Vowel length from Latin to Romance [First edition.] 9780191630538, 0191630535, 9780191779749, 0191779741

Table of contents :
1. Introduction
2. Vowel length in the Latin-Romance transition
3. The development of VL in Romance
4. The analysis of Northern Romance vowel length
5. Dialect variation and comparative reconstruction
6. In lieu of a conclusion
Appendix: Language and dialect area maps

Citation preview

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Vowel Length from Latin to Romance

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

OXFORD STUDIES IN DIACHRONIC AND HISTORICAL LINGUISTICS general e di to rs Adam Ledgeway and Ian Roberts, University of Cambridge a dv i so ry ed i to r s Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; David Willis, University of Cambridge rece ntly p ubl is he d in t he s e rie s 5 The History of Negation in the Languages of Europe and the Mediterranean Volume I: Case Studies Edited by David Willis, Christopher Lucas, and Anne Breitbarth 6 Constructionalization and Constructional Changes Elizabeth Traugott and Graeme Trousdale 7 Word Order in Old Italian Cecilia Poletto 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent 9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Vowel Length from Latin to Romance M I C H E L E LO P O R C A R O

1

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

3

Great Clarendon Street, Oxford, ox2 6dp, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries # Michele Loporcaro 2015 The moral rights of the author have been asserted First Edition published in 2015 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2013956954 ISBN 978–0–19–965655–4 Printed and bound by CPI Group (UK) Ltd, Croydon, cr0 4yy Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

For Giovanna and Laura: they know why

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Contents Preface List of figures List of abbreviations and notational conventions Latin authors and works cited in abbreviation 1 Introduction 1.1 The starting point: vowel length in Classical Latin 1.2 The long-term trend: Latin harbingers of the loss of CVL 1.3 Structure of the argument and aims of the book 2 Vowel length in the Latin–Romance transition 2.1 Change in Latin VL: metalinguistic testimonies and the rise of OSL 2.2 Why OSL must be Proto-Romance: excluding conceivable alternatives 2.2.1 OSL at a later date, and only in some Romance languages 2.2.2 OSL in Republican Latin? 2.3 The rise of OSL and the regional diversification of Latin 2.3.1 Evidence from metrical inscriptions, part 1: Herman (1982) 2.3.2 Evidence from metrical inscriptions, part 2: Adams (1999) 2.3.3 Latin and Romance in Africa 2.4 Quality is not quantity, after all 2.5 Intermediate summary and provisional conclusion 3 The development of VL in Romance 3.1 3.2 3.3 3.4

Three types of distribution of VL in the Romance languages In defence of Open Syllable Lengthening in modern Standard Italian The eastern and western peripheries of Romance Northern Romance 3.4.1 Northern Italo-Romance 3.4.2 Gallo-Romance 3.4.3 (The rest of ) Rhaeto-Romance 3.5 Summing up: VL and OSL from Proto-Romance to the modern languages 4 The analysis of Northern Romance vowel length 4.1 Competing analyses of the rise of CVL in Northern Italo-Romance 4.1.1 Formal accounts of the rise of VL in Milanese 4.1.2 Competing explanations of the rise and status of VL in Friulian 4.1.3 Alternative formal accounts for the rise of VL in Cremonese

ix xiii xiv xvii 1 1 9 12 18 20 25 25 30 40 41 46 47 51 57 61 61 65 75 80 82 101 108 115 121 121 124 129 132

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

viii

Contents

4.2 Diachronic phonology, generative grammar, and method in historical linguistics 4.3 Too much synchrony into diachrony, too much diachrony into synchrony 5 Dialect variation and comparative reconstruction 5.1 At the vanguard of change: the fading of contrastive VL in Northern Italo-Romance 5.1.1 Loss of CVL in peripheral Friulian dialects 5.1.2 The areal pattern of CVL in Liguria and Piedmont 5.1.3 The fading of CVL in Alpine and Eastern Lombard 5.1.4 The position of Venetan 5.2 Apocope and the rise of contrastive VL in Northern Romance 5.2.1 On the non-co-occurrence of CVL and apocope 5.2.2 The gradual spread of apocope in Northern Italo-Romance 5.3 The rearguard of change: at the source of VL in Northern Romance 5.3.1 Geminate consonants and VL in Alpine Lombard 5.3.2 On the southern periphery of Northern Romance 5.3.3 Phonetic gradience in vowel and consonant length and the change from PRom gemination to NRom CVL 5.3.4 A closer look at Western Romance degemination 5.3.5 The odd one out: CVL without degemination south of the Apennines? 5.4 Making sense of the comparative picture 5.4.1 A phonetic constraint on vowel length: rhythmical compensation 5.4.2 Romance oxytones and VL 5.5 Taking stock of the reconstructive evidence

133 138 145 147 148 149 152 163 164 165 168 172 174 184 190 194 200 203 208 214 223

6 In lieu of a conclusion

227

Appendix: Language and dialect maps References Index of languages Index of names Index of subjects

241 250 287 293 300

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Preface This book provides a global reappraisal of a central topic of Romance historical linguistics: the development of vowel length from Latin to the Romance languages and dialects. Vowel length was contrastive in Latin, and no daughter language has inherited that contrast, though several have established novel vowel length distinctions at different stages of their history and in different ways, which will be addressed in what follows. While all those developments deserve attention per se, the main question the present monograph focuses on is whether a diachronic link, via a series of subsequent changes, can be established between at least some of the length contrasts documented throughout Romance, today or in the past, and Latin distinctive vowel quantity. The answer will be positive: it will be argued that the most satisfactory reconstruction of the rise of contrastive vowel length in a substantial number of the Romance systems displaying that property (nowadays or in the past), scattered over the central-northern part of the Romance-speaking territory from the Apennines to the North Sea, is one in which this development is analysed as a diachronic successor of a Proto-Romance open syllable lengthening process, whose rise can be in turn held responsible for the loss of the contrastive vowel length that used to characterize Classical Latin. The topic is a much-debated and highly controversial one, in all its facets: to cite just one aspect, the demise of Latin contrastive vowel quantity has been explained variously, and dated in a widely diverging way, ranging from the 3rd century bc to the 4th ad. The topic also has wide empirical implications, since discussion of vowel quantity entails addressing the relationship with (the historical development of ) stress assignment and syllable structure, consonant gemination (word-internal as well as at word boundary, through raddoppiamento fonosintattico) and degemination, as well as with apocope, diphthongization, etc. This list includes some of the major isoglosses distinguishing subgroups within the Romance language family: thus, dealing with this topic boils down to coming close to providing a comprehensive account of the diachronic development of Latin–Romance phonology, at the prosody-to-segments interface. The implications for theoretical linguistics (in particular, for the theory of sound change) are also significant: discussion of the most effective explanation of the empirical phenomena at hand will require addressing the division of labour between synchronic and diachronic linguistics, in terms of methods and explanatory scope, as well as, even more broadly, the issue of the interplay between form and substance in linguistic sound systems, or the phonetics/ phonology interplay. A number of more specific issues could be mentioned here: for

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

x

Preface

instance, the development of vowel length contrasts in Romance has often been considered in the context of cross-linguistic discussions of the nature of compensatory lengthening and its impact on a language’s phonological system or, more broadly, when debating whether there are universal constraints on the paths by which languages may acquire vowel length contrasts. On all the issues covered in the following pages, an extensive literature is already available. Yet the present study can boast novelty in that it is the first to reconsider the development of vowel length (VL) from Latin to Romance drawing systematically on evidence from fields as diverse as, say, the study of Latin inscriptions from Rome and the rest of the Western Roman Empire, on the one hand, and on the other, firsthand experimental phonetic analysis of data from modern Romance dialects, collected in some remote villages on the Alps or the Apennines in the early 21st century. The possibility of relying on such multifarious evidence is the hallmark of Romance historical linguistics, since we do not have access to such a varied database for many other language families: the Latin–Romance continuum, with its two and a half millennia of documented history and with the great amount of scholarly work that has been carried out on it from many different perspectives, still presents the historical linguist not only with unsolved problems, which is obvious, but also, less trivially, with the challenging opportunity to apply to these problems the method of historical linguistics in its fullest form. This wealth is also a considerable burden, though, and puts severe constraints on the analyst, since any respectable reconstruction in this field must prove capable of reconciling the results gained from those different sources of evidence. This is the fundamental challenge this monograph tries to face in reconstructing the evolution of VL from Latin to Romance. The structure of the book will be outlined at the end of the introductory chapter. Let me add here some preliminary logistical notes. Cross-references throughout the book point by default to numbered examples or footnotes within the same chapter, unless another chapter/section number is specified. Linguistic data are provided in IPA transcription, including data presented by the original authors using different conventions. Whenever unreferenced, dialect data stem from my own fieldnotes. For the citation of Latin authors and texts, I follow the Oxford Latin Dictionary (OLD) conventions; translations stem from the Loeb classical collection, unless specified otherwise. In quoting Latin linguistic materials, different conventions are used, depending on whether these are cited per se or rather as a diachronic source of their Romance descendants. In the latter case, following traditional practice in Romance linguistics, Latin etyma are given in small caps and (for nouns and adjectives) in the accusative form (e.g. villam), which became generalized in the Romance languages except Romanian. By contrast, Latin forms, phrases, and/or sentences are given in italics whenever they are considered per se rather than from a Romance angle (e.g. uillam), and are given in italicized small caps when stemming from inscriptions (e.g. oppidum). (As seen in the above examples, vowel quantity

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Preface

xi

is not normally marked for Latin etyma, except when directly relevant to the argument—for example for metrical structure or the vowel qualities of Romance outcomes.) In glossing, Leipzig-style abbreviations are used, with some adjustments to avoid redundancy. Thus, for nouns and adjectives, grammatical feature values are omitted whenever the quoted form corresponds to the default: singular for nouns, masculine singular for adjectives (for Latin, the case value ‘nominative’ is also left out); for verb forms in the present indicative, only person is specified (e.g. Milanese nas ‘be born.3sg’); and whenever English translation is shorter than the grammatical gloss without being ambiguous, this is preferred. As I take leave of the book, I must thank a number of persons and institutions for their help and support. Over the past few years, the following friends and colleagues took time to discuss with me several aspects of the present research, either in person or in print, and/or to point me to important references (or even to send them to me): my thanks go to Pier Marco Bertinetto, Giovanni Bonfadini, Kathleen Coleman, Tom Cravens, Stefania Ferrari, Jürg Fleischer, Marco Mancini, Stefania Maina, Peggy Renwick, Emanuele Saiu, Mary Stevens, Paul Videsott, and M. Teresa Vigolo: with respect to them, of course, usual disclaimers apply and the same goes for three anonymous OUP reviewers who provided constructive criticism on the project. With other friends and colleagues, discussion took the form of joint work on the topic, partly within the framework of a research project carried out at the University of Zürich (‘At the Source of Northern Italo-Romance: Quantity and Related Phenomena’, http://www.research-projects.uzh.ch/p5703.htm): Rachele Delucchi, Lorenzo Filipponio, Tania Paciaroni, Stephan Schmid. Last but not least, the book would not have been the same if Yves Charles Morin had not generously taken time to point to passages where the text could be improved and the discussion made more thorough. A further kind of scientific conversation I have to acknowledge here is that with students who wrote under my supervision in Zürich theses dealing with vowel length and related phenomena in Romance varieties: Carla Biasini, Camilla Bernardasci, Debora Nardini, Cristina Pifferi, and Steffi Gredig. Also, the audiences of the lectures I gave on the topic in Zürich, especially in the spring of 2010 and autumn of 2011, greatly helped me to fine-tune the presentation of the arguments. As is common practice, the reference list that closes the book indicates the interlocutors in dialogue with whom this book has materialized: however, one name I would like to single out in this context, since my own ideas on the fate of vowel quantity in Late Latin could not possibly have taken their present form without the constant inspiration provided by the work of the late József Herman. I am also indebted for feedback and comments to the audiences to whom I have had the opportunity to presenting parts of this research over the last few years, at the Scuola normale superiore in Pisa, the Universities of Graz, Roma Tre, Cluj-Napoca, Geneva, Naples Federico II, the Collegio Ghislieri in Pavia, and the Circolo filologico linguistico padovano. Invited talks at the 17th International Conference on Historical

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

xii

Preface

Linguistics (2005) and the Philological Society in London (2012) were important landmarks in the ripening process of this book, for which I warmly thank the organizers. To the editors of the Cambridge History of the Romance Languages, Martin Maiden, Adam Ledgeway, and J. C. Smith, I owe an important impulse to start thinking seriously and systematically about several of the topics discussed in this monograph; but without the decisive intervention of John Davey, the most congenial linguistics editor one could dream of, this thinking would hardly have materialized into a book. To the rest of the OUP team who dealt with the manuscript at different stages, as well as to Mitch Cohen, I am indebted for the constructive joint work on improving it. The Faculty of Philosophy of the University of Zürich has to be mentioned in the acknowledgements, not only for providing me with the best imaginable workplace on the planet for a historical Romance linguist (with the Jakob Jud Library and a series of other facilities) but also, more specifically, for partly funding the fieldwork sessions in Val Bregaglia (2002), on the Emilian Apennine (2004, 2008), and in Val Marebbe and Val Müstair (2011), from which I have drawn some of the data discussed in Chapters 3 and 5. Chapter 5 especially, where the areal distribution of the different VL systems investigated here is focused on as providing crucial clues for diachronic reconstruction, has benefited from discussion with colleagues within the framework of—and is intended as a contribution to—the University Research Priority Program of the University of Zürich on ‘Language and Space’. Finally, two more Zürich colleagues should be gratefully mentioned: Curdin Derungs, who prepared the maps, and Vincenzo Faraoni, who assisted me with the preparation of the indexes (with the cooperation of Massimo Bellina, Stefano Negrinelli, M. Chiara Paravicini Bagliani, and Luca Willi). Thanks are due, moreover, to Magdalen College Oxford, which granted me a visiting fellowship during Michaelmas term 2012, and hosted me in a most inspiring setting for the final writing-up, at Holywell Ford Mill, in the middle of the unforgettable deer park and yet just ten minutes’ walk away from the superb Oxonian libraries: this book would not have taken its present shape without the days and evenings spent in the Bodleian, the Taylorian, and the Sackler. It would take a very long time for me to thank by name all the persons—rather dully called, in our jargon, ‘informants’—who have helped me collect the first-hand linguistic data discussed throughout the book, sharing with me their insights on their native languages/dialects. I would like however to mention at least some of them explicitly, as a sign of gratitude for the hours spent working together: Luciano Persico (Crotta d’Adda, Cremona), Benito Biagi, Gualtiero Bonucchi, Sergio Polmonari, Maurizio Poli (Lizzano in Belvedere), Ferdinando Lunardi, Lina Manattini, Emiliano Mazzoni (Piandelagotti), Giorgio M. Daledo (Travacò Siccomario), Marco Morandotti (Lungavilla), Claudio Cervati (Sairano), and Danko, Liliana, and Reto Giovanoli, Fiorella Willy, Anna Nunzi (Soglio).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

List of figures 3.1 Stressed vowel duration before consonants in open and closed syllables in Standard Italian.

67

3.2 Vowel durations before singleton consonants, tauto-, and hetero-syllabic clusters in Standard Italian.

67

3.3 The five levels of representation of Boersma’s BiPhon model.

69

5.1 [ˈʃtel:ɐ] ‘drop’ (dialect of Soglio, male speaker, aged 73), sonagram and sound wave.

178

5.2 [ˈʃte:lɐ] ‘star’ (dialect of Soglio, male speaker, aged 73), sonagram and sound wave.

178

5.3 Mean duration values for the stressed vowel and the following consonant in the minimal pair [ˈʃtel:ɐ] ‘drop’ vs [ˈʃte:lɐ] ‘star’ (dialect of Soglio).

179

5.4 Mean duration values for the geminate vs singleton consonants and the stressed vowels preceding geminate vs singleton consonants in the dialect of Lizzano in Belvedere. 5.5 [aˈvi:vu kanˈta na ˈbεla kanˈtsÕ :] ‘did you sing a beautiful song?’ (dialect of Lizzano in Belvedere, male speaker, aged 57), sonagram and sound wave.

187

188

5.6 [ˈtʃεrta ke ˈvu: kanˈta: ˈprOprjo ˈma:le] ‘indeed, you sing very poorly!’ (dialect of Lizzano in Belvedere, male speaker, aged 57), sonagram and sound wave.

188

5.7 Rhythmical compensation in Standard Italian.

209

5.8 Durations of word-final long vs short stressed vowels in utterance-final vs utterance-internal position in the dialect of Cavanna. 5.9 Durations of word-final stressed vowels in secondary and primary oxytones, in utterance-final vs utterance-internal position in the dialect of Stabiazzoni.

222

223

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

List of abbreviations and notational conventions* abl

ablative

acc

accusative

Arab.

Arabic

BS

Bresciano

Cat.

Catalan

CL

Classical Latin

CoL

compensatory lengthening

CVL

contrastive vowel length

dat

dative

DO

direct object

Fr.

French

gen

genitive

Germ.

Germanic

Gk.

Greek

Goth.

Gothic

GRom

Gallo-Romance

IE

Indo-European

impf

imperfect (tense)

int

interrogative

IO

indirect object

It.

Italian

Lat.

Latin

lit.

literally

Log.

Logudorese (Sardinian)

μ

mora

me

Middle English

MHG

Modern High German

ModE

Modern English

* Grammatical abbreviations are listed here only when they are not included in the Leipzig Glossing Rules (http://www.eva.mpg.de/lingua/resources/glossing-rules.php). Bibliographical abbreviations, omitted here, are included in the final reference list.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

List of abbreviations and notational conventions NHG

New High German

NItRom

Northern Italo-Romance

NRom

Northern Romance

OCS

Old Church Slavonic

OE

Old English

OFr.

Old French

OGen.

Old Genoese

OHG

Old High German

OIt.

Old Italian

ONorse

Old Norse

OSL

open syllable lengthening

OSp.

Old Spanish

OT

Optimality Theory

Pg.

Portuguese

PIE

Proto-Indo-European

PRom

Proto-Romance

pt

point

PU

phonological utterance

PW

phonological word

RF

raddoppiamento fonosintattico

Ro.

Romanian

RRom

Rhaeto-Romance

σ

syllable

Sard.

Sardinian

sc.

scilicet [to rectify mistakes within quotations]

SI

Standard Italian

Sic.

Sicilian

Skr.

Sanskrit

Sp.

Spanish

SRC

Strong Rhyme Constraint

VL

vowel length

WFC

Well-Formedness Condition

WGerm.

Western German

WRom.

Western Romance

!

becomes (by synchronic derivation)

xv

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

xvi

List of abbreviations and notational conventions derives from (synchronically)

>

becomes (diachronically)


[i:] (early 3rd century bc), and [ou̯] > [u:] (late 3rd c. bc; cf. Meiser 2010: 57–60). 8 Terentius Scaurus GL 7.33.7 prescribes complementarity of the two devices (‘super i tamen litteram apex non ponitur’ [‘you should not put an apex on the letter i’]. However, Christiansen (1889: 12–17) shows that, though less frequently than on other vowels, apices were used on ’s as well (e.g. CIL XII 1357, 1st c. ad, fı´lio´, mı´litvm, qvı´ etc.), and occasionally even on I longae: sabı´no CIL VI 3940, svÍs VI 7527, etc. 9 The use of apex, for instance, is recommended (by Quint. 1.7.2-3, Terent. Scaur. GL 7.33.5-7) only to discriminate between words forming minimal pairs such as those in (1), explicitly discouraging as ‘very silly’ (ineptissimum, Quintilian says) across-the-board use.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Vowel length in Classical Latin

5

‘to drive’), or Slavic, where conversely only short /a/ was velarized (e.g. OCS mati ‘mother’ vs ostrŭ ‘sharp’, cf. Lat. acer < PIE *h2ak´ro-). Within Italic, evidence from Oscan is of special relevance since Oscan extant writings, in particular those in the Central Oscan epichoric alphabet) display notation of vowel length in word-initial root syllables—which also happen to be stressed—much more systematically and at a much earlier date than the corresponding, only desultory, reduplication of vowel graphemes in Latin: attestations of geminatio uocalium in Oscan start around the middle of the 3rd century bc, i.e. more than one century before their earliest Latin counterparts (cf. Buck 1928: 25–6; Lejeune 1975: 240–45).10 Thus, Oscan maatreís ‘mother.gen.sg’, faamat ‘have at one’s orders.3sg’ vs actud ‘do.fut.imp.3sg’, akrid ‘sharply’ (abl.sg) directly confirm the quantities of the corresponding vowels in the Latin counterparts mātris, fāma ‘rumour’, ăgĭto, and ăcrī ‘sharp.abl.sg’ (cf. de Vaan 2008: 22, 30, 367; Meiser 2010: 55–6). In addition to the comparative evidence, one can also adduce proofs attesting to the Latin VL contrast from borrowing into other IE languages which are known to possess or to have possessed such a contrast: thus, the OHG forms for ‘cat’ and ‘street’ (respectively kazza vs strāza > MHG Katze vs Straße) witness to the Latin contrast between the stressed vowels of cattus and (uia) strāta (cf. Sihler 1995: 73). Similarly, borrowings into Celtic—in which Latin short vs long vowels are systematically kept distinct (cf. Jackson 1994: 84, 88, and }2.5 below)—attest to the shortness of the ŭ in the final syllable of e.g coccum ‘scarlet (colour)’, scamnum ‘stool’, as shown by Campanile (1973: 403) because a long ū—which should have arisen in Lüdtke’s (1965: 487) opinion as a consequence of final m-deletion—would have triggered metaphony in e.g. Welsh coch ‘red (colour)’, ysgafn ‘bed’, whose stressed vowel did not, however, undergo metaphony (otherwise they would be **cych, **ysgyfn). Evidence from sound change also points to the existence of CVL, whenever not recognizing the contrast would lead to the impossibility of describing economically either the change itself or its conditioning environment. Similar cases are legion throughout the (pre)history of Latin. Thus, a word-final -/d/ was deleted by the mid3rd century bc (Meiser 2010: 100) after long vowels only (e.g. mēd > mē ‘me’ vs sĕd ‘but’, pōplicōd > pūblicō ‘public.abl.sg’ vs apŭd ‘at’); other changes which applied in Archaic Latin, such as -ss-degemination, iambic shortening (}1.2) or the so-called ‘Latin ablaut’ (e.g. făcio ‘do.1sg’ vs perfĕctum ‘accomplished’ vs perfĭcio ‘accomplish.1sg’) must be stated with reference to VL;11 and most of the Romance 10 Indeed, the geminatio uocalium has been argued by Lazzeroni (1956: 292) not to have ever been a feature of Latin orthography stricto sensu, but rather to be ‘tipica del latino scritto dagli italici’ [‘typical of the Latin written by Italics’; emphasis original], since it is virtually confined to Latin inscriptions which either stem from Italic territory or otherwise involve people of Italic origin. 11 The term ‘ablaut’, though traditional (see below), is actually a misnomer, since the change involved was sensitive to the phonological context. In fact, by far the simplest generalization accounting for the application of change in the root vowel in, say, occĭdo ‘fall.1sg’ (compare the simplex cădo ‘fall.1sg’), vs its

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

6

Introduction

developments too—on which see below, Chapter 2, (15), (23)–(25)—would be obscured entirely if one could not refer to CVL in their statement (in the wake of e.g. Foerster 1878), to discriminate between the outcome of, say, long vs short /i/ or /u/ in Italian, such as in mille ‘thousand’ < mı¯lle vs secco ‘dry’ < sı˘ccum, nulla ‘nothing’ < nu¯llam vs rotto ‘broken’ < ru˘ptum. The same holds true of the synchronic characterization of Classical Latin stress, which implies reference to VL as one determinant of syllable weight: (4)

Classical Latin stress: In words of more than two syllables, stress falls on a heavy (i.e. bimoraic) penult, otherwise on the antepenultimate.

This stress rule is believed to have come into being early in the 4th century bc, replacing an older initial stress which Latin inherited from Proto-Italic and thus used to share with the neighbouring cognate languages such as Oscan-Umbrian, but also with unrelated Etruscan.12 During that previous stage (first reconstructed by Dietrich 1852: 546), vowel reduction resulted in alternations—improperly labelled sometimes, especially in the French scholarly tradition, as ‘(Latin) ablaut’ (apophonie, cf. e.g. Kerlouégan 1978: 35): amicus ‘friend’/inimicus ‘enemy’, arma ‘weapon’/inermis ‘unarmed’, dedi ‘give.prt.1sg’/reddidi ‘give back.prt.1sg’—whose outcomes survived into the modern Romance languages: e.g. Fr. ami ‘friend’ < a˘mı¯cum vs ennemi ‘enemy’ < inı˘mı¯cum < *ı´na˘mı¯cum, with /a/ > /i/ in an unstressed open syllable. The change testifies to an original VL contrast, since only short, and not long, unstressed word-internal vowels underwent reduction. The reality of the rule in (4) for CL is attested by the persistence of stress position into the daughter languages, in spite of the fact that the stress rule has been radically modified (or even eliminated altogether, as in French): in, say, canta¯tum ‘sung.acc.sg’, stress fell on the long /a:/ in the penult, and it is the continuant of that vowel that still carries stress in Sp./Pg. cantado, Fr. chanté, It. cantato, Ro. cântat.13 Like

non-application in, say, occāsus ‘sunset’, is that the change targeted short, not long, vowels which had come to follow stress as this became word-initial in Proto-Italic (see n. 12). See e.g. Lepschy (1962: 216–22), Leumann (1977: 246–7), Morani (2000: 174–84). On the protosyllabic stress of Oscan, see e.g. Bottiglioni (1954: 24), Kuryłowicz (1958: 381); on the prehistory of Latin stress in the context of its neighbour languages, see e.g. Prosdocimi (1986), Weiss (2009: 109–10). This Proto-Italic firstsyllable stress was an innovation with respect to the PIE stress system, in which stress was free to occur on any syllable of the word, as first demonstrated by Verner (1877: 97), though with a default tendency towards initial stress which emerged to determine main word stress in Proto-Italic (just as in Proto-Celtic and Proto-Germanic; see e.g. Kiparsky 1973: 819; Halle 1997: 299; Viti 2013). 13 All this might seem obvious, but it actually is not, as the operation of rule (4) for Classical Latin has been doubted by some—most notably, Pulgram (1975: 91), who claims that a rule referring to VL such as (4) existed for ‘Written (Classical) Latin’ only, but not for the spoken language which had in his view no CVL as early as the 3rd century bc. 12

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Vowel length in Classical Latin

7

accent, Latin metrics (see e.g. Allen 1973; Boldrini 1992; Fortson 2011) is not describable without reference to VL. Let us now move on to consider metalinguistic testimonies by Latin authors, who comment on vowel length, and occasionally on minimal pairs distinguished exclusively by that feature. Some such passages have already been mentioned (nn. 2 and 3); consider, in addition, the following: (5)

a. palus aliud priore syllaba longa, aliud sequenti significat [‘palus means one thing if the first syllable is long and another if the second is long’, i.e. pālŭs ‘post’ vs pălūs ‘swamp’]. (Quint. Inst. orat. 1.7.3) b. alia cum intellexerimus nomen esse, primam syllabam longam esse existimamus [ . . . ], cum uero pronomen est, breuis erit [‘when we understand alia as a noun (= ‘garlic.pl’), we take the first syllable to be long, but when it is a pronoun (= ‘other.npl’), it (i.e. the first syllable) will be short’] (De ultimis syllabis liber ad Caelestinum, GL 4.260.10–13)

At a certain point in time, such comments were reiterated out of sheer tradition, after CVL had ceased to be part of the natively acquired linguistic competence. This might well have been the case for the (anonymous) treatise De ultimis syllabis (5b), dating probably to the 4th century ad (cf. Kaster 1988: 348–50; Stok 1997: 15–19) and transmitted under the name of one Probus, to be distinguished from M. Valerius Probus from Berytos (1st century ad, cf. Suet. Gramm. 24). Much scholarly effort was devoted to speculating on how and precisely when this change may have taken place, with proposed dates ranging from the 3rd century bc to the 4th–5th century ad, a point which is of course crucial for the (pre)history of the sound system of the Romance languages. As of now, suffice it to observe that during the Republican age and the early Imperial period (Quintilian, quoted in (5a), wrote in the second half of the 1st century ad), metalinguistic testimonies seem to indicate that the VL contrast was perceived by native speakers, independent of education level. This is emphasized by Cicero: (6)

In uersu quidem theatra tota exclamant si fuit una syllaba aut breuior aut longior; nec uero multitudo pedes nouit nec ullos numeros tenet nec illud quod offendit aut cur aut in quo offendat intellegit; et tamen omnium longitudinum et breuitatum in sonis sicut acutarum grauiumque uocum iudicium ipsa natura in auribus nostris collocauit [‘the whole audience will hoot at one false quantity. Not that the multitude knows anything of feet, or has any understanding of rhythm; and when displeased, they do not realize why or with what they are displeased. And yet nature herself has implanted in our ears the power of judging long and short sounds as well as high and low pitch in words’] (Cic. Orator 173)

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

8

Introduction

This passage has been much discussed, not only for VL, but also in the context of determining the exact phonetic nature of Latin accent, a thorny issue we will not dwell on here.14 The controversy on that point, however, was stirred by the circumstance that evidence from other sources does not seem to support the assumption of a pitch accent, whereas for VL this and similar metalinguistic comments unambiguously converge with the linguistic evidence.15 Furthermore, in (6) the remark on illiterate audiences hooting concerns specifically VL, rather than pitch. A very similar wording is encountered in another passage by Cicero: (7)

Quotus enim quisque est qui teneat artem numerorum ac modorum? At in his si paulum modo offensum est, ut aut contractione breuius fieret aut productione longius, theatra tota reclamant [‘For what proportion of people understands the science of rhythm and metre? Yet all the same if only a slight slip is made in these, making the line too short by a contraction or too long by dwelling on a vowel, the audience protests to a man’] (Cic. de Orat. 3.196).

In this case too, the passage is introduced with reference to the power of ‘nature’ (‘magna quaedam est uis incredibilisque naturae’ [‘nature has a vast and indeed incredible power’], Cic. de Orat. 3.195), and seems to indicate natively acquired knowledge of language, independent of literacy. All the different sources of data briefly recapitulated up to now converge in pointing to CVL being a genuine part of the phonological system of CL. As we shall see, however, not even that much is uncontroversial, since several scholars have entertained a very early date for the demise of CVL in spoken Latin (see }2.2.2). Occasionally, furthermore, doubt has been cast on Latin CVL on speculative grounds, as is the case for Kaye (1989: 151), claiming that in Latin ‘length distinctions can be removed from considerations of phonemic status and assigned to syllable structure, where they belong’. 14 Whether or not CL had a pitch accent, as opposed to stress accent in the archaic period, has been debated at length (cf. e.g. Lepschy 1962; Leumann 1977: 248–54 for a recapitulation of different opinions on the matter). Nowadays, the view seems to prevail that stress accent was there throughout and that descriptions in terms of acutus/grauis were due to ‘close imitation of Greek models’ (McCullough 2011: 90), whereby ‘the grammarians have slavishly misapplied the Greek system to the description of Latin’ (Allen 1973: 151). (Though one must mention that there are scholars who, still relatively recently, take these descriptions at face value, and claim that CL had a pitch accent (e.g. Bernardi Perini 1986: 13; Boldrini 1992: 17); this claim is often reiterated without discussion in handbooks of Romance linguistics, especially by French scholars: e.g. Joly 1995: 22.) Greek imitation possibly had an impact on literary performance, a cultural practice in which Greek-style pitch differences may have been superimposed on ordinary stress by cultivated people in reading verse aloud (Allen 1973: 151–69). That the description of Latin stress with Greek terms never corresponded to phonetic reality is shown, inter alia, by the shift in meaning seen in Pompeius (5th–6th century, GL 5.126.5–26), who uses acutus for ‘stressed (vowel) in a closed syllable’ (e.g. árma ‘weapon’), circumflexus for ‘stressed (vowel) in an open syllable’ (e.g. mêta ‘goal’), and gravis for ‘unstressed’. 15 And indeed, there is a long tradition of taking this information at face value (e.g. Herman 1998: 9), even if scepticism has been expressed by some (e.g. Pulgram 1975: 226, n. 23).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The long-term trend

9

This is the conclusion of Kaye’s analysis of Latin VL, which requires the reader to share several assumptions about phonological representation within one specific theoretical framework, that of Government Phonology. In other words, it is a theory-internal conclusion, and I shall not take issue with this here. For anybody not sharing those assumptions there is no compelling reason to distrust what the Latin evidence itself shows, with minimal pairs like those in (1)–(2): long vs short vowels, namely, could occur in exactly the same environments (syllabic and otherwise, although with some restrictions, to be addressed in }1.2) and are therefore contrastive, as acknowledged by studies on Latin phonemics in the classical structuralist tradition (e.g. Kerlouégan 1978: 33). This observation provides the starting point for my discussion of the Latin–Romance transition and the further development of VL in the Romance languages, to be developed in what follows.

1.2 The long-term trend: Latin harbingers of the loss of CVL Though, as we saw in }1.1, CVL was contrastive in Latin and hence in one sense independent of contextual factors such as stress and syllable structure, this observation must be hedged with qualifications since considerations of a quantitative nature allow one to realize that Latin phonology already contained the seeds of the Romance development. Firstly, it must be mentioned that during the (pre)history of Latin several shortening processes dramatically reduced the functional load of the VL contrast in unstressed, especially final, syllables. The list includes changes with a quite restricted structural scope (if with vast impact in terms of number of lexemes), such as the shortening of 1st class nominative and neuter nominative/accusative plural endings -ă, which stem from PIE *-ā (< -eh2; cf. Leumann 1977: 110): e.g. infera ‘inferior.nom.fsg or nom/acc.npl’ instead of *inferā < *n dh-erā, cf. Skr. ádharā.16 The scope of iambic shortening (correptio iambica) was much wider. This process affected word-final originally long vowels after a light syllable in words like bĕnĕ, mo˘do˘, căvĕ, pŭtă. It ‘can hardly have been much earlier than 200 b.c.’ according to Allen (1973: 182), as the terminus post quem for its dating is provided by deletion of -d after a long, but not after a short, vowel, which is still found in the conservative orthography of some inscriptions from the early 2nd century bc (poplicod ‘public. abl.sg’ CIL I2 581, 186 bc), while its deletion is already attested in 189 bc (in tvrri lascvtana ‘in the L. tower’, ea tempestate ‘in that moment’ CIL I2 614, with -d > 16 Though the fsg ending is customarily written as (reconstructed) *-ā, the original quantity is still attested in Archaic Latin, e.g. in Pl. As. 762: Ne epistula quidem ulla sit in aedibus ‘that there be no one single letter [scanned as epistulā] in the house’ (cf. Wedding 1902: 2). The change observed in these inflections has been interpreted alternatively as a morphological substitution of the original vocative ending (PIE *-ă) for the original nominative one (Meiser 2010: 77). This explanation has the drawback of leaving the shortening in the homophonous neuter endings unexplained, since the vocative can hardly be invoked there.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

10

Introduction

Ø in *turrīd, *eād) and must consequently have applied somewhat earlier.17 Had d-deletion applied later than iambic shortening, it would have been bled by the latter, resulting in **bĕnĭd, **mo˘dŭd, instead of attested bĕnĕ, mo˘do˘, from earlier (reconstructed) *dwĕnēd, *mo˘dōd (Kuryłowicz 1958: 338–9; Allen 1973: 181–2). Iambic shortening concerned wordforms from several grammatical categories, like verbs in the 1sg (e.g. puto ‘estimate’, rogo ‘ask’), or ablatives (see above), etc. In authors of the early Empire one finds short final -o extended analogically to noniambic forms such as Pōlli(o˘) Verg. Ecl. 3.84 (Leumann 1977: 110). For the 1sg verbal ending, short -o˘ was prescribed as standard for non-monosyllabic verbs by the 4th century ad (cf. Charisius, GL 1.16.20, Marius Victorinus, GL 6.28.21; Pompeius, GL 5.232.21–37). In Archaic Latin (about 200 bc) unstressed non-word-final vowels preceding a consonant other than -s also underwent shortening (except in monosyllables), which gave rise to a series of alternations within inflectional paradigms: e.g. 3sg amăt, uidĕt (vs 2sg amās, uidēs), or nominative animal, uicto˘r vs the oblique cases (animālis, uictōris etc.). Long vowels before -t -r -l are still attested in early Latin poetry (e.g. arāt Pl. As. 874, morōr Pl. Rud. 1248, pōnēbāt Enn. Ann. 371), whereas before -m shortening applied in pre-literary times, although not reaching as far back as common Italic: cf. quăm vs Oscan paam (Meiser 2010: 77). Shortening of prevocalic long vowels also contributed a reduction in the number of long vowels occurring in unstressed syllables (in e.g. fuēre ‘be.prt.3pl’ < fū-): it was completed in the pre-literary period for ē (dĕus < *dēos), whereas there is still some textual evidence from Archaic Latin for prevocalic ī and ū (fūimus ‘be.prt.1pl’ in Enn. Ann. 377, Dīāna Enn. Ann. 62; cf. Meiser 2010: 76). The handful of exceptions which persisted into the Classical period was fed by borrowing: among Greek loanwords, even sparse minimal pairs occurred, such as Cŏus ‘Cos (an island in the Dodecanese)’ vs Cōus ‘pertaining to Cos’ (Biville 1995: 21). By and large, however, it is true that in CL, as a product of the sound change just mentioned, the prevocalic position was, for both stressed and unstressed vowel, a neutralizing environment (cf. e.g. Kerlouégan 1978: 37). All the cases of restrictions on the contrastiveness of VL considered so far involved shortening of unstressed vowels: a symmetrical eventuality has been entertained in some marginal cases, involving Greek again. Since words like mōrum ‘blackberry’ (Gk. μόρον), rāpum ‘turnip’ (Gk. ῥάπυϛ), līnum ‘flax plant’ (Gk. λίνον) display a stressed long vowel in Latin as opposed to a short one in Greek, some have hypothesized borrowing from Greek with lengthening of the short vowel under stress (see Biville 1995: 23), though the most common interpretation is that those words were borrowed independently into both Latin and Greek from some substrate

17

An earlier date of iambic shortening (4th-3rd century bc) is given in Meiser (2010: 76).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The long-term trend

11

language. Whatever the case here, these sparse putative exceptions do not impinge on the basic generalization: in most cases, limitations on contrastiveness came about via shortening, which normally targeted unstressed vowels, from the origins of Latin. As a consequence of these changes, by the Classical period the VL contrast was largely limited to stressed syllables, with a skewing analogous to that observed for consonant gemination (see }1.1). Herman’s (1968: 197, n. 5, 199) text counts show that the ratio of long to short vowels in Classical Latin was 1:3 under stress, while the text frequency of long vowels dropped to 1:4 in unstressed syllables. Thus, already by Classical Latin times VL had ceased to be (at least statistically) totally independent from stress. This change is not surprising, neither in a general-typological perspective nor with specific reference to the language family Latin belongs to. Other Italic languages, in fact, followed a similar path. In Oscan, ‘la prévalence de l’initiale [ . . . ] a eu pour conséquence . . . un amenuisement des différences de quantité dans les parties subséquentes du mot’ [‘the prevalence of the initial [syllable] had as a consequence a diminution of the quantity differences in the following parts of the word’] (Lejeune 1975: 245). Oscan and Umbrian probably preserved Proto-Italic initial stress ‘jusqu’en pleine époque historique’ [‘well into the historical period’] (Kuryłowicz 1958: 381). Both Sabellian and Latin, consequently, shifted away, though at different paces, from the Late PIE free distribution of CVL, in the direction predicted by what has been claimed to be a cross-linguistically valid generalization: ‘There are languages with long vowels and short vowels where all long vowels must be stressed, but there are no languages with long and short vowels where all short vowels must be stressed’ (Blevins 2004: 9). In addition to stress, the further contextual factor that played a role in the Late Latin merger of CVL was (as will be illustrated in Chapter 2) syllable structure. Also here a ‘flash forward’ is in order, since some changes which applied already in archaic Latin can be viewed, again, as forerunners of later development. Consider the distribution in (10a): (10)

a. examples b. % stressed % unstressed

i. light CV rgd f c l 7% 24.5%

open closed ii. heavy (no coda) iii. heavy (coda) iv. superheavy CVː CVC CVːC st ll s pr u t c nt ng t t ct s d p n p ru rs m 10.3% 12.8% 2.3% 11.3% 26.5% 5.3%

TOTAL

32.4% 67.6%

While all four combinations of VL and syllable structure were admissible, the set of superheavy syllables was sizeably depleted by change.18 For instance, geminate

18 The term ‘superheavy’ is employed here to single out (10iv)-type syllables, since their having been selectively targeted by change shows that Latin phonology treated them as a natural class. Unlike in other languages, however, this had no impact on any synchronic rule—like the stress rule in (4) above—nor on

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

12

Introduction

-ss- was degeminated after a long vowel or diphthong around 100 bc (Leumann 1977: 181): fūssus ‘poured’ > fūsus, cāssus ‘fall’ > cāsus, caussa ‘cause’ > causa. This led to differentiation between forms which once were identical, as in s-perfect formation, where gĕssit ‘carry.prt.3sg’ retains -ss- after a short vowel, while in mīsit ‘release. prt.3sg’ < mīssit (CIL I2 1216) degemination occurs because of the root long /i:/. At a more recent date, strings consisting of /V:/ + geminate lateral (and, more marginally, /nn/) must also have become problematic, as shown by their elimination through contrasting strategies in different Romance branches: a *ste¯la (< ste¯lla) has to be reconstructed for Gallo-Romance (Fr. étoile ‘star’; cf. the open syllable diphthongization as in toile ‘canvas’ < te¯la) and Rhaeto-Romance (Engadinian staila, again with the same open-syllable diphthongization of e¯ as in taila), whereas the Romanian cognate stea goes back to *ste˘lla (compare şa ‘saddle’ < *se̯a < se˘lla), with vowel shortening. However widespread, these changes did not result in complete eradication of /V:ll/ strings, as witnessed by It. m[i]lle ‘thousand’ which, given the diachronic correspondences in stressed vowel qualities discussed below in Chapter 2, (15), attests to preservation of both the long vowel and the geminate consonant. But on the whole they contributed to the statistical imbalance illustrated by Kiss (1971: 14), whose counts (performed on an 814-syllable passage from Caes. Gal. 3.3–5) are reported in (10b). As is readily apparent, superheavy syllables (10iv) were a clear minority, both in stressed (2.3%) and unstressed position (5.3%). Symmetrically, light syllables (10i) were much less frequent under stress (7%) and occurred much more frequently in unstressed position (24.5%). Thus, the establishment of a complementary distribution of VL with respect to stress position and syllable structure, which was to reach completion in the Latin–Romance transition (see Chapter 2), as Wüest (1979: 117) nicely put it, ‘apparait déjà en filigrane dans l’ancien système’ [‘already appeared as a watermark in the old system’].19 It is now time to move on, in the next chapter, to scrutinizing the evidence pointing to this transformation. Before this, the plan of the book will be summarized.

1.3 Structure of the argument and aims of the book Chapter 2 will be devoted to expounding the philological evidence available on the transition from the Latin quantity system, featuring both contrastive VL and contrastive consonant gemination, to the early Romance one, in which VL was no longer contrastive. Sifting the evidence—which is intricate and has been the subject poetic metre: in other words, Latin had no synchronic ternary weight distinction; this contrasts with e.g. Hindi, where CVVC and CVCC syllables count as superheavy and depart from heavy syllables CVV and CVC for purposes of stress assignment (Broselow et al. 1997: 19; Davis 2011: 113; Zec 2011: 1346). Many others (including e.g. Lüdtke 1956: 127–31; Herman 1968: 202; Porzio-Gernia 1976–7: 149f.; A. Castellani 1991: 19) have stressed the role of the increasing subordination of vowel quantity to stress as a forerunner of the demise of CVL in Late Latin/early Romance. 19

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Structure and aims of the book

13

of unceasing debate—I shall show that recent research on Late Latin has come up with a plausible story about how things may have gone: these conclusions from research on Late Latin cannot be ignored when advancing hypotheses on how the individual Romance CVL systems came into being. While in Chapter 2 modern Romance data will be mentioned only inasmuch as they provide clues for reconstruction (a necessary move, since Proto-Romance, by definition, is not attested), these data come to the fore in Chapter 3, which consists of a tour d’horizon of VL in modern Romance languages and dialects. These can be classified in several ways, as far as VL is concerned. A first subdivision is between languages displaying CVL and languages lacking it, whereby all the contemporary standard languages belong to the latter category. However, between, say, Italian, Romanian, and French, there are clear differences: while Italian has no CVL but has an allophonic rule deriving long vowels in stressed open syllables, Romanian, also lacking CVL, does not display such an allophonic rule. French is like Romanian today, in this respect, but contrary to Romanian, there is evidence that it must have gone through a stage in which VL was distinctive, under conditions similar to those observed to this day in several Gallo-, Rhaeto- and Northern Italo-Romance dialects (i.e. in an area usually dubbed ‘Northern Romance’, see Map 2 in Appendix). These three types display a certain areal coherence: absolute lack of phonological specification of VL, either phonemic or allophonic, is found not only in the Balkans but also in the entire Iberian peninsula (though with a few exceptions),20 whereas the part of Central Romance complementary to the northern area just mentioned, namely central–southern Italy and Sardinia, has VL ruled allophonically like Standard Italian: the Italian facts will be paid special attention, since (a) the open syllable lengthening at work there plays a major role in the reconstruction of the Latin–Romance transition advocated here, (b) its existence has been insistently denied in recent work in experimental phonetics and theoretical phonology, and (c) its relevance has often been ignored in recent studies on the diachrony of CVL in Northern Romance. After describing the different Romance VL systems in Chapter 3, I shall move on in Chapter 4 to considering competing hypotheses on the rise of secondary CVL in Northern Romance, focusing especially on the varieties for which this issue has been most debated. The same chapter also discusses the method of diachronic phonology, in an attempt to pave the way for Chapter 5, in which I put forward my own reconstruction. This crucially capitalizes on a detailed inspection of dialect variation, which (as I intend to show) is of paramount importance for fully assessing the competing explanations for the sound changes under discussion.

In Old Portuguese, derived long vowels developed via coalescence (e.g. [’sO:] ‘alone.fsg’ < soa < solam, [’sε:ta] ‘arrow’ < saéta < sagittam; cf. Naro 1971: 382, de Haas 1988: 175–8), but these did not give rise to any phonemic contrast (cf. Mateus and d’Andrade 2000: 30). 20

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

14

Introduction

The comparative reconstruction to be outlined in Chapter 5 results in a hypothesis about how the VL systems observed in different dialects relate to each other and about how this relationship arose over time. This is further confirmation of an old tenet in historical linguistics, according to which geographical variation across related dialects is liable to mirror the stepwise progression of change over time and space. In the final part of Chapter 5 (}5.4), I shall propose an explanation for the sequence of diachronic changes thus reconstructed: this appeals to the interplay between substantive phonetic forces affecting vowel duration and phonological constraints on VL, which are in part motivated by the phonetics but in part also independent from it, as phonology is in general (at least, under the view of phonology espoused here, as made explicit in }3.2). Chapter 6, finally, will draw some conclusions, and address two more collateral topics, providing a discussion of the effectiveness of accounts in terms of foot structure of the Romance facts investigated here, and a sketchy comparison with similar data from West Germanic, another language (sub-)family which on the one hand displays CVL and on the other has undergone open syllable lengthening. The aim of the book is declared by its title: I intend to describe the way in which VL changed from Latin to the Romance languages and dialects. This could be done, in principle, in many different ways, as attested by the extremely varied literature on the topic, ranging from work in theoretical phonology (e.g. Mester 1994) to studies in Latin philology (e.g. Allen 1973) and the historical sociolinguistics of the Latin language (e.g. Adams 2013: 37–51). The references just cited, all pieces of scholarship of the highest order, could hardly differ more strongly from each other in terms of treatment of the primary data, focus, priorities, expository style, etc. Yet they all deal with VL in Latin, and comparably diverging references could be adduced for any Romance languages touched upon in what follows. For contemporary varieties, to the types of text just mentioned we should add experimental phonetic studies (e.g. Renwick’s 2012 recent comprehensive study on Romanian vowels). While literature from these different genres will be taken into account in what follows, the present book is an essay in historical Romance linguistics, whose basic concern is to provide a detailed survey of the relevant Romance data and to contribute to the understanding of how these data have evolved in diachrony, and of why they evolved the way they did. This is reflected in the tenor of the following discussion. Even if this (admittedly) hardly meets the expectations laid upon a monograph in phonology and/or phonetics in the current academic landscape, the results of the discussion will also be of relevance for theoretically interested phonologists. One first reason is ex negativo, and concerns the issue of the limits of synchronic explanation, an issue addressed avant la lettre (i.e. before the establishment of modern synchronic linguistics) by Paul (1880: 20)—according to whom synchronic consideration of language is merely descriptive, while only reconstruction is explanatory—and influentially revived in more recent times by Blevins (2004: 311),

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Structure and aims of the book

15

who maintains: ‘Where historical explanations are available [for cross-linguistic tendencies—M.L.], there is no need to encode these tendencies within synchronic grammars, either as markedness constraints or structural primitives’. Theoretical phonology is primarily a synchronic discipline, which develops tools for analysing sound systems as part of language competence. Applications to diachrony are mostly parasitic on the synchronic method—whatever this is pro tempore, from structuralist phonology to OT. Much has been written on whether this is fruitful, and I shall briefly take up the issue in }}4.2–3, drawing on the evidence discussed in this book. What my case study clearly shows is that straightforward projections of synchronic accounts of CVL in Romance onto the explanation of its origin in this or that system, which are current in the literature, are doomed to failure basically because of the intrinsic limits of internal reconstruction. This is, therefore, the first reason (albeit a negative one) why scholars working on synchronic phonology should take an interest in this monograph, if only in order to become aware of these limits when venturing onto the ground of diachrony: following the Socratic motto, it is useful to know what you do not know. More positive reasons why this book could be of interest to readers not specializing in historical Romance linguistics concern a number of empirical results of the present research. For instance, if, as argued here, most of the Romance systems displaying CVL first acquired it through the phonologization of the output of an earlier allophonic rule lengthening stressed vowels in open syllables (see, in Chapter 2, (6), and Chapter 3, (3)), this conclusion flies in the face of claims such those of de Chene (1979: ix), according to whom ‘there is only one mechanism by which a language without a distinction of vocalic length commonly introduces such a distinction. This mechanism is the coalescence of vowel sequences’. This hypothesis (see also de Chene and Anderson 1979: 523) has been discussed by specialists in Romance phonology, who point to the fact that Romance data crucially disconfirm it (see Hualde 1990: 31; Morin 1994: 144–7). The present monograph adds further weight to the burden of proof, for any future proponents of such a view. The same applies to the results of any discussion of specific changes, which intersect necessarily with the cross-linguistic database on which general hypotheses in the relevant domain are tested. Thus, in a cross-linguistic study of compensatory lengthening (CoL), Kavitskaya (2002: 108–17, 160–63) devotes much space to Friulian, which is inventoried among the languages in which CVL has arisen anew through CoL (Kavitskaya 2002: 108): C[ompensatory] L[engthening] is the sole source of length oppositions in Friulian, which thus constitutes one of a few counterexamples to the claim that C[ompensatory] L[engthening] is a structure-preserving sound change.

This is indeed what her sources claim about Friulian (as discussed in }4.1.2 below): clearly, if it turns out that a more economical explanation for the rise of CVL in

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

16

Introduction

Friulian traces it back to a previous allophonic open syllable lengthening, rather than to CoL, the place of this variety in the database for such cross-linguistic studies as the one just quoted will have to be revised. To an extent, this in the nature of things: room for conflicting interpretations of the dataset, in phonology as elsewhere, will never be exhausted, so any conclusions, including those of the present work, have to be regarded as provisional. However, one way in which I hope this monograph will serve the scientific community, whatever the acceptance of the specific ideas and interpretations aired in what follows, is by presenting a reasonably detailed and reliable inventory of the facts which any further study of the same empirical domain will have to consider in order to provide a sensible diachronic account of the changes which affected VL from Latin to the Romance languages. Obvious though this may seem at first glance, the relevance of such an endeavour does not seem negligible to me, especially given the circumstance that in the present academic landscape, one often meets with studies in theoretical phonology and/or experimental phonetics which apply their synchronic machinery to the explanation of sound change without a clear notion of the historical changes to be thereby ‘explained’, as though knowledge of the data to be accounted for were not a precondition for the soundness of the analysis. To illustrate, consider the following case. In a cross-linguistic study of the phonetics and phonology of vowel reduction within Optimality Theory, a section ‘On the Historical Development of Vowel Reduction’ contains the following passage: Historically, Italian tense /e, o/ derive from etymological short vowels (including Proto-Romance short /e,o,i,u/), while lax /ε, O/ derive from etymological long vowels. The long~short distinction was ultimately lost in Italian, but was apparently retained in stressed position longer than it was retained in unstressed positions [ . . . ] When the long~short distinction was finally lost in stressed as well as unstressed syllables, the contrast between previous /e/ vs /e:/ and /o/ vs /o:/ was recast in terms of a quality distinction: /e/ vs /ε/ and /o/ vs /O/. (Crosswhite 2001: 116–17)

It is true that quantity distinctions were recast in terms of quality distinctions, and there is much that is far from uncontroversial, in this area, as shown below in }} 2.2–2.4. Yet anybody with a basic acquaintance with the data mentioned in this passage will realize that the author, notwithstanding the merit of having included in her database linguistic evidence from a domain in which she is not a specialist, nevertheless cannot be credited with checking appropriately the extensive literature which has established the historical facts she mentions (see especially }2.2.2, (15) below): contrary to her claim, Latin short /e, o/ vs long /e:, o:/ gave rise to Italian lax vs tense mid vowels, respectively. Crosswhite’s book focuses on the cross-linguistic constants observed in vowel reduction processes, and applies the results of synchronic study in phonetics and phonology to the investigation of sound change in the past—a well-established procedure with illustrious predecessors ranging from Osthoff and Brugmann’s

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Structure and aims of the book

17

(1878: 163–4) remarks on Winteler (1876), to Labov (1974), Ohala (1989), and others. In order to apply such results to past stages of whatever language—and, consequently, in order to be entitled to use the adjective ‘historical’ (or the adverb ‘historically’)—one has to first acquaint oneself with the facts about the past of that language. Needless to say, as in other historical disciplines, the past in historical linguistics is construed, as pointed out effectively by Lass (1997: 27): To the extent then that history is not observational but argumentative, it is necessarily constructivist; the historian participates actively in making his subject matter.

This by no means implies, however, that any narrative will do, à la Barthes (1967). Historical linguistics has not been infected by the sophisms of postmodernism: rather, it has a well-established method, and in order to be able to talk competently about language change one first has to tune in to that method, and then to sift the available evidence through it. In the specific case of sound change from Latin to Romance, the scholar who wants to tackle it seriously is indeed confronted with an extremely rich and multifarious body of evidence. This will come as no surprise to anybody trained in historical Romance linguistics. If readers from outside this tradition get at least a flavour of how serious a task it is to claim anything sensible about sound change from Latin to Romance when venturing into this domain, this book will fully have reached its goal.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

2 Vowel length in the Latin–Romance transition The Latin–Romance transition is by no means a straightforward or easy-to-define concept, as attested by the endless discussions stirred by notions such as Vulgar Latin or Proto-Romance.1 The general issue is not for me to rediscuss in depth in this context, since this would demand a separate monograph. Rather, I shall proceed inductively and sift in what follows the evidence from various sources bearing on the maintenance vs loss of CVL: as a by-product, it is hoped that this will result in a contribution to the debate on the chronology of the Latin–Romance transition, in which the demise of CVL—a property of the phonology distinguishing CL from all the Romance languages—features prominently. According to some scholars, the description of the phonology of CL—including CVL—sketched in Chapter 1 basically mirrors a conservative written variety which had ceased to exist in the spoken language (so Pulgram 1975: 287f.), or at least in its basilectal varieties (e.g. Vineis 1984b; 1993: xlii–v; Giannini and Marotta 1989: 272), as early as the 3rd century bc. If this was, according to such views, the situation in Rome, it is only natural for it to be mirrored in the provinces too: thus, for Alarcos Llorach (1951: 13), in the 3rd century bc Roman legionaries and colonists imported into Spain a variety of Latin without contrastive VL. For other, less extremist views, VL was gone by the 1st (e.g. Bonfante 1968; 1983; Väänänen 1966: 18f.) or 2nd century ad (A. Castellani 1991: 21).2 This is the background to provocative claims such as those of Bonfante (1983: 449), according to whom in Pompeii and in general in the cities of Italy in the 1st century ad people spoke ‘una lingua in cui si diceva . . . otto per octō . . . . Questo non è latino: è italiano’ [‘a language in which one said otto instead of

1 See e.g. Banniard (1992: 17–29), Zamboni (2000: 88–92), Coseriu (2008: 40–177) for insightful recapitulations of the different opinions on the chronology of the Latin–Romance transition, and Adams (2013: 3–27) for a recent critical discussion of the notion ‘Vulgar Latin’. 2 The list could be easily continued, with datings to the 3rd, 4th, or 5th c. ad, to show that virtually all possibilities have been exhausted: e.g. ‘etwa im 3. Jhdt. n. Chr.’ (Leumann 1977: 55); ‘au cours du IIIe et IVe siècle’ (Bourciez 1946: 42); ‘probably the disappearance of the old distinction was not general before the fourth and fifth centuries’ (Grandgent 1907: 76).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The Latin–Romance transition

19

octō. This is not Latin: it is Italian’]. (A similar view of an early rise of ‘Italian’ is maintained e.g. by Křepinský 1958.) As suggested by reference to /’okto :/ > /’ɔtto/, such a view implies that the transition to the Romance non-quantitative system is something which happened early on, by the 1st century of the Empire.3 A radically different take on the issue characterizes the studies on the Latin– Romance transition by many specialists in Late and Medieval Latin. In general, these have been inclined to recognize a continuity in spoken Latin (and thus a correspondence between the written record and the spoken language) until very late, even up to the end of the 8th century ad for Müller (1921), with many in-between shades, such as those represented by E. Löfstedt (1959), Norberg (1958), and the other references mentioned in Banniard (1992: 21–2), who locate the transition between 600 and 800 ad, or Pfister (1987: 329–32), who sets the divide around 600 ad for lexicographical purposes, or Zamboni (2000: 90), according to whom the transition starts—with the ‘divorzio non registrato tra lingua scritta e lingua parlata’ [‘the unrecorded divorce between the written and the spoken language’]—in the 4th–5th century and is accomplished by the 8th. According to this view, only after this date do Romance languages clearly emerge as linguistic realities, not only in the trivial sense of emerging to written documentation. Among specialists in Late Latin, a particularly elaborate and well thought-out scenario was proposed by Herman (1985: 88–9; 1998: 9, 21), who distinguishes three waves of differentiation/dialectalization of Latin. After the first, which took place during the Romanization of the Italic peninsula as Latin spread to eventually oust the pre-existing languages in the conquered regions, the second wave, which he dubs ‘pre-Romance differentiation’, ‘s’est déroulé sous l’Empire et [ . . . ] a duré jusqu’aux décennies qui ont suivi la chute de l’Empire d’Occident’ [‘took place under the Empire and lasted until the decades which followed its fall in the West’] (Herman 1985: 89). The demise of CVL is part of this second wave and, according to Herman (1998: 21), starts between the 2nd and the 3rd centuries and comes to completion towards the end of the Empire. This will be the main focus of the present chapter. In the subsequent ones I shall move on to consider the outcomes of the final differentiation (in Herman’s terms, the ‘Romance dialectalization’), which took place in a third wave of change, which puts an end to the history of Latin, dissolving the linguistic unity of the Latin-speaking world (Herman 1996), between the end of the Empire and the emergence of the earliest Romance texts.4 3 Strictly speaking, this example might indicate loss of CVL just in unstressed position, where merger started earlier, as seen in }1.2. However, Bonfante (1968: 24; 1983: 449) himself inferred from such examples that CVL was gone altogether by the 1st century ad (see also the discussion on (20) below). 4 An important turning point, in Herman’s (1996) account of the ‘end of the history of Latin’ (as a spoken language), is the years around 700 ad, when apocope in northern Gaul (cf. }3.4), which applied shortly after the deletion of unstressed word-internal vowels, reshaped the prosody of the spoken vernacular of that province in such a way as to render Latin unintelligible for illiterate speakers, whereas

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

20

The Latin–Romance transition

2.1 Change in Latin VL: metalinguistic testimonies and the rise of OSL Compared with those examined in }1.1 (6)–(7), metalinguistic judgements by Latin authors change significantly during the last two centuries of the Western Empire, as several writers report uncertainty about and/or mispronunciation of long vs short vowels, which seems to point to the Romance development. Some of the evidence available seems to lend itself to contradictory inferences, which explains why rivers of ink have flowed on the interpretation of the crucial passages. I shall try to accomplish the difficult task of making sense of testimonies by ancient authors in a way that is compatible with what subsequent Romance development tells us, so us to pave the way for further discussion of the latter in the following chapters. The grammarian Sergius (4th century ad) observes that: (1)

syllabas natura longas difficile est scire; sed hanc ambiguitatem sola probant auctoritatis exempla, cum versus poetae scandere coeperis [‘it is hard to know which syllables are long by nature; this ambiguity, however, can be checked solely through examples drawn from the good authors, as one starts to scan the poets’ verses’] (GL 4.522.24–6).

For him, it is metric scansion which helps to calculate quantities (in prose), rather than the other way around. Somewhat later, a series of testimonia on both the phonology and the phonetics of VL is provided by two rather special authors, viz. Augustine (354–430 ad), and Consentius (probably a Gaul, writing in the first half of the 5th century; cf. Loyen 1943: 78–81; Holtz 1981: 83; Kaster 1988: 396–7; Fögen 1997–8: 168–72). The well-known passage by Augustine in (2) reports neutralization of /o :/ and /o/, formerly contrasting in ōs ‘mouth’ vs o˘ s ‘bone’, in the context of proclaiming the need for intellectuals to accommodate their lexical choices so as to be intelligible to the mass of uncultivated people. Thus, he argues, since os has become ambiguous, a preacher who wants to be understood by people is advised to use the (substandard) doublet ossum for ‘bone’: (2)

cur pietatis doctorem pigeat imperitis loquentem, ossum potius quam os dicere, ne ista syllaba non ab eo, quod sunt ossa, sed ab eo, quod sunt ora, intellegatur, ubi Afrae aures de correptione uocalium uel productione non iudicant? [‘Why should a teacher of piety when speaking to the uneducated have regrets about

their ancestors could still understand Latin natively until about the mid-7th century (cf. Banniard 1992: 487ff.), i.e. prior to the application of those changes. This ‘second prosodic revolution’ (Loporcaro 2011c: 67)—the first being the demise of CVL—put an end to the transmission of Latin as a natively acquired language. It broke the ‘horizontal’ mutual intelligibility among different regions in what thereby became the Romance linguistic continuum, as well as the ‘vertical’ intelligibility within the linguistic repertoire, between the local vernacular and Latin, learned and used for official purposes. Other parts of the former Western Roman Empire followed the same path after some delay, due to the lesser impact of changes in the prosodic system.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Change in Latin VL

21

saying ossum (‘bone’) rather than os in order to prevent that monosyllable (i.e. os ‘bone’) from being interpreted as the word whose plural is ora (i.e. ōs ‘mouth’) rather than the word whose plural is ossa (i.e. o˘ s), given that African ears show no judgement in the matter of the shortening of vowels or their lengthening?’] (Augustine, De doctr. christ. 4.10.24, trans. Adams 1999: 115). Augustine identifies this insensitivity to VL as a trait of African Latin, concurring in this with other sources, as we shall see shortly. In another passage Augustine provides a description of the phonetic process through which the phonological merger was carried out: (3)

itaque uerbi gratia cum dixeris cano uel in uersu forte posueris, ita ut uel tu pronuntians producas huius uerbi syllabam primam, uel in uersu eo loco ponas, ubi esse productam oportebat; reprehendet grammaticus, custos ille uidelicet historiae, nihil aliud asserens cur hunc corripi oporteat, nisi quod hi qui ante nos fuerunt, et quorum libri exstant tractanturque a grammaticis, ea correpta, non producta usi fuerint. [‘And so, for example, when you say cano or happen to use it in verse, such that you either lengthen in pronunciation the first syllable of this word or place it in verse in a position where it should be long, the grammarian, that guardian of tradition, will find fault with you, giving no other reason why it should be shortened except that those who have come before us and whose books survive and are handled by the grammarians have treated it as short not long.’] (Augustine, De musica 2.1.1, trans. Adams 2013: 44).

The context, much like that of (1) above, is one in which the narrating voice, in dialogue with a pupil, says that quantities at that time and place depend on authority only, rather than on the speakers’ native judgement; and similar passages (discussed in Koller 1981) occur elsewhere in De musica (3.3.5; 5.5.10; 6.12.35). Augustine is referring here not only to errors in scansion (use of cano with a light first syllable in a verse position where a heavy syllable is required), but also to the phonetic implementation and/or the prosodic precondition thereof: a pronunciation [’ka :no], with lengthening of the originally short stressed vowel. It can hardly be considered coincidental that exactly the same phonetic description is delivered, slightly later, by Consentius, whose remarks point to a lengthening of originally short vowels in stressed open syllables (4a, c, d) as characteristic of African Latin, also contributing the complementary information that originally long vowels were shortened whenever unstressed (4b, d): (4)

a. Consentius, Ars de barbarismis et metaplasmis, GL 5.392.3–4: ut quidam dicunt piper producta priore syllaba, cum sit brevis, quod vitium Afrorum familiare est [‘just as some people say piper lengthening the first syllable, when it is short, which vice is characteristic of Africans’]

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

22

The Latin–Romance transition b. GL 5.392.11–12: ut siquis dicat orator correpta priore syllaba, quod ipsum vitium Afrorum speciale est [‘as if anyone were to say orator with the first syllable shortened, which vice is particular to Africans’] c. GL 5.392.18 [exemplifying barbarismus per immutationem temporis] ut si quis pices dicens priorem extendat [‘as if anyone, saying pices, were to lengthen the first syllable’] d. GL 5.392.25–6: ut siquis dicens pices producta priore et correpta sequenti pronuntiet [‘as if anyone, uttering pices, were to pronounce it lengthening the first syllable and shortening the following one’]

While there are other passages by grammarians reporting vacillation of CVL in the Latin of the Low Empire, those by Augustine and Consentius are of special importance.5 Augustine’s remarks on the issue, which are a polemic against grammarians, centre on the spoken language and the native competence of uneducated speakers.6 Similarly, Consentius, while writing a treatise of which two parts survive (the passages in (4) stem from one of those, De barbarismis and metaplasmis), was not a full-time grammarian: rather, his style [ . . . ], his readiness to quote from the spoken Latin of his day [ . . . ], and his independence in organization and judgment [ . . . ] all combine to distinguish his work from that of the professional grammatici [ . . . ]. On the evidence available, he should be placed in the class of learned amateurs. (Kaster 1988: 396)7

Finally, another later testimony by a professional grammarian also from Africa, Pompeius (5th–6th century) also reports uncertainty on VL:

5 Deviation from the CL norm along the same lines as in (3) ([’dε :os] instead of CL [’dεo :s]) is reported, without reference to Africans, by Aelius Donatus (mid-4th c.) (see (i.a)), who was perhaps himself of African origin (Kaster 1988: 275–8), but also (see (i.b)) by grammarians neither of African origin nor mentioning peculiarities of African accent, such as Marius Plotius Sacerdos (probably 2nd half of the 3rd c. ad: see Kaster 1988: 352–3), who mentions a mispronunciation Cērĕs for CL Cĕrēs:

(i)

a. Donatus, Ars grammatica, Holtz (1981: 654.5–6): ut si quis deos producta priore syllaba et correpta posteriore pronuntiet [‘as if anyone were to say deos lengthening the first syllable and shortening the following one’]. b. Sacerdos, Ars, GL 6.451.13: per immutationem accentuum, ac si dicas Ceres ce longa, cum breuis sit, et res breui, cum sit longa [‘(a barbarism may arise) through stress modification, as if you were to say Ceres with a long ce, whereas it is short, and short res, while it is long’].

6 Proclaimed contempt towards grammarians is a topos of Roman philosophy: e.g. Seneca’s Ep. 95.65 sarcasm on the ‘gramatici, custodes Latini sermonis’ [‘grammarians, guardians of Latin speech’], as well as ep. 108.30 (cf. Kaster 1988: 53). 7 This repeats judgements by Keil (GL 5.333), Barwick (1922: 36), Loyen (1943: 80), and Holtz (1981: 85–6) who point to Consentius’ autonomy from previous grammarians: ‘Er entnimmt seine Beispiele zu den einzelnen Barbarismen nicht der auctoritas scriptorum, sondern ausschließlich dem “usus cotidie loquentium”; er begründet sein Verfahren 391.26ff ’ [‘He does not draw examples of specific barbarisms from the authority of writers, but exclusively from usage of everyday speakers; he justifies his procedure at 391.26ff ’] (Barwick 1922: 36).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Change in Latin VL (5)

23

est alter [sc. barbarismus] qui fit in pronuntiatu. plerumque male pronuntiamus et facimus uitium, ut breuis syllaba longo tractu sonet aut iterum longa breuiore sono: siqui uelit dicere Ruoma, aut si uelit dicere aequus pro eo quod est equus, in pronuntiatione hoc fit [‘there is another [barbarism] which is committed in pronunciation. Often we utter a bad pronunciation and commit the fault of sounding a short syllable long or, again, a long syllable shorter: as if anyone wanted to say Ruoma, or wanted to say aequus instead of equus, this (is a barbarism which) is committed in pronunciation’] (Pompeius, Commentum artis Donati, GL 5.285.5–9, trans. partly from Adams 2007: 263)

Marco Mancini (2001: 331) regards this passage as bearing on the phonology of African Latin, and specifically as attesting to the occurrence of free variation in vowel length under stress (rather than to systematic vowel lengthening in open stressed syllables). The argument is based on Mancini’s (1994) conjecture that the incongruous be read as Ruma and phonetically interpreted, ‘secondo i canoni del latino parlato “italico” ’ [‘by the canons of ‘Italic’ spoken Latin’] (on which label see (22) below) (Mancini 2001: 317) as [’roma] (instead of CL [’ro :ma]). On this view, Pompeius is held to describe here free variation of VL unconstrained by syllable structure, which disproves in Mancini’s (2001: 318) opinion the occurrence of OSL: under this reading of (5), in an open syllable, a shortening (such as in putative [’roma]) is just as possible as a lengthening (as in equus pronounced as [’ε :kwʊs], instead of CL [’εkwʊs]).8 According to Adams (2007: 263), however, the 1st person plural used in the passage in (5) does not imply reference to African speech, since ‘Pompeius alternates between the first and second person throughout the work in describing general features of the language’. Under this interpretation, Pompeius in (5) is reporting in general about uncertainties regarding VL, not specifically about African Latin. And it is undeniable that such testimonies (cf. (3) and (4), and (i.a,b) in n. 5) mostly report OSL (in keeping with the evidence from later Romance development) rather than shortening of long vowels in stressed open syllables.9 8 A shortening is implied also by Seelmann’s (1890: 41) alternative conjecture , with originating via erroneous copying of the shortness sign as if it were a . 9 That a putative testimony about open syllable shortening under stress, if this were indeed the sense of the passage about the pronunciation of Roma in (5), is (or rather, would be) isolated, was argued in Loporcaro (1997: 66). Mancini (2001: 316–18) objects to this, and he is of course right in pointing out that the examples chosen show that Pompeius was drawing from a traditional set of customary examples (or ‘parole bandiera’ [‘flag-words’], as Filipponio 2006: 53 puts it), which recur several times in the corpus of Latin grammarians. Among these, however, the only other place in which the word Roma is mentioned to exemplify an error in vowel quantity is the following:

(i)

Servius, in Donat., GL 4.444.4–14: barbarismus autem dicitur eo, quod barbari praue locuntur, ut siqui dicat Rumam pro Roma. [ . . . ] fit autem barbarismus principaliter modis duobus, pronuntiatione et scripto: si aut naturaliter longas syllabas breuiter proferamus, ut Romam, aut si naturaliter breves producamus, ut rosam [‘it is called barbarism because barbarians speak poorly, as if

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

24

The Latin–Romance transition

Among those testimonies, as we have seen, those by Augustine and Consentius are of special relevance for our issue. This said, we still have to agree on their exact import. Schuchardt (1866–8: 3.43–4) inaugurates a tradition of taking Consentius’ remarks at face value: ‘Darnach haben sie [sc. Africans] am frühesten romanisch gemessen, d.h. betonte Vokale bei folgendem einfachen Konsonanten lang, unbetonte kurz gesprochen’ [‘according to this testimony, it was Africans who first scanned in the Romance way, i.e. pronounced stressed vowels as long when followed by singleton consonants, and unstressed vowels as short].10 Under this interpretation, the scenario is supposed to be one in which CVL is lost earlier in African than in Roman Latin, through the rise of an OSL rule such as the following: (6)

V !V :/____ ]σ [+stress]

A similar rule—as we shall see in }3.2—has to be assumed for modern Standard Italian and several other Romance varieties: its origin, in this view, is directly rooted in Late Latin, perhaps African Late Latin. That OSL has to be inferred from the descriptions in (3) and (4) is not uncontroversial, though (see especially Schürr 1970, discussed in }2.2.1),11 and there is no consensus on the ascription of the merger and/or of OSL to African Latin either. Consentius’ remarks are held to show only ‘dass im Allgemeinen der Tonvokal in freier Stellung gelängt wird’ [‘that in general [i.e. not specifically in African speech—M.L.] the stressed vowel was lengthened in free position’] (von Wartburg 1950: 81). Under this interpretation, mention of Africans in this connection is argued to have been motivated by the commonplace of African Latin as a paramount representative of ‘bad’ pronunciation. anyone were to say Rumam instead of Roma. [ . . . ] a barbarism can be produced mainly in two ways, in pronunciation or in writing: in pronunciation, either if we were to pronounce as short syllables which are long by nature, as in Romam, or if we prolong vowels which are short by nature, as in rosam’] Note that here the formulation is ambiguous, as Servius (c.400 ad) does not claim to be reporting on actually occurring mispronunciations, unlike Augustine and Consentius in (3) and (4). Thus, he might well be mentioning exempla ficta (made up using the traditional set of words), meant to illustrate how such errors could in principle arise. As for the other passages mentioned by Mancini (cf. (12) and (13) below), they do contain the same examples (Roma among them), but do not discuss errors in vowel length and hence are not relevant to the interpretation of Pompeius’ Ru(o)ma in (5). 10 The same interpretation of Consentius’ testimonies is given e.g. by Bonfante (1956: 354), Acquati (1971: 158), and Pisani (1975: 168). 11 Mancini (2001: 314) argues, like Lüdtke (2005: 202), that the demise of CVL took the form of generalized free variation of [V( :)] rather than OSL and in addition, for reasons unclear to me, defines Consentius’ remark on orator (4b) as ‘non [ . . . ] rilevante’ [‘not relevant’] to the issue. The same example is ‘adjusted’ by A. Castellani (1991: 20–21), in accordance with his conviction that CVL was lost throughout the Empire by the 2nd or at latest the 3rd c. ad. If so, then neither Consentius nor Augustine could possibly have contrasted (conservative) varieties with CVL with innovative ones having neutralized it: what is at stake in the orator passage in (4b), then, must be the quality of the stressed vowel. To argue this, however, Castellani has to propose that orator should be emended to orat. As we shall see in }2.3.1–2, such speculations are proved false by the undeniable existence, at least as late as the 3rd c. ad, of ‘dialects or accents of Latin [ . . . ] preserving distinctions of vowel quantity’ (Adams 1999: 117).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

25

As to the exact realization of the lengthening in piper, according to von Wartburg Consentius must have meant [’pe :per], with the common Romance development (ĭ > /e/). Weinrich (1958: 24–5) and Lüdtke (2005: 491–2), on the other hand, assume that OSL was already established not only in Africa but throughout the Empire by that time, and therefore take passages (4a,c) to prove (indirectly) that African Latin had a vowel system of the Sardinian type. Consentius’ remark is argued to have been motivated by observation of the difference between his own pronunciation (which must have been [’pe :per]), characteristic of the Latin of Gaul, with ĭ > [e :], and the African pronunciation [’pi :per], where /i/ was not lowered to [e], so that the deviation from the classical norm consisted only in a lengthening.12 More recently, Adams (2007: 264) too concludes that ‘the two tendencies brought up by Consentius (lengthening of short vowels in accented syllables and shortening of long vowels in unaccented syllables) are not confined to Africa’. This amounts to ascribing OSL to (Late) Latin as such, under the form of the ‘two complementary tendencies, for short vowels under the accent to be lengthened, and long vowels in unstressed syllables to be shortened’ (Adams 2011: 275), which is expressed synthetically in rule (6). The occurrence of OSL is a cornerstone of the argument to be elaborated here regarding later Romance development. For this cornerstone to stand, one has to make sure that this view is preferable to alternatives that have been proposed and argued for at length, in diachronic studies on both Romance and Latin. According to other interpretations, OSL arose either much later or much earlier than the late Imperial period to which the testimonies mentioned in (1)–(4) refer. For proponents of a later date, this process concerned only specific branches of the Romance language family, and thus cannot be intertwined with the demise of CVL. Proposing an earlier date, on the other hand, often goes hand in hand with locating the demise of CVL well into the history of early Imperial or even Republican Latin (though, as we shall see, the reverse implication does not hold).

2.2 Why OSL must be Proto-Romance: excluding conceivable alternatives 2.2.1 OSL at a later date, and only in some Romance languages A later date for the rise of the OSL has been maintained by several scholars. Such a later dating follows automatically from the assumption that, when CVL collapsed, all Lausberg (1976: 204), on the other hand, does not subscribe to the hypothesis of an early OSL, but does agree with Weinrich and others that African Latin had a ‘Sardinian’ vowel system of the type displayed in (25) below, mentioning the passages in (3) and (4) as supporting evidence. For later discussion of whether African Latin had a Sardinian vowel system, see e.g. Fanciullo (1992: 178–9), Herman (1985: 192–3), and Adams (2007: 648), who concludes: ‘The inscriptional evidence from Africa is consistent with a vowel system of the Sardinian type’. Indeed, there is decisive proof, from a much later source than the passages in (3) and (4), that settles the matter and shows that the Romance of Africa did resemble Sardinian throughout its history: see }2.3.3. 12

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

26

The Latin–Romance transition

stressed vowels (irrespective of syllable structure) became ‘einigermassen long’ [‘somewhat long’] (Lüdtke 2005: 202; see also M. Mancini 2001: 314), i.e. longer than unstressed ones. An influential representative of this line of thought was Schürr (1956: 111; 1970: 5–6) who defended this view as instrumental to his general claim (cf. Schürr 1936) that Romance diphthongization of Latin stressed e˘ and o˘ (cf. }3.3), standardly regarded as a further development of OSL in languages such as French and Italian, arose instead throughout Romance because of the metaphonic effect of final -ı¯ and -u˘. This metaphonic effect was assumed by Schürr even for those Romance varieties which show diphthongization of stressed e˘ and o˘ not conditioned by metaphony (like Spanish or Tuscan-based Standard Italian) and for those in which metaphony took the form of vowel raising, with no traces of diphthongization (like Sardinian and Portuguese). To argue for this claim, he had to exclude the possibility that open syllable diphthongization as found in (Old) French (e.g. fier ‘fierce’ < fe˘rum vs fer ‘iron’ < fe˘rrum) or Tuscan (e.g. fiero ‘fierce’ vs ferro ‘iron’) could be a further development of an earlier OSL, as currently maintained. In order to achieve this goal, he had to subscribe to a very early dating of diphthongization (following e.g. Richter 1934: 138), capitalizing on epigraphic evidence which has since proved inconclusive (cf. }2.2.2 below). He also comments on Consentius’ testimony in (4) and reduces his phonetic (allophonic) specifications (‘lengthening the first syllable’) to Augustine’s phonemic ones in (2): ‘il faut réduire ces témoignages à leur juste valeur. Ils parlent de la confusion dans l’observation des quantités qui régnait parmi les Africains parlant latin, ni plus ni moin’ [‘these testimonies must be reduced to their real value. They talk about the confusion about quantities that prevailed among Latin-speaking Africans, nothing more and nothing less’] (Schürr 1970: 5). However, this hasty dismissal is at odds with what the source actually tells us, since Consentius explicitly speaks of lengthening under stress, as does Augustine in (3).13 Schürr also adds an areal argument, claiming that OSL (in Italian) and/or further developments thereof (in French) are limited to a central part of the Romancespeaking territory, which, by Bartoli’s (1943) argument from the conservative nature of ‘lateral areas’, must be regarded as innovative in this respect. OSL is not observed—Schürr (1970: 6) claims—in the rest of the Romània, ‘qui se trouve notamment en position latérale (portugais, espagnol, catalan, occitan, sarde, patois italiens méridionaux, roumain’ [‘that lies in a lateral position (Portuguese, Spanish, Catalan, Occitan, Sardinian, southern Italian dialects, Romanian)’].

See also Schürr (1956: 111). Schürr’s reductive interpretation of Consentius’ passage was followed by e.g. Lancel (1981: 277), A. Castellani (1991: 20), Crevatin (1992: 30), Fanciullo (1992: 178), and M. Mancini (1994: 625; 2001). 13

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

27

This list contains a number of inaccuracies. Sardinian, for one thing, does display OSL, as shown by the following Logudorese examples (and as systematically documented by Wagner’s 1941 transcriptions):14 (7)

[’kan :a] [’dɔ :mɔ]

‘reed’ ‘house’

vs vs

[’ka :na] [’drɔm :ɔ]

‘grey-haired.fsg’ ‘sleep.1sg’

Interestingly, in Logudorese this complementary distribution obtains regularly in a deeper layer of the lexicon, in indigenous words which underwent regular sound change, whereas in loanwords singleton intervocalic obstruents of the model language generally tend to be replicated like indigenous geminates, as shown in (8) (cf. Wagner 1941: 248–60): (8) a. intervocalic obstruents in b. indigenous geminates c. indigenous singleton loanwords consonants [a’t :ɔ :p :ɔ] ‘meet.1sg’ = [’a :pɔ] ‘have.1sg’ 6¼ [nε’βɔ :ðε] ‘nephew’ (Sp. topar) [’t :sɔt :a] ‘whip’ (Sp. azote) = [’ɔt :ɔ] ‘eight’ Systematic application of this adaptation strategy results in the peculiar rhythm of the regional variety of Italian spoken on the island, in which OSL is virtually absent because, in a sense, every single word in that code is a loanword. Compare the Sardinian regional Italian sentence in (9a) with its Logudorese Sardinian counterpart in (9b): (9)

a. [ti ’dˑikˑɔ ’kˑε nˑɔn tʃi ’sˑɔˑnɔ ’statˑɔ] b. [ti ’na :rɔ ’ɣi nɔ b :i ’zˑɔ i’sta :ðu] ‘I tell you that I have not been there’

In the latter, [’na :rɔ] and [i’sta :ðu] show OSL, whereas regional Italian [’dikˑɔ] and [’statˑɔ] do not. If anything, the evidence from Sardinian shows that OSL was at work there originally, and that it has been losing ground progressively, partly as a consequence of the increase in borrowed lexical items in the subsequent development of this Romance language. OSL occurs also in southern Italian dialects, as exemplified with one variety for each of the three main dialect subdivisions (‘area mediana’, ‘alto Meridione’, and ‘Meridione estremo’, respectively) in (10a–c) (see Map 1 for localization):

14 The Logudorese data in (7)–(9) are given in the phonetic form they have in the dialect of Bonorva (province of Sassari). Though Wagner (1941) does not provide instrumental evidence, he constantly transcribes Sardinian paroxytonic words with a stressed open syllable as [’CV :CV]. Conversely, Contini (1987) omits transcribing allophonic vowel length, though OSL is evidenced by the sonagrams he shows (e.g. that for [’ka :ra] ‘face’ in M. Contini 1987: 62).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

28

The Latin–Romance transition

(10) a. Macerata (T. Paciaroni, p.c.)

[’pa :tʃe] ‘peace’ [’fa :me] ‘hunger’

6¼ [’fat :ʃa] ‘face’ 6¼ [’fam :e] ‘do.imp.2sg=1sgIO’

b. Naples (AIS pt 721) [’ʃu :r@] ‘flower.pl’ 6¼ [’kur :@] ‘run.2sg’ [’ka :n@] ‘dog’ 6 [’kan :@] ‘reed.pl’ ¼ c. San Giovanni in Fiore (province of Cosenza; cf. Mele 2009: 171): [’mI :na] ‘beat.imp.2sg’ 6¼ [’mIn :a] ‘tit’ [’ka :rʊ] ‘dear’ 6¼ [’kaɽ̥ :ʊ] ‘wagon’ The original allophonic differentiation, preserved in (10), has further evolved in the dialects spoken on the Adriatic coast, from the southern Marche to Apulia, in which the lengthened allophone was modified via several diphthongization and/or colouring processes, as shown by selected examples, representative of the whole area, in (11a–c):15 (11)

a. Diphthongization of PRom /o/ in Bitontino (province of Bari, Merlo 1911–12: 916–17; Saracino 1957)

/o/ nepotem

utrem

bursam

pulvis

muscam

*flundam

currere

nəˈpautə

ˈɔtrə

ˈvɔrsə

ˈpɔlvə

ˈmɔskə

ˈjɔndə

ˈkɔrːə

‘nephew’

‘leather bottle’

‘bag’

‘dust’

‘fly’

‘sling’

‘run.inf’

b. Diphthongization of PRom /e/ in Cerignolese (province of Foggia, Zingarelli 1899: 84–5) /e/ seram ˈsεirə

*pullitram vir(i)dem --pəˈdːεtrə

‘evening’ ‘filly’

*moll+iscam

implere

illam

ˈvεrdə

---

muˈdːεskə

ˈεnɟə

ˈεdːə

‘green.f’

---

‘soft.f’

‘fill.inf’

‘she’

c. Diphthongization of PRom /o/ in Vastese (province of Chieti, Rolin 1908) /o/ florem

supra

bursam

---

augustum *compti- verecundiam

ˈfjaurə

aˈsːæprə

ˈværdzə

---

aˈɦːæstə

ˈkændʒə

‘flower’

‘above’

‘bag’

---

‘August’

‘prepared’ ‘shame’

vriˈvæɲːə

15 These data are ordered in the sequence motivated in }3.2, (6) below. Note in passing that in these data, only the ’CVCV environment, displayed in the leftmost column, counts as open syllable, as highlighted by the divide, since even ‘muta cum liquida’ clusters—usually syllabified as V.CRV in the standard Romance languages (cf. e.g. the evidence from French and Tuscan in }3.2, (9))—always pattern with heterosyllabic ones in these southern Italian varieties. Cf. Loporcaro (1996; 2005b; 2008), where these and similar data are discussed, elaborating on an early remark by Merlo (1912: 2) on the Apulian dialect of Andria: ‘la tonica + muta + r trattata come in sillaba chiusa’ [‘stressed vowel + stop + r evolves as in closed syllable’]. This reconstruction has been criticized by some scholars: e.g. Zamboni, in Filipponio (2013: 88–95), as well as Mańczak (2012: 187), who argues that the southern Italian data in (11) must represent an innovation in syllabification restricted to that Romance area. What this objection neglects is the convergence—discussed at length in Loporcaro (2005b: 424–6, 2007b: 120–2, 2008: 340f.)—between the Romance outcomes in (11) and Late Latin evidence, from both metrical practice and metalinguistic judgements, which unanimously points to the heterosyllabication of muta cum liquida in the Latin of the late Empire period.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

29

All these dialects display a differentiation between open and closed syllable comparable with that observed in northern Gallo-Romance (cf. French toile ‘cloth’ < te¯lam vs vert ‘green’ < vĭridem, and }3.2, (9a)), Rhaeto-Romance (cf. e.g. Lower Engadinian [’sai ̯rɐ] ‘evening’ < se¯ram vs [sa’jεt :ɐ] ‘thunderbolt’ < sagĭttam, Pult 1897: 33–5), several northern Italian dialects (cf. e.g. Piedmontese in }5.1.2, (6)), or Dalmatian (cf. [’pai ̯ra] ‘pear’ < pĭram vs [fa’maʎa] ‘family’ < famĭliam, [’krau̯k] ‘cross’ < cru˘cem vs [’buka] ‘mouth’ < bu˘ccam; Windisch 1998: 918; Bernoth 2008: 2737).16 In other words, if the French, Rhaeto-Romance, or Dalmatian evidence is valid proof of a previous differentiation in VL between vowels occurring in a closed vs open syllable, so is the evidence in (11a–c). In addition, comparison of (10) and (11) shows that, for the purpose of assessing which modern Romance varieties support the assumption of a Proto-Romance—and hence originally shared—OSL, both types must be counted together as evidence in support, contrary to what Schürr does, ignoring both types of dataset. This is indeed what Lausberg (1976: 207–8) did too, with the same purpose of claiming that OSL was an innovation restricted to a central-northern area. Like Schürr, Lausberg puts together Daco- and Ibero-Romance with Occitan, Sardinian, and the western part of southern Italy, claiming that they all represent a more conservative stage, preserving ‘la indipendenza arcaica della quantità vocalica’ [‘archaic independence of VL’ (from syllable structure—M.L.)] (Lausberg 1976: 208), with respect to northern GalloRomance, Rhaeto-Romance, and northern Italian (including Tuscan). For Sardinia and southern Italy, this has already been shown to be factually incorrect (note that both (10b) and (10c) stem from the western part of southern Italy, not from Adriatic varieties). For Occitan as well, the evidence to be considered in }3.4.2.2 suggests that OSL might have been at work in a previous stage. Thus, the modern evidence from all these languages converges with that from Late Latin reviewed in }2.1 to demonstrate that the grouping proposed by Schürr and Lausberg is spurious, and that the whole central-southern part of the Romance-speaking world (all southern Italian dialects, Sardinian, and possibly Occitan) has to be annexed to the OSL area. There is also compelling evidence that this isogloss stretched further south, encompassing the African ‘Romània submersa’ (see }2.3). Therefore, Schürr’s and Lausberg’s argument for pushing the rise of the OSL into the later history of Gallo-Romance and just a few other varieties of the central-northern Romània does not withstand closer inspection. Indeed, the whole diachronic reconstruction of the development of the Romance vowel systems to which the dismissal of an early Proto-Romance OSL was meant to be subservient does not stand on any firmer ground. Schürr’s assumption of generalized, metaphony-driven diphthongization of Latin stressed e˘ and o˘ has been proved to be

16 With the proviso that, unlike these other Romance varieties, consonant degemination did not apply in central–southern Italy.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

30

The Latin–Romance transition

misguided. As for diphthongization, A. Castellani (1970a; 1970b; 1970c) demonstrated the implausibility of assuming a pre-literary stage with just metaphonic diphthongs (e.g. buono vs *bòna) in the history of Florentine (pace Schürr 1965; 1972; Lausberg 1976: 207, 228, 230; Maiden 1995: 54–5). As for metaphony, Barbato (2009) and Loporcaro (2013: 125–6; 2011b: 128–35) showed that the metaphonic dipththongization of e˘ and o˘ found today in Neapolitan and much of southern Italy (e.g. Neapolitan [’pje :r@] ‘foot.pl’ vs [’pε :r@] ‘foot’, [’mwort@] ‘dead.m’ vs [’mɔrt@] ‘dead.f’) is best interpreted as a secondary development, whereas in a first stage e˘/o˘-metaphony uniformly started out as raising (ε ɔ ! e o). It is still so in Logudorese Sardinian (e.g. [’bentu] ‘wind’ vs [’bεntɔs] ‘wind.pl’, [’ortu] ‘kaleyard’ vs [’ɔrtɔs] ‘kaleyard.pl’), in most dialects of central Italy (except Tuscan), in some varieties of southern Italy and in Portuguese (n[o]vu ‘new’ vs n[ɔ]va ‘new.fsg’), and there is compelling evidence that it must have been so in pre-literary Neapolitan as well (cf. Barbato 2009: 283). In sum, Schürr’s arguments against a Late Latin/Proto-Romance OSL can be dismissed as inconclusive. The same is true of the alleged evidence from loanwords invoked earlier, to the same end, by Meyer-Lübke (1890: 524; 1920: 142). He proposed the 6th century as terminus a quo for the rise of the OSL, on the evidence of its application to Late Latin loans into Germanic and of the application of diphthongization to Frankish loans into Old French. The argument is even weaker here. On the one hand, the dating of Latin borrowings such as Old High German scuola ‘school’, fiebar ‘fever’ is disputed: according to Mackel (1896), followed by Meyer-Lübke, they date from a period between the 6th and 9th centuries, whereas in EWA 3.204 the borrowing of Lat. febris (feminine) into West Germanic is regarded as older, on the strength of the gender change uniformly attested by the WGerm. outcomes (*fēbr masculine/neuter). Moreover, as pointed out by Brüch (1921: 574), whatever the dating of such loans into Germanic, it is quite possible that their stressed vowels underwent lengthening (*schōla, *fēbre) already in the donor language, (Late) Latin. 2.2.2 OSL in Republican Latin? Having excluded a late dating of the OSL (which would have confined it to only some branches of the Romance family), we are left with the thorny problem of determining when it arose in Latin. We saw in passing in }2.1 that for some Romance scholars (e.g. Grandgent 1907: 76), OSL must have been operating throughout the Empire by the late 4th/early 5th century ad. Others have dated the rise of OSL earlier. For Straka (1953: 268), the lengthening must have been already at work by the 3rd century ad, a conclusion which is forced by the high chronology he assumes for the diphthongization of Lat. short e˘ and o˘ (mid-3rd century for e˘ > je, slightly later, by the early 4th century for o˘ > wo). In placing diphthongization so early, Straka follows

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

31

Richter (1934: 155f.),17 but the arguments invoked were not solid, as they rest on erroneous readings of the evidence, metalinguistic and epigraphic. Richter (1934: 138) takes the following passage by Servius as evidence for diphthongization: (12)

uocales sunt quinque, a e i o u. ex his duae, e et o, aliter sonant productae, aliter correptae. nam o productum quando est, ore sublato vox sonat, ut Roma; quando correptum, de labris vox exprimitur, ut rosa. item e quando producitur vicinum est ad sonum i litterae, ut meta; quando autem correptum, vicinum est ad sonum diphthongi, ut equus [‘Vowels are five, a e i o u. Out of these, two, viz. e and o, sound differently when long vs short. In fact, when o is long, the voice resounds with raised tongue (lit. ‘mouth’),18 as in Roma; when short, the voice is emitted from the lips, as in rosa.19 Likewise, when e is long, it is close to the sound of the letter i, while when it is short, it is close to the sound of the (ae) diphthong, such as in equus’] (Servius, in Donat. (about 400 ad), GL 4.421.16–21)

However, as agreed on by e.g. Battisti (1946: 64), Fouché (1958: 194), von Wartburg (1950: 81), Spore (1972: 269), Franceschi (1976: 263 n.), M. Mancini (1994: 617), and Filipponio (2006: 52), sonum diphthongi here does not mean a phonetic diphthong from short stressed e˘ in equus ‘horse’ (a recurring misinterpretation, found e.g. in Goidànich 1906: 41; Wright 1982: 59–60), but rather refers to [ε :], i.e. the sound of the monophthongized CL diphthong (/ai ̯/ > /ae ̯/ > /ε :/) still written , as in aequus ‘right’. This passage does not therefore constitute evidence for early diphthongization, but rather (much like other similar ones: cf. (13) below) for a phonetic difference in tenseness between long and short vowels. Similarly, of the three alleged early examples of Romance diphthongization in Latin inscriptions on which Richter (1934), Straka (1953: 264), Schürr (1970: 6) etc. base their claim of an early diphthongization, two have been shown to be nonexistent: see Herman (1970: 30), who discusses dieo for deo CIL VIII 9181 (emended to deeo in the volume’s supplement, VIII 20820), and vobit (for obiit ‘died’ from a Christian African inscription, dated 419 ad, no. 3436 in Renier 1855–8: 409), the product of incorrect word division of a string that must have read in fact as vixit beniv obit ‘lived two years (and then) died’.20 An even earlier date than Straka’s is assumed for OSL by Pulgram (1975), Vineis (1984b; 1993: xlii–v), and Giannini and Marotta (1989: 272). According to Pulgram He is followed by several others: e.g. Weinrich (1958: 40, n. 53), Schürr (1970: 6). I am following Filipponio (2006: 45). A different translation (‘in der weiten mund höhle’ [‘in the enlarged oral cavity’]) in Seelmann (1885: 208). 19 The exact meaning of the phonetic description of the contrast between short and long o has been the subject of much discussion. Whatever the literal meaning, it is possible that what is intended is a contrast in height/tenseness, parallel to front vowels. 20 The only non-illusory example of spelling instead of expected thus remains niepos for Nepos in CIL XV 1118b (from Rome, ad 120). 17 18

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

32

The Latin–Romance transition

(1975: 251), spoken Latin by the 3rd century bc ‘lacked prosodemic vocalic quantity’. This amounts to upsetting the standard account, as summarized in Chapter 1 above, according to which the phonological system of Latin has five vowel qualities, contrastively long vs short. There can be little doubt that, at a certain point, a distinction in quality between long and short vowels must have arisen, as reported, specifically for mid vowels, by several sources from late antiquity. This is illustrated by the following passage (and the very similar one in (12) above): (13)

uocales sunt quinque. hae non omnes uarios habent sonos, sed tantum duae, e et o. nam quando e correptum est, sic sonat, quasi diphthongus, equus; quando productum est, sic sonat, quasi i, ut demens. Similiter et o quando longa est, intra palatum sonat, Roma orator; quando brevis est, primis labris exprimitur, opus rosa’ [‘Vowels are five, of which not all but just two have different sounds, viz. e and o. In fact when e is short, it sounds almost like the diphthong, equus; when it is long, it sounds almost like i, as in demens. Likewise, also o when it is long, resonates within the palate, as in Roma orator; when it is short, it is articulated with the tip of the lips,21 as in opus rosa’] (Sergius, Explanat. in Donatum, GL 4.520.27–31)

This has often been taken as evidence that contrastively long vs short vowels came to differ phonetically, at some point in the history of Latin, in a way that presaged the later Romance development, as shown in (14) and (15) (where three Romance languages illustrate the most widespread development, the so-called common Romance vowel system):22 (14)

Classical Latin vowel system

/ iː [ iː filum ‘thread’

(15) iː

i i pira ‘pear’

e ε venio ‘come.1sg’

a a male ‘badly’

aː aː amatus ‘loved’

o ɔ porcus ‘pig’

oː oː vox ‘voice’

u ʊ ursus ‘bear’

uː / uː] durus ‘hard’

‘Common Romance’ vowel system i

i filo hilo fil

e eː stella ‘star’

eː

e ε

a

ɔ

stella estrella étoile

ferro hierro fer

male mal mal

porco puerco porc

e pera pera poire

a

o

oː

u o

voce voz voix

uː u

orso oso ours

duro duro dur

Italian Spanish French

See the translation in Seelmann (1885: 208). The allophonic realizations of long vs short vowel phonemes assumed in (14)—as has long been recognized in studies of the Latin–Romance transition (cf. e.g. Straka 1959: 174)—are in keeping with evidence from studies in experimental phonetics (cf. e.g. Lehiste 1970: 30–33; Hillenbrand 2003: 122–5), showing that, ceteris paribus, long vowels tend to be more peripheral than short ones. 21 22

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

33

This is not accepted unanimously: for instance, Mignot (1975 : 208) states that a difference in quality (tenseness) is conceivable but cannot be demonstrated with certainty, while Weiss (2009: 64) denies any such difference for CL, bolstering this rejection with the observation that the vowel system of Sardinian (see (25) below) does not display any difference in quality between the outcomes of long vs short vowels. However, as will be shown in }2.4, the Sardinian evidence need not lead to such a conclusion, and can be reconciled with the overwhelming evidence in support of a differentiation in quality of long vs short vowels. Even if this differentiation is granted, there is still a matter of dispute as to the point when long vs short vowels, and exactly which ones among them—whether only mid or also high vowels— differed by quality: some have restricted the contrasts either chronologically or structurally, or both. Thus, some scholars believe that the contrast may have concerned all non-low vowels throughout the history of Latin (cf. e.g. Grandgent 1907: 75; Meillet 1909–10: cccxxi; Straka 1959: 178; Tekavčić 1972: 1.17, 21; A. Castellani 1991: 11; M. Mancini 1994: 617; Sihler 1995: 72). Others have claimed that this situation must have come into being by the 1st century ad (Meyer-Lübke 1904–6: 467), and yet others have maintained that this difference in quality arose only later: no earlier than the 2nd century, according to Biville (1995: 27), or between the 2nd and the 4th century, according to Battisti (1946: 61), since it is in the 4th century that one begins to find unambiguous testimonies such as (13).23 As for the vowel qualities involved, that only mid vowels were affected by a differentiation by tenseness is maintained e.g. by Clackson and Horrocks (2007: 273), and the same is claimed, though with a chronological qualification, by Fouché (1958: 195–6), according to whom ĕ and o˘ became more lax in the 3rd century, ĭ and ŭ considerably later. However, both the (widespread) idea that passages such as (13) deliver the first clear evidence of the differentiation in quality and the suggestion that, since these remarks concern e and o, the contrast must have been limited to mid vowels, are untenable, because the testimonies to a differentiation are not quite so late and indeed concern high vowels too. The Roman-born Aulus Gellius (c.123–after 169 ad) reports that even cultivated people had difficulties in telling the quantity of the stressed a in actus ‘activity’ (CL [’a :ktʊs]): Haec quosdam non sane indoctos uiros audio ita pronunciare, ut primam in his litteram corripiant [‘These verbs—i.e. actito, actitavi, derived from actus—I have heard some men, and those not without learning, pronounce with a shortening of the first syllable’] (Noctes Atticae 9.6).

On the contrary, no such difficulties were encountered—he goes on—in telling apart the stressed vowels of dĭctus ‘said’ vs scrīptus ‘written’, ‘a causa evidentemente 23 Two different qualities for long vs short e and o are reported also by Marius Victorinus (2nd half of the 4th c. ad, GL 6.33.3–8; cf. Seelmann 1885: 182).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

34

The Latin–Romance transition

del diverso colore di ĭ e ī; cfr. ital. detto e scritto’ [‘evidently owing to the different colour of ĭ and ī; cf. It. detto ‘said’ and scritto ‘written’’] as Bonfante (1956: 356, n. 2) aptly put it.24 Indeed, there is no other reasonable explanation, even if this need not imply that only colour made the difference for the competence of a Roman like Gellius, an implication that Bonfante endorsed but which can be refuted, as shown in }2.4 below. To conclude, direct evidence for a phonetic tense vs lax distinction, concomitant with CVL as shown in (14), is available for as early as the 2nd century ad. On this issue, Pulgram (1975) adopts an extreme position. Not only does he project back the quality differentiation in (14) onto the Latin of the Republican period, but he also claims that at that time, at least as early as the 3rd century bc, qualitative differences had fully replaced VL as far as phonemic contrastiveness is concerned. Consider Pulgram’s (1975: 251) scheme, partially reported in (16): (16) a. Proto Latin

iː

i

eː

e

aː a

o

oː

u

uː /

b. Written Latin

/

1

1

e

e

a

a

o

o

u

u

c. Spoken Latin A /

i

i

e

ɛ

a

a

ɔ

o

ʊ

u

[ d. Spoken Latin B /

iː

i

eː

ɛ

aː

a

ɔ

oː

ʊ

uː ]

i

i

e

ɛ

a

a

ɔ

o

ʊ

u

/

[

i

i

e

ɛ

a

a

ɔ

o

ʊ

u

]

/

‘Written Latin’, as the label says, mirrors—in writing and in the literary practice (verse) based on the written language—the original phonological system, including CVL, which Pulgram (1975: 251) ascribes to ‘prehistoric Latin’. By the time of the earliest documents, however, he takes the phonological system of Latin to have already shifted to what he labels ‘Spoken Latin A/B’ (whereby A vs B do not instance a chronological succession but rather, it seems, synchronic variation). In the former, ‘quality is prosodemic and quantity is an accompanying, predictable feature’ (Pulgram 1975: 252), whereas in Spoken Latin B ‘quantity has disappeared completely in both the phonemic and the phonetic statement. But it seems that in some regions open syllables tended to have a phonetically longer, closed syllables a shorter vowel’ (i.e. OSL occurred). These varieties of spoken Latin Pulgram (1975: 253) ‘date[s] as concurrent with the spread of Latin over alloglot regions outside of Latium, that is to say, they date back to a very early period of the history of the language rather than toward the end of the Roman Empire’.25 Pulgram does not declare the reasons that lead him to assume OSL 24 Note also that the case reported as problematic by Gellius involves a long vowel in a closed syllable (/’a :k.tus/), an environment in which shortening (eliminating superheavy syllables) is attested, in several cases, since Archaic Latin (see }1.2). 25 In a way, Pulgram is retelling an old story here: his theory is reminiscent of that argued for centuries ago by the Florentine humanist Leonardo Bruni, who, in a debate with Biondo Flavio at the court of Pope Eugene IV (1435), claimed that modern diglossia was already implanted in ancient Rome, where only the literate spoke Latin, whereas ordinary people spoke volgare, just as in the Renaissance (cf. Vitale 1953; Bruni 1984: 46–7; Tavoni 1984: 3–41, 197–221; 1992: 60–62; Charlet 1993: 242).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

35

for his ‘Spoken Latin B’, yet although it is so vague, his position requires discussion, since it was received favourably even by outstanding scholars (e.g. A. Castellani 1991: 14: ‘il Pulgram, nel fondamentale volume’ etc. [‘P., in his seminal volume’]) and is still credited until recently (cf. M. Mancini 2001: 238). I shall first address a general issue about quantity sensitivity, and then discuss some specific points about sound changes, which will suffice to illustrate that Pulgram’s theory is untenable. Pulgram claims that the sensitivity to VL of Latin metrics was due to a ‘resuscitating influence of a nascent concern with literature [ . . . ] in the third and second centuries B.C., in that crucial period when Latin came under the tutelage of Greek’ (Pulgram 1975: 256). This obviously cannot explain how Latin speakers of that time (however cultivated), being by hypothesis speakers of a language having long lost CVL, in order to imitate the Greek for verse practice might have come up with long vs short vowels precisely in the environments where these must be reconstructed to have occurred, on the strength of comparative evidence, in prehistoric Latin (and late PIE): say, in flōs ‘flower’ < *bhlohs-s vs po˘ tis ‘able’ < *póti-s, where the quantity contrast is confirmed by IE comparison, as is the case (cf. }1.1) e.g. for the o vs a outcomes in Germanic (cf. Goth. bloma ‘flower’ vs [bruþ-]faþs ‘bridegroom’, Meiser 2010: 56–7). The only reasonable explanation is direct inheritance, and it is clear that the role of Greek models is more sensibly appreciated by scholars such as Herman (1998: 9): ‘la facilité même avec laquelle la métrique classique fut “importée” de Grèce, employée avec maîtrise déjà par Plaute, témoigne de la présence d’un paramètre quantitatif parfaitement en vigueur dans la langue de tout le monde’ [‘the very ease with which classical metrics was “imported” from Greece, employed with mastery already by Plautus, witnesses to the presence of a quantitative parameter which was in full operation in everybody’s speech’]. True, Pulgram was careful to put up hedges around his claims, as exemplified in his insisting on the fact that ‘Written’ (Latin) does not indeed ‘imply that this kind of dialect was never spoken: but it must have been restricted to certain formal or solemn occasions’ (Pulgram 1965: 146–7; cf. also Pulgram 1950). However, it is hard to see how the register used orally on such ‘formal or solemn occasions’ could account for sound change. In order to investigate change, one has to look at everyday spoken language, and if everyday spoken Latin was by Plautus’ time as depicted in (16c,d), it becomes unclear how, in non-quantitative spoken Latin A/B, one could accommodate changes such as iambic shortening. This, on the standard view (see }1.2), applied before 200 bc, and requires a statement of its environment in terms of quantities which is hard to rephrase if 3rd-century spoken Latin indeed had either of the vowel systems in (16c,d). Presumably, one should restate it in terms of vowel harmonization (through changes such as *[’bεne( :)] > [’bεnε], *[’mɔdo( :)] > [’mɔdɔ]) involving the feature specification [tense]/ [ATR]. This would not be unheard of: cf. e.g. the vowel harmony by tenseness attested in many Bantu languages, such as Mongo-Nkundo (Guthrie C61), which has a seven-vowel system with alternations such as [-ét-el-] ‘call-applicative’ vs [-kɔt-εl-]

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

36

The Latin–Romance transition

‘cut-applicative’, where the vowel in the applicative suffix harmonizes by the ATR feature with the root vowel (cf. Hyman 1999: 241). Note, however, that this would not work for /a:/ > /a/ (e.g. in pŭtā > pŭtă), since low long vs short vowels were never differentiated in quality, nor for diphthongs, whose patterning with long vowels is only natural under the standard view (cf. e.g. Odden 2011b: 467–8 for cross-linguistic evidence), but would become a mystery given a non-quantitative system.26 The same argument now expounded for iambic shortening can be repeated for virtually any other quantity-sensitive change known to have applied in Republican Latin: e.g. the relevant environment for preclassical degemination in cāssus ‘fall’ > cāsus, or caussa ‘cause’ > causa etc. (cf. }1.2) could not be stated sensibly given (16). Not only does the general architecture of Pulgram’s theory leave much to be desired, but it is also clear that, when it comes to Romance, his speculations are worked out taking liberties with the actual data, as shown eloquently by his account of change in the stress system. Having sacrificed Latin CVL, he must claim that stress was already ‘prosodemic’ (i.e. lexicalized) by the 3rd century bc. To bolster this claim, he gathers a series of Latin ‘accentual doublets’, which are meant to show that the CL stress rule standardly assumed (see Ch. 1, (4)) was no longer operating, and that lexicalization of stress resulted in systematic vacillation of stress assignment between the penult and the antepenult. These doublets are drawn on the one hand from the well-known series of words with muta cum liquida clusters between the penult and the last syllable, exemplified in (17a) by palpebra and tenebras, and on the other hand from a series of further putative examples, some of which are listed in (17b) reproducing exactly the form in which they appear in Pulgram (1975: 168–9), including stress marks on Latin, Romance, and German forms: (17)

a. Italian pálpebra Portuguese trévas

pa´lpebra te´nebras

French paupière Spanish tiniéblas

‘eye-lid’ ‘darkness’

b. Sicilian (a)síd.d.a27 Spanish centella Italian finèstra Italian giovénco Italian maéstro Italian tavèrna

axı´lla scintı´lla fene´stra iuue´ncum magı´strum tabe´rna

German Áchsel Romanian scî´nteie German Fénster Romanian junc French maî´tre German Zábern

‘armpit’ (Sic. ‘wing’) ‘sparkle’ ‘window’ ‘bullock’ ‘master, teacher’ ‘tavern/place name’

26 This is not enough of an obstacle for Pulgram (1975: 252), who claims that ‘nothing prevents one from believing that the two [i.e. long vs short /a(:)/—M.L.] were slightly different through Spoken Latin B, with [a:] being higher and fronter than [a]’. However, this claim about ‘Spoken Latin B’ is at odds with historically attested spoken Latin: the evidence preventing such speculations is provided by Aulus Gellius’ passage quoted earlier in this section (Noctes Atticae 9.6), where Gellius reports that the two as were easily confused in closed syllables, while this was not the case for long vs short /i(:)/, which were neatly kept distinct. This would not have been the case if long and short /a/s had not shared exactly the same quality. Nor is there any evidence for such a contrast from Romance, where long and short /a(:)/ merge without residue. 27 Sic: actually Sic. scíd . d.a ‘armpit’ (Traina 1868: 887).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

37

The derivatives on the right-hand side, it is claimed, ‘do not [ . . . ] go back to the accentuation of the central column’ (Pulgram 1975: 168), and therefore provide evidence for the lexicalization of stress and its not obeying the penult stress rule currently assumed (see Ch. 1, (4)). This claim, in turn, requires a long discussion of the shift from the prehistoric initial stress to the historical stress rule (Pulgram 1975: 92–122), which need not concern us here, since it is easy to show that the evidence in (17) is inconclusive. First, stress reassignment before muta cum liquida is a Late Latin phenomenon, which arose as a consequence of the heterosyllabication of those clusters in late antiquity (cf. Loporcaro 2007b: 121–2), still observable to this day in southern Italy (cf. (11) and n. 15 above). Therefore, the stress shift reflected in Fr. paupière is irrelevant to the issue of determining whether stress doublets were already around in 3rd-century bc Latin, as are German loans in general (as Pulgram halfheartedly admits elsewhere (p. 172) but only for Fenster and Zabern), because they are later and above all because their initial stress complies with the Germanic stress rule.28 Most of the remaining Romance examples suffer from the same problem, as their stress shift is simply due to regular sound change. Romanian junc, where an intervocalic consonant has been deleted, displays vowel coalescence (Nandriș 1963: 21), as does Fr. maître: the stress shift arose in OFr. máistre after deletion of -g-, whereas Pulgram’s insertion of this example (like others) in the right-hand column wrongly implies that this form is incompatible with CL stress as computed by rule (4) in }1.1 in magı´strum. Of course, if one disregards the well-established relative chronology of the sound changes involved, one may come up with an exponential growth of (putative) examples of violations of the Latin stress rule. In Genoese, for instance, [’maju] ‘husband’ or [’pai ̯ze] ‘village’, by Pulgram’s standards, would point to *ma´rı¯tum, *pa´gensem, which is of course not the case, since intervocalic consonant deletion in Genoese (cf. Toso 1997b: 34) applied prior to the stress shift which took place later in the newly created diphthongs. Last but not least, in the case of the Romanian word for ‘sparkle’ in (17b), the argument is built on a factually wrong form. Actually, Romanian scânteie (in modern orthography, or scînteie, sch- in older ones) has penultimate stress [skn’tεjε] (the circumflex is not a stress, but a vowel quality mark) and thus should not occur in the third column in (17) in the first place. Indeed, Pulgram forces this word into the argument’s mould despite the evidence (from sound laws and etymology), since Romanian sound laws make it clear that scânteie cannot be a regular outcome of Latin scintilla. In fact, (-)sce/i- gives [ʃt(e/i)], as shown by e.g. crescis > crești ‘grow.2sg’, scis > ştii ‘know.2sg’, piscem > peşte ‘fish’ (cf. Sala 1970: 139), so that a

28 Indeed, in studying what Germanic loans can tell us about Latin–Romance phonology, scholars working with method are careful to select just those words ‘die den Akzent auf dergleichen Silbe belassen haben, also solche mit Anfangsbetonung’ [‘which have maintainted stress on the same syllable, i.e. those with initial stress’ (i.e. those which had initial stress already in Latin—M.L.)] (Lüdtke 1956: 132–3).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

38

The Latin–Romance transition

perturbation (e.g. a cross with *excandere, Cioranescu 1958–66: 734; Nandriș 1963: 21) must be assumed, as well as a change of suffix (*(ex)scantilia) since -llwas deleted in Romanian (e.g. stellam > stea ‘star’, sellam > șa ‘saddle’), with some exceptions such as ollam > oală ‘pot’, but never changed to [j] which is the regular outcome of -l(l)i ̯- instead: mulierem > muiere ‘woman’, palliolam > păioară ‘veil (used to cover dead people’s face)’ (cf. Sala 1970: 88–9). In sum, it would be unwise to take too seriously claims about the operation of OSL as early as the 3rd century bc proposed within the context of a work that takes such liberties with the data it handles.29 With Vineis (1984b: 49; 1993: xlii–v), by contrast—as well as with Giannini and Marotta (1989: 272) and Benedetti and Marotta (2014: 31–3) who follow him—we are back in the province of reliable scholarship. Vineis too places OSL in the late (or even mid-) 3rd century bc, at least for a basilectal variety of Latin which had already given up CVL, replacing it through a vowel quality (tenseness) contrast. Vineis’ idea capitalizes on the further Romance development (15), whose harbingers, as far as vowel quality is concerned, were observed by Latin grammarians ((12), (13)) and which is squeezed into the verbal repertoire of the 3rd century bc. Thus, he assumes that ‘nei registri più bassi’ [‘in the lowest registers’] (Vineis 1993: xliv) the following vowel system must have obtained: (18)

i

e

ɛ

a

a

ɔ

o

ʊ

u

[

iː iː

eː

ɛː

aː

aː

ɔː

oː

ʊː

uː ]

[

i

e

ɛ

a

a

ɔ

o

ʊ

u

/ open syllable closed syllable

i

i

/ ]

This parallels Pulgram’s Spoken Latin B (16d), especially as far as OSL is concerned: ‘la quantità, non più pertinente, tende a disporsi in distribuzione comple in sillaba aperta mentare in dipendenza dalla struttura sillabica, secondo il modulo V e V̆ in sillaba chiusa’ [‘VL, no longer phonemic, tends to settle into a complementary  in open vs V̆ distribution depending on syllable structure, according to the pattern V in closed syllables’] (Vineis 1984b: 49). The evidence Vineis adduces for the coexistence of this low variety with the high one, which in his view still preserved CVL, is twofold: on the one hand, Romance outcomes which deviate from the correspondences in (15); on the other, some deviations from the orthographic norm which are found in the epigraphic evidence as early as the 3rd century bc. The Romance outcomes considered are examples like It. fèsta, Sp. fiesta, which go back to *fe˘sta REW 3267 rather than to CL fēsta (cf. fēriae, whose ē is ensured by metrical evidence as well as from borrowing into Germanic, which keeps apart Latin ē vs ĕ: cf. OHG fîra

29 Indeed, Pulgram’s speculations about Latin stress and VL are mentioned dismissively in a solid reference work such as Leumann (1977: 56): ‘umgekehrt wieder Pulgram’ (1965). The reason why I had nonetheless to elaborate on this, as noted introductorily, is that Pulgram’s reconstruction is still taken seriously in the early third millennium in some of the literature on the loss of Latin CVL (see }2.3.1 below).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Why OSL must be Proto-Romance

39

‘feast’ EWA 3.304–5 vs fieber ‘fever’, from Lat. fe˘bris). This is explained as the emergence of a ‘low’ pronunciation, with the stressed vowel shortened in a closed syllable, and the same explanation is provided for a series of similar instances, which all reconstruct with a short stressed vowel in Proto-Romance as opposed to a long vowel in CL (e.g. It. fréddo ‘cold’ REW 3512.2 vs CL frı¯gidus, férmo ‘steady’ REW 3320 vs CL fı¯rmus, lètto ‘bed’ REW 4965 vs CL le¯ctus). The argument goes as follows: given a long tense vowel in a closed syllable, the popular system recategorized this vowel as allophonically short because of the non-application of OSL. But then, by hypothesis, this shortness induced laxing, resulting in *fI rmus, *lεktus, and the like. The traditional explanation, which treats these forms as isolated exceptions, has the advantage of also covering deviations in the opposite direction such as It. tórno ‘return.1sg’ (Fr. tourne, Sp. torno, not **tuerno), or It. fórse ‘maybe’. These Romance outcomes are unexpected under Vineis’ theory: they are not compatible with CL to˘ rnō ‘make round’, fo˘ rsit, but rather presuppose, on the traditional view, *tōrnō, *fōrsit.30 In principle, the Italian, French, and Spanish forms would be compatible with a change in quality rather than quantity, of the kind hypothesized by Adams (1995: 92) for turtas, a variant of torta ‘a type of twisted loaf ’ (a word ˙ occurring several times in the Vulgate, from Lat. torqueo ‘twist’), attested in the Vindolanda writing tablets (180.20). Adams assumes here vowel raising induced by the following coda rhotic (to˘ rta > tŭrta > tórta REW 8802, cf. Fr. tourte, It. Sp. Pg. Occ. torta, with [o] rather than *[ɔ] in Italian and rather than *[we] In Spanish, etc.). However, this would be no viable alternative in the case of *tōrnō, *fōrsit, because the Sardinian outcomes, given the diachronic correspondences in (25), indeed point to *o¯ and exclude *u˘: cf. Log. [’tɔr:ɔ] ‘return.1sg’ (with regular -rn-> [r:]), as opposed to [’tur:ε] ‘tower’ < tu˘rrem, and [’fortsizi] ‘maybe’ (not **[’furtsizi]), with ‘adverbial’ -s. As for inscriptions, Vineis (1984b: 49) argues that ‘una grafia come tempestatebus (accanto al “normale” tempestatibus conservatoci dalla tradizione colta) può testimoniare la discesa timbrica di ĭ conseguente alla sua pronuncia rilassata’ [‘a spelling like tempestatebus (alongside “normal” tempestatibus preserved by learned tradition) may attest to the lowering of ĭ ensuing from its laxened pronunciation’]. Since this tempestatebus is carved on the grave of one of Rome’s rulers (L. Cornelius Scipio, consul in 259 bc, CIL I2 2, 9), it is itself an integral part of the ‘learned tradition’ and hardly represents a low variety (pace Vineis, who invokes it to illustrate ‘la compresenza—e l’interferenza—di queste due strutture fonologiche già assai per tempo in epoca repubblicana’ [‘the coexistence—and interference—of these two phonological structures [i.e. CL (14) and the ‘low’ variety (18)—M.L.] already very early on in the Republican age’]. It simply has no bearing on the hypothesis of an Other ad hoc explanations have been proposed: thus, e.g. Màfera (1958: 149) assumes that [o] in [’forse] in the dialects of Veneto is due to raising in sentence-internal position. Of course, such a hypothesis cannot extend to tórno and similar cases. 30

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

40

The Latin–Romance transition

early OSL, like the many other instances of similar graphical exchanges to be discussed in }2.4, but shows that the realization as [I] (and perhaps, variably, already [e]) of short ĭ) had been around, in all registers of Latin, from very early on. In sum, while the existence of a diastratic differentiation is in itself plausible (for Latin, as for any language spoken in a sociolinguistically complex community), there is no evidence that this differentiation might have involved an OSL in lower registers as early as the 3rd century bc, nor that such an OSL might have ousted CVL completely in spoken Latin by that date. Speculations of this kind, in whatever flavour they may come, can be confidently dismissed.

2.3 The rise of OSL and the regional diversification of Latin Reference to African Latin in }2.1 has already introduced the topic of the regional diversification of Latin. This is, for a change, a thorny and much-debated issue; but it has to be addressed here, since from the evidence for regional diversification one can arguably gain more solid arguments for the chronology of the rise of OSL and the demise of CVL than the purely speculative ones refuted in }2.2. In his monumental study on the regional differentiation of Latin, Adams (2007: 725) concludes: ‘Regional variety, albeit difficult to identify because of the paucity of the evidence, is there from the time of almost the earliest records, though its patterning changed because of historical events’. A more specific question is whether this differentiation can be grasped through comparative study of Latin inscriptions from different regions of the Empire: again, ‘the answer is a guarded yes’ for Adams (p. 676). This is in line with Herman’s (2000: 133) response to the same question: ‘sì, differenze sistematiche nel parlato delle diverse regioni esistevano, ed è possibile provarlo’ [‘yes, systematic differences in speech among different regions [of Italy, during the Empire—M.L.] did exist, and it is possible to prove it’] through the study of different error rates in inscriptions. In several studies since the 1960s, Herman has consolidated a method which has brought about a quantum leap in research in this area. The method, in short, requires that one does not simply count deviations from the classical norm which occur in a given epigraphic text, or in a set of texts from a given area, because absolute frequency of errors is an index of extralinguistic facts: ‘il est évident que la fréquence d’une “faute” dans un ensemble de textes ou un texte donné ne reflète directement que la “force de résistance” de la tradition orthographique, donc un facteur culturel, extralinguistique’ [‘it is evident that the frequency of a “mistake” within a set of texts or in one given text does not directly reflect anything but the “force of resistance” of orthographic tradition, hence a cultural, extralinguistic, factor’] (Herman 1965: 20). Rather, one must count relative frequencies comparing deviations in different areas of the linguistic system, because ‘une fois [ . . . ] que le rédacteur ou le graveur s’écartait de la norme, il n’y avait pour lui aucune raison extralinguistique de s’écarter

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rise of OSL

41

des règles nettement plus souvent sur un point que sur un autre’ [‘once the writer and the engraver had departed from the norm, there was no extralinguistic reason for them to depart from rules distinctly more often on one point than on another’] (Herman 1965: 18).31 While several ancient sources report the existence of regional accents across the Roman Empire (see }2.3.3 below on African accent), only thanks to this method could these be pinned down, providing evidence—in our specific case—that retaining or losing CVL might have played a role in this regional diversification and, even more to the point, that the rise of OSL might have been involved in this loss. In the following sections (}}2.3.1 and 2.3.2), I will review two studies on African Latin: while much has been written on the topic, and especially on whether this variety of Latin displayed OSL (see e.g. Bonfante 1956; Acquati 1971; M. Mancini 2001), these studies stand out for their innovatory force, their systematic approach, and their neat results. 2.3.1 Evidence from metrical inscriptions, part 1: Herman (1982) Herman (1982) put his method to use in order to check whether evidence from African metrical inscriptions could confirm Consentius’ and Augustine’s testimonies (see (2)–(4) above). He analysed a corpus of 279 inscriptions dating from between the 1st and the mid-4th centuries ad, and compared this with two control corpora from Rome, one contemporary and one later, in order to ascertain similarities and dissimilarities in the patterns of deviation from the Classical Latin norm. Given that many of the ‘metrical’ inscriptions considered indeed follow the dactylic pattern only loosely, deserving to be called ‘quasi-verse’, Herman (1982: 223f.) is careful to take into account the two final feet on a hexameter or the second hemistych’s end in a pentameter, as well as the first verse foot ‘dans les cas où on observe un effort systématique tendant à donner à ce pied un caractère régulier’ [‘in cases where one observes a systematic effort aiming to grant regularity to this foot’] (p. 224). As he himself acknowledges, this rests on subjective judgement: yet the procedure adopted was the same for the three subcorpora, ensuring comparability.32

More recently, Adams (2007: 629–30) discusses at length Herman’s method for evaluating what inscriptions can contribute to the knowledge of pre-Romance changes that might have been ongoing in local varieties of spoken Latin. Adams comes up with some amendments. In particular, he recommends that, in a first count, the number of errors for a given feature be averaged against the total number of correct occurrences. Only in a second step, then, should ratios (error/correct) concerning distinct features be compared. This is surely an improvement, but does not detract from the importance of Herman’s ground-breaking work. 32 Though subjective, this method rests on established practice in the study of Latin popular verse, from Galletier (1922: 302) up to Adams (1999: 113f.), who applied the same method to the Bu Njem poems (see below, }2.3.2), observing ‘recognizable snatches of hexameter rhythm’ and then proceeding to an ‘examination of the last two feet of the verses’. 31

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

42

The Latin–Romance transition The results are summarized in (19a–c):

(19)

Errors on stressed vowels a. Africa (1st–early 4th c.) b. Rome (1st–early 4th c.) c. Rome (late 4th-6th c.)

total 28 7 16

% 27 8.6 29

Errors on stressed vowels consist, overwhelmingly, of the use of a short vowel in open syllables which should be heavy for the metre: e.g. in tĭtulo clārum ‘bright in the inscription’ CIL VIII 9080, vĕtula dormit ‘the old woman sleeps’ CIL VIII 20277. Herman’s method also implies that the observed figures are contrasted with those that would be expected given a random distribution (i.e. if the variable at issue—in this case stressed vs unstressed—were uninfluential). Since in Latin texts the ratio of stressed to unstressed vowels is about 1:3 (i.e. 37.2% vs 62.8%, Herman 1968: 197, n. 5), the error rate in Africa ((19a)) closely approaches randomness, as opposed to the data from Rome in the same period, where the errors cluster on unstressed vowels, with 91.4%, i.e. with far more than chance frequency. This proves that the VL contrast, in Rome, was endangered in unstressed position but much better preserved under stress, a finding which is in keeping with the long-term trend revealed by the Archaic and Classical Latin evidence reviewed in }1.2. On the strength of these data, one must question the earlier procedure—common before Herman’s studies, but whose (unwarranted) results are still pretty much around in the relevant literature—of simply pointing to metrical errors in inscriptions in order to diagnose loss of CVL. Thus, for example, Bonfante (1968: 24) quotes the hexameter in (20), with the comment ‘l’ablativo forma finisce in una breve’ [‘the ablative forma ends in a short vowel’]: (NB: feet are separated by |. Length vs shortness signs do not indicate syllable weight but vowel length according to the CL prosodic norm, so as to make metrical errors evident.) (20)

hīc ĕgo˘ | nūnc fŭtŭ|ē fōr|mōsā(m) | fōrmā pŭ|ēllă(m) CIL IV 1516 ‘now, here I have fucked (with -ē < -ei, = CL futuī) a beautiful girl’

This is used as an argument to conclude that VL was, as such (i.e. in all environments), no longer contrastive in the writer’s competence. However, this need not have been the case, as Herman’s results show through the skewed distribution of errors on stressed vs unstressed vowels in (19b), with the latter by far outnumbering the former.33 The fact 33 Which is not, of course, to say that deviations on stressed vowels, with OSL of short vowels, did not occur at all in Pompeii. That they did is shown e.g. by Ve¯nerem and v¯bi in the iambic senarii (ll. 2 and 5) in CIL IV 5092 (Väänänen 1966: 19; Leumann 1977: 56):

(i) magi(s) | propera|res, ut uide|res Ve|nerem ‘you would hurry up even more in order to see Venus’ Pompei|os de|fer, u|bi dul|cis est | amor. ‘take (me) to Pompeii, where (my) sweetheart is’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rise of OSL

43

that no such imbalance is observed in Africa suggests that in the time-span covered by the examples (19a,b), the prosody of African differed from that of urban Latin in that distinctive VL had been eliminated altogether in the former, not yet in the latter.34 On the other hand, (19c) shows that after the mid-4th century, the situation in Rome changes radically, as the urbs too reaches ‘African’ figures, with confusions evenly distributed over stressed and unstressed vowels. Herman also addresses the related issue of whether this epigraphic evidence supports the hypothesis that the merger of CVL co-occurred with the rise of OSL. As said above, errors on stressed vowels mostly involve erroneous occurrence of a short vowel where a long one would be required, as in the quoted hexameter clausulas vĕtula dormit CIL VIII 20277, in tĭtulo clārum CIL VIII 9080, or in ămor adēgit ‘love led’ CIL VIII 20808. Symmetrically, errors on unstressed vowels involve the replacement of long vowels with short ones with more-than-chance frequency: 68% in Africa, 47% in Rome, as against an expected random distribution of about 20%, the ratio of long to short vowels in unstressed position being 1:4 (cf. Herman 1968: 199). In sum, the results of the analysis of metrical evidence correspond exactly to Consentius’ and Augustine’s description in (2)–(4), and thus support the hypothesis of an early rise of OSL in the Latin of Africa and of a delay in its spread to the rest of the Empire: in Rome, OSL must have been operating by the late 4th century. Herman’s reconstruction was not met with unanimous agreement. One problem, of which Herman himself was well aware, is statistical significance, since the study elaborates on very low absolute figures (see n. 35). Perhaps this was one of the reasons for Adams’ (2007: 667, n. 61) comment on Herman (1982): ‘I do not find the statistical distinctions made between the two corpora in this paper decisive.’35 While not lingering further on the statistics, Adams (1999: 117, n. 47) voices another reservation concerning the necessity to focus on the authorship of every single epigraphic poem, since a carmen written in Rome could well stem from a nonRoman author. This is a fully justified caution, to be sure, except that the fact remains that figures from the Roman subcorpora (19b,c) change in the 4th century, while there is no evidence that the set of authors involved may have been different in the

But crucially, they occur more seldom, as already noted by Galletier (1922: 302). Quantification is decisive here, and even if the figures tested negatively for statistical significance, given the database available no credible alternative to Herman’s (1982) procedure is in sight, if with the amendments suggested by Adams (2007: 629–30) (cf. n. 31 above). Without such quantifications, Bonfante (1956: 356, n. 2) equated Italy and Africa, contrasting both of them with Gaul, a conclusion which is no longer tenable. See also n. 59 below for further evidence drawn from the use of quantity diacritics, suggesting an earlier loss of CVL in Africa than elsewhere in the central and western provinces of the Empire. 35 In another passage, Adams (2007: 632) observes that the numbers of tokens involved in Herman’s studies are too small to be statistically meaningful. On the other hand, Adams himself (p. 632) does regard the evidence on change in intervocalic /w/ he gains from Bu Njem ostraca as reliable ‘though the number of tokens is small’. 34

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

44

The Latin–Romance transition

two periods. Be that as it may, Adams (1999: 117) quotes Herman’s results and concludes: ‘It is distinctly possible that the modification of the classical quantitative system progressed more quickly in Africa than in other parts of the Empire or was more obvious there to casual observers’. Note, finally, that his analysis of the Bu Njem poems (to be reviewed in }2.3.2) is in keeping with Herman’s results. Marco Mancini’s (2001) criticism is more radical. He, like Adams, does not specifically address Herman’s results about the imbalance in the distributions of errors over stressed vs unstressed vowels in Africa vs Rome in the 1st–early 4th century data (19a,b).36 On the other hand, he adds a series of arguments meant to demonstrate that in African Latin, as described by Consentius and Augustine, CVL was neutralized but, crucially, without OSL, as he concurs with Schürr’s (1970) claim that OSL arose only later (Mancini 2001: 321). One of these arguments, concerning the (later) passage by the grammarian Pompeius, has been discussed above (}2.1, (5)). Another argument against OSL is the stress-based metric of African poets like Commodianus. Note that this was judged by Bonfante (1956: 354, n. 2) as fully compatible with the assumption of OSL in African Latin: for Bonfante, the fact itself that African poets, within the Latin-speaking world, were first to switch to a new metric system fits in well with an earlier demise of CVL in Africa than elsewhere, however this exactly came about.37 Note further that Mancini (2001: 230) bases his own argument on Pulgram’s (1975) speculations about Republican spoken Latin—a shaky foundation indeed, as demonstrated in }2.2.2. In addition, it is somewhat surprising for such a connoisseur of the Latin grammatical tradition to elaborate on the speculations about the structural history of Latin contained in a book whose arguments rest not only on primary datatwisting of the kind exemplified in (17) or n. 26 above, but also on contempt for linguists who take grammarians’ testimonies seriously: It is strange to observe how a critical scholar who just as soon would not believe what he reads in today’s newspaper, is willing to swear upon the words of a palpably incompetent plagiarist who lived centuries ago and described the accents of a language he never heard spoken natively, and does so in terms of another language (i.e. Greek), of which he knows even less. (Pulgram 1975: 88)38 36 In doing this, in my opinion, Mancini (2001: 322) avoids the burden of proof incumbent on him, leaving a fundamental question unanswered: if Herman’s method is not sound, what else can explain the distributional contrast between Africa and Rome seen in the figures in (19b) vs (19a)? 37 A more radical opinion on the matter was issued by Meyer-Lübke (1920: 142): ‘doch folgt natürlich daraus, daß Commodian in der Mitte des 3. Jahrhunderts seine Hexameter nach dem Akzent baute, für die Frage nach der Quantität nichts’ [‘however, nothing follows, for the issue of (vowel) quantity, from the fact that Commodian, in the mid-3rd century, based his hexameters on stress’]. 38 The passage refers to stress, but reflects a general stance which Pulgram also adopts in his discussion of testimonies on CVL. This contempt for the evidence is to be compared with the more respectful (and appropriate) stance voiced by specialists in Late Latin such as B. Löfstedt (1983: 467) (who warns against underestimating ‘die Intelligenz und die Sachverständigkeit der antiken und mittelalterlichen Grammatikern’ [‘the intelligence and expertise of ancient and medieval grammarians’]), or Banniard (1992: 34): ‘rhéteurs et grammairiens [ . . . ], de Varron à Consentius, ont apporté une mine de renseignements non seulement sur les faits proprement philologiques mais aussi sur la perception qu’en avaient les

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rise of OSL

45

But even if Pulgram were right, it would be inconsistent to endorse at the same time Pulgram’s (p. 252) reconstruction, according to which OSL was already at work (in Spoken Latin B, at least ‘in some regions’) in the 3rd century bc, and Schürr’s (1970: 5–6), according to whom OSL is a later phenomenon of just some Romance languages which had not yet arisen in the 5th century ad. One last argument is adduced by Mancini (2001: 334, n. 10) to impugn the existence of OSL in African Latin: È decisiva, infine, la testimonianza—sfuggita a tutti—di voci latine trascritte in carattteri greci in frammenti papiracei di provenienza africana, nelle quali compaiono là ove, in sillaba tonica aperta, ci attenderemmo . . . : γνοβεϛ per nōbis (Kramer 1983: 82), ιστορουμ per istōrum (Kramer 1983: 102) [‘Finally, the testimony—which escaped everybody’s attention—of Latin forms transcribed in Greek characters on papyri fragments of African provenance is decisive, in which ’s occur where, in open stressed syllables, one would expect : γνοβεϛ for nōbis (Kramer 1983: 82), ιστορουμ for istōrum (Kramer 1983: 102)’].

Even though Kramer (1983: 15) classifies ‘γνοβεϛ statt nōbis’ [‘γνοβεϛ instead of nōbis’] among the reflexes in the glossaries of the demise of CVL in Latin, the argument is hardly convincing. Such forms would indeed constitute reliable evidence if they occurred in texts written by speakers of a quantity-sensitive language like Classical Greek. As a matter of fact, however, γνοβεϛ occurs in a Latin–Greek glossary from the 4th century ad and ιστορουμ in a Latin–Greek–Coptic book of prayer dated to the 5th–6th century, which seems to stem ‘von einem koptischen Schreiber’ [‘from a Coptic scribe’] (Kramer 1983: 83, 97). One has to ask what kind of Greek those texts represent, crucially, as far as CVL is concerned. The exact chronology of the loss of CVL in Greek is not easy to establish: Allen (1987: 94) places it in the 2nd–3rd centuries ad and seems to represent a majority view (see also Leumann 1977: 55: ‘etwa im 3. Jhdt. n. Chr.’, i.e. at the same time as in Latin, in his view; cf. n. 2 above). Also much earlier datings circulate in the literature: thus, based on the evidence from loanwords into Latin, Biville (1995: 29–32) assumes a loss of CVL in Greek ‘dès les débuts de la koinè hellénistique’ [‘since the beginnings of the Hellenistic koine’] in the 4th century bc. On the contrary, according to Sturtevant (1968: 47), a VL contrast might have persisted as late as the 4th century ad in the standard of Constantinople, since Greek loanwords into Ulfilas’ Gothic (mid-4th century) keep reflexes of Greek vs systematically distinct (see e.g. Jellinek 1926: 87–8, 122, 193). But even proponents of a late date for the merger agree that the Greek spoken in Hellenistic Egypt seems to have been at the vanguard of this change: ‘the mistakes in papyri become so common by the second century b.c. that we must infer identical quality [i.e. merger—M.L.] for Egyptian Greek’ (Sturtevant 1968: 47). professionnels de la parole et de la communication’ [‘rhetoricians and grammarians . . . , from Varro to Consentius, contributed a wealth of information not only on specifically linguistic facts but also on the perception thereof by speech and communication professionals’].

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

46

The Latin–Romance transition

Thus, it seems that Kramer’s texts date back to a period in which, even for Greek monolinguals in Athens or Constantinople, CVL had disappeared, whereas the chances are that for Hellenized Egyptians the demise had been even earlier. Consequently, for those speakers/writers the signs / denoted one and the same phoneme and, though still kept distinct graphically in writing Greek out of sheer orthographic tradition (just like, say, and in English homophones such as cue and queue), could be employed interchangeably to notate Latin vowel sounds of comparable quality, regardless of their length. In other words, such forms as γνοβεϛ and ιστορουμ occurring in Egyptian papyri from the 4th–6th centuries are not valid evidence for a putative lack of OSL in whatever variety of Latin at that time. 2.3.2 Evidence from metrical inscriptions, part 2: Adams (1999) Adams (1999) analysed the verses by two centurions at the Bu Njem military outpost (in Tripolitania) in the early 3rd century ad: Q. Avidius Quintianus, who wrote, or reviewed, a carmen in iambic senarii dated 202–3, and M. Porcius Iasucthan, whose carmen stems from early 222 ad. Both centurions—though with differences thoroughly analysed by Adams—had been trained in spelling but did not receive a full literary education, even if Avidius’ poem ‘displays traces of literary phraseology’ (Adams 1999: 125). Thus, their verses coincide in showing many deviations from the CL norm, substandard features in syntax and morphology, etc. But they differ crucially in one important respect. While Avidius’ senarii basically comply with CL prosody, Iasucthan’s verses read as hexameters only ‘on a charitable view’ (Adams 1999: 109), as exemplified in (21): (21) a. iām nūnc | cōntēn|dūnt fĭĕ|rī cĭto˘ | mīlĭtēs | ōmnēs (line 13) ‘now all the soldiers strive that it be quickly done’ b. căpĭtă | uērsō|rūm rĕlĕ|gēns ād|gnōscĕ cū|rāntĕm (line 33) ‘reading the start of the verses, identify him who saw to it’ By CL standards, mīlĭtēs (21a) and căpĭtă, cūrāntĕm in (21b) are placed wrongly, and in the whole poem only one line is entirely correct. Inspection of the last two feet in the verses, however, shows that they all scan ‘correctly’ under the assumption that short vowels are lengthened under stress (căpĭtă) and long unstressed vowels are shortened (mīlĭtēs, cūrāntĕm). An external explanation of the contrast between Iasucthan’s metrical chaos and Avidius’ correctness seems unlikely (Adams 1999: 117): if Avidius’ awareness of the difference between long and short vowels was entirely artificial, with no basis in his speech whatsoever, it is odd, given his relatively low cultural level as marked by the substandard syntax and morphology of the piece [ . . . ], that he was able to avoid mistakes almost entirely.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rise of OSL

47

The two were of different origin: Avidius probably came from central Italy (Rebuffat 1987: 102), whereas Iasucthan’s cognomen qualifies him as a Libyan. This regional difference is deemed relevant by Adams (1999: 117), who concludes that it is possible ‘to hear in the two poems two distinct dialects or accents of Latin, the one preserving distinctions of vowel quantity, the other based on a system which [ . . . ] was characterized by the loss of these distinctions’. Phonetically, on the evidence provided by deviations from the CL norm such as those exemplified in (21), this loss must have been realized as OSL, with complementary shortening of long unstressed vowels, a result which coincides with Herman’s.39 Iasucthan seems thus to lend a face to the vowel system which may be inferred from Herman’s African corpus. A potential problem impinging on this conclusion comes from Adams’ (1999: 110) contention that Iasucthan might have been a non-native speaker of Latin, which could cast doubt on the notion that this testimony bears on the native Latin of African provincials. In fact, as Clackson and Horrocks (2007: 260) put it, If Iasucthan was a non-native speaker of Latin [a thesis they subscribe to—M.L.], we cannot be sure whether his failure to hear Latin long vs short vowels was a result of his imperfect command of Latin phonology, or whether it actually reflects a more general breakdown in the vowel quantity system in the spoken Latin of Africa at this date.

The issue thus requires closer inspection. And the issue actually concerns not just Iasucthan’s Latin, but rather the Latin of Africa more generally. 2.3.3 Latin and Romance in Africa There can be little doubt that, by the early 3rd century ad, a native provincial variety of Latin was fully established in Africa. After the destruction of Carthage (146 bc), Romans gained control of Africa for over half a millennium, and Latin (and then Romance) was spoken there by generations of speakers for about fifteen centuries, as demonstrated by the fact that African-born Latin–Romance speakers continued to produce Latin inscriptions until the first half of the 11th century ad (some of the latest stem from Tripolitania, by the way: see Lancel 1981: 286–7), as well as by the evidence for a spoken Romance variety which developed locally out of Latin and persisted, in rural areas of Tunisia, as late as the last two decades of the 15th century, as attested by

Adams (1999: 116) hedges this finding with qualifications: ‘The impression should not be given that a vowel system in which, as Consentius noted, stressed vowels were lengthened and unstressed vowels shortened, was peculiar to Africa. Inscriptional verses from (e.g.) Rome, Spain, Gaul, and Moesia, as well as Africa, show signs of lengthening under the accent and shortening of unstressed vowels.’ Yet he himself recognizes that shortening of unstressed long vowels (involved in most of the examples he adduces) is a much older and more widespread phenomenon (cf. n. 33 above), and the results of his comparative study of the two centurions’ verses support Herman’s conclusion that Africa was at the vanguard in the demise of CVL through systematic application of OSL. 39

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

48

The Latin–Romance transition

the Roman humanist Paolo Pompilio (1455–91).40 We also know that this spoken variety, in the late 15th century, was perceived as similar to Sardinian, thus confirming spectacularly that the local developments of Latin in Africa and Sardinia displayed noteworthy parallelisms. The evidence comes from the second book, written in 1485 or slightly later, of Pompilio’s Notationes, where he recounts (ch. 6.1) that a Catalan merchant from Girona named Riaria had come recently from Tunis to Rome after spending thirty years in North Africa. Riaria, having visited villages in the mountainous region of Gebel Oresc (Aurès), told Pompilio: ubi pagani integra pene latinitate loquuntur et, ubi uoces latinae franguntur, tum in sonum tractusque transeunt sardinensis sermonis, qui, ut ipse noui, etiam ex latino est [‘where villagers speak an almost intact Latin and, when Latin words are corrupted, then they pass to the sound and habits of the Sardinian language which, as I myself know, also comes from Latin’] (Charlet 1993: 243)

As Charlet (1993: 245) and Varvaro (2000: 498) observe, this testimony is reliable, given that Sardinia at that time had been under Catalan rule for more than a century, so that Riaria would surely have had the opportunity to trade in Sardinia as well as Africa, or at least to pass through the island on the route to Tunis. This rural persistence seems to have been the last rampart of the Romance variety that some centuries earlier was still documented in towns such as Kafsa/Gafsa: the latter had a sizeable Christian community as late as the 12th century which is reported by the Arab geographer Al-Idrîsî (in 1154) to speak ‘al-Latini al-Afriki’ (see Lewicki 1951–2: 417). This reported perception of similarity between the two Romance varieties can be pinned down to specific properties of their phonetics/phonology. The two most evident linguistic features separating Sardinian from the rest of Romance are the lack of palatalization of velar stops before front vowels and the pairwise merger of long and short non-low vowels (cf. }2.4, (25)). Now, there is evidence that both isoglosses were shared by the Latin of Africa. Velar stops remain unaffected in the Latin loanwords into Berber (e.g. tkilsit ‘mulberry-tree’ < (morus) celsa, i-kīk@r ‘chickpea’ < cicer, ig(e)r ‘field’ ager, cf. Schuchardt 1918: 22, 24, 50) and are 40 Unlike Pompilio’s testimony (on which see directly below), reference to a spoken Romance variety in North Africa by Leo Africanus, in the 16th c., does not rest on direct evidence but rather on tradition (cf. Varvaro 2000: 496). Leo was born in Granada about 1485 into a prominent Muslim family that moved to Fez (Morocco) in 1492 because of the Spanish reconquista. Having come to Rome as a prisoner, he wrote there his Descrittione dell’Africa, completed in 1526 for Pope Leo X, where he reports that the inhabitants of north Africa retained their own language (which he calls ‘Italian’) after Muslim conquest (cf. Lewicki 1951–2: 431). But even if this 16th-c. mention of indigenous Romance in Africa refers to earlier periods, we still have direct evidence from a few decades earlier. Thus, one has to be careful not to claim that ‘Africa did not produce a Romance language’ (Adams 2007: 260), in the same sense as it can be legitimately claimed about Britain (p. 583), nor to presume that loanwords into Berber, if borrowed ‘at a later date’ after the fall of the Western Empire, must have come from one of the now extant Romance languages (p. 247), rather than from the now submerged popular outgrowth of the Latin of Africa.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rise of OSL

49

consistently notated , in Latin inscriptions from Tripolitania as late as the 10th–11th centuries, diverging from contemporary European usage in writing Latin (cf. forms such as dikite ‘say.imper.2pl’, iaket ‘lie.3sg’, with as in dilektus ‘beloved’, karus ‘dear’ from the inscriptional corpus analysed by Lorenzetti and Schirru 2010: 311). Also, the evidence bearing on the vowel system of African Latin/Romance is compatible with its being similar to Sardinian, as argued by many (cf. Wagner 1941: 10; Weinrich 1958: 24; Lausberg 1976: 204; Adams 2007: 262), with pairwise neutralization of short vs long vowels: Augustine’s testimony on the merger of ōs ‘mouth’ and o˘ s ‘bone’ (cf. (2) above) would not make sense if African Latin had a common Romance vowel system of type (15), where the two vowels were kept distinct as /o/ vs /ɔ/. On the other hand, this testimony is perfectly in keeping with the assumption of an earlier demise of CVL and an early rise of OSL, of the kind indicated by the metrical evidence reviewed in }}2.3.1 and 2.3.2. Both linguistic features just mentioned are likely candidates for a recognizable accent, allowing Latin speakers from other parts of the Empire to spot Africans. Latin sources indeed corroborate this: thus Augustine, De ordine 2.17.45 (386 ad) reports his acquaintances in Mediolanum (Milan) finding fault with his African accent: me enim ipsum, cui magna necessitas fuit ista perdiscere, adhuc in multis uerborum sonis Itali exagitant et a me uicissim, quod ad ipsum sonum attinet, reprehenduntur [‘For I myself, upon whom there has been a great compulsion to learn these things thoroughly, am still criticised by the Italians in the matter of many sounds within words, and they in their turn are criticised by me in the matter of sound’]. (Adams 2007: 193)

Somewhat earlier, the author of the biography of Septimius Severus in the Historia Augusta recounts that the emperor, born in Leptis Magna in 146 ad into a Roman equestrian family, was never able to get rid of his accent: ‘Afrum quiddam usque ad senectum sonans’ [‘He retained an African accent right up to old age’] (Spartianus, SHA 19.10; see Schuchardt 1866–8: 1.98). This remark may be just derogatory, but if reliable it surely does not mean that Septimius did not master Latin natively: ‘an African accent need not be a Punic accent’ (Barnes 1967: 96). Rather, as Birley (1999: 35) put it, the emperor’s African accent ‘was a provincial Roman accent, not a foreign one’. This is worth emphasizing, since Adams (2007: 260) is inclined to see ‘ “learners’ Latin”, which is not the same thing as a regional variety’ even within that emperor’s family, based on another passage from Historia Augusta. This reports Septimius sending back home his own sister, Septimia Octavilla, from Asia, where he was leading a military campaign in 198 ad, because she was ‘uix latine loquens, ac de illa multum imperator erubesceret’ [‘scarcely capable of speaking Latin, and the emperor was greatly embarrassed by her’] (Spartianus, SHA 15.7). The reliability of this testimony has been questioned by historians for various reasons (cf. Birley 1999: 131). But even assuming that the biographer, writing more than two centuries later,

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

50

The Latin–Romance transition

elaborated on contemporary reports rather than inventing ‘a mere slander, directed at an emperor reputedly boorish’ (Barnes 1967: 96), the passage cannot possibly imply that Octavilla was an L2 speaker of Latin (pace Adams 2003: 237, 289), since it is unthinkable for a Roman equestrian family to have raised their children in a language other than Latin in 2nd-century Leptis Magna.41 Against this background, one can now evaluate Adams’ conclusion that Iasucthan’s poem, too, is an instance of learner’s Latin. First of all, Adams himself has shown elsewhere (1994: 90) that the Latin written at the Bu Njem outpost—even at a much lower level than Iasucthan’s poem, such as that preserved by the ostraca (many of them written by simple members of the garrison)—cannot be legitimately considered a creole, as suggested by earlier interpretations (cf. Marichal 1992: 48). In addition, a centurion would be much more tightly integrated into Roman culture than the ordinary troops, and Adams (1999: 130–33; 2003: 551–4) gathers ample evidence of bilingualism in the imperial centurionate.42 The linguistic evidence from Iasucthan’s verses does not seem to attest with certainty to his being a nonnative speaker (pace Adams as well as Clackson and Horrocks 2007: 256–62), since for most of his oddities parallels from non-L2 Latin texts are adduced by Adams himself. There are various hanging constructions, which are characteristic of the written production of substandard speakers at all times, and examples of wrong uses of lexemes, collocations, and verb tenses/moods. Take for instance tertia Augustani ‘the third Augustans’ (or, ‘those of the legio tertia Augusta’, line 8), which is odd because ‘logically tertia augusta would produce tertiani augustani’ (Adams 1999: 111, n. 16) whereas, if the writer wanted to build a compound, this would have needed a linking -o (like mulo-medicus, Leumann 1977: 390). This is true, but one must consider that this seldom-occurring kind of compound may challenge the competence even of literate native speakers. Take for instance Italian compounds with tosco‘Tuscan’ as a first member: this shorter form of toscano- is acceptable for all speakers in coordinative compounds such as (Appennino) tosco-emiliano ‘TuscanEmilian Apennine’, (campionato) tosco-ligure ‘Tuscan-Ligurian championship’, but for many speakers (including myself) it is ungrammatical in subordinative compounds. Nevertheless, forms like (area) tosco-occidentale (cf. e.g. Franceschini 1991: 262) and tosco-orientale (cf. e.g. Tristano 2006: 269), ungrammatical in my variety, do

41 This was already judged correctly by Schuchardt (1866–8: 1.98): ‘Es lässt sich wohl denken, dass eingeborene Afrikanen das Latein nicht lernten, allein nicht, dass in Afrika angesessene Römer dasselbe verlernten’ [‘It is conceivable that African natives did not learn Latin, but surely not that Romans settled in Africa forgot it’]. A more plausible account is Birley’s (1999: 132), who surmises that, if this piece of information is genuine, Octavilla might have married a ‘Greek speaker and her Latin had become rusty from lack of practice’. 42 Thus, even if for some Latin writers at the Bu Njem outpost Adams (1999: 123) invokes the parallel of ‘[l]inguistically reduced forms of communication, based on French, [ . . . ] typically used by recruits to the French foreign legion’, an officer like Iasucthan would not have been in that category of low-level writers.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Quality is not quantity

51

occur even in academic writing.43 Another putatively L2 error is the subjunctive form dederit used instead of the required indicative dedit in quidam amplius uoluntatem suam dederit ‘one gave his enthusiasm’ (line 23). If similar errors in native speakers’ use of verb tenses/moods were unheard of, one could not explain the rise of such preterite forms as OFr. auret ‘had.3sg’, pouret ‘was able to.3sg’, which derive from Lat. pluperfect habuerat and potuerat respectively (cf. Roncaglia 1971: 158). All in all, it seems that there is no compelling reason to believe that Iasucthan was a non-native speaker of Latin. Assuming that he was, perfect similarity in his treatment of vowel quantities with the writers of the African inscriptions analysed by Herman (1982) (see }2.3.1) would be quite astonishing: if, however, he was a Latin-speaking African, like those writers, then we have one more piece of evidence confirming that African Latin merged CVL, establishing OSL, earlier than the Latin spoken in Italy, which is represented in the Bu Njem poem by Iasucthan’s colleague, Q. Avidius Quintianus.

2.4 Quality is not quantity, after all The evidence discussed in }2.3 impels us to reconsider recurrent claims of a very early demise of CVL in (Roman) spoken Latin. Re-examining such claims, it becomes clear that they rest, fundamentally, on just one piece of evidence: the recurrent graphic exchanges between vowels of different qualities. In particular, exchanges between / occur in Rome (and in other parts of Italy) as early as the 3rd century bc and become more and more frequent during the Empire, in the epigraphic Latin of central and northern Italy and Gaul above all. Thus, for instance Bonfante (1968: 25) adduces for ĭ in Pompeii inscriptions (like in μηνοϛ, veces for minus ‘less’, uices ‘times’), combined with metrical evidence of the kind commented on above in (20), in order to claim that ‘[l]a pèrdita delle distinzioni quantitative è già sicuramente attestata a Pompei (dunque prima del 79 d. C.)’ [‘the loss of quantity distinctions is surely documented already in Pompeii (hence before 79 ad)’]. The same argument is often used for the exchange of with . According to Väänänen (1966: 19, 24) the fact that this exchange involves much more frequently ĕ than ē (e.g. aedeo = edō ‘eat.1sg’ CIL IV 6892) attests to the fact that the phoneme notated was already in the process of merging with short ĕ and that, more generally, ‘l’opposition de timbre commençait déjà à l’emporter sur l’opposition de quantité des voyelles’ [‘the quality contrast was starting to override the vowel quantity contrast’] (Väänänen 1966: 19).

43 In the first article quoted, on the same page one reads another ungrammatical compound: ‘nei testi antico-pisano’ [‘in Old Pisan texts’]. This time the problem does not lie in the form of the first member but in the lack of agreement of the second with the head noun: grammatical alternatives would be antico-pisani (compound) or antichi pisani (phrase). While this is very likely to be a misprint, it does show that such compounds challenge the competence—at least—of contemporary Italian typesetters and proofreaders.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

52

The Latin–Romance transition

This seemed evident to many, given that in Romance the two did eventually merge without residue: e.g. It. presto [’prεsto] ‘soon, early’ < praesto = veste [’vεste] ‘dress, garment’ < vestem (cf. e.g. the direct link between the graphic / merger— with the implication that ē was not concerned—and the Romance development established by Sturtevant 1968: 127–8). But alluring though this view is, with regard to 1st-century Pompeiian Latin the argument is inconclusive, since examples of for ē do occur as well (e.g. aegisse = ēgisse ‘to have done’ CIL IV 2413f add., Väänänen 1966: 24), in spite of the fact that the corresponding phonemes never merged in Romance (except in Sardinian, cf. (25) below).44 The imbalance in distribution—with more instances of / than of / confusions—is simply due to the fact that, as Herman (1998: 9, n. 7) observes, ĕ was altogether much more frequent than ē: all these deviations from the CL norm attest to nothing more than the interchangeable use of the graphemes and . Of course, these graphical exchanges are proof enough that changes in vowel quality were taking place. However, as exemplified above, such evidence has been generally overinterpreted as an indication that the change from the CL vowel system (14) to the Romance one (15) took place as schematically shown in (22): (22)

From the Classical Latin to the ‘Italic qualitative system’ (e.g. Lausberg 1976: 202)

a. iː b. i

i

eː e

e

a(ː)

o

ɛ

a

ɔ

oː

u o

uː u

Schemes such as that in (22), standardly included in reference works on the Latin– Romance transition (e.g. Rohlfs 1966–9: 1.6; Lausberg 1976: 202; Väänänen 1981: 30), imply that the changes in quality undergone by short vowels ([i] > [I] > [e], [u] > [ʊ] > [o]) are sufficient evidence for the collapse of Latin VL, and for a ‘transformation de la durée latine en timbre’ [‘transformation of Latin length into timbre’] (Straka 1956: 199).45 Note that the output system (22b) is labelled as ‘Italic qualitative system’ because the Italic languages show a similar shift in vowel qualities long before Latin; the evidence is particularly clear for Oscan, but the change is assumed to be shared by Umbrian, and hence to be proto-Sabellian (cf. Lejeune 1975; Seidl 1994). Seelmann (1885: 183, 190, 225) already provides ample documentation of / exchanges involving both ĕ and ē. 45 In this passage, Straka follows Väänänen (1966: 18–25)—quoted of course in the 1st edn (1937: 25–40)—in claiming that this ‘transformation’ must have been completed, at least ‘dans le parler populaire’ [‘in popular speech’] (Väänänen 1966: 18), by the time of Pompeii’s destruction. In Straka (1959: 178–9, n. 24), on the other hand, the Czech scholar takes issue with the handbook presentation in (22): ‘On enseigne généralement que la durée vocalique s’est transformée en timbre, mais on oublie qu’entre les deux stades il y a eu inévitablement un stade intermédiaire où la durée et le timbre ont coexisté pendant un certain temps’ [‘It is generally taught that vowel duration was transformed into timbre, but one forgets that between the two stages there has unavoidably been an intermediate stage where duration and timbre have co-existed for a certain time’]. See Haudricourt and Juilland (1949: 30) for a similar view. 44

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Quality is not quantity

53

But careful evaluation of the diachronic phonology of Italic shows that the scheme in (22) is oversimplified. In fact, even with the said rearrangement of the length-toquality mapping, Italic had preserved contrastive VL, if only under stress. Thus, in Oscan, the reflex of PIE *ĭ in e.g. pís ‘who’ [’pes] still contrasted with that of PIE *ē (in e.g. trííbúm ‘house.acc’ [’tre :bom]).46 In this vein, even if short /i/ changed to [I]/[e], this vowel may still have been kept distinct from long /e :/, just as in today’s popular Québecois French, brique ‘brick’, realized variably as [’brIk] or [’brek] (cf. Santerre 1971: 557), contrasts for length with the English loanword [’bre :k] ‘brake’ (Y.C. Morin, p.c.). Likewise, monophthongization of /ae/ might have resulted in /ε :/, still contrasting with /ε/. Thus, to conclude reliably for merger, evidence from vowel quality is not sufficient: rather, evidence on quantity is indispensable, though difficult to get. And this evidence, as seen in }2.3, supports a later dating of the demise of CVL in Italy, as far as stressed vowels are concerned. Stressed vs unstressed positions have to be carefully distinguished also as one evaluates ancient testimonies by grammarians. In fact, all the passages quoted above in }2.1, attesting to differences in quality between short and long es, and thus also to long /e :/ approaching short /i/ in articulatory space, are compatible with a system like (14), as none reports merger. Note, however, that they all discuss stressed vowels. The picture changes when one takes unstressed vowels into account. In fact, an earlier passage by Terentius Scaurus, writing in Hadrian’s reign in the 2nd century ad, proposes that adjectives like facilĭs ‘easy’, docilĭs ‘tame’ be written with in the nominative singular non-neuter form, whereas for the nominative/accusative plural non-neuter (facilēs) he recommends (GL 7.32.21–33.2) (cf. e.g. Seelmann 1895: 75).47 In unstressed position, therefore, the need was felt to distinguish between two sounds which had formerly been contrasting, in both quantity and quality, but were apparently converging. Note that, when it comes to stressed vowels, Terentius himself keeps recommending apex and I longa (7.33.4–10): of course, this amounts to following established tradition (cf. }1.1, n. 8). But interestingly, in the case of the graphical distinction facilis vs facileis an innovation was perceived as needed, whereas Terentius’ being content with traditional vénit ‘come.prt.3sg’ vs venit ‘come.3sg’, pIla ‘pile’ vs illa ‘that.fsg/npl’ suggests that merger of the stressed vowels of vēnit vs ĭlla (It. v[e]nne = qu[e]lla) was not yet a real threat in his time.

46 In his study of the Oscan vowel system, Lejeune (1975: 249–50) states clearly that the loss of quantitative oppositions is in no sense a necessary consequence, at least in the short term, of the transformation of the system of vowel quality. The same can be assumed for the phonological development of vowel quality in Latin, as pointed out e.g. by Spence (1965: 12), Franceschi (1976: 277), Loporcaro (1997: 68–9), Seidl (1998: 377), and Adams (2013: 43). 47 Under an alternative interpretation of the passage (Goidànich 1906: 51), Terentius is reporting on current usage rather than putting forward a prescription himself. Also in this case, the essential point is that—whoever felt that need—a need for graphical distinction was first felt for unstressed ĭ vs ē, in spite of the fact that the doubtless (near-)merger of qualities concerned stressed vowels as well.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

54

The Latin–Romance transition

Furthermore, epigraphic evidence from different regions and provinces of the Latin-speaking world shows that this merger progressed asymmetrically, with the lowering of ĭ to [e] preceding that of ŭ to [o]. Analysing an epigraphic corpus from the mid-2nd until the 4th century, Herman (1985a: 75–6) found that graphic confusions between balance those between only in northeastern Italy (Regio X), and nearly so in parts of Campania (see however Bonfante 1983: 417 on lack of merger of vs in Pompeii), whereas in the rest of Italy and in Gaul confusions occur more seldom than would be expected given a random distribution. Herman (1985a: 76) also observes that on the Adriatic coast south of Ancona (the area including Apulia), such confusions do not appear before the Christian era. Moving eastwards, Dalmatia shows also confusions, although not so frequent as (Herman 1971: 139–43). In nearby Pannonia, they occur generally for whereas for they are restricted to the southern fringe of the province, bordering on Dalmatia; in Dacia, on the other hand, only very rare occurrences of confusions are found (Herman 1983: 103). This squares with the evidence from later Daco-Romance, in which, as shown in (23) with examples from Standard Romanian, the outcomes of Latin ˘ı and e¯, but not those of u˘ and o¯, have merged (the first line displays the Latin source vowel):48 (23)

Romanian vowel system iː

i

eː

i e fir leg cred ‘thread’ ‘tie.1sg’ ‘believe.1sg’

e

a

ɛ piatra ‘stone’

a vaca ‘cow’

o

oː

u

ɔ roata ‘wheel’

uː u

soare ‘sun’

gura ‘mouth’

cur ‘arse’

Beyond Daco-Romance, this kind of asymmetrical merger occurs also in a small area of western Lucania—a fact first revealed by AIS (see also Rohlfs 1937; Lausberg 1939: 44–6):49 (24)

Asymmetric (‘Romanian-like’) vowel system (dialect of Castelmezzano, province of Potenza, AIS pt 733, see Map 1 in Appendix for localization) iː

i

eː

e

a

o

i ˈfiːl ‘thread’

ˈseːtə ‘thirst’

e ˈseːra ‘evening’

ˈmeːlə ‘honey’

a ˈaːkə ‘needle’

ˈkoːrə ‘heart’

oː

u

uː

ˈsoːlə ‘sun’

u ˈvuddə ‘boil.3sg’

ˈmuːrə ‘wall’

o

48 Some instances of u˘ > /o/ lowering do occur in Romanian as well: e.g., scoate < excu˘tere ‘to shake’, roşu < ru˘sseum ‘red’ (cf. Sánchez Miret 2001: 377); this shows that the claim (by e.g. Straka 1959: 180) that the change first arose after complete separation of Romanian from the rest of the Latin-speaking world is overstated. 49 Unlike Daco-Romance, in addition, in this area the outcomes of proto-Romance /ε/ merge with those of /e/, as illustrated in (24).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Quality is not quantity

55

The occurrence of such an asymmetric vowel system in this small area of southern Italy prompted the inference that a substantial part of the Romance-speaking territory—apart from Sardinia (see (25) below) and possibly Africa—went through a ‘Romanian’ stage. Thus, according to Lüdtke (1956: 97–8), ‘Apulien und nicht der Balkan Ursprungsland des östlichen Vokalsystems ist’ [‘Puglia and not the Balkans is the place of origin of the eastern vowel system’] and according to Bonfante (1983: 417) the entire Italian peninsula once had the vowel system in (23).50 Further evidence for the broader geographic extension of such an asymmetric vowel system seems to come from Dalmatian, whose northern variety was the last to become extinct in 1898, with the death of the last speaker of Vegliote. In Vegliote the outcomes of e¯ and ĭ merged in all syllabic contexts (e.g. [’mai ̯k] < me¯cum ‘with me’= [’fai ̯d] < fĭdem ‘faith’= [’stale] < ste¯llae ‘stars’ = [’laŋga] < lĭnguam ‘tongue’) whereas o¯ and u˘ merged in open syllables, as they were both affected by diphthongization ([’bau̯d] < vo¯cem ‘voice’ = [’nau̯k] < nu˘cem ‘(wal)nut’), but remained distinct in checked syllables ([’samno] < *so¯mnium instead of so˘mnium ‘dream’ vs [’buka] ‘mouth’ < bu˘ccam ‘cheek’; cf. Bartoli 1906: 2.336–7).51 Further proof of the independence of the collapse of distinctive VL and the qualitative mergers of ˘ı vs e¯ and of u˘ vs o¯ is provided by Sardinia and the other Romance areas which share the Sardinian vowel system. Herman (1985c) demonstrated that in the island’s epigraphic Latin, / confusions are extremely rare (cf. also Herman 2000: 131; Lupinu 2000: 32f.) and tend to be limited to inflections, which further Romance development suggests have been reshaped analogically. Thus, for instance, qvescet ‘rest.3sg’ CIL X 7550 or dvcet ‘bring.3sg’ CIL X suppl. 772, addvcet CIL X suppl. 798 are not instances of phonetic merger but rather of the substitution of an inflectional affix, as confirmed by modern Logudorese [’dʒu :ɣεðε] ‘bring.3sg’ (< ducet  iugum), which represents the generalization of the ending -/εt/ to the regular subclass of 2nd conjugation (contrast 3rd conjugation -/it/ in [’fi :niði] ‘finish.3sg’ < finit). Once morpheme substitutions of this kind are factored out, Herman shows, vowels, both stressed and unstressed, are basically stable in the 50 Lüdtke (1956: 175–85) pushed this line of argument to the extreme, claiming that even Portuguese, at the western periphery of the Romània, shows traces of the same successive layers in the development of stressed vowels, so that [ε ɔ] from e˘ o˘ in e.g. bod[ε]ga, f[ɔ]rma evidence a ‘Sardinian’ stratum, whereas [u] from u˘ in e.g. jugo, sulco, cruz are compatible with either a ‘Sardinian’ or a ‘Romanian’ layer, upon which the common Romance four-height vowel system was superimposed through lexical diffusion, as the product of later waves of colonization. There are alternative explanations for this evidence: for instance, the metaphonic alternation in the -oso/-osa suffix, though explained by Lüdtke as a remnant of a Sardinian system, may just as well be attributed to analogy on paradigms in which an etymological short o˘ had developed regularly into an alternating vowel (e.g. n[o]vo vs n[ɔ]va ‘new.m/f ’). 51 Although this fact has been emphasized by many scholars (e.g. Lüdtke 1956: 293–5; Bonfante 1998: 12), Bartoli himself observed that the evidence for [a] from checked o¯ is scanty. At any rate, a later stratum with ‘common Romance’ vowel system can be inferred for Dalmatian, as the Dalmatian loanwords into Slavic and Albanian (cf. Muljačić 2000: 331–3) all show a common treatment of o¯ and u˘ (> /o/ > /u/) as opposed to u¯ (> /y/).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

56

The Latin–Romance transition

epigraphic evidence from Sardinia, which is in keeping with the vowel system of the Romance varieties which have arisen on the island. Here, the allophonic differentiation [i :/I], [e :/ε] etc. occurring in CL (cf. (14) above) was apparently given up, so that no merger of [I] with [e( :)] ever occurred. The final outcome, i.e. the Sardinian vowel system (25), shows a plain merger of CL long and short counterparts into one phoneme, yielding a five-vowel system (examples from Logudorese): (25)

Sardinian vowel system iː

i

eː

ˈpiːra ‘pear’

ˈsεːrɔ ‘evening’

a(ː)

ˈbεːnε ‘well’

ˈkaːnε ‘dog’

ɛ

i ˈfiːlu ‘thread’

e

o

oː

u

ɔ

a ˈbɔːna ‘good.fsg’

uː u

ˈbɔːɣε ‘voice’

ˈruːɣε ‘cross’

ˈluːɣε ‘light’

This kind of vowel system is shared today by Logudorese (the most conservative variety, spoken in the central-northern part of the island) and Gallurese (in the northeast), but there is evidence that the remaining parts of the island also once displayed the same vowel system.52 The same Sardinian system (25) stretches into nearby southern Corsica as well (e.g. [’filu] ‘thread’ = [’pilu] ‘hair’, [’mezi] ‘month’ = [’pedi] ‘foot’, [’kori] ‘heart’ = [’fjori] ‘flower’, [’fur :u] ‘oven’ = [’mulu] ‘mule’, cf. Barbato 2008: 145; Dalbera-Stefanaggi 2001: 99–100) and it also occurs in an area of southern Italy (first described in detail by Lausberg 1939, also based on AIS data, who dubbed it Mittelzone [‘Middle Zone’]), which covers the northernmost part of Calabria (north of the rivers Crati and Coscile) and southern Lucania (south of the river Agri). In the dialect of Trebisacce (province of Cosenza), for instance, the outcomes of Latin e¯ and e˘ merge (e.g. [’tε:ʁ@] < te¯lam ‘cloth’ = [’tε:n@] < te˘net ‘holds, has’) and contrast with that of Latin ĭ (e.g. [’pi:p@] < pĭper ‘pepper’); symmetrically, for back vowels, one finds merger of o˘ o¯ [’nɔ:v@] < no˘uam ‘new.f’ like [’sɔ:ʁ@] < so¯lem ‘sun’, contrasting with [’nʊ:tʃ@] < nu˘cem ‘walnut’ (see Pace 1993–4).

52 In the southern half of the island, Campidanese nowadays has a seven-vowel system, due to the opacification of metaphony, which applies regularly to stressed mid vowels before high vowels in Logudorese: e.g. [’be :ni] ‘come.2sg.imp’< veni vs [’bε :nε] ‘well’ < bene. In Campidanese, on the other hand, post-tonic /ε ɔ/ were raised, a change which spread from Cagliari towards the centre of the island from the 11th c. on (cf. Wagner 1941: 36–7; Loporcaro 2002–3: 192–3): as a consequence, minimal pairs such as [’be :ni] ‘come.2sg.imp’< vs [’bε :ni] ‘well’, or [’ol :u] ‘oil’ < oleum vs [’ɔl :u] ‘want.1sg’ < *voleo occur nowadays. This led scholars in the structuralist tradition (Virdis 1978: 26; 1988: 900; Piras 1994: 208–17; Ferguson 1976: 107 etc.) to assume phonologization of the /ε 6¼ e/ and /ɔ 6¼ o/ contrasts, whereas generative analyses represent both Campidanese metaphony and unstressed vowel raising as synchronic processes, so that underlyingly the same five-vowel system is assumed as for Logudorese (see Bolognesi 1998: 19–22). As to the northwestern part of Sardinia, Sassarese nowadays displays an asymmetrical system in which the outcomes of Latin ĭ and ŭ have been lowered to /ε/ and /ɔ/ respectively, unlike in Sardinian (cf. Loporcaro 2013: 170, n. 144), but especially its rural variety spoken in Sorso (cf. Gartmann 1967: 26–8) has a fair amount of lexical residue of the original Sardinian vowel system: e.g [’kiɖ :u] ‘that.m’ < ecc(u)- ĭllum, [’pud :a] prune.3sg < pu˘tat.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Intermediate summary and conclusion

57

While for Sardinia and Corsica there is hardly any doubt that (25) is a primary development out of the Latin vowel system (14), for Lausberg’s Middle Zone it has been argued that this is due to secondary convergence, capitalizing on the fact that in some dialects of this area the outcomes of e˘ e¯ and o˘ o¯ do not merge under metaphony: e.g. in Senise (province of Potenza) metaphonic [’pjeɾ@] ‘feet’ < pe˘des is kept distinct from [’mIs@] ‘months’< me¯nses (contrast the merger of non-metaphonic [’pεr@] < pe˘dem ‘foot’ and [’mεs@] < me¯nsem ‘month’, in compliance with (25)); symmetrically, [’fwok@] < fo˘cum ‘fire’ vs [n@’pʊt@] < nepo¯tes ‘nephews’ remain distinct, while [’rɔt@] < ro˘tam ‘wheel’ and [n@’pɔt@] < nepo¯tem ‘nephew’ merge (cf. Fanciullo 1988: 676–7). The distinct outcomes under metaphony, thus, have been taken as evidence that the area might have gone through a stage with a vowel system of the common Romance type in (15). On the other hand, this evidence is counterbalanced by the existence of several other dialects of the area in which the metaphonic outcomes of e¯ o¯ and e˘ o˘ do merge: e.g. in Cersòsimo (province of Potenza) [a’tʃi@̯t@] < ace¯tum ‘vinegar’ = [’pi@̯t@] < pe˘de¯s ‘feet’ and [’su@̯r@tʃ@] < so¯rice¯s ‘mice’ = [’fu@̯k@] < fo˘cu(s) ‘fire’. From such data, Savoia (1997: 371) concludes that this situation is primary, and hence that the dialects of the Lausberg area did originally preserve a Sardinian vowel system. Finally, to Sardinia and the other Romance areas just mentioned, where the vowel system in (25) is preserved to this day, must be added Africa, which seems to have developed such a system in the Latin–Romance transition, given the evidence reviewed in }2.3.3 above.53 The evidence provided by the Romance varieties reviewed, displaying the Romanian and the Sardinian vowel system, demonstrates that the scheme in (22) is oversimplified, and that the demise of VL cannot be inferred directly from changes in vowel qualities.

2.5 Intermediate summary and provisional conclusion The study of change in vowel quality and quantity, for Late Latin, must be done in close connection, but without drawing unwarranted inferences from one to the other. Once one introduces some opportune distinctions, it emerges that the available evidence from all relevant sources either confirms (metalinguistic evidence from Consentius, Augustine etc.; epigraphical evidence from metrical inscriptions) or does 53 Some doubts about the postulation of a Sardinian vowel system for Africa have been voiced by e.g. Fanciullo (1992: 178–80) and M. Mancini (2001). True, epigraphic evidence does not provide, for Africa, as strong support as for Sardinia, since epigraphic Latin offers here many examples of confusion (see Acquati 1971: 159–65). However, in the most recent comprehensive evaluation of the epigraphic evidence, Adams (2007: 648) concludes for the plausibility of reconstructing the vowel system in (25) for African Latin, and the 15th-c. testimony on the perceived similarity of African and Sardinian Romance, mentioned in }2.3.3, further strengthens this conclusion.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

58

The Latin–Romance transition

not contrast (evidence from changes in vowel quality) with Herman’s (1998: 21) placement of the VL collapse within the first of the two main rounds of change he assumes to have taken place in the Latin–Romance transition (see }2.1). This first round was completed before the fall of the Western Empire, and affected a linguistic system that was basically still unitary in speech, as well as in writing (though of course with diatopic, diastratic, and diaphasic variation).54 Even though some harbingers of the change appear as early as the 2nd century, the collapse of CVL was completed towards the end of the Empire. Note that the issue as to whether OSL first arose in Africa or elsewhere (or perhaps polygenetically in different parts of the Empire), to which I devoted several pages in }2.3, while intriguing per se, is not crucial for the discussion of the Romance facts in the following chapters. The really crucial point is that the studies reviewed in }}2.3.1 and 2.3.2 make it plausible that at some time under the Roman Empire distinct regional dialects of spoken Latin could differ by displaying either CVL— well preserved at least under stress—or OSL, the former a conservative feature, the latter its diachronic successor. If this is true, we gain a powerful argument for reclaiming linguistic reconstruction in this structural area from the fog of sheer speculation, and anchoring it historically. There was indeed a point in time—a variable one, from region to region—in which CVL was ousted by OSL, and this change spread gradually to different Latin-speaking areas, possibly as part of Herman’s (1985) second wave of ‘dialectalisation latine’ or ‘pre-Romance differentiation’ of spoken Latin. On the strength of this evidence, one can exclude alternative scenarios which have been extensively argued for (but with essentially speculative arguments) in Romance linguistics. Thus, it is surely not the case that Roman legionaries and colonists imported into the conquered provinces a non-quantitative variety of Latin, as argued e.g. for Spain in the 3rd century bc by Alarcos Llorach (1951: 13).55 Alternatively, one might speculate that, even if CVL were there in the mother tongue of the conquerors, it was immediately lost in the provinces as the locals acquired Latin, in a process of creolization (such as that argued for by de Dardel and Wüest 1993 or Goyette 2000).56 Yet much evidence speaks against such a hypothesis for the Latin–Romancespeaking world. For Spain, Seidl (1995) demonstrated that a creolization scenario is, on the whole, implausible: this would imply, for instance, loss of case distinctions in 54 This is why Herman dubs this the second wave of Latin dialectalization, whereas ‘Romance dialectalization’ starts only after the fall of the Western Empire. 55 See however Grandgent (1907: 76) for an early stance to the contrary (‘It is possible that the quantity of unstressed vowels was better kept in the provinces than in Italy’), which, as will be shown shortly, appears to be borne out by subsequent research. 56 Indeed, Creole languages usually have a simple phonological system: on the non-occurrence of CVL—and its loss, in the case of the lexifier language possessing it—in pidgins and creoles, cf. Heine (1979), Holm (1989: 554; 2000: 146–7), Mühlhäusler (1997: 139). In Klein’s (2006: 10) corpus, there is just one creole language displaying CVL, Sri Lanka Creole Portuguese.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Intermediate summary and conclusion

59

nominal inflection (a point de Dardel and Wüest 1993: 48 insist on, in order to stress the creole-like nature of the early Romance varieties especially of Iberia and Sardinia), whereas there is evidence that two-case noun inflection persisted well into the Latin–Romance of Visigothic Spain.57 The same argument is available for Dacia, at the northeastern periphery of the Empire, and the latest province (conquered by Trajan in 106 ad), since Romanian preserves a two-case distinction to this day. Needless to say, the same is true of medieval Gallo-Romance. For the Latin spoken in the northern provinces, in addition, one finds positive evidence for maintenance of CVL. This has been argued on the strength of graphical use by Flobert (1989; 1990), who studied the occurrence of apices and I longae in inscriptions from Vienne and Lyon dating back to the High Empire. A more focused study of apices in the Vindolanda writing tablets was carried out by Adams (1995). The tablets, from the early 2nd century ad, were produced by scribes at a military outpost in Britannia which was manned by Batavian and Tungrian units, so that they might reflect the Latin ‘in use among Romanized natives of areas such as Gallia Belgica and Germany’ (Adams 1995: 87). The result of Adams’ analysis is that the use of apex demonstrates ‘that the writer was aware of the correct length of the vowel’ (p. 98). Of course, this awareness—like writing skills per se—must have been trained under the supervision of grammatici, but the pattern of occurrence of the apex makes it highly implausible that, however literate, the military personnel at that garrison employed apices notating quantity just because of grammatical prescription, with no native sense for quantity. This emerges clearly from the high incidence (76%) of apices on vowels in final syllables, and, even more to the point, on final -o in 1st person verb inflections: rogó 291, exoró 307, cupió, putó, scribó 319. This converges with other evidence (cf. }1.2) showing that long vs short vowels in unstressed syllables, even in Rome, tended to merge earlier than under stress.58 In particular, 1st person -o was generally recategorized as short, independently from iambic shortening, by the late Empire as attested by the grammarians Charisius (GL 1.16.20, mid-4th century ad) and Marius Victorinus (GL 6.28.15–22); cf. Allen (1973: 179–85), Leumann (1977: 109–10). Literate military personnel at Vindolanda are likely to have used apices in obedience to grammarians’ prescription that, in general, an apex is needed where there might be a doubt or the risk of confusion (in particular in minimal pairs, as seen in Chapter 1, n. 9). But, in following the general rule, they are unlikely to have been concerned about specific vowels and/or environments (say, 1st person singular -o) for cultural reasons. Thus, their practice must reflect their native linguistic intuition; in particular, it ‘can be interpreted as a reaction against the shortening of long vowels,

57

Seidl’s (1995) demonstration is ignored by Goyette (2000). More recently, scepticism towards the creolization theory has been voiced by Maiden (2011: 699, n. 1) and Ledgeway (2012: 333–4). 58 In some of those verb forms (putó, rogó), iambic shortening had applied, unlike in e.g. scribó.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

60

The Latin–Romance transition

particularly -o, in that position’, a ‘habit which was spreading in colloquial speech’ (Adams 1995: 98). This conclusion is strengthened by comparing the occurrence of diacritics in other provinces: Flobert (1989: 5), having observed a fairly correct use of apices in his corpus from Gaul, contrasts this result with the exceptionality of quantity diacritics in inscriptions from Africa (cf. Christiansen 1889: 19). This may suggest that diacritics are least used in the province where CVL was lost earliest.59 That the Latin speakers from the northern provinces of the Empire had CVL in their native competence is further confirmed by borrowings into the Celtic and Germanic languages, which show unambiguous reflexes of the Latin VL contrast, as mentioned above in }1.1. Thus Jackson’s (1994: 88) study of Latin loanwords into Britain’s insular Celtic languages concludes: ‘There is no certain trace of this [i.e. of the demise of CL CVL] in British; with one or two possible exceptions, the loanwords show that the Classical quantity system was preserved, even with unstressed vowels, in the Latin of Roman Britain’. Jackson’s observation was part of a more general claim—substantiated through a list of twelve linguistic traits—to a particularly conservative character of Britain’s Latin. Adams (2007: 586–94) calls into question the validity of Jackson’s conclusions, but has nothing to object on this specific point; nor has Smith (1983: 940), who observes on VL: ‘I have no ground on which to offer a different view’. Having anchored historically the change from CVL to OSL to a Late Imperial Latin stage has consequences for the reconstruction of Proto-Romance: in addition to the well-established generalization that no Romance languages inherited the Latin VL contrast (a quite uncontroversial fact), one might venture—after the present discussion—that all may have inherited OSL, since this is demonstrably an innovation which made its way into the Latin of the late Western Empire. So equipped, we can now move on to consider the further development of vowel length in the Romance languages, where we shall see that assuming a Proto-Romance OSL proves instrumental in assessing competing hypotheses on the further development of VL in the Romance languages.

59 Out of the 10,988 inscriptions of CIL vol. VIII, only one or two display apices. However, the argument should be somewhat relativized, given that no single instance of apex occurs in the inscriptions from Britain either (CIL VII, 1,355 inscriptions in all; cf. Christiansen 1889: 19), in spite of the fact that retention of CVL in the Latin of Britain is evidenced by Latin loanwords into Celtic (see below).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

3 The development of VL in Romance The Romance languages have been mentioned in the foregoing pages mainly as sources of evidence for the reconstruction and analysis of Latin at different stages of its development. From this chapter on, Romance data will be considered in their own right, beginning with a general overview of the further developments of quantity in the different Romance branches.

3.1 Three types of distribution of VL in the Romance languages It follows from what has been said in Chapter 2 that the Latin VL contrast did not survive into any of the Romance languages, which, at a first level of approximation, can be classified into three types, as shown in (1a–c): (1)

i. ’CV.CV [’la :to] ‘side’ 6¼ [’laðo] ‘side’ 6¼ < latus c. Northern Italo-Romance [’pa :n] ‘bread’ 6¼ (Cremonese) < panem a. Italian b. Spanish

ii. ’CVC.CV [’gat :o] ‘cat’ [’gato] ‘cat’ < cattum [’pan] ‘cloth’ < pannum

The three types differ with regard to whether or not, in the languages at issue, the stressed vowel is treated differently in the environments indicated in the first line: (1.i) vs (1.ii). This contextual difference, as made clear by the etymological forms in small capitals, refers to Latin and survives to this day into (Tuscan-based) Standard Italian, where /’la.to/ vs /’gat.to/ in (1a) still differ for the stressed vowel occurring in an open vs closed syllable. Descriptions of Italian in fact report lengthening of stressed vowels in open non-final syllables, which for instance in minimal pairs such as those in (2a,b) occurs in complementary distribution with the gemination of the following consonant:

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

62 (2)

The development of VL in Romance a. ’CV.CV [’ka :ne] [’va :le] [’fa :to] [’ka :de]

‘dog’ ‘be worth.3sg’ ‘fate’ ‘fall.3sg’

b. ’CVC.CV [’kan :e] [’val :e] [’fat :o] [’kad :e]

‘reed.pl’ ‘valley’ ‘done’ ‘fall.prt.3sg’

A stressed vowel in a word realized in isolation can last up to 200 ms. if it occurs in an open word-internal syllable (2a), whereas before a geminate consonant (2b) it is unlikely to last more than 100 ms. (cf. the experimental phonetic studies by Fava and Magno Caldognetto 1976, Bertinetto 1981, Marotta 1985, Farnetani and Kori 1986, D’Imperio and Rosenthall 1999, etc.).1 In other words, Italian displays Open Syllable Lengthening (OSL), which can be stated in the form of a phonological rule like the one formulated for Late Latin in (6), }2.1, except for the specification of the condition preventing final stressed vowels from undergoing lengthening:2 (3)

V → V/

]σ [+stress]

Open Syllable Lengthening (Italian) where ]σ ≠ ]PW

This can be restated, within a no-rule approach to phonology such as OT, in the form of the OT tableau in (4), where a structural constraint Acc(ent)-μμ overrides a faithfulness constraint Dep(endence)-μ-io and thus forces every stressed syllable to be bimoraic:3 (4) / kasa/ a. [ kasa] b. [ ka sa]

Superordinate constraints

ACC-µµ

DEP-µ-IO

non-selective constraints

∗! ∗

With this vocabulary, the change discussed in }2.2, which brought about OSL in Late Latin and thus determined the loss of CL CVL, is not modelled as the addition of an allophonic rule to the phonological component but rather as constraint reranking, since in Classical Latin the faithfulness constraint Dep(endence)-μ-io used to dominate Acc(ent)-μμ, thus allowing for e.g. Lat. /’kasa/ ‘hut’ to surface without lengthening of the stressed vowel. The notational difference between (3) and

Of course, absolute duration varies as a function of speech rate (see n. 19 below). See e.g. Bertinetto (1981: 75, 254), Vayra (1994), and van Santen and D’Imperio (1999) on the shortness of stressed word-final vowels in Italian (e.g. [tʃi’t :a] ‘city’), to which I shall return in }5.4.2. 3 This parallels Löhken’s (1997) treatment of vowel lengthening in Middle High German. 1 2

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Three types of distribution of VL

63

(4) is immaterial to my present concern: consequently, in what follows I will stick by the rule metaphor.4 The same distribution of VL as in Standard Italian (1a), or (3), with OSL, is found—not surprisingly—in Tuscan dialects as well as in the vast majority of central–southern ItaloRomance dialects, from the central Marche, Umbria, Lazio down to Sicily, as already illustrated in }2.2.1, (10).5 Type (1b) is exemplified with Spanish: there is no contrastive VL and all stressed vowels have approximately the same duration, regardless of the original syllable structure in the Latin input (reflected in the contrast in (1.i,ii)), and regardless of any other conceivable factor, for that matter (apart from low-level phonetic constraints of the kind mentioned in nn. 8 and 11 below). The same goes for the rest of Ibero-Romance and for Romanian. In fact, going through descriptions of the phonology of those languages, one realizes that vowel length is of no concern. Thus, Hualde’s (1992: 384) introduction to Catalan, in the section on suprasegmentals, mentions underlying length for geminate sonorants (nasals and laterals: e.g. ametlla ‘almond’ [a’mεʎ :@]) but, when it comes to vowels, speaks of long vowels only to refer to morpheme-internal sequences: ‘Morpheme-internally long vowels are rare. When they occur they can be optionally shortened in fast speech: e.g. vehement [b@@’men], [b@’men].’ (See also the vowel coalescence process attested in Old Portuguese, mentioned in Ch.1, n. 20). Another well-known case in which secondary CVL is reported to have arisen within Ibero-Romance is exemplified with the minimal pairs from Miami Cuban in (5a,b) (from Resnick and Hammond 1979: 86): (5)

a. pescado [pe :’kaðo] ‘fish’ busque [’bu :ke] ‘look for.prs.sbjv.1/3sg’

b. pecado [pe’kaðo] ‘sin’ buque [’buke] ‘boat’

This variety, like southern Peninsular and many American Spanish dialects, has undergone deletion of syllable-final /s/ (via /s/ ! [h]), which elsewhere results in contrasts in quality rather than quantity, as reported for Andalusian e.g. by Zamora Vicente (1967: 292): [’beso] ‘kiss’ vs [’bεsOh] ‘kisses’, [’sorðo] ‘deaf ’ vs [’sOrðO] ‘deaf. mpl’.6 For Miami Cuban, on the other hand, Resnick and Hammond’s (1979: 86) perceptual and instrumental study shows that the vowels after which /s/ has been deleted ((5a)) do not differ in quality from their counterparts in ((5b)), and are 4 I am reassured in this choice by Odden’s (2011a: 37) recent conclusion: ‘The general ideas of rule-based and constraint-based grammar are sufficiently open-ended that neither can be per se reasonably judged superior to the other.’ 5 In the same section, I have also mentioned southern Italian dialects (mostly from the eastern half of southern Italy, along the Adriatic coast), where the allophonic distribution (1a) gave way to a stronger differentiation of stressed vowels in open vs closed syllables owing to sound change (dipththongization and colouring of various different kinds) affecting stressed vowels in open syllables. For the (diachronically based) classification in (1a–c), those dialects are a subtype of (1a). 6 On s-deletion in Andalusian Spanish, see more recently Hernández-Campoy and Trudgill (2002).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

64

The development of VL in Romance

fully neutralized in word-final position, so that e.g. comprendes and comprende ‘understand.2sg vs 3sg’ are not discriminated by native speakers out of context, whereas word-internally /s/-deletion in (5a) results in a lengthening by average 36.3%. Several other similar cases have been reported for varieties of American Spanish (see Seklaoui 1989: 59–61 for an overview). But in the present context, neither long vowels arisen via coalescence, such as in the Catalan examples above, nor derived long vowels of the Andalusian/South American kind will be dwelt upon in themselves, since none of these is in a diachronic relationship with either CL CVL or the Late Latin OSL.7 Daco-Romance falls under the same rubric as Ibero-Romance, for our present purposes, since in Romanian, too, stressed vowels are phonologically non-distinct and no allophonic rule differentiates them: thus, stressed vowels have phonetically the same duration in e.g. [’sarε] ‘salt’ < salem, [’kas@] ‘home’ < casam (etymological context (1i)), on the one hand, and [’mam@] ‘mum’ < mammam, [’vak@] < vaccam ‘cow’ (etymological context (1ii)), on the other. The same goes for e.g. [’nas] ‘nose’ < nasum vs [’kal] < caballum ‘horse’, which differ synchronically from the above examples because of their being monosyllabic (owing to apocope), but instance the same diachronic contrast (1i) vs (1ii).8 The third option in (1c) is exemplified there with a Northern Italo-Romance (henceforth NItRom) variety, the dialect of Cremona (southern Lombardy; see Map 3 in Appendix for localization). This kind of distribution is found—in different subtypes—in most of NItRom and in several transalpine Romance varieties from the Gallo- and Rhaeto-Romance areas. In this area, sometimes called Northern Romance, a novel VL contrast was established, as apparent from the minimal pair in (1c). The basic question to be addressed in what follows is that of the historical relationship between these three different kinds of Romance developments of VL. I will explore the hypothesis that the OSL found nowadays in Standard Italian derives in a straight line from the lengthening which, as we saw in }2.2, arose in Late Latin and that, in turn, the same OSL lies at the basis of the secondary CVL of type (1c). Elaborating on this basic idea, it has been hypothesized that such an OSL was shared, originally, by all Romance varieties. This hypothesis, as shown in }2.2, yields a straightforward explanation for the demise in Late Latin of Classical Latin VL, 7 In what follows, long vowels which arose through vowel coalescence, monophthongization, or compensatory lengthening will be mentioned only if they occur within systems which otherwise possess CVL of type (1c) or at least within dialect areas in which this kind of CVL is documented in neighbouring varieties (cf. }3.4.1.3 on Genoese, }3.4.2.2, and n. 70 on the Occitan patois of Ambialet). 8 The experimental phonetic study on Romanian vowels by Renwick (2012: 172) does report duration differences between stressed vowels as a function of syllable structure (stressed vowels are longer in CV syllables than in CVC ones) and the occurrence in word-final vs word-internal position (to which we shall return below, }5.4.1). Within the model of phonology discussed below in }3.2, (7), this kind of difference belongs in the realm of low-level phonetic implementation constraints ((7d)). Anyway, none of these differences correlates with the diachronically based distinction in (1i) vs (1ii).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In defence of OSL in modern Standard Italian

65

which was contrastive and independent of syllable structure (as seen in }1.1, (1) and (2)), but came gradually to depend on it (}1.2) and eventually could not survive the establishment of the allophonic rule (3)—or the reranking of the relevant constraints in (4).9 As for the diachronic relationship between the Standard Italian type of allophonic distribution (1a) and the other two (1b,c), the null hypothesis, given what has been said so far, consists in viewing the latter as further developments of an original state (1a). To pave the way for exploring this hypothesis, I will first (}3.2) comment on recurrent claims that the statement of an allophonic distribution such as (3) is not really an accurate description of the modern Italian facts. Once (1a) (and (3)) are established as the second stronghold—together with the rise of OSL in Late Latin—of my reconstruction, I shall move on to consider the other options attested by the Romance languages and dialects. As for the systems of type (1b), nowadays found at the eastern and western peripheries of the (European) Romània (}3.3), the Romance varieties spoken there are not internally differentiated for the phenomena we are considering, and have not been so throughout their documented history: thus the assumption that they must have shared, at the outset, an OSL process must remain speculative, though comparative arguments in support of this hypothesis can be drawn—as we shall see later (in Chapter 5)—from dialects in which the same distribution (1b) came into being through the loss of CVL in more recent times. The dialects providing such evidence are found in the subdivision which is currently labelled Northern Romance (see Map 2 in Appendix), to which in }3.4 I will proceed, in concluding the present chapter, and which the following chapters will further discuss.

3.2 In defence of Open Syllable Lengthening in modern Standard Italian In order to proceed further in my diachronic reconstruction, it is first necessary to take a closer look at the synchronic system of (Tuscan-based) Standard Italian. The characterization of VL as determined by an allophonic rule, as given in commenting on (1a) and (2) above, has been contested from two complementary perspectives in synchronic analyses of the phonetics/phonology of modern Standard Italian. If such analyses were on the right track, this would have remarkable repercussions on the present argument: as a consequence, they have to be discussed here. On the one hand, Saltarelli (1970a,b) maintained that minimal pairs such as [’fa :to] ‘fate’ 6¼ [’fat :o] ‘done’ or [koɱ’vi :to] ‘banquet’ 6¼ [koɱ’vit :o] ‘boarding school’ contrast underlyingly by the length of the stressed vowel, whereas stress placement and consonant gemination, occurring both word-internally and at word boundary by raddoppiamento fonosintattico (cf. }4.2) are derived by rule. That a postlexical rule such as 9

A link between the rise of OSL and the loss of CVL in Late Latin has been assumed by many, previously, including e.g. Schuchardt (1866–8: 3.44), Grandgent (1907: 77), Bonfante (1956: 354), Weinrich (1958), and Tekavčić (1972: 1.22).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

66

The development of VL in Romance

raddoppiamento is derived is obvious: but this view on lexical gemination is a major departure from established linguistic scholarship, which regards Italian as a textbook example of contrastive consonant gemination (e.g. Laver 1994: 437). While Saltarelli does not address the diachronic implications of his analysis, the assumption of CVL—obviously, derived from PRom OSL—makes Standard Italian a variety of type (1c), like the Northern Italo-Romance dialects to be discussed in }3.4.1.10 Saltarelli’s (1970a; 1970b) analysis did not find many followers (though a similar view is reiterated by Burzio 1994), due to evident shortcomings both of a phonological and a phonetic nature, as discussed in Bertinetto (1981: 115–18, 146). The idea that VL rather than stress is specified underlyingly is at odds with the non-occurrence of the VL contrast in word-final position; this is to be compared with its occurrence in systems which do display CVL, considered from }3.4.1 onwards. Furthermore, Saltarelli’s assumption that consonant gemination is derived by rule (in the context ’V_) leaves unexplained the occurrence of contrastive geminates before stress (e.g. [ka’m :ino] ‘walk.1sg’ 6¼ [ka’mi :no] ‘fireplace, chimney’). As for the phonetic manifestation of length, there is abundant evidence that consonant length distinctions are signalled more prominently in Italian than VL ones. Also perceptually, Bertinetto’s (1981: 146) experimental evidence shows that consonant length is much more salient for Italian speakers/listeners than VL, and is better suited to serve as the phonetic cue for an underlying contrast. While Saltarelli’s (1970a; 1970b) proposal nowadays belongs, so to speak, to the history of the discipline, another potential objection to the reconstruction defended here can be derived from a more current line of research. In fact, there are recurrent claims that there is not actually conclusive evidence for an OSL rule operating in Standard Italian—or, to rephrase it for the last time in OT terms, not evidence for a highly ranked structural constraint such as Acc-μμ forcing systematic bimoraicity of all stressed syllables. Were it indeed so, it would be misguided to build a reconstructive argument upon a synchronic regularity that does not actually exist in the first place. That Italian indeed displays no OSL has been claimed among others by Luschützky (1984: }}10–11) and, more recently, by McCrary (2002; 2003; 2004), McCrary Kambourakis (2007). The latter scholar, in her thorough experimental study of vowel durations in the productions of Tuscan speakers from Pisa, observed that vowel duration varies gradually, as a function of both the segmental nature and the number of the consonants following the stressed vowel. Consider for instance Fig. 3.1 (after McCrary 2003: no. 25, McCrary Kambourakis 2007: 148):

10 This is symmetrical to analyses of northern dialects which posit underlying consonant gemination, such as that of Montreuil (1991) (cf. }4.1.1). Under Saltarelli’s (1970a; 1970b) view, Standard Italian would resemble even more closely those NItRom, Franco-Provençal and Romansh dialects which display gemination after contrastively short stressed vowels, to be discussed in (27)–(30), (75), and }}5.3.1–3 below.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In defence of OSL in modern Standard Italian

67

Stressed Vowel Duration before Segments in Coda vs. Onset 250 200 150 100 50 0

V/

T

V/

l

V/

n

V/

C in coda

134

156

159

167

C in onset

171

189

195

212

r

FIG. 3.1 Stressed vowel duration before consonants (T = plosive) in open and closed syllables in Standard Italian (McCrary Kambourakis 2007: 148).

The graph shows that vowel duration increases gradually depending on the following consonants: stressed vowels are shortest before stops and longest before rhotics, both in a closed and in an open syllable. Figure 3.2, on the other hand (no. 18 in McCrary 2003, fig. 4.8 in McCrary Kambourakis 2007: 146), shows that if one measures the average duration of vowels in open stressed syllables followed by a singleton consonant (leftmost box), of those followed by a tautosyllabic consonant cluster (complex onset; the central box), and of stressed vowels before a heterosyllabic consonant cluster (i.e. in closed syllables; the box on the right-hand side), one comes up with a gradient durational increase, with considerable overlap in duration between the three categories, rather than a clear-cut binary contrast between the first two categories (i.e. vowels in stressed open syllables, whatever the complexity of the following onset) and the last one (vowels in closed syllables). 240 220

Units

200 C

180

tauto CC

160

hetero CC

140 120 100

Vowel Duration (ms)

FIG. 3.2 Vowel durations before singleton consonants, tauto- and hetero-syllabic clusters in Standard Italian (McCrary Kambourakis 2007: 177).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

68

The development of VL in Romance

From these experimental results, McCrary (2003: 15) (= McCrary Kambourakis 2007: 177) concludes that ‘[t]he conditioning factors [ . . . ] are segmental, contrastbased conditions’ and that ‘[s]yllable structure is not implicated in these phenomena’. In this view, stressed vowel durations are exhaustively determined by the durational trade-off between stressed vowels and the following consonants: this trade-off is trans-syllabic, and its existence is argued to disprove that syllable structure plays a role here, under the form of OSL. The fact that stressed vowel duration in Italian, at the phonetic surface, does not display a neat complementary distribution (long vs short) but rather a finegrained continuum is in itself not surprising and has been known for a long time. This is shown in (6), which reports data from an experimental study by Fava and Magno Caldognetto (1976) (stressed vowel durations in milliseconds are indicated below each segmental skeleton, where C = consonant, V = vowel, T = plosive, S = sibilant, N = nasal, R = trill, L = lateral): (6)

’CVCV 208.4 >

’CVTRV 184.1 >

’CVRTV 177.6 >

’CVLTV 121.7 >

’CVSTV 112.7 >

’CVNTV 98.6 >

’CVC :V 85.3

This gradient, however, is not in itself conclusive proof that OSL does not exist: and indeed, Fava and Magno Caldognetto (1976) did not regard their results as disconfirming the existence of OSL. McCrary’s conclusion crucially depends on the model adopted, which is phonetically grounded OT, an output-oriented model that conflates phonology and phonetics (cf. e.g. Kirchner 1997; Flemming 2001). Consider, however, the more conservative view of the phonological component displayed in (7) (cf. e.g. Kiparsky 1985): (7)

a.

/underlying representation/

b.

postlexical (allophonic) processes

c.

d.

e.

derived phonological representation (or ‘categorical phonetic representation’, Keating 1990: 324)

phonetic implementation constraints

acoustic constraints

[output]

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In defence of OSL in modern Standard Italian

69

This model—one of the conceivable instantiations of what Bermúdez-Otero (2007: 502) dubs the ‘classical modular feedforward architecture of phonology’—differentiates between postlexical allophonic processes (7b) and phonetic constraints (7d): the former are part of the phonology and hence cognitive rather than physical in nature (Pierrehumbert 1990: 379), and operate on discrete symbols like phonological feature-values, syllables, and syllabic constituents, etc.; phonetic constraints (7d), on the other hand, are gradual in nature and operate in terms of ‘continuous values over the many simultaneous physical parameters’ (Keating 1990: 324).11 The same basic idea underlies the model of Natural Phonology (cf. Stampe 1979; Dressler 1984)—for this model, in (7) one would have to add prelexical processes, shaping underlying phoneme inventories (an aspect immaterial to our present concerns). In this view, phonological processes are motivated by phonetic constraints, but do not reduce to them. As Dressler (1984: 31) puts it, Anderson (1981) attacks a straw-man who would reduce phonology to its phonetic basis, e.g. phonological cover features of a specific language [ . . . ] to phonetic features measured experimentally.

Phonetically grounded OT is this kind of straw-man, though within OT as well there are multi-layered models of the phonetics–phonology interface which resemble (7) in providing for a principled distinction between phonological and phonetic representations. A case in point is Boersma’s (1998; 2007) BiPhon model, which provides for the five representation levels seen in Fig. 3.3:

phonological representations phonetic representations

lexicon

|underlying form| /surface form/ [auditory form]

phonetics–phonology interface

[articulatory form]

FIG. 3.3 The five levels of representation of Boersma’s (2007: 1) BiPhon model.

11 There are lots of implementation constraints on vowel duration which have been uncovered by research in experimental phonetics, and will play an important role in the present discussion, including e.g. the effects of coarticulation with adjacent segments (e.g., lengthening before voiced consonants: cf. Jespersen 1932: 182; Fowler 1991; Maddieson 1997: 625; Piroth and Janker 2004: 102–3), those of stress (stressed vowels being longer than unstressed), those of syllable structure (cf. Maddieson 1985), as well as the intrinsic durational differences correlating with vowel height, with lower vowels intrinsically longer than higher ones (cf. Jespersen 1932: 181; Lehiste 1970: 18–19; Maddieson 1997: 623–4). None of these must necessarily determine regularities at deeper levels of grammar ((7a–b)), though all of them of course may. Some authors call the gradual operations taking place at (7d) ‘phonetic rules’, with a terminological symmetry with respect to ‘phonological rules’ (cf. Bermúdez-Otero 2007: 502).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

70

The development of VL in Romance

In this model too, there is strucural room for a principled distinction between (the effects of) allophonic rules—whatever they are called within OT—and phonetic implementation. For expository simplicity, though, I will stick by (7) for the present discussion. The layering in (7) allows for the traditional (e.g. Trubetzkoyan) distinction of phonetics and phonology (cf. Trubetzkoy 1939) which, though rejected by approaches to phonology like the one McCrary adheres to—ultimately traceable to the ‘no-interface’ view of phonetics/phonology authoritatively advocated by John Ohala (e.g. Ohala 1990)—12 is current in many models and analyses, and is applied to both synchrony and diachrony. To quote a few examples, Blevins’ (2004: 38) theory of sound change assumes distinct levels of representation, labelling them traditionally as /x/ vs [y] (e.g. /ut/ ! [yt]). Or consider the following description of vowel duration and (phonological) VL in English by Odden (2011b: 465): Because of the influence of voicing on vowels—they have shorter duration before a voiceless obstruent—English has a cline of four durational patterns illustrated in bead, beat, bid, bit. Phonetic duration does not automatically translate into phonological length, and durational differences, especially ternary oppositions, are not necessarily of a continuous type (1, 2, 3, 4 units), but may instead reflect multiple, intersecting phonological phenomena.

Back to Romance VL, under a model such as (7), the Italian facts in (6) (and McCrary’s findings in Figures 3.1–3.2) can be interpreted as follows. First, allophonic OSL applies, deriving lengthened stressed vowels in open syllables, which is then encoded in the intermediate representation in (7c) (variously labelled in the literature, e.g. as ‘derived phonological’ or ‘categorical phonetic’ representation). Then, in production, coarticulation between sounds in the speech chain intervenes as schematized on the left-hand side in (7d), so that the contrast in length becomes blurred at the surface, and the continuum in (6) eventually emerges, which can be measured in the signal. On the perception side, on the other hand, acoustic contraints operating on ‘acoustic parametric representation’ (Keating 1990: 324) (or ‘auditory form’, Boersma 2007) allow the listener to trace back the phonetic input to the higher-level representation (7c). The layering in (7b-d) allows one to make sense of still other factors which influence vowel duration. In fact, recent research in experimental phonetics has shown that durational differences depending on segmental interactions seen in Figures 3.1–3.2 and in (6) do not account for the entire spectrum of variation in vowel duration in Italian. These segment-driven adjustments, in fact, cumulate with the effect of stress, which is more complex than presented introductorily in (2a-b) above, following the traditional description of VL in Italian. This has it that lengthening applies in (8a)

Cf. e.g. Browman and Goldstein’s (1990: 411) plea for a model of phonetics/phonology in which ‘the importance imparted to such dichotomies is abandoned’ (cf. also the objections by Myers 2000: 246, n. 2) and, more recently, Ladd’s (2011) critique of the concepts ‘phone’ and ‘systematic phonetics’. 12

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In defence of OSL in modern Standard Italian

71

while vowels occurring in all the remaining environments stay short (cf. (8b–d), where the relevant vowels are underlined): (8)

a. b. c. d.

’CV.CV]PW ’CVC.CV]PW ’CV]PW V . . . ’V . . . ]PW

[’ka :sa] [’kas :a] [bon’ta] [ka’set :a] [ka’s :et :a]

‘house’ ‘crate’ ‘goodness’ ‘little house’ ‘little crate’

As shown by Stevens (2012: 309–11), based on her own experimental study and on discussion of recent literature, the elsewhere case in (8b–d) is actually internally differentiated in a structured way. In particular, stressed vowels in closed syllables ((8b)) last consistently longer than unstressed vowels ((8d)):13 in Stevens’ corpus of controlled speech, for instance, the average duration is of 88 ms for the former and of 61 ms (closed syllable) and 62 ms (open syllable), for the latter, with a difference between the two categories just above the perceptual threshold (or JND). Thus, in addition to the fine-grained variation in Figures 3.1–3.2 and (6), there seems to be at least a three-way duration contrast. The issue is whether these phonetic facts have to be encoded in the phonology, perhaps under the form of a ‘three-way mora count distinction’, as proposed for Italian by Hajek (2000: 115)—followed, with modifications, by Stevens (2012: 312)—14 or whether, given a model such as that shown in (7), the phonetic three-way duration contrast can be accounted for at the phonetic implementation level, distinct from phonological structure. I think nothing prevents us, in principle, from exploring the second possibility. Of course, stress has an impact on segmental duration even in languages where no allophonic rule of OSL such as (3) is at work: in fact, vowels in stressed syllables have been shown to be systematically longer than vowels in unstressed syllables also in Romance languages of type (1b) like Spanish (Ortega-Llebaria and Prieto 2007: 162) or Romanian (Renwick 2012: 172), and Occitan dialects lacking CVL (cf. Müller and Martín 2012: 154, 156–7 on the Vivaroalpenc dialect of the Ubaia valley). Thus, purely phonetic lengthening of stressed vowels must arise at some point in the phonological derivation, and since this is a good candidate for a very general 13 The duration increase documented in the individual experimental studies ranges from 37.6% (Pickett et al. 1999) to 88.6% (McCrary Kambourakis 2007), with an average, according to Stevens’ (2012: 309) calculations, of 59.65%. There is also a significant effect of syllable structure for unstressed vowels, with vowels in open syllables lasting consistently longer (Dell’Aglio et al. 2002: 55). 14 Hajek (2000: 117) proposes the representation in (i) for stressed vowels in open syllables:

(i)

p

σ

σ

μ μ μ

μ

a

n

e

pane ‘bread’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

72

The development of VL in Romance

phonetic constraint, one is tempted to place it in (7d). Under this view, in Italian, stressed vowels in open syllables are targeted by both an allophonic OSL (at (7b)) and a lower level, gradual phonetic lengthening which affects stressed vowels independently of syllable structure (at (7d)), whereas in contemporary Spanish or Romanian the duration of stressed vowels is affected only by the latter.15 A similar distinction between ‘the longer duration of stressed syllables’ determined by cross-linguistic general phonetic factors, and the change that brought about OSL in early Modern German, is drawn by Page (2007: 340). According to his hypothesis, the latter may have arisen as listeners reanalysed the purely phonetic lengthening of short stressed vowels in open syllables as a phonological process, which implied allotting two moras to the stressed vowel. In other words, the change implied, in Ohala’s (1989) terms, a hypocorrection through which listeners failed to correct for coarticulatory lengthening, and which eventually resulted in recategorization. Crucial evidence in support of the classical modular architecture of phonology in (7) and, more specifically, for the view of Italian VL just expounded, comes from diachrony. To see how, it suffices to consider virtually any one of the syllablerelated sound changes reported in handbooks of Romance historical linguistics, like /a/-fronting in (Old) French (9a) or /ε/-diphthongization in (Old) Tuscan (9b): (9)

a. /a/-fronting in (Old) French

/a/

CLAVEM

LABRA

PARTEM

SALTUM

PASTAM

PLANTAT

PASSUM

clef

lèvre

part

saut

pâte

plante

pas

‘key’

‘lip’

‘part, side’

‘jump’

‘dough’

‘plant.3sg’ ‘step’

b. /ε/-diphthongization in (Old) Tuscan /ε/

HERI

PETRAM

PERDIT

CELS(AM)

VESTEM

CENTUM

TERRAM

ieri

pietra

perde

gelso

veste

cento

terra

‘yesterday’

‘stone’

‘lose.3sg’

‘mulberry’ ‘dress’

‘hundred’

‘earth’

The standard account of such changes—and of an endless series of comparable ones, attested throughout Romance (cf. e.g. (88)–(89) below, or Ch. 5, (6) and (109)– (111))—implies that there was an allophone lengthened via OSL in the first place, and that this allophone (or its diachronic successor)16 underwent the change while the

Not only stress, but also syllable structure affects the fine-tuning of vowel duration in Romanian, as shown by Renwick (2012: 171–5) (cf. }5.4.1, (99)): the magnitude of the open syllable effect, though, is much smaller than in Italian, as expected if Italian has an allophonic OSL rule, as distinct from lower level, universally valid, phonetic constraints, which are the only factors conditioning (at level (7d)) the duration of vowels in Romanian. 16 This qualification is necessary for /a/-fronting in Old French, which was actually the fronting of an /a :/ whose length had been phonologized before, as illustrated in the Proto-French system in (66) (cf. also }3.4.3.1 for a similar point concerning Central Ladin). 15

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In defence of OSL in modern Standard Italian

73

non-lengthened one remained unaffected. The examples in (9a–b), which contain the Latin etyma as well as the Romance outcomes and are displayed in the same order as the decreasing stressed vowel durations along the continuum in (6), show that stressed vowels not occurring in an open syllable were not affected, whatever the consonant cluster (or geminate consonant) following them. To this, one has to add that unstressed vowels too normally remained unaffected: cf. French palais ‘palace’ < palatium, OFr. lavez ‘wash.2pl’ < lavatis (not **levez) vs. lef ‘wash.1sg’ < lavo (Berschin et al. 1978: 136)—whereas je lave in modern French is due to analogical reshaping, and, conversely, extension to pretonic position is sporadically observed (OFr. meniere, alongside manière ‘manner’)—and Italian venire ‘come’ < venire (not **vienire), as opposed to viene ‘come.3sg’ < venit. Thus, unstressed vowels patterned uniformly with stressed vowels in closed syllables as the elsewhere case for those—and a host of other—sound changes. In other words, application vs non-application of sound change reveals a binary choice which can be described accurately if one assumes that OSL was operating in those varieties prior to change.17 On the other hand, these elementary generalizations about sound change would be missed under the conflated view of phonology–phonetics. If vowel duration really depended exclusively on segmental coarticulation effects, and on the phonetic gradients they determine, then the statement of the changes in (9a–b) could not make reference to either syllable structure or to a lengthened allophone. And no sensible alternative is in sight. Clearly, the nature of the following sound does not play any role here, as shown by the fact that in (9b) diphthongization applies in ieri, not in terra. Thus, the only possibility left would be to assume that speakers, at a certain point, applied colouring or diphthongization to just those stressed vowels whose actual phonetic duration was, say,  165.4 milliseconds.18 This is inconceivable, however, since experimental phonetic results show that there is an overlap in absolute durations across different contexts and, besides, that duration is contingent upon speakers, speech rate and style.19 This in no ways detracts from McCrary’s

17 This is a kind of argument standardly used in work in phonetics which does not deny the phonetics vs phonology distinction: cf. e.g. Lehiste (1970: 45) on the analysis of long consonants in Finnish, which ‘have, at least at an earlier stage of the language, been true geminates and have contained a syllable boundary. The evidence for this is primarily historical’. Cf. also Bermúdez-Otero (2007: 503) for the point that sound change provides crucial evidence for the classical modular feedforward architecture of phonology. 18 This solution would parallel, for vowel quantity, the set of constraints assumed by Kirchner (1997), which introduce into the phonology of vowel quality direct reference to phonetic substance replacing e.g. the feature specification [+high] with the ‘abstract value’ [–V(owel)h(eigh)t > 33] etc. Even more directly, Boersma (1998: 280) builds formant values into acoustic faithfulness constraints (e.g. F1 > 600 Hz) selecting over candidates such as ‘550 Hz’, ‘600 Hz’, and the like. 19 The influence of speech rate on segmental duration is a well-investigated topic (cf. the early study on Italian by Bertinetto 1980 and, most recently, the review of previous literature in Stevens 2012: 47–9). For Italian, Pickett et al. (1999) observe variation in absolute durations as a function of speech rate, and conclude that the duration ratio between the stressed vowel and the following (singleton or geminate) consonant provides a reliable acoustic cue of the C vs C : contrast, as it ‘remains stable across speaking rate

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

74

The development of VL in Romance

account of the phonetics of stressed vowel duration in modern Standard Italian. Yet, there is no genuine case here against OSL: the phonology of vowel length cannot be reduced to phonetics alone, as the evidence from diachrony eloquently reminds us. The same evidence also suggests that the phonetic gradient consisting in the decreasing durations across the environments in (8a–d) should not be written into the phonological representation. The output of change in this domain is always a binary contrast such as the occurrence of a long vs short vowel, or the application vs non-application of /a/-fronting (e.g. in (9a)), diphthongization (e.g. in (9b)), or the like. This is not to say that three-way phonemic VL contrasts are completely non-existent. Indeed, they do occasionally arise, as exemplified with the following triplet from Burgundy regional French (cf. Morin 1994: 144 and the previous literature—Taverdet 1989: 19—quoted there):20 (10)

Burgundy regional French a. [’kry] ‘raw.m’ < OFr. cru b. [’kryˑ] ‘believed.m’ < OFr. creu [kr@’y] c. [’kry :] ‘raw.f’ < OFr. crue [’kry@] ‘believed.f’ < OFr. creue [kr@’y@]

< crudum < *credutum < crudam < *credutam

In this Gallo-Romance variety, such a three-way contrast, which occurs for /y/ only, did not arise via one and the same change targeting three distinct stretches along one and the same pre-existing phonetic gradient (of the kind seen in (6) above). Rather, as shown by the etyma on the right-hand side, it precipitated from the convergence of different changes. Thus, the (seldom) occurrence of such more than binary contrasts supports, rather than disproves, the idea that phonological change is generally actualized by imposing binary contrasts on phonetic gradience. Note that what I have been arguing so far for allophonic features such as VL in modern Standard Italian can be generalized also to (the phonetic realization of) features/contrasts which are encoded underlyingly. Thus, there can be little doubt (see above, pace Saltarelli 1970a; 1970b; Burzio 1994) that consonant gemination, not VL, is contrastive in Italian—and hence must be encoded in (7a)—and that it has been so ever since Proto-Romance. Yet, much like VL, also the surface realization of consonant gemination is context-sensitive, as geminates, for instance, last longer after stressed than after unstressed vowels (cf. Stevens 2011: 27; Dmitrieva 2012: 128). and speaker’ (Pickett et al. 1999: 156). This C/V ratio, thus, seems to be another candidate to replace the syllable-based generalization in (9a–b). One should first calculate this ratio taking into account not just geminate vs singleton consonants but rather consonant intervals in general, along the lines of studies on speech rhythm such as Ramus et al. (1999). Whether or not it is feasible to model sound changes such as those mirrored in (9a–b) in this alternative way is an empirical question: the difficulty of the task is not to be underestimated, though, especially in view of the fact that a much simpler alternative is available, based on syllable structure and OSL. 20

Three-way VL contrasts are rare cross-linguistically: cf. Ch. 1, n. 1.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The eastern and western peripheries of Romance

75

Here too, the fine-tuning of duration will be a matter of surface implementation constraints (i.e. at (7d)), which one can reasonably assume to be at work in language after language. Under certain circumstances, however, like those discussed in }5.3 below, gemination can become an allophonic process: in such systems, thus, consonant length will be specified not at (7a) but only at (7b).21 The same seems to have happened to VL in the transition from CL to Proto-Romance.

3.3 The eastern and western peripheries of Romance The changes in the distribution and status of VL responsible for the options in (1a-c) are intertwined with a series of other sound changes that feature in the list of criterial isoglosses distinguishing major Romance subareas. Two of them, which I shall touch upon repeatedly in what follows, are the degemination of Latin geminate consonants, and apocope. The former applied in all Romance languages except the Italo-Romance varieties spanning Tuscany and the Centre–South of the Italian Peninsula. The northern border of this area preserving phonemic geminates is part of the bundle of isoglosses called, in the wake of von Wartburg (1936; 1950), the La Spezia–Rimini line which separates—in Wartburgian terms—Western from Eastern Romània. In addition to degemination, another isogloss concurs (among others) to form this line, viz. lenition of intervocalic singleton stops, not shared by Daco-Romance, as shown schematically in (11): (11)

Spanish French

Italian

Romanian

Latin < sapone ‘soap’22

a. jabón

savon

sapone

sapun

lenition

b. copa

coupe

coppa

cupa

degemination < cuppam ‘goblet’

Comparison of (11a,b) shows that, in Western Romance, lenition co-occurred with degemination in a chain shift.23 While Martinet (1955) conceived of it as a 21 Actually, in a language like Standard Italian consonant gemination is determined by specifications at three different levels, since it is not only, as just described, lexically specified ((7a)). It also arises as a product of a postlexical phonological rule, raddoppiamento fonosintattico (cf. }4.2, (18)): this clearly has to be taken care of at (7b) in categorical terms (plus/minus application), to be then adjusted at the surface like any other phonological material. Indeed, Marotta (1986) has shown experimentally that initial consonants lengthened via raddoppiamento fonosintattico last longer if they immediately precede main stress, and decrease in duration as this becomes more distant. 22 This word is an ancient Germanic loan (Germ. *saipo¯n), which entered spoken Latin during the Empire (cf. A. Castellani 2000: 42). 23 There are sparse exceptions to either (11a) or (11b) in some WRom. dialects: thus, in the so-called ‘Elcock zone’ straddling the Pyrenees, between Aragón and Gascogne, one finds many instances of nonvoiced obstruents: e.g. in several dialects of Béarn [ma’ty] ‘ripe’ < maturum, [ka’pano] ‘hut’ < capannam, [ur’tikos] ‘nettle’ < urticas (cf. Elcock 1938: 63, 87, 112). Aragón also hosts dialects for which preservation of geminate sonorants has been described (see }5.3.4 below).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

76

The development of VL in Romance

push-chain, the change seems more appropriately defined as a drag-chain, since lenition is demonstrably older. In fact, the philological evidence discussed in Politzer (1951) and Campanile (1971: 394–6) shows that intervocalic lenition was already at work in Latin. The problem with lenition is that the harbingers of the Romance process are documented almost everywhere in the Roman Empire, not only in the West but also in the East, as shown in (12): (12)

a. Lusitania: imudavit (= immutavit ‘borrowed, took away’) CIL II 462, from Augusta Emerita (Mérida, Spain), before 400 ad; b. Pannonia Inferior: extricado (= -ato, a consul’s name.abl) CIL III 3620, from Kovács (near Aquincum), 217 ad; c. Egypt: audem (= autem ‘but’) CPL 237.II.5, from Fayoum, Neronian period; peccadis (for peccatis ‘sin.abl.pl’) CPL 45.b.1, from Karanis, before 115 ad; galliga (= gallica ‘Gaulish.f’) CPL 114.1, 130 ad.

In Eastern Romance, however, lenition did not eventually succeed, as shown in (11a) by the voiceless stop in Romanian săpun (and roată ‘wheel’ < rotam, foc ‘fire’ < focu(s), etc.) and Italian sapone (as well as ruota, fuoco, etc.). The explanation proposed by Cravens (1991) for this seeming contradiction is the following: Latin had an allophonic voicing process, just like American English voicing and flapping of intervocalic -t- which also emerges in graphical notation in misspellings such as equidable, heredical, extracted by Cravens (1991: 59) from written productions of students of his, and which he compares with Late Latin epigraphic evidence such as that in (12).24 This process, then, was phonologized in the West, leading to restructuring, but was lost in the East. The loss of allophonic obstruent voicing in Daco-Romance provides the blueprint of a possible argument to accommodate the total lack of reflexes in Ibero- and Daco-Romance of the original (by hypothesis) Late Latin OSL. Its fate may well have been of the same kind: if degemination applies in a phonological system in which OSL is in operation, two scenarios are conceivable a priori: either the rule output (the lengthened vowel) is preserved in spite of the opacity of the environment in (formerly) CVCiCiV vs CVCV minimal pairs; or else the rule is suppressed altogether. The former scenario would give rise to a phonemic VL contrast,25 whereas the latter would result in a generalized lack of allophonic variation of the kind observed throughout the documented history of Daco- and Ibero-Romance.26 In fact, even syllable-structure-sensitive processes which applied 24 Varvaro (1984: 17) also observes that voicing occurred in Latin ‘con evidenza a volte maggiore a volte minore in tutta l’area imperiale’ [‘with varying degrees of prominence all over the Empire’], which seems warranted by the evidence (pace Weinrich 1960). 25 This position was upheld in many analyses, especially, though not exclusively, in structuralist diachronic studies: cf. }5.2, (32b) below. 26 Another comparable case of rule suppression in Romance, as argued by Cravens (2002: 134), is the one concerning i-prosthesis before initial s+C clusters, which is widely attested in Late Latin and must have

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The eastern and western peripheries of Romance

77

very early during the medieval pre-literary period of most Romance languages, such as diphthongization of e˘ o˘, are not sensitive to syllable structure in either Romanian or Spanish. Consider the Romanian data in (13): (13)

i. Open syllable

ii. Closed syllable

a.

iapă ‘mare’ < equam, iarbă ‘grass’ < herbam, ţară ‘country’ piatră ‘stone’ < petram < terram, ORo. ţărmure ‘shore, bank’ < *tiérmine < termine

b.

fiere ‘gall’ < fele, ieri ‘yesterday’ < heri

fier ‘iron’ < ferrum, piept ‘chest’ < pectus

In Romanian, the difference between the two outcomes in (13a,b) /ja/ vs /je/, both from Latin stressed e˘, depends on the following vowel:27 through the application of metaphony, an originally non-differentiated diphthong */jε/ was raised before high vowels (cf. Loporcaro 2011b: 129). This also provides a relative chronology argument for the early application of diphthongization in Daco-Romance. In fact, while datings of e˘- (and o˘-) diphthongization to as early as the 3rd century ad (and hence its ascription to Latin) discussed in }2.2.2 are unwarranted, it is true that diphthongization of the low-mid Proto-Romance vowels /ε O/ stemming from Latin e˘- and o˘- must be dated to a fairly remote period of the Latin–Romance transition. For Tuscany, Arrigo Castellani (1961: 95) has gathered compelling evidence from the rendering of place names in Latin texts written under Longobardic rule. The evidence, combined with the relative chronology argument drawn from the non-application of diphthongization to the outcome of Lat. au (e.g. causa > [’kO :sa]), allowed him to date diphthongization to the first half of the 7th century ad. In Gallo-Romance, diphthongization of /ε O/ > /jε wO/ appears to have applied somewhat earlier than in Tuscan. As I said above while commenting on }2.2.2, (9), its placement in the 3rd–4th centuries ad in e.g. Richter (1934: 138, 155–6) and Straka (1953: 268) (still repeated in e.g. Zink 1999: 53 or Joly 1995: 33) is based ultimately on untenable arguments, but it is true that diphthongs are found in the earliest Old stretched throughout the Empire (with the likely exception of Britain; cf. Sampson 2010: 61), on the strength of the epigraphic evidence. This vowel prosthesis has permanently affected the word shape of Ibero- and Gallo-Romance varieties (cf. spatham > Sp. espada, OFr. espée > Fr. épée ‘sword’), with the exception of Walloon, where the prosthetic vowel was later deleted ([’spɒ :l] ‘shoulder’ < spat(u)lam), giving way to a phonosyntactic epenthesis rule: Liégeois [in si’pɒ :l] ‘a shoulder’ (see Stasse 2007: 406; thanks to Y.C. Morin for pointing this out to me). Prosthesis remains today as a living phonological rule only in Spanish, for whose speakers it still represents a genuine constraint on pronounceability (emerging in foreign language acquisition), while it has completely disappeared elsewhere, through different stages examined in detail in Sampson (2010: 53–145). 27 Outcomes of o˘ are not discussed here for simplicity, given the asymmetric vowel system of Romanian (cf. }2.4, (23)).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

78

The development of VL in Romance

French texts (e.g. buona < bo˘na, ciel < caelum in the sequence of St. Eulalia, late 9th century, cf. Hilty 2001: 62–3) so that diphthongization must have applied before that date (in the 6th century, according to Bourciez 1937: 94, or in the 7th, according to Schuchardt 1866–8: 1.101–2),28 and it is also widely held to have taken place before degemination (although for criticism of this traditional argument see Morin 2003), since geminate consonants prevented it, which would not have been the case had they already become singleton: cf. e.g. OFr. cesse ‘cease.3sg’ < ce˘ssat like feste ‘feast’ < *fe˘stam and unlike ciel ‘sky’ < caelum. Indeed, current datings of degemination (for obstruents; cf. }5.3.4 below on sonorants) oscillate around the 8th century (Richter 1934: 250; Politzer and Politzer 1953: 41).29 As for Ibero-Romance, as in Romanian, no contrast is observed in the duration of the stressed vowels in, say, lado ‘side’ < la˘tus and arado ‘plough’ < ara¯trum. These examples are adduced by Menéndez Pidal (1952: 44) in order to illustrate that CL VL is gone without residue, but they also show that Late Latin OSL—if present in the spoken Latin of Hispania as well—disappeared without leaving any trace in IberoRomance. In general, no traces of sensitivity to syllable structure are found, even in an old change such as the diphthongization of PRom /ε O/: (14)

i. Open syllable

ii. Closed syllable

a.

nuevo ‘new’ < novum puerto ‘haven’ < portum suelo ‘soil, ground’ < solum cuello ‘neck’ < collum huesped ‘guest’ < hospitem

b.

cielo ‘sky’ < caelum hielo ‘ice’ < gelum

hierro ‘iron’ < ferrum ciento ‘one hundred’ < centum

Diphthongization seems to have applied later in Ibero-Romance than in GalloRomance. While Menéndez Pidal (1986: 504) places it in the Visigothic period (5th–7th centuries), the analysis of graffiti on the Visigothic slates from the 6th–8th centuries has brought Herman (1995: 67) (followed by Velázquez Soriano 2004: 482–3) to conclude that there are no sure examples of diphthongization in those texts, since all cases of apparently corresponding to etymological e˘ occur in just one morphologically defined context (the participial suffix -entem, e.g. ualiente(s), slate 102), and may consequently be explained as analogical on 4th-class participles such as audientem. This analogy is attested elsewhere, and earlier: Herman mentions abiente[s] for habentes, occurring several times in the Tablettes Albertini (from

28 In Schuchardt’s view, diphthongization came to completion in the 7th c., and was still a shared phenomenon within a linguistically unitary Late Latin speech community. 29 According to Straka (1964: 237) and Zink (1999: 154), degemination took place at an earlier date (7th c.); for Bourciez (1946: }271) at a later one (9th c.).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The eastern and western peripheries of Romance

79

Vandalic Africa, late 5th century; cf. Väänänen 1965: 27). The analogical extension of ie from 4th-class verbs was enhanced by the extensive paradigmatic coincidence brought about by sound change, as habeo ‘have.1sg’ or ualeo ‘be strong.1sg’ came to be pronounced [’abjo], [’valjo], like [’au̯djo]. Other Late Latin texts from Spain show similar examples of -iente(m) (e.g. curriente on an inscription from Córdoba, 682 ad), whereas the earliest examples of diphthongs in contexts other than this participial ending, gathered by Menéndez Pidal (1986: 113–52), stem from as late as the 10th century. Given the distribution of diphthongization in (14), it is readily apparent that IberoRomance, like Daco-Romance, does not offer crucial evidence bearing on the chronology of the establishment of OSL. Thus, any discussion of Ibero-Romance from this angle is bound to remain speculative. As an example, consider Grandgent’s (1907: 77) claim: ‘In Spain all vowels [i.e. not only in open but also in closed syllables; emphasis original—M.L.] were apparently long: te̜ m ̄ pus, po̜ r̄ ta.’ This claim, which refers to the variety of Late Latin spoken in Spain around the 5th century, rests solely on the observation that diphthongization applied in (14ii) too. Along slightly different lines, Lapesa (1968: 56) assumes, like Grandgent, a generalized Late Latin OSL and specifies that En Hispania, estas diferencias de duración debieron de ser menores que en otras zonas de la Romania, pues la misma suerte han corrido ĕ, ŏ en pe˘-dem, no˘-vum, que en se˘p-tem, po˘r-tam: unas y otras han dado ié, ué (pie, nuevo, siete, puerta) [‘In Spain, these differences in duration [according to syllable structure—M.L.] must have been smaller than in other areas of the Romània, since the same fate befell ĕ, ŏ in pe˘-dem, no˘-vum as in se˘p-tem, po˘r-tam: both the former and the latter turned to ié, ué’].

While one can agree on the assumption that Spanish diphthongization indicates that lengthening must also have applied in closed syllables at a previous stage—as argued by Morin (2003: 137–8), dismissing analyses which consider a syllabification po.rta as a necessary precondition for Sp. puerta ‘door’ (Bourciez 1946: 142; Straka 1956: 198)—the relatively recent date of diphthongization in Spanish, compared with French and Italian, casts doubt on the relevance of this circumstance for the discussion of a phonological rule (OSL) which is independently known for sure to have applied, in other parts of the Western Empire, some five or six centuries earlier. It is quite possible that this innovation spread not only to Africa, Italy, and Gaul but also to Hispania, to then be lost before diphthongization set in. This scenario has been envisioned by Hilty (1969: 96), among many others, and, given what we know, it is in principle as possible as the opposite one, according to which the Iberian Peninsula (like the Daco-Romance territory) was never attained by OSL: So allgemein kann indes diese Tendenz nicht sein, da ihr doch Sprachen wie das Spanische und Rumänische mit normierten (isochronen) Vokalquantitäten so erfolgreich widerstehen [‘This tendency, however [i.e. OSL—M.L.], cannot have been so general, since languages such

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

80

The development of VL in Romance

as Spanish and Romanian resist it so successfully by equating the length of all vowels’]. (Weinrich 1958: 181)

The empirical inadequacy of Lausberg’s (1976: 207–8) similar claims on this point has already been discussed in }2.2.1 above. As for Weinrich (1958: 181), this contention is at odds with his asserting (throughout his book) that OSL was the Proto-Romance (‘gemeinromanisch’) cause of the demise of CL CVL, as illustrated by the following quotation, just one page after the previous one: Die Kürze des Tonvokals in lat. gŭla beispielsweise setzt sich nirgendwo in der Romania direkt fort, weil schon beim gemeinromanischen Kollaps der phonologischen Vokalquantitäten daraus ein Langvokal entsteht: gọla. ̄ An diese Länge nun schliessen die romanischen Sprachen an [‘The shortness of the stressed vowel in Lat. gŭla, for instance, does not persist directly anywhere in the Romània, because a long vowel arose here already in the process of the common Romance demise of CVL: gọla. ̄ From this length the development in the Romance languages sets in’]. (Weinrich 1958: 182)

From the above, it becomes clear that any conclusion regarding the issue of whether or not stressed vowels in Daco- and Ibero-Romance were originally affected by OSL must not be judged per se but rather for its degree of consistency with the more conclusive evidence provided by Late Latin (already scrutinized in Chapter 2) and the other Romance branches, to which we now turn.

3.4 Northern Romance The Northern Romance area is a subdivision of Western Romance, which spans the territory from the North Sea to the Apennines and, possibly, the Pyrenees (Appendix, Maps 2–3). To the core of Northern Romance one can surely assign French (with the Oïl dialects), Franco-Provençal, Rhaeto-Romance, and Northern Italian dialects, whereas the position of Occitan is controversial: for Lausberg (1976: 299), for instance, Occitan belongs to the southwestern Romània, together with Catalan, Spanish, and Portuguese, whereas for Zamboni (2000: 102–3) Northern Romance includes the Gallo-Romance varieties in their entirety; solomonically, Gsell (1996: 577) adds a parenthesized ‘North’ in listing, as components of Northern Romance, ‘(Nord-)Gallien und Oberitalien’. This classificatory issue is ultimately unsolvable, since clear-cut dialect borders notoriously do not exist, as maintained in a long series of studies from Schuchardt (1870) to Chambers and Trudgill (1980) and beyond. Any classification depends largely on the isoglosses selected as crucial: for my present purposes, I shall use the label ‘Northern Romance’ in the extensive sense, and in what follows it will turn out that this area coincides with the one in which secondary CVL of type (1c) is either observed today or can be argued to have arisen in the past, even if it is not preserved to this day: the fact that Occitan is the sub-area in which this secondary CVL is least attested (though with some exceptions) fits well with its diverging classification as either within or outside Northern Romance.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

81

By this definition, secondary CVL is one of the criterial isoglosses for the individuation of Northern Romance (see e.g. Videsott 2001: 151). Another isogloss, which is related both spatially and (under some analyses, as we shall see in }}4.1 and 5.2) structurally with secondary CVL, is the loss of final unstressed vowels, which concerns, in different forms and to different degrees, all the Northern Romance area including Occitan. A first exemplification is provided in (15), where Italian is added for comparison (Lausberg 1976: 262–5): (15) French Occitan Surselvan Italian

viginti vingt vint vegn venti ‘twenty’

canem chien can tgaun cane ‘dog’

rotam roue roda roda ruota ‘wheel’

octo huit uech otg otto ‘eight’

caballum cheval caval cavagl cavallo ‘horse’

Italian contrasts with the other languages in (15) in preserving final unstressed vowels (though merging -u and -o > -o), whereas all other languages delete non-low final vowels. Apocope is most radical in French, where also the outcome of Latin -a was eventualy lost. This reduction of the unstressed vowel system happened in two steps: in Northern Gallo-Romance, non-low vowels were first centralized and then deleted, a process completed around 700 ad (cf. Richter 1934: 243–6; Sampson 1980: 30; Herman 1996: 376). At this point, word-finally, French had a binary contrast  6¼ /@/, the latter regularly derived from -a, as well as from non-low vowels under specific conditions (e.g. in etymological proparoxytones: OFr. jue(f )ne < iuvenem). This is the situation mirrored in the earliest extant records, where all non-low etymological vowels have been deleted—cf. e.g. dreit ‘right’ < directum, nul plaid ‘no covenant’ < nullum placitum in the Serments de Strasbourg (ad 842), or mort ‘death’ < mortem in the St. Eulalia (c.880–900) 20, Krist 24 < christum; cf. Hilty (2001: 60–66)—except if this deletion would have resulted in a disallowed final consonant cluster, in which case the final vowel was retained as [@]. This was often written (e.g. veintre ‘win’ < vincere, diaule ‘devil’ < diabolum, St. Eulalia 3 and 4) but occasionally displayed vacillation (e.g. Karle/Karlo ‘Charles’, fradre/fradra ‘brother’ < fratrem in the Serments), which provides evidence for its realization as a centralized vowel. By similar reasoning, one can be sure that the diachronic successor of Latin final -a, notated or e.g. in domnizelle ‘maiden’ < dom(i)nicella (St. Eulalia 23) alongside pulcella ‘maiden’ < *pulicella (St. Eulalia 1)—both with the same suffix: -[’εl :@] < -ella—was pronounced [@] at that stage.30 This

Likewise, in the Serments one mostly finds for Latin unstressed -a(-) in the final syllable (e.g. cadhuna cosa ‘every thing’), but also or : suo part ‘his side’ < suam partem, fazet ‘do.prs.sbjv.3sg’ < faciat. 30

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

82

The development of VL in Romance

final vowel was then deleted in modern French between the 16th and 17th centuries (Zink 1999: 47, 181–3). None of the other NRom varieties was reached by the loss of -a,31 though the quality of this final vowel was variously modified in several areas, through fronting (e.g. to -[e] in Central Friulian, cf. Marcato 2008: 2752) or backing (e.g. in most of Occitan, cf. Ronjat 1930–41: 1.206), often resulting in reduction (e.g. -[@]/-[ɐ] in Romansh, cf. Grünert 2012: 127).32 By contrast, non-low unstressed vowels were deleted outright, and there is evidence that in other NRom varieties this deletion occurred later than in Old French.33 From Northern France, it expanded gradually and eventually reached most of Northern Italy, though sparing several peripheral dialect areas. Two major dialect subdivisions lacking apocope are Ligurian, in the west, and Veneto, in the east (with the exception of the Northern Veneto varieties in the province of Belluno, which did drop final unstressed non-low vowels: cf. Zamboni 1974: 56–7).34 Lack of apocope recurs also on the southern fringe of Northern Italy, in the Apennines: the consequences of lack of apocope and of its gradual spread for the reconstruction of the rise of CVL proposed here will be explored in }5.2 below. 3.4.1 Northern Italo-Romance In Northern Italy, CVL occurs in several varieties spanning the whole area from the Apennines to the Alps, and from Liguria to Friuli (see Appendix, Map 3 for an overview). This is not to say that this contrast is documented in every single dialect in this territory: there are gaps, the larger ones being represented by Veneto, Eastern Lombard, and Piedmontese, which generally lack CVL. In }}3.4.1.1–5, CVL is illustrated with some selected dialects from Northern Italy. It is important to stress that these dialects, while providing a first glimpse of the data to be discussed, do not exhaust the gamut of structural (micro)variation documented in this area. This finegrained variation will be tackled in Chapter 5, where it will be shown to provide useful clues for diachronic reconstruction. 31 Loss of final -/a/ does occur occasionally in Northern Italo-Romance, but this is always confined to specific lexemes/morphemes. Thus, in Old Milanese, alexandrine hemistichs like in la scrigiura divina ‘in the holy Bible’ (Bonvesin da la Riva, Vita Beati Alexii 32) show that the final -a of scriptura(m) has been dropped (cf. Salvioni 1911: 376), but this is a peculiarity of that derivational affix, whereas -/a/ is otherwise preserved in Old and modern Milanese (cf. }3.4.1.4, (46)). 32 See Delucchi (2012) for an overview of reduction processes affecting the outcome of final -a in many of the Romance varieties in which it was not deleted. 33 Apocope also spread beyond Northern Romance. It crossed the Pyrenees and became established in Catalan (Badía i Margarit 1984: 174), whereas Old Castilian went through a stage (between the 12th and the mid-13th centuries) in which so called ‘apócope extrema’ applied optionally (e.g. nieve/nief ‘snow’, adelant (e) ‘forward’; cf. e.g. Harris-Northall 1991; Pensado 2001; Wright 2003: 213), an option that was then lost in the last quarter of the 13th c. (Cano Aguilar 2008: 3068). 34 This is not necessarily a linear evolution: as discussed below in }5.2.2, Old Veronese had final vowel deletion as a synchronic rule sensitive to sentence phonetics, whereas the modern dialect has restored final unstressed vowels throughout, after the model of Venetian (cf. Zamboni 1974: 48).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

83

The choice of the varieties in }}3.4.1.1–5 for a first illustration is for both internal and external reasons. Structurally, the dialects mentioned here—which are localized in Appendix, Map 3—display prima facie cases of a VL contrast, which is ultimately due to the specific diachronic development of their vowel system.35 In addition, these dialects have been the subject of thorough investigation, specifically as far as both the synchronic status and the diachronic development of VL are concerned, resulting in several competing analyses. Discussion of those analyses will be postponed to Chapter 4, where the data presented in the present chapter will be referred to. 3.4.1.1 Cremonese I shall begin with the dialect of Cremona, in southern Lombardy, a transitional variety between Lombard and Emilian (Merlo 1960–61: 6; Lurati 1988: 494) which has the following vowel phonemes in stressed syllables:36 (16)

i iː

y yː e eː

u uː o oː

ø øː ɔ ɔː

ε εː a aː

As seen in (16), stressed vowels, may be either short or long.37 Vowel length in Cremonese is discussed by Oneda (1964: 8–9; 1965: 34), and Rossini (1975: 187–92), and further data are drawn, in what follows, from the entries of Heilmann and Oneda (1976) and Magri (1995). Consider the following (sub)minimal pairs: (17) a. [’ri :s] [’y :s] [’me :s] [’pe :s] [’sε :t] [’pa :s] [’na :s] [’la :k] [’na :t]

‘rice’ ‘custom’ ‘month’ ‘weight’ ‘be.2sg’ ‘peace’ ‘nose’ ‘lake’ ‘born’

6¼

b. [’ris] [’ys] [’mes] [’pes] [’sεt] [’pas] [’nas] [’sak] [’mat]

‘curl(y)’ ‘door’ ‘half ’ ‘worse’ ‘seven’ ‘step’ ‘be born.inf’ ‘sack’ ‘crazy’

35 This is not to say that alternative analyses have not been put forward, implying that those dialects do not possess CVL as a property of the underlying representation: such analyses will be addressed in }4.3. 36 Sanga (1993: 195; 1997b: 257) classifies Old Cremonese as Eastern Lombard. Present-day Cremonese, though, does not share any of the features characterizing Eastern Lombard, listed e.g. in Merlo (1960–61: 3), Bonfadini (1990: 47–8; 1993: 105–7), Sanga (1997b: 257–9) (on one of these, viz. the lowering of stressed /i/ and /y/, cf. }5.1.3, (18)–(19), below). 37 The system in (16) is given by Oneda (1964: 9). Maddalon and Miotto (1986: 147) report a different vowel system for Cremonese, with several asymmetries between long and short vowels: e.g. /y :/ vs /ø/ and /e :/ vs /ε/ but no */y ø : e ε :/. This is at odds with the data: cf. (17)–(21) and, in addition, [ni’sø :n] ‘nobody’ vs [’løm] ‘lamp’ (cf. Heilmann and Oneda 1976: 171, 203). Several contributions, to be discussed in }4.1.3, have been devoted to the analysis of the phonology of VL in Cremonese: e.g. Repetti (1992: 180), Prieto (1994: 92), Morin (2003; 2012), Loporcaro (2005a; 2007a).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

84

The development of VL in Romance

(18) a. [’pe :l] [’ka :r] [’pa :n] [’mO :l]

‘hair’ ‘dear’ ‘bread’ ‘pier’

6¼

b. [’pel] [’kar] [’pan] [’mOl]

‘skin’ ‘chariot’ ‘cloth’ ‘soft’

In the first member of these pairs, (17a)/(18a), the long vowel derives from a stressed vowel which occurred originally in an open syllable (e.g. [’pa :s] ‘peace’ < pacem), whereas the short stressed vowel in (17b)/(18b) stems from a PRom vowel occurring in a closed syllable (e.g. [’pas] ‘step’ < passum). This applies both to (17), where the stressed vowel is followed by an obstruent, and (18), where it is followed by a sonorant.38 Minimal or subminimal pairs are to be found also in paroxytonic words such as the following: (19) a. [’fa :la] [’sta :la] [’la :na] [’ve :der] [’po :di]

‘do.int.3fsg’ ‘stay.int.3fsg’ ‘wool’ ‘glass’ ‘can.1sg’

6¼

b. [’fala] [’stala] [’kana] [’veder] [’podi]

‘fault (in a cloth)’ ‘stable, cowshed’ ‘reed’ ‘see.inf’ ‘prune.1sg’

In Cremonese, as in most varieties to be considered in what follows, this kind of contrast is limited to the stressed syllable, a fact that in itself would be compatible with an analysis in terms of syllable cuts (cf. e.g. Vennemann 2000) rather than vowel quantity (see Uguzzoni 1974: 247–9; Uguzzoni et al. 2003).39 While this restriction is a fact for most NItRom dialects, and is explained straightforwardly if one assumes that CVL is the diachronic successor of OSL (which applied in PRom to stressed vowels), an important difference with respect to what has traditionally been described as a syllable-cut contrast—for languages such as German—is that VL in NItRom dialects occurs also in word-final open syllables, where there is no question of a syllable-structure conditioning: (20)

a. [’me :] [’le :] [’asε :] [’pO :] [’sta :] [’fa :]

‘my’ ‘she’ ‘enough’ ‘Po river’ ‘stay.inf ’ ‘do.inf ’

[’me] [’le] [’trε] [’pO] [’sta] [’fa]

‘me’ ‘there’ ‘three.f’ ‘then’ ‘stay.3sg ‘do.3sg

38 There are certain distributional restrictions, described by Rossini (1975: 189–91). For instance, vowels followed by [tʃ m ɲ] are always short, and there is no VL contrast before some consonant clusters: thus, stressed vowels before sonorant + obstruent clusters are always long (e.g. [’ma :rs] ‘March’/‘rotten.m’, [’ma :rsa] ‘rot’), whereas before /sC/ clusters the stressed vowel is long word-internally but not in the final syllable (e.g. [’pa :sta] ‘pasta’ vs [’past] ‘meal’). 39 Cf. Filipponio (2012: 83–5, 247–51) for a recent survey of the discussion on this point. Filipponio suggests that Bolognese might be the NItRom variety which has advanced furthest on its way to changing to a syllable-cut prosodic system.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

85

As argued by Martinet (1956: 75; 1975: 205) for Franco-Provençal, this shows that the contrast, though limited to stressed position, is really one of VL rather than a more general property of syllable structure.40 There are also reported cases, albeit rare, in which CVL has been extended to unstressed syllables (cf. }3.4.1.3, (37) and (38)), where, again, no difference in syllable structure can be spotted. In Cremonese, one can observe a series of further distributional restrictions which are widespread across NItRom. One such restriction is that, apart from word-final position (see (20b)), short vowels normally occur before voiceless obstruents, while long vowels tend to precede voiced obstruents: (21)

a. [’ro :da] [’spu :za] [’rO :za] [’pe :za] [’pa :ga] [’ne :gi]

‘wheel’ ‘bride’ ‘pink’ ‘scales’ ‘pay.3sg’ ‘deny.1sg’

6¼

b. [’rota] [’rusa] [’grOsa] [’pesa] [’paka] [’beki]

‘broken.fsg’ ‘red.fsg’ ‘big.fsg’ ‘rag, cloth’ ‘slap’ ‘peck.1sg’

This distributional restriction is for a diachronic reason: since in (21a) the vowel usually occurred in an open syllable from the outset, the following singleton obstruent was voiced, as illustrated above for Western Romance in (11a): e.g. rotam > [’ro :da], pacat > [’pa :ga]. Conversely, short vowels in (21b) precede voiceless consonants because these arose via degemination of PRom voiceless geminates which, given the chain shift commented on in (11), resisted voicing: e.g. siccam > [’sεka] ‘dry.fsg’, ruptam > [’rota] ‘broken.fsg’. However, the correlation between VL and voicing of the following obstruent is far from exceptionless synchronically, which is due to several factors. First, voiced geminate obstruents did occur in Latin, albeit much less frequently, and some arose anew via sound change in PRom. Stressed vowels preceding the (degeminated) successor of these original geminates were not lengthened, as shown in (22a) (cf. Rossini 1975: 27, 33): (22)

a. [’freda] ‘cold.fsg’ < *fred :a < frig(i)dam [’meza] ‘half.fsg’ < *mεd :za < mediam [’guba] ‘hunchback’ < *gob :a < *gubb(am)

b. [’fret] ‘cold.m’ [’mes] ‘half.m’ [’gop] ‘hunchbacked.m’

In (22b), on the other hand, it is shown that intervocalic voiced obstruents alternate with voiceless ones in paradigmatically related forms, where due to apocope the consonant has become final and has been subsequently subjected to final devoicing, a process which applied widely throughout Northern Italo-Romance CVL has not been established in this environment in some peripheral dialects of NItRom (cf. }}5.3.1, 5.4.2), which document a more conservative stage in the development of CVL, with respect to the dialects considered in this section. 40

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

86

The development of VL in Romance

and which I shall be mentioning repeatedly in what follows. These alternations thus provide prima facie evidence—as argued by Sanga (1988: 295), Morin (2003: 126; 2012: 131–4)—that VL does not automatically depend on the voice of the following consonant, in spite of the wide correlation fed by sound change, observed in (21a, b), which obtains in a large share of the dialect’s native lexicon because of sheer diachronic heritage.41 However, first of all, the minority cases in (22a) do also occur; and in addition, once the VL contrast had been established, it was extended to other environments, which resulted in a further blurring of the correlation between length and voice in (21a,b), as one also finds word-internal long vowels before voiceless consonants, as shown in (a), which contrast (sub)minimally with short vowels in the same context:42 (23)

a. [’pa :pa] [’tʃO :sa] [’mO :ka] [’pi :ta]

‘Pope’ ‘broody hen’ ‘grimace’ ‘irresolute person’

6¼

b. [’papa] [’Osa] [’mOkol] [(en) ’pit]

‘pap, mush’ ‘shoehorn’ ‘curse’ ‘(a) little bit’

In some cases, the unexpected /V : + C[–voice]/ string occurs in a loanword, such as the Italianism [’pa :pa], which escaped intervocalic p-lenition. This is evidence against the postulation of a productive constraint barring such configurations in the competence of native speakers of today’s Cremonese, and is also evidence for the vitality of the VL contrast, since this is projected from the native lexicon (as exemplified in (17)–(21)) onto loanwords. This happens unpredictably, as seen in (24a,b): (24)

a. [’na :no] ‘dwarf ’; cf. the synonymous [’na :n], vs [’nano] ‘child’ b. [’myzo] ‘muzzle’; cf. the synonymous [’my :s]

The word for ‘dwarf ’ (24a), like the one for ‘muzzle’ (24b), occurs in two variants, the one ending in a consonant, as required by regular sound change, the other with Italian-like final -o: however, among the latter one finds either contrastively long or short stressed vowels, unpredictably. While the examples adduced up to now displayed long vowels in either oxytonic or paroxytonic words, VL occurs in Cremonese also, if more restrictedly, in proparoxytones ((25a)) as well as in words that were proparoxytonic before apocope ((25b)): 41 A fortiori, no contextual determinant for VL can be spotted in minimal pairs in which the stressed vowels are followed by sonorants, such as those in (18). See }4.3 for further discussion of this point. 42 The data in (22a) are at odds with Repetti’s (1992: 180) statement that, in Cremonese, ‘in the underlying representation short vowels may be followed by either voiced or voiceless consonants but long vowels are always followed by voiced consonants’: this cannot be accurate, given the (sub)minimal pairs in (23a,b). There are indeed several such examples, a systematic source of which is e.g. strings containing monophthongized /O :/ < au, in which the following obstruent escaped intervocalic voicing. This happened in Cremonese (cf. e.g. [’O :ka] ‘goose’ < aucam, Morin 2012: 130), as in many other Western Romance varieties (cf. e.g. Benedetti 1995; Burdy 2006: 130).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance (25)

87

a. [’ta :vula] ‘table’ < tabulam, [’ma :zena] ‘grinder, grindstone’ < machinam, [kwa’re :zima] ‘Lent’ < quadr(ag)esimam b. [’a :zen] ‘donkey’ < asinum, [’ma :zer] ‘rettery’ (cf. SI màcero), [’ba :ver] ‘collar’ (cf. SI bàvero)

The vowel [a :] especially occurs as long in (etymological) proparoxytones, as shown in (25a). On the whole, it is much more common for proparoxytones to host a stressed short vowel ((26a)), whose shortness has been left untouched also after apocope ((26b)): (26) a. [’fregula] ‘heat, itch’, [’fabula] ‘tale’; b. [’ryvit] ‘rough’ < rugidum, [’strOlek] ‘fortune-teller’ < (a)strologum, [sal’vadek] ‘wild’ < salvaticum (for CL silv-) Short stressed vowels in (26a) are more in keeping with the general tendency observed across NItRom for stressed antepenults to be short (e.g. in Milanese or Friulian), though with some exceptions, notably in Emilian (cf. }5.4 and Filipponio 2010: 24; 2012: 63–4). Thus in this respect, Cremonese appears more similar to its southern neighbour than to Lombard dialects. 3.4.1.2 Emilian This similarity, indeed, extends to other aspects of the phonology which are relevant to our present discussion. In particular, the dialects of Emilia, like Cremonese, display CVL not only in the final syllable but also in the penult, as illustrated with examples from the dialect of Bologna in (27) and (28) (cf. Coco 1970: 88; Hajek 1997: 135): (27)

a. [’me :l] [’sa :k]

(28)

a. [’fa :ta] [’la :ɲa] [’me :ter]

‘honey’ ‘sack’ ‘done.f’ ‘whine’ ‘metre’

6¼

b. [’melˑ] [’sakˑ]

‘thousand’ ‘dry’

6¼

b. [’fatˑa] [’laɲˑa] [’metˑer]

‘slice’ ‘wood’ ‘put.inf’

As apparent in the transcriptions in (27) and (28), an important difference with respect to Cremonese lies in the fact that several dialects of Emilia, especially as one approaches the Apennines in the southern part of the region (Appendix, Map 6), display at the surface a consistent durational distinction not only on stressed vowels but also on the consonants following them, with a complementary distribution. This is further exemplified for the dialect of Benedello (province of Modena; from Uguzzoni 1974: 241): (29)

a. [’re :d] [’pe :z] [’tø :t] [’pO :s] [’ma :t]

‘network.pl’ ‘weight’ ‘take.prs.int.2sg’ ‘can.1sg’ ‘crazy’

6¼

b. [’red :] [’pez :] [’tøt :] [’pOs :] [’mat :]

‘laugh.1sg’ ‘worse’ ‘all’ ‘pit’ ‘put.2sg’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

88 (30)

The development of VL in Romance a. [’me :la] [’bO :ta] [’pa :na]

‘honey’ ‘blow’ ‘cream’

6¼

b. [’mel :a] [’bOt :a] [’pan :a]

‘thousand’ ‘barrel’ ‘pen’

Since consonants are lengthened exclusively after short stressed vowels, this is usually regarded as an allophonic feature, whereas the phonological contrast is one of VL.43 Indeed, being context-dependent, consonant length, though described, is not notated in the transcription in much descriptive work on Emilian, such as Coco (1970: xxxi).44 The interplay between vowel and consonant duration in Emilian and in other dialects at the periphery of Northern Italo-Romance will be addressed in }}5.3.1–3, where it will be shown that the change from phonemic consonant gemination to phonemic vowel length proceeded gradually, through several steps which are still visible today in some dialects. Note that, while the examples in (27)–(30) illustrate the occurrence of a VL contrast synchronically, they do not directly mirror the application vs non-application of OSL to one and the same PRom stressed vowel, since many dialects of Emilia (and Bolognese in the most extreme form: cf. Filipponio 2012: 276–86) underwent a vowel shift. Thus, for instance, stressed /a :/ vs /a/ in the Bolognese examples in (28) stem respectively from PRom /a/ and /e/, both in closed syllables, as shown in (31b) and (32b) respectively (cf. Coco 1970: 3–5, 10–12): (31)

(32)

Outcomes of PRom /a/ a. ’CVCV

b. ’CVCCV

[’mε :r] < mare ‘sea’, [’lε :der] < latro ‘thief ’, [’pjε :z] < placet ‘please.3sg’

[’va :ka] < vaccam ‘cow’, [’la :t] < lacte ‘milk’, [’bra :s̪] < brachium ‘arm’

Outcomes of PRom /e/ a. ’CVCV

b. ’CVCCV

[kan’dai ̯lɐ] < candelam ‘candle’, [’nai ̯v] < nivem ‘snow’, [’sai ̯d] < sitim ‘thirst’

[sa’jatˑa] < sagittam ‘thunderbolt’, [’krasˑ] < crescit ‘grow.3sg’, [’sˑak] < siccum ‘dry

43 This lengthening is not as homogeneous over different prosodic structures as mirrored in the transcriptions in (29) and (30): see }5.3.3, (75a) below for experimental measurements on the dialect of Benedello. 44 Coco (1970: 113) notates half-length (in e.g. [’melˑ] ‘thousand’, [’metˑer] ‘put.inf ’) only when focusing explicitly on it, but omits its notation in the remaining phonetic transcriptions of Bolognese materials provided throughout the book. There are, however, several studies describing this gemination in Bolognese (e.g. Coco 1971: 165–7; Hajek 1995; 1997).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

89

Of course, the differentiation in the outcomes of PRom stressed /a/ in (31a,b) and /e/ in (32a,b) is evidence for the PRom OSL, since the different changes have affected allophones which must have been distinct: for stressed /a/, the lengthened allophone of PRom /a/ (! [a :]) was fronted to /ε :/ in open stressed penultimate syllables ((31a)),45 whereas it remained unchanged in closed syllables ((31b)). Later, this /a/ also underwent lengthening, a process that affected low and low-mid vowels in the history of Bolognese. PRom /e/, on the other hand, gave rise to a long /e :/ in open stressed word-internal syllables ((32a)) which underwent diphthongization. On the other hand, the same PRom vowel became contrastively short in closed syllables, and later lowered to /a/ ((32b)). As a final output, the vowel shifts of the kind exemplified in (31) and (32)—in Bolognese and across Emilian—resulted in a synchronic situation in which not all vowel qualities occur as both long and short. Thus, in Bolognese /i :/ and /u :/ lack a short counterpart (Coco 1970: 112; Kaze 1992: 331), whereas in the dialects of central Frignano exemplified in (29) and (30) four vowel qualities occur as contrastively short vs long, viz. /e( :) ø( :) a( :) O( :)/ and the remaining five are only long: /i : y : ε : o : u :/ (cf. Uguzzoni 2000: 227).46 3.4.1.3 Genoese The dialect of Genoa, which is representative of a situation that obtains widely (though not without exception, cf. }5.1.2, (3) and (4)) across Liguria, has the following stressed vowel system (Ricciardi 1975: 60):47 (33)

i iː

y yː e eː

u uː o oː

ø øː ε εː

a aː

Contrastive VL in Genoese is exemplified in the following minimal or subminimal pairs (data on Genoese VL are drawn from Ageno 1957: 12ff.; Forner 1975: 50; 1988: 458; Ricciardi 1975: 60–69; Toso 1997a: 16, 26, and passim): (34)

a. [’pO :su] [’pu :su] [’fa :su] [’va :zu] [’la :gu] [’ri :ku] [’fi :tu] [’fry :tu]

‘relax.1sg’ 6¼ ‘wrist’ ‘false’ ‘vase’ ‘lake’ ‘Henry’ ‘quick, soon’ ‘fruit’

b. [’pOsu] [’pusu] [’fasu] [’mazu] [’vagu] [’riku] [’fitu] [’brytu]

‘can.1sg’ ‘pit’ ‘do.1sg’ ‘May’ ‘go.1sg’ ‘rich’ ‘rental’ ‘ugly’

45 Arguably, this happened after VL became contrastive: thus, palatalization affected a phonemically long /a :/. 46 The asymmetry in the inventory of long vs short vowels is discussed for Bolognese in Kaze (1992: 331) and for the Romagnolo dialect of Galeata (province of Forlì) by Baroni (1993–4: 13–17). 47 Forner (1988: 458) also gives eight vowel qualities, but with /œ O/ instead of /ø o/.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

90

The development of VL in Romance

Note that, as these examples show, Ligurian dialects—unlike the other NItRom varieties dealt with in the present section—did not undergo generalized apocope of non-low final unstressed vowels, so that (sub)minimal pairs of shape ’CV :C vs ’CVC are extremely rare. They occur only in words like the following, where a final vowel (-/e/ or -/o/) has been deleted after a nasal consonant:48 (35)

a. [a’sε :ŋ] [’sε :ŋ]

‘steel [to light fire]’ ‘be.cond.3pl’

6¼

b. [’frεŋ] [’bεŋ]

‘brake’ ‘well’

This distinguishes Ligurian from the other NItRom varieties reviewed in this chapter. On the other hand, in Genoese, as elsewhere in NItRom, stressed vowels contrast by length also word-finally, which confirms the lack of contextual determination of VL already seen in (34): (36)

a. [’da :] [dur’mi :] [’di :] [pa’pε :]

‘give.inf’ ‘sleep.inf’ ‘say.inf’ ‘paper’

6¼

b. [’da] [’mi] [’ti] [ka’fε]

‘give.3sg’ ‘I’ ‘thou’ ‘coffee’

A peculiarity which singles out Ligurian from the rest of NItRom is that the VL contrast occurs in unstressed position as well, before stress, as seen in the following Genoese pairs (cf. Forner 1975: 50; Ricciardi 1975: 66–7): (37) a. [ka :’seta] [spe :’dʒeti] [afi :’ta] [i :’sa :] [fry :’ta] [to :’metu]

‘little sock’ ‘mirror.dim.pl’ ‘tan.inf’ ‘lift.inf’ ‘benefit.inf’ ‘alert child’

6¼

b. [ka’seta] [spe’dʒeti] [afi’ta] [fi’sa :] [by’ta] [o’metu]

‘little ladle’ ‘spectacles’ ‘let/rent.inf’ ‘stare.inf’ ‘throw.inf’ ‘man.dim’

This is not only a feature of Genoese but occurs in other Ligurian dialects, as seen in the following examples from the variety of Compiano (province of Parma; Vitali and Rulli 2010: 10):49 (38)

a. [vu :’ta :] [a :’sa :] [ka :’setu]

‘turn.inf’ ‘lift.inf’ ‘little sock’

6¼

b. [vu’ta :] [a’sa :] [ka’setu]

‘vote.inf’ ‘steel’ ‘drawer’

48 Final vowel deletion in Genoese took place also after singleton -r- and -l- (cf. Toso 1997b: 31), but these consonants were deleted themselves in turn, so that they do not occur word-finally today: e.g. [’ma :] ‘sea’ < mare = ‘badly’ < male. 49 The occurrence of contrastively long vowels also before stress in the Ligurian dialects of the upper Val di Taro (province of Parma) is described in Petrolini (1983: 232f.). The same happens in Tabarchino (cf. e.g. [ka :’setˑa] ‘little sock’, [fa :’detˑa] ‘little rock’, Toso 2005: 38, 48).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

91

The diachronic sources of CVL both under stress and in unstressed position are diverse. Let us begin from the latter context in which, obviously, OSL is not involved, since this applied only under stress. Much as in Gallo-Romance (}3.4.2), VL was fed by vowel coalescence (39a) and compensatory lengthening for the loss of a coda consonant (39b) (Parodi 1902–5: 159–60):50 (39) a. Vowel coalescence: [gwa :’ɲa :] ‘earn.inf’ < Germ. *waidanjan [sko :’sa :] ‘apron’ < Germ. *skauz+alem (rew 7986) [go :’di :] ‘enjoy.inf’ < gaudere (*-ire) b. Coda C deletion and CoL: [ka :’seta] ‘little sock’ < *calce(am)+ittam [sa :’ta :] ‘jump.inf’ < saltare [aso :’da :] ‘weld.inf’ < *(ad)sol(i)dare Clearly, what happened in Ligurian is that the long vowels that had arisen from the processes in (39) were retained in spite of their being unstressed. This, by the way, offers a powerful argument against an anlysis of VL in NItRom in terms of syllable cuts (cf. }3.4.1.1), since there is no syllable structure contrast correlating with the VL distinction in (37a,b), (38a,b), nor is VL dependent on stress. Even more surprising, from a cross-linguistic perspective, are the data in (40a) (cf. Toso 1997a: 43, 59, 92f.), which evidence the rise (via CoL) and maintenance of long vowels as a consequence of the loss of an onset, rather than coda, /l/: (40)

a. [u :/a :/i :/e : ’pOrtu] [u :/a :/i :/e : ’maŋdu ’vi :a]

‘I bring him/her/them.m/f ’ ‘I send him/her/them.m/f away’

b. [u/**u : ’kaŋ], [i/**i : ’kε :ŋ] [a/**a : ’ka :za], [e/**e : ’ka :ze]

‘the dog.sg/.pl’ ‘the house.sg/.pl’

c. [u ’fraŋku u/**u : ’veŋ du’maŋ] [sta ’dOna a/**a : l ’ε so ’sø :]

‘Franco is coming tomorrow’ ‘this woman is his sister’

Note that the forms of the definite article (40b) and the subject clitic (40c), which share the same etymon as the DO clitic (i.e. (il)lum, (il)lam, etc.), contrast minimally with the latter in displaying a short vowel in Genoese.51

50

CoL for /l/-deletion, occurring in (39b), applied to stressed vowels when followed by /l/ + coronal consonant (Ageno 1957: 13, Toso 1997b: 34). 51 In other Ligurian varieties, DO clitics display no lengthening and are homophonous with subject clitics and definite articles (e.g. in the dialect of Compiano, cf. Vitali and Rulli 2010: 11, 17, 23). The kind of derived VL observed for Genoese in (40a) has been documented also in ItRom dialects otherwise lacking contrastive VL, such as Romanesco (cf. Loporcaro 1991; 2007c), where deletion of onset /l/ has brought about lengthening in both the definite article and the DO clitic, bringing about contrasts such as [a : ’ro :ma] ( /la ’roma/) ‘the A.S. Roma football club’ vs [a ’ro :ma] ( /a ’roma/) ‘in Rome’.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

92

The development of VL in Romance

What has been said about the diversity of the diachronic inputs to long vowels before stress can be repeated for the stressed long vowels in (34a), (35a), and (36a). Yet, in many cases such as the following, there is still a recognizable relation between VL and PRom syllable structure, in the sense that long vowels occur in originally open syllables whereas short vowels occur in originally closed syllables, as shown in the examples in (41), where etymological information is provided (Parodi 1902–5: 333; Ageno 1957: 12–16): (41) a. [’ka :za] [’va :zu] [’di :ʒe] [’ta :ʒe]

‘house’ < casam 6¼ b. [’pasa] ‘pass.imper.2sg’ < *passa ‘vase’ < vasum [’mazu] ‘May’ < mai(i)um ‘say.3sg’ < dicit [fi’niʃe] ‘end.3sg’ < fin(*isc)it ‘be silent 3sg’ < tacet [’naʃe] ‘be born.inf’ < nascere

The examples also show that, while in several instances long stressed vowels precede voiced (lenited) consonants and short vowels occur before voiceless (degeminated) consonants, this is not always the case (cf. e.g. [’mazu] ‘May’). The nonexistence of a categorical synchronic correlation between voice and VL is further illustrated by the following examples:52 (42) a. [’di :tu] [’krε :tu] [’tʃO :su] [’ε :se] [’dO :ta] [’pO :ku]

‘said’ ‘believed’ ‘closed’ ‘be.inf’ ‘twice’ ‘few’

b. [’dadu] [’vede] [’kaze] [’treze] [’vagu] [’digu]

‘dice’ ‘see.inf’ ‘fall.3sg’ ‘thirteen’ ‘go.3sg’ ‘say.3sg’

These all can be explained diachronically. The long vowels in (42a) arose mostly via coalescence or CoL (e.g. [’tʃO :su] ‘closed’ < clausum, [’di :tu] ‘said’ < dictum, where -c- was vocalized to [j], etc.; cf. Parodi 1902–5: 156–9). Short vowels in (42b), on the other hand, are sometimes the expected result of regular sound change, as is the case for [’treze] ‘thirteen’ (and [’duze] ‘twelve’, [’seze] ‘sixteen’), where both the stressed short vowel and the voiced consonant are expected given the originally proparoxytonic structure: tredecim > *’tred(e)ze > [’treze] (cf. Parodi 1902–5: 358, and Toso 1997b: 34, who dates the deletion to the second half of the 12th century). More often, however, short stressed vowels before voiced consonants occur in words whose development suffered some irregularities. For instance, Latin intervocalic -d- and -twere originally deleted in Old Genoese (cf. Parodi 1902–5: 361): Lat. videre > OGen. ver (Nicolas 1994: 578), which nowadays survives as a rural variant ([’vei ̯], Toso 1997a:

52 The contrast is also found before sonorants, where voice stays constant: [’falu] ‘shortcoming’ vs [’fa :lu] ‘do.inf it’ (Ricciardi 1975: 65).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

93

200), alongside urban Genoese [’vede]. Also [’kaze], which already occurs in OGen. (caze), is not a regular outcome of cadit (cf. OGen. caer/cair < cadere, Nicolas 1994: 551). In some cases there is evidence that shortening is secondary: e.g. [’digu] ‘say.3sg’, as opposed to OGen [’di :gu], which had a long stressed vowel, as guaranteed by the rhyme with nemigo ‘enemy’ (Parodi 1898: 101; cf. contemporary Genoese [a’mi :gu], [ne’mi :gu]). The regularity of the correlation between VL and PRom syllable structure, arising originally via OSL, was further blurred by several lengthening processes which affected stressed vowels before certain consonants: the list usually provided includes /g gw v z ʒ r/ and /ɹ/ or /ɾ/, according to dialects (cf. Ghini 2000: 183 on the dialect of Miogliola, province of Savona, and Toso 1997a: 16; Forner 1975: 51–2, 250–51 on Genoese), though only before /ʒ r/ is the lengthening really exceptionless. For instance, before /z/ one more often encounters long vowels, but there are nonetheless pairs such as Genoese [’mazu] ‘May’ vs [’va :zu] ‘vase’ (Toso 1997a: 16). 3.4.1.4 Milanese (43)

i iː

Milanese has the following stressed vowel system:

y yː ε eː

u uː ɔ ɔː

ø øː a ɑː

While descriptive sources differ in several details concerning the relationship between vowel quality and quantity,53 all descriptions (cf. Salvioni 1884: 42ff.; Nicoli 1983: 45, 49f.; Sanga 1984a: 60–64; 1988: 291–3; 1999: 141–7; Gökçen 1990) agree on the occurrence of minimal pairs contrasting by VL in the ultima, both in the presence (44) and in the absence (45) of a following final consonant:

53 See Gökçen (1990) for a recapitulation and comparison of previous descriptions. One point on which all agree is that for the mid front non-round vowels length correlates with tenseness: long lax [ε :] is an allophone of /ε/ occurring only before /r/ and liquid + consonant clusters (e.g. [’fε :r] ‘iron’, [’vε :rs] ‘(animal) cry, way’; cf. Sanga 1984a: 63), whereas short stressed [e] occurs very seldom, in loanwords (like [dʒi’le] ‘waistcoat’ < Fr. gilet) and verb forms with pronominal enclitics (like [’fel] ‘do.imp.2pl=it’; cf. Sanga 1984a: 63). Several descriptions observe a (purely allophonic) difference in quality, in terms of tenseness, for most other vowels, which is disregarded here. Sources also differ as to the extent to which they admit contrastive VL for individual vowels. Sanga (1984a: 63), for instance, does not admit of a contrast /ø :/ 6¼ /ø/, adducing the reason that only one minimal pair can be spotted ([’tø :] ‘your.mpl’ 6¼ [’tø] ‘take.inf ’). Note however that [ø :] and [ø] are both free to occur in the final stressed syllable, either open or closed: e.g. [’pø] ‘then’ vs [’bø :] ‘oxen’, [’øtʃ] ‘eye’ vs [’fø :g] ‘fire’, etc. It must be kept in mind that (as aptly stressed e.g. by Morin 2012: 124) minimal pairs are just a practical way to show that a given contrast is not constrained by the context (and hence, not allophonic). Thus, if the environment can be shown to be uninfluential, the lack (or paucity) of minimal pairs can be disregarded as a mere lexical accident.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

94

The development of VL in Romance

(44) a. [’na :z] [’ka :l] [’fi :g] [’tu :z] [’y :z] [’rø :d] [’me :z] [’pe :l] (45)

a. [an’da :] [’di :] [ri’dy :] [’tø :] [’pe :]

‘nose’ ‘loss’ ‘fig’ ‘boy’ ‘usage’ ‘wheel’ ‘month’ ‘hair’

6¼

b. [’nas] [’kal] [’rik] [’tus] [’ys] [’røt] [’mεz] [’pεl]

‘be born.3sg’ ‘corn’ ‘rich’ ‘cough’ ‘door’ ‘burp’ ‘half ’ ‘skin’

‘went’ ‘said’ ‘laughed’ ‘your.mpl’ ‘feet’

6¼

b. [an’da] [’di] [ri’dy] [’tø] [’pε]

‘go.inf’ ‘say.inf’ ‘reduce.inf’ ‘take.inf’ ‘foot’

On the other hand, in all non-final stressed syllables (as well as in unstressed position), VL is non-contrastive. Thus, whenever the syllable hosting a contrastively long vowel ((44)) becomes non-final due to affixation, the stressed vowel ceases to be long, as shown in (46a) (in (46b), similar alternations involving contrastively short vowels are added for comparison):54 (46) a. Root /V :/ [’di :z]/[’dizi] [’dy :r]/[’dyra] [’pe :l]/[’pela] [’nø :f]/[’nøva] [’na :z]/[’naza] [’mO :d]/[’mOda]

‘say.3/1sg’ ‘hard.m/fsg’ ‘hair/peel.3sg’ ‘new.m/fsg’ ‘nose/sniff.3sg’ ‘fashion.pl/sg’

b. Root /V/ [’rik]/[’rika] [’myf]/[’myfa] [’mεz]/[’mεza] [’gøb]/[’gøba] [’gat]/[’gata] [’mOs]/[’mOsa]

‘rich.m/fsg’ ‘rotten.m/fsg’ ‘half.m/fsg’ ‘hunchbacked.m/fsg’ ‘cat/she-cat’ ‘moved.pl/sg’

In words of ’CVCV shape, vowel duration may be affected by the following consonant (cf. Gökçen 1990: 253). Crucial for our present purposes is the fact that no VL contrast is allowed to occur in this context. Neutralization of VL also occurs whenever the vowel becomes unstressed due to affixation (Sanga 1984a: 55–7): (47)

[’na :z] [’sak]

‘nose’ ‘sack’

! !

[na’za] [sa’kεt]

‘sniff.inf’ ‘little sack’

Stressed vowels, finally, are invariably short in proparoxytonic words (cf. Nicoli 1983: 47–58):

Note that tenseness distinctions, which accompany redundantly VL contrasts in stressed final syllables, become thereby distinctive in non-final ones (cf. Sanga 1999: 145): [’peza] ‘pitch’ vs [’mεza] ‘half.fsg’. 54

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance (48)

95

[’pegura] ‘sheep’, [’legura] ‘hare’, [’strOlega] ‘gipsy(f)’, [’nivula] ‘cloud’ (also in etymological proparoxytones like [’azen] ‘donkey’ < asinum)

For Milanese too, as for the other NItRom varieties discussed in the previous sections, a number of generalizations emerge. One is that short vowels in (44b) tend to be followed by voiceless obstruents, whereas long vowels in (44a) can be followed by voiced obstruents.55 The latter have been transcribed as such because final devoicing, applying more extensively in other NItRom dialects, is reported to occur only variably in today’s Milanese (Sanga 1984a: 52–3; 1988: 295), so that the realizations [’fø :g]/ [’fø :g̊ ]/[’fø :k] ‘fire’ co-occur.56 The reason for this correlation lies, once again, in the WRom. chain shift in (11). However, here too, this is not an absolute constraint, as observed by e.g. Sanga (1984a: 65) and Morin (2003: 130), so that it cannot be argued that VL depends on devoicing, either synchronically or diachronically, as made clear by examples like the following, to be added to [’mεz] and [’gøb] in (46b): (49)

[’frεtʃ]/[’frεdʒa] [’vεtʃ]/[’vεdʒa]

‘cold.m/fsg’ ‘old.m/fsg’

The independence of VL from consonant voice is confirmed by the occurrence of minimal pairs before sonorants, already seen in (44a,b). (The examples concern only /l/, for diachronic reasons.)57 On the other hand, a genuine generalization is that long vowels only occur in the final syllable, and that this is often a consequence of apocope: this crucial point will be addressed in Chapter 4. In its basics, the distribution of VL now described for Milanese is representative of the situation found in all the Western Lombard area (Appendix, Maps 3 and 4), as exemplified in (50) for the dialect of Premana (province of Como; cf. Sanga 1984b: 31–4; Bellati and Bracchi 2007: 67): (50)

a. [’ka :r] [’pε :s] [’fO :] [ves’ti :] [’kro :s]

‘dear’ 6¼ ‘weight’ ‘outside’ ‘dressed.m’ ‘cross’

b. [’kar] [’pεs] [’fO] [ves’ti] [’ros]

‘chariot’ ‘fish’ ‘do.1sg’ ‘dress.inf’ ‘red’

(dialect of Premana)

This correlation has been observed as early as Biffi (1606: 189–90). This seems to follow from a relatively recent undoing of an earlier change, as Salvioni (1884: 158) describes final devoicing as being categorical in Milanese: ‘ogni sonora mediana che venga a trovarsi all’uscita vi diviene sorda’ [‘any word-internal voiced consonant which comes to occur word-finally gets devoiced’]. 57 While -m- is out of the question, due to its systematic gemination throughout NItRom (cf. }5.3.4), /n/ is also problematic since, when singleton -n- became word-final via apocope, it was deleted with nasalization of the preceding vowel (e.g. [’kã :] ‘dog’ < canem), though in contemporary Milanese the nasal stop is being reintroduced due to the pressure of Standard Italian (Sanga 1999: 159). Final singleton /l/ and /r/ were deleted too (e.g. [’sa :] ‘salt’ < salem, [mi’e :] ‘wife’ < mulierem), but with exceptions. However, whenever spared, final /r/ has lengthened a preceding short stressed vowel, as seen in [’ka :r] ‘chariot’ < carrum, homophonous with [’ka :r] ‘dear’ < carum (cf. Salvioni 1884: 188; Sanga 1984a: 62). 55 56

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

96

The development of VL in Romance

The following alternations show that, much as in Milanese, in Premanese too VL is not contrastive in the penult (cf. Bellati and Bracchi 2007: 783, 993): (51)

[’pa :l] [’pale]

‘shovel.pl’ ‘shovel’

6¼ =

[’spal] [’spale]

‘shoulder.pl’ ‘shoulder’

(dialect of Premana)

Another example of contrastive VL from Western Lombard is provided in (52) for the dialect of Casale Corte Cerro (province of Verbania; cf. Weber Wetzel 2002: 31–4): (52)

a. [pu’li :t] ‘well’ 6¼ b. [pu’lit] ‘turkey.pl’ [’ta :z̥] ‘be silent.3sg’ [’tas] ‘badger’ [’vø :j] ‘empty’ [’vøj] ‘want.1sg’ [’ly :j] ‘he’ [’lyj] ‘July’

(dialect of Casale Corte Cerro)

While some details differ (e.g. in the dialect of Premana one finds minimal pairs for long vs short vowels preceding /r/, like [’kar] ‘chariot’ vs [’ka :r] ‘dear’ in (50), whereas in the dialect of Casale Corte Cerro /r/ induces lengthening of the preceding vowel, and hence those minimal pairs are neutralized into [’ka :r] ‘chariot’ = ‘dear’, as in Milanese (cf. n. 57 and Weber Wetzel 2002: 39, 169)), in all these dialects the VL contrast is restricted to the ultima and is neutralized in paroxytonic and proparoxytonic words. Weber Wetzel (pp. 35–7) describes the phonetic implementation of this neutralization in ’CVCV words as a degree of length intermediate between long (as in (52a)) and short (as in (52b)), regardless of whether the following consonant was originally singleton (53a) or geminate (53b):58 (53)

a. [’buˑnɐ] [’teˑlɐ] [’lyˑnɐ] [’seˑdɐ]

‘good.f’ (cf. fpl. [’bu :n]) = b. [’kaˑnɐ] ‘reed’ ‘cloth’ [’bεˑlɐ] ‘beautiful.f’ (cf. fpl. [’bεl]) ‘moon’ [’piˑɲɐ] ‘pine cone’ ‘silk’ [’gaˑtɐ] ‘she-cat’ (cf. pl. [’gat])

Together with that of Weber Wetzel, the other, relatively recent, most detailed description of the phonology of VL in a Western Lombard variety is by Bonfadini (1997) (based on data from Massera 1985) on the dialect of Novate Mezzola (province of Sondrio). Though this dialect borders on Valtellinese, where VL is retreating (see }5.1.3), it has contrastive VL with high functional load for seven of its eight vowel phonemes: (54)

i iː

y yː e eː

u uː

(dialect of Novate Mezzola)

ø øː ɔ ɔː

ε a aː

58 Vowels preceding /ts/ and /dz/, however, are categorically short: e.g. [’matsɐ] ‘club’ < *matteam, [’mεdzɐ] ‘half.f’ < mediam.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

97

This is exemplified by pairs such as the following, which show that CVL occurs in the last syllable, both word-internally and word-finally: (55)

a. [’li :s] [’gy :s] [’ka :l] [’pO :s] [’tu :s]

‘worn out’ ‘squirrel.pl’ ‘loss’ ‘positions’ ‘boy’

6¼

b. [’lis] [’gys] [’kal] [’pOs] [’tus]

‘smooth’ ‘shell’ ‘corn’ ‘stale’ ‘cough’

(56)

a. [’di :] [pru’dy :] [fa’zø :] [’sa :]

‘finger’ ‘produced’ ‘bean’ ‘salt’

6¼

b. [’di] [pru’dy] [gri’zø] [’sa]

‘day’ ‘produce.inf’ ‘blueberry’ ‘know.3sg’

The same situation is described for other nearby dialects in the western part of the province of Sondrio (excluding Valtellina, to the east, on which see }5.1.3): Samòlaco (e.g. [’pa :s] ‘peace’ ¼ 6 [’pas] ‘step’; Scuffi 2005: 291, 295), and Villa di Chiavenna ([’sa :] ‘salt’ ¼ 6 [’sa] ‘know.3sg’; Bracchi 2010: 29; Giorgetta and Ghiggi 2010: 603, 608). This distribution of VL, identical to that of Milanese, is described, furthermore, for other subareas of Western Lombard: for the variety of Comano, near Lugano (Bosoni 1995: 357), Mendrisio (Lurà 1987: 31), Balerna (Pifferi 2004), Tortona (Cabella 1999), Voghera (Maragliano 1976), Lecco (on the border with Eastern Lombard) (Bosoni 1995: 359; 2001: vi), etc. Thus, for instance in Vogherese one finds [’pa :z] ‘peace’ vs [’pas] ‘step’, [’na :z] ‘nose’ vs [’nas] ‘be born.inf’, [’pe :z] ‘weight’ vs [’pεs] ‘piece’, while phonemic VL does not occur in paroxytones: e.g. [’bala] ‘ball’ < Germ. balla has the same short vowel as [’pala] ‘shovel’ < palam (as opposed to [’pa :l] ‘post’ < palum). 3.4.1.5 Friulian Moving eastwards in northern Italy (see Appendix, Map 3), one finds large areas without contrastive VL (Eastern Lombard and Veneto, to be discussed in }} 5.1.3–4). Skipping here the Central Ladin area, to be addressed in }3.4.3.1,59 one finds VL again in Central Friulian (on which the standardized regional koiné is based), where it has a distribution very similar to what has now been illustrated for Western Lombard. This aspect of Friulian phonetics and phonology has been the subject of a number of studies, from several angles: alongside more traditional dialectological investigations (e.g. Francescato 1966: 130–43; Frau 1984: 18–30), one finds theoretically-oriented phonological analyses (Bender et al. 1952; Hualde 1990; Morin 1992; Repetti 1992; Vanelli 1979; 1998; Finco 2009; Iosad 2012) and studies in descriptive (Miotti 2002) and experimental (Baroni and Vanelli 2000; Finco 2007) phonetics. Drawing on this literature, I will sketch here some basic facts concerning the Central variety, before discussing and comparing different analyses in }4.1.2. Peculiarities of local varieties

59

See also }3.4.3 for a commentary on the choice of addressing Friulian here, rather than in that section.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

98

The development of VL in Romance

will be mentioned only sparingly here, whereas variation across Friulian dialects will be covered in more detail later on (}5.1.1). Central Friulian has seven vowel qualities, all occurring as contrastively short and long (cf. Finco 2007: 121):60 (57)

i iː

u uː e eː

o oː ɔ (ɔː)

ε (eː) a aː

The context in which all Friulian dialects with CVL show these contrasts is the by now familiar position in which the stressed vowel now occurs in a final closed syllable due (historically) to apocope: (58)

a. [’li :s] [’pe :s] [’mε :s] [’la :t] [’dO :s] [’vo :s] [’lu :s]

‘worn out’ ‘weight’ ‘my.fpl’ ‘gone’ ‘two.f ’ ‘voice’ ‘light’

6¼

b. [’lis] [’ves] [’pεs] [’lat] [’tOs] [’fos] [’lus]

‘smooth’ ‘have.impf.sbjv.1/3sg’ ‘fish’ ‘milk’ ‘cough’ ‘be.impf.sbjv.1/3sg’ ‘luxury’

In addition, descriptions of Central Friulian (cf. Frau 1984: 20, 77; Miotti 2002: 244, Finco 2007: 121) report the occurrence of the VL contrast in word-final position as well: (59)

a. [’di :] [mu’ri :] [’je :] [’ve :] [can’ta :]

‘say.inf’ ‘die.inf’ ‘she’ ‘have.inf’ ‘sing.inf’

6¼

b. [’di] [mu’ri] [’je] [’ve] [can’ta]

‘day’ ‘die.prt.3sg’ ‘it to her/him’ (dat+acc clitic) ‘there/here’ (presentative)61 ‘sing.prt.3sg’

These descriptions often add that the occurrence of long vowels in this position is rare (and always has been), being confined to verb infinitives and a few other lexemes. In addition, perhaps as a consequence of this rarity, the phonetic realization of length in this context is sometimes said to be confined to the peripheral dialects of the northern half of Friuli (Carnia) and to have been obliterated in Central Friulian (cf. Francescato 1966: 21–3; Vanelli 1998: 71, n. 5; Finco 2009: 55).62 Indeed, Finco’s 60 Low-mid /ε :/ and /O :/ are marginal according to Frau (1984: 19). Francescato (1959: 15; 1966: 10) gives a five-vowel system, with no contrast but rather complementary distribution between high- and low-mid (i.e. /ε e : O o :/). 61 E.g. in [’velu ’la] ‘there he/it is’ (Nazzi and Saidero 2000: 724). 62 The somewhat labile status of CVL in word-final position in Central Friulian is mirrored in the divergent transcriptions across the sources: thus e.g. feminine possessives are transcribed [’me :] ‘mine.f ’,

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

99

(2007: 129–30) experimental study, comparing two varieties of Central Friulian (those of San Daniele and Tarcento) with two of Carnico (those of Preone and Pradumbli), shows that in the latter the stressed /’a :/ in cjatâ ‘find.inf’ is consistently longer (351 ms. in Preone, 278 in Pradumbli) than that of cjatât ‘found.msg’ (312 ms. in Preone, 251 in Pradumbli), whereas the reverse is true in the Central dialects of San Daniele (cjatâ 278 ms. vs cjatât 286 ms.) and Tarcento (cjatâ 231 ms. vs cjatât 304 ms.). This confirms that vowel length in (59a) may be (slightly) less perceptible in Central Friulian dialects, but at the same time shows that the contrast with (59b) is maintained, since the duration of phonologically long vowels in absolute word-final position (in cjatâ) is sharply distinct from that of phonologically short /’a/’s in words like sofà ‘sofa’ (165 ms. in San Daniele, 151 ms. in Tarcento), which demonstrates that the contrast, in these Central dialects, is alive and well even in absolute word-final position ((59a) vs (59b)). In non-final syllables, on the contrary, VL is usually non-contrastive (but see (62) below). Thus, one observes paradigmatic alternations such as the following (see e.g. Baroni and Vanelli 2000: 14–17): (60) a. [’vi :f]/[’vive] [’fre :t]/[’frede] [’la :t]/[’lade] [’lo :f]/[’love] [’kru :t]/[’krude]

‘alive.msg/fsg’ b. [’frit]/[’frite] ‘cold.msg/fsg’ [’sεk]/[’sεce] ‘gone.msg/fsg’ [’lat]/[’late] ‘wolf/she-wolf ’ [’rOs]/[’rOse] ‘raw.msg/fsg’ [’brut]/[’brute]

‘fried.msg/fsg’ ‘skinny.msg/fsg’ ‘milk/breast feed.3sg’ ‘red.msg/fsg’ ‘ugly.msg/fsg’

Both contrastively long ((60a)) and short ((60b)) vowels alternate with vowels which are phonologically unmarked for length (qua non-contrastive) and usually transcribed as short. Actually, if one excludes contrastively long vowels, the remaining stressed vowels (a set comprising contrastively short as well as non-contrastive ones) are realized with different degrees of duration, as shown in the detailed phonetic description by Finco (2009: 57–8): (61)

Realization of phonemically non-long vowels in Friulian a. Extra-short [ă] i. word-final closed syllables, before voiceless obstruents: [măt] ‘crazy’, [trŎp] ‘much’ ii. proparoxytones before voiceless obstruents: [’lε̆tare] ‘letter’, [’scătule] ‘box’

[’to :] ‘yours.f ’, [’so :] ‘his/her.f ’ in Miotti (2002: 244) but [’me], [’to], [’so] in Frau (1984: 75). No such lability is ever reported in studies on the dialects of Carnia: cf. e.g. Trumper (1975: 9) on the Carnico of Chiusaforte ([’ʃta :] ‘stay.inf’ vs [’ʃta] ‘stay.3sg’ etc.), Leonard (1972: 66) on the dialect of Pesariis, and Finco’s (2007: 129–30) experimental measurements on the Carnian varieties of Preone and Pradumbli.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

100

The development of VL in Romance b. Short [a] i. in open penult before voiceless obstruents: [’vace] ‘cow’, [’strake] ‘tired.fsg’ ii. in open antepenult before voiced consonants: [’pOdine] ‘tub’, [’ɾεgule] ‘rule’ iii. word-finally: [a’mi] ‘friend’, [so’fa] ‘sofa’ iv. in final closed syllables before sonorants or /s/: [’maŋ] ‘hand’, [ca’val] ‘horse’, [par’tis] ‘leave.impf.sbjv.3sg’ c. Half-long [aˑ] i. in open penult before voiced consonants: [’caˑze] ‘house’, [’paˑle] ‘spade’ ii. before tautosyllabic sonorants or /s/: [’uˑltime] ‘last.fsg’, [’caˑrte] ‘paper’, [’veˑndi] ‘sell.inf ’, [’paˑste] ‘noodles’, [’frεˑsk] ‘fresh’, [’caˑns] ‘dogs’ iii. in the nucleus of a falling diphthong: [’fraˑI ̯de] ‘rotten.fsg’, [’joˑI ̯be] ‘Thursday’

Before closing this short survey of VL in Friulian, one has to add that it is not entirely accurate, in a synchronic description, to claim that contrastively long vowels are barred from the stressed syllable of paroxytones. In fact, long vowels do occur in paroxytonic words like the following (cf. e.g. Benincà 1989: 56–7): (62) a. [’po :re] ‘fear’ < *paúra . . . < pavor(em) b. [’pa :ri] ‘father’ < *padri . . . < patrem, [’vo :li] ‘eye’ < *ogli . . . < oc(u)lum, [so’re :li] ‘sun’ < *sorégli . . . < *solic(u)lum In these words, either monophthongization ((62a)) or the deletion of a following obstruent have occurred ((62b)),63 the latter being evidenced by the retention of this consonant in paradigmatically and/or etymologically related forms such as [pa’driŋ] ‘godfather’, [vo’gloŋ] ‘big eye’. Thus, diachronically it is clear that the rise of CVL in (62) has to be explained separately from that in (58a) and (59a). Indeed, many analyses of Friulian VL (e.g. Hualde 1990: 33; Vanelli 1998: 70) leave these cases out of consideration in analysing VL, because they treat the occurrence of length in examples such as (58a) as derived synchronically, rather than as underlying, which can hardly be the case for long vowels in (62). Yet, as will be argued in }}4.1.2 and 4.3, it seems preferable to analyse CVL in today’s Friulian as underlying. If this option is taken, it then becomes unjustifiable to refuse to acknowledge that CVL can occur also in paroxytonic words, such as those in (62).64 63 See e.g. Miotti (2002: 243), according to whom ‘the difference [in duration—M.L.] between cjase with [VV]) and pôre with [VˑV]) may be clearly heard in more conservative pronunciations’. 64 Vanelli (1998: 70–71, n. 4) reports that VL in (62) is not as widespread across Friulian dialects as that in (58a) (which she terms ‘canonical’), a claim which seems to be contradicted by the evidence presented in R. Castellani (1980: 89–91) (cf. Ch. 6, (12)). In addition, she claims that the long vowels in (62) ‘non danno mai origine a contrasti fonologicamente rilevanti’ [‘never give rise to phonemically relevant contrasts’]. While it is true that it is difficult to find minimal pairs (though cf. Ch. 6, (11) and (12)), in Friulian dictionaries one does come across words which could form near-minimal pairs with those in (62): e.g. [kon’trari] ‘contrary’, [’Ore] ‘hour’, [’sOlit] ‘solid’ (Faggin 1985: 212, 884, 1318).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

101

3.4.2 Gallo-Romance 3.4.2.1 Standard French and Oïl dialects Gallo-Romance is another subdivision of the Romance language family in which VL contrasts occur in several dialects. Within the Oïl subgroup of Gallo-Romance, Standard French is the only major Romance standard language for which CVL has been described, until a relatively recent date. This circumstance explains the difference in structure between the present subsection and the preceding ones: while in the discussion of Italo-Romance data in }}3.4.1.1–5 the primary focus was on synchronically observed contrasts, with remarks on their diachronic sources kept in the background, for Gallo-Romance not only synchronic dialect variation but also diachrony must be addressed from the outset. This necessity is readily apparent if one considers the recent development of VL in Standard French. In fact, in conservative late 19th-century Parisian French, as reflected for example in Michaelis and Passy (1914: xxii–xxiii), contrasts such as the following occurred (cf. Martinet 1959: 178; Morin 2006: 138): (63) a. gîte fête brûle goûte veule côte pâte

[’ʒi :t] [’fε :t] [’bʁy :l] [’gu :t] [’vœ :l] [’ko :t] [’pA :t]

‘dwelling’ ‘feast’ ‘burn.3sg’ ‘taste.3sg’ ‘weary’ ‘coast’ ‘dough’

6¼

b. frite faite nulle goutte veulent cotte patte

[’fʁit] [’fεt] [’nyl] [’gut] [’vœl] [’kOt] [’pat]

‘fried.f’ ‘done.f’ ‘invalid.f’ ‘drop’ ‘want.3pl’ ‘skirt’ ‘paw’

At that stage, these contrasts were residual: they were banned from the final position, where only short vowels occurred, and were already neutralized in more innovative varieties of Parisian French (Morin 2008: 2922).65 Michaelis and Passy (1914: 320) refer to this variation in their list of ‘différences omises’, where they mention, on the one hand, the ‘nivellement de toutes les différences de durée [Néologisme]’, and on the other the persistence of VL contrasts, even word-finally (e.g. [’bu :] boue ‘mud’ vs [’bu] bout ‘end’) qualifying it as ‘Archaïsme: Belgique, Suisse’. In the 20th century, neutralization progressed. The first investigation into the phonemics of modern French by Martinet (1933) still reports several among the contrasts in (63) as general features of the standard language (e.g. /ø :/ vs /ø/ in filleule ‘goddaughter’ vs filleul ‘godson’), whereas he later recognized that several such contrasts (like the one just quoted, originally arising via the change in (67v) below), rather than being general features of French phonology at that time, were typical of his own regional variety (Martinet was born in Savoy) and, even in his own competence/speech, they had disappeared by the 1940s (Martinet 1959: 171; 1990: 16). Martinet (1959: 184–6) also describes the fading, across age classes of Parisian French 65 Between the 16th and the 18th c., on the contrary, VL contrasts were also alive and well word-finally, as a product of the changes mentioned below in (67iii,iv).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

102

The development of VL in Romance

speakers, of the /a/ vs /A :/ contrasts in e.g. patte ‘paw’ 6¼ pâte ‘dough’, rat ‘rat’ 6¼ ras ‘smooth, shaved’. In sum, the development of the French vowel system, with VL contrasts gradually yielding ground, for the different vowel qualities, from the late 17th century onwards, can be schematized as follows (after Morin 2008: 2922, elaborating on Martinet 1959: 175–9; nasal vowels are omitted for simplicity): (64)

a.

late 17th century

b.

i iː y yː ʊ ʊː e eː ø øː ε εː ɔ ɔː a aː

transition

i iː y yː e eː øː ε εː œ a

mid 20th century

c. u uː oː

ɔ ɑː

i

y

u ø

e ε εː a

o œ

ɔ ɑ

By the 1940s the only contrast left in the standard was /ε :/ 6¼ /ε/, and even that was merged in the linguistic competence of nearly half the speakers surveyed by Martinet (1945) (cf. Walter 1976: 120–21) and further faded across age classes in the subsequent decades (Walter 1993: 215). Regional varieties of French as well as several Gallo-Romance primary dialects preserve VL contrasts much better to this day. The regional French of Wallonia, for instance, displays the VL contrast (cf. Walter 1982: 111–12), which has been preserved—within the linguistic repertoire of the same speech community—in Walloon dialects as well, as exemplified in (65) for the patois of Tenneville (Francard 1980: 277):66 (65) a. [’ki :] [’ly :] [’bu :] [’ba :ʃ]

‘summon.inf’ ‘shine.3sg ‘ox’ ‘canvas cover’

6¼

b. [’ki] [’ly] [’bu] [’baʃ]

‘who’ ‘he’ ‘end, tip’ ‘lower.3sg’

Much has been written on VL contrasts in French, and quite controversially so.67 For my present purposes, it suffices to observe that what is in focus in the present investigation is not what has attracted most interest in this area. To illustrate this,

66 Cf. Walter (1982: 110–13, 119–27, 130–38, 161–4, 194–7) on the Gallo-Romance varieties with VL contrasts in her survey. Several further Gallo-Romance dialects, including regional varieties of French (which are secondary dialects in Coseriu’s 1981 terms), display quality differences in the development of one and the same Latin/Proto-Romance vowel which can be interpreted as the diachronic successor of previous VL contrasts, as argued by Morin and Dagenais (1988: 155) with respect to cases such the three distinct outcomes of Latin stressed -a´- in the Picard patois of Mesnil-Martinsart (after Flutre 1955: 4), along with several others. 67 Morin and Dagenais (1988: 153) observe that the quarrel on the distinctiveness of VL has accompanied the whole history of French grammar, quoting controversies on the topic from Meigret (1550) vs Peletier (1555) up to Straka (1981: 183, n. 115) vs Martinet and Walter (1973). More recently, Gess (2001: 145) argued that in French ‘the length distinction [ . . . ] was lost by the 16th century’, a view radically opposed to the one synthesized in (64a-b).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

103

consider the following statement in Posner (1997: 288), where pairs like those in (63)—and their merger in contemporary French—are discussed, concluding: Today vowel length distinctions as such are not part of most French speakers’ internalized systems [ . . . ] Viewed sub specie aeternatis [sc. aeternitatis] the introduction and demise of vowel length was a blip in the history of French which might have been overlooked if it had occurred in a period when documentation was less abundant and interest in language less intense.

This ‘blip’ is the period from late medieval/early modern French up to the present, and the ‘intense interest’ is attested by the unceasing discussions among grammarians and linguists hinted at in n. 67 above. However, were the facts on VL in the history of French exhaustively described in this way, then there would indeed be no meaningful intersection with the topic of the present book, since none of the long vowels in (63) owes its origin to Late Latin OSL. What is involved here is, rather, a series of much later processes of CoL, like that applying in the late Middle Ages, compensating for the deletion of coda /s/ in OFr. giste > gîte, feste > fête etc. (from jacitam FEW 5.4, and festam respectively), and the like. However, Posner’s description of the fate of French VL leaves out completely a more remote stage, which can still be reconstructed on the strength of comparative GalloRomance evidence. In this previous stage, which is labelled as Proto-French, ‘duquel on peut faire provenir l’ensemble des parlers romans d’oïl’ [‘from which one can derive the whole of the Romance oïl dialects’] (Morin 2008: 2908), or even as ‘common GalloRomance’ (Berschin et al. 1978: 92), the phonological system must have possessed a general VL contrast, as shown in (66):68 (66)

i iː

u uː e eː

o oː ɔ wɔ

ε iε a aː

By the earliest documented stage of Central French the system had undergone several changes: among them, a neutralization of VL, [u] > [y], [o] > [ʊ] and the change of long [a :] to [e :] (Morin 2008: 2910–11). Phonetically, the only difference in

68 A phonemic system with CVL is assumed for Northern Gallo-Romance ( francien) by Haudricourt and Juilland (1949: 37). As seen in (66), this system corresponds neatly to the PRom seven-vowel system seen in Ch. 2, (15), with the addition of CVL and with the phonetic diphthongs [iε] [uO] still analysable as the realization of long /ε :/ and /O :/ respectively (Morin 2008: 2908). This boils down to the same underlying system assumed for the Romance variety of ‘Nordgallien’ in the stage after consonant degemination by Hilty (1969: 96): according to Hilty, open-syllable diphthongization must have followed the phonologization of CVL in Northern Gallo-Romance.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

104

The development of VL in Romance

length occurring at that stage must have been [e] vs [e :] (the latter from [a :], Wüest 1979: 154), no longer a distinction of VL for some, like Berschin et al. (1978: 91), who contest that one isolated contrast may be considered phonemic. Then, however, VL was resurrected in Middle French, and was described by almost all grammarians between the 16th and 18th centuries (Thurot 1881–3: 2.568) until the late 19th-century stage synthesized above in (63). This was the product of the subsequent rounds of vowel lengthening distinguished as follows by Morin (2006: 148–9), which came after OSL, i.e. the ‘first layer’ change responsible for phonemic VL in (66):69 (67)

i. First layer: ‘a general Romance allophonic process of vowel lengthening in open stressed syllables [ . . . ], traces of which are still observed in Walloon, Franco-Provençal and, more frequently, Alpine Provençal. . . . This early lengthening is the source of the diphthongization that developed in many Oïl, Romansh and Gallo-Italian dialects, e.g. de¯be˘t > Proto French [’de :vet] > OFr. deit [dei ̯t] / doit [du̯εt] “(he) owes”’. ii. Second layer: ‘medieval lengthening’ via coda consonant deletion (goste [’gʊ :t@] ‘savour.3sg’ < gustat), monophthongization (croule [’krʊ :l@] < *[’krOu̯l@] ‘crumble.3sg’ < *corrotulat) and a series of other sound changes. iii. Third layer: vowel coalescence (/’V + @/ > /’V :/), starting in Middle French and accomplished earlier in Northern and Eastern Oïl dialects, later elsewhere. iv. Fourth layer: after the 17th c., VL in Latinisms ending in -Vs: e.g. virus [vi’ʁy :s]. v. Fifth layer: 18th c. CoL for the loss of final -[@] resulting in a half-long vowel (in e.g. vile ‘vile.f’ [’viˑl] vs vil ‘vile.m’ [’vil], cruelle ‘cruel.f’ [kʁy’εˑl] vs cruel ‘cruel.m’ [kʁy’εl], according to some 18th-c. testimonies such as Mauvillon 1754; cf. Thurot 1881–3: 2.643).

The first layer in (67i), corresponding to Late Latin OSL, may be regarded as having given rise to a phonological VL contrast in the prehistory of French (or even of Gallo-Romance as a whole after degemination), under the assumption that ‘les differences de durée allophonique se maintiennent après la dégémination (ou la simplification des groupes consonantiques) et après l’apocope’ [‘allophonic length differences are preserved after degemination (or consonant cluster simplification) and apocope’] (Morin 2003: 121). This is schematized in (68):

69

Especially work by Yves Charles Morin over the last two decades or so has cast light on many aspects of VL in the history of French and Gallo-Romance dialects: in addition to the already quoted essays, see also Morin (1989; 1994; 2003; 2012).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

105

(68) pı˘lum

eccı˘llum te¯lam eccı˘llam

/’pelo/ [’pe :lo]

/et’tsello/ [et’tsello]

/’tela/ [’te :la]

/et’tsella/ [et’tsella]

a. Degemination

[’pe :lo]

[et’tselo]

[’te :la]

[et’tsela]

b. Apocope

[’pe :l]

[et’tsel]

[’te :la]

[et’tsela]

c. French diphth.

[’pei ̯l]

[et’tsel]

[’tei ̯la]

[et’tsela]

d. Old French

peil

icel

teile

icele

The relative chronology in (68) (Morin 1994: 146; 2003: 121; 2008: 2009) entails the assumption of the Proto-French vowel system in (66), in which all vowels, before the application of diphthongization (and a series of other changes affecting stressed vowels in open stressed syllables), could occur as distinctively either long or short. This view, also endorsed by e.g. Hilty (1969: 96) and Wüest (1979: 157), is alternative to that of Straka (1953: 291) and Zink (1999: 124, 154), according to whom ‘French’ diphthongization of PRom mid high stressed vowels (peil ‘hair’ < pilum, goule ‘throat’ < gulam) must have preceded apocope, and hence also degemination. Under that view there was no stage such as (68a,b), involving a phonological VL contrast, since diphthongization is assumed to have modified vowels which were still only allophonically lengthened before non-geminate intervocalic consonants. 3.4.2.2 The evidence from dialect variation Decision as to which reconstruction is more appropriate is not easy, but the arguments from the relative chronology of diphthongization and syncope pointed to by Morin (2003: 120–35) to back up the reconstruction in (68) seem compelling. In addition to relative chronology, a powerful argument in support of the phonologization of the VL contrast through degemination is provided by the fact that, sifting the Gallo-Romance dialect evidence, Morin (1994: 145; 2003: 132) succeeded in pointing to some residues of VL contrasts which can still be traced back to the first layer (67i) and, hence, to a phonologization of OSL of the kind seen in (68a). Since in most cases later changes have obscured these contrasts, they are not easy to spot in present-day dialects. One case in point is that of the outcomes of Latin -a´- before /l/, which in several Gallo-Romance dialects still display a VL contrast: (69) (70)

Tenneville Walloon (Francard 1980: 31–2) [’pa :l] ‘spade’ < palam 6¼ [’bal]

‘bundle, ball’ < Germ. balla

Le-Pont-de-Beauvoisin (Haut-Dauphiné Franco-Provençal: Devaux and Duraffour 1935: 168, 63, 222, 230) [’sa :] ‘salt’ < salem 6¼ [θi’va] ‘horse’ < caballum [’pa :la] ‘spade’ < palam [@’pala] ‘shoulder’ < spat(u)lam

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

106

The development of VL in Romance

While the examples in (69) stem from an Oïl dialect, those in (70) are from the Franco-Provençal domain, and no data have been mentioned so far from the third major subdivision of Gallo-Romance, Occitan. Now, leaving aside sparse cases where secondary VL contrasts occur, which are clearly unrelated to Late Latin OSL,70 it is indeed the case that Occitan is the GalloRomance subgroup which offers less evidence for CVL. This general lack of evidence has led many scholars to posit for Occitan a radically divergent development with respect to Oïl dialects and Franco-Provençal. Thus, in one of the masterpieces of structuralist diachronic phonology, Haudricourt and Juilland (1949: 35) defined the kind of VL in (67i), whose traces are still directly observable in contrasts such as those in (69) and (70), as the ‘corrélation de longueur vocalique francienne (ā:ă)’[‘the Francien vowel length correlation’].71 As for Occitan, on the other hand, Haudricourt and Juilland (1949: 48) explicitly exclude that the same might have been the case: ‘La situation devait être toute différente dans les parlers provençaux qui, n’ayant pas acquis de nouvelles correlations de longueur vocalique . . . ’ [‘the situation must have been completely different in the Provençal dialects which, not having acquired any new correlation of vowel length . . . ’]. This is a widely held opinion (e.g. von Wartburg 1950: 82, n. 1; Hilty 1969: 96; Zufferey 2008: 3000). However, while it is true that Occitan dialects generally lack contrastive VL and that Late Latin/PRom syllable structure, in most of them, left no traces in VL, it has been argued that in some peripheral varieties of Occitan, VL contrasts did arise, whose original diachronic nucleus can be traced back to OSL. In particular, Morin (2003: 127, 131–2) points to such evidence from the southeastern fringe of the Occitan domain, mentioning the patois of Breil and Briga/La Brigue, described in Dalbera (1994a: 125–9), spoken in the Roya valley a few kilometres north of Nice: these areas came under French rule in two rounds in 1860 and 1947, whereby the shift in state borders cut them off from the Ligurian dialects immediately neighbouring to the east (on the Ligurian Apennine, from Pigna to Ormea). These have the stressed vowel system in (71) (where nasal vowels and diphthongs are omitted for simplicity), exemplified with data from the patois of Breil in (72):

For instance, in the patois of Ambialet (dépt. of Tarn; cf. Maurand 1974: 111; Morin 1994: 147–50), deletion of a coda [ʃ] may result in a secondary contrast, e.g. in [la : r :ama’ʃãm] ‘we pick them.fsg up’ (with /laʃ/ ! [la :]) vs [la r :ama’ʃãm] ‘we pick it.fsg up’. This is reminiscent of the secondary contrast arising from coda /s/-deletion in Spanish dialects (cf. (5) above) and, like this, it is not germane to our present concern. 71 This, in their view (as already schematized in (68)), ousted the earlier ‘corrélation de longueur consonantique latine vulgaire’ [‘Vulgar Latin consonant length correlation’] when degemination affected the strings ătta, ălla (> ăta, ăla) so that they started to contrast with āta, āla by VL. 70

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance (71)

i iː

y yː

107

u uː

e eː

o oː ε εː

œ a aː

(72)

Breil (Dalbera 1994a: 126–9) [’na :z] ‘nose’ < nasum 6¼ [’bras] ‘arm’ < brac(c)hium [’tʃe :g] ‘fold’ < deverbal from plicare [’sek] ‘dry’ < siccum [’lO :g] ‘hire.prs.sbjv.3sg’ < locet [’tOk] ‘touch.prs.sbjv.3sg’ < *toccet

Since no final devoicing is at work in these varieties, Dalbera (1994a: 126) analyses VL in (72) as allophonically determined by the voice of the following consonant. However, that VL is contrastive in the language is demonstrated by the occurrence of minimal pairs in word-final position: (73)

[’di :] [’se :] [’pε :] [’ka :]

‘say.inf ’ ‘sky’ ‘skin’ ‘dear’

6¼ [’di] [’se] [’pε] [’ka]

‘day’ ‘thirst’ ‘foot’ ‘house’

As Yves Charles Morin pointed out to me (p.c., December 2013), the problem with this piece of evidence is that, though ascribed to Provençal by several scholars (see e.g. the recent map of the dialect subdivision of Occitan in Sumien 2006: 158; Müller and Sidney Martín 2012: 150, where roiasc is at the easternmost tip of Provençal, though marked with the stripes characterizing a mixed area; see also Telmon and Ferrier 2007: 5, 13, 34, numbering the dialect of Briga among Occitan patois), these dialects are best classified as the westernmost exponent of Alpine Ligurian (cf. the map in Forner 1988: 455). Evidence for this classification is provided by the series of distinctive isoglosses which, as shown in the map in Forner (2010: 49), separate roiasc from niçart/niçois. Among these, one finds criterial features singling out Ligurian such as the palatalization of the Latin consonant clusters pl, bl, and fl (cf. Bouvier 1979: 60, Dalbera 2013: 505): e.g. [tʃu’vy] (Briga)/[tʃu’vyu̯] (Tende and Fanghetto) ‘rain.ptp.msg’ (Forner 2010: 80). Even if the evidence for CVL in (71) and (72) has to be redeployed from the present section into }3.4.1.3 or }5.1.2, dealing with Ligurian, there still are peripheral Occitan dialects in which stressed vowels show the effects of an original differentiation by syllable structure, as is the case in the North Alpine Provençal of Val Germanasca (see Appendix, Map 3 for localization): (74)

Val Germanasca, North Alpine Provençal (Pons and Genre 1997; Morin 2003: 131) [’pe :no] ‘punishment’ 6¼ [’p@n :o] ‘pen’ [’pa :lo] ‘shovel’ [ei ̯’pal :o] ‘shoulder’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

108

The development of VL in Romance

Note that in this dialect, a short stressed vowel co-occurs with a following geminate consonant, while long stressed vowels are followed by singleton consonants. Given this distribution, one could in principle entertain the idea that it is consonant gemination which is contrastive here, rather than VL—an issue which was settled by Martinet (1975: 205) by pointing to the fact that, in Franco-Provençal, ‘il est des positions, la finale tonique absolue par example, où la voyelle brève s’oppose à la non brève sans que la question se pose de géminer une consonne suivante’ [‘there are environments, such as the absolute final position, where short vs long vowels contrast without there being any question of geminating a following consonant’]. Given this criterion (mentioned in }3.4.1.1 above), in the Franco-Provençal examples in (75a), the phonemic contrast is the one between /O :/ and /O/—whereas consonant gemination is allophonic—as evidenced by the fact that the same contrast occurs word-finally as well (cf. (75b)):72 (75) Franco-Provençal, Hauteville (Martinet 1956: 74–5) a. [’bO :la] ‘bundle’ 6¼ [’bOl :a] ‘ball’ b. [’mO :] ‘evil, bad’ [’mO] ‘word’ If one considers the evidence from Occitan and Franco-Provençal in (74) and (75), it seems that a strict limitation of the phonologization of VL to northern France is unwarranted. Even for Occitan, it might be that, at a very early stage, this too underwent the same change and that CVL was then lost, leaving only some very scarce traces such as those considered above in (74). One can conclude that Gallo-Romance too, like Northern Italo-Romance—though in a less transparent way, if one considers only present-day dialect variation—upon closer inspection does provide at least some evidence for the reconstruction of a VL contrast diachronically rooted in Late Latin OSL. This contrast must have obtained, in past stages of this Romance branch, much more widely than is the case today. 3.4.3 (The rest of ) Rhaeto-Romance A classificatory dispute has raged ever since Ascoli (1873) proposed that Friulian, Central Ladin, and Romansh varieties should be gathered into a separate branch of the Romance language family, which he dubbed ‘Ladin’ and which was called 72

Occurrence of geminate consonants after short stressed vowels is described for several FrancoProvençal varieties: e.g. [’sOp :a] ‘soup’ < suppam, [’gOt :a] ‘drop’ < guttam in the patois of Vaux (Duraffour 1932: 4), [’bεn :@] ‘mat’ < benna, [’fʎæm :@] ‘flame’ < flammam in that of Certoux, Geneva (O. Keller 1919: 153) (see also, more generally, H.-E. Keller 1958: 103). The only description of a FrancoProvençal patois I know of in which the opposite view is argued for explicitly is Fassò (1974: 236) on Valdôtain. Here, minimal pairs such as [’krap :a] ‘residue of melted butter’ vs [’kra :pa] ‘crevice’, [’sal :a] ‘that.fsg’ vs [’sa :la] ‘pantry’, [’pat :a] ‘paw, foot’ vs [’pa :ta] ‘noodles’ are analysed as involving a /C :/ vs /C/ underlying contrast and no phonemic VL (cf. also the vowel system on p. 247). This choice is based on the rather weak argument that /mm/ and /nn/ are variably realized either as [m :], [n :] or as [ŋm], [ŋn] respectively (e.g. in [’kxaŋma] ‘milk skim, cream’, [’laŋna] ‘wool’). The regional varieties of French spoken in the Franco-Provençal domain also mostly retain CVL, as shown by Walter’s (1982: 194–6) survey.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

109

Rhaeto-Romance (Raetoromanisch) by Gartner (1883). The legitimacy of this classificatory sub-unit has been questioned by many (e.g. Salvioni 1917; Battisti 1929; Pellegrini 1991), including the author of the reference chart of Italian dialects (Pellegrini 1977a), in which Friulian and Central Ladin are included under the heading ‘Italo-Romance’ (cf. also Pellegrini 1972a). This thorny classificatory issue is orthogonal to my present concerns: in any case I have dealt with Friulian—like Pellegrini—in the section devoted to (Northern) Italo-Romance, and I could have done the same for Central Ladin. The choice of presenting the data in this order (with Ladin in the present subsection) is motivated by the fact that Friulian resembles closely the other varieties of northern Italy discussed in }3.4.1 in showing not only contrastive vowel length but, more specifically, CVL in a distribution that can easily be traced back to an original OSL not unlike CVL in other Romance varieties of northern Italy. On the contrary, neither of these conditions is often fulfilled in Central Ladin and Romansh, where—much like in Gallo-Romance—VL has either faded away completely or, in the dialects in which it persists, no longer shows a distribution corresponding to what can be thought to have been the original one, on the strength of the diachronic data expounded in Chapter 2 and of the comparative reconstruction to be sketched in Chapter 5. 3.4.3.1 Central Ladin The label ‘Central Ladin’ is currently used either in a narrow sense, to denote ‘the dialects spoken in the Dolomites around the Sella massif ’ (Salvi 1997: 286), or in a broad sense, stretching eastwards to include also AmpezzanoCadorino: the latter position is mirrored in Pellegrini’s (1977a) reference chart of Italian dialects, which labels Sella Ladin ‘ladino atesino’, joining it to ‘Cadorino’ in the same group.73 Pellegrini extends the label ‘ladino’ also to the bordering transitional area of upper Veneto to the south (‘ladino-veneto’, comprising Agordino, Zoldano, etc.), which departs from the rest of Veneto in a series of features, notably in displaying two characteristic isoglosses of Northern Romance that are missing in Central and Lagoonal Veneto: apocope and (remnants of) CVL. Central Ladin dialects display a fuzzy situation, in several respects. Starting from the Dolomitan varieties,74 while it is widely assumed that OSL has applied across Central Ladin (cf. Salvi 1997: 287) and that at some later stage CVL used to occur homogeneously over the whole area (cf. Gsell 2008: 2774), CVL is preserved today only in Badiotto and Marebbano (cf. Mair 1973: 29; Kramer 1981: 57f.; Videsott and

73 Note that, even in the broader sense, the qualification as ‘Central’ is motivated by the intermediate position between Romansh, to the northwest, and Friulian, to the southeast, which together with Central Ladin constitute Ascoli’s (1873) ‘unità ladina’. 74 These are, proceeding clockwise from the northwest corner, at the German linguistic border, Gardenese, Marebbano, and Badiotto (in the province of Bolzano), Fodóm (or Livinallese, in the province of Belluno), and Fassano (in the province of Trento).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

110

The development of VL in Romance

Plangg 1998: 30; Videsott 2001: 154). Marebbano has the following stressed vowel system (Mair 1973: 29):75 (76)

i iː

y yː e eː

u uː o oː

ø øː

ɔ ɔː

ε εː a aː

The VL contrast is exemplified with the following minimal pairs (data from San Vigilio di Marebbe, Craffonara 1971–2: 52–3; shown in Appendix, Map 3):76 (77) a. [’ma :] [’ma :t] [’va :l] [’a :ra] [’e :rt] [’tε :] [’o :res] [’my :ʃ]

‘only’ ‘only’ ‘something’ ‘threshing floor’ ‘trade, craft’ ‘tea’ ‘work.pl’ ‘face.pl’

6¼

b. [’ma] [’mat] [’val] [’ara] [’ert] [’tε] [’ores] [’myʃ]

‘May’ ‘crazy’ ‘be worth.3sg’ ‘wing’ ‘steep’ ‘take.imp.2sg’ ‘hours’ ‘donkey’

As pointed out by Craffonara (1971–2: 90; 1977: 83–4), these minimal pairs cannot be explained exhaustively by appealing to an original ’CV.CV vs ’CVC.CV contrast, of the type seen in (1c) above. Other alternative explanations—such as invoking the lengthening effect of a final voiced obstruent (to be reviewed in Chapter 4)—are also, in themselves, insufficient, given pairs like [’a :ra] < aream vs [’ara] < alam or [’e :rt] < artem vs [’ert] < *erctum ( *er(i)gere), where there is no question of a voicing-induced lengthening. The rest of Central Ladin has no contrastive VL, but even the dialects which nowadays display no VL contrast at all preserve traces of such a contrast for previous stages in the qualities of stressed vowels, much as is the case in French (}3.4.2). Consider the following data from Fassano in (78a) (from Elwert 1943: 26–30), compared with their Standard Italian counterparts (78b): (78)

i. stratam palum placet a. Fassano [ˈʃtrεdɑ] [ˈpεl] [ˈpjεʃ] b. Italian [ˈstraːda] [ˈpaːlo] [ˈpjaːtʃe] ‘road’

‘post’

nasum [ˈnεs] [ˈnaːso]

‘please.3sg’ ‘nose’

ii. *stalla [ˈʃtalɑ]

vallem brachium [ˈval] [ˈbratʃ]

[ˈstalːa]

[ˈvalːe]

[ˈbratːʃo]

‘stable’

‘valley’

‘arm’

Francescato (1963: 74, n. 22) expressed a diverging opinion, on which see the conclusion of }3.4.3.2. On CVL in Marebbano there is a recent experimental study by Bernardasci (2013a) (on which see also, }5.3.3, (77) and (78)). 75 76

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

111

These data clearly show that OSL must have been at work, in a previous stage of the language, since Fassano /ε/ regularly corresponds to the lengthened allophone [a :] of Standard Italian.77 In principle, palatalization could also have affected a PRom allophonically lengthened [a :] in Fassano (and Sella Ladin in general). However, this is clearly incompatible with the absolute chronology warranted by the historical record. In fact, a-palatalization, which affected the outcomes of Latin stressed -ain open syllables, is known to have started to apply towards the end of the 16th century (Kuen 1923: 68f.; Battisti 1926: 77f.; von Wartburg 1950: 137; Kramer 1981: 55, 57),78 whereas degemination—which destroyed the difference between the two environments (78i) vs (78ii)—applied many centuries earlier. We have to admit an intermediate stage, previous to palatalization, in which the stressed vowels in (78a.i) vs (78a.ii) contrasted phonemically by length.79 Indeed, inspection of the dialects commonly held to lack the VL contrast also reveals traces thereof. Heilmann’s (1955: 279) description of Moenese presents VL as non-contrastive, though indicating one minimal pair, for /a :/ vs /a/ only: l’opposizione di quantità (breve ~ lunga) che si potrebbe sospettare in una coppia come [’tʃar] «carro» ~ [’tʃa :r] «caro», si rivela in realtà illusoria se consideriamo queste due parole nel contesto parlato: [l e ’tʃa :r] «è caro»—[ke bεl ’tʃa :r] «che bel carro» (tono di meraviglia) [‘the quantity contrast (short vs long) that one might suspect in a pair like [’tʃar] ‘wagon’ vs [’tʃa :r] ‘dear, expensive’ in fact reveals itself as illusory as soon as we consider these two words in a spoken context (= in connected speech—M.L.): [l e ’tʃa :r] ‘it’s expensive’—[ke bεl ’tʃa :r] ‘what a beautiful wagon’ (tone of astonishment)’].

Heilmann concludes from this that VL is non-distinctive, and that ‘[l]e differenze quantitative hanno invece solo valore stilistico’ [‘quantitative differences rather have only stylistic value’], whereas his example suggests that they exist (at least for Craffonara (1977: 77–8) compares cognates with etymological stressed -a´- in six varieties of Central Ladin, three with CVL (the Marebbano of San Vigilio, and the Badiotto varieties of San Martino and La Villa) and three without (the dialects of Ortisei in the Val Gardena, of Canazei in Val di Fassa and of Arabba in Livinallongo), arguing that the distribution of the palatalized outcomes (e.g. in the three latter dialects [’pεʃ] ‘peace’ < pacem vs [’fat] ‘done’ < factum) is to be explained appealing to ‘ein früheres romanisches langes a :’ [‘an earlier Romance long a :’] (1977: 77) and that a CVL contrast has to be reconstructed for the whole of Central Ladin. 78 Palatalization started earlier in Badiotto and Marebbano, then spread to the rest of Dolomitan Ladin. As for its environment of application, Craffonara (1977: 97–8) objects to its common characterization as ‘open-syllable’, and indeed there are some complications, as other factors contributed to the set of long /a :/s which underwent the change. Yet the traditional label (as found e.g. in Pellegrini 1972b: 60) is motivated by the fact that a substantial proportion of the items displaying /ε :/ (< /a :/) ultimately owe it to OSL (pace Craffonara). 79 Evidence such as that in (78) and (79) is hard to reconcile with the alternative explanation for the rise of CVL in Ladin as due to devoicing of the final obstruent following the stressed vowel (cf. Videsott 2001). The distribution of different vowel qualities in Fassano and Cadorino is identical to that of long vs short vowels in Cremonese, Emilian, and Ligurian (}}3.4.1.1–3) and in Friulian and Milanese (}}3.4.1.4 and 3.4.1.5) respectively. To argue that this is mere coincidence would strain credulity. A unitary account is called for, such as that to be proposed in Chapters 4 and 5. 77

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

112

The development of VL in Romance

/a/ vs /a :/, which is the most resistant contrast, whenever CVL is lost: cf. (80) below as well as the discussion on (13) and (17) in }5.1.3), but can be neutralized prepausally under emphasis, as distinctively short vowels, contrasting in isolation with long ones, can be lengthened if occurring in an emphatic utterance (a well-known fact, documented e.g. for Friulian in the experimental study by Finco 2007: 123–4). A similar situation, with traces of a previous stage with CVL, characterizes Ladin lato sensu, with the inclusion of Cadorino and the Ladino-Veneto transition area. Cadorino generally lacks CVL nowadays (cf. Zamboni 1984: 57): for instance, the dialect of San Vito di Cadore (cf. Menegus 1980–81: 13–17) has a seven-vowel system like that of Standard Italian (/i e ε a O o u/), so that [’pas] ‘peace’ < pacem has the same stressed vowel as [’fas] ‘bundle’ < fascem (cf. It. fascio [’faʃ :o], as opposed to [’pa :tʃe] ‘peace’). The same goes for the dialect of Selva di Cadore (cf. Nicolai 2000: 436), where [’tas] ‘badger’ (< Late Lat. taxum) = [’tas] ‘be silent.3sg’ (< tacet) are homophonous. However, traces of a previous VL contrast are evident in vowel qualities, as in San Vito; the reflexes of PRom /o/ and /e/ are diphthongs in the same etymological contexts in which Central Friulian has contrastively long vowels ((79a)), and are monophthongs where Friulian has contrastive shorts ((79b,c)) (cf. Menegus 1980–81: 211, 214; Zamboni 1984: 56). (See }3.4.1.5 (58), (60), for the relevant Friulian data.) (79)

a. cotem, acetum

≠ b. gulam, seram

= c. ruptum, siccum

PRom /o/ [ˈkwoðe] ‘whetstone’

[ˈgola] ‘throat’

[ˈroto] ‘broken’

PRom /e/

[ˈsera] ‘evening’

[ˈseko] ‘dry’

[aˈzjeðo] ‘vinegar’

Furthermore, some isolated remnants of VL contrasts crop up in some other dialects of the area. At the northeastern fringe of Cadorino, the dialect of Upper Comelico has some miminal pairs, within morphological series such as [’sta :di] vs [’stadi] ‘been.mpl vs fpl’, [’du :di] vs [’dudi] ‘gone.mpl vs fpl’, or in [’ntε :ndi] ‘dye. inf’ vs [’ntεndi] ‘intend.inf’, [’me :da] ‘half.fsg’ vs [’meda] ‘half-litre measure’ (cf. Zamboni 1984: 57; De Lorenzo Tobolo 1977: xxiv, xxvii, 205–6, 240).80 Similar evidence is provided, further south, by the Ladino-Veneto transition area. Here, VL is mostly lacking, as in the rest of Veneto (cf. }5.1.4), so that for example in Agordino (cf. Rossi 2008: 1151–2), [’tas] ‘badger’ (< Late Lat. taxum) and [’tas] ‘be silent.3sg’ (< tacet) are homophonous. However, in the next valley (some 20 km northeastwards, bordering on Cadore), Zoldán (the dialect of Valle di Zoldo) displays some minimal pairs, if only for stressed /a/ and /e/ (Croatto 2004: 324, 375): (80) a. [’na :s] ‘nose’ [’pe :s] ‘weight’

< nasum < pe(n)sum

6¼

b. [’nas] ‘be born.3sg’ [’pes] ‘fish’

< nascit < piscem

80 While these are surely remnants of previous CVL, much as in Marebbano (cf. (77)) they cannot be traced back in a straight line to OSL.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Northern Romance

113

Note that, given contrasts such as those in (80), the original distribution of long vowels as successors of earlier allophones lengthened in open stressed syllables can still be recognized, as seen in the Latin etyma. There is every reason to believe that CVL once stretched over all the Central Ladin territory. Where the contrast is still observable, its distribution today sometimes does not make sense etymologically (as in Marebbano and Badiotto, or in the sparse remnants found in the dialect of Comelico), because a series of lengthening and shortening processes, described in detail in Videsott (2001: 155–66), have obscured the original distribution and the link with OSL. But sometimes it does make sense, and in that case the link with the OSL is transparent, as shown in (80). 3.4.3.2 Romansh The argument now developed for Central Ladin can be repeated for Romansh. Here too, there are varieties (see Appendix, Map 2 for localization) in which VL is contrastive to this day. Among the major subgroups, this holds for Surmeiran and Engadinian, as illustrated in (81) and (82) (cf. Lüdtke 1956: 378; Haiman and Benincà 1992: 32–3; Videsott 2001: 154). The contrast is exemplified for Surmeiran with the following minimal pairs (cf. Thöni 1969: 16, 275): (81) a. [’e :r] ‘field’ < ager 6¼ [’go :t] ‘forest’ < Germ. wald [’bo :t] ‘early’ < Germ. bald

b. [’er] ‘also’ < *era [’got] ‘drop’ < gutt(am) [’bot] ‘hill’ < *bott- (DRG 2.447)

To exemplify VL in Engadinian, consider the following data from the lower Engadinian dialect of Scuol (Gredig 2000: 59): (82) a. [’fi :t] ‘trust.1sg’ < fı¯do 6¼ b. [’fit] ‘rental’ [’fy :s] ‘spindle’ < fu¯sum [’fys] ‘be.impf.sbjv.3sg’ [’mø :ts] ‘wat/fashion.pl’< modos [’møts] ‘bushel’ [’tʃe :l] ‘sky’ < caelum [’tʃel] ‘that.msg’ [’tɕa :r] ‘dear’ < ca¯rum [’tɕar] ‘wagon’ [’bo :f] ‘ox’ < bovem [’bof] ‘gust of wind’

< fı¯ctum < fu(i)sset < modium < *ecc(e)-illum < carrum < *buff

One more piece of evidence for the contrastiveness of VL in this variety is that VL contrasts occur also word-finally: (83) a. [’be :] ‘beautiful’ [’ma :] ‘never’

< bellum < ma(gi)s

6¼

b. [’be] ‘only’ [ma] ‘but’

< bellum DRG 2.278 < ma(gi)s

As shown by the etyma in (82), the contrasts can be manifestly traced back to PRom OSL, though in several instances this is not the case. Not only in the word-final syllable but also in paroxytones, the VL contrast may arise from a series of further changes, as illustrated in (84): (84) a. [’fO :sa] ‘false.fsg’ < falsam 6¼ b. [’fOsa] ‘grave’ < fo˘ssam [’ku :la]‘ball’ < Germ. kugel DRG 4.354 [’kula] ‘flow.3sg’ < co¯lat

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

114

The development of VL in Romance

The long vowels in (84a) are not diachronic successors of allophones lengthened by OSL, and also the short vowel in [’kula] in (84b) is not the result one would expect, if CVL were to reflect transparently an older OSL. Yet there are also several cases that are in keeping with such an assumption (which can hardly be explained away as coincidental), like those in (82a–b)—as apparent from the etyma—or the following: (85) a. [’dy :r] [’no :f] [’la :t] [’tʃa :r]

b. [’dy :ra] [’no :va] [’la :da] [’tʃa :ra]

‘hard.m/f’ ‘new.m/f ’ ‘large.m/f ’ ‘clear.m/f ’

< du¯rum/-am < no˘vum/-am < la¯tum/-am < cla¯rum/-am

The occurrence of a long vowel also in the paroxytonic alternants in (85b) shows that the dialect of Scuol resembles Cremonese, Emilian, and Ligurian (}}3.4.1.1–3), while differing from Milanese and Friulian (}}3.4.1.4 and 3.4.1.5) in this respect. In view of the contrasts seen in (82)–(84), one can hardly subscribe to the claim that ‘in Puter and Vallader, length is largely, if not entirely, predictable’ (Haiman and Benincà 1991: 33). In spite of this, there has been a tradition of denying that VL in Engadinian is contrastive (cf. e.g. von Essen 1964: 141). Gredig’s (2000: 60) experimental study on the dialect of Scuol (the same investigated by von Essen 1964) shows clearly that the phonetic realization of long vs short vowels in (82a,b) differs sharply, in a way that is consistent with the hypothesis that the contrast is phonemic: the average durational values (for her four speakers) for the stressed vowels in the pairs in (82a) vs (82b) are 201 vs 93 ms. The situation now illustrated in (82)–(84) is observed in the whole of Engadinian, down to Val Müstair, at the easternmost fringe of Romansh, on the linguistic border with Tyrolean German. This is exemplified in (86):81 (86) a. [(ma) ’my :t] [’tɕa :r] [’la :t] [(ma) ’fi :t] [’tɕe :l] [’be :]

‘change dress.1sg’ ‘dear’ ‘broad’ ‘trust.1sg’ ‘sky’ ‘beautiful’

6¼

b. [’myt] [’tɕar] [’lat] [’fit] [’tɕel] [’be]

‘dumb’ ‘wagon’ ‘milk’ ‘rental’ ‘that’ ‘only’

As for the other major dialect group, Surselvan, this is generally analysed as lacking CVL (Kramer 1972: 354; Liver 1999: 128). However, the variety spoken in the Tavetsch valley, on the western border of the Surselvan territory, does display CVL (data from Caduff 1952: 23–9, 71, 130, 162f.; see also Plangg 1987: 121): (87) a. [’pa :lɐ] [’na :s] [’a :lɐ] [’kla :f ]

‘grassy slope’< palam 6¼ ‘nose’ < nasum ‘wing’ < alam ‘key’ < clavem 81

b. [’balɐ] [’pas] [’val] [’ɟat]

‘ball, bale’ < Germ. balla ‘step’ < passum ‘valey’ < vallem ‘cat’ < *gattum

The data stem from my own fieldwork (June 2011).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Summing up

115

Sutselvan too, the demographically weakest variety of Romansh, is normally described as lacking CVL: cf. e.g. Solèr (1991: 55) on the dialect of Andeer (Schams valley). This description is contradicted by Solèr’s report of minimal pairs parallel to those given above in (82) (e.g. [’bo :f] ‘ox’ < bovem vs [’bof] ‘gust of wind’ < *buff), which he claims not to be relevant phonemically because words with [’V :] can have their vowel optionally shortened (e.g. [’bo :f] ‘ox’ can be also pronounced [’bof]). However, much as we have seen in }3.4.3.1 for Heilmann’s (1955: 279) description of Moenese, these observations indeed show that, at least residually, a contrast is still observable, although it can be neutralized. Summing up, it seems fair to subscribe to Lüdtke’s (1954–5: 237) conclusion, according to which inspection of the Rhaeto-Romance data suggests that in this whole area ‘ebenso wie im Französischen und Oberitalienischen betonte Vokale in freier Stellung gedehnt wurden und dadurch eine vokalische Quantitäts- oder Anschluß(-Silbenschnitt)-korrelation entstand’ [‘as in French and Northern Italian, stressed vowels in open syllables were lengthened and through this a vowel quantity, or syllable-cut, contrast arose’]. Because of this areal coherence, this view seems to be preferable to others which have been entertained in the literature. The present discussion also shows that, while there is abundant evidence showing that CVL was a common feature, originally, of the whole of Rhaeto-Romance,82 it is hardly justified to consider CVL as a defining isogloss of Rhaeto-Romance alone, singling it out from the rest of the Romance languages, as argued by e.g. Craffonara (1977: 77). Craffonara takes issue with Francescato (1963: 74, n. 22), who in turn restricts even further the geographical/classificatory scope of CVL, regarding it as a feature originally shared by all Friulian varieties, but distinguishing them from Central Ladin and Romansh which, according to him, ‘non presentano il fenomeno della quantità vocalica fonologicamente pertinente’ [‘do not display the phenomenon of phonologically relevant VL’]. None of these views seems tenable, given the evidence discussed in the preceding pages.

3.5 Summing up: VL and OSL from Proto-Romance to the modern languages With this, our first tour d’horizon through the basic facts of VL in Romance has been completed. This has shown that alongside many Romance varieties lacking CVL, considered in the first part of the chapter (}}3.1–3), there are several Romance dialects 82 This is also the conclusion of Leonard (1972), who ascribes CVL to ‘Proto-Rhaeto-Romance’, based on the comparison of four varieties, two lacking CVL (Gardenese and the Romansh dialect of Trun, Upper Vorderrhein) and two displaying it (the Lower Engadinian dialect of Scuol and the Friulian variety of Pesariis, Carnia). Like Lüdtke, Leonard (1972: 67) traces back this CVL to previous OSL: ‘The reflexes of earlier Romance stressed vowels in checked syllables appear as shorts, while those in free syllables emerge as longs.’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

116

The development of VL in Romance

which possess VL contrasts. Such secondary phonemic VL, never stemming from preservation of the CL contrast, has arisen by disparate paths: for instance, dialects with secondary surface VL contrasts fed exclusively by CoL for the loss of an adjacent consonant are found in Ibero-, Gallo-, and Italo-Romance, as exemplified respectively in (5), n. 51, and n. 70 above. While such cases are scattered, there is a clear areal pattern involving the Northern Romance subdivision, since many Romance varieties spoken from the North Sea down to the Apennines display a series of VL contrasts which, albeit with some differences, in part still resemble each other synchronically. This also extends to dialects which have no CVL today but show clear traces of its presence at an earlier stage. This family resemblance becomes all the more striking if one considers diachrony, since it becomes evident that one possible explanation for the rise of CVL in Northern Romance, wherever it occurs, is that this is a diachronic successor to an OSL rule of the kind still found in modern Standard Italian.83 The rise of OSL in Late Latin was the subject of Chapter 2. Having now introduced more Romance data, we can add that Romance varieties provide cues which allow one to further refine the reconstruction of its origin. In }3.2, we dealt with OSL as though it were a strictly word-level process. However, the experimental evidence of modern Standard Italian shows that this is not so. While there is clear evidence for a roughly 1:2 ratio for stressed vowel durations in ’CVCCV vs ’CVCV words, when uttered in isolation (Fava and Magno Caldognetto 1976; Farnetani and Kori 1986; D’Imperio and Rosenthall 1999: 6), it has also been demonstrated that OSL is utterance-bound. Bertinetto (1981: 132–7, 263) shows that the effects of OSL dramatically reduce (to a ratio very close to the threshold of perceptibility) when ’CVCV words are uttered in connected speech in non-prepausal position. This result has been replicated by many other studies: for instance, Farnetani and Kori (1990: 55) show that while stressed /’a/ in nove mani ‘nine hands’ lasts over 200 ms., in mani corte ‘short hands’ it is only around 120 ms. long. The conclusions of D’Imperio’s (2000: 72) study on the acoustic-perceptual correlates of sentence prominence in Italian are in line with these findings; and Dell’Aglio et al. (2002) show that, in the corpus of spontaneous speech they analyse, the only condition in which stressed vowel durations in closed vs open syllables differ significantly is when main-clause stress cumulates with final prominence. On the whole these durational effects on OSL are but one specific facet of a well-known phonetic universal: prepausal lengthening (cf. e.g. Maddieson 1997: 631–2; Pompino-Marschall 2010: 413–14).

83 Indeed, contrastively long vowels occur in many Northern Romance varieties in the same contexts where a lengthened stressed allophone occurs in modern Standard Italian (cf. (2a)). Arguably, whenever this is not the case, this is due to later changes which affected VL obscuring the original distribution, as will be shown at several places in Ch. 5.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Summing up

117

These experimental results for Standard Italian have a special value for the reconstruction of (variation in) vowel duration in Proto-Romance, since Italian (the standard language as well as its central and many of its southern dialects, cf. } 2.2.1, (10)) is the only variety—with Sardinian (cf. }2.2.1, (7))—that retains OSL as a synchronically active allophonic process. Is there any reason to project these conditions back into Proto-Romance? A general reason to envisage this possibility is that prepausal lengthening appears to be quite widespread cross-linguistically. Further relevant evidence can be gained from descriptions of Italo-Romance dialects in which the original (purely allophonic) lengthening evolved into phonemic changes. This happened in two different ways. In southeastern Italian dialects, none of which has developed secondary VL contrasts, prepausal position is where all diphthongization and/or colouring processes apply most systematically. We saw in }2.2.1, (11) that such processes are sensitive to syllable structure. In addition, in many of these dialects diphthongization and/or colouring are also utterance-bound. Thus, in the dialect of Bisceglie (province of Bari; see De Gregorio 1939: 34–5, 37), stressed -a- was coloured to [O], whereas high vowels diphthongized (-ı¯- > [ø i ]̯ , -u¯- > [iu̯]) in open non-antepenultimate syllables: e.g. [’kOp@] ‘head’ < caput, [’fOv@] ‘broad bean’ < fabam; [ga’d :øi n̯ @] ‘hen’ < gallı¯nam, [’føi ̯k@] ‘fig’ < fı¯cum; [’kriu̯t@] ‘raw’ < cru¯dum, [’miu̯t@] ‘funnel’ < imbu¯tum. All these processes are blocked when the word concerned is not prepausal, as seen in (88a):84 (88) a. [f@’k :a ’ind@] [a’s :i ’fo :r@] [’c :u ’b :rut :@]

‘to stick into’ ‘to get out’ ‘uglier’

b. [f@’k :O] [a’s :øj@] [’ci :w@]

‘poke.inf ’ ‘exit.inf’ ‘more’

The same happens in nearby Bitonto (cf. Merlo 1911–12: 909, 919), as illustrated with /i/ ! [øi ̯] diphthongization in (89) (see Appendix, Map 1 for the localization of these and other southern Italian dialects in this section):85 (89) a. [r@ ’tʃim@ d@ ’reu̯p] ‘the turnips’ b. [r@ ’tʃøi m ̯ ] ‘the cauliflowers’ [’ε ’ʃ :u̯t a ʃ :ʊ’kwew@] ‘s/he went to play’ [s@ n ’ε ’ʃ :iu̯t] ‘s/he went away’ Nor are the Apulian dialects the only ones concerned: this kind of positional variability is described for dialects spoken further north as well, like the Molisano dialect of Agnone (province of Isernia; cf. Ziccardi 1910: 417):

84 Prepausal [’b :rut :@] in (88a) exemplifies non-application of the process to stressed vowels in closed syllables. The relation between the syllable structure condition and the prosodic one (antepenult vs elsewhere) is going to be addressed in Ch. 5, (102). 85 Experimental phonetic evidence for the positional variability of some of the diphthongs of another Apulian dialect, Altamurano, is produced in Loporcaro (2001: 296–9).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

118

The development of VL in Romance

(90) a. [’ji l@ ’vai̯d@] ‘I see him’ [’pwo m@’ni ad :@’me̯an@] ‘you can come tomorrow’

b. [l@ ’ved@ ’joj :@] ‘it’s me who see him’ [n@m ’bOt :s@ m@’noj :@] ‘I cannot come

The Lucanian dialect of Matera displays similar alternations: (91) a. [nan dʒ ’I b :@’nIt@ n@’ʃ :i-i ̯n] b. [nan dʒ ’I b :@’ni-i ̯t] ‘s/he didn’t come’ ‘nobody came (here)’ [’fo :r@ ’tar :] ‘out of town’ [’I d :ʒI :t@ ’fai ̯r] ‘he went to the countryside’ (lit. ‘outside’ < foras) Contrary to (88)–(90), in this variety diphthongization under main-utterance stress affected closed syllables too: (92) a. [na tʃ@’pOd :a ’b :jOŋg]

‘a white onion’ b. [’la tʃ@’pau̯d :] ‘the onion’ ‘smear well’ [’jau̯ndʒ] ‘smear’ ‘there are many dogs’ [’na n :@ ’stau̯n :] ‘there aren’t any’ [nʊ tras’kIrs@ ’lʊɲ :@ ’lau̯ɲ :] ‘a very long discourse’ [’lau̯ɲ :] ‘long.m’ [’jOndʒ@ ’b :øu̯n] [’stOn : a’s :ε ’kε :n]

Similar examples of diphthongization are found, albeit less frequently, in the western part of southern Italy. Rohlfs (1938) discusses two such dialects, the Campanian variety of Pozzuoli (province of Naples; see now also Abete 2006) and the northern Calabrian variety of Belvedere Marittimo (Cosenza):86 (93) a. [’tεla ’jaŋg@]

‘white cloth’ b. [’tai ̯l@] ‘cloth’ (Pozzuoli, province [na ’nOtʃ@ ’rOs :@] ‘a big walnut’ [a ’nau̯tʃ@] ‘the walnut’ of Naples) [nu ’fil@ ’føi ̯n@] ‘a thin thread’ [’føi ̯l@] ‘thread’

(94) a. [u ’vinu ’jaŋku] ‘white wine’ b. [’vai ̯nu] ‘wine’ (Belvedere Marittimo) [a ’luna ’nOva] ‘new moon’ [’lau̯na] ‘moon’ (province of Cosenza) [a ’b :utʃe ’avuta] ‘aloud’ [’vau̯tʃe] ‘voice’ As Rohlfs points out, the diphthongizations under discussion are the product of relatively recent changes: all over southern Italy, they must be placed in relative chronology after metaphony (see Weinrich 1958: 175–7; Papa 1986 [1981]; Loporcaro 1988a: 26). Therefore, they do not directly mirror a Proto-Romance situation. Nevertheless, since these diphthongizations are diachronic successors of an OSL of the Tuscan kind, the fact that they first arise in—and, in some dialects, remain limited to— utterance-final position provides comparative evidence to be added to experimental 86 As in Materano, in Pozzolano too utterance-bound diphthongization affects stressed vowels also in stressed closed syllables (see Abete 2006: 390):

(i)

a. [i ’r :et :s@ s@ t@’rav@n a ’b :rat :ʃ@] b. [a ’ser@ van : a ’m :et :@r i ’r :UI t̯ :s@] ‘the nets used to be hauled in by hand’ ‘in the evening, they go to cast the nets’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Summing up

119

measurements for contemporary Standard Italian pointing to the sensitivity of OSL to prepausal vs non-prepausal position. Similar evidence comes also from NRom dialects of the type considered above in }3.4, as some descriptions emphasize that contrastive VL is fully realized phonetically when the word is uttered in isolation and/or prepausally, but is heavily reduced elsewhere. Thus, for the Emilian dialect of Fiorenzuola, Casella (1922: 51) observes that the long vowels in (95b) are shortened when not occurring prepausally (see (95a)): (95) a. [’pa :ga]

‘pay.2sg.imp’ b. [’pag ti ’to ’debit] ‘pay your debts’ (Fiorenzuola, [a’mi :k kun tyt | a’mi :k ku ’nsõ] [a’mi :k] ‘friend’ province of ‘everybody’s friend, nobody’s friend’ Piacenza)

Rossini (1975:183–4) makes a similar point for Cremonese, discussing the minimal pair /’mi :a/ ‘not’ vs /’mia/ ‘mine.fsg’, as exemplified in the following utterances: (96) a. b.

[’ke :sta ’ki l ’ε ’mi :a ’mia] [’ke :sta ’ki l ’ε ’mia ’mia]

‘this one(f) is not mine’ Cremona ‘this one(f) is really mine’ (lit. ‘mine mine’)

He observes that in the utterance in (96a) the quantity of the short stressed vowel in /’mia/ is not subject to variation: non è possibile sotto alcuna condizione una pronuncia lunga della -i- del possessivo; qualora ciò si verificasse, mía verrebbe percepito dall’ascoltatore come imprestito dalla lingua, conferendo alla frase un bizzarro aspetto bilingue. (Rossini 1975: 184) [‘under no condition is a long realization of -i- in the possessive possible; should this happen, mía would be perceived by the listener as a loan from Italian, thus bestowing on the utterance a bizarre bilingual flavour [or, more technically, lengthening would boil down to code mixing, i.e. insertion of Standard Italian [’mi :a]—M.L.].]

Long vowels are treated differently, so that the negation /’mi :a/ ‘not’ is normally pronounced as homophonous with possessive /’mia/, i.e. as [’mia], whenever not in focus. Thus, if realized with neutral intonation (96a) becomes homophonous with (96b), whereas phonemic /i :/ would surface as [i :] under narrow focus (‘it’s not mine’): ‘Ossia: “Quante volte devo ripetere che questa non è mia?”’ (Rossini 1975: 184) [‘That means, “How many times must I repeat that this one is not mine!?”’]. In sum, the contrast is there lexically, but its realization is suspended whenever the word containing a long vowel is not pronounced in isolation nor occurs prepausally and/or in focus.87 We have seen that not only OSL in Standard Italian but also several other phonological processes/properties arguably arising from PRom OSL, like the southeastern Italo-Romance diphthongizations/colouring processes considered in (88)–(94) 87 Similar descriptions are provided in the literature for several other Northern (Italo-)Romance dialects, e.g. the varieties of Friulian considered in R. Castellani (1980: 31), in which ‘nel discorso corrente’ [‘in connected speech’] long vowels are shortened. This need not be the case everywhere, though: cf. Figs 5.5 and 5.6 below for the illustration of an Emilian dialect in which VL contrasts are realized phonetically also in utterance-internal position.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

120

The development of VL in Romance

and the Northern Italo-Romance CVL in (95) and (96)—in some varieties, at least— are sensitive to the position within the utterance, as well as to focus. All of the evidence now reviewed suggests that OSL may have been so conditioned from the outset in Proto-Romance too. This carries a further implication for the whole issue, discussed in Chapter 2, of the loss of Latin distinctive VL: the VL contrast may well have been lost in prepausal position first, through OSL, yielding to a conflict between word-level phonology, providing for VL contrasts (at least under stress), and sentence-level phonetics, with its tendency to enhance stressed segments occurring immediately before boundaries, and especially in prepausal position. This said, I here conclude this addition to the reconstruction of the rise of OSL in Late Latin/early Romance (as a complement to Chapter 2) and turn to the reconstruction of the rise of the novel CVL in Northern Romance. This is far from uncontroversial, and indeed a host of differing diachronic explanations have been proposed, which I shall now review in Chapter 4 before proposing my own account in Chapter 5.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

4 The analysis of Northern Romance vowel length In }}3.4.1–3, data from several Romance varieties were reviewed, most of which display contrastive VL in their present stage, while some (French, among the major standard languages: cf. }3.4.2) did at some past stage. I shall now concentrate on Romance dialects which have CVL today and look at the different analyses of VL which have been put forward, in traditional dialectological work and above all in more recent studies in theoretical phonology. As will become apparent in }4.1, much analytical effort has been expended in the formalization of the synchronic working of VL in these dialects. All such formal accounts have diachronic implications, many of them address diachrony explicitly, and some even announce this from the title, like Prieto’s (1994) paper on ‘Historical vowel lengthening in Romance’ [emphasis added—M.L.]. In considering these different analyses, I shall keep two basic issues in mind: first, the question of how well they fit in with the long-term Latin–Romance story I am dealing with, which started with the rise of OSL in Late Latin considered in Chapter 2; secondly, the issue of whether such analyses are capable of establishing a relationship between the different kinds of CVL reviewed in Chapter 3. After discussing these competing accounts of (the rise of) VL, in the final part of the present chapter (}}4.2–3) I shall discuss some issues of method concerning the relationship between synchronic and diachronic linguistics, which will prove relevant for assessing the relative merits of competing approaches by weighing them against a finer-grained inspection of dialect variation.

4.1 Competing analyses of the rise of CVL in Northern Italo-Romance For various reasons which will become apparent as the argument proceeds, it is useful to concentrate particularly on the different kinds of VL system found across northern Italy. The study of VL, like that of many other linguistic phenomena, demonstrates the truth of Lausberg’s dictum: ‘La dialettologia italiana [ . . . ] è il fuoco nel sistema ottico della linguistica romanza’ [‘Italian dialectology is the focus in the optical system of

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

122

Analysis of Northern Romance vowel length

Romance linguistics’] (Lausberg 1974: 252). In other words, in no other branch of Romance is a similar structural differentiation to be observed: all structural options documented in the whole of the Romània occur in one or other of the dialects of Italy. Let us start from the difference in the distribution of VL which distinguishes the varieties addressed in }}3.4.1.1–3 from those discussed in }}3.4.1.4 and 3.4.1.5. The two types are synthesized in (1) and (2) with data from Cremonese and Milanese respectively:1 (1)

Cremonese (Southern Lombard; cf. }3.4.1.1) a. ’CV.CV i. Vowel quantity contrast [’pa :n] ‘bread’ in oxytones [’pe :l] ‘hair’ ii. . . . and paroxytones [’pa :la] ‘shovel’ [’la :na] ‘wool’

b. ’CVC.CV [’pan] ‘cloth’ [’pel] ‘skin’ 6 ¼ [’spala] ‘shoulder’ [’kana] ‘reed’ 6¼

(2) Milanese (Western Lombard; cf. }3.4.1.2) a. ’CV.CV i. Vowel quantity [’ka :l] contrast in oxytones [’fy :z] ii. . . . but not in [’pa(ˑ)la] paroxytones [’pi(ˑ)la]

b. ’CVC.CV ‘loss’ 6¼ [’kal] ‘spindle’ [’fys] ‘shovel’ = [’spa(ˑ)la] ‘mortar’ [’spi(ˑ)la]

‘corn’ ‘be.impf.sbjv.3sg’ ‘shoulder’ ‘pin’

While in (1) VL contrasts occur in both oxytones (1i) and paroxytones (1ii), in (2) they are restricted to oxytones, to the exclusion of paroxytones (and, a fortiori, of proparoxytonic words: see }5.4 below). Comparing these two different distributions, the question in (3) naturally arises: (3)

Same diachronic source for contrastive VL in (1) and (2)? a. Yes. Two stages in the same development. Weinrich (1958: 188); Morin (2003); Loporcaro (2007a)2 b. No. Two distinct developments. Francescato (1966: 130–43); Vanelli (1979), Zamboni (1984: 56); Montreuil (1991); Repetti (1992); Hualde (1992); Benincà (1995: 51); Bonfadini (1997); Baroni and Vanelli (1999; 2000); Prieto (1994; 2000); Videsott (2001); etc.

1 All the words involved in these pairs stem from Latin disyllables as shown by the CV skeleton on top, which also shows that the contrast, in Latin, used to be one of gemination, which was neutralized through Western Romance degemination (}3.3, (11b)). The words now ending in a consonant underwent apocope (}3.4). 2 Filipponio (2012: 66–8, n. 105) admits the link for Milanese, while denying it for Friulian, in which the distribution of VL is similar to that found in Milanese (though with the difference addressed while commenting on }3.4.1.5, (62)).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Competing analyses of the rise of CVL

123

In (3a,b) some proponents of the two opposite conceivable answers are listed. Neither list is exhaustive, and both gather contributions which deal with the issue from disparate angles and with huge differences in scope and coverage. Thus, for instance, Weinrich (1958: 188), while extensively discussing dialects of EmiliaRomagna of type (1), devotes just a quick remark to Friulian, which is however sufficient for us to range him in (3b).3 Symmetrically, Francescato (1966) is a thorough book-length survey of dialect variation across Friulian dialects, which outlines a full reconstruction of the diachronic development of their phonology, without addressing comparison with NItRom varieties outside Friuli. Though far from exhaustive, the lists are a representative sample in the sense that they mirror the circumstance that the second answer (3b) enjoys more favour in current studies in the field: (3b) lists just a selection of some relatively recent contributions which are only compatible with an answer in the negative. Some of them deal with both types (1) and (2), proposing entirely disjunct accounts for VL in the two kinds of system (Repetti 1992; Prieto 1994); most of them, on the other hand, only deal with VL in varieties of type (2) (especially Friulian and Milanese), but in such a way that their treatment cannot possibly be extended to cover the systems of type (1). Repetti (1992: 180), summing up the main results of her comparative analysis of northern Italian dialects, makes this point explicitly, listing the divorce between the two types of development among the advantages of her reconstruction: I have shown how similar synchronic structures (long vowels) in related languages (northern Italian dialects) may have different origins. In some dialects, the long vowels are the result of vowel lengthening in open syllables (bimoraic norm), while in others the long vowels arose through a process of compensatory lengthening due to apocope of word final vowels [respectively, Cremonese vs Milanese; emphasis added—M.L.].

Thus, for Repetti, long vowels in the Cremonese type (1) go back in a straight line to Late Latin OSL—the bimoraic norm, in her terms—whereas in the Milanese type

3 Here is the relevant passage: ‘In den friaulischen Mundarten werden, wie überhaupt in Oberitalien und in weiten Gebieten der westlichen Romania, die Quantitätsoppositionen der Konsonanten in den Vokalismus verlegt. Wie vorher die Konsonanten, so unterscheiden sich nun die Tonvokale nach Länge und Kürze, dergestalt daß eine alte Konsonantenlänge durch eine neue Vokalkürze, eine alte Konsonantenkürze durch eine neue Vokallänge vertreten wird. Dieser Quantitätsunterschied ist phonologisch relevant. Hinter dem vokalisch-konsonantischen Lautwandel steht also eine Oppositionsverschiebung’ [‘In the dialects of Friuli, as generally in northern Italy and large parts of the western Romània, the consonant quantity contrasts are shifted onto the vocalism. As earlier the consonants, now the vowels contrast by length and shortness, in such a way that older consonant length is represented by a new vowel shortness, and older consonant shortness is represented by new vowel length. This difference in quantity is phonemically relevant. Hence, behind the sound change involving consonants and vowels is a shift in phonemic contrast’] (Weinrich 1958: 188).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

124

Analysis of Northern Romance vowel length

(2) they are the product of an entirely different change. This basic idea is shared by most of the studies listed in (3b), which are reviewed in what follows. Another trait shared by most of those studies, from Vanelli (1979) on, is that they tend to consider VL in the varieties involved as a derived property, not encoded in the underlying representation. In other words, whatever change may have caused long vowels to arise, this is taken to be still part of the synchronic phonology of those dialects. Both assumptions, about the derivative and synchronic nature of VL, are scrutinized in the present chapter. As it turns out, VL is more economically accounted for as the product of a diachronic change which has affected phonological representation. 4.1.1 Formal accounts of the rise of VL in Milanese Repetti’s (1992: 175) account of the rise of contrastive VL in Milanese is couched within Moraic Phonology. It assumes that VL arose via CoL for the loss of final vowels, as shown in (4): (4)

Example: focu > *[’fOgo] > [’fø :g]/[’fø :k] ‘fire’ a. Input form

b. Apocope

σ

σ

σ

σ

μ

μ

μ

μ

f ɔ ɡ o

→

c. Parasitic delinking

f ɔ ɡ

→

d. Compensatory lengthening

σ μ

f ɔ ɡ σ

μ

μ μ →

f ɔ ɡ

Exploiting the resources of the formalism, which provides for a representation of timing units (μs) as separate from segmental content, it is assumed that apocope ((4b)) cancelled only the latter, while sparing the final mora which was then reassociated with the stressed vowel resulting in CoL. This explains why no contrastively long vowels are found in the stressed syllables of paroxytonic words, even in forms whose stressed vowels alternate with /V :/ in ’CVC (cf. Ch. 3, (44a): e.g. [’nø :f ]/ [’nøva] ‘new.m/f’, [’di :z]/[’dizi] ‘say.3sg/1sg’. Repetti’s account is a formalization within Moraic Phonology of a view that had been upheld long before, including the first scientific description of Milanese: ‘la vocale d’uscita, sparendo dopo consonante sonora, rese lunga la vocale accentata di penultima’ [‘the final vowel, disappearing after a voiced consonant, lengthened the

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Competing analyses of the rise of CVL

125

stressed penultimate vowel’] (Salvioni 1884: 159).4 It implies that, prior to the change in (4), Milanese and Friulian had no distinctive VL nor any allophonic lengthening. Formally, they had only monomoraic vowels underlyingly, and acquired long vowels through compensatory lengthening. This resulted in the contrast which is standardly represented as (5a,b) in Moraic Phonology: (5)

a.

b. μ

c. μ

d. μ

μ

V

V

C

C

[V]

[Vː]

[C]

[Cː]

While (5a,b) formalize a VL contrast, (5c,d) show the standard underlying representations of short vs long (geminate) consonants. This is the contrast capitalized on in another formal treatment of Milanese VL by Montreuil (1991: 43ff.), who develops an alternative mora-based account. For minimal pairs like those in }3.4.1.4, (44a,b), he assumes the structural representations in (6a,b): (6)

a. [’fys] ‘be.impf.sbjv.3sg’ input (and output) form

b. [’fy :z] ‘spindle’ input form

c. output (SRC)

σ

σ

σ

μ μ

μ

μ μ

f y s

f y z

→

f y z

Under this view, stressed short vowels are assumed to be followed by moraic consonants underlyingly, whereas long vowels are followed by non-moraic codas. Given the standard moraic representations in (5c,d), this is tantamount to claiming that consonant gemination is still phonemically relevant in Milanese, despite its complete disappearance at the surface, due to the WRom. chain shift seen in Ch. 3 (11), which affected Milanese just like French, Spanish, and the rest of WRom. Now, once consonant gemination (or, technically, moraic weight on consonants) is assumed to be underlying, vowel length can be derived, as shown in (6c), through the enforcement of a Strong Rhyme Constraint (SRC) like the one proposed for standard Italian—a language with overt contrastive gemination—by Chierchia (1986). The SRC decrees that all stressed syllables be bimoraic, and therefore prevents the underlying structure (6b) from emerging as a well-formed surface structure, so that lengthening of the stressed vowel in [’fy :z] takes place: 4 Contini (1935a: 59) extends Salvioni’s explanation to the Western Lombard dialects of Pavia and Voghera. More recently, the same view is adopted by Sanga (1997b: 255) for Milanese, and Bonfadini (1997: 599–600) for the dialect of Novate Mezzola (another Western Lombard variety spoken in the province of Sondrio, cf. }3.4.1.4, (54)–(56)).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

126

Analysis of Northern Romance vowel length

There are essentially two ways in which the SRC could rectify (11) [= (6b) here—M.L.]: by assigning a mora to the consonant, i.e. by somehow forcing the Weight-by-Position process to apply and thus destroying the moraic vs non-moraic coda contrast; or by lengthening the vowel. Milanese opts for the latter, and we can view lengthening as the process that turns (11) into (15) [= (6b) ! (6c) here—M.L.]. (Montreuil 1991: 44)

Contrary to Repetti’s approach, this treatment could be extended to dialects of the Cremonese type (1), though the comparative issue is not addressed by Montreuil. It would suffice to assume that paroxytones in (1ii) also contrast underlyingly by consonant gemination, and that the SRC applies there too to derive VL at the surface. Something along similar lines (abstracting away from differences in technical details) is indeed proposed for Cremonese by Prieto (1994), to be discussed in }4.1.3. Obviously, this analysis is costly, as it postulates an underlying representational contrast which never appears at the surface. It is also empirically unsatisfactory, since it is unclear how such a treatment could cope with the occurrence of CVL in word-final position (e.g. [kan’ta :] ‘sung’ vs [kan’ta] ‘sing.inf’: cf. }3.4.1.4, (45a,b)), a piece of evidence not addressed by Montreuil. The same applies to the CoL-based account by Repetti. In addition, as noted by Vanelli (1998: 78f.), under this account, which Repetti proposes for both Friulian and Milanese, it is unclear what may have prevented CoL from applying in proparoxytones ((7a)) or in paroxytones containing a geminate or a consonant cluster ((7b)), since there as well a final vowel has been deleted and there as well, in principle, the final mora could have been reassociated to the preceding vowel, which however was not the case, as shown in (7a,b): (7)

a. σ

σ

σ

μ

μ

μ

u

b. l

m

i

d

o

σ

σ

μ

μ

a

t

→

e

σ

σ

μ

μ

μ

i

d

u

m

→

l

σ

μ

μ

μ

i

d

u

σ →

σ

m σ

μ

μ

a

t

→ l

μ

μ

a

t

Symmetrically—Vanelli (1998: 78) further argues—there is no principled reason accounting for the fact that reassociation of the floating mora to the preceding consonant (seen in (7a,b)) was not the option selected in (4) (the analysis which Repetti proposes for lengthening in words going back to a /’CVCV/ Proto-Romance etymon, in both Milanese and Friulian). Consider now Prieto’s (2000) analysis of Milanese vowel length, already partly sketched in Prieto (1994: 101). This is couched within Optimality Theory, and derives the Milanese facts making use of the three constraints in (8a–c):

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Competing analyses of the rise of CVL (8)

127

a. F(oo)t-Bin(arity): ‘Feet should be analyzable as binary.’5 b. Hnuc: ‘A higher sonority nucleus is more harmonic than one of lower sonority.’ c. Fill: ‘Syllable positions are filled with segmental material.’

The appropriate ranking of these constraints accounts for the Milanese data. In (9) I reproduce the inputs assumed by Prieto. Note that, again, VL is not included in these inputs, though at the surface [’nø :v] has a long vowel, whereas the remaining words in (9) have short vowels: (9)

/nøv/ ‘new’

6¼

/myf/ ‘mouldy’, /nøva/ ‘new.fsg’, /uni/ ‘unique.fsg’

Consider first the account proposed for polysyllabic words (Prieto 2000: 267): (10)

i.

/nøva/ a. [(ˈnøːva)] b. [(ˈnøva)]

Ft-Bin *!

ii.

/unika/ a. [(ˈuːni)ka]

Ft-Bin *!

b. [(ˈuni)ka]

As already said, all inputs have short vowels throughout. Given the extrametricality assumed for the last syllables of proparoxytones (as in /uni/), both cases are alike, metrically. In the absence of lengthening, a bimoraic trochee emerges as the metrical structure of the winning candidates (10b). Lengthening, on the other hand, would yield a less harmonic foot (a trimoraic trochee), which incurs a fatal violation of Ft-Bin. Consider now monosyllables—or, more generally, oxytonic words, which provide the only environment in which VL is contrastive at the surface. Here, Prieto (2000: 264) reverses Montreuil’s account: [W]e are not dealing with a process of optimization at the syllable level (through Bimoraic Enforcement), but at a higher level, namely, at the level of the foot.

At the foot level, Ft-Bin does part of the job, but in order to discriminate between monosyllables with long vs short stressed vowels, a representational adjustment is needed, which is opposite to the one assumed by Montreuil in (6a,b): Only voiced consonants will be able to license a moraic unit, with the exception of special cases like final /n/ or /l/, in line with the crosslinguistic observation that sonorous consonants tend to have more prosodic weight. (Prieto 2000: 267)

5 This is Prince’s (1990: 360) formalization of the ban against monomoraic feet (trochee: bimoraic [σμμ] or [σσ] > trimoraic [σμμσ] > monomoraic [σ]). For (8b) see Prince and Smolensky (1993: 16; 2004: 20).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

128

Analysis of Northern Romance vowel length

In this way, the voiced obstruent in /nøv/ ((11a)) gets a mora, while the voiceless one in /myf/ ((11b)) gets none: (11)

a.

b. µ µ

µ

/nøv/

/myf/

Given this representational contrast, the tableaux in (12) derive the correct result: (12) i.

μ /nøv/

Ft-Bin

Hnuc

a. [(ˈnøːv)]

ø–ø

b. [(ˈnøv)] *!

ø–v ø

Fill

Ft-Bin

c. [(ˈnø)v]

ii.

Hnuc

/myf/ a. [(ˈmyː)f] b. [(ˈmy)f] c. [(ˈmyf)]

*!

y–y *!

y y–f

In /nøv/ the selective constraint is HigherSonorityNucleus: the mora contributed by the voiced consonant has to be parsed into the structure, and this constraint states that the vowel is a better candidate for this parsing (as in (12i.a)). In /myf/, conversely, the final consonant carries no mora, by hypothesis. Therefore, lengthening (shown in the (12ii.a) candidate) would crucially violate Fill, by adding segmental material in the output. Prieto’s conclusion, as for VL, is the following: The analysis offers a principled motivation for the fact that vowels are only lengthened in final syllables. In particular, the constraint explains why vowels are lengthened in this context and stay short when in penultimate or antepenultimate positions. (Prieto 2000: 270; emphasis added)

As highlighted in the italicized part of this quotation, this analysis assumes a vowel lengthening, which applies synchronically in Milanese, thus paralleling the other approaches just discussed. This common trait will have to be kept in mind for the discussion in }4.3. Consider also that the enforcement of lengthening via Ft-Bin implies that this lengthening may only have arisen after apocope took place. This is a feature common to Montreuil’s and Repetti’s accounts as well, with which Prieto’s

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Competing analyses of the rise of CVL

129

also shares the impossibility of coping with the occurrence of CVL word-finally, where Ft-Bin could not possibily intervene in an undifferentiated input /kan’ta/ to select both [kan’ta :] ‘sung’ and [kan’ta] ‘sing.inf’, which both occur (and contrast) at the surface, as seen in }3.4.1.4, (45). In sum, none of these treatments provides for a link between Milanese VL and PRom OSL. Under this kind of account, the fact that (as shown in Chapter 2) in Late Latin the word novum came to be pronounced as [’nO :vum] (as opposed to CL [’nOvum]), and the fact that its Milanese outcome [’nø :v] displays a long vowel in the same position, are unrelated—a mere coincidence, which indeed mostly goes unnoticed in the cited papers. 4.1.2 Competing explanations of the rise and status of VL in Friulian In studies on Friulian there is also a wide consensus that CVL in the first members of minimal pairs such as those seen in }3.4.1.5, (58) is not to be traced back to Late Latin OSL, as stated explicitly by Repetti (1992: 165): ‘the bimoraic norm is not part of Friulian historical grammar’. More often, however, this disjunction is left implicit in the treatment of Friulian VL, both in studies in comparative and diachronic dialectology (first and foremost Francescato 1966: 130-35) and in work in theoretical phonology (most recently, Iosad 2012), which converge in sharing this implication in spite of the many differences among them, highlighted in Vanelli’s (1998: 74-89) thorough review of the literature on the topic. Much as has been seen in reviewing the different analyses proposed for Milanese, for Friulian too VL is derived through some mechanism which implies that it arose anew in Friulian: crucially, after final vowels were deleted. Francescato (1966: 134) assumes that, after WRom. intervocalic obstruent voicing (latum > *lado), apocope and degemination (lacte > *latte > *late) applied, without however neutralizing minimal pairs such as those in (13), which he reconstructs for a preliterary stage of Old Friulian: (13) Proto-Friulian according to Francescato (1966) a. [’lat] ‘gone’ 6¼ b. [’laT] ‘milk’ [’pas] ‘peace’ [’paS] ‘step’ [’det] ‘finger’ [’deT] ‘said’ In his notation, upper case stands for a consonant with ‘stronger articulation’, which kept the preceding vowel short whereas ‘weaker’ consonants left the vowel free to lengthen, in a following step, thus resulting in today’s contrast [’la :t] vs [’lat]. Empirically, this coincides with Iosad’s (2012: 922) OT account, according to which ‘obstruents which are the outcome of final devoicing are phonologically distinct from true voiceless obstruents, being completely unspecified for laryngeal features’. In Iosad’s view, it is the ‘representational deficiency of such delaryngealized

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

130

Analysis of Northern Romance vowel length

obstruents [ . . . ] which opens the way to vowel lengthening’ before such obstruents, given ‘their inability to license a mora’. This lengthening, in turn, is represented (much as in Prieto’s 2000 analysis of Milanese discussed in }4.1.1) as complying with a Ft-Bin constraint. Note that, since VL is contrastive before (some) sonorants as well, an entirely different explanation is proposed to account for VL in [’va :l] ‘cost.3sg’ < valet (vs [’val] ‘valley’ < vallem),6 since the lengthening is analysed here ‘as stemming from a binarity requirement, which rules out monomoraic candidates. Given a choice between two bimoraic candidates, the computation selects the candidate with both morae associated to the vowel’ (Iosad 2012: 933). In neither case is VL in the input: rather, long vowels are held to ‘derive from underlyingly short vowels’ (Iosad 2012: 933). Several other analyses proposed earlier for Friulian VL—despite differences in the details, discussed in Vanelli (1998: 74–89)—concur in claiming that what is contrastive in Friulian is not VL, but rather some property of the following consonants, however defined exactly, which is held to determine the length of the preceding vowel. This property, in work in this line of research, has been defined in very different ways. Thus, for Iosad (2012) vowel lengthening is ultimately due to the representational impossibility for the final consonant in words such as [’la :t] ‘gone.msg’ to bear a mora (see above), whereas for Vanelli (1979) and Hualde (1990), on the contrary, the same lengthening is triggered because the stressed vowel inherits a mora from the following underlyingly voiced obstruent as the latter gets devoiced at the surface (cf. Hualde 1990: 43–4), as shown in (14): (14)

[’la :t] ‘gone’

/’lad/ Derivation:

devoicing

reassociation

μ μ l a d O

R

μ μ →

l

a

O

R

t

μ μ → [ˈlaːt]

This does not happen in words like [’lat] ‘milk’, whose voiceless final consonant is assumed not to bear any mora.

6 This is the same problem which encumbers Vanelli’s (1979: 72–5) analysis: whilst she maintains that VL before obstruents was triggered by final devoicing, she has to admit that before /l/ the very same effect followed from an entirely unrelated cause, i.e. that CVL before /l/ is the diachronic successor of the PRom allophonic length difference arisen via OSL. Obviously, since OSL can explain the rise of VL before obstruents as well, it is more economical to have just one explanation for one and the same VL contrast in different environments—and one, into the bargain, which accounts for all Northern Romance varieties where VL appears, whereas CoL for final devoicing cannot. Haiman and Benincà (1992: 42) call this descriptive problem ‘one relative implausibility’ of the explanation appealing to final devoicing, which they nonetheless stick to. I would rather term this implausibility ‘absolute’ rather than ‘relative’, as expounded in }4.3 below.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Competing analyses of the rise of CVL

131

Vanelli (1998: 102) and Baroni and Vanelli (2000) ascribe the length of the stressed vowel in [’la :t] ‘gone’ to the fact—demonstrated through experimental phonetic measurements—that final obstruents in these words are shorter than those in words whose final consonant has been voiceless throughout (like [’lat] ‘milk’). This is in turn conceived of as an effect on the phonology of Friulian of a well-known general phonetic constraint, since it has been shown in several studies of many languages that vowels before voiced consonants tend to be longer, ceteris paribus, than vowels preceding voiceless obstruents (cf. e.g. Jespersen 1932: 182; Chen 1970; Kluender et al. 1988; Fowler 1991; Maddieson 1997: 625; Piroth and Janker 2004: 102–3). This constraint is reported to have given rise to contrasts comparable to that occurring in Friulian: for instance, in Appalachian English bed vs bet contrast as [’be :t] vs [’bet] after the application of final devoicing (Hock 1991: 139; Blevins 2004: 274–5). Still another view on the rise of CVL in Friulian, defended by Repetti (1992: 165) and Prieto (1992), assumes CoL, though not for the loss of voicing of the final consonant but rather ‘as a direct prosodic consequence of the deletion of the final non-low vowels’ (Prieto 1992: 206), along lines similar to what has been seen for Milanese in (4) above.7 In some cases, the two putative causes for VL, deletion of the final vowel and devoicing of the final consonant, are lumped together, as seen in Miotti (2007: 80, n. 11): ‘Com’è noto, la caduta delle vocali finali latine (eccetto -a) avrebbe provocato l’allungamento della vocale accentata, se seguita da consonante sonora’ [‘As is well known, deletion of Latin final vowels (except -a) would have induced lengthening of the stressed vowel, if followed by a voiced consonant’]. His ‘com’è noto’ is symptomatic for the general consensus among specialists of Friulian, on the assumption that long stressed vowels derive from a lengthening which took place after the final vowels were deleted, and hence has nothing to do, historically, with Late Latin OSL.8 In comparison with the substantial number of studies just reviewed, those that maintain that Friulian contrastive VL ultimately goes back to OSL are a tiny minority (Lüdtke 1956: 268; Leonard 1972: 67; Morin 2003: 129; Loporcaro 2007a).9 A corollary of this contrast in the analysis of the data is the following: the assumption that Friulian VL derives from OSL automatically implies that this is a diachronic change which effected, centuries ago, a restructuring in the underlying

See Vanelli’s (1998: 76–81) criticism of the explanations assuming CoL for the loss of the final vowel in Friulian. 8 Among the many studies on Friulian endorsing this view, see also Pellegrini (1982: 17), Rizzolatti (1981: 20), or more recently, Heinemann (2003: 45–50), Marcato (2008: 2751). 9 Earlier, Morin (1992: 82) ventured yet another explanation, claiming that a final unstressed -[a] must have prevented lengthening: ‘Early Romance vowels were lengthened in paroxytones when they were: a. in an open syllable, and b. not followed by a low vowel.’ This position was proved untenable by Vanelli (1998: 81–2). 7

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

132

Analysis of Northern Romance vowel length

form (as no open syllable is there, synchronically, in, say, [’la :t] ‘gone’). On the other hand, most of the more recent generative analyses for which VL depends on the following consonant postulate lengthening as a part of the synchronic competence of speakers of Friulian, similarly to what we have seen for Milanese in }4.1.1.10 We shall return to this distinction in }4.3. 4.1.3 Alternative formal accounts for the rise of VL in Cremonese As a final illustration of the oblivion which has befallen OSL in the literature of theoretical phonology on Northern Romance varieties over the past two decades or so, consider Prieto’s (1994) account of VL in Cremonese. Here too, VL is treated as derived synchronically, rather than as underlying. According to her proposal, as schematized in (15), length in Cremonese arose as a product of a prosodic change (which she terms Foot Expansion) within the history of that variety: (15) Prieto (1994: 92): Foot Expansion in early Cremonese [μ] [μ μ] a. Foot structure á. ‘wing’ mó. bi. ‘mobile’ b. Foot Expansion [μμ] n.a. [μ] ! [μμ] á. la Output: [á :la] [móbile]

[μμ] vák. ‘cow’ n.a. [vákka]

Note that the synchronic derivation in (15b) is similar to OSL in Italian (Ch. 3 (3)), except that it excludes categorically proparoxytones. Both derivations operate on a monomoraic input, i.e. on an underlyingly short stressed vowel (see (5a) above). In order to achieve this result, (15) has to assume an underlying geminate consonant in / ’vakka/ (actually, phonetically [’vaka]; cf. Heilmann and Oneda 1976: 379). Even so, the account of VL in (15) seems empirically inadequate, since it cannot possibly cope with the fact (not considered at all by Prieto) that contrastively long vowels do occur, albeit rarely, in the stressed syllable of Cremonese proparoxytones, as seen in (16a) (cf. Heilmann and Oneda 1976: 102, 226, 353): (16)

a. contrastive VL in Cremonese proparoxytones: [’ta :vula] ‘table’ vs [’fabula] ‘tale’; b. context-dependent long vowels: [’ta :rtera] ‘a cake (made of eggs, milk, and sugar)’, [’pa :skuli] ‘pasture.1sg’.

In addition, as shown in (16b), stressed long vowels occur in proparoxytones also in segmental environments which force lengthening (on which see }3.4.1.1, n. 38). Neither (16a) nor (16b) should be the case if long vowels in Cremonese were indeed 10

This does not apply to classical structuralist analyses like that of Bender et al. (1952), nor to the early generative treatment by Trumper (1975: 69), the latter assuming that Friulian long vowels are underlyingly geminated /VV/.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Diachronic phonology and generative grammar

133

synchronically derived by Foot Expansion, since under Prieto’s account, the input representations to be assumed for the proparoxytones in (16a,b)—given (15a)—are legal and hence do not need to be repaired via Foot Expansion. Obviously, VL in [’ta :vula] (16a) must be specified in the input, and once this is done for a proparoxytonic word, it is only natural to conclude that [’a :la] also has an underlying long (or bimoraic) vowel, unlike in Prieto’s analysis. Otherwise, one must multiply the structural mechanisms accounting for long vowels in different words/environments, in violation of Occam’s razor. With (15), this review of analyses dismissing, explicitly or implicitly, a diachronic link between Late Latin OSL and NRom contrastive VL, has come full circle. I started in }4.1.1 by discussing Repetti’s (1992: 180) treatment, which assumes CoL to explain length in Milanese and Friulian ((2)) while accepting that VL in Cremonese ((1)) goes back to OSL, and I concluded with Prieto’s hypothesis, which proposes a specific lengthening mechanism (Foot Expansion) for Cremonese as well. Now, once it is established that OSL was PRom, it follows that length in open syllables in words like [’a :la] may well be inherited and consequently it is uneconomical to assume lengthening (due to Foot Expansion or any other cause) in early Cremonese. This lack of economy, though, does not seem to bother any of the proponents of the analyses reviewed so far. To see why, I shall have to say something about method in diachronic linguistics (}4.2), to then indicate the ways in which work on sound change in generative phonology departs from that well-established method (}4.3). In Chapter 5, I shall further demonstrate that a unitary account of the rise and fall of CVL in NRom is preferable.

4.2 Diachronic phonology, generative grammar, and method in historical linguistics The different and partly alternative proposals reviewed in }4.1 share some basic traits. As already pointed out, none of them provides for a structural and historical connection between the two types of VL system in (1) and (2). As for type (2), in addition, they all assume that VL arose, in one way or another, after apocope, if not directly because of it. This conclusion is reached through a procedure which is common in generative phonology, a research area in which, ever since King (1969), diachronic explanation tends to reduce to the application of internal reconstruction. This applies to diachrony the general attitude towards linguistic analysis which is aptly summarized in the motto ‘Theory and notation are one’.11 Although rigorous formalization is of course desirable, the (undoubted) virtues of synchronic

11 While this mirrors a widespread general stance, the specific wording stems from Morris Halle’s answer to a question posed by Bernard Comrie—the question whether the particular notation adopted was really essential to the explanation—in the discussion of the plenary paper given at the DGfS 21 conference in Konstanz (Feb. 1999), published as Halle (1997).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

134

Analysis of Northern Romance vowel length

formalization should not prevent one from recognizing that this kind of procedure, reducing the study of language change to internal reconstruction, does not do justice to the established method of historical linguistics. This, as pointed out as early as de Saussure (1922 [1979]: 291), does not coincide completely with that for synchronic analysis (even though during the 20th century the former undeniably benefited from importing procedures and concepts from the latter):12 [T]andis que la linguistique synchronique n’admet qu’une seule perspective, celle des sujets parlants, et par conséquent une seule méthode, la linguistique diachronique suppose à la fois une perspective prospective, qui suit le cours du temps, et une perspective rétrospective, qui le remonte [‘While synchronic linguistics allows only one perspective, that of speakers, and as a consequence only one method, diachronic linguistics simultaneously requires both a prospective perspective, which follows the course of time, and a retrospective perspective, which goes back in time.’].

Since diachronic linguistics has two perspectives, any diachronic account must reconcile the evidence from reconstruction (the Saussurean perspective rétrospective)— which in turn consists of two operations, internal and comparative reconstruction—with the evidence from philological inspection of extant relevant records (the perspective prospective). This produces the following checklist for relevant evidence: (17) a. philological evidence; b. comparative reconstruction; c. internal reconstruction. Note that the ordering in (17a–c) reflects a hierarchy of relevance. In particular, as Ringe (2003: 244) puts it, ‘IR [= internal reconstruction] is of limited use in historical linguistics; CR [= comparative reconstruction] is so much more reliable that it is preferred whenever possible.’13 Formally-oriented diachronic analyses such as those reviewed in }4.1 are not in line with (17) as, first, they do not check the philological evidence from past stages of the languages at issue, and so, for instance, they flatly ignore the evidence on Old Milanese (cf. }5.2.2 below for discussion of that evidence) when formalizing an alleged change from Latin to modern Milanese; and second, they do not have recourse to comparative reconstruction at all. The latter is understandable, given that generative

12 To mention just two examples: Hoenigswald (1950) introduced into diachronic linguistics, for the reconstruction of proto-languages, the discovery procedure employed by American structuralism to work out the phonemes of a language in synchronic analysis. Kiparsky (1965), on the other hand, as Blevins (2004: 66) puts it, ‘by distinguishing innovation from restructuring [ . . . ] introduced the synchronic distinction of competence and performance into the realm of sound change’. 13 The priority of comparative reconstruction is argued for as early as e.g. Paul (1880: 20).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Diachronic phonology and generative grammar

135

phonology is primarily a method for syncrhronic analysis, and that ‘IR replicates phonological analysis point for point’ (Ringe 2003: 246, on final devoicing in German). Ignoring the evidence from past stages of the languages, on the other hand, is just a matter of lack of competence, and patently leads to incorrect conclusions, as shown for Milanese in }5.2.2. In fact, any claims in the realm of diachrony should be checked against the different sources of evidence listed in (17a–c), considering, first, that any conclusion reached on reconstructive grounds ((17b,c)) must be compatible with the evidence delivered by the philological record ((17a)), and second, that no conclusion reached through IR alone ((17c)) can be valid if it is contradicted by evidence from higherranked (17a,b). By way of illustration, consider the history of a phonological process occurring in Standard Italian, which has attracted much interest and been subjected to several different formalizations within generative phonology, raddoppiamento fonosintattico (RF). In Italian, gemination of an initial consonant is regularly triggered when the preceding word ends in a stressed vowel ((18a)). On the other hand, RF is also triggered by a closed list of unstressed monosyllables whose Latin etymon ended in a consonant that got assimilated in external sandhi ((18b)):14 (18) a. regular RF: tu [d :]ici < tu dicis ‘you say’ [/ all stressed monosyllables _ ] b. irregular RF: e [t :]u < et tu ‘and you’ [/ some unstressed monosyllables _ ] Work on RF in Generative Phonology (e.g. Saltarelli 1970a; 1970b; 1983; Vogel 1978; 1982: 66ff., Chierchia 1986; Kaye et al. 1990: 206) focused on regular stressconditioned RF (18a), and derived the RF facts by means of a Well-Formedness Condition (WFC) on the structure of stressed syllables, the same responsible for word-internal OSL, or, more precisely, for the lack of lengthening in closed syllables, with the ensuing complementary distribution of vowel length considered above in }3.2, (2). This synchronic analysis has been extended to diachrony by Vincent (1988) and Repetti (1991). Focusing on regular RF, they put forward a formal account thereof and, based on that formalization, claim that RF arose as a by-product of the collapse of Latin contrastive VL. They assume that strings consisting of a stressed short vowel plus a short consonant became illegal also at word boundaries, in the same way as happened word-internally, as seen above in }1.2, (10i). The WFC account is therefore directly transferred from the synchronic analysis of modern Standard Italian to the explanation of the sound change responsible for the rise of RF. According to Repetti (1991: 311), for instance, RF in tu [d :]ici ‘you.sg say’ started to apply as the second mora associated with long /u :/ in Latin /’tu :/ ‘you.sg’ was delinked from the segmental content of the vowel, thus becoming available for reassociation with a following

14

This description is simplified for the sake of expository clarity.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

136

Analysis of Northern Romance vowel length

initial consonant. As argued in Loporcaro (1997: 50–52), this is at odds with the available evidence: had Lat. /’tu :/ become monomoraic /’tu/ before the collapse of CL CVL, the Italian 2sg personal pronoun would have ended up as (non-existent) **[’to], by the diachronic evolution of the vowel system seen in }2.2.2, (15). More generally, in Loporcaro (1988b; 1997) I have shown that this kind of internal reconstruction, which centres on (Tuscan-based) Standard Italian, is demonstrably ill-founded. In fact, going through the checklist of the relevant kinds of evidence in (17a–c), one discovers that the diachronic source of irregular RF, viz. full assimilation of final consonants in external sandhi, is widely attested in the Latin record (e.g. suc cura CIL VI 9502-(3), instead of sub cura ‘by’) while there is no trace in Latin of RF after stressed final vowels not followed by a final consonant (e.g. **tu ffacis, **qui ddicit). In addition, it turns out that apart from Tuscan (and Tuscan-influenced Corsican and Romanesco), the remaining dialects showing RF all over southern Italy and Sardinia lack stress-conditioned RF (18a). Thus, both the philological evidence (17a) and comparative reconstruction (17b) militate in favour of an alternative explanation for the rise of RF.15 First of all, irregular RF (18b) must have arisen earlier than regular RF. The latter, consequently, cannot date back to the Latin– Romance transition, otherwise it would be shared by all dialects of central–southern Italy and Sardinia as well, which preserve consonant gemination and irregular RF. Rather, it is best conceived of as a later development, peculiar to Tuscan (and, possibly, parts of Western Romance: see }5.3.1, (59)–(62) and }5.3.5, n. 56), which arose via the reanalysis of the pre-existing irregular RF in the trigger context where final stress co-occurred with a final consonant. Elementary though this reasoning might appear to anyone trained in historical linguistics, one should not underestimate the difficulty of convincing zealous advocates of the ‘generative enterprise’ of its soundness. This difficulty stems from a notorious misconception of the circuit between data and theory, as illustrated in Ludlow’s (2011: 85) recent dismissal of the above account of the rise of RF. In a section ‘On the alleged priority of the data’, he mentions the argument according to which Sardinian and southern Italian data disconfirm the claim that RF arose in the Latin– Romance transition as a product of a WFC of the sort mentioned above. His comment is the following: when Loporcaro (1989: 343)16 contends that ‘although descriptive simplicity is a desirable goal, it should not be attained at the cost of contradicting actual linguistic data,’ he is mistaken. 15 Formally-oriented studies seem to have no sense for this kind of empirical argument, as attested by all sorts of wrong claims in the literature. Thus e.g. Nespor and Vogel (1979: 479) claim: ‘While the specific phonological conditions vary to some extent according to region, the one condition that always causes RS in all the dialects in which it occurs is that w1 ends in a vowel that bears the primary word stress.’ It is only possible to make such a claim by ignoring the available primary data on central–southern dialects, as abundantly described in the published literature on the dialects of Italy. 16 Actually Loporcaro (1988b); dates too are data, and, to Ludlow, do not appear to matter exceedingly.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Diachronic phonology and generative grammar

137

The mistake lies allegedly in not perceiving the following: If the ‘descriptive simplicity’ enables the successful reduction of linguistics to more basic sciences then we should be prepared to take the ‘contradictory data’ as puzzles remaining to be explained rather than grounds to abandon the theory.

I find it hard to see what kind of ‘reduction of linguistics to more basic sciences’ would be in sight if one chose one hypothesis on the rise of RF over another, competing one. In this specific case, contrary to Ludlow’s claims, it is hardly at issue that one should accord priority to raw data over theories organizing them, given that, as Ludlow (2011: 83) rightly remarks, there is no ‘theory-free data gathering’ but rather ‘all observation must be for or against some view if it is to be of any service’ (a statement by Charles Darwin in a letter to Henry Fawcett). This is precisely the point: the fact that only irregular RF (18b) is documented in the Latin sources and in most of the dialects today preserving phonemic consonant gemination argues for the view that regular RF ((18a)) is a later innovation of Tuscan, and against the alternative view which conceives of it as a by-product of the demise of CL CVL. In other words, this is not a question about theories, nor, even worse, a matter of having a theory rather than gathering data just for the sake of it, as Ludlow implies:17 it is a choice between conflicting hypotheses, based on the evidence available. Such bold claims as Ludlow’s simply mix up general theory (or, even, ‘science’) with specific hypotheses, and in doing so they free specific hypotheses from the burden of complying with observational data, in obedience to the principles of Aristotelian logic. The method synthesized in (17a–c) is meant to ensure that arguments in diachronic linguistics avoid this kind of logical fallacy and do not incur contradictory statements. Indeed, as I intend to show in Chapter 5, application of the standard method of historical linguistics, as synthesized in (17a–c), leads to conclusions similar to those now summarized for the origin of RF, also as far as the rise and fall of CVL in Northern Romance is concerned: it shows that in those systems CVL is unlikely to have first arisen through any of the changes formalized in the proposals reviewed in }4.1, in spite of the fact that this or that formalism may elegantly describe the way a lengthening process could in principle have applied (via CoL for the loss of a final vowel, Foot Expansion, the inability for final consonants unspecified for laryngeal features to license a mora, or the like). The crucial reason to maintain this is that both comparative reconstruction and inspection of the philological record—i.e. both (17a) and (17b)—tell us that (phonetic) vowel length was there from the outset. The philological evidence has been already scrutinized while discussing the Latin– Romance transition in Chapter 2. There, it was shown that OSL arose in Late Latin He airs the common misrepresentation according to which ‘[s]ociologically, linguistic theory in general has been split since the 1950s between those linguists that are “data first” people [the others, of course—M.L.] and those that advocate “data in the service of theory” ’ (Ludlow 2011: 82). 17

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

138

Analysis of Northern Romance vowel length

and became established before the fall of the Western Empire, and that this conclusion—crucially supported by the evidence from metrical inscriptions discussed in }}2.3.1 and 2.3.2 following Herman (1982) and Adams (1999)—yields the most economical explanation for the demise of CL CVL. Thus, OSL can be ascribed to PRom, which in this respect is faithfully mirrored to this day by Romance varieties of type (1a) (Chapter 3), like Italian (including Tuscan as well as most of the central– southern Italian dialects) and Sardinian. The next step, crucial to the establishment of CVL, was degemination, as hinted at in }3.3 and seen in more detail, for GalloRomance, in }3.4.2.1 (a point to be resumed in }5.3.4). Once one assumes that long vowels must have been there from the outset, in the positions where we find them today in at least some Northern Romance varieties, the next step is a chronological ordering of the two distributions of VL, (1) vs (2), contrasted earlier in this chapter: under this scenario, (1) must be the more conservative. Indeed, closer inspection of dialect variation across NItRom strongly confirms that (2) is a further evolution of (1), and that both are part of a broader picture, representing—as said in (3a) above—two successive steps along one and the same diachronic path, to be illustrated in detail in Chapter 5.

4.3 Too much synchrony into diachrony, too much diachrony into synchrony Before moving forward to the demonstration in Chapter 5, however, a final point of method needs to be addressed here. One crucial difference between the account of VL in Northern Romance proposed here and the alternative ones discussed earlier in this chapter (}4.1) is that most of the latter choose to represent VL in the varieties at issue as a property which is not present underlyingly but is either derived by rule—in older, rule-based accounts such as Vanelli (1979), Montreuil (1991), Repetti (1992)—or is encoded in the winning candidate output but is not specified in the input, in more recent OT analyses, from Prieto (2000) to Iosad (2012). The view defended here is much more traditional: VL in these dialects is contrastive at the surface, and hence must be encoded underlyingly as a /V :/ vs /V/ contrast. To the reader familiar with the history of the debate stirred by the generative turn in phonology in the 1960s, it might seem that the choice whether to add complexity to the representation or to rules (or, in more theory-neutral terms, to the computation) is just a matter of taste, or of the priorities set by the model adopted. My contention is that this is not the case, and that the reconstruction advocated here has to be preferred since it is, on all accounts, more economical than alternative ones. Several arguments have been already hinted at in }4.1, while reviewing the different analyses. These are now put together so as to show that an alternative account is desirable.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Too much synchrony, too much diachrony

139

The idea that CVL in Northern Romance is ultimately rooted in Proto-Romance OSL, and that it arose via phonologization of the output of this lengthening, has a number of implications. One is that CVL must have first arisen word-internally, where the stressed vowel was followed by consonant segments. Indeed, as we shall see in }5.4.2, the contrast occurs in word-final position in some Northern Romance varieties but not in others, and where it occurs word-finally, it does so in ways that differ widely across dialects. This comes as no surprise, under my account: once /V :/ vs /V/ became established as an underlying contrast, the system was liable to exploit it also in new contexts, different from the word-internal one where it first arose. On the contrary, in the competing accounts reviewed in }4.1, for which VL is still derived synchronically, these facts are hardly explicable. Assuming that the long vowel in, say, Milanese [’fy :s] ‘spindle’ or Friulian [’la :t] ‘gone’ is underlyingly short (or unspecified for length) and then becomes lengthened synchronically, either because of the devoicing of the final obstruent (e.g. Vanelli 1979; 1998; Yamamoto 1991: 652; Baroni and Vanelli 1999; 2000) or because of an SRC (Montreuil 1991) or from any other contextual factor, precludes the possibility of accounting jointly for the contrastively long vowel in, say, Milanese [ri’dy :] ‘laughed’ or Friulian [can’ta :] ‘sing. inf’, where no contextual conditioning factor is at work. This is clearly a step backwards with respect to Martinet (1956: 75; 1975: 205), who recognized the occurrence in word-final position as criterial for the diagnosis of the phonologization of CVL—i.e. in post-structuralist terms, of its becoming a property of underlying representation via restructuring. Formally refined though those generative treatments are, their empirical scope is narrower than that of classical structuralist analyses (see e.g. Morin 2012: 124–5 for a similar point on the ignorance of previous structuralist work on the topic, specifically by de Chene 1979 and de Chene and Anderson 1979). Another positive implication of the hypothesis that CVL is the diachronic successor of OSL is that this accounts uniformly for CVL in all words which originally shared the ’CVCV shape. As I showed in }4.1.2, n. 6, this simple result is beyond the reach of hypotheses appealing to the influence of the following consonant, which have to make up different stories in order to account for CVL before sonorants and before obstruents as well as for CVL before different classes of obstruents. Thus, all accounts claiming that CVL in Friulian (a) is derived synchronically and (b) depends on the devoicing of the obstruent following the stressed vowel (see }4.1.1) have to propose a special explanation for the lack of VL alternation in (19a,b) (see Finco 2009: 56): (19) a. [’mjetʃ] [’vjatʃ] [avan’tatʃ]

‘half.msg’ ‘journey’ ‘advantage’

b. [’mjedʒe] [vja’dʒa :] [avanta’dʒa :]

‘half.fsg’ ‘travel.inf’ ‘advantage.inf’

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

140

Analysis of Northern Romance vowel length

As is readily apparent from the alternants in (19b), these morphemes contain an underlying voiced affricate, which is devoiced word-finally in (19a) without this resulting in vowel lengthening. Vanelli (1998: 77) and Baroni and Vanelli (2000: 15, 19) assume that in [’mjetʃ] etc. lengthening does not apply because the affricate is longer than other voiced obstruents, which could appear plausible from a crosslinguistic perspective. Yet, this claim is at odds with the fact that the same affricate does not prevent lengthening of the preceding vowels in Italian loanwords in Central Friulian like [’re :tʃ] ‘hold.3sg’, [’fri :tʃ] ‘fry.3sg’ (vs [’redʒi] ‘hold.inf’, [’fridʒi] ‘fry. inf’; cf. Baroni and Vanelli 2000: 24; Finco 2009: 56). Since Baroni and Vanelli regard devoicing and lengthening as synchronic processes, these data—as noted by Morin (2012: 132, n. 13)—are at odds with their appeal to the intrinsic length of affricates as a phonetic explanatory factor for the shortness of the stressed vowels in (19a). But even if this account of vowel shortness in (19a) were on the right track, the problem is that this would add yet another distinct ad hoc explanation, as seen in (20): (20) a. b. c. d.

Phenomena Long vowel in [’la :t] ‘gone’ (Ch. 3, (58a)) Long vowel in [’va :l] ‘valley’ (cf. }4.1.2 and n. 6) Short vowel in [’lat] ‘milk’ (Ch. 3, (58b)) Short vowel in [’mjetʃ] ‘half. msg’ ((19a))

Causes CoL for devoicing of final /d/ OSL (cf. Vanelli 1979: 72–5) Underlying voicelessness of the final obstruent Intrinsic length of the following affricate, albeit devoiced

Thus, even considering that (20a,c) are the two opposite specifications of one and the same factor, we have here three unrelated explanations for VL in Friulian in different environments, in blatant violation of Occam’s razor. Into the bargain, these explanations cannot handle the parallel data from related dialects, like the Milanese ones in Chapter 3, (46b) and (49), here repeated in (21): (21)

a. [’mez] [’gøb] [’fretʃ] [’vetʃ]

‘half.msg’ ‘hunchbacked.msg’ ‘cold.msg’ ‘old.msg’

b. [’meza] [’gøba] [’fredʒa] [’vedʒa]

‘half.fsg’ ‘hunchbacked.fsg’ ‘cold.fsg’ ‘old.fsg’

This is a widespread situation which occurs in dialect after dialect across Northern Italo-Romance, as shown by the similar data already mentioned for Cremonese in }3.4.1.1, (22), and by those from rural Bresciano to be discussed in }5.1.3, (26). In sum, as argued by Morin (2003: 129–30), these data are much more economically explained under the hypothesis that CVL in Northern Italo-Romance derives originally from OSL: all short vowels in (19)–(21) uniformly go back to a stressed vowel

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Too much synchrony, too much diachrony

141

in a Proto-Romance closed syllable, in which Proto-Romance allophonic lengthening did not apply in the first place.18 This is not to say that the voice of the following consonant never plays a role in determining VL. Of course, as has been shown in }3.4.1 for many dialects, the affinity between length of the stressed vowel and a following voiced obstruent (contrasting with that between shortness of the stressed vowel and a following voiceless obstruent) is a genuine one, and there is indeed evidence that occasionally, speakers of northern Italian varieties reanalysed VL as a necessary correlate of the voice of a following consonant. This was the case, for instance, in several dialects spoken in the province of Pavia (southwestern Lombardy), which otherwise show a distribution of CVL identical to Milanese. Consider the data from the three varieties of Lungavilla, Travacò Siccomario, and Sairano in (22)–(24) (see Appendix, Map 3): (22) Lungavilla (northeast of Voghera, Oltrepò Pavese) i. ’CVCV ii. ’CVCCV a. [’lɜ̃ˑnɐ] ‘nose’ [’kaˑnɐ] ‘be born.inf’ [’paˑlɐ] ‘spade’ [’spaˑlɐ] ‘shoulder’

(23)

b.

[’na :z̥] [’fy :z@]

‘nose’ ‘spindle’

[’nas] [’fys]

‘be born.3sg’ ‘be.impf.sbjv.3sg’

c.

[’me :z̥]

‘month’

[’me :z̥]

‘half.m’

Travacò Siccomario (south of Pavia) i. ’CVCV a. [’laˑnɐ] ‘wool’

ii. ’CVCCV [’kaˑnɐ]

‘reed’

b.

[’pa :z̥] [’ta :z̥] [’pe :z̥] [’se :d̥]

‘peace’ ‘be silent.imp.2sg’ ‘weight’ ‘thirst’

[’pas] [’tas] [’pes] [’set]

‘step’ ‘badger’ ‘fish’ ‘seven’

c.

[’me :z̥]

‘month’

[’me :z̥] [’gø :b̥] [un ’rO :b̥] [’ve :dʒ]

‘half.m’ ‘hunchbacked.m’ ‘one thing’ ‘old.m’

18 Though in Latin the stressed vowels of, say, medius ‘half, middle’, uetulus ‘old’ or frigidus ‘cold’ all occurred in an open syllable, a combination of changes modified syllable structure making all of those stressed syllables checked, so that the Romance forms involved went through intermediate *med(d)ju, ueclus ‘old.msg’ (attested in the Appendix Probi tertia, Rome, mid 5th c., lemma 5) and fricda ‘cold.fsg’ (Appendix Probi tertia, lemma 54): on the heterosyllabication of the relevant consonant clusters in ProtoRomance, see Pensado Ruiz (1988); Vennemann (1988: 46); Cull (1995: 122–3); Loporcaro (2011c: 99–102).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

142 (24)

Analysis of Northern Romance vowel length Sairano (an outlying part of Zurasco, southwest of Pavia) i. ’CVCV a. [’laˑnë] b. [’pa :z̥] [’ta :z̥] [’se :d̥] c. [’me :z̥] [’pe :z̥]

‘wool’ ‘peace’ ‘be silent.imp.2sg’ ‘thirst’ ‘month’ ‘weight’

ii. ’CVCCV [’kaˑnë] [’pas] [’tas] [’set] [’me :z̥] [’pe :z̥] [’gø :b̥]

‘reed’ ‘step’ ‘badger’ ‘seven’ ‘half.m’ ‘worse’ ‘hunchbacked.m’

As seen in the (a) examples, paroxytones do not host a VL contrast, which is also neutralized, as in Milanese, in oxytones before /r/: [’ka :r] ‘dear’ = ‘wagon’. Again as in Milanese, the VL contrast is observed in oxytones ((22b)–(24b)), where it correlates with the voice of the following obstruents, which is only partially devoiced prepausally. Unlike Milanese, however, as seen in (22c)–(24c), words ending in a voiced obstruent have systematically a long stressed vowel, including those whose voiced obstruent goes back to a PRom geminate consonant (the (c.ii) case): here too, the stressed vowel must be long, so that e.g. **[’gøb̥] ‘hunchbacked.m’ or **[’mez̥] ‘half ’ are rejected as totally ungrammatical. If all dialects of northern Italy were like this, Francescato’s and Vanelli’s account of Friulian could be extended succesfully to the whole of Northern Italo-Romance. But crucially, this is not the case: the hypothesis that VL first arose because of the devoicing of the following obstruents cannot account for the Milanese ((21)) nor for the Cremonese data, which cannot be explained away by assuming an ad hoc shortening in (21a) either, since this would lack any contextual motivation. Conversely, the hypothesis that CVL arose out of OSL is compatible with all the data: under this hypothesis, the distribution in (21) is the original one, and that attested in the dialects near Pavia ((22)–(24)) is derived straightforwardly from (21), assuming that speakers reanalysed the long vowels in (22b–c.i)–(24b–c.i), first arisen through OSL, as dependent on voice.19 In addition to the impossibility of coping with (21), there are still a number of further empirical shortcomings of the analyses reviewed in }4.1, which have been 19 The change may have gone through a stage in which different speakers categorized VL differently, through what King (1988: 22) calls an ‘ambiguous projection’ in his analysis of OSL in German. While this started out as a lengthening in open stressed syllables (e.g. MHG trăge > early NHG trāge), long vowels came to occur usually before voiced obstruents, so that ‘speakers projected two generalizations [ . . . ] from the data’ (King 1988: 22):

(i) a. Lengthening1: V ! [+ long] in open syllables b. Lengthening2: V ! [+ long] before voiced obstruent Through subsequent changes, then, in King’s (p. 29) reconstruction, ‘Lengthening1 was lost, leaving only the rule lengthening vowels before voiced obstruents (Lengthening2)’.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Too much synchrony, too much diachrony

143

pointed to above: thus, for instance, Prieto’s (1994) analysis of VL in Cremonese excludes categorically the possibility that CVL may ever occur on the stressed syllable of proparoxytones; however, while it is indeed the case that this happens seldom, it does happen (cf. (16a) above and Chapter 3, (25a)), thus flying in the face of Prieto’s categorical ban and calling for a subtler explanation (to be provided in }5.4.1), which may account for what has to be modelled as a gradual, phoneticallybased preference rather than as a categorical property built into the phonological representation. Put briefly, these analyses achieve poor descriptive accuracy, and yet do so with a costly formal machinery: for some, VL depends on the fact that—by assumption— one mora is carried by voiceless, not by voiced obstruents (Montreuil 1991; Prieto 1992; Iosad 2012); for some others, on the contrary, it is the other way round, so that voiced obstruents carry a mora whereas voiceless ones do not (Hualde 1990; Vanelli 1998). These and similar arbitrary assumptions go hand in hand with likewise arbitrary representational adjustments, abundantly exemplified in }4.1 above, like assuming geminate (i.e. underlyingly moraic) consonants (Montreuil 1991; Prieto 1994) where there are none at the surface, and the like. The reason why this is the case is often that, refusing to acknowledge that changes (such as degemination, devoicing, and suchlike) have happened in diachrony and resulted in restructuring, such theories keep on squeezing sound change into the synchronic phonology, in the way familiar from classical analyses in generative phonology from Chomsky and Halle (1968) onwards. Notoriously, in that sort of analysis, because of the conflation of morphophonology with phonology proper, the early Middle English Great Vowel Shift was camouflaged as a synchronic phonological rule of modern English, so as to account for alternations like div[aI]ne/div[I]nity. I have shown in Loporcaro (2011a) that this is not a thing of the past, since this type of abstract analysis is still proposed by practitioners of generative phonology (under the Kiparskyan label ‘opacity’: see e.g. McCarthy 2007: 108; Baković 2011), and is still perceived as the method for phonological analysis in neighbouring fields such as that of psycho/neurolinguistics (Levelt et al. 1999: 37, n. 6): the same underlying word form will surface in rather drastically different ways, depending on the morphonological context (as in period/periodic or divine/divinity), a core issue in modern phonology. These and other phenomena [ . . . ] require rather abstract underlying form representations.

The excesses of this method are notorious: consistent application of the procedure unavoidably leads to undoing sound change up to the earliest reconstructible stage of the language. A telling example is Lightner’s (1975: 621) critique of Chomsky and Halle’s (1968: 233–4) underlying representation /rixt/ for right, which is too concrete in Lightner’s view: ‘A more reasonable lexical representation of the root in right(eous) is -reg- with a meaning something like ‘lead straight, guide, conduct’.’

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

144

Analysis of Northern Romance vowel length

Summing up, the treatments of Northern Romance CVL reviewed in }4.1 suffer from two apparently counterposed yet deeply intertwined shortcomings. On the one hand, they cram too much diachrony into synchrony, in terms of data, since their underlying representations—with no CVL where the language has long acquired one by Martinet’s criterion, or with contrastive geminates where the language has long lost them—squeeze diachronic change into the synchronic structure of the language at issue, be it modelled as derivation by rule or as an input–output relationship. On the other hand, they put too much synchrony into diachrony as well, in terms of method, in that they extend synchronic analysis (in the form of internal reconstruction, which, as we saw in }4.2, ‘replicates phonological analysis point for point’: Ringe 2003: 246) far beyond its legitimate province, without bothering to check the superordinate evidence (17a,b). Once one accords precedence to comparative over internal reconstruction, though, it becomes clear that it is desirable to identify one explanation, for example for the facts in (19)–(21) above, rather than construct a list of disparate causes for what are evidently manifestations of one and the same empirical phenomenon within the same variety or across dialects. Likewise, once one agrees on the elementary procedural maxim that ascertained facts cannot be dismissed (pace Ludlow 2011: 85, cited in }4.2) and that features documented for past stages of the language(s) at issue must be accommodated in any sensible reconstruction—and that, conversely, reconstructions at odds with such attested facts must be discarded—then it becomes clear that the occurrence of PRom OSL puts certain constraints on the reconstruction of the further development of VL in Northern Romance. Having clarified this, I can now proceed to work out such a detailed reconstruction in Chapter 5.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

5 Dialect variation and comparative reconstruction In this chapter, I shall consider in more detail dialect variation in VL in NRom. Though other Romance varieties will be referred to as well, much as in Chapter 4 the focus will remain primarily on Northern Italo-Romance: indeed, it will turn out that dialect variation across northern Italy is crucial for reconstruction, since it arguably preserves traces of all the stages in the development of CVL. Once this variation is considered in sufficient detail, it becomes increasingly clear that the different systems observed in different dialect areas belong together, in the sense that each of them represents a chapter in a coherent and consistent story, under the title ‘Rise and fall of secondary CVL’. In other words, detailed inspection of dialect variation—and of the spatial distribution of the systems involved—adds up to the arguments already discussed in }4.3 in support of an answer in the positive to the question as to whether the different kinds of VL system nowadays observed are structurally and historically related (}4.1, (3a)). The question was posed there with regard to just two kinds of VL system, the Friulian-Milanese and Cremonese-Emilian types. But I shall show here that CVL comes in many different shades, and that all of them fit into the picture and thus contribute to enhancing its resolution (and with it its plausibility). Thorough inspection of dialect variation over time and space (the latter is visualized by the maps in the Appendix) also has advantages for the analysis, in that data from different dialects at different stages enlighten each other. For instance, while a CoL-based approach to the rise of VL in Milanese (cf. }4.1.1, (4)) may seem at first sight plausible considering the distribution of long vowels in the modern dialect (cf. }3.4.1.4, (46)), comparison with Eastern Lombard, as shown in }5.1.3, (18) and (19), demonstrates beyond any doubt that such an explanation is on the wrong track, as does inspection of the evidence about apocope and VL in older stages of Milanese to be considered in }5.2.2. Another example concerns the occurrence of CVL word-finally: comparison of vowel qualities across Rhaeto- and Northern Italo-Romance shows that these whole groupings must have originally shared with Tuscan the condition (seen in }3.1, (3)) barring lengthening of word-final stressed vowels. Detailed inspection of the dialects spoken on the Tuscan–Emilian border (}5.4.2) shows the steps by which this restriction may have been lost. And the examples could easily be multiplied.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

146

Dialect variation and comparative reconstruction

Let us start by considering one crucial prediction that follows from the different treatments of VL reviewed in }4.1 above, which was not considered in the comparative discussion in }4.3. These treatments assume several entirely distinct changes through which VL may have arisen anew, here and there, within the structural history of the systems under scrutiny. The alternative view defended here, on the contrary, assumes that CVL in all these varieties first arose from one single change: the phonologization of the lengthened output of OSL.1 Under this view, we already have a diachronic source for length in northern Italian dialects (and Northern Romance on the whole), and it is uneconomical to propose alternative ones, unrelated to OSL. PRom was like today’s Standard Italian: subsequently, WRom. degemination applied, levelling out the difference between the two environments in Chapter 3(1i and 1ii), (i.e. open vs closed syllable), and the difference in length became contrastive in the ’CV(:)CV environment. This implies that, of the two types of distribution contrasted in Chapter 4 (1) and (2), here repeated more synthetically in (1a,b) for convenience, the latter must be innovative with respect to the former, which preserves contrastive length in paroxytones (1ii.a): (1)

a. Cremonese b. Milanese i.

σ´

]pw

+

+

ii.

σ´σ

]pw

+

–

[+ = contrastive VL in the given environment] If this is so, then what we have to explain is not lengthening in Milanese or Friulian, which never took place, but rather shortening. Viewed more generally, the two alternative views under scrutiny make two counterposed predictions concerning cross-dialectal variation in VL and the diachronic drift across Romance. If the disparate accounts discussed in }}4.1, 4.3 were on the right track, one could in principle expect VL to be on its way to expansion, since it arose polygenetically in several systems from many dialect areas via several distinct changes which did not 1 This is not to deny that other changes may have contributed to the VL contrast: this can be seen clearly e.g. for Genoese in }3.4.1.3, (37)–(39), where CVL can occur also before stress, a context in which it cannot possibly have been fed by OSL. Similarly, in many Gallo-Romance dialects CoL for coda consonants brought about VL contrasts before stress, as seen e.g. in the Franco-Provençal patois of Montana (mid-Wallis; cf. Gerster 1927: 138), where [ε:’fIn:a] ‘thorn’ < spinam had a long pretonic /e:/ due to coda /s/-deletion, contrasting with short /e/ in e.g [e’fan:a] ‘span’ < spannam. The rise of contrastively long vowels before stress is witnessed also in the history of French (cf. Ch. 3, (67ii)), where vowels became lengthened before /r/ giving rise to (sub)minimal pairs such as OFr. parrin [pa:’rin] vs parent [pa’ɾent], where the contrast in VL ‘deviendra distinctif lorsque [ɾ] > [r]’ [‘will become contrastive when [ɾ] > [r]’] (Morin 2008: 2916). Thus, vowel lengthening may in principle have several different causes in addition to OSL, including CoL for the loss of some other segment, the influence of neighbouring (usually following) consonants, and vowel coalescence, an issue to be returned to in Ch. 6.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

147

affect the underlying representation but rather, for most of these analyses, resulted in productive synchronic phonological rules (or constraint rankings) deriving VL at the phonetic output. Under the view advocated here, on the contrary, CVL first arose as the /’CVCV/ vs /’CVC:V/ distinction was lost, a change which affected underlying representation about one millennium ago. To be sure, there was a first stage of expansion, which is still evidenced to this day by the diverging distributions of CVL in absolute word-final position (cf. }5.4.2). However, in dialects of type (1b), contrastive VL must have withdrawn from the penultimate syllable: this suggests that VL is a recessive rather than an expanding feature. If this is so, we would expect a systematic sifting of geolinguistic variation to reveal more cases of retreat of CVL, both in the spatial dimension (retreat across dialects) and in the structural one (retreat across structural environments). This chapter will be devoted to a quest for more evidence which bears on the issue, by searching Romance varieties not so far considered, in order to compare them with (1a,b) and the others considered up to this point. I shall start with a scrutiny of dialects in which VL can be shown to be beating a retreat (}5.1), turning then to consider the (alleged) relationship between the rise of secondary CVL and apocope (}5.2). Having ascertained that apocope cannot have played the role several authors ascribe to it in the rise of CVL, and hence that degemination must have been responsible for it instead, I shall move on in }5.3 to considering Northern Romance dialects which still to some extent preserve geminate consonants, and the cues they offer for reconstruction of the intermediate steps through which CVL came into being. Section 5.4 will propose a comprehensive reconstruction of the rise and fall of CVL in Northern Romance, accounting for the relationship between all the dialects touched upon: }5.4.1 provides a phonetically-based rationale for the retreat of VL from (1a) to (1b) and, further, for its demise in the dialects considered in }5.1, while }5.4.2 focuses on the extension of CVL to the word-final position.

5.1 At the vanguard of change: the fading of contrastive VL in Northern Italo-Romance It is easy to see which of the two predictions compared above—concerning the expansion vs recessiveness of CVL—is borne out by the inspection of dialect variation across space and time. When introducing the Gallo-Romance data in }3.4.2 above, I have already said that for French, a pre-literary stage with CVL arising from OSL can be reconstructed, and that some scanty remnants of this original layer of VL (the first layer in Morin’s 2006: 148–9 terms; cf. }3.4.2.1, (67i)) can still be spotted in some modern dialects, while they have disappeared completely, over time, in modern Standard French. Likewise, for Rhaeto-Romance and Central Ladin we have already seen in }3.4.3 that some varieties still have CVL to this day whereas others do not, and that CVL is

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

148

Dialect variation and comparative reconstruction

standardly posited as a common feature of the whole of these two dialect groups, prior to internal differentiation (cf. Gsell 2008: 2774; Leonard 1972: 67). In northern Italy too, every single region whose dialects display CVL provides evidence for its gradual retreat. This is clearly visible from the areal pattern observable in Friulian (}5.1.1) and Ligurian and Piedmontese (}5.1.2), as well as in Lombard (}5.1.3). 5.1.1 Loss of CVL in peripheral Friulian dialects While Central Friulian preserves CVL, dialects spoken at the eastern and western fringes of Friuli have lost it, and everything suggests that this is ‘una semplificazione relativamente recente’ [‘a relatively recent simplification’] (Francescato 1963: 74, n. 23), whereas at a previous stage CVL spanned the whole of Friuli. The case of Friuli is somewhat special, because here loss of CVL can possibly be ascribed to contact with Veneto: in fact, CVL is missing in Western Friulian, bordering on Veneto, where the stressed vowels of [’stat] ‘been’ and [’fat] ‘done’ have become homophonous, as in Veneto (cf. Lüdtke 1957: 123). The influence of Veneto was exerted, on the one hand, via sheer geographic contact—with the Veneto-Friulian linguistic border progressively shifting eastwards over time (cf. Lüdtke 1957: 125)—and on the other hand via the ‘parachuting’ of the Venetian prestige model into the main towns of the region, Udine, Pordenone, and Gorizia, which all nowadays include (and have done so for some centuries) an urban variety of Veneto in their verbal repertoire.2 This resulted in a ‘horseshoe’ shape (Francescato 1964: 62) of the isoglosses delimiting Friulian, with Venetan features surrounding the linguistically Friulian area from southwest and southeast. Gorizia is the main town of eastern Friuli, 30 km east of Udine, on the Slovenian border. The Friulian dialect spoken in Gorizia, like Western Friulian at the other end of the region, nowadays has a five-vowel system /i ε a O u/ without CVL, so that the stressed vowels in (2a,b) are identical (cf. Francescato 1966: 359–60; Trumper 1975: 17–18), whereas they contrast in Central Friulian, as seen above in }3.4.1.5: (2) a. [’nas] [’fil] [vi’dut] [’pεl] [’sεt]

‘nose’ < nasum b. [’bras] ‘arm’ < brachium ‘thread’ < filum [’mil] ‘thousand’ < mille ‘seen’ < *vidutum [’dut] ‘all’ < *tot(t)um (cf. It. tutto) ‘hair’ < pilum [u’zεl] ‘bird’ < *au(i)cellum ‘thirst’ < sitim [’frεt] ‘cold’ < frig(i)dum

2 This resulted in outright replacement in Pordenone as well as, on the Adriatic coast, in Trieste and Grado (cf. Zamboni 1980: 85; Frau 1984: 8). Recall that Friuli was for several centuries (1420–1796) under the rule of the Republic of Venice, apart from Gorizia and the easternmost fringe of the region, which belonged to Austria in that period. See Francescato (1966: 20) for the delimitation of the Western and Eastern Friulian areas without CVL.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

149

This merger is already documented in the Goriziano verses by Gian Giuseppe Bosizio (1660–1743), who rhymes cantat ‘sung’ (which has a short vowel in today’s Goriziano but a long one in Central Friulian, [can’ta:t]) and fat ‘done’, which has a short /a/ in both (Francescato and Salimbeni 2004: 204). The same was the case for 18th-century Pordenonese, judging from the rhymes (e.g. nas ‘nose’ : bas ‘low’) in the verses by G. Comini (1722–1812). Here too this may have happened under the pressure of Veneto, which subsequently ousted Friulian in Pordenone (cf. Francescato 1963–6: 63).3 That Goriziano, like the rest of Friulian, must have originally kept apart the pairs of vowel phonemes occurring in (2a) and (2b) (two originally different environments, neutralized by sound change) is demonstrated by the different outcomes of the diphthongized reflexes of PRom /ε O/: thus, for PRom /ε/ one finds /i/ in an originally open syllable (e.g. [’sil] ‘sky’ < caelum) vs /jε/ in an originally closed syllable (e.g. [’pjεl] ‘skin’ < pellem). 5.1.2 The areal pattern of CVL in Liguria and Piedmont In Liguria too, while Genoese (cf. }3.4.1.3) and Central Ligurian display CVL, the dialects spoken on the western fringe of the region (i.e. the Intemelio group, centring on Ventimiglia) lack it, as illustrated by the following data from Ventimigliese (cf. Azaretti 1982: 25): (3)

a. [’sonu] [’nasu]

‘play/sound.1sg’ ‘nose’

< sono < nasum

b. [’sonu] [’galu]

‘sleep’ ‘cock’

< somnum < gallum

According to Forner (1988: 458), this lack of CVL characterizes not only Intemelian, at the western end of the region, but also its eastern neighbour, Western Ligurian (with Imperia, Albenga, etc.): for both groups, VL is said to be subject to a contextual condition: ‘Langvokal vor v, r, ž, z1 (nicht: z2), n1 (nicht: n2 < NN . . . ) und vor Vokal; Kurzvokal sonst’ [‘long vowel before v, r, ž, z1 (not z2), n1 (not n2 < NN . . . ) and prevocalically; short vowel elsewhere’].4 Loss of CVL even creeps, under somewhat special conditions, into Central Ligurian, affecting a ‘displaced’ variety, viz. the dialect of the Genoese enclave of Bonifacio, at the southern tip of Corsica (cf. Ricciardi 1975: 19, 23f.). In Bonifacino, for instance, verb infinitives consistently have a short stressed vowel (e.g. [kan’ta] ‘sing.inf’, [’di] ‘say.inf’, ALEIC II 200, 205) as opposed to Genoese [kaŋ’ta:], [’di:] (Toso 1997a: 170, 205). This dialect was transplanted into Corsica with the Genoese conquest in 1195, but the town was then partly repopulated from Genoa after a plague in 1528 3 Lack of CVL is observed also in the contact varieties of Venetan in the verbal repertoire of Friulian towns (see Francescato 1954–7: 14 on Udine’s Venetan). 4 However, Alpine Intemelian—in a peripheral position in the Alps, bordering on Occitan—does preserve CVL, as shown in Ch. 3, (71) and (72) with examples from the Roiasc dialect of Briga. This contrasts with Intemelian and Western Ligurian, where the contextual conditions on VL just described ousted CVL (though in Western Ligurian too, Alpine dialects preserve CVL: e.g. [’fry:tu] ‘fruit’ vs [’brytu] ‘ugly.msg’ in the dialect of Ormea, province of Cuneo; see Schädel 1903: 25). As said in the conclusion of }3.4.1.3, such contextual conditions on VL are to be found in Central Ligurian (Genoese) as well, though coexisting there with CVL.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

150

Dialect variation and comparative reconstruction

(cf. Bottiglioni 1928: 2; Toso 2008a: 38–9). Since at that time CVL in Genoese had been established for centuries, it seems plausible that Bonifacino shared this feature and then lost it (see also }5.2.1).5 Note that Bonifacino nowadays displays short vowels also in contexts in which Genoese and the rest of Ligurian have long vowels which arose through later changes (distinct from OSL), as shown in (4) (Dalbera 1994b: 98–9): (4)

Bonifacino [’fatu] [’kada] [’ata] [’su]

Genoese [’fæ:tu] [’ka:da] [’a:ta] [’su:]

‘done’ ‘warm.fsg’ ‘tall.fsg’ ‘sun’

< factum < cal(i)dam < altam < solem

Moving northwards toward Piedmont, one finds transitional Ligurian dialects without CVL, like Cairese (spoken in Cairo Montenotte, province of Savona), with the same VL in e.g. [’sura] ‘above’ < supram and [’suta] ‘below’ (< subtus, reshaped by analogy to the former), as well as in [’paŋ] ‘bread’ = [’pAn] ‘cloth’ < pannum (Parry 2005: 74, 94, 103). Crossing the border into Piedmont, one finds a mirrorimage distribution of CVL with respect to Liguria. In fact, here CVL is preserved in just a handful of peripheral dialects, while it is missing in most varieties including Turinese (cf. Soffietti 1949: 20, 46–7; Berruto 1974: 14), which has the following stressed vowel phonemes: (5)

i

y e

u ø

ə

o

a

In Turinese, the vowels of, say, [’vol] ‘flight’ (backformed from volare) and [’vuz] ‘voice’ < vocem are phonemically non-distinct from those of [’mol] ‘soft’ < mollem and [’forta] ‘strong’ < fort(e)m, or [’dus] ‘sweet’ < dulcem and [’gula] ‘throat’ < gulam, respectively. Their duration may vary: Soffietti (1949: 45) lists [a/a:], [y/y:] etc. as free variants (e.g. [’spala/’spa:la] ‘shoulder’, [’pjyma/’pjy:ma] ‘feather’), but there is evidence for a complementary distribution of long vs short stressed vowels in a previous stage. In fact, much like in French (}3.4.2), the occurrence of diphthongs in (6a.i) (from PRom /e/), contrasting with the monophthongs in (6a.ii), suggests that those diphthongs developed out of a previously lengthened vowel ([e:]), as occurring to this day in the Standard Italian cognates in (6b.i) which preserve the original syllable structure ’CVCV:

5 Other colonial varieties of Genoese of more recent origin, like Tabarchino (originating in the early 16th c. through migration from Pegli to the island of Tabarka, off the Tunisian coast, and later redeployed to southwestern Sardinia—Carloforte and Calasetta—in the mid-18th c.), have preserved CVL: e.g. [’po:su] ‘stale’ vs [’posu] ‘can.1sg’ (Toso 2005: 38, 204).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change (6)

i. me(n)sem

151

(of)fellam × -ittam

seram

telam

ii. siccam

a. Turinese

[ˈmεis]

[ˈsεira]

[ˈtεila]

[ˈsəkːa]

[ˈfətːa]

b. Italian

[ˈmeːse]

[ˈseːra]

[ˈteːla]

[ˈsekːa]

[ˈfetːa]

‘month’

‘evening’

‘cloth’

‘dry.fsg’

‘slice’

While the lack of the VL contrast in Turinese is representative of the vast majority of Piedmontese, some rural dialects of Piedmont still preserve the contrast, albeit with restrictions in some cases. Thus, while the dialects of Canavese in general do not have contrastive VL (see Zörner 1998: 18, 21 on the varieties spoken in Cuorgnè and Forno), in the Upper Canavese dialect of Trausella (see Vignola Saffirio 1978), all vowel qualities are contrastively long or short under stress only in the final syllable, as in nearby Western Lombard (cf. type (1b)):6 (7)

a. [pu’li:t] [’fy:s] [’mu:t] [’pε:s] [’nO:] [’pa:s]

‘clean’ ‘spindle’ ‘milked’ ‘lost’ ‘knot’ ‘peace’

6¼

b. [pu’lit] [’ys] [’mut] [’pεs] [’nO] [’pas]

‘chick.pl’ ‘door’ ‘dumb’ ‘wilted’ ‘no’ ‘pace’

That this is a residual situation among Piedmontese dialects is shown by the fact that another rural variety of the region, that spoken in Valle d’Andorno, in the province of Biella (Grassi 1968: 158; Berruto 1970: 14–15; 1974: 28), shows the VL contrast, but only for /a/ 6¼ /a:/ and /i/ 6¼ /i:/: (8)

a. [’di:] [’na:s]

‘finger’ ‘nose’

6¼

b. [al ’di] ‘say.3msg’ [al ’nas] ‘be born.3msg’

Conversely, no such limitations on VL contrasts are observed in the dialects of Valsesia, at the northeastern edge of Alpine Piedmontese, whose vowel system is very similar to that of the Western Lombard dialects spoken some 50 km further to the east (see e.g. the examples for Casale Corte Cerro given above in Ch. 3, (52)). Thus, for instance in Rossa (Valsesia, province of Vercelli: cf. Dell’Aquila 2010: 71–6), one finds (sub)minimal pairs with all vowel timbres contrasting by length both in the final syllable (e.g. [’mes] ‘half.m’ vs [’me:s] ‘month’, [’Ot] ‘eight’ vs [’sO:t] ‘solid.m’) and also in absolute final position (e.g. [ca’me] ‘call.inf’ vs [’me:] ‘honey’, [sau̯’ta] ‘jumped.msg’ vs [sau̯’ta:] ‘jumped.fsg’). Likewise, for nearby Campertogno, Molino, and Romano (2008: 31, 36, 58) report minimal pairs such as [’pas] ‘step’ vs [’pa:s] ‘peace’, [’fra] ‘friar’ vs [’fra:] ‘grating’, [’di] ‘day’ vs [’di:] ‘say.inf’, etc. 6

As usual, here too some of the long vowels arose via CoL for the loss of coda consonants ([’mu:t] < *[’munt] < [’munto] < munctum; [’pε:s] < *[’pεrs] < [’pεrso] < *persum for CL perditum), but others clearly can be traced back to OSL ([’pa:s] < pacem, [pu’li:t] < politum).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

152

Dialect variation and comparative reconstruction

5.1.3 The fading of CVL in Alpine and Eastern Lombard Lombardy too hosts dialects which behave differently with respect to CVL. I showed in }3.4.1.4 that Milanese is representative for most of Western Lombard in having contrastive VL only in the final stressed syllable, while displaying no contrast in paroxytones (type (1b) distribution). The remaining subdivisions (Eastern and Alpine Lombard; see Appendix, Maps 4–5) show a more variegated picture, which can be capitalized on for purposes of reconstruction. In Alpine Lombard, some peripheral varieties have lost CVL. Thus, the dialects of the upper and middle Val Leventina (northern Tessin), at the northern periphery of the Lombard dialect area, have lost contrastive VL. One finds there, e.g. in Airolo, the same short vowel in [’set] ‘thirst’, [’kret] ‘believe.3sg/inf’ (originally ’CVCV), as in [’net] ‘clean’, [’met] ‘put.3sg/inf’ (originally ’CVCCV) (cf. Sganzini 1924–6: 100, 103). However, while the stressed vowels in Airolese [’car] ‘wagon’ < carrum and [’cεr] ‘dear’ < carum nowadays have the same duration, as shown by Bosoni’s (1995: 361) measurements, their difference in quality can only be explained by assuming fronting of the lengthened allophone in open syllables (prior to degemination) or of the distinctively long */a:/ (after degemination). Conversely, there is no way to explain this contrast if we assume that OSL never arose at all in this area. Pifferi (2004) provides experimental measurements for another nearby variety of Leventinese, viz. the dialect of Giornico, also showing that contrastive VL is lost: (9)

a. [’pes] ‘weight’ < pe(n)sum [’mes] ‘month’ < me(n)sem

=

b. [’pes] ‘fish’ < piscem [’mes] ‘put.fpl’ < missae

While these come from a dialect from the subdivision labelled ‘lombardo alpino’, further south, in Canton Tessin, another area in which CVL is missing is that of the Corticiasca and Valcolla valleys (near Lugano), whose dialects belong to Western Lombard. Here one finds short stressed vowels throughout, including words which have a contrastively long vowel in the rest of Western Lombard: e.g. [in’trek] ‘entire. m’, [’mez] ‘month’, [’nøv] ‘nine’, [’nas] ‘nose’, [’pas] ‘peace’, [pa’is] ‘village’ (all words stemming from a ’CVCV etymon from the dialect of Signôra, Valcolla: cf. Quaderno fonetico del VSI, no. 63).7 Absence of CVL is documented also for the Alpine Lombard varieties of Valtellina. Thus, in Aprica (middle Valtellina) [’kar] with a short stressed vowel is the outcome of both carrum ‘wagon’ and carum ‘dear’, and the same goes for the nearby dialects of Chiuro and Castionetto (where [’nas] ‘nose’ = ‘be born.3sg’, cf. Della Ferrera 2008: 185), Teglio, Grosio and other parts of Valtellina (cf. Bonfadini 2010: 24, 31).8

Thanks to Mario Vicari and Nicola Arigoni (p.c., March 2012) for pointing out these facts to me. In some of these dialects, a previous VL contrast left its traces in differences in quality: e.g. [’pas] ‘step, pace’ vs [’pæs] ‘peace’ in Grosino (Antonioli and Bracchi 1995: 601–2). 7 8

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

153

A particularly telling case is that of the dialects of Campo and Tartano in Val Tartano, a side-valley of mid-Valtellina. Bianchini and Bracchi (2003: l) describe a VL contrast with the same distribution as in Milanese, as seen in minimal pairs like [kan’ta] ‘sing. inf’ vs [kan’ta:] ‘sung’), or in the following ones, extracted from the dictionary: (10) càar nàas pàas gràam gròos bòots gòot téeč péet

‘dear’ ‘nose’ ‘peace’ ‘worried’ ‘big’ ‘scarabaeus’ ‘enjoy.inf’ ‘stained’ (with soot) ‘a little bit’

6¼

car nas pas gram gròs (a)bòt gòp tèč pès

‘wagon’ ‘be born.inf ‘step’ ‘gram’ ‘10 grams’ (measure) ‘enough’ ‘hunchbacked’ ‘roof ’ ‘fish’

However, in transcribing several other words they include in brackets the second half of the long vowel: e.g. pè(e)l ‘skin’, cö(ö)s ‘cook.inf’, nö(ö)f ‘new’, fö(ö)k ‘fire’, lü(ü)s ‘light’, bà(a)k ‘silkworm cocoon’, ò(o)r ‘gold’, trò(o)t ‘trot’, mu(u)t ‘summer pasture in the mountains’, murù(u)s ‘boy-friend’. This results in the loss of contrast, for example if the words in (11a) are realized with a short stressed vowel, whereby they become homophonous with those in (11b). (In the traditional, Milanese-centred orthography of Lombard dialects, a double consonant indicates shortness of the preceding vowel.) (11)

a. mö(ö)f mö(ö)t

‘move.inf’ ‘manner’

(6¼)

b. möff möt

‘mouldy’ ‘dumb’

The change has been already completed in Upper Valtellina, in the dialects of Livigno and Trepalle, where CVL is gone completely, so that the stressed vowels in the pairs in (12) have the same durations (cf. Bosoni and Mambretti 2011: 133–4): (12) a. [’paʃ] [’pes] [’taʃ]

‘peace’ ‘weight’ ‘be silent.3sg’

=

b. [’pas] [’peʃ] [’tas]

‘step’ ‘fish’ ‘badger’

Bosoni and Mambretti (2011: 133–4) adduce experimental evidence for this lack of contrast, and also remark that previous studies on those dialects reported long vowels, in some cases, which no longer seem to be extant.9

9 Huber (1954–5: 255; 1960–61: 407) transcribes a long vowel in [’ka:r] ‘dig.inf ’, which represents the coalescence of the two homophonous vowels in previous *[ka’(v)ar]. On the other hand, his transcriptions show that the words for ‘dear’ and ‘wagon’, which so often contrast in Northern Italo-Romance varieties, were already, by the mid 20th c., fully homophonous ([’kar] < carrum and carum).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

154

Dialect variation and comparative reconstruction

It has long been observed (cf. Merlo 1951: 1370 and, more recently, Bonfadini 2010) that the dialects of Valtellina display a mix of Western and Eastern Lombard features. And indeed, Eastern Lombard is characterized by the absence of contrastive VL, as is readily apparent from the following examples from urban Bergamasco (Bernini and Sanga 1987: 75; Sanga 1987c: 37, n. 1): (13) a. [’nas] < nasum [’tas] < tacere [’pas] < pacem [’kar] < carum

‘nose’ ‘be silent.inf’ ‘peace’ ‘dear’

b. [’nas] < nascit [’tas] < taxum [’pas] < passum [’kar] < carrum

‘be born.3sg’ ‘badger’ ‘step’ ‘wagon’

The pairs of words listed in (13) do not contrast (any more) in this dialect, which has lost VL altogether. Note, however, that the corresponding words do form minimal pairs in Milanese (Sanga 1984a: 62–3; 1988: 292–3): (14) a. [’na:z] [’pa:z]

< nasum < pacem

‘nose’ ‘peace’

=

b. [’nas] < nascere [’pas] < passum

‘be born.inf’’ ‘step’

For mid vowels, the quantity contrast in Milanese combines with a tenseness contrast, along the lines familiar from cross-linguistic surveys: long vowels are tense, short vowels are lax (cf. e.g. Lehiste 1970: 30ff.): (15) a. [’pe:s/-z] [’me:s/-z]

< pe(n)sum < me(n)sem

‘weight’ ‘month’

6¼

b. [’pεs] [’mεs/-z]

< piscem ‘fish’ < mediu ‘half ’

The vowel system of Bergamasco is best analysed as a further evolution of the Milanese type: i.e. VL must have been there in a previous stage. As this disappeared, the contrast in quality still kept the words distinct if they had a stressed mid vowel ((16)), whereas for low vowels, on the other hand, no difference in quality was there to rescue the distinctions, and merger took place, as seen in (13): (16) a. [’set] [’pes]

< sitim(-em) < pe(n)sum

‘thirst’ ‘weight’

6¼

b. [’sεt] [’pεs]

< septem < peius

‘seven’ ‘worse’

In principle, considering the data in (16) in isolation, one might also avoid the assumption of CVL in a previous stage, since the quality contrast in (16a,b) could be a direct outcome of the identical contrast already occurring in the common Romance vowel system (cf. }2.2.2, (15)). However, internal reconstruction of an earlier phonemic VL contrast for Bergamasco is proved correct by a series of other facts: by further internal evidence from Bergamasco (cf. (18) and (19) below on /i y/-lowering); by comparison with Western Lombard ((14)–(15)); and, finally and most decisively, by philological evidence as well as by comparative evidence from closer Eastern Lombard dialects. In fact, in the 19th century the orthographic conventions adopted in Tiraboschi’s (1873: 34) dictionary of the dialect of Bergamo still kept long vs short

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

155

/a/s distinct, while no diacritic was used for any other vowel, implying that the length contrast in pairs such as (16a,b) had already been abandoned, whereas it was still preserved in pairs like those in (13a,b): Coll’accento circonflesso (^) noto le vocali, che hanno un prolungamento di suono. Nâs, Naso—Pâs, Pace—Tâs, Tacere [‘By a circumflex I mark vowels whose sound is prolonged. Nâs, nose—Pâs, peace—Tâs, to be silent’].

While â is used for long /a:/, no diacritic is used for the short-vowel counterparts nas ‘be born.inf’, pas ‘step, pace’, tas ‘badger’ (Tiraboschi 1873: 840, 930, 1337; cf. Sanga 1987b: 19). Thus, in the 19th century, the VL contrast survived for /a:/ 6¼ /a/ only, which was probably last to be given up because—as Bonfadini (1997: 590) remarks while describing VL in the Western Lombard dialect of Novate Mezzola (cf. }3.4.1.4, (54)–(56))—‘/a:/ ~ /a/ è sicuramente l’opposizione con maggior rendimento funzionale’ [‘/a:/ ~ /a/ is surely the contrast with the highest functional load’] (for all the dialects of the area; cf. another case in point in (8) above, illustrating that /a/ vs /a:/ may resist longer than the contrast involving other vowel qualities). Finally, the VL contrast is still recorded today in the rural Bergamasco dialects spoken in the nearby Val Cavallina, east-northeast of Bergamo (cf. Bonfadini 1987: 333, 375), just for one minimal pair involving /a:/ vs /a/:10 (17) a. [’na:h] ‘nose’ < nasum

6¼

b. [’nah] ‘be born.3sg’ < nascit

Moving further eastwards, the easternmost province of Lombardy, that of Brescia, bordering on Veneto, offers interesting data. Bresciano, spoken on a territory which was for over three and a half centuries (1426–1796) under Venetian rule, is in many respects an ‘anello di congiunzione fra le parlate lombarde e quelle venete’ [‘a missing link between the Lombard and Venetan dialects’] (Bonfadini 1990: 41). Like Bergamasco, to the west, and most dialects of Veneto, to the east, Bresciano lacks CVL. Absence of this contrast is observed in the urban dialect as well as in those of the entire province (Bonfadini 1990: 53), where e.g. [’paŋ] ‘bread’ and [’paŋ] ‘cloth’ are homophonous, whereas south-southwest of Brescia, beyond the Oglio river, in Cremona and its province the two contrast by VL (with /a:/ in [’pa:n] ‘bread’, cf. Ch. 3, (1c)). Much like urban Bergamasco, Bresciano still preserves traces of an earlier VL contrast in the qualities of stressed vowels. All Eastern Lombard dialects, for instance, today have /e/ and /ø/ (18i.a–b) corresponding etymologically to Western Lombard short /i/ and /y/ respectively ((18ii.a–b)), whereas Eastern Lombard /i/ and /y/ (18i.c)

10 For the rest of the lexicon, a short [a] is nowadays generalized where Tiraboschi’s Bergamasco had (and Western Lombard still has) long /a:/s: [’hal] ‘salt’ < sale, [’pal] ‘post’ < palum etc. (Bonfadini 1987: 377). This also holds generally for rural Bergamasco and rural Bresciano (cf. e.g. [’pas] ‘peace’ = ‘pace’ in Val di Scalve: Piffari and Tagliaferri 2011: 295; or in Cigole: Sanga 1979: 407), with the few exceptions to be discussed below in (20)–(26).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

156

Dialect variation and comparative reconstruction

correspond to Western Lombard long /i:/ and /y:/ ((18ii.c)) (cf. Merlo 1960–61: 3; Bonfadini 1990: 47; Sanga 1997b: 257).11 (BS = Bresciano, MI = Milanese.) (18)

filum

mu¯ rum

i. BS

[ˈres]

[ˈbrøt]

[ˈke]

[ˈpjø]

[ˈfil]

[ˈmyr]

ii. MI

[ˈris]

[ˈbryt]

[ˈki]

[ˈpy]

[ˈfiːl]

[ˈmyːr]

‘curl’

‘ugly’

‘here’

‘more’

‘thread’

‘wall’

a. (e)ricium bru¯ t(t)um

b. eccum-hi¯c plu¯s

c.

The environment in which lowering has applied in Eastern Lombard is often described simply as ‘in sillaba chiusa’ (Sanga 1993: 197) or, more analytically, as ‘in sillaba chiusa, davanti a m e in finale assoluta’ (Bonfadini 1993: 105; see also Bonfadini 1990: 47) [‘in checked syllables, before m and word-finally’]. This list of contexts is accurate from the point of view of the Latin etyma of the words affected by the change: in the etymon of Bresciano [’res] and [’brøt] in (18a) the stressed vowel indeed occurred in a PRom closed syllable (cf. the Italian counterparts riccio [’rit:ʃo], brutto [’brut:o]), and in (18b) the change did apply word-finally. In addition, it applied also before (the outcome of ) Lat. -m-, e.g. in Bresciano [’lema] ‘file’, [’føm] ‘smoke’ (vs Milanese [’lima], [’fym]) whereas it did not affect stressed vowels in originally open syllables ((18c.i)). However, this list can hardly be an accurate phonological description of the input to the change: what all these contexts have in common is that they contain a short stressed vowel in Western Lombard. In the by now familiar environment (18a), this is historically due to the non-application of OSL—which e.g. in [’res]/[’ris] ‘curl’ depends originally on the heterosyllabicity of the PRom /k.j/ cluster (cf. Ch. 4, n. 18)—whereas shortness in oxytones ((18b)) and before [m] are specific features of Northern Italo-Romance (cf. }}5.3.4 and 5.4.2).12 Given what we know about the relationship between Western and Eastern Lombard, it is hence fair to assume that the vowels which underwent lowering, in (18a,b) and before [m], were defined underlyingly by their shortness (contrasting with length) in an earlier stage of Eastern Lombard too.13 Note that Eastern Lombard lowering yields a decisive argument to settle the issue concerning the distribution of CVL in Milanese. As we saw in }4.1.1, in many

11 In addition to the quoted linguistic studies, Bresciano forms are drawn from the dictionaries by Ruggeri (1970) and Scaramella (1986). 12 While many Northern Italo-Romance varieties have the VL contrast also in word-final position, it can be argued that at an earlier stage word-final stressed vowels must have been uniformly short all over this area (cf. }5.4.2). 13 The lowering of /i y/ seems to have started from absolute word-final position, subsequently generalizing to all short vowels. This can be argued based on the rustic Bresciano dialects of Val Camonica (cf. Bonfadini 1990: 61, 63), where lowering occurs only in (18b) ([’hø] ‘on, above’ < *sy < su¯(rsum), [’he] ‘yes’ < sı¯c; compare It. su, sì) but not in (18a) ([’bryt] ‘ugly’, [’fit] ‘rental’).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

157

accounts, this has been claimed to have arisen just in words such as those in (18c), because of CoL for the loss of the final vowel, whereas such a VL is claimed never to have arisen in ’CVCV words which did not undergo apocope and, indeed, do not show CVL today. It is easy to see that such a view is at odds with the data on Eastern Lombard lowering. In fact, as shown in (19a) vs (19b), lowering did not apply across the board in all ’CVCV words which did not undergo apocope: (19)

a. lu¯ nam su¯ dat

fi¯la

b. *(cu)cutiam *acu¯ tiat

vi¯neam

i. BS

[ˈlyna]

[ˈsyda]

[ˈfila]

[ˈsøka]

[ˈgøsa]

[ˈeɲa]

ii. MI

[ˈlyna]

[ˈsyda]

[ˈfila]

[ˈsyka]

[ˈgysa]

[ˈviɲa]

‘moon’

‘sweat.3sg’

‘row’

‘pumpkin’

‘sharpen.3sg’

‘vineyard’

Rather, it applied in (19b.i) but not in (19a.i). This implies that, in the previous stage common to Eastern and Western Lombard, the vowels in (19b) were short, while those in (19a) were long. This is predicted under the view defended here, according to which CVL arose as a phonologization of the allophone lengthened by OSL all over this territory, and hence must have been there, originally, in words like [’ly(:)na], [’fi(:)la] ((19a)), in Lombard too, in the same way as is still observable today in words of similar structure in Emilian, Ligurian, etc. or with a different status (lengthened allophone) in Standard Italian. In other words, there has been a change from (1a) to (1b). On the contrary, assuming accounts of VL in Milanese like those of Repetti (1992), Montreuil (1991), or Prieto (2000) reviewed in }4.1.1, according to whom the vowels in (19a.ii) were never lengthened—and, consequently, the systems (1a) vs (1b) are the product of two distinct and independent developments—would preclude an economical statement of the facts about vowel lowering in Eastern Lombard. Eastern Lombard (see Appendix, Map 4) provides still more interesting evidence for our present discussion. Focusing on one peripheral mountain variety of Bresciano, that of Val Rendena, Bonfadini (1993: 111) concludes that more isolated dialects of this province often share features with Western Lombard (Milanese, in particular) which distinguish them from urban Bresciano and the rest of Eastern Lombard: whenever this is the case, it is Brescia (and the rest of Eastern Lombard) which has innovated, whereas rural Bresciano (and Milanese) preserve an earlier stage. Similarly, Sanga (1997b: 253) depicts Eastern Lombard as characterized by innovations spread from Bergamo and Brescia, while a series of peripheral valleys in the east of the region ‘still speak substantially Milanese-type dialects’. Mention of this well-established consensus in the interpretation of the geolinguistic dynamics of eastern Lombardy allows us to place in the right perspective the data from some particularly conservative varieties of Eastern Lombard, to which we now turn. One case in point is the dialect of Sant’Apollonio, a small village in the comune of Lumezzane (in Val Gobbia, province of Brescia). The phonology of this dialect

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

158

Dialect variation and comparative reconstruction

was recently subjected to a thorough phonological analysis by Ferrari (2007–8),14 who was able to show that in this variety VL is (still) contrastive in final stressed syllables:15 (20) a. [’brø:t] [’na:h] [’tʃa:h]

‘broth’ ‘nose’ ‘key’

6¼

b. [’brøt] [’nah] [’tʃah]

‘ugly’ ‘be born.3sg’ ‘noise, hubbub’

To interpret these data, the geographical position of this variety is crucial (see Appendix, Map 4). Val Gobbia is a 12-km-long mountain valley running west to east to the northeast of Brescia. At its western end, towards Brescia, it debouches into the Val Trompia, whereas its eastern extremity communicates with Val Sabbia, whence— going further eastwards—one reaches Lake Garda and, on its eastern shore, the province of Verona (Veneto). Now, in Val Gobbia, among the villages of the comune of Lumezzane, CVL as exemplified in (20) is to be found only in the village of Sant’Apollonio, near the mountain pass (Passo del Cavallo) which leads into the Val Sabbia. Clearly, this is a relic area where one observes the persistence of a feature which must have been shared, in the past, by the dialects spoken in the surrounding region. The interesting fact is that the distribution of VL in this dialect—spoken in an area in which we have independent evidence to conclude that VL is retreating—exactly parallels the distribution found in Milanese (with the whole of Western Lombard) and Central Friulian. Long vs short vowels, in fact, also contrast word-finally:16 (21) a. [’ka:] [’be:] [’we:]

‘dog, dogs’ ‘well’ ‘come.3sg’

6¼

b. [’ka] [te] [’de]

‘house, houses’ ‘you.sg’ ‘day’

14 Sant’Apollonio di Lumezzane is covered by the AIS (as pt 258), whose transcriptions (by Paul Scheuermeier) do register VL (e.g. [’na:h] ‘nose’ I 168) although obviously the atlas provides no study of VL. On the other hand, in the phonetic study by von Ettmayer (1902: 436), one finds [’peh] ‘weight’ and similar transcriptions with a short vowel where, given Ferrari’s description, one would expect a long vowel. Given what we independently know about the distribution of VL in the area, nowadays and in the past, this is probably best explained as due to inaccuracy in Ettmayer’s transcriptions. 15 Minimal pairs are few, also because the distribution of these long vowels is subject to restrictions. Ferrari (2007–8: 70) remarks, for instance, that /u:/ and /y:/ have no short counterpart, since earlier */u/ and */y/ lowered through a change which applied generally in Eastern Lombard (cf. (18) and (19) above). 16 Ferrari (2007–8: 69, n. 29) is reluctant to admit that VL is underlyingly contrastive (see directly below), and therefore proposes various representational adjustments, in order to explain away the VL contrasts in word-final position, at least for some minimal pairs. Thus, for [’ka:] ‘dog’ she assumes an underlying representation /’kaŋ/, mirroring the diachronic source canem, and, more generally, for the /e:/ vs /e/ contrast, she suggests that [e:] could be lengthened by rule from /e/, whereas surface short [e] could be underlyingly /i/. However, she admits herself that ‘in sincronia risulta [ . . . ] difficile individuare sistematicamente i contesti in cui si produrrebbe l’allofono abbassato e’ [‘synchronically, it is difficult to systematically detect the contexts in which the lowered allophone e would be derived’]. A descriptively simpler alternative consists in taking the data in (21) at face value: by Martinet’s (1956: 75; 1975: 205) criterion (cf. }}3.4.1.1 and 3.4.2.2), occurrence of minimal pairs at word boundary implies that VL is contrastive.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

159

On the other hand, no contrast occurs in paroxytones. In this context, one finds a half-long vowel which alternates with contrastively long vowels in the final syllable (22a), and a short vowel (22b) in roots which also appear with a contrastively short vowel in the ultima: (22) a. [’nø:f]/[’nøˑa] ‘new.m/fsg’ 6¼ [kan’ta:t]/[kan’taˑða] ‘sung.m/fsg’ [’wi:h]/[’wiˑwa] ‘alive.m/fsg’

b. [’roh]/[’roha] ‘red.m/fsg’ [’kOt]/[’kOta] ‘cooked.m/fsg’ [’hεh]/[’hεka] ‘dry.m/fsg’

Following Vanelli’s (1998) account of Friulian, Ferrari (2007–8: 71) analyses VL as derived rather than underlying, as shown in the derivation of the minimal pair [’hεt] ‘seven’ vs [’he:t] ‘thirst’: (23) a. Underlying representation b. Final devoicing and vowel lengthening c. Surface form

/’hεt/ – [’hεt]

/’hed/ ’he:d [’he:t]

In the latter word, which historically derives fom Lat. sitim via intervocalic voicing (> *[’se:de]) and apocope (> *[’se:d])—plus the rural Eastern Lombard change s- > [h]—she assumes that the final obstruent is (still) voiced underlyingly, and that it is its devoicing which, synchronically, lengthens the preceding stressed vowel. In sum, VL is claimed not to be phonemic, much as in the analyses discussed in }4.3; rather, the long vowel in [’he:t] is lengthened synchronically. This lengthening Ferrari (2007–8: 57) calls ‘allungamento fonologico’ [‘phonological lengthening’], contrasting it with the more moderate (half-)lengthening (which she terms ‘allungamento fonetico’ [‘phonetic lengthening’]), applying in ’C1VC2V strings where C2 is voiced, as seen in (24) as well as in the ’CVCV alternants in (22a) above: ‘omelette’ ‘silk’ ‘nettle’

(24) [fre’taˑða] [’heˑða] [ur’tiˑga]

< -ata(m) < seta(m) < urtica(m)

This latter lengthening does not apply when C2 is voiceless, as seen in (25) (cf. also the ’CVCV alternants in (22b) above): (25) [’baha] [’waka] [’kOha] [’wεtʃe]

‘low.fsg’ ‘cow’ ‘thigh’ ‘old.fpl’

< bassam < vaccam < coxam < vet(u)lae

A problem for Ferrari’s analysis is that phonetic lengthening is categorically excluded in the words in (26a), although C2 is voiced just as in (24): (26) a. [’mεða] [’frεða] [’gOba]

‘half.fsg’ ‘cold.fsg’ ‘hunchback’

‘half.m’

b. [’mεh] ? ?

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

160

Dialect variation and comparative reconstruction

This is unexpected given (24) and (25), as is the fact that ‘phonological lengthening’ does not apply in [’mεh] ((26b)) either. Rather, in the paradigm [’mεh]/[’mεða] the stressed vowel stays consistently short. This is at odds with the derivation in (23), based on the assumption that VL is not underlying. Note that although Ferrari does not provide the (26b) masculine counterparts to [’frεða] and [’gOba], given the rest of her description, one can predict that the words for ‘cold.m’ and ‘hunchbacked.m’ are liable to surface as *[’frεh] and *[’gOp] respectively.17 While this remains to be checked, the series in (26) is the same as we have already met in Cremonese (cf. }3.4.1.1, (22)), Milanese (cf. }3.4.1.4, (9)) and Friulian (cf. }4.3, (19)). Indeed, in all dialects of northern Italy the counterparts of (26b) have a short vowel, which is at odds with those accounts of VL which assume that this is due to final consonant devoicing and is derived synchronically by rule. As seen in }4.3, (20) above, these accounts have to resort to additional ad hoc hypotheses to accommodate such data. In this vein, Vanelli (1998: 77) and Baroni and Vanelli (2000: 15, 19) assume that in Friulian [’mjεtʃ] ‘half.msg’ lengthening does not apply because of the nature of the following consonant: an affricate, being longer than other voiced obstruents, prevents lengthening. Ferrari too has to adopt an ad hoc hypothesis here, though not the same as for Friulian, because former *[dz] < *[d:z] (from PRom -di-̯ ) changed to [h] in Lumezzanese: gli allungamenti (sia quello ‘medio’ in sillaba interna che quello ‘extra’ di marca fonologica in sillaba finale chiusa da desonorizzata) escludono come target almeno le V /ε, O, o/ [‘the lengthenings (both the “intermediate” one in internal syllables and the “stronger” one, phonological in nature, in final syllables ending with a devoiced consonant) exclude as targets at least the vowels /ε, O, o/]. (Ferrari 2007–8: 73)

The ad hoc nature of this explanation is evident from the fact that it involves no natural class, since high mid [e:] does display lengthening, cf. (23c), contrary to [ε], [O] and [o]. As argued in }4.3, under the alternative view according to which all these vowel contrasts go back to the phonologization of OSL, a unitary (and hence more economical) explanation is available here: no lengthening ever occurred in the Friulian and Lumezzanese outcomes of Latin medium > PRom *[’mεd.(:)ju] (cf. Ch. 4, n. 18) simply because the vowel concerned occurred in a closed stressed syllable. So, the shortness is due to sheer inheritance, an explanation which holds true for all the dialects of northern Italy as argued by Morin (2003: 22–3), and is obviously more economical, as shown while discussing Chapter 4, (19)–(21), than assuming distinct and mutually independent causes in different dialects for what is clearly the same phenomenon. 17 Indeed, AIS chart 5.1038 reports [’frεt] for pt 258 (Mosniga, Lumezzane-Sant’Apollonio). This differs from the variety described by Ferrari in showing final devoicing without weakening of the stop to [h] but has, as predicted, a short stressed vowel. For chart 1.187, ‘gobbo’ [‘hunchbacked’], the atlas unfortunately does not report any answer from pt 258, on which I had no opportunity to do fieldwork myself.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change

161

Also in this detail, like in the overall working of VL, there is an identical phonological common core to Lumezzanese and to Milanese (and, more generally, Western Lombard), on the one side, and to Central Friulian on the other. Assuming that this is just by chance, and that novel phonological rules such as the one in (23b) have arisen polygenetically in all these varieties, due to different causes, giving rise to exactly the same surface distribution of the relevant phenomena, strains credulity. It seems more plausible to consider Lumezzanese, in terms of the geographic distribution of the VL systems (shown in Appendix, Maps 3–4), as the missing link between Friulian and Western Lombard, attesting to a previous stage in which this common innovation must have spanned the whole Northern Italo-Romance area from west to east. Since CVL occurs in Central Ladin too, and is standardly reconstructed as a property originally shared by the whole of that group (cf. }3.4.3.1), there is a territorial continuity, via Ladin, between the CVL systems of Friulian and of Western Lombard, if Eastern Lombard too was involved in the past, as the traces gathered here clearly show.18 Note that the dialect of Sant’Apollonio di Lumezzane can legitimately be called a missing link also in a structural sense. In fact, preservation of half-length in (22a), as distinct from shortness in (22b), seems a natural successor of the stage reconstructed for the whole of Lombardy based on the evidence from Eastern Lombard high vowel lowering in (18)–(19)—with CVL also in [’CV(:)CV] —and a natural precursor of the situation found today in Western Lombard dialects like that of Casale Corte Cerro (see }3.4.1.4, (53)), where all paroxytones have a half-long vowel regardless of etymology (whether [’CVCV] or [’CVC:V]) and voice of the following consonant. The distribution in (22a–b) is closer to the one examined for Friulian in (61), with short [’V] before voiceless intervocalic obstruents (e.g. [’vace] ‘cow’, (61b.i)) vs halflong [’Vˑ] before all voiced intervocalic consonants (e.g. [’caˑze] ‘house’, (61c.i)). Further crucial evidence is provided by the dialect of Malcésine (in the province of Verona), on the eastern shore of Lake Garda. This dialect shows clear traces of what might have been, in the past, a situation comparable with the remnants of Eastern Lombard VL as still observed in Sant’Apollonio di Lumezzane. In fact, though spoken in the Veneto region, Malcesinese shares many features with its Lombard neighbours (first of all, the fronting of PRom /O/ and /u/ to /ø/ and /y/ respectively; cf. Bonfadini 1983: 25, 41–2) and is often defined as a Lombard dialect tout court (e.g. in von Ettmayer 1902: 336; Bertoletti 2005: 37).19 18 From Central Ladin, as seen in }3.4.3.1, CVL shades into Upper Veneto. On Central and Lagoonal Veneto, which are exempt from CVL, see below, }5.1.4. 19 Trimeloni (1995: 6) mirrors the identity perception by the natives, rather than airing a scientific argument, in claiming that Malcesinese is Venetan, not Lombard. Interestingly, he puts Malcesinese’s ‘anomaly’ with respect to the common Veneto linguistic outfit on the bill of the village’s isolation: ‘Nel suo semi-isolamento Malcesine maturò un’evoluzione profondamente originale della sua parlata, piegando il suo “veneto” a toni [ . . . ] che sembrano singolarmente riflettere la durezza [ . . . ] di una vita più di montagna che di lago’ [‘In its semi-isolation, Malcesine underwent an anomalous and deeply original development of its speech, bending its “Venetan” to tones . . . which oddly seem to mirror the harshness . . . of a mountain, rather than lacustrian, life’]. It is well known that several Lombard features used to stretch further east, into Veronese, in the Middle Ages, justifying the classification of Old Veronese as a Lombard

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

162

Dialect variation and comparative reconstruction

Trimeloni’s (1942–3) diachronic sketch of the phonetics of Malcesinese does not address VL explicitly. However, length is notated in the transcriptions, which never indicate a long vowel in words where the stressed vowels occurred in a closed syllable in the PRom antecedent: e.g. [’tet] ‘roof ’ < *tetto < tectum, [’sek] ‘dry’ < *sekko < siccum, [’dʒas] ‘ice’ < *gjattso < glaci(em), [’Ot] ‘eight’ < *Otto < octo.20 As for the outcomes of vowels in PRom stressed open syllables, there is some vacillation, as one finds unexpected shortening in some lexical items (27b), while in a majority of instances a long vowel is notated (27a): (27) a. [fu’rjo:s] ‘furious’, [go’lo:s] ‘gluttonous’ b. [nume’ros] ‘numerous’ < -osum [’fø:k] ‘fire’ < focu, [’zø:k] ‘game’ [’møt] ‘manner’ < modum < iocum [’de:s] ‘ten’ < decem [’mes] ‘month’ < me(n)sem Generally, in words going back to etyma of shape ’CVCV, one finds a long stressed vowel if apocope has occurred ((28a)), whereas if the final syllable is preserved, the stressed vowel is never long ((28b)):21 (28)

a. [’tʃa:f] ‘key’ < clavem, [’na:f] ‘ship’ < navem, [’no:s] ‘walnut’ < nucem, [’ø:f] ‘egg’ < ovum b. [a’mizi] ‘friend.pl’ < amici, [’fava] ‘fava bean’ < fabam, [’f(r)eve] ‘feber’ < febrem; [’gosa] ‘drop’ < *gutteam, [’gula] ‘throat’ < gulam, [mi’Ola] ‘marrow’ < medulla

This results, apparently, in some near-minimal pairs: (29) a. [’de:s] [’na:s] [’no:s]

‘ten’ ‘nose ‘walnut’

< decem < nasum < nucem

b. [’mεs] [’las] [’tos]

‘half ’ ‘lace’ ‘cough’

< medium < laq(u)eum < tussim

Long vowels occur also word-finally in Malcesinese: e.g. [fa’la:] ‘fail, err.inf’ and all oxytonic verb infinitives (cf. the lists of forms with long vowels, extracted from Trimeloni 1995, in Marcone 2011–12: 46–9). This is to be compared with the occurrence of long stressed vowels in infinitives in some peripheral Western Lombard dialects, as shown in (106b), }5.4.2. The similarity with Lumezzanese is striking, as similar restrictions on VL obtain in the two dialects. For instance, for Malcesinese, Trimeloni (1942–3) never transcribes a long vowel before a final /l/, whatever the original syllable structure:

dialect in Pellegrini (1966: 70), Sanga (1993: 197, n. 56). As Bonfadini (1983: 25, 41–2) shows, several isoglosses contrasting Bresciano and Veronese nowadays run just east of Malcesine (e.g. apocope and, as already mentioned, the occurrence of the fronted outcomes /ø/ and /y/ of PRom /O/ and /u/) respectively). The same applies to the dictionary of Malcesinese by the same author (Trimeloni 1995). This is confirmed systematically by the entries in Trimeloni (1995): e.g. cana ‘reed’ < cannam = luna ‘moon’ < lunam (pp. 54, 132). 20 21

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

At the vanguard of change (30)

a. ’CVCV [’pel] ‘hair’ [’tal] ‘such’ [’fjøl] ‘son’

< pilum < talem < filiolum

b. ’CVC:V [kor’tεl] ‘knife’ [ka’val] ‘horse’ [’mOl] ‘soft’

163

< *cult(r)ellum < caballum < mollem

This corresponds perfectly to the restriction described for Lumezzanese by Ferrari (2007–8: 75), where a lateral induces categorical shortening of the preceding stressed vowel: (31)

a. ’CVCV [’pel] ‘hair’ [’hal] ‘salt’ [’kyl] ‘arse’ [’fjøl] ‘son’

< pilum < salem < culum < filiolum

b. ’CVC:V [’pεl] ‘skin’ [’gal] ‘cock’ [’kOl] ‘neck’ [’mOl] ‘soft’

< pellem < gallum < collum < mollem

Thus this dialect, spoken on the western fringe of the Veneto, seems to warrant the reconstruction of a previous stage in which at least rural Veronese joined Eastern Lombard in displaying CVL: this was later submerged in the whole of Eastern Lombard, leaving some scanty traces to this day in Val Cavallina (cf. (17)), in the province of Bergamo, and in Val Gobbia, in the province of Brescia. 5.1.4 The position of Venetan Note that, considering the urban dialects of Brescia and Bergamo in their present state, one might in principle have been misled to speculate that all these NItRom varieties failed to develop contrastive VL, or that they possibly never even had an OSL rule. The evidence reviewed in (16)–(22), however, shows without any doubt that this would have been the wrong guess: even if we did not have direct proof from the marginal dialects in (17), contrasts in vowel qualities such as those in (18i.a, b) vs (18i.c) guarantee that a phonemic VL contrast has been there in the past. For Veneto proper, however, excluding the rural Veronese variety just mentioned, any such evidence from vowel qualities is lacking. As for VL, it is non-distinctive everywhere in Veneto (cf. Zamboni 1988: 527 and, more specifically, e.g. Wanner 1971: 72 on Venetian; Trumper 1972: 9 on Padovano and Polesano),22 and has been so throughout the documented history of these dialects (cf. e.g. Stussi 1965 on Old Venetian; Bertoletti 2005 on Old Veronese). This, in principle, may indicate that VL in the dialects of Veneto developed, from the outset, in an entirely different way with respect to Friulian, to the east, Lombard, to the west, and Emilian, to the south, without ever becoming contrastive. However, the discussion just developed for

22

In the territory of Veneto, the only dialects with CVL are on the northernmost fringe (Pellegrini’s 1977a ‘ladino-veneto’), as seen in }3.4.3.1, (80). Otherwise, even the rest of upper Veneto lacks it (see e.g. Migliorini and Pellegrini 1971: xvii on Feltrino; Màfera 1958: 143–9 on Trevigiano-Bellunese).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

164

Dialect variation and comparative reconstruction

Eastern Lombard, and the fact that some isoglosses characterizing the Lombard/Veneto border have moved further west during the documented history of these dialects, suggest that one would be well advised to follow Uguzzoni’s (1975: 59) cautious conclusion on this point: Non è dato sapere se questa correlazione fonematica di quantità vocalica fu un tempo presente in tutti i dialetti dell’Italia settentrionale e poi abbandonata dalla maggior parte o se essa fu impedita in molti dialetti da fenomeni di livellamento quantitativo che si verificarono nelle vocali toniche contemporaneamente alla degeminazione delle consonanti postoniche [‘It is not for us to know whether this phonemic VL correlation was once present in all the dialects of northern Italy and was then abandoned by most of them, or whether it was prevented in many dialects by phenomena of quantitative levelling which took place in stressed vowels at the same time as the degemination of post-tonic geminate consonants’].

More precisely, this caution is appropriate for Veneto, since for all other areas nowadays lacking CVL (French, many Rhaeto-Romance dialects, most of Piedmontese, Eastern Lombard, and several other peripheral varieties of Lombard, Ligurian, and Friulian) there is conclusive evidence that the VL contrast did arise and was then lost, as seen above in }}5.1.1–3.

5.2 Apocope and the rise of contrastive VL in Northern Romance We have seen in }4.1 that many explanations of the rise of CVL in NItRom establish a causal link, either direct (Repetti 1992; Prieto 1992) or indirect (Francescato 1966; Vanelli 1979 etc.), between VL and the loss of final unstressed non-low vowels, which co-occurs—or used to co-occur—with CVL in many NRom dialects. A non-exhaustive list of proponents of this kind of causal link is provided in (32a), whereas (32b) lists some proponents of the alternative view endorsed here, according to which it was degemination, rather than apocope, that caused VL to become contrastive in Northern Romance:23 (32)

Origin of CVL in Northern Romance a. Caused by (or at least applying only after) apocope: Baroni and Vanelli (2000); Bonfadini (1997: 599f.); Contini (1935b: 59); Francescato (1966); Hualde (1990); Montreuil (1991); Pellegrini (1982: 17); Prieto (1994; 2000); Repetti (1992: 174f.); Vanelli (1979); Videsott (2001)

23 The two lists resemble those provided in }4.1, (3), since appeal to apocope prevents a unitary account of the different vowel systems to be found across Northern Romance, whereas the assumption that VL arose through degemination affords a unitary account. Criterial for inclusion in (32a) is the claim that the rise of CVL follows apocope in at least some NRom varieties (e.g. as seen in }4.1, Repetti 1992 makes the claim for Milanese and Friulian, but not for Emilian).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Apocope and the rise of contrastive VL

165

b. Due to degemination: Haudricourt and Juilland (1949: 35); Lüdtke (1956: 135–6, 257–62); Weinrich (1958: 183–5); Uguzzoni (1975: 58); Morin (2003); Loporcaro (2005a; 2007a; 2007b); Filipponio (2012: 41–57) In the present section, I will show that the view in (32a) is untenable, for both empirical and theoretical reasons. On the latter front, this explanation of the rise of CVL is less economical than the one defended here because it accounts for only some of the systems displaying CVL (say, for Friulian or Milanese) but has nothing to say about Cremonese or Emilian, which show VL also in contexts where apocope did not apply, nor for the dialects of Liguria or of the upper Emilian Apennine, which developed CVL without having been affected by apocope at all. Obviously, the claim that CVL needs apocope precludes a unitary account of the rise (and fall) of CVL. Even more seriously, these hypotheses leave a sizeable proportion of the data unexplained, even for the very same varieties for which they are put forward: thus, Repetti’s (1992) or Montreuil’s (1991) account of Milanese is meant to explain the VL contrast in, say, [’na:z] ‘nose’ vs [’nas] ‘be born.inf’ but cannot cope with the one in [kan’ta:] ‘sung’ vs [kan’ta] ‘sing.inf’. On the empirical side, on the other hand, it is easy to demonstrate that this explanation is factually wrong, for the dialects whose written documentation has sufficient time depth. Milanese, again, is a case in point, as I will show considering Old Milanese in }5.2.2, and the same goes for Old Ligurian, to be addressed in }5.2.1. 5.2.1 On the non-co-occurrence of CVL and apocope There is hardly any doubt that the rise of CVL cannot be traced back either to apocope or to final devoicing ensuing from apocope in a deterministic way: counterexamples are legion, even in the Romance languages which did develop CVL via other changes. Consider e.g. French [’nœf] ‘new’, whose alternation with feminine [’nœ:v] could lend itself to an analysis parallel to those discussed for Friulian in }4.1.2, if [œ] were lengthened: but in fact it is not.24 And indeed, among the many causes for lengthening in the history of French, inventoried in Morin’s (2006: 148–9) list reported in }3.4.2.1, (67), final devoicing does not feature, though apocope does (cf. (67v)). However, Spanish (e.g. finid [fi’niθ] ‘finish.imp.2pl’ < finide < finite), Catalan (cap [’kap] ‘head’ < *kabo < capu(t)) etc. show that there is no causal necessity for apocope to result in vowel lengthening.

24 Of course, the diphthong in OFr. nuef, nueve was itself the diachronic successor of a Proto-French phonemically long vowel, to be traced back in turn to an allophone lengthened by OSL (cf. }3.4.2.1, n. 68). The present argument, however, is panchronic: it aims to show that devoicing and lengthening need not necessarily co-occur, at any given stage. In modern French, by the way, it is the vowel preceding voiced final [v], rather than voiceless [f], which is lengthened.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

166

Dialect variation and comparative reconstruction

Even within Northern Italo-Romance, where apocope has been invoked by many (cf. }}4.1.1 and 4.1.2), the correlation between apocope and the rise of CVL is far from general. This can be seen at first glance by comparing the four main urban dialects of Northern Italy: (33) degemination CVL apocope

a. Turin + – +

b. Milan + + +

c. Genoa + + –

d. Venice + – –

Soffietti (1949: 20)

Sanga (1988)

Forner (1975: 50)

Lepschy (1962)

All of these, of course, share consonant degemination, which caused the rise of CVL according to (32b).25 As has been said above (}5.1.4), the place of Venetan in this connection can be debated, whereas concerning Piedmontese, as shown in }5.1.2, there is no doubt whatsoever that the whole of it must have possessed CVL at an earlier stage. Taking now the two varieties displaying CVL to this day ((33b,c)), the existence of Genoese—and Ligurian, more generally—which has CVL without having lost final unstressed vowels (cf. }3.4.1.3 above) suffices to demonstrate that apocope cannot be held to be a necessary precondition for the rise of CVL. Note that Ligurian has shown the constellation in (33c) all along its documented history. This can be argued for Old Genoese, since the corpus also includes verse, whose rhymes provide indications on VL. Indeed, in the first systematic linguistic study of Old Genoese, Parodi (1898: 100–102) argued that in 13th-century Genoese poetry vowels rhyme with each other, only provided that VL (as assessed independently on the basis of modern Genoese, though in individual cases changes have occurred in the meantime) has the same value: ‘la vocal breve del dialetto non rima che con vocale breve; vocale lunga non rima che con vocale lunga’ (Parodi 1898: 100).26 Thus, one finds, for example, the two distinct series in (34a) vs (34b): (34)

a. /’e:/: De ‘God’, pe ‘foot.sg/pl’, re ‘guilty’, ve ‘come.prs.2sg’ b. /’e/: de ‘must.3sg’/‘give.pret.3sg’, fe ‘faith’/‘do.pret.3sg’, re ‘king’, ve ‘see.3sg’

From the fact that the vowels in (34a) rhyme with each other, but not with those in (34b), and Parodi infers that the former series represents a phoneme /e:/, the latter /e/, and this strict separation holds for all vowel qualities, not only wordfinally but word-internally as well: ‘[a]nche le vocali interne sono soggette alla regola’

25 To the best of my knowledge, there is only one exception to this generalization: the dialect of the Emilian enclave of Colognora in Valleriana, in northwestern Tuscany, mentioned below in }5.3.5. 26 Parodi’s remarks refer to the verses by the Anonimo Genovese (later identified as Luchet(t)o— Nicolas 1983: xvi; 1994: cxviii), who wrote civil and religious verses between the late 13th and the early 14th centuries. On CVL in Old Genoese see more recently Formentin (2002: 102).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Apocope and the rise of contrastive VL

167

(Parodi 1898: 101). One finds Margarita ‘Margaret’ : vita ‘life’ : saita ‘thunderbolt’ (with /i/ in modern Genoese), for example, rhyming with each other, and distinct from e.g. descumfita ‘defeat’ : scrita ‘written.fsg’ (cf. Nicolas 1994: 161; both have /i:/ in modern Genoese). Word-internally, VL is never reflected in the orthography, whereas this occasionally happens in word-final position, where e.g. the fsg past participle ending, written as -aa or -a, rhymes with the phonologically and etymologically identical suffix of adjectives and nouns such as the two in (35a), but never with the final stressed vowel in the 3sg ending of the preterite in (35b), which must have been realized as a short [’a] (cf. Nicolas 1994: cxxxviii–cxxxix):27 (35)

a. naa ‘born.fsg’ : santificà ‘sanctified.fsg’ (3, 5–8); devorgaa ‘famous.fsg’ : renomaa ‘renown’ (12, 658–9) b. resuscità ‘be resurrected.pret.3sg’ : montà ‘rise.pret.3sg’ (12, 222–3)

Note that (34) and (35) attest to the occurrence of CVL word-finally, an environment to which the contrast must have spread after arising in word-internal position. In addition, there is reason to believe that the contrast already occurred also before stress, since the processes feeding VL in this position (vowel coalescence and coda /l/deletion, cf. }3.4.1.3, (39)) had already applied: e.g. cazinha ‘lime’ < calcinam, fazon ‘skythe’ < *falce-onem (cf. Cocito 1970: 34). Thus there is evidence that VL has been contrastive ever since the earliest extant Genoese texts, while apocope never applied in this variety. As for degemination, on the other hand, whose diagnosis does not depend on rhyme, one can go further back, to the earliest vulgar documents from Liguria, dating back to the mid- to late 12th century, such as the fully vulgar Dichiarazione di Paxia (Savona, ad 1178–82; see A. Castellani 1976: 171–87) or the Latin–Genoese inventory annexed to Raimondo Piccenado’s last will and testament (Genoa, ad 1156; see Parodi 1898: 18–19). In these texts, degemination of intervocalic obstruents ((36)) is reflected regularly, whereas geminate sonorants ((37)) are still largely preserved graphically at this time, a circumstance which must reflect actual phonetic realization according to A. Castellani (1976: 177): (36) a. Dichiarazione di Paxia: enapo 9 ‘cup’ < Germ. *hnapp (cf. It. nappo), capa ‘cloak’ < cappam, bruneta 4–5 ‘brown fine cloth’ < Germ. *brun- + -ittam, rota 9 ‘broken.fsg’ < ruptam, sachon 5 ‘mattress’ (It. saccone), çoculi 6 ‘clog.pl’ (cf. It. zoccoli), gabie 12 ‘fish trap.pl’ (literally ‘cage.pl’) < *gabbie < caveas b. Raimondo Piccenado’s inventory: enapo 25 ‘cup’, bota 23/botam 22 ‘barrel’, pl. botas 2

27

The references are to poem number and verse in Nicolas’ (1994) edition.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

168

Dialect variation and comparative reconstruction

(37) a. Dichiarazione di Paxia: sepellir 2 ‘bury.inf’, agnello 4 ‘lamb’, pelle 4 ‘pelt/ leather’, barril 6, 11 ‘barrel’, canne 10 ‘reeds’, gonnelle 4 ‘type of dress.pl’, scudelle 8 ‘bowl.pl’, galleda 9 ‘jar’ < galletam b. Raimondo Piccenado’s inventory: pelle 28 ‘pelt/leather’, mantellum 8 ‘mantle’, ferro 10 ‘iron’, pannos 7 ‘cloth.pl’ (alongside anelos 7 ‘ring.pl’ with -l- < -ll-) There is abundant evidence demonstrating that in all Western Romance languages and dialects, degemination affected sonorants centuries later than obstruents (cf. }5.3.4): thus, one possible scenario is that CVL was first established as obstruent gemination was lost completely. Another fact that must be kept in mind (and will be elaborated on in }5.3.3) is that for the phonologization of CVL only the phonemic gemination contrast is required, not the complete elimination of all phonetic geminates at the surface. The examples from living dialects to be discussed below will show that the persistence of sonorant geminates in (36) is not incompatible with CVL, which is evidenced, a century and a half later, by Old Genoese rhymes. Whilst the link between CVL and (phonemic) degemination is thereby confirmed, the mutual independence of CVL and apocope is further demonstrated even by the Northern Italo-Romance dialects for which apocope has been crucially invoked as the trigger for (alleged) vowel lengthening, as I shall now move on to show in }5.2.2. 5.2.2 The gradual spread of apocope in Northern Italo-Romance As already pointed out in }3.3, Northern Romance apocope took place first in Northern Gallo-Romance around 700 ad, to then gradually spread southwards. For Romansh, evidence for such deletion is available from the 12th century onwards, under the form of vacillating notations for placenames like Airfurnu < agro furno, written Agisfurn in a chart dating back to 1156 (von Ettmayer 1919: 10). In Northern Italy, it is clear that apocope applied even later, at least in some well documented and well investigated dialects. While in the oldest extant Eastern Lombard texts (of Bergamasco and Bresciano, which date back to the 14th century) non-low final unstressed vowels are fully dropped (Contini 1935a: 143; 1935b: 50; Formentin 2002: 105), in 13th-century Milanese, final nonlow vowels were dropped by synchronic rule only in non-prepausal position, while being retained prepausally. This was demonstrated by Contini’s (1935b: 46–7) seminal study through the inspection of the metric and rhymes of the late 13th-century verses by Bonvesin da la Riva (c.1250–1314/15). Consider the two alexandrines in (38) (from Bonvesin’s Disputatio rose cum viola, 9–10; Contini 1941: 77; 1960: 1.620): (38) Tutọ zo che la vïora devrave inanzẹ parlar, Perk’ella nass inanze ki vol raxon cercar ‘Although the violet should speak first, Because it is born earlier (if one wants to look for rhyme and reason) . . . ’

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Apocope and the rise of contrastive VL

169

The metric ensures that inanz in line 9, though spelled inanze in the manuscript due to a copying mistake, causing hypermetry (which the editor signals through expunction), scans as dysyllabic [i’nants], whereas the same word must be read [i’nantse] in the following line, where it occurs at the end of the first hemistich.28 Contini (1935b: 47) shows that in this text line-final unstressed vowels are neither deleted nor centralized, on the evidence that endings like ‘-ento (-empo), -ente (-ende), -enti (-empi) [ . . . ] rimano sempre e solo ognuna con se stessa’ [‘ . . . each rhyme always and exclusively with the identical string’]. Had merger or deletion already taken place, one would expect graphically different strings like, say, -ento and -ente to rhyme freely with each other, the same way as differently spelt eight/fate/great/straight/trait do in modern English. Data such as those in (38) justify the assumption of an apocope rule for Old Milanese which deletes final unstressed vowels synchronically, except if occurring in prepausal position (cf. Loporcaro 2005–6: 84): (39)

[+syll, –low] ! / _ ]pw

X]pu

[where X 6¼ ]

Consider now the implications of (39) for accounts of Milanese VL which appeal to apocope, like Repetti’s (1992) (cf. }4.1.1 above). Since apocope (e.g. locum > logo > [’lø:g] ‘place’) is claimed to be the cause of the length of the stressed vowel, this account predicts that at a stage in which, say, logo could not have its final vowel deleted, it might not have had a long stressed vowel either, as CVL may not have occurred before apocope. This prediction is disconfirmed by the Old Milanese data, not considered in Repetti (1992)—whose account reduces to internal reconstruction, along the lines discussed in }4.2—nor by the other generative phonology accounts of Milanese VL reviewed in }4.1.1. In fact, another seminal study on Old Milanese based on Bonvesin’s rhymes, Salvioni (1911: 176–7), showed that VL was already contrastive in the late 13th century. This can be argued on the basis of rhyme series like those in (40): (40)

28

a. utilitae ‘usefulness’ : caritae ‘charity’ : amistae ‘friendship’ : celestial ‘celestial’ (Vulgare de elymosinis 28–4), veritae ‘truth’ : cugial ‘spoon’ (De quinquaginta curialitatibus ad mensam 65–6) b. descenderà ‘come down.fut.3sg’ : montar ‘climb.inf’ : guardar ‘watch.inf’ : remirar ‘gaze.inf’ (De quindecim miraculis 9–12); va ‘go.3sg’ : retornar ‘return.inf’ : dexdegniar ‘disdain.inf’ : far ‘do.inf’ (Laudes de Virgine Maria 117–20)

The alexandrine as used in this text consists of two hemistichs, each one a settenario, i.e. a verse defined by the fact that the last beat falls on the sixth syllable, and comprising seven syllables in all if the last word is paroxytonic, or six or eight if the last word is oxytonic or proparoxytonic respectively.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

170

Dialect variation and comparative reconstruction

The endings in (40a,b), though still spelt in a variety of ways due to their respective etyma—e.g. veritae < veritatem, celestial < cælestialem in (40a), -ar < -are, -à < -(h)at in (40b)—must have been homophonous to rhyme together. Crucially, items from the two series never intermingle within the same rhyme, and the fact that they correspond respectively to word-final /’a:/ ((40a)) vs /’a/ ((40b)) in modern Milanese (cf. }5.4.2 below) leads Salvioni to the incontrovertible conclusion that the contrast must have been there already in the 13th century. Comparison among Northern Italo-Romance varieties—I shall argue in }5.4.2— shows that CVL in absolute word-final position must have arisen later than wordinternally, the context in which it was first triggered (in the reconstruction upheld here) by degemination. Old Milanese data are compatible with this assumption, as word-internal stressed vowels may rhyme together only if they have exactly the same quantity and quality—judging, again, by modern Milanese outcomes (Salvioni 1911: 158–63). (41)

a. /’ε/: pegio ‘breast’ : legio ‘bed’ : despegio ‘annoyance’ : maladegio ‘cursed’ (Libro delle Tre Scritture 121–4) b. /’e:/: greve ‘heavy’ : vorreve ‘want.cond.1sg’ : porreve ‘can.cond.1sg’ : peccareve ‘sin.cond.1sg’ (De peccatore cum Virgine 105–8) c. /’ø:/: provo ‘close’ (< prope) : logo ‘place’ (De quinquaginta curialitatibus ad mensam 117–18) d. /’O:/: poco ‘little’ : corpo ‘body (De quinquaginta curialitatibus ad mensam 31–2)

Summing up, if, say, logo in (41c) already was, phonemically, /’lø:go/ at a stage at which its final vowel could not yet be deleted prepausally, this is further proof that apocope did not cause any compensatory lengthening of the stressed vowel in this variety. The same synchronic apocope rule (39) reconstructed by Contini for Old Milanese has been shown to be at work in Old Veronese in Bertoletti’s (2005: 116, 134) analysis of documentary texts from the 13th–14th centuries. One can be all the more confident that the reconstruction is accurate since there are several modern dialects spoken at the southern periphery of the Northern Italo-Romance area (more precisely, on the northern slope of the Apennines, from Lunigiana, to the west, up to the province of Bologna, to the east) which display a similar situation today (cf. Loporcaro 2005–6: 89–94 and Appendix, Map 6). This is illustrated in (42) and (43) with the apocope of word-final /e/ and /o/ in the two dialects of Sassalbo (province of Parma, Lunigiana; cf. Giannarelli 1913: 285–6) and of Lizzano in Belvedere (province of Bologna; cf. Malagoli 1930: 151; field data from Lizzano’s outlying part of Pianaccio):29 See Loporcaro (2005–6: 88–9) for a review of the modern dialects displaying a similar contextsensitive vowel deletion rule. The southernmost dialect of this type is that of Sillano (in the province of Lucca; Giannelli 1984: 13–14), for which Pieri’s (1893b: 347–54) phonetic transcriptions omit schwas only 29

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Apocope and the rise of contrastive VL (42) a. __ ## [ma’ɲar dl̩ ’kaʒo/**’kaʒ] ‘to eat cheese’ [un ra’gats ’bravo/**’brav] ‘a good boy’ (43) a. __ ## [e ’skris:e/**’skris] [e j ’a ’skrit:o/**’skrit]

171

b. __ ] pw [ . . . ]pw . . . [’kaʒ ’boŋ] ‘good cheese’ [um ’brav ra’gatso] ‘a good boy’

b. __ ] pw [ . . . ]pw . . . [e ’skris na ’let:ra] ‘he wrote (a letter)’ [e j ’a ’skrit na ’let:ra] ‘he has written him (a letter)’

Like Old Milanese, the dialect of Lizzano has CVL, as shown by subminimal pairs like the following (cf. Malagoli 1930: 139–41 and the acoustic measurements in Loporcaro et al. 2006: 512–14, reported in (73)): (44)

a. [kaŋ’ta:] [aŋ’da:] [’se:] [’so:] [so’ti:] [’du:]

‘sing.2pl’ ‘go.2pl’ ‘six’ ‘his/her.mpl’ ‘thin.mpl’ ‘two’

6¼

b. [kaŋ’ta] [aŋ’da] [a’ʒe] [’dʒo] [ak’si] [’nu]

‘sung’ ‘gone’ ‘vinegar’ ‘down’ ‘so’ ‘we’

A number of other dialects from the Emilian Apennine confirm that CVL does not need apocope. Thus, in the dialect of Piandelagotti, as shown in (45), CVL occurs although final unstressed non-low vowels have been merged to /@/ without being deleted completely:30 (45)

a. [’ka:r@] [al ’na:z@] [’di:]

‘dear’ ‘the nose’ ‘finger’

6¼

b. [’kar:@] [al ’nas@] [’di]

‘wagon’ ‘be born.3msg’ ‘day’

The dialects of Lizzano and Piandelagotti will be addressed again in }}5.3.2–3, where I shall discuss the relationship between the rise of CVL and degemination, since, like several other dialects of the Emilian Apennine, they deliver crucial evidence for reconstruction, as pointed out as early as Weinrich (1958: 166–70). Summarizing, a considerable bulk of evidence has been lined up, in this and the previous section, showing that recurrent claims that the rise of CVL in Northern ItaloRomance depends upon previous apocope, either directly (via CoL for the loss of the final vowel) or indirectly (via compensation for the devoicing of the following consonant) are incompatible with a comprehensive account of (the rise of ) CVL. In particular, such claims have been laid for Milanese (Repetti 1992) based on internal reconstruction. Applying the classical method synthesized in }4.2, (17), I have shown in the

clause-internally: e.g. [e d: a’l:ora i’n:ants la dʒus’titsja int a ’reɲ@] ‘and from then on, justice in his kingdom . . . ’ (with [i’n:ants] [i’n:ants@], much as in Bonvesin’s first line in (38)). 30 Though they can be deleted variably, in connected speech and even in isolation (see the experimental study by Loporcaro et al. 2007: 71–4), they are still retained in the underlying representation.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

172

Dialect variation and comparative reconstruction

present section that both comparative reconstructive evidence—from peripheral dialects like those in (43)–(45)—and philological evidence from Old Milanese ((38)–(41)) militate against this kind of internal reconstruction.31

5.3 The rearguard of change: at the source of VL in Northern Romance In }5.1 we saw that several dialects of northern Italy (Eastern Lombard, Piedmontese, Eastern Friulian) have lost CVL while there is evidence that they were like Western Lombard or Central Friulian at an earlier stage. Bringing these dialects into the picture, then, turns the binary contrast between the Milanese and Cremonese types in (1a,b) into a three-step scale:

a. Cremonese b. Milanese

(46)

c. Bergamasco

i.

σ´

]pw

+

+

–

ii.

σ´σ

]pw

+

–

–

[+ = contrastive VL in the given environment] As we saw, there is cogent evidence that the above-mentioned dialects now at stage (46c) shifted along this scale, and similar evidence is available for the preceding step too. In fact, while analysing lowering in Eastern Lombard in (18) and (19), I have argued that this provides evidence for a phonological representation */’ly:na/ ‘moon’, */’fi:la/ ‘row’ in Common Lombard, the joint predecessor of both Eastern and Western Lombard dialects. In other words, comparison with a dialect at the vanguard of change (Bergamasco, (46c)) demonstrates that what are today stressed vowels not specified contrastively for length in Milanese /’lyna/ and /’fila/ must have been contrastively long vowels earlier: this boils down to saying that Milanese was once like Cremonese, Emilian, Ligurian, etc. ((46a)). Just as for the retreat of CVL at the vanguard of the change ((46b) > (46c)), for its rearguard too one can point to supporting evidence from the geographical distribution of the dialects at issue across the Northern Italo-Romance territory. The dialects of Valtellina have been seen, in }5.1.3 (10)–(12), to offer examples of (46b) > (46c). However, Valtellina also hosts dialects in which CVL has an ‘Emilian’ distribution, occurring also in paroxytones ((46ii)). This is the case in the dialect of Val d’Ambria—a side-valley of Valtellina south of Sondrio (see Appendix, Map 4)—for which Bosoni (2003: 212, n. 31) reports the following data: (47)

31

CVCV a. [’mO:ra] b. [’rø:da] [bu’fa:va]

‘blackberry’ ‘wheel’ ‘blow.impf.3sg’

CVC:V [’mOra] [’rota] [’gata]

‘morra’ (a game) ‘broken.fsg’ ‘she-cat’

This converges with comparative evidence from dialects nowadays lacking CVL (see the discussion on the Eastern Lombard data in }5.1.3, (19)).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

173

While for the examples in (47b) one could invoke the voice(lessness) of the following consonant as a discriminating factor, (47a) is a bona fide case of a pure VL contrast. Furthermore, all the examples on the left-hand side transparently go back to Proto-Romance open stressed syllables, while all those on the right go back to a checked syllable (cf. e.g. the standard Italian counterparts of (47a), mora vs morra). Of course, it would hardly be sensible to claim that Valdambrino is ‘Emilian’ because of this distribution of VL: rather, it is a variety of Valtellinese which lags behind those in (10)–(12) (}5.1.3) along the same path of change as modelled by the scale in (46). This is not an isolated case among Alpine Lombard dialects, as shown by the dialect of Val d’Antrona (province of Verbania; see Appendix, Map 4), where CVL occurs in paroxytones too: [’pa:la] ‘shovel’ vs [’spala] ‘shoulder’ (see Nicolet 1929: 161, 185). Other Alpine Lombard dialects provide further clues to refine our reconstruction. If, as argued here, CVL arose in general through degemination, a potentially interesting source of evidence is provided by those Northern Romance dialects which preserve, to some extent, geminate consonants. Indeed, in several descriptions of NRom varieties one occasionally comes acrosss remarks such as the following, concerning the upper Engadinian dialect of Celerina/ Schlarigna (Walberg 1907: 67): Il nostro parlare, come in genere i dialetti ladini e quelli dell’Alta Italia, non possiede consonanti lunghe alla toscana. [ . . . ] Le doppie originarie latine sono ridotte a semplici, cioè brevi—seanche [sic] meno brevi delle semplici primitive—davanti alle quali si pronunzia breve la vocale [‘Our dialect, like Ladin and Northern Italian dialects in general, has no long consonants à la Tuscan. [ . . . ] Latin original geminates were reduced to singleton, i.e. short— even though not so short as the original short ones (emphasis added—M.L.)—in front of which the vowel is pronounced short’].

The observation that degemination is not complete is backgrounded in a concessive clause, which fits in well with the fact that Walberg’s transcriptions do not record any difference in length (not even as half length) with respect to original singleton consonants: e.g. [’tɕapa] ‘cloak’, [’kOpa] ‘goblet’, [’tεra] ‘earth’, [’an] ‘year’. This transcription habit fostered the widespread impression that degemination applied without any residue all over the WRom. territory, an impression further backed up by several descriptions which, contrary to Walberg’s, explicitly exclude that long consonants may have persisted anywhere in Northern (Italo-) Romance: see e.g. Schädel (1903: 22 n. 1), who claims that whenever double consonants occur in the notation of Northern Italian dialects, this kind of spelling ‘bedeutet nur Kürze des vorhergehenden Vokals’ [‘only indicates shortness of the preceding vowel’, emphasis original]. This is indeed very often, but not always, the case, and there is now a substantial bulk of evidence for partial preservation of consonant length on the fringes of Northern Romance, concerning several different dialect areas. Suffice it to mention the few references in (48):

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

174 (48)

Dialect variation and comparative reconstruction a. Alpine Lombard: Val Chiavenna (Maina 2001–2); Val S. Giacomo (Merlo 1932: 268); Val Bregaglia (Stampa 1934); Val Calanca (Urech 1946) b. Other Alpine dialects from northwestern Italy: Valsesia (Northern Piedmontese, Spoerri 1918: 687–8: e.g. [’wak:a] ‘cow’, [’ʒmaɟ:a] ‘spot’ < *ex+mac(u)lam; cf. also Molino and Romano 2008: 18 [’kan:a] ‘reed’ < cannam); Borgomanero (Western Lombard, province of Novara; cf. AIS pt 129, Loporcaro 1997: 81) c. Romansh: lower Engadinian: see Pult (1897: 115) on Sent ([’ʃtop:a] ‘tow’, [’sel:a] ‘saddle’, [’fjer:] ‘iron’); Gredig (2000: 16, 61) on Scuol; Sutselvan (Luzi 1904: 54–5; Rupp 1963: 103 on Domat, Trin, and Flem);32 Surmeiran: see Lutta (1923: 284–5 on Bravuogn) d. Franco-Provençal: a generally widespread feature (cf. (77) below, and }3.4.2.2, n. 72) though absent in some dialects: cf. Martinet (1975: 205), singling out the patois of Bramans, Haute Maurienne, for lacking any ‘trace de gémination’ e. Alpine Provençal: Oulx and Pragelà (see Sibille 2007: 338) [’vaʃ:O] ‘cow’ < vaccam, [’bat:u] ‘beat.1sg’ vs [’ba:tu] ‘put on the packsaddle.1sg’ (that at least some of these geminates are non-etymological is shown by e.g. [’vit:e] ‘vine’ < vitem); see also the examples from Val Germanasca in Ch. 3, (74); f. Apennine Emilian: see Malagoli (1930: 131) and Loporcaro et al. (2006) on the dialect of Lizzano in Belvedere (Bologna); Malagoli (1954: 9) on the dialect of Collagna (Reggio Emilia); Uguzzoni (1974: 241) on the dialect of Benedello, and Uguzzoni and Busà (1995) on that of Crocette (Pavullo nel Frignano, Modena); Bernardasci (2013b) on the dialect of Piandelagotti (Modena)33

In what follows, while drawing occasionally from (48b–f ) as well, I shall focus primarily on Alpine Lombard (}5.3.1) and Southern Emilian (}5.3.2), two varieties on which I have carried out fieldwork for the present project and which provide interesting data for our diachronic reconstruction. 5.3.1 Geminate consonants and VL in Alpine Lombard Within Northern Italo-Romance, occurrence of non-reduced geminate consonants was observed, at least as early as Salvioni (1907: 729), for the dialects of the Swiss (Grigionese) valleys Calanca and Bregaglia, as well as for an Italian appendix of the latter, Val Chiavenna and Val San Giacomo. As a first illustration, consider the 32 Several of these sources describe the realization of gemination as dependent on sentence phonetics. Thus, according to Rupp (1963: 103) gemination is ‘nach dem Hauptton vernehmbar, schwindet aber, wenn das Wort im Satz gesprochen wird’ [‘is perceptible after main (utterance) stress, but disappears when the word is pronounced within a sentence’]. 33 Actually, as we have already seen in }3.4.1.2, (27) and (28), the occurrence of long consonants is not confined to Apennine dialects, being observed in Bolognese and other dialects of the Emilian plains as well.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

175

following data from the dialect of Villa di Chiavenna (province of Sondrio), from Maina’s (2001–2: 51) study on gemination in these Alpine varieties: (49)

a. [’na:s] [’va:k] [’ka:r] [’me:s]

‘nose’ ‘go.prs.1sg’ ‘dear’ ‘month’

b. [’nasˑ] [’vakˑ] [’karˑ] [’mesˑ]

‘be born.inf’ ‘cow.pl’ ‘wagon’ ‘put.fpl’

Parallel to what has been shown for the Emilian dialects considered in }3.4.1.2, (27)–(30) above, here as well a long stressed vowel is followed by a short consonant, whereas a short stressed vowel is followed by a half-long consonant. Consonant length is fully retained in words which did not undergo apocope, as well as when sonorants are involved: (50)

a. [’grOs:e] [’mat:e] [’strak:e] [’bεl:e] [’tεr:e]

‘big.fsg’ ‘girl’ ‘tired.fsg’ ‘beautiful.fsg’ ‘land’

b. [’grOsˑ] [’matˑ] [’strakˑ] [’bεlˑ] [’tεrˑ]

‘big.msg’ ‘boy’ ‘tired.msg’ ‘beautiful.msg’ ‘land.pl’

All non-degeminated consonants mentioned thus correspond to geminates in the respective etyma (e.g. [’tεr:e] < terram, [’vak:e] < vaccam). And indeed, Salvioni (1907: 729) regards such geminates occurring in the Alpine Lombard dialect of Calanca as a direct outcome of the Latin ones, whereas Merlo (1932: 268), on the other hand, considers the geminates of the dialect of Soglio (Val Bregaglia) as ‘una alterazione relativamente recente, dovuta alla brevità della vocale che precedeva’ [‘a relatively recent change, triggered by the shortness of the preceding vowel’]. More recently, Maina’s (2001–2; 2006: 446–7) comprehensive study seconds this latter opinion. Part of the motivation for this assumption may lie in the circumstance that degemination has otherwise affected the whole of Western Romance, and at a very early date, so that it might have appeared—to Merlo and others—more reasonable to assume some sparse secondary departures from this general situation than the occurrence of true exceptions to it from the outset. In-depth consideration of the phonetics and phonology of consonant gemination (and, of course, VL) in the dialect of Soglio (Val Bregaglia) will help solve this controversy, and will cast light on the origin of CVL (see Appendix, Map 5 for the localization of this and other dialects in this section). The dialect of Soglio was reported to display geminate consonants in the description by Stampa (1934: 130–33). In this description, several non-etymological geminates are recorded, as summarized in (51):34

34 Further literature on the dialect confirmed the occurrence of geminates: see also, before Maina’s quoted studies, the data collected in the unpublished Quaderni fonetici (by C. Salvioni and P. E. Guarnerio for the Vocabolario dei dialetti della Svizzera italiana, VDSI), and those reported by AIS (Soglio is pt 45), Stampa (1939), Schaad (1936; 1939), and Rinaldi (1985: 148).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

176 (51)

Dialect variation and comparative reconstruction a. [’bas:ɐ] ‘short.f’, [’tεr:ɐ] ‘earth’, [ɐl ’bof:ɐ] ‘blow.3msg’, [ɐl ’kop:ɐ] ‘kill.3msg’, [’pen:ɐ] ‘feather’, [’mam:ɐ] ‘mum’, [ɐl ’got:ɐ] ‘drip.3msg’, [’vak:ɐ] ‘cow’ b. [lɐ ’filˑɐ] ‘spin.3fsg’, [’εlˑɐ] ‘wing’, [’fOr:ɐ] ‘out’, [ɐl ’kyr:ɐ] ‘look after.3msg’ c. [’lim:ɐ] ‘(iron) file’, [ɐl ’fym:ɐ] ‘smoke.3msg’ d. [’εz:ɐn] ‘donkey’, [sul’vad:ɐk] ‘wild’, [’lar:ɐʃ] ‘larch’, [’dod:ɐʃ] ‘twelve’, [’tred:ɐʃ] ‘thirteen’

Whilst the geminates in (51a) are etymological, those in (51b–d) are not. Thus, these data could at first appear to support Merlo’s (1932) and Maina’s (2001–2; 2006) view. However, a moment’s reflection shows that this is not the case. First, the nonetymological geminates in (51c,d) have a regular explanation, since both gemination after stress in proparoxytones (of whatever consonant) and categorical gemination of Latin -m- (with merger with -mm-) are general features across Northern ItaloRomance, as can be seen from inspection of ancient texts, from data from modern dialects still retaining, at least in part, geminates, and, for the rest of modern Northern Romance dialects, in which degemination is nowadays complete, from syllable-related processes (cf. }5.3.4 below): thus, both [’fym:ɐ] ‘smoke.3msg’ < fumat and [’man:ɐk] < manicum ‘handle’, [’pyl:ɐʃ] < pulicem ‘flee’ etc. are regular. Note that this is far from being a trivial point, if no less an authority than Rohlfs(1966–9: 1.322, n. 1) considers intervocalic geminate [m:] in Bregagliotto, Valsesiano, or Genoese (quoting respectively Stampa 1934: 133, Spoerri 1918: 688, and Parodi 1902–5: 154) as evidence for the claim that post-tonic geminates in Alpine Lombard are, on the whole, not inherited but rather derive from ‘un allungamento secondario’ [‘a secondary lengthening’]. Of course, given the general process of -m-gemination across Northern ItaloRomance (}5.3.4), this is not a valid argument to refute Salvioni’s (1907) claim that geminates were retained throughout in alpine Lombard. Once the data on -m-gemination and on gemination in proparoxytones ((51c,d)) have been stripped away, we are left with the geminates in (51b). However, Stampa’s description is in need of emendation on this point, as is apparent from the transcriptions in (52a), which stem from my own fieldnotes (June 2002): (52)

a. [la ’fi:lɐ] [’ε:lɐ] [’fo:rɐ] [al ’ky:rɐ]

< filat < alam < foras < curat

‘spin.3fsg’ ‘wing’ ‘out(side)’ ‘look after.3msg’

b. [pa’dεl:ɐ] [’kwel:ɐ] [’kor:ɐr] [’tεr:ɐ]

< patellam < *eccu+illam < currere < terram

‘frying pan’ ‘that.fsg’ ‘run.inf’ ‘earth’

Crucially, there is no trace of the gemination of -l- and -r- in paroxytones, recorded by Stampa in (51b). Rather, as shown in (52b), geminate [l:] and [r:] occur exclusively in (proparoxytonic) words in which they stem from etymological

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

177

-ll- and -rr-. Since the phonetic distinction is preserved, there are (sub)minimal pairs such as the following:35 (53) a. [’ʃte:lɐ] ‘star’ < *stelam b. [’ε:lɐ] ‘wing’ < alam [’ve:lɐ] ‘old.fsg’ < ve(tu)lam [’kε:rɐ] ‘dear.fsg’ < caram

[’ʃtel:ɐ] [’εl:ɐ] [’vel:ɐ] [’tεr:ɐ]

‘drop’ < stillam ‘is.int.fsg’ < es=-(il)lam ‘sees.int.fsg’ < videt=(il)lam ‘earth’ < terram

This excludes the hypothesis that by the time of Stampa’s description the geminates in (51b) might have been there, to disappear later: had this been the case, the minimal pairs in (53a,b) would have merged and the contrast could not possibly have been restored. This leads us to conclude that the geminates in (51b) simply derive from erroneous transcriptions. Stampa was from another village of Val Bregaglia, and since only the dialect of Soglio displays gemination, people from elsewhere in the area tease people from Soglio claiming that they put geminates everywhere, including in words like those in (51b)36 (clearly, a case of hypercharacterization). It appears that Stampa fell himself victim to this sociolinguistic stereotype (cf. Loporcaro et al. 2005: 601). Once Stampa’s transcription errors have been emended, we can move on to analysing gemination and its relationship with VL in the Bregagliotto of Soglio. In an experimental study (Loporcaro et al. 2005), we have shown that both short and long stressed vowels and, respectively, long vs short consonants following the stressed vowel, are clearly distinct phonetically (see the two sonagrams in Figs 5.1 and 5.2). As in Italian, in such a paroxytonic word, vowel and consonant length are in complementary distribution, as shown in Fig. 5.3. This, of course, is not in itself enough to decide whether it is VL or consonant gemination which is contrastive in this variety. And indeed, the dialect of Soglio yields somewhat contradictory evidence on this point. On the one hand, as in all dialects of Northern Romance—including those listed in (48)—geminates before stress have been lost completely as shown in (54b) (to be compared with the alternants in (54a), where gemination is retained after stress): (54) a. [’kaʃ:ɐ] [’ʒbas:ɐ] [i ’kor:ɐn] [i la ’briʃ:ɐn]

‘case’ b. [ka’ʃOt:ɐ] ‘lower.imp.2sg’ [ʒba’sε] ‘run.prs.mpl’ [la ko’re:ɐ] ‘they.mpl open.prs.3pl it.fsg’ [bri’ʃε]

‘case.dim’ ‘lower.inf’ ‘run.impf.3fsg’ ‘open.inf’

35 *ste¯lam (instead of CL ste¯llam) has to be reconstructed on Gallo- and Rhaeto-Romance evidence (cf. }1.2 above). [’εl:ɐ] and [’vel:ɐ] are the int(errogative) counterparts to 3fsg present indicative [l ’ε] ‘(she) is’, [la ’ve] ‘(she) sees’: e.g. [ma ’le ndu ’εl:ɐ] ‘and she, where is she?’. 36 This was pointed out to me by my informants in Soglio, when I was asking for their judgement on the items in (51b).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

178

Dialect variation and comparative reconstruction

Frequency (Hz)

5000

0 0.515261

0 Time (s)

FIG. 5.1 [’ʃtel:ɐ] ‘drop’ (dialect of Soglio, male speaker, aged 73), sonagram and sound wave.

Frequency (Hz)

5000

0 0

Time (s)

0.543878

FIG. 5.2 [’ʃte:lɐ] ‘star’ (dialect of Soglio, male speaker, aged 73), sonagram and sound wave.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

179

300 250 200

170

91

C V

150 100 50

144 92

0 ˈʃtelːa

ˈʃteːla

FIG. 5.3 Mean duration values for the stressed vowel and the following consonant in the minimal pair [’ʃtel:ɐ] ‘drop’ vs [’ʃte:lɐ] ‘star’ (dialect of Soglio, female speaker, aged 60; averaged on 10 occurrences each; after Loporcaro et al. 2005: 605).

As a result, they now occur only after short stressed vowels, which implies that gemination can be regarded as non-distinctive—contrary to Standard Italian, where alternations such as canna [’kan:a] ‘reed’/cannella [ka’n:εl:a] ‘faucet’, becco [’bek:o] ‘peck’/beccuccio [be’k:ut:ʃo] ‘spout’ show that gemination is independent from stress, phonemically. At the phonetic level, however, a correlation with stress obtains in Standard Italian too, as geminates immediately following stressed vowels have been shown to be consistently longer than those occurring after unstressed vowels (cf. Stevens 2011: 27; Dmitrieva 2012: 128). Thus, phonetic degemination, as exemplified in (51b), has followed a path designed by phonetic constraints, a fact that has long been recognized in studies in Romance diachronic phonology like Lüdtke (1956: 136), who addresses the perceptual side of the coin in saying, on the evidence of peripheral dialects such as those in (48), that WRom. degemination affected ‘zunächst nur die vortonigen, die sowieso schwerer zu unterscheiden sind’ [‘at the outset only pretonic [geminates], which are anyway more difficult to discriminate’].37 While complete degemination before stress is the rule throughout this area (and in Northern Romance in general), the neighbouring dialects of Val Calanca seem to reveal some exceptions. Thus, Urech (1946: 110), in addition to singling out retention of etymological geminates as a distinctive feature of Calanchino (with examples from the dialects of Augio and Rossa, cf. (55a)), also provides not only examples of non-

37 Position within the utterance (i.e. prepausal vs non-prepausal) may also have played a role, as argued in }5.3.3 below.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

180

Dialect variation and comparative reconstruction

etymological geminates of the kind already seen for Soglio above (cf. the instance of -m-gemination in (55b), paralleling those from the dialect of Soglio in (51c)) but also examples of unexpected geminates before stress ((56a)), alongside expected instances of degemination observed for instance within morphological paradigms with alternating stress ((56b)): (55)

a. [’grOs:o] ‘big.fsg’, [’tεr:a] ‘earth’, [’vak:a] ‘cow’, [’ak:u] ‘water’ b. [’fam:] ‘hunger’ < famem, [o ’fym:a] ‘smoke.3msg’ < fumat (= [’fcam:a] ‘flame’ < flammam, [’mam:a] ‘mum’ < mammam)

(56)

a. [sa’r:ada] ‘closed.fsg’, [fi’d:ʒidi] ‘escaped.fsg’, [ga’l:ini] ‘hen’, [kra’p:a] ‘die.inf’, [na’t:a] ‘clean.inf’, [sa’s:anta] ‘sixty’, [sa’t:anta] ‘seventy’ (Urech 1946: 10–15) b. [ki’lO] ‘here’ < *eccu+illoc, [ga’tiŋ] ‘cat.dim’, [’mam:a] ‘mum’ ! pl. [ma’maŋ], [so’rεl:a] ‘sister’ ! pl. [sore’laŋ] (Urech 1946: 40, 45, 53)

Maina (2001–2: 31) regards these as due to analogical extension, following an established interpretation pattern.38 This move is instrumental to her claim that none of these geminates has been preserved all the way from Latin. Analogy surely is possible for geminates occurring within verb paradigms such as (57), which is an Italianism (otherwise, -p- in crepat would have resulted in -v- by WRom. lenition):39 (57)

[o ’krεp:a] ‘die.3msg’ < It. crepa/[kra’p:a] ‘die.inf’ < It. crepare

However, appeal to analogy is not absolutely necessary for the remaining items in (56a), and it is implausible for paradigmatically isolated words like [a’s:ε] ‘enough’ < *ad+satis, [py’s:ε] ‘more’, or for [a’l:εgɐr] ‘merry’, where no related word containing the etymological geminate after stress in the same lexical morpheme can be spotted (Urech 1946: 110–11). It seems that one cannot exclude completely the possibility that even pretonic geminates left at least some traces in modern Northern Romance dialects. Comparable evidence comes from Gallo-Romance. For the Provençal dialect of Ambialet, Maurand (1974: 153–70) describes phonemic gemination contrasts occurring also before stress, 38 Lutta (1923: 284), describing the Romansh dialect of Bravuogn/Bergün, observes that etymological geminates retain ‘eine relative Länge’ (i.e. they last longer than singleton consonants, but less than true geminates such as found e.g. in Italian), ‘Aber auch diese relative Länge ist vor dem Hauptton in der Regel verloren gegangen [ . . . ] abgesehen von analogischen Fällen wie z.B. tus : ekr̥ “husten” nach la twës : , εl twës : a, tat : e:r “saugen” nach εl tεt : a, la tεt : a’ [‘but also this relative length was lost before main stress (emphasis original—M.L.) apart from analogical cases such as e.g. tus : ekr̥ ‘cough.inf ’ after la twës ‘the cough’, εl twës : a ‘cough.3msg’, tat : e : r ‘suck.inf ’ after εl tεt : a ‘suck.3msg’, la tεt : a ‘the tit’’]. Note that here too these are all instances of etymological geminates (Lat. tussire, tittam). 39 It is normally the case that Italian loans with intervocalic voiceless singleton consonants are borrowed into northern Italian dialects with a short stressed vowel and, in dialects with geminates, with a voiceless geminate consonant (cf. Maina 2001–2: 37–8 reporting examples like [sa’lat:a] ‘salad’, [’vit:a] ‘life’ from Calanchino and Bregagliotto).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

181

for example in the two proper names [ka’nak] vs [ka’n:ak] (which, again, is not part of a paradigm, and thus cannot owe its [n:] to analogy), or in [ʃa’na] ‘geld.inf’ vs [ʃa’n:a] ‘bleed.inf’, [a’nado] ‘gone.fsg’ vs [a’n:ado] ‘year’, [ru’lan] ‘Roland’ vs [ru’l:ãn] ‘roll. prs.1pl’, [ʃa’ra] ‘salt.inf’ vs [ʃa’r:a] ‘close.inf’, etc., all cases in which the geminate is inherited (e.g. [a’n:ado] < *annatam, [ʃa’r:a] < serrare etc.). Pretonic -rr- > [r:] (also from assimilated -tr- and -dr-) preserved its length, both before and after stress, also in the Franco-Provençal patois of Val d’Illiez and Montana, by the time of Fankhauser’s (1911: 137) and Gerster’s (1927: 139) descriptions respectively: thus, Fankhauser recorded not only [’tε:r:a] ‘earth’ < terram, [’sεr:@] ‘close.3sg’ < serrat but also [tsε’r:e] ‘wagon’ < *carr+ittum and [sε’r:a̟] ‘close.inf’ < serrare).40 Likewise, Gerster (1927: 139) recorded [ar:œ’va:] ‘arrive.inf’ < *arripare, [fε’r:a:] ‘shoe.inf’ < *ferrare, [’kʎu:r:@] ‘close.inf’ < claudere, contrasting with -r- > [r] in e.g. [’u:ra] ‘hour’ < horam, [’pʎu:r@] ‘cry.3sg’ < plorat. In the Bregagliotto of Soglio, on the other hand, geminates were completely lost before stress (as shown in (54b)), but other features of this dialect depart from the average Northern (Italo-)Romance varieties in which VL has clearly become distinctive. One such feature is the non-occurrence of long vowels word-finally, as exemplified for several word classes in (58): (58)

a. Nouns and pronouns: [’ʃta ’ka] ‘this.f house’, [um ’pε] ‘a foot’, [’me/’te/’se ’frε] ‘my/your.sg/his/her brother’, [’dʒe] ‘I’, [’ty] ‘thou’, [’ly] ‘he’, [’le] ‘she’ b. Verbs: [tru’ε] ‘find.inf’, [’di] ‘say.inf’, [i nu ’sa] ‘(I) don’t know’, [a ɱ ’va] ‘I feel like’, [’le l ’ ε] ‘she is’, [a s ’fa/’pO] ‘one does/can’, [a m ’se] ‘(we) are’; [u ’ve] ‘(you.pl) have’, [u av’de] ‘(you.pl) see’, [bri’ʃe] ‘open.imp.2pl’ c. Adverbs: [’la] ‘there’, [ki’lO] ‘here’, [’ɟa] ‘already’, [’p(l)y] ‘more’

This implies that the contrast seen for the word-internal position in (53) cannot occur in the absolute word-final position, a distribution which parallels that of Tuscan-based Standard Italian (cf. }3.2, (8c)). Another Tuscan-like feature of this dialects is the occurrence of raddoppiamento fonosintattico (RF), though only with enclitics, as exemplified with object clitics hosted by imperatives in (59b) and with interrogative clitics in (60b): (59)

a. [sen’ti] ‘hear.imp.2pl’ [ba’ve] ‘drink.imp.2pl’ [bri’ʃe] ‘open.imp.2pl’

b. [sen’til:ɐ] ‘hear.imp.2pl=it.3fsg’ [ba’vel:ɐ] ‘drink.imp.2pl=it.3fsg’ [bri’ʃel:ɐ] ‘open.imp.2pl=it.3fsg’

Etymological singleton -r- was geminated only after stress in the patois of Val d’Illiez: e.g. [’dyr:a] ‘hard.fsg’ < duram, [’kjy:r:@] ‘clean the stable.3sg’ < curat vs [kʷʊ’ra :̟ ] ‘clean the stable.inf ’ < curare, [mʷʊ’rεi ̯] ‘die.inf ’ < morire (Fankhauser 1911: 137). On the contrary, non-etymological geminate sonorants used to occur in the long-extinct dialect of Certoux, which must hence have undergone pretonic gemination: e.g [s@’l:i] ‘barn’ < solarium, [s@n:’Ou] ‘play/sound.inf ’ < sonare, [fε’n:œy] ‘fennel’ < fenuculum (cf. O. Keller 1919: 153–5). 40

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

182

Dialect variation and comparative reconstruction

(60)

a. [l ’ε] ‘(s/he) is’ [la ’da] ‘(she) gives.f’ [la da’ra] ‘(she) will (she) give.f’ [’ty ty ’ve] ‘you.sg see’

b. [’εl:ɐ] ‘is (she) . . . ?’ [’dal:ɐ] ‘does (she) give . . . ?’ [da’ral:ɐ] ‘will (she) give . . . ?’ [al ’vet:ɐt] ‘do you.sg see it?’

Among the dialects listed in (48) this is not unparalleled, as RF with enclitics is attested, for example in nearby Val Calanca ((61a)) and in other Lombard dialects ((61b)), as well as in Franco-Provençal ((61c)): (61)

a. Augio and Rossa (Val Calanca): [’dim:el] ‘to tell me’, [’dit:el] ‘to tell you’, [’din:el] ‘to tell us’ (Urech 1946: 75) b. Borgomanero: [j ’o sa’pjœl:ʊ] ‘I got to know it’ (AIS pt 129, 8.1652), [j ’O pi’tsal:ʊ] ‘I lit it’ (5.913) c. Hauteville: [’leva] ‘rise.imp.2sg’ ![le’vat:e] ‘stand up.imp.2sg’ (Martinet 1939: 57), [’beta] ‘put.imp.2sg’ ![b@’tal:o] ‘put.imp.2sg it’ (Martinet 1956: 86)

If one considers medieval varieties of Northern Italo-Romance, RF occasionally occurs, even without being restricted to enclitics. For instance, in her linguistic notes to the 13th-century verses by the Anonimo Genovese, Cocito (1970: 36) observes that RF in examples like (62) ‘compare con notevole frequenza e rappresenta quasi certamente un riflesso della pronuncia’ [‘appears with noteworthy frequency and almost certainly represents a reflex of pronunciation’]: (62) e llo ‘and it.m’ (63.47) etc., a rreme ‘with oars/rowing’ (93.88), e rrixi ‘and smiles’ (53.17). In general, in modern NRom varieties RF has been uniformly lost with degemination, or, more precisely, with the application of degemination before stress—which must have preceded degemination of consonants following the stressed vowel. This is confirmed by the dialects spoken immediately south of the border between Western and Eastern Romània (Wartburg’s La Spezia–Rimini line, or Pellegrini’s Carrara– Fano line), where—as shown with the following examples from the dialect of Castello di Sambuca (province of Pistoia; cf. Weinrich 1958: 167–9; Loporcaro 1997: 95)— geminates are preserved after stress ((63a)) but lost before it ((63b)), and so is RF ((63c)):41 (63)

41

a. [’bεl:a] ‘beautiful.fsg’, [’dOn:a] ‘woman’ b. [ka’ti:vo] ‘bad’, [ra’konta] ‘tell.3sg’ c. [a ’me] ‘to me’, [la ’sta ’ma:le] ‘she’s feeling bad’, [’tre ga’li:ne] ‘three hens’

The dialect of Castello di Sambuca belongs to the area further illustrated below (}5.3.2, (67)) where singleton post-tonic consonants in proparoxytones have been geminated: e.g. [’peg:ora] ‘sheep’, [’feg:ato] ‘liver’, [’tev:edo] ‘lukewarm’.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

183

Given this link between degemination and loss of RF, it is natural that NRom varieties which have preserved, at least to some extent, geminate consonants are the only ones which are liable to show at least traces of RF to this day, as happens in Soglio. All in all, the phonology of the dialect of Soglio looks as though this dialect were still in a transitional stage between a system with contrastive gemination (like ProtoRomance) and one with CVL. The phonetics confirms this diagnosis, as shown in Loporcaro et al. (2005: 609). In fact, if one compares the mean short-to-long duration ratios [V]/[V:] for stressed vowels with those of the consonants following the stressed vowel (geminate vs singleton), the resulting ratios for the dialect of Soglio are not too far from those for Standard Italian (such as those resulting from the durations measured by Farnetani and Kori 1986: 25): (64)

Short/long stressed vowels a. Dialect of Soglio b. Standard Italian

ˈV/ˈVː 0.85 0.62

Short/long consonants after stress C/Cː 0.64 0.37

In both varieties, vowel and consonant length are in complementary distribution in stressed syllables (as seen for Italian in (2a,b), }3.1 and for Bregagliotto in (53a,b) above), and in both, length contrast is signalled at the surface more prominently for consonants than for vowels. For Standard Italian, this squares well with the fact that consonant gemination is contrastive, hence specified underlyingly, whilst VL is determined allophonically depending on stress and syllable structure.42 By the same token, one could argue that the ratios in (64a) support the conclusion—concurring with the phonological evidence reviewed in (58)–(60)—that the change from contrastive consonant gemination to CVL has not yet been completed in the dialect of Soglio. As we shall see in }5.3.3, ‘well-behaved’ dialects from northern Italy in which VL has become contrastive display a significantly different picture. This yields an argument in favour of the idea that post-stress geminates in this dialect are inherited (as assumed by Salvioni 1907). Merlo (1932: 268) and Maina (2006: 446–7) suggested, on the contrary, that these arose via a ‘rigeminazione’, i.e. a novel gemination which lengthened, after short stressed vowels, consonants which had been originally degeminated as in all of Western Romance. This assumption, though, can only be made for dialects in which VL has become contrastive: otherwise, minimal pairs like those in (53a,b) could not possibly have been kept distinct but rather should have merged as soon as Proto-Romance geminates—by Merlo’s and Maina’s hypothesis—were degeminated. We know independently that the

42 In other studies, the average ’V/’V: ratio shows still less inbalance: for instance, a ratio of 0.70 results from duration measurements in D’Imperio and Rosenthall (1999: 4).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

184

Dialect variation and comparative reconstruction

gemination contrast for -l(l)-, -r(r)-, and -n(n)- was preserved much longer than for obstruents in all WRom. varieties (see }5.3.4 below): in the dialect of Soglio, this preservation reaches unbroken into the 21st century.43 Note also that, if this analysis is on the right track, the dialect of Soglio seems to be the mirror image of Genoese, where apocope did not apply, yet CVL arose. Soglio, on the contrary, shares the application of apocope with all Lombard dialects (with the few exceptions discussed in Contini 1935b: 55–9; Loporcaro 2005–6: 83),44 as illustrated in (65): (65)

[’so:r] [’be:f] [’vet:ʃ] [’pεŋ] [’bεɲ]

< soror < bibo < video < panem < bene

‘sister’ ‘drink.1sg’ ‘see.1sg’ ‘bread’ ‘well’

[’fø:k] [’mεŋ] [’let:ʃ] [’Os:] [’om:ɐn]

< focu(s) < manum < lectum < *ossi < homin(es)/-i

‘fire’ ‘hand’ ‘bed’ ‘bon.pl’ ‘man.pl’

5.3.2 On the southern periphery of Northern Romance The situation observed in Soglio (Fig. 5.3 above), whereby a complementary distribution obtains between the duration of the stressed vowel (short vs long) and that of the following consonant (geminate vs singleton, respectively), is quite widespread in a series of eccentric dialects of Northern Romance. Indeed, most of the varieties listed in (48) display this distribution, including Franco-Provençal (see Martinet 1939: 21, 55–7; 1975: 198), as exemplified for the patois of Hauteville in (66) (see also }3.4.2.2, (75) above): (66)

a. [e’tA:la] [’py:ðO]

‘star’ ‘flee’

< *ste¯lam < pulicem

b. [’gOt:a] [’f@n:a]

‘drop’ ‘female’

< guttam < feminam

43 Of course, this need not imply that the same must be the case in every single dialect from the areas listed in (48) above. For several Franco-Provençal dialects, for instance, it is evident that a secondary gemination applied after stressed vowels: e.g. [’vOl:@] ‘fly.3sg’ < volat vs [vo:’le] ‘fly.inf ’ < volare in the Valdôtain patois of St Nicolas (H.-E. Keller 1958: 101), or [ε:’fIn:a] ‘thorn’ < spinam, [a:r’mun:a] ‘alms’ < alemosynam, with gemination, alongside [ε’fan:a] ‘span’ < spannam, with -nn- preserved in the patois of Montana (mid-Wallis; cf. Gerster 1927: 138). In the patois of Val d’Illiez (lower Wallis), not only sonorants but also obstruents are geminated after stress (Fankhauser 1911: 137): [’tõn:a] ‘thunder.3sg’ < tonat, [’lãn:a] ‘wool’ < lanam, [’ra ̟ : v:a] ‘turnip’ < rapam (where -p- > [v] lenition preceded gemination), [’la ̟ : v@] ‘wash.3sg’ < lavat. 44 Contini (1935a: 55–9) gathers AIS data showing partial retention of final non-low vowels, stretching from Valsesia (in the northwest) to Upper Brianza, southwestern Lombardy, Val Poschiavo, and Comelico (see also Rohlfs 1966–9: 1.186). For instance, from AIS 1.162 ‘knee’: Selveglio di Riva Valdobbia (pt 124) [dʒ@’nøɟʊ], Pianezza (pt 126) [dʒy’nøɟu], Bienate (pt 250) [dʒy’nødʒu], Sant’Angelo Lodigiano (pt 274) [ze’nødʒe]. In some of these Lombard dialects, final -us are centralized, and only variably deleted: e.g. in the dialect of Travacò Siccomario, 3 km south of Pavia (southwestern Lombardy, a transitional area to Emilian): [’ka:r(@)] ‘dear’ = ‘wagon’, [’stras(@)] ‘rag’, [’pje:n(@)] ‘full’ ([@] is usually deleted clause-internally, but may be realized prepausally).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

185

As shown in (66a), long stressed vowels are followed by a short consonant (within an etymological ’CVCV string), whereas after short stressed vowels, which have arisen in ’CVCCV strings or in proparoxytones, geminates are preserved or created anew ((66b)). If one further considers that in the same dialects (a) geminates were generally lost before stress and (b) VL has become contrastive in absolute word-final position (Martinet 1975: 205), this results in a system in which consonant gemination has lost distinctiveness and is reduced to an allophonic concomitant of the shortness of the stressed vowel. The situation nowadays observed in Franco-Provençal or in the dialect of Soglio— with degemination before stress, but preservation of geminates (at least phonetically) after stress—is widely documented on the southern fringe of Northern Romance from the Middle Ages. Medieval texts from the belt of territory immediately south of the La Spezia–Rimini line, spanning the northern Marche, northern Umbria, and eastern Tuscany, show systematic degemination only before stress (see A. Castellani 1972: 44ff. for a delimitation of the area), a situation which persists in the same transition area to this day, stretching west to the Appennino Pistoiese (cf. Weinrich 1958: 167–72).45 This sort of system was seen above in (63) and is further exemplified in (67) with data from the dialect of Senigallia (province of Ancona; cf. A. M. Mancini 1986: 222–30): (67)

a. [’rik:@] ‘rich’, [su’rεl:a] ‘sister’, [a’tʃet:a] ‘hatchet’ (cf. It. accetta) b. [ka’ti:vo] ‘bad’, [ga’li:ne] ‘chickens’, [bu’ta] ‘throw.inf’, [a’tʃet:a] ‘hatchet’ c. [’feg:@t@] ‘liver’, [’mεd:ik@] ‘doctor’, [’stup:id@] ‘stupid’

As shown in (67a), geminates following a stressed penult are preserved, whereas degemination applies before stress ((67b)) and new geminates arise after the stressed vowel in proparoxytones ((67c); cf. also Vincent 1988: 427). These dialects can be regarded as the southernmost outcrop—south of the La Spezia–Rimini (or Carrara–Fano) line—of Western Romance degemination, which has percolated into this territory, phonologically, though without eliminating all surface manifestation of phonetic gemination. More recent research on early texts from northern Italy has shown that the same situation—with systematic degemination only before stress and at least traces of gemination in proparoxytones—is documented also for varieties like 12th–13thcentury Romagnolo (Formentin 2007: 153–6) or 13th–14th-century Veronese (Bertoletti 2005: 194–200), spoken in areas whose modern dialects today show complete phonetic degemination, both before and after stress, both in paroxytones

45 Still further west, in the Carrarese—at the Tyrrhenian end of the divide between Western and Eastern Romània—degemination occurs before stress and gemination variably affects consonants immediately following the stressed vowel, though here this is the case for both proparoxytones and paroxytones: [pu’tan:a] ‘whore’, [sal’vat:iko] ‘wild’, [ta’pet:o] ‘carpet’ (Barbera 2012: 221–2).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

186

Dialect variation and comparative reconstruction

and in proparoxytones.46 Thus, what are today ‘average’ Northern Italo-Romance dialects were once like the varieties of ‘die Grenzzone der Toskana’, which renders ‘die Konsonantenquantitäten [ . . . ] akzent- oder vokalabhängig’ [‘the area on the periphery of Tuscany, which renders consonant length dependent on stress or on the [preceding] vowel’] (Weinrich 1958: 172). Let us now take a closer look at one dialect on this southern periphery of Western (and Northern) Romance (see Appendix, Map 6). The Southern Emilian dialect of Lizzano in Belvedere (province of Bologna), a village near the Apennine watershed (Malagoli 1930; Weinrich 1958: 170), has been already mentioned in (44) above to show that it has CVL. I have also shown in (43) that apocope applies only variably in this dialect, subject to a sentence phonetics condition. Prepausally, final vowels are not dropped, as shown in (68): (68)

a. [’lu:vo] ‘wolf ’, [’nvo:do] ‘nephew’, [a’mi:go/-a] ‘friend.m/f’, [ka’de:na] ‘chain’, [’ra:va] ‘turnip’ b. [’fritˑo] ‘fried’, [’kreʃˑe] ‘grow.3sg’, [u’ʒεlˑo] ‘bird’, [ka’valˑo] ‘horse’, [’nOtˑe] ‘night’, [’olˑa] ‘jar’ < ollam, [’ʃutˑo] ‘dry’

Comparison of (68a,b) also shows that geminates are still there, phonetically, although somewhat weakened, as Malagoli (1930: 131) says in his description: ‘la consonante lunga lizzanese suona come una consonante e mezzo toscana’ [‘the Lizzanese long consonant sounds like one and a half Tuscan consonants’]. As in the dialect of Soglio, geminate vs singleton consonants occur in complementary distribution with the preceding short vs long vowel, as has been shown experimentally by Loporcaro et al. (2006) and Filipponio (2012: 246). Fig. 5.4 displays the mean duration of consonants (on the left-hand side) in words pronounced in isolation, whereas on the right-hand side one can see that the durations of short vs long stressed vowels significantly differ, occurring in complementary distribution with respect to the duration of the following consonant.47 As shown in (44) above, this contrast in vowel duration also occurs in word-final position and is also realized when the word occurs non-prepausally, as exemplified in Figs 5.5 vs 5.6, with two sentences which contain the minimal pair [kan’ta] ‘sung’ vs [kan’ta:] ‘sing.prs.2pl’. This also shows that CVL is not only realized prepausally (a privileged context for lengthening, as discussed above in }3.5) but utteranceinternally as well.

46 Remnants of RF in Medieval NRom varieties such as those cited in (62) above point to a still more conservative stage with respect to the one analysed by Formentin and Bertoletti. 47 All measurements were carried out, in Loporcaro et al. (2006), on paroxytonic words (574 items in all) pronounced in isolation by 3 male speakers.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change

187

Duration of consonant (ms)

250

200

150

100

50

Duration of geminate [Cː] (ms)

Duration of singleton [C] (ms)

Duration of stressed V before [Cː] (ms)

Duration of stressed V before [C] (ms)

Duration of stressed vowel (ms)

250

200

150

100

50

FIG. 5.4 Mean duration values for the geminate vs singleton consonants (above) and the stressed vowels preceding geminate vs singleton consonants respectively (below) in the dialect of Lizzano in Belvedere (average values for three male speakers; after Loporcaro et al. 2006: 505–9).

In this dialect too, as in most other dialects considered in this section (and in general in Northern Romance), degemination applied categorically before stress, as shown by comparison of (69a) with (69b) (Malagoli 1930: 136ff.): (69)

a. [la’ʃa:re] [fu’ma:re] [do’ni:na]

‘leave’ ‘smoke’ ‘woman.dim’

b. [la’ʃˑo] [’fumˑo] [’donˑa]

‘leave.1sg’ ‘smoke.1sg’ ‘woman’

< <
(a)jebe ‘alum’), whereas final singleton -b was sometimes preserved (árabe), sometimes changed to some other consonant (alacrán ‘scorpion’, almotacén ‘weights and measures officer’). A similar contrast is observed for sonorants: final geminate -rr receives an epithetic vowel (Arab. ħurr > horro ‘free, freedman’), singleton -r does not (e.g. alcázar ‘fortress’ < Arab. al-qasr, aljófar < Arab. al-ĝau̯har ‘set ˙ of (small) pearls’). Further Ibero-Romance evidence sifted by Pensado (1993: 201), apart from loans, shows that geminate sonorants survived in Spanish, however marginally, in the Middle Ages and beyond. Indeed, a later date of degemination can be inferred for Western Romance as a whole from the fact that the outcomes of Lat. -ll-, -rr-, and -nn- are generally kept apart from those of their singleton counterparts (cf. e.g. Martinet 1955: 275–83; Weinrich 1958: 186–7; Cravens 2002: 93–115), as shown by the following examples, where one finds differences in consonant articulation instead of the original length contrast:51 (79)

a. jura ‘swear.3sg’ Sp./Cat./Pg. [ɾ] b. cielo [l] ‘sky’ cel [l] ‘sky’ ceo [] ‘sky’ c. man [n] ‘hand’ lluna [n] ‘moon’ mão [] ‘hand’

< iurat

6¼

< caelum < caelum < caelum < manum < lunam < manum

6 ¼ 6¼ 6 ¼ 6 ¼ 6¼ 6¼

torre ‘tower’ < turrem Sp./Cat. [r], Pg. [ʀ] (Lisbon [å]) sello [ʎ] ‘stamp’ < sigillum Sp. sell [ʎ] ‘stamp’ < sigillum Cat. selo [l] ‘stamp’ < sigillum Pg. año [ɲ] ‘year’ < annum Sp. any [ɲ] ‘year’ < annum Cat. ano [n] ‘year’ < annum Pg.

51 Cf. e.g. Mateus and d’Andrade (2000: 21), Recasens i Vives (1996: 254–67, 305–28) on the realization of the quality contrasts in Portuguese and Catalan respectively. For some WRom. varieties, descriptions vacillate as to whether the inherited contrast persists as a qualitative or as a quantitative one. Thus, in his description of the Franco-Provençal patois of Hauteville, Martinet (1956: 64) gives [ba’rõ] ‘baron’ vs [ba’ʀõ] ‘window bar’, whereas in Martinet (1970: 202), though denying it explicitly (‘le phonème /r/ n’y est pas non plus affecté par la gémination’ [‘the /r/ phoneme is not affected by the gemination either’]), he transcribes [’bOr:a] ‘fluff, flock’ < burram with a geminate [r:], as opposed to [’jOra] ‘now’ < horam.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

196

Dialect variation and comparative reconstruction

In addition to differences in quality such as those in (79), descriptions of some IberoRomance dialects report the persistence of long sonorants [n:], [l:] well into the 20th century. This is the case for Belsetán (the dialect of Bielsa, Aragón), where Badía Margarit (1950: 87–8) reports [’pεn:a] ‘rock’, [’bεl:a] ‘beautiful.f’, etc., with geminates occurring also before stress (e.g. [eska’n:arse] ‘choke.inf’ < *ex+cann-are).52 In northern Italy too, unlike obstruents, geminate sonorants were still retained in medieval texts, as shown above in }5.2.1, (37) for Old Genoese and Savonese. In Northern Italo-Romance, degemination of sonorants is usually held to have applied as late as the 13th century (cf. A. Castellani 1976: 127; Pellegrini 1972a: 70; 1975a: 103–6; Rohlfs 1966–9: 1.323; Zamboni 1976: 326–7; Formentin 2002: 98). This explains why in the Gallo-Italic enclaves of southern Italy, which were populated by settlers coming after the Norman conquest from northwestern Italy between the early 12th (Sicily) and the 13th (Lucania) centuries (cf. Pfister 1991: 102; 1994: 13; Vàrvaro 1983: 142), the outcome of Latin geminate -ll- (> [ɖ:]) corresponds to the development of the geminate (as opposed to that of singleton -l-) in the neighbouring southern Italian dialects from which the retroflex sound has been borrowed: cf. e.g. [ja’ɖi:na] ‘hen’, [tsi’vuɖa] ‘onion’ in the dialect of Piazza Armerina (province of Enna), [mo’ɖ:i:ga] ‘crumb’, [vu’ɖ:u:ð@] ‘boiled’, [ga’ɖ:i:na] ‘hen’ in that of Trécchina (province of Potenza). This, as argued by Rohlfs (1966–9: 1.323), indicates that, by the time of their redeployment in southern Italy, those northern Italian speakers must have still pronounced a geminate [l:]. The modern dialects of northern Italy also preserve indirect traces of a later date for sonorant degemination. This is the case in Venetian, where final unstressed mid vowels have been deleted after /n/ and liquids ((80a)), unless these were geminated ((80b), cf. Zamboni 1974: 26, n. 33):53 (80)

a. [’graŋ] ‘grain’ < granum 6¼ [’val] ‘be worth.3sg’ < valet

b. [’a:no] ‘year’ < annum [’va:e] ‘valley’ < vallem

In this case as well, dialects from the area stretching from the central Marche to northern Umbria and the Tuscan Apennines provide evidence for a stage which is long past in northern Italy. The dialect of Fano (province of Ancona), for instance, described in Nardini Marigliano (2000: 45–7), shows a full degemination of obstruents, both before ((81a)) and after ((81b)) stress:

52 Badía Margarit (1950: 88) also remarks that, by that time, geminate sonorants in such words were already in free variation with degeminated [n], [l]. Furthermore, Vázquez Obrador (2009: 246, n. 15) observes that, in Badía Margarit’s transcriptions, several words only occur with degeminated sonorants: e.g. anullo ‘one/two year old calf ’ < annuc(u)lum. 53 As seen in [’va:e] ‘valley’, intervocalic /l/, from both -l- and -ll-, was then vocalized in Venetian. Especially after /r/, deletion targeted specific suffixes, but the details are orthogonal to the present argument.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

The rearguard of change (81) a. [fe’broŋ] [a’dOs] [avi’ε] [ma’dʒor]

‘big fever’ ‘on (one’s back)’ ‘start/send.inf’ ‘greater’

b. [’febre] [’rota] [’fata] [’madʒ]

197

‘fever’ ‘broken.fsg’ ‘done.fsg’ ‘May’

Among sonorants, full degemination has affected -ll- protonically ((82a)) and, in rustic Fanese, also after stress, contrary to the urban dialect, which preserves [l:] after stress ((82b)): (82) a. [pa’li:na] [ga’li:na]

‘little ball’ ‘hen’

b. [’pala] (rustic)/[’pal:a] (urban) [’ʃpola] (rustic)/[’ʃpol:a] (urban) [su’rεla] (rustic)/[su’rεl:a] (urban)

‘ball’ ‘onion’ ‘sister’

As for /n/ and /r/, the whole of Fanese, both rustic and urban, shows degemination after stress irregularly, which means that the change at present is on its way to lexical diffusion in this dialect spoken at the eastern end of the linguistic border between Western and Eastern Romance: (83) a. [’dzan:a] ‘tooth/fang/tusk’ [’pan:a] ‘skin (of milk)’ [’pen:a] ‘feather’ [’kur:a] ‘run.inf’ (vs [ku’riŋ] ‘run.1pl’, [ku’ri:va] ‘run.impf.3sg’)

b. [’kana] [ka’pana] [’nOna] [’tεra] [’gwεra]

‘reed’ ‘hut’ ‘grandma’ ‘earth/soil’ ‘war’

Contrary to the liquids and to -nn-/-n-, on the other hand, no distinction is generally observed in the outcomes of the bilabial nasal. This is exemplified with Ibero-Romance in (84): (84) llama chama clama

‘call.3sg’

< clamat

= = =

llama chama flama

‘flame’

< flammam

Sp. Pg. Cat.

In Northern Italo-Romance as well, there is evidence for a merger of the outcomes of Lat. -mm- vs -m-. Phonetically, this must have been realized as gemination (Rohlfs 1966–9: 1.311), so that in the stage (lasting up to the 13th century) when all this area preserved a gemination contrast for sonorants, all intervocalic bilabial nasals must have been geminated. Indirect evidence for this comes from syllable structuresensitive processes in dialects which thereafter lost gemination completely. Consider the examples in (85), drawn from Tuttle (1991: 55–8):54 54 Tuttle proposes an explanation grounded in phonetics: cross-linguistic experimental evidence shows that [m] is intrinsically longer than other nasals. This phonetic tendency was brought to an extreme in Northern Italo-Romance, resulting in phonemic gemination. The data in (85) originally stem from Tuttle (1981–2).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

198 (85)

Dialect variation and comparative reconstruction Fanzolo a. [’bõŋ] [’pãŋ]

Barcón [’bõŋ] [’pãŋ]

<
*[mu’ʃ:i:no] > 57 There may be indeed some at least partial parallels, like the dialect of Castello di Sambuca Pistoiese, as mentioned in n. 85 below (though the evidence is scanty). For better-described marginal dialects at the southern periphery of Northern Romance, it generally holds true that whenever degemination did not apply, CVL is not found either. This is for example the case in the other archaic Emilian enclave in the province of Lucca, that of Gombitelli, as described by Pieri (1893a: 320–21), which preserves geminates also before stress (e.g. [fri’t:ada] ‘omelette’, [p:e’di] ‘appetite’, compare SI frittata, appetito) and has no CVL.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

203

[mu’ʃĩ:]. When geminates were later reintroduced owing to the contact pressure of surrounding Tuscan, these lexical items escaped the change, since degeminated [ʃ] had been meanwhile reanalysed as the intervocalic realization of /tʃ/ and, into the bargain, those words were paradigmatically isolated contrary to, say, [la’ʃ:a:va] ‘leave.impf.3sg’, where the pretonic geminate must have been restored on the model provided by paradigmatically related forms like [’laʃ:a] ‘leave.prs.3sg’. The same restoration must have taken place regularly in words like those listed in (89b). If this hypothesis is on the right track and pre-stress geminates were restored secondarily, the co-occurrence of these geminates with CVL in present-day Cológnoro ceases to constitute a problem for the general reconstruction of the rise of Northern Romance CVL advocated here: for Cológnoro too, in spite of later Tuscanization, one can reconstruct an original stage in which this dialect, like Lizzanese (}5.3.2), acquired CVL through (phonemic) degemination, though with retention of phonetic geminates after main word stress.

5.4 Making sense of the comparative picture In considering data from marginal dialects of Northern Italo-Romance, we have often seen that proparoxytones have to be singled out as a category per se, as far as the development of vowel and consonant length are concerned (see (51d) on Soglio, (67c) on Senigallia, (70) on Lizzano in Belvedere, etc.). Indeed, many scholars (e.g. Weinrich 1958 and Vincent 1988) have capitalized, for diachronic reconstruction, on the circumstance that lack of vowel lengthening in proparoxytones is sometimes concomitant with gemination of the following consonant. To illustrate this, both Weinrich and Vincent focused on dialects spoken in the area immediately south of the La Spezia–Rimini line, as exemplified above with data from the varieties of Castello di Sambuca and Senigallia (cf. (63) and n. 41, as well as (67) respectively). Gemination in proparoxytones, however, is only one of the symptoms of a clear dispreference for stressed vowel lengthening in the antepenult, of which one finds several manifestations across Romance. Some scholars, such as Fouché (1958: 214), projected back a preference for short vowels in the antepenult onto Classical (even Republican) Latin based on shortenings which must be reconstructed on comparative Romance evidence, whenever it is the case that several daughter languages independently show an outcome which, by the diachronic correspondences in (15), Chapter 2, is not compatible with the CL stressed long vowel: e.g. lu¯rı˘dum > *lu˘rı˘du(m) OFr. lort [’lort], It. lordo ‘dirty’, su¯cı˘dum > *su˘cı˘du(m) (REW 8414.1) > It. sozzo ‘dirty’, frı¯gı˘dum > *frı˘gı˘du(m) (REW 8512.2) > It. freddo, OFr. freit, Engadinian fraid, arthrı¯tı˘cum > *arthrı˘tı˘cum ‘gouty’ > Sp. Pg. OIt. artético and the other examples discussed in Fouché (1958: 177–84). There is no consensus, however, that these Romance facts really point to sound change in Republican Latin: for instance, no mention of

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

204

Dialect variation and comparative reconstruction

proparoxytonic shortening is found in the sections on vowel shortening in the Latin reference grammar by Leumann (1977: 105–11). Actually even at a much later date, one finds also lengthening of short stressed vowels in proparoxytones, as is the case in the hexametric clausula perspĭcere possit occurring in Sacerdos (probably a Roman, writing in the 3rd century: Kaster 1988: 352–3) GL 6.493.24 (cf. the recent discussion by Adams 2013: 46). Apparently, the evidence resists postulation of any strictly lautgesetzlich regularity, as to the length of the stressed vowel of Latin proparoxytones. Yet it is true that several later sound changes which applied in various Romance languages provide evidence for the above-mentioned dispreference against length. Consider open syllable diphthongization in Italian and French (seen above in }3.3), which is a phonologized extreme version of the original allophonic lengthening. In both languages, it applied regularly in paroxytones: French (il) vient ‘come.3sg’ < ve˘nit, coeur < OFr. cuer ‘heart’ < *co˘r(e), Italian viene, cuore and thousands of further examples, as opposed to just a handful of exceptions like bene ‘well’, nove ‘nine’. But in proparoxytones, diphthongization was far less systematic: one finds for PRom /ε/ e.g. It. tiepido ‘lukewarm’, Fiesole (a place-name) vs pecora ‘sheep’, mèdico ‘physician’, whereas for PRom /O/ the diphthong occurs in suocero/-a ‘father-/mother-in-law’ < so˘cerum/-am and uomini ‘men’ (related paradigmatically to the paroxytonic singular uomo < ho˘mo), and is otherwise restricted to verb forms (e.g. nuocere ‘hurt.inf’, suonano ‘sound.3pl’, muovono ‘move.3pl’ etc.), which coexist with paroxytonic forms within the same paradigms (e.g. nuoce ‘hurt.3sg’, suona ‘sound.3sg’, muove ‘move.3sg’ etc.). In the rest of the lexicon, the [wO] diphthong does not occur in proparoxytones: e.g. còfano ‘coffer’ < co˘phinum, mònaco ‘friar’ < mo˘nachum, stòmaco ‘stomach’ < sto˘machum, vòmere ‘ploughshare’ < vo˘merem (see Rohlfs 1966–9: 1.133–5). In Old French as well, there is on the one hand diphthongization in oeuvre ‘work’ < o˘peram, friente ‘noise’ < fre˘mitum, fiertre ‘coffin’ < fe˘retrum, along with the vacillation documented in e.g. tiede/tede/tieve/teve < te˘pidum (cf. Straka 1956: 254; Morin 2003: 137–49). In Italian and French, therefore, OSL and its further effects are less pervasive in the stressed antepenult than they are in paroxytonic words. This need not be the case everywhere, as shown by the Rhaeto-Romance dialect of the Tavetsch valley (Western Surselvan). This Romansh variety, as seen in }3.4.3.2, (87), has CVL, still observable for some vowels, which arose via the phonemicization of the output of OSL. On the other hand, OSL also resulted in subsequent modifications of lengthened vowels, which affected (original) proparoxytones ((92b)) on a par with paroxytones ((92a)). The following examples illustrate this with the outcomes of PRom /ε/ (data from Caduff 1952: 33–7):58

58 Note that the original proparoxytones in (92b) and (92d) have been reduced to paroxytones through apocope. However, by the time OSL applied, they still were proparoxytones, as shown by both absolute and relative chronology (cf. }5.2.2 on the date of apocope in Romansh).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture (92)

205

a. ’CV.CV

[’ei ̯r] < heri ‘yesterday’, [’lei ̯f] < levem ‘light’, [’mei ̯l] < mel(e) ‘honey’ b. ’CV.CV.CV [’mei ̯der] < metere ‘harvest.inf’, [’vei ̯der] < vetere ‘old, used’ c. ’CVC.CV [’mjazɐ] < mediam ‘half.f’, [’sjat] < septem ‘seven’, [’jarvɐ] < herbam ‘grass’ d. ’CVC.CV.CV [’pjarder] < perdere ‘lose.inf’

PRom /ε/ diphthongized to /ei ̯/ in originally open syllables, both in penult and antepenult ((92a,b)), whereas it diphthongized to /ja/ in closed syllables, both in penult and antepenult ((92c,d)). Thus, it is only syllable structure, not occurrence in penultimate vs antepenultimate position, that proves relevant for the two different types of diphthongization here. Before diphthongization to /ei ̯/, there must have been OSL in the same contexts, (92a-b), given (a) that this dialect has CVL, still manifested elsewhere, e.g. in [’pa:lɐ] ‘grassy slope’ < palam vs [’balɐ] ‘ball, bale’ < Germ. balla, mentioned above in }3.4.3.2, (87); and (b) that, as said in that section, all of RhaetoRomance can be reconstructed as having gone through a common stage for which CVL, arisen from OSL, has to be postulated. Thus, we have seen on the one hand languages in which OSL must have applied regularly to proparoxytones as well as to paroxytones (Tavetsch Surselvan),59 alongside languages in which OSL has regularly affected paroxytones, but less so proparoxytones (Tuscan, French). In this respect, dialects of northern Italy are like Italian and French, except that the dispreference for long vowels in proparoxytones becomes more pronounced. In some of them, like Milanese (as seen above in Ch. 3, (48)), all etymological proparoxytones have a short stressed vowel, whether they stayed as such (when ending in -a, (93a)) or reduced to paroxytones through apocope ((93b)) (cf. Nicoli 1983: 47–58): (93) a. [’pegura] ‘sheep’, [’legura] ‘hare’, [’strOlega] ‘gipsy(f)’, [’nivula] ‘cloud’ b. [’azen] ‘donkey’ < asinum In Ligurian dialects as well, the stressed vowel of proparoxytones usually did not lengthen (cf. (94a), from Ghini 2001: 171–2), although in some cases lengthening did take place ((94b)):60

As seen in }4.1.3 above, in her treatment of VL in Cremonese Prieto (1994: 92) assumes that ‘lengthening (or Monomoraic Foot Expansion) can only apply in paroxytones, but not in proparoxytones, which are already bimoraic’. Although she mentions the Tavetsch data (p. 87), Prieto does not address the issue about how OSL in proparoxytones could be handled under her approach. 60 In Loporcaro (2007a: 329; 2011c: 72) I followed Ghini (2001: 171–2) in claiming that in Ligurian dialects long vowels are categorically excluded from the antepenult, and extended this claim to the whole of Northern Italo-Romance. More accurately, this generalization should have been limited to Western Lombard, as the data in (93) and (94b)–(96) show. 59

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

206 (94)

Dialect variation and comparative reconstruction a. Ligurian: Genoese [’zuvenu] ‘young man’, [’karegu] ‘load.1sg’, [’navegu] ‘be at sea.1sg’; Savonese [u ’navega] ‘is at sea.msg’, [’avidu] ‘greedy’, [’arabu] ‘Arab’ b. Genoese [’pe:gwa] ‘sheep’ (< Old Genoese [’pe:guɹa] < pecora), [’la:grima] ‘tear’ (Parodi 1902–5: 157–8)

In the Standard Italian counterparts of (94a,b), one does find OSL, in spite of recurrent claims to the contrary, discussed in }5.4.1 ([’dʒoˑvane] ‘young man’, [’kaˑriko] ‘load.1sg’, [’aˑvido] ‘greedy’, [’aˑrabo] ‘Arab’, etc.), though realized less markedly than in paroxytones. In Genoese, on the other hand, though diachronically OSL tended not to apply in the antepenult (as seen in (94a)), it is clear from (94b) that there is no categorical filter against /V:/ in that context, as further testified by the fact that, when arising through vowel coalescence or other processes, such long vowels were occasionally tolerated, e.g. [’mε:ʒimu] ‘same’ (= It. medésimo, Toso 1997a: 14). Also, Genoese tolerates long vowels in proparoxytonic (postlexical) phonological words arising in verb + clitic strings (Toso 1997a: 16): (95)

[’spu:zilu] [’sε:rilu] [’pa:gime] [’la:vite]

/’spu:za + lu/ /’sε:ra + lu/ /’pa:ga + me/ /’la:va + te/

‘marry him’ ‘close it’ ‘pay me’ ‘wash yourself ’

In Emilian dialects too, the stressed syllable of proparoxytones sometimes hosts vowels which have undergone OSL, as shown by the fact that the stressed vowels in the words in (96a) show the open-syllable outcome, comparable with that found in paroxytones (as seen in (97a)), whereas some other proparoxytones ((96a)) show the outcome of the stressed vowel otherwise found in paroxytones with a closed stressed syllable (compare (97b)):61 (96)

Bolognese, proparoxytones (Filipponio 2010; 2012) a. [’tε:vla] ‘table’ < tabulam, [’frε:vla] ‘strawberry’ < fragulam, [’ε:zen] ‘donkey’ < asinum, [’rau̯ver] ‘sessile oak’ < roborem, [’po:ver] ‘poor’ < pauperum, [’pri:gwel] ‘danger’ < periculum, [’dzau̯ven] ‘young’ < iuvenem b. [’ma:ndga] < manicam ‘sleeve’, [’a:nma] < animam ‘soul’, [’gra:mla] < gramulam ‘flax brake’, [’ʂa:bet] ‘Saturday’ < sabbatum, [ʂal’va:dg] ‘wild’ < *salvaticum (for CL silvaticum), [’tavd] ‘lukewarm’ < tepidum

61

Note that the outcomes of PRom low-mid vowels were lengthened independently in Bolognese, a process which has nothing to do with (and is much later than) PRom OSL. Prieto (1994: 96) argues for a different view. According to her, a-lengthening in (96b) above (and }3.4.1.2, (31b)) applied before degemination, so that the intermediate step *[’va:k:a] (for today’s Bolognese [’va:ka] ‘cow’) must be posited—a ‘false step’, in Zwicky’s (1974) sense. However, while degemination is common to the whole of WRom. and is demonstrably an early change (see }3.3)—earlier for obstruents than for sonorants (cf. }5.3.4)—the lengthening of low and low-mid vowels observed in Bolognese is restricted to just some dialects of central Italy (see e.g. Merlo 1929 on [Borgo] Sansepolcro) and surely took place later in absolute chronology.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

207

(97) Bolognese vowel system (Coco 1970) iː i

i

ˈʂpiːga ‘ear (of corn)’ ˈdesː b. ‘say.prt.3sg’ a.

eː e ˈpail ‘hair’ ˈpaʂː ‘fish’

e ε ˈmeːl ‘honey’ ˈʂεːt ‘seven’

a a ˈlεːg ‘lake’ ˈbaːʂ ‘low’

o oː u ɔ o ˈaura ˈroːda ‘wheel’ ‘hour’ ˈkoːl ˈtarː ‘neck’ ‘tower’

uː u ˈluːz ‘light’ ˈʂot ‘dry’

Latin Common Romance PRom open σ PRom closed σ

The lexically idiosyncratic distribution of the two outcomes in (96a,b) points to a vacillation, comparable with that found in Tuscan for /ε O/-diphthongization (pècora, stòmaco vs tièpido, suòcero). It can hardly be by chance that such a vacillation is found across northern Italy, in dialects like Emilian and Ligurian (cf. (94a,b) above), whereas it is not in Milanese (cf. (93)). And indeed, a straightforward explanation is available, which is part of the general explanation proposed here for the development of CVL in the different NRom varieties compared so far. To outline this explanation, it is necessary to complete the scale in (46), inserting a further prosodic context, viz. proparoxytones, and consequently a further step: (98)

Prosodic environment: a. σ´ ]PW b. σ´σ

]PW

c. σ´σσ

]PW

i. +

ii. +

iii. +

iv. –

+

+

–

–

+

–

–

–

Length(ening): More natural Less natural

[+ = VL in the given environment]

In the review of previous accounts of NRom CVL in }4.1.–4.2, I stated the aim of showing that the different VL systems considered in this study are related, and correspond to different steps along one and the same path. This path is now schematically displayed on the scale in (98). I started in }4.1 by comparing just the two types on which the current literature in theoretical phonology has been most thoroughly focusing lately (Cremonese vs Milanese), which are now indicated as (98ii,iii), but after drawing up an inventory of the many other different systems earlier in this chapter, we have obtained a fuller picture, and a clearer idea of how the different steps relate to each other. The step schematized as (98iv) has been exemplified in }3.4.2.1 with modern Standard French and in }5.1 with a number of Italo-Romance varieties (including most dialects of Piedmont and eastern Lombardy, several dialects of Valtellina, and scattered peripheral varieties of Ligurian and Friulian).62 In these dialects, CVL has retreated from all contexts, including the most resistant one, viz. ’CV:C vs ’CVC 62 For Standard French, this is true in a diachronic, not in a synchronic sense, as the language no longer shows the different stress patterns in (98b,c), after the reduction of the residual two-syllable window for stress assignment which took place with the deletion of final unstressed schwas (e.g. homme ‘man’ [’Om@] > [Om]) between the 16th and 17th c. (cf. Zink 1999: 47, 181–3; Loporcaro 2011c: 76). Conversely, Central and

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

208

Dialect variation and comparative reconstruction

oxytones. In some of these varieties, neutralization of pairs of such shape is still only optional (cf. e.g. (11)) or is on its way towards generalization across the lexicon (cf. (17)). This means that all those dialects must have been like Milanese ((98iii)) at a previous stage, before VL started to withdraw from its last stronghold. Milanese, in turn, was like Cremonese, Emilian, and Ligurian (i.e. used to display CVL in paroxytones as well, (98b)) at a pre-documentary stage, as demonstrated in }5.1.3 through the comparison with Eastern Lombard when analysing the /i y/-lowering process which took place in those dialects. Now, the scale in (98) as a model for the gradual retreat of CVL in Northern ItaloRomance provides a straightforward explanation for the contrast observed earlier in this section between the absolute ban on long vowels in proparoxytones which seems to obtain in Milanese ((93)) and their marginal occurrence in Cremonese, Emilian, and Ligurian (cf. (94)–(96) as well as }4.1.3, (16)): Milanese is a type (98iii) system,63 which bans CVL even from paroxytones, so that its occurrence in proparoxytones would be a breach of the implicational scale, whereas the other dialects mentioned are between (98ii) and (98i). One could symbolize this transitional stage with a ‘’ sign in (98c). The same sign, turning from Northern Romance CVL to its ancestor, PRom OSL, can be used in order to qualify the non-pervasiveness of OSL in proparoxytones in Tuscan, as mirrored to this day in the irregular distribution of the /jε/ diphthong (< PRom /ε/) in that environment. 5.4.1 A phonetic constraint on vowel length: rhythmical compensation The scale in (98), which has been built up inductively through dialect comparison, has a straightforward phonetic motivation, which is schematically represented by the arrow pointing upwards on the right-hand side in (98). As for Standard Italian, the experimental phonetic literature shows that ‘rightward word extension has systematic shortening effects on the first vowel, when it is stressed’ (Farnetani and Kori 1986: 32). This decrease in stressed vowel length, as the number of syllables to the right of the stressed one increases, is illustrated iconically in Fig. 5.7, with data from D’Imperio and Rosenthall (1999: 4–8; see also Marotta 1985: 54; Farnetani and Kori 1986: 27–8; 1990: 55). This is the effect of a general phonetic constraint which has to be placed at level (7d), Chapter 3, but which, like all phonetic forces, may have an impact on the phonology.64 It is a substantive factor at work cross-linguistically, and as such it Lagoonal Venetan are synchronically like (98iv), though there is no conclusive evidence for their ever developing CVL out of PRom OSL (cf. }5.1.4). 63 In Friulian too, CVL is never found in proparoxytones (cf. }3.4.1.5, (61a.ii, b.ii, c.ii)), though with CVL in paroxytones there is a contrast between a diachronic tendency to erase it and the synchronic occurrence of contrasts, fed by different changes (cf. Ch. 6, (11) and (12)). 64 One may conceive of this constraint as related to the structure of the foot, and indeed the very label ‘rhythmical compensation’ refers to the foot, the building block of rhythm. However, the predominant

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

209

0.22

Duration (sec)

0.20 0.18 0.16 0.14 0.12 0.10 fate

fatele

fatevele

fatte

FIG. 5.7 Rhythmical compensation in Standard Italian (from D’Imperio and Rosenthall 1999: 4–8: mean [a] duration in stressed open syllables, one speaker).

has been documented for languages which have otherwise different phonological systems. Limiting ourselves to Romance, one finds evidence for it in Standard Italian, in which VL is allophonic ((7b), Chapter 3) but also, say, in Romanian—which has neither CVL nor allophonic OSL and in which, consequently, vowel durations are exclusively sensitive to more superficial factors (level (7d) in Ch. 3)—or in Friulian, where VL is contrastive (and, in the view defended here, underlying: i.e. specified at level (7a) in Ch. 3). For Romanian, the experimental study by Renwick (2012: 171–5) provides evidence that stressed vowels last longer when word-final than when occurring word-internally. (The data in (99), referring to high vowels, are averaged on the productions by fourteen female speakers; Renwick does not disaggregate duration data in non-final position by paroxytones vs proparoxytones.) (99)

Monosyllable

Final

Non-final

’V#

’VC#

’V#

’VC#

’V.

’VC.

Duration (ms.)

158

125

149

94

103

88

s.d. (ms.)

49

46

49

24

24

34

view of foot structure in current theoretical phonology seems to be disfunctional for effective analysis of the Romance data considered here, especially in view of strict foot binarity, a basic tenet which is generally upheld within this line of research, where it is maintained ‘that the inventory of foot templates can be limited to binary and unbounded, with ternary templates excluded’ (Hayes 1995: 55; emphasis original; see also Prince 1990: 361; Kager 1995: 401). Obviously, framing the constraint schematized in (98) in terms of the foot domain would imply the assumption of well-formed ternary feet, as assumed e.g. in Bafile (1996), since if the final syllable is analysed as extrametrical in a Romance proparoxytone, there is no obvious reason to expect it to have any influence on the length of the stressed vowel in the head of the preceding foot.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

210

Dialect variation and comparative reconstruction

As seen from the figures, other conditioning factors are at play as well, like syllable structure—in particular vowels in open syllables being consistently longer—but it is nonetheless clear that adding more syllables to the right has a shortening effect on the stressed vowel. Turning now to Friulian, consider (100), which reports the results of the measurements of vowel durations in Central Friulian by Finco (2007: 125):65 (100) Before a voiceless obstruent

Before /nC/ cluster

’CVCV#

’CVCVCV#

’CVCCV# ’CVCCVCV#

[’take]

[’takule]

[’sante]

‘notch’

‘spot’

Duration (ms.)

170

s.d. (ms.)

29

Before a voiced obstruent ’CVCV#

’CVCVCV#

[’santule]

[’spade]

[’spadule]

‘saint(f)’

‘godmother’

‘sword’

‘pork shoulder’

135

184

125

202

175

31

36

34

28

30

Clearly, occurrence in penult vs antepenult contributes to determining stressed vowel duration, together with other contextual factors such as the laryngeal specification of the following consonant or syllable structure (open vs closed).66 While Finco’s results in (100) all concern phonemically short vowels, there is evidence that rhythmical compensation may affect differently, within systems with CVL, short vs long stressed vowels. This is shown by the results of Bernardasci’s (2013b: 40) study on the Emilian dialect of Piandelagotti. While the mean average durations are clear evidence for a phonemic contrast (187.92 ms. for long vs 90.46 ms. for short vowels), word structure affects durations selectively, as shown in (101):67 (101) /’V:/   ]PW  ]PW

/’V/ ]PW

  ]PW

 ]PW

 ]PW

Duration (ms.)

146.53

178.90

214.58

96.41

93.12

84.12

s.d. (ms.)

22.97

42.64

76.29

15.16

16.76

20.41

The figures give the average length of the stressed vowels produced by speakers of four different varieties of Friulian. 66 Both factors are well known from the literature on phonetic universals: cf. the references quoted above while commenting on }4.1.2, (14) for the former, and Maddieson (1985) for the latter, which is universally at work, independently of whether a given language has an allophonic OSL rule. 67 The data refer to all vowel qualities and are drawn from a corpus of semi-spontaneous speech recorded with 3 speakers. The figures reported in (101) are not disaggregated by syllable structure in the quoted source. 65

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

211

As is apparent, long stressed vowels are affected by rhythmical compensation, getting progressively shorter as more unstressed syllables are added to the right, whereas short vowels are not affected, and even increase slightly in duration as the stressed syllable moves away from the end of the word. This shows that short vowels have a clearly different behaviour in Piandelagottese than in Friulian, and may suggest that in this variety they are already so short that they are not prone to further context-sensitive temporal compression, whereas this is observed, as expected, with their long counterparts. While the scale in (98) is motivated phonetically, there is evidence that it can have an impact on the phonology, in the sense that both lengthening phonological rules and the distribution of CVL seem to obey, ceteris paribus, the implications encoded in (98): there do not seem to be Romance systems, for instance, in which CVL occurs in proparoxytones, while being excluded from paroxytones, nor systems in which CVL occurs only in the penult, but not in the last stressed syllable (absolute wordfinal position has a special status, on which see }5.4.2). The same goes for allophonic OSL: there is no reported variety that applies it to the antepenult while disallowing it in the penult. There are many cross-linguistic parallels for the phonological consequences of the tendency to rhythmical compensation, the most renowned probably being English trisyllabic shortening (cf. e.g. Lahiri and Fikkert 1999). This applied both in the native lexicon (e.g. OE sūðerne > sŭðerne > ModE. southern [’sUð@n] vs south [’sao̯θ]) and to later borrowings, in particular in latinate vocabulary, giving rise to alternations such as chastity [’tʃæst@ti] vs chaste [’tʃeI ̯st]. While in modern English this shortening is nowadays a morphophonological correspondence, there are languages in which such constraints are seen at work synchronically, such as Chamorro (an Austronesian language spoken in Guam and the Northern Mariana Islands). Chamorro has OSL, which affects penultimate stressed syllables (e.g. /’nana/ ! [’nɒˑnA] ‘mother’), whereas the antepenult resists lengthening (/’higadu/ ! [’igadu] ‘liver’), like penultimate closed syllables (/’manha/ ! [’mɒn:A] ‘green coconut’; cf. Chung 1983: 37).68 Also in the dialects of southern Italy, not considered so far in this connection, the antepenultimate syllable—whether open or closed—always patterns with the closed penult, as opposed to the open penult, in the environment of phonological rules. For instance, the environment of diphthongization processes of the sort reviewed in }2.2.1, (11), and }3.5, (88)–(93), is normally described as ‘open-syllable’. This is just shorthand, though, since proparoxytones generally do not qualify for application of those processes throughout southern Italian dialects, irrespective of syllable structure. Thus, proparoxytones are exempt from the diphthongization processes exemplified above in }2.2.1, (11), and }3.5, (88)–(93): cf. e.g. [’kOt@k@] ‘rind’ < cu˘ticam 68 Word-final stressed vowels, occurring only in loanwords, are also immune to lengthening in Chamorro: /sju’da/ ! [sju’dɒ] ‘city’ (< Sp. ciudad).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

212

Dialect variation and comparative reconstruction

not **[’kau̯t@k@] in Bitontino (Merlo 1911–12: 916), [’sεm@n@] ‘sow.inf’ < se¯minat not **[’sεi ̯m@n@] in Cerignolese (Zingarelli 1899: 85), [’liv@t@] ‘bruise(d)’ < lı¯vidum not **[’løi ̯v@t@] in Biscegliese (De Gregorio 1939: 34), to be compared with the opensyllable diphthongs in }2.2.1, (11a,b) and }3.5, (88b).69 Consequently, the environment of such processes is more accurately stated as follows (cf. Loporcaro 1996: 172; 2001: 262):70 (102)

V

!

VV̯/_μ01 ]PW

Given (102), proparoxytones form a natural class of environments with closedsyllable paroxytones. Since checked syllables universally tend to host shorter vowels than open ones, this affinity can be ultimately traced back to the stronger resistance put up by the antepenultimate syllable against lengthening. In other words, the ban on diphthongization from the antepenult in (102) is but one manifestation of the phonetic pressure discussed in (98) (rhythmical compensation), breaking through the phonology in this case. Such examples can be adduced from the synchronic grammar of Northern Italo-Romance varieties as well. Consider the Western Emilian dialect of Fiorenzuola (province of Piacenza, cf. Casella 1922: 19, 43). In this dialect, the successors of Latin proparoxytones all have a short vowel ((103a)), thus paralleling original paroxytones with a closed stressed syllable ((103b)), and unlike original paroxytones with an open stressed syllable ((103c)): (103)

a. [’salaz] ‘willow’ < salicem, [sal’vadag] ‘wild’ < silvaticum = [’pasar] ‘sparrow’ < passer(um), [’sabat] ‘Saturday’ < sabbatum b. [’gat] ‘cat’ < cattum, [’sapja] ‘know.sbjv.3sg’ < *sappja < sapiat c. [’a:la] ‘wing’ < alam, [’na:z] ‘nose’ < nasum, [’la:g] ‘lake’ < lacum

In the same dialect, the ban on long vowels in the antepenult is still seen at work synchronically, when a verb with root /V:/ ((104a)) is followed by two clitics ((104c)), since this creates a postlexically proparoxytonic string, which falls under the same constraint as lexical proparoxytones. Conversely, strings consisting of the verb + one clitic do not undergo shortening, as seen in (104b):71

69 This is a general situation across southeastern Italy: see e.g. Savoia’s (1990: 352) description of the Abruzzese dialects of Popoli, Palmoli, and Tufillo, where one finds (examples from Palmoli, in the province of Chieti) e.g. [’sεu̯r@] ‘sister’ < soror vs [’sOr@m@] ‘my sister’, where enclisis induces proparoxytonic stress and prevents diphthongization, so that the stressed syllable has the same vowel as [’Ort@] ‘vegetable garden’ (closed syllable, paroxytonic). 70 The statement in (102) is a revision of that proposed in Savoia (1987: 240) for a similar diphthongization in Stiglianese, a dialect of Lucania, viz. _μ11. Savoia’s formalization is at odds with the fact that, in all varieties of southeastern Italy including those just mentioned and the dialects Savoia analyses, such diphthongizations also apply to word-final stressed vowels: this is a general property of southeastern Italo-Romance, as first recognized by Merlo (1926: 87). 71 This constraint must have already been operating in the stage preceding apocope, when all the words in (103a) still had the shape ’CV(C)CVCV.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture (104)

213

a. [’ba:za] ‘kiss.imp.2sg’, [’trø:va] ‘find.imp.2sg’ b. [’ba:zla] ‘kiss her’, [’trø:vla] ‘find her’ c. [’bazamla] ‘kiss her for me’, [’pagatla] ‘pay it.f for yourself ’

The facts reviewed in the present section provide one more demonstration of the effectiveness of the classical distinction between phonetics and phonology, as argued for in }3.2 above. We must be careful, in fact, not to confuse cases such as (102)–(104), where the phonetic tendency has precipitated a categorical phonological condition, and cases where this did not happen. Standard Italian is a case in point. This is often described, in influential studies in theoretical phonology such as Prince (1990: 374–5) or Hayes (1995: 148), as categorically lacking OSL in proparoxytones, and this description is then subjected to a formalization along the following lines: Prince suggests that the mechanism in such cases is a late application of extrametricality, which forces lengthening to preserve a well-formed foot [ . . . ] This mechanism cannot derive lengthening in an antepenultimate syllable, since to make the rightmost two syllables extrametrical would require illegal chained extrametricality. (Hayes 1995: 148)

For the empirical observation to which this analysis applies, Prince (1990: 374) refers to Calabrese (1984–5: 19; see also Calabrese 2000: 69, reiterating the same description): there is a constraint on lengthening and diphthongization in Italian; no lengthening or diphthongization occurs when the stressed vowel is in the antepenultimate syllable, i.e. in the case of antepenultimate stress.

However, this statement is incorrect, given the coexistence (seen above) of tièpido, Fièsole, suòcero, uòmini with the (admittedly more frequent) cases when PRom /ε O/ did not undergo diphthongization. Irregular application of this sound change, which insisted on previous OSL, serves as a litmus test confirming that the stressed vowel of proparoxytones, in Tuscan, actually occupies an intermediate position, as for VL, between vowels in open and those in closed stressed syllables, as predicted by a combination of the syllable-structure parameter with the scale in (98). The precondition for diphthongization itself, viz. OSL, still seen at work for the remaining vowels, shows a similar behaviour: commenting on the data reproduced in Fig. 5.7 above, D’Imperio and Rosenthall (1999: 4) remark that the duration difference between the stressed vowels of /’fate/ ‘do.imp.2pl’ /’fate+le/ ‘do.imp.2pl + them(f)’ ‘does not reach significance’.72 On the contrary, the statistics run on the mean values measured for the third subject analysed in Farnetani and Kori (1986: 35–8) do reveal significant differences between the stressed vowels in /’skala/ ‘ladder’ (268 ms.) vs /’skala+la/ ‘climb.imp.2sg it(f)’ (212 ms.) vs /’skalda/ 72

This is the empirical motivation for their formallization which assumes the winning output candidate for the input /krédito/ ‘credit’ to be {cré.di.}to, without OSL, rather than *{cré:.}di.to (D’Imperio and Rosenthall 1999: 15).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

214

Dialect variation and comparative reconstruction

‘heat.3sg’ (147 ms.). Clearly, there are idiolectal differences at play: in any case, claiming that in Italian the stressed vowel in an open syllable of a proparoxytone simply does not undergo OSL is factually incorrect. The phonetic tendency towards rhythmical compensation, in Tuscan-based Standard Italian, is still observable as such and has not made it into a categorical phonological property, unlike what has been suggested by formalizations such as the one just quoted (a ban on ‘illegal chained extrametricality’, Hayes 1995: 148). 5.4.2 Romance oxytones and VL As to the scale in (98), there is still one important point to be discussed: the special status of oxytones in which the stressed vowel occurs in word-final position. With respect to oxytones, in fact, we have up to now been considering evidence which may seem contradictory. On the one hand, there is cogent experimental evidence that vowel duration tends to decrease as more syllables are added to the right of the stressed one (cf. Fig. 5.6 and (99)–(101)); on the other, we have seen that lengthening phonological rules like the OSL at work in Standard Italian (}3.1, (3)) may be blocked precisely in the ultima, whereas they apply in stressed syllables further left in the word.73 The contradiction is only apparent, though: the situation becomes clear as soon as one distinguishes between the phonetics and the phonology of VL. Under a view of sound systems like that defended in }3.2 above, which recognizes a phonology–phonetics interaction, rather than unifying the two, the substantial/phonetic motivation—the arrow on the right in (98)—is expected to constrain the phonology (the box in the middle), without determining it exhaustively, though (cf. the discussion on }3.2, (7)). In fact, not only does the phonology impose binary choices on phonetic gradients, it also maintains room for self-organization—as Hyman (2001) convincingly argues—and can even impose constraints of its own, not (directly) grounded phonetically and, at times, even contradicting phonetically motivated tendencies. In the case of NRom CVL, proofs that this has long been conventionalized—and hence has autonomized itself to some extent from phonetic forces such as the one symbolized by the arrow in (98)—are legion. For instance, I have mentioned in (97) above that in Bolognese—as well as in many other dialects of Emilia-Romagna— originally short (qua originally in PRom closed syllables) stressed low and lowmid vowels became lengthened, whereas high-mid and high short vowels stayed short. This is in keeping, as noticed as early as Schürr (1919: 60), with a general phonetic constraint which causes lower vowels, ceteris paribus, to be intrinsically

As already mentioned in Ch. 3 n. 2, there is solid evidence for the shortness of the stressed final vowel in e.g. It. [ve’r:a] ‘come.fut.3sg’: Bertinetto (1981: 75, 254); Vayra (1994), van Santen and D’Imperio (1999); Stevens (2012: 309–11). 73

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

215

longer than higher ones (Lehiste 1970: 18–19; Maddieson 1997: 623–4). Consider however the following examples. In the Eastern Lombard dialect of Sant’Apollonio di Lumezzane (Ferrari 2007–8: 77–8) (cf. (20)–(26) above), vowels are always long before /r/ + C clusters: e.g. [’we:rda] ‘green.fsg’, [’fu:rma] ‘shape’, [’hu:rda] ‘deaf.fsg’, [’ky:rta] ‘short.fsg’. Only /ε/ resists this lengthening ([’εrba] ‘grass’): the fact that a low–mid vowel did not undergo lengthening in exactly the same environment in which higher vowels did is not in line with phonetically based expectations and its result is, synchronically, a phonological idiosyncrasy, as is the distribution of long vs short vowels before /sC/ clusters in Cremonese, mentioned above in Chapter 3, n. 38: one finds long stressed vowels wordinternally but not in the final syllable (e.g. [’pa:sta] ‘pasta’ vs [’past] ‘meal’), which does not make sense phonetically, given the universal tendency for stressed vowels to shorten as further syllables are added within the same word (and foot). An environment in which substantive phonetic constraints and constraints on linguistic form imposed by the phonology often happen to divorce is the word-final position. Many languages of the world impose special restrictions on quantity in oxytones and, more broadly, there seems to be a conspiracy against assigning prominence to the final syllable in terms of stress, vowel length, and heaviness.74 Latin and Standard Italian are two cases in point: the former has no final stress, the latter does, but does not allow vowel lengthening word-finally. Mester (1994) captured this by means of a markedness scale, further refined by D’Imperio and Rosenthall (1999), to account for Italian as well: (105)

Markedness scale for final syllables (Mester 1994; D’Imperio and Rosenthall 1999) extrametrical, weak branch > stress > heavy

This hierarchy does not follow from any substantive constraints, but is purely phonological in nature. According to the hierarchy, allowing for final stressed heavy syllables is the most marked option for a phonological system, whereas the most In Choctaw (a Muskogean language spoken in Mississippi and Oklahoma), for instance, final syllables, independently of stress, cannot host a word-final long vowel (cf. Lombardi and McCarthy 1991: 18). Cf. also the ban on the application of OSL word-finally in Chamorro (n. 68). On the other hand, phonological constraints may also have the symmetric effect of barring word-final short vowels, especially from monosyllables, in languages in which ’V]PW strings are incompatible with prosodic minimality requirements. Thus, for instance, in Mayo (an Uto-Aztecan language spoken in Mexico, between the states of Sonora and Sinaloa) no monosyllabic words of shape *[CV] may occur at the surface, though in Hagberg’s (1990: 160–61) analysis there are several monosyllables which have that shape underlyingly, as seen in examples like (ia): 74

(i)

a. kaá kó’okore ‘he is not sick’ neé bítcha ‘he sees me’

b. ka-tím kó’okore ne-chím bítcha

‘they are not sick’ ‘they see me’

The negation and the 1st person pronoun are realized as kaá and neé respectively when occurring as separate words, whereas they surface with a short vowel ((ib)) whenever they are followed by an enclitic.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

216

Dialect variation and comparative reconstruction

unmarked one is for the ultima to be the weak branch of the word-final foot or to be skipped altogether by the stress-assignment algorithm, as happens in Latin. Italian has a more permissive option on this scale, although it does not allow—at least within the native lexicon—for a final stressed syllable to be heavy. This constraint is overcome completely in most dialects of Northern Italo-Romance, which nowadays allow for long vs short vowels to contrast in word-final position, as shown repeatedly while reviewing the relevant data in }3.4.1.75 To exemplify, consider now the two Western Lombard varieties in (106a,b): (106) a. Milanese b. Dialect of Casale Corte Cerro

‘sing.inf’ [kan’ta] [kan’ta:]

‘sing.ptp.m’ [kan’ta:] (Salvioni 1884) [kan’ta] (Weber Wetzel 2002)

As is apparent, both contrast long vs short vowels word-finally, but with a mirrorimage distribution.76 Comparable divergences occur widely across Northern ItaloRomance, and are evidence for the fact that CVL did not simultaneously originate word-internally and word-finally. Rather, under the view adhered to here, CVL first arose word-internally when consonant degemination (first applying before stress: cf. }}5.3.1 and 5.3.2) blurred the phonological difference between the environments ’CV:CV vs ’CVC:V.77 Once this novel /V:/ vs /V/ phonemic contrast was established, it was then extended to the word-final position as well,78 with a distribution differing from dialect to dialect, as seen with the example of the infinitive vs past participle endings in (106a,b), where Milanese and Casalese diverge. Further Western Lombard dialects take yet another option. In Tortonese, for instance (Tortona lies in the linguistically Lombard province of Alessandria, in Piedmont), both infinitives and past participles end in a short stressed vowel ((107a)), though in word-internal position CVL with the Milanese kind of distribution is observed (i.e. (107b) vs (107c); cf. Cabella 1999: 142f., 152, 215, 273):79

75

Of course, these dialects also acquired, via apocope, many oxytones ending in a consonant, which however are a different case, less heavily penalized according to the markedness scale in (105). 76 Sanga (1997b: 255) lists the occurrence of a long final vowel in the participle ending among the characteristic features of Western Lombard, though note that the one in (106b) is a Western Lombard dialect as well, though spoken in a town (Casale Corte Cerro, in the province of Verbania) belonging to Piedmont administratively. 77 As noted in }5.3.4, this is likely to have happened at a stage in which geminates were phonetically preserved after stress: at that stage, a string [’CVC:V] was reanalysed as /’CVCV/, contrasting with /’CV:CV/. 78 As explained in }5.3.5, the ‘displaced’ Emilian dialect of Cológnora in Valleriana (an enclave in the province of Lucca) is the only apparent exception to this generalization, among the dozens of systems scrutinized for the present research. 79 In general, in searching Cabella’s (1999) dictionary, I could not find any oxytones with a long wordfinal vowel: e.g. [’lε] ‘she’, [’ly] ‘he’, [’lu] ‘wolf ’, [g ’ø/’e/’a] ‘have.1/2/3sg’. Needless to say, this should be tested through dedicated fieldwork, which I could not carry out on Tortonese for the present research. The same caveat holds in principle for the data drawn, in what follows, from other sources in which the length of the final stressed syllable is not discussed explicitly.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture (107)

217

a. [kan’ta] ‘sing.inf’ = ‘sing.ptp.m’, [fi’ni] ‘finish.inf’ = ‘finish.ptp.m’ b. [’na:z] ‘nose’ < nasum 6¼ [’nas] ‘be born.inf’ < nascere c. [’pala] ‘baker’s shovel’ < palam = [’spala] ‘shoulder’ < spat(u)lam

As we saw in }5.3.1, when lack of contrast in word-final position cumulates with retention of geminates (as well as with remnants of RF and a more pronounced phonetic signalization of consonant gemination than of VL word-internally), as in the dialect of Soglio, it is justified to claim that in such a system the change from contrastive gemination to CVL has not been completed yet. But this cannot be claimed for a dialect like Tortonese, given minimal pairs such as (107b). Cross-dialectally diverging distributions, as now exemplified in (106) and (107) for Western Lombard, are observed in other regions as well, within Northern ItaloRomance. In Central Emilian, one normally finds CVL word-finally, with the same distribution as in (106a): e.g. [’frε:] ‘shoe.inf’ vs [’frε] ‘wounded’ in Benedello (Pavullo nel Frignano, province of Modena; cf. Uguzzoni 1971: 127). In the Western Emilian dialect of Fiorenzuola, on the other hand, stressed vowels occurring wordfinally seem to be consistently short, according to Casella’s (1922: 22ff.) description: [’mi] ‘I’, [’ki] ‘here’, [’li] ‘there’, [ak’si] ‘so’, [’pO] ‘then’, [pe’rO] ‘but’, [ki’mO] ‘here’, [ʃa’mO] ‘already’, [ba’ty] ‘beaten’, etc. At the opposite end of the region, all wordfinal stressed vowels are short in Romagnolo as well—as highlighted by Weinrich (1958: 221), drawing on data from Schürr (1919)—including those which occur exclusively as long word-internally (low-mid and low vowels): cf. Riminese [’bO] ‘ox’, [li’mε] ‘rasp.inf’.80 This applies not only to original oxytones, like those just mentioned, but also to secondary ones: e.g. [’fjø] ‘son’ < filiolum, [’fja] ‘breath’ < flatum, [’fra] ‘friar’ < fratrem. As to this aspect, this dialect resembles Tuscan, in which secondary oxytones which have arisen through apocope have also been equated to the original ones, as seen for example in their triggering RF: [’ka ’f:oskari] ‘Ca’ Foscari’ (a Venetian building’s name, pronounced this way in the Tuscan-based standard), [’fra m:ar’ti:no] ‘brother Martin’, etc. The vowel in these forms was long at the surface, before apocope ([’ka:sa], [’fra:te]), but retaining a long vowel wordfinally was simply not an option, neither in Tuscan nor in the dialect of Fiorenzuola. In general, it is fair to say that, even in dialects which did generalize the /’V/ vs /’V:/ contrast word-finally, final short stressed vowels have a higher type frequency. Consider, as a further illustration, the Western Lombard dialect of Novate Mezzola, described by Bonfadini (1997: 596–8). In this variety, all original oxytones have /’V/ ((108a)), whereas long stressed word-final vowels occur only if vowel coalescence or apocope have applied ((108b)), and not even always then ((108c)): 80 In Bolognese, on the other hand, primary oxytones generally have a short /’V/, even when the same vowel would occur as categorically long word-internally (e.g. [’la] ‘there’, [’pjo] ‘more’), except where vowel coalescence has applied, as in [’pi:] ‘feet’ < *pedi, [’si:] ‘six’ < *sεi < sex (see Coco 1970: 5–26 and (113a) below).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

218

Dialect variation and comparative reconstruction

(108) a. [’trε] ‘three’, [’sta/’sa/’va/’da] ‘stay/know/go/give.3sg’, [’la] ‘there’, [’ʃa] ‘here’, [’pjy] ‘more’, [’sø] ‘above’, [’dʒu] ‘down’ b. [’ly:] ‘he’ < (il)lui, [’fjø:] ‘son’ < filiolum, [’bø:] ‘ox’ < bovem c. [ma’rø] ‘ripe’ < maturum, [’di] ‘day’ < diem, [’ni] ‘nest’ < nidum, [’krø] ‘raw’ < crudum All in all, it can be assumed that the superimposition of the VL contrast on wordfinal stressed vowels superseded an earlier layer in which, throughout Northern ItaloRomance, such vowels were uniformly short, as in Tuscan. This hypothesis seems to be confirmed by the outcomes of the stressed vowels, whose quality tends to be identical in word-final position and in closed stressed non-final syllables, as opposed to open stressed ones. This is illustrated by the outcomes of proto-Romance /i u/ in Bolognese in (109) and with the development of proto-Romance /e o/ in a Central Ladin dialect, Fassano, in (110): (109)

Bolognese (Gaudenzi 1889; Coco 1970) a. ’V

(110)

b. ’V]PW

c. ’VC

/i/

[’vi:d] ‘see.prt.3sg’ [’de] ‘day’ [’spi:ga] ‘ear (of corn)’ [ak’se]‘thus’

[’des] ‘say.prt.3sg’ [’mel] ‘thousand’

/u/

[’kru:d] ‘raw’ [’lu:z] ‘light’

[’brot] ‘ugly’ [’sot] ‘dry’

[’pjo] ‘more’ [vir’to] ‘virtue’

Fassano (Elwert 1943: 44–5, 52–3) a. ’V

b. ’V]PW

c. ’VC

/e/

[’tεi ̯la ‘cloth’ [’mεi ̯s] ‘month’

[’me] ‘me’ [’te] ‘thee’

[’sek] ‘dry’ [ka’peʃe] ‘understand.1sg’

/o/

[’krou̯ʃ] ‘cross’ [a’lo] < adillo¯c ‘there’ [’sou̯la] ‘alone.fsg’ [ni’o] < nec u˘bi ‘nowhere’

[’rot] ‘broken’ [’sort] ‘deaf ’

If one ignores the fact that the (a) vs (b),(c) contrast has resulted in phonemic quality differences here, this is the same distribution as observed in Tuscan. But this is not a general situation across Romance, and not even across NRom. In French (cf. (111)) stressed vowels in final position pattern with those in open rather than closed syllables, and the same is true of the dialects of southeastern Italy (see nn. 69 and 70), as exemplified in (112) with the Apulian dialect of Altamura (province of Bari) (both examples refer to the outcomes of PRom /e/):

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture (111)

French a. ’V. /e/

(112)

219

b. ’V]PW

toile ‘canvas’ toi ‘thee’ poil ‘hair’ moi ‘me’

c. ’VC. dette ‘debt’ vert ‘green’

Altamurano (province of Bari; Loporcaro 1988a: 31–4) a. ’V. /e/

b. ’V]PW

[’pai̯p]‘pepper’ [’tai̯] ‘thee’ [’mai̯s]‘month’ [’mai̯]‘me’

c. ’VC. [’pεʃ:]‘fish’ [’sεk:]‘dry.f’

All of the changes which took place in open syllables in (109)–(112) have affected an allophone previously lengthened via OSL. Consequently, the occurrence of the different distributions in (109) and (110) vs (111) and (112), as argued in Loporcaro (1997: 70-72; 2011c: 73-5), makes clear that the application or non-application of OSL to the word-final position (depending on the absence or presence in the rule of the condition obtaining today in languages like modern Standard Italian)81 cannot be reconstructed as holding uniformly for Proto-Romance. Conversely, the evidence from Northern Italo-Romance is compatible with its convergence with Tuscan in this respect: original, long word-final stressed vowels must have been excluded.82 Indeed, microvariation across the dialects spoken on the Tuscan–Emilian border, straddling the southern limit of Northern Romance on the La Spezia–Rimini (or Carrara–Fano) line, provides telling evidence for the stepwise extension of the set of long vowels in word-final position, and thus for the steps through which a VL contrast came into being (in, say, Bolognese, Milanese, etc.) in that context. This areal microvariation has been investigated recently by Filipponio and Nocchi (2010) and Filipponio (2012: 47–9, 245–6) for the villages forming the outlying parts

81 This is the condition specified in }3.1, rule (3), which bars lengthening word-finally in Sardinian as well: cf. Logudorese [g:a’f:ε] ‘coffee’, [’mε] ‘me.obl’, [g:a’t:O] ‘almond sweetmeat’. 82 In Loporcaro (2011c: 687, n. 27) I surmise that the occurrence of RF in Proto-Romance— demonstrated in Loporcaro (1997) on the strength of comparative reconstruction and philological evidence from (Late) Latin—may have played a role in the subdivision between languages/dialects of type (109), (110) and those of type (111), (112). Geminates, in Proto-Romance as in modern Italian, required shortness of the preceding stressed vowel, and this must have held also for those derived via RF. Assuming that the restriction in Ch. 3, rule (3), was not there uniformly from the outset, one might speculate that OSL was originally free to apply to word-final stressed vowels occurring prepausally, while being inhibited within the utterance, whenever an initial consonant followed. Then, varieties of type (111), (112) generalized the (formerly) prepausal allophone as a new underlying representation of the vowel involved, whereas in those of type (109), (110) the short allophone, conditioned by RF, prevailed. The dialect of Cológnora in Valleriana, described by Saiu (2012) (cf. }5.3.5 above), provides some interesting clues in this respect, since RF, as shown in (90) and (91) above, occurs exclusively after distinctively short stressed vowels.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

220

Dialect variation and comparative reconstruction

of the comune of Sambuca Pistoiese (province of Pistoia: see Map 7 in Appendix), a territory which belonged to Etruria in antiquity and was then disputed for centuries between Tuscany and Emilia. This is a focal point (as already recognized by Weinrich 1958: 167–9, who discusses Sambucano at length) in which the Tuscan quantity system, corresponding to PRom as argued in Chapter 2 and }3.2, gradually yields to the innovatory NRom system. Filipponio and Nocchi investigated the outcomes in these dialects of the words which resulted in secondary oxytones ending in a long vowel in Emilian, as exemplified in (113a) with the rural Bolognese variety of Lizzano in Belvedere, spoken some 18 km northwest of Sambuca (cf. Malagoli 1930: 139–50; urban Bolognese displays the same distribution: cf. Coco 1970: 5–26):83 (113) i. SEX > *ˈsεi a. Lizzano in B.

ˈseː

b. Castello

ˈVː#

ii. PEDES > *ˈpεdi ˈpeː

iii. -ATIS ˈkantaː

f. Standard It.

ˈsεˑi ‘six’

fraˈdeː

paˈjɔːli

fraˈdεlːi

ˈkantaːte

paˈjɔːli

fraˈdεlːi

ˈpjeːdi

ˈkantaːdi

paˈjoːli

fraˈdεlːi

ˈpjεːdi ‘feet’

ˈkantaːte ‘sing.2pl’

paˈjɔːli ‘cauldron’

fraˈtεlːi ‘brothers’

ˈV#

e. Lagacci

paˈroː

v. *FRATELLI fraˈdεjˑi

c. Cavanna d. Stabiazzoni

iv. *PARIOLĪ

The scale has been built up inductively, classifying the outcomes of the counterparts of Emilian secondary oxytones in the dialects at issue.84 At the other end of the scale with respect to the representative of Emilian, Lizzanese, one finds the Standard Italian non-oxytonic counterparts of the same words ((113f)). As shown in (113b–e), the frazioni of Sambuca Pistoiese shift gradually from (113f), the most conservative stage, towards (113a), following a south–north progression (see Appendix, Map 7): not unexpectedly, the closer to Pistoia, the more conservative (Lagacci), and the closer to

83 Similarly to what has been shown above for other NItRom dialects, in Lizzanese too primary oxytones invariably have a contrastively short stressed vowel: e.g. [’ε] ‘is’ < est, [’pO] ‘then’ < post, [’di] ‘say.imp.2sg’’ < dic, [’li] ‘there’ < illic, [’ʃi] ‘yes’ < sic, [’dʒo] ‘below’ < deo(rsum), [’su] ‘above’ < su(rsum), [’tu] ‘you.sg’ < tu, etc. The same applied to oxytonic loanwords such as [ka’fε] ‘coffee’, [dʒi’lε] ‘waistcoat’, and the like, showing that shortness is still the default option for word-final stressed vowels. 84 I have reported just a selection of the data discussed in Filipponio and Nocchi (2010: 232) and Filipponio (2012: 48). Each word in (113i–v) stands for a whole class: e.g. (113i) also includes [’du:] ‘two.m’, [’le:] ‘she’, [’lu:] ‘he’, [fa’re:] ‘do.prs.cond.1sg’, (113ii) also the masculine plural possessives [’me:], [’to:], [’so:] as well as [’fjo:] ‘son’ (all Lizzanese examples). I repeat that these classes are defined in a purely bottom-up way, through inspection of the distribution of oxytonic vs non-oxytonic outcomes in the dialects in (113b–e). Since the change has evidently progressed via lexical diffusion, one finds words whose etyma had similar phonological properties distributed over different subclasses: e.g. in Stabiazzoni and Cavanna one finds *pario˘lı¯ > [pa’jO:li] vs *filio˘lı¯ > [’fjo:].

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Making sense of the comparative picture

221

Bologna (Castello) the more innovative. Thus, in the dialect of Castello di Sambuca ((113b)), only words stemming from llı¯-ending etyma ((113b.v)) among those resulting in oxytones in Bolognese (or in Lizzanese) stayed paroxytonic, whereas in Cavanna also those originally ending in -Vlı¯ did ((113c.iv,v)), and so on. Significantly, in the dialects in which fewer lexical types have become oxytones, the stressed final vowel is short. In Lagacci and Stabiazzoni the word for ‘two.m’, which must be traced back to *duı¯ in this area (cf. Malagoli 1930: 140), is realized as [’du] ((113d,e.i)). In a sense, this parallels what happens synchronically in Florentine, when a word-final postvocalic vocoid gets deleted clause-internally, as in /’due ’kani/, which is realized as [’du ’ha:ni] ever since the deletion of a postvocalic (semi)vowel started to apply, in the mid-13th century (as demonstrated by A. Castellani 1952: 106–7). Given that this language has no CVL, the phonetic length of the stressed vowel in [’duˑe] cannot resist oxytony, where VL is barred by a general condition. Thus, in the dialect of Lagacci, even if apocope in (113e.i) has become non-conditioned, unlike Tuscan, the vowel has remained categorically short, as in Tuscan. Conversely, in the dialects in which more sub-classes among those in (113i–v) became oxytonic, this determined a gang-effect, as it were, fostering preservation of the VL which arose through vowel coalescence. This is how the dialects of Castello and Cavanna acquired a VL contrast in word-final position. Since all the dialects in (113b–e) have undergone pretonic degemination and gemination in proparoxytones while retaining phonetic geminates after stress (as shown for Castello di Sambuca in (63) and n. 41 above), those in which secondary oxytones did not tolerate long vowels (113d–e) are identical, as for the consonant/ vowel quantity relationships, to the Alpine Lombard variety of Soglio (as remarked by Filipponio 2012: 49, n. 74).85 On the other hand, those in (113b,c) are more similar to Lizzanese structurally, except for the somewhat less pervasive distribution of final long vowels (given that oxytony was fed by fewer classes of words) and for the phonetics of word-final VL, whose details are discussed in Filipponio and Nocchi (2010: 238–44). Figs 5.5 and 5.6 above illustrate the persistence of word-final CVL in Lizzanese in utterance-internal position. On the contrary, in the dialect of Cavanna, as shown in 85 Actually, the dialect of Castello di Sambuca displays at least some pretonic geminate consonants ((ia)), alongside many instances of degemination ((ib)), as recorded by Bonzi (1972–3: 71–2) (thanks to Emanuele Saiu for pointing this out to me):

(i) a. seccado ‘dried’, chiacchierone ‘chatter-box’, lettigare ‘quarrel’, cuccumelli ‘primrose.pl’ b. sofritto ‘browned’, canèlla ‘rolling-pin’, fritèlla ‘pancake’, fritada ‘omelette’, alòro ‘laurel’, garétto ‘heel’, arodín ‘knife-grinder’ (compare SI soffritto, canna + the suffix -èlla, frittèlla, frittata, allòro, garrétto, arrotino) Here too, just as in Cológnora, one may assume the pretonic geminates in (i.a) either to be remnants of a previous situation—given that gemination was there also pretonically in the first place—or to have been reintroduced via lexical diffusion from the neighbouring Tuscan dialects of the Pistoiese area, supported over many centuries by the prestige of the Florentine-based standard.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

222

Dialect variation and comparative reconstruction

Fig. 5.8, phonemically long word-final vowels are shortened utterance-internally (see the second box) and thus become non-distinct from word-final short vowels (third box), similarly to what is reported (for CVL in general, not just in word-final position) for dialects like those of Fiorenzuola (cf. Casella 1922: 51) or Cremona (cf. Rossini 1975: 183–4), as exemplified in }3.5 (95) and (96). This is one more piece of evidence for the interplay between CVL, stored lexically and hence observable at word level, and the phonetic implementation of vowel duration, showing that even the manifestation of phonologically relevant features may be obscured by conflicting phonetic forces of the kind discussed above in }}3.2 and 3.5 (shortening in utterance-internal position, as opposed to prepausal lengthening). Comparison with the dialects of type (113d–e) is also revealing as to phonetic detail, as shown in Fig. 5.9. Measurements of the duration of final stressed vowels in word-final position confirm the perceptual impression by the transcriber (and the native speakers’ judgement) that they never are distinct phonetically by length, either utteranceinternally or prepausally. These dialects are therefore just like Tuscan in this respect, for they have not established a novel CVL.86 Those spoken further north ((113b,c)), on the other hand, clearly did, as shown in Fig. 5.8. Filipponio and Nocchi’s (2010: 220 200 180 160 140 120 100 80 60 40 N=

3

3

3

3

ˈVː/ __ ]pu

ˈVː/ __ X]pu

ˈV/ __ ]pu

ˈV/ __ X]pu

where X contains at least one PW

FIG. 5.8 Durations of word-final long (left) vs short (right) stressed vowels in utterance-final (first and third boxes) vs utterance-internal (second and fourth boxes) position in the dialect of Cavanna (after Filipponio and Nocchi 2010: 239). 86 At least, not word-finally. Since pretonic degemination has occurred, a margin of indeterminacy remains, as in the dialect of Soglio, with regard to the analysis of [’CV:CV] vs [’CVC:V] strings.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Taking stock of the reconstructive evidence

223

220 200 180 160 140 120 100 80 60 40 N=

4

4

4

4

ˈVː/ __ ]pu

ˈVː/ __ X]pu

ˈV/ __ ]pu

ˈV/ __ X]pu

where X contains at least one PW

FIG. 5.9 Durations of word-final stressed vowels in secondary oxytones (displaying long vowels in Bolognese), on the left-hand side, and in primary oxytones (displaying short vowels in Bolognese), on the right-hand side, in utterance-final vs utterance-internal position in the dialect of Stabiazzoni (after Filipponio and Nocchi 2010: 240).

239) and Filipponio’s (2012: 49, 245) analysis of the facts is that the dialects of Castello and Cavanna (Fig. 5.8) do not feature CVL, just like those of Stabiazzoni (Fig. 5.9) and Lagacci. However, since these dialects (a) underwent pretonic degemination and (b) display word-final long vowels, contrasting with short ones when the word is uttered in isolation (e.g. [’du:] ‘two.m’ vs [’tu] ‘you.sg’, as reflected in Fig. 5.8 in the first and third boxes, respectively), this conclusion seems unwarranted to me. What the data do show is that CVL has become less strongly entrenched in the phonological system of those varieties than it has in Lizzanese, where the manifestation of the length contrast persists utterance-internally, as shown in Figs 5.5 and 5.6 above. This is in keeping with the observation that, in Lizzanese, the words displaying final stressed long vowels are more numerous, so that the VL contrast in word-final position has acquired a higher functional load.

5.5 Taking stock of the reconstructive evidence In this chapter I have shown that, in Northern Romance, CVL is beating a retreat on all fronts. Geographically it once embraced a larger territory, and then progressively shrank (as illustrated in }5.1), structurally, it could once occur in contexts in which it is

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

224

Dialect variation and comparative reconstruction

no longer found today. Both observations contradict the different explanations for the rise of secondary CVL reviewed in }4.1, for two reasons. First, these explanations, as pointed out at the beginning of this chapter, can in no way predict the recessiveness of CVL. For them, it has been brought into existence by several distinct processes (CoL, foot optimization, or the like) and is still encoded as synchronic lengthening in the phonological derivation (or successors thereto) in the relevant varieties; thus, in principle, it could just as well be an expanding feature, which I have shown it is not. Secondly, and more importantly, these competing treatments all devise explanations of the rise of CVL which are crucially dependent on its distribution as observed today, but are incompatible with prior stages reconstructible via comparison and the philological evidence: for example, one cannot possibly maintain that CVL in Milanese arose as a compensation for apocope, and conversely never arose whenever apocope did not apply (as in [’lyna] ‘moon’), given the evidence discussed firstly in }5.2, showing that apocope had not yet been completed by the time CVL is first documented (through rhymes) in 13th-century Old Milanese texts; and secondly in }5.1.3, where it was shown that lack of /y/-lowering in Eastern Lombard [’lyna] (not **[’løna]) implies that that vowel must have been long in Common Lombard (just as predicted, given the occurrence of OSL in Proto-Romance).87 None of the previous accounts discussed in Chapter 4 ever addressed the areal relationship of the different Northern Romance CVL systems (see Appendix, Maps 2–7) and its import for reconstruction. Also, none has focused on the stepwise retreat of CVL; indeed, a principled explanation thereof does not seem to be within reach of such analyses. Conversely, the explanation of the rise of CVL through the loss of the phonemic consonant gemination contrast ((32b)) is compatible with the comparative evidence discussed here. This change took place in all Northern Romance varieties, and the resulting contrast was later affected by subsequent changes which, however, were not able to totally obscure the original relatedness of the different CVL systems. This emerges especially in the many similarities in the distribution of the VL contrast observed between territorially discontinuous dialects: for instance, it can hardly be by chance that Central Friulian, on the one extreme, and Western Lombard, on the other, share—as shown in }5.1.3—with just one dialect spoken on the mountains north-northeast of Brescia (amid Eastern Lombard, which has no CVL) and with 87 This convergence of evidence from different sources, comparative and philological, not considered in any of the alternative accounts of Milanese VL, should be enough to dispel the objection raised by one anonymous reviewer, who commented on my elucidation of the relationship between the Bergamasco and the Milanese vowel systems (provided more synthetically in the book project), that ‘the whole argument rests on a largely aesthetic appeal to pattern’. Of course, what has been conducted here is reasoning on linguistic data based on qualitative, rather than quantitative, arguments. However, qualitative reasoning in historical linguistics rests on firm methodological ground, as I tried to show in }4.2. By those standards, I believe I have demonstrated beyond any reasonable doubt that the present reconstruction of the development of VL in, say, Milanese and the present account of its historical relationship with, say, the Eastern Lombard systems with no CVL is the right one, and that alternative proposals on the market are wrong. This has hardly anything to do with aesthetics.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Taking stock of the reconstructive evidence

225

another isolated variety on the eastern shore of the Garda Lake (surrounded by Venetan dialects, which again have no CVL) the restriction of CVL to the final stressed syllable as well as a number of finer details, such as the contrastive shortness of the vowel in the outcome of Latin medium: Milanese [’mεz], Central Friulian [’mjεtʃ], Lumezzanese [’mεh] ‘half.msg’. Those four areas are separated by dialects with no CVL at all, and neither parachuting nor polygenesis is conceivable: shared retention is the only reasonable explanation, which means, in the framework outlined in }5.4.1, retention of stage (98iii) along the scale modelling the gradual retreat of CVL. The same goes for the retention of CVL also in penultimate stressed syllables—contrary to the dialects just mentioned—in, say, Central Ligurian (}3.4.1.3), Lower Engadinian (}3.4.3.2), or some isolated Alpine Lombard dialects of Valtellina (cf. (47) above), surrounded nowadays by rather different systems. The rise and further development of CVL in all of these—and other—systems, with their different distributions of VL, has been shown to be best understood as parts of the same line of diachronic development. The most remote stage attainable through comparison of extant cognate dialects is one in which CVL has not yet been firmly established through degemination in the first place. While the phonological catastrophe which generated the Northern Romance CVL system (via degemination) must be conceived of as punctual, this switch has been prepared for gradually by a series of converging phonetic modifications in the V to C quantity relationships. This gradual shift has been reconstructed in }5.3, through comparison of selected dialects mostly from the periphery of the Northern Italo-Romance area. Focusing on the impact of stress position and word prosody (σ´ ]PW vs σ´ σ]PW) on vowel and consonant durations, it has been shown that these dialects bear witness to this day to the path by which contrastive gemination yielded to CVL. After becoming part of the phonological representation, as gemination was completely lost underlyingly, CVL survived in several Northern Romance dialects, but on the whole it has been losing ground: in }5.4.1 I showed that this retreat obeys a general phonetic constraint (rhythmical compensation), which is at work cross-linguistically and has an impact on the synchronic phonology of several Romance varieties, as exemplified in (102)–(104). In keeping with this constraint, the retreat of CVL went through the structural stages predicted by the scale in (98). Beyond the scope of that scale is what happened to stressed vowels in word-final position: this is a critical position for CVL, as, more generally, for heaviness and prominence, where special constraints—purely phonological in nature—have been shown to be at work in language after language. The absolute final position also provides important reconstructive evidence, given the fact that in word-final position VL cannot be dependent on context, so that when long vs short vowels occur there, one can diagnose contrastiveness with certainty, contrary to word-internal position, where contextual factors determining VL may be at play. The changes affecting word-final stressed vowels in different varieties were discussed in }5.4.2,

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

226

Dialect variation and comparative reconstruction

reconstructing the path which led Northern Romance dialects to acquire CVL wordfinally. This was shown to be a gradual process, leading through different steps— illustrated with one case study from the Tuscan–Emilian border in (113)—from a system in which final stressed vowels are categorically short (as in Tuscan) to a system in which this position hosts a VL contrast (as in 19th-century French, FrancoProvençal dialects, Genoese, and most of the further Northern Romance dialects reviewed in }3.4): the cohort of minimal pairs establishing the contrast in that position has been shown in (113) to have been fed stepwise by several changes. To put it in a nutshell, detailed inspection of dialect variation has afforded a better understanding of the gradualness of the overall change that, progressing along different yet converging lines, led to the transformation of the Proto-Romance into the Northern Romance phonological system.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

6 In lieu of a conclusion The most unmarked opening for a final chapter is through some formulation such as ‘The foregoing discussion has shown that . . . ’. Alternatively, if the final chapter were not declared as a concluding one, one could take the liberty of addressing at least some of the issues the reader might have legitimately expected to find discussed, but which were not, in the main body of the monograph. I shall adopt a mix of both strategies: the chapter provides a concise overview of (some of ) the main results of the study which is now approaching conclusion, complemented, however, by coverage of a few selected topics which have not so far been touched upon (although some may feel they should have been). This is one of the reasons why this chapter is not simply entitled ‘Conclusion’, the other being that it is unthinkable for the foregoing discussion to really have ‘concluded’ anything, on such complex and long-debated topics as those dealt with here. As stated in }1.3, the best I can hope to have achieved is to have added further weight to the burden of proof for competing analyses of the empirical data considered in this book. One notion the reader familiar with current literature on the phonology of VL may have missed, throughout the book, is that of the foot, a notion that has been playing a prominent role in discussions of suprasegmentals over the last few decades. In the foregoing pages I have analysed the Latin–Romance transition and a number of different kinds of distribution of VL across Romance without resorting to formalization of quantity in terms of foot structure, and it is an empirical question whether appealing to it would yield better descriptive results.1 As already explained in Chapter 5, n. 64, this depends on the fact that current assumptions in metrical stress theory—like strict foot binarity—prove disadvantageous, in my opinion, for a satisfactory account of the data analysed here.2 This has been seen in passing, when 1 This is actually no exemplum fictum: ‘I was struck to see no mention of the term foot’ featured among the comments by one anonymous reviewer on the book proposal. 2 The problem lies in those specific assumptions, not in prosodic analysis in terms of feet itself: indeed, I referred to feet, implicitly, in capitalizing on the notion ‘rhythmical compensation’ (cf. Ch. 5, n. 64), as well as explicitly, in mentioning the markedness scale constraining prosodic prominence on the final syllable in Ch. 5, (105). I believe this is a sound generalization, even if proposed within work some of whose analyses do not seem satisfactory to me.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

228

In lieu of a conclusion

discussing alternative explanations for the sound changes at issue here, which crucially refer to foot structure and elaborate on such current assumptions: cf. e.g. the discussion of analyses of VL in Milanese in terms of foot optimization (}4.1.1) or of Prieto’s (1994) ‘Foot Expansion in early Cremonese’ (}4.1.3). I have shown that these alternative analyses fare worse, as far as empirical scope and descriptive economy are concerned, and several similar cases could be added: I shall limit myself to just two examples, one concerning Latin, the other Italian. Formalization of the Latin stress rule in Chapter 1, (4) in terms of foot structure (e.g. Jacobs 1990: 83; Hayes 1995: 91–2) has become standard: this capitalizes crucially on the extrametricality of the final syllable. However, generalization of such an account beyond the province of the synchronic computation of the CL stress rule, appealing though it may appear at first glance, can lead to wrong predictions about Latin–Romance diachronic phonology. Consider Mester’s (1994: 37–43) elegant account of syncope in Latin. He posits two distinct rounds of syncope: ‘early syncope’ as exemplified in (1a), contrasting with ‘late syncope’ as exemplified in (1b): (1) a. *per+regō > pergō ‘proceed.inf’ (cf. perfect perrexī), porrigō > porgō ‘stretch out.inf’, jūrigō > iurgō ‘quarrel.inf’ b. ca˘lı˘dam ‘hot.fsg’ > caldam > Fr. chaude, It./Sard. kalda, Ro. caldă vı˘rı˘dem ‘green’ > virdem > Fr. vert, It. verde, Log. Sard. bilde, Ro. verde o˘cu˘lum ‘eye’ > oclum > Fr. œil, It. occhio, Log. Sard. ’o :ʒu, Ro. ochi Assuming right-to-left foot formation, and under the hypothesis that prehistoric and Classical Latin had a maximally bimoraic foot (‘quantitative trochee’), early syncope is explained as the erasure of a ‘trapped’ (i.e. unmetrified) syllable: (2) [ó]̄ ó̆ ̄

>

[ó]̄ ̄

This formalization of early syncope, which affords a unified analysis of this process with iambic and trochaic shortening, crucially depends on the heaviness of the first syllable involved: since the examples in (1b) do not match the input specified in (2), they are taken to be instances of a structurally independent, later wave of (Vulgar Latin) syncope, that is assumed to have applied when the system had already become quantity-insensitive. Within such a system, trisyllabic strings with a light initial syllable were also affected by vowel deletion, which could not have been the case, by hypothesis, in Archaic and Classical Latin. Note that Mester exemplifies this claim only with instances of post-stress syncope, occurring originally in words with antepenultimate stress like those reported in (1a), but claims in a footnote that prestress early syncope (as seen in the examples discussed e.g. in Burger 1928) ‘mostly also occur in trapping configurations’ (Mester 1994: 39, n. 45). Actually, closer inspection of the evidence (as argued in Loporcaro 2011c: 62–3) shows that early syncope applied in pre-literary and Archaic Latin, both after and

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion

229

before stress, to strings which could have been metrified perfectly, under Mester’s own assumptions, made explicit by bracketing in the following examples (drawn from Rix 1966; Niedermann 1931: 47; Leumann 1977: 95–9; Meiser 2010: 66):3 (3)

a. (călĕ)( făcĭ) > calfaciō ‘warm.inf’, *(sŭsĕ) > sūmō ‘assume/take up. inf’, *(dŏkĭ) > doctus ‘learned’, *(uĭrŏ) > uirtūs ‘courage’ b. monstrum ‘portent’ < *mŏ(nes) (cf. mŏneō ‘warn.1sg’), fenstra < fĕ(nes) ‘window’ (both documented in Plautus—cf. Meiser 2010: 66; Morani 2000: 179—whereby only the non-syncopated form survives into Romance)

Syncope with loss of what could have been, under Mester’s assumptions, the stressed vowel of a perfectly well-formed foot took place in bă(lĭnĕ) ‘bath’, which was nonetheless syncopated to balneum from Varro (116–27 bc) and Cicero onwards (cf. also bal(i)neātor ‘bather’ in Plautus: Väänänen 1981: 43). Examples like those in (3b), where syncope affects the nucleus of a heavy syllable, fit Mester’s picture even worse, since under trochaic moraicity the syllable which lost its nuclear vowel should have been footed by itself as *mŏ(nes), fĕ(nes) (parallel e.g. to metrically wellformed ă(mī) ‘friend’ (cf. Hayes 1995: 92). For e.g. monstrum, sūmō, or doctus, these are the only attested forms, which have hence ousted their non-syncopated counterparts completely. For other words, like fen(e)stra, vacillation is documented: (4)

columen, both Archaic and Classical, alongside more widespread culmen (attested since the 1st c. bc, in Caesar, Livy, Virgil etc.); officīna ‘workshop’ < opificīna (Pl. Mil. 880); frigdaria ‘refreshing’ (cf. frīgidus ‘cold’), caldarium ‘room for hot baths’ (cf. calidus ‘warm’), in Lucilius (2nd c. bc).

Thus, the evidence shows clearly that a host of examples in which syncope is not motivated by prosodic trapping are much too early to be dubbed as instances of ‘Late Latin syncope’ (as those in (1b)). Nor can syncopated variants, in Archaic and Classical Latin, all be regarded as necessarily ‘low’ in sociolinguistic terms, which would be an obvious escape route:4 the emperor Augustus reportedly dubbed as pedantic the pronunciation calidus (Quint. 1.6.19) so that by the late 1st century bc, syncopated caldus could not be qualified as ‘low’. Also, several of those examples show syncope under stress, and so cannot be counted as evidence for an early

3 Mester (1994: 38) is misled by his sources here, as he quotes Lindsay (1894: 173) for the (incorrect) generalization that ‘post-tonic syncope under the penultimate accent law “[ . . . ] seems, during the Republic and early Empire, to occur only when the accented vowel is long” ’. This is at best a quantitative tendency, but no absolute ban on the occurrence of syncope in other configurations, as the examples in (3) and (4) demonstrate. 4 This is suggested by Mester (1994: 39, 41) himself, as he says that late syncope of type (1b), as in uălĭdē > ualdē, ‘began in Vulgar Latin and became fully general in Late Latin and Romance’.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

230

In lieu of a conclusion

establishment of a quantity-insensitive initial foot preceding main stress (cf. Lahiri et al. 1999: 387–8, 394–5), long before VL ceased to be contrastive altogether. In conclusion, claims of a sharp distinction between early and late syncope (à la Mester 1994) appear unwarranted. More accurately, Pensado Ruiz (1984: 234), for example, observes that one cannot speak of a ‘start’ to Romance syncope, but only of generalization of a pre-existing phenomenon, given the non-discontinuity of the conditions under which syncope had occurred ever since Archaic Latin. Since the distinction of the alleged two rounds in (1a,b) has proved ill-founded, so has explanation of this alleged distinction in terms of sensitivity vs non-sensitivity to foot structure. A similar point has been made in }5.4.1 as regards the modelling of Italian OSL and diphthongization in terms of foot structure and extrametricality. The idea that extrametricality of the final syllable carried over from Latin to Italian, and that this explains why OSL and diphthongization allegedly occur in the penult but are excluded from the antepenult, leads to a patently mistaken explanation (Prince 1990: 374–5; Hayes 1995: 148) which can be put forward, in the first place, only at the cost of misrepresenting the empirical data (Calabrese 1984–5: 19; 2000: 69). The truth of the matter is as follows. Let us start by schematizing the facts on OSL reviewed in }5.4.1 (based on Bertinetto 1981; Farnetani and Kori 1986: 35–8; 1990: 54–7; D’Imperio and Rosenthall 1999: 3–10; Dell’Aglio et al. 2002): (5)

σ structure

PW structure

a. open b. closed

i. σ́σ ]pw

ii. σ́σσ ]pw

[ˈskaːla] ‘ladder’ [ˈskalda] ‘heat.3sg’

[ˈskaˑlala] ‘climb.imp.2sg it(f)’ [ˈskaldala] ‘heat.imp.2sg it(f)’

In prepausal position, stressed vowels are lenghtened in (5a), though the phonetic effect of lengthening is consistently less pronounced in proparoxytones ((5a.ii)). The explanation for the facts defended here is that an allophonic rule of OSL is at work in modern Standard Italian ((5a)), whose surface effects are then minimized when counteracted by phonetic constraints such as rhythmical compensation (seen at work in (5a.ii) vs (5a.i)). Evidence from sound change confirms that this is correct, as shown in (6) (based on the data discussed in }5.4):

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion (6)

σ structure

231

PW structure i. σ́σ ]pw

ii. σ́σσ ]pw

a. open

[ˈtjεːne] ‘keep.3sg’

[ˈtjεˑpido] [ˈpεˑkora]

‘lukewarm’/ ‘sheep’

b. closed

[ˈtεŋgo] ‘keep.1sg’

[ˈtεŋgono]

‘keep.3pl’

Diphthongization of *[ε:], previously lengthened via OSL, applied categorically in (6a.i), but only desultorily, in a lexically idiosyncratic manner, in (6a.ii). This evidence from sound change is especially welcome if one considers that the phonetic data themselves are often regarded as ambiguous. As seen in Fig. 5.7, the steps between fate and fàtele, on the one hand, and fàtele and fatte, on the other, are of unequal magnitude in D’Imperio and Rosenthall’s (1999) results, and the statistics run by the authors even show the latter difference to be non-significant: they therefore conclude that there is no genuine difference between (5b.i) and (5a.ii), and this non-distinction is reflected in their OT account of OSL, in which the latter is shown not to apply in proparoxytones (cf. D’Imperio and Rosenthall 1999: 15 and Chapter 5, n. 72). However, as mentioned in }5.4.2, other studies have come up with different results: for Farnetani and Kori (1986: 35–8), all the contrasts between [’ska:la], [’skaˑlala], and [’skalda] are statistically significant. These different results square well with the diachronic data in (6), and support the hypothesis that the phonetic basis (lengthening) for diphthongization in (6a.ii) was shakier, but still existed, otherwise diphthongization would not have applied at all. Clearly, assuming flatly that the final syllable in (5a.ii), (6a.ii) is extrametrical, in compliance with strict foot binarity (Hayes 1995: 58; Kager 1995: 401), is of little help here.5 What the data show is that rhythmical compensation is best understood under a simpler and more traditional view of the foot as the string consisting of a stressed syllable plus the following non-stressed ones (cf. e.g. Bertinetto 2004: 588). Only if all the syllables in, say, Italian [’pεˑkora] or [’skaˑlala] are part of one and the same foot is it possible to analyse rhythmical compensation as foot-related. Adopting the terminology used in Ritt’s (1994: 75) account of Middle English OSL, the problem with accounts of Italian OSL and diphthongization which capitalize on

5 The assumption of extrametricality is a widespread move in generative treatments of stress in Romance, often put to work to accommodate the otherwise recalcitrant behaviour of etymological proparoxytones (cf. discussion in Loporcaro 2011c: 89). Extrametricality, thus, is not a general property of the system but is resorted to only in some cases, in order to adjust representations when needed to accommodate otherwise deviant data. Though for Latin extrametricality is assumed much more generally, for that language too it is not a general built-in property of the system, as shown e.g. by its suspension in words undergoing iambic shortening such as hŏmō > (hŏmŏ) (cf. Mester 1994: 16, 23).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

232

In lieu of a conclusion

extrametricality and binary foot structure is that they formalize the impact of prosodic factors on lengthening and diphthongization in strictly digital, rather than scalar, terms. This is the reason why I avoided framing the generalizations on changes in the VL system from Latin to the Romance varieties considered here in terms of foot structure under current assumptions, as now explained in the discussion of Latin syncope and Tuscan diphthongization: the foregoing discussion has shown that both scalar and (quasi-)digital factors are relevant for effective modelling of those changes. Another fact which emerged clearly from the previous chapters is that VL has been linked tightly to stress ever since pre-literary and Archaic Latin, where contrastiveness of the long/short distinction was progressively eroded by a conspiracy of changes shortening formerly long vowels in unstressed positions (}1.2). This connection has been entertained as a candidate for a universal of phonological structure since, as Blevins (2004: 9) puts it, ‘[t]here are languages with long vowels and short vowels where all long vowels must be stressed, but there are no languages with long and short vowels where all short vowels must be stressed’. This generalization is confirmed by the Late Latin facts (Ch. 2) as well as by all the Romance secondary CVL systems reviewed here, none of which has CVL limited to unstressed position: though some varieties do have it also in that context (Old French and Central Ligurian: cf. Chapter 5, n. 1 and Chapter 3, (37–39) respectively), in the vast majority of them CVL only occurs under stress.6 The reason for this distribution in Romance is a historical one, since Romance CVL is a diachronic successor of Late Latin OSL (Chapter 2), which affected stressed syllables and started to apply, in the reconstruction advocated here, by the end of the Western Empire (during the 4th century in Italy, earlier in Africa, perhaps slightly later in Gaul, if lengthening in African pronunciation still struck Consentius’ ears as non-standard, in early 5th-century Narbonne, southern Gaul; cf. }2.1, (4)). It is a universal phonetic tendency for vowels to last longer, other things being equal, under stress than in unstressed position, and in open syllables than before tautosyllabic consonants (cf. the references cited above, while commenting on }4.1.2, (14), and in Chapter 5, n. 66)). Further, we have seen that all Romance languages, including those nowadays lacking an OSL rule, provide evidence for the effectiveness of these phonetic constraints (see e.g. Renwick 2012 on Romanian, cited in }5.4.1, (99), and Ortega-Llebaria and Prieto 2007: 162 on Spanish, cited in }3.2). Yet it is a fact that in a language like modern Standard Italian the magnitude of the lengthening of stressed vowels in open syllables is much greater than in languages where no OSL phonological 6

As Y.C. Morin points out to me, a further exception to this regularity is observed in Occitan varieties where coda-/s/ was deleted resulting in CoL of the preceding vowel: thus, in the patois of Vinzelles (as described by Dauzat 1897: 62, 91; see also Morin 2000), one finds e.g. [’pu:za:] ‘pause.pl’ < pausas, [tsa:’te] ‘castle’ < castellum; in that of Oulx and Pragelà, [la: ’vat∫a:] ‘the cows’ < (il)las vaccas (Sibille 2007: 325). As in the Ligurian and Old French examples already mentioned above, in this case as well the diachronic source for these long vowels is distinct from OSL.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion

233

rule is at work. As I have argued, modern Standard Italian (and the Italian dialects of Tuscany and the centre–south, as well as Sardinian) may still be considered to mirror the Proto-Romance syntagmatic distribution of quantity. Given what has been said so far, then, it is natural to conceive the rise of Late Latin OSL as an instance of stabilization, i.e. ‘a process of regular categorical change that creates a new phonological counterpart for an existing phonetic rule’ (Bermúdez-Otero 2007: 506). This account assumes a quite traditional architecture of phonology, of the kind shown in Chapter 3, (7). In this specific case, the phonetic rules (or implementation constraints) providing the preconditions for the rise of Late Latin OSL were those just mentioned; in addition, it is possible (as argued in }3.5) that another universal phonetic constraint, prepausal (or pre-boundary) lengthening, may have played a role, favouring an earlier superseding of VL contrasts in prepausal position. On the whole, the present study on change in VL from Latin to Romance can be considered as evidence in support of modular ‘feed-forward’ models of phonology, drawing a distinction between phonology and phonetics. Interaction between phonetics and phonology played a major role, as an explanatory factor, in Chapter 5, where the diversity of structural conditions under which CVL nowadays occurs, especially in the southern part of Northern Romance (spanning Romance and Northern Italo-Romance dialects), has been shown to be considerable; and comparison between those widely diverging systems has been capitalized on for the reconstruction of the changes which led to the rise of secondary CVL and, subsequently, to present-day differences in CVL systems. Comparative reconstruction, and especially the inspection of the areal distribution of the different CVL systems over the Northern (Italo-)Romance territory, has shown that a joint account of the diachronic development of those systems is desirable, since otherwise one would have to assume extensive polygenesis in a way that would strain credulity. This point has already been made in }5.5, showing that the coincidences in the arrangement of VL between, say, Central Friulian, Western Lombard, and some isolated dialects spoken between eastern Lombardy and the western fringe of Veneto, on the one hand, or the striking distributional parallelisms between Lower Engadinian, Emilian, and some isolated dialects of Valtellina, on the other, can hardly be due to mere chance. The same goes for the similarities—e.g. in the preservation of phonetic consonant length— observed between marginal dialects spoken at the northern and southern fringes of the Italo-Romance space. Instead of dismissing them as coincidental, I have capitalized on them—in }5.3—in order to gain insights into the early steps of the process which brought about the rise of secondary CVL, showing that this can be effectively modelled as a phonetically gradual shift in distinctiveness from consonant to vowel duration. Once established throughout Northern Romance territory from northern France to the Apennines, then (with the possible exception of Central and Lagoonal Veneto and perhaps parts of Occitan, on which see }}5.1.4 and 3.4.2.2 respectively), CVL became increasingly differentiated from area to area. It is not only the conservative dialects just mentioned which are responsible for this differentiation: rather, the linguistic chart of

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

234

In lieu of a conclusion

VL—as evident from Maps 2–7 in the Appendix—is nowadays very colourful, displaying side by side dialects lacking CVL and dialects which contrast long and short vowels, and, among the latter, dialects which do so under subtly different conditions. An inventory of these different systems, with their different distributions of CVL, showed (}5.4.1) that they could be lined up along a scale—that in (98)—which has a straightforward phonetic motivation. Accordingly, the major factor responsible for the gradual retreat of CVL in Northern Romance must have been the increasing difficulty of maintaining length as the number of unstressed syllables following the stressed one increased. Like all phonetic factors, by definition, this acts synchronically, whereas for phonological properties this is not to be taken for granted. Trivial though it may appear, one has to recall that, under a classical feed-forward model of phonology like the one assumed here, an allophonic phonological rule is part of the synchronic phonological derivation, whereas contrasts which have arisen through sound change and have made it into the underlying representation via restructuring are lexicalized.7 This is an issue I had to dwell on in Chapter 4, since alternative current accounts of CVL in Northern Romance varieties mostly analyse VL as rules (or as the effect of output candidate selection through constraint ranking) in the present-day phonology of the languages/ dialects at issue, not as encoded in the input, the way I have argued they should.8 At this point, after having introduced the scale in Chapter 5, (98), it is possible to provide some final illustrations of why this approach yields a better account of the Romance data than the alternatives previously available. Consider the Engadinian dialect of Scuol (cf. Gredig 2000). This variety, as shown in Chapter 3, (82)–(85), displays CVL in the word-final syllable—both closed and open ((7a-b))—as well as in paroxytones ((7c)): (7)

a. [’tɕa:r]

‘dear’

6¼

b. [’ma:]

‘never’

[ma]

‘but’

c. [’fO:sa]

‘false.fsg’

[’fOsa]

‘grave’

b. [’tɕar]

‘wagon’

This distribution has come into being through a series of sound changes, which have fed the contrast in the different positions where it nowadays occurs. One of those processes—which arguably represents the first diachronic layer, by analogy with what has been shown for Old French in Chapter 3, (67i)—is PRom OSL, to which the long vowels in words such as those in (8) (repeated from Chapter 3, (85)) can be traced back in a straight line:

7 In between, there is a gradual transition, with synchronic rules (morphophonological rules and allomorphic-morphological rules in Dressler’s 1985 early systematic account of morphonology) in which morphological (and lexical) factors increasingly play a role. 8 The argument in favour of maintaining this is elementary and, again, traditional: contrastiveness, or distinctiveness, which I use as synonymous expressions denoting the property of distinguishing at the surface words of different meaning. (I am not adopting e.g. the differentiation between ‘contrast’ and ‘distinctiveness’ drawn by Flemming 2004.)

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion (8) a. [’dy:r] [’no:f] [’la:t] [’tʃa:r]

b. [’dy:ra] [’no:va] [’la:da] [’tʃa:ra]

‘hard.msg/fsg’ ‘new.msg/fsg’ ‘large.msg/fsg’ ‘clear.msg/fsg’

235

< du¯rum/-am < no˘vum/-am < la¯tum/-am < cla¯rum/-am

Occurrence of the long vowels in (8b) shows that the phonological system, in this area, resisted the pressure exerted by the phonetic tendency towards rhythmical compensation mirrored in the scale in Chapter 5, (98): thus, CVL was not eradicated from the penult. Consider however the minimal pair in (9): (9)

a. [’ku:la] ‘ball’ < Germ. kugel DRG 4.354 b. [’kula]

‘flow.3sg’ < co¯lat

The short stressed vowel in [’kula] should be long, just like those in (8b), since it goes back to a stressed vowel in a PRom open syllable. The fact that it is short— whereas the same vowel in the same root stayed long in apocopated [’ku:l] ‘sieve’ < co¯lum (DRG 4.340)—shows that rhythmical compensation had some impact on paroxytones in this variety too, though by no means to the same extent as in the systems which banned CVL from the penult altogether. In spite of (arguable) shortening in [’kula] ‘flow.3sg’, length in (8b) remained the dominant option in this Engadinian variety, and in this specific case this word entered a VL contrast with [’ku:la] ‘ball’, a German loanword whose long vowel is not inherited from Late Latin/ Proto-Romance. Yet the system provides for this structural possibility—i.e. the possibility for CVL to occur in proparoxytones—which was exploited, for this specific pair of lexical items, in this dialect.9 The account of CVL provided here, appealing to the distinction and the interplay between lexicalized CVL (as a diachronic successor of PRom OSL) and synchronic phonetic constraints on vowel duration threatening its preservation, is as flexible as is required to cope with these facts: on the contrary, assuming that CVL in Romansh derives from CoL for either the apocope of the final vowel or the devoicing of the final consonant (cf. Haiman and Benincà 1992: 42; Videsott 2001) would manifestly preclude a sensible explanation of the data. Rhythmical compensation can impact on the system to a greater extent than just exemplified for Engadinian. Milanese, which today has no minimal pairs in the contexts exemplified in (7c) and (9), illustrates this impact, leading to a categorical distribution in this prosodic environment (paroxytones).10 Central Friulian, in the literature discussed in }4.1.2, is usually regarded as identical to Milanese in this respect, though the data already introduced in }3.4.1.5, (62), suggest that this view is oversimplified. On the one hand, it is true that Central Friulian, just like Milanese and unlike Engadinian (or Ligurian, Emilian, etc.), nowadays completely lacks long In other varieties of Romansh, the same German loan has a short stressed vowel: [’kula] (DRG 4.354). Exemplification is here limited to steps (98ii,iii) along the scale discussed in }5.4.1, but it could be repeated for each one of them: cf. discussion on the rarity, but crucially not non-occurrence, of long vowels in proparoxytones in Ligurian and Emilian dialects in Ch. 5, (94)–(96). 9

10

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

236

In lieu of a conclusion

vowels in words which go back to PRom etyma of ’CVCV shape and did not undergo apocope. In other words, in paradigms where Engadinian has a contrastively long vowel in both alternants ((8a,b)), Friulian has a VL alternation: (10) a. [’vi:f] [’fre:t] [’la:t] [’lo:f] [’kru:t]

b. [’vive] [’frede] [’lade] [’love] [’krude]

‘alive.msg/fsg’ ‘cold.msg/fsg’ ‘gone.msg/fsg’ ‘wolf/she-wolf ’ ‘raw.msg/fsg’

Given what has been said so far, the occurrence of short stressed vowels in (10b) can be confidently ascribed to rhythmical compensation. Under the scenario put forward here, CVL arose, in Friulian as elsewhere in Northern Romance, when degemination equated the environments in, say, vallem ‘valley’ > *[’val:e] vs valet ‘be worth/cost.3sg’ > *[’va:le], or lacte ‘milk’ > *[’lat:e] vs *amb(u)latum ‘gone’ > *[’la:do], which subsequently became, through apocope and final devoicing, Friulian [’val] vs [’va:l] and [’lat] vs [’la:t] respectively.11 In a subsequent stage, CVL must have withdrawn from the penult, as in Milanese, yielding to the phonetic pressure portrayed in (98), which is and was of course at work in Friulian too, as evidenced by the results of the experimental study on Friulian reported in Chapter 5, (100), illustrating the effects of rhythmical compensation. However, the fact that rhythmical compensation eroded, in paroxytonic words, the specific VL contrasts which had arisen from the phonologization of the output of OSL does not in itself imply that no VL contrast at all is allowed to occur in the penult. And indeed, long vowels created by other sound changes—like CoL for the loss of a following consonant—were retained in Friulian, in that position as well, giving rise to minimal (or near-minimal) pairs (cf. R. Castellani 1980: 91; Faggin 1985: 378): (11)

a. [’spa:la] ‘shoulder’ < spatulam 6¼ [’ma:ri] ‘mother’ < matrem [’ne:re] ‘black.fsg’ < nigram

b. [’pala] ‘spade’ < palam [’stala] ‘stable’ < *stalla [kon’trari] ‘contrary’ (a learnèd word; cf. It. contrario) [’sere/’sεre] ‘evening’ < seram

The stressed vowel in [’sere/’sεre] is notated with either one or the other quality in Pirona (1871: 378), and in general for mid vowels there is disagreement (as reported in Chapter 3, n. 60) as to whether Central Friulian has length contrasts for both high-mid and low-mid stressed vowels. However, for /a/ the phonemic contrast is straightforward and undisputable. For Carnico (Northern Friulian), R. Castellani (1980: 89) 11 The starred forms, except the last one (which shows the application of WRom. lenition, -t- > [d], in a lexical item not occurring in Tuscan), are identical to their modern Standard Italian counterparts. This is of course not by chance, given what has been argued here about the conservative character of Tuscan with respect to the syntagmatic quantity relationships inherited from Proto-Romance.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion

237

reports that such minimal pairs are even popularly exploited for tongue-twisters, such ˆ = [V:]): as the following (which I report in the author’s orthography, where V (12) a. Un cjăr di cjăr di cjâra cjăra ‘a wagon of expensive goat meat’ b. La sêla su la sĕla ‘the bucket on the stool’ c. Una zumièla di mŏris mōris ‘a handful of black blackberries’ Long vowels in (12a–c) arose via CoL, like those in (11a) ([c]âra ‘goat’ < *[’kavra] < capram, sêla ‘bucket’ < *[’sedla] < sit(u)lam), or via vowel coalescence (mōris ‘black. fpl’ < mauras). Apparently, although the impact of rhythmical compensation was sufficient to eradicate long vowels from the first layer, the subsequent changes mirrored in (11a), (12) reintroduced CVL—which had persisted throughout in the final syllable— in the penult as well. Again, the alternative proposals reviewed in Chapter 4 would be at a loss in facing these Friulian data: they all assume that long vowels in (10a) are derived synchronically by rule (or, in Iosad’s 2012 recent OT account, are a property of the winning output candidates, not encoded in the corresponding inputs) because of the devoicing of the following obstruents (or the loss of the final vowel: cf. Prieto 1992; Repetti 1992). In addition to the problems of descriptive economy discussed in Chapter 4, (19) and (20), these approaches are unable to analyse long vowels in (10a), on the one hand, and in (11a), on the other, as surface manifestations of one and the same series of underlying entities.12 Conversely, in the analysis proposed here, CVL, first arising via degemination, has become a property of underlying representation which may be retained or lost in different contexts. The fact that CVL may be fed by later sound changes, which—seen in (11)—even restore the contrast itself in an environment from which it had been previously lost, does not force us to complicate unnecessarily the account of Friulian VL by positing for example ‘different’ sorts of VL, derived via entirely distinct formal devices. Recognition of the possibility for CVL to have multiple causes brings me to the very final point. When discussing Engadinian or Friulian, like any other of the Northern Romance varieties considered here, I have claimed that it is possible to spot an original layer of long vowels: in other words, a central claim in this book— following a line of research which includes Haudricourt and Juilland (1949), Weinrich (1958), and Morin (2003)—has been that contrastive VL was acquired throughout this subdivision of Romance through one shared sound change,

12 As pointed out in Ch. 3, n. 64, Vanelli (1998: 70–71, n. 4) plays down CVL in (11a) and (12), denying that there are any minimal pairs in this context, a claim that is proven false by the data in (11)–(12).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

238

In lieu of a conclusion

degemination, with subsequent phonologization of the lengthened output of the PRom OSL rule.13 Many other changes have been reviewed in the preceding chapters—like CoL or vowel coalescence in Friulian, now mentioned while commenting on (11a) and (12)—which however are not so widespread as the first trigger of CVL, and have been argued to have applied when the dialects already possessed the contrastively long vowels of the first layer. Obviously, this need not universally be the case, as it may well be that the evidence available does not afford a unitary explanation of the kind proposed here. This can be illustrated by comparing the case of Germanic. So far, data from Germanic have been mentioned in two different ways. In Chapters 1 and 2, evidence from Germanic languages—in the form of comparison with either loans or cognate words—was mentioned among the sources of established knowledge about Latin CVL. In subsequent chapters, on the other hand, specific analyses of VL in Germanic languages (like that of Page 2007 in }3.2, King 1988 in Ch. 4, n. 19, or Ritt 1994 cited earlier in this chapter) were mentioned in order to compare them with some specific aspects of the present argument. Comparison with Germanic, however, provides a more serious challenge to the account of the historical development of CVL from Latin to the Romance languages proposed here. The overall picture, first of all, bears some striking similarities. In Germanic, too, there are data (in particular from West Germanic) which have been interpreted assuming the operation of an OSL—though the latter intervened, unlike in Proto-Romance, in a system with inherited CVL: (13)

a.

MHG

hăben nĕmen

> >

NHG NHG

b.

ME

măken hŏpen

> >

māken hōpen

hāben nēmen

‘have.inf’ ‘take.inf’

(Page 2007: 337) (Ritt 1994: 1)

For West Germanic as well, the analysis of these data is not uncontroversial: for the ME facts in (13b), while OSL has been assumed by many authors (e.g. Prokosch 1933; 1939; Hogg 1996; Lahiri and Dresher 1999; McMahon 2000: 151ff.), other scholars (cf. Minkova 1982; Lass 1985; Hayes 1989) have assumed that non-high long vowels in ME words like tāle arose via CoL for the loss of final [@]. The availability of those competing analyses is reminiscent of what has been seen while comparing diverging views on the rise of CVL in Milanese or Friulian, in }}4.1.1 and 4.1.2. In addition, it is clear that OSL is not the only source of long vowels in West Germanic. German, for instance, has lengthened stressed vowels not only via OSL in polysyllabic words but also in monosyllables: e.g. MHG [’glas] > [’gla:s] ‘glass’, [’wεk] > [’ve:k] ‘way’ (cf. e.g. Seiler 2009: 230), though in many such cases other wordforms 13

Though structurally homogeneous, the change cannot have occurred at exactly the same time all over the area involved, as attested by the fact that some peripheral dialects—as seen in }5.3—are still at the rearguard of change to this day.

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

In lieu of a conclusion

239

in the same paradigms did have an open syllable (e.g. MHG wec, pl. we.ge), so that lengthening might have arisen regularly here to then spread analogically to monosyllabic forms. Other lengthening processes were triggered by the segmental environment in which the vowel occurred: Middle Dutch has long vowels in nāme (Modern Dutch naam) which, in the absence of documentation for older stages of the language, can be argued to have undergone lengthening based on comparison with OE nama; alongside this OSL, however, we find lengthening before clusters of /r/ + dental consonant (e.g. Dutch paard ‘horse’, staart ‘tail’; see Lahiri and Dresher 1999: 683). Likewise, in English too, originally short vowels were lengthened before clusters of sonorant + stop (with /r/ as C1, also in case C2 was a voiced fricative or a nasal): e.g. cild > cīld, word > wōrd (Ritt 1994: 82–93). Germanic also resembles Romance in displaying not only changes conspiring to create the same output (long vowels) but also changes that are in contradiction to each other: some dialects of German, in fact, have undergone open syllable shortening, as exemplified in Bernese German [’I ̯æ.rIg̊ ] ‘yearly’, [’hy.z̊ @r] ‘house.pl’ < MHG [’I ̯æ:.rek], [’hy:.s@r] (Seiler 2005: 477). This is a very intricate picture, and though several scholars have proposed joint explanations for at least some of the changes just mentioned (cf. e.g. Prokosch 1933; 1939; Lahiri and Dresher 1999, who put forward a unified account of OSL in West Germanic, albeit along different lines), according to others this body of evidence seems to point just to a ‘contingent overlap between quite independent processes of lengthening’ below which no unity can be spotted. This is an argument an anonymous referee brought to bear, claiming that comparison with Germanic suggests that one ‘has every reason to be sceptical’ towards the unified account proposed here for the history of Romance VL. Though it is clear that in Northern Romance, just as in West Germanic, CVL was fed by distinct and subsequent changes,14 the evidence gathered and sifted in the present study has demonstrated that this objection is not justified. The many later changes which affected VL in Northern Romance varieties were not able to obscure completely the oldest one, which can still be traced through comparative reconstruction, viz. the rise of secondary CVL from the loss of Late Latin OSL—whose rise in turn is not only reconstructible via comparison (Chapter 5) but also emerges from inspection of the philological record (Chapter 2). In other words, the evidence shows that there is a straight line of diachronic development, leading from Latin CVL, through its demise via the rise of an OSL allophonic rule in Late Latin, to the establishment of Northern Romance secondary CVL. 14 Several of these changes have been discussed in the preceding pages. The inventory is the usual one, featuring the rise of long vowels through vowel coalescence (cf. (12c) above and Ch. 3, (39a), (42a), (67iii); CoL for the loss of an adjacent consonant (cf. (11) and (12a,b) above as well as (39b), Ch. 3, (67ii), or of a following vowel Ch. 3, (67v)); and the influence of a following consonant (see the lengthening induced by a following voiced obstruents in the rural Pavese dialects in Ch. 4, (22c), (23c), and (24c)) or consonant cluster (cf. the lengthening before sonorant + obstruent clusters in Cremonese, discussed in Ch. 3, n. 38).

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Appendix: Language and dialect maps

Spanish Catalan

French Occitan Francoprovençal Provençal Italo-Romance Sardinian Rhaeto-Romance Galician

Veglia/Krk

Daco-Romanian Sansepolcro

Senigallia Macerata

Popoli

Bielsa Rome

Vasto Palmoli Tufillo Bisceglie Agnone Cerignola Bito nto Altamura Nap les Matera Castelmezzano Pozzuoli Stiglian o Senise Trecch ina Belvedere Marittimo

0 0

Miles

Kilometres

300 400

Map 1 The Romance Languages. The map shows the places whose local varieties are referred to in the book, except for those lying in the Northern Romance territory (for which see Maps 2–7)

Piazza Armerina

Cersosimo San Giovanni in Fiore

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Portuguese

CVL only in oxytones CVL only in oxytones, with restrictions (only certain vowels and/or only certain lexemes and/or variably neutralized) No CVL CVL also in paroxytones No information on CVL Upper limit of Romance dialect area Southern border of Northern (and Western) Romània

WALLONIA

E LV

AN

SU

N IA D

Scuol

EN Celerina/Schlarigna Ambialet

Montana River R hô ne

Certoux

Bravuogn/Bergün

BÉARN

Val d’Illiez

Vinzelles Vaux (Ain)

BURGUNDY FRENCH

IN

Sent

A

TS

Flem N VA L E Trun RS Tavetsch Andeer

Domat

G

SU

FRANCOPROVENÇAL

OÏL

Trin

SURMEIRAN

River Rhin e

Mesnil- Tenneville Martinsart

FRANCOPROVENÇAL

Le-Pont-deBeauvoisin

Bramans

PROVENÇAL Jausiers Barcelonnette

Carrara Massa Fano Senigallia

MAP 2 NORTHERN ROMANCE. THE MAP SHOWS THE PLACES WHOSE LOCAL VERNACULARS ARE REFERRED

TO IN THE BOOK, EXCEPT FOR THOSE OF NORTHERN ITALY

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Hauteville

56 113 85

47

126 92 90

7 25 21

4

103 36

34

105

65

101

TAIN74 DÔ L VA

60

86 16

73

13

111

68

115

80

D

M

O

N

E

MB

70

79 82

FRIULIAN

A

10 42

99

41

97 19

116

77

VENETAN

59

D

91

9

114 123

125 69

96 37

87

58

110 43

EMILIAN

62

R

20

52 76

L

IG

U

66

18

1

121

12

AR

122

107

5

39

112

89 128

31 109

108 119 127

P

IE

46

LO

63

118

61

S TE

23

55

8

22

78

71

95

100

29

81

117

53 6

67 24

17 50

27 88

L

64 2

51 124

3

IA

N

45

35

14

120 38

32 98 104

72

11

RO MA G N O L O

84 57

26 33

30 28 54 106

83 44 40

48

TUSCAN

102

49

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

I N75 D93 94

15

CVL only in oxytones CVL only in oxytones, with restrictions (only certain vowels and/or only certain lexemes and/or variably neutralized) No CVL Agordo Airolo Albenga Antronapiana Aprica Arabba Augio Balerna Barcon Belluno Benedello Bergamo Bienate Bologna Bonifacio Borgomanero Bormio Breil Brescia Cairo Montenotte Calasetta Campertogno Campo (Val Tàrtano) Canazei Carloforte Carrara Casale Corte Cerro Castello di Sambuca Castionetto Cavanna Chiuro Collagna

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

Colognora Comano Compiano Corticiasca Cremona Crocette Darfo Boario Terme Fano Fanzolo Feltre Fiorenzuola d’Arda Galeata Genoa Germanasca Valley Giornico Gombitelli Gorizia Grosio Imperia La Brigue/Briga La Villa (Val Badia) Lagacci Lecco Livigno Lizzano in Belvedere Lungavilla Malcesine Mendrisio Milan Miogliola Moena Müstair Valley

Map 3 Vowel length in Northern Italo-Romance

65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96

Novate Mezzola Ormea Ortisei Oulx Padua Pavia Pesariis Piandelagotti Pianezza Borgosesia Piedicavallo Pieve di Marebbe Pigna Pordenone Poschiavo Pradumbli Pragelà Premana Preone Rimini Riolunato Rossa (Calanca) Rossa (Valsesia) Rovigo Saint Nicolas Sairano Samòlaco San Daniele San Giacomo Filippo San Martino (Val Badia) San Vigilio San Vito di Cadore Sant’Angelo Lodigiano

97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128

Sant’Apollonio di Lumezzane Sassalbo Schilpario Selva di Cadore Selveglio di Riva Valdobbia Senigallia Signôra Valcolla Sillano Soglio Stabiazzoni Tarcento Tartano Teglio Tortona Trausella Travacò Siccomario Trepalle Treviso Turin Udine Upper Comelico Val Cavallina Val d’Ambria Val di Taro Val Rendena Valle di Zoldo Venice Ventimiglia Verona Villa di Chiavenna Vilminore di Scalve Voghera

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

CVL also in paroxytones CVL in oxytones, marginally also in paroxytones (Friulian) Transition between contrastive consonant gemination and CVL No information on CVL

Müstair Valley

La Villa (Val Badia)

Giornico

Trepalle

Rossa (Calanca) Augio San Giacomo Filippo Samòlaco

Soglio

Campo

Comano Corticiasca

Moena Poschiavo

Villa di Chiavenna Novate Mezzola

Signôra Valcolla

Grosio Chiuro Teglio

Castionetto Tartano Val d’Ambria

Premana

Casale Corte Cerro

Arabba

Canazei

Bormio

Aprica Val Rendena Schilpario

Feltre

Vilminore di Scalve Dar fo Boario Term e

Mendrisio Balerna

Lecco

WESTE RN L OMB ARD Bienate Milan

Pavia Sairano

Travacò Siccomario Lungavilla

Voghera Tortona

Map 4 Remnants of CVL in Eastern Lombard

Bergam o

Val Cavallina

Val Sabbia

Malcesine

Sant’Apollonio di Lumezzane

EAS TE

Sant’Angelo Lodigiano

Val Trompia

Brescia

RN LOMBARD

V

E

N

E

TO Vicenz a

Verona Padua

CVL only in oxytones Cremona

CVL also in paroxytones CVL only in oxytones, with restrictions (only certain vowels and/or only certain lexemes and/or variably neutralized) No CVL Transition between contrastive consonant gemination and CVL

Rovigo

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Airolo

Borgomanero

Ortisei

Livigno

Müstair Valley Airolo

A

L

P

Livigno

I

N

Rossa (Calanca) Augio

Giornico

Signôra Valcolla

L

Soglio

San Giacomo O Filippo Villa di Chiavenna Samòlaco Novate Mezzola

M

Premana

B

R APoschiavo

Chiuro Campo

Corticiasca

Antronapiana

Trepalle

E

Tartano

Val Rendena

Mendrisio

Dar fo Boario Term e Lecco

Balerna

Malcesine Bergam o

Borgomanero

Val Cavallina Sant’Apollonio di Lumezzane

Bienate Brescia CVL only in oxytones

Milan

CVL also in paroxytones CVL only in oxytones, with restrictions (only certain vowels and/or only certain lexemes and/or variably neutralized) No CVL Transition between contrastive consonant gemination and CVL

Map 5 Alpine Lombard

Sant’Angelo Lodigiano

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Pianezza Borgosesia

Teglio Aprica

Schilpario

Campertogno Piedicavallo

Grosio

Vilminore di Scalve

Casale Corte Cerro Rossa (Valsesia)

Bormio

Castionetto Val d’Ambria

Comano

Selveglio di Riva Valdobbia

D

E

L I G U R I A N

M I L I A N

Modena

Compiano

Bologna

Val di Taro

Pavullo Benedello Crocette Sillano

Riolunato Pianaccio

Piandelagotti

Lizzano in Belvedere

Colognora Gombitelli

T U S C A N CVL also in paroxytones No CVL Transition between contrastive consonant gemination and CVL

Map 6 The Emilian Apennine

Firenzuol a

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Parma

OUP CORRECTED PROOF – FINAL, 10/7/2015, SPi

Modena

Bologna

Re

no

EMILIA ROMAGNA

Cavanna

a

Pistoia

E. L im

ella ntr

Stabiazzoni

en t r a

Lagacci

W. Lim ent r Lim e

Re

no

Castello di Sambuca

TUSCANY

CVL also in paroxytones No CVL Transition between contrastive consonant gemination and CVL

Florence

Map 7 Transition from contrastive gemination to CVL in the territory of Sambuca Pistoiese (after Filipponio and Nocchi 2010: 228, with modifications)

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References Abete, Giovanni (2006). Sulla questione della sillaba superpesante: i dittonghi discendenti in sillaba chiusa nel dialetto di Pozzuoli. In Savy and Crocco (2006: 92 [summary], 379–98 [CD-ROM]). Acquati, Anna (1971). Il vocalismo latino-volgare nelle iscrizioni africane. Annali della Facoltà di lettere e filosofia dell’Università degli studi di Milano 24: 155–84. Adams, James N. (1994). Latin and Punic in contact? The case of the Bu Njem ostraca. Journal of Roman Studies 89: 87–112. Adams, James N. (1995). The language of the Vindolanda writing tablets: an interim report. Journal of Roman Studies 85: 86–134. Adams, James N. (1999). The poets of Bu Njem: language, culture and the centurionate. Journal of Roman Studies 89: 109–34. Adams, James N. (2003). Bilingualism and the Latin Language. Cambridge: Cambridge University Press. Adams, James N. (2007). The Regional Diversification of Latin 200 bc–ad 600. Cambridge: Cambridge University Press. Adams, James N. (2011). Late Latin. In Clackson (2011: 257–81). Adams, James N. (2013). Social Variation and the Latin Language. Cambridge: Cambridge University Press. Ageno, Gian Carlo (1957). Studi sul dialetto genovese. Genoa: Istituto internazionale di studi liguri, Sezione di Genova. AIS = Jaberg and Jud (1928–40). Alarcos Llorach, Emilio (1951). Esbozo de una fonología diacrónica del español. In Estudios dedicados a Ramón Menéndez Pidal, vol. 2. Madrid: CSIC, 9–39. ALEIC = Bottiglioni (1933–42). Allen, W. Sidney (1973). Accent and Rhythm: Prosodic Features of Latin and Greek. A Study in Theory and Reconstruction. Cambridge: Cambridge University Press. Allen, W. Sidney (1987). Vox Graeca: A Guide to the Pronunciation of Classical Greek, 3rd edn. Cambridge: Cambridge University Press. Anderson, Stephen (1981). Why phonology isn’t ‘natural’. Linguistic Inquiry 12: 493–540. Antonioli, Gabriele, and Bracchi, Remo (1995). Dizionario etimologico grosino: con annotazioni di carattere etnografico e storico e repertorio italiano–grosino. Grosio: Biblioteca comunale. Ascoli, Graziadio Isaia (1873). Saggi ladini. Archivio glottologico italiano 1: 1–556. Azaretti, Emilio (1982). L’evoluzione dei dialetti liguri esaminata attraverso la grammatica storica del ventimigliese, 2nd edn. Sanremo: Edizioni Casabianca. Badía Margarit, Antonio (1950). El habla del Valle de Bielsa (Pirineo aragonés). Barcelona: Instituto de estudios pirenaicos. Badía [i] Margarit, Antonio (1984). Gramàtica històrica catalana, 2nd edn. Valencia: Climent. Bafile, Laura (1996). Sulla rappresentazione delle strutture metriche ternarie. Quaderni del dipartimento di linguistica dell’Università di Firenze 7: 2–24. Baković, Eric (2011). Opacity and ordering. In Goldsmith et al. (2011: 40–67).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

251

Banniard, Michel (1992). VIVA VOCE: communication écrite et communication orale du IVe au IXe siècle en Occident latin. Paris: Institut des études augustiniennes. Barbato, Marcello (2008). Sistemi vocalici a contatto in area italo-romanza. In Sabine Heinemann (ed.), Sprachwandel und (Dis-)Kontinuität in der Romania. Tübingen: Niemeyer, 139–52. Barbato, Marcello (2009). Metafonia napoletana e metafonia sabina. In Alessandro De Angelis (ed.), I dialetti italiani meridionali tra arcaismo e interferenza. Palermo: Centro di studi filologici e linguistici siciliani, 275–89. Barbera, Massimiliano (2012). Toscani a metà: degeminazione e geminazione nell’italiano di Carrara. In Cristina Albizu, Larissa Birrer, Marie Burkhardt, Hans-Jörg Döhla, Marc Duval, Alain Schorderet, Marie-Florence Sguaitamatti, and Vera Ziswiler (eds), Kontaminationen/ Contaminations/Contaminazioni/Contaminaciones: atti del IV Dies Romanicus Turicensis. Aachen: Shaker, 215–37. Barnes, Timothy D. (1967). The family and career of Septimius Severus. Historia 16: 67–107. Baroni, Marco (1993–4). La relazione tra struttura segmentale e costituenza moraica. Graduation thesis, University of Padua. Baroni, Marco, and Vanelli, Laura (1999). Il contrasto di lunghezza vocalica in friulano. In Paola Benincà, Alberto M. Mioni, and Laura Vanelli (eds), Fonologia e morfologia dell’italiano e dei dialetti d’Italia. Rome: Bulzoni, 291–317. Baroni, Marco, and Vanelli, Laura (2000). The relationship between vowel length and consonantal voicing in Friulian. In Lori Repetti (ed.), Phonological Theory and the Dialects of Italy. Amsterdam: Benjamins, 13–44. Barthes, Roland (1967). Le discours de l’histoire. Social Science Information 6: 65–75. Trans. Stephen Bann as ‘The discourse of history’, Comparative Criticism 3 (1981): 7–20. Bartoli, Matteo G. (1906). Das Dalmatische: Altromanische Sprachreste von Veglia bis Ragusa und ihre Stellung in der Apennino-Balkanischen Romania. 2 vols. Vienna: Hölder. Bartoli, Matteo (1943). Lineamenti di linguistica spaziale. Milan: Le Lingue estere. Barwick, Karl (1922). Remmius Palaemon und die römische Ars grammatica. Leipzig: Dieterich’sche Verlagsbuchhandlung. Battisti, Carlo (1926). Le premesse fonetiche e la cronologia dell’evoluzione di á in é nel ladino centrale. L’Italia dialettale 2: 50–84. Battisti, Carlo (1929). Sulla pretesa unità ladina. Archivio glottologico italiano 22/23: 409–45 (= Benvenuto A. Terracini (ed.), Silloge linguistica dedicata alla memoria di Graziadio Isaia Ascoli). Battisti, Carlo (1946). La crisi del latino: lezioni universitarie sul latino volgare. Florence: Universitaria Editrice. Bellati, Antonio, and Bracchi, Remo (2007). Dizionario dialettale etnografico di Premana. De cént in-t-üne, la favola di una comunità. Premana (Lecco): Il corno. Bender, Byron W., Francescato, Giuseppe, and Salzmann, Zdeněk (1952). Friulian phonology. Word 8: 216–23. Benedetti, Marina (1995). Le consonanti dopo -au- tra lenizione e rafforzamento: un capitolo di fonetica storica latina e romanza. Pisa/Rome: Istituti editoriali e poligrafici internazionali. Benedetti, Marina, and Marotta, Giovanna (2014). Monottongazione e geminazione in lationo: nuovi elementi a favore dell'isocronismo sillabico. In Piera Molinelli, Pierluigi Cuzzonlin, and Chiara Fedriana (eds), Latin vulgaire-latin tardif X: actes de Xe colloque international sur

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

252

References

le latin vulgaire et tardif. Bergamp, 5–9 Septembre 2012. Bergamo: Bergamo University Press/ Sestante Edizioni, 25–43. Benincà, Paola (1989). Friulano. Evoluzione della grammatica. In LRL, vol. 3: Die einzelnen romanischen Sprachen und Sprachgebiete von der Renaissance bis zur Gegenwart: Rumänisch, Dalmatisch/Istroromanisch, Friaulisch, Ladinisch, Bündnerromanisch, 563–85. Benincà, Paola (1995). Il friulano. In LRL, vol. 2(2): Die einzelnen romanischen Sprachen und Sprachgebiete vom Mittelalter bis zur Renaissance, 42–61. Benincà, Paola, and Vanelli, Laura (2005). Linguistica friulana. Padua: Unipress. Bermúdez-Otero, Ricardo (2007). Diachronic phonology. In de Lacy (2007: 497–517). Bernardasci, Camilla (2013a). La quantità vocalica distintiva in marebbano: un’analisi sperimentale. Ladinia 37: 125–46. Bernardasci, Camilla (2013b). Un’analisi fonetica di alcuni aspetti del sistema vocalico e consonantico del dialetto di Piandelagotti (MO). Master’s thesis, University of Zürich. Bernardi Perini, Giorgio (1986). L’accento latino: cenni teorici e norme pratiche. Bologna: Pàtron. Bernini, Giuliano, and Sanga, Glauco (1987). Fonologia del dialetto di Bergamo. In Sanga (1987a: 37–63). Bernoth, Anja (2008). Interne Sprachgeschichte des Dalmatischen. In Ernst et al. (2008: 2731–49). Berruto, Gaetano (1970). Dialetto e società industriale nella Valle d’Andorno. Turin: Istituto dell’Atlante linguistico italiano. Berruto, Gaetano (1974). Piemonte e Valle d’Aosta. Pisa: Pacini. Berschin, Helmut, Felixberger, Josef, and Goebl, Hans (1978). Französische Sprachgeschichte: Lateinische Basis, interne und externe Geschichte, sprachliche Gliederung Frankreichs. Mit einer Einführung in die historische Sprachwissenschaft. Munich: Hueber. Bertinetto, Pier Marco (1980). La quantità vocalica in italiano: verifica spettrografica. In Atti del XIV Congresso Internazionale di Linguistica e Filologia Romanza, vol. 3. Naples: Macchiaroli/Amsterdam: Benjamins, 75–85. Bertinetto, Pier Marco (1981). Strutture prosodiche dell’italiano. Florence: Accademia della Crusca. Bertinetto, Pier Marco (2004). Piede. In Gian Luigi Beccaria (ed.), Dizionario di linguistica e di filologia, metrica, retorica, 2nd edn. Turin: Einaudi, 588. Bertoletti, Nello (2005). Testi veronesi dell’età scaligera: edizione, commento linguistico e glossario. Padua: Esedra. Bianchini, Giovanni, and Bracchi, Remo (2003). Dizionario etimologico dei dialetti della Val Tartano, 2nd edn. Madonna di Tirano (Sondrio): Istituto di dialettologia e di etnografia valtellinese e valchiavennasca. Biasini, Carla (2010). Fonetica e fonologia del dialetto di Piandelagotti sull’Appennino modenese. Lavor0 di licenza, University of Zürich. Biffi, Ambrogio (1606). Prissian da Milan della parnonzia Milanesa. In Giulio Lepschy (ed.), Una fonologia milanese del 1606: il Prissian da Milan della parnonzia Milanesa. L’Italia dialettale 28 (1965): 143–80. Repr. in Lepschy (1978: 177–215). Birley, Anthony R. (1999). Septimius Severus: The African Emperor, 2nd edn. London: Routledge.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

253

Biville, Frédérique (1995). Les emprunts du latin au grec: approche phonétique, vol. 2: Vocalisme et conclusions. Louvain: Peeters. Blevins, Juliette (2004). Evolutionary Phonology: the Emergence of Sound Patterns. Cambridge: Cambridge University Press. Boersma, Paul (1998). Functional phonology. Dissertation, University of Amsterdam. Boersma, Paul (2007). Cue constraints and their interactions in phonological perception and production. ROA 944; available at: http://roa.rutgers.edu/view.php3?roa=944. Boitani, Piero, Mancini, Mario, and Vàrvaro, Alberto (eds) (2001). Lo spazio letterario del medioevo, 2: Il medioevo volgare, vol. 1.2: La produzione del testo. Rome: Salerno. Boitani, Piero, Mancini, Mario, and Vàrvaro, Alberto (eds) (2002). Lo spazio letterario del medioevo, 2: Il medioevo volgare, vol. 2: La circolazione del testo. Rome: Salerno. Boldrini, Sandro (1992). Fondamenti di prosodia e metrica latina. Rome: Carocci. Bolelli, Tristano (1997). Per una storia della ricerca linguistica, 2nd edn. Naples: Morano. Bolognesi, Roberto (1998). The Phonology of Campidanian Sardinian. Dordrecht: HIL. Bonfadini, Giovanni (1983). Il confine linguistico veneto-lombardo. In Manlio Cortelazzo (ed.), Guida ai dialetti veneti, vol. 5. Padua: Unipress, 23–59. Bonfadini, Giovanni (1987). Il dialetto della Val Cavallina e zone adiacenti. In Sanga (1987a: 317–95). Bonfadini, Giovanni (1990). Il dialetto bresciano: modello cittadino e varietà periferiche. Rivista italiana di dialettologia 14: 41–92. Bonfadini, Giovanni (1993). Per una ridefinizione del lombardo orientale. Atti del Sodalizio glottologico milanese 31: 103–15. Bonfadini, Giovanni (1997). Le opposizioni vocaliche di durata nel dialetto di Novate Mezzola (So). In Renato Arena, M. Patrizia Bologna, M. Luisa Mayer Modena, and Alessandro Passi (eds), Bandhu: Scritti in onore di Carlo Della Casa in occasione del suo settantesimo compleanno. Alessandria: Edizioni dell’Orso, 583–604. Bonfadini, Giovanni (2010). Tra Valtellina e Val Camonica: il dialetto di Aprica. In Gabriele Iannàccaro, Massimo Vai, and Vittorio Dell’Aquila (eds), ‘Féch, cun la o cume fuguus’: per Romano Broggini in occasione del suo 85 compleanno, gli amici e allievi milanesi. Alessandria: Edizioni dell’Orso, 21–33. Bonfante, Giuliano (1956). Il sardo kenápura e la quantità latina. La parola del passato 11: 347–58. Bonfante, Giuliano (1968). Quando si è incominciato a parlare italiano? Criterii fonologici. In Kurt Baldinger (ed.), Festschrift Walter von Wartburg zum 80. Geburtstag, vol 1. Tübingen: Niemeyer, 21–46. Bonfante, Giuliano (1983). La lingua latina parlata in età imperiale. In Temporini and Haase (1983: 413–52). Bonfante, Giuliano (1998). The Origin of the Romance Languages: Stages in the Development of Latin, ed. Larissa Bonfante. Heidelberg: Winter. Bonzi, Maria Luisa (1972–3). Ricerche lessicali a Castello di Sambuca. Graduation thesis, University of Florence. Bosoni, Giorgio (1995). Dialettologia lombarda: un esempio di approccio strumentale allo studio delle opposizioni di quantità vocalica in sillaba tonica. Studi italiani di linguistica teorica e applicata 4: 245–364.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

254

References

Bosoni, Jørgen Giorgio (2001). Fonologia e fonetica del lecchese. In Anderlo Biella, Virginia Favaro Lanzetti, Luciana Mondini, and Gianfranco Scotti, Vocabolario italiano–lecchese, lecchese–italiano. Oggiono (Lecco): Cattaneo, v–xiii. Bosoni, Jørgen Giorgio (2003). Una proposta di grafia unificata per le varietà linguistiche lombarde: regole per la trascrizione. Bollettino storico Alta Valtellina 6: 195–297. Bosoni, Jørgen Giorgio, and Mambretti, Emanuele (2011). Osservazioni sulla fonologia, fonetica e grafia dei dialetti di Livigno e Trepalle. In Mambretti and Bracchi (2011: 129–78). Bottiglioni, Gino (1928). L’antico genovese e le isole linguistiche sardo-corse. L’Italia dialettale 4: 1–78. Bottiglioni, Gino (1933–42). Atlante linguistico-etnografico italiano della Corsica. 10 vols. Pisa: Simoncini. Bottiglioni, Gino (1954). Manuale dei dialetti italici: Osco, Umbro e dialetti minori. Bologna: STEB. Bourciez, Édouard (1937). Précis de phonétique française, 8th edn. Paris: Klincksieck. Bourciez, Édouard (1946). Eléments de linguistique romane, 4th edn. Paris: Klincksieck. Bouvier, Jean-Claude (1979). L’occitan en Provence: le dialecte provençal, ses limites et ses variétés. Revue de linguistique romane 43: 46–62. Bracchi, Remo (2010). Profilo del dialetto di Villa di Chiavenna. In Giorgetta and Ghiggi (2010: 25–244, 751–956). Broselow, Ellen, Chen, Su-I, and Huffman, Marie (1997). Syllable weight: convergence of phonology and phonetics. Phonology 14: 47–82. Browman, Catherine P., and Goldstein, Louis (1990). Representation and reality: physical systems and phonological structure. Journal of Phonetics 18: 411–24. Brüch, Josef (1921). Zur Entwicklung der betonten Vokale im Volkslatein. Zeitschrift für romanische Philologie 41: 574–82. Bruni, Francesco (1984). L’italiano: elementi di storia della lingua e della cultura. Turin: UTET. Buck, Carl D. (1928). A Grammar of Oscan and Umbrian, 2nd edn. Boston: Ginn. Repr. Hildesheim: Olms, 1974. Burdy, Philipp (2006). Untersuchungen zu lat. au und seinen romanischen Nachfolgern. Hamburg: Buske. Burger, André (1928). Etudes de phonétique et de morphologie latines. Recueil de travaux publiés par la Faculté des lettres, University of Neuchâtel, 13. Burzio, Luigi (1994). Principles of English Stress. Cambridge: Cambridge University Press. Cabella, Maurizio (1999). Vocabolario del dialetto tortonese. Alessandria: Edizioni dell’Orso. Caduff, Léonard (1952). Essai sur la phonétique du parler rhétoroman de la Vallée de Tavetsch (Canton des Grisons—Suisse). Bern: Francke. Calabrese, Andrea (1984–5). Metaphony in Salentino. Rivista di grammatica generativa 9–10: 3–140. Calabrese, Andrea (2000). The feature [advanced tongue root] and vowel fronting in Romance. In Repetti (2000: 57–88). Camodeca, Giuseppe (1999). Tabulae Pompeianae Sulpiciorum (TPSulp.): edizione critica dell’archivio puteolano dei Sulpicii. Rome: Quasar. Campanile, Enrico (1971). Due studi sul latino volgare. L’Italia dialettale 34: 1–64. Repr. in Campanile (2008: 337–400).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

255

Campanile, Enrico (1973). Sulla quantità della vocale che precede -m in latino. L’Italia dialettale 36: 1–6. Repr. in Campanile (2008: 401–6). Campanile, Enrico (2008). Latina et Italica: scritti minori sulle lingue dell’Italia antica, ed. Paolo Poccetti. 2 vols. Pisa: Serra. Cano Aguilar, Rafael (2008). Historia interna del español: fonología y sistema gráfico. In Ernst et al. (2008: 3062–78). Casella, Mario (1922). Studi sui dialetti della Valdarda: fonologia del dialetto di Fiorenzuola. Studj romanzi 17: 5–71. Castellani, Arrigo (1952). Nuovi testi fiorentini del Dugento. 2 vols. Florence: Sansoni. Castellani, Arrigo (1961). Sulla formazione del tipo fonetico italiano: fenomeni vocalici. Studi linguistici italiani 2: 24–45. Repr. in Castellani (1980: 1.73–95). Castellani, Arrigo (1970a). Dittongamento senese e dittongamento aretino nei dialetti dell’Italia mediana (in epoca antica). In I dialetti dell’Italia mediana con particolare riguardo alla regione umbra: atti del V Convegno di studi umbri. Gubbio: Centro di studi umbri presso la Casa di Sant’Ubaldo-Perugia, 311–80. Repr. in Castellani (1980: 1.358–422). Castellani, Arrigo (1970b). Note sul dittongamento toscano. In Mille: i dibattiti del Circolo linguistico fiorentino. Florence: Olschki, 41–53. Repr. in Castellani (1980: 1.146–55). Castellani, Arrigo (1970c). Ancora sul dittongamento italiano e romanzo: seconda risposta a Friedrich Schürr. Cultura neolatina 30: 117–30. Repr. in Castellani (1980: 1.156–76). Castellani, Arrigo (1972). Frammenti d’un libro di conti castellano del Dugento. Studi di filologia italiana 30: 5–58. Castellani, Arrigo (1976). I più antichi testi italiani: edizione e commento, 2nd edn. Bologna: Pàtron. Castellani, Arrigo (1980). Saggi di linguistica e filologia italiana e romanza (1946–1976). 3 vols. Rome: Salerno. Castellani, Arrigo (1991). Sulla scomparsa dell’opposizione di quantità vocalica in latino volgare. In Dieter Kremer (ed.), Actes du XVIIIe Congrès International de Linguistique et Philologie Romanes, vol. 3, section 5: Grammaire diachronique et histoire de la langue. Tübingen: Niemeyer, 10–21. Castellani, Arrigo (2000). Grammatica storica della lingua italiana, vol. 1: Introduzione. Bologna: Il Mulino. Castellani, Riccardo (1980). Il friulano occidentale. Modena: Del Bianco. Chambers, J. K., and Trudgill, Peter (1980). Dialectology. Cambridge: Cambridge University Press. Charlet, Jean-Louis (1993). Un témoignage humaniste sur la latinité africaine et le grec parlé par les ‘Choriates’: Paolo Pompilio. Antiquités africaines 29: 241–47. Chen, Matthew (1970). Vowel length variation as a function of the voicing of the consonant environment. Phonetica 22: 129–59. Chene, Brent Eugene de (1979). The historical phonology of vowel length. Bloomington: Indiana University Linguistics Club. Chene, Brent Eugene de and Anderson, Stephen R. (1979). Compensatory lengthening. Language 55: 505–35. Chierchia, Gennaro (1986). Length, syllabification and the phonological cycle in Italian. Journal of Italian Linguistics 8: 5–33.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

256

References

Chomsky, Noam, and Halle, Morris (1968). The Sound Pattern of English. New York: Harper & Row. Christiansen, Jacobus (1889). De apicibus et I longis inscriptionum latinarum. Ph.D dissertation, University of Kiel. Chung, Sandra (1983). Transderivational relationships in Chamorro phonology. Language 59(1): 35–66. CIL = Corpus Inscriptionum Latinarum. Consilio et auctoritate Academiae Litterarum Regiae Borussicae, editum Berolini: apud Georgium Reimerum MDCCCXCIII–. Cioranescu, Alexandre (1958–66). Diccionario etimológico rumano. La Laguna: Universidad de la Laguna. Clackson, James (ed.) (2011). A Companion to the Latin Language. Oxford: Wiley-Blackwell. Clackson, James, and Horrocks, Geoffrey (2007). The Blackwell History of the Latin Language. Oxford: Wiley-Blackwell. Cocito, Luciana (ed.) (1970). Anonimo genovese, Le rime volgari. Poesie. Edizione critica, introduzione, commento e glossario. Rome: Edizioni dell’Ateneo. Coco, Francesco (1970). Il dialetto di Bologna. Bologna: Forni. Coco, Francesco (1971). Effetti della degeminazione consonantica nel dialetto bolognese. In Atti del VII Convegno del Centro per gli Studi Dialettali Italiani. Turin: Rattero, 162–7. Contini, Gianfranco (1935a). Commento ai testi bresciani. L’Italia dialettale 11: 133–51. Part 2 of ‘Antichi testi bresciani, editi da Giuseppe Bonelli e commentati da Gianfranco Contini’, L’Italia dialettale 11: 115–51. Contini, Gianfranco (1935b). Per il trattamento delle vocali d’uscita in antico lombardo. L’Italia dialettale 11: 33–60. Contini, Gianfranco (1941). Le opere volgari di Bonvesin da la Riva. Rome: Società filologica romana. Contini, Gianfranco (1960). Poeti del Duecento. 2 vols. Milan: Ricciardi. Contini, Michel (1987). Etude de géographie phonétique et de phonétique instrumentale du sarde. 2 vols. Alessandria: Edizioni dell’Orso. Coseriu, Eugenio (1981). Los conceptos de “dialecto”, “nivel” y “estilo de lengua” y el sentido propio de la dialectología, Lingüística española actual 3: 1–32. Published in German translation as ‘Die Begriffe “Dialekt” “Niveau” und “Sprachstil” und der eigentliche Sinn der Dialektologie’, in Jörn Albrecht, Harald Thun, and Jens Lüdtke (eds), Energeia und Ergon: sprachliche Variation, Sprachgeschichte, Sprachtypologie: studia in honorem Eugenio Coseriu, vol. 1: Schriften von Eugenio Coseriu (1967–1987). Tübingen: Narr, 1988, 15–43. Coseriu, Eugenio (2008). Lateinisch–Romanisch: Vorlesungen und Abhandlungen zum sogenannten Vulgärlatein und zur Entstehung der romanischen Sprachen. Tübingen: Narr. CPL = Robert Cavenaile (ed.), Corpus papyrorum Latinarum. Wiesbaden: Harrassowitz, 1956–8. Craffonara, Lois (1971–2). Le parlate di San Vigilio di Marebbe, di San Martino e la Villa in Val Badia. Graduation thesis, University of Padua. Craffonara, Lois (1977). Zur Stellung der Sellamundarten im romanischen Sprachraum. Ladinia 1/2: 73–120. Cravens, Thomas D. (1988). Consonant strength in the Romance dialects of the Pyrenees. In David Birdsong and Jean-Pierre Montreuil (eds), Advances in Romance Linguistics. Dordrecht: Foris, 371–409.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

257

Cravens, Thomas D. (1991). Phonology, phonetics and orthography in Late Latin and Romance: the evidence for early intervocalic sonorization. In Roger Wright (ed.), Latin and the Romance Languages in the Early Middle Ages. London: Routledge, 52–69. Cravens, Thomas D. (2002). Comparative Historical Dialectology: Italo-Romance Clues to Ibero-Romance Sound Change. Amsterdam: Benjamins. Crevatin, Franco (1992). Intorno al vocalismo ‘protoromanzo’. In Etymologie und Wortgeschichte des Italienischen/Genesi e dimensioni di un vocabolario etimologico. Wiesbaden: Reichert, 26–31. Croatto, Enzo (2004). Vocabolario del dialetto ladino-veneto della Valle di Zoldo (Belluno). Costabissara (Vicenza): Colla. Crosswhite, Catherine M. (2001). Vowel Reduction in Optimality Theory. London: Routledge. Cull, Naomi (1995). Reconstruction of the Proto-Romance syllable. In Henning Andersen (ed.), Historical Linguistics 1993: Selected Papers from the 11th International Conference on Historical Linguistics. Amsterdam: Benjamins, 117–32. Dalbera, Jean-Philippe (1994a). Les parlers des Alpes-Maritimes: étude comparative, essai de reconstruction. London: Association internationale d’études occitanes. Dalbera, Jean-Philippe (1994b). Systèmes en contact et dyamique évolutive: le cas de Bonifacio, isolat ligurien de Corse. Orbis 37: 97–112. Dalbera, Jean-Philippe (2013). Le ligurien. In Georg Kremnitz (ed.), Histoire sociale des langues de France. Rennes: Presses universitaires de Rennes, 503–11. Dalbera-Stefanaggi, Marie-José (2001). Essais de linguistique corse. Ajaccio: Piazzola. Dardel, Robert de, and Wüest, Jakob (1993). Les systèmes casuels du protoroman: les deux cycles de simplification. Vox Romanica 52: 25–65. Dauzat, Albert (1897). Etudes linguistiques sur la Basse Auvergne. Phonétique historique du patois de Vinzelles (Puy-de-Dôme). Paris: Félix Alcan. Davis, Stuart (2011). Quantity. In Goldsmith et al. (2011: 103–40). Deaton, Karen, Noske, Manuela, and Ziolkowsky, Michael (eds) (1990). Papers from the Twenty-Sixth Regional Meeting of the Chicago Linguistic Society, vol. 2: The Parasession on the Syllable in Phonetics and Phonology. Chicago: Chicago Linguistic Society. Degrassi, Attilio (1963). Inscriptiones Latinae liberae rei Publicae, vol. 2. Florence: La Nuova Italia. De Gregorio, Iolanda (1939). Contributo alla conoscenza del dialetto di Bisceglie (Bari). L’Italia dialettale 15: 31–52. Della Ferrera, Tarcisio (2008). Vocabolario dialettale di Chiuro e Castionetto: profilo del dialetto di Chiuro e Castionetto di Remo Bracchi. Chiuro (Sondrio): Istituto di dialettologia e di etnografia valtellinese e valchiavennasca. Dell’Aglio, Monica, Bertinetto, Pier Marco, and Agonigi, Maddalena (2002). Le durate dei foni vocalici in rapporto al contesto nel parlato di locutori pisani: primi risultati. In Agostino Regnicoli (ed.), La fonetica acustica come strumento di analisi della variazione linguistica in Italia. Atti delle XII Giornate di studio del Gruppo di fonetica sperimentale. Rome: Il Calamo, 53–8. Dell’Aquila, Vittorio (2010). Appunti di fonetica storica del valsesiano con particolare attenzione al vocalismo nella parlata di Rossa. In Gabriele Iannàccaro, Massimo Vai, and Vittorio

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

258

References

Dell’Aquila (eds), ‘Féch, cun la o cume fuguus’. Per Romano Broggini in occasione del suo 85° compleanno, gli amici e allievi milanesi. Alessandria: Edizioni dell’Orso, 69–87. De Lorenzo Tobolo, Élia (1977). Dizionario del dialetto ladino di Comèlico superiore, provincia di Belluno. Bologna: Tamari. Delucchi, Rachele (2012). Esiti di -a finale e armonia vocalica: i dialetti della Svizzera italiana in prospettiva romanza e generale. Ph.D dissertation, University of Zürich. Devaux, André, and Duraffour, Antonin (1935). Les patois du Dauphiné. Lyon: Bibliothèque de la Faculté catholique des lettres. Dietrich, A. (1852). Zur geschichte des accents im lateinischen. Zeitschrift für vergleichende Sprachforschung 14: 543–56. D’Imperio, Mariapaola (2000). Acoustic-perceptual correlates of sentence prominence in Italian. OSU-WPL 54: 59–77. D’Imperio, Mariapaola, and Rosenthall, Sam (1999). Phonetics and phonology of main stress in Italian. Phonology 16: 1–28. Dmitrieva, Olga (2012). Geminate typology and the perception of consonant duration. Ph.D dissertation, Stanford University. Dressler, Wolfgang U. (1984). Explaining natural phonology. Phonology Yearbook 1: 29–50. Dressler, Wolfgang U. (1985). Morphonology: The Dynamics of Derivation. Ann Arbor, Mich.: Karoma. Dressler, Wolfgang U., Prinzhorn, Martin, and Rennison, John R. (eds) (1994). Phonologica 1992: Proceedings of the 7th International Phonology Meeting. Turin: Rosenberg & Sellier. DRG: Robert von Planta, Florian Melcher, and Chasper Pult (1939–). Dicziunari rumantsch grischun. Bischofberger: Societad Retorumantscha Cuoira. Duraffour, Antonin (1932). Phénomènes généraux d’évolution phonétique dans les dialectes franco-provençaux étudiés d’après le parler de la commune de Vaux (Ain). Revue de linguistique romane 8: 1–280. Elcock, William D. (1938). De quelques affinités phonétiques entre l’aragonais et le béarnais. I: La conservation des occlusives sourdes entre voyelles. II: La sonorisation des occlusives sourdes après nasale ou liquide. Paris: Droz. Elwert, Wilhelm Theodor (1943). Die Mundart des Fassa-Tals. Heidelberg: Winter. Ernst, Gerhard, Gleßgen, Martin-Dietrich, Schmitt, Christian, and Schweickard, Wolfgang (eds) (2008). Romanische Sprachgeschichte/Histoire linguistique de la Romania, 3. Berlin: Mouton de Gruyter. Essen, Otto von (1964). Vokalquantitäten im Rätoromanischen des Unterengadins. Phonetica 11: 133–43. Ettmayer, Karl R. von (1902). Lombardisch-ladinisches aus Südtirol. Romanische Forschungen 13: 321–673. Ettmayer, Karl R. von (1919). Zur Kenntnis des Altladinischen. Zeitschrift für romanische Philologie 39: 1–17. EWA: Lloyd, Albert L., Springer, Otto, and Lühr, Rosemarie (1988–). Etymologisches Wörterbuch des Althochdeutschen. Göttingen: Vandenhoeck & Ruprecht. Faggin, Giorgio (1985). Vocabolario della lingua friulana. Udine: Del Bianco. Fanciullo, Franco (1988). Lucania. In LRL, vol. 4: Italienisch, Korsisch, Sardisch, 669–88.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

259

Fanciullo, Franco (1992). Un capitolo della Romania submersa: il latino africano. In Dieter Kremer (ed.), Actes du XVIIIe Congrès International de Linguistique et Philologie Romanes, I: Romania submersa—Romania nova. Tübingen: Niemeyer, 162–87. Fanciullo, Franco (1997). Anticipazioni romanze nel latino pompeiano. Archivio glottologico italiano 82: 186–98. Fankhauser, Franz (1911). Das Patois von Val d’Illiez (Unterwallis). Halle a.S.: Ehrhardt Karras. First published in Revue de dialectologie romane 2 (1910): 189–344; 3 (1911): 1–76. Farnetani, Edda, and Kori, Shiro (1986). Effects of syllable and word structure on segmental durations in spoken Italian. Speech Communication 5: 17–34. Farnetani, Edda, and Kori, Shiro (1990). Rhythmic structure in Italian noun phrases: a study on vowel duration. Phonetica 47: 50–65. Fassò, Andrea (1974). Analisi fonematica della parola. In Celestino Guichardaz and Andrea Fassò, La parlata francoprovenzale di Cogne (Val d’Aosta). Turin: Giappichelli, 213–46. Fava, Elisabetta, and Magno Caldognetto, Emanuela (1976). Studio sperimentale delle caratteristiche elettroacustiche delle vocali toniche e atone in bisillabi italiani. In Simone et al. (1976: 35–79). Ferguson, Thaddeus (1976). A History of the Romance Vowel Systems through Paradigmatic Reconstruction. The Hague: Mouton. Ferrari, Stefania (2007–8). Fenomeni fonologici in bresciano: un caso di allungamento vocalico. Graduation thesis, University of Padua. FEW = Wartburg (1928–2002). Filipponio, Lorenzo (2006). Simpliciter sonat: a proposito di una didascalia di Pompeo grammatico. In Renato Oniga and Luigi Zennaro (eds), Atti della I Giornata di linguistica latina. Venice: Cafoscarina, 41–70. Filipponio, Lorenzo (2010). La quantità vocalica nei proparossitoni etimologici al confine tra toscano e gallo-italico. In Maria Iliescu, Heidi M. Siller-Runggaldier, and Paul Danler (eds), Actes du XXVe Congrès International de Linguistique et Philologie Romanes, vol. 2. Berlin: de Gruyter, 67–76. Filipponio, Lorenzo (2012). La struttura di parola dei dialetti della Valle del Reno: profilo storico e analisi sperimentale. Bologna: Forni. Filipponio, Lorenzo (2013). Alberto Zamboni a proposito di muta cum liquida in latino e romanzo: l’occasione per un ricordo. Archivio Glottologico Italiano 98: 80–102. Filipponio, Lorenzo, and Nocchi, Nadia (2010). Diagnostica fonetica e diagnosi fonologica: ossitoni lunghi di sillaba libera a Sambuca Pistoiese (PT). In Stephan Schmid, Michael Schwarzenbach and Dieter Studer (eds), La dimensione temporale del parlato. Torriana: EDK, 225–48. Finco, Franco (2007). La durata delle vocali friulane: risultati di un’indagine fonetica. In Vicario (2007: 119–40). Finco, Franco (2009). Fonetiche e fonologije dal furlan centrâl. Gjornâl Furlan des Sciencis 11: 53–85. Fleischer, Jürg, and Schmid, Stephan (2006). Zürich German. Journal of the International Phonetic Association 36(2): 243–53. Flemming, Edward (2001). Scalar and categorical phenomena in a unified model of phonetics and phonology. Phonology 18: 7–44.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

260

References

Flemming, Edward (2004). Contrast and perceptual distinctiveness. In Bruce Hayes, Donca Steriade, and Robert Kirchner (eds), Phonetically-Based Phonology. Cambridge: Cambridge University Press, 232–76. Flobert, Pierre (1989). Le vocalisme de la Lyonnaise et de la Viennoise d’après les ‘apices’ et les ‘I longae’ des inscriptions latines. In La langue des inscriptions latines de la Gaule. Lyon: Boccard, 1–8. Flobert, Pierre (1990). Le témoignage épigraphique des apices et des I longae sur les quantités vocaliques en latin impérial. In Gualtiero Calboli (ed.), Latin vulgaire–latin tardif II: actes du IIème Colloque International sur le Latin Vulgaire et Tardif. Tübingen: Niemeyer, 101–10. Flutre, Louis-Ferdinand (1955). Le parler picard de Mesnil-Martinsart (Somme). Geneva: Droz. Foerster, Wendelin (1878). Bestimmung der lateinischen Quantität aus dem Romanischen. Rheinisches Museum für Philologie 33 (Neue Folge): 291–9. Fögen, Thorsten (1997–8). Der Grammatiker Consentius. Glotta 74: 164–92. Formentin, Vittorio (2002). L’area italiana medievale. In Boitani et al. (2002: 97–147). Formentin, Vittorio (2007). Poesia italiana delle origini. Rome: Carocci. Forner, Werner (1975). Generative Phonologie des Dialekts von Genua. Hamburg: Buske. Forner, Werner (1988). Ligurien. In LRL, vol. 4: Italienisch, Korsisch, Sardisch, 453–69. Forner, Werner (2010). Le brigasque: Occitan? La France latine 151: 45–92. Fortson, Benjamin W. IV (2011). Latin prosody and metrics. In Clackson (2011: 92–104). Fouché, Pierre (1958). Phonétique historique du français, vol. 2: Les voyelles. Paris: Klincksieck. Fowler, Carol A. (1991). Vowel duration and closure duration in voiced and unvoiced stops: there are no contrast effects here. Haskins Laboratories Status Report on Speech Research 107/108: 123–40. Francard, Michel (1980). Le parler de Tenneville: introduction à l’étude linguistique des parlers wallo-lorrains. Louvain-la-Neuve: Cabay. Francescato, Giuseppe (1954–7). Il bilinguismo friulano-veneto. Atti dell’Accademia di scienze, lettere e arti di Udine 14 (series 4): 3–29. Francescato, Giuseppe (1959). La dittongazione friulana. L’Italia dialettale 23: 43–54. Repr. in Francescato (1970: 14–21). Francescato, Giuseppe (1963). Il dialetto di Erto. Zeitschrift für romanische Philologie 79: 492–525. Repr. in Francescato (1970: 65–96). Francescato, Giuseppe (1963–6). Uno scrittore friulano del Settecento e il suo dialetto. Atti dell’Accademia di scienze, lettere e arti di Udine 5 (series 7): 61–83. Francescato, Giuseppe (1964). Il confine friulano-veneto a oriente. In Giovan Battista Pellegrini and Luigi Ciceri (eds), 41 Congresso della Società Filologica Friulana. Udine: Doretti, 58–62. Repr. in Francescato (1970: 58–64). Francescato, Giuseppe (1966). Dialettologia friulana. Udine: Società filologica friulana. Francescato, Giuseppe (1970). Studi linguistici sul friulano. Florence: Olschki. Francescato, Giuseppe, and Salimbeni, Fulvio (2004). Storia, lingua e società in Friuli. Rome: Il Calamo. Franceschi, Temistocle (1976). Sull’evoluzione del vocalismo dal latino repubblicano al neolatino. In Scritti in onore di Giuliano Bonfante, vol. 1, Brescia: Paideia, 259–79. Franceschini, Fabrizio (1991). Note sull’anafonesi in Toscana occidentale. In Luciano Giannelli, Nicoletta Maraschio, Teresa Poggi Salani, and Massimo Vedovelli (eds), Tra Rinascimento e strutture attuali: atti del Primo Convegno SILFI. Turin: Rosenberg & Sellier, 259–72.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

261

Frau, Giuseppe (1984). Friuli. Pisa: Pacini. Galletier, Édouard (1922). Etude sur la poésie funéraire romaine d’après les inscriptions. Paris: Hachette. Gartmann, Christian (1967). Die Mundart von Sorso (Provinz Sassari, Sardinien). Zürich: Juris. Gartner, Theodor (1883). Raetoromanische Grammatik. Heilbronn: Henninger. Gaudenzi, Augusto (1889). I suoni, le forme, le parole dell’odierno dialetto della città di Bologna. Turin: Loescher. Gerster, Walter (1927). Die Mundart von Montana (Wallis) und ihre Stellung innerhalb der frankoprovenzalischen Mundarten des Mittelwallis. Aarau: Sauerländer. Gess, Randall S., and Arteaga, Deborah (eds) (2006). Historical Romance Linguistics: Retrospective and Perspectives. Amsterdam: Benjamins. Gess, Randall S. (2011). Distinctive vowel length in Old French. Evidence and implications. In Laurel Brinton (ed.), Historical linguistics 1999. Selected papers from the 14th International Conference on Historical Linguistics, Vancouver, 9–13, 1999. Amsterdam: Benjamins, 145–56. Ghini, Mirco (1997). Segments, prosodic units, and ambisyllabicity in Ligurian. In Pier Marco Bertinetto, Livio Gaeta, Georgi Jetchev and David Michaels (eds), Certamen phonologicum III: Papers from the Third Cortona Phonology Meeting. Turin: Rosenberg & Sellier, 181–204. Ghini, Mirco (2001). Asymmetries in the Phonology of Miogliola. Berlin: Mouton de Gruyter. Giannarelli, Dante (1913). Studi sui dialetti lunigianesi compresi fra la Magra e l’Appennino reggiano. Revue de dialectologie romane 5: 261–311. Giannelli, Luciano (1984). Il dialetto di Sillano rivisitato: riflessioni su una situazione ‘difficile’. Rivista di archeologia, storia e costume 12: 3–20. Giannini, Stefania, and Marotta, Giovanna (1989). Fra grammatica e pragmatica: la geminazione consonantica in latino. Pisa: Giardini. Giorgetta, Giovanni, and Ghiggi, Stefano (2010). Vocabolario del dialetto di Villa di Chiavenna: profilo del dialetto di Villa di Chiavenna di Remo Bracchi. Grosio: Istituto di dialettologia e di etnografia valtellinese e valchiavennasca. GL = Keil (1856–80). Goidànich, Pier Gabriele (1906). Note di esegesi e critica di testi grammaticali latini. Rivista di filologia e d’istruzione classica 34: 35–56. Gökçen, Adnan (1990). Della quantità nel vocalismo milanese. Quaderni d’italianistica (Toronto) 11(2): 238–64. Goldsmith, John, Riggle, Jason, and Yu, Alan (eds) (2011). The Handbook of Phonological Theory, 2nd edn. Oxford: Wiley-Blackwell. Goyette, Stéphane (2000). From Latin to Early Romance: a case of partial creolization. In McWhorter (2000: 103–31). Grandgent, Charles (1907). An Introduction to Vulgar Latin. Boston: Heath. Grassi, Corrado (1968). Che cosa ne pensa Chiaffredo Roux. In Cesare Segre (ed.), Linguistica e filologia: omaggio a Benvenuto Terracini. Milan: Il Saggiatore, 149–63. Gredig, Stephanie (2000). Fonetica e fonologia del vocalismo tonico nel romancio di Scuol (Bassa Engadina) con uno studio sperimentale delle durate vocaliche. Lavoro di licenza, University of Zürich. Gröber, Gustav (ed.) (1904–6). Grundriss der romanischen Philologie, 2nd edn, vol. 1. Strassburg: Trübner.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

262

References

Grünert, Matthias (2012). Remartgas introductivas a la reediziun dals texts rumantschs. In Elvira Glaser and Michele Loporcaro (eds), Stimmen der Schweiz = Voix de la Suisse = Voci della Svizzera = Vuschs da la Svizra. In historischen Aufnahmen. In text und Ton herausgegeben für die Landi 1939. Frauenfeld: Huber, 124–7. Gsell, Otto (1996). Chronologie frühromanischer Sprachwandel. In LRL, vol. 2(1): Latein und Romanisch: Historisch-vergleichende Grammatik der romanischen Sprachen, 557–84. Gsell, Otto (2008). Interne Sprachgeschichte des Dolomitenladinischen. In Ernst et al. (2008: 2770–90). Haas, Wim de (1988). A Formal Theory of Vowel Coalescence: a Case Study of Ancient Greek. Dordrecht: Foris. Hagberg, Larry (1990). Syllabification of long vowels in Mayo. In Deaton et al. (1990: 159–83). Haiman, John, and Benincà, Paola (1992). The Rhaeto-Romance Languages. London: Routledge. Hajek, John (1995). A first acoustic study of the interaction between vowel and consonant duration in Bolognese. Rivista italiana di acustica 19: 3–9. Hajek, John (1997). Analisi acustica delle quantità segmentali in area bolognese. Rivista italiana di dialettologia 21: 133–47. Hajek, John (2000). How many moras? Overlength and maximal moraicity in Italy. In Repetti (2000: 111–35). Halle, Morris (1977). Tenseness, vowel shift, and the phonology of the back vowels in Modern English. Linguistic Inquiry 8: 611–26. Halle, Morris (1997). On stress and accent in Indo-European. Language 73(2): 275–313. Harris-Northall, Ray (1991). Apocope in Alfonsine texts: a case study. In Thomas Cravens and Ray Harris-Northall (eds), Linguistic Studies in Medieval Spanish. Madison, Wis.: Hispanic Seminary of Medieval Studies, 29–38. Haudricourt, André Georges, and Juilland, Alphonse (1949). Essai pour une histoire structurale du phonétisme français. Paris: Klincksieck. Hayes, Bruce (1989). Compensatory lengthening in moraic phonology, Linguistic Inquiry 20: 253–306. Hayes, Bruce (1995). Metrical Stress Theory: Principles and Case Studies. Chicago: University of Chicago Press. Heilmann, Luigi (1955). La parlata di Moena, nei suoi rapporti con Fiemme e con Fassa. Bologna: Zanichelli. Heilmann, Luigi, and Oneda, Romano (1976). Dizionario del dialetto cremonese. Cremona: Libreria del Convegno. Heine, Bernd (1979). Some linguistic characteristics of African-based pidgins. In Ian F. Hancock, Edgar Charles Polomé, Morris F. Goodman, and Bernd Heine (eds), Readings in Creole Studies. Ghent: E. Story-Scientia, 89–98. Heinemann, Sabine (2003). Studien zur Stellung des Friaulischen in der nördlichen Italoromania. Bonn: Romanistischer Verlag. Herman, József (1965). Aspects de la différenciation du latin: problèmes généraux. Bulletin de la Société de linguistique de Paris 60: 53–70. Repr. in Herman (1990: 10–28). Herman, József (1968). Statistique et diachronie: essai sur l’évolution du vocalisme dans la latinité tardive. Word 24: 242–51. Repr. in Herman (1990: 196–203).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

263

Herman, József (1970). Les particularités de l’évolution du latin provincial. In Actele celui de-al XII-lea Congres Internaţional de Lingvistică s,i Filologie Romanică, vol. 1. Bucharest: Editura Academiei, 125–30. Repr. in Herman (1990: 29–34). Herman, József (1971). Essai sur la latinité du littoral adriatique à l’époque de l’empire. In Eugenio Coseriu and Wolf-Dieter Stempel (eds), Sprache und Geschichte: Festschrift Harri Maier. Munich: Fink, 199–226. Repr. in Herman (1990: 121–46). Herman, József (1982). Un vieux dossier réouvert: les transformations du système latin des quantités vocaliques. Bulletin de la Société de linguistique 77: 285–302. Repr. in Herman (1990: 217–31). Herman, József (1983). Le latin dans les provinces danubiennes. In Temporini and Haase (1983: 1089–1106). Herman, József (1985a). La différenciation territoriale du latin et la formation des langues romanes. In Linguistique comparée et typologie des langues romanes: actes du XVIIe Congrès international de linguistique et philologie romanes, Aix-en-Provence, 29 août– 3 septembre 1983. Aix-en-Provence: University of Provence, vol. 2, 15–62. Repr. in Herman (1990: 62–92). Herman, József (1985b). Témoignage des inscriptions latines et préhistoire des langues romanes: le cas de la Sardaigne. In Mélanges Skok. Zagreb: Jugoslavenska Akademija Znanosti i Umjetnosti, 207–16. Repr. in Herman (1990: 183–94). Herman, József (1990). Du latin aux langues romanes: études de linguistique historique. Tübingen: Niemeyer. Herman, József (1995). Les ardoises wisigothiques et le problème de la différenciation territoriale du latin. In Louis Callebat (ed.), Latin vulgaire–latin tardif IV: actes du 4e colloque international sur le latin vulgaire et tardif. Hildesheim: Olms–Weidmann, 63–76. Herman, József (1996). The end of the history of Latin. Romance Philology 49: 364–82. Repr. in Herman (2006: 195–213). Herman, József (1998). La chronologie de la transition: un essai. In Herman and Mondin (1998: 5–25). Herman, József (2000). Differenze territoriali nel latino parlato dell’Italia tardo-imperiale: un contributo preliminare. In József Herman and Anna Marinetti (eds), La preistoria dell’italiano. Tübingen: Niemeyer, 123–35. Herman, József (2006). Du latin aux langues romanes II: Nouvelles études de linguistique historique. Tübingen: Niemeyer. Herman, József, and Mondin, Luca (eds) (1998). La transizione dal latino alle lingue romanze: atti della Tavola rotonda di linguistica storica, Università Ca’ Foscari di Venezia, 14–15 giugno 1996. Tübingen: Niemeyer. Hernández-Campoy, Juan Manuel, and Trudgill, Peter (2002). Functional compensation and southern peninsular Spanish /s/-loss. Folia Linguistica Historica 23: 31–57. Hillenbrand, James M. (2003). American English: Southern Michigan. Journal of the International Phonetic Association 33: 121–6. Hilty, Gerold (1969). Zur diphthongierung im Galloromanischen und im Iberoromanischen. In Wolf-Dieter Lange and Joseph Maria Piel (eds), Philologische Studien für Joseph M. Piel. Heidelberg: Winter, 95–107. Also (in Spanish translation) in Hilty (2007: 133–40). Hilty, Gerold (2001). I primi testi romanzi. In Boitani et al. (2001: 57–89).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

264

References

Hilty, Gerold (2007). Íva·l con la edat el coraçón creçiendo: estudios escogidos sobre problemas de lengua y literatura hispánicas. Madrid: Iberoamericana/Frankfurt a.M.: Vervuert. Hock, Hans Heinrich (1991). Principles of Historical Linguistics. Berlin: Mouton de Gruyter. Hoenigswald, Henry M. (1950). The principal step in comparative grammar. Language 26: 357–64. Hogg, Richard (1996). Old English open syllable lengthening, Transactions of the Philological Society 94(1): 57–72. Holm, John (1989). Pidgins and Creoles, vol. 2: Reference Survey. Cambridge: Cambridge University Press. Holm, John (2000). An Introduction to Pidgins and Creoles. Cambridge: Cambridge University Press. Holtz, Louis (1981). Donat et la tradition de l’enseignement grammatical: étude sur l’Ars Donati et sa diffusion (IVe–IXe siècle) et édition critique. Paris: CNRS. Hualde, José I. (1990). Compensatory lengthening in Friulian. Probus 2: 31–46. Hualde, José I. (1992). Catalan. London: Routledge. Huber, Joseph (1954–5). Texte in der Mundart von Livigno. Vox Romanica 14: 243–68. Huber, Joseph (1960–61). Zur Mundart von Trepalle. Zeitschrift für romanische Philologie 76: 376–445; 77: 470–513. Hyman, Larry M. (1999). The historical interpretation of vowel harmony in Bantu. In JeanMarie Hombert and Larry M. Hyman (eds), Bantu Historical Linguistics: Theoretical and Empirical Perspectives. Stanford, Calif.: CSLI, 235–95. Hyman, Larry M. (2001). On the limits of phonetic determinism in phonology: *NC revisited. In Elisabeth Hume and Keith Johnson (eds), The Role of Speech Perception Phenomena in Phonology. San Diego, Calif.: Academic Press, 141–85. Iosad, Pavel (2012). Final devoicing and vowel lengthening in Friulian: a representational approach. Lingua 122: 922–51. Jaberg, Karl, and Jud, Jakob (1928–40). Sprach- und Sachatlas Italiens und der Südschweiz. 8 vols. Zofingen: Ringier. Jackson, Kenneth H. (1994). Language and History in Early Britain: a Chronological Survey of the Brittonic Languages 1st to 12th century ad, 2nd edn. Dublin: Four Courts Press. Jacobs, Heike (1990). On markedness and bounded stress systems. Linguistic Review 7: 81–119. Jellinek, Max Hermann (1926). Geschichte der gotischen Sprache. Berlin: de Gruyter. Jespersen, Otto (1932). Lehrbuch der Phonetik, 5th edn. Leipzig: Teubner. Joly, Geneviève (1995). Précis de phonétique historique du français. Paris: Colin. Joseph, Brian D., and Janda, Richard D. (eds) (2003). The Handbook of Historical Linguistics. Oxford: Blackwell. Kager, René (1995). The metrical theory of word stress. In John Goldsmith (ed.), The Handbook of Phonological Theory. Oxford: Blackwell, 367–402. Kaster, Robert A. (1988). Guardians of Language: The Grammarian and Society in Late Antiquity. Berkeley: University of California Press. Kavitskaya, Darya (2002). Compensatory Lengthening: Phonetics, Phonology, Diachrony. New York: Routledge. Kaye, Jonathan (1989). Phonology: a Cognitive View. Hillsdale, NJ: Erlbaum. Kaye, Jonathan, Lowenstamm, Jean, and Vergnaud, Jean-Roger (1990). Constituent structure and government in phonology. Phonology 7: 193–231.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

265

Kaze, Jeffery W. (1992). Five underlying vowels in Bolognese: evidence from metaphony. In Ruth M. Brend (ed.), The Eighteenth LACUS Forum 1991. Lake Bluff, Ill.: Linguistic Association of Canada and the United States, 330–9. Keating, Patricia (1990). Phonetic representations in a generative grammar. Journal of Phonetics 18: 321–34. Keil, Heinrich (1856–80). Grammatici latini. 7 vols. Leipzig: Teubner. Repr. Hildesheim: Olms, 1981. Keller, Hans-Erich (1958). Etudes linguistiques sur les parlers valdôtains: contribution à la connaissance des dialectes franco-provençaux modernes. Bern: Francke. Keller, Oskar (1919). Der Genferdialekt dargestellt auf Grund der Mundart von Certoux, 1: Lautlehre. Zürich: Leemann. Kerlouégan, François (1978). Les voyelles longues du latin classique: étude phonologique. Langages 50: 32–7. King, Robert D. (1969). Historical Linguistics and Generative Grammar. Englewood Cliffs, NJ: Prentice Hall. King, Robert D. (1988). A problem of vowel lengthening in early New High German. Monatshefte 80(1): 22–31. Kiparsky, Paul (1965). Phonological change. Ph.D dissertation, MIT. Kiparsky, Paul (1973). The inflectional accent in Indo-European. Language 49: 794–849. Kiparsky, Paul (1985). Some consequences of Lexical Phonology. Phonology Yearbook 2: 85–138. Kirchner, Robert (1997). Contrastiveness and faithfulness. Phonology 14: 83–111. Kiss, Sándor (1971). Les transformations de la structure syllabique en latin tardif. Debrecen: Studia Romanica Universitatis de Ludovico Kossuth nominatae. Klein, Thomas B. (2006). Creole phonology typology: phoneme inventory size, vowel quality distinctions and stop consonant series. In Parth Bhatt and Ingo Plag (eds), The Structure of Creole Words: Segmental, Syllabic and Morphological Aspects. Tübingen: Niemeyer, 3–21. Kluender, Keith R., Diehl, Randy L., and Wright, Beverly A. (1988). Vowel-length differences before voiced and voiceless consonants: an auditory explanation. Journal of Phonetics 16: 153–69. Koller, Hermann (1981). Die Silbenquantitäten in Augustinus’ Büchern De Musica. Museum Helveticum 38: 262–7. Kramer, Johannes (1972). Abbozzo di una fonematica del sursilvano letterario. Revue roumaine de linguistique 17: 345–57. Kramer, Johannes (1981). Historische Grammatik des Dolomitenladinischen, 2nd edn. Würzburg: Lehmann. Kramer, Johannes (1983). Glossaria bilinguia in papyris et membranis reperta. Bonn: Habelt. Křepinský, Maximilian (1958). Romanica II: la naissance des langues romanes et l’existence d’une période de leur évolution commune (latin vulgaire, période romane). Rozpravy Československé Akademie Věd (Prague) 68(13): 1–78. Repr. in Reinhold Kontzi (ed.), Zur Entstehung der romanischen Sprachen (Darmstadt: Wissenschaftliche Buchgesellschaft, 1978), 301–65. Kuen, Heinrich (1923). Zur Chronologie des Uebergangs von a > e im Grödnischen. ZRPh 43: 68–77. Kuryłowicz, Jerzy (1958). L’accentuation des langues indo-européennes, 2nd edn. Wrocław: Zakład Narodowy imienia Ossolińskich/Kraków: Wydawnictwo Polskiej Akademii Nauk.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

266

References

Labov, William (1974). On the use of the present to explain the past. In Proceedings of the 11th International Congress of Linguists, vol. 2. Bologna: Il Mulino, 825–51. Lacy, Paul de (ed.) (2007). The Cambridge Handbook of Phonology. Cambridge: Cambridge University Press. Ladd, D. Robert (2011). Phonetics in phonology. In Goldsmith et al. (2011: 348–73). Lahiri, Aditi, and Dresher, Elan (1999). Open syllable lengthening in West Germanic. Language 75(4): 678–719. Lahiri, Aditi, and Fikkert, Paula (1999). Trisyllabic shortening in English: past and present. English Language and Linguistics 3(2): 229–67. Lahiri, Aditi, Riad, Tomas, and Jacobs, Heike (1999). Diachronic prosody. In Harry van der Hulst (ed.), Word Prosodic Systems in the Languages of Europe. Berlin: Mouton de Gruyter, 335–422. Lancel, Serge (1981). La fin et la survie de la latinité en Afrique du Nord: état des questions. Revue des études latines 59: 269–97. Lapesa, Rafael (1968). Historia de la lengua española, 7th edn. Madrid: Escelicer. Lass, Roger (1985). Minkova noch einmal: MEOSL and the resolved foot. Folia Linguistica Historica 4: 245–65. Lass, Roger (1997). Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Lausberg, Heinrich (1939). Die Mundarten Südlukaniens. Halle a.S.: Niemeyer. Lausberg, Heinrich (1974). Noterelle di dialettologia italiana, 1: Nachrichten der Akademie der Wissenschaften in Göttingen (Philologisch-Historische Klasse) 7: 251–60. Lausberg, Heinrich (1976). Linguistica romanza, 2nd edn, vol. 1: Fonetica. Milan: Feltrinelli. Laver, John (1994). Principles of Phonetics. Cambridge: Cambridge University Press. Lazzeroni, Romano (1956). La ‘geminatio vocalium’ nelle iscrizioni latine. Annali della Scuola normale superiore di Pisa 25 (series 2): 124–35. Repr. in Tristano Bolelli and Saverio Sani (eds), Scritti scelti di Romano Lazzeroni (Pisa: Pacini, 1997), 277–92. Ledgeway, Adam (2012). From Latin to Romance: Morphosyntactic Typology and Change. Oxford: Oxford University Press. Lehiste, Ilse (1970). Suprasegmentals. Cambridge, Mass.: MIT Press. Lejeune, Michel (1975). Réflexions sur la phonologie du vocalisme osque. Bulletin de la Société de linguistique de Paris 70: 233–51. Leonard, Clifford S. Jr (1972). The vocalism of Proto-Rhaeto-Romance. Orbis 21: 61–100. Lepscky [= Lepschy], Giulio (1962). Il problema dell’accento latino. Annali della Scuola normale superiore di Pisa 31 (series 2): 199–246. Lepschy, Giulio (1962). Fonematica veneziana. L’Italia dialettale 25: 1–22. Lepschy, Giulio (1978). Saggi di linguistica italiana. Bologna: Il Mulino. Leumann, Manu (1977). Lateinische Laut- und Formenlehre, 5th edn. Munich: Beck. Levelt, Willem J. M., Roelofs, Ardi, and Meyer, Antje S. (1999). A theory of lexical access in speech production. Behavioral and Brain Sciences 22: 1–38. Lewicki, Tadeusz (1951–2). Une langue romane oubliée de l’Afrique du Nord: observations d’un arabisant. Rocznik orientalistyczny 17: 415–80. Lightner, Theodore (1975). The role of derivational morphology in generative grammar. Language 51: 617–38.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

267

Lindsay, Wallace M. (1894). The Latin Language: An Historical Account of Latin Sounds, Stems, and Flexions. Oxford: Clarendon Press. Liver, Ricarda (1999). Rätoromanisch. Tübingen: Narr. Löfstedt, Bengt (1983). Rückschau und Ausblick auf die vulgärlateinische Forschung: Quellen und Methoden. In Temporini and Haase (1983: 453–79). Löfstedt, Einar (1959). Late Latin. Oslo: Aschenhoug. Löhken, Sylvia C. (1997). Deutsche Wortprosodie. Abschwächungs- und Tilgungsvorgänge. Tübingen: Stauffenburg. Lombardi, Linda, and McCarthy, John (1991). Prosodic circumscription in Choctaw morphology. Phonology 8: 37–72. Loporcaro, Michele (1988a). Grammatica storica del dialetto di Altamura. Pisa: Giardini. Loporcaro, Michele (1988b). History and geography of raddoppiamento fonosintattico: remarks on the evolution of a phonological rule. In Pier Marco Bertinetto and Michele Loporcaro (eds), Certamen phonologicum: Papers from the 1987 Cortona Phonology Meeting. Turin: Rosenberg & Sellier, 341–87. Loporcaro, Michele (1991). Compensatory lengthening in Romanesco. In Pier Marco Bertinetto, Michael Kenstowicz, and Michele Loporcaro (eds), Certamen phonologicum II: Papers from the 1990 Cortona Phonology Meeting. Turin: Rosenberg & Sellier, 279–307. Loporcaro, Michele (1996). On the analysis of geminates in Standard Italian and Italian dialects. In Bernhard Hurch and Richard Rhodes (eds), Natural Phonology: the State of the Art. Berlin: Mouton de Gruyter, 153–87. Loporcaro, Michele (1997). L’origine del raddoppiamento fonosintattico: saggio di fonologia diacronica romanza. Basel: Francke. Loporcaro, Michele (2001). Distinctive features and phonological change: vowel fronting and gravity interactions in Altamurano. Rivista di linguistica 13: 255–308. Loporcaro, Michele (2002–3). Di una presunta reintroduzione preromanza di -us di accusativo plurale in Sardegna. Studi e saggi linguistici 40–41: 187–205. Loporcaro, Michele (2005a). La lunghezza vocalica nell’Italia settentrionale alla luce dei dati del lombardo alpino. In Max Pfister and Gabriele Antonioli (eds), Itinerari linguistici alpini: atti del convegno di dialettologia in onore del prof. R. Bracchi. Sondrio: Istituto di dialettologia e di etnografia valtellinese e valchiavennasca, 97–113. Loporcaro, Michele (2005b). La sillabazione di muta cum liquida dal latino al romanzo. In Sándor Kiss, Luca Mondin, and Giampaolo Salvi (eds), Latin et langues romanes: études de linguistique offertes à József Herman à l’occasion de son 80ème anniversaire. Tübingen: Niemeyer, 419–30. Loporcaro, Michele (2005–6). I dialetti dell’Appennino tosco-emiliano e il destino delle atone finali nel(l’italo-) romanzo settentrionale. L’Italia dialettale 66–67: 69–122. Loporcaro, Michele (2007a). Facts, theory and dogmas in historical linguistics: vowel quantity from Latin to Romance. In Salmons and Dubenion-Smith (2007: 311–36). Loporcaro, Michele (2007b). L’Appendix Probi e la fonetica del latino tardo. In Francesco Lo Monaco and Piera Molinelli (eds), L’Appendix Probi: nuove ricerche. Florence: SISMEL–Edizioni del Galluzzo, 95–124. Loporcaro, Michele (2007c). Osservazioni sul romanesco contemporaneo. In Claudio Giovanardi and Franco Onorati (eds), Le lingue der monno. Rome: Aracne, 181–96.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

268

References

Loporcaro, Michele (2008). La tendenza alla sillaba chiusa in latino tardo. In Wright (2008: 336–44). Loporcaro, Michele (2011a). Morphological complexity without abstractness: Italo-Romance metaphony. Poznań Studies in Contemporary Linguistics 47(2): 177–202. Loporcaro, Michele (2011b). Phonological processes. In Maiden et al. (2011: 109–54, 689–98). Loporcaro, Michele (2011c). Syllable, segment and prosody. In Maiden et al. (2011: 50–108, 684–89). Loporcaro, Michele (2013). Profilo linguistico dei dialetti italiani, 2nd edn. Rome: Laterza. Loporcaro, Michele, Delucchi, Rachele, Nocchi, Nadia, Paciaroni, Tania, and Schmid, Stephan (2006). La durata consonantica nel dialetto di Lizzano in Belvedere (Bologna). In Savy and Crocco (2006: 98 [summary], 491–517 [CD-ROM]). Loporcaro, Michele, Delucchi, Rachele, Nocchi, Nadia, Paciaroni, Tania, and Schmid, Stephan (2007). Schwa finali sull’Appennino emiliano: il vocalismo del dialetto di Piandelagotti. In Veronica Giordani, Valentina Bruseghini, and Piero Cosi (eds), Scienze vocali e del linguaggio: metodologie di valutazione e risorse linguistiche. Torriana (RN): EDK, 5–6 [summary], 57–76 [CD-ROM]. Loporcaro, Michele, Paciaroni, Tania, and Schmid, Stephan (2005). Consonanti geminate in un dialetto lombardo alpino. In Piero Cosi (ed.), Misura dei parametri: aspetti tecnologici ed implicazioni nei modelli linguistici. Brescia: EDK, 67 [summary], 579–618 [CD-ROM]. Lorenzetti, Luca, and Schirru, Giancarlo (2010). Un indizio della conservazione di /k/ dinanzi a vocale anteriore nell’epigrafia cristiana di Tripolitania. In Tantillo and Bigi (2010: 303–11). Loyen, André (1943). Sidoine Apollinaire et l’esprit précieux en Gaule aux derniers jours de l’empire. Paris: Les Belles Lettres. LRL = Günter Holtus, Michael Metzeltin, and Christian Schmitt (eds) (1988–2001). Lexicon der Romanistischen Linguistik. 8 vols. Tübingen: Niemeyer. Ludlow, Peter (2011). The Philosophy of Generative Linguistics. Oxford: Oxford University Press. Lüdtke, Helmut (1954–5). Zur Lautlehre des Bündnerromanischen. Vox Romanica 14: 223–42. Lüdtke, Helmut (1956). Die strukturelle Entwicklung des romanischen Vokalismus. Bonn: Romanisches Seminar an der Universität. Lüdtke, Helmut (1957). Inchiesta sul confine dialettale fra il veneto e il friulano. Orbis 6: 122–5. Lüdtke, Helmut (1965). Die lateinischen Endungen -im, -um, -unt und ihre romanischen Ergebnisse. In Omagiu lui Alexandru Rosetti la 70 de ani. Bucharest: Editura Academiei Republicii Socialiste România, 487–99. Lüdtke, Helmut (2005). Der Ursprung der romanischen Sprachen: eine Geschichte der sprachlichen Kommunikation. Kiel: Westensee. Lupinu, Giovanni (2000). Latino epigrafico della Sardegna: aspetti fonetici. Nuoro: Ilisso. Lurà, Franco (1987). Il dialetto del Mendrisiotto: descrizione sincronica e diacronica e confronto con l’italiano. Mendrisio: Unione di Banche Svizzere. Lurati, Ottavio (1988). Lombardia e Ticino. In LRL, vol. 4: Italienisch, Korsisch, Sardisch, 485–516. Luschützky, Hans Christian (1984). Remarks on segmental quantity in Italian. Wiener linguistische Gazette 33–4: 105–20. Lutta, Conrad Martin (1923). Der Dialekt von Bergün und seine Stellung innerhalb der rätoromanischen Mundarten Graubündens. Halle a. S.: Niemeyer.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

269

Luzi, Johann (1904). Lautlehre der Subselvischen Dialekte. Erlangen: Junge & Sohn. Mackel, Emil (1896). Zur romanischen Vokaldehnung in betonter freien Silbe. Zeitschrift für romanische Philologie 20: 514–9. Maddalon, Marta, and Miotto, Attilio (1986). Gli sviluppi del sistema vocalico veneto a confronto con altri sistemi gallo-italici. Rivista italiana di dialettologia 10: 129–60. Maddieson, Ian (1984). Patterns of Sounds. Cambridge: Cambridge University Press. Maddieson, Ian (1985). Phonetic cues to syllabification. In Victoria A. Fromkin (ed.), Phonetic Linguistics. Orlando, Fla.: Academic Press, 203–21. Maddieson, Ian (1997). Phonetic universals. In William J. Hardcastle and John Laver (eds), The Handbook of Phonetic Sciences. Oxford: Blackwell, 619–39. Màfera, Giovanni (1958). Profilo fonetico-morfologico dei dialetti da Venezia a Belluno. L’Italia dialettale 22: 131–84. Magri, Riccardo (1995). Dialetto cremonese di città e dei paesi. Dizionario italiano–cremonese. Ortografia e grammatica. Cremona: Turris. Maiden, Martin (1995). A Linguistic History of Italian. New York: Longman. Trans. into Italian as Storia linguistica dell’italiano (Bologna: Il Mulino, 1998). Maiden, Martin (2011). Morphological persistence. In Maiden et al. (2011: 155–215, 699–707). Maiden, Martin, and Parry, Mair (eds) (1997). The Dialects of Italy. London: Routledge. Maiden, Martin, Smith, John Charles, and Ledgeway, Adam (eds) (2011). The Cambridge History of the Romance Languages, vol. 1: Structures. Cambridge: Cambridge University Press. Maina, Stefania (2001–2). Relazione tra lunghezza vocalica e rigeminazione consonantica in alcuni dialetti del Canton Grigioni (Svizzera) e della Val Chiavenna (So). Graduation thesis, University of Padua. Maina, Stefania (2006). Le consonanti geminate nei dialetti di due valli grigionesi. Quaderni grigionitaliani 75(4): 442–7. Mair, Walter (1973). Ennebergische Morphologie: Analyse eines dolomitenladinischen Flexionssystems. Innsbruck: Institut für Romanistik. Malagoli, Giuseppe (1930). Fonologia del dialetto di Lizzano in Belvedere (Appennino bolognese). L’Italia dialettale 6: 125–196. Repr. in Malagoli (2011: 15–86). Malagoli, Giuseppe (1954). Intorno ai dialetti dell’alta montagna reggiana, 1: Note fonològiche del dialetto di Collagna. L’Italia dialettale 19: 1–29. Malagoli, Giuseppe (2011). Il dialetto di Lizzano in Belvedere, ed. Lorenzo Filipponio and Michele Loporcaro. Vidiciatico: Gruppo Studi Capotauro. Mambretti, Emanuele, and Bracchi, Remo (eds) (2011). Dizionario etimologico-etnografico dei dialetti di Livigno e Trepalle. Comune di Livigno: Istituto di dialettologia e di etnografia valtellinese e valchiavennasca. Mancini, Anna Maria (1986). Note sul dialetto di Senigallia. Quaderni dell’Istituto di linguistica dell’Università di Urbino 4: 195–248. Mancini, Marco (1994). Un passo del grammatico Pompeo e la dittongazione protoromanza. In Palmira Cipriano, Paolo Di Giovine, and Marco Mancini (eds), Miscellanea di studi linguistici in onore di Walter Belardi, vol. 2. Rome: Il Calamo, 609–27. Mancini, Marco (2001). Agostino, i grammatici e il vocalismo del latino d’Africa. Rivista di linguistica/Italian Journal of Linguistics 13: 309–38.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

270

References

Mańczak, Witold (2012). Muta cum liquida et origine des langues romanes. In Frédérique Biville, Marie-Karine Lhommé, and Daniel Vallat (eds), Latin vulgaire, latin tardif IX. Actes du IXe colloque international, Lyon, 2–6 septembre 2009. Lyon: Maison de l’Orient et de la Méditerranée – Jean Pouilloux (Collection de la Maison de l’Orient et de la Méditerranée, 49; Série linguistique et philologique, 8), 183–87. Maragliano, Alessandro (1976). Dizionario dialettale vogherese, ed. Virginio Giacomo Bono. Bologna: Pàtron. Marcato, Carla (2008). Storia interna del friulano. In Ernst et al. (2008: 2750–69). Marcone, Valentina (2011–12). I dizionari dialettali moderni del veronese (1986–2005). Graduation thesis, University of Venice (Ca’ Foscari). Marichal, Robert (1992). Les ostraca de Bu Njem (Suppléments de Libya Antiqua 7). Assraya al Hamra/Tripoli: Département des Antiquités, Grande Jamahiriya Arabe, Libyenne, Populaire et Socialiste. Marotta, Giovanna (1985). Modelli e misure ritmiche: la durata vocalica in italiano. Bologna: Zanichelli. Marotta, Giovanna (1986). Rhythmical constraints on ‘syntactic doubling’. Journal of Italian Linguistics 8: 35–52. Martinet, André (1933). Remarques sur le système phonologique du français. Bulletin de la Société de linguistique de Paris 34: 191–202. Martinet, André (1939). Description phonologique du parler franco-provençal d’Hauteville (Savoie). Revue de linguistique romane 15: 1–86. Martinet, André (1945). La prononciation du français contemporain. Paris: Droz. Martinet, André (1947). Note sur la phonologie du français vers 1700. Bulletin de la Société de linguistique de Paris 43: 13–23. Repr. with modifications as ‘La phonologie du français vers 1700’ in Martinet (1969: 155–67). Martinet, André (1955). Economie des changements phonétiques. Bern: Francke. Martinet, André (1956). La description phonologique, avec application au parler francoprovençal d’Hauteville (Savoie). Geneva: Droz/Paris: Minard. Martinet, André (1959). L’évolution contemporaine du système phonologique français. Free University Quarterly (Amsterdam) 7(2): 1–16. Repr. with modifications in Martinet (1969: 168–90). Martinet, André (1961). Review of Keller (1958). Erasmus 14: 530–34. Repr. in Martinet (1975: 203–7). Martinet, André (1969). Le français sans fard. Paris: Presses Universitaires de France. Martinet, André (1970). Analyse phonématique et analyse prosodique. Revue des langues romanes 79: 149–56. Repr. in Martinet (1975: 195–203). Martinet, André (1975). Remarques sur la phonologie des parlers francoprovençaux. In Evolution des langues et reconstruction. Paris: Presses Universitaires de France, 195–207. Martinet, André (1990). Remarques sur la variété des usages dans la phonie du français. In John N. Green and Wendy Ayres-Bennett (eds.), Variation and Change in French. Essays Presented to Rebecca Posner on the Occasion of her Sixtieth Birthday. London: Routledge, 13–26. Martinet, André, and Walter, Henriette (1973). Dictionnaire de la prononciation française dans son usage réel. Paris: France-Expansion.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

271

Massera, Sandro (1985). Vocabolario del dialetto di Novate Mezzola. Chiavenna: Centro di studi storici valchiavennaschi. Mateus, M. Helena, and d’Andrade, Ernesto (2000). The Phonology of Portuguese. Oxford: Oxford University Press. Maurand, Georges (1974). Phonétique et phonologie du parler occitan d’Ambialet (Tarn). Toulouse: EUS. Mauvillon, Éléazar (1754). Cours complet de la langue françoise, distribué par exercices, à l’usage des personnes pour qui cette langue est étrangère. Dresden: Libraire du Roi. Mazzola, Michael (ed.) (1994). Issues and Theory in Romance Linguistics. Washington, DC: Georgetown University Press. McCarthy, John (2007). Derivations and levels of representation. In de Lacy (2007: 99–117). McCrary, Kristie (2002). Syllable structure vs. segmental phonotactics: geminates and clusters in Italian revisited. In Proceedings of the Texas Linguistics Society, University of Texas, Austin. Available at: http://uts.cc.utexas.edu/%7Etls/2002tls/TLS_2002_Proceed ings.html. McCrary, Kristie (2003). La fonetica e la fonologia della durata segmentale nell’italiano pisano. Paper presented at the workshop ‘Italiano Parlato’, 13–15 February 2003, Naples. McCrary, Kristie (2004). Prosodic structure and segment duration in Italian revisited. Paper presented at 34th Linguistic Symposium on Romance Languages, 12–14 March 2004, Salt Lake City. McCrary Kambourakis, Kristie (2007). Reassessing the Role of the Syllable in Italian Phonology: An Experimental Study of Consonant Cluster Syllabification, Definite Article Allomorphy and Segment Duration. New York: Routledge. McCullough, Matthew (2011). The sounds of Latin: phonology. In Clackson (2011: 83–91). McMahon, April M. S. (2000). Lexical Phonology and the History of English. Cambridge: Cambridge University Press. McWhorter, John (ed.) (2000). Language Change and Language Contact in Pidgins and Creoles. Amsterdam: Benjamins. Meigret, Louis (1550). Trętté de la grammęre françoęze. Paris: Wechel. Meillet, Antoine (1909–10). Review of Édouard Bourciez, Eléments de linguistique romane (1910). Bulletin de la Société de linguistique de Paris 16: cccix–cccxxiii. Meiser, Gerhard (2010). Historische Laut- und Formenlehre der lateinischen Sprache, 3rd edn. Darmstadt: Wissenschaftliche Buchgesellschaft. Mele, Biagio (2009). Fonetica e fonologia del dialetto di San Giovanni in Fiore. Tübingen: Francke. Menegus, Giuseppina (1980–81). Fonologia del dialetto di San Vito di Cadore: aspetti sincronici e diacronici. Graduation thesis, University of Padua. Menéndez Pidal, Ramon (1952). Manual de gramática histórica española, 9th edn. Madrid: Espasa Calpe. Menéndez Pidal, Ramon (1986). Orígenes del español, 10th edn. Madrid: Espasa Calpe. Merlo, Clemente (1911–12). Note fonetiche sul parlare di Bitonto (Bari), parte 1a: Vocalismo. Atti della Reale accademia delle scienze di Torino 47: 907–32. Merlo, Clemente (1912). Note fonetiche. In Clemente Merlo and Riccardo Zagaria, ‘Lessico etimologico del dialetto di Andria (Bari)’. Apulia 2: 1–20.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

272

References

Merlo, Clemente (1926). Il vocalismo tonico del dialetto di Carbonara di Bari. L’Italia dialettale 2: 85–99. Merlo, Clemente (1929). Consonanti brevi e consonanti lunghe nel dialetto di Borgo S. Sepolcro. L’Italia dialettale 5: 66–80. Merlo, Clemente (1932). breg. štęla n. l. L’Italia dialettale 8: 268. Merlo, Clemente (1951). Profilo fonetico dei dialetti della Valtellina. Akad. der Wiss. und der Literatur in Mainz–Abhandlungen der geistes- u. sozialwissenschaftlichen Klasse 151/15: 1369–98. Merlo, Clemente (1960–61). I dialetti lombardi. L’Italia dialettale 24: 1–12. Mester, R. Armin (1994). The quantitative trochee in Latin. Natural Language and Linguistic Theory 12: 1–61. Meyer-Lübke, Wilhelm (1890). Grammatik der romanischen Sprachen, I: Lautlehre. Leipzig: Reisland. Meyer-Lübke, Wilhelm (1904–6). Die lateinische Sprache in den romanischen Ländern. In Gröber (1904–6: 451–97). Meyer-Lübke, Wilhelm (1920). Einführung in das Studium der romanischen Sprachen, 3rd edn. Heidelberg: Winter. Michaelis, Hermann, and Passy, Paul (1914[1897]). Dictionnaire phonétique de la langue française, 2nd edn. Hanover: Meyer. Migliorini, Bruno, and Pellegrini, Giovan Battista (1971). Dizionario del feltrino rustico. Padua: Liviana. Mignot, Xavier (1975). Phonologie pragoise et phonologie générative dans la description du latin. Bulletin de la Société de linguistique de Paris 70: 203–31. Minkova, Donka (1982). The environment for open syllable lengthening in Middle English. Folia Linguistica Historica 3: 29–58. Miotti, Renzo (2002). Friulian. JIPA 32: 237–47. Miotti, Renzo (2007). Le varietà di Dignano, Flaibano e Sedegliano nel contesto dei dialetti friulani: aspetti fonologici. In Vicario (2007: 71–118). Molino, Gianni, and Romano, Antonio (2008). Il dialetto valsesiano nella media Valgrande—Area linguistica di Campertogno, Mollia e Rassa. Alessandria: Edizioni dell’Orso. Monachesi, Paola (1995). A Grammar of Italian Clitics. Tilburg: ITK. Montreuil, Jean-Pierre (1991). Length in Milanese. In Dieter Wanner and Douglas A. Kibbee (eds), New Analyses in Romance Linguistics. Amsterdam: Benjamins, 37–47. Morani, Moreno (2000). Introduzione alla linguistica latina. Munich: Lincom Europa. Morin, Yves Charles (1989). Changes in the French vocalic system in the nineteenth century. In M. E. H. Schouten and Pieter van Reenen (eds), New Methods in Dialectology. Dordrecht: Foris, 185–97. Morin, Yves Charles (1992). What are the historical sources of lengthening in Friulian? Probus 4: 81–4. Morin, Yves Charles (1994). Phonological interpretations of historical lengthening. In Dressler et al. (1994: 135–55). Morin, Yves Charles (2000). Le parler de Vinzelles revisité: observations phonologiques. In Élie Fayette (ed.), Actes du Colloque Albert Dauzat et le patrimoine linguistique auvergnat (5–7 novembre 1998, Thiers). Thiers: Parc naturel régional Livradois-Forez, 231–55.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

273

Morin, Yves Charles (2003). Syncope, apocope, diphthongaison et palatalisation en galloroman: problèmes de chronologie relative. In Fernando Sánchez Miret (ed.), Actas del XXIII Congreso internacional de lingüística y filología románica, vol. 1. Tübingen: Niemeyer, 113–69. Morin, Yves Charles (2006). On the phonetics of rhymes in classical and pre-classical French: a sociolinguistic perspective. In Gess and Arteaga (2006: 131–62). Morin, Yves Charles (2008). Histoire des systèmes phonique et graphique du français. In Ernst et al. (2008: 2907–26). Morin, Yves Charles (2012). Sources et évolution des distinctions de durée vocalique: l’éclairage du gallo-roman. In Mario Barra-Jover, Guylaine Brun-Trigaud, Jean-Philippe Dalbera, Patrick Sauzet, and Tobias Scheer (eds), Études de linguistique gallo-romane. Saint-Denis: Presses universitaires de Vincennes, 117–48. Morin, Yves Charles, and Dagenais, Louise (1988). Les normes subjectives du français et les français régionaux: la longueur vocalique depuis le XVIe siècle. In Karin van Reenen-Stein and Pieter van Reenen (eds), Distributions spatiales et temporelles, constellations des manuscrits: études de variation linguistique offertes à Anthonij Dees à l’occasion de son 60e anniversaire. Amsterdam: Benjamins, 153–62. Mühlhäusler, Peter (1997). Pidgin and Creole Linguistics, 2nd edn. London: University of Westminster Press. Müller, Daniela and Martín, Sidney (2012). A preliminary acoustic study of the Occitan vowel system. In Mario Barra-Jover, Guylaine Brun-Trigaud, Jean-Philippe Dalbera, Patrick Sauzet, and Tobias Scheer (eds), Études de linguistique gallo-romane. Saint-Denis: Presses universitaires de Vincennes, 149–59. Müller, Henry François (1921). When did Latin cease to be a spoken language in France? Romanic Review 12: 318–34. Muljačić, Žarko (2000). Das Dalmatische: Studien zu einer untergegangenen Sprache. Cologne: Böhlau. Myers, Scott (2000). Boundary disputes: the distinction between phonetic and phonological sound patterns. In Noel Burton-Roberts, Philip Carr, and Gerard Docherty (eds), Phonological Knowledge: Conceptual and Empirical Issues. Oxford: Oxford University Press, 245–72. Nandriş, Octave (1963). Phonétique historique du roumain. Paris: Klincksieck. Nardini Marigliano, Debora (2000). [akˈsi se ˈperla a ˈfaŋ]: grammatica del dialetto di Fano (PS). Lavoro di licenza, University of Zürich. Naro, Anthony J. (1971). Resolution of vocalic hiatus in Portuguese: diachronic evidence for binary features. Language 47: 381–93. Nazzi, Gianni, and Saidero, Deborah (2000). Friulan Dictionary: English–Friulan/Friulan–English. Udine: Ente Friuli nel Mondo. Nespor, Marina, and Vogel, Irene (1979). Clash avoidance in Italian. Linguistic Inquiry 10: 467–82. Nicolai, Luigi (2000). Il dialetto ladino di Selva di Cadore: dizionario etimologico. Selva di Cadore: Unión de i Ladìgn de Sélva. Nicolas, Jean (ed.) (1983). Anonimo genovese, Le poesie storiche. Genoa: Compagna. Nicolas, Jean (ed.) (1994). Anonimo genovese, Rime e ritmi latini. Bologna: Commissione per i testi di lingua.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

274

References

Nicolet, Nellie (1929). Der Dialekt des Antronatales: Lautlehre, Formenlehre, Texte, Glossar. Hallea. S.: Niemeyer. Nicoli, Franco (1983). Grammatica milanese. Busto Arsizio: Bramante. Niedermann, Max (1931). Historische Lautlehre des Lateinischen, 2nd edn. Heidelberg: Winter. Norberg, Dag (1958). Le développement du latin en Italie de saint Grégoire le Grand à Paul Diacre. In I caratteri del VII secolo in Occidente. Spoleto: Centro italiano di studi sull’alto Medioevo, 485–503, 519–37. Odden, David (2011a). Rules vs. constraints. In Goldsmith et al. (2011: 1–39). Odden, David (2011b). The representation of vowel length. In Marc van Oostendorp, Colin J. Ewen, Elizabeth V. Hume, and Keren Rice (eds), The Blackwell Companion to Phonology, vol. 1. Malden, Mass.: Wiley-Blackwell, 465–90. Ohala, John J. (1989). Sound change is drawn from a pool of synchronic variation. In L. Egil Breivik and Ernst Håkon Jahr (eds), Language Change: Contributions to the Study of its Causes. Berlin: Mouton de Gruyter, 173–98. Ohala, John J. (1990). There is no interface between phonology and phonetics: a personal view. Journal of Phonetics 18: 153–171. Oneda, Romano (1964). Fonologia del dialetto cremonese. Strenna dell’ADAFA (Cremona) 1964, 5–13. Oneda, Romano (1965). L’opposizione di quantità vocalica e il dialetto cremonese. Strenna dell’ADAFA (Cremona) 1965, 23–34. Ortega-Llebaria, Marta, and Prieto, Pilar (2007). Disentangling stress from accent in Spanish: production patterns of the stress contrast in deaccented syllables. In Pilar Prieto, Joan Mascaró, and Maria-Josep Solé (eds), Segmental and Prosodic Issues in Romance Phonology. Amsterdam: Benjamins, 155–76. Osthoff, Hermann, and Brugmann, Karl (1878). Morphologische Untersuchungen auf dem Gebiete der indogermanischen Sprachen. Leipzig: Hirzel. Pace, Anna (1993–4). Ricerche di morfosintassi sui dialetti di Trebisacce e Castrovillari. Graduation thesis, University of Calabria. Page, B. Richard (2007). On the irregularity of open syllable lengthening in German. In Salmons and Dubenion-Smith (2007: 337–50). Papa, Eugene (1986). Two Studies on the History of Southern Italian Vocalism (Ph.D dissertation, Indiana University, 1981). Ann Arbor, Mich.: University Microfilms. Parodi, Ernesto Giacomo (1898). Studi liguri, 1: Le carte latine. 2: Il dialetto nei primi secoli. Archivio glottologico italiano 14: 1–110. Parodi, Ernesto Giacomo (1901). Studi liguri, 2: Il dialetto nei primi secoli (continuazione). Archivio glottologico italiano 15: 1–82. Parodi, Ernesto Giacomo (1902–5). Studi liguri, 3: Il dialetto di Genova dal secolo XVI ai nostri giorni. Archivio glottologico italiano 16: 105–61, 333–65. Parry, Mair M. (2005). Parluma ’d Còiri: Sociolinguistica e grammatica del dialetto di Cairo Montenotte. Savona: Società savonese di storia patria. Paul, Hermann (1880). Prinzipien der Sprachgeschichte. Halle a.S.: Niemeyer. 8th edn. Tübingen: Niemeyer, 1968. Peletier, Jacques (1555). Dialogue de l’ortografé et prononciation françoęse. Lyon: Jean de Tournes.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

275

Pellegrini, Giovan Battista (1966). La posizione del veronese antico. In Vittore Branca (ed.), Dante e la cultura veneta. Florence: Olschki, 95–107. Repr. in Pellegrini (1977b: 75–82). Pellegrini, Giovan Battista (1972a). I cinque sistemi linguistici dell’italo-romanzo. Revue roumaine de linguistique 18: 105–29. Repr. in Pellegrini (1975b: 55–87). Pellegrini, Giovan Battista (1972b). Saggi sul ladino dolomitico e sul friulano. Bari: Adriatica. Pellegrini, Giovan Battista (1975a). Fonetica e fonematica. In Pellegrini (1975b: 88–141). Pellegrini, Giovan Battista (1975b). Saggi di linguistica italiana: storia, struttura, società. Turin: Boringhieri. Pellegrini, Giovan Battista (1977a). Carta dei dialetti d’Italia. Pisa: Pacini. Pellegrini, Giovan Battista (1977b). Dialetti veneti antichi. In Pellegrini (1977c: 33–88). Pellegrini, Giovan Battista (1977c). Studi di dialettologia e filologia veneta. Pisa: Pacini. Pellegrini, Giovan Battista (1982). Osservazioni di sociolinguistica italiana. L’italia dialettale 45: 1–34. Pellegrini, Giovan Battista (1991). La genesi del retoromanzo (o ladino). Tübingen: Niemeyer. Pensado Ruiz, Carmen (1984). Cronología relativa del castellano. Salamanca: Universidad de Salamanca. Pensado Ruiz, Carmen (1988). How do unnatural syllabifications arise? The case of consonant + glide in Vulgar Latin. Folia Linguistica Historica 8: 115–42. Pensado, Carmen (1993). Consonantes geminadas en la evolución histórica del español. In Ralph Penny (ed.), Actas del primer congreso anglo-hispano, Asociación de Hispanistas de Gran Bretaña e Irlanda, vol. 1: Lingüística. Madrid: Castalia, 193–204. Pensado, Carmen (2001). El valor de la toponimía en la reconstrucción de la fonética sintáctica. In Emili Casanova and V. M. Rosselló (eds), Congrés internacional de toponímia e onomàstica catalanes (València 18–21 d’abril 2001). Valencia: Denes, 863–78. Petrolini, Giovanni (1983). Sul carattere ligure delle parlate altovaltaresi. In Lorenzo Coveri and Diego Moreno (eds), Studi di etnografia e dialettologia ligure in memoria di Hugo Plomteux. Genoa: Sagep, 229–47. Pfister, Max (1987). Die Abgrenzung von Latein und Romanisch im Lessico Etimologico Italiano (LEI). In Wolfgang Dahmen (ed.), Latein und Romanisch: Romanistisches Kolloquium I. Tübingen: Narr, 323–38. Pfister, Max (1991). Gerhard Rohlfs e le colonie gallo-italiche nella Basilicata. In Nicola De Blasi, Paolo Di Giovine, and Franco Fanciullo (eds), Le parlate lucane e la dialettologia italiana (Studi in memoria di Gerhard Rohlfs). Galatina: Congedo, 91–106. Pfister, Max (1994). Indizi cronologici e linguistici del lessico gallo-italico siciliano nel contesto storico-culturale dell’alto Medioevo. In Migrazioni interne: i dialetti galloitalici della Sicilia. Padova: Unipress, 5–36. Pickett, Emily R., Blumstein, Sheila E., and Burton, Martha W. (1999). Effects of speaking rate on the singleton/geminate consonant contrast in Italian. Phonetica 56: 135–57. Pieri, Silvio (1893a). Il dialetto gallo-romano di Gombitelli nella provincia di Lucca. Archivio glottologico italiano 13: 309–28. Pieri, Silvio (1893b). Il dialetto gallo-romano di Sillano. Archivio glottologico italiano 13: 329–54. Pierrehumbert, Janet (1990). Phonological and phonetic representation. Journal of Phonetics 18: 375–94.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

276

References

Piffari, Sergio, and Tagliaferri, Sergio (2011). Scalvi: primo dizionario della lingua locale della Val di Scalve. Bergamo: Lubrina. Pifferi, Cristina (2004). Analisi sperimentale della durata vocalica nei dialetti di Balerna e Giornico. Lavoro di licenza, University of Zürich. Piras, Marco (1994). La varietà linguistica del Sulcis: fonologia e morfologia. Cagliari: Della Torre. Pirona, Jacopo (1871). Vocabolario friulano. Venice: Antonelli. Piroth, Hans Georg, and Janker, Peter M. (2004). Speaker-dependent differences in voicing and devoicing of German obstruents. Journal of Phonetics 32: 81–109. Pisani, Vittore (1975). Testi latini arcaici e volgari con commento glottologico, 3rd edn. Turin: Rosenberg & Sellier. Plangg, Guntram (1987). Phonemstruktur im Surselvischen (Tavetsch). In Guntram A. Plangg and Maria Iliescu (eds), Akten der Theodor Gartner-Tagung (Rätoromanisch und Rumänisch) in Vill/Insbruck. Innsbruck: Institut für Romanistik, 105–21. Politzer, Frieda N., and Politzer, Robert (1953). Romance Trends in 7th and 8th Century Latin Documents. Chapel Hill: University of North Carolina Press. Politzer, Robert (1951). On the chronology of the simplification of geminates in northern France. Modern Language Notes 66: 527–31. Pompino-Marschall, Bernd (2010). Einführung in die Phonetik. In Ludger Hoffmann (ed.), Sprachwissenschaft: Ein Reader. Berlin: de Gruyter, 405–30. Pons, Teofilo G., and Genre, Arturo (1997). Dizionario del dialetto occitano della val Germanasca. Alessandria: Edizioni dell’Orso. Porzio-Gernia, Maria Luisa (1976–7). Lo stato attuale degli studi di fonologia latina. Incontri linguistici 3: 137–52. Posner, Rebecca (1997). Linguistic Change in French. Oxford: Clarendon Press. Prehn, Maike (2012). Vowel quantity and the fortis–lenis distinction in North Low Saxon. Dissertation, University of Amsterdam. Utrecht: LOT. Prieto, Pilar (1992). Compensatory lengthening by vowel and consonant loss in early Friulian. Catalan Working Papers in Linguistics 2: 205–44. Prieto, Pilar (1994). Historical vowel lengthening in Romance: the role of sonority and foot structure. In Michael Mazzola (ed.), Issues and Theory in Romance Linguistics. Washington, DC: Georgetown University Press, 87–107. Prieto, Pilar (2000). Vowel lengthening in Milanese. In Repetti (2000: 255–72). Prince, Alan S. (1990). Quantitative consequences of rhythmic organization. In Deaton et al. (1990: 355–98). Prince, Alan S., and Smolensky, Paul (1993). Optimality Theory: constraint interaction in generative grammar. MS, Rutgers University. Malden, Mass.: Blackwell, 2004. Prokosch, Eduard (1933). An Outline of German Historical Grammar. New York: Oxford University Press. Prokosch, Eduard (1939). A Comparative Germanic Grammar. Philadelphia: Linguistic Society of America. Prosdocimi, Aldo L. M. (1986). Sull’accento latino e italico. In Annemarie Etter (ed.), o-o-pe-rosi. Festschrift für Ernst Risch zum 75. Geburtstag, Berlin: de Gruyter. Pulgram, Ernst (1950). Spoken and written Latin. Language 26: 458–66.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

277

Pulgram, Ernst (1965). The accentuation of Greek loans in spoken and written Latin. American Journal of Philology 86: 138–58. Pulgram, Ernst (1975). Latin–Romance Phonology: Prosodics and Metrics. Munich: Fink. Pult, Gaspard (1897). Le parler de Sent (Basse-Engadine). Lausanne: Payot. Quaderno fonetico del VSI, no. 63 [n.d.]. Comune di: Cortisciasca, Insome, Scareglia, Signora, Colla. MS, Bellinzona: Centro di dialettologia e di etnografia. Ramus, Franck, Nespor, Marina, and Mehler, Jacques (1999). Correlates of linguistic rhythm in the speech signal. Cognition 73: 265–92. Rebuffat, René (1987). Le poème de Q. Avidius Quintianus à la déesse Salus. Karthago 21: 93–105. Recasens i Vives, Daniel (1996). Fonètica descriptiva del català (Assaig de caracterització de la pronúncia del vocalisme i consonantisme del català al segle XX), 2nd edn. Barcelona: Institut d’Estudis Catalans. Renier, Léon (1855–8). Inscriptions romaines de l’Algérie. Paris: Imprimerie Impériale. Renwick, Margaret E. L. (2012). Vowels of Romanian: historical, phonological and phonetic studies. Ph.D dissertation, Cornell University. Repetti, Lori (1991). A moraic analysis of raddoppiamento fonosintattico. Rivista di linguistica 3: 307–30. Repetti, Lori (1992). Vowel length in northern Italian dialects. Probus 4: 155–82. Repetti, Lori (ed.) (2000). Phonological Theory and the Dialects of Italy. Amsterdam: Benjamins. Resnick, Melvyn C., and Hammond, Robert M. (1979). The status of quality and length in Spanish vowels. Linguistics 156: 79–88. REW = Meyer-Lübke, Wilhelm (1935). Romanisches Etymologisches Wörterbuch, 3rd edn. Heidelberg: Winter. Ricciardi, Joseph Sergio (1975). A brief phonology of three varieties of Ligurian Romance. Ph.D dissertation, Cornell University. Richter, Elise (1934). Beiträge zur Geschichte der Romanismen, 1: Chronologische Phonetik des Französischen bis zum Ende des 8. Jahrhunderts. Halle a.S.: Niemeyer. Rinaldi, Sascha (1985). Das Bergell: Tal des Übergangs. Bern: Francke. Ringe, Don (2003). Internal reconstruction. In Joseph and Janda (2003: 244–61). Ritt, Nikolaus (1994). Quantity Adjustment: Vowel Lengthening and Shortening in Early Middle English. Cambridge: Cambridge University Press. Rix, Helmut (1966). Die lateinische Synkope als historisches und phonologisches Problem. Kratylos 11: 156–65. Rizzolatti, Piera (1981). Elementi di linguistica friulana. Udine: Società filologica friulana. Rohlfs, Gerhard (1937). Sprachliche Berührungen zwischen Sardinien und Süditalien. In Paul Scheuermeier, Gerhard Rohlfs, Max L. Wagner, and Jakob Jud (eds), Donum natalicium Carolo Jaberg messori indefesso sexagenario oblatum a sodalibus Atlantis italico-helvetici. Zürich: Niehans, 27–75. Rohlfs, Gerhard (1938). Der Einfluß des Satzakzentes auf den Lautwandel. Archiv für das Studium der neueren Sprachen und Literaturen 174: 54–6. Rohlfs, Gerhard (1966–9). Grammatica storica della lingua italiana e dei suoi dialetti. 3 vols. Turin: Einaudi.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

278

References

Rolin, Gustav (1908). Die Mundart von Vasto in den Abruzzen. Prager deutsche Studien 8: 477–504 (Prague: Bellmann). Roncaglia, Aurelio (1971). La lingua d’oïl: avviamento allo studio del francese antico. Rome: Edizioni dell’Ateneo. Ronjat, Jules (1930–41). Grammaire istorique des parlers provençaux modernes. 4 vols. Montpellier: Société des langues romanes. Rossi, Giovanni Battista (2008). Vocabolario dei dialetti ladini e ladino-veneti dell’Agordino: lessico di Cencenighe, San Tomaso, Vallada, Canale d’Agordo, Falcade, Taibon, Agordo, La Valle, Voltago, Frassenè, Rivamonte, Gosaldo, 2nd edn. Belluno: Istituto bellunese di ricerche sociali e culturali. Rossini, Giorgio (1975). Capitoli di morfologia e sintassi del dialetto cremonese. Florence: La Nuova Italia. Ruggeri, Santo (1970). Dizionario bresciano italiano. Brescia: Tipografia pavoniana. Rupp, Theodor (1963). Lautlehre der Mundarten von Domat, Trin und Flem: zur Bestimmung der Lautgrenzen am Flimser Wald und beim Zusammenfluss des Vorder- und Hinterrhein. Dissertation, University of Zürich. Saiu, Emanuele (2012). Isole linguistiche in provincia di Lucca: Cológnora in Valleriana e Gombitelli. MS, Scuola Normale Superiore di Pisa. Saiu, Emanuele (in preparation). Il dialetto di Cológnora in Valleriana. Ph.D. dissertation, Scuola Normale Superiore di Pisa. Sala, Marius (1970). Contribuţii la fonetica istorică a limbii române. Bucharest: Editura Academiei Republicii Socialiste România. Salmons, Joseph C., and Dubenion-Smith, Shannon (eds) (2007). Historical Linguistics 2005: Selected Papers from the 17th International Conference on Historical Linguistics. Amsterdam: Benjamins. Saltarelli, Mario (1970a). A Phonology of Italian in a Generative Grammar. The Hague: Mouton. Saltarelli, Mario (1970b). La grammatica generativa trasformazionale: con introduzione alla fonologia, sintassi e dialettologia italiana. Florence: Sansoni. Saltarelli, Mario (1983). The mora unit in Italian phonology. Folia Linguistica 17: 7–24. Salvi, Giampaolo (1997). Ladin. In Maiden and Parry (1997: 286–94). Salvioni, Carlo (1884). Fonetica del dialetto moderno della città di Milano. Turin: Loescher. Repr. Sala Bolognese: Forni, 1979. Salvioni, Carlo (1907). Lingua e dialetti della Svizzera italiana. Rendiconti dell’Istituto lombardo di scienze e lettere 40 (2nd series): 719–36. Salvioni, Carlo (1911). Osservazioni sull’antico vocalismo milanese desunte dal metro e dalla rima del cod. berlinese di Bonvesin. In Studi letterari e linguistici dedicati a Pio Rajna. Florence: Ariani, 367–88. Repr. in Salvioni (2008: 3.157–78). Salvioni, Carlo (1917). Ladinia e Italia. Rendiconti dell’Istituto lombardo di scienze e lettere 50 (2nd series): 41–78. Repr. in Salvioni (2008: 2.406–45). Salvioni, Carlo (2008). Scritti linguistici, ed. Michele Loporcaro, Lorenza Pescia, Romano Broggini, and Paola Vecchio. 5 vols. Bellinzona: Edizioni dello Stato del Cantone Ticino. Sampson, Rodney (1980). On the history of final vowels from Latin to Old French. Zeitschrift für romanische Philologie 96: 23–48.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

279

Sampson, Rodney (2010). Vowel Prosthesis in Romance: A Diachronic Study. Oxford: Oxford University Press. Sánchez Miret, Fernando (2001). Proyecto de gramática histórica y comparada de las lenguas romances. 2 vols. Munich: Lincom Europa. Sanga, Glauco (1979). Dialetto e folklore. Ricerca a Cigole. Milan: Silvana. Sanga, Glauco (1984a). La tensione nei dialetti lombardi. In Dialettologia lombarda: lingue e culture popolari. Pavia: Aurora, 45–67. Sanga, Glauco (1984b). Premana: note di storia linguistica minore. In Dialettologia lombarda: lingue e culture popolari. Pavia: Aurora, 69–127. Sanga, Glauco (ed.) (1987a). Lingua e dialetti di Bergamo e delle valli. 3 vols. Bergamo: Lubrina. Sanga, Glauco (1987b). Introduzione. Bergamo tra Venezia e Milano. In Sanga (1987a: 17–35). Sanga, Glauco (1987c). Fonetica storica del dialetto di Bergamo. In Sanga (1987a: 37–63). Sanga, Glauco (1988). La lunghezza vocalica nel milanese e la coscienza fonologica dei parlanti. Romance Philology 41: 290–97. Sanga, Glauco (1993). Il Breviarium attribuito a Bonaventura da Bergamo e il problema del lombardo orientale antico. Atti del Sodalizio glottologico milanese 31: 187–98. Sanga, Glauco (1997a). La lunghezza vocalica nel milanese e la coscienza fonologica dei parlanti. Romance Philology 41: 290–97. Sanga, Glauco (1997b). Lombardy. In Maiden and Parry (1997: 253–59). Sanga, Glauco (1999). Il dialetto di Milano. RID 23: 137–64. Santen, Jan van and D’Imperio, Mariapaola (1999). Positional effects on stressed vowel duration in standard Italian. In John J. Ohala, Yoko Hasegawa, Manjari Ohala, Daniel Granville, and Ashlee C. Bailey (eds), Proceedings of the XIVth International Congress of Phonetic Sciences. University of California: Linguistics Dept, 241–4. Santerre, Laurent (1971). Les voyelles orales dans le français parlé à Montréal. Doctoral thesis, Université des sciences humaines, Strasbourg. Saracino, Giacomo (1957). Lessico dialettale bitontino: nuova edizione con annotazioni etimologiche a cura del dott. Vincenzo Valente. Bari: STOS. Saussure, Ferdinand de (1922). Cours de linguistique générale, 2nd edn. Paris: Payot. Critical edn by Tullio De Mauro, 1979. Savoia, Leonardo M. (1987). Teoria generativa, modelli fonologici e dialettologia: la propagazione di u in una varietà lucana. Rivista italiana di dialettologia 11: 185–263. Savoia, Leonardo M. (1990). Proprietà vocaliche e struttura metrico-sillabica in alcune varietà abruzzesi. In Studi in memoria di E. Giammarco. Pisa: Giardini, 331–80. Savoia, Leonardo M. (1997). Il vocalismo a tre gradi dell’area calabro-lucana. In Amalia Catagnoti (ed.), Studi linguistici offerti a Gabriella Giacomelli dagli amici e dagli allievi. Padua: Unipress, 363–75. Savy, Renata, and Crocco, Claudia (eds) (2006). Analisi prosodica: teorie, modelli e sistemi di annotazione. Torriana (RN): EDK. Scaramella, Giovanni (1986). Nuovo vocabolario ortografico bresciano: Bresciano–italiano, italiano–bresciano. Montichiari: Zanetti. Schaad, Giacomo (1936). Terminologia rurale in Val Bregaglia. Bellinzona: Arti Grafiche Arturo Salvioni & Co. Schaad, Giacomo (1939). I nomi popolari della flora prativa in Val Bregaglia. Vox Romanica 4: 48–64.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

280

References

Schädel, Bernhard (1903). Die Mundart von Ormea. Halle a. S.: Niemeyer. Schuchardt, Hugo (1866–8). Der Vokalismus des Vulgärlateins. 3 vols. Leipzig: Teubner. Schuchardt, Hugo (1870[1900]). Über die Klassifikation der romanischen Mundarten. Graz: Universitäts-Buchdruckerei Styria. Schuchardt, Hugo (1918). Die romanischen Lehnwörter im Berberischen. Vienna: Hölder. Schürr, Friedrich (1918, 1919). Romagnolische Dialektstudien. 2 vols: Lautlehre alter Texte (1918); Lautlehre Lebender Mundarten (1919). Vienna: Hölder. Schürr, Friedrich (1936). Umlaut und Diphthongierung in der Romania. Romanische Forschungen 50: 275–316. Schürr, Friedrich (1956). La diphthongaison romane. Revue de linguistique romane 20: 107–44, 161–248. Schürr, Friedrich (1965). Grundsätzliches zu den Fragen der romanischen, insbesondere italienischen Diphthongierung. Archiv für das Studium der neueren Sprachen und Literaturen 201: 321–39. Schürr, Friedrich (1970). La diphthongaison romane. Tübingen: Narr. Schürr, Friedrich (1972). Epilogo alla discussione sulla dittongazione romanza. Revue de linguistique romane 36: 311–21. Scuffi, Sergio (2005). Nü ’n cuštümáva: vocabolario dialettale di Samolaco. Note sul dialetto del ‘brí’, scene di vita, detti tipici, sentenze, proverbi. In Remo Bracchi, Profilo del dialetto di Samolaco. Sondrio: Istituto di dialettologia e di etnografia valtellinese e valchiavennasca. Seelmann, Emil (1885). Die Aussprache des Latein nach physiologisch-historischen Grundsätzen. Heilbronn: Henninger. Seelmann, Emil (1890). Volkslatein. Kritischer Jahresbericht über die Fortschritte der Romanischen Philologie 1: 38–64. Seidl, Christian (1994). Gemeinsabellisch und Vulgärlateinisch: der Vokalismus. In George E. Dunkel, Gisela Meyer, Salvatore Scarlata, and Christian Seidl (eds), Früh-, Mittel-, Spätindogermanisch. Akten der IX. Fachtagung der Indogermanischen Gesellschaft vom 5. bis 9. Oktober 1992 in Zürich. Wiesbaden: Reichert, 349–70. Seidl, Christian (1995). Le système acasuel des protoromans ibérique et sarde: dogmes et faits. Vox Romanica 54: 41–73. Seidl, Christian (1998). Lingua latina in bocca italica? Uno sguardo critico dal punto di vista del sostrato. In Giovanni Ruffino (ed.), Atti del XXI Congresso di linguistica e filologia romanza, vol. 1: Grammatica storica delle lingue romanze. Tübingen: Niemeyer, 371–83. Seiler, Guido (2005). Open syllable shortening in Bernese German. In Rebecca T. Cover and Yuni Kim (eds), Proceedings of the 31st Annual Meeting of the Berkeley Linguistics Society: General Session and Parasession on Prosodic Variation and Change. Berkeley: Berkeley Linguistics Society, 477–88. Seiler, Guido (2009). Sound change or analogy? Monosyllabic lengthening in German and some of its consequences. Journal of Comparative Germanic Linguistics 12: 229–72. Seklaoui, Diana R. (1989). Change and Compensation: Parallel Weakening of [s] in Italian, French and Spanish. Frankfurt a.M.: Lang. Serbat, Guy (1975). Les dérivés nominaux latins à suffixe médiatif. Paris: Sorbonne. Sganzini, Silvio (1924–6). Fonetica dei dialetti della Val Leventina. L’Italia dialettale 1: 190–212; 2: 100–155.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

281

Sibille, Jean (ed.) (2007). Marcellin Richard, La Passion de Saint André. Édition critique suivie d’une étude linguistique comparée. Paris: Champion. Sihler, Andrew L. (1995). New Comparative Grammar of Greek and Latin. Oxford: Oxford University Press. Simone, Raffaele, Vignuzzi, Ugo, and Ruggiero, Giulianella (eds) (1976). Studi di fonetica e fonologia. Rome: Bulzoni. Smith, Colin (1983). Vulgar Latin in Roman Britain. In Temporini and Haase (1983: 893–948). Soffietti, James P. (1949). Phonemic Analysis of the Word in Turinese. New York: Columbia University. Solèr, Clau (1991). Romanisch im Schams. Zürich: Phonogrammarchiv der Universität. Sommer, Ferdinand, and Pfister, Raimund (1977). Handbuch der lateinischen Laut- und Formenlehre, 1: Einleitung und Lautlehre. Heidelberg: Winter. Spagnoletti, Carmela, and Dominicy, Marc (1992). L’accent italien et la cliticisation de la terminaison verbale -no. Revue québecoise de linguistique 21: 9–32. Spence, Nicol C. W. (1965). Quantity and quality in the vowel system of Vulgar Latin. Word 21: 1–18. Spoerri, Teofilo (1918). Il dialetto della Valsesia. Rendiconti dell’Istituto lombardo di scienze e lettere 51 (2nd series): 391–409, 683–98, 732–52. Spore, Palle (1972). La diphtongaison romane. Odense: Odense University Press. Stampa, Gian Andrea (1934). Der Dialekt des Bergell, 1: Phonetik. Aarau: Sauerländer. Stampa, Gian Andrea (1939). Due testi bregagliotti con alcune considerazioni d’ordine foneticoproposizionale. Vox Romanica 4: 270–87. Stampe, David (1979). A Dissertation on Natural Phonology. Bloomington, Ind.: IULC. Stasse, Simon (2007). Dictionnaire populaire de wallon liégeois. Aywaille: Société Royale Littéraire ‘La Wallonne’. Stevens, Mary (2011). Consonant length in Italian: gemination, degemination and preaspiration. In Scott M. Alvord (ed.), Selected Proceedings of the 5th Conference on Laboratory Approaches to Romance Phonology. Somerville, Mass.: Cascadilla, 21–32. Stevens, Mary (2012). A Phonetic Investigation into ‘raddoppiamento sintattico’ in Sienese Italian. Frankfurt a.M.: Lang. Stok, Fabio (1997). Appendix Probi IV. Naples: Arte tipografica. Straka, Georges (1953). Observations sur la chronologie et les dates de quelques modifications phonétiques en roman et en français prélittéraire. Revue des langues romanes 71: 247–307. Straka, Georges (1956). La dislocation linguistique de la Romania et la formation des langues romanes à la lumière de la chronologie relative des changements phonétiques. Revue de linguistique romane 20: 249–67. Repr. in Straka (1979: 193–211). Straka, Georges (1959). Durée et timbre vocaliques: observations de phonétique générale appliquées à la phonétique historique des langues romanes, Zeitschrift für Phonetik und allgemeine Sprachwissenschaft 12: 276–300. Repr. in Straka (1979: 167–91). Straka, Georges (1964). L’évolution phonétique du latin au français sous l’effet de l’énergie, et de la faiblesse articulatoire. Travaux de linguistique et de littérature 2: 17–98. Repr. in Straka (1979: 213–94). Straka, Georges (1979). Les sons et les mots: choix d’études de phonétique et de linguistique. Paris: Klincksieck. Straka, Georges (1981). Sur la formation de la prononciation française d’aujourd’hui. Travaux de linguistique et de littérature 19: 161–248.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

282

References

Sturtevant, Edgar H. (1968). The pronunciation of Greek and Latin, 2nd edn. Groningen: Boumas Boekhuis. Stussi, Alfredo (1965). Sui fonemi del dialetto veneziano antico. L’Italia dialettale 28: 125–42. Sumien, Domergue (2006). La standardisation pluricentrique de l’occitan: nouvel enjeu sociolinguistique, développement du lexique et de la morphologie. Turnhout: Brépols. Tantillo, Ignazio, and Bigi, Francesca (eds) (2010). Leptis Magna: una città e le sue iscrizioni in epoca tardoromana. Cassino: Edizioni dell’Università. Taverdet, Gérard (1989). Le français régional parlé en Bourgogne, 2nd edn. Dijon: CRDP. Tavoni, Mirko (1984). Latino, grammatica, volgare: storia di una questione umanistica. Padua: Antenore. Tavoni, Mirko (1992). Il Quattrocento. Bologna: Il Mulino. Tekavčić, Pavao (1972). Grammatica storica della lingua italiana, I: Fonematica. Bologna: Il Mulino. Telmon, Tullio, and Ferrier, Consuelo (2007). Le minoranze linguistiche piemontesi nel 2006. In Enrico Allasino, Consuelo Ferrier, Sergio Scamuzzi, and Tullio Telmon (eds), Le lingue del Piemonte. Turin: Quaderni di ricerca 113, IRES Istituto Ricerche Economico Sociali, Regione Piemonte, Assessorato alla Cultura, 7–60. Temporini, Hildegard, and Haase, Wolfgang (eds) (1983). Aufstieg und Niedergang der römischen Welt, II: Principat. Berlin: de Gruyter. ThLL = Thesaurus linguae Latinae. Editus auctoritate et consilio academiarum quinque Germanicarum Berolinensis Gottingensis Lipsiensis Monacensis Vindobonensis. Leipzig: Teubner 1900ff. Thöni, Gion Peder (1969). Rumantsch–Surmeir: grammatica per igl idiom surmiran. Coira: Ligia Romontscha. Thornton, Anna M. (1999). Diagrammaticità, uniformità di codifica e morfomicità nella flessione verbale italiana. In Paola Benincà, Alberto M. Mioni, and Laura Vanelli (eds), Fonologia e morfologia dell’italiano e dei dialetti d’Italia. Rome: Bulzoni, 483–502. Thurot, Charles (1881–3). De la prononciation française depuis le commencement du XVIe siècle, d’après le témoignage des grammairiens. 3 vols. Paris: Imprimerie nationale. Tiraboschi, Antonio (1873). Vocabolario dei dialetti bergamaschi antichi e moderni, 2nd edn. Bergamo: Fratelli Bolis. Toso, Fiorenzo (1997a). Grammatica del genovese: varietà urbana e di koinè. Recco: Le Mani. Toso, Fiorenzo (1997b). Storia linguistica della Liguria, 2nd edn. Recco: Le Mani. Toso, Fiorenzo (2005). Grammatica del tabarchino. Recco: Le Mani. Toso, Fiorenzo (2008a). Aspetti del bonifacino in diacronia. Bollettino di studi sardi 1: 147–77. Repr. in Toso (2008b: 37–63). Toso, Fiorenzo (2008b). Linguistica di aree laterali ed estreme: contatto, interferenza, colonie linguistiche e isole culturali nel Mediterraneo occidentale. Recco: Le Mani. Traina, Antonino (1868). Nuovo vocabolario siciliano–italiano. Palermo: Pedone Lauriel. Repr. Milan: SAS, 1991. Trimeloni, Giuseppe (1942–3). Il dialetto di Malcesine. Graduation thesis, University of Padua. Trimeloni, Giuseppe (1995). Dizionario etimologico del dialetto di Malcesine. Malcesine: Comitato del Museo Castello Scaligero di Malcesine. Tristano, Caterina (2006). Cultura grafica ad Arezzo agli albori dello ‘Studium’. In Francesco Stella (ed.), 750 anni degli Statuti universitari aretini. Florence: SISMEL, 245–72.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

283

Trubetzkoy, Nikolaj S. (1939). Grundzüge der Phonologie. Prague: Cercle Linguistique de Prague. Trumper, John (1972). Il gruppo dialettale padovano-polesano. Padua: Rebellato. Trumper, John (1975). Obiezioni sistematiche all’uso dei tratti teso/rilassato nell’analisi di sistemi vocalici e di rotazioni vocaliche. Lingua e contesto 1(2): 1–86. Tuttle, Edward F. (1981–2). Un mutamento linguistico e il suo inverso: l’apocope nell’Alto Veneto. Rivista italiana di dialettologia 5: 15–35. Tuttle, Edward F. (1991). Nasalization in northern Italy: syllabic constraints and strength scales as developmental parameters. Rivista di linguistica 3: 23–92. Uguzzoni, Arianna (1971). Quantità fonetica e quantità fonematica nell’area dialettale frignanese. L’Italia dialettale 34: 313–33. Uguzzoni, Arianna (1974). Sulla struttura della parola dei dialetti emiliani: aspetti sincronici e aspetti diacronici di un problema. Deputazione di storia patria per le antiche provincie modenesi (Modena: Aedes Muratoriana): 239–52. Uguzzoni, Arianna (1975). Appunti sull’evoluzione del sistema vocalico di un dialetto frignanese. L’Italia dialettale 38: 47–76. Uguzzoni, Arianna (2000). Aspetti -emici e aspetti -etici del fenomeno della ‘quantità vocalica’ nei dialetti dell’Italia settentrionale. In Donatella Locchi (ed.), Atti delle X giornate di studio del Gruppo di fonetica sperimentale. Naples: Istituto universitario orientale, 227–35. Uguzzoni, Arianna, Azzaro, Gabriele, and Schmid, Stephan (2003). Short vs. long and/or abruptly cut vowels: new perspectives on a debated question. In Daniel Recasens (ed.), Proceedings of the XVth International Congress of the Phonetic Sciences, vol. 3. Barcelona: Universitat Autònoma de Barcelona, 2717–20. Uguzzoni, Arianna, and Busà, M. Grazia (1995). Correlati acustici della opposizione di quantità vocalica in area emiliana. Rivista italiana di dialettologia 19: 7–39. Urech, Jakob (1946). Beitrag zur Kenntnis der Mundart der Val Calanca. Biel: Graphische Anstalt Schüler. Vaan, Michiel de (2008). Etymological Dictionary of Latin and the Other Italic Languages. Leiden: Brill. Väänänen, Veikko (1965). Etude sur le texte et la langue des Tablettes Albertini. Helsinki: Suomalainen Tiedeakatemia. Väänänen, Veikko (1966). Le latin vulgaire des inscriptions pompéiennes, 3rd enlarged edn. Berlin: Akademie. Väänänen, Veikko (1981). Introduction au latin vulgaire, 3rd edn. Paris: Klincksieck. 1st edn. Helsinki: Imprimerie de la Société de littérature finnoise, 1937. Vanelli, Laura (1979). L’allungamento delle vocali in friulano. Ce fastu? 55: 66–76. Vanelli, Laura (1998). Le vocali lunghe del friulano. Quaderni della grammatica friulana di riferimento 1: 69–108. Repr. in Benincà and Vanelli (2005: 159–98). Varvaro, Alberto (1983). Sulla nozione di area isolata: il caso della Lucania. In Federico Albano Leoni, Daniele Gambarara, Franco Lo Piparo, and Raffaele Simone (eds), Italia linguistica: idee, storia, strutture. Bologna: Il Mulino, 149–66. Repr. in Alberto Varvaro, La parola nel tempo: lingua, società, storia (Bologna: Il Mulino, 1984), 127–44. Varvaro, Alberto (1984). Omogeneità del latino e frammentazione della Romània. In Vineis (1984a: 11–22).

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

284

References

Varvaro, Alberto (2000). Le notizie di Paolo Pompilio sulla diffusione delle lingue romanze nel Quattrocento. Romanische Forschungen 112: 485–99. Vayra, Mario (1994). Phonetic explanations in phonology: laryngealization as the case for glottal stops in Italian word-final stressed syllables. In Dressler et al. (1994: 275–93). Vázquez Obrador, Jesús (2009). De fonética histórica aragonesa: a propósito de la evolución de -nn- y -nd- en la toponimia de localidades de la cuenca alta del río Gállego (Huesca). Zeitschrift für romanische Philologie 125: 243–72. Velázquez Soriano, Isabel (2004). Las pizarras visigodas: entre el latín y su disgregación. Burgos: Instituto Castellano y Leonés de la Lengua. Vennemann, Theo (1988). Preference Laws for Syllable Structure and the Explanation of Sound Change. Berlin: Mouton de Gruyter. Vennemann, Theo (2000). From quantity to syllable cuts: on so-called lengthening in the Germanic languages. Rivista di linguistica/Italian Journal of Linguistics 12: 251–82. Verner, Karl (1877). Eine Ausnahme der ersten Lautverschiebung. Zeitschrift für vergleichende Sprachforschung auf dem Gebiete der indogermanischen Sprachen 23: 97–130. Vicario, Federico (1998). Il quaderno di Odorlico da Cividale: contributo allo studio del friulano antico. Udine: Forum. Vicario, Federico (ed.) (2007). Ladine loqui: IV colloquium retoromanistich. Udine: Società filologica friulana. Videsott, Paul (2001). Vokallängen im Norditalienischen und im Dolomitenladinischen. In Peter Wunderli, Iwar Werlen, and Matthias Grünert (eds), Italica–Raetica–Gallica: Studia linguarum litterarum artiumque in honorem Ricarda Liver. Tübingen: Francke, 151–68. Videsott, Paul, and Plangg, Guntram A. (1998). Ennebergisches Wörterbuch. Ennebergischdeutsch, mit einem rückläufigen Wörterbuch und einem deutsch-ennebergischen. Innsbruck: Universitätsverlag Wagner. Vignola Saffirio, Vittoria Carla (1978). Spunti per un’analisi sincronica della parlata di Trausella (Alto Canavese). In Gianrenzo P. Clivio and Giuliano Gasca Queirazza (eds), Lingue e dialetti nell’arco alpino occidentale. Turin: Centro studi piemontesi, 105–15. Vincent, Nigel (1988). Non-linear phonology in diachronic perspective: stress and wordstructure in Latin and Italian. In Pier Marco Bertinetto and Michele Loporcaro (eds), Certamen phonologicum: Papers from the 1987 Cortona Phonology Meeting. Turin: Rosenberg & Sellier, 421–32. Vineis, Edoardo (ed.) (1984a). Latino volgare, latino medioevale, lingue romanze: atti del convegno della Società italiana di glottologia. Pisa: Giardini. Vineis, Edoardo (1984b). Problemi di ricostruzione della fonologia del latino volgare. In Vineis (1984a: 45–62). Vineis, Edoardo (1993). Preliminari per una storia del latino parlato. In Friedrich Stolz, Albert Debrunner, and Wolfgang P. Schmid (trans. Carlo Benedikter), Storia della lingua latina, 4th edn. Bologna: Pàtron, xxxvii–lviii. Virdis, Maurizio (1978). Fonetica storica del dialetto campidanese. Cagliari: Della Torre. Virdis, Maurizio (1988). Sardo: aree linguistiche. In LRL, vol. 4: Italienisch, Korsisch, Sardisch, 897–913. Vitale, Maurizio (1953). Le origini del volgare nelle discussioni dei filologi del ’400. Lingua nostra 14: 64–9.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

References

285

Vitali, Daniele (2007). Il dialetto di Porretta Terme. Nuèter 65: 52–8. Vitali, Daniele, and Rulli, Ettore (2010). Grammatica del dialetto di Compiano: parlata ligure della Valle del Taro. Compiano: Compiano Arte e Storia/Rupe Mutevole. Viti, Carlotta (2013). Die Rekonstruktion der urindogermanischen Phono-Morphologie am Beispiel der Betonungsparadigmen. MS, University of Zürich. Vogel, Irene (1978). Raddoppiamento as a resyllabification rule. Journal of Italian Linguistics 3: 15–28. Vogel, Irene (1982). La sillaba come unità fonologica. Bologna: Zanichelli. Wachter, Rudolf (1987). Altlateinische Inschriften: sprachliche und epigraphische Untersuchungen zu den Dokumenten bis etwa 150 v. Chr. Bern: Lang. Wagner, Max Leopold (1941). Historische Lautlehre des Sardischen. Halle a.S.: Niemeyer. Walberg, Emanuel (1907). Saggio sulla fonetica del parlare di Celerina-Cresta (Alta Engadina). Lund: Ohlsson. Walter, Henriette (1976). La dynamique des phonèmes dans le lexique français contemporain. Paris: France-Expansion. Walter, Henriette (1982). Enquête phonologique et variétés régionales du français. Paris: PUF. Walter, Henriette (1993). L’évolution phonologique du français depuis le XVIIIe siècle. In Jürgen Schmidt-Radefeldt and Andreas Harder (eds), Sprachwandel und Sprachgeschichte: Festschrift für Helmut Lüdtke zum 65. Geburtstag. Tübingen: Narr, 211–20. Wanner, Dieter (1971). Phonematik eines venezianischen Idiolektes. Zürich: Juris. Wanner, Dieter, and Kibbee, Douglas A. (eds) (1991). New Analyses in Romance Linguistics. Amsterdam: Benjamins. Wartburg, Walther von (1928–2002). Französisches etymologisches Wörterbuch: eine Darstellung des galloromanischen Sprachschatzes. 25 vols. Bonn: Klopp [subsequently Bâle: Zbinden Druck und Verlag. Ed. Otto Jänicke 1972–8, Margaretha Hoffert 1976, Carl Theodor Gossen 1979–83, Jean-Pierre Chambon 1986–90, Jean-Paul Chauveau 1997]. Wartburg, Walther von (1936). Die Ausgliederung der romanischen Sprachräume. Zeitschrift für romanische Philologie 56: 1–48. Wartburg, Walther von (1950). Die Ausgliederung der romanischen Sprachräume. Bern: Francke. Weber Wetzel, Elena (2002). Il dialetto di Casale Corte Cerro: contributo alla conoscenza delle parlate del Cusio. Alessandria: Edizioni dell’Orso. Wedding, Georgius (1902). De vocalibus productis Latinas voces terminantibus. In Adalbert Bezzenberger and Walther Prellwitz (eds), Beiträge zur Kunde der indogermanischen Sprachen, vol. 27. Göttingen: Vandenhoek & Ruprecht, 1–62. Weinrich, Harald (1958). Phonologische Studien zur romanischen Sprachgeschichte. Münster Westfalen: Aschendorff. Weinrich, Harald (1960). Sonorisierung in der Kaiserzeit? Zeitschrift für romanische Philologie 76: 205–18. Weiss, Michael L. (2009). Outline of the Historical and Comparative Grammar of Latin. Ann Arbor, Mich.: Beech Stave Press. WH: Alois Walde and Johann Baptist Hofmann (1965). Lateinisches etymologisches Wörterbuch. 4th edn. 2 vols. Heidelberg: Winter.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

286

References

Windisch, Rudolf (1998). Die historische Klassifikation der Romania, 2: Balkanromanisch. In LRL, vol. 7: Kontakt, Migration und Kunstsprachen: Kontrastivität, Klassifikation und Typologie, 907–37. Winteler, Jost (1876). Die Kerenzer Mundart des Kantons Glarus in ihren Grundzügen dargestellt. Leipzig: Winter. Wright, Roger (1982). Late Latin and Early Romance in Spain and Carolingian France. Liverpool: Cairns. Wright, Roger (2002). A Sociophilological Study of Late Latin. Turnhout: Brepols. Wright, Roger (2003). A sociophilological approach to the earliest Romance texts: [-t], -/t/ and -t in Castile. In Michèle Goyens and Werner Verbeke (eds), The Dawn of the Written Vernacular in Western Europe. Leuven: Leuven University Press, 201–14. Also in Wright (2002: 282–94). Wright, Roger (ed.) (2008). Actes du VIIIe colloque international sur le latin vulgaire et tardif. Hildesheim: Olms-Weidmann. Wüest, Jakob (1979). La dialectalisation de la Galloromania. Bern: Francke. Yamamoto, Shinji (1991). Alcuni ampliamenti dei casi dell’allungamento vocalico nel friulano. In Laura Vanelli and Alberto Zamboni (eds), Per Giovan Battista Pellegrini: scritti degli allievi padovani. Padva: Unipress, 645–55. Zamboni, Alberto (1974). Veneto. Pisa: Pacini. Zamboni, Alberto (1976). Alcune osservazioni sull’evoluzione delle geminate romanze. In Simone et al. (1976: 325–36). Zamboni, Alberto (1980). Elementi extralinguistici nella definizione dell’italia dialettale. In Edoardo Vineis (ed.), Per la storia e la classificazione dei dialetti italiani: atti del convegno della Società italiana di glottologia. Pisa: Giardini, 79–101. Zamboni, Alberto (1984). I dialetti cadorini. In Giovan Battista Pellegrini and Sergio Sacco (eds), Il ladino bellunese. Belluno: Istituto bellunese di ricerche sociali e culturali, 45–83. Zamboni, Alberto (1988). Veneto. In LRL, vol. 4: Italienisch, Korsisch, Sardisch, 517–38. Zamboni, Alberto (2000). Alle origini dell’italiano: dinamiche e tipologie della transizione dal latino. Rome: Carocci. Zamora Vicente, Alonso (1967). Dialectología española, 2nd edn. Madrid: Gredos. Zec, Draga (2011). Quantity sensitivity. In Marc van Oostendorp, Colin J. Ewen, Elizabeth V. Hume, and Keren Rice (eds), The Blackwell Companion to Phonology, vol. 2. Malden, Mass.: Wiley-Blackwell, 1335–61. Ziccardi, Giovanni (1910). Il dialetto di Agnone: la fonetica e la flessione. Zeitschrift für romanische Philologie 34: 405–36. Zingarelli, Nicola (1899). Il dialetto di Cerignola. Archivio glottologico italiano 15: 83–96, 226–35. Zink, Gaston (1999). Phonétique historique du français, 6th edn. Paris: PUF. Zörner, Lotte (1998). I dialetti canavesani di Cuorgné, Forno e dintorni: descrizione fonologica, storico-fonetica e morfologica. Cuorgné: CORSAC. Zufferey, François (2008). Histoire interne de l’occitan. In Ernst et al. (2008: 2998–3020). Zwicky, Arnold (1974). Taking a false step. Language 50: 215–24.

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of languages Languages and dialects are identified either with the glossonym (language name, e.g. Badiotto = the Ladin dialect of Val Badia) or with the place name (e.g. Augio = the Alpine Lombard dialect of Augio, Val Calanca), especially in cases where the glossonym is not used as widely or has no univocal denotation (e.g. Marche, for the dialects of the Marche region, that does not possess one ‘Marchigiano’ dialect). Some languages are subdivided geographically (e.g., under Latin one finds Latin of Africa, of Britannia etc.), whereas none are subdivided diachronically (thus, Latin encompasses all references to Archaic, Classical, and Late Latin). Abruzzese 212 Adriatic coast 28–9, 54, 63, 148 Agnonese 117 Agordino 109, 112 Agri 56 Airolese 152 Albanian 3, 55 Altamurano 117, 218–19 Ambialet 64, 106, 180 Ampezzano 109 Andalusian 63, 64 Andeer 115 Andriese 28 Apennine Emilian 171, 174, 189, 192 Apennine Tuscan 196 Appalachian English 131 Appennino Pistoiese 185 Aprica 152 Apulian 28, 55, 117, 218 Arabba 111 Arabic 195 Augio 179, 182 Badiotto 109, 111, 113 Balerna 97 Bantu 35 Béarnais 75, 189 Belsetán 196 Benedello 87–8, 174, 191–2, 217 Berber 48 Bergamasco 154–5, 157, 163, 168, 172, 224

Bergün see Bravuogn/Bergün Bernese German 239 Bienate 184 Biscegliese 117, 212 Bitontino 28, 117, 212 Bolognese 84, 87–9, 174, 199, 20–7, 214, 217–21, 223 Bonifacino 149–50 Bonorvese 27 Borgo Sansepolcro see Sansepolcro, Borgo Borgomanero 182 Bramans 174 Bravuogn/Bergün 174, 180 Bregagliotto 174–7, 180–1, 183 Breil 106–7, 199 Bresciano 140, 155–7, 162–3, 168 Brianza 184 Briga/La Brigue 106–7, 149, 199 British (Celtic) 60 Cadorino 109, 111–12 Cagliari 56 Cairese 150 Calabrian 118 Calanchino 174–5, 179–82 Calasetta 150 Campanian 118 Campertogno 151 Campidanese see Sardinian Campo Val Tàrtano 153 Canavese 151

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

288

Index of languages

Canazei 111 Carloforte 150 Carnico 98–9, 115, 236 Carrarese 185 Casale Corte Cerro 96, 151, 161, 216 Castello di Sambuca Pistoiese 182, 202–3, 220–1, 223 Castelmezzano 54 Castilian 82, 195 Castionetto 152 Catalan 26, 63–4, 80, 82, 165, 195, 197 Cavanna 220–3 Celerina/Schlarigna 173 Celtic 5–6, 60 Cerignolese 28, 212 Cersòsimo 57 Certoux 108, 181 Chamorro (Austronesian; Guam and Northern Mariana Islands) 211, 215 Chiuro 152 Chiusaforte 99 Choctaw (Muskogean, SE USA) 215 Collagna 174 Cológnoro 166, 199–203, 216, 219, 221 Comano 97 Comelico 112–13, 184 Compiano 90–1 Constantinople 45–6 Coptic 45 Corsican 56–7, 136, 149 Corticiasca 152 Cremonese 61, 64, 83–7, 111, 114, 119, 122–3, 126, 132–3, 140, 142–3, 145–6, 155, 160, 165, 172, 194, 205, 207–8, 215, 222, 228, 239 Creole languages 58 Crocette 174, 191–2 Cuban 63 Cuorgnè 151 Daco-Romance 29, 54, 64, 75–7, 79–80 Dalmatian 29, 55 Danish 190 Dolomitan 109

Domat 174 Dutch 238 Emilian 83, 87–9, 111, 114, 119, 123, 145, 157, 163–6, 171–5, 184, 186, 189–92, 200, 202, 206–8, 210, 212, 216–17, 219–20, 226, 233, 235 Emilia-Romagna 123, 214 Engadinian 12, 29, 113–15, 173–4, 203, 225, 233–5, 237 English 1, 46, 53, 70, 76, 131, 143, 169, 211, 231, 238 Estonian 1 Etruscan 6 Fanese 196–7 Fanghetto 107 Fanzolo Barcón 198 Fassano 109–11, 218 Feltrino 163 Finnish 73, 190 Fiorenzuola 119, 212, 217, 222 Flem 174 Florentine 30, 202, 221 Fodóm see Livinallese/Fodom Forno 151 Franconian (German) 1 Franco-Provençal 66, 80, 85, 104–6, 108, 146, 174, 181–2, 184–5, 195, 226 Frankish 30 French 3, 6, 12–13, 26, 28–30, 32, 36–7, 39, 50–1, 72–5, 77–82, 93, 101–6, 108, 110, 115, 121, 125, 146–7, 150, 164–5, 190, 195, 203–5, 207, 218–19, 226, 228, 232, 234 Burgundy regional French 72 regional French of the Franco-Provençal area 101, 108 Québecois 72 Wallonia regional French 102 Frignano 89 Friulian 15–16, 82, 87, 97–100, 108–9, 111–12, 114–15, 119, 122–3, 125–6, 129–33, 139–40, 142, 145, 146, 148–9, 159–61, 163–5, 172, 199, 207–11, 224–5, 233, 235–8

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of languages Galeata 89 Gallo-Italian 104, 196, 200 Gallo-Romance 12–13, 29, 59, 64, 74, 77–8, 80–1, 91, 101–6, 108–9, 116, 138, 146–7, 168, 177, 180, 196, 200–1 Gallurese 56 Gardenese 109, 115 Geneva 108 Genoese 37, 64, 89–93, 146, 149–50, 166–8, 176, 184, 196, 206, 226 German 5, 30, 36–8, 62, 72, 84, 109, 114, 135, 142, 235, 238–9 Germanic 1, 4, 6, 14, 30, 35, 37–8, 75, 91, 97, 105, 113–14, 167, 205, 235, 238–9 Giornico 152 Gombitelli 202 Goriziano 148–9 Gothic 35, 45 Grado 148 Greek 3, 4, 8, 10, 35, 44–6, 50 Grosino 152 Haut-Dauphiné 105 Hauteville 108, 182, 184, 195 Hindi 12 Hopi (Uto-Aztecan, SW USA) 1 Ibero-Romance 29, 59, 63–4, 76–80, 116, 195–7 Illyrian 3 Imperia 149 Indo-European 2, 4–6, 9, 11, 35, 53 Intemelian 149 Italian 3, 6, 9, 12–13, 16, 19, 24, 26–9, 32, 34, 36, 38–9, 48, 50–4, 58, 61–8, 70–6, 79, 81, 86, 95, 109–12, 116–17, 119, 122, 125, 132, 135–6, 138, 140, 146, 148, 150–1, 156–7, 167, 173, 177, 179–81, 183, 185, 190–2, 201, 203–6, 208–9, 213–16, 219–20, 228, 230–2, 236 Sardinian regional Italian 27 Italic 4–6, 10–11, 23, 52–3 Italo-Romance 75, 101, 109, 116–17, 207, 233 Central 30, 206 Central-Southern 13, 63, 136, 138

289

Northern 13, 29, 54, 61, 64, 66, 80, 82–100, 108–9, 115, 119–203, 205, 207–8, 212, 216–20, 225, 233 South-Eastern 119, 212 Southern 26–30, 37, 55–6, 63, 117–18, 136, 138, 196, 211–12, 218 KiKamba (Bantu, Tanzania) 2 Ladin 72, 97, 108–13, 115, 147, 161, 173, 192, 218 Lagacci 220–1, 223 Latin 1–14, 16–20, 22–6, 28–38, 40–2, 44–54, 56–8, 60–5, 73, 75–81, 85, 92, 102–8, 111, 113, 116, 120–3, 129, 131, 133–8, 141, 151, 156, 159–60, 167, 173, 175–7, 180, 195–7, 200, 203–4, 206–7, 212, 215–16, 219, 225, 227–33, 235, 238–9 regional varieties Africa 21–6, 31, 40–51, 55, 57, 60, 79, 232 Britain 59–60, 77 Campania 54 Central Italy 51, 54 Dacia 54, 59 Dalmatia 54 Egypt 76 Gallia Belgica 59 Gaul 19, 25, 43, 47, 51, 54, 60, 79, 232 Germany 59 Hispania 18, 47, 58–9, 78–9 Italic peninsula 19, 23, 51 Leptis Magna 50 Lusitania 76 Moesia 47 Northern Italy 51, 54 Pannonia 54, 76 Pompeii 3, 42, 51–2, 54 Rome 24, 42–3, 47, 51 Sardinia 48, 55–6 Southern Italy 54 Latin of Tripolitania 49 Tuscany 77 Visigothic Spain 59, 78 Le-Pont-de-Beauvoisin 105 Lecco 97

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

290

Index of languages

Leventinese 152 Ligurian 50, 82, 90–1, 106–7, 111, 114, 148–50, 157, 164–6, 172, 199, 205–8, 225, 232, 235 Lisbon 195 Livigno 153 Livinallese/Fodom 109 Lizzanese 170–1, 174, 186–92, 199, 203, 220–1, 223 Logudorese see Sardinian Lombard 82–3, 87, 96–7, 122, 125, 145, 148, 151–9, 161–4, 172–4, 176, 182, 184, 198–9, 205, 208, 215–17, 221, 224–5, 233 Lumezzanese 157–63, 225 Lungavilla 141 Macerata 28 Malcesinese 161–2 Marche 28, 63, 185, 196 Marebbano 109–13, 192–4 Materano 118 Mayo (Uto-Aztecan, Mexico) 215 Mendrisio 97 Mesnil-Martinsart 102 Milanese 82, 87, 93, 95–7, 111, 114, 122–35, 139–42, 145–6, 152–4, 156–8, 160–1, 164–6, 168–72, 194, 205, 207–8, 216, 219, 224–5, 228, 235–6, 238 Miogliola 93 Moenese 111, 115 Molisano 117 Mongo-Nkundo (Bantu, Congo) 35 Montana 146, 181, 184 Mosniga 160 Neapolitan 28, 30 Niçart 107 Novate Mezzola 96, 125, 155, 217 Occitan 26, 29, 39, 64, 71, 80–2, 106–8, 149, 232–3 Oïl 80, 101, 104, 106 Old Norse 4 Ormea 149

Ortisei 111 Oscan 5–6, 10–11, 52–3 Oscan-Umbrian see Italic Oulx 174, 232 Padovano 163 Palmoli 212 Pavese 125, 239 Pavullo nel Frignano 174, 191, 198, 217 Pesariis 99, 115 Pianaccio 170 Piandelagottese 171, 174, 189–91, 199, 210–11 Pianezza 184 Piazza Armerina 196 Picard 102 Piedmontese 29, 82, 148, 151, 164, 166, 172, 174, 207 Pisan 66 Polesano 163 Popoli 212 Pordenonese 149 Portuguese 6, 13, 26, 30, 36, 39, 55, 58, 63, 80, 195, 197, 203 Pozzolano 118 Pradumbli 99 Pragelà 174, 232 Premanese 95–6 Preone 99 Provençal 104, 106–7, 174, 180 Puter 114 Rhaeto-Romance 12–13, 29, 80, 108–9, 115, 145, 147, 164, 177, 204–5 Riminese 217 Roiasc 107, 149 Romagnolo 89, 123, 185, 217 Romance of Africa 25, 47–9, 57 Central 13, 26, 29 Eastern 65, 75–9, 182, 185, 197 Northern 13, 29, 64–5, 80–116, 121–47, 164, 168, 172–3, 176–7, 179–81, 184–7, 189–90, 192, 194, 199, 201–3, 208, 219, 223

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of languages

291

Proto- 13, 16, 18, 25–40, 54, 60, 66, 74–5, 77–8, 80, 84–5, 88–9, 92–3, 102–3, 105–6, 111–13, 115–20, 126, 129–30, 138–9, 141–2, 144–6, 149–50, 156, 160–2, 173, 183, 190–1, 194–5, 198–9, 204–8, 213–14, 218–20, 224, 226, 232, 234–8 Western 75–80, 85–6, 95, 122, 125, 129, 136, 146, 173, 168, 175, 179–80, 182–6, 189, 194–5, 197, 206, 236 Romanesco 91, 136 Romanian 6, 12–14, 26, 36–8, 54–5, 57, 59, 63–4, 71–2, 75–80, 209, 232 Romansh 66, 82, 104, 108–9, 113–15, 168, 174, 180, 204, 235 Rossa (Val Calanca) 179, 182 Rossa (Valsesia) 151

Seri (isolate, Sonora, Mexico) 1 Sicilian 36 Sillano 170 Slavic 5, 55 Slavonic 5 Soglio 175–6, 190–2, 199, 201, 203, 217, 221–2 Sorso 56 Spanish 3, 6, 26–7, 32, 36, 38–9, 61, 63–4, 71–2, 75, 77, 79–80, 106, 125, 165, 195, 197, 203, 232 Stabiazzoni 220–1, 223 Stiglianese 212 Surmeiran 113, 174 Surselvan 81, 114, 204–5 Sutselvan 115, 174 Swiss German 190

Sabellian 11, 52 Sairano 141 Samòlaco 97 San Daniele del Friuli 99 San Giovanni in Fiore 28 Sansepolcro, Borgo 206 Sanskrit 9 Sant’Angelo Lodigiano 184 Sant’Apollonio 157–8, 160–1, 215 Sardinian 3, 13, 27, 29, 33, 39, 48–9, 52, 55–7, 59, 117, 136, 138, 219, 228, 232 Campidanese 56 Logudorese 3, 27, 30, 39, 56, 219, 228 Sassalbo 170 Sassarese 56 Savonese 167, 196, 206 Saxon (German) 1 Schlarigna see Celerina/Schlarigna Scuol 113–15, 174, 234 Sella Ladin 109, 111 Selva di Cadore 112 Selveglio di Riva Valdobbia 184 Seneca (Iroquois, NE USA and Canada) 1 Senigallia 185, 203 Senise 57 Sent 174 Serbo-Croat 190

Tabarchino 90, 150 Tarcento 99 Tàrtano 153 Teglio 152 Tende 107 Tenneville 102, 105 Tortonese 97, 216–17 Trausella 151 Travacò Siccomario 141, 184 Trebisacce 56 Trécchina 196 Trepalle 153 Trieste 148 Trun 115 Tufillo 212 Turinese 150, 151, 166, 194 Tuscan 26, 28–30, 50, 61, 63, 65–6, 72, 75, 77, 118, 136–8, 145, 173, 181, 186, 192, 200–3, 205, 207–8, 213–14, 217–22, 226, 231–2, 236 Tuscan-Emilian border 145, 219, 226 Tyrolean (German) 114 Ubaia valley 71 Umbria 63, 185, 196 Umbrian 6, 11, 52 Uto-Aztecan 215

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

292

Index of languages

Val Camonica 156 Val Cavallina 155, 163 Val Chiavenna 174 Val d’Antrona 173 Val d’Illiez 181, 184 Val di Scalve 155 Val di Taro 90 Val Germanasca 107, 174 Val Gobbia 157–8 Val Leventina 152 Val Poschiavo 181 Val Rendena 157 Val San Giacomo 174 Valcolla 152 Valdambrino 172–3 Valdôtain 108, 184 Vallader 114 Valle d’Andorno 151 Valsesiano 151, 174, 176, 184, 199 Valtellinese 96–7, 152–4, 172–3, 207, 225, 233 Vastese 28

Vaux 108 Vegliote 55 Venetan 39, 82, 97, 109, 112, 148–9, 155, 161, 163–4, 166, 198, 208, 225, 233 Venetian 82, 148, 155, 163, 166, 196, 198 Ventimigliese 149 Veronese 161–3, 170, 185 Villa di Chiavenna 97, 175 Vinzelles 232 Vivaroalpenc 71 Vogherese 97, 125, 141 Wallis 146, 184 Walloon 77, 102, 104–5 Welsh (Celtic) 5 Wichita (Caddoan, SE USA) 1 Yurok (Algic, SW USA) 2 Zoldano 109, 112 Zurasco 142

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of names Abete, G. 118 Accius, Lucius 4 Acquati, A. 24, 41, 57 Adams, J. 14, 18, 21, 23–5, 39–41, 43–4, 46–50, 53, 57, 59–60, 138, 204 Africanus, Leo see Leo Africanus Ageno, G. C. 89, 91–2 Alarcos Llorach, E. 18, 58 Al-Idrîsî 48 Allen, W. S. 7–10, 14, 45, 59 Anderson, S. 15, 69, 139, 201 Anonimo Genovese 166, 182 Antonioli, G. 152 Arigoni, N. 152 Ascoli, G. I. 108–9 Augustinus, Aurelius Hipponensis (St. Augustine) 20–2, 24, 26, 41, 43–4, 49, 57 Augustus, Octavianus 229 Avidius (Quintus Avidius Quintianus) 46–7, 51 Azaretti, E. 149 Badía i Margarit, A. M. 82, 196 Bafile, L. 209 Baković, E. 143 Banniard, M. 18–20, 44 Barbato, M. 30, 56 Barbera, M. 185 Barnes, T. D. 49, 50 Baroni, M. 89, 97, 99, 122, 131, 139–40, 160, 164 Barthes, R. 17 Bartoli, M. 26, 55, 194 Barwick, K. 22 Battisti, C. 31, 33, 109, 111 Bellati, A. 95–6 Bender, B. 97, 132 Benedetti, M. 38, 86 Benincà, P. 100, 113–14, 122, 130, 235

Bermúdez-Otero, R. 69, 73, 232 Bernardasci, C. 110, 174, 190, 193, 210 Bernardi Perini, G. 8 Bernini, G. 154 Bernoth, A. 29 Berruto, G. 150–1 Berschin, H. 73, 103–4 Bertinetto, P. M. 62, 66, 73, 116, 214, 230–1 Bertoletti, N. 161, 163, 170, 185–6 Bianchini, G. 153 Biasini, C. 189, 199 Biffi, A. 95 Biondo Flavio 34 Birley. A. R. 49–50 Biville, F. 10, 33, 45 Blevins, J. 1–2, 11, 14, 70, 131, 134, 232 Boersma, P. 69–70, 73 Boldrini, S. 7–8 Bolognesi, R. 56 Bonfadini, G. 83, 96, 122, 125, 152, 154–7, 161–2, 164, 217 Bonfante, G. 18–19, 24, 34, 41–4, 51, 54–5, 65 Bonvesin da la Riva 82, 168–9, 171 Bonzi, M. L. 221 Bosizio, G. G. 149 Bosoni, J. G. 97, 152–3, 172 Bottiglioni, G. 6, 150 Bourciez, E. 18, 78–9 Bouvier, J. -C. 107 Bracchi, R. 95, 152–3 Broselow, E. 12 Browman, C. P. 70 Brüch, J. 30 Brugmann, K. 16 Bruni, F. 34 Bruni, L. 34 Buck, C. D. 5 Burdy, P. 86

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

294

Index of names

Burger, A. 228 Burzio, L. 66, 74 Busà, M. G. 174, 191 Cabella, M. 97, 216 Caduff, L. 114, 204 Caecilius Statius 2 Caesar, Gaius Iulius 4, 12, 229 Calabrese, A. 213, 230 Camodeca, G. 3 Campanile, E. 5, 76 Cano Aguilar, R. 82 Casella, M. 119, 212, 217, 222 Castellani, A. 12, 18, 24, 26, 30, 33, 35, 75, 77, 167, 185, 196, 202, 221 Castellani, R. 100, 119, 236 Chambers, J. K. 80 Charisius, Flavius Sosipater 10, 59 Charlet, J. -L. 34, 48 Chen, M. 131 Chene, B. E. de 15, 139, 201 Chierchia, G. 125, 135 Chomsky, N. 1, 143 Christiansen, J. 4, 60 Chung, S. 211 Cicero, Marcus Tullius 7–8, 229 Cioranescu, A. 38 Clackson, J. 3, 33, 47, 50 Cocito, L. 167, 182 Coco, F. 87–9, 199, 207, 217–18, 220 Comini, G. 149 Commodianus 44 Comrie, B. 133 Consentius, Publius 2, 20–2, 24–6, 41, 43–5, 47, 57, 232 Contini, G. 125, 164, 168–70, 184 Contini, M. 27 Coseriu, E. 18, 102 Craffonara, L. 110–11, 115, 193 Cravens, T. D. 76, 189, 195 Crevatin, F. 26 Croatto, E. 112 Cull, N. 141

D’Imperio, M. 62, 116, 183, 208–9, 213–15, 230–1 Dagenais, L. 102 Dalbera, J. -P. 106–7, 150, 199 Dalbera-Stefanaggi, M. -J. 56 Dardel, R. De 58–9 Darwin, C. 137 Dauzat, A. 232 Davis, S. 12 De Gregorio, I. 117, 212 Degrassi, A. 4 Dell’Aglio, M. 71, 116, 230 Dell’Aquila, V. 151 Della Ferrera, T. 152 Delucchi, R. 82 Devaux, A. 105 Dmitrieva, O. 74, 179 Dominicy, M. 189 Donatus, Aelius 22 Dresher, E. 238–9 Dressler, W. U. 69, 234 Duraffour, A. 105, 108 Elcock, W. D. 75, 189 Elwert, W. T. 110, 218 Ennius, Quintus 10 Essen, O. von 114 Ettmayer, K. R. von 158, 161, 168 Eugene IV, Pope 34 Eunus, Gaius Nouius 3 Faggin, G. 100, 236 Fanciullo, F. 3, 25–6, 57 Fankhauser, F. 181, 184 Farnetani, E. 62, 116, 183, 190, 208, 213, 230–1 Fassò, A. 108 Fava, E. 62, 68, 116 Fawcett, H. 137 Ferguson, T. 56 Ferrari, S. 158–60, 163, 215 Ferrier, C. 107 Fikkert, P. 211 Filipponio, L. 23, 28, 31, 84, 87–8, 122, 165, 186, 191–2, 200, 206, 219–23

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of names Finco, F. 97–9, 112, 139–40, 210 Fleischer, J. 190 Flemming, E. 68, 234 Flobert, P. 3–4, 59–60 Flutre, L. -F. 102 Fögen, T. 20 Formentin, V. 166, 168, 185–6, 196 Forner, W. 89–90, 93, 107, 149, 166 Fortson, B. W. IV 7 Fouché, P. 31, 33, 203 Fowler, C. A. 69, 131 Francard, M. 102, 105 Francescato, G. 97–8, 110, 115, 122–3, 129, 142, 148–9, 164 Franceschi, T. 31, 53 Franceschini, F. 50 Frau, G. 97–9, 148 Galletier, E. 41, 43 Gartmann, C. 56 Gartner, T. 109 Gaudenzi, A. 218 Gellius, Aulus 33–4, 36 Genovese, Anonimo see Anonimo Genovese Genre, A. 107 Gerster, W. 146, 181, 184 Gess, R. 102 Ghiggi, S. 97 Ghini, M. 93, 205 Giannarelli, D. 170 Giannelli, L. 170 Giannini, S. 2, 18, 31, 38 Giorgetta, G. 97 Goidànich, P. G. 31, 53 Gökçen, A. 93–4 Goldstein, L. 70 Goyette, S. 58–9 Grandgent, C. 18, 30, 33, 58, 65, 79 Grassi, C. 151 Gredig, S. 113–14, 174, 234 Grünert, M. 82 Gsell, O. 80, 109, 148 Guarnerio, P. E. 175

295

Haas, W. De 13 Hadrian (Publius Aelius Hadrianus) 4, 53 Hagberg, L. 215 Haiman, J. 113–14, 130, 235 Hajek, J. 71, 87–8 Halle, M. 1, 6, 133, 143 Hammond, R. M. 63 Harris-Northall, R. 82 Haudricourt, A. G. 52, 103, 106, 165, 200, 237 Hayes, B. 209, 213–14, 228–31, 238 Heilmann, L. 83, 111, 115, 132 Heine, B. 58 Heinemann, S. 131 Herman, J. 8, 11–12, 19, 25, 31, 35, 40–4, 47, 51–2, 54–5, 58, 78, 81, 138, 190 Hernández-Campoy, J. M. 63 Hillenbrand, J. 32 Hilty, G. 78–9, 81, 103, 105–6 Hock, H. H. 131 Hoenigswald, H. M. 134 Hogg, R. 238 Holm, J. 58 Holtz, L. 20, 22 Horrocks, G. 3, 33, 47, 50 Hualde, J. I. 15, 63, 97, 100, 122, 130, 143, 164 Huber, J. 153 Hyman, L. M. 36, 214 Iasucthan, Marcus Porcius 46–7, 50–1 Iosad, P. 97, 129–30, 138, 143, 237 Jackson, K. H. 5, 60 Jacobs, H. 228 Janker, P. M. 69, 131 Jellinek, M. H. 45 Jespersen, O. 69, 131 Joly, G. 8, 77 Juilland, A. 52, 103, 106, 165, 200, 237 Kager, R. 209, 231 Kaster, R. A. 7, 20, 22, 204 Kavitskaya, D. 15 Kaye, J. 8–9, 135

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

296

Index of names

Kaze, J. W. 89 Keating, P. 68–70 Keil, H. 22 Keller, H. -E. 108, 184 Keller, O. 108, 181 Kerlouégan, F. 6, 9–10 King, R. D. 133, 142, 238 Kiparsky, P. 6, 68, 134, 143 Kirchner, R. 68, 73 Kiss, S. 12 Klein, T. B. 58 Kluender, K. R. 131 Koller, H. 21 Kori, S. 62, 116, 183, 190, 208, 213, 230–1 Kramer, J. 45–6, 109, 111, 114 Křepinský, M. 19 Kuen, H. 111 Kuryłowicz, J. 6, 10–11 Labov, W. 17 Ladd, D. R. 70 Lahiri, A. 211, 229, 238–9 Lancel, S. 26, 47 Lapesa, R. 79 Lass, R. 17, 238 Lausberg, H. 25, 29–30, 49, 52, 54, 56–7, 80–1, 121–2 Laver, J. 66 Lazzeroni, R. 5 Ledgeway, A. 59 Lehiste, I. 1, 32, 69, 73, 154, 190, 192, 215 Lejeune, M. 5, 11, 52–3 Leo Africanus 48 Leo X, Pope 48 Leonard, C. S. Jr. 99, 115, 131, 148 Lepschy, G. 6, 8, 166 Leumann, M. 3–4, 6, 8–10, 12, 18, 38, 42, 45, 50, 59, 204, 229 Levelt, W. J. M. 143 Lewicki, T. 48 Lightner, T. 143 Lindsay, W. M. 229 Liver, R. 114 Livy (Titus Livius) 229 Löfstedt, B. 44

Löfstedt, E. 19 Löhken, S. C. 62 Lombardi, L. 215 Loporcaro, M. 20, 23, 28, 30, 37, 53, 56, 77, 83, 91, 117–18, 122, 131, 136, 141, 143, 165, 169–71, 174, 177, 179, 182–4, 186–7, 190–2, 201, 205, 207, 212, 219, 228, 231 Lorenzetti, L. 49 Loyen, A. 20, 22 Luchet(t)o see Anonimo Genovese Lucilius, Gaius 229 Ludlow, P. 136–7, 144 Lüdtke, H. 5, 12, 24–6, 37, 55, 113, 115, 131, 148, 165, 179 Lupinu, G. 55 Lurà, F. 97 Lurati, O. 83 Luschützky, H. C. 66 Lutta, C. M. 174, 180 Luzi, J. 174 Mackel, E. 30 Maddalon, M. 83 Maddieson, I. 1, 69, 116, 131, 210, 215 Màfera, G. 39, 163 Magri, R. 83 Maiden, M. 30, 59 Maina, S. 174–6, 180, 183 Mair, W. 109–10 Malagoli, G. 170–1, 174, 186–7, 191, 199, 220–1 Mambretti, E. 153 Mancini, A. M. 185 Mancini, M. 23–4, 26, 31, 33, 35, 41, 44–5, 57 Mańczak, W. 28 Maragliano, A. 97 Marcato, C. 82, 131 Marcone, V. 162 Marichal, R. 50 Marotta, G. 2, 18, 31, 38, 62, 75, 208 Martín, S. 71, 107 Martinet, A. 75, 85, 101–2, 108, 139, 144, 158, 174, 182, 184–5, 195 Massera, S. 96 Mateus, M. H. 13, 195 Maurand, G. 106, 180

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of names McCarthy, J. 143, 215 McCrary Kambourakis, K. 66–8, 70–1, 73 McCullough, M. 8 McMahon, A. M. S. 238 Meigret, L. 102 Meillet, A. 33 Meiser, G. 3–5, 9–10, 35, 229 Mele, B. 28 Menegus, G. 112 Menéndez Pindal, R. 78–9 Merlo, C. 28, 83, 117, 154, 156, 174–6, 183, 206, 212 Mester, R. A. 14, 215, 228–31 Meyer-Lübke, W. 30, 33, 44 Michaelis, H. 101 Migliorini, B. 163 Mignot, X. 33 Minkova, D. 238 Miotti, R. 97–100, 131 Miotto, A. 83 Molino, G. 151, 174 Monachesi, P. 189 Montreuil, J. -P. 66, 122, 125–8, 138–9, 143, 157, 164–5 Morani, M. 6, 229 Morin, Y. C. 15, 53, 74, 77–9, 83, 86, 93, 95, 97, 101–7, 122, 131, 139–40, 146–7, 160, 165, 199, 204, 232, 237 Mühlhäusler, P. 58 Muljačić, Z. 55 Müller, D. 71, 107 Müller, H. F. 19 Myers, S. 70 Nandriş, O. 37–8 Nardini Marigliano, D. 196 Naro, A. J. 13 Nazzi, G. 98 Nespor, M. 136 Nicolai, L. 112 Nicolas, J. 92–3, 166–7, 184 Nicolet, N. 173 Nicoli, F. 93–4, 205 Niedermann, M. 229

297

Nocchi, N. 219–23 Norberg, D. 19 Ockham, W. of see William of Ockham Octavilla, S. 49–50 Odden, D. 1, 36, 63, 70 Ohala, J. J. 17, 70, 72 Oneda, R. 83, 132 Ortega-Llebaria, M. 71, 232 Osthoff, H. 16 Pace, A. 56, 155 Paciaroni, T. 28 Page, B. 72, 238 Papa, E. 118 Parodi, E. G. 91–3, 166–7, 176, 206 Parry, M. M. 150 Passy, P. 101 Paul, H. 14, 134, 158 Peletier, J. 102 Pellegrini, G. B. 109, 111, 131, 162–4, 182, 196 Pensado, C. 82, 141, 195, 230 Petrolini, G. 90 Pfister, M. 3, 19, 196 Piccenado, R. 167–8 Pickett, E. R. 71, 73–4 Pieri, S. 170, 202 Pierrehumbert, J. 69 Piffari, S. 155 Pifferi, C. 97, 152 Piras, M. 56 Pirona, J. 236 Piroth, H. G. 69, 131 Pisani, V. 24 Plangg, G. 110, 114 Plautus, Titus Maccius 2, 4, 9–10, 229 Politzer, R. 76, 78 Pompeius 8, 10, 22–4, 44 Pompilio, Paolo 48 Pompino-Marschall, B. 116 Pons, T. G. 107 Porzio-Gernia, M. L. 12 Posner, R. 103 Prehn, M. 1

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

298

Index of names

Prieto, P. 71, 83, 121–3, 126–8, 130–3, 138, 143, 157, 164, 205–6, 228, 232, 237 Prince, A. S. 127, 209, 213, 230 Probus, Marcus Valerius 7 Prokosch, E. 238–9 Prosdocimi, A. L. M. 6 Pulgram, E. 6, 8, 18, 31, 34–8, 44–5 Pult, G. 29, 174 Quintilianus, Marcus Fabius 4, 7, 229 Ramus, F. 74 Rebuffat, R. 47 Recasens Vives, D. 195 Renier, L. 31 Renwick, M. E. L. 14, 64, 71–2, 209, 232 Repetti, L. 83, 86, 97, 122–4, 126, 128–9, 131, 133, 135, 138, 157, 164–5, 169, 171, 237 Resnick, M. C. 63 Riaria 48 Ricciardi, J. S. 89–90, 92, 149 Richter, E. 26, 31, 77–8, 81, 190 Rinaldi, S. 175 Ringe, D. 134–5, 144 Ritt, N. 231, 238–9 Rix, H. 229 Rizzolatti, P. 131 Rohlfs, G. 52, 54, 118, 176, 184, 196–7, 204 Rolin, G. 28 Romano, A. 151 Roncaglia, A. 51 Ronjat, J. 82 Rosenthall, S. 62, 116, 183, 208–9, 213, 215, 230–1 Rossi, G. B. 112 Rossini, G. 83–5, 119, 222 Ruggeri, S. 156 Rulli, E. 90–1 Rupp, T. 174 Sacerdos, Marius Plotius 22, 204 Saidero, D. 98 Saiu, E. 199–202, 219, 221 Sala, M. 37–8 Salimbeni, F. 149

Saltarelli, M. 65–6, 74, 135 Salvi, G. 109 Salvioni, C. 82, 93, 95, 109, 125, 169–70, 174–6, 183, 216 Sampson, R. 77, 81, 190 Sánchez Miret, F. 54 Sanga, G. 83, 86, 93–5, 125, 154–7, 162, 166, 216 Santen, J. van 62, 214 Santerre, L. 53 Saracino, G. 28 Saussure, F. de 134 Savoia, L. M. 57, 212 Scaramella, G. 156 Schaad, G. 175 Schädel, B. 149, 173 Scheuermeier, P. 158 Schirru, G. 49 Schmid, S. 190 Schuchardt, H. 24, 48–50, 65, 78, 80 Schürr, F. 24, 26, 29–31, 44–5, 214, 217 Scipio, Lucius Cornelius 39 Scuffi, S. 97 Seelmann, E. 23, 31–3, 52–3 Seidl, C. 52–3, 58–9 Seiler, G. 238 Seklaoui, D. R. 64 Seneca, Lucius Annaeus 22 Septimius Severus 49 Serbat, G. 2 Sergius 20, 32 Servius (Maurus Servius Honoratus) 23–4, 31 Sganzini, S. 152 Sibille, J. 174, 232 Sidney, M. 107 Sihler, A. L. 5, 33 Smith, R. 60 Smolensky, P. 127 Soffietti, J. P. 150, 166 Solèr, C. 115 Sommer, F. 3 Spagnoletti, C. 189 Spartianus, Aelius 49 Spence, N. C. W. 53 Spoerri, T. 174, 176, 199

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of names Spore, P. 31 Stampa, G. A. 174–7 Stampe, D. 69 Stasse, S. 77 Stevens, M. 71, 73–4, 179, 214 Stok, F. 7 Straka, G. 30–3, 52, 54, 77–9, 102, 105, 204 Sturtevant, E. H. 45, 52 Stussi, A. 163 Suetonius (Gaius Suetonius Tranquillus) 7 Sumien, D. 107 Tagliaferri, S. 155 Taverdet, G. 74 Tavoni, M. 34 Tekavčić, P. 33, 65 Telmon, T. 107 Terentius Scaurus, Quintus 4, 53–4 Thöni, G. 113 Thornton, A. M. 189 Thurot, C. 104 Tiraboschi, A. 154–5 Toso, F. 37, 89–93, 149–50, 206 Traina, A. 36 Trajan 59 Trimeloni, G. 161–2 Tristano, C. 50 Trubetzkoy, N. S. 70 Trudgill, P. 63, 80, 99, 132, 148, 163 Trumper, J. 99, 132, 148, 163 Tuttle, E.F. 197–8

299

Vennemann, T. 84, 141 Virgil (Publius Vergilius Maro) 10, 229 Verner, K. 6 Vicari, M. 152 Victorinus, Marius 4, 10, 33, 59 Videsott, P. 81, 109–11, 113, 122, 164, 235 Vignola Saffirio, V. C. 151 Vincent, N. 135, 185, 203 Vineis, E. 18, 31, 38–9 Virdis, M. 56 Vitale, M. 34 Vitali, D. 90–1 Viti, C. 6 Vogel, I. 135–6 Wachter, R. 3, 4 Wagner, M. L. 27, 49, 56 Walberg, E. 173 Walter, H. 102, 108 Wanner, D. 163 Wartburg, W. von 24–5, 31, 75, 106, 111, 182 Weber Wetzel, E. 96, 216 Wedding, G. 9 Weinrich, H. 25, 31, 49, 65, 76, 80, 118, 122–3, 165, 171, 182, 185–6, 188, 195, 200, 203, 217, 220, 237 Weiss, M. L. 1, 3, 6, 33 William of Ockham 133, 140 Windisch, R. 29 Winteler, J. 17 Wright, R. 31, 82 Wüest, J. 12, 58–9, 104–5

Uguzzoni, A. 84, 87, 89, 164–5, 174, 191, 198, 217 Urech, J. 174, 179–80, 182

Yamamoto, S. 139

Vaan, M. de 2, 5 Väänänen, V. 18, 42, 51–2, 79, 229 Valerius Probus see Probus, Marcus Valerius Vanelli, L. 97–100, 122, 124, 126, 129–31, 137–40, 142–3, 159–60, 164 Varro, Marcus Terentius 44–5, 229 Varvaro, A. 48, 76, 196 Vayra, M. 62, 214 Vázquez Obrador, J. 196 Velázquez Soriano, I. 78

Zamboni, A. 18–19, 28, 80, 82, 112, 122, 148, 163, 196 Zamora Vicente, A. 63 Zec, D. 12 Ziccardi, G. 117 Zingarelli, N. 28, 212 Zink, G. 77–8, 82, 105, 207 Zörner, L. 151 Zufferey, F. 106 Zwicky, A. 206

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of subjects ablaut 5–6 acutus (stress mark) 8 affix, affixation 55, 82, 94 allophony 68–70, 75–6, 88, 93, 108, 158, 185, 189, 202, 234 allophonic vowel length 13, 15–16, 26–8, 32, 39, 56, 62–5, 70–4, 76, 89, 93, 104–5, 107, 111, 113–14, 116–17, 124–5, 130, 141, 152, 157, 165, 183, 204, 209–11, 219, 230, 239 analogy 10, 55, 73, 78–9, 150, 180–1, 234, 238 apex (stress mark) 4, 53, 59–60 apócope extrema 82 apocope 19, 64, 75, 81–2, 85–7, 90, 95, 98, 104–5, 109, 122–4, 126, 128–9, 131, 133, 145, 147, 157, 159, 162, 164–71, 175, 184, 186, 189–93, 196, 198, 204–5, 212, 216–17, 221, 224, 235–6 final /@/ 81, 104, 171, 184, 207, 238 Appendix Probi 141 architecture of phonology 69, 72–3, 232 article, definite 91 bilingualism 50, 119 bimoraic 6, 62, 66, 123, 125, 127, 129–30, 133, 205, 228 BiPhon model 69 chain shift 75–6, 85, 95, 125 chronology 30, 40, 79 absolute 45, 111, 204, 206 relative 37, 77, 105, 118, 138, 195, 199, 204 circumflexus (stress mark) 8 clitic 98, 181–2, 189, 201, 206, 212, 215 object 91, 93, 98, 181–2, 189, 201, 206, 212, 215 subject 91, 201 coalescence 13, 15, 37, 63–4, 91–2, 104, 146, 153, 167, 201, 206, 217, 221, 237, 239

coarticulation 69–70, 72–3 coda 11, 39, 67, 91, 103–4, 106, 125–6, 146, 151, 167, 232 colour(ing) 5, 28, 34, 63, 73, 117, 119, 233 competence 7, 15, 22, 34, 42, 50–1, 60, 86, 101–2, 132, 134–5 complementary distribution 12, 27, 38, 47, 61, 63, 65, 68, 87, 98, 135, 150, 177, 183–4, 186 connected speech 111, 116, 119, 171 consonant 2, 10, 24, 27, 37, 66–9, 73–5, 85–6, 88, 90, 92–3, 95–6, 100, 104–5, 108, 116, 122–7, 131–2, 135–6, 139–41, 146, 151, 153, 160–1, 171, 173, 175–6, 179–80, 182–7, 190–5, 201–2, 210, 216, 219, 225, 233, 236, 239 final (relevant for VL) 85–7, 91, 93–5, 100, 107, 110–11, 128, 129–31, 137, 139–40, 159–60, 162, 165, 191, 235–6, 239 geminate see formant values, geminates consonant cluster 73, 84, 126, 198, 239 final 81, 87, 91, 94, 107, 110–11, 128–30, 137, 159, 162, 165, 235 heterosyllabic 28, 37, 67, 141, 156 initial /s/+C 76 /l/ + coronal 91 muta cum liquida see muta cum liquida /nC/ 210 obstruent + /l/ 107 /r/ + C 215 /r/ + dental 238 /r/ and liquid + C 93 /sC/ 84, 215 simplification 104 sonorant + obstruent 84, 239 sonorant + stop 239 tautosyllabic 67, 100, 232 consonant gemination 2–3, 11, 61, 65–6, 74–5, 88, 108, 122, 125–6, 135–7, 174–203, 217

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of subjects

301

consonant length 66, 75, 88, 106, 123, 173, 175, 177, 183, 186, 189–90, 192, 194, 200, 203, 221, 233 constraint-based models 63 constraint 2, 15, 62, 65, 73, 77, 86, 95, 126–8, 144, 147, 209, 213–16, 226, 234 acoustic 68, 73 F(oo)t-Bin(arity) 127–30 faithfulness 62, 73 Fill 127–8 HNuc 127–8 implementation 64, 68–9, 75, 232 non-selective 62 phonetic 63–4, 68–9, 72, 131, 179, 194, 208, 211–12, 214–15, 225, 230, 232–3, 235 phonological 14, 215, 226 strong rhyme (SRC) 125–6, 139 structural 62, 66 contrast see vowel length, contrastive binary see vowel length, binary three-way (phonemic) see vowel length, three-way vowel quality see vowel quality contrastive, underlying 12, 66, 74, 88, 125, 137, 183, 194, 224 correptio iambica see iambic shortening creolization 58–9

intervocalic consonant see syncope onset /l/ 91, 100 prosthetic vowel 77 word-internal vowel see syncope dialectalization 19, 58 Dichiarazione di Paxia 167–8 diphthong 4, 12, 30–2, 36–7, 77, 79, 100, 103, 106, 112, 117, 150, 165, 204, 212 diphthongization 12, 26, 28, 30–1, 55, 63, 73–4, 77–9, 89, 104–5, 117–18, 204–5, 211–13, 230–1 of PRom /e O/ 26, 29–30, 72–4, 77–9, 105, 149, 195, 204–5, 207–8, 213, 231 open syllable 12, 26, 55, 103–4, 119, 204, 230–1 utterance-bound 118 distributional restriction 2, 71, 84–5 duration 67–71, 73, 75, 79, 99–100, 116, 183, 210 absolute 62, 73 consonant 87–8, 183, 186–7, 190, 192–4, 225 phonetic 70, 73–4 segmental 71, 73 stressed syllable 72 stressed vowel 63–4, 67–8, 72–4, 78–9, 87, 99, 114, 116, 150, 152–3, 179, 183–4, 186–7, 190, 193, 209–11, 213–14, 222–3, 225 vowel 14, 52, 66, 69, 70–3, 94, 117, 194, 222, 235

deaffrication 201–2 degemination 5, 12, 29, 36, 75–6, 78, 85, 92, 103–6, 111, 122, 129, 138, 143, 146–7, 152, 164–8, 170–1, 173, 175–6, 182–3, 185, 192, 194–6, 199–203, 206, 216, 225, 236–7 post-tonic degemination 164, 179, 182–3, 185, 196–7, 199–200, 202–3 pretonic degemination 179–80, 182, 185, 187, 189, 196–7, 199–202, 216, 221–3 delaryngealization 129 deletion coda consonant 63–4, 91, 103–4, 106, 146, 151, 167, 232 final /@/ see apocope, final /@/ final consonant see apocope final vowel see apocope

emphasis 112 epenthesis 77 experimental phonetics 13–14, 16, 32, 62, 64, 66, 68–71, 73, 75, 88, 97, 99, 110, 112, 114, 116–19, 131, 152–3, 171, 177, 186, 191, 193, 197, 208–9, 214, 236 extrametricality 127, 209, 213–15, 228, 230–1 final devoicing 85, 95, 107, 111, 129–31, 135, 139–40, 142–3, 159–60, 165, 171, 235–7 focus 119–20 foot 7, 14, 41–2, 46, 127, 132, 208–9, 213, 215–16, 224, 227–31 binary 209, 227, 231 ternary 209

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

302

Index of subjects

Foot Expansion 132–3, 137, 205, 228 formalization 121, 124, 127, 133–5, 212–14, 227–8 formant values 73 contrastive 66, 75, 132, 144, 180 etymological 75, 85, 173, 175–6, 179–80, 196 geminates 2–3, 11–12, 27, 62–3, 73–4, 78, 85, 125–6, 143, 147, 167–8, 153, 173–7, 180–1, 183, 185–7, 189–92, 194–6, 198–200, 202–3, 217, 219 non-etymological 174–5, 181, 185 post-tonic 2, 96, 108, 164, 176, 179, 182–5, 187–8, 199–200, 216, 219, 221 pretonic 177, 179–81, 185, 196, 200, 202–3, 221 Proto-Romance 85, 142, 183, 219 geminatio uocalium 4–5 gemination see consonant gemination generative grammar 56, 132–9, 143, 169, 231 Government Phonology 9 grammar 15, 63, 69, 102, 129, 204, 212 gravis (stress mark) 8 heterosyllabification 28, 37, 141 homophony/homophonous 9, 46, 91, 95, 112, 119, 148, 153, 155, 170 hypercharacterization 177 hypermetry 169 hypocorrection 72 I longa 4, 53, 59 iambic shortening 5, 9–10, 35–6, 59, 228, 231 infinitive 98, 149, 162, 216 intervocalic consonant lenition see lenition isogloss 29, 48, 75, 80–1, 107, 109, 115, 148, 162, 164, 189, 200 koiné 45, 97 La Spezia-Rimini (Carrara-Fano) line 75, 182, 185, 200, 203, 219 language acquisition 77 Latin grammarians 2, 4, 8, 20–3, 38, 44–5, 53, 59

Latin metrics 4, 7, 20, 35, 41–4, 46, 49, 51, 57, 138, 204, 229 Latin orthography 3–5, 9, 38, 40 Latin prosody 4, 19, 21, 43, 46 Latin-Romance transition 18 lengthening allophonic 117, 125, 141, 204 coarticulatory 72 compensatory (CoL) 15–16, 64, 91–2, 103–4, 116, 123–6, 130–1, 133, 137, 140, 145–6, 151, 157, 170–1, 224, 232, 235–9 consonant see consonant gemination open syllable (OSL) 13–16, 20–51, 58, 60–76, 78–80, 84, 88–9, 91, 93, 103–6, 108–9, 111–21, 123, 129–33, 135, 137–40, 142, 144, 146–7, 150–2, 156–7, 160, 163, 165, 203–6, 208–11, 213–15, 219, 224, 230–9 phonetic 71–2, 159, 231 prepausal 116–17, 186, 219, 222, 230, 233 voicing-induced 69, 110, 131, 140, 142, 159–60, 165, 239 vowel 10, 21–3, 25–6, 62, 64, 79, 89, 91, 93, 96, 104, 113, 119, 121, 125–8, 130–3, 137, 140, 145–6, 159–60, 165, 168, 176, 198, 203–6, 211–15, 219, 224, 230–2, 238–9 lenition 75–6, 86, 180, 184, 189, 199–200, 236 lexicalization 36–7, 201, 234–5 loanword 3, 5, 10, 27, 30, 37–8, 45, 48, 53, 55, 60, 75, 86, 93, 119, 140, 180, 195–6, 211, 220, 235, 238 metaphony 5, 26, 29–30, 55–7, 77, 118 metric 4, 28, 127, 168–9, 227 minimal pair 2–4, 7, 9–10, 56, 59, 61, 63–5, 76, 83–4, 86, 89–93, 95–7, 100, 103, 107–8, 110–15, 119, 122, 125, 129, 149, 151, 153–5, 158–9, 162, 177, 179, 183, 186, 208, 217, 226, 234–7 monophthongization 4, 31, 53, 64, 86, 101, 104, 112, 150 monosyllable 2, 10, 21, 64, 127, 135, 201, 209, 215, 238 mora 62, 71–2, 124–8, 130, 132, 135–7, 143, 212, 229

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of subjects Moraic Phonology 124–5 morphological (vs. sound) change 9, 78 morphology 9, 46, 78, 234 derivational 82 morpheme 55, 63, 69, 82, 140, 180 morphological paradigms 112, 180 morpho(pho)nology 143, 211, 234 muta cum liquida 28, 36–7 nasalization 95 Natural Phonology 69 neutralization 10, 20, 24, 44, 49, 64, 94, 96, 101, 103, 112, 115, 119, 122, 129, 142, 149, 192, 208 nucleus 100, 106, 127–8, 229 onset 67, 91 Optimality Theory 15–16, 62, 66, 68–70, 126, 129, 138, 231, 237 orthographic norm 3–5, 9, 38, 40, 46, 153–4, 167 output oriented models 68 oxytones 86, 122, 127, 142, 156, 169, 191–4, 200–1, 208, 214–23 primary 217, 223 secondary 200, 217, 220–1, 223 palatalization 48, 107, 111, 202 /a/-palatalization 89, 111 paroxytones 27, 84, 86, 96–7, 100, 113–14, 122, 124, 126, 131, 142, 146, 152, 159, 161, 169, 172–3, 176–7, 185–6, 189, 191–4, 204–6, 208–9, 211–12, 221, 234–6 perceptual threshold/JND 71, 116 performance 8, 134 phonetic gradient 67–8, 74, 190–3, 214 phonetics-phonology interface 69–70 phonological feature 7, 12, 18, 49, 58, 69, 101, 109, 115, 129, 137, 144, 148, 150, 158, 176, 179, 216, 222, 224 phonological process 72, 119, 135 phonological rule see rule, phonological phonological system 1, 3, 8, 32, 34, 58, 76, 103, 135, 209, 215, 223, 226, 234

303

phonological utterance 169, 222–3 phonological word 62, 71, 146, 169, 171–2, 189, 193, 206–7, 210, 212, 215, 218–19, 222–3, 225, 230 phonologization 15, 56, 103, 105, 108, 139, 146, 157, 160, 168, 194, 236–7 phonosyntactic see sentence phonetics pitch accent 8 position antepenultimate 128, 205 intervocalic 27, 37, 43, 75–6, 85–6, 92, 105, 129, 159, 161, 167, 176, 180, 196–7, 199–200, 202–3 non-prepausal 116, 117, 119, 168, 179, 186, 192 penultimate 128, 205 prepausal 112, 116–17, 119–20, 142, 168–70, 179, 184, 186, 192, 219, 222, 230, 233 pretonic 73, 146, 179–81, 201–3, 221–3 stressed 16, 53, 85, 90 unstressed 12, 19, 42–3, 53, 90–1, 94, 232 utterance-final 118, 222–3 utterance-internal 119, 192, 221, 222–3 word-final 64, 66, 85, 98–9, 107, 126, 139, 147, 156, 158, 167, 170, 181, 185–6, 192–3, 200–1, 214–19, 221–3, 225 word-internal 64, 167, 216, 225 possessive 98, 119, 220 prepausal lengthening 116–17, 119–20, 186, 222, 230, 233 process allophonic 69, 75, 104, 117 phonological 69, 72, 119, 135 postlexical allophonic 68 prelexical 69 production 50, 66, 70, 76, 209 pronunciation 20–5, 33, 39, 100, 119, 182, 229, 232 proparoxytones 86–7, 92, 94, 96, 99, 122, 126–7, 132–3, 143, 169, 176, 182, 185–6, 189, 192, 203–9, 211–14, 221, 230–1, 235 etymological 81, 87, 95, 204–5, 212, 231 proparoxytonic shortening 204 prosody 4, 19, 21, 42–3, 46, 131, 225, 227

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

304

Index of subjects

prosodic change 20, 132 prosodic environment 207, 231, 235 prosodic minimality 215 prosodic structure 88, 117, 194 prosodic trapping 229 prosodic weight 127 prospective perspective 134 prosthesis 76–7 proto-language 134 quantity 2–3, 7, 9, 11–12, 16, 19–21, 26, 34–6, 39, 43–5, 51–61, 63, 111, 123, 184, 190, 195, 215, 220, 224–5, 227–9, 232, 236 consonant 12, 65–6, 87–8, 123, 125–6, 175, 177, 179, 183, 186, 189, 225 vowel 3–5, 9, 12, 23–4, 29, 32–3, 38, 44, 47, 51, 58, 73, 79–80, 84, 93, 115, 119, 122, 154, 164, 170, 192, 221, 225 raddoppiamento fonosintattico 65–6, 75, 135–7, 181–3, 186, 192, 201–2, 217, 219 reassociation 124, 126, 130, 135 reconstruction 14–15, 29, 109, 133–4, 136–7, 144, 154, 169, 171–2, 194, 203, 219, 224, 233, 239 reduplication 3, 5 representation 69–71, 124–5, 127–30, 133, 137–8, 143, 158, 208, 231 acoustic parametric 70 categorical phonetic 68, 70 phonetic 69–70 phonological derived 9, 68–9, 74, 124, 143, 172, 225 underlying 56, 63, 65–6, 68–9, 74, 83, 86, 100, 103, 108, 124–6, 130–3, 138–40, 143–4, 147, 156, 158–60, 171, 183, 189, 201–2, 209, 215, 219, 225, 234, 237 retrospective perspective 134 rhyme 93, 125–6, 139, 149, 166–70, 224 rhythm 7, 8, 27, 41, 74, 208 rhythmical compensation 208–14, 225, 228, 230–1, 234–7 rule 11, 59, 62, 66, 77, 138, 142, 144, 158, 160, 179, 234, 237

allomorphic-morphological 234 allophonic 13, 15, 62, 64–5, 70–2, 202, 230, 239 morpho(pho)nological 211, 234 OSL 24–5, 62, 66, 72, 76, 116, 163, 210, 219, 232, 237, 239 phonetic 69, 82, 232 phonological 62, 69, 75, 77, 79, 143, 147, 161, 211, 214, 232, 234 postlexical 65, 68–9, 75 stress see stress placement suppression 76, 168–70, 189, 192 rule-based models 63, 138 sentence phonetics 82, 174, 186, 189, 191, 194 sentence prominence 116, 120 Serments de Strasbourg 81 shortness/shortening 5, 9–12, 21–3, 25, 33–6, 39, 42, 47, 59, 62, 80, 87, 93, 113, 123, 140–2, 146, 153, 156, 160–3, 173, 175, 185, 203–4, 208, 210–12, 214, 219, 220, 222, 225, 228, 231–2, 235, 239 sociolinguistics 3, 14, 40, 58, 177, 229 speech 22–4, 35, 40, 45, 52, 58, 60, 74, 101, 111, 116, 119, 161, 171 controlled 71 fast 63 speech chain 70 speech community 78, 102 speech rate 62, 73 spontaneous 116, 210 St. Eulalia 78, 81 stress placement 6, 11, 36–7, 65, 228 structuralism 9, 15, 56, 76, 106, 132, 134, 139 subminimal pairs 83–4, 86, 89–90, 146, 151, 171, 177 suprasegmentals 63, 227 syllabic constituents 69 syllabification 28, 79 syllable antepenultimate 6, 36, 100, 117, 128, 203–6, 210–13, 228, 230

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

Index of subjects closed 2, 8, 11, 28–9, 34, 36, 38–9, 61, 63, 67, 71, 73, 77–9, 84, 88–9, 92–3, 98–100, 116–18, 135, 141, 146, 149, 156, 160, 162, 180, 198–9, 205–7, 210–14, 218, 230, 234, 237 final 2, 5, 9, 36, 59, 62–3, 81, 84, 87, 93–100, 113–14, 127–8, 151–2, 158–60, 162, 190, 201, 209, 215–16, 228, 230–1, 234 heavy 6, 11–12, 21, 42, 215–16, 229 light 9, 11, 12, 21 non-antepenultimate 117 non-final 2, 94, 97, 99, 218, 225 open 8, 11–14, 16, 20–51, 55, 58, 60–80, 84–5, 88–9, 91–3, 100, 103–6, 108–9, 111–21, 123, 129–33, 135, 137–42, 144, 146–7, 149–52, 156–7, 160, 163, 165, 195, 198–9, 204–15, 218–19, 224, 230–9 open non-final 2, 61–2, 89 open stressed 2, 13, 15, 21, 23, 27, 45, 61, 63, 67, 70–3, 89, 93, 104–5, 111, 113, 115, 142, 156, 162, 173, 209, 212–14, 218, 232, 235 open unstressed 6, 71 penultimate 6, 36–7, 87, 89, 96, 100, 124–5, 128, 147, 185, 205, 210–11, 225, 229–30, 234–7 stressed 2, 4–5, 8, 11, 13, 15, 21, 23, 27, 45, 61–3, 66–7, 70–3, 83–4, 89, 93–4, 100, 104–5, 111, 113, 115, 124–5, 132, 135, 141–3, 152, 156, 158, 160, 162, 173, 183, 206, 209, 211–14, 216, 218, 225, 231–2, 235 superheavy 11–12, 34 unstressed 10, 11, 16, 25, 59, 71, 85, 190, 211, 233 word-final open 84 syllable boundary 73 syllable cut 84, 91, 115 syllable structure 8, 9, 11–12, 23, 26, 29, 38, 63–5, 68–9, 71–4, 76–9, 84–5, 88, 91–3, 106–7, 117, 125, 127, 135, 141, 150, 162, 183, 194, 197–8, 205, 210–11, 213, 230 syllable weight 6, 12, 42, 125–6 synchrony 6, 11–12, 14–16, 34, 56, 64–6, 70, 82–3, 85, 88–9, 92, 95, 100–1, 116–17, 122–4, 128, 132–5, 138–40, 143–4, 147, 158–60, 168–70, 189, 192, 202, 207–8, 211–12, 215, 221, 224–5, 228, 238–45, 237

305

syncope 19, 37–8, 64, 81, 91–2, 100, 105, 189, 228–31 synonymous 86, 234 systematic phonetics 70 timbre 39, 51–2, 151 transition(al) dialects 83, 109, 112, 150, 183–5, 192, 208 trisyllabic shortening 211 trochee 127, 228 utterance 112, 116–20, 174, 179, 186, 192, 219, 221–3 variation 3, 34, 70–1, 73, 82, 98, 101, 117, 119, 147 allophonic 76 dialect 13–14, 101, 105, 108, 121, 123, 138, 145–226 free 11, 23–4, 150, 196 social 40, 58 structural (micro)variation 82, 219 vocalization 92, 196 voice 31, 69, 85–6, 92, 95, 100, 107, 110, 124, 127–8, 130–1, 140–3, 159–61, 165, 173, 189, 210, 239 voicing 70, 76, 85–6, 110, 129, 131, 159 vowel centralized 81, 169, 184 contrastively short/long stressed 66, 86, 89–90, 94, 98–9, 100, 112, 116, 124, 132, 146, 151–2, 159, 172, 201, 220, 235, 237 extra-long 2 extra-short 99 final stressed 62, 145, 156, 167, 191, 200, 209, 211–12, 214, 217–23, 225–6 final unstressed 10, 26, 59, 81–2, 86, 90, 104, 123–4, 126, 129, 131, 136–7, 157, 164, 166, 168–71, 181, 184, 186, 190, 192, 196, 198, 207, 215–17, 221–3, 235, 237–8 half-long 100, 104, 159, 161 high 33, 56, 73, 77, 98, 117, 161, 198, 209, 215 high mid 98, 105, 160, 214, 236

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

306

Index of subjects

vowel (cont.) long 1–13, 16, 20–6, 31–9, 42–3, 46–9, 52–3, 56, 59, 63–4, 67–9, 71, 74, 79–80, 83–7, 89–96, 98–100, 103–6, 108, 111–14, 116, 119, 123–5, 127, 129–33, 135, 138–40, 142, 145–6, 149–59, 162, 165, 169, 172, 175, 177, 179, 181, 183–6, 192, 194, 199–201, 203, 205–6, 208–12, 215–17, 219–25, 229, 232–9 low 69, 131, 154, 214, 217 low mid 77, 89, 98, 206, 215, 217, 236 mid front non-round 93 nasal 95, 102, 106 non-low 33, 48, 81–2, 90, 131, 164, 168, 171, 184, 190, 192 short 1, 2, 4–11, 16, 20–2, 24–5, 30–6, 39–40, 42–3, 46–9, 51–3, 55–6, 59, 62, 66–8, 70–2, 74, 80, 83–9, 91–101, 105, 108, 111–16, 119, 123, 125, 127–32, 135, 139–40, 146, 149–61, 167, 173, 175, 177, 179–80, 183–6, 188–9, 191–2, 194, 201, 203–5, 210–12, 214–23, 225–6, 232–3, 235–6, 238–9 short stressed 21, 31, 39, 66, 72, 84, 86–8, 92–3, 95, 108, 119, 127, 132, 149–50, 152–3, 156, 160, 175, 179, 180, 183, 185, 188–9, 191–2, 194, 204–5, 216–17, 219–20, 235–6 tautosyllabic 1 unstressed 6, 8, 10–12, 16, 19, 21, 24, 26, 42–4, 46–7, 53, 55–6, 58, 60, 69, 71, 73–4, 81–2, 85, 90–1, 94, 131, 135, 164, 166, 168–9, 171, 179, 190, 196, 201, 207, 232 unstressed non-word-final 10 velarized 5

vowel harmony 35 vowel height 69 vowel length 1–3, 5–9, 11–14, 16, 18, 20–5, 27, 29, 35, 38, 42, 57, 60–203, 208, 214–23 binary 2, 67, 74 contrastive 1–16, 18–20, 22, 24–5, 32, 34–6, 38, 40–7, 49, 51, 53, 55, 58–60, 62–6, 70–1, 74, 76, 80–7, 89–91, 93–4, 96–117, 119–33, 135–73, 175, 183–4, 186, 189–94, 199–205, 207–11, 214, 216–17, 221–6, 232–9 three-way 1, 2, 71, 74 vowel quality 1, 12, 32–3, 37–8, 52–3, 57–8, 73, 89, 93, 98, 102, 111–12, 145, 151, 155, 163, 166, 210 vowel reduction 6, 16, 81–2 vowel shift 88–9, 143 vowel system 25, 29, 33–6, 38–9, 44, 47, 52–3, 57, 73, 83, 89, 93, 102–3, 105–8, 110, 151, 154, 164, 207, 224, 228 asymmetric 54–6, 77 five-vowel 32–3, 49, 55–7, 98, 148 seven-vowel 32, 35, 49, 52, 55–7, 103, 112, 136, 154, 207 unstressed 81, 190 vowel tenseness 1, 16, 31, 33–5, 38–9, 93–4, 154 Well-Formedness Condition 135–6 word boundary 65, 120, 135, 158, 233 word shape 77, 90, 94, 139, 162, 208, 212, 215, 235 word-level 116, 120, 191–2, 222

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

OXFORD STUDIES

IN

DIACHRONIC

AND

HISTORICAL LINGUISTICS

General editors Adam Ledgeway and Ian Roberts, University of Cambridge Advisory editors Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; David Willis, University of Cambridge Published 1 From Latin to Romance Morphosyntactic Typology and Change Adam Ledgeway 2 Parameter Theory and Linguistic Change Edited by Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar 3 Case in Semitic Roles, Relations, and Reconstruction Rebecca Hasselbach 4 The Boundaries of Pure Morphology Diachronic and Synchronic Perspectives Edited by Silvio Cruschina, Martin Maiden, and John Charles Smith 5 The History of Negation in the Languages of Europe and the Mediterranean Volume I: Case Studies Edited by David Willis, Christopher Lucas, and Anne Breitbarth 6 Constructionalization and Constructional Changes Elizabeth Traugott and Graeme Trousdale 7 Word Order in Old Italian Cecilia Poletto 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent 9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro

OUP CORRECTED PROOF – FINAL, 9/7/2015, SPi

In preparation Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden The History of Negation in Low German Anne Breitbarth Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti The Historical Dialectology of Arabic: Linguistic and Sociolinguistic Approaches Edited by Clive Holes A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τος and its Implications for Language Change and Structure Brian D. Joseph Gender from Latin to Romance Michele Loporcaro The Syntax and Semantics of Vedic Particles John J. Lowe Syntactic Change and Stability Joel Wallenberg The History of Negation in the Languages of Europe and the Mediterranean Volume II: Patterns and Processes Edited by David Willis, Christopher Lucas, and Anne Breitbarth