261, 262, 264–272, 274 Late Old Swedish 258, 260, 263, 272

Latin 131–137 Latin stress 77–79 Latinate stress 75 latir 261, 262, 267 lexicon 136 light verbs 188 London sociolects 87 låta 257, 258, 261, 262, 264–266, 268–272, 274

M Mainland Scandinavian 1 marker 159–162, 164–171, 173–175 medial laxing 82 medieval Portuguese culture 49 Meillet 131, 134 metaphor 141, 143, 145, 148, 155 metrical stress theory 241 metrical structure 76, 83 middle voice 291, 292, 295 Middle English 78, 81, 88 Middle Low German 1 mismo 200, 207 modal verb 133 Modern Swedish 258, 270, 272 monoflexional 4 monotransitivization 181 morphologization 106, 111, 112

N negative raising 33 negation 133–135, 137 non-configurational 14 nonprestigious 65, 71 Nordic prefix loss 241 normative orthographic pressure 52

O official documents 57 Old English 75, 76, 79–83 Old Spanish 103, 104, 108 orthographic standardisation 55

Index 

P parameters 79, 82 passive voice 280, 291 past participle 279, 280 periphrasis 279, 281, 282, 292 pertinacity 75, 76 Pforzen 244 phonetic phenomena 47 polysemy 153 Portuguese 47, 50 prestigious varieties 71 preverb 241 progressive 301–314 Prokosch’s Law 246 prosodic repair strategies 242 Q quantitative 201, 215 R raising verbs 31, 32, 35, 36, 39–41, 43 reanalysis 31, 41, 43, 75, 81, 84, 132, 135, 181, 210, 215–219, 235, 237, 257, 258, 264, 266, 267, 269 reconstructions 119 reflexive 199–207, 210–220 reflexive pronouns 105 Reistad inscription 242 resultative 301, 302, 305–314 Romance stress 76, 77 S selective marking 3 semantic bleaching 188 semantic roles 141–143, 145 shared identity 71 sí 105, 186, 199–202, 204–211, 213–221, 291 sí mism- 199–202, 204–211, 213–221, 291

si 162, 164, 165, 173, 174 ¸si 165, 168–172, 174 sibilants 103–105, 111 sic 159, 161, 164, 165, 168, 169, 173 simplex 202 sociolinguistic perspective 57 Southern American English 64 SOV 13 Spanish 179–197, 199–201, 203–208, 211, 213–215, 218, 219, 279–282, 285, 287, 291, 292, 295 speaker involvement 210, 212–214 Stentoften inscriptions 243 stress 75–84 subordinator 225–235, 237 SVO 13 Swedish 225–227, 229, 231, 237

T target audience 87 theta-role 271, 272 transitive 115–118, 120–122, 124–129 trisyllabic shortening (TSS) 81

U unfooted syllables 246 unstressed vowels 49

V variation 88 varieties of Portuguese 47 verbal constructions 115, 117 verbo-nominal structure 182

W weak ditransitive structure word order change 88

179

CURRENT ISSUES IN LINGUISTIC THEORY

E. F. K. Koerner, Editor

Zentrum für Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Berlin [email protected] Current Issues in Linguistic Theory (CILT) is a theory-oriented series which welcomes contributions from scholars who have signiﬁcant proposals to make towards the advancement of our understanding of language, its structure, functioning and development. CILT has been established in order to provide a forum for the presentation and discussion of linguistic opinions of scholars who do not necessarily accept the prevailing mode of thought in linguistic science. It oﬀers an outlet for meaningful contributions to the current linguistic debate, and furnishes the diversity of opinion which a healthy discipline must have. A complete list of titles in this series can be found on the publishers website, www.benjamins.com 221 D’HULST, Yves, Johan ROORYCK and Jan SCHROTEN (eds.): Romance Languages and Linguistic Theory 1999. Selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999. 2001. viii, 406 pp. 222 HERSCHENSOHN, Julia, Enrique MALLÉN and Karen ZAGONA (eds.): Features and Interfaces in Romance. Essays in honor of Heles Contreras. 2001. xiv, 302 pp. 223 FANEGO, Teresa, Javier PÉREZ-GUERRA and María José LÓPEZ-COUSO (eds.): English Historical Syntax and Morphology. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 1. 2002. x, 306 pp. 224 FANEGO, Teresa, Belén MÉNDEZ-NAYA and Elena SEOANE (eds.): Sounds, Words, Texts and Change. Selected papers from 11 ICEHL, Santiago de Compostela, 7–11 September 2000. Volume 2. 2002. x, 310 pp. 225 SHAHIN, Kimary N.: Postvelar Harmony. 2003. viii, 344 pp. 226 LEVIN, Saul: Semitic and Indo-European. Volume II: Comparative morphology, syntax and phonetics. 2002. xviii, 592 pp. 227 FAVA, Elisabetta (ed.): Clinical Linguistics. Theory and applications in speech pathology and therapy. 2002. xxiv, 353 pp. 228 NEVIN, Bruce E. (ed.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 1: Philosophy of science, syntax and semantics. 2002. xxxvi, 323 pp. 229 NEVIN, Bruce E. and Stephen B. JOHNSON (eds.): The Legacy of Zellig Harris. Language and information into the 21st century. Volume 2: Mathematics and computability of language. 2002. xx, 312 pp. 230 PARKINSON, Dilworth B. and Elabbas BENMAMOUN (eds.): Perspectives on Arabic Linguistics. Papers from the Annual Symposium on Arabic Linguistics. Volume XIII-XIV: Stanford, 1999 and Berkeley, California 2000. 2002. xiv, 250 pp. 231 CRAVENS, Thomas D.: Comparative Historical Dialectology. Italo-Romance clues to Ibero-Romance sound change. 2002. xii, 163 pp. 232 BEYSSADE, Claire, Reineke BOK-BENNEMA, Frank DRIJKONINGEN and Paola MONACHESI (eds.): Romance Languages and Linguistic Theory 2000. Selected papers from ‘Going Romance’ 2000, Utrecht, 30 November–2 December. 2002. viii, 354 pp. 233 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume I: Segmental structure. 2003. x, 308 pp. 234 WEIJER, Jeroen van de, Vincent J. van HEUVEN and Harry van der HULST (eds.): The Phonological Spectrum. Volume II: Suprasegmental structure. 2003. x, 264 pp. 235 LINN, Andrew R. and Nicola McLELLAND (eds.): Standardization. Studies from the Germanic languages. 2002. xii, 258 pp. 236 SIMON-VANDENBERGEN, Anne-Marie, Miriam TAVERNIERS and Louise J. RAVELLI (eds.): Grammatical Metaphor. Views from systemic functional linguistics. 2003. vi, 453 pp. 237 BLAKE, Barry J. and Kate BURRIDGE (eds.): Historical Linguistics 2001. Selected papers from the 15th International Conference on Historical Linguistics, Melbourne, 13–17 August 2001. Editorial Assistant: Jo Taylor. 2003. x, 444 pp. 238 NÚÑEZ-CEDEÑO, Rafael, Luis LÓPEZ and Richard CAMERON (eds.): A Romance Perspective on Language Knowledge and Use. Selected papers from the 31st Linguistic Symposium on Romance Languages (LSRL), Chicago, 19–22 April 2001. 2003. xvi, 386 pp.

239 ANDERSEN, Henning (ed.): Language Contacts in Prehistory. Studies in Stratigraphy. Papers from the Workshop on Linguistic Stratigraphy and Prehistory at the Fifteenth International Conference on Historical Linguistics, Melbourne, 17 August 2001. 2003. viii, 292 pp. 240 JANSE, Mark and Sijmen TOL (eds.): Language Death and Language Maintenance. Theoretical, practical and descriptive approaches. With the assistance of Vincent Hendriks. 2003. xviii, 244 pp. 241 LECARME, Jacqueline (ed.): Research in Afroasiatic Grammar II. Selected papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000. 2003. viii, 550 pp. 242 SEUREN, Pieter A.M. and Gerard KEMPEN (eds.): Verb Constructions in German and Dutch. 2003. vi, 316 pp. 243 CUYCKENS, Hubert, Thomas BERG, René DIRVEN and Klaus-Uwe PANTHER (eds.): Motivation in Language. Studies in honor of Günter Radden. 2003. xxvi, 403 pp. 244 PÉREZ-LEROUX, Ana Teresa and Yves ROBERGE (eds.): Romance Linguistics. Theory and Acquisition. Selected papers from the 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002. 2003. viii, 388 pp. 245 QUER, Josep, Jan SCHROTEN, Mauro SCORRETTI, Petra SLEEMAN and Els VERHEUGD (eds.): Romance Languages and Linguistic Theory 2001. Selected papers from ‘Going Romance’, Amsterdam, 6–8 December 2001. 2003. viii, 355 pp. 246 HOLISKY, Dee Ann and Kevin TUITE (eds.): Current Trends in Caucasian, East European and Inner Asian Linguistics. Papers in honor of Howard I. Aronson. 2003. xxviii, 426 pp. 247 PARKINSON, Dilworth B. and Samira FARWANEH (eds.): Perspectives on Arabic Linguistics XV. Papers from the Fifteenth Annual Symposium on Arabic Linguistics, Salt Lake City 2001. 2003. x, 214 pp. 248 WEIGAND, Edda (ed.): Emotion in Dialogic Interaction. Advances in the complex. 2004. xii, 284 pp. 249 BOWERN, Claire and Harold KOCH (eds.): Australian Languages. Classiﬁcation and the comparative method. 2004. xii, 377 pp. (incl. CD-Rom). 250 JENSEN, John T.: Principles of Generative Phonology. An introduction. 2004. xii, 324 pp. 251 KAY, Christian J., Simon HOROBIN and Jeremy J. SMITH (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume I: Syntax and Morphology. 2004. x, 264 pp. 252 KAY, Christian J., Carole HOUGH and Irené WOTHERSPOON (eds.): New Perspectives on English Historical Linguistics. Selected papers from 12 ICEHL, Glasgow, 21–26 August 2002. Volume II: Lexis and Transmission. 2004. xii, 273 pp. 253 CAFFAREL, Alice, J.R. MARTIN and Christian M.I.M. MATTHIESSEN (eds.): Language Typology. A functional perspective. 2004. xiv, 702 pp. 254 BALDI, Philip and Pietro U. DINI (eds.): Studies in Baltic and Indo-European Linguistics. In honor of William R. Schmalstieg. 2004. xlvi, 302 pp. 255 MEULEN, Alice ter and Werner ABRAHAM (eds.): The Composition of Meaning. From lexeme to discourse. 2004. vi, 232 pp. 256 BOK-BENNEMA, Reineke, Bart HOLLEBRANDSE, Brigitte KAMPERS-MANHE and Petra SLEEMAN (eds.): Romance Languages and Linguistic Theory 2002. Selected papers from ‘Going Romance’, Groningen, 28–30 November 2002. 2004. viii, 273 pp. 257 FORTESCUE, Michael, Eva Skafte JENSEN, Jens Erik MOGENSEN and Lene SCHØSLER (eds.): Historical Linguistics 2003. Selected papers from the 16th International Conference on Historical Linguistics, Copenhagen, 11–15 August 2003. 2005. x, 312 pp. 258 AUGER, Julie, J. Clancy CLEMENTS and Barbara VANCE (eds.): Contemporary Approaches to Romance Linguistics. Selected Papers from the 33rd Linguistic Symposium on Romance Languages (LSRL), Bloomington, Indiana, April 2003. With the assistance of Rachel T. Anderson. 2004. viii, 404 pp. 259 CARR, Philip, Jacques DURAND and Colin J. EWEN (eds.): Headhood, Elements, Speciﬁcation and Contrastivity. Phonological papers in honour of John Anderson. 430 pp. Expected Winter 04-05 260 NICOLOV, Nicolas, Kalina BONTCHEVA, Galia ANGELOVA and Ruslan MITKOV (eds.): Recent Advances in Natural Language Processing III. Selected papers from RANLP 2003. 2004. xii, 402 pp. 261 KAY, Christian J. and Jeremy J. SMITH (eds.): Categorization in the History of English. 2004. viii, 268 pp. 262 VAJDA, Edward J. (ed.): Languages and Prehistory of Central Siberia. 2004. x, 275 pp. 263 BRANCO, António, Tony McENERY and Ruslan MITKOV (eds.): Anaphora Processing. Linguistic, cognitive and computational modelling. 2005. x, 449 pp.