The Transmission of Anglo-Norman: Language History and Language Acquisition [Hardcover ed.] 9027208263, 9789027208262

Citation preview

The Transmission of Anglo-Norman

Language Faculty and Beyond Internal and External Variation in Linguistics Language Faculty and Beyond (LFAB) focuses on research that contributes to a deeper understanding of the properties of languages as a result of the Language Faculty and its interface with other domains of the mind/brain. While the series will pay particular attention to the traditional tension between descriptive and explanatory adequacy, the series will also address issues such as the level of linguistic design, through new lines of inquiry often referred to as ‘physiological linguistics’ or ‘biolinguistics’. LFAB aims to publish studies from the point of view of internal and external factors which bear on the nature of micro- and macro-variation as, for example, understood in the minimalist approach to language. For an overview of all books published in this series, please see http://benjamins.com/catalog/lfab

Editors Kleanthes K. Grohmann University of Cyprus

Pierre Pica CNRS, Paris

Advisory Board Paola Benincà

Anders Holmberg

Cedric Boeckx

Lyle Jenkins

Guglielmo Cinque

Richard K. Larson

Noam Chomsky

Andrew Ira Nevins

University of Padova, Italy ICREA/University of Barcelona, Spain University of Venice, Italy Massachusetts Institute of Technology, Cambridge, USA

Stephen Crain

Macquarie University, Sydney, Australia

Marcel den Dikken

CUNY Graduate Center, New York, USA

Naama Friedmann

Tel Aviv University, Israel

Volume 9

University of Newcastle, UK Biolinguistics Institute, Cambridge, USA Stony Brook University, USA University College London, UK

Alain Rouveret

University of Paris VII, France

Esther Torrego

University of Massachusetts, Boston USA

Anna Papafragou

University of Delaware, Newark, USA

Akira Watanabe

University of Tokyo, Japan

The Transmission of Anglo-Norman. Language history and language acquisition by Richard Ingham

The Transmission of Anglo-Norman Language history and language acquisition

Richard Ingham Birmingham City University

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of the American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Ingham, Richard, 1955The transmission of Anglo-Norman : language history and language acquisition / Richard Ingham. p. cm. (Language Faculty and Beyond, issn 1877-6531 ; v. 9) Includes bibliographical references and index. 1. Anglo-Norman dialect--History. 2. Anglo-Norman dialect--Grammar. 3. AngloNorman dialect--England. 4. Language acquisition--History. 5. Discourse analysis--History. 6. Language and culture--England--History--To 1500. 7. Historical linguistics. I. Title. PC2943.I54   2012 447’.942--dc23 2012026109 isbn 978 90 272 0826 2 (Hb ; alk. paper) isbn 978 90 272 7334 5 (Eb)

© 2012 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Preface chapter 1 Introduction to key issues 1.1 The research problem  1 1.2 Anglo-Norman and second language transmission  3 1.3 Language transmission  6 1.4 The critical period hypothesis and second language acquisition  7 1.5 Early child bilingualism  9 chapter 2 Anglo-Norman and L2 varieties of medieval French 2.1 Introduction  13 2.2 An outline sketch of early C13 A-N grammar   14 2.3 Old French as an L2 in southern Italy  20 2.4 Law French as an L2  23 2.5 Prospects  25

ix 1

13

chapter 3 The context of transmission 27 3.1 Previous interpretations  27 3.2 The historical evidence for the status of A-N  29 3.3 The elementary school as a context of first exposure to Anglo-Norman  33 3.4 Implications  35 chapter 4 Rationale and design of the study 4.1 Initial considerations  39 4.2 Target areas of medieval French  41 4.3 Data sources  43 4.4 Design of the syntax study corpus  46 4.5 Additional data sources  47 4.6 Overall approach  49

39

vi

The Transmission of Anglo-Norman

chapter 5 Anglo-Norman phonology 5.1 Introduction  53 5.2 Acquisition target properties  54 5.3 Variation and change in OFr phonology  56 5.4 Influence of English: vowels  58 5.5 Influence of English: consonants  60 5.6 Loss of sound contrasts in A-N or underspecification in rhymes?  61 5.7 Investigating neutralisation   63 5.8 Results  64 5.9 Summary of findings and implications  68 chapter 6 The syntax of quantifiers in Anglo-Norman 6.1 Introduction  73 6.2 Acquisition target properties: Old French quantifiers  74 6.3 Diachronic developments in the Old French quantifier system  76 6.4 A comparison with Middle English  78 6.5 An investigation of quantifier syntax in Anglo-Norman  79 6.6 Implications  84 chapter 7 Noun gender marking in Anglo-Norman 7.1 L2 gender acquisition background  89 7.2 Acquisition target properties  90 7.3 Gender in A-N: previous research  91 7.4 Gender marking in the A-N Hub textbase: findings from possessive determiner contexts  95 7.5 Implications  97 chapter 8 Verb second and null subjects in Anglo-Norman 8.1 Introduction  101 8.2 Acquisition target properties in OFr: V2 and null subjects  102 8.3 A comparison with Middle English  105 8.4 Investigating the maintenance and loss of V2 in Anglo-Norman  107 8.5 V2 after selected initial adverbs in main clauses  109 8.6 V2 with initial Objects versus initial Adjuncts  111 8.7 Null subjects in subordinate clauses  114 8.8 Implications  116

53

73

89

101



Table of contents vii

chapter 9 The order of Attributive Adjective and Noun in Anglo-Norman 121 9.1 Overview   121 9.2 Attributive Adjective-Noun order in medieval English and French   122 9.3 Acquisition target properties  125 9.4 Anglo-Norman attributive adjective position  129 9.5 Adjective syntax in north-eastern French dialects  133 9.6 Interpretation  135 chapter 10 The syntax and pragmatics of discourse particles in Anglo-Norman 10.1 Discourse particles as a target of acquisition  139 10.2 Target properties of the Old French particle si  140 10.3 Data sources and analysis of particle si  144 10.4 Target properties of the discourse connectives ainz and mes in Old French  148 10.5 A comparison with Middle English  150 10.6 The connectives ainz and mes in Anglo-Norman  151 10.7 Implications  155 chapter 11 Conclusions 11.1 Overview of results  159 11.2 Two discontinuities in transmission  160 11.3 Envoi  162

139

159

Bibliography

165

Name index

175

Subject index

177

Preface

The idea that the writings of later medieval Anglo-Norman clerks might have something to tell us about the workings of the human language faculty might at first seem far-fetched. Yet thanks to the unusual circumstances in which the language they conveyed to us arose, it has much to tell us of relevance to current thinking on the nature of language. It allows us, in brief, to observe what happens when the ordinary environment for language acquisition is absent. Conventionally, language acquisition takes place through the coming together of a species-specific capacity for language, of linguistic input from the earliest years of life, and of interaction with caregivers in a nurturant home environment. In the case of later Anglo-Norman, however, it appears that the second and third were not generally present in the experience of its users, so that the first factor, the language faculty itself, can be studied under a different set of circumstances from the usual case. As with linguistic studies of the deaf and of language deprivation, though in a different way, we can ask how far the outcome departs from the language phenotype constituted by the regular form-meaning correspondences of a conventionally acquired mother tongue. In the period with which we are mainly concerned in this book, Anglo-Norman was a second language, though not one learned by explicit instruction. Its users acquired it in middle childhood, approximately between 5 and 7 years of age, in what appears to have been an immersion environment at school. Their mother tongue would normally have been English. Thanks to the very substantial amount of material extant in Anglo-Norman, spanning nearly 300 years, we will be able to see how far it was influenced by its speakers’ mother tongue English. In this study the question at issue, then, is how robust the capacity for acquiring a second language independently of a mother tongue is at that stage of life. I also aim in this book to address a long-standing problem in the linguistic history of England: Anglo-Norman is widely thought to have died out as a native speaker variety early in the 13th century. Yet from then on we find extensive and growing evidence of its influence on English, raising the problem of how a language variety that had virtually expired could have been so influential. Bringing to bear recent research into language acquisition, particularly the acquisition of a second language in childhood, the status of Anglo-Norman as a source language for contact with English is re-thought and put on a new basis.

x

The Transmission of Anglo-Norman

The book is laid out as, first, three chapters that situate the problem of later Anglo-Norman in relation to its historical context and to current issues in the transmission of language, a chapter that sets out the methodology and resources used, and then six chapters that analyse specific areas of language for the outcomes of its acquisition in atypical circumstances. In considering language acquisition, I have always been guided by Steven Pinker’s question: ‘What is acquired?’ (Pinker 1984), and I believe the equivalent question must be answered in diachronic study, in the form: ‘What changes?’ To answer it means proposing specific analyses of the linguistic properties that change. Therefore, the approach taken here is formal, in that I concentrate on linguistic forms, but it is not formalist, in the sense of arguing for a single theoretical analysis to the exclusion of others. The analysis chapters take a broadly descriptive approach to the linguistic features of medieval French (including Anglo-Norman) and English. The terms and concepts employed in the chapters on phonology, as well as those on the structure of nominals and of clauses can be found in standard textbooks. Where formal structural analyses are introduced, as in Chapters 8 and 9, explanations are given. The origins of the book lie in my curiosity about the nature of language acquisition in atypical circumstances, assisted by the ‘rich input’ on issues to do with second language learning provided by former colleagues at the University of Reading. Along the way, supervising students researching language change and language acquisition constantly refreshed my intention to consider what the limits are to acquiring a language successfully, and what maintains the transmission of language across time. Because the book assimilates research results from more than one subject area, I have been selective in the sources I have drawn on, and have limited my coverage to the perspectives and findings I judged most relevant to the central topic under discussion in this volume. This has probably meant in some cases that key issues in those subject domains, taken in their own right, may have not been given their due. This applies in particular to questions of what conditions ultimate attainment in L2 acquisition, and also to matters of syntactic theory. In both areas I have tried to adopt a stance that represents what would command fairly general assent in the field, for example invoking a ‘sensitive period’ facilitating acquisition, rather than a ‘critical period’ beyond which the human capacity for language acquisition is abruptly switched off. Accordingly, the book is not a guide to current state of research in the respective fields of enquiry, but draws somewhat eclectically on what contributes to a better understanding of the problems discussed.



Preface

A wide-ranging piece of research such as this has incurred debts of gratitude to a wide range of people, which it is a most pleasurable duty for me to acknowledge. Going way back, I would like to thank Neil Smith and Roy Harris, who in their very different respects, gave me the knowledge and the curiosity that eventually led to this book, Paul Fletcher, who showed me how to do language acquisition, David Wilkins, who sparked my interest in L2, James Hiddleston, who trusted me to do French, and more recently David Roberts, who trusted me to do English. My indebtedness to David Trotter for the incomparable research opportunities his Anglo-Norman Hub facility provided, as well as for his vast knowledge of the subject area, which will be evident on virtually every page of this book. Others from whom I have gained many insights on Anglo-Norman are Andres Kristol, Jocelyn Wogan-Browne, Serge Lusignan, Brian Merrilees, Ian Short, William Rothwell, and Marianne Ailes. At key moments, the help and encouragement of Eric Haeberli, David Lightfoot, David Birdsong, Pierre Larrivée, Christiane Marchello-Nizia, Tony Lodge, Paul Brand, Paul Kerswill, Laura Wright, Dominique Nouveau, Herbert Schendl, and Ann Curry have all been greatly appreciated. A research leave semester granted to me by Birmingham City University in 2011 made it possible for me to write up the book. Special thanks go to the series editors for their constant support, and their determination to see this book through to its final state, and to Anke de Looper of John Benjamin’s, who in the publication process was there at the right time and knew just what to do. I thank two anonymous reviewers who read the draft of the book in 2011–2012 and gave me much valued feedback. Any remaining errors and misjudgments are my own. Finally I should like to record my gratitude for the endless patience and good humour of my wife Fusae, which allowed me to reach the finishing line still in good spirits.

xi

A language is not some gradually and imperceptibly changing object which smoothly floats through time and space, as historical linguistics based on philological material all too easily suggests. Rather, the transmission of language is discontinuous, and a language is recreated by each child on the basis of the speech data it hears. Kiparsky (1968: 175)

chapter 1

Introduction to key issues

1.1

The research problem

Languages without native speakers die out. Unremarkable though this proposition may sound, it challenges us to consider some of the most fundamental issues in the study of the human language faculty. What is a native speaker, and what does it mean for a language to live? It is fairly easy to deal with such questions, at least for pedagogical purposes, by appealing to clearly contrasting examples, such as the different cases of Latin and of Icelandic. Latin now has to be learnt by a process of explicit instruction, and furthermore does not develop across time: the classical models and rules for instruction in Latin remained the same today as 500 years ago. Conversely, those who speak Icelandic have mostly learned it as a mother tongue, and it is a living language in the sense that it changes, as can be demonstrated by comparing its modern manifestations with Old Icelandic, which will show that there have been systemic developments in its morphology and syntax, as well as many lexical changes. Furthermore, native speakers acquiring it sound in certain ways a little different from their parents, a sign that a language is gradually evolving. Yet although the contrasting cases of Latin and Icelandic may seem to illustrate the issues in question well enough, clear examples do not establish principles. The language variety to be studied in this volume, Anglo-Norman, poses the problem that it appears to have continued to show systemic development, long after losing its native speakers. Far from dying out, it flourished to the point of causing concern in some quarters in the early 14th century that its continued use was threatening the position of English. Importantly, for upwards of 200 years, virtually all those who used it were bilingual, usually with English as their mother tongue. The question, then, is how to relate Anglo-Norman, as the second language in this bilingual setting, to currently viable notions of linguistic nativeness.1 The status of Anglo-Norman has long been debated, though often by historians and literary scholars whose specialisms lay at some remove from issues in linguistics. Nonetheless, it is of linguistic interest because it clearly raises issues of the nature of a form of language that is no longer a monolingual native speaker variety. In this study criteria will be sought on which to base our assessment of

2

The Transmission of Anglo-Norman

what counts as a native language variety, and which can be deployed in order to resolve the problem posed by Anglo-Norman. We return below to the challenges (and opportunities) inherent in addressing the matter in relation to a long-disappeared historical variety. First it should be noted that the core of the problem posed above is very much in the forefront of the linguistic study of contemporary societies. To look no further than English, there are overseas varieties, Singaporean English for example, that children learn as a mother tongue, even though sociolinguists might consider such varieties to have a different status from e.g. British or Australian mother-tongue English. The grounds on which such a distinction would be drawn would typically be that a variety such as Singaporean English is a second language. But nothing in principle prevents a person from being a native speaker of two or more languages. In practice, being a native speaker is generally related to how early the second language is acquired. Much recent research has taken place into this particular question, precisely because so much uncertainty has existed as to the distinction between a monolingual native speaker of a language variety, and a bilingual speaker of the same variety who acquired it in childhood. A central point at issue is whether the second language is acquired independently of the other language in the child’s entourage; in the case of Singaporean English this is usually a Malay variety of Chinese, typologically very different from English in its grammatical and phonological systems. Mutatis mutandis, the present study will focus on how far Anglo-Norman, typologically a Romance variety, was acquired independently of the originally Germanic system properties of Middle English. Some of the fundamental issues in linguistics can thus be examined by considering how a language is maintained as a second acquired variety. They are empirical questions, with undoubtedly major theoretical consequences. The generative linguistics precept that the object of study in linguistics is native speaker competence is invoked. So is the distinction drawn by Labov (1994, 2007) between the transmission and the diffusion of a language variety, which involves the difference between linguistically conditioned properties acquired in childhood, and features of a new dialect acquired by an adult who picks them up in adulthood. Both approaches to language, however different they are in terms of approach, have at their heart the concept of a natural process of acquisition. It is the replication of this natural process of native speaker acquisition that is crucial to the survival of a language. The key factor in language maintenance or death is the persistence of opportunities for a significant number of speakers to acquire the language within the time period of early childhood that seems biologically designed for that purpose, and language acquisition studies in the last half-century or so have provided a wealth of detail as to how this process takes place. Essential though this work has been, there is a sense in which it tells us only what we know already, that the human language faculty, in



Chapter 1.  Introduction to key issues

the sense of Chomsky (2000) works extremely well: children usually acquire something very much like their parents’ language. Far rarer have been studies that push at the boundaries of what is possible for the human language faculty. How robust is it, and how far can it cope with suboptimal conditions?2 This is the problem space in which the present study operates, the transmission of language ‘under exceptional circumstances’ (De Graff 1999). We ask what can be deduced about the human language faculty from studying an acquisition scenario that diverges in key respects from the conventional home language/mother tongue scenario. How far is language learnable under non-normal conditions that modify input characteristics, learner characteristics or both? The language variety studied in this volume was transmitted in a context that allows answers to be provided taking both set of factors into account. First, by the period with which we are concerned, it was almost never a mother tongue L1 acquired from birth in a nurturant home setting: it was acquired naturalistically from informal input in an educational setting. Second, learners were typically exposed to it in middle childhood, at a point when a mother tongue L1 had been acquired, but while the opportunity for naturalistic language learning, at least concerning some aspects of language, is generally thought to be still available. There are major debates over the nature/nurture question in language acquisition, especially over the concept of a critical period for it, but the usual view that ‘earlier is better’ in this respect is hardly seriously questioned. This is especially the case when it comes to the system properties of a language, notably its syntactic and phonological features. It will be seen that in Anglo-Norman a dissociation is evident by the late C13 between the transmission of inherited forms in phonology and in syntax. Whereas the latter had been successfully transmitted for about 200 years after the Conquest, phonology evidently had not. The analysis chapters of this volume provide empirical substantiation of this point. 1.2

Anglo-Norman and second language transmission

Let us now consider how a long-dead language variety can be studied in relation to the issues of language maintenance and acquisition posed above. What sources of evidence are there, and what can they tell us of the means by which the forms of language were reproduced? Anglo-Norman is the conventional name for the variety of French used in the British Isles between the Norman Conquest of England in 1066 and the beginning of the 15th century. Largely terminological debates have been conducted on the propriety of this label, and in particular as to whether a later ‘Anglo-French’ period should be distinguished from an earlier ‘Anglo-Norman’ period of ­insular

3

4

The Transmission of Anglo-Norman

French, but for convenience the designation ‘Anglo-Norman’ is retained here as it has institutional prestige associated with the Anglo-Norman Dictionary project and the major source of published texts, the Anglo-Norman Text Society. The choice of the label ‘Anglo-Norman’ in this study will be given no theoretical significance. Because those who acquired this variety left a rich written record, extending over hundreds of years, its evolution over time can be studied, and equally plentiful data sources from France will allow a comparison to be made with (varieties of) mother tongue mediaeval French.3 It will thus provide empirical evidence permitting the synchronic and diachronic study of language varieties distinguished by ordinary versus exceptional acquisition circumstances. While of course the experimental rigour of a laboratory setting can never be attained, it seems that the opportunity afforded by Anglo-Norman may be somewhere near-unique in the above respects. It can provide answers to the question: ‘Under what circumstances can a language live?’, as much as ‘Under what circumstances will a language die?’ Evidently, it lived, without native speakers in the conventional sense, for hundreds of years, and then, under well documented circumstances, it died. In this study the aim is to identify the ‘support system’ that permitted its extended survival, and whose removal caused its death. Answers will be provided by analysing the system properties of Anglo-Norman, especially its grammar and phonology, the levels that most decisively allow conclusions to be drawn about the psycholinguistic status of speaker varieties, i.e. whether they were native-like or not. Our sources of evidence are essentially of two kinds: the language produced by Anglo-Norman users extant in written texts, and the socio-historical information that is available regarding chiefly the contexts in which Anglo-Norman was used. It is the latter which will substantiate a view of the setting of acquisition which we have labelled ‘exceptional’. Admittedly, the school setting is found in many places in the contemporary world (especially in so-called language immersion schemes such as that practised in Canada from the 1970s onwards), so it could be studied directly, without considering data from the 13th or 14th centuries. But the studies made of such contemporary equivalents are essentially synchronic snapshots. They do not offer the opportunity to study over an extended time-frame the evolution of a language acquired in this way, nor to identify the factors potentially responsible for its dissolution. Above all, Anglo-Norman offers the opportunity to investigate over time the effect of language transmission where initial exposure to a language takes place in middle childhood. In particular, outcomes can be observed when the initial exposure, at around age 5, probably came too late for phonology to be native-like, but was in time for syntax acquisition to be native-like. In this volume those two dimensions are investigated systematically.



Chapter 1.  Introduction to key issues

There is a quite widespread belief among previous authors that AngloNorman (henceforth A-N) survived as a school subject in which older child learners received instruction. Accordingly, the possibility that A-N was transmitted as an instructed L2 will be treated as a plausible eventuality, even though the historical record offers no evidence for formal instruction until too late a date for it to be of explanatory value. Our approach will be to test this claim by looking at the language produced by those who acquired A-N, to see which account is better supported by linguistic analysis: A-N as an instructed L2 taught to older learners, or as a naturalistic variety acquired in mid-childhood. If the outcomes tend more towards showing the effects of naturalistic acquisition than of instructed learning, this will be counted against the claim that it was ‘taught’, in the sense of being the object of conscious instruction. To return to the exceptional scenarios envisaged by De Graff (1999), AngloNorman constituted an exceptional case of first language acquisition in the sense that almost none of its users acquired it as a conventional primary L1 in a home setting. By the period in which we have a large body of textual sources, the late 12th century, they were native speakers of English as well as speakers of Anglo-Norman. As shown by Short (1980), earlier distinctions between French and English ethnic communities when referring to historical events in England were replaced by using the term ‘English’ to refer to all subjects of the English crown, regardless of Norman or Saxon ancestry. Once it was no longer clear from individuals’ language what their ethnic origin was, the situation ceased to be one of a societally bilingual but individually monolingual country. Individuals enjoying superior status by birth or by education knew French, in its insular incarnation, at least as well as English. Lower status individuals generally knew only English. The primary first language of virtually everyone was English, it is thought, because even aristocratic infants were attended by English-speaking wet nurses. But higher-status children subsequently acquired French, to the extent of being able to provide the audience for a substantial literary culture, as well as providing the producers and end-users of an extensive central and local government bureaucracy conducted in AngloNorman. A key issue in this book will be to understand how this was made possible for so long. In other societies, cases of a non-native higher status language being used for administrative and literary purposes are common enough (presentday Francophone Africa for example). But this state of affairs has been made possible because French has in these contexts been an instructed language imparted as a school subject. As noted above, A-N does not appear to have been passed on in this way. This is the reason why it poses an interesting puzzle: it offers the prospect of understanding the outcomes of language acquisition, and the continued transmission of a language, in unusual circumstances, and in these respects may contribute to the study of the human capacity for language.

5

6

The Transmission of Anglo-Norman

1.3

Language transmission

The modern linguistic concepts deployed in this volume include the notion of the reliable transmission of a language in the context of child language acquisition, a point common to generative approaches (Lightfoot 1999, 2006) and sociolinguistic ones (Labov 1994, 2007). Both perspectives highlight the ability of young children to acquire linguistic systems accurately, though both maintain that variable input may be a catalyst for language change. The present research draws on these insights to shed light on the problem of the linguistic status of A-N. Transmission is defined by Labov (2007: 7) as ‘the product of the acquisition of language by young children’. It is characteristic of systems transmitted in such fashion that richly structured distinctions are observed, and that as systems become modified they are still linguistically constrained, albeit that in a subsequent state of language such constraints operate differently than in an earlier one. Transmission can operate as a process of change, in the form of ‘incrementation’ of a feature by children, who ‘reproduce and advance their parents’ system’ (Labov 2007: 51). Young children accentuate the direction of change, taking up a variant feature witnessed in their caregivers’ language and systematising it. The transmission of language in childhood is contrasted with the spread of a feature by diffusion, typically the learning of new forms by adults via contact with another language variety. Where modifications to language arise by diffusion, they tend to efface linguistic constraints on the distribution of forms.4 ‘The common marks of adult language learning [are] the loss of linguistic configurations that are reliably transmitted only by child language learners’ (Labov 2007: 14). This is dubbed ‘generational learning’. A study by Tagliamonte & D’Arcy (2009) of new grammar and discourse particle features in contemporary English showed that incrementation peaks in adolescent use, so young children may not yet drastically modify the frequencies of a new variant as compared with the input they hear. The change gathers momentum in the speech community itself. This is important for our purpose in this research because it means that the observed phenomena noted in our analysis chapters, of conformity with changes in continental French, do not have to be seen as somehow ‘imported’ into England from France, but rather can be interpreted as normal language acquisition in the transmission scenario of Labov and others. The approach taken by Lightfoot (1999, 2006), in common with much other work in principles and parameters research, is to see the role of the child in grammar acquisition as identifying cues in input to successful parameter-setting, thanks to an innate endowment of some kind which favours the acquisition of language naturalistically (i.e. without instruction) in early childhood. Language change may take place when modifications to input arise, as for example in



Chapter 1.  Introduction to key issues

dialect­ or language contact through population migrations (see e.g. Kroch & Taylor 1995, Kroch 2001). Although research into the acquisition of Universal Grammar parameters has in recent years been exposed to conceptual changes in how generative theorists formalise syntactic properties, it may be taken that some such approach to the issues in language acquisition and change is appropriate. The cues for acquiring syntactic properties, such as verb movement, that have been studied in recent years within the framework of such assumptions, however, remain unsettled and it would be premature to adopt them in the present enquiry. Nevertheless it will be taken that cues in input in some form are exploited by young children in L1 acquisition, though L2 learners are not able to make use of them systematically (White 1996). The term ‘language transmission’ as employed in this study therefore refers to the naturalistic process of acquiring the system properties of a language from the linguistic environment on the basis of some innate endowment. At issue in this volume will be whether the linguistic phenomena of A-N can be understood in terms of this conceptual framework, located as it is at the convergence of recent lines of enquiry into language change and language acquisition. This study draws on more than one research domain, where, as with any field of enquiry, important questions are always subject to further debate, so it will not be assumed that what are here presented as plausible conclusions will always remain so. Nevertheless, the extent of detailed information now available in second language acquisition studies, and the vast increase in searchable corpus resources, are so incomparably greater now than only a few decades ago that it seems opportune to use these approaches to take a perspective on Anglo-Norman which would not have been open to earlier authors. 1.4

The critical period hypothesis and second language acquisition

A fundamental issue for this research is whether proficiency in a second language is correlated in a meaningful way with the age at which it is acquired. There is a very large literature on this subject, by no means always pointing in the same direction. A recent summary of research findings by Dowden, Sanz and Stafford (2005: 109) concludes that research has shown that the ability to acquire language deteriorates with age, and states that ‘age on arrival, (a.k.a. age at first exposure) has been shown to be the best predictor of L2 performance.’ This conclusion was based on studies of non-mother tongue English immigrants to the USA, taking into account their age on arrival, and hence of their first exposure to English. These findings strengthened the critical period hypothesis, originating in the 1950s and 60s on the basis of findings in neuroscience, which proposed that

7

8

The Transmission of Anglo-Norman

children have a specific capacity for language learning which disappears later. However, when tested empirically using data from second language learners, research did not clearly uphold the notion of a cut-off point beyond which nativelike second language acquisition is impossible. Much controversy has surrounded the nature of the measures used in this research field by participants on both sides of the debate. Because of the existence of exceptions to it, some authors speak of ‘age effects’, and of a ‘sensitive period’, rather than a critical period (Meisel 2009), and this perspective will be taken in the present work. Support for age effects is provided by the findings of Johnson and Newport (1989), and of De Keyser (2000), in which L2 English speakers varying in their age of arrival in the United States were tested on a grammaticality judgement task.5 White and Genesee (1996) also found clear-cut age effects in their results. They surveyed French-speaking English adult learners with very high proficiency, and then investigated whether the subjects displayed native speaker-like performance on syntactic traits such as the empty category principle and subjacency. In the most native speaker-like subjects, no age effects were found, but among less proficient ones they were. Taking the results as a whole, in a group of advanced proficiency second language learners some – probably gifted – individuals acquire a second language to something like a native speaker-like level, while the others do not. The outcome relevant to the present research is that a high-proficiency second language-speaking population is likely to show some departures from native speaker norms, even though certain individuals may attain them. This predicts more variability in the grammar of A-N than would be expected if it was acquired by a native speaker population. As regards pronunciation, Flege (1999) found that production accuracy declines in linear fashion with age of learning, as the child fixes the pronunciation of his or her L1.6 The process of fixing the properties of the native sound system begins very early, as is known from Jusczyk (1997), so it is no surprise to see age effects in second language acquisition having an early impact. In Flege, Munro and Mackay (1995) a perceived degree of foreign accent characterised Italian speakers of L2 English if their age of arrival in the English speaking country was in the 2 to 4 years bracket. Flege (1999) reported that after age of first exposure c. 3 years, the chances of acquiring a nativelike accent are severely compromised. Long (1990) surveyed previously published studies that tended to indicate an early decline in native-like ability in pronunciation, whereas native-like ability in morpho-syntax is possible after a somewhat later age of arrival. In similar vein, Dowden et al. (2005: 111) point out that ‘Second language acquisition is not monolithic: the different components of L2 development (e.g. morphosyntax and phonology) may be affected differently by age, and asymmetrical outcomes are thus to be expected.’



Chapter 1.  Introduction to key issues

Much recent research on the critical period, notably by Birdsong (1999, 2006), seeks to discover whether native-like performance in a second language is possible at all, if only among certain exceptional individuals. This is not the issue under consideration in the present enquiry, however. What matters here is not the ultimate attainment of possibly exceptional individuals, but the general tendency reported in research on second language learners. It may be assumed that a second language-speaking population, such as those using Anglo-Norman in the mediaeval period, would have shown an average range of abilities in respect of learning a second language, had the process of learning it been subject to the usual conditions of learning a second language after early childhood, i.e. varying proficiency depending on aptitude, motivation and extent of exposure to the second language. Alongside a probably near-native level of performance in some cases, significant departures from native-like performance on systemic levels such as grammar would be expected. The alternative hypothesis, which we propose in the present work, holds that Anglo-Norman was acquired in middle childhood, and that morpho-syntax was acquired to a native-like level of proficiency across all or most members of the language community in question, exposed to the same kind of input in the same kind of setting. In the next section, the acquisition research literature relevant to these alternative scenarios is reviewed. 1.5

Early child bilingualism

Do early child bilinguals typically and reliably attain native-like competence in a language? In this section we consider answers to this question provided by research into the acquisition of contemporary languages by children aged three to six, and ask what the implications of these studies are for the problem of explaining the transmission of Anglo-Norman. The phenomenon of what is now called successive bilingualism has often been studied from the point of view of outcomes when a second language is added to a mother tongue during the later childhood years, especially with children attending school. A large literature has been devoted to the problems faced by the second language learner as compared with the ease of acquisition of a first language. In recent years, however, researchers have examined what happens when the children younger than six acquire two languages successively. Yip & Matthews (2007: 28) raise two central epistemological questions: how far the difference between the bilingual child’s dominant and non-dominant languages resembles that between a first and second language in second language acquisition, and how far apart age of first exposure can be before we speak of first and second language acquisition, rather than simultaneous bilingualism.7

9

10

The Transmission of Anglo-Norman

Meisel (1990, 2004, 2009) has argued that up to a certain age of first exposure, language acquisition by a bilingual, of both languages, is like first language acquisition, a theme developed further in Meisel (2011: 206): ‘If the age of onset happens after the age of approximately 6–7 years, the course of acquisition as well as the acquired grammatical knowledge increasingly resembles of adult L2 acquisition.’ Meisel (2011: 212) cites Hyltenstam & Abrahamsson (2003) in support of his conclusion that ‘the evidence in support of an [age of onset] around age 7 as a critical phase for L2 acquisition is solid enough to be accepted as such without further discussion.’ Recent research findings on child bilingualism provide further support for this view, pointing to the ability of a child to acquire the syntax of a second language independently of the first, as long as exposure to the second does not come too late. Rothweiler (2006) investigated Turkish children learning German, finding that the rather complex basic sentence structure rules of German were acquired accurately in a fairly short space of time – a number of months at most – by children beginning the acquisition of German at about age 3. Similarly, research by Dimroth (2007: 67) showed that of two Russian sisters acquiring German as an L2, one aged 8 years at first exposure, and the other 14 years, the younger acquired the V2 rule accurately, but the older one did not. These findings provide evidence for the sensitive or critical period hypothesis: environmental factors for these children were very similar, so that the key variable related to successful attainment was that of age of first exposure to the language. However, onset of exposure is not the only factor to determine outcomes. Language dominance in one or other of the bilingual’s languages will reflect factors such as the country or milieu in which the child is growing up, the parental language practices etc. Schlyter (1993) considers that a bilingual child’s weaker language is like an L2, showing incomplete or variable acquisition. Bernardini (2003) found that weaker Swedish-dominant children among Swedish-Italian bilinguals incorrectly placed Italian adjectives routinely before nouns, whereas Italian-dominant ones did not. Aside from the dominance question, Döpke (1997) and Yip & Matthews (2007) adopt an ‘interactionist’ view whereby a bilingual’s languages influence one another. In English-Cantonese bilingual children, placement of Cantonese locative PPs, they found, was influenced by English, whereas the formation of English Wh-questions was influenced by Cantonese. However, very early differentiation of phonological systems was in place, since there was apparently no cross-language influence in phonology. On the basis of the foregoing sketch of recent research on bilingual acquisition, it is now possible to set out some expectations that might be plausibly entertained as regards the form to be expected of A-N if it was acquired in childhood as a second language. As we saw in the previous chapter, the period of initial exposure would have been located, in the great majority of its users, in the middle­



Chapter 1.  Introduction to key issues

childhood years, at a point where early cognitive development is no longer a confounding issue with language development, and before the biological capacity for naturalistic native-like language acquisition ceases to be generally available. Anglo-Norman was not, generally speaking, a home language, transmitted from parent to child, but rather transmitted from other adults to a child, where the other adult appears to have been a member of the secular clergy acting in a pedagogical role. The distinctive properties of this variety are therefore that it was acquired as a second language, beginning in middle childhood, from input belonging to some kind of institutional register and typically directed to a group of children rather than to the individual child. The latter facts are relevant to a view of acquisition featuring adult-child interaction (Snow 1978, Tomasello 2001) in which caregivers facilitate acquisition by providing individually tailored feedback and by focusing on shared attention with the individual child (the ‘shared attention’ hypothesis of Tomasello 2001). Although proponents of an interactional view of language acquisition might see the absence of input from parental caregivers as a problem with the present account, the institutional setting nevertheless provided for individual adult-child interaction and was thus compatible with the successful acquisition of systemic aspects of language. Notes 1. The concept of ‘native speaker’ is not a straightforward matter. It has been widely problematised in British applied linguistic circles, though often on sociolinguistic rather than psycholinguistic grounds (see e.g. Davies 2003), and is often discussed in relation to questions of ‘Global’ English. In this study, we take nativeness to relate to exposure to a language from very early childhood, as generally in the studies to be surveyed in Chapter 2. 2. One might frame the question in the following way: Given that acquisition normally takes place very successfully on the basis of incomplete and imperfect input data, what would count as sub-optimal conditions? We have in mind here variation in environmental factors, rather than physical/pathological ones such as hearing loss. 3. This in itself opens up numerous problematic aspects of what can be counted as ‘French’ which cannot be pursued here. As is well-known, as late as the French Revolution it was estimated by the Abbé Grégoire that only a minority of the inhabitants of France were to be considered speakers of French, rather than of patois. 4. Labov illustrated this with the loss of constraints on a sound change, the tensing and raising of short-a, which originated in New York City, then spread to other cities in the USA, e.g. Cincinnati and New Orleans. In its original form, still transmitted in New York City, the sound change took place in closed but not in open syllables, though it did not affect function words. When diffused to the cities mentioned, the sound change was applied much more generally, without respecting these contraints.

11

12

The Transmission of Anglo-Norman

5. On the other hand, research by Birdsong (1992), and Birdsong and Mollis (2001) found no such correlation up to age 16 among Spanish speakers of L2 English. From that point onwards, however, there was a decline in performance in the latter study. 6. For example, L1 Voice Onset Time values of stop consonants influence L2 production (Flege & Hildebrand 1984). 7. According to an oft-quoted distinction proposed by McLaughlin (1978), the acquisition of two languages beginning before three is treated as simultaneous bilingualism while after three it is successive bilingualism.

chapter 2

Anglo-Norman and L2 varieties of medieval French

2.1

Introduction

The analysis chapters of this monograph take a diachronic perspective on later A-N starting at c. 1200–1250, so it will be useful to gain a picture of what this variety was like at around that time. In this chapter we first present the relevant features of Anglo-Norman morpho-syntax as far as can be established from texts attributed to the first half of the C13. In this way, a characterisation is provided of the target properties of the language at the stage where its transmission during the time-period studied began. At the same time it allows the reader unfamiliar with Anglo-Norman, or any variety of Old French, to gain an acquaintance with the main constructional possibilities of the language, in terms both of clausal and of nominal structures. Next, a comparison is drawn with two other varieties of French used in the medieval period, which were used by 2nd language speakers and were clearly quite heavily influenced by an L1. The aim is to focus attention on aspects of morpho-syntax which we might expect to betray the influence of English if A-N had been an imperfectly acquired L2. Much attention will be subsequently given in later chapters to how AngloNorman, as a type of Old French, differed from medieval English. It is important at the outset, however, to be aware that Old French also differed substantially from Modern French, and indeed in ways that made it rather similar to Modern English. An example of how different medieval French was from modern French can be seen in the following uses of the Old French verb + particle construction, e.g. corir sus (‘attack’) illustrated from the work of Chrétien de Troyes, e.g.: (1) a. …vergoingne/ Li cort sus.  shame to-him ran upon ‘Shame overcame him’

LionMsH (p. 91d.3157)

b. Corent li un as autres sus. ran the ones to the others upon ‘They attacked each other’

The verb + particle combination (V + prt) could be transitive:

Cliges (p. 53)

14

The Transmission of Anglo-Norman

(2) Il m’an metent le blasme sus  they me of-it put the blame upon ‘They blame me for it’

Erec p. 78

The Germanic syntax of V + prt is no doubt a relic of Frankish superstrate influence on Old French (Hilty 1968). Another respect in which Old French resembled English was its ability to use change of position verbs such as lever ‘(get up’) as pure intransitives, rather than with a reflexive pronoun such as se required in Modern French, e.g. leva in: (3) Li prodon par matin leva/au lit au valet s’an ala/la ou il le trova gisant.  Conte du Graal p. 367 The good man by morning got up to the bed to the servant self of-it went there where he him found lying ‘The good man got up and went off to his servant’s bed, where he found him lying…’

Thus care has to be taken in an enquiry of this nature to identify the properties of A-N that will be revealing when it comes to the larger question at issue, that of evidence for the status of A-N and how it was transmitted. It has not been unknown for authors in the past to find fault with A-N by judging it to be incorrect French, when the trait in question was in fact perfectly normal in the medieval state of the language. The next section therefore spells out the characteristics of C13 A-N that belonged to ordinary Old French of the period, but also considers some areas in which it presented idiosyncratic features. The two need to be carefully distinguished in order for the enquiry undertaken in this volume to proceed. 2.2

An outline sketch of early C13 A-N grammar

This study is particularly concerned with the observance in A-N of the properties of Old French syntax, as this is a domain in which English influence would be strongly expected if it was in its later stages an imperfectly learned second language (see § 1.4 above). The first half of the C13 is the jumping-off point for this enquiry, and to establish key features of A-N syntax at this period, three prose sources have been used: the Tripartita chronicle, written in the 1190s, but surviving in a manuscript of the early C14, the Reis des Engleis, a chronicle written in the 2nd quarter of the C13, and the Miroir de Seynte Eglise a religious instruction text attributed to the 2nd quarter of the C13, and published from a manuscript of the late 13th.



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

In terms of basic syntactic typology, the canonical S(v)VOX order in main clauses was as in Middle English, and is found in examples such as the following:

(4) a. Nus devuns aver quatre choses en chescun oreysun. we must have four things in each prayer ‘We must have four things in every prayer’ b. Deus fist plusurs miracles pur luy.  God made many miracles for him ‘God worked many miracles for him’

Mirour p. 50

Tripart l. 6

In (4) the pronominal S, nus (‘we’), is clause initial, the finite auxiliary (notated above by ‘v’) devuns (‘must’) precedes the non-finite main verb V aver, the Object (O) quatre choses (‘four things’) stands after the main verb and the X element, the adjunct PP en chescun oreysun (‘in each prayer’), stands last. Other than lacking an auxiliary, a clause such as (4b) also shows this clause constituent order. The same analysis in terms of grammatical relations can be applied to subordinate clauses, e.g.: (5) …si nus ne fasuns vostre pleysir en terre.  if we neg do your pleasure in earth ‘… if we don’t do Your will on earth’

Mirour p. 50

In (5b), after the complementiser si (‘if ’), the constituent order is once again SVOX. The main negator ne precedes the verb. These examples illustrate the basic typology of A-N as regards default clause order, but as with Old French generally, various displacement operations could alter default clause order. A clause-initial non-subject constituent typically triggered inversion of subject and verb, the V2 phenomenon found in modern­ Germanic languages but also to varying extents in Old Romance languages (Benincà 2004). Preposed direct objects, indirect arguments, and adjuncts normally produced VS order, as in the following examples, respectively: (6) a. La meite de tuz ses dreiz porchaz dunat il as abbeies. the half of all his right gains gave he to-the abbeys ‘Half his rightful gains he gave to the abbeys’ b.

Reis Eng p. 17

A la cunusance de vus meymes poez vus venir par suverene. to the understanding of you same can you come by sovereign meditacion Mirour 8 meditation ‘You may arrive at a better understanding of yourself by superior meditation’

15

16

The Transmission of Anglo-Norman

(6) c. Apres la quinte bataille firent il pais. after the fifth battle made they peace ‘After the fifth battle they made peace’

Tripart l. 15

Nominal subjects inverted in these configurations, as well as pronominal ones. Unlike Modern German, and Middle English, Old French allowed subject pronoun omission, but normally only if the subject, if expressed, would be postverbal, e.g.: (7) a. A la conusance de vus meymes poez venir en ceste manere: to the knowledge of you same can-2pl come in this way Mirour 8 ‘In this way you may arrive at a better understanding of yourself ’ b. Aprés ço devez savoir queus sunt les set vertuz.  after that must-2pl know which are the seven virtues ‘After that, you need to know what the seven virtues are’

Mirour 32

Note the minimal pair on the clause left periphery, comparing (7a) with Example (6b). Null subjects could also be used in subordinate clauses, but normally only when a constituent stood after the complementiser and before the tensed verb:

(8) a. Mes li Daneys le hasterent tant, ke IX fez en un an li rendirent bataille champestre entre autres assauz.  Reis Eng p. 15 ‘But the Danes harassed him so much that nine times in a year they gave pitched battle against him as well as other attacks’ b. Ci devez penser si vus avez Deu leaument ahuré, si sur totes choses le avez servie, si vos promesses li avez rendu, Mirour p. 28 ‘Here you should think whether you have faithfully worshipped God, if above all else you have served Him, if you have kept your promises to Him’

With a few discourse adverbs, such as nepurquant (‘nevertheless’) and d’autre part (‘in addition’), V3 order was usual, as in continental Old French: (9) a. Nepurkant il les volent bien aver.  ‘Nevertheless they did want to have them’

Mirour 44

b. De autre part vus estes si chaunjable ke…  ‘Besides, you are so changeable that…’

Mirour 10

It was common in A-N, as with OFr generally, for a Direct Object or a Prepositional Phrase Adjunct constituent in a clause containing a tensed auxiliary to be ‘embraciated’ between the auxiliary and the main Verb, as with the Direct Object constituents autri vie and sa vie in (10a) and the Prepositional phrase de Deu in (10b):



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

(10) a. Ço est droit, kar ki vot autri vie esparnier, il ne doit pas tolir chose ke doit sa vie sustenir. Mirour 30 ‘That is right, for whoever wishes to spare another person’s life must not take away his means of sustaining it’ b. Chekun hore ke vus n’avez de Deu pensé si avez perdu. Mirour 8 ‘Every hour when you have not thought about God you have wasted’

However, as the same was true of Early Middle English, this area of syntax will not be given further consideration in this study. In subordinate clauses with pronoun subjects, or subject relatives, we find complements or adjuncts preposed before the finite verb. These can be adverbs or non-finite forms, e.g.: (11) a. Ço est encontre glutunie, ke tuz jurs pense de charnauté.  Mirour p. 26 ‘This is against gluttony; which always thinks of the flesh’ b. Pernez garde tant tost si vus oez rien ke valer vus pusse a edificacion.  Mirour p. 22 ‘Take care right away if you hear anything that can be of moral worth to you’

The term Stylistic Fronting is often applied to this phenomenon in languages such as Icelandic. Again, as Stylistic Fronting also occurs in OE and EME (Fischer 2010), it will not concern us here. In A-N, as in continental Old French, the discourse particle si was used to link clauses, e.g.: (12) a. Quant il ne ot eir, si envea il le erseveske Roberd de Catoreberi en Normandie. Tripart l. 139 ‘When he had no heir, he sent Archbishop Robert of Canterbury to Normandy’ b. Li reis vint e se seignat de la croiz, si feri en la presse. Reis Eng 16 ‘The King came and made the sign of the cross, and charged into the mêlée’

A pronoun subject was typically non-overt, as in (12b), but if expressed had to be postverbal, as in (12a). The discourse particle si was also used to mark a topicalised constituent on the left periphery of the clause, e.g.: (13) a. Li reis de Nordhumberlond si aveit tute la terre de utre le Humbre jeske en Escoce. Reis Eng 12 ‘The King of Northumberland had all the land beyond the Humber as far as Scotland’

17

18

The Transmission of Anglo-Norman

(13) b. Apres Richard si regna Johan, sun frere.  ‘After Richard reigned John his brother’

Reis Eng 25

The syntax of the nominal, which in keeping with current generative practice we label Determiner Phrase (DP), resembled English in some but not all respects. Adjectives generally preceded the head Noun, as in Old French generally (Buridant 2000), though, as we shall see in Chapter 9, certain types of adjective increasingly tended towards post-head position. The determiner system included articles, demonstratives, possessives and quantifying words, all positioned on the left edge of the DP. One difference from medieval English was the existence of the so-called absolute genitive with animate possessors, as in: (14) Le seme comandement est itel: vus ne cuveyterez pas la chose vostre prome.  Mirour p. 32 ‘The seventh commandement is as follows: you shall not covet your neighbour’s goods’

In continental usage, this was being replaced in the C13 by the modern possessive de + nominal construction. Modifiers within the DP agreed with the head Noun in number and gender: in early C 13 A-N, agreement within the DP was respected as in Continental French, e.g. the appropriate plural -s and feminine -e inflections on totes cestes meseises in (15a). However, some adjectives, especially those derived from the Latin 3rd declension class, were invariable in form, as with esperitel in (15b):1 (15) a. De totes cestes meseises vus [ad] ja delivré vostre espus.  ‘Of all these hardships, your spouse has now delivered you’

Mirour 10

b. … baruns puissanz ke Jesu Crist a sei eslist a faire ceste espiritel bataille.  Sermon on Joshua, p. 12 ‘… powerful barons that Jesus Christ chose for himself to do this spiritual battle’

In Middle English, adjectives did not inflect for noun gender, now moribund in any case, or for number, although vestiges remained of the Old English/Germanic weak versus strong inflectional marking on adjectives, depending on the presence or absence of a determiner. We conclude this section with some points on the morphosyntax of functional elements, notably pronouns. Old French object pronouns normally stood preverbally by the C13 (though see Hirschbühler & Labelle 2005 for older uses). A-N respected this usage, including the need for clitic climbing to take place in an auxiliary + infinitive verb complex, as in le volez amer and les devez ayder below:



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

(16) a. (Pensez avanaument des biens k’il vus a fet suvent,) e plus fra si vus le volez amer enterement. Mirour 12 ‘(Think well on the good things He has often done for you) and will do again if you will love him fully’ b. Vus les devez ayder sulunc vostre poer.  ‘You must help them according to your power’

Mirour 30

Weak form object pronouns2 depending on non-finite verbs in A-N stood after them, as in OFr generally until the C13 (de Kok 1985), e.g.: (17) a. Pensez cum ço est grant pusance de fere teu chose de nyent, … e grant saver de ordiner les en si grant beauté.  Mirour p. 20 ‘Think that is great strength to do such a thing from nothing… and great wisdom to arrange them in such great beauty’ b. Si vout li abbes Aeiluuard…. remuer de la u il gut e mettre le aillurs.  Reis Eng 21 ‘And (he) wanted to move the Abbot Aelward from where he lay and put him elsewhere’

Strong forms could precede the non-finite verb, however: (18) Ço est a saver, prosperité de ço secle pur luy deceyvre par fauces suavitez.  Mirour 40 ‘That is to say, prosperity in this world to deceive him by false blandishments’

In addition to the clause negator ne A-N made considerable use of the reinforcing adverb pas in an ordinary negative clause, e.g.: (19) Le second comandement est itel: vus ne prendrez pas le nun vostre Deu en veyn.  Mirour 28 ‘The second commandment is as follows: thou shalt not take the name of thy God in vain’

But when the negated verb was followed by de + a nominal, the adverb point was used: (20) Le tierz comandement est cest: vus ne frez point de lecherie.  ‘The third commandment is this: you shall not do lechery’

Mirour 30

Ingham (2006a, 2006b, 2008) has shown how reliably C13 A-N followed continental Old French usage with respect to subject and object pronoun position and the use of pas or point in negative clauses.

19

20 The Transmission of Anglo-Norman

Although in the respects surveyed above early C13 A-N behaved much as any other variety of OFr, in other respects it displayed some individual characteristics. Most noticeably, the spelling forms are highly idiosyncratic, showing no evidence of regularization.3 There are other grammatical features in which A-N diverged form continental French nevertheless. For example, the subjunctive was routinely used in the apodosis of a conditional clause, e.g. seit in: (21) Onc ne porra ester c’ome ne face le sacrement de l’auter si il ne seit prestre.  Mirour 39 ‘…it can never be that anyone performs the sacrament of Mass unless he is a priest’

Strong form pronouns can sometimes be found in preverbal position normally reserved for clitics, e.g. the form li (for lui) in the following, where the clitic form le would be expected in OFr: (22) Les Jeus li pristrent et li tiendrent. ‘The Jews took him and held him’

Mirour 61

In addition, the distinction between the subject and object relative pronoun forms qui and que was quite often neutralised, e.g.: (23) Deus choses sunt ke desturbent hom de bien fere. ‘There are two things that prevent one form doing good’

Mirour 26

However, the second trait is also found in Old Walloon (Larrivée and Ingham 2010), and the last can be observed in Old Franco-Provençal, so such features in A-N can hardly be attributed to a lack of native-like competence in a language. To find such characteristics, one has to examine varieties of medieval French which are clearly unlike native speaker varieties, two of which are studied in the next sections of this chapter. 2.3

Old French as an L2 in southern Italy

In the previous section we have surveyed properties of A-N visible in the mid13th century demonstrating its adherence to the structural patterns of what Buridant (2000) called ‘common’ Old French, that is, the structures shared by the various dialects of the language he surveyed. Some of them do, and others do not, distinguish medieval French from medieval English. In the present work, we shall be concerned especially to observe outcomes where the former scenario was the case, so as to be able to observe possible mother-tongue English influence on A-N. Some empirical evidence is available to allow us to judge what to expect



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

when medieval French was heavily influenced by L1 transfer. In the late 13th century and beyond, French had the status of a language of culture in Italy, and was used for a range of literary purposes. However, it did not always follow Old French usage norms. A clear example of how its syntax diverged from common Old French under these circumstances can be seen in a work known as L’Hystoire de li Normant, extant in a manuscript estimated to date from about 1300. It is a translation of a Latin chronicle into French by a monk of Monte Cassino in Italy sometime after 1288, perhaps in the final years of the C13, if the manuscript dating is reliable. Its first editor (Champollion-Figeac 1835: xci, ff.) commented on the numerous Italianisms used in this translation, especially lexis and phrasing, and the syntax also displays features not found in other varieties of Old French, as will now be shown. In the syntax of function words, there are striking departures from Old French usage. For example, instead of possessive determiners, the translator most commonly employed possessive words as post-modifiers, as was possible in Italian, e.g.: (24)

Les choses soes HdLN 37; li compagnion lor HdLN 150; ‘His affairs’ ‘their companions’ lo intention mie HdLN 63; les armes nostres HdLN 156 ‘my intention’ ‘our weapons’

Whereas the C13 A-N texts surveyed above used personal pronouns, subject and object, much as Old French did, the HdlN translator departed significant from Old French usage in these domains. He used clitic object pronouns as French did not, making them resumptive on relative pronouns such as lequel: (25) a. … liquel Dieu tout puissant lo avoit fait ricche. ‘… whom almighty God had made (him) rich’

HdLN 231

b. …laquelle Guaymere, lo pere de cestui inique et malvaiz prince, l’avoit amée come sa sorur.  HdLN 132 ‘… whom (fem.) G., the father of this evil and bad prince, had loved (her) as his sister.’

These traits may be related to the syntax of Southern varieties of Italian which employ resumptive pronouns in a similar way. The use of non-pronominal clitics y and en was likewise overextended by comparison with common Old French, e.g.: (26) En lo monastier de Mont de Cassin tant petit de moines i remestrent… que…  HdLN 261 ‘In the Monte Cassino monastery so few monks remained (there) that …’

21

22

The Transmission of Anglo-Norman

The use of null subjects was overgeneralised to a context after the conjunction car, e.g.: (27) Quar avoient fet contre la loi del prince…  ‘For (they) had acted against the law of the prince’

HdLN 131

In Old Italian, subject-verb inversion was much freer than in Old French (Benincà 2004). Whereas in common Old French, VS order after a relative expression or subordinator is avoided, it was commonly used in the HdLN in subordinate clauses, with a relative expression or subordinator treated as a triggering context for V2, as in (28a) and (28b) respectively: (28) a. … laquelle amistié lui consenti Richard. ‘… which friendship Richard accorded him.’

HdLN 121

b. Ja estoit faite Casete et Magdelone, coment ai-je dit en l’Hystoire de li Longobart.  HdLN 59 ‘C. and M. had already been built, as I said in the History of the Longobards.’

The conjunction mais (‘but’) never triggered subject verb inversion in Old French, but the Italian translator overgeneralised VS to this context too, e.g.: (29) Il non manda epistole ne non manda message, mais vint il en persone.  HdLN 509 ‘He did not send a letter, nor did he send a messenger, but came in person’

In Old French, VS order could not be used to place a clitic + verb combination in initial position: this is known as the Tobler-Mussafia Law, after the philologists who first stated the principle. But in HdlN this is frequently violated, e.g. in the second sentence of the following example, where the Subject constituent lo misere follows the verb, in the absence of a constituent in clause-initial position before the clitic + verb combination se esforza: (30) Toutez foiz la malice de la supplantation de li judee vainchi la devotion de la religion de la foi. Se esforza lo misere de traire de mente ceste cogitation.  HdlN 301 ‘Nevertheless the malice of the Jews’ wiles overcame (his) devotion towards the religion of faith. The wretch tried to drive this thought from his mind’

A further illustration of Italian influence is the very general application of the se moyen impersonal construction of French, along the lines of Italian quasi-passive construction with ci, cf.



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

(31) Et de ces se sot le nombre quant furent ‘And of these it is known how many of them there were’

HdlN 509

As noted in Section 1 (see Example (3)), Old French tended to underuse reflexive pronouns with verbs with se, by comparison with the modern language, so the Italian translator’s tendency in the opposite direction is noteworthy; this quasipassive use of se savoir is unknown in OFr. The translator uses adverbs as noun premodifiers by analogy with Italian in ways not permitted in Old French, e.g. trop boiz in: (32) Mes por ce que estoit fort a monter et i avoit trop boiz… HdlN 492 ‘But because it was hard to climb and there were too many woods…’

Morphologically speaking, a very prominent trait of the translator’s L2 Old French is his avoidance of the -s plural inflexion after the plural definite article li. In ‘classical’ Old French texts of the C12, a case system was in operation whereby the -s plural inflection was used only when the DP in question had Object function. By 1300, this system had largely broken down in OFr, and the Italian translator of HdlN does not respect it either. Instead, he fairly consistently used -s as a noun inflection very generally, after all types of premodifiers, but not after li, cf. the following noun plural expressions: (33) a. les tors, celles choses, autres engins, nos pechiés, ses filz… b. li mur, li castel, li mont, li sarrazin, li chevalier, li autre…

This pattern seems to imitate, in part at least, the Italian li plural article and the absence of an -s plural noun inflection in Eastern Romance varieties’ evolution from Late Latin (Herman 1990). Such features as those surveyed here give us clear evidence of the non-native like quality of L2 French in the relevant period. They indicate the powerful influence of the grammar of Italian on users of French who can be considered as imperfect L2 learners, rather than balanced bilinguals.4 Although little information is available as to the context in which Italians acquired French at this time, is beyond doubt that it was one that favoured very substantial interference from mother-tongue Italian. 2.4

Law French as an L2

A second domain where L2 French syntax can be clearly observed is in the curious language variety called Law French, the last vestige of French used in mediaeval England. It was transparently not a native speaker variety of French, but

23

24

The Transmission of Anglo-Norman

was acquired as adults by lawyers in England in the 15th to the 17th centuries. Numerous texts written in Law French in the later 15th and early 16th centuries, especially law reports and treatises, testify to its peculiarities. Its morphosyntax departed radically from any other historical French variety in many respects, notably the levelling of grammatical contrasts. Gender contrasts were levelled even with possessive determiners, whereas in A-N they remained native-like, as we shall see in Chapter 7. Both masculine and feminine singular nouns took the masculine possessive modifier form: (34) a. son cryme, son damage, mon corps, mon profit, etc. his crime his damage my body my profit b. son cheyne dore, son mayn, mon liberte, mon robe, etc.  YBHVIII, Trinity Term (1520) his gold chain his hand my freedom my robe

Another non-nativelike grammatical feature of Law French concerned object pronouns. Where French would have used a clitic object with a finite verb, Law French used the strong form, usually postverbally, e.g.: (35) a. Come si home bate moy et devie.  ‘For instance, if someone beats me and dies’ b. Si jeo enfeoffe vous sur condition de payment. ‘If I enfeoff you upon condition of payment’

YBHVIII, 47 (1520) YBHVIII, 153 (1523)

Verb morphology in Law French is very clearly affected by the influence of English. The latter had no future/conditional verb form contrast, nor did Law French: (36) a. Sil ad tiel communication entre vous et moy que vous averez mon chyvall.  YBHVIII, I, 155 (1523) ‘If there is such a communication between you and me that you will have my horse’ b. Quar jeo ne fuy content que vous averes mon chyvall forsque pur £9.  YBHVIII, 160 (1523) ‘For I was not happy that you would have my horse for only £9’

Another striking instance of English influence concerns the loss of the conditional versus imperfect tense form contrast with modal verbs, where French clearly distinguished conditional forms such as pourrait from imperfect pouvait. This contrast is lacking in English, where could, might were originally past tense forms which then lost their past-referring nature, and refer to potential future eventualities. This levelling is reflected in Law French:



Chapter 2.  Anglo-Norman and L2 varieties of medieval French

(37) a. Issint fuit tenus que il poit prender cest issue.  ‘Thus it was held that he could take this issue’

YBHVIII, 42 (1523)

b. Le visconte poit arguer nostre autorite.  ‘The sheriff may dispute our authority’

YBHVIII, 43 (1523)

An extreme case of verbal morphology reduction is seen with the loss of verb inflections, such as the markers of infinitive, past participle and past tense endings: (38) a. …que il poit optayn le volounte le primer lessour. YBHVIII, 151 (1523) ‘That he could obtain the goodwill of the first lessor’ b. Ore il ad port son action sur sur cest endenture. YBHVIII, 47 (1523) ‘Now he has brought this action upon this indenture’ c. La verite del mater est que le defendant arrest le pleyntife.  YBHVIII, 42 (1523) ‘The truth of the matter is that the defendant arrested the plaintiff ’

Law French was acquired by young adults as they learnt the legal profession from the practice of the law. Training in legal disputation took place in the Inns of Court in London, notably in the form of ‘moots’, that is, sessions where established and trainee lawyers would engage in legal argumentation on particular cases. Law French was in no sense a natural language, used across a range of communicative contexts, but was a professional jargon, capable of great subtlety in its use of specialised lexis, but prone to blur grammatical contrasts not deemed essential for successful communication of a restricted range of meanings. As pointed out by e.g. Brand (2000), Law French is not to be confused with Anglo-Norman. Its value for our purposes is that it reveals plainly what was to be expected if medieval French were learned as a professional jargon, not as a naturalistically acquired variety. Perhaps even more than the French used by the writer of HdLN, its grammar diverged quite radically from native speaker norms and shows unmistakable L1 influence. 2.5

Prospects

The examination conducted in the last two sections of this chapter, of what forms medieval French might have taken when learned as an L2 and subject to L1 influence, gives some important pointers to what can be expected if A-N had been similar in nature. Grammatical contrasts not observed in the L1 ought to have

25

26 The Transmission of Anglo-Norman

been ­vulnerable, as with gender in Law French, and constructions extant in the L1 with a wider domain of application than in the L2 over-extended, such as VS order and null subject properties in Italian. While conducting our survey of A-N prose texts up to 1250, we found no evidence of any such features as regards the French of England, which in the grammatical areas investigated aligned itself with common Old French, suggesting that Middle English and A-N were acquired independently of each other. The case for regarding A-N as an instructed L2 thus appears implausible, at least until the early-to mid-C13. The task undertaken in this study is to see whether this remained the case subsequently. The question of the linguistic quality of later A-N is clearly a crucial one to the argument at hand, which one might have expected to feature prominently in discussions of its status by earlier authors, but this line of enquiry has rarely been adopted, as we shall see in the next chapter. Instead, commentators on the status of A-N have tended to base their interpretations on extrapolation from historical anecdote and the like. We next survey what is available in the literature taking such perspectives, before returning to more properly linguistic considerations in Chapter 4. Notes 1. Note also the pre-head position of the adjective esperitel, which was in line with earlier Old French order, as mentioned in the text. 2. The inventory of direct object pronouns in OFr was as follows:

Strong: 1sg moi 2sg toi 3sg lui/li 3pl eux Weak: 1sg me 2sg te 3sg le/la 3pl les 1pl no(u)s and 2pl vo(u)s belonged to both series.

3. A good indication, in fact, that it was not an instructed L2, as orthography is typically standardised in such cases, as with Latin. 4. In fact the lexis is also very frequently unlike that of Old French, often showing Latin influence, an indication that the translator was insecure in his command of Old French vocabulary.

chapter 3

The context of transmission

3.1

Previous interpretations

The means by which A-N was maintained in use has long been a matter of keen debate. In the early 20th century the view was current that extensive bilingualism prevailed in England (Legge 1941, 1950, 1980, Suggett 1946), and that French became a widely-used vernacular, remaining in use as such until the late medieval period. The pendulum was sent swinging back in the opposite direction by the work of Rothwell (1976), Berndt (1972) and Kibbee (1991, 1996), who claimed that outside a tiny aristocratic élite French was by the 13th and 14th centuries a language taught at school as an instructed language, and that it ceased to be a ‘true vernacular’ by the end of the 12th century at the latest. This notion has gained extensive acceptance in subsequent decades (Short 1980, Mensah 2008), and has become the prevalent orthodoxy among historians of English (Lass 1987, Thomason & Kaufman 1988), who understandably felt able to follow Romance scholars as regards what is a Romance language variety. Yet, as we shall see in this study, that interpretation is open to serious question. First, the qualitative estimates of A-N available to researchers rested largely on the work of Pope (1934) who considered that by the mid-C13 the French of England had ‘decayed’, insofar as it was no longer adhering to the phonology of continental Old French, but had undergone English influence. While this latter point was no doubt correct, and will be adopted here (see Chapter 5), to say that the grammar was in a ‘state of disintegration’, as Pope (1934) did of verb morphology, is well wide of the mark in the early to mid-C13. Secondly, the picture of Anglo-Norman as in its death throes in the 13th and 14th century ignores the fact that it was in this period that A-N saw a dramatic expansion in the range of its uses. In the second half of the C13, uses of French multiply in business recordkeeping, private correspondence, government business etc. The C14 became the heyday of insular French in terms of its use as a language of record. Any satisfactory account of the use of French in medieval England must therefore provide an explanation of the extent and quality of French as used in these later sources. In addition, much of that debate rested, it seemed to Trotter (2003), on a false premise: protagonists often assumed that once English was most people’s mother-tongue, French could not also be a competently acquired language. But

28

The Transmission of Anglo-Norman

in principle­ mother-tongue English speakers could also have learned French to a highly competent standard. The maintenance of French in medieval England relied on the existence of many less socially elevated individuals able to communicate effectively in the language. As also pointed out by Trotter (2003), for the ruling elite of a country to have conducted much of their day-to-day business in A-N must mean that those who executed their instructions in writing were trusted to do so competently and reliably. The effect of denying vernacular status to Anglo-French for most of its later career has been to deepen the mystery of how it survived so long without mother tongue speakers. The commonly held view that later insular French was ‘degenerate’ and ‘corrupt’, indeed merely ‘bad French’ (Price 1984), typically fails to look beyond superficial phonological and morphological idiosyncrasies. Such features, however, also characterise certain continental varieties, such as Picard and Walloon, and are not in themselves diagnostic of native speaker status or otherwise. In the first generations after the Conquest, the descendants of the invaders must have grown up in communities still using varieties of French as a mother tongue. By the beginning of the 15th century, on the other hand, French had unquestionably become an instructed second language in England, as we see from the numerous compilations of grammar, vocabulary, manières de langage, etc. found in manuscripts dating from around the end of the 14th century. Between 1066 and 1400, however, there is a gap of over 300 years in which the French language appears to have been passed on across about ten to twelve generations. If it was not a vernacular, how did it survive? What kind of knowledge base did the users of this supposedly ‘artificial’ language (Kibbee 1991) build upon? The psycholinguistic status of A-N has not received much attention from earlier generations of scholars. The only serious attempt to address the issue, Rothwell’s (1976) proposal that by the mid-C13 French had become an instructed second language in England, engaged directly with the problem of how Anglo-Norman was acquired, if it was not a widely used mother tongue. He argued that glosses, spelling and grammatical treatises, vocabulary manuals (esp. Bibbesworth’s Tretiz de langage) functioned as ‘manuals of different kinds to promote the teaching of French’ (Rothwell 1976: 463). An institutional setting of some kind must have been generally responsible for fostering the transmission of French, since the home environment of the young child is unlikely to have done so, outside knightly and aristocratic circles. However, the notion that in the 13th century and for most of the 14th century Anglo-Norman constituted an instructed foreign/second language taught in schools rests on little or no empirical foundation (Ingham 2007), and thus remains only a hypothetical solution to the language maintenance issue. It has nevertheless gained widespread acceptance thanks to a misinterpretation,



Chapter 3.  The context of transmission

or over-literal interpretation, of Rothwell’s (1976) hypothesis. Rothwell was reacting to an overly favourable view of the social spread of French within England propounded by earlier authors such as Suggett, Legge and others earlier in the 20th century. Rejecting their claims, Rothwell opposed not only the contention that French was used by large numbers of ordinary people, but also the notion that it was a naturally acquired variety for any significant segment of the English population. These two quite distinct propositions were not kept separate, either by Rothwell or by previous authors. Yet it is important to recognise two different issues: what sort of French was transmitted in England 1250 to 1400? How extensive was its use? An outcome can be envisaged in which Rothwell (1976) was largely right with a fairly restrictive answer to the second question, but wrong in his answer to the first. His instructional hypothesis on the first issue will be tested in this research in the way that seems most appropriate, by examining whether later A-N showed the hallmarks of imperfect language learning. Recent discussion of the subject continues to avoid taking into account the psycholinguistic dimension of how later A-N was passed on, and how it came to disappear, preferring to opt for socio-political accounts of its demise. Rightly, in our view, Miller (2000) identified the post-Black Death period, not the loss of Normandy to the English Crown in 1204, as a turning point, but he considered the key factor to have been the opportunity that the Black Death afforded the middle and lower classes to ‘acquire power’. He likewise called the upper classes ‘the last stronghold of French’, and Anglo-Norman an ‘elitist’ language whose demise was accelerated by the rise of the middle class. Yet by Miller’s own acknowledgement French was used for over half a century after the Black Death by such nonaristocratic groups as the London Grocers’ Company, to which we should add the fact that municipal documents continued to be drawn up in Anglo-Norman until around 1430.1 They were likewise the work of burghers, not of aristocrats. The Black Death, as we shall see, had a crucial effect on A-N, but this was especially because of its impact on its psycholinguistic status, and not so much because it triggered reorganization in society generally. 3.2

The historical evidence for the status of A-N

Contemporary testimony regarding the ordinary circumstances of language learning in our period of study is fairly sparse, presumably for the usual reason that writers typically do not need to mention what people already know. Still, there are some pointers. Some idea of the aristocratic context is provided by a roughly 1200-line poem written in the late 13th century by Walter de Bibbesworth for an aristocratic lady Dyonise de Monchesney, offering her a wide range of general and

29

30

The Transmission of Anglo-Norman

special purpose vocabulary to use in educating her son. The semantic fields it featured include a rich profusion of natural kind names, as well as those of artefacts. Just how the poem was supposed to be used as a teaching aid is unclear; although surviving manuscripts have some words glossed into English, there is no evidence that this glossing was the work of Bibbesworth. The lexis of the poem ranges from words for parts of the body that only a very young child would have needed to be told, to terms for brewing techniques and architectural features that would be relevant only as specialist training for adult life. Although puzzling in certain respects, Bibbesworth’s poem provides valuable information in that it depicts the mode of transmission in this aristocratic milieu as a mother-child dyad. It shows that the learner in question had already acquired the basics of grammar without which the lexical information would have been unintelligible, and makes it clear that insular French was at the time a viable mode of communication in a medieval estate setting. The latter point is sociolinguistically important: though presumably not used to address the peasantry, French was the language a young aristocrat was expected to use to communicate with those engaged in managing a feudal estate.2 A fair amount of documentary material is also available to clarify the status of A-N in its final stages. By the last quarter of the 14th century a different means of passing on insular French is in evidence: the formal teaching of French written usage was taking place in commercial establishments, i.e. private schools, run by teachers associated with Oxford University, in which instruction was given in writing letters and other documents in French (and Latin). A collection of manuscript material edited by Richardson (1942) shows that learners aimed to master the epistolary conventions used when writing in French for business and private purposes. Right at the end of the 14th century, conversation manuals were produced which model appropriate discourse in communicative settings, e.g. how to ask one’s way, converse with innkeepers, and so forth, and one of these manuals was intended for people wishing to travel in France. So by the late 14th century, for those in England with the means to engage in foreign travel, French cannot have been generally known as a medium of everyday communication. These late C14 developments, featuring model letter books and conversation manuals, attest to a change in the status of A-N by this time. Both correspondence manuals and conversation manuals present the learning of French in terms of advanced proficiency: they presuppose that ordinary grammar and lexis have already been acquired. They would have been of little use to learners unable to conjugate verbs or structure sentences, so the question is how those more basic abilities were acquired. No genuine French grammar book intended for English learners exists prior to one written about 1415 by John Barton. Although one cannot conclude from the absence of proof prior to the late 14th century that French



Chapter 3.  The context of transmission

was an instructed second language, there is a good deal of converging evidence that just at this time something qualitatively different happened as regards the status of French in England. Towards the end of the 14th century, vocabulary lists that paired Latin with French words, spelling rules, and grammar notes, and a conversation manual all appear in the same manuscripts with a clear pedagogical function (Kristol 1990). This coincided with other indications. English negotiators at the truce of Leulinghem in 1393 refused to use French because they were unhappy with the nuances their French counterparts employed in their use of French (Lusignan 2006). In peace negotiations prior to that date between the two countries, such as those of 1359, English negotiators had been content to use French. It seems that the proficiency of the English participants at Leulinghem was no longer up to the task of ensuring the wording was satisfactory to them: the truce agreement was written in Latin instead. Just at this time, private correspondence begins to appear in English, as do petitions.3 The late 14th century, then, saw a marked shift in the status of French in England. During the decades before and after the Black Death, it went from being a language that people lower down the social scale aspired to learn, even threatening the future of English as the national language, to being brusquely abandoned by the school system and parliament in the decades after 1350 as the everyday language for conducting business. The later decades of the 14th century were a time when children who survived the Black Death, which raged periodically between 1349 and the early 1360s, grew to adulthood and, if favoured by birth or education, assumed key roles in economic and cultural production. Writing before the Black Death, the chronicler Higden commented on an unusual aspect of English education: school pupils learned the curriculum (Latin rhetoric and logic in particular) not through their mother tongue, but via the medium of French. A monk known as Holcote, slightly earlier in the 14th century, confirms this latter point (Orme 1973). But Trevisa’s later 14th century translation of Higden, made around 1385, explicitly states that French is no longer used as a medium language in schools. In the late C14 and early C15, then, a syllabus of works for the teaching of French did in fact come into being, when we find grammars, manières de language and adaptations of Bibbesworth’s poem on vocabulary that overtly teach French using English. But by then the game is up for Anglo-Norman; the quality of the French declines just at the point we have clear evidence of its status as an instructed second language; nor is the association of those two developments particularly remarkable. The clear implication is that French needed teaching once it was no longer naturally acquired. As far as the noble élite was concerned, their use of insular French would probably not have been threatened by the disruption of schooling via French in

31

32

The Transmission of Anglo-Norman

public grammar schools. French remained the badge of ancestral superiority: tracing one’s lineage to the Conqueror and his companions meant membership of an elite class, so a strong incentive would have spurred them on to maintain facets of a tradition associated with their distinguished position, which included language. The resources at their disposal would have given them the means to afford resident tutors from whom they could have acquired French at an early age, and also through direct interaction with their parents, as was apparently the case with Dyonise de Monchesney and her offspring. It seems unlikely, then, that any catastrophic event comparable to the Black Death affected the transmission of Anglo-Norman in aristocratic circles. The loss of their lands in Normandy in the early 13th century, a factor often quoted in textbooks, may have played a role but not all nobles by that time had land in Normandy anyway. The onset of the Hundred Years War in 1337 may have deprived English aristocrats of opportunities for peaceful interaction with local French people that they would have had when living some of the time in Normandy as landowners. But texts initiated by English aristocrats and gentry in Anglo-Norman are still plentiful in the late 14th century. One of them, the Duke of Lancaster, composed an outstanding work of religious piety in A-N in the 1350s, another distinguished English leader in the Hundred Years War, the Black Prince, had his life commemorated in French, and John of Gaunt’s business correspondence was still in French in the 1370s, so not too much weight should be placed on the notion that the outbreak of the wars with France caused a shift in the language attitudes of the elite classes. The historian Froissart, who visited England in the 1370s commented that the nobility used English at court, though whether he meant exclusively is not clear. He also noted that the English nobility still ‘taught’ their children French, about 100 years after Bibbesworth wrote his poem for the use of Dionyse de Montchesney. Further evidence of the aristocratic setting as a context for acquisition, almost as a finishing school, can be seen in the 14th century Livre de Courtoisie, an A-N poem intended for the education of young page boys, which expounds in French the good manners expected in a high-status household. However, historical testimony about the educational practices of the aristocratic elite does not suffice to explain how the ordinary users of A-N, those responsible for the written texts in which it has survived, acquired their competence in the language. This class of individuals were the ones who, as clerks, were responsible for the creation of the Anglo-Norman manuscripts that stand as witnesses to that language, so the question of how they came by their French stands at the heart of the transmission problem studied in this work. The next section therefore pursues what can be gleaned as to the context in which the nonaristocratic scribal class came by their knowledge of A-N.



3.3

Chapter 3.  The context of transmission

The elementary school as a context of first exposure to Anglo-Norman

School pupils (always boys) began grammar school at seven, at which age they began to learn Latin grammar. As mentioned above, two early 14th century contemporary witnesses, the monks Holcote and Higden (Orme 1973) noted that French was used as the medium language through which the grammar school syllabus was taught. Evidently, schoolboys knowing no French could scarcely have learned their Latin grammar lessons: to imagine that teachers imparted one unknown, Latin, through another unknown, French, is a pedagogical absurdity. Rothwell (1976), aware of the problem, opined that in reality teachers and pupils must commonly have ignored the proscription of English. To some unquantifiable extent, this may have happened, though Rothwell’s view remains no more than conjectural. However there is another possibility we would like to propose as to how school pupils coped with French-medium Latin classes. The grammar school was not in fact the first school pupils typically attended. According to Orme (1973), they initially went to a school run by the Church, probably held in the church building itself, in which they learned basic subjects like the alphabet and singing. The name ‘song school’ (schola cantus) was often used for this kind of elementary school, which pupils seem to have started at about age 5, when abstract formal subjects such as Latin grammar would have been beyond them, but the cognitive tasks involved in learning letter shapes and sounds (linguistic or musical) would not have been. These first schools took charge of teaching basic literacy, and thus provided the foundation on which the grammar school curriculum and hence the training of medieval England’s literate class were built. Leach (1915) compared song schools to elementary schools, whereas grammar schools formed the equivalent of secondary education. The modern comparison is not exact, since pupils usually seem to have started grammar school at about age 7. However, there is no doubt that the grammar school curriculum was not the first educational experience of an English child. We suggest that in order to cope with French as a vehicle language for the teaching of Latin at grammar school, children were initiated into the everyday use of French at their first school. A-N texts were produced by people who attended such schools. These schools were run by the clergy: the Lateran Council in the early C13 promoted the establishment of schools associated with the church in many localities, large and small, which would necessarily have included the provision of initial literacy. We know that the English clergy in the C13 and C14 made extensive use of French, from regulations and correspondence written in that language. If they did so when interacting with pupils from the start of their school experience, the latter would have received a kind of language immersion at a very young age, when they would have been more likely to pick it up without conscious effort than when they

33

34

The Transmission of Anglo-Norman

were older. Furthermore, knowing that in a few years’ time pupils would need to be able to understand French used in spoken form as a vehicle language in order to succeed in their grammar school studies, the clergy would surely have had an incentive to provide them with initial familiarity with French. In addition to having the motive, the clergy (secular and regular) had the means: 14th century monasteries enjoined the use of French, rather than English, with presumably senior members having to lead by example. The senior clergy are known to have corresponded in French. Elementary schooling at that period left little or no formal record, so it is impossible to know for sure which language was actually used as a medium of instruction, but if it was A-N, and if pupils assimilated enough of it to get by for day-to-day communicative purposes, the problem of how English grammar school pupils coped with French-medium Latin classes may be solved. Before going to the grammar school, they would already have had about two years of A-N classroom discourse, giving them a basis for communicative needs, such as the ability to follow a teacher’s explanations. The acquisition of A-N under such circumstances would of course have been an incremental process, taken further with the experience of using French for ordinary classroom communication at grammar school. The proposal here is thus not that the proficiency of a seven year-old in England entering grammar school would have been the same as his counterpart in France. However, the initial period of naturalistic exposure to French for about two years prior to that point would have enabled him to acquire the linguistic means for expressing the principal cognitive domains interfacing with language. Among these one might list the determination of entities, marking of temporal relations, the clausal construction of a proposition, propositional linking devices such as conjunctions and relatives, discourse markers of various kinds and no doubt numerous other aspects required for ordinary communication. For non-aristocratic users of French, the scribal class who became clerks and administrators, we argue that initial exposure to the language and the acquisition of its basic grammatical properties must have taken place in the elementary school context. In adulthood, those in authority over them, especially nobles and prelates who no doubt acquired French privately, would have initiated the correspondence, charters and other text types that scribes dealt with. But clerks could surely not have carried out these tasks efficiently unless they already possessed the capacity to understand what they were supposed to write. Indeed, they would presumably not have been trusted to do so, when another language, Latin, existed, which allow those with a formal education to communicate reliably with each other. Any notion that clerks writing Anglo-Norman were uncomprehendingly taking down the texts dictated to them therefore cannot be taken seriously.



Chapter 3.  The context of transmission

The English school system was thus the principal factor in transmitting insular French, we contend, but not in the way envisaged by earlier authors who invoked the teaching of French. The roughly 300 years in which French was maintained as a vehicle language at school were the time when Anglo-Norman remained viable as a variety used for a wide range of written purposes in England. After the practice was abandoned, its use gradually faded out within a couple of generations. By the early C15, it can be seen from John Barton’s Grammar of 1415, French has become a formal object of study, a subject to be taught with some sort of pedagogical apparatus as a foreign language, in effect. Likewise, in the 1460s, one of the Paston family copied out the verb conjugations in French, perhaps during his time at university. The tradition of using French as a medium of ordinary spoken communication in schools had been comprehensively broken by the time John Trevisa translated Higden a generation after the Black Death. The reason for this change in the C14 in the status of A-N from being a medium language in grammar schools, to being essentially a foreign language acquired through conscious learning, is to do with the Black Death. With its heavy incidence in main population centres, it struck hard at the school system, since grammar schools were generally located near cathedrals or churches in major towns (Orme 1973). It is also known to have taken a heavy toll among the clergy, required by their calling to attend to those sick and dying of the plague. Soon, the church was having to ordain aspirants to the clergy whose Latin was not up to pre-Black Death standards, and who are unlikely to have been proficient in A-N. Since teachers were members of the clergy, the consequence in many places was that the use of A-N as a vehicle language in school could hardly be sustained. The transmission system of Anglo-Norman had collapsed, and its disappearance from the scene in England (except in law) was not long delayed. 3.4

Implications

In this chapter, evidence has been reviewed for what can be ascertained of how Anglo-Norman was transmitted in the three centuries or so after 1066. Three main loci are identified for this process: high status households, religious institutions, and educational establishments (schools and universities).4 In all of these, some kind of hierarchy of authority can be discerned by which decisions were implemented regarding language choice, and in which senior members were in a position to provide a linguistic model for junior members to follow. Thus a framework was provided for language transmission across generations. As regards the matrix of language transmission conventionally taken as the norm in modern-day language acquisition studies, the infant-caregiver pair, little or no information is available.5

35

36

The Transmission of Anglo-Norman

The school has been identified as a setting in which A-N was routinely used for communicative purposes up to the mid-C14. Inculcating a subject such as Latin grammar requires a medium language in which explanations referring to quite complex and quite abstract concepts would have needed to be understood, as a glance at any traditional Latin primer will confirm. So the ultimate linguistic attainment of learners of A-N would have had to reach a level where cognitively sophisticated mental operations could be expressed. On the other hand, as regards the starting point of the process, teaching children how to draw and write shapes and sing musical notes may be feasible using initially very limited means of communication, as long as the learner can perceive an object of cognitive attention that the unfamiliar language relates to. The L1 acquisition experience involves precisely this dynamic, according to just about every theory of acquisition, whatever position one might take as regards a richly structured innate language faculty. A bootstrapping process operates under which the learner’s cognitive organisation of perceived experience facilitates identification of linguistic features of the ambient language. The nature of the song school context could thus have provided English learners with an acquisitional ‘entry wedge’ into the properties of French, which they would have picked up without explicit formal teaching, thanks to an orientation towards the well-recognised ‘here and now’ context that is characteristic of L1 acquisition. Simple, repetitive, cognitively salient activities accompanied by child-directed discourse by an adult, and young children’s own linguistic and other cognitive abilities, would have provided the environmental conditions for acquisition of Anglo-Norman. We posit, in keeping with the position taken in the introductory chapter, that young children have a cognitive capacity permitting, from a very early age, the identification of linguistically significant traits available in the primary linguistic data, the ambient language. This capacity will be referred to as the language faculty, without commitment to a theoretical or neurolinguistic stance on the architecture of cognition. In the relevant respects, this cognitive capacity functions extremely efficiently (in non-pathological cases) when exposure to the language in question takes place in the first few years of life, but then diminishes in efficiency the later such exposure occurs. The decisive event affecting the status of A-N that has been identified is the Black Death and its impact on language used in the school system, which undoubtedly disrupted the transmission mechanism of Anglo-Norman, even though it carried on being used in England for some time afterwards. Evidence will be sought in this study that Anglo-Norman texts produced after the Black Death reveal qualitative changes from those written before it. If so, these texts would allow us to infer a change in the nature of the grammatical system used by



Chapter 3.  The context of transmission

their writers, and that change could then be related to the disrupted context of transmission. There are many unknowns and imponderables to be faced, but the hypothesis to be pursued in the rest of this volume can now be stated. Up to about 1350, the hypothesis that A-N was a naturalistically acquired variety should be reflected in accurately transmitted linguistic properties; thereafter evidence should be available of a loss of L1 status, that is, evidence of error patterns pointing to L1 influence from English. The design of the research undertaken in order to test these predictions is set out in the next chapter. Notes 1. Factual errors likewise affect Miller’s account of the 1250–1350 period: Bibbesworth did not write a ‘textbook’ to ‘teach French’ as a native language, as Miller claimed, but a poem to teach hard words to a child who already had a very good understanding of French: see the next section. 2. The fact that the Bibbesworth manuscripts were being glossed into English in the early 14th century is also significant. It is difficult to imagine any use for the work except as a pedagogic instrument of some kind – information on such things as brewing techniques was doubtless available in more practical form than via a poem. No doubt Anglophone learners and possibly teachers were already utilising the poem as a more general way into French, focusing on the hard words. Whether the glosses testified to informal self-instruction or to a more institutionalised process cannot be determined. 3. A little earlier, French started to lose ground in formal contexts in England, cf. in particular the opening of parliament in English (1362) and the statute of pleading, which explicitly claimed that French was not very well known in England, and hence the courts should not use it for purposes of pleading (Ormrod 2005). Local courts (manor courts) are thought to have used English anyway throughout the Middle Ages in order to deal with petty offenders, so the statute of pleading was relevant only to higher courts, notably the King’s Bench, where more important and financially substantial cases were heard. Almost by definition, these involved higher social levels: gentry, nobility and upper clergy. Even though the statute seems to have been disregarded for some time, its passing is significant. By the late 14th century, French was poorly enough known among these classes to make it desirable to modify statute law. The ­Anglo-Norman period was entering now a final twilight zone where its transmission was faltering and would soon cease altogether. 4. Since the clergy took charge of education, there is an overlap between the church and the school contexts. 5. Héroard’s diary (Foisil 1989) may offer suggestions on how a high-status child’s linguistic interaction with caregivers proceeded a few centuries earlier.

37

chapter 4

Rationale and design of the study

4.1

Initial considerations

The rationale for the analyses to be undertaken in this study is that a good deal is now known about the characteristics of early second language acquisition versus instructed L2 learning, and that Anglo-Norman can be fruitfully investigated from this perspective. In Chapter 1 the main findings on child second language acquisition and bilingualism, especially regarding syntax and phonology, were reviewed. If, alternatively, A-N was an instructed L2, it should have exhibited influence from L1 English, on the basis of contemporary L2 acquisition research. The key notion here is that of transfer, defined by Paradis & Genesee (1996: 3) as ‘The incorporation of a grammatical property into one language from the other’. If A-N users acquired French as a foreign language, we expect to find that areas where medieval French diverged from medieval English should have been vulnerable to transfer errors.1 There were indeed various aspects of Old French syntax which differed from Middle English, such as the behaviour of pronoun subjects after an initial constituent, the marking of noun gender, the placement of spatial particles immediately after the finite verb, the use of an imperfect tense distinct from the preterite, different verb constructions, plural agreement on adjectives, and discourse particles which could trigger subject-verb inversion. In all these cases, negative transfer might have occurred, in the form of word order following the English pattern, the inappropriate use of tense distinctions and gender or number agreement, or of verb conjugations or of discourse particles. In the domain of morphology, transfer would take the form of erratic observance of a distinction, such as noun gender, where it was not drawn in Middle English. These are clear pointers therefore, that can be searched for, so as to decide whether later Anglo-Norman was an L2 learner variety. If it was, negative transfer phenomena should provide evidence that L1 English was influencing the acquisition of insular French. Even then, caution is required. The language of bilingual speakers is known to reveal the influence of one language on the other to some extent, a phenomenon studied under the rubric of contact language varieties (see e.g. Mougeon & Beniak 1991). Some cases of negative transfer from English to A-N might therefore be compatible with a naturalistic acquisition scenario. What would not be

40 The Transmission of Anglo-Norman

compatible with it, however, would be widespread and systematic disregard of specifically the morpho-syntactic traits that were vulnerable to influence from the native language (English, in this case). If such an outcome were observed – as in Chapter 2 we saw that it was in Italo-French and in Law French – it would be hard to deny the status of A-N as an L2 learner variety. The procedure adopted in the present work will therefore rely heavily on investigating areas of language where the properties of medieval French and medieval English contrasted, so as to see how far they were vulnerable to influence from English. Certain challenges that were encountered as regards continental Old French need to be identified. Issues have arisen in the nature and interpretation of historical corpus resources, to do especially with the selection of materials, but in part with their editing as well, which is sometimes rather uneven in character, as happens when works edited in line with contemporary practice are included in a corpus together with those edited according to the practices of the 19th century. There is much to be said about the desirability of enhancing the quality and genre balance of historical corpora of French, but for present purposes we can do nothing but acknowledge their limitations, and elaborate strategies for working around them. Another dimension that poses a challenge is the extent of variation in past states of language. Historical linguistic treatments of French, in particular, tended to show rather limited appreciation of the interaction of diachronic and diatopic variation, a respect in which historians of English, especially of the Middle English period, have shown greater awareness, establishing extensive dialect variation in pronunciation and morphology and to a more limited extent in syntax.2 Points of non-uniformity in medieval English will be mentioned in individual chapters as they arise. Chapter 9 addresses an area of linguistic variation in medieval French, using new continental data to clarify a particular problem; otherwise, however, the orthodox position is followed that regional variation in Old French syntax can be treated as insignificant (Buridant 2000: 742–3).3 Indeed, a great deal of the time, the research strategy in the present volume is to seek evidence precisely of a lack of variation between insular and continental Old French, asking whether Anglo-Norman patterned more in keeping with continental French or with Middle English, on linguistic traits which would be informative as regards its transmission. Framed in this way, the research issue inevitably idealises each of the two medieval languages, whereas in reality they displayed variable traits, as all languages do. With phonology, we deal with this problem by addressing dialectal variability in pronunciation directly, as will be seen in Chapter 5. As regards syntax, however, the usual consensus among scholars of Old French is that there was in fact very little dialectal variability in this domain, so a policy of idealization, in this respect at least, seems relatively benign.



4.2

Chapter 4.  Rationale and design of the study

Target areas of medieval French

In this chapter, the design of the research into areas where mediaeval English and mediaeval French contrasted is set out, providing areas where research could be undertaken so as to clarify whether the two languages, the latter in the form of A-N, were acquired independently of each other, or whether English influenced French. Medieval French forms and constructions are taken to constitute the target of acquisition, and medieval English the mother tongue that could have been responsible for transfer errors. The concept ‘target-like’ is not straightforward: in Chapter 2, we considered non-native varieties of a foreign language whose presumed target could be taken as a kind of French used in France. But this presupposes that some kind of educational process was followed by which a pedagogical authority modelled learners’ language on an authorised target variety. Whereas with the teaching of Latin this is a justified assumption, as regards mediaeval French it is not. Practitioners of Law French surely did not set out to learn French as used in Paris, and the same may have been true of Italo-French, so in these cases a ‘Gallic’ target of learning cannot be assumed as a conscious objective. For any particular generation of A-N users, the target in a practical sense would have been their local variety. While this is true, the analytic issue for our purposes still remains the one discussed at length in previous treatments of A-N: how far it patterned, or did not pattern, with continental Old French, and for this task, if only as a heuristic device, a standard of comparison is unavoidable. The present enquiry cannot proceed merely by itemising the features of Anglo-Norman; it seeks an answer to whether it differed from varieties of mediaeval French in ways that would be expected if it were not a naturally acquired language. The standard of comparison must therefore be what was undoubtedly L1 French, that is, French produced by people who we may reasonably assume learned it as an L1, and were located in the langue d’Oïl area of what is now Northern France. Continental dialects of OFr differed considerably, but mainly in phonology and morphology (Pope 1934); syntax seems to have shown little dialectal variation. There is of course no initial assumption that Anglo-Norman was definitely not L1 French, merely that at the outset of the investigation its status is uncertain. For phonology, comparison will be made with the range of variation reported for this domain in continental varieties, particularly bearing in mind the fact that A-N originated as a Western dialect of French thanks to the preponderance of Normans and other Western French speakers among those participating in the Norman Conquest and subsequent settlement of French native speakers in England. As regards lexis, it might be worthwhile to consider L1 transfer from English in this area, but lexis as conventionally understood is different from the other

41

42

The Transmission of Anglo-Norman

levels­in that it does not possess system properties. Lexical information is standardly treated as itemised knowledge: a word’s meaning is generally not predictable by rule. L1 transfer, both positive and negative, is undeniably a fact in the domain of lexis, given the existence of cognate forms which may or may not mean the same in the two languages. Today, false friends such as library and actually commonly trip up their L2 learners.4 In principle, therefore, L1 influence on A-N lexis could be investigated. For example, there were perhaps Latinate English words, already established with a particular sense before the coming of the Normans, which had cognate Latinate Old French words with a different sense. If A-N used such Old French words in a way that reflected the sense in which it had been borrowed into English, a case could be made for L1 influence on A-N in lexis too.5 But there are many practical difficulties, involving extensive lexical research on individual words and their senses, which were judged insuperable within the confines of the present work. Although in principle lexical research into A-N is possible and desirable, the parameters of enquiry, notably etymological and semantic, are so different in nature from those envisaged in enquiring into syntax and phonology that in the present research lexical transfer as a measure of English influence on A-N will not be pursued. In the analysis chapters of this volume, then, evidence is sought as to whether A-N system levels, its phonology and syntax, showed a tendency towards acquiring stable systems in childhood, versus the unstable variability found in later acquired second language varieties. We initially considered morphology as a further domain of enquiry, since morphology is a well-known area where child learners establish regular paradigms without particular difficulty (Berko 1958), whereas post-childhood learners generally have greater problems: De Graff (1999) noted that child learners impose regularity on primary linguistic data while adult learners introduce ‘unstable innovations’, partly stemming from ‘transfer strategies’. To assess how far morphology showed evidence of one or other of these tendencies would certainly have been of interest, but serious obstacles stood in the way. Continental Old French morphology is rather variable anyway (see e.g. Pope 1934, Buridant 2000), so variability or instability per se would not necessarily be very informative as regards native speaker status. Secondly, in written forms of inflections, phonological confusion, rather than variable morphology, seems to have been responsible in A-N for the numerous apparent deviations from continental Old French. A final obstacle is that the investigation of age-ofonset effects in second language inflectional morphology acquisition has been less well documented than the equivalents for phonology and syntax. For these reasons, morphology has not been given much consideration in this study, other than in relation to phonological and syntactic issues.



Chapter 4.  Rationale and design of the study

Let us now consider what properties would provide evidence as to the independence of later Anglo-Norman from mediaeval English, and thus of its ­acquisitional status. English was the presumed mother tongue of virtually everyone producing the later Anglo-Norman prose texts we possess. A methodological requirement already stated is that a trait selected for study should have had a form in mediaeval English that was different from mediaeval L1 French, in order for us to see whether Anglo-Norman followed the former or the latter. In practice, the construct ‘mediaeval English’ is hardly less problematic than ‘mediaeval French’, and in some ways even more so. For one thing Middle English showed great dialectal diversity, and the more conservative southern dialects retained archaising forms, as well as word order patterns that sometimes made them resemble L1 French.6 Secondly, it was evolving across time, morpho-syntactically speaking. Word order patterns involving pronoun position and inflected verbs appear very different in a 14th century text from one of the 12th century. Thirdly, Middle English is very unevenly represented in terms of surviving prose texts, our preferred data source for syntactic analysis. After the first half of the 13th century, there is a gap of 100 years with virtually nothing except a couple of translations from French (an obviously undesirable text type with which to identify independent properties of the two languages reliably). In the late 14th century, prose sources once again become more plentiful, though still in some cases translations from French. Perhaps the biggest methodological problem in the path of the investigator is the growing evidence of French influence on Middle English in the 13th century. Thus it would be misleading to seek to establish the properties of English during the period 1250 to 1400, as we intend to, as if it were an entirely independent entity from French. For this reason, parameters will be selected that, in Middle English of the early 13th century, which happens to be reasonably well supplied with surviving prose texts, show contrastive properties with these of Old French. At least at the beginning of the period of study, the L1 English and L1 French grammatical systems can in these respects be established separately, allowing the possibility of transfer from English to L2 French to be meaningfully investigated.

43

44 The Transmission of Anglo-Norman

4.3

Data sources

Historical linguistics can be thought of as a kind of archaeology, in that the evidence for the analyses proposed comes essentially from material remains. With pottery and other artifacts, the features of how earlier generations lived can be decoded: similarly, texts surviving in the form of manuscripts can bear witness to the features of their language. Yet texts of earlier eras were never an immediate representation of speech, any more than ancient religious artifacts constitute a belief system: interpretation is always required in order to reconstruct the properties of the human cognitive capacity, and even more so when reconstructing its manifestations in societies without living representatives to be consulted. Deriving from artifacts properties of a cognitive system such as languages, or religious beliefs for that matter, is fraught with the additional problem that the human language faculty is only one of the factors at work in the creation of a text. Cultural practices, the conventions of a particular genre, possibly the physical limitations of the medium (as with epigraphy) will all be involved in determining the characteristics of any particular piece of writing that has survived. Interfaces with social organisation, technology etc. will determine much of the language used, specifically its lexis. However, all this does not mean that it is ­hopeless to try to identify the role played by the language faculty and its instantiation in the writers of the texts in question. Within the conventions of particular genres, the language phenotype evolves across time and as we saw in Chapter 2 shows important variation between L1 and L2 French. These dimensions of variation are orthogonal, involving morpho-syntactic systems that constitute autonomous linguistic properties.7 Given that a number of specific morpho-syntactic properties will be the target of enquiry in this book, an important issue is the appropriateness of the sources which will provide the information required. Certain text types tend to offer contexts for particular grammatical forms more than others.8 Rhymed verse, for example, is good for determining phonological properties, especially of vowels, but of questionable use as regards syntax. Choices therefore often have to be made so as to allow as far as possible for contexts to be provided for the traits under investigation. Published Anglo-Norman texts exist in reasonably plentiful quantities. They fall, however, into two broad periods with sharply differing characteristics (Pope 1934). The earlier period, in terms of presumed date of composition, extends from about 1120 to about 1250, and is characterised chiefly by verse texts, ranging from epic poetry to hagiography and romances. The second period, beginning in the 1250s, sees an increasing preponderance of non-fictional prose: treatises,



Chapter 4.  Rationale and design of the study

a­ dministrative documents, correspondence, chronicles and petitions. Later period texts are on the whole reliably dated, but this is rarely the case with earlier period works. Indeed there is a fair number of texts whose editors have not managed to establish even the century in which they were written.9 The broad divide between an earlier period of verse fiction, and a later period of prose non-fiction, will do at a first pass. It is not absolute, since there were verse works written post-1250 and a small number of roughly datable prose works written prior to that date. Still, on the whole, a study of securely dated pre-1250 Anglo-Norman prose sources would find relatively little to draw on. In general, the most rewarding genre is that of narrative prose, in which texts are reasonably plentiful from the later 13th century onwards (mainly chronicles, saints’ lives and the occasional prose romance). Petitions extend over roughly the same time-span, and involve an element of past tense narrative. Technical treatises in law and land management enjoyed a burst of activity in the late 13th century. On the whole, then, the approach to be taken is that particular syntactic traits are best analysed using texts known to favour their occurrence, while others will be analysed using all available prose texts, unless there is good reason not to. To investigate discourse syntax in dialogue, for example, as will be done in Chapter 10, the only suitable data source was the law Yearbooks, which contain long stretches of dialogue in Anglo-Norman (not yet Law French) from the last quarter of the 13th century throughout most of the 14th century. These reports of legal pleading in dialogue form offer opportunities for interactive language phenomena that may be absent in other genres. Linguistic investigation of historical texts is nowadays greatly aided for several languages by the existence of electronic corpora, which permit searches by lexical item or phrase, and sometimes by grammatical construction. In 2005, when the data collection for the present research was begun, no such electronic resource existed for Anglo-Norman. It was thus judged essential to narrow the field of data mining to the genre most likely to yield positive results. In 2007, however, the Anglo-Norman Hub website made available a large body of electronically searchable texts, which greatly facilitated matters, offering a very substantial amount of texts, both early and late. In this study, material from the A-N Hub textbase (Trotter 2007) will be relied on most of the time. Much of the material is non-literary in nature, and often possible to date and localise, whereas literary manuscripts typically passed through a number of hands and may have beeen partially rewritten before the version that happens to have come down to us. This makes them problematic for the purposes of diachronic study, where it is desirable to know where and when the text originated, and the date of composition of the text. A great deal of securely datable and localisable A-N source material is

45

46 The Transmission of Anglo-Norman

now electronically searchable through the Anglo-Norman Hub as well as PROME (Given-Wilson et al. 2005), and the Anglo-Norman Correspondence Corpus (Ingham 2008). With such data it is at least possible to be confident that data sources come from the period or periods that are being examined, and have not been rewritten subsequently. In the case of Anglo-Norman, another concern is to ensure that the texts genuinely represent insular French, and specifically, that they were written in England. In practice this is not a serious cause for concern, as the spelling forms used in insular French, in addition to certain grammatical features, make it clear that a given text was indeed Anglo-Norman, especially in the later period. As we go back in time to the 12th century, however, the surface differences between insular and continental written language are far less evident. However, the texts in the Anglo-Norman Hub text base, created by Trotter (2007) are undoubtedly representative of the insular variety. Since the A-N Hub textbase was not conceived as a diachronic corpus, ways had to be found of making the best possible use of it, by selecting from files which fitted the categories of analysis intended, and were most likely to provide adequate data in the domains being researched. Although it is common to find diachronic studies which are able to provide analyses that draw on a single large corpus, especially in the history of English, this was not possible in the present work. For one reason, corpus resources on Anglo-Norman have not yet attained the stage of development of their counterparts in the history of English. For another, the nature of the present volume made it necessary to encompass a number of different areas of linguistic competence, for which different kinds of corpus resources needed to be examined. It was at one point considered whether the research would gain from a direct comparison with equivalent continental Old French corpora. However, these were not available in the kind of quantities and with the kind of text types that would allow a direct comparison with the later Anglo-Norman prose texts, many of which are non-fiction. Historical French corpora available at the time when data-collection was being conducted (2008–9) still overwhelmingly consisted of literary texts. 4.4 Design of the syntax study corpus A syntax study corpus of prose texts was constituted using the A-N Hub textbase. It consists of the following 27 items, tabulated by genre, and abbreviated as in the A-N Hub textbase (see Table 1 which also states the word length, where this information is available to the author):



Chapter 4.  Rationale and design of the study

Table 1.  A-N Hub texts used as syntax study corpus Correspondence Reg Peckham 8681

Roy Lett 3413

Westm 16456

Chronicles

Ann Burt 5227

Anon Chr2

Cron Lond Reis Britt

Ann Lond

Gov’t statutes and treaties

Stats 1. 134380

Foedera 711218

Ordinances

Dom. Gip. 27015

Red Bk 19305

Lib. Cust. 24239

Oak Book Liber Al16079 bus 63685

Other documents

Private Indentures 36825

Percy Ch Ireland 13664 38265

Religious prose

St Sard 72542

Blk Bk 29221

Stapledon 2806

Rough 22104

YMB 42602

Sz Med part 1

Fiction

Fouke

Legal treatises

Britton 158298 (but only part 1 was used)

Other documentary

Man. Lang.

In terms of time periods, these sources are attributed to the 13th or 14th centuries as follows: Table 2.  Periods of origin of texts in syntax study corpus, A-N Hub textbase 1250–1299

1300–1349

1350–1399

Westm Reis Britt Roy Lett Ann Burt Reg Peckham Percy ch 1290s 39–144; Foedera 1,339–1,911; Britton pt 1

Cron Lond Stapeldon Oak Book Ann Lond Blk Bk Private Indentures Red Bk Dom Gipp. 1–22 Percy Charters 145–193 Fouke St Sard Lib Cust Foedera 1,911–3,196 Stats 1,136–309

Foedera 3,196–4,167 Statutes 1,310– Percy Ch: 199–end. Rough Man Lang Liber Albus Sz Med Anon Chr2 Ireland YMB

47

48 The Transmission of Anglo-Norman

For further bibliographical and philological details the A-N Hub may be consulted (www.anglo-norman.net/). 4.5

Additional data sources

For certain purposes, the syntax study corpus needed to be augmented by other A-N resources. In Chapter 6, on quantifiers, all texts in the A-N Hub textbase were deemed suitable for use, for reasons explained in the chapter. Chapter 8 takes in clausal syntax in the earlier C13, but only two A-N Hub prose texts represent this period (Reis Eng. and Mirour de Seinte Religion). They were therefore supplemented by prose works outside the A-N Hub, namely two ANTS editions, the Sermons on Joshua and the Livre des Régions. These sources were searched manually for target constructions. Chapter 5, on phonology, required an analysis of C12 poetic texts so that evidence from rhymes could be studied. In this chapter, use was made of various verse texts in the ANH textbase. Divided into a ‘12th century period’, and a ‘later 13th century’ period, they were as follows: ‘12th century’ GAIMAR, Saint Brendan, Saint Franceys, Comput, Salemon ‘Later 13th century’ Kalender, Secr1, Apocalypse, Hospitaller’s Rule, Lumière as Lais, Folie Tristan

These sources were augmented by three datable Anglo-Norman verse texts from the Campsey Collection of Saints’ Lives (Russell 2007): Saint Richard, La Vie Seinte Audree, and Sainte Catherine. The first two fall into the 12th century period, and Saint Richard into the later 13th. Two additional sources were used in Chapter 10. To obtain data for a relatively rare discourse marker which the A-N Hub did not instantiate in sufficient quantity, the electronic version of the Parliament Rolls of Mediaeval England (Given-Wilson et al. 2005) was used. This is a lexically searchable CD-ROM database consisting of petitions, many of which were written in Anglo-Norman between the late 13th and the early 15th centuries. In the same chapter, analysis was undertaken of a discourse marker whose use in dialogic contexts needed to be studied. It was searched in another CD-ROM database, the Anglo-Norman Yearbooks corpus (Larrivée & Ingham 2010). Altogether, then, the following six linguistic domains were searched, using the resources shown listed here:



1. 2. 3. 4. 5. 6.

Chapter 4.  Rationale and design of the study

Quantifiers: ANH textbase Gender: ANH textbase, prose Phonemic contrasts: ANH textbase, verse, and Campsey corpus texts Verb second and null subjects: ANH textbase, prose, early C13 ANTS texts Adjectives: ANH textbase, prose Discourse markers: ANH textbase, prose, PROME, and law Yearbooks

Following the approach set out in Section 4 above, comparisons needed to be made throughout with existing studies of continental Old French so that any divergences in A-N could be determined. This was not generally difficult, as studies of Old French in most of the areas to be investigated are reasonably adequate. Its phonology has been well-studied, and its gender norms are reliably recorded in dictionaries. Old French discourse markers and the syntax of quantifiers, verbs and adjectives have also been researched in some detail. This being so, fresh corpus analysis of continental Old French texts has generally not been pursued. In one case, adjective syntax, however, existing research has been supplemented by our own investigation of peripheral dialect syntax, for particular reasons to be explained in Chapter 9. 4.6 Overall approach Using a range of available sources, the present study has thus been able to obtain data bearing on the target variables relevant to the present investigation, both for A-N and for continental Old French. It is the first empirical study of insular French conducted on this scale, using these methods. Our approach to diachronic corpus research is worth stating at this point, since a great deal will hinge on the interpretation of what is found in the corpus resources to be used here. First, we assume that the object of linguistic study is not the collection of forms in the corpus, or their distribution, but the language systems constituting the competence of the speakers who produce those forms and give rise to their distribution. In this we follow the principles of modernday generative linguistics. A corpus is a collection of performance data, not the object of study itself. For historical stages of the language, a corpus provides us with necessary information on what the broad tendencies of the language were, from which the linguistic system, whose construction is the work of the linguist, has to be determined, and this is where the difficulties begin. Any collection of material in a historical corpus will be circumscribed by accidental limitations, and increasingly, as the time of origin of the corpus recedes further into the past,

49

50

The Transmission of Anglo-Norman

by the accidents of survival. A historical corpus can therefore not be treated as a systematic record of usage. In pre-modern societies only certain kinds of text, often of a more formal type, were deemed worthy of preservation. Very often, the more formal kinds of language do not record ongoing changes in the ordinary vernacular, with the result that a corpus can in a very real sense be anachronistic as well as incomplete. These are aspects of the well-known ‘bad data’ problem in historical linguistics mentioned by Labov (1994). Despite its acknowledged limitations, a historical corpus nevertheless remains indispensable as a source of quantifiable information on the broad outlines of the principal characteristics of language at particular periods.10 If a phenomenon in past states of the language is of importance, instances of it should be picked up by a properly constructed type of investigation, looking at contexts in which it is thought likely to arise. This is analogous to research procedures in psycholinguistic domains, where sample data are obtained in order to cast light on the object of study, such as the grammar of aphasia or of early grammar acquisition. As in any empirical research domain, corpora of observational data, be they historical or otherwise, provide a means of pursuing specific questions. They need to be interrogated in ways that treat them as a potential source of answers to hypotheses on the object of study, the linguistic competence of those that produced the data. This is the approach to be followed in the present work, where the aim is not to survey corpora of Anglo-Norman and report the findings pre-theoretically, but to use them to test a set of theoretically derived hypotheses, arising from issues in the domains of acquisition research reviewed above. The alternative research hypotheses are stated, for each linguistic domain of enquiry analysed, as: (i) that the distribution of corpus data will reflect the influence of L1 English on A-N and (ii) that it will not. Linguistic data have been chosen that in each case patterned differently in Middle English and in continental French, thus making it possible to discriminate between these alternatives. Further domains of analysis may be identifiable which would modify the conclusions that will be drawn in the present work. However, we believe that the criteria adopted here of clearly demarcated system contrasts (phonological or syntactic) make the six domains of enquiry set out in the preceding section a wellmotivated set of linguistic topics to investigate, in order to assess the psycholinguistic status of later Anglo-Norman by comparison with continental French.



Chapter 4.  Rationale and design of the study

Notes 1. In fact, Old French often resembled Middle English, so positive transfer would have enabled an English adult learner of French to find the right constructions. For example, the basic main clause word order was SVO. If the object was placed first a nominal subject had to follow the verb. Auxiliary verbs ordinarily preceded lexical verbs, articles and other determiner words preceded head nouns, and spatial particles followed the verb as they did in Middle English. 2. Lass (2006: 71) conveniently summarises dialect variation in morphology by stating that ‘everything new begins in the north and percolates down though the East Midlands’. Taylor & Kroch (1997) showed that Northern versus Southern varieties differed syntactically as regards the applicability of Verb Second (Chapter 8). 3. In terms of case morphology, this was not so, however: Northern texts are known to have conserved Subject case marking longer than central varieties. 4. However many of these words, of Latinate origin, may have entered English by French and reflected older meanings of the French words (e.g. Montaigne’s librairie). 5. Another possibility arises with words or Germanic origin which were borrowed into French, where the old English cognate possessed a different sense. If the French word was used in England in a sense reflecting the old English cognate, we would again have negative transfer. As stated in the text, however, such lines of research go beyond the limits of what can be undertaken in this volume. 6. Such as preverbal placement of object pronouns. 7. They are thus unlike lexical innovations denoting e.g. new types of military equipment or social organisation. If the latter vary across time and place or between L1 and L2 varieties, we will be informed about how awareness of new technology was spreading. But this could happen as individual lexical borrowings from a foreign language, and indeed did, in 16th and 17th century English, when numerous items of French military terminology were borrowed into English and other languages. Lexical borrowing does not engage the human language acquisition capacity as such. 8. Wills, for example, disfavour past tense narratives while historical chronicles do not favour future tenses and modals. 9. The problem of dating mediaeval fictional or devotional works which lack internal clues as to their time of origin may be insuperable. 10. Individual instances, though valued in lexicography, will be of problematic status unless supported by a substantial body of similar instances elsewhere in the corpus.

51

chapter 5

Anglo-Norman phonology

5.1

Introduction

Researching the phonology of a dead language is a hazardous enterprise: whereas the combinatorial possibilities of its morphology and syntax can to a great extent be gauged from what is visible in the textual record, its sounds can no longer be recovered. Erratic spelling practices in pre-modern texts often mean that reconstruction of sounds from graphemic evidence is difficult.1 Nevertheless, the phonology of A-N is of central importance to the present enquiry, so consideration must be given to identifying its sound values as far as they can be ascertained. Virtually the only tools at one’s disposal are spelling forms – despite the reservation just stated – and rhyming evidence. Words that rhyme tell us when sounds were treated as ‘the same’ by versifiers, from the point of view of their vowels and consonants supporting the rhyme. When sounds do not rhyme, the inference to be made is that a phonological distinction was involved. Spelling forms can be useful as evidence that a former sound contrast has been neutralised: when the same spelling form is used both for word A and for word B, phonological equivalence in this respect is a reasonable assumption, even if not a necessary conclusion. In this chapter we analyse aspects of the phonology of Anglo-Norman using these two kinds of information, though with appropriate caution. In terms of textual resources with which to carry out this work, the position is better than at any time in the past to conduct research into the subject. The very substantial material in the A-N Hub textbase includes numerous electronically searchable verse texts, and the Anglo-Norman Text Society has published a large and well edited body of verse texts which can throw light on phonemic similarities and differences, on the basis of rhymes. These have been put to good use in Short’s (2007) Manual of Anglo-Norman, which provides a well-documented survey of the main segmental phonemic contrasts. No assurance can ever be given, of course, that the sounds of a dead language have been accurately recaptured, and doubt has been cast on whether rhyming patterns, Short’s principal analytic resource, provide evidence for them. Nevertheless, as we shall later on in this chapter, such challenges can be dealt with, and the analysis offered by Short seems to us to hold up as a reasonably well-established account of how the systemic properties of the A-N sound system operated.

54

The Transmission of Anglo-Norman

Complete coverage of the phonology of Anglo-Norman is not our objective here: the approach is contrastive, addressing sound changes that took place in continental Old French, and those that took place in Anglo-Norman, apparently under the influence of English, which are not paralleled by continental Old French. The chapter goes about these aims as follows: first, Early Middle English and Old French phonology are compared. We then ask how far Anglo-Norman respected the phoneme contrasts of Old French not present in Middle English, using evidence from rhyming verse, and then where appropriate from spelling forms. This allows an overall conclusion to be reached on whether Anglo-Norman can be considered independent of mediaeval English phonology, or whether it showed the influence of English. It will be shown that the answer partly depends on the time period, whether before or after 1200, and that convergence on English phonology was well-established by the 13th century, long before A-N syntax showed any tendency to converge on English. French speakers who participated in the 1066 conquest of England did not all speak the same variety, and certainly did not speak ‘Parisian’ or Central French to any great extent. The great majority of names identifiable in the Domesday Book have links with Normandy, although some are associated with Northern France and Brittany (Keats-Rohan 1995).2 It was thus the properties of Old French as spoken in Normandy that would determine the development of Old French in England. The Norman dialect belonged to the Western dialects of Old French (Pope 1934) and had certain particular features which will be mentioned where relevant in what follows. 5.2

Acquisition target properties

The properties of A-N phonology to be addressed in this chapter are essentially those where Early Middle English and Old French, especially its Norman variety, contrasted, since it is here that contact influence is most obviously detected. In particular, items in the sound inventory of Old French not present in EME are of interest, as they would normally have represented the main challenge from the point of view of acquiring a second language. Middle English phonology has been relatively well-studied. There is a fair number of 13th century Middle English texts, especially in verse in the later part of the century, which has allowed historians of the language to gain a reasonably well-informed picture of its segmental properties. According to Lass (2006), Early Middle English had the vowel phoneme inventory shown on the left in Table 1 below, which we compare with central Old French (Pope 1934) on the right:



Chapter 5.  Anglo-Norman phonology

Table 1.  Early Middle English and Old French simple vowel systems compared3 Early Middle English

Old French (Central variety)

High Mid

Low

Front Back Front Back Central Front Back

/i/ /u/ /e/ /o/ – /a/ /a:/

/i:/ /u:/ /e: / /ε/ /o:/

/i/ /u/

/y/ /e/ /o/ /ò/ /schwa/ /a/

/ε/

NB: The symbol /ò/ represents approximately the Cardinal 6 vowel.

Leaving aside the fact that vowel length was phonemic in Middle English but not in Old French, the main differences between these two were the existence of a high front rounded vowel /y/ in Old French but not in Middle English, and the status of a weak ‘e’ (schwa) vowel. The latter enjoyed phonemic status in Old French, serving for example to contrast cop ‘a blow’ and cope ‘cuts’, or ‘gold’ and ore ‘now’, oie ‘(sense of) hearing’, oi ‘heard’.4 The loss of final weak ‘e’ in French appears to have been a post-mediaeval development. In Middle English, short vowels in weak syllables merged to /e/ or schwa (Lass 2006), and by the 14th century this final weak vowel was becoming unstable (Minkova 1991). Its vulnerability to omission earlier in the history of English than in the history of French will be important in this chapter and in later ones. As regards diphthongs, the situation was quite unlike the present-day states of the languages. Whereas Modern French lacks diphthongs, Old French actually had more than Middle English, as can be seen in Table 2: Table 2.  Early Middle English and Old French diphthongs compared Early Middle English

Old French (Central variety)

Level

/iu/

/iu/

Rising

/ai/ /au/ /eu/ /òu/

/ai/ /ei/ /òi/ /oi/ /au/ /eu/

Falling

–

/ue/

Old French had certain palatalised semivowels, not shown above, which English lacked, specifically /je/, as in pied, and the semivowel in nuit, which we shall notate for typographical convenience as /yi/. Again, these sounds will provide an important point of comparison for the present study.

55

56

The Transmission of Anglo-Norman

The respective consonant inventories showed few major mismatches in contrasting sounds. Stops, nasals, and liquids showed essentially the same oppositions. Both had the semi vowel /w/, though only English had /j/. Among the fricatives, both languages had /f, s, š / but the absence of voiced counterparts in OE seems to have been maintained into early ME, whereas in OFr /v, z/ were present. Both languages had the affricates /tš/ and /dž/. However, the interdental /θ/ and /ð/, present in early Old French, soon dropped out. The cases where OFr possessed consonant sounds absent from the medieval English phonemic inventory, and which will be of importance in this study, were the palatalised nasal and lateral consonants, notated here as /nj/ and lj/. 5.3

Variation and change in OFr phonology

In this section four sound changes taking place in Old French after 1100, according to the textual record, are presented. It will be seen that only the earliest of these found its way into A-N. Particular attention is paid in this section to cases where western Old French, including the Norman dialect, varied from central Old French. In the discussion in this section the key differences will to some extent be visible to a non-specialist reader, if rather haphazardly, from differing spelling forms, especially those for vowel sounds. The grapheme in words such as françois and loi became normal in central OFr texts by the early C13, while western sources retained here. The graphemes and initially involved different vowels, and continued to do so in western Old French, but they merged in central varieties. The grapheme frequently occurs in central Old French, e.g. fleur (‘flower’) to indicate a central vowel distinct from /u/, the latter being written , e.g. tour (‘tower’). In western texts, however, the grapheme is quite often used for /u/. However, because of the vagaries of spelling practices both in England and on the continent, these tendencies were very far from consistent and no reliance can be placed on spellings by themselves. For example, the grapheme often appears in flor (‘flower’) and tor (‘tower’) but its phonological value for the originator of the text is uncertain.5 The diphthongs /ai/ and /ei/ contrasted in early Old French,6 but in Western dialects the two diphthongs had also merged by the time of the earliest continental rhyming verse evidence, e.g.: Franceis:pais (= Lat. pax) (CDN 10649), estoit : lait (St Eustache 918).7 Rhymes between them are found in 12th century A-N verse from before the middle of the century onwards, e.g.: retrait : dreit (Gaimar 139). However, they are absent from the earliest verse, such as Brendan and Comput, written around 1120–1130. This seems to show that A-N incorporated the merger only after the initial implantation of French in England, but while it still remained



Chapter 5.  Anglo-Norman phonology

a branch of the Western French dialect area. Numerous early A-N spelling forms such as neistre, funteine, and paleis (Short 2007: 76) also indicate the merger. The Old French vowels represented by and spellings coalesced, according to Sampson (1999), in the later C12. This is no doubt true of more Eastern varieties: Chrétien’s champenois variety rhymes vivant with les quatre vant (Cheval. Charr. 1953), as well as vilmant with devant (ibid. 2735). But Western OFr took longer to run the two together. Langlois (1914) showed that central varieties had done so by the first half of the C13, on the basis of rhymes such as dormant : durement etc. But evidence for it in Western rhyming verse texts up to the early C13 is absent. The merger did not spread to the French of England, a point first established by Meyer (1886). Although at that time only a limited range of material was available to him, the evidence of verse composed in the later C13, and now published by the ANTS, has amply borne out his observation. Thus for example, La Vie de Seint Richard (1276) rhymes tant, avant, chant etc. with present participles, e.g. demurant, but never with the frequent adverbs or nouns ending in -ment, and the same is true of the Lumière as Lais attributed to the same author.8 Spellings of -ment adverb endings in the A-N Hub textbase show virtually no confusion between -ment and -mant, attesting to the maintained separation of the two vowel sounds. The third sound change to have occurred in Old French was a vowel split, mentioned above, distinguishing the vowels /u/ and /œ/. These vowel sounds were not phonologically distinct in Western texts from the later C12, where we find rhymes such as seignor : jor (énéas 5999–6000) and demore : secore (énéas 8107–8). Saint Eustache (Normandy, early C13) still shows no split, rhyming e.g. veneor: jor (253), jor : labor (811), and nos : besoignos (1335). Bourciez (1958) considered that the split originated in North-Eastern France and spread to more central areas during the C12. It never reached England (Short 2007). The early Old French /u/ words which on the continent shifted to the /œ/ vowel, such as pour (‘fear’), continued in A-N to rhyme with those that did not undergo the shift, such as ModFr tour /tur/.9 Where the vowel split occurred, continental Old French spelling forms usually showed a distinction, but in A-N the persistence of /u/ in words such as seulement and plusieurs is reflected in the continuation of spellings in later A-N for words that in continental OFr would have /œ/, e.g.: soulment, plusours.10 The final vowel change of significance is the merger between the diphthongs /òi/11 and /ei/. In continental verse until the later C12, words such as oi (‘heard’), soi (‘knew’), esjoi (‘delight’), poi (‘few’), bois (‘wood’), chois (‘choice’) and vois (‘go’) rhymed only with each other, and never with words such as moi, cortois and vois (‘see’) etc. which had the /ei/ diphthong. This changed in the last few decades of the C12; rhymes from later C12 verse from the North & North-East

57

58

The Transmission of Anglo-Norman

show that the diphthong in former /ei/ words, such as moi, voi and doi, now rhymed with the old dipthong /òi/, as in oi (‘heard’) and soi (‘knew’), e.g.: oi : moi (Eracle p. 34) oi : voi (Charette 6551), cortois : bois (Jean Renart, Roman de la Rose p. 3). The initial vowel of the merged diphthong is likely to have been central (Fouché 1956). In A-N, however, such rhymes continued to be next to non-existent, except in a few works influenced by continental practice (Short 2007: 77). A search of seven original /òi/ words, viz. oi, soi (‘knew’), esjoi, poi, bois (‘wood’), chois and vois (‘go’) in the ANH verse sources (33 texts) was carried out for the purposes of the present research to test this point. It found only a single example of such a rhyme, bois: demanois, in a poem by the no doubt significantly named ‘Matthew of Paris’. Otherwise, A-N poets continued to treat the /òi/ and /ei/ diphthongs as distinct, and spelling forms continued to show in words like mei, curteis, and veis long after continental scribes had gone over to the spelling. Even in the last period from which there survives a substantial amount of French verse written in England, the third quarter of the C 13, there are virtually no signs that the merger was taking hold.12 Words in /ei/ rhymed with original /ai/, e.g.: cunterai : moy (Manuel des Péchés 5869), but not with original /òi/. The first part of the /ei/ diphthong in e.g. mei thus retained its front quality in A-N, rather than acquiring a more central realisation as in continental Old French. Although in later A-N spellings for /ei/ become common, this seems to have been a graphemic matter encouraged by continental practice as recommended in spelling treatises such as Orthographia Gallica (c. 1300). In sum, of these four Old French sound changes, A-N seems to have incorporated only the /ai/–/ei/ merger, the one which was clearly under way at an early stage, before 1160 on the basis of continental verse texts. The changes that spread across the Langue d’Oïl rather later never took hold in insular French. From the later C12 onwards, A-N users were thus not part of the wider francophone speech community as regards processes of phonological change. 5.4

Influence of English: vowels

Major alterations to vowels took place both on the continent and in AngloNorman, but the pathways of change were very different. The OFr high front rounded vowel /y/ did not fare well in A-N. Short (2007: 60) calls the rhyming of /u/ and /y/ ‘by far the most easily recognisable dialectal characteristic of insular French’. By the later C13 /u/–/y/ admixtures ‘had become the rule’, to quote Short, though they are found sporadically in the preceding century too, where they first become common in pre-consonantal contexts, especially before /r/ and /m/ as in Criat[u]r : dur (Adam 231), and sumes : fumes (Brendan 520), but then at the



Chapter 5.  Anglo-Norman phonology

start of the C13 spread to other contexts such as rendu : nevu (Pope 1903: 19–20). It would thus appear that from a fairly early stage continental /y/ was gradually being replaced by a /u/ vowel, as attested by later A-N spellings with , e.g.: (1) purvou (pourvu) Gt Yarmouth 1272 droitoures (droitures)  Fountains charter 1292 louter (lutter) Earliest Prose Brut 168 mour (mur) Reis Britt 306

Another realisation turned it into a glide, apparently from a high front to a high back vowel, e.g.: (2) Et moi du primer qe jeo unqes fiu de age qe jeo poai peccher…  Sz Med 178 ‘And I, from the first when I was ever of age to be able to sin…’

A third type of spelling features or again suggesting a glide realisation, e.g.: (3) vewe Gt Yarmouth 1272; pleus tard Crisis 1297, 102

The Middle English phoneme inventory did not contain a high front rounded vowel, so the pronunciations attested by the spellings in these examples show a tendency to replace the continental French vowel sound by the nearest English ones, either a high back rounded vowel, or a glide from a central or high front position to high back. The OFr semivowel which we have notated /yi/, as in puis, suis, nuit, contained a high front rounded element, and likewise fell victim to alteration in insular speakers’ pronunication, indicated by spellings and sometimes : (4) nut Turpin 40; pus Ann Burt 476; angulles Exeter Custumal; ennue (ennuyé) Earliest ProseBrut 341; lusoit (luisait) Earliest ProseBrut 2617; siut (suit) Reis Engl 2, 17

The sounds /y/ and /yi/ were cases where no known changes took place to vowel quality in Old French. There were also cases of changes that took place in Old French, but where different ones occurred in A-N. On the continent, the OFr diphthong /oe/, as in poet, noef, quoer shifted in the direction of the rounded central vowel quality of modern French /œ/ (Fouché 1956).13 This does not seem to have happened in A-N. Two realisations of /oe/ are found in later Anglo-Norman, one as /u/, attested by the rhyme esprueve : truve (St Edward 141), and by spellings such as buf (Brendan) and espruf (St Edward), the other as /ε/ as attested by rhymes such as cuer (‘heart’) : ver (‘see’) Lumlais 2, 7; cuer : fer Resurr 119; and suef : chef (Apocalypse 2679). Spelling forms such as pet for poet (Becket 36 and Romphil 23), treve for trueve (Fabliaux 34), also suggest that reduction to a monophthong /ε/ occurred.

59

60 The Transmission of Anglo-Norman

In Continental OFr the dipthong /eu/, deriving from Latin /oCu/ e.g. locus → lieu, nodus → noeud, developed into /ö/ in open syllables. In A-N, however, it became /u/: lieu, feu and geu, and as such can be found rhyming with /y/, pronounced /u/ in A-N as we saw above. Words with /eu/ in Old French are generally found in A-N in spellings such as merveillous (Bibbesworth 916, rhyming with vous) and ambedouz (Resurrec. 244, rhyming with clous).14 In continental Old French, as we saw above, /ei/ underwent some kind of centring process so that it no longer contrasted with /oi/. In Anglo-Norman, however, from the end of the 12th century onwards, the diphthong /ei/ underwent a monothongisation process so that it no longer contrasted with /ε/ (Merrilees 1982), as shown by rhymes such as creire : terre, and crestre : estre (Short 2007: 77). The spellings crere for creire and here for heire, which reflect the monophthongisation of /ei/, are very common in the ANH, likewise the reduction of the -eir infinitive ending to -er, e.g. aver (aveir), poer (poeir) remaner (remaneir). Finally, Old French still generally maintained an opposition between the palatalised diphthong /je/ and the vowel /e/.15 In continental verse, items ending in /je/ could normally only rhyme with each other, and not with items ending only in /e/. This constraint was most noticeable with verb forms such as the 1st conjugation infinitive, where verbs such as e.g. aidier, laissier, treitier, etc. could not rhyme with regarder, penser, rester etc. In A-N, this contrast began to disappear at a fairly early stage, and rhymes between the two classes of verbs became quite common in the late C12 (Short 2007: 65). Later A-N hypercorrect spellings such as arrestier, occupier and gardier instead of the normal -er endings testify to the breakdown of a phonological contrast here in later A-N. 5.5

Influence of English: consonants

Old French consonant sounds identified above as absent from Middle English were principally the palatal consonants /lj/ and /nj/. Old French words ending in -gne such as baigne, baraigne, overaigne and words ending in -ne such as domaine, semaine humaine did not normally rhyme. Nor did /lj/ words such as fille, famille, faucille rhyme with /l/ words such as ville, évangile, mille. In AngloNorman, however, it has generally been accepted at least since Pope (1934: 450) that the palatal consonants /lj/ and /nj/ tended to be replaced by the equivalent sounds /l/ and /n/, shared with the English phoneme inventory. This was an early development: already in the mid-C12 we find fedeil : conseil (Sanson 2837) and a little later merveilles : esteilles (Ipomédon 5575). For Short (2007: 102), there is good evidence for ‘general non-palatalisation of /l/’ in A-N. In the early C12 there are already signs that a similar loss of palatal versus non-palatal opposition was



Chapter 5.  Anglo-Norman phonology

taking place in /nj/ vs. /n/, with rhymes such as vaine : bargaigne and semaine : cumpaigne (Short 2007: 100). The traditional view, espoused by Pope (1934) and Short (2007: 100), is that both the palatal consonants /lj/ and /nj/ lost their distinctiveness from /n/ and /l/ under the influence of English. 5.6

Loss of sound contrasts in A-N or underspecification in rhymes?

In the foregoing sections, very substantial evidence has been assembled, taken from earlier studies and supported in some places by further data of our own, that A-N pronunciation lost a considerable number of phonemic oppositions present in Old French. This conclusion has generally been taken as unassailable, certainly since Pope (1934) and indeed earlier. However, another interpretation of the facts has recently been proposed. In Floquet (2010), the value of rhymes as an indication of the phonological system of past states of language has been called into question. He argued that in A-N /nj/ and /n/ probably remained distinct in ordinary pronunciation, but that the phonological distinctive features that differentiated the two sounds – [+ palatal] with /nj/, [+ dental] with /n/ – could be neutralised at the rhyme. The shared features allowing a rhyme between e.g. semaine and compaigne were [+ nasal, + coronal], where coronal was underspecified for dental or palatal. Such underspecification, he claimed, was not normally permitted in Continental OFr verse. Floquet’s argument rests on the observation that it is not unusual for phonemic oppositions normally respected in ordinary speech to be neutralised in some verse forms in more recent times, for example in popular songs or nursery rhymes. Consider the apparent rhymes in: (5) a. Je ne suis pas si vilaine … Puisque le fils du roi m’aime. b. L’alouette est sur la branche… Faites-nous trois pas de danse.16

One hopes that, say, a 29th century diachronic phonologist will not infer that /m/ and /n/, and the palatal and post-alveolar fricatives /š/ and /s/, failed to contrast in ordinary Modern French on the basis of distinctive features neutralised in these rhymes. Floquet’s work makes an important methodological point which has probably been unjustly neglected in the study of past states of language. The assumption, tacit or otherwise, that only when sounds failed to rhyme can we say they contrasted in the language of the period underpins much theorising on historical phonology. We must, therefore, take seriously the possibility that some freedom with rhyme operated in medieval versifying practices too, and not assume that because rhymes can be found involving the sounds X and Y, X and Y were not phonemically contrasted in the language of that period.

61

62

The Transmission of Anglo-Norman

It is a logically compelling argument, and as such would appear to pose a threat to the methodology of Short (2007) and his predecessors, whose analyses of sound changes, especially mergers, were based on the observations of cases where sounds no longer contrast at the rhyme. Given Floquet’s critique, the Old French sound contrasts described above, assumed by previous researchers to have been lost in A-N, could have continued in use. We shall refer to his approach as a ‘poetic licence’ account of the rhyming phenomena, allowing rhyming on the basis of underspecified vowel values. Where earlier researchers have reported a growing tendency for A-N not to respect sound contrasts in OFr, in terms of Floquet’s approach this would merely show that A-N versifiers became increasingly given to poetic licence, producing the impression of increasing divergence from expected Old French phonological norms. In other words, the ambient language did not necessarily change under English influence. The matter is not considered diachronically, however, by Floquet (2010), which is a relatively short research article. It therefore seemed advisable to undertake a diachronic investigation of the depalatalisation phenomenon in A-N, and of other cases of loss/neutralisation of phonemic contrasts. This was carried out for the present research by comparing sound contrasts where English shared the contrast in question, and those where it did not. If an increasing tendency towards loss/neutralisation is found across both categories, the poetic licence interpretation will be upheld. If the two categories of sounds behave differently, however, it will seem it far less cogent, especially if the level of loss/ neutralisation of sound contrasts shared by English and French remains constant over time, but the observation of sound contrasts not existing in English shows a marked decline. In that eventuality we would want to conclude that an interpretation of the divergence between insular and continental rhymes in terms of poetic licence is not satisfactory, but rather, that English phonological influence was at work, as hypothesised in the present research and assumed in previous studies. Such an undertaking is not straightforward, first because Floquet’s source, the ANH, does not organise texts by period and also because the dating of A-N literature is not always possible. Nevertheless, the difficulty is not insurmountable. A substantial number of works were undoubtedly composed in the C12, while for others, dates of composition can be securely placed in the early, mid- or later C13. Even though individual authors’ styles certainly varied, making use of the rhymes used in such texts allows the researcher to get an overall picture of how the phonological properties of A-N evolved.



5.7

Chapter 5.  Anglo-Norman phonology

Investigating neutralisation

We thus sought to establish whether the observance of phonemic contrasts in A-N poetry in the 2nd half of the C13 differed from those of C 12 verse. The phonemic contrasts involving palatal versus non-palatal sounds affected primarily two consonant pairs /nj/–/n/, /lj/–/l/, and the vowel contrast /yi/–/u/. If neutralisation occurred, sets of words having the first sound of each pair should be seen entering freely into rhymes with sets of words having the second sound of each pair. For example, /nj/ words such as montaigne, gaaigne, remaigne would be found rhyming with /n/ words such as vaine, humaine, laine. On the other hand, if the sounds in each pair normally remained distinctive, and underspecification allowed poetic licence to be sometimes taken by the writer, the two series should not have commonly co-occurred in rhyme. First we investigate the status of the palatal nasal /nj/. Occurrences at the rhyme of a set of /nj/ words17 were tracked, and the quality of the nasal consonant noted in the corresponding rhyme word. The target /nj/ words were spelling variants of items conventionally ending in -aigne or -eigne in continental OFr, where, as in A-N, /ai/ and /ei / had merged, as we saw above. The third vowel opposition concerned the high front rounded semivowel /yi/. The distinction between the high front rounded vowel /y/ and the high back rounded vowel /u/ collapsed in A-N (Short 2007: 61), so the semivowel /yi/ was vulnerable as it involved the same place of articulation as /y/. It will next be considered, therefore, whether /yi/ remained distinct at the rhyme from /u/. For this purpose, three electronic corpora were drawn on: the ANH textbase, the Campsey Corpus (Russell 2007) and the University of Ottawa’s Laboratoire de Français Ancien. Because of the well-known problem in dating many AngloNorman texts, only a minority of the sources complied with our datability criterion, and only one text from the Laboratoire de Français Ancien did so. For the later 13th century, the electronic corpus resources proved insufficient and two further texts were consulted, as shown below. To aid comparability, only poems written in rhyming couplets were considered. Table 3.  Sources for diachronic study of A-N rhymes C12 Campsey: Vie Sainte Catherine ANH: Gaimar, Salomon, Hospitaler’s riwle, Comput LFA: Ipomédon C13 Campsey: Vie S. Richard ANH: Kalender, Secreta secretorum, Apocalypse, LumLais, Sermon on shrift, Vie S. François Other: Grossteste’s Chastel, Waddington’s Manuel des péchés

63

64 The Transmission of Anglo-Norman

Texts whose date of composition can be placed with reasonable certainty in the 12th century, and another set attributable to the mid-to late C13, were compared (Table 3). A wide gap was intentionally left between the earlier and later periods so as to identify clearly any diachronic trends that might be visible between them. Pope (1934) considered that the period around 1230 marked a division between an earlier stage of rough approximation to continental Old French norms and a later one of clear departures from them. The two date ranges adopted here were well before and somewhat after that date, respectively. We looked for evidence of pure or mixed rhymes, found in rhyming couplets, which either upheld or did not uphold a phonemic contrast in continental Old French. For purposes of practical convenience, the texts in the Anglo-Norman Hub textbase were taken as a single source. That is, if two words were found to rhyme in any one of the texts in this selection, the same rhyming pair was not counted from other texts in the A-N Hub. Other sources were counted separately, however. 5.8

Results

Table 4 shows frequencies of occurrences and percentages of the three target sounds in pure and mixed rhymes in the two periods investigated. For the mixed rhymes, the OFr target sound and its usual A-N substitute are shown in brackets in the left-hand column. Table 4.  Evolution of rhymes involving palatal sounds not existing in EME C12

later C13

Pure rhymes

mixed rhymes

Pure rhymes

mixed rhymes

/nj/ (/n/)

N %

  27   68

13 32

 5 16

  26   84

/ui/ (/u/)

N %

  47   77

14 23

14 24

  45   76

/je/ (/e/)

N %

207   95.4

10   4.6

74 38.1

120   61.9

In the first half of the C12, there was a minority incidence (between 5–32%) of mixed rhymes. In the later C13 the situation was very different. Now, mixed rhymes predominated, accounting for 62–84% of cases.



Chapter 5.  Anglo-Norman phonology

Mixed rhymes in the C12 showed different levels of use in the three variables, with /nj/ : /n/ as the most common and /je/ : /e/ as the least common. In particular, /nj/ : /n/ rhymes accounted for a third of the earlier C12 data, a fact which is consistent with Floquet’s interpretation of neutralisation. Nevertheless, all three variables showed a very sizable increase in frequency in the later C13. To maintain a poetic licence interpretation, one would have to suppose that poets became far more prone to neutralising phonemic distinctions at the rhyme than before. To test this prediction, we consider three cases where segments remained phonemically distinct in both languages, but could nevertheless rhyme by poetic licence in Old French. These were rhymes involving final /m/, and the consonant clusters /rs/ and /str/. It has been observed that in Old French final /m/ sometimes rhymed with final /n/ (prodom : meson, 2ème Continuation Perceval, p. 108), -estre sometimes rhymed with -este (teste : fenestre, Chev. Charr., 1127) and -rs with -s (fors : dos, Béroul, p. 63). This also concerned items ending in -ums/oms and their rhymes, since spellings with or without final , especially in the 1st plural verb ending, did not seem criterial for the rhyme. Old French was tending to lose consonant clusters in word-final position,18 especially those involving /s/. However, in Middle English consonant cluster reduction was not in evidence. The same texts were used as before to carry out this analysis. The results showed no temporal evolution, as can be seen from Table 5. Table 5.  Evolution of rhymes, sounds existing in EME C12

C13

Pure rhymes

Mixed rhymes

Pure rhymes

Mixed rhymes

/um/ (/un/) N %

85 55.6

68 55.4

65 56.0

51 44.0

/str/ (/tr/)

N %

47 84

 9 16

34 97

 1  3

/rs/ (/s/)

N %

65 89

 8 11

61 87

 9 13

The results in Tables 4 and 5 are displayed in graph form in Figures 1 and 2 respectively: The divergence between the two categories of sound contrasts could hardly have been clearer. Pure rhymes involving sounds not present in English phonology showed a sharp fall from the 12th to the later 13th century, in favour of mixed rhymes which used the corresponding English sounds. On the other hand,

65

66 The Transmission of Anglo-Norman

100

80

60 /nj/ /yi/ /jé/ 40

20

0 C12

Later C13

Figure 1.  Percent pure rhymes, sounds not existing in EME* *Alternatively, rhyme in /nj/ replaced by /n/; rhyme in /yi/ replaced by /u/; rhyme in /jé/ replaced by /é/. 100

80

60 /rs/ /str/ /um/ 40

20

0 C12

Later C13

Figure 2.  Percent pure rhymes, sounds existing in EME* *Alternatively, rhyme in /rs/ replaced by /s/; rhyme in /str/ replaced by /st/; rhyme in /uns/ replaced by /un/.



Chapter 5.  Anglo-Norman phonology

rhymes involving sounds shared with Middle English showed no diachronic tendency: two of them remained more or less flat while the third actually showed a trend against mixed rhymes, i.e. a trend away from poetic licence. The observation by earlier scholars that Anglo-Norman rhyming tendencies were ­different from continental usage thus cannot be explained by attributing an increase in poetic licence to A-N authors. They showed no tendency towards employing greater freedom with the three sounds shared by Middle English, but did so only with the three sounds that did not exist in Middle English. While some rather limited effects of poetic licence can be seen in these three areas in earlier C12 Anglo-Norman verse, by the last period a dramatic change had taken place making mixed rhymes very common. Yet no such shift took place in the set featuring sounds shared with Middle English. The only explanation of the very different results found in these two sets of rhymes that seems satisfactory is that sound changes were taking place in the variables belonging to the first set, but not in the second. The results are so starkly defined that statistical analysis is unnecessary in order to distinguish these sets. To maintain, with Floquet (2010), that non-observance of strict rhyming in Anglo-Norman was not a matter of English influence, but was created by poetic licence on the part of the versifier, who exploited underspecification, offers no account of the sharp differences observed between matching and non-matching sounds. Later poets showed no tendency towards greater neutralisation of sound contrasts when the sounds in question existed in English phonology. But they did show neutralisation in those areas where the sounds did not exist in English phonology, so to explain the latter in terms of neutralisation looks to be ad hoc. The breakdown of OFr sound contrasts accords, in fact, with earlier caricatures of Anglo-Norman pronunciation in continental texts. Converging evidence, then, from caricature by contemporaries (questionable in itself because of the risk of exaggeration) and from rhyming verse leaves little doubt that phonemic contrasts involving a palatal feature had been lost from Anglo-Norman by some time in the 13th century. It is worth returning at this point to the question of how far spelling forms are informative as to historical phonology. If the sound contrasts absent from English had been neutralised merely as a matter of rhyming practice, and not as a reflection of ordinary A-N pronunciation, there would be no reason to expect orthography to treat the sounds as equivalent in other contexts. Yet in fact A-N spelling forms other than at rhymes also seem to show plenty of confusion between the /n/ and /nj/ phonemes of OFr, for example, as illustrated by the following forms in verse texts in the A-N Hub texbase:

67

68 The Transmission of Anglo-Norman

muntaine (St Modw) cf. OFr montagne ‘mountain’ cumpaine (Lumlais) cf. OFr compaigne ‘company’ chastaine (Sanson) cf. OFr chastaigne ‘chestnut’ uvraine (Becket) cf. OFr ovraigne ‘task’ champaine (A-N Alexander) cf. OFr campagne ‘fields’ plainent (Prothes., St Modw) cf. OFr plaignent ‘complain’ gainent (Secré des secrez) cf. OFr gaaignent ‘win’ ateine (A-N Alexander) cf. OFr atteigne ‘reach’ enseine (Prothes., Lumlais) cf. OFr enseigne ‘teach’ busoine (St Edward, Prothes.) cf. OFr besoigne ‘need’

Therefore, the indications that A-N phonology differed from continental French in respect of these sounds are not limited to neutralised rhymes. 5.9

Summary of findings and implications

In this chapter, we have first presented well-established findings on the phonology of Anglo-Norman as compared with continental French, and then our own investigation of whether the collapse of continental sound contrasts, as attested by rhymes in A-N verse, can be handled as a form of poetic licence. It was first shown that only the earliest of four continental OFr vowel changes impinged on A-N. Phonological developments in OFr from the later C12 onwards did not, it seems, make their way into insular phonology. Next, the abundant evidence collected by Short (2007) and his predecessors for insular developments in phonology were reviewed, showing that rhyming and spelling data attest to confusion of phonemic oppositions, specifically on points where English lacked the oppositions in question. The argument that later A-N rhymes showed increasing toleration for poetic purposes of underspecified sound contrasts, rather than reflecting genuine sound changes, was shown to be inadequate. It can thus be concluded that AngloNorman users pronounced Old French in a way that markedly distinguished it from any known variety of continental Old French, which is indeed the impression that contemporaries appear to have had by the 13th century and even before. The question crucial to this study is what can be deduced from the characteristics of this idiosyncratic phonological system that might shed light on the circumstances in which it was acquired. The earliest rhymed verse shows only the beginnings of the /ai/–/ei/ merger, which becomes generalised around the middle third of the 12th century. The palatal phonemes are generally kept distinct from their non-palatal counterparts. At this time, it seems Anglo-Norman was being transmitted in an acquisitional context which resembled that of continental French: there is no reason to assume that the mechanism for the spread of these



Chapter 5.  Anglo-Norman phonology

sound changes differed either side of the Channel. By the late 12th and early 13th century period, however, when the other three sound changes studied were in progress on the continent, it appears that there was now some factor at work in the insular context of acquisition that prevented them from becoming established in England in the same way as before. The evidence of the data surveyed here is that it was the influence of English. Phonologically, then, Anglo-Norman belonged to the mediaeval Francophone continuum up to around the middle of the 12th century and experienced changes taking place within it. Typically, phonological changes spreading across a speech community are attributed to ease of articulation, changing prestige associated with variant forms and suchlike factors. Reduction of consonant groups and of vocalic distinctions clearly belongs to the first kind of explanation, and is routinely claimed by e.g. Pope (1934) to have been responsible for numerous changes in the development of French and proto-romance phonology. We know that by the late 12th century the pronunciation of the Île de France was starting to enjoy a greater prestige than other varieties, so this would be a reason why some changes initiated in that region were taken up elsewhere. So the failure of the /an/–/en/ and /ei/–/oi/ mergers and of the /ur/–/œr/ split to establish themselves in AngloNorman, despite the prestige of central French, which by the 13th century had adopted them, shows that England stood outside the domain of sociolinguistic influence of continental French. The phonemic contrasts which distinguished French from English concerned principally the /y/ vowel and the four palatal sounds /je/, /nj/, /yi/, and /lj/. Here the verdict of the philological tradition has been clear: Anglo-Norman was characterised by a tendency to identify all five sounds with a close equivalent in the Middle English phonological system. To that consensus we can add the finding in Section 5.8 that this tendency had a decisive effect between the mid-C12 and the mid-C13 century, at least with /nj/, /je/, and /yi/; in the earlier 12th century, rhymes still strongly tended to respect the existence of the Old French palatal phonemes, whereas in the later 13th century they strongly tended not to. The fate of the palatal vowels and consonants in Anglo-Norman thus showed sound changes taking place under L1 influence outside the mediaeval French dialectal continuum, and at a time when A-N had apparently ceased to undergo regular sound changes affecting the rest of the medieval Francophone community. As was noted earlier, this does not mean that all Anglo-Norman users were wholly ignorant of continental French phonology. A degree of bi-dialectalism, especially amongst the most highly educated and well-travelled members of the speech community, is most likely to have been the case. In addition to their knowledge of continental French, they were certainly very familiar with, and probably used, insular French. But most users of A-N at this time probably did

69

70 The Transmission of Anglo-Norman

not have lengthy experience of residing in France and familiarity with the French of France. Their phonological system reflected the acquisition circumstances available in England, and later Anglo-Norman verse reveals some of the properties of that system. These fall into two categories: the retention of some earlier continental French phonological traits, including changes implemented prior to the later 12th century, and the levelling of contrasts absent from English. The former category of items is not very informative about the acquisition of AngloNorman, but the latter suggests that Anglo-Norman was generally acquired at an age after the phonological systems of a simultaneous bilingual child’s languages are acquired. In the latter case they are acquired independently of one another. But if the second language begins to be acquired after early childhood, contrasts in the second-acquired language absent from the first-acquired are not mastered in naturalistic acquisition.19 We return to this issue in Chapter 11. For now, what has been demonstrated is that A-N phonology, on several key points, underwent strong influence from English well before the mid-13th century. This chapter has re-examined the thesis familiar from Pope (1934) and earlier scholars that the phonology of A-N showed extensive influence from English, and finds that the traditional view is securely upheld by a substantial body of corpus data. In diachronic phonology, as in historical linguistics more generally, verse evidence will not tell us when changes took place in vernacular speech. It does, however, indicate that by the time of such evidence A-N had already carried through the sound change in question, to the point where the tradition of treating the sounds separately in verse was abandoned. By the late C 13 there can be little doubt that Anglo-Norman phonology had diverged very significantly from continental Old French varieties. Notes 1. This remains true, despite the fact that, in pre-standardised languages, they may seem to offer a better window on pronunciation than standard language spellings. 2. This concurs with contemporary chroniclers’ descriptions of William the Conqueror’s army as having been made up of mainly of Normans, with some Bretons and Picards. 3. Throughout this chapter we use phonetic symbols available on a conventional keyboard, in preference to those of the International Phonetic Association. It will be seen that this affects palatal sounds in particular: phonetic values of the relevant symbols used will be indicated in the text. 4. In modern Standard French, the schwa vowel no longer has phonemic status, although a trace of it survives in certain modern dialects, such as Norman, in the form of a lengthened stem vowel (Leppelley 1999).



Chapter 5.  Anglo-Norman phonology

5. Unless, that is, these words are found at the rhyme. If they are, a pronunciation more akin to /u/ than to /œ/ may be assumed, on the indirect evidence that reflexes of tor (from Lat. turris) never have an /œ/ vowel in modern dialects. 6. As we can tell from the vowel assonances in the Chanson de Roland (pre-1100). 7. In early Old French the /ai/-diphthong also contrasted with a monophthong /ε/ (Pope 1934), but in Western dialects the difference was being lost by the time of the earliest extant texts from that region, dating from the middle third of the C12, as is seen in rhymes such as plaist : forest (Enéas 283). C12 A-N verse likewise rhymes original /ai/ : /ε/, e.g.: fere : cyvere (Amis 910)  apres : lais (Tristan de Thomas, ms. Sneyd 642). 8. A very rare occurrence of an /an/ : /en/ rhyme occurs in the mid-C13 poem Corset attributed to Robert de Gretham (Short 2007). 9. In C14 charters, especially from about 1330 onwards, however, spelling in words such as Monsieur, seigneur and leur are found, though whether they betoken a phonological alignment of A-N on CF at this late date is not clear: the virtual disappearance of A-N poetry by this stage means rhyming evidence is lacking. 10. The A-N Hub has only 21 instances of seul(e) spelling forms as against 171 of soul(e). 11. This notation represents a glide from a mid-low back vowel, to a high front vowel (Short 2007: 81). 12. Gower’s Mirour de l’homme poem, probably written in the 1370s, does show numerous /oi/ : /ei/ rhymes, but Gower’s awareness of continental poetic practice is well-known, and can hardly be taken as conclusive as regards later A-N pronunciation. 13. Old English had this vowel, but it had died out by the Middle English period. 14. Some continental spellings are found, but in the ANH religieus gets 5 citations, and gracieus 6, as compared with 28 for religious and 46 for gracious. 15. We follow Short (2007). Alternatively, one could regard the former as a complex segment /je/ formed by a palatal approximant and a mid-high front vowel. 16. From the traditional French folk-songs ‘En passant par la Lorraine’ and ‘L’Alouette est sur la branche’, respectively. 17. Identified by using the ‘ending with’ facility for concordance searching on the ANH. This produced the following /nj/ words: campaigne compaigne complaigne funtaigne gaaigne montaigne overaigne plaigne remaigne suveraigne atteigne desdeigne enseigne. 18. More precisely, as word final consonant segments, preceding schwa, as in meistre, conoistre etc. 19. Again, this does not rule out late mastery of L2 phonology by highly motivated assidous learners with a high language aptitude level, enjoying prolonged residence among nativespeakers­. But the capacity of exceptional individuals to learn an L2 falls outside the problem space considered in this work.

71

chapter 6

The syntax of quantifiers in Anglo-Norman

6.1

Introduction

Research has uncovered substantial variation in the syntax and semantics of quantifying words, both within and across languages (May 1977, Barwise & Cooper 1981, van Benthem 1986). Semantically, quantifiers, such as English all, many, few, each, enough, convey the extent to which membership of a set of entities is denoted, and broad uniformity may be said to exist across languages as to the range of meanings found in this domain.1 In syntactic terms, however, there is a great deal of variability. Although some discussions of the subject, in historical linguistics and elsewhere, treat quantifiers as a grammatical category, in practice there is no such grammatical class, at least in a language such as English. Quantifiers may be noun modifiers, e.g. many people, or else function as pronouns, e.g. many of the people, that is, they have the distribution of noun phrases (cf. a large number of the people). Furthermore, modern English has separate modifier and pronoun forms in some cases, e.g. no and none, as well as pairs of forms, one of which cannot function as a pronoun, e.g. every, while the other form can be either a pronoun or a modifier, e.g. each. These examples taken from within a single language suffice to show that quantifying words do not behave as a single grammatical class. When semantic equivalence across languages is considered, further variation in syntax-semantics correspondences can be expected, and this was indeed the case when we compare Middle English and Old French. In this chapter a contrastive analysis of quantifying words in these two languages will first be offered, after which an analysis will be conducted of how A-N behaved on points where they differed. As will be seen, a further dimension of the problem is that the divergence between the two languages grew as time went on, since continental Old French quantifier syntax shifted somewhat during the course of the Old French period. It should be noted that for comparative purposes quantifier use in early Middle English will be considered only up to c. 1225, as by the 2nd quarter of the 13th century Middle English showed widespread signs of influence from French (cf. Trotter 2003c), so by this time it may be problematic to take ME data as a potential influencing factor independent of French. This seemed a desirable precaution­

74

The Transmission of Anglo-Norman

to take, even though such influence is normally thought to have affected lexis only. The aim of this investigation is to present evidence for whether A-N syntax was acquired independently of English in the domain of quantifiers. 6.2 Acquisition target properties: Old French quantifiers Old French possessed a considerable range of mono-morphemic lexical items that can be classed as quantifiers semantically. However, their syntactic category status varied. Most could be found as pre-modifiers agreeing with the head noun, e.g.: (1) a. Aucunes gens dient…  ‘Some people say…’

Roman de la Rose 1

b. Ne savez vos nules noveles de vostre mere?  Queste del Saint Graal p. 74 ‘Don’t you know any news of your mother?’ c. De chascune flor issoit arbres portant fruit a grant plenté.  Queste del Saint Graal p. 171 ‘From each flower came a tree carrying abundant fruit’

They could also be used as pronouns taking a PP complement, headed by the preposition de, in which case the Noun phrase governed by de was semantically definite: (2) a. Ne je ne tu n’avrons la terre: aucuns de ces autres l’avra.  Roman de Thèbes p. 116 ‘Neither you nor I will have the land: one of the others will have it’ b. Or me dites… se vos trovastes puis nul de nos compagnons…  Queste del Saint Graal p. 147 ‘Now tell me… if afterwards you didn’t find any of our companions…’ c. Car chascune des colors avoit en soi une vertu.  Queste del Saint Graal p. 202 ‘For each of the colours possessed a power’

Whether or not quantifiers of the nul class premodified the noun directly, they agreed with it in gender, cf. (1c)–(2c) above and: (3) a. Et nonpourquant ele aleta chascun de ces enfanz.  Miracles de Saint Louis p. 154 ‘And nevertheless she breast-fed each of her children’ b. Uns hermites i chantoit chascun jor.  ‘A hermit sang there each day’

Mort le roi Artu p. 246



Chapter 6.  The syntax of quantifiers in Anglo-Norman

The quantifiers considered so far belong to a group to which we shall refer as the nul class. Another group of quantifiers, to which belonged molt, tant and poi, behaved somewhat differently. They could premodify the head noun directly, in which case molt and tant normally agreed with it in gender and number (but see (4c)):2 (4) a. Escuz unt genz de multes cunoisances.  ‘They have splendid shields distinctively marked’ b. Tanz jurs t’ai desirret.  ‘I have desired you for so many days’

Roland 3089

St Alexis 471 (Buridant 2000)

c. Lores l’ensewisent poi humes.  ‘Then few men followed him’

Elucidaire iii p. 6

But when used before de, molt class items were always invariable in form (poi has been included for completeness in (5c), though it was invariable in the premodifier construction as well): (5) a. Molt de riens dont mainte gent/ont oeus.  ‘Many things from which many people benefit’

Eracle p. 124

b. Tant i avrat de besanz esmerz.  Roland 132 (Buridant 2000: 168) ‘There will be so many pure gold coins’ c. Il avoit… pris auoc soi poi de moines.  ‘He had taken with him a few monks’

DialGreg 71

In this construction the NP was generally indefinite, as in these examples. This was not possible with the nul class: examples such as *nul d’amis or *aucun d’amis are not found. It could be also be definite, as the following examples show, though with poi and tant at least, this was very unusual: (6) a. Cil a Keus navré mortelmant/Et ocis a molt de sa gent. Wace, Brut p. 142 ‘He mortally wounded Kay/And slew many of his men’ b. Tu… meis en toi tant des vertuz de l’ anemi com Nostre Sires i avoit mis des soes.  Queste del Saint Graal p. 126 ‘You put in yourself as many powers of the devil as Our Lord put in his own’ c. Un pou de lor gent/Tienent ci un chastel. ‘A few of their men hold a castle here’

Adenet, Buevon 2220

One or two more rarely used quantifiers, such as alquant and auques (‘some’, ‘quite a number’) had only the pronominal, not the modifier use (Buridant 2005: 180), e.g.:3 (7) a. Et molt grant peine ont sufferte/Alquant de noz bons anceisurs.  Wace Brut 6188 ‘And some of our good ancestors suffered very great hardship’

75

76

The Transmission of Anglo-Norman

Saisnes A 2380

(7) b. Auques de ses homes a ça outré remés.  ‘Quite a few of his men stayed on the other side’

Tout normally stood before the definite article, i.e. as a premodifier, as in (8a) below, or else it directly modified the head noun, as in (8b). It not attested with a de PP (Buridant 2000: 161–66), unlike the other quantifiers so far discussed. Roland 1800

(8) a. Es destriers muntent tuit li barun de l’ost. ‘All the lords in the army mount their chargers’

Roland 394

b. Et tutes terres met en chalengement. And they claim all the lands’

Although tout could also appear alone in what one might consider a pronominal use, this is probably best handled separately as quantifier-float (Kayne 1975): Roland 702

(9) Vers dulce France tuit sunt achiminez.  ‘They set off for sweet France’

In short, the main characteristics of the Old French system lay in the alternative syntactic frames in which most quantifiers could appear, and the ability of the molt class, but not the nul class, to introduce an indefinite nominal in the de PP frame. These features are displayed in Table 1: Table 1.  Early Old French quantifier syntax (up to later C12)

Molt class Nul class Tout

6.3

Premodified noun

Pronominal + de + def. nom

Pronominal + de + indef. nom

Y Y Y

Y Y N

Y N N

Diachronic developments in the Old French quantifier system

The situation described in the previous section holds for the early part of attested Old French, up to the mid-to-late C12, but changes soon become apparent in the textual record. The ability of quantifiers of the molt class to modify nouns directly did not long remain the case. As shown by Carlier (2009) using the case of the quantifier molt, these items increasingly tended to become detached from the noun phrase and to appear with a de PP, as in (5a). Carlier argued for a grammaticalisation account of the quantifier molt, which derives from the Latin adverb multum. From having been an adverb modifying a clause, the Romance descendants of multum (molt in OFr) took on a new syntactic function within a nominal constituent, with



Chapter 6.  The syntax of quantifiers in Anglo-Norman

which they were frequently juxtaposed. She regarded the detached uses of molt as having adverb status, combining with de to form a ‘complex nominal determiner’: (10) ‘A partir du XIIIe siècle, la forme fléchie, déjà relativement rare en très ancien français, disparaît… Seul subsiste moult invariable. De répond donc à une contrainte syntaxique : il permet qu’un quantifieur de statut adverbial puisse se rapporter à un nom. À la suite de la disparition de la forme fléchie, il contribue à former un déterminant nominal complexe, pouvant quantifier un nom non intégré dans le syntagme verbal.’  (Carlier 2009: 43–4) ‘As from the 13th century onwards, the inflected form, already relatively rare in early Old French, disappears… Only the invariable form remains. De thus meets a syntactic constraint: it permits a quantifier having adverbial status to be related to a noun. After the disappearance of the inflected form, it contributes to creating a complex nominal determiner, able to quantify a Noun that is not part of the Verb phrase.’  [author’s translation]

Whether or not one adopts a complex determiner analysis, rather than taking the grammaticalisation process to lead to a new categorization of molt as pronominal, does not affect the immediate point that a change in the surface syntax took place. The direct nominal modifier use is lost, whether inflected or otherwise. Similarly, modifier uses of tant are rare by the later 12th century, and by the mid-13th century they are not found at all in continental French prose works consulted.4 Only the pronominal uses remain, e.g.: (11) a. b.

Queste p. 40

Tant de puceles  ‘So many girls’ Tant de biens  ‘So many goods’

Artu p. 162

Poi has very few attested modifier uses, even in the earliest texts, so by the early C13 the class consisting of molt, tant and poi retained only the syntactic construction with a de PP, as in (11). These trends gave rise to the later Old French system shown in Table 2: Table 2.  Later Old French quantifier syntax (C13)

Molt class Nul class Tout

Premodified noun

Pronominal + de + def. nom

Pronominal + de + indef. nom

N Y Y

Y Y N

Y N N

Concomitantly, the forms of molt and tant inflected for plural also died out.

77

78

The Transmission of Anglo-Norman

6.4 A comparison with Middle English Early Middle English quantifiers resembled those of Old French to some extent. Quantifiers such as fele, manie (‘many’), sume (‘some) and fewe (‘few’) could directly modify a plural head noun: (12) a. Ure elderne hadden fele fon. ‘Our elders had many enemies’  b. Þo waren manie childre dede.  ‘Then many children died’ c. To sume men cumeð ure loured. ‘Our Lord comes to some men’ d. Fewe friend we scule finden.  ‘We shall find few friends’

TrinHom p. 194 TrinHom IV TrinHom V Prov Alfr XIV

They could also take an of PP, comparable to the Old French de PP construction. However, here they differed from the equivalent French items in that the nominal expression following of in such instances had to be definite, cf.: (13) a. Monie of þas wimmen  ‘Many of those women’ b. Ani of ðine friend  ‘Any of your friends’

TrinHom V VV, 65

c. Elch of hem wrot his uers  ‘Each of them wrote his verse’

TrinHom IV

d. Sume of þe wordes þe weren speken  ‘Some of the words that were spoken’

TrinHom IX

If the quantified nominal was indefinite, the quantifier had to be used as a direct head modifier, as in (12). Another difference was that the morphological form of plural quantifiers in ME was the same in both functions, whereas in Old French only the modifier use showed agreement.5 The quantifiers sume, eni, elch, fele/mani, fewe corresponded to OFr aucun, nul, chacun, molt, and poi. Swa mani/swa fele corresponded to OFr tant, of which there was no mono-morphemic equivalent. Middle English alle corresponded to tout in OFr not only semantically, but also syntactically, as it either modified the quantified noun directly, or acted as a predeterminer, preceding the definite article, e.g.: (14) a. Iwisliche þa clennesse iwelt alle unþeawes.  ‘Wisely (?) purity overcomes all faults’

Lamb Hom. sermon 10



Chapter 6.  The syntax of quantifiers in Anglo-Norman

(14) b. Alle þe fughelas þe flughen bi þan lufte. ‘All the birds that flew through the air’

Lamb. Hom. sermon 12

Unlike modern English, alle did not allow an of-PP, any more than OFr allowed quantifier tout to be followed by a de PP. In sum, there was a superficial similarity between early Old French and Middle English syntax, in that both had alternative constructions with Q + N as well as Q de N/of N strings. The syntactic similarity changed when the OFr molt class quantifiers lost the ability to directly premodify nouns, but no such change took place in Middle English, where fewe and mani remained adjectival modifiers when quantifying a nominal. Thus the syntactic category status in the two languages of words with quantifier meanings corresponding to the OFr molt class became clearly contrasted in the two languages, which provides us with our principal domain of enquiry in the following section. 6.5

An investigation of quantifier syntax in Anglo-Norman

The domain of quantifiers was clearly one in which the two languages presented rather differently in terms of syntactic distinctions, especially from the C13 on. If the transmission of A-N is seen as a matter of proficient L2 learning, potential pitfalls were numerous. Semantic translation equivalents did not always share the same syntactic category properties, and even where they did, as with all and tout, the fact that a de PP construction was available with other quantifiers could have led a learner to overgeneralise and incorrectly use tout with de PPs. L1 transfer from English would be a potential source of error, inducing the L1 English user of French to continue to construct quantifiers of the molt class as adjectival modifiers of nouns. As in the clausal domain investigated in later chapters, therefore, English L2 learners of mediaeval French would have been faced by acquisitional challenges likely to give rise to error if L1 English influence made itself felt. As its members did not pose this challenge, little needs to be said about nul class quantifiers, where the two alternative constructions in OFr, with and without de, are abundantly illustrated in A-N, e.g.: (15) a. Car chescun de ces pechies est mortiel.  ‘For each of these sins is a mortal one’

Lib Cust 22 (c. 1300)

b. A cel tens fu un tronk mis en chescune eglise de Engletere.  Reisbritt 218 (c. 1275) ‘At that time a trunk was placed in every church in England’

79

80 The Transmission of Anglo-Norman

(15) c. … ke nuls de noz Bailliz ren n’offrent ne promettent  ‘… that none of our bailiffs offers or promises anything’ d. Si nul dettour voudra faire fyn… ‘If any debtor shall wish to make settlement...’ e. Se aucuns noz subgiz faisoit le contrere  ‘If any of his sworn affidavits are received by us’

Roy Let 2, 131

Foedera 2,1050 (1338) Foedera 3 754 (1364)

f. Nonobstant aucun serment que fait nous avetz au contraire  Foedera 3 618 (1361) ‘Notwithstanding any sworn affidavit that you have made to us to the contrary’

But such conformity with continental norms is uninformative for present purposes. This class of quantifiers continued to modify head nouns throughout OFr as their equivalents did in English, and they took de with definite NPs only, again as with English. Thus A-N users would have used them as illustrated in (15) if they had acquired A-N independently of Middle English, or if their A-N was calqued on Middle English. The crucial area to be considered was quantifiers of the molt class: Old French quantifiers of this class lost their ability to premodify nouns directly, so it can be determined whether Anglo-Norman did likewise, or whether they kept the modifier use that their English counterparts still retained. This question was pursued using the full resources of the Anglo-Norman Hub up to the mid-14th century.6 There are very few prose texts in the Anglo-Norman Hub that predate the mid-13th century, and on the basis of Carlier (2009) the timespan between the mid-12th to the mid-13th century was the crucial time when the molt class of quantifiers lost modifier status in continental French, at least according to the textual record. The chief linguistic variables were whether in A-N the molt class quantifiers quantified over indefinites in the construction with de, and whether A-N followed the continental Old French change towards exclusively using de + N in the case of the molt class. For this purpose, data were divided into four periods using dating information provided via the ANH. These have texts dated c. 1150 which have been included in the first half of the 12th century, which is otherwise rather thinly provided for. Thereafter the time periods used were 1160–1250, the second half of the 13th century, and the first half of the 14th century. Ideally, it would have been preferable to separate the 1160–1250 period in two, roughly in keeping with the other analysis time periods. Unfortunately, a good number of the texts in the Anglo-Norman Hub within this timespan, especially poetic ones, have rather imprecise estimates of composition date, so this was simply not possible.



Chapter 6.  The syntax of quantifiers in Anglo-Norman

Texts were searched for uses of the molt class quantifier items in the various spellings current in A-N. Syntactic patterns with the less common quantifiers auques and asez were also tabulated, but given the parallel syntactic behaviour of ME alle and OFr tout mentioned above, findings from tout could not be expected to shed light on the question at issue in this chapter, and were not analysed. The nul class was left aside, for the same reason. The results are set out in Table 3: Table 3.  Syntax of A-N quantifiers, molt class With de Without de TOTAL

N % N % N

C12 1st half

1160–1249

1250–1299

1300–1349

10 13 68 87 78

  34   21 120   79 154

132   77   40   23 172

  97   94    6    6 103

At the outset, in the 1st half of the C12, uses of molt class quantifiers with de were relatively rare in A-N. By the last half of the 13th century, they occurred over three-quarters of the time, and by the first half of the 14th century over 90% of the time. No more than a residual handful of instances by this time retained the modifier syntax found with the equivalent English words. A-N clearly developed on the same lines as continental French, and differently from English A related question is whether, as happened in continental French with molt (Carlier 2009), the shift to syntactically detached status went along with a move to a morphologically invariable form, i.e. whether the plural inflection was lost as molt and tant tended to become detached. Accordingly, frequencies were tabulated of agreeing versus invariable quantifier forms in the two contexts, with and without de, using the A-N spelling mult. Table 4.  Occurrences of mult and tant in ANH showing plural agreement With de Without de

N % N %

C12 1st half

1160–1249

1250–1299

1300–1349

– – 47/65 72

  6/31 19 73/118 62

84/128 66 18/39 46

  72/90   80   4/4 100

As can be seen, the continental loss of overt agreement with the molt class of quantifiers was not followed through in A-N. Though it was clearly under way in

81

82

The Transmission of Anglo-Norman

the C12, the trend was then reversed and agreement marking apparently strengthened in the later C13 and beyond. This is seen in Figure 1, where the overall percentage frequencies of plural marking on the quantifier are plotted in graph form against the rise in the percentage frequency of the de construction: 100 80 60

Forms with de Plural inflec.

40 20 0

Early C12

1160– 1249

1250– 1299

1300– 1349

Figure 1.  Syntax and plural morphology of molt class quantifiers7 (%)

The tendency towards recovery of plural inflection marking on the molt class quantifiers is a rather unexpected finding. Although A-N followed the syntactic trend of Old French, it did not do so morphologically, retaining and indeed more strictly observing plural marking on quantifiers even when they stood before de. We return to this point below. The use of de + quantifier in A-N was distinct from of + quantifier in English: in almost every case, the reading of the nominal quantified by molt, tant or poi was clearly indefinite. The only exceptional case, where either a definite or an indefinite reading was plausible, was the following example: (16) … Hereward … ki eschapa par les mareis oue poi des genz ki [li] swirent.  Reisbritt 152 ‘… Hereward… who escaped across the marshes with a few (of the) people who followed him’

Here, the indirect object [li] is an editorial insertion. With or without it, the sentence can be read as stating that some of his followers (definite) went with Hereward, or that some people (indefinite ) followed him. Putting this unclear case aside, it can be said that in the other cases an indefinite reading was unquestionably intended, e.g.:



Chapter 6.  The syntax of quantifiers in Anglo-Norman

Stfrancis 274

(17) a. Ové taunt de lermes me apelastes!  ‘You called me with so many tears!’

b. Les gentz de Noef Chastell’ paient tant de dismes et autres taxes com…  Northumberland 258 ‘The people of N. pay as many tithes and other taxes as…’ Mirour 82

c. Co poez vus ver en muz de maneres. ‘This you can see in many ways’

Secré 28

d. Kar mut des mals suffert avez. ‘For you have suffered many evils’

An apparent complication arises in that in later A-N confusion took place between de and des spellings, as can be seen from (17d) above. Normally, however, the context made it clear that an indefinite reading was intended. For direct comparability with Carlier’s (2009) study of molt, the results for molt alone are presented below, in Table 5: Table 5.  Frequencies of syntactic and morphological patterns with inflectional forms of molt in A-N, 12th–14th centuries

C12 1st half N % 1150–1249 N % 1250–1299 N % 1300–1349 N %

mult sg + Npl.

mult pl. + Npl.

mult sg. + de + Npl.

mult pl. + de + Npl.

TOTAL

18 28 32 40 16 12  0  0

36 56 26 32 18 13  3  4

10 16 16 20 23 17  9 11

 0  0  6  8 79 58 68 85

  64   80 136   80

Even though A-N was slower to abandon mult + N entirely than Carlier (2009) found for continental Old French, the impact of the change can be seen here almost as strongly in the case of mult as with the molt class overall (Figure 1). In the earliest period, up to the mid-12th century, over 80% of the data showed molt directly pre-modifying the head noun, without the proposition de. In the last period, the first half of the 14th century, over 95% of the data show the new pattern with de. The change has clearly occurred by the last half of the 13th century, when nearly 3/4 of the data show the new syntactic pattern. Accompanying the rise in de use, there was a sharp increase in plural agreement marking, as also occurred with tant for which data were included in Table 4 above.

83

84

The Transmission of Anglo-Norman

The syntactic trend followed with molt and tant strongly suggest the absence of English influence. The direct modifier construction with the molt class, the only one available in Middle English with indefinite nominals, was virtually ­eliminated in later A-N, just as it was in the course of Old French. Anglo-Norman was evidently behaving as a normal dialect of mediaeval French here, though the change apparently needed somewhat longer to take hold in England. 6.6 Implications The syntax of quantifiers in Anglo-Norman indicates the independence of AngloNorman syntax from Middle English, and of its faithfulness to ongoing developments in mediaeval French. The absence of contact effects is quite clear-cut, and suggests that the transmission of Anglo-Norman must have taken place in such a way that syntax was acquired independently in the two languages. In fact, evidence of contact influence appears to have gone in the opposite direction. The common use of the preposition of in Middle English quantified noun phrases was undoubtedly strengthened by the existence of de in AngloNorman. Furthermore, occasional Late Middle English examples can be found showing quantifiers plus of with an apparently indefinite nominal: (18) a. So went Iulius Cesar oute of this lande with a fewe of Romayns that were lefte a-lif.  Brut p. 32 (c. 1400) b. Soo many of children fatherless and soo many churches wasted.  Foure Sonnes of Aymon Ch 1. (late C15)

This construction is not found in Early Middle English, as noted above. The examples quoted above both come from texts translated from either insular or continental French, and may well be an imitation of the syntax of the original. Although Q + of with an indefinite nominal, as in (18), seems to have been very marginal and did not take hold in later English, it testifies to the effect that intense contact through bilingualism could have on the syntax of quantifiers. It is thus all the more remarkable that no contact influence in the opposite direction, from English on A-N, has been detected in this investigation. Of course, such influence, had it occurred, would not have been blatant in the sense of producing constructions that were ungrammatical in Old French. The constructions possible in Middle English were all possible in early Old French anyway, and they can all be found in A-N, taking the data surveyed as a whole. What is significant is the near-elimination as time goes on in the use of one of them, the Q + N construction, with the molt class. Middle English offered no basis for this development, but continental Old French did.



Chapter 6.  The syntax of quantifiers in Anglo-Norman

In general terms, the acquisition challenge to a bilingual A-N/ME speaker was a matter of relating particular lexemes to particular meanings in the two ­languages. This could present at least two kinds of problem. The semantic domain covered by one item in one language could be carved up between two or more in the other. This arose in cases where there was a grammatical contrast in English but not in French, e.g. (un)countable nouns and negative indefinites, e.g. molt = manie/much, nul = anie/non. Contrasts such as these, existing in the source language but not in the target language, did not pose a learnability problem to learners of A-N, or at least not one that produced observable errors. The problem cases addressed in this volume all go in the opposite direction, i.e. where a contrast in the target language, OFr, had no counterpart in the source language, English. In this chapter we have surveyed the behaviour of A-N in just such a scenario: a contrast developed between the syntax of the nul class of quantifiers and that of the molt class in C13 Old French, but not between English quantifiers expressing the same quantifying notions. A-N users respected the evolving contrast in French. Though using many as a direct modifier in Middle English with an indefinite nominal, they ceased using molt as a direct modifier and used only molt de. Also, had Middle English been a source of transfer influence, we would expect to see the Q + de + N sequence reserved for definite noun phrases, which was not the case. Anglo- Norman evolved in keeping with continental French, crucially in a direction taking it away from earlier congruence with semantically equivalent English items in earlier stages of Old French. As with the preceding chapter, then, it has been found that Anglo-Norman displayed a syntactic change in progress in mediaeval French, despite the fact that it went against the equivalent syntactic pattern of Middle English.8 To return to the points made at the outset of this chapter, the grammatical category assignment of quantifying items vis-à-vis their semantics is to a large extent arbitrary, so this was a domain in which a sizeable challenge could be expected to learners of Old French as a foreign language. No evidence, however, that those acquiring Anglo-Norman at this time were induced to make errors attributable to English influence was discovered, and even more conclusively, the diachronic trend in French was found to have been followed even when it ran against the distribution of quantifiers in English. It thus appears that Anglo-Norman was acquired in its own terms, without translation influence, and at an age where learners can accurately acquire syntax to a native-like level. Postulating an acquisition mechanism for the grammar of Anglo-Norman that in essentials was like that of a native language has received further support in this chapter. Its users seem to have behaved, and evolved, as members of a Francophone linguistic community, not as instructed second language learners of Old French.

85

86 The Transmission of Anglo-Norman

There remains the intriguing finding that while following the development of mediaeval French quantifier syntax, Anglo-Norman users apparently did not do so as regards plural morphology on quantifiers. We say ‘apparently’ because there is an inevitable caveat over the interpretation of the data as they appear in texts, in particular as to whether the graphemic evidence offers reliable evidence of overt realisation in spoken use. For now, we will adopt a precautionary position, to the effect that final plural marking on quantifiers was pronounced, and thus that A-N users did not follow the trend in continental French whereby plural marking by an overtly pronounced final -s died out probably by the 15th century (Fouché 1958). In Middle English no such change took place. From this perspective, then, the propensity of Anglo-Norman users after 1250 to mark plural morphology even on detached (pronominal) quantifiers could be taken as evidence of English influence, or at the very least, of the absence of French influence, as regards the realisation of final -s. Putting the problem in this way may in fact be crucial to understanding what was happening: matters of phonology were affecting grammar transmission. In the next chapter, we shall see that this problem becomes of central importance when considering whether the transmission of Anglo-Norman respected mediaeval French noun gender norms. When word forms need to be taken into account, i.e. when the level of morphology has to be envisaged, the realisation of morphemes in phonological form complicates matters. In the next chapter a way is found of addressing such issues. Notes 1. Though they may or may not be lexicalized, e.g. the absence of a modern French lexeme meaning ‘both’. 2. Poi was rarely used in this construction, however. 3. ‘Pronominal’ it should be remembered, is our own term, not used in traditional historical French approaches, which consider ‘detached’ molt and tant as adverbs. 4. In LFA. 5. In some OFr texts, systematic respect for the noun case system affected the position: molt could be a nominative plural form if the noun phrase whose head it modified had a subject function. The early demise of the case system in A-N meant this was not an issue in our investigation of insular French. 6. Verse is known for greater freedom in the order of clause constituents than prose. However, there seemed no reason to expect versification to produce any real distortion of the plural morphology variable. Also, the use or avoidance of the construction with the preposition de in verse would in practice offer the poet the no doubt convenient alternatives of one more or one



Chapter 6.  The syntax of quantifiers in Anglo-Norman

fewer syllable in the line, so it cannot be seen as a systematic biasing factor in favour of one or the other syntactic form. 7. Frequencies of plural inflections for mult and tant only, as in Table 4. 8. It also went against the pattern of the other available linguistic model in mediaeval England, Latin. This is a language which would also have been a potential source of interference, given the similarity of the word forms in Latin and French and the fact that in Latin the equivalent quantifiers multi, tanti and pauci directly pre-modified nouns.

87

chapter 7

Noun gender marking in Anglo-Norman

7.1

L2 gender acquisition background

Research into the acquisition of noun gender in contemporary L1 and L2 acquisition has pointed up clear differences in outcomes in this domain. L1 learners successfully register the gender properties of nouns on the basis of their distribution in input – young children’s abilities in this domain are well documented (see e.g. Clark 1985, Granfeldt 2005), and gender errors are rare. Hawkins & Franceschina (2004) postulated that there is a critical period for gender acquisition, past which learners will not acquire gender assignment as a fixed property of nouns as L1 learners do. This is particularly the case if the L1 lacks gender features. In Hawkins & Franceschina’s ‘failed functional feature’ (FFF) approach, the functional feature of gender will not remain active in L2 acquisition if it has not played a part in L1 acquisition. Hence English native speaker learners of Dutch gender fared particularly badly compared with speakers of Romance languages, as these authors showed. However, Swedish-speaking learners of French fared a lot better, thanks to the presence, so it is argued, of gender as a functional feature in both languages. The occurrence of gender errors is notorious as a sign that an English speaker does not have nativelike command of a language such as French. Speakers of English, a language without grammatical gender, appear from L2 acquisition research to be particularly challenged here. Dewaele & Véronique (2000) documented persistent problems with noun gender marking in anglophone learners of French, even among those with otherwise high proficiency in the grammar of the language. The quantitative extent of gender errors need not be dramatically high, however, for the conclusion to be drawn that noun gender represents a major challenge to L2 learners. Herschensohn & Arteaga (2006) found that, measured by token frequency, two advanced adult American learners of French had gender concord error rates of 5% and 16%, in spoken data recordings. They had error rates of 2% and 9% respectively on gender marking by article forms. These findings suggest that gender marking among proficient L2 French speakers can best be regarded as comparatively infrequent, yet not negligible. These two studies measured noun tokens. Granfeldt (2005), measured error in relation to noun type frequency, in samples taken at pre-school age, and reported a level of gender errors by parallel child bilingual learners (French and Swedish) ranging

90 The Transmission of Anglo-Norman

between 4%–8%. The corresponding figure for Swedish untutored adult L2 learners of French averaged about 25%. This sharp difference indicates the importance of childhood onset of acquisition in acquiring gender. Similarly, Guillelmon & Grosjean (2001) found that early bilinguals show responses to incorrect gender marking that are indistinguishable from those of monolinguals. The picture that thus emerges from contemporary research into noun gender acquisition by L2 learners is that those who acquire the L2 in childhood fare better than adult learners, and that if the L1 lacks noun gender, L2 performance on this measure is impaired. Both these factors, as we shall see, are important for understanding the transmission of noun gender among other grammatical properties of A-N. 7.2

Acquisition target properties

Old French nouns belonged normally to one of two genders, masculine or feminine, for which no consistent phonological or morphological properties offered entirely reliable cues. Nevertheless, a number of sub-regularities applied, especially with respect to derivation from Latin etymons, e.g. nouns ending in -é derived from Latin -tas were feminine. This applied also to items derived from Latin -or nouns, such as honneur, feminine in OFr, though masculine in ModFr; other such items, such as amour, could be either masculine or feminine. Gender variability was likewise found in comté (‘county’) and duchié (‘duchy’). Buridant (2005: 60–61) notes a number of further cases where noun gender was either variable or different from ModFr. In this chapter a distinction will be observed between gender assignment – i.e. the inherent gender property of a given noun – and gender agreement marking, i.e. how the gender assignment of that noun may be reflected in agreeing forms elsewhere in the clause in which the word appears. Old French typically distinguished gender agreement in the choice of determiner (article, possessive and demonstrative items, among others) and in the anaphoric pronoun used to refer to a noun in a text. The following example will illustrate these aspects of gender marking in OFr. (1) Quant il orent veüe la plaie ensi come il la porent veoir, il disent qu’ele estoit moult perilouse a garir. Artu p. 151 ‘When they had seen the wound, as they could see it, they said that it was very hard (lit. dangerous) to heal’



Chapter 7.  Noun gender marking in Anglo-Norman

The object pronoun la in la porent veoir (‘could see it’) is co-referential with the feminine noun plaie and agrees with it in gender, as do the definite article la and the adjective perilouse. However, unlike in ModFr, the past participle has the feminine -e ending, agreeing with its direct object, even though the latter stands after it. Gender assignment was not reflected in gender agreement in the same way in all dialects of OFr. In Picard, the definite article le was generally used for feminine nouns as well as masculines (Pope 1934: 488). As will be seen in this chapter, this notion of a gender contrast that is neutralised in one context but not in others will be important to consider in relation to A-N. As regards Middle English, the Old English grammatical gender system had almost entirely broken down by the early 13th century thanks to the general collapse of inflectional marking on nominals; forms distinguished by gender, as with pronouns he/ha (‘he’/’she’) and possessive forms his/hire (‘his’/’her’), survived only as markers of natural, not grammatical, gender. So in terms of L1 background, those using A-N by the 13th century would be counted as having an L1 system without gender, in terms of the contemporary studies cited in Section 1, and thus to have been posed with a very substantial acquisitional challenge. 7.3

Gender in A-N: previous research

Gender marking in A-N should be a fruitful domain to investigate, in order to throw light on its nativelike status or otherwise. Earlier studies have seen noun gender marking in later A-N as problematic. Tanquerey (1916: lv) found sporadic gender errors with determiners that were ‘dus à l’ignorance’ as he put it, as well as failures of adjective-noun gender agreement that he took to be either spelling mistakes or due to English influence. In his edition of the Shropshire Bills in Eyre (c. 1290), in which he commented adversely on the standard of French, Bolland (1914) drew attention in particular to gender marking errors, e.g. cele Thomas, cely Alice. Gender errors were also noted by Kibbee (1996: 10) as a factor reflecting English influence. English influence is a plausible-looking explanation of these gender-marking violations, since Old English noun gender distinctions had largely disappeared by Early Middle English. If it operated, L1 influence would be expected to have neutralised gender marking, either by using a single form to replace the two members of a gender opposition, or by using the two forms indiscriminately. Now, at first sight, indiscriminate gender marking abounds in later A-N from the late C13 onwards, especially in the form of cases where the modifying words such as articles and adjectives disregard the gender marking appropriate to the Noun they modify in continental French. Masculine modifying forms are used with feminine nouns, e.g.:

91

92

The Transmission of Anglo-Norman

(2) a. en cel prison ‘In that prison’ b. sicum en acun rivere ‘As in some river’ c. par certeyn enchesoun  ‘For a particular reason’

Bolland (1914: 37) Britton 404 Gippswich 22

Conversely, determiner and adjective forms with final ‘-e’ can be found with masculine nouns, e.g.: (3) a. de ceste trepas  ‘Of this death’ b. de ceste escrit ‘Of this writ’ c. en le haute chemin vers Donestaple ‘On the high road towards Dunstable’

Bolland (1914: 12) Leics 194 Tanquerey (1916: 50)

These phenomena were noted in a pilot study reported by Ingham (2010), which analysed data from a period ranging from the later 13th century to 1326, published in the Parliament Rolls Of Medieval England (Given-Wilson et al. 2005). A-N was found to be losing the un/une distinction, with forms such as un enqueste, un espeye, un meyn, une bon boys, une payement. Similarly gender marking with cest/ceste (‘this’) and cel/cele (‘that’) was neutralised: cel ferme, cel foiz, cel prison, cele heritage, cele compassement, cest chose, cest fausete, ceste terme, ceste bail. Such examples show that the forms stigmatised by Bolland (1914) as evidence of provincial ignorance of French were also common by this time in the higher echelons of central government administration. However, these apparent erroneous forms may not be what they initially seem. They all involve the same phonological element: the weak final vowel spelt in Old French, usually assumed to have had something like a schwa realisation in continental Fr. It played a key role in OFr gender marking, since its presence signalled feminine agreement, while its absence signalled masculine agreement. Likewise, schwa in the masculine singular le contrasted with the /a/ vowel of the feminine singular la. As observed by Pope (1934: 438),1 A-N showed attrition of the final schwa vowel in the 13th century. The instability of schwa in A-N must therefore have had the consequence that it no longer reliably served the functions of gender marking. If final ‘-e’ spellings no longer corresponded to an audible phonemic distinction, their use as gender agreement markers would then have become vulnerable to confusion at the graphemic level, producing the erratic inflectional forms observed.



Chapter 7.  Noun gender marking in Anglo-Norman

The result of these phonological developments is that Anglo-Norman nouns could have retained their gender properties, even though the overt means to signal them on articles and adjectives, using schwa, had been lost. Apparent gender errors such as (2)–(3) would then not have been a matter of imperfect grammar, but rather would have stemmed from the erosion of a particular phonological means of making gender distinctions. This concept is already familiar in medieval French dialect phonology, as was noted above in the case of Picard. The hypothesis that the problem of marking gender in A-N in cases such as (2)– (3) was phonological, not grammatical,2 was addressed in Ingham (2010) by examining gender agreement on possessive modifiers mon/ma, ton/ta, son/sa, where schwa does not enter into the contrast. An analysis was made of these forms in the late 13th and 14th century Parliament Rolls of Medieval England (Given-Wilson et al. 2005), henceforth referred to as ‘PROME’, between the years 1310–1399. These two forms, in addition to their differing vowel qualities, were distinguished by the nasal consonant in the masculine, which is absent in the feminine. Consequently, if gender errors cropped up here as well, this would surely have been good evidence for imperfect grammar learning.3 In fact, until 1376, only three cases of gender marking diverging from continental French could be found, e.g.: (4) Qe la baillie ly soit livre a profit le roi et a son suffisante seurte. PROME 1328 ‘That the bailiwick be delivered to him to the king’s profit and with his sufficient guarantee.’

In the late 14th century, quite frequent gender errors were noted, e.g. the uses of son with feminine nouns in the following: (5) a. Come s’il l’eust dit au roi de son propre bouche.  ‘As if he had said it to the king in his own words’

PROME 1376

b. Et Maistre Henry retournast en son dit chambre.  ‘And Master Henry went back into his room’

PROME 1383

Errors of sa used with a masculine noun also start to occur in the later 14th century: (6) …estoit empeschez de sa voiage nadgaires fait vers les parties de Bretaigne.  PROME 1376 ‘(And he) was… impeached concerning his voyage formerly made to parts of Brittany’

To answer the question of whether this observed increase in errors was indicative of a change in the status of A-N, one must consider how to interpret the frequency data obtained. A methodological issue that then arises is the question of what

93

94 The Transmission of Anglo-Norman

measure­ to take as the denominator, in order to represent the frequency of error. In fact, to measure the incidence of error in any kind of frequency terms poses a problem in that grammatical errors among high proficiency L2 users would not be expected to be at all frequent, so even if those who wrote the petitions were advanced L2 learners rather than bilinguals, the frequency of errors as a proportion of all uses, e.g. of possessive determiners might still be quite small. Conventionally, what distinguishes native speakers from advanced L2 users is that under normal circumstances the former do not make grammatical system errors of the type in question here. On the very rare occasions when they do, such occurrences would be treated as inadvertent slips of the tongue or of the pen. On that assumption, distinguishing advanced L2 users and bilingual users of A-N would amount to finding a difference between a fairly low incidence of errors, and virtually no errors at all. This is not a scenario which lends itself readily to numerical analysis. Ingham (2010) considered PROME gender marking errors with son and sa in relation to the number of sessions of parliament in which they occurred. They are extremely rare indeed until the last few decades of the 14th century, averaging just one in every twelve sessions. Thereafter, they become more common, with 34 errors occurring in the 29 sessions between 1371 and 1399. Before, then grammatical gender properties of nouns must have been very reliably known to insular users of French, despite their English-influenced phonology and perhaps the frequent errors involving weak -e arose not from A-N users’ lack of awareness of gender properties, but for reasons to do with the phonological neutralisation of morphological contrasts involving weak -e. A subset of the data was accordingly searched to find out if the rate of weak -e errors also increased in the late fourteenth century. Petitions from around 1300 were compared with petitions written in the 1370s. Specifically the masculine premodifier forms cel, cest, nul and tout, on which errors had been previously noted, were investigated. Around 1300, cel, cest, nul and tout co-occurred with a total of 55 masculine and 28 feminine nouns, the latter counting as errors and constituting 33.7% of instances. In the 1370s the corresponding figure was 60 masculine and 34 feminine nouns (36.2%). So Ingham (2010) concluded that no great change took place in gender marking with weak ‘e’ forms in the intervening 70 years. Errors where weak ‘e’ was involved were already common at the beginning of the period studied, unlike the very low incidence of errors at that time with preposition + article forms. Yet the same data sources showed virtually error-free uses of gender agreement with possessive determiners, for nearly 100 years after the marking of these properties in agreeing forms featuring weak -e became unreliable. Such results strongly indicated the maintenance of the awareness of noun gender properties among A-N users until the later 14th century. In other words, inaccurate gender marking using the weak -e forms was not a matter of the loss of gender assignment.



7.4

Chapter 7.  Noun gender marking in Anglo-Norman

Gender marking in the A-N Hub textbase: findings from possessive determiner contexts

The previous research was carried out on the single-genre PROME corpus.4 Although the results seemed quite compelling, it could be argued that the research needed to be extended to a much wider range of later A-N texts belonging to different genres. For one thing, petitions make repeated use of certain themes and concepts, as is likely to happen when a single genre is selected, and this could have served to inflate accurate performance on gender with possessive determiners, whereas a fuller range of noun types might have revealed a higher rate of error. For the present research, the Anglo-Norman Hub syntactic study corpus was used, so as to investigate gender-marking of possessive determiner forms in a wider range of texts and periods. The study corpus texts (see Chapter 4) for the three periods, 1250–1299, 1300–1349 and 1350–1399 were searched for all uses of 3rd person singular possessive determiners. The masculine form turned up in the spellings son, sun, soen and suen, the feminine form only in the normal spelling sa. Instances were set aside where the modified noun has natural gender, e.g. fils, femme etc., or where institutional roles in male-dominated mediaeval society identified a noun with a particular natural gender, as with bailiff, attorne, seneschal, clerc, messager etc. This was necessary to ensure that targetlike performance was not artificially boosted by the presence of language-external clues to gender assignment, so that users of A-N can be seen to have made linguistic choices unsupported by realworld contingencies. The results of this analysis are shown in Table 1: Table 1.  Token frequencies of gender marked possessive forms in Anglo-Norman Hub study corpus texts 1250–

1300–

1350–

N

%

N

%

N

%

son + masc son + fem

742    3 745

99.6   0.4

1044     1 1045

99.9   0.1

699   61 760

92.0   8.0

sa + masc sa + fem

   2 814 816

  0.2 99.8

    2   943   945

  0.2 99.8

  13 546 559

  2.3 97.7

It can be seen that on this measure noun gender agreement errors are next to nonexistent until the second half of the 14th century. From 1350 onwards, however,

95

96 The Transmission of Anglo-Norman

a low but appreciable level of errors occurs. The A-N Hub findings are thus in line with those from PROME, first in that for most of the time gender errors are vanishingly rare, and also in that the time-frame where they become noticeably more common is the later C14. The convergence of the two studies using differing databases but the same methodology leaves little doubt that gender assignment in users of A-N was largely intact until the 1370s. It should be noted that the results are analysed in terms of token frequencies, and are thus directly comparable with results of modern-day L2 acquisition studies cited in Section 2 above that use this measure. Low but non–negligible error rates are usual among high proficiency L2 learners, but in these A-N data the incidence of error until 1350 is a mere 6/3551, or 0.2%. The belief put about by certain earlier authors mentioned in 7.3 that later A-N users did not adequately acquire noun gender cannot stand up to the evidence presented here that many hundreds of noun types had been successfully acquired with their gender property. Had this not been the case, one would expect random gender marking, or a default to one 3rd sing. possessive form, probably son, since the masculine was the default form used to mark English inanimate nouns when used in French texts. In fact, random gender marking with possessives is absent until the later C14, when we begin to see nouns in the A-N Hub used with both genders in the same texts, e.g.: (7) a. b. c. d.

toute sa seignourie Kingscouncil 19; son seignourie de Balymore Kingscouncil 131 soun comissioun Rough 16; sa commissioun Rough 24 soun Nef Rough 97; ma nef susdite Rough 138 sa service Rough 25; soun service Rough 242

These data uphold the finding of Ingham (2010), this time using a much broader database than the single-genre petitions used in that study. Whether the data are taken from formal petitions or from a range of different sources, the outcome is the same: until then, A-N users seem to have experienced no problem in identifying the gender of nouns. The mid-point of the C14 is thus shown to have been a clear watershed in the transmission of noun gender, after which errors such as the following become noticeable: (8) son sepulture Cronlond 5; son bosoigne SeyntzMed 79; soun dit eglise Kingscouncil 217; sa pucellage Manlang 41; sa roialme Manlang 23; sa paiement Albus 395

We return in a later chapter to the significance of the timing of the onset of such errors.



7.5

Chapter 7.  Noun gender marking in Anglo-Norman

Implications

Knowledge of gender properties constitutes a discriminating factor indicating whether A-N was acquired in a naturalistic fashion to a native-like standard, or whether it was acquired post-childhood as an L2. These new results confirm that noun gender marking errors are almost vanishingly rare in later Anglo-Norman, once phonological interference is controlled for. The contrast between the negligible incidence of gender agreement errors with son/sa and its frequent manifestation with modifiers where the presence or absence of weak -e marks gender agreement, analysed in Ingham (2010), is very telling. One might wonder why this point had not been previously noted in earlier studies of A-N. In our view it is attributable to the absence until recently of large-scale database resources, which would allow quantitative data to be obtained in such a way as to control for phonology in marking noun gender. One might argue that the presence of any errors at all in the earlier periods, before 1350, is a sign of non-nativeness in language learning, but this would be over-strong. Continental French texts very occasionally show non-conventional use of noun gender, and in modern studies of noun gender acquisition native speaker errors have also been reported.5 The same logic can be applied to insular French and would account for the tiny incidence of gender errors with possessive determiners until the second half of the fourteenth century. Then the results would be consistent with imperfect grammar learning by A-N users only from the 1370s onwards. Though extremely rare, the gender errors that do occur are nevertheless qualitatively interesting. But such tiny frequencies of occurrence surely need to be considered not so much in relation to the assumed competence of native speakers as to actual performance, in other words allowing for the possibility of mere slips (see note 5).6 To summarise, Anglo-Norman users were clearly not challenged by the acquisition of French noun gender as such, or at least not until the later 14th century. What gives rise to the perception that they were is the disappearance of reliable gender agreement marking in contexts involving weak -e. Thus an acquisitional context must be envisaged where gender marking on possessives could be very reliably acquired.7 It was not that of the foreign language class. If A-N users had been learning French as a foreign language, some non-trivial level of gender assignment errors would be expected across various contexts: the gross imbalance discovered between possessive determiner contexts and weak ‘e’ contexts would have no reason to arise. However, if A-N was still being learned naturalistically

97

98 The Transmission of Anglo-Norman

from spoken interaction, the outcome observed is readily explained. Because final -e was vulnerable to phonetic reduction and thus often unrealised, child learners would have heard gender marking consistently realised in possessive determiner forms, but not with the weak ‘e’ forms. These are indeed the patterns we see in the later A-N data analysed. At a certain point, however, this naturalistic context could no longer be relied upon to transmit knowledge of gender properties adequately, and gender errors even on possessive determiners begin to proliferate. It must be concluded that a change took place in the transmission of A-N such that incorrectly agreeing possessive determiner forms became used sufficiently for the cue reliability (Lightfoot 1998) of son as mark of masculine and sa as a mark of feminine gender to be compromised in their turn. The findings on gender agreement in this and previous work converge on a similar time for when that process took place. By the 1370s, gender marking errors with possessive determiners, i.e. outside weak ‘e’ contexts, were becoming quite common in the writings of A-N users who would have been exposed to such error-prone input some time earlier, at the time when they were acquiring the language. The overall picture, then is one in which nouns were until then reliably acquired with their gender properties, to all intents and purposes up to the same level of accuracy as with monolingual native speakers, but that the domain of application of gender marking in A-N was much more restricted than in continental French, thanks to the phonological factor of weak -e erosion. The evidence from the first half of the 14th century is that this system could be successfully transmitted, at least for a time. The fact that by then English had no surviving linguistic gender properties does not seem to have hampered the successful acquisition of these properties in A-N. Once again, then, it is seen that influence from English did not affect the transmission of grammatical features of A-N until a rather late stage of its existence. Notes 1. Short (2007: 39) makes the same point. 2. No detailed study of noun gender marking in A-N is known to us, and the editors of the A-N dictionary (Rothwell et al. 1977) did not indicate noun gender as a fixed or variable property of entries, so direct comparison with continental French usage cannot be made via this means. 3. Determined with reference to a standard Old French Dictionary (Greimas 1992). 4. The extant text of petitions in PROME is of uncertain authorship: they may have been drawn up in the locality where they originated, or they may have been put together by local government administrators once the petitioner arrived in London to deliver the petitions to Parliament (Dodd 2009).



Chapter 7.  Noun gender marking in Anglo-Norman

5. A study reported in Franceschina (2005) show Spanish L1 speakers making gender errors on about 1% of occasions (tokens). 6. Gender marking slips in adult French is not a topic where research is known to us. 7. It might be thought that knowing the gender of corresponding Latin nouns by itself would have permitted A-N users to achieve a high level of successful performance on French noun gender. Although knowledge of corresponding Latin items would certainly have been accessible to them, this fact alone does not fully explain what we observe with A-N gender marking, in particular the failure to mark gender using articles and adjectives, where knowledge of Latin should in principle also have applied. Therefore, A-N users cannot be seen as having simply reproduced the gender of corresponding Latin words.

99

chapter 8

Verb second and null subjects in Anglo-Norman

8.1

Introduction

In this chapter it is shown that until the later C14 A-N largely respected two leading requirements of Old French, one in main and one in subordinate clauses. Main clause syntax remained strongly verb-second, and null subjects in subordinate clauses were normally licensed only after an initial clause constituent preceding the finite verb. Both traits were good discriminants of French versus English syntax. Null subjects were common in OFr, but in Middle English were limited to impersonal contexts. Although Middle English also had Verb second properties, it clearly differed from the patterns seen in Continental French, allowing predictions to be made as to how Anglo-Norman syntax should have diverged from continental French, if L1 transfer from English had been operating on Anglo-Norman. In modern speech communities the properties and acquisition of V2 have by now received a great deal of attention. It is known that young children acquiring a verb second language do not normally go through a stage of making errors from which they recover, but acquire V2 from the outset in clauses providing suitable contexts, e.g. the following utterance in child Norwegian: (1) No kan æ ikke drikke det ‘Now I cannot drink it’

Ole 2;10 (Westergaard 2008)

Norwegian is a V2 language, requiring inversion of subject ae (‘I’) and finite verb kan (‘can’) after the initial adverb no (‘now’) in the child’s utterance in (1). Early bilingual German/French-speaking children invert subject and verb after initial XPs in German but not in French (Meisel 1990), indicating that sensitivity towards the cues for V2 grammars (Lightfoot 1999) is to be seen as part of the human language faculty. This chapter is laid out in three main sections, first summarising the differences between early Middle English and Old French with respect to verb second and null subjects. Next, empirical investigations are presented of the occurrence and evolution of verb second in Anglo-Norman in two different contexts, as

102 The Transmission of Anglo-Norman

found in the A-N Hub textbase. Then an investigation of null subjects is laid out drawing on the same resource. The chapter thus provides an opportunity for wellexplored formal syntactic properties of OFr to be investigated as to their status in A-N. The advantage they offer in this respect is that, as we saw in Chapter 2, such formal properties of ordering and (non)overtness have been researched in an L2 acquisition context, as well as in diachrony. They are therefore particularly advantageous areas in which to look for evidence that the transmission of A-N was either naturalistic, or a matter of post-childhood instruction, in the light of relevant modern findings on language acquisition. 8.2

Acquisition target properties in OFr: V2 and null subjects

Old French showed a very strong tendency towards Verb second syntax in main clauses (Adams 1987, Roberts 1993), in which initial constituents of various types triggered raising of the finite verb to C in root declarative clauses.1 The initial constituent triggering V2 could be a focused constituent, e.g. de vous in: (2)

(Cist autre me dient ce qu’il sevent… mais) de vous n’ai je mie oï ce que vos en savez. ‘(Those others tell me what they know but) from you I haven’t heard what you know of it’

LL III 132

It could also be a topicalised referential expression, e.g. por ce in: (3) Por ce li guerredonerai ge de mon povoir. ‘For this reason, I want to reward her as much as is in my power’

LL I 170

V2 could also be triggered by adverbial expressions, e.g.: (4) a. Or ai jo quant qu’il m’a mestier. ‘Now I have whaver I need’

Courtois d’Arras v. 482

b. Encore n’en savoit il riens, einz cuide que ce soit por la reïne…  Artu p. 130 ‘He still knew nothing of it, but thought that it was for the Queen’

Expressions of illocutionary force such as mar (‘woe betide’, ‘at one’s peril’) and voirement likewise triggered inversion (Buridant 2000: 524), e.g.: (5) a. Mar conquerron nos l’autre terre.  ‘We conquer the other land at our peril’

Roman de Thebes 1, 99

b. Voirement sont ce des aventures del saint Graal. ‘Truly these are the adventures of the Holy Grail’

Queste 52



Chapter 8.  Verb second and null subjects in Anglo-Norman 103

Given the very wide range of elements having very different discourse properties that triggered V2 in Old French, it was clearly a syntactic requirement of a similar order to V2 in modern German, rather than being a discourse configurational matter, as seems to be the case in some varieties of Northern Italian (Poletto 2004). Although the precise structural analysis of V2 clauses in OFr remains quite controversial (Lemieux & Depuis 1995, Vance 1997, Kaiser 2002, Labelle 2007, Kaiser & Zimmermann 2010), the key descriptive facts are reasonably well established. In a main clause, if a non-subject constituent begins the clause, the subject noun phrase or pronoun follows the finite Verb. The situation in subordinate clauses is far less easy to summarise, however, especially when, as some researchers have done, data from poetry is taken as relevant to the problem to be accounted for. C12 verse, as noted by Labelle & Hirschbuehler 2005, displays considerable parallelism with main clause syntax as regards V2. In C13 prose however, this is not the case.2 Certain types of initial elements did not normally trigger V2 (Foulet 1930). They include a small class of discourse and negative polarity adverbs, such as certes and ja, and left dislocated elements, e.g.: (6) a. Certes il n’a pas en vos tant de bien…  ‘Certainly there isn’t so much good in you’

Lancelot II 144

b. Ja li set frere ne fussent ocis par vos ne par vostre aide. Queste p. 54 ‘Never would the seven brothers have been killed by you or with your help’ c. Rex Chilperings il se fud morz. ‘King Chilpering had died.’

St Léger 116

To that extent, V2 in Old French was less strict than in a modern verb second language, such as German. Nevertheless, until the late 13th century, V2 is so prevalent in prose texts as to constitute a genuine syntactic constraint, on contexts other than those illustrated in (6). In Middle French, beginning in the 14th century, strong signs are evident that the V2 constraint fell into decline in continental French. It appears to have been lost in syntactically definable stages, according to previous studies. Vance (1997: 339) analysed the process in terms of a growing number of adverbial items that joined the class of non-inverting items. She refers to an ‘unstable class’ containing adverbial expressions that increasingly diplayed V3 when clause initial (aussi, lors, encore, ore) (Vance 1997: 264ff).3 In her analysis of data from the early C13 prose romance La Queste del Saint Graal, she found that tant (‘so much’), autrement (‘otherwise’) and donc (‘therefore’) also show V3. Ingham (2008) likewise found that certain adverbial constituents were losing V2 more quickly than others in the 14th century, whereas initial direct objects still consistently ­triggered inversion. Figures for Middle French by Zwanenburg (1978) show the same phenomenon.

104 The Transmission of Anglo-Norman

The reasons for V2 loss are complex and cannot be explored in any detail in the present research. Subject pronoun cliticisation has been proposed, but Vance (1997: 65) showed that nominal as well as pronominal subjects are found in V3 order in Middle French, e.g. in the early 14th century author Joinville (our examples): (7) a. Et lors je me parti de Joinville.  ‘And then I left Joinville’

Joinville, Vie de Saint Louis §122

b. A solleil levant mon seigneur Geoffroy de Sergines ala dans la vile. ‘At sunrise my lord G. de S. went into the town’  Joinville, Vie de Saint Louis §369

She also documented the increasing use, towards the end of the OFr period and beyond, of V3 with certain adjuncts, especially time adverbials, a finding replicated by Ingham (2008). For the purpose of what should be expected in later A-N, if it observed properties of continental French, the following should be noted: – Generalised V2 until around 1300 – V3 with some adverbs, beginning before 1300, but gaining momentum thereafter – V2 observance with both nominal and pronominal subjects – Signs of V2 loss occurring with both nominal and pronominal subjects Turning next to null subjects in Old French, the orthodox position (Thurneysen 1892, Foulet 1930, Roberts 1993) is that clause subjects could be left null in a configuration where they could otherwise appear after the finite verb, if overt. In other words, null subjects were licit if a non-subject constituent occupied the clause-initial position, e.g.: (8) Noveles vos aport moult merveilleuses.  ‘(I) bring you very remarkable news’

Queste 5, 12

The presence of the preposed Direct Object noveles in Spec CP licenses a null subject in (8); alternatively a postposed pronominal subject je would have been grammatical, cf. the postposition of the subject pronouns in the examples in (2)– (5) above. Although main clauses normally form the domain studied with respect to null subjects in OFr, subordinate clauses also showed a variant of the null-subject property. When a non-subject constituent stood before the finite verb of a subordinate clause, a pronominal subject could be left unexpressed, e.g.:

Chapter 8.  Verb second and null subjects in Anglo-Norman 105



(9) Por ce vos di ge que en nule maniere m’en tendroie que ge ne guerroie encontre cex…  Artu p. 135 ‘Therefore I tell you that I have no intention of not fighting against those…’

Here, in the first subordinate clause introduced by que, the XP en nule maniere precedes the verb complex ne entendroie, and the subject pronoun is omitted. In the second, however, no XP precedes the verb complex ne guerroie, and the subject pronoun is overt. If a subject pronoun was expressed with a preverbal XP, it had to stand to its left, immediately after the complementiser, e.g.: (10) … que je devant mon frere viegne ‘… that I might come before my brother’

Queste 260, 25

This is unlike the position in main clauses, where a subject pronoun cannot precede a preverbal XP. Note that a nominal subject, unlike a subject pronoun, could not stand between the complementiser and the preverbal XP. In structural terms, according to Vance (1997), the subject pronoun in (10) can be treated as enclitic on C, which would explain the limitation to pronominal subjects. Null subjects in main clauses were typically treated by earlier generative researchers as being licensed under government by a finite verb in C. This account clearly could not be adopted for subordinate clauses, where a complementiser stands in C. An analysis of null subjects in subordinate clauses, developing work by Vance (1997), was presented by Labelle & Hirschbühler (2005).4 In 13th century Old French, they proposed, the SpecIP position in subordinate clauses was obligatorily filled. This requirement was satisfied either by a nominal subject, or else by a scrambled non-subject constituent such as en nule maniere and devant mon frere in (9)–(10), and pronominal subject were enclitic on Complementisers, as in Vance (1997). If these analyses of subordinate clause structure are adopted, the null subjects in an example like (9) above cannot be a null pronominal standing in SpecIP, which is occupied by the preverbal XP, but is a null clitic on C. In sum, the task facing a learner of Old French as regards null subjects was to acquire the syntactic conditions under which a pronoun subject could be left null, in main clauses and also in subordinate clauses with preverbal XPs. 8.3

A comparison with Middle English

Main clauses beginning with a non-subject constituent in Old and Early Middle English normally displayed inversion of verb and subject if the latter was nominal, e.g.:

106 The Transmission of Anglo-Norman

(11) Þis shrud haueð ech man on him after his fulcninge. ‘Each man has this shroud upon him after his baptism’

Tr Hom 95, 23

Pronominal subjects, at least in Southern and Midlands dialects of ME, did not take part in inversion, however:5 (12) Ure lif we ledeð richtliche togenes ure louerd. ‘We lead our lives rightly towards our lord’

Tr Hom 9, 3

This was a stable system found from the earliest Old English prose texts onwards. However a small number of time adverbials, notably þonne (‘then’) and nu (‘now’), triggered inversion with both nominal and pronominal subjects. Van Kemenade (2010) has shown how the structural position of nominal subjects with respect to other clause elements, especially these adverbs, varied according to their ­discourse-new or discourse-old status, but the generalisation that nominal subjects inverted whereas pronominal ones generally did not, to which we refer here as the ‘asymmetric’ V2 pattern, remains valid. In later Middle English texts, as shown by Haeberli (2000), the variation in V2 realisation is such that it is difficult to infer from them a stable state of language in this respect. It seems most likely that, as Old English/early Middle English V2 became lost as a feature of ordinary language, it survived for a while as a stylistic and/or dialectal trait with certain writers. However, up to the early 13th century, it is clear that the asymmetric V2 pattern of early English is predominant across all text types, at least in dialects attested in that period. Thus the change that English seems to have undergone in ordinary usage, though in what time-frame cannot confidently be discerned from written data, is from an asymmetric V2 system in which nominal subjects tended to be post-verbal after a nonsubject initial XP, to one in which all nominal subjects were preverbal in declarative clauses, regardless of discourse status, or of the presence of an initial XP, i.e. the modern English system. It should be noted that whether the initial XP was an argument, such as a direct object, or an adverbial such as a time adjunct, never affected the status of V2 in medieval English. If A-N did experience substantial English influence on the syntax of V2, it should therefore have shown evidence of favouring V2 with nominal subjects, and not with pronominal ones. By the mid-14th century V2 was breaking down in Middle English, so if English influenced later A-N syntax, it should have contributed pressure to remove V2 altogether.6 Null subjects in early Middle English, as in Old English, were licit only with impersonal verbs, e.g.: (13) … þat naht ne scolde reinin.  ‘… that it should not rain’

V&V 143, 30



Chapter 8.  Verb second and null subjects in Anglo-Norman 107

There was a clear distinction between the grammars of later medieval French and Middle English as regards the possibility of a null subject in the subordinate clause. In Old French, this construction was productive where an XP preceded the finite verb, as we saw above, but not otherwise. In Middle English the fact that only an impersonal subject could be null operated regardless of the presence of an initial XP in the structure, whether in a main or in a subordinate clause. In the C13, English was therefore quite distinct from Old French in the conditions on null subject licensing and on V2. Although the two languages were eventually to converge in both respects, disallowing null subjects entirely and disallowing V2, these developments were only at an early stage in French in the C14, whereas in later C14 English they were going to completion in both respects. So the hypothesis of L1 English influence on A-N makes empirical predictions as to the forms that such influence would take: a collapse in V2 and null subjects in the C14, following a tendency towards patterning V2 and null subject use along the fault lines of English in the C13. In the subsequent sections of this chapter these predictions will be addressed. 8.4 Investigating the maintenance and loss of V2 in Anglo-Norman In the C14 signs of the impending demise of V2 became apparent in continental French. Analysing the status of V2 in Anglo-Norman, therefore, matters are complicated by the fact that the final period of its existence was when V2 was starting to be lost. Whereas L1 learners of French up to about 1250 acquired a fairly strict V2 grammar, by the 14th century they were acquiring a grammar with V2 after preposed objects, but with optional V2 following initial adverbials. This means that to assess the nativeness of Anglo-Norman, it has first, until around 1300, to be measured against the Continental French of an earlier period displaying regular V2, and then considered vis-à-vis a later Continental French period displaying partial V2 loss.7 In this chapter we seek to establish whether insular texts showed this evolution. V2 was therefore investigated over four time periods: the first and second halves of each of the 13th and 14th centuries. For this purpose the A-N Hub data is supplemented for the first period by two ANTS texts, and a short prose chronicle from the last decade of the C12. For the last period, some early C15 material is also included to supplement later C14 texts. Since the Anglo-Norman Hub textbase is unparsed, in order to allow syntactic structure to be studied, a procedure needed to be adopted of searching by lexical items that could provide data relevant to syntactic properties. Verb second contexts were identified in two ways, first by selecting lexical adverb items that in Old French

108 The Transmission of Anglo-Norman

functioned as triggers for V2, and also by searching forms of personal pronoun subjects and inspecting whether or not they appear with V2 syntax in clauses having a non-subject initial constituent. The former allowed V2 with SNom and Spro to be compared. The latter allowed the incidence of V2 after particular kinds of adjunct, notably time adverbials (Vance 1997), to be assessed. In Early Middle English as in Old English, V2 took place normally only in clauses having a nominal subject, not those with personal pronoun subjects, as we saw in Section 3. Therefore, if English influence was at work we would expect V2 to be observed noticeably more commonly in Anglo-Norman clauses with nominal subjects than in those with personal pronoun subjects. OFr V2 syntax with pronominal subjects had to be acquired in its own terms, unsupported by L1 transfer. It could thus be expected to show errors if used by late instructed learners, as in contemporary L2 case studies of V2. The analysis procedure also allowed clauseinitial arguments to be distinguished from clause-initial adjuncts, the dimension found to be important as regards V2 loss in late OFr and early Middle Fr. This investigation seeks to chart the decline of V2 in A-N by asking: given a particular type of initial non-subject constituent, how did the frequencies of the syntactic variants VS and SV order evolve over time? If A-N was influenced by English syntax, the following can be expected: i. in the 13th century, when the asymmetric V2 pattern of Middle English was still strongly apparent, V2 occurs regularly with nominal subjects but rarely with pronominal subjects ii. no argument versus adjunct dissociation should be apparent If A-N was influenced by medieval French syntax, on the other hand, the following can be expected: iii. no nominal versus pronominal subject dissociation with V2 iv. a dissociation should become apparent in the 14th century between a decline in V2 after initial adjuncts and V2 maintenance after initial objects These alternative hypotheses – (i) versus (iii), and (ii) versus (iv) – are investigated below in Sections 8.5 and 8.6 respectively. It will be seen that the issues were addressed by focusing on predicted outcomes in terms of the A-N data themselves, not on drawing quantitative comparison betwen OFr and A-N. Because the corpus resources of OFr and for A-N differ substantially in terms of spread across genres, and also in terms of time frame of available prose works, it would clearly have been unsatisfactory to use frequency data across the two varieties, in order to establish whether continental French and Anglo-Norman were different or alike in the relevant respects. The most informative way of proceeding, we believe, is the dissociation approach: did A-N ­dissociate the observance and non-observance of V2 in the same ways that have



Chapter 8.  Verb second and null subjects in Anglo-Norman 109

been independently established for Middle English on the one hand and continental French on the other? It was hypothesised that if A-N retained its status as a syntactic dialect of OFr, later Anglo-Norman would dissociate initial objects and initial adjuncts for the purpose of V2, but would not dissociate the inversion of nominal and pronominal subjects. 8.5

V2 after selected initial adverbs in main clauses

The C13–C14 time periods are first investigated for V2 maintenance or loss after initial adverbs, the context in which V2 loss first became noticeable in French. A way of addressing this research issue based on lexical searching by specified adverbs was adopted, as stated above, so as to deal with the fact that the A-N Hub is not syntactically parsed. Retaining from an initial search for designated adverbials only those instances where the target items stood in initial position produced contexts which in continental Old French of the early 13th century required V2 order. Frequencies were then calculated of V2 realisation or otherwise in these contexts. The expectation was that V2 should remain, up to the late 13th century, much as it had been in the early part of the century (see Chapter 4), but that during the 14th century V2 should show a clear decline. Data were analysed with four adverbs (ores, donque, encore, aussi) in main clause initial position. Two of these, ore and encore, were those studied in C13 Old French prose by Ingham (2005a) to which was added aussi (Ingham 2010). The adverb donc was also used in view of its observed high frequency in initial position. Initially it was planned to use lors as well, on the basis of its changing status with respect to V2 in Middle French (Vance 1997) but this adverb turned out not to appear in the 13th century AngloNorman Hub data, and was therefore dropped. A similar problem rose with après, which according to Vance, was one of those adverbs which led the way towards V3. However, in Anglo-Norman data it was almost always used as a preposition or conjunction, only three examples being identified of its use as an initial adjunct in the first half of the 14th century. Its extremely rare adjunct use in Anglo-Norman was somewhat disappointing, but a satisfactory amount of data was obtained by using other adverbs, as will be seen. In this analysis coordinating conjunctions were ignored, especially et, mais and car, which did not normally count for V2 at this time. Nor were instances scored where, as well as an occurrence of the targeted adverbs, another adverbial element stood before the finite verb. It was essential to know whether inversion, if it occurred, was individually triggered by the clause-initial item being ­investigated. Extensive spelling variation in Anglo-Norman required spelling variant forms shown in the AND for each adverb to be utilised in the search procedure

110 The Transmission of Anglo-Norman

A key question was whether the nominal and pronominal subjects would show different likelihoods of V3. If in general English was a strong influence on Anglo-Norman by the 13th century, a substantially higher rate of V2 with nominal than with pronominal subjects would be expected. The results of this analysis procedure are displayed in Table 1: Table 1.  A-N main clause order after four selected initial adverbs, 13th and 14th centuries C13 1st half

C13 2nd half

C14 1st half

C14 2nd half

VS SV Total Nom. Subj.

57  (100%)  0 57

  18  (82%)    4   22

23*  (70%) 10 33

  37**  (36%)   65 102

VS SV Total

39   (93%)  3 42

  97  (84%)   18 115

50*  (78%) 14 64

  36**  (57%)   27   63

Pron. Subj.

* p = 0.4572 (2-tailed), ** p = 0.0102 (2-tailed)

In the last two periods trends start to diverge, as regards V2 with pronominal and with nominal Subjects. A chi-square analysis (Fischer’s exact test) was performed on the figures for these two periods. The difference between the frequencies of VS with pronoun and nominal subjects in the C14 1st half is non-significant. However, the corresponding difference in the C14 2nd half is highly significant. The above results are shown graphically in Figure 1: 100 80 60

Pers. pron. Nominal

40 20 0 1190– 1249

1250– 1299

1300– 1349

1350–

Figure 1.  Rate of supply of V2 after four initial adverbs in main clause, by Subject type



Chapter 8.  Verb second and null subjects in Anglo-Norman

Until the mid-C14 insular French users were clearly following continental practice, not distinguishing nominal and pronominal Subjects for V2; in other words, they showed no influence of L1 transfer. This outcome is unexpected on the premises of the position adopted by some commentators on later A-N, to the effect that the grammar of French was imperfectly acquired as a result of English influence. It is compatible, however, with accounts of later Anglo-Norman which see it as a variety that tended to conform syntactically to the mainstream of continental French. The overall picture with these four initial adverbs is one of an initially very high level of observance of V2, followed by a gradual decline as far as the first half of the C14. In the second half of the 14th century, the decline becomes more sharply apparent, especially in the case of pronominal subjects, which as we have seen became significantly more likely to appear in V3 order. The personal pronoun subject/ nominal subject asymmetry in medieval English is not there in the Anglo-Norman data until the later 14th century. Until then, Anglo-Norman was clearly not converging with English as regards V2. Although Table 2 shows somewhat diverging rates of V2 observed after 1350, which could count in favour of Middle English influence, by this time V2 was virtually lost anyway in English. Until that point, the case for English influence on A-N V2 was non-existent on the basis of these results. 8.6 V2 with initial Objects versus initial Adjuncts The next investigation examined whether later A-N also showed the dissociation that would be expected, if A-N was following the early Middle French trend, between V2 retention after a preposed Direct Object, and V2 decline after an initial adjunct. The same sources, periodised in the same way as before, were used as for the previous analysis. The search terms used here were the subject personal pronoun forms ge, tu, il, ele, nous and vous, in their various spellings attested in the AND. As before the independent variable was SV or VS order, in clauses where the subject pronoun appeared following an initial nonsubject constituent, e.g.:

(14) Apres deivent il mercier le roi de France de… 

‘Afterwards they must thank the King of France for…’

The results of this analysis procedure are shown in Table 2:

St Sard 182

111

112 The Transmission of Anglo-Norman

Table 2.  SV or VS order with personal pronoun subjects in MCs, following preposed Objects or Adjuncts Dir Obj 1190–1250 1250–1300 1300–1350 1350–1420

N % N % N % N %

Adjunct

  VS

SV

Total

  VS

  SV

Total

  27 100   38   91   16   89   33   81

 0  0  4  9  2 11  8 19

27

  81   95 134   76   73   54 199   51

   4    5   42   24   61   46 189   49

  85*

42 18 41

176* 134*,** 388**

* p =