Generative Theory and Corpus Studies: A Dialogue from 10 ICEHL [Reprint 2011 ed.] 9783110814699, 9783110166873

Table of contents :
1. Structure
1.1. Continuity versus discontinuity
Obsolescence and sudden death in syntax: The decline of verb-final order in early Middle English
On the history of relative that
The complementation of verbs of appearance by adverbs
On the use of current intuition as a bias in historical linguistics: The case of the LOOK + -ly construction in English
The indefinite pronoun man: “nominal“ or “pronominal”?
1.2. Form and function
Coordinate deletion, directionality and underlying structure in Old English
The position of the adjective in Old English
On the history of the s-genitive
The passive as an object foregrounding device in early Modern English
Reinforcing adjectives: A cognitive semantic perspective on grammaticalisation
2. Text types
Variation and change: Text types and the modelling of syntactic change
The progressive form and genre variation during the nineteenth century
The conjunction and in early Modern English: Frequencies and uses in speech-related writing and other texts
3. Sociolinguistics and dialectology
Processes of supralocalisation and the rise of Standard English in the early Modern period
The rise and fall of periphrastic DO in early Modern English, or “Howe the Scots will declare themselv ’s”
Grammatical description and language use in the seventeenth century
Geographical, socio-spatial and systemic distance in the spread of the relative who in Scots
Inversion in embedded questions in some regional varieties of English
Putting words in their place: An approach to Middle English word geography
4. Phonology
HappY-tensing: A recent innovation?
Syllable ONSET in the history of English
Name index
Subject index

Citation preview

Generative Theory and Corpus Studies

W G DE

Topics in English Linguistics 31

Editors

Elizabeth Closs Traugott Bernd Kortmann

Mouton de Gruyter Berlin · New York

Generative Theory and Corpus Studies A Dialogue from 10 ICEHL

Edited by

Ricardo Bermüdez-Otero David Denison Richard M. Hogg C. B. McCully

Mouton de Gruyter Berlin · New York 2000

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

© Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability. Library of Congress

Cataloging-in-Publication-Data

Generative theory and corpus studies : a dialogue from 10 ICEHL / edited by Ricardo Bermüdez-Otero ... [et al.]. p. cm. — (Topics in English linguistics ; 31) Includes bibliographical references and index. ISBN 3-11-016687-9 (alk. cloth) 1. English language-Grammar, Generative. 2. English language—Discourse analysis. I. Bermüdez-Otero, Ricardo, 1970II. Series. PE1097.G46 2000 425-dc21 00-030389

Die Deutsche Bibliothek -

Cataloging-in-Publication-Data

Generative theory and corpus studies : a dialogue from 10 ICEHL / ed. by Ricardo Bermüdez-Otero — Berlin ; New York : Mouton de Gruyter, 2000 (Topics in English linguistics ; 31) ISBN 3-11-016687-9

© Copyright 2000 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Printing: Werner Hildebrand, Berlin — Binding: Lüderitz & Bauer-GmbH, Berlin Printed in Germany

Preface Ricardo Bermüdez-Otero, David Denison, Richard M. Hogg, C. B. McCully

The Tenth International Conference on English Historical Linguistics (10ICEHL) was held at the University of Manchester, England, between 21st and 26th August 1998. The occasion of the tenth anniversary of the ICEHL series was marked by a record number of participants, in excess of 260, attesting to the burgeoning growth of English historical linguistics as a discipline since the first conference took place at Durham in 1979. In total, more than 130 papers were presented during the six days of the Manchester meeting, requiring four parallel sessions. A distinguished panel of invited speakers gave plenary talks: Cynthia Allen, Elan Dresher, Paul Kiparsky, Roger Lass, David Lightfoot, Donka Minkova (the Tenth Anniversary Conference Speaker), Terttu Nevalainen, Nikolaus Ritt, Elizabeth Traugott and Wim van der Wurff. Several events were held in conjunction with the main programme; notably the Fourth G. L. Brook Symposium, convened by Christian Kay, ran during the last three days of the conference, with papers on lexis and semantics, demonstrations of on-line resources, and a discussion of the plans for OED 3 . Christiane Dalton-Puffer and Merja Kytö organised a workshop and teaching session on the use of statistics in historical linguistics, under the leadership of Roeland van Hout; there was a short parasession on Optimality Theory convened by Ricardo Bermudez-Otero; and, on the last day of the conference, a workshop was held on social network analysis and the history of English, with Ingrid Tieken-Boon van Ostade as organiser and moderator. This volume offers a selection of the papers presented in the general sessions; the proceedings of the Brook Symposium (including papers on lexis and semantics) and of the workshop on social network analysis will appear as separate publications: Coleman and Kay (forthcoming), Tieken-Boon van Ostade and Nevalainen (forthcoming). The fact that ICEHL had reached its tenth instalment called for celebration. Appositely, the title of the Tenth Anniversary Conference Speaker's plenary address, "As good as it gets", simultaneously referred obliquely to the

vi

The Editors

optimality-theoretic content of her paper and paid overt compliment to the achievement and proud tradition of ICEHL. The tone of the meeting was suitably festive: though a bad turn in the weather did its best to dampen delegates' spirits on Sunday 23rd (their day off), there was dancing in a marquee specially erected in the grounds of Hulme Hall, and participants were entertained amidst the pictorial delights of the Whitworth Art Gallery, the grandeur of the Town Hall, and the luxury of the Palace Hotel. Yet for the very same reason that celebration was in order, the organisers felt, when it came to publishing the proceedings, that their duties should go beyond selecting and editing a representative sample of the best contributions to the conference — not that this was in itself an easy task, for the number of papers submitted for inclusion in this volume exceeded 50. Rather, it seemed appropriate, when preparing this book, to ask what it could tell us about the current state of English historical linguistics. In the prevailing millennial climate, a measure of soul-searching appeared almost de rigueur. How much synergy is there between the efforts of the individual scholars who gather biennially at ICEHL? How successful have we been as a scientific community in delivering progress in the understanding of the history of English? The overall organisation of this volume is, to a large extent, an attempt to answer these questions, however partially. When one reflects upon the collective nature of English historical linguistics as a scientific enterprise, one cannot ignore the fact that, though in principle united by its object of study, the ICEHL community stands sharply divided by method and theoretical outlook. It therefore seems pertinent to ask to what extent English historical linguistics suffers from the ailments diagnosed by Kuhn (1970) as besetting "immature science". According to Kuhn, mature scientific communities are characterised — in all but a few sporadic revolutionary episodes — by the sharing of a single paradigm. A paradigm, in the broad sense of "disciplinary matrix" (Kuhn 1970: 182), is a coherent set of ontological assumptions, methodological guidelines and epistemic values, embodied in the actual examples of successful scientific practice on which a group of researchers model their work. 1 When united by a single paradigm, a scientific community is able to progress beyond debate over fundamentals, and so to focus its efforts on the steady and cumulative resolution of narrowly defined and relatively esoteric puzzles. An immature scientific discipline, in contrast, will according to Kuhn be in a chronic state of conflict between rival

Preface

vii

paradigms. Although each competing school of thought may be able to foster highly technical and advanced research within its particular confines, none is able to deliver a fully satisfying account of the shared object of study. In such circumstances, progress is hampered by continuous wrangling over foundational issues; principled decisions as to what datasets are relevant cannot be made or, when made, are hotly contested; and, rather than build on the secure groundwork laid by his predecessors, each young scholar is forced, at the outset of his career, to agonise over the very constitution of the discipline. In Masterman's (1970: 74) chilling summation, each sub-field as defined by its technique is so obviously more trivial and narrow than the field as defined by intuition, and also the various operational definitions given by the techniques are so grossly discordant with one another, that discussion on fundamentals remains, and long-run progress (as opposed to local progress) fails to occur. Does this sombre picture accurately reflect the current state of English historical linguistics? On the evidence of the array of papers presented at 10ICEHL, there is no gainsaying the theoretical polarisation of the field. The generative tradition, grammaticalisation theory and sociolinguistics were all strongly represented; the points of contention between these three approaches are well known and need little rehearsing here. •

For generative theorists, linguistic change is primarily driven by the language learner as, during the process of acquisition, he or she adopts innovative settings for one or more of the highly abstract parameters supplied by Universal Grammar. The resetting of a parameter involves an abrupt discontinuity which typically manifests itself as a cluster of superficially disparate innovations, punctuating an otherwise enduring equilibrium. The first half of van der Wurff s paper in this volume provides a particularly clear illustration of this approach. Drawing on proposals made by Bobaljik and Thräinsson (1998) he suggests that, roughly between 1400 and 1500, the Split-IP parameter was reset in English. As AgrSP, TP and AgrOP were collapsed into a single functional projection (the IP), English lost no less than five different grammatical possibilities: transitive expletive constructions, intransitive expletive constructions in which the logical subject precedes the lexical

viii

•

•

The Editors verb, preverbal objects, adverbials intervening between a verb and a non-heavy postverbal object, and discrete tense and agreement inflection on verbs. Proponents of grammaticalisation theory, in contrast, insist on the crucial role played by the communicative strategies and pragmatic inferences adult speakers resort to as they use language in ordinary conversational exchanges; change is regarded as essentially gradual, advancing through the routinisation of discourse strategies. This approach was very strongly represented at 10ICEHL, with papers on grammaticalisation running continuously in one of the parallel sessions for more than two days;2 with the exception of Paradis's, however, those dealing specifically with semantic change have been absorbed into the proceedings of the Brook Symposium. In this volume, the emphasis on discourse strategies is represented, for example, in the paper by Rosenbach, Stein and Vezzosi, who show that the historical decline in the use of the ^-genitive in English (under pressure from o/-phrases) came to a halt during the late Middle and early Modern periods; these authors suggest that the s-genitive, with the concomitant ordering of possessor before possessum, was redeployed to serve certain pragmatic and textual functions. For historical sociolinguists, in turn, the course of language change is ultimately determined by the attitudes which members of the community (differing in terms of age, sex, class, etc.) consciously or unconsciously adopt in respect of particular linguistic variants. Indeed, a sharp distinction is drawn between "innovation" and "change", where the latter is said to occur only when a new variant has been incorporated into the norm of some subgroup of the speech community. In this volume, Nurmi's paper clearly illustrates the importance which sociolinguists attach to the social evaluation of linguistic variants; Nurmi suggests that, following Elizabeth I's death, the arrival of James's Scottish entourage at Whitehall changed courtly fashion, precipitating a sudden decline in the use of periphrastic DO in affirmative clauses.

In preparing this volume we have made no attempt to downplay such theoretical dissension; on the contrary, the sequence of papers directly mirrors the polarisation of the field. There is thus an overall progression from articles primarily addressing aspects of internal structure to contributions

Preface

ix

with a predominant emphasis on external dimensions of variation. Within section 1, "Structure", the initial papers consider elements of syntax essentially in isolation; thereafter, attention shifts gradually to connections of syntactic phenomena with semantics, pragmatics and psycholinguistics. In section 2, "Text types", the focus on external factors widens further, as three articles consider how the varying pragmatic circumstances surrounding particular genres determine the frequency and distribution of syntactic constructions in the textual record. Finally, in section 3, "Sociolinguistics and dialectology", the emphasis is firmly placed on the embedding of linguistic variants within a geographic and social matrix. Does this mean that, in our view, Kuhn and Masterman's gloomy appraisal of multiparadigmatic science applies with full force to English historical linguistics? Not necessarily. 1OICEHL did indeed witness solemn restatements of principle, and controversy over fundamentals; there was no sign of an agreed standard concerning which data, from the "too sizable and inchoate pool of information" provided by the textual record (paraphrasing Kuhn 1970: 17), practitioners of the discipline should be held accountable to. Yet in preparing this volume we were impressed by a few encouraging signs: •

A substantial proportion of the papers depend on evidence obtained by performing quantitative analyses upon suitable corpora. This is symptomatic of the fact that the quantitative methods associated with corpus studies are fast becoming prevalent in the field. As has often been observed, a number of purely logistical factors have greatly contributed to this development: notably, a wide and ever expanding range of texts, from all periods in the history of English, available in machine-readable form, together with increasingly powerful automatic search tools. Yet, nurtured by ever more refined corpus techniques, a new climate of empirical responsibility has also been fostered in the discipline. The study of word order in Old English is a case in point: in this area, a lively synergetic relationship has developed between generative theorising and quantitative methods which might appear at first blush better suited to variationist paradigms. The papers by Allen and van Bergen in this volume show how, on the basis of corpus evidence, generative proposals about Old English word order are

χ

•

•

The Editors

routinely subjected to minute scrutiny, paving the way for criticism and further articulation. There also appears to be a growing awareness that in order to attain a comprehensive understanding of language change in its various aspects, one must attempt to integrate the partial insights afforded by competing paradigms. Variationist studies, for example, have consistently shown that syntactic innovations propagate at different speeds in different registers or text types; yet in his paper, van der Wurff acknowledges the charge that generative accounts of syntactic change often sidestep such observations. He nonetheless suggests that a synthesis is possible: noting that functional pressures can determine the incidence of particular syntactic constructions in a given register, he proposes the hypothesis that the various superficial developments associated with a reset parameter will be felt earlier in those registers where the incidence of factors triggering resetting is greater. Programmatic though van der Wurff s proposals are, their spirit is as important as their substance: his paper is a plea for serious work on how to integrate the fruits of work carried out within different research traditions. Finally, as one surveys the papers in this volume, one is struck by the fact that, though the authors may work within widely different theoretical frameworks, they can nonetheless be seen as grappling with a fairly narrow range of recurrent issues. A perennial source of difficulty, for example, lies in mapping categories onto the labile mixture of stasis and change displayed by the historical record. Another long-standing problem concerns the interplay between linguistic structures and the circumstances of language use. To a certain extent, the profile of these broad questions has been raised by the seemingly counter-intuitive answers which they have received in the generative tradition: e.g. that linguistic categories are discrete and reanalysis abrupt, and that at the core of grammar there lies an autonomous formal system whose design does not straightforwardly respond to the requirements of communicative functions. Whether or not one espouses these answers, the questions which they address pervade the day-to-day labours of historical linguists; the grouping of contributions in this volume attempts to highlight these broad themes.

Preface xi A recurrent topic in section 1.1 is that of continuity versus discontinuity — whether in the implementation of linguistic changes or in the categorisation of forms subject to historical transformation. Allen's paper, for example, challenges Lightfoot's (1991: chapter 3) claim that verb-final word order in subordinate clauses suffered a sudden death in early Middle English; she shows that the evidence of southern texts reveals a gradual decline, whilst in the Midlands there are no suitable late Old English texts with which to compare the Middle English evidence provided by the continuations of the Peterborough Chronicle. Continuity of a different kind is at issue in Seppänen's paper: he argues that the modern relative that must be viewed as retaining syntactic properties of both the Old English pronoun pcet and the Middle English complementiser that. Just as Seppänen's paper addresses the uncertain status of relative that as pronoun or complementiser, both Gisborne and Killie tackle the problematic nature of -ly derivatives used in construction with verbs of appearance (as in it looks most solemnly): Gisborne regards them as adverbs, performing a syntactic function intermediate between adjunct and predicative complement; Killie, in contrast, opts for an adjectival analysis, suggesting continuity with the use of deadjectival -lie in Old English to form adjectives referring to (visual) appearance. Van Bergen's paper presents another case study in problematic categorisation: she shows that, in Old English, indefinite man behaves like personal pronoun subjects, in that it fails to invert with the finite verb in clauses with a topicalised constituent; but it also behaves like nominal subjects, in that object pronouns can intervene between a subordinator and subject man. The papers in section 1.2 are broadly concerned with the interplay of syntax with semantics, pragmatics and psycholinguistic processing. It is heartening to see that none of these contributions evokes the acrimony of the theoretical debate surrounding the "autonomy of syntax", to which we briefly referred above; rather, an integrating spirit prevails. Perez Lorido, for example, observes that, given Subject-Object-Verb order in Old English subordinate clauses, Ross's (1970) Directionality Constraint predicts that, in subclauses subject to coordinate reduction, verb deletion should target the left conjunct (backward gapping); this prediction is, however, roundly refuted by corpus data. With brilliant panache, Perez Lorido turns the

xii

The Editors

argument around by showing that, for psycholinguistic reasons to do with parsing, an alternative strategy of verb repetition was preferable to backward gapping; in confirmation, he shows that the incidence of verb repetition declined as Subject-Verb-Object order (prompting forward gapping) became established in subclauses. A recurrent motif in the papers by Fischer, Seoane Posse, and Rosenbach, Stein and Vezzosi is the way that semantic and pragmatic constraints on sequencing affect the distribution of syntactic structures. Fischer, for example, argues that Old English strong adjectives, being associated with indefiniteness and "new" information, were common in postnominal position; weak adjectives, in contrast, implied definiteness and "given" information and hence normally occurred prenominally. Similarly, Rosenbach, Stein and Vezzosi suggest, as we noted above, that during late Middle and early Modern English the opposition between s-genitives and o/-genitives acquired new semantic and pragmatic significance associated with the opposition between possessor-possessum and possessum-possessor orderings. In turn Seoane Posse investigates factors which can confer topicality on an object, leading to its foregrounding through passivisation; intriguingly, she finds that early Modern English differs from present-day English (and appears to resist crosslinguistic tendencies) in according a relatively minor role to inherent object topicality in triggering passivisation. In the final paper within section 1.2, Paradis addresses the semantic and structural prerequisites for the grammaticalisation process whereby adjectives may acquire a reinforcing reading, losing their propositional content. She suggests that this development may only affect premodifying adjectives with a potential degree reading in collocation with degree nouns. We have already referred in previous paragraphs to van der Wurff s attempt at integrating the variationist focus on register with generative approaches to syntactic change. His proposal crucially relies on the premise that variables such as field, tenor, mode and purpose can affect the incidence of particular syntactic constructions in texts. This assumption is clearly borne out in section 2 by the evidence of Smitterberg's and Culpeper and Kytö's papers. Smitterberg finds that the frequency of the progressive form became an increasingly genre-specific feature in nineteenthcentury English: the use of the progressive declined in scientific writings, whilst its incidence in letters, drama, fiction and historiography was high and/or increasing. Similarly, Culpeper and Kytö show that the use of and in

Preface xiii written texts (particularly at clause level) has steadily declined since the mid-to-late seventeenth century; they correlate this development with changes in writing conventions, notably punctuation. As pointed out above, section 3 is primarily concerned with the embedding of linguistic variants within a geographic and social matrix. In this connection the rise of Standard English during the early Modern period attracts particular attention. Nevalainen's paper demonstrates that the linguistic features eventually incorporated into this supralocal sociolect had diverse regional and social origins: features of northern dialects were selected alongside others characteristic of the East Midlands, and change from below was in evidence alongside change from above. Nurmi's contribution focuses more narrowly on the decline of periphrastic DO in affirmative clauses after 1600; as we pointed out above, Nurmi attributes an instrumental role in this development to the arrival in London of King James's Scottish entourage. Moessner in turn provides an interesting angle on prescription by comparing and contrasting the metalinguistic pronouncements of seventeenth-century grammarians with their actual linguistic usage. The papers by Meurman-Solin and Filppula share a Celtic connection. In her survey of the spread of relative who (quha) in late Medieval and Renaissance Scots, Meurman-Solin argues for the kind of integrative work which was advocated above; she suggests that, to understand the propagation of who, one must take into account systemic, spatial, social and genrespecific factors. Filppula examines various hypotheses concerning the origins of inversion in embedded questions in certain dialects of English; he points towards influence from the Celtic substratum as the most plausible account. Black's concerns are also dialectological. The Linguistic Atlas of Late Mediaeval English provides reliable localisations for a large corpus of fourteenth-century and fifteenth-century texts; Black argues, however, that these texts cannot be used to reconstruct Middle English word geography without investigating the treatment of lexis by particular scribes. The volume closes with two papers on phonology: Beal's and Minkova's. Their somewhat odd position in relation to the preceding sections reflects the continuing expansion of syntax at the expense of phonology over the past decades.3 Beal presents compelling evidence against the prevailing view that so-called "HappY-tensing" is a recent innovation, spreading to the Northern English cities from the South. Minkova, in turn, argues that in Old English stressed syllables complied

xiv

The Editors

with a constraint requiring that their onset be filled; in the absence of an underlying onset consonant, a glottal stop was epenthesised. She suggests that this constraint became demoted during Middle English as a result of contact with French, creating the environment for resyllabification, elision, and the loss of alliteration between non-identical vowels. It is fitting that Minkova's paper should close the volume, not only in recognition of her status as Tenth Anniversary Conference Speaker, but also because we can look to her contribution, alongside others in this book, as a further model of work straddling research traditions. Whilst the topic of her paper already occupied the minds of nineteenth-century philologists, with whom she is in continuing dialogue, her conclusions are framed in — and informed by — the latest trends in generative phonology. Let us hope that, through such breadth of vision, the ICEHL community will in future succeed in rendering itself immune to charges of Kuhnian immaturity. We close with thanks to many individuals and organisations who helped make 10ICEHL a success. Without those who offered and delivered papers there would have been no conference: our first debt is of course to all of them. Sponsorship was gratefully received from the British Academy, the British Council (Poland, South Africa), Cambridge University Press, the Linguistics Association of Great Britain, Oxford University Press, and the University of Manchester Research and Graduate Support Unit. Enormous help was provided by the staff of Hulme Hall (Ian Ordish and others), by the accounts section of the University's Finance Department, by the Faculty Accountant and Printing Officer (Lucy Crompton, John Cochrane), and by Manchester Computing (Louis Blanchet, Jackie Boyer, Alistair Bamber). The hardworking organisers of workshops were noted above; here we thank Christian Kay, Christiane Dalton-Puffer, Merja Kytö and Ingrid Tieken Boon-van Ostade collectively for their commitment to making the conference a useful and enjoyable experience. A number of (anonymous) colleagues acted selflessly as referees of conference abstracts, and another group refereed full submissions for this collection; the conference was, and the book is, greatly the better for their efforts. Elizabeth Traugott has been as assiduous and supportive a General Editor as we could have wished for. Emma Moore has put this book together with cheerful efficiency, right up to camera-ready copy. And finally we must acknowledge the local ICEHL team. The organising committee which

Preface

xv

worked for up to two years before and during the conference included Debby Collier, Tolli Eythorsson, Christina Lee, Petra Storjohann and Joanne Townsend. For the conference itself we also enjoyed the enthusiasm and competence of these helpers: Alison Cort, Fernanda Miguens, Karolina Shaw, Helen Thomas and Mark Wharmby. The boys — actually, most of them were girls — done good.

Notes 1. Kuhn (1970: 187) refers to such instances of scientific achievement as "paradigms in the narrow sense" or "exemplars". Within our field, for instance, Labov's (1963) work on diphthong centralisation in Martha's Vineyard would be a paradigm-defining exemplar in sociolinguistics. For an elucidation of the polysemy of the term "paradigm" in Kuhn's work, see Masterman (1970). 2. As a piece of anecdotal evidence, the word "grammaticalisation" was explicitly mentioned in the title of five of the papers presented during the conference. 3. Adamson explicitly referred to this phenomenon ten years ago in her preface to the proceedings of 5ICEHL.

References Bobaljik, Jonathan David and Höskuldur Thräinsson 1998 Two heads aren't always better than one. Syntax 1: 37-71. Coleman, Julie and Christian Kay forthcoming Lexicology, Semantics and Lexicography: Papers from the Fourth G. L. Brook Symposium. (Current Issues in Linguistic Theory 194.) Amsterdam and Philadelphia: John Benjamins. Kuhn, Thomas S. 1970 The Structure of Scientific Revolutions, Second edition, enlarged. Chicago: University of Chicago Press. Labov, William 1963 The social motivation of a sound change. Word 19: 273-309. Lightfoot, David 1991 How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Masterman, Margaret 1970 The nature of a paradigm. In: Imre Lakatos and Alan Musgrave (eds.), Criticism and the Growth of Knowledge. Proceedings of

xvi

The Editors the International Colloquium in the Philosophy of Science, London· 1965, Volume 4. Cambridge: Cambridge University Press.

Ross, John 1970

Gapping and the order of constituents. In: Manfred Bierwisch and Karl Heidolph (eds.), Progress in Linguistics, 249-259. The Hague: Mouton. Tieken-Boon van Ostade, Ingrid and Terttu Nevalainen forthcoming Special issue of European Journal of English Studies on Social network analysis and the history of English.

Contents 1. Structure

1

1.1. Continuity versus discontinuity

1

Obsolescence and sudden death in syntax: The decline of verb-final order in early Middle English Cynthia Allen

3

On the history of relative that Aimo Seppänen

27

The complementation of verbs of appearance by adverbs Nikolas Gisborne

53

On the use of current intuition as a bias in historical linguistics: The case of the LOOK + -ly construction in English Kristin Killie

77

The indefinite pronoun man: "nominal" or "pronominal"? Linda van Bergen

103

1.2. Form andfunction

123

Coordinate deletion, directionality and underlying structure in Old English Rodrigo Perez Lorido

125

The position of the adjective in Old English Olga Fischer

153

On the history of the s-genitive Anette Rosenbach, Dieter Stein and Letizia Vezzosi

183

xviii

Contents

The passive as an object foregrounding device in early Modern English Elena Seoane Posse

211

Reinforcing adjectives: A cognitive semantic perspective on grammaticalisation Carita Paradis

233

2. Text types

259

Variation and change: Text types and the modelling of syntactic change Wim van der Wurff

261

The progressive form and genre variation during the nineteenth century Erik Smitterberg

283

The conjunction and in early Modern English: Frequencies and uses in speech-related writing and other texts Jonathan Culpeper and Merja Kytö

299

3. Sociolinguistics and dialectology

327

Processes of supralocalisation and the rise of Standard English in the early Modern period Terttu Nevalainen

329

The rise and fall of periphrastic DO in early Modern English, or "Howe the Scots will declare themselv 's" Arja Nurmi

373

Contents xix

Grammatical description and language use in the seventeenth century Lilo Moessner

395

Geographical, socio-spatial and systemic distance in the spread of the relative who in Scots Anneli Meurman-Solin

417

Inversion in embedded questions in some regional varieties of English Markku Filppula

439

Putting words in their place: An approach to Middle English word geography Merja Black

455

4. Phonology

481

HappY-tensing: A recent innovation? Joan Beal

483

Syllable ONSET in the history of English Donka Minkova

499

Name index

541

Subject index

551

1. Structure 1.1. Continuity versus discontinuity

Obsolescence and sudden death in syntax: The decline of verb-final order in early Middle English Cynthia Allen

1.

Introduction

It is a widely held assumption among students of historical syntax that language learners play a central role in syntactic change. I make the usual generativist assumption that children learn a language by internalising a grammar based on the linguistic data to which they are exposed in the form of the speech of older members of the community. 1 Adult speakers clearly play a very important role too because they determine the sort of linguistic data to which these language learners are exposed. For example, the adults of one generation may use a given construction with a particular frequency but the adults of another generation may use the same construction with a greater frequency for some reason — such as being exposed to foreigners who have learned English but have only imperfectly learned the use of this construction and therefore use it in situations where native speakers would not. This second generation has internalised the same grammar as the earlier generation and has changed only the frequency of the use of a construction. However, this change in frequency may have profound effects on the grammars of the next generation as children make a new assessment of the nature of the construction. To give a simple example, in Old English the verb could come at the very end of a clause or there could be material "extraposed" beyond the verb, giving Verb-Object order on the surface. "Verb (projection) raising" could also apply with auxiliary verbs to give surface Verb-Object order (see for example van Kemenade 1987 and Pintzuk 1991). One widely held idea is that the evidence for underlying Object-Verb order became obscured in late Old English or early Middle English, possibly because of a change in the frequency of some constructions, and language learners reanalysed the medial position as the more basic position for the verb. In the earlier days of generative grammar, such a change was interpreted

4

Cynthia Allen

as a change to which order was the basic order given by the phrase structure rules of the language and which one was the derived one, created by transformations. With the introduction of the Principles and Parameters approach, however, syntactic change of this sort was viewed as a change to the language-specific parameter settings which are assumed by a child learning the language. The assumption is that a change in the nature of the linguistic data may cause the children of one generation to set these parameters differently from the way they are set in the grammars of older speakers. More recently, within the Minimalist Program, syntactic change has largely been viewed by generativists as typically changes to such things as the inventory of functional projections or the weak or strong nature of a feature. For simplicity's sake, I will couch my discussion here in terms of parameter setting. However, the points that I want to make apply equally to any account in which the adoption of a new analysis is assumed to involve the immediate replacement of an old one. This parameter setting view of linguistic change raises a very interesting possibility for the obsolescence of a construction. It predicts that it is possible for a construction to die a very sudden death shortly after it has been a productive and robust construction. This could happen as a result of a new parameter setting making the construction impossible. How could such a resetting happen in the face of apparently robust evidence for the old setting? Lightfoot (1991, see especially chapter 2) argues that children do not ignore robust data from main clauses but they do not have access to any data from subordinate clauses, except for the very front part of the clause. So the evidence may appear robust in subordinate clauses but a construction may die suddenly because the evidence is for some reason now only found in subordinate, not in main, clauses. Lightfoot (1991: chapter 3) presents the loss of verb-final order in early Middle English as a striking example of the sudden death of a construction which results from this sort of development in main clauses. 2 The idea that Object-Verb order disappeared quite abruptly in the early Middle English period is far from new, 3 although Lightfoot (1991) offers a new and interesting suggestion for why such an apparently sudden change should happen. In this paper, I will examine the evidence, including some new evidence, for the commonly held belief that Object-Verb suffered a sudden death in early Middle English.

The decline of verb-final order in early Middle English 2.

5

Word order in Old and Middle English

Let us first briefly outline current conceptions of Old and early Middle English word order. It is an indisputable fact that Old English had frequent verb-final order in subordinate clauses, although there are differing reports on just how common this order was. It is also indisputable that verb-final order was more common in subordinate clauses than in main clauses. Finally, it must be agreed that verb-final order disappeared sometime in the Middle English period or shortly thereafter. There is general agreement on these facts, but the interpretation of the facts of course varies a great deal according to theoretical orientation. There is widespread agreement that a fundamental change to the grammar took place in early Middle English which caused the loss of Object-Verb order; following the work of Canale (1978), this change is usually dated around or before circa 1200. There has been considerable debate about what the exact nature of this fundamental change was. Until recently it has generally been accepted that in Old English the verb phrase was head-final, i.e. the verb was generated verb-phrase-finally, and any sentences with the verb in any other position showed the effects of movement, either of the verb or of some other elements of the sentence.4 In Middle English the situation changed and the verb was generated to the left of its complements, although it might undergo subsequent movement. Any sentence-final verbs at this time must be accounted for by movement either of the verb or of some other element. In contrast to this position, Roberts (1997) adopts Kayne's (1994) theory of phrase structure in which heads must always precede their complements and argues that the loss of morphological case in Middle English led to a loss of the ability of objects to move into the Specifier position of A G R Q P which had in Old English resulted in surface ObjectVerb order (Roberts 1997: 420).5 It is not possible to discuss here all the possible variants that have been proposed, and for my purposes it is not important exactly what devices in the grammar are supposed to be responsible for these orders. In the following discussion, when I refer to "reanalysis" 1 will be treating it as a reanalysis in the base position of the verb, but the basic points I will make apply equally well to other accounts which assume a fundamental change to the grammar in early Middle English, whether it be a change to headedness

6

Cynthia Allen

or the nature of a functional category. What is of importance here is that such accounts often lend themselves to the interpretation that the old verbfinal surface pattern went from being fairly vigorous in late Old English to being nearly non-existent in early Middle English. Indeed, the supposedly nearly overnight disappearance of verb-final sentences as a productive pattern is often adduced as evidence for a fundamental change to the grammar. In the following sections, I will argue that the evidence which has been adduced for the sudden demise of verb-final order in the early Middle English period is seriously flawed. The evidence suggests that the loss of this order was gradual and that some mechanism in the grammar is needed to account for the verb-final orders found in Middle English, regardless of whether we assume a change to the underlying position of the verb or not.

3. Problems with the evidence It is important to realise that in looking at word order in early Middle English we are severely hampered by the paucity of original texts — as opposed to copies of earlier texts — in the twelfth century. There is no dearth of English texts in twelfth-century manuscripts, but the problem is that nearly all of them are either known or suspected copies of earlier works. In these copies, we simply do not know whether we have the syntax of the original or the syntax of a later scribe who copied the text, or (more likely) a mishmash of archaic and innovative syntax. This problem is well known and it is for this reason that people looking at early Middle English syntax rely heavily on the Peterborough Chronicle Continuations of the twelfth century, and these texts are indeed a great help to our knowledge of the syntax of the time, although they are unfortunately brief. But a problem with the Continuations is that they are written in a very different dialect from the texts which we have from late Old English; late Old English texts are mainly in the West-Saxon dialect, while the language of the Peterborough Chronicle belongs to the East Midlands, on the northern border of the South Midlands area. Therefore we have the very real possibility than when we compare these texts with, say, earlier parts of the Anglo-Saxon Chronicle we are really comparing apples and pears. Thus when we find a complete lack of some orders familiar from late Old English texts, we

The decline of verb-final order in early Middle English

7

simply cannot be certain whether this represents a rapid change or not. We simply do not know how frequent verb-final order or, for that matter, any other order was in the Peterborough area in late Old English, and so the Peterborough Chronicle Continuations have nothing to tell us about how sudden the loss of verb-final order was in this area. It may well be that there was a rather sudden change in the Peterborough area. Contact with Danish in this area might have caused a massive and sudden change to the data heard by language learners and so the construction by one generation of very different grammars from those which were constructed by previous generations. But if this is what happened in this area, it would of course not be an instance of language learners failing to notice the evidence in subordinate clauses for verb-final order, because in fact they would be basing their grammars on data which did not have this evidence. Therefore the Peterborough Chronicle Continuations also provide no evidence one way or the other on whether children ignore certain types of evidence in their acquisition data.

4. Word order in the southerly texts What we need to do instead is to look at more southerly texts. As mentioned above, there is a problem here in that most of the southern texts of the twelfth century are probably copies of earlier works and so we could not trust the figures that we would get from them. However, texts which were composed in the twelfth century in the south are not completely lacking, although there are only a few. I have looked at the word order in two of these. The first is the Festis Marie, a sermon for the festival of the Virgin Mary, which was translated from Latin into English circa 1125 and is found in a manuscript which was written before 1150 (Cotton Vespasian D. xiv). The other is the Elucidarium, found in the same manuscript and translated into English at about the same time.6 This manuscript was written in a southeastern dialect which has generally been identified as Kentish7 and so offers a better comparison to the West Saxon documents of the tenth and eleventh centuries than does the Peterborough Chronicle. I've also looked at two texts from the thirteenth century to evaluate the evidence for the continued use of verb-final order in that period. Vices and Virtues is from the very beginning of the century and is written in a

8

Cynthia Allen

southeastern dialect.8 I have used only a portion of this text, namely the first 50 subordinate clauses found in the Penn-Helsinki Parsed Corpus of Middle English selections.9 The second text is the Kentish Sermons of circa 1275. The Kentish Sermons are unfortunately quite short so I have used the entire 55 subordinate clauses found in them.

4.1. Infl-final Order It turns out that when we compare verb-final order for late Old English texts with the twelfth-century text and the two thirteenth-century texts used here, the evidence suggests a gradual decrease of verb-final word order in subordinate clauses, in sharp contradiction to the claims of an abrupt halt to verb-final order in these clauses which has frequently been suggested. Before looking at the figures, we must define what we mean by verbfinal order. I will first present statistics concerning what Pintzuk (1991) refers to as "Verb Final" and "Verb Medial" word order. I have used Pintzuk's system because she produces valuable figures for Old English and our comparisons of early Middle English and Old English are only meaningful if we are using the same criteria for both periods. Pintzuk (1991: 71-124) argues for a "double base" in Old English; that is, that there were two possible underlying structures. One of these had Infi (the node for verbal inflection) at the end of the sentence and the other was Infl-medial. The surface position of the verb is in Infi, whether it is medial or final. Pintzuk assumes that the underlying order within the verb phrase was usually Object-Verb, although she argues that Verb-Object already occurred at a low rate in Old English.10 For Pintzuk, the sentences which are termed "Verb Final" are not those in which the main verb is generated in final position in the verb phrase — the case for nearly all clauses — but rather those in which Infi is generated in final position. I will therefore use Infl-final

and Infl-medial

instead of "Verb Final" and "Verb Medial". I

have used Pintzuk's criteria to class clauses as Infl-final or Infl-medial and to exclude structurally ambiguous clauses. Like Pintzuk, I have limited my corpus to clauses in which there is both a finite and a non-finite verb. So (1) is an example of the sort of sentence which I have counted as Infl-final for early Middle English:

The decline of verb-final order in early Middle English

(1)

9

pcethie naure on heuenrich wunijen ne mujen that they never in heaven dwell not may 'that they may never dwell in Heaven' (VICES 1,13.100)

Pintzuk found a good deal of variation in the use of medial and final Infi in Old English. Table 1 presents a selection of Pintzuk's figures for late Old English. The general trend is towards an increase in Infl-medial (IM), but it is not a simple progression. There is not even consistency between the works of the same author. Pintzuk found that jElfric used Infl-medial order 74 per cent of the time in subordinate clauses in his Catholic Homilies, but only 59 per cent of the time in his Lives of Saints. Clearly, there was a good deal of variation, not only between individuals but by the same individual. Table 1. Some figures from Pintzuk (1991) for late Old English Date η IM Text 912 45 BIHom 51% ChronA, Scribe 2 930 23 57% jECHom 991 43 74% 995 172 JELS 59% 49 WHom 1013 47% 64 ApT 1025 61% ChronA, Scribes 7-9 1100 10 60%

Let us compare these figures with my figures for the twelfth-century and thirteenth-century texts, presented in table 2. Table 2. Overall Infl-medial (IM) and Infl-fmal (IF) underlying order in subordinate clauses for early Middle English Date η IM IF Text Fest is Marie c. 1125 24 21==87.5% 3== 12.5% c. 1125 22 Elucidarium 17==77% 5=•23% c. 1125 Total Southern twelfth century 46 38==83% 8= 17% Vices and Virtues c. 1200 50 42==84% 8= 16% c. 1275 55 54==98% 1 == 2% Kentish Sermons

As in Old English, we find variation. The Fest is Marie and the Elucidarium were composed around the same time and are found in the same manuscript, but the Elucidarium has a considerably higher percentage of Infl-final word

10

Cynthia Allen

order than does the Festis. It is clear enough that by 1100 Infl-medial order had the upper hand even in the most conservative southern areas. By this time, Infl-medial order seems to be in the majority in subordinate clauses in every text, although there is variation in just how large the majority is. However, it appears that Infl-flnal was definitely in retreat by at least the end of the tenth century, since only one text from the eleventh century shows a percentage of less than 50 per cent for Infl-medial and no text has less than 60 per cent Infl-medial after 1025. This fact does not fit very well with the idea of a reanalysis sometime in the twelfth century. It furthermore does not seem possible to say that there was a sudden collapse of Infl-final order at any time. Looking at the last two figures from Pintzuk, it looks like Infl-medial order increased from around 60 per cent in 1100 to at least 77 per cent in the Elucidarium. This looks like a fairly large increase but when we look at Pintzuk's figure for the Catholic Homilies the increase between it and the texts from about 130 years later is small. Infl-final order was clearly on the decline in the twelfth century, but the frequency of this order in at least one twelfth-century text is only slightly less than that found 130 years earlier in the Catholic Homilies. So the most we can say is that Infl-medial word order was established as the more common one by 1100. We cannot establish a dramatic collapse at any time. It is true that there is a fairly substantial difference (14 per cent) between the Vices and Virtues and the Kentish Sermons. It is quite clear that by the time of the Kentish Sermons Infl-final order maintained only a tenuous existence. But this cannot be taken as evidence for a sudden change. The difference is in fact slightly less than that found in two works of a single Old English writer, namely ^lfric. Also, we can see by looking at the Old English figures that we must be careful not to read too much into a difference found between any two texts, such as the Vices and Virtues and the Kentish Sermons. Finally, it is important to realise that the Kentish Sermons were translations from French, and they show a good deal of the influence of French syntax. Thus any conclusions about word order based on this text must be used with caution.

The decline of verb-final order in early Middle English

11

4.2. Verb-Auxiliary order The figures we have looked at so far do not by themselves tell us whether the inflected verb actually followed the uninfected verb in surface structure or not. This is because Pintzuk included as Infl-final examples like (2): (2)

gif dcet underfangne ondgiet to ryhte tide bid if that received understanding to right time is fordbroht brought-forth 'if the received understanding is brought forth at the right time' (CP 366/11)

For Pintzuk, sentences like this must be counted as Infl-final because they have two heavy constituents before the tensed verb; this means that the noun phrase has been moved into the topic position, leaving no room for the prepositional phrase. Thus the prepositional phrase is part of the verb phrase and the Auxiliary-Verb order must be derived from an Inflfinal structure with verb-raising. Such sentences become very unusual in early Middle English, but I have found at least one example in the twelfthcentury Festis Marie: (3)

And for pan pe mcegdhad stranglice mceg widstandan And for that that maidenhood strongly can withstand pcere galnysse the licentiousness 'And because maidenhead can strongly withstand licentiousness' (Festis Marie 17.29)

This is a place where our counts depend very much on our analysis, because another possible analysis is that these sentences are Infl-medial with scrambling. So it is interesting to take a look at exactly how often we get Verb-Auxiliary surface order in these texts. Looking at table 3, we find that unambiguously Infl-final order is still found with a respectable frequency even at the beginning of the thirteenth century in the South, although by the middle of the thirteenth century it clearly has only a

12

Cynthia Allen

tenuous existence. It is clear from this table that language learners were still constructing grammars which allowed them to produce Infl-final word order for a considerable period after this word order became the minority one. So we must certainly account somehow for the continued existence of inflected verbs in final position in the twelfth and even the thirteenth century. Table 3. Verb-Auxiliary (VAux) order in subordinate clauses in early Middle English — Southern texts η Text Date AuxV VAux 24 Festis Marie c. 1125 22==92% 2== 8% c. 1125 22 Elucidarium 17==77% 5==23% c. 1125 46 39==85% 7== 15% Total Southern twelfth century Vices and Virtues c. 1200 50 44==88% 6==12% 54==98% 1 = 2% Kentish Sermons c. 1275 55

It is of interest to note that Infl-final surface order is still found in the early thirteenth century even in the Ancrene Wisse, which, as I will discuss a bit later, generally shows less verb-final syntax than more southern or eastern texts: (4)

kinge & keisers habbad hare liueneö of ower large kings and emperors have their livelihood of your large relef p je ileauet habbad relief that you left have 'kings and emperors have their livelihood out of your generous leaving that you left behind' (AW f.45b2 [Tolkien: 87, 22])

(5)

as brid hwen itfleo wule as bird when it fly will 'like a bird when it wants to fly' (AW f.36a.6 [Tolkien: 69, 29])

Thus it seems that Infl-final word order was still a possibility in the whole Midlands and southern area in the early thirteenth century, although it was definitely not very common in the West Midlands.

The decline of verb-final order in early Middle English

13

4.3. Infl-final and verb-final To summarise, the historical record does not give evidence for a sudden and dramatic collapse of Infl-final order, although it does give us some evidence for when the Infl-medial order became the dominant one. But this is not the whole picture about the shift away from verb-final order, because we must look at the position of the uninflected verb also. As mentioned above, it is generally assumed by generativists that the base position of the uninflected verb in Old English was verb-phrase-final, with complements and adjuncts preceding it. This final position is shown by the fact that objects, particles, etc. can precede the final verb in Old English even when the tensed verb is in medial position. Then, it is assumed, at some point a reanalysis or parameter resetting took place in which the basic position of the verb was verb-phrase-initial. Thus the grammar may have remained head-final in the verb phrase for a period after it became predominantly head-initial in the inflectional phrase (IP). Let us now consider some evidence concerning how this order with a final uninflected verb disappeared. It is well known that verb-final order did not disappear entirely in early Middle English. For example, Pintzuk (1991: 369) mentions the continued existence of the "brace" construction late into Middle English and argues that Object-Verb order only gradually disappeared, being ousted by a competing Verb-Object order. Example (6) is an example of this "brace" construction, as the object pet water is sandwiched between the auxiliary and main verb. (6)

ac po sergeanz wel hit wiste pet hedde pet water ibrocht but the servants well it knew that had the water brought 'but the servants who had brought the water knew it well' (KS, 216.54)

Stockwell and Minkova (1991: 397) also comment on the continued existence of the "verb brace" construction and comment that Anglicists have always emphasised the gradualness of linguistic change to their students. Van Kemenade (1987: 178) comments that Object-Verb order did not become ungrammatical as soon as the parameter was reset to generate Verb-Object order, an event which she places around 1200. Van der Wurff

14

Cynthia

Allen

(1997) and Foster and van der Wurff (1997) show that verb-final order was still found in late Middle English However, so far there has been very little discussion of just how common head-final verb phrases were in early Middle English, so I present some relevant figures in table 4. In this table, which covers only subordinate clauses, two types of clause are treated as verb-final: those in which the auxiliary verb is actually final," e.g. (1), and those which have a constituent (either a complement or an adjunct) before the non-finite verb which could not appear in that position in Modern English, e.g. (6). I have not counted as verb-final examples with an element like an adverb intervening which would also be possible in Modern English, such as (7); rather I have counted such examples as Auxiliary-Verb. (7)

swa swa we hit mugen scortlice belucan so as we it may shortly close 'as we may shortly conclude it' (Festis 29.129)

Table 4. Verb-final (VF) order (Auxiliary-X-Verb clauses Date Text c. 1125 Festis Marie Elucidarium c. 1125 c. 1125 Total Southern twelfth century Vices and Virtues c. 1200 c. 1275 Kentish Sermons

and Verb-Auxiliary) in subordinate η 24 22 46 50 55

AuxV 16==67% 10==45% 26==57% 32==64% 46==84%

VF 8==33% 12==55% 20==43% 18==36% 9==16%

Table 4 clearly demonstrates that even as late as the mid-thirteenth century, examples with a head-final verb phrase were by no means rare. It should also be noted that there are numerous examples which are not included in table 4 because they do not have an auxiliary verb but which give evidence of being verb-final because a stranded preposition occurs before the verb, rather than after it. A small number of such examples is found even in the Midland texts which only exhibit a low frequency of verb-final order, see (8). It is quite clear that verb-final order was still a very real option even well into the thirteenth century in the south. Somehow, these examples must be accounted for, either by assuming the retention of the older grammar as an option, albeit a less favoured one than

The decline of verb-final order in early Middle English

15

the newer option, or by the assumption that some other mechanism allowed language learners to produce the Object-Verb orders that they heard as they were learning English. (8)

p je cussen ... jpeo pe he ow wid beated that you kiss ... those that he you with beats 'that you kiss those that he beats you with' (AW f. 49b.28 [Tolkien: 96, 20])

5. Reconciling the claims How can we reconcile these figures to recent claims of a sudden and dramatic drop in verb-final order in subordinate clauses? Let us now look at the sort of evidence that people have relied on to support this idea. One study which is frequently quoted in support of the idea of dramatic and sudden changes to the position of the verb in subordinate clauses is Kohonen's (1978) study of word order in three pieces of religious prose: jElfric's Catholic Homilies (CH), Sawles Warde (SW) and Vices and Virtues ( W ) . The first text is from circa 1000 while the other two are from circa 1200. Kohonen (1978: 125) states that "the developments were more dramatic in dependent clauses". But what often fails to get mentioned is that Kohonen also states repeatedly that the Vices and Virtues seems to be much more archaic in this respect than the other text from the early thirteenth century which he studied, the West Midlands Sawles Warde. It is apparently Sawles Warde, and not Vices and Virtues, which the word "dramatic" applies to, since Kohonen (1978: 125) states that "VV had high proportions of medial objects ... but the tendency is nevertheless clear". And exactly how dramatic were the changes in the Sawles War del Kohonen (1978: 124) states that "[t]he development was dramatic in direct objects, where the percentage of medial objects in CH was over 60 per cent, but was reduced to some 20 per cent in SW". This is certainly a highly significant shift, but we must not forget that Kohonen was comparing texts which were at least 200 years apart. He did not look at any intermediate texts. A change of even this magnitude over a period of 200 years cannot be construed as indicating a sudden and

16

Cynthia

Allen

dramatic shift at any one time. There is a tendency for people working on historical syntax to think of 200 years as a very short time. If we use a figure of 25 years per generation, then we have about eight generations between iElfric and the writer of Sawles Warde. It seems clear that we do not really have any evidence here that any generation of language learners reset their parameters in a way that caused a dramatic drop in verb-final order. A second study which is often cited as providing evidence for a sudden decline in verb-final order is that by Hiltunen (1983). Hiltunen (1983: 111) looked at verbal prefixes and phrasal verbs in English and found that "[t]here is a steady growth of the pattern V(...)a in the 10E and eME periods, especially in the dependent clauses, which catch up with the main clauses in eME". Hiltunen's "a" stands for what he referred to as a "phrasal adverb". Examples of a(...)V and V(...)a are given in (9) 12 and (10) respectively: (9)

...pa hit up of helle geate asprong ...when it up of hell's gate sprang 'when it erupted from the entrance of hell' (Or 88.30)

(10)

On pam geare asprong up Epne fyr In that year sprang up Etna's fire 'In that year, Etna erupted' (Or 220.15)

There can be no doubt that there was a substantial increase in the latter pattern at the expense of the former by the early Middle English period, but the question is again whether there was a sharp increase at any one period. Figure 1 (Hiltunen 1983:111, figure 2) does seem to show a dramatic change. But Hiltunen states that this figure for early Middle English is based on his findings for Midland texts. It is true that he includes Vices and Virtues as a Midlands text, but the fact is that Hiltunen found only nine examples of phrasal adverbs in this text, of which three, or one third, had the old preverbal order for the phrasal adverb. In contrast, Hiltunen's corpus included 59 examples from Ancrene Riwle, in which the old order was very rare (Hiltunen found only one example of it). In fact, the majority

The decline of verb-final order in early Middle English

17

of Hiltunen's examples come from West Midland texts. Thus, although Hiltunen's figures show a very big difference between the West Saxon texts of late Old English and the West Midlands texts of early Middle English, they do not prove a sudden change in any dialect.

Figure 1. The Development of V(...)a order (after Hiltunen 1983:111) We might pause for a moment to ask why the West Midlands texts should show such a difference with the more southerly texts. A possible explanation has to do with contact with Scandinavian. Kroch and Taylor (1997) argue that there was a significant difference between Northern and Southern English in verbal syntax in Middle English, which they attribute to the northern contact with Scandinavian. There is some reason to believe that contact with Scandinavian may have played a role in the early loss of Object-Verb order in the West Midlands. First, there is evidence of substantial Scandinavian settlement in the area where our West Midlands texts were probably composed. Dobson (1976: 116, 121) points out that the substantial Scandinavian element of Ancrene Wisse presents no problem to localising the composition of the text in the West Midlands because Cnut settled a large proportion of his disbanded army there. Secondly, Weerman (1993) argues that Scandinavians learning English as a second language played a crucial role in the shift from Object-Verb to Verb-Object order, based on findings that people acquiring an Object-Verb language with verb-seconding tend to produce erroneous Verb-Object surface patterns which would have led to a shift to underlying Verb-Object order earliest in the areas with a substantial Scandinavian population. Whatever the reason behind the lower incidence of Object-Verb order in the West Midlands than in the Southeast, the differences are sufficient by

18

Cynthia Allen

themselves to show that there are problems assuming a sudden decrease in Object-Verb order using West Midland materials from early Middle English. This fact does not detract from the value of the studies which I have discussed; they have presented us with some very useful data. But the interpretation which some observers have put on these data, namely that they prove a sudden and dramatic collapse in the use of verb-final structures, cannot be maintained.

6.

Conclusions

It is clear enough that verb-final order was still robust in subordinate clauses in southern early Middle English, although decreasing in frequency. It also still existed in the Midlands although it was greatly reduced. I see no alternative to assuming that early Middle English speakers' grammars allowed for both possibilities. My findings are consistent with the idea that syntactic reanalyses do not result in the immediate replacement of one analysis by another; rather the two analyses co-exist for an extended period as one gets favoured over the other. This is certainly not a new idea and it is not incompatible with the generative approach, as shown by works such as Kroch (1989), Santorini (1989, 1992) and Pintzuk (1991, 1997), who assume that speakers have access to more than one grammar. I am not concerned here with how the grammar might allow for both possibilities — we could have competing grammars of the sort first introduced by Kroch (1989) or optionality in a single grammar. The point is that there seems to be no reason to assume that the Object-Verb orders which we find are different in nature from Verb-Object orders — that Verb-Object sentences are generated by "core grammar" and Object-Verb sentences are somehow outside the "real" grammar. Whatever analysis we adopt, we have to account for the fact that both orders are a grammatical possibility. It has frequently been suggested that old surface patterns are maintained after a fundamental grammatical change through the use of special sorts of rules, such as the "adaptive" rules of Andersen (1973) or the "peripheral" rules of Weerman (1993). But a problem with this approach is that we would be justified in expecting that a radically different sort of grammatical device would generate a rather different sort of output from that of the core grammar. 13 If so, it is incumbent upon anyone who seeks to

The decline of verb-final order in early Middle English

19

explain variation by such devices to demonstrate a difference in the sort of sentences found in the periods directly before and after a reanalysis is supposed to have taken place. But in the case under consideration here, it does not appear that there is a real difference in kind between the ObjectVerb sentences found in early Middle English, after the reanalysis of the base order is assumed by many people to have taken place, and the ObjectVerb sentences of the Old English period. It seems reasonable to expect that "adaptive" or "peripheral" rules which were devised by language learners to try to match the surface output of speakers with a fundamentally different core grammar would result in some examples of constructions not found in the earlier period. Weerman (1993: 909) explicitly mentions that the "peripheral" rule strategy for second language learners learning a verbsecond language results in some very odd output which is not grammatical for the target language, but this does not seem to be what happens in Middle English. Rather, we simply seem to get a reduced frequency of some old sentence patterns. A second conclusion that we can draw is that there is no evidence of a dramatic falling-off of Object-Verb order in subordinate clauses that was triggered by some change to the grammar. There is no evidence from the Midlands because we just do not know what the word order was like there in late Old English. Mo evidence has been presented so far from the south for a sudden collapse of verb-final word order and the evidence which has been presented here suggests a more gradual change. Of course, the texts do not necessarily reflect the spoken language very well and one could always claim that the southern texts simply fail to reflect a sudden change that took place in the spoken language. But one can claim anything about the spoken language. The texts are all we have to work with and the point is that we cannot use the supposedly sudden collapse of verb-final order in English as an argument for any theoretical viewpoint because no such sudden collapse can be demonstrated. Lightfoot (1991: 56-75) argues that the dramatic decline of Object-Verb order in subordinate clauses shows that children do not have access to evidence from subordinate clauses, since the evidence for Object-Verb order was supposedly robust in subordinate clauses at the time when children reset their parameter. But the facts presented in this paper lend no support to the idea of "degree-0 learnability", i.e. that children learning their first language have access only to main clauses and a small part of

20

Cynthia Allen

subordinate clauses. The obsolescence of verb-final order was a gradual affair in the area for which we have evidence on this question. A good deal of recent work on other constructions which have at some time been claimed to involve abrupt loss supports my belief that a closer examination of the facts will generally show that the loss of a given construction was implemented gradually. For example, Allen (1995) and Fischer and Van der Leek (1987) show that the impersonal constructions did not all disappear as soon as the grammatical innovations which were supposed to have caused their demise entered the language, in contrast with some earlier treatments such as Lightfoot (1979: 229-239) which suggested that the impersonal constructions disappeared quite suddenly. Similarly, earlier treatments of the development of the English auxiliary system (e.g. Lightfoot 1979: 98-113) assumed an abrupt change in the word class of the verbs which are now auxiliary verbs — a change from verb to auxiliary. But van Kemenade (1992) presents evidence that even within a generative model we must make provisions for certain verbs sometimes acting as auxiliaries, sometimes as main verbs, as far back as Old English, and Warner (1993) (in addition to other publications on the subject) shows how the gradual development of the auxiliary and the substantial changes which took place in the sixteenth century can be reconciled. The introduction of a new analysis is by its nature abrupt, but the evidence shows that a syntactic innovation may co-exist for a considerable period with the older syntax. I think it most probable that there is always a period of co-existence, although it may be fairly brief. The case histories of supposed sudden death which have been presented from English are seriously flawed. Unless better evidence for the sudden disappearance of a previously robust construction can be brought forward, any model which predicts that such an event is possible must be viewed as making a prediction which is not borne out by the available facts.

Appendix: Sources from which examples are cited AW

The English Text of the Ancrene Riwle: Ancrene Wisse. Edited by J.R.R. Tolkien. (Early English Text Society 249.) London: Oxford University Press [1962]. Cited by folio and line number as well as page and line number in Tolkien's edition.

The decline of verb-final order in early Middle English CP

Elucidarium

Festis Marie

KS

Or

VICES

21

King Alfred's West-Saxon Version of Gregory's Pastoral Care. Edited by Henry Sweet. (Early English Text Society 45, 50.) London: Oxford University Press [1871, 1872], Cited by page and line number. The Old English Honorius. In: Early English Homilies from the Twelfth Century MS Vesp. D. xiv. Edited by Rubie Warner. (Early English Text Society 152.) London: Oxford University Press [1917]. The texts used here are numbers XLV and XLVI of Warner's edition. Cited by page and line number. Die spätmittelenglische Übersetzung der Pseudo-Anselmschen Marien Predigt. Edited by Μ. Förster. Palaestra 148, 15-40. [1925]. Kentish Sermons. In: Selections from Early Middle English 11301250. Edited by Joseph Hall. Oxford: Clarendon Press [1920]. Examples are cited as they appear in the Penn-Helsinki Parsed Corpus of Middle English (see note 9). The Old English Orosius. Edited by Janet Bately. (Early English Text Society Supplementary Series 6.) London: Oxford University Press [1980], Cited by page and line number. Vices and Virtues. Edited by Ferdinand Holthausen. (Early English Text Society 89, 159.) London: Oxford University Press [1888 and 1920]. Examples are cited as they appear in the PennHelsinki Parsed Corpus of Middle English.

Notes 1. I do not wish to exclude the possibility that speakers make significant changes to their grammars beyond the initial acquisition stage and I think that it remains to be demonstrated that wholesale restructuring is beyond the capacity of the average adult, as suggested by Halle (1962). However, I think that there can be no doubt that the construction of a grammar by a child on the basis of the linguistic data surrounding them must be an important entry point for syntactic change. 2. Space does not allow a discussion of Lightfoot's arguments that the evidence for Object-Verb order in main clauses diminished in the early Middle English period, but for our purposes the important fact is that Lightfoot assumes that Object-Verb order disappeared quite abruptly in subordinate clauses at some point. 3. For example, Stockwell and Minkova (1991: 367-368) point out that Kellner (1892: 290) refers to a "sudden stop" in the use of verb-final order to signal a subordinate clause.

22

Cynthia Allen

4. This position is taken, for example, by Lightfoot (1979) and van Kemenade (1987). Pintzuk (1991, 1997) argues that a very small number of clauses in Old English already had a head-initial verb phrase. 5. In the Minimalist system which Roberts adopts, A G R Q is the node which in Old English contained the strong feature which checked the object for case. 6. It is somewhat unfortunate that these pieces are translations, rather than original compositions, since the possibility of influence on the syntax by the original language always exists. I have compared the English Festis Marie with the Latin original and have concluded that although the Latin has had some slight influence on the English syntax, the position of the verb in the English cannot be predicted on the basis of the position of the verb in the Latin. The translation is not a slavish one and the English often differs considerably from the Latin. 7. However, Kitson (1993: 43, footnote 116) disputes Ker's (1957: 276-277) identification and says that the manuscript "is definitely from further north in the southeast than Kent". For our purposes, a more precise location than southeastern is unnecessary. 8. Laing (1993: 106) reports that in a personal communication, Michael Samuels put this text in Essex. 9. The examples from VICES and KS are cited as they appear in the PennHelsinki Parsed Corpus of Middle English (PPCME), which was created by Anthony Kroch and his colleagues at the University of Pennsylvania. The PPCME is a somewhat expanded and syntactically annotated version of the Helsinki Corpus, which was compiled under the direction of Matti Rissanen and Ossi Ihalainen at the University of Helsinki. I gratefully acknowledge the use of these corpora. I have checked the examples used in these texts against the editions listed. 10. For a more recent summary of the arguments for the development of a headinitial verb phrase in the Old English stage, see Pintzuk (1997). 11.1 include here examples of the type Verb-Auxiliary, where Verb has no complement or adjunct. The reason is that the order Verb-X-Auxiliary is never found in Old or Middle English (see for example Pintzuk 1991: 180). As Pintzuk states, a final Auxiliary in Old English implied a head-final verb phrase. Pintzuk (1997: 235) states that "head-initial VPs within a head-final IP are not attested in any language". 12. It is unfortunate that Hiltunen included examples like (9) in his examples of a(...)V, since up could be part of the prepositional phrase in such examples, rather than an independent constituent. 13. For a discussion of the problems which would be involved in trying to deal with the Yiddish facts using adaptive rules, see Santorini (1992).

The decline of verb-final order in early Middle English

23

References Allen, Cynthia L. 1995 Case Marking and Reanalysis: Grammatical Relations from Old to Early Modern English. Oxford: Clarendon Press. 1986 Reconsidering the history of like. Journal of Linguistics 22: 375-409. Andersen, Henning 1973 Abductive and deductive change. Language 49: 765-793. Bean, Marian C. 1983 The Development of Word Order Patterns in Old English. London and Canberra: Croom Helm. Canale, William 1978 Word order change in Old English: Base reanalysis in generative grammar. PhD dissertation, McGill University, Toronto. Dobson, Eric 1976 The Origins of'Ancrene Wisse". Oxford: Clarendon Press. Fischer, Olga C.M and Frederike van der Leek 1987 A "case" for the Old English impersonals. In: Willem Koopman, Frederike van der Leek, Olga Fischer, and Roger Eaton (eds.), Explanation and Linguistic Change, 79-12. (Current Issues in Linguistic Theory 45.) Amsterdam/Philadelphia: John Benjamins. Foster, Tony and Wim van der Wurff 1997 From syntax to discourse: the function of object-verb order in late Middle English. In: Jacek Fisiak (ed.), Studies in Middle English Linguistics, 135-156. (Trends in Linguistics Studies and Monographs 103). Berlin/New York: Mouton de Gruyter. Halle, Morris 1962 Phonology in a generative grammar. Word 18: 54-72. Hiltunen, Risto 1983 The Decline of the Prefixes and the Beginnings of the Phrasal Verb in English. Turku: Turun Yliopisto. Jucker, Andreas H. 1990 Word order changes in early Middle English: Some evidence against the conservatism of subordinate clauses. Studia Anglica Posnaniensia 23: 31—42. Kayne, Richard 1994 The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press. Kellner, Leon 1892 Historical Outlines of English Syntax. London: Macmillan. van Kemenade, Ans 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris.

24

Cynthia Allen 1992

Ker, Neil R. 1957 Kiparsky, Paul 1996

Kitson, Peter 1993 Kohonen, Viljo 1978

Structural factors in the history of English modals. In: Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen, and Irma Taavitsainen (eds.), History of Englishes: New Methods and Interpretations in Historical Linguistics, 287-309. (Topics in English Linguistics 10.) Berlin/New York: Mouton de Gruyter. Catalogue of Manuscripts Clarendon Press.

Containing

Anglo-Saxon. Oxford:

The shift to head-initial VP in Germanic. In: Höskuldur Thräinsson, Samuel David Epstein, and Steve Peter (eds.), Studies in Comparative Germanic Syntax, Volume II, 140-179. Dordrecht/Boston/London: Kluwer Academic Publishers. Old English dialects and the stages of the transition to Middle English. Folia Linguistica Historica 11: 27-87.

On the Development of English Word Order in Religious Prose around 1000 and 1200 A.D.: A Quantitative Study of Word Order in Context. Turku: Abo Akademi Foundation. Kroch, Anthony S. Function and grammar in the history of English: Periphrastic 1989 "do". In: Ralph W. Fasold and Deborah Schiffrin (eds.), Language Change and Variation, 133-72. Amsterdam/Philadelphia: John Benjamins. Kroch, Anthony S. and Ann Taylor Verb movement in Old and Middle English: Dialect variation and 1997 language contact. In: Nigel Vincent and Ans van Kemenade (eds.), Parameters of Morpho-Syntactic Change, 297-325. Cambridge: Cambridge University Press. Laing, Margaret Catalogue of Sources for a Linguistic Atlas of Early Medieval 1993 English. Cambridge: D.S. Brewer. Lightfoot, David Principles of Diachronic Syntax. Cambridge: Cambridge University 1979 Press. 1991 How to Set Parameters: Arguments from Language Change. Cambridge, Mass.: MIT Press. Shifting triggers and diachronic reanalyses. In: Nigel Vincent and 1997 Ans van Kemenade (eds.), Parameters of Morpho-Syntactic Change, 253-272. Cambridge: Cambridge University Press.

The decline of verb-final order in early Middle English Mitchell, Bruce 1964 Pintzuk, Susan 1991

1997

Roberts, Ian 1997

25

Syntax and word-order in the Peterborough Chronicle 1122-1154. Neuphilologische Mitteilungen 65: 113-144. Phrase structures in competition: Variation and change in Old English word order. Ph.D. dissertation, Department of Linguistics, University of Pennsylvania. Post-verbal complements in Old English. In: Richard M. Hogg and Linda van Bergen (eds.) Historical Linguistics 1995, Volume 2, 233-246. Amsterdam/Philadelphia: Benjamins.

Directionality and word order change in the history of English, In: Nigel Vincent and Ans van Kemenade (eds.), Parameters of Morpho-Syntactic Change, 397-426. Cambridge: Cambridge University Press. Rohrbacher, Bernhard W. 1994 The Germanic VO languages and the full paradigm: A theory of V to I raising. PhD dissertation, University of Massachusetts. Santorini, Beatrice 1989 The generalization of the verb-second constraint in the history of Yiddish. PhD dissertation, University of Pennsylvania. 1992 Variation and change in Yiddish subordinate clause word order. Natural Language and Linguistic Theory 10: 595-640. Stockwell, Robert P. and Donka Minkova 1991 Subordination and word order change in the history of English. In: Dieter Kastovsky (ed.), Historical English Syntax, 357-408. (Topics in English linguistics 2.) Berlin/New York: Mouton de Gruyter. Vennemann, Theo 1984 Verb-second, verb late, and the brace construction: Comments on some papers. In: Jacek Fisiak (ed.), Historical Syntax, 627-686. (Trends in Linguistics: Studies and Monographs 23). Berlin/New York: Mouton de Gruyter. Warner, Anthony R. 1993 English Auxiliaries. (Cambridge Studies in Linguistics 66.) Cambridge: Cambridge University Press. Weerman, Fred 1993 The diachronic consequences of first and second language acquisition: The change from OV to VO. Linguistics 31: 903-931. van der Wurff, Wim 1997 Deriving object-verb order in late Middle English. Journal of Linguistics 33: 485-509.

On the history of relative that Aimo Seppänen

1.

Introduction

In descriptions of relative that the word has long been treated in two ways, classed either as a relative pronoun which is basically comparable to the w/z-pronouns, or analysed as a conjunction (complementiser) which functions here as a mere marker of subordination exactly as in ordinary content clauses. This controversy has its counterpart in discussions of the historical origin of the word, which similarly either trace it to the Old English pronoun se, seo, pcet, or view it as representing the Old English and Middle English conjunction pcet (that). Between these two positions, a third view has sometimes been suggested which sees the modern relative that as "intermediate between a relative pronoun and a conjunction" (Zandvoort 1957: 163), or as an item which functions as "a highly pronominal relativizer" (van der Auwera 1985: 171); in historical accounts a similar compromise position has been suggested which views the modern relative that as going back in some sense to both the pronoun and the conjunction. With the diversity of views presented in the current literature, it seems clear that there is room for a new examination of the historical origin of that. From the full history of the word, the present paper proposes to select for examination some central aspects, in order to determine to what extent certain somewhat baffling properties of relative that can be understood in the light of one or other of the different views of its history found in current analyses.

2.

The basic facts of relative structures in Old and Middle English

As a basis for the discussion, let us briefly review the types of relative structures that are commonly recognised in grammars of Old and Middle English. In the following sets of examples, they are represented by four different types of clauses, arranged in parallel ways for both periods of the language.

28

Aimo Seppänen

(1)

a. Adam gestrynde sunu, pone he nemde Seth Adam begat son whom he called Seth 'Adam begat a son whom he called Seth' (Gen 4.25) b. Wa dam

men

durh

pone

pe

byd mannes sunu

woe the [DAT] man [DAT] through whom that is man's

son

belcewed. betrayed 'Woe to the man through whom the son of man is betrayed' (Mt 26.24) c. scedon him da uundra 0 dyde se hcelend said them the wonders did the Saviour 'told them the miracles the Saviour did' (JnGl [Li] 11.46) d. sio scir hatte Halgoland pe he on bude the shire was-called Halogaland that he in lived 'the shire in which he lived was called Halogaland' (Or 16.1) (2)

a. She shall han right hym on whom hire herte is set (Chaucer PF 629) b. I am dwellyng with the god of thonder, which that men callen Jupiter (Chaucer HF 606) c. det oder gostliche shrud 0 ich embe spece is mildhertnesse the other spiritual garment I of speak is clemency 'the other spiritual garment I speak of is clemency' (Lamb.Hom 95.28) d. I was come ayen into the place that I of spak (Chaucer PF 296)

On the history of relative that

29

In the (a) sentences the clause is introduced by an item, derived from either the demonstrative th-stem or the interrogative w/z-stem, which is in both cases easily recognised as a pronoun by virtue of its clear morphological and/or syntactic properties. Being a noun phrase, it is an element in the functional structure of the clause, but occurs in a pre-subject position (the Spec-CP position of modern generative accounts), accompanied by a gap (or sometimes by a resumptive pronoun) in the normal position of that clause element. In the (b) sentences this initial element is followed by another word, whose grammatical status is somewhat less clear. Traditional grammarians have tended to describe it simply as "redundant" or "pleonastic" (Kivimaa 1967: 8) without a closer analysis of its status; in generative treatments it is represented as a subordinator occupying the complementiser node (Grimshaw 1975, Lightfoot 1979: 314-316, Allen 1980a: 92). In the (c) sentences we have relative clauses with no overt marking either by a pronoun or a complementiser; in grammatical terms this clause type is commonly described as containing a non-overt or zero pronoun (Chomsky and Lasnik 1977, Huddleston 1984: 397, Quirk et al. 1985: 366, 12481256). In the last examples the clause is introduced by a word which occurred as a pleonastic element or complementiser in the structure of the (b) sentences. The proper analysis of these words in the (d) sentences is a controversial issue: for the traditional grammarians the words are in this case not pleonastic but function in the same way as the pronouns of the (a) sentences; for the modernists they are here the same complementisers as in the (b) sentences, and the clauses contain a zero relativiser in the way that the (c) sentences do. 1 Table /.Types of relative clause in Old and Middle English Spec-C C pcet ipe) (a) and (b) wh(that) (c)

0

Cd,)

Pe that

(d2)

0 0

-

IP

...a... ...a... ...a...

-

...a... ...a...

pe that

...a... ...a...

-

30

Aimo Seppänen

The analysis of the different clause types distinguished is summed up in table 1, where the (d) sentences are given in two versions, and where the sign π is used to indicate the logical position of the fronted relative constituent. In the light of this survey, consider now the current accounts of the history of relative that. For the fully traditional account, such as that offered by Sweet (1891: 80, 354) and Mustanoja (1960: 188-189), the history is essentially an account of the survival of the Old English pronoun in its neuter singular nominative and accusative form, and the pleonastic use of the word that in older English plays no part in it. For the modernists, such as Grimshaw (1975) and Allen (1980a: 241), who see present-day English that as a complementiser, the history is essentially a stoiy of the replacement of the old pronoun by the complementiser. The various intermediate positions proposed are not fully worked out, but generally they see that as a descendant of the old pronoun which has somehow been influenced by the conjunction (Shearin 1903: 86, Kivimaa 1967: 4 3 ^ 4 , Mitcheiri985: 108), or they derive it from the old conjunction which has been largely pronominalised (van der Auwera 1985), or they suggest that there may have been a conflation of the two sources (Traugott 1972: 153). In a new examination of the question it will obviously be necessary to consider both the pronoun and the complementiser as in principle equally relevant to the discussion.

3.

Some new data on Old and Middle English relatives

From the preceding survey it will be clear that the weak point of the current accounts of Old and Middle English relatives is in the analysis of clauses of type (d), where the evidence presented does not suffice to determine the grammatical status of the word. The ideal solution would therefore be to find new evidence which helps to clinch the issue. A search through the existing descriptions does indeed bring to light a few examples which might offer a clue, but they are too few to be significant and have therefore never been properly incorporated into the accounts reviewed above. In this respect, the easier access to more material made possible by modern technology has created a situation where the old problem may be tackled from a partly new perspective. Approaching the question in this way it is possible to complement the picture summed up above by adding to it a further clause type. Beginning with Old English, consider the structure of the following clauses.

On the history of relative that

(3)

31

a. To pcem Icede us pe leofce Crist, pe pe is sop wisdom to that lead [SUBJ] us the dear Christ who that is true wisdom and sowie lif; pe pe ... leofcep and rixap a on ecenesse. and soul's life who that... lives and rules for ever in eternity 'May the dear Christ lead us to that, who is true wisdom and the life of the soul, who ... lives and rules for ever in eternity' OEHomM 1 [Bel 9] 151.291) b. pes mon is eower sunce, pe pe ge secgap pcet wcere soplice this man is your son who that you say that was truly blind acenned? blind born 'Is this man your son, who you say was truly born blind?' OEHom Μ 2 [Irv 3] 26.53)

The crucial point about these structures is the occurrence of the pe pe sequence to introduce the relative clause. This is similar to the se pe relatives, and must obviously be analysed in the same way, with the first pe as a pronoun and the second pe as a pleonastic complementiser. Solving a controversial issue in the analysis of pe, this discovery has implications which are relevant even for the analysis of relative pcet and will be taken up later. Apart from this, a further discussion of the pe pe relatives must be left to a separate study (Seppänen forthcoming a). The theoretical significance of the pe pe pattern will automatically raise the question whether a comparable pattern can also be found for the Middle English relatives with that. The answer to the question is in the affirmative, as is shown by examples like (4). With the sequence that that, these clauses are comparable to the older pe pe structures and the Middle English which that or who that relatives, with a fronted pronoun followed by a complementiser. The pronominal status of the first that is further evidenced by the occurrence of the piedpiped preposition before it in (4b); the preposition + pronoun structure, without a following complementiser, is attested more generally in Middle English, and affords a further unambiguous indication of the grammatical status of that. The examples in (5) offer a selection from those found; for further examples, see Ancr 10.25, Vsp.D.Hom 46.1, RomR 3041, WBible (1) Lev 15.22, RParl 5.125a.

Α im ο Seppänen a. Nimad jeme ... hwilche jife he us jefed pet pet ear us take heed what gifts he us gives who that earlier us bohte deore. bought dearly 'Note what gifts he gives us who has bought us dearly' (Lam.Horn 19.11) b. panne whoso wil go fro the lond of Galilee of pat pat I have then whoever will go from the land of Galilee of that I have spoke... men comenajainbe Damasce. spoken ... men come again by way of Damascus 'then anyone who wants to go from the land of Gallilee of which I have spoken, they travel again by way of Damascus' (Mandev [1] 81.13) c. And went and determyned her self for to fulfllle the commandement of her sayd suster Elysse, and to do all by ordre that that she had charged her for to do. (Caxton, Eneydos 87.18) a. pcet he ne do ne ne quede, ne ne denche no ping for that he not do nor not speak nor not think no thing for pcet he bie unwurdere gode. which he be unworthier of-God 'that he shall not do nor speak nor think anything for which he might be less worthy of God' (Lamb.Hom 95.25) b. Forsothe the day stood, in that Ester... shulde go (WBible [1], Esth 2) c. We byseche God Almizty that he have evremore youre roial persone in hys kepynge with encrece [of] all Manere of Honours, after that your noble and roial hert desires (Ellis, Original Letters, II. 1.66)

On the history of relative that

33

d. As for fastynge ... the days that by the chirche were appoynted she kept them diligently & ferously, & in especyall the holy lent, throughout that she restrayned her appetyte tyl one mel (Fisher, Engl.Wks 294.5) In the light of examples like these, covering the whole period from the earliest to the latest texts that one might wish to designate as Middle English, it is clear that we must accept the traditional view of the existence of that as a relative pronoun in Middle English. On the other hand, since the word that can occur in the complementiser position both with whpronouns and with that, this same usage must be recognised as a possibility even with the zero pronoun, as generally assumed in generative accounts. Applying the same reasoning to Old English, we can then represent the relative structures of earlier English as consisting of three main types, all of them found in two variant forms: Table 2. Types of relative clauses in Old and Middle English (revised) Middle English Old English wh- (that) π se (pej π that (that) π pe (pe) a 0 (that) α 0 (f>e) π This display of forms brings out two facts which are potentially relevant to the analysis of the relative clauses in Old and Middle English. The first concerns the complete parallelism of Old English pe and Middle English that in their two uses as both relative pronouns and as complementisers which can occur in all of our three types of relative clauses. This finding suggests the possibility that the grammar of Old English pe may be relevant to the grammar of the relative that which came to replace it in the later stages of the language. The second interesting point here is the finding that relative clauses overtly introduced solely by Old English pe and Middle English that were ambiguous between two structural descriptions, in that the introductory word could be either a pronoun or a complementiser. What is more, although the grammatical analysis must distinguish between the two structures, from the point of view of semantics or pragmatics there was no genuine ambiguity between them. But, since the speaker of Old English and Middle English was thus never forced by ordinary communicative needs to make a clear choice between the two grammatically

34

Aimo Seppänen

different structures, it is not to be expected that the formal differences between them were kept strictly distinct. If the Modern English relative that shows both pronominal and non-pronominal features, then one must recognise that this state of affairs may go back to the Old and Middle English structural ambiguity between two semantically equivalent structures. On the basis of these observations, our next task will be to examine how far and in what particular way the grammatical properties of the modern relative that might be traced back to the two structures of Old and Middle English, and to consider whether these properties might be understood as reflecting the grammar of Old English pe in the same two uses of the word.

4.

The pronoun pcet/that in Old and Middle English

Beginning our examination with the pronominal use of pcet/that, let us return to the examples quoted in (4) and (5) above. From the usage recorded it becomes clear that the relative that, for all its pronominal status, is invariable, showing neither gender nor case nor even number inflection. In this way it looks like a conjunction, and this lack of inflection is indeed an essential part of the motivation for the analysis which treats it in both Middle and Modern English as a complementiser rather than as a pronoun. This question thus concerns the most crucial and controversial issue in the history of the word, and needs to be considered in some detail. The process leading to the invariability of that had its origin in Old English and is well documented in current literature. In a lucid account of the data known to her, Allen (1980a: 102-105, 1980b: 274-276) makes a distinction between two different treatments of the form pcet in Old English. On the one hand, the form was used in the way of the other gender forms as an unmistakable pronoun, agreeing in number and gender with its antecedent, appearing in the case required by its syntactic function, and pied-piping its preposition (cf. example [lb] above). On the other hand, it had a use which deviated from that normal behaviour on every point mentioned: it violated the gender and number concord, occurred with a stranded preposition, and appeared in the form poet instead of taking the case normally required by the preposition. These points are illustrated by Allen with the following sentences, which are said to show violation of the

On the history of relative that

35

normal concord in gender (6a) and number (6b),2 and of the regular case government of prepositions (6c). (6)

a. Nu is se tima diet deos woruld is gemcencged mid now is the time that this world is confused with meanigfealdan mane manifold evil 'Now is the time when this world is confused with many evils' (WHom 5.24) b. hwcer hie landes hcefdon pcet hie mehten an gewician where they land had that they might on camp 'where they had land that they might camp on' (Or 46.15) c. da for he ford bi deem scrcefe dcet he oninnan wees then went he forth by the cave that he within was 'then he passed by the cave in which the man was' (CP 197.13)

Given these facts, Allen interprets the situation as implying that the word pcet in these deviant cases is grammatically distinct from the pronoun and is in fact a complementiser. This analysis is presented as providing an overall explanation of all the apparent irregularities of its grammatical behaviour, and it is this complementiser which is then said to have been generalised in the later history of the relatives and thus to have taken the place of the old pronoun. In an account of the relatives in Chaucerian English built on the same theoretical position, Grimshaw (1975) thus has no place for a relative pronoun that, the language having in this respect already reached the present-day situation. The description presented by Allen and Grimshaw is immediately appealing because of its simplicity, but on a closer examination it turns out to be beset with problems which make it difficult to accept. A first objection to it comes from an examination of the general distribution of these apparently deviant Old English forms. In her account, Allen points out that this special invariable pcet was mainly used in three cases: with neuter heads, with temporal heads, and with the antecedent eall. But given

36

AimoSeppänen

that its occurrence is in this way dependent on the antecedent, one would not expect to see it treated as a complementiser. In fully general terms, it is relative pronouns and adverbs which, being coreferential with the antecedent, may copy some features of the antecedent, such as grammatical gender, animacy (cf. the who/which contrast), or some other semantic property (cf. the place where, the day when, the reason why). By contrast, the complementisers, being mere non-referential clause links, cannot be expected to show that kind of agreement. Thus, from the usual distribution of Allen's would-be complementiser poet one is led to conclude that the word does not behave like a complementiser but like a relative pronoun. What partly obscures the picture here is the occurrence offacetwith temporal nouns, as in (6a). But the crucial fact about this use, not noted by Allen, is that in cases like these pcet functions as an adverbial and is therefore best analysed as a relative adverb: like modern when, it is selected on the basis of the meaning of the antecedent regardless of its gender, and as an adverb does not itself inflect for gender.3 Leaving the adverb out of the discussion here as not relevant to the analysis of the pronoun, we see that the invariant pcet is generally found with neuter antecedents (nouns or neuter forms of pronouns, like eall),4 and we are thus dealing here with an instance of gender concord, which clearly argues that the word is a pronoun, not a complementiser. A somewhat similar situation is recorded by Mcintosh (1947-1948) from early Middle English, where the old gender classification was reflected in a somewhat fuzzy semantic contrast between animates and inanimates, the former tending to be relativised by the, representing the older non-neuter forms se and seo, and the latter relativised more commonly by that. This usage as such is recognised by Allen (1980a: 240-241), but the two forms are treated by her as complementisers without any discussion of the theoretical implications of the distribution of the forms. A second and more concrete objection to Allen's analysis of the data derives from the fact that there are actually attested occurrences of that which go against the assumed neat distinction between a fully regular inflected pronoun and an invariant complementiser. These awkward uses fall into two different types illustrated in (7) and (8). In (7) pcet is found with a neuter singular antecedent and shows no deviation from the usual requirements of agreement or government, and thus appears to be a pronoun in spite of its stranded preposition. As pointed

On the history of relative that

37

out in note 2, the example quoted under (6b) may be a similar case, Allen's different analysis notwithstanding. The examples of (8) then illustrate the opposite situation: the form pcet must here be a pronoun because it is followed by the complementiser or preceded by a preposition, but the form of the pronoun is pcet rather than the expected masculine in (8a) or feminine in (8b). (7)

(8)

se pe pcetded pcet icymbe spece.se ded himsylfitm he who that does that I about speak he does himself mycle pearfe great harm 'He who does that which I am speaking of does himself a great deal of harm' (WHom 13.78) a. Hwa is pcet pe all da yfel... asecgean mcege who is who that all those evils tell can 'Who is there who can tell all those evils?' (Or 27.26) b. pcet is seo lufe embe pcet he witegode 'That is the love of which he prophesied' (Sol 341.32)

The type of cases illustrated above would appear to be relatively infrequent in Old English, but they become highly relevant to the historian of English when they are connected with the two structures illustrated from Middle English under (4) and (5): the that + that pattern, which does not seem to have been noted by earlier grammarians, and the preposition + that pattern, whose existence in Middle English has been explicitly denied by them (Mustanoja 1960: 189, Grimshaw 1975: 42-43, Lightfoot 1979: 320). In these patterns Middle English that is clearly a pronoun in spite of its total lack of any inflection, and in spite of its more normal occurrence with stranded prepositions. What these Middle English forms thus show is that the "deviant" Old English structures were early indications of a change affecting the morphosyntax of the pronoun pcet but not its status as a pronoun (cf. Mcintosh 1947-1848: 21, Mitchell 1985: 102, 155), and that it

38

Α im ο Seppänen

was a mistake to interpret them as indicating that the old pronoun was now replaced by a complementiser. In treating inflection and pied-piping as criterial properties of pronouns, Allen is of course merely following the trend which has dominated generative analyses of relative clauses ever since the 1970s. I have elsewhere argued that such a view does not fit the facts of present-day English (Seppänen 1997a), and I see the case at hand as an indication that that indefensible view of the pronoun/complementiser distinction necessarily leads to an altogether false picture of the most central issue in the history of relative that. The data illustrated in the examples of (3) above provide further material which clashes still more radically with generative views of these relative structures: viz. sentences with the pe pe pattern, where the first pe can only be a pronoun although it never showed any trace of inflection and never occurred with a pied-piped preposition.5 In modern historical studies within the generative framework (cf. Grimshaw 1975, Allen 1980: 224-236), it has been claimed that preposition stranding as a concomitant of a movement rule was introduced into English relative structures only with the Middle English vcA-pronouns. The traditional analysis of pe as a pronoun, in contrast, automatically implied that stranding was a fully regular part of Old English grammar with this fronted pronoun, and came to be extended to the pronoun that in Middle English. With the new data presented here, the traditional analysis is clearly the only view which can do justice to the data. In generalising the original neuter pcet as an invariable pronoun, and in changing its syntactic behaviour from pied-piping to preposition stranding, the pronoun that was adopting grammatical features associated with the older pronoun pe, which thus seems to have influenced the behaviour of its successor. The change was particular rapid in the case of preposition stranding: in Old English pied-piping was the rule with se, seo, pcet, and stranding was a rare exception limited to the form pcet; in Middle English stranding was the rule, and pied-piping a relatively infrequent exception. In this case the pace of the change was apparently influenced by the introduction of the form that as a complementiser in Middle English. Since the complementiser could not be preceded by a preposition, the formal identity of the pronoun and the complementiser apparently strengthened the tendency to abandon pied-piping with the pronominal that, and came to spread the stranding alternative even to the wÄ-pronouns when they were

On the history of relative that

39

introduced into relative structures. With the pronoun that, pied-piping as an alternative to stranding disappeared completely during the early Modern English period (cf. Ward [1765] 1967: 357 for an early grammarian's record of this), whereas the pronouns who and which retained pied-piping and stranding as alternative patterns (cf. Bergh and Seppänen, forthcoming). Only what, less commonly found in bound relatives, ultimately reached a position where its prepositional usage is almost the same as the usage with the pronoun that (Seppänen 1997: 130-131 and forthcoming b). In addition to the influence of pe and the complementiser that, a further factor has been pointed out by some scholars as having shaped the development of the pronoun that. This is the not at all unusual situation illustrated below where a dependent clause can be interpreted either as an adverbial clause of purpose or as a relative clause. (9)

Vton sendan sceaweras diet sceawion dcet land 'Let us send observers that reconnoitre the land' (Deut. 1.22)

In structures like these, it was suggested by Kock (1897: 32) that the clause was perhaps originally an adverbial clause introduced by the conjunction poet, with the subject unexpressed, but came later to be interpreted as a relative clause, with the conjunction reinterpreted as a pronominal subject of the clause. This development, further discussed by Shearin (1903: 86) and Mitchell (1985: 107), seems altogether plausible, and it would explain why the morphosyntactic changes under discussion started with the neuter form which was identical with the conjunction, and spread to the other categories only with the generalisation of the invariable that.

5.

The Middle English complementiser that

Leaving now the pronoun poet/that, let us proceed to consider the complementiser that in Middle English and its possible role in the history of the modern relative pronoun. As noted earlier, the complementiser was found with both the wA-pronouns and the pronoun that in the combinations wh- + that and that + that, and may also be assumed to combine with the zero pronoun in the sequence 0 + that. Of these forms, the repetitive that that pattern was

40

Α im ο Seppänen

rare in Middle English and is not to my knowledge attested at all in later use. The wh- + that pattern enjoyed a brief period of popularity in the fourteenth and early fifteenth centuries, but appeared then to fall into disuse and has been generally assumed to have become obsolete (Jespersen 1927: 167, Mustanoja 1960: 192, 197, Lightfoot 1979: 320-333). As regards the grammatical status of the complementiser, the historical studies of Grimshaw (1975) and Allen (1980a, 1980b) take for granted the standard generative view that the complementiser was identical with the conjunction that of content clauses. However, in later generative work, this identification of the two uses of that was seen to be problematic because of the different status of structures derived by the application of short or long movement of the zero relativiser in subject function. The relevant structures are illustrated below with examples quoted, in a strongly simplified form, from Haegeman (1991: 424-425): (10) a. the letter [ π that 0 will surprise Poirot] b. *a letter [ π that John said [that 0 would surprise Poirot ]] The crux of the matter is that, while the assumed sequence of a conjunction/complementiser that followed by a subject gap makes the sentence ungrammatical in (10b), the same sequence in (10a) is perfectly normal. The difference between the two cases corresponds with the traditional distinction between the conjunction in (10b) and the pronoun in (10a) and clearly requires a recognition of that difference in the grammar. To come to terms with this contrast, a rule was proposed by Pesetsky (1982: 306) to effect a merger of the zero relativiser and the complementiser into one element in (10a), but not in (10b), where the zero pronoun is not adjacent to the complementiser. As a result of the rule, the word that comes to have the referential index of the zero and is thus a pronoun, but it differs from the w/z-pronouns in that it stays under the Complementiser C node. The recognition that relative that is in reality not identical with the conjunction is of great interest, bringing as it does the generative approach, or at any rate one generative school, to an almost traditional view on this point. However, the way in which the case is handled — by means of what is essentially a feature-changing or category-changing rule — is theoretically awkward, and is in any case empirically inadequate because it can only

On the history of relative that

41

deal with the zero relatives, such as (10b). The wh- + that structure was not considered by the generativists because they accepted the traditional view that the wh- + that structures had become obsolete, dealt with in generative accounts by the Doubly Filled COMP filter of Chomsky and Lasnik (1977). From a recent corpus study of the question (Seppänen and Trotta 2000) it becomes clear, however, that the filter is inappropriate because the pattern is still part of English. In standard usage the wh- + that structure has a merely marginal status in bound relative clauses, but it has a more secure position in dependent interrogative clauses and especially in free relatives (for some relevant statistics, see Seppänen and Trotta 2000). In the following examples, this use is illustrated in bound and free relatives in both subject and non-subject positions. (11) a. I'd take this opportunity to answer a few of the trickier questions which that π often come my way. (CDC: today) a'. / think it's probably moved into the policy field faster than any other area of research which that I can think ofπ (CDC: ukspok) b. The hills were seared to an ugly brown. What air that π stirred was hot and muggy. (CDC: ukbooks) b'. Whatever measure that one uses a, government has grown rapidly during this century. (BNC: P. Furlong et al., Power in capitalist society, London 1986) As sentences (a) and (b) contain the sequence of that + a subject gap, they should be ungrammatical exactly as the corresponding structures with the zero pronoun, but clearly their status does not distinguish them from the non-subject cases. It is thus obvious that the view of the complementiser that as identical with the conjunction must here too be rejected as mistaken, but since the wA-pronoun and the that are clearly two distinct items, the merging rule assumed in generative analyses must here be replaced by a different approach, ideally one which can be applied to the wh- and the zero relatives alike.

42

Aimo Seppänen

To develop a different approach, let us start by noting that a subject gap after the complementiser was possible in older forms of English, as is illustrated in examples like the following, quoted here from Bergh and Seppänen (1994: 132). (12) a öcet hrcegl was geboden dcet π scolde bion geworht the robe was commanded that should be made of... twispunnenum linenum o f . . . twice-spun linen cloth 'The dress was commanded to be made of twice-spun linen cloth' (CP 75.16) b. grymbert who wolde ye that π sholde goo and daye hym to Grymbert who would you that should go and persuade him to come come 'Grimbert, who would you like to go and persuade him to come' (Caxton, Reynard 23.27) c. She hath left thee whom no man can be sure that π will not leave him (Queen Eliz, Boethius 21.38) As all of the examples quoted are ungrammatical today, we are faced here with a historical process of change. In a corpus-based study of this change, Bergh and Seppänen (1994) showed how the process had already started in Old English and was continued in Middle English so as to be virtually completed by the beginning of the early Modern English period. This development should obviously have affected the status of clauses with a wh- + that sequence and, as noted above, there was indeed in the fifteenth century such a sharp drop in the use of these constructions that the pattern has been thought to have disappeared from the language. Since it has not in reality completely disappeared, something must have happened which saved the wh- + that structures even when the that was followed by a subject gap. What was it? At this point let us return to the point made earlier of the two different structural analyses of clauses introduced by that in Middle English.

On the history of relative that

43

Table 3. Structural analyses of clauses with relative that in Middle English Spec-S C IP (i) Antecedent that a.... a.... that (ii) Antecedent 0 -

The ban on the sequence of complementiser that + subject gap was bound to rule out the second of these structures but would not affect the status of the first one. Given the availability of the structural analysis of (i) during the Middle English period when the ban affecting (ii) was being introduced, the speakers had an easy way of coping with the effect of the ban by simply reinterpreting the complementiser that in the second structure as a pronoun on the model of the first structure. Such a reinterpretation must be seen as a very natural process because the semantic equivalence of the two structures gave the speakers no motivation to keep the two formally different structures strictly distinct. The result of this assumed historical process is a structure which is identical with that represented as the result of Pesetsky's merging rule, with that as a pronoun but occurring in the complementiser slot. What distinguishes the two approaches is that the historical reinterpretation rule can equally apply even to the wh- relatives, which in Middle English occurred in the following structures: Table 4. Wh-relative in Middle English Spec-C (i) Antecedent whwh(ii) Antecedent

C that

IP a.... a

Here again the (ii)-structure should have become ungrammatical, and its attested survival even in modern use, in examples like those illustrated in (11) above, may be interpreted as an indication that a process of reinterpretation has taken place. However, while the sentence then does not contain the banned sequence of a conjunction + a subject gap, it now contains an antecedent apparently followed by two relative pronouns, and this may be the reason for its awkward status in the language. The picture becomes very different when we proceed to consider the reinterpretation and its effects in the free relative structures in which the wh- + that pattern is most commonly found today. As suggested in several modern discussions of free relatives, such as Bresnan and Grimshaw

44

Aimo Seppänen

(1978), Quirk et al. (1985: 1056-1057) and Huddleston (1988: 402-404), free relatives differ from other wA-structures in that they are strictly speaking not at all clausal but phrasal structures, with the w/z-phrase functioning as an antecedent head which is postmodified by a clause which is fundamentally analogous to a bound relative clause. This difference is particularly clear from the form of the verb in examples like the following: (13) a. [What books (that) they had left] was not clear to us. b. [What books [(that) they had left]] were all translations. If the postmodiiying clause contains the original complementiser, as indicated in the examples in (13), that word — if it is reinterpreted as a pronoun — can be very naturally integrated into the clause structure as subject, object, etc., and will in this way make the clause even more like a regular bound relative clause. This, I will suggest, is the reason why the free relatives with that retained are so much more natural and so much more frequent than the other wh- + that structures in present-day use. This difference as such is thus a strong argument for the assumed process of reinterpretation, but the free relatives offer a more convincing piece of evidence for it: structures in which that is replaced by a wÄ-pronoun exactly as in any ordinary bound relative clause. (14) a. Their idea of a family way was just literally that that (sic) the husband the wife the children whoever who was involved in that family living in a house should earn money. (CDC: ukspok) b. 'Victoria, you're a sensible girl, you go first', was an order she heard often and with pleasure, for whatever ordeal into which she had to lead the rest ... was always rewarded by authority's approval. (BNC: C. Brayfield, The prince, London 1990) Thoughout the discussion of the reinterpretation of the complementiser, we have been mainly concerned with the status of that before a subject gap, and it remains to consider the non-subject cases, such as any area of

On the history of relative that

45

research which that I can think of π (cf. [1 la']) and whatever measure that one uses α (cf. [1 lb']). Pesetsky's merging rule will automatically apply to any occurrence of the zero relative immediately followed by the complementiser, and will thus introduce a fully general distinction between the conjunction that and all instances of what has traditionally been considered the relative pronoun. Within the framework of Head-driven Phrase Structure Grammar (HPSG), it has, however, been suggested by Pollard and Sag (1994: 220-224) that present-day English has a relative pronoun that which has only a nominative form and is therefore limited to subject function, while other occurrences of that in relative clauses are instances of the complementiser. It seems to me that such a view is suspect on fully general grounds because the lack of formal distinction between the different case forms makes it possible to use the non-distinct forms in certain types of coordinated structures where one occurrence of the form represents the nominative in one clause and the accusative in the other. By way of example, consider which and that in the following sentences. (15) a. That is a sentence which/that Peter appears to dislike π but Mary feels π is perfectly all right. b. At last another date was suggested which/that Gregson said he was not too happy about π but all the others felt α suited them perfectly. Coordinations like these clearly argue that the word that must be treated as the same pronominal element regardless of its syntactic position in the relative clause, and this conclusion receives further confirmation from the observation that the wh- + that pattern produces the same effect on the status of the clause regardless of the syntactic position of the elements: awkward in bound relative clauses and more or less natural in free relatives, where that can also be replaced by a w/z-pronoun in subject as well as non-subject positions, as can be seen from (14). Altogether, the distinction must thus be made between the relative pronoun that and the conjunction that, and in fact even within the HPSG framework the pronominal status of that has been extended more recently to all instances of that in relative clauses (Sag 1997).

46 6.

AimoSeppänen Two positions for the pronoun that in Modern English

In the discussion above we first claimed that the relative pronoun inherited from Old English is still preserved in English, and then claimed that the old complementiser that has been reinterpreted as a pronoun. This implies that there are two syntactic positions for the pronoun in Modern English, and, unexpected as the claim may appear, I believe it is correct and defensible. As the availability of the C position for that is uncontroversial, we need only discuss the question of possible evidence for the Spec-CP position. First, consider the following structures. (16) a. This standard figure is called Bogey, which if you have beaten you are a good player. (Nicholson 1957:646) b. Write a list down of all the animals that if you ran over you'd have to report to the police. (CDC: npr) c. Shakespeare's mind may be likened to that modern machine into which if a thousand voices speak it will treasure up and redeliver the words (Wendt 1914: 171) The sentences represent a usage more commonly associated with somewhat earlier forms of English (cf. Jespersen 1927: 201-203), but (16a) is actually taken from a usage book where it is described as "right but unusual" (Nicholson 1957: 646), and the structure is certainly attested in modern corpora, even with that in place of the w/z-pronoun, as in (16b). In an i/-clause the conjunction occupies the C node (for a recent statement of this, see Radford 1997: 296), so that the pronoun before it, whether a weitem or that, can only be positioned under the Spec-C node. As a second example, I will quote a structure which is fully modern — so recent in fact that it is recorded only from the last hundred years and is at least so far used only by a minority of speakers. The form is the genitive that's, best attested in regional speech, but also recorded from colloquial standard English in speech and occasionally even in writing (cf. Seppänen and Kjellmer 1995, Seppänen 1997b and 1999).

On the history of relative that

47

(17) a. This is not the house that's price they were interested in n. b. There was very little in it that's meaning we thought π was fully clear. c. This is not the house in whose price they were interested. Functioning as a non-subject in (17a) and as a subject moved into the matrix clause in (17b), that's price/meaning can only occupy the same Spec-CP position which is the normal landing site of fronted w/z-elements. The two uses illustrated here are both infrequent, and although other similar structures do exist, it may still be the case that the occurrence of that outside the C position is rather marginal in present-day usage. Even as an infrequent form it is of interest, however, as the clearest instance of a survival of the Old English pronoun in modern use. Furthermore, having again acquired a genitive form, the word has at least in some varieties of the language come closer to Old English than it had been in the intervening centuries. At the same time its grammar is, however, not fully identical with that of the old se, seo, pcet because it retains even here one peculiarity which it acquired in Old and Middle English: its impossibility of occurring as the complement of a preposed preposition. Thus, while a wA-pronoun can have a preposition in front of it in our sentences, as shown in (16c) and (17c), there is no corresponding structure into that if a thousand voices speak or the house in that's price they were interested, even in this very clearly pronominal use. Being preceded by a preposition is normal for pronouns, but having lost that property in the course of its history, relative that retains its idiosyncratic restriction even when it is again inflected for case.

7.

Conclusion

At the end of a long discussion, what position is proposed here on the controversial issue of the historical origin of the present-day English relative that? The first obvious point is that those descriptions which derive the word simply from either the Old English pronoun or the Old English conjunction both contain an element of truth, but oversimplify the situation to the point of becoming unilluminating. Much the same applies even to the

48

Aimo Seppänen

views which speak of a pronoun which has been influenced by the conjunction, or of a conjunction which has been partly pronominalised; both views are partly correct but fail to capture anything like the full history of the word. The general characterisation which comes closest to the view here proposed is the suggestion made by Traugott (1972: 153) which speaks of a conflation of two different sources. As understood here, those two sources are on the one hand the Old and Middle English pronoun pcet/that, on the other hand the Middle English complementiser that. This double origin of the word is still reflected in the availability of two syntactic positions in which the word occurs. However, in the course of the complex history of the word, the two originally distinct uses have influenced each other in ways which have made them identical in both grammatical and semantic terms, although some of their shared properties may still be traced back to one or the other of the original sources; thus, the syntactic and semantic status of that as a pronoun clearly goes back to the old pronoun, and the inability of that to be the complement of a preposed preposition derives basically from the old complementiser. The picture is further complicated by other factors which seem to have shaped the grammar of the word: above all the Old English pe in its use both as a pronoun and as a complementiser, and apparently even the Old English conjunction pcet in adverbial clauses.

Notes 1. The position of the zero is here assumed to be the same Spec-CP position as with w/j-relatives. While this is the most common view, an alternative view is proposed by Grimshaw (1975) and Allen (1980a: 92, 1980b: 272-273), which assumes that the zero stays in its logical position. 2. In her account Allen categorises landes as nominative/accusative plural, and presents this as a violation of number concord. But this is surely a mistake: the plural of land was land in Old English, and the form landes is a genitive singular, used here in a partitive sense, as is particularly clear from the continuation of the sentence not quoted by her (odde wceteres pcet hie mihten him durst of adrincari). There is therefore no violation of number concord here, and the exact status of the sentences depends on whether the preposition an can govern an accusative here rather than a dative. If the answer is yes, as I suspect that it might be, then there is no violation of government here either.

On the history of relative that

49

3. It is of interest to note that in Dutch a distinction is made between the variable pronoun (masculine and feminine die, neuter dat), and the invariable adverb dat, used without regard to gender. Consider the following contrast: (i)

Dat was een dag die ik nooit zal vergeten. "That was a day I shall never forget'

(ii) Ze kwamen ons opzoeken op de dag dat Alice uit het ziekenhuis kwam 'They came to visit us on the day that Alice came out of the hospital' (I am indebted to Bas Aarts, Olga Fischer and Elly van Gelderen for help with the Dutch examples, and will discuss the English data, with some comparative material from Dutch, German and Swedish, in a separate paper now in preparation.) 4. When Allen points out the use offacetwith the antecedent eall, this remark must be restricted to the single form, as in (i): (i)

Abraham sealde eallpcet he ahte (Gen 25.1)

In the plural form there is, according to Mitchell (1985: 164), a tendency to employ the pronoun pe rather than pcet. The only point that is remarkable about this usage is the very clear dominance of pat over pe in the singular. The usage is similar to the later preference for what in all what (with parallels in German, Dutch and Swedish), which once more shows the influence of semantics on the choice of relative pronoun. 5. In actual fact, there are a few recorded instances of a combination of preposition + pe in which the complement of the preposition might possibly be the relative pronoun. For a discussion, with references to the different interpretations proposed by various scholars, see Mitchell (1985: 157-159).

References Allen, Cynthia 1980a 1980b

Topics in Diachronic English Syntax. New York/London: Garland. Deletion and movement in Old English. Linguistic Inquiry 11: 261-323. van der Auwera, Johan 1985 Relative that — a centennial dispute. Journal of Linguistics 21: 149-179.

50

Aimo Seppänen

Bergh, Gunnar and Aimo Seppänen 1994 Subject extraction in English: the use of the ίΑαί-complementiser. In: F. Fernandez, M. Fuster and J. J. Calvo (eds.), English Historical Linguistics 1992, 131-143. Amsterdam/Philadelphia: John Benjamins. forthcoming Preposition stranding with wA-relatives: A historical survey. Bresnan, Joan and Jane Grimshaw 1978 The syntax of free relatives in English. Linguistic Inquiry 9: 331-391. Chomsky, Noam and Harold Lasnik 1977 Filters and control. Linguistic Inquiry 8: 425-504. Grimshaw, Jane B. 1975 Evidence for relativisation in Chaucerian Middle English. In: Jane Grimshaw (ed.), Papers in the History and Structure of English, 35-43. (University of Massachusetts Occasional Papers in Linguistics 1.) Massachusetts: Amherst. Haegeman, Liliane 1991 Introduction to Government and Binding Theory. Oxford: Blackwell. Huddleston, Rodney 1984 Introduction to the Grammar of English. Cambridge: Cambridge University Press. 1988 English Grammar: An Outline. Cambridge: Cambridge University Press. Jespersen, Otto 1927 A Modern English Grammar on Historical Principles, Volume 3. Copenhagen: Munksgaard. Kivimaa, Kirsti 1967 t>e and j)at as Clause Connectors in early Modern English with Especial Consideration of the Emergence of Pleonastic )je. Helsinki: Societas Scientiae Fennica. Kock, Ernst Albin 1897 The English Relative Pronouns: A Critical Essay. Lund: Möller. Lightfoot, David W. 1979 Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Mcintosh, Angus 1947-1948 The relative pronouns de and dat in early Middle English. English and Germanic Studies 1: 73-87. Mitchell, Bruce 1985 Old English Syntax, Volume 2. Oxford: Clarendon. Mustanoja, Tauno F. 1960 A Middle English Syntax. Helsinki: Societe Neophilologique.

On the history of relative that

51

Nicholson, Margaret 1957 A Dictionary of American-English Usage. London: Oxford University Press. Pesetsky, David 1982 Complementiser trace phenomena and the nominative island condition. Linguistic Review 11: 297-343. Pollard, C. and I. Sag 1994 Head-driven Phrase Sructure Grammar. Chicago: Chicago University Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Radford, Andrew 1997 Syntactic Theory and the Structure of English: A Minimalist Approach. Cambridge: Cambridge University Press. Sag, Ivan 1997 English relative clauses. Journal of Linguistics 1997: 431—483. Seppänen, Aimo 1997a Relative that and prepositional complementation. English Language and Linguistics 1:111-133. 1997b The genitives of the relative pronouns in present-day English. In: Jenny Cheshire and Dieter Stein (eds.), Taming the Vernacular: From Dialect to Written Standard Language, 152-169. London/New York: Longman. 1999 Dialectal variation in English relativization. Lingua 109: 15-34. forthc.a The Old English relative pe . forthc.b The pronoun what in bound relative clauses. Seppänen, Aimo and Göran Kjellmer 1995 The dog that's leg was broken: On the genitive of the relative pronoun. English Studies 38: 3 8 9 ^ 0 0 . Seppänen, Aimo and Joe Trotta 2000 The wh- + that pattern in present-day English. In: John M. Kirk (ed.), Corpora Galore: Analyses and Techniques in Describing English, 161-176. Amsterdam/Atlanta GA: Rodopi. Shearin, Hubert Gibson 1903 The Expression of Purpose in Old English Prose. (Yale Studies in English.) New York: Henry Holt. Sweet, Henry 1891 A New English Grammar Logical and Historical. Oxford: Clarendon Press. Traugott, Elizabeth Closs 1972 A History of English Syntax. New York: Holt, Rinehart and Winston.

52

Α im ο Seppänen

Ward, William 1967 Wendt, G. 1914

An Essay on Grammar. Menston: Scolar Press. First published London: Robert Horsfield [1765], Die Syntax des heutigen Englisch. Teil 2: Die Satzlehre. Heidelberg: Carl Winter.

Zandvoort, R. W. 1957 A Handbook of English Grammar. London/New York: Longman.

The complementation of verbs of appearance by adverbs1 Nikolas Gisborne

1. Introduction In this paper, I discuss some data presented in Nevalainen (1997), where Middle English verbs of appearance — verbs like APPEAR, LOOK, SEEM, SOUND — are complemented by adverbs. There is an example in (1). (1)

he founde moche more than he said to Cirus that he shuld haue in his tresure, if he him selfe had gadred and kept it. And whan all appiered sufficiently, Cirus than said, Howe thinke you, Cresus, haue I nat tresure? (El IS/EX EDUC ELYOT 155: Heading)2

In (1), the verb of appearance, appiered, occurs with the adverb sufficiently, which is apparently its predicative complement.3 The data present two major problems: first, what are the possible complementation patterns of these verbs in Middle English (in particular, how come they can be complemented by adverbs)?; and, second, what are the routes by which the subsequent possible patterns of complementation emerge? The complementation of verbs by adverbs violates normal expectations about what verbs subcategorise for and suggests that grammatical relationships do not fall into discrete categories. Such a pattern opens up the possibility that there are continuities between adjunction and complementation, or that the word-classes adverb and adjective in the Middle English and early Modern English periods were not discrete. In this paper, I explore a construction in Modern English which demonstrates just such a continuity between adjunction and complementation,4 and which therefore offers a path by which the grammaticalisation from adverb to complement could happen. That is, I argue first that there are sufficient

54

Nikolas Gisborne

similarities between adjunction and complementation for there to be a development whereby regularly collocating adjuncts become complements, and, second, that the path from adjunct to complement in the case of these verbs involves an intermediary construction, which can still be identified in Modern English. In particular, the construction I identify as intermediary between predicative complementation and adjunction accounts for the distribution of adverbs with these verbs. I have called the intermediary construction the "attributary" construction; this is a construction which has the syntactic characteristics of complements and the semantic characteristics of adjuncts. Following Nevalainen (1997), I claim that the word-classes adverb and adjective were sufficiently discrete for it not to be possible to attribute the pattern observed in (1) to confusion between the word classes. It is the existence of the attributary construction that licences adverbs as predicative complements. The paper is divided into five main parts: section 1 is by way of introduction; in section 1.1,1 introduce the data; in section 1.2, I introduce an earlier account for the data; and in section 1.3,1 identify some analytical problems the data present. In section 2, I offer a solution to the problems raised by these data, based on an account of the same verbs in contemporary English. Section 3 returns to the problem of the word-class of the predicative complement in Middle English and early Modern English. Section 4 generalises the account to a pair of related constructions, and section 5 concludes the paper.

1.1. Nevalainen's data The account I present resolves a problem described in Nevalainen (1997). The example in (2) is typical of such examples in Chaucer; here, the adjective holwe is conjoined with the adverb sobrely, which provides clear evidence that adverbs could be predicative complements in fourteenthcentury English. In addition to the fourteenth-century data, Nevalainen has adduced further examples from the early Middle English subsections of the Helsinki Corpus. In addition to sobrely being the adverb predicative complement of looked in (2), sweetly in (3) is similarly the predicative complement of smelling. Examples (4) and (5) give further examples of adverbs as the predicative complements of verbs of appearance.

The complementation of verbs of appearance

55

(2)

As leerte was his hors as is a rake, And he was not right fat, I undertake, But looked holwe, and therto sobrely. (Gen. Prologue to C. Tales 286-288)

(3)

I warrant you Coach after Coach, letter after letter, gift after gift, smelling so sweetly; ... (E2 COME SHAKESP 45)

(4)

and there being another house pretty close to it high built with such a tower and lanthorn also, with the two churches towers and some other buildings pretty good made it appear nobly at a distance; ... (E3 TRAV FIENNES 151-2)

(5)

it can receive no Light but at the Doors and Window of the Porch, whereby it looks more solemnly; ... (E3 TRAV FRYER 1, 186)

In these cases, the adverb cannot be an adjunct of the verb because the postverbal element is compulsory: the strings that have adverbs in them are uninterpretable unless the adverb is present — or at least they have different meanings.5

1.2. Nevalainen 's account Nevalainen's account is part of a general account of adverbialisation in English. In dealing with the data in (2) to (5) she argues that the adverbs are indeed adverbs (1997: 149-152). She observes that adjectives can be formed in Middle and early Modern English with the -ly suffix, but points out that the adjectives that are formed in -ly are de-adjectival, and that they have specialised senses — in particular that they have metaphorical and moral meanings. Marchand (1960: 267) provides the evidence Nevalainen draws on. She also observes that deadjectival adjective formation of this kind is a limited process of low productivity in Middle and early Modern English. She then goes on to conclude that a distributional approach is to be

56

Nikolas Gisborne

preferred to a formal approach, because lexemes "like NOBLY and SOLEMNLY typically occur in grammatical environments requiring an adverb, but they are not used as premodifiers of nouns, for instance. On distributional grounds, they are prototypical adverbs, and should retain their word-class label even when used as predicate complements of a small number of verbs." (Nevalainen 1997: 151). It should be noted that the distribution of adverbs as the predicative complement of verbs of appearance persists to the nineteenth century, as the example in (6) shows. The examples from Quirk et al. (1985: 407-408) in (7), taken from Nevalainen (1997: 150), also demonstrate that this distribution can possibly be found in present-day English, although these examples are marginal for some native speakers. (6)

(7)

On the whole, however, things as yet looked not unfavourably for James (1849, from the Oxford English Dictionary) a. The flowers smelt sweet/sweetly b. She felt bad/badly about it

There are two things to account for: the distribution of adverbs as predicative complements and the subsequent development of, or preference for, adjectives as predicative complements. In this paper, I am primarily concerned with explaining the distribution of adverbs; however, it is possible to show some reasons for the subsequent development of the complementation of verbs of appearance. My approach is consonant with most accounts of grammaticalisation, where change is observed to be gradual and to exploit indeterminacy in the syntax.

1.3 Criteria for the analysis Given that Middle and early Modern English verbs of appearance show a course of grammaticalisation from adjunct to predicative complement, it is necessary to establish criteria for being a complement and for being an adjunct. These criteria can then be applied to objects (as highly prototypical complements), predicative complements, depictives and resultatives, and

The complementation of verbs of appearance

57

sentence adjuncts. Setting the different grammatical relations out in a matrix as in Denison (1990: 128) establishes the common elements between them. Some criteria which distinguish a complement from an adjunct are these: (I) (II) (III) (IV) (V) (VI)

whether its form is determined by its head (e.g. an object must be a noun phrase), whether it is obligatory, whether its maximum number is determined by its head,6 whether its semantic relation is determined by its head, whether it is a semantic argument of the head,7 whether it has its own (semantic) argument.8

So, for example, objects do have their form and semantic relation determined by their head; are obligatory; do have a maximum number; are semantic arguments of their head; but do not have their own semantic argument. Thus Peter in Jane kissed Peter is obligatory; it is an noun phrase; you can only have one noun phrase of this type; it is the patient of "kiss"; it fills the semantic slot of "kiss-ee". But Peter does not have an argument of its own. The form of adjuncts is not determined by their head; they are optional; their maximum number is not determined by their head; they do not fill a semantic relation of their head; they are not semantic arguments of their head; and they do take an argument of their own, typically their head. So, for example, energetically in energetically, Jane kissed Peter can be replaced by a prepositional phrase (in an energetic way)', is optional; it is recursive: energetically, enthusiastically, happily, Jane kissed Peter, it does not fill a semantic role of the verb's but rather, being an adverb expressing an attitude, determines its own semantic relation to the verb; and the verb is an argument of the adverb (i.e. the event of Jane kissing Peter is an argument of "energetically"). Between these two poles, there are other possibilities. Resultative and depictive adjuncts, as in (8), are different from sentence adjuncts in that they do not take their head as their argument. In (8), the bold type resultative adjunct (8a) or depictive adjunct (8b) takes the plain type word or phrase as its argument. In this respect, these adjuncts are like predicative complements, which also take an argument of their head as their argument: in Peter seems silly there is a predicational relationship between Peter and silly.

58

Nikolas Gisborne

(8)

a. Jane kissed Peter silly b. The waiter served the meat naked

Note also that resultative adjuncts share with predicative complements the quality of being examples of control or raising (Carrier and Randall 1992; Simpson 1983; and Wechsler 1997).9 In this respect, resultative adjuncts are the same as predicative complements; what makes them adjuncts rather than predicative complements is that they are not subcategorised for. Resultatives have a fixed form — they must be adjectives — which is a complement-like property. It is moot whether the semantic relationship is determined by the verb, or by the resultative adjunct — on the one hand, a resultative adjunct expresses a result, by definition; on the other, given that causatives are incapable of having a resultative predication (Rappaport Hovav and Levin 1998: 122), it must be the case that a verb licenses a resultative semantically. Like predicative complements, resultatives and depictives are arguments of their heads whilst having an argument of their own, which is one of the noun phrases in the sentence. Predicative complements, by comparison, have properties unlike other kinds of complement. The complements of SEEM have a range of properties which are different from the properties of objects, as the examples in (9) show. (9)

a. b. c. d. e.

Jane seems nice Jane seems a fool Jane seemed to be going Jane seemed over the moon It seemed that Jane was drunk

Examples (9a-d) suggest that the form of the predicative complement of SEEM is not fixed by the verb, except that SEEM does not permit a present participle to be its predicative complement.10 So the generalisation must be that predicative complements are freer in form than, say, objects but restrictions are still possible. The relatively free form of the predicative complements in (9) is atypical of complements. Example (9e) shows that there is only one semantic argument of SEEM and that it is a proposition. The argument of SEEM does not have a clearly defined semantic role — it certainly is not an agent, or a patient, or a beneficiary, for example. However, the predicative complement of SEEM is

The complementation of verbs of appearance

59

obligatory, it does fit a semantic slot in its head, and its maximum number is determined by its head. In these respects, predicative complements are like other kinds of complements. However, predicative complementation has one feature in common with adjunction: the complement of SEEM has its own argument. But, unlike standard adjunction, and like the argument of resultative adjuncts, the argument of a predicative complement is not the verb, but one of the arguments of the verb. The feature matrix below identifies the common features of these different kinds of relationship according to criteria (I) to (VI). Table 1. The properties of complements and adjuncts Predicative Property Object Resultative Complement Adjunct + (I) + + (II) + + (III) + + +/(IV) + + + (V) + + (VI) -

Sentence Adjunct

-

-

-

-

-

-

-

-

+

Properties (I)—(III) in the table deal with restrictions on the formal qualities of the constructions. Properties (IV)-(VI) deal with the range of argument relations. The table shows that there are a number of continuities between predicative complementation, resultatives and sentence adjuncts. These continuities most clearly establish themselves in the set of argument relations. The crucial point is that, like adjuncts, resultatives and predicative complements have arguments, whereas objects do not.

2.

Using the present-day structures to explain the past

It is possible to account for Nevalainen's data by looking at the uses of these verbs in contemporary English. Careful examination identifies the attributary construction, which is the clue to the development of these verbs in earlier varieties of English, as it is probably a remnant of an intermediate stage of grammaticalisation. An interesting feature of the attributary construction is that it is marginal between adjunction and complementation. However, it shows a different kind of in-between status from resultatives.

60

Nikolas Gisborne

There are three possible structures for these verbs: raising, control, and the attributary structure. Raising and control are different variants of straightforward predicative complementation. The attributary construction falls between straightforward predicative complementation on the one hand, and adjunction on the other. I think that syntactically the attributary structure is the same as predicative complementation, but that semantically it behaves like adjunction. The control structure has an evidential meaning where the subject has properties that provide the evidence for the evaluation as in (10)." (10) a. b. c. d. e.

he sounds foreign he looks foreign the fabric feels foreign the wine smells foreign the food tastes foreign

In these examples, the patterning of syntactic and semantic relations is similar to examples like Jane tried to go. What makes them instances of control is that there is a semantic relation between the verb and its subject. These examples can be paraphrased by examples like to judge by his sound, he is foreign. Such examples confirm that it is he in (10a) who sounds, and that these verbs have an evidential meaning. The raising structure has also an evidential meaning, but the subject is not the source of the evidence for the proposition that the predicative complement expresses. In these examples, the pattern of syntactic and semantic relations is like that in Jane seemed happy. There are examples in (11).

(11) a. (I've heard the forecast and) tomorrow's weather sounds unsettled b. (I've seen the forecast and) tomorrow's weather looks unsettled What makes these examples instances of raising is that there is no semantic relation between the verb and its subject. One problem that these verbs create for a traditional clausal approach to the raising/control distinction is that they are often ambiguous or indeterminate between the two structures. In these examples, the subject is not an argument of the verb and it is not the case that the quality of tomorrow's weather is evaluated on the basis of

The complementation of verbs of appearance

61

the sensory impression created by the weather. Instead, the sensory modality expressed by the verb identifies the means by which the speaker comes to have the information which leads to the judgement. These examples can be paraphrased as in (12). (12) a. to judge by what I've heard, tomorrow's weather will be unsettled b. to judge by what I've seen, tomorrow's weather will be unsettled Again, the to judge by phrase shows that this use encodes a speaker judgement. Extraposed subjects of verbs that have predicative complements are often taken to be good evidence of raising structures as in the examples with SEEM, LOOK and SOUND in (13). (13) a. it seems unlikely that she will ever visit now b. it looks unlikely that she will ever visit now c. it sounds unlikely that she will ever visit now The extraposition shows that the subject is not an argument of the verb. This diagnostic will be important in section 4. The final structure is the attributary one as in (14). This structure is, I suspect, a fossil of the initial stages of the grammaticalisation of this class of verbs. By recognising the attributary structure we can reconstruct the path from adjunction to raising and so account for Nevalainen's data. (14) a. b. c. d.

Peter's face looks lived in this music sounds lovely this cloth feels sticky this food smells spicy

In the attributary structure, the adjective modifies the meaning of the verb and the subject is an argument of the combination of verb and adjective. 12 Therefore the attributary construction has a rather different semantic structure from the examples in (10) and (11). In the examples in (14), the argument of the predicative complement is the verb and not the subject of the verb. That is, in (14b) it is the sound of the music that is lovely, not some other quality. It would be reasonable to say (14b) when the referent of this music was the score and not the sound of the music. In these cases,

62

Nikolas Gisborne

the semantic structure is just the same as the semantic structure in Peter ran quickly. In Peter ran quickly it is the running that is quick. In this cloth feels sticky, it is the feel of the cloth that is sticky. In the same way, Peter looks lovely is not an assessment of Peter's inherent qualities, or his character, or some other lovely aspect of Peter. It is a description of his appearance.13 The chief evidence for this position is found in the observation that attributary examples can be paraphrased by the examples in (15). (15) a. b. c. d.

Peter's face has a lived-in look the music has a lovely sound the cloth has a sticky feel Peter had a quick look at the answer book

The examples in (15) are light-verb constructions with HAVE. The examples in (15a)-(15c), which paraphrase (14a)-(14c), show that the verb and its predicative complement form a semantic unit where the predicative complement modifies the meaning of the verb. This is because they work semantically like (15d), where quick modifies look as quickly modifies look in Peter looked quickly at the answer book. The subject is then the argument of the whole semantic unit composed over the verb and the predicative complement. That is to say that the relationship between the adjective and the verb in the examples in (14) is the same as the semantic relationship between a verb and a verb phrase adverb. I have called these patterns attributary because semantically they resemble the attributive adjective + noun patterns in (15). The attributary structure differs from both raising and control in that the subject of the verb is not an argument of the predicative complement. Furthermore, the attributary uses cannot be paraphrased by a to judge... string as the examples in (16) show. (16) a. \to judge by its look, Peter's face is lived in b. \to judge by its sound, this music is lovely c. \to judge by its feel, this cloth is sticky Judgements may vary for the examples in (16), but to the extent that they are genuinely semantically anomalous, they suggest that the attributary examples are, therefore, not evaluative and do not encode a speaker judgement.

The complementation of verbs of appearance

63

Additional evidence for the attributary structure comes from factivity differences between the evidential uses and the attributary use. The raising and control uses of verbs of appearance are non-factive, as is shown by (17) and (18). (17) a. he looks ill but he's as fit as a flea a1, he looks ill and he is b. he sounds foreign but he isn't b'. he sounds foreign and he is (18) a. he looks a nice man but he isn't a', he looks a nice man and he is b. he sounds a nice man but he isn't b\ he sounds a nice man and he is The attributary structure, on the other hand, does not have a factivity value. The ability to discuss factivity shows whether there is a subordinate proposition or not. If I show you some sheet-music and utter (19), I am making a nonsensical assertion. (19)

! this music sounds lovely but it's horrible

We can account for the absence of a factivity value because, as the subject is not an argument of the predicative complement in these examples, there is no subordinate proposition. A verb only has a factivity value if one of its arguments denotes a proposition. This means that the attributary structure is different from standard predicative complementation because it does not involve a semantic argument relation between the adjective and the subject of the verb. The judgement for (19) is the same as for examples like IPeter ran quickly but he was slow. The attributary construction has the syntax of a predicative complement and the semantics of an adjunct. It constitutes a kind of mismatch relation, in that it's a grammatical relation whose syntax and semantics are out of step. It is only in the attributary construction that an adjective gets to modify a verb semantically; it is precisely this mismatch that leads to the indeterminacy that licences adverbs in this position.

64

Nikolas Gisborne

In summary, we can identify three separate uses of verbs of appearance. The attributary/evidential distinction can be decided according to factivity. The control/raising distinction is made on the basis of whether there is a semantic relation between the sense of the verb and the referent of its subject or not. The crucial distinction here is between the attributary structure and the other kinds of predicative complement. Given the semantic account assumed here, I would predict that the development from control to raising is a form of subjectification as discussed in the case of English modals in Traugott (1989). The attributary construction fits into the feature matrix of table 1 like this: Table 2. The properties of complements and adjuncts, with the attributary construction Object Attributary Resultative Sentence Property Predicative Complement Adjunct Adjunct 14 + + (I) + + + (II) + + + (III) + + +/(IV) + + + (V) + + + + (VI) -

-

-

-

-

-

-

-

-

-

-

-

As the table shows, the attributary construction forms a kind of staging post between adjunction and complementation. In terms of properties (I)-{III) it works in the same way as objects. In terms of criteria (IV)-(VI), it works just like adjuncts. As properties (I)-(III) are syntactic, we can say that the attributary construction has the syntax of adjuncts and the semantics of complements. In this way it is distinct from resultatives, which have two out of three of the syntactic properties of adjuncts (criteria [I]-[II]) and one out of three of the criteria for the semantics of complements (criterion [V]). The attributary construction is also different from predicative complementation in terms of criterion (IV). It is clear, then, that the attributary construction is midway between adjunction and predicative complementation. In the next section, I discuss how the attributary construction helps account for Nevalainen's data, and in the subsequent section, I extend the account to some other data, also involving verbs of appearance.

The complementation of verbs of appearance

65

3. The classification of the predicative complement Now I shall return to Nevalainen's data. Nevalainen (1997: 150) discusses examples like (20)-(23), repeated from (2)-(5). (20)

As leene was his hors as is a rake, And he was nat right fat, I undertake, But looked holwe, and therto sobrely. (=[2] Gen. Prologue to C. Tales 286-8)

(21)

I warrant you Coach after Coach, letter after letter, gift after gift, smelling so sweetly;... (=[3] E2 COME SHAKESP 45)

(22)

and there being another house pretty close to it high built with such a tower and lanthorn also, with the two churches towers and some other buildings pretty good made it appear nobly at a distance; ... (=[4] E3 TRAV FIENNES 151-2)

(23)

it can receive no Light but at the Doors and Window of the Porch, whereby it looks more solemnly;... (=[5] E3 TRAV FRYER 1, 186)

Recall that Nevalainen argues in favour of treating the highlighted elements as adverbs, and states that they are quite clearly some kind of predicative complement in a "weakly encoded area of grammar". Of example (20), she says that Mustanoja considers this to be a Latin-influenced literary form in Chaucer, but that Jespersen ([1927] 1961: 367-368, as in Nevalainen, 1997: 150) notes that "predicative adverbs may occur instead of adjectives with verbs of appearance. This is common with verbs such as LOOK in Shakespeare." She also says that this usage is common in the early Modern English section of the Helsinki Corpus (Kytö 1991). Further examples from the Middle English Dictionary are given in (24) and (25). (24)

Greesse-growen as α gälte, full gryslych he lukez. (Morte D'Arthur 1101)

66

(25)

Nikolas

Gisborne

there set was ...A tabernacle ...offoure pillers vp pight all ofpure gold: Like ymages were all ... Lohend full lyuely as any light angels. (Destruction of Troy 8742)

Having identified the attributary structure, it is now possible to account for these data: they are all examples of this construction. The adverbs are selected for by the verbs and are, therefore, complements, but the semantics are the same as you find in verb phrase adjunction. None of the examples in (20)-(23) could be included in a to judge by ... paraphrase, as the examples in (26) show. (26) a. \to judge by their smell, they are sweet b. \ to judge by its appearance, it was noble c. I/o judge by its appearance, it is solemn On the other hand, all of the examples could be paraphrased by adjective + noun strings, as the examples in (27) show. (27) a. they have a sweet smell b. it has a noble appearance c. it has a solemn look In each of the cases in (20)-(23), the verb and its predicative complement form a semantic unit which is then, in turn, predicated of the subject, as the examples in (26) and (27) demonstrate. Nevalainen's early Modern English examples therefore show that this construction involves an adverb as predicative complement where the semantic relationship between adverb and verb is similar to the usual pattern of adjunction, except that in this case the subject is an argument of both the verb and the adverb together, whereas in adjunction the verb, with all its arguments instantiated, is in the scope of the adjunct. The same account extends to the data in (24) and (25). Some early examples are ambiguous. In (20) above, an adjective and an adverb are conjoined as the predicative complement of looked. Assuming the attributary analysis of looked... sobrely, it is necessary to assume an attributary analysis of looked holwe. In Chaucer, as now, an adjective predicative complement of these verbs can be attributary. Furthermore,

The complementation

of verbs of appearance

there are examples like that in (28), taken from the Oxford Dictionary, where adverbs are conjoined with adjectives. (28)

67

English

And thanne cam Covetiese, I kan hym naght discryve — So hungriliche and holwe sire Heruy hym loked (Piers Plowman, B-text, Passus V, 186-187)

In this example, sire Heruy and Covetiese must be one and the same: there is nobody else present. The second line should probably be translated as 'Sir Hervey had a hungry and hollow look to him'. The example in (28) shows, then, that it was possible in the fourteenth century for adverbs to modify entities other than verbs and sentences. In the Helsinki Corpus, there are 50 examples of LOOK with a percept subject (that is, evidential or attributary LOOK). Of these, 24 have an adverb as their predicative complement, 15 an adjective. (There are further examples with clausal complements.) There are no examples of LOOK with extraposition, which would give some certainty to an analysis of raising, but there are examples which clearly instantiate some kind of evidential meaning; these typically address an evaluation of health or prosperity (as in "looks well"). Adverbs can be found as the predicative complement of LOOK in all of the Middle English and early Modern English subsections of the Helsinki Corpus·, with two exceptions, adjectives are only found in the early Modern English subsections, and they increase in number through the period. The data for the final early Modern English subsection are skewed by one particular author, Fiennes (responsible for [22] above), who uses these constructions very frequently. He is responsible for five of the examples with adjectives, and seven of those with adverbs. The numbers are too small to make any statistical predictions, but they do suggest a tendency to an increased use with adjectives in the early Modern English period, and stable usage with adverbs. Further quantitative research is necessary, but these data offer further evidence that the attributary construction antedates the evidential or raising/control construction. There is a complication to the story, however. In the nineteenth century, there is a brief flurry of verbs of appearance with adverbs as their predicative complements, but which have evidential senses.15 By the nineteenth century, adverbs have, in all other contexts, become restricted to modifying only verbs and sentences.

68

Nikolas Gisborne

(29) a. On the whole, however, things as yet looked not unfavourably for James (1849) b. It tasked all the art of Kneller to make her look tolerably on canvass (1855) c. Things had, by that time, begun to look badly for all concerned (1891) The examples in (29) are taken from the Oxford English Dictionary. Example (29b) is interpretable as an attributary example, but the other examples are clearly control or raising examples. The first example, (29a), must be evidential because things refers to matters of politics, which do not have a physical appearance, and the favourable nature of political circumstances. The final example must be evidential because ftjhings also refers to general circumstances and what is bad is the nature of those circumstances — so the example constitutes an evaluation of something which is abstract. I cannot account for this curiosity. It was a short-lived possibility and one that, I think, may have been limited to literary language. Phillipps (1970: 183-184, cited in Denison 1998: 232) assumes that the kinds of example where adverbs complement verbs of appearance as in you look very nicely indeed (Denison 1998: 232, example [391a]) is a kind of hypercorrect reaction to prescriptive teaching. I think that the primary reason for this possibility can be attributed to the general possibility of the attributary construction. It is still possible now to have an adverb as the attributary complement of a verb of appearance, given the evidence in (7). And given that the semantics of the attributary construction is more typical of the semantics of adverbs as the dependents of verbs than it is of the semantics of adjectives as the dependents of verbs, there might well be functional pressures for the retention of adverbs in this position. On the whole, however, adverbs with this distribution were becoming disfavoured, because of the increasing determinacy of the construction and the clear preference for adjectives as predicative complements. However,

The complementation of verbs of appearance

69

the attributary construction remains to the present day; if it accounts for the earlier distribution of adverbs in this position, there can be no clear grammatical reason why adverbs should not, for a conservative speaker, still be available. The story of how adverbs become dispreferred is a story of sociohistorical linguistics — of the transmission of change. The primary point of this paper is to explain how the grammatical environment made a certain distribution of adverbs possible, and how that environment is set up in such a way that it predisposes the verbs concerned to establish a set of standard predicative complements in their grammar.

4.

Further data

The attributary analysis extends to two further sets of data. One is a case study reported in Bender and Flickinger (1999), where they discuss how verbs of appearance are complemented by clauses headed by as if and like. The other concerns data like the example in (30). (30)

Whereby it appereth clerely, that there is but one substaunce of the same good, and ofblessednes or felicitie. (El XX PHILO BOETHCO 77:Heading)

In (30), the adverb clerely occurs after the verb in a position typical of a predicative complement as in it appears clear that there is but one substance of the same good. In the Helsinki Corpus, adverbs relevant to sensory experience, such as clerely in (30), often occur, but not in a postverbal position. I have found six examples of APPEAR like (30). In these examples the adverbs — those like plainly, evidently, and clearly — are ambiguous: they are either relevant to visual appearance or are sentenceadverbs which have a speaker-oriented attitude interpretation. The example in (30) involves a verb which has an extraposed clause which is co-referential with the empty pronoun subject of the verb. This example is important because clerely is in a position which is incompatible with sentence adverbs and canonical for predicative adjectives. However the extraposition shows that this is a raising construction. The issue, then, is whether this is a case of a sentence adverb — which would have to be marked out by intonation — or another predicative adverb. If it is a

70

Nikolas

Gisborne

sentence adverb, there is no problem. If it is a predicative adverb, then in these early Modern English data the adverb form was encroaching on the raising environment. This option could give rise to the ambiguity that makes examples like those in (29) possible. Bender and Flickinger (1999) investigate examples like (31), their (24b) and (28). These examples consist of cases where a verb of appearance is complemented by a subordinate clause headed by as i f , as though, or like. In the following discussion, I refer to all these three kinds of clausal complement as as if clauses. They claim that "a historical change took place sometime in the 17^ century, where adjuncts introduced by as if were reanalyzed as complements for certain classes of verbs" (Bender and Flickinger 1999: 209). I think that this same reanalysis can be accounted for by treating at least the initial examples of as //clauses as occurrences of the attributary structure. (31) a. The Fellow looks as if he were broke out of Bedlam. (G. Farquhar, The Beaux Strategem 1707) b. Does not all this look as if some unseen power who guides our actions had set a... (F.C. Sheridan, Memoirs of Miss Sidney Biddulph 1761) The reason why these examples are important is that the as //complement clauses have the distribution of adverbial clauses in all other contexts. Bender and Flickinger claim that the semantics of as if clauses are compatible with them becoming predicative complements because they are two-place predicates which mediate a relationship between their dependent clause and their host clause. This, however, is no reason to assume that the change from adjunct to complement is straightforward. As we have seen, present participles have a semantic structure which is compatible with their being predicative complements, but there are verbs like SEEM which no longer select present participles to be their predicative complements. There is, therefore, no reason to assume that a word which has compatible semantics will become a predicative complement. Bender and Flickinger argue that the development of as if involves the creation of a new lexeme for each of the verbs of appearance that takes the as //clause as a complement, in an account which also requires them to

The complementation of verbs of appearance

71

assume that a verb can select for adjuncts. The idea that a verb can select for adjuncts is fairly similar to the attributary analysis in this paper, but the attributary analysis offers a clearer relationship between adjunction and predicative complementation. It is certainly not incompatible with the main thrust of Bender and Flickinger's account. The attributary analysis suggests a path by which a clausal adjunct could be reanalysed as the complement of a verb of appearance. On this analysis, an as //clause can be the predicative complement of a verb of appearance because its occurrence in the attributary structure permits a measure of ambiguity: it is unclear whether the as if clause is an adjunct or a complement. In due course, the as //clause is reanalysed as a complement and, rather than taking its head as its semantic argument, which is the usual state of affairs with adjuncts and attributaries, it takes the subject of its head as its argument, thus becoming a more usual kind of predicative complement. This approach generalises across different kinds of adjuncts which become complements and has the virtue of improving on Bender and Flickinger in that a construction type that is motivated elsewhere is exploited — rather than assuming that an adjunct can be added to the list of words a verb subcategorises for, with some other argument eventually being removed, and therefore extending the argument-taking properties of that verb for a short period of its history.

5.

Conclusions and prospects

In this paper I have provided an account of the Middle English and early Modern English distribution of adverbs as predicative complements that Nevalainen (1997: 150) noted. Providing a reason for this distribution entailed an analysis of verbs of appearance in contemporary English in order to identify the attributary construction, which is responsible for licensing adverbs as predicative complements. The attributary construction has the syntax of a predicative complement, and the semantics of an adjunct, which combination allows it to provide a bridge between adjunction and predicative complementation. It is a kind of "mismatch" grammatical relation. The fact that the attributary construction shares its syntax with predicative complementation provides a measure of indeterminacy

72

Nikolas Gisborne

between this construction and both raising and control. It is this indeterminacy, I think, which permits the examples seen in (29); the same indeterminacy accounts for the data that Bender and Flickinger (1999) introduce. This paper provides an initial account of the grammar of these examples. There is a need for further work to establish the diachronic progress of these constructions. For example, given Traugott's (1989: 34-35) Tendencies 1 and 3 , 1 would expect control structures to emerge out of the attributary construction and then for raising structures to follow. However, this proposal remains a hypothesis for further investigation.

Notes 1. Thanks to an anonymous reviewer, the editors, Dick Hudson, Jasper Holmes, Steve Matthews, John Stonham and audiences at Helsinki University and the Tenth International Conference on English Historical Linguistics for help with this paper. The University of Hong Kong provided financial support. 2. The references to examples taken from the Helsinki Corpus follow the practice of Kytö (1991). 3. I am using the term "predicative complement" as a cover term for both "raising" and "control" predicates. It is a syntactic relation, of precisely the same kind as LFG's "xcomp" (Bresnan 1982: 287). In this paper I do not really address the differences between raising and control structures, so I wish to adopt a single term to cover both possibilities. I have preferred the term "predicative complement" because it follows the practice of Denison (1998: 228-238), whilst specifiying that the predicative element under consideration is a complement and not an adjunct. 4. In this paper, the class of adjuncts that I am discussing is verb phrase adverbs. I retain the terminology "adjunct" and "adjunction" because adjunction as a phenomenon extends beyond the domain of verb phrase adverbs in a fashion which is grammatically uniform: verb phrase adverbs and attributive adjectives display aspects of their grammar in common. This choice maintains terminological continuity with Bender and Flickinger (1999) discussed in section 4. 5. An anonymous reviewer wonders if this statement is true for (4). I take it that (4) means that the building has a noble appearance, and that if nobly were a speaker-attitude adverb it would appear either at the beginning or end of the clause it modifies, not verb-phrase-internally. For discussion of position within the clause as a diagnostic of adverb type, see Palander-Collin (1997: 371-372).

The complementation of verbs of appearance

13

6. You can only have one of each type of complement, whereas adjuncts are typically recursive. However, recursion of adjuncts is subject to various constraints. So, for example, it is only possible to have one causal adjunct. In he left because he was drunk because the punch was spiked the second because-clause modifies the first: it is not the case that they both modify he left. 7. It becomes clear that (IV) and (V) are distinct criteria when you look at indirect objects, which are semantic arguments of ditransitive verbs, but which fix their own semantic relation in that they are always "beneficiary" or "recipient". 8. Argument-taking is a property of adjuncts. Parsons (1990: 13-15, 40-67) discusses this motivation in a general account of event semantics. 9. For a note on control and raising see Denison (1993: 166-168) 10.Denison (1998: 231) shows that it was possible for SEEM to have a present participle as recently as 1945 as in his example (389e) fejveryone seemed milling around, banging into furniture. It would appear from this that the restriction preventing present participles from being predicative complements is a recent one. 11.1 assume that these structures are syntactically the same, and that they are distinguished semantically. This is similar to Langacker's (1995) position on raising and control, and it is supported by the entirely semantic nature of the diagnostics employed by Radford (1988: 317-324, 435-^36). 12. Anderson (1971: 148) sketches the lineaments of an account similar to this for SMELL when it occurs with a predicative complement. 13.Tony Kroch (personal communication) has suggested that this construction constitutes a subdomain of control because it seemed to involve adjectives which expressed qualities only accessible through the sensory modalities denoted by the verbs concerned. However, this is not true of the examples in (14a)-(14b), and there is further evidence from factivity below which also shows that these examples are unlike both raising and control. 14. The form of an attributary is essentially determined by its head in that in contemporary English it has to be an adjective. However, the fact of the adverb distribution in this position suggests that, at least when there was free variation between adverbs and adjectives in this construction, the verb was not responsible for the form of the attributary. 15. David Denison, Teresa Fanego and Christain Mair have all made helpful suggestions about this "flurry", and Nigel Vincent has kindly provided yet more examples. One likely suggestion (David Denison's) is that these examples, which are all literary, were influenced by Jane Austen's literarylinguistic conservatism.

74

Nikolas Gisborne

References Anderson, J.M. 1971

Towards a Grammar of Case. Cambridge: Cambridge University Press. Bender, Emily and Dan Flickinger 1999 Diachronic evidence for extended argument structure. In: Gosse Bouma, Erhard Hinrichs, Geert-Jan M. Kruiff and Richard T. Oerhle (eds.), Constraints and Resources in Natural Language Syntax and Semantics, 3-19. Stanford, CA: Center for the Study of Language and Information. Bresnan, Joan 1982 Control and complementation. In: Joan Bresnan (ed.), The Mental Representation of Grammatical Relations, 282-390. Cambridge MA: MIT Press. Carrier, Jill, and Janet Randall 1992 The argument structure and syntactic structure of resultatives. Linguistic Inquiry 23:173-234. Denison, David 1990 The Old English impersonals revived. In: Sylvia Adamson, Vivien Law, Nigel Vincent and Susan Wright (eds.), Papers from the Fifth International Conference on English Historical Linguistics, 111-140. Amsterdam/Philadelphia: John Benjamins. 1993 English Historical Syntax. London/New York: Longman. 1998 Syntax. In: Suzanne Romaine (ed.), The Cambridge History of the English Language, Volume 4: 1776-1997, 92-329. Cambridge: Cambridge University Press. Jespersen, Otto 1961 A Modern English Grammar on Historical Principles, Part 3. London: Allen and Unwin. First published Heidelberg: Carl Winters Universitätsbuchhandlung [1927]. Kytö, Merja 1991 Manual to the Diachronic Part of the Helsinki Corpus of English Texts: Coding Conventions and Lists of Source Texts. Helsinki: Department of English, University of Helsinki. Langacker, Ronald W. 1995 Raising and transparency. Language 71: 1-62. Marchand, Hans 1960 The Categories and Types of Present-Day English WordFormation. Wiesbaden: Harrassowitz. Middle English Dictionary 1952Hans Kurath, Sherman M. Kuhn, John Reidy, Robert E. Lewis (eds.). Ann Arbor: University of Michigan Press.

The complementation of verbs of appearance

75

Nevalainen, Terttu 1997 The process of adverb derivation in late Middle and early Modern English. In: Matti Rissanen, Merja Kytö, and Kirsi Heikkonen (eds.), Grammaticalization at Work, 145-189. Berlin/New York: Mouton de Gruyter. Oxford English Dictionary 1989 Second edition. Prepared by J.A. Simpson and E.S.C. Weiner. Oxford: Clarendon Press. Palander-Collin, Minna 1997 A medieval case of grammaticalization, methinks. In: Matti Rissanen, Meija Kytö, and Kirsi Heikkonen (eds.), Grammaticalkation at Work, 371-403. Berlin/New York: Mouton de Gruyter. Parsons, Terence 1990 Events in the Semantics of English. Cambridge, MA: Massachusetts Institute of Technology Press. Phillipps, K.A. 1970 Jane Austen's English. London: Andrd Deutsch. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Radford, Andrew 1988 Transformational Grammar. Cambridge: Cambridge University Press. Rappaport Hovav, Malka and Beth Levin 1998 Building verb meanings. In: Miriam Butt and Wilhelm Geudar, (eds.), The Projection of Arguments, 97-134. Stanford, CA: CSLI Publications. Simpson, Jane 1983 Resultatives. In: L. Levin, M. Rappaport and A. Zaenen (eds.), Papers in Lexical-Functional Grammar, 143-158. Bloomington, Indiana: Indiana University Linguistics Club. Traugott, Elizabeth Closs 1989 On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65: 31-55. Wechsler, Stephen 1997 Resultative predicates and control. In: R. Blight and M. Moosally (eds.), The Syntax and Semantics of Predication, 307-321. (Texas Linguistic Forum 38.) Austin, Texas: University of Texas Department of Linguistics.

On the use of current intuition as a bias in historical linguistics: The case of the LOOK + -ly construction in English1 Kristin Killie

1. Introduction: subject matter and organisation In his recent monograph, Roger Lass (1997: chapter 2) provides a discussion of historical language data and their interpretation. Among other things, he draws attention to rhymes of the type Robin : sobbing and ruin : doing, which are quite commonly used in earlier English. These have been claimed to be false rhymes on the assumption that their phonetic representations are /-in/ and /-iq/, respectively. However, there is ample evidence that from the sixteenth century and well into the nineteenth unstressed -ing was commonly pronounced /in/ in more colloquial styles; hence the characterisation "false rhyme" is probably incorrect. This erroneous interpretation is, according to Lass (1997: 75), due to "an ignorant editor's projection from later norms". The present paper discusses another case where the same kind of misinterpretation may have taken place, i.e. structures of the type you look sadly. More precisely, it draws attention to collocations of copular LOOK with a -ly element, where the characteristic denoted by the -ly form is the result of a concrete visual impression of the appearance — i.e. facial expression, posture or other characteristic — of some person or object. The main point to be addressed is the status of the -ly derivations that occur in this type of collocational pattern. It has been claimed that they are instances of overprocessing by the adverbial component, i.e. they are adverbs and anomalies. However, on the basis of a fairly large corpus I will claim that the LOOK + -ly construction is in fact a regular feature of the grammar at some stage, and that they are not necessarily adverbs either, but may be adjectives formed on the basis of an Old English semantic distinction. The structure of this paper is as follows. Section 2 briefly recounts the few comments on the pattern that I have been able to find. Section 3

78

Kristin

Killie

presents the corpus data, while section 4 discusses the status of the -ly element in some more detail. Finally, section 5 discusses the implications of studies such as this to the discipline.

2. Views expressed in the literature To my knowledge, the structure in question has not been given much attention in the literature, but has been discussed in a rather cursory manner. In addition to my own (1993) excursus on this topic (see note 1), I have found seven brief comments: Heuer's (1932) comment on Chaucer's usage, Mustanoja (1960) on Middle English, Jespersen ([1946] 1961) on Shakespeare, Nevalainen (1997) on late Middle and early Modern English, Western (1906) on nineteenth-century English, Phillipps (1970) on Jane Austen's English and The Oxford English Dictionary (second edition) on the whole history (cf. also Gisborne this volume). None of these provide any quantitative data, but they nevertheless make claims about the status of the -ly element, as shown in the following. Heuer (1932: 107-108) regards Chaucer's use of -ly forms in collocation with copular LOOK as "illegitimate" adverbs. According to him, the fact that the verb LOOK in addition to its copular meaning has a more dynamic one (as in look at) may have caused a certain functional indeterminacy ("Beziehungsunsicherheit"). This, he conjectures, may have induced the use of the adverb even when the copular meaning of the verb is intended and we would thus expect an adjective. In practice, this means that LOOK + -ly clauses may have been formed by analogy with structures such as that in (1) below: (1)

He looked (at her) wearily/merrily (at her).

Mustanoja (1960: 314-316, 648-650) regards LOOK + -ly constructions as ungrammatical in Middle English. He suggests that LOOK + -ly constructions are the result of a certain formal as well as functional indeterminacy as regards adjectives and adverbs. The formal indeterminacy in question is assumed to have been caused by the well known levelling process, which probably started in late Old English (cf. for example Pyles and Algeo 1982: 152-153). This process, which eventually did away with a

The case of the LOOK +

-ly

construction in English

79

substantial part of the unstressed syllables in English, radically changing the morphological and syntactic structure of the language, also left the adverbial -e suffix unpronounced. As this suffix was what typically distinguished adjectives and adverbs, the levelling process rendered the two parts of speech formally indistinguishable, thus Old English -lie and -lice both ended up as -lie, later to become -ly? According to Mustanoja (1960: 315), the shared form of adverbs and adjectives caused a certain confusion with respect to their use: Owing largely to the loss of final -e and the resulting assimilation in form ..., adjectives are often used in adverbial function. The matter is further complicated by the existence of certain functional areas where the choice between an adjective and an adverb is a matter of subtle and elusive differences in mental attitude. Thus a modal adverb may be found in an expression where modern usage would prefer an appositive or predicative adjective... 3

According to Mustanoja, the formal and functional indeterminacy described included the use of copular look with what he takes to be a "modal" (i.e. manner) adverb. Jespersen (1961: 416) notes that the LOOK + -ly pattern exists in Shakespeare and he seems to take it for granted that the -ly element must be an adverb. Unlike Heuer and Mustanoja, however, he does not really discuss the legitimacy of the construction. Nevalainen (1997: 150-151), discussing late Middle and early Modern English, considers the possibility that the -ly forms under discussion may be adjectives. According to her, this is in principle possible since adjectives could at this stage be derived from other adjectives by means of the -ly suffix, examples being lowly, meetly and sickly from Middle English and weakly and poorly from early Modern English. However, Nevalainen rejects the adjective hypothesis for two reasons. Firstly, the process of deriving deadjectival adjectives by way of -ly was much more limited in Middle and early Modern English than in Old English. Secondly, "most deadjectival -ly adjectives have specialised senses — metaphorical and moral meanings attributed by Guimier (1985: 164) to the 'virtual character' of the adjective suffix" (Nevalainen 1997: 151). Western (1906: 92), like Jespersen, simply refers to the -ly element as

80

Kristin

Killie

an adverb without further discussion. Strangely enough, he appears to find the LOOK + -ly construction perfectly acceptable in nineteenth-century English, if the absence of a comment to the opposite effect can be interpreted that way. Phillipps (1970: 183-184) notes that in Jane Austen's language, LOOK is found with "adverbs". Again, the construction is viewed as ungrammatical, and yet again its use is explained in terms of a functional indeterminacy between adverbs and adjectives. The occurrence of the construction is ascribed to the influence of the grammarians, who are said to have created "a misconceived notion of the right use of predicative adjectives", with the result that "[a]djectives were felt to be incorrect even when they were warranted" (Phillipps 1970: 183). The Oxford English Dictionary devotes a separate section to the LOOK + -ly pattern (under the verb look, 9.b.), providing a fairly large number of examples. With respect to the status of the -ly word, the dictionary provides the following comment (where the relevant passages are given emphasis by me): 9. a. intr. To have the appearance of being; to seem to the sight, b. with adv. of manner (tor advb. phrase): To have a certain look or appearance. ... In some earl^ instances the apparent adv. may possibly be an adj. in -LY1. It is interesting to note that the lexicographers do not preclude the possibility that the -ly element may in some cases be an adjective. I will come back to this view in section 4 below, but before that I will present the result of my empirical study.

3. The grammatically of the -ly form 3.1 Investigating the

LOOK + - l y

construction: database and method

The corpus consists of several components: a collection of sentences from The Oxford English Dictionary, a Chaucer corpus, a Shakespeare corpus and a mixed corpus. The latter consists of a Middle English, an early Modern English and an eighteenth-century corpus and comprises a number of texts taken from the Internet (mostly prose) as well as the Middle and

The case of the LOOK + -ly construction

in English

81

early Modern English sections of the Helsinki Corpus of English texts.5 (For more information on the specific texts used, see the Appendix). In a corpus each subpart should of course ideally consist of approximately the same number of words, authors and texts or extracts. This rule is not strictly observed in the present study. This is because I wished to include the Helsinki Corpus, which does not cover the eighteenth century. If we disregard the Helsinki Corpus, the early Modern and eighteenth-century corpora are balanced with respect to the number of writers represented, i.e. ten. By keeping the Helsinki figures apart from the rest of the data, a certain measure of comparability is obtained. The number of texts is not the same in the two corpora. As I wanted to use whole texts rather than extracts, and as the length of texts varies, some authors are represented by two texts; hence the early Modern English corpus comprises fourteen texts and the eighteenth-century corpus only ten. However, if we include the Helsinki material, the two corpora are roughly of the same size, i.e. approximately one million words.6 (Shakespeare's work is kept completely outside the rest of the corpus, due to its large size.) The Middle English corpus is, however, larger (approximately 1.5 million words, Chaucer excepted), simply because copular LOOK is so scarce in this period. The Chaucer Corpus consists of The Canterbury Tales, Troilus and Criseyde and A Treatise on the Astrolabe, and the Shakespeare Corpus comprises his complete works. There are two main reasons why I have included so much text by Chaucer and Shakespeare. Firstly, it is always interesting to study how the language of these writers relates to that of their contemporaries, as in linguistic studies it has been quite common to derive grand generalisations about Middle and early Modern English on the basis of their work. Thus, the language of Chaucer and Shakespeare is commonly taken to be representative of the English of their time. This is a hypothesis I wished to test in relation to the LOOK + -ly pattern. Secondly, I wanted to check the truth of Heuer's (1932: 107-108) claim that LOOK + -ly is really to be considered an anomaly in Chaucer's work. The method applied is the following. Using a concordance program, I scanned the various texts, selecting all instances of copular LOOK (as well as APPEAR and SEEM; I will return to this in section 4) + modifier, i.e. including -ly forms as well as plain forms. I checked all the variant spellings listed in The Oxford English Dictionary, including all tenses and aspects.

82

Kristin Killie

Having done this, I checked the Oxford English Dictionary entries for all the -ly forms that collocate with copular LOOK in the corpus, as well as those provided under the entry for LOOK in the Oxford English Dictionary. This was done in order to find out (i) if the Dictionary supplies more attested uses of these forms with LOOK (APPEAR and SEEM) than those listed under LOOK, and (ii) if the -ly forms in question are shown, in addition to occurring with copular LOOK, to have had other prototypical adjective uses. The examples of LOOK + -ly provided by The Oxford English Dictionary were also made part of the database, providing evidence of the relatively wide currency of the pattern; however, these examples could naturally not be used in determining the relative distribution of -ly and plain forms, and were hence kept separate from the corpus specially compiled for this study. Excluded from the data were comparatives and superlatives, well, and adjectives ending in -ly, which are known normally not to allow -ly derivation for phonotactic reasons.

3.2 The data Table 1 below shows the distribution of the LOOK + -ly pattern in the corpus. The figures include the examples from the corpus as well as those supplied by The Oxford English Dictionary ? Table 1. Number of suffixed modifiers in the corpus Chaucer Shakespeare Other corpus Helsinki 1 6 3 Middle English Early Modern English Eighteenth century Total

27

6

-

-

-

7

6

27

OED Total 2 12

7

23

63

10

7

17

20

32

92

The case of the LOOK + -ly construction

in English

83

As shown in the table, there are 92 examples of the LOOK + -ly pattern in the material. Only 12 occurrences are from Middle English. Six of these are found in Chaucer,8 one in the Helsinki Corpus, and two in the Oxford English Dictionary. Scanning the 1.5 million-word Middle English corpus compiled for this study yielded only three occurrences. On the basis of these figures, it seems reasonable to conclude that the pattern was not particularly widespread in the Middle English period. In view of the fact that LOOK began to be used as a copula only late in the Middle English period (see note 5), this is hardly surprising. The figures may indicate that Chaucer is a pioneer in his use of the LOOK + -ly pattern; however, we have to be careful in drawing any conclusions to this effect since this poet is over-represented compared to the other Middle English writers. Nevertheless, the data seem to suggest that LOOK + -ly was not an anomaly in Chaucer's language, as was claimed by Heuer. The figures from the early Modern English period contrast sharply with those from Middle English. There are as many as 63 LOOK + -ly sentences in the early Modern English material. Of course, the Shakespeare figures skew the data considerably, the works of this writer containing as many as 27 occurrences; however, there are also 36 examples from writers other than Shakespeare. Of these, 23 are found in the Oxford English Dictionary, six in the Helsinki Corpus and seven in the body of texts compiled for the purposes of this study. Hence the LOOK + -ly pattern appears to be a fairly regular feature of early Modern English, and particularly of Shakespeare's language. With respect to the eighteenth-century figures, these are fairly high. There were seven instances of LOOK + -ly in the Oxford English Dictionary and ten in the other corpus. The latter figure is actually higher than the corresponding figure for early Modern English. This could be due both to the increased use of copular LOOK in this century ( c f . the discussion of relative distribution below) and to the larger size of the eighteenth-century corpus, corresponding to the size of both the Helsinki material and the specially composed early Modern English corpus. To arrive at a more complete picture of the use of LOOK + -ly, however, it is necessary to study the relative distribution of -ly and bare forms. Table 2 below is a survey of the individual usage of the writers in the corpus. The Chaucer and Shakespeare figures are kept separate in order to prevent them from unduly influencing the totals. As for the Middle English corpus, a

84

Kristin Killie

large majority of the writers do not have the LOOK + -ly pattern, nor copular LOOK with a plain adjective. For more information about these writers and works, the reader is referred to the Appendix. For the Helsinki Corpus I only provide the totals. As mentioned in section 3.1, the Oxford English Dictionary examples are kept out of the discussion of relative distribution. Table 2. Relative distribution of -ly and bare forms in the corpus texts Middle English -ly non -ly (i) Book of the knight 3 1 (ii) Helsinki Corpus 1 2 Total (i)-(ii) 4 3 Chaucer 4 5 Total Middle English 8 8 Early Modern English (iii) Ascham 0 0 (iv) Bacon 0 0 (v) Behn 4 1 (vi) Cavendish, George 1 2 (vii) Cavendish Margaret 0 0 (viii) Chettle 0 0 (ix) Elyot 0 0 (x) Helsinki Corpus 6 1 (xi) Jonson 2 0 (xii) Middleton 1 3 (xiii) More 1 0 Total (iii)-(xiii) 13 9 Shakespeare 27 93 Total early Modern English 40 102 Eighteenth century (xiv) Brooke 0 1 (xv) Burney 1 6 (xvi) Defoe 1 11 (xvii) Fielding 1 12 (xviii) Gay 2 0 (xix) Goldsmith 0 2 (xx)Schaw 1 4 (xxi) Sheridan 4 38 (xxii) Woolman 0 0 (xxiii) Wollstonecraft 0 0 Total eighteenth century (xiv)-(xxiii) 10 74

The case of the LOOK + -ly construction in English

85

The figures are highly interesting, as they reveal individual differences that would not be visible if the individual rates had simply been conflated. It appears that the form of the complement used with copular LOOK is highly variable in all the periods, though the Middle English data are too scanty to supply a sufficient basis for discussion. With regard to the Middle English period, it is interesting to note that -ly forms and plain forms occur at approximately the same rate in Chaucer. This must be seen as further counter-evidence against the claim that LOOK + -ly is not a part of the grammatical inventory of this writer. 9 As for the early Modern English data, only four of the ten writers represented in the specially compiled corpus make use of the LOOK + -ly construction. However, this is not evidence against its grammaticality, for if we study the data more closely, we see that five of these writers do not use copular LOOK with a plain adjective either. Three of the four authors who use LOOK + -ly also have copular LOOK in collocation with a bare adjective. Only one writer, Jonson, uses plain adjectives only. In the Helsinki Corpus, the number of -ly forms is six times higher than the number of plain forms. With respect to Shakespeare's usage, there are 27 -ly forms and 93 bare forms in his works, i.e. the ratio is approximately one to 3.5. Thus the famous writer seems to prefer bare forms; yet the number of -ly forms is large enough to justify the claim that -ly must be an option for him as well. 10 Disregarding the Shakespeare figures, there are 13 -ly forms and 9 plain forms; hence, -ly forms are in fact more common than plain forms in this period. If we compare the specially compiled corpora from the early Modern English period and the eighteenth century, we see that the number of authors to use LOOK + -ly is in fact higher in the latter period. While in the early Modern English material six of the ten writers do not use the construction, in the eighteenth-century corpus six out of ten use it. This is, however, hardly due to an increase in the productivity of the pattern, but is more likely an effect of a notable rise in the use of LOOK as a copular verb. (In addition, the eighteenth-century corpus is larger, which may have played a role too.) The figures show that the relative proportion of -ly forms to bare forms drops considerably in this period, the ratio being one to seven; hence, the bare alternative is in the process of ousting the -ly form. Of the four eighteenth-century writers who do not use LOOK + -ly, two do not use copular LOOK with bare adjectives either, while two only use the

86

Kristin

Killie

bare form. One writer, Gay, uses -ly only. It is interesting to note, however, that even those writers who clearly prefer bare forms (e.g. Burney, Defoe, Fielding and Sheridan) use -ly forms occasionally. Thus also in this century -ly must have been an option generally, though perhaps a more marked one. On the basis of the figures in tables 1 and 2, I would claim that the LOOK + -ly pattern is productive in earlier English, and particularly in early Modern English. There are simply too many recorded uses for it to be written off as an anomaly. Two other factors which may also serve as indicators of productivity are (i) the large number of writers to use the construction and (ii) the wide range of semantic and morphological types represented. As to the latter point, there are 49 attested forms in the corpus material, given in the list below. (Where there is more than one attested example of a form, token frequencies are indicated in parentheses). agedly amiably angrily (2) awkwardly barrenly bloodily charmingly (2) cheerfully (2) cheerly debonairly demurely (3) dirtily dismally (2)

dispitously dustily enchantingly fiercely foolishly freshly (2) frowningly (2) gloriously (2) greenly grimly (3) heavily (3) lightly lothsomely

merrily (7) meekly mildly nobly (2) piteous ly (3) pleasantly (2) prettily (3) sadly (5) saucily scornfully scurvily (3) shockingly (3) simperingly (2)

smugly (2) soberly (2) solemnly stedfastly strangely successfully sweetly wearily wildly (6) youngly (2)

With respect to the number of writers represented, there are as many as 43 recorded users of the construction in the material. To go along with the overprocessing view, we would have to assume that all these writers are not fully competent users of the language. This is obviously a doubtful assumption. In (2)—(11) below I provide some examples from the corpus, all the different periods being represented.11

The case of the LOOK +

-ly

construction in English

87

Middle English: (2)

...for there is mani that lokit soberly, stedfastly, and mildely in her fals lokes, that women werten that thei be full of thought for distresse of loue, and it is done but to make a fals semblaunt to deseyue hem. (Anon, Book of the knight..: 57)

(3)

He loked grymly andfyersly in his vysage for grete wrath. (c.1489 Caxton Sonnes of Aymon ii. 61, Oxford English Dictionary)

(4)

Who looketh lightly now but palamoun? {Canterbury Tales, group 1: lines 1869-1870)

Early Modern English: (5)

Me thinketh, by the masse, by your countenance you loked so wildly when you came in," quoth this good wife, "that somthing was amis. (Harman, A Caveat or Wareningfor Commen, Helsinki Corpus)

(6)

... Mistress Minx ... that looks as simperingly as if she were besmeared... (Nashe, Pierce the Penniless·. 69)

(7)

Wherefore looke ye so sadly to day? (Bible, Gen: xl. 7, Oxford English Dictionary)

(8)

Looks he as freshly as he did the day he wrestled (Shakespeare, As You Like It: 3, 2, 244-245)

Eighteenth century: (9)

You look most shockingly today, my dear friend. (1768 Goldsm. Goodn. Man 1, Oxford English Dictionary)

88

Kristin Killie

(10)

... says Sophia, looking more foolishly than ever ... (Fielding, The History of Tom Jones)

(11)

And her face look wond 'rously smugly. (Gay, Beggar's Opera)

Having claimed that copular LOOK could select both bare and -ly complements in earlier English, we may ask what motivated the use of either form. It seems plausible that one or other of the forms may in many cases have been chosen because of the requirements of poetry. As is well known, writers relied much more heavily on metre and rhyme in earlier times than is the case today. This concerns practically any type of text or genre, and regardless of the nature of the topic treated. Studying the language in older texts one therefore has to consider how the poetic devices in question may have affected the linguistic feature under scrutiny (cf. Killie forthcoming: chapter 1 for a discussion). This may be particularly important when the object of study is a structure whose use has been or is subject to change or is unstable for other reasons, as is the -ly suffix (cf. Killie forthcoming: chapter 2), as such linguistic features may well be easier for authors to manipulate than more stable ones. In (12) and (13) below, where two of the examples from Chaucer are put into context, it may well be that the choice of a -ly complement is motivated by poetic factors. In other cases, presumably, such considerations may prompt the use of bare forms. (12)

As leene was his hors as is a rake, And he nas nat right fat, I undertake, But looked holwe, and therto sobrely. Ful thredbare was his overeste courtepy; {Canterbury Tales·, group 1, Prol., lines 287-290)

(13)

I sawgh hyr daunce so comlily, Carole and synge so swetely, Laughe andpleye so womanly, And loke so debonairly. {Book of the Duchess: 851, [Mustanoja 1961: 650])

The case of the LOOK +

-ly

construction in English

89

Some would probably claim that formal requirements may in such cases have caused the adverbial component of the grammar to overprocess; however, this appears to me a questionable assumption. I would rather argue that for the choice of LOOK + -ly to be an option, the construction must have been available in the first place. The high number of recorded uses suggest that this was indeed the case. It is also interesting to note that a large number of the examples occur in prose, where metre and rhyme would not have played a role.

4. The word-class status of the -ly form 4.1 The adjective hypothesis In the preceding section I presented evidence that the LOOK + -ly pattern was once a feature of English grammar. The aim of the present section is to discuss the word class status of the -ly element in the construction. As mentioned in sections 1 and 2 above, it seems commonly to be taken for granted that we are dealing with an instance of adverbial -ly. In the following, however, I put forth an alternative analysis, suggesting that the element in question may be an adjective. Having done this, I evaluate both alternative hypotheses. Recall the suggestion by the Oxford English Dictionary that in some of the earlier instances of the LOOK + -ly pattern, the -ly element may have been an adjective. This ties in neatly with Mcintosh's (1991) analysis of Old English adjectives. It is well known that Old English has a double set of adjectives, one with a bare stem, the other with the adjectival -lie suffix attached, as in wis and wislic. In his (1991) paper, Mcintosh asks whether there are any semantic differences between the two members of such sets. He concludes that the choice of form is governed by some basic semantic distinctions. Normally, the base form is used to qualify animate beings while the -lie derivative has only "secondarily animate reference"; it is used to qualify inanimate beings with respect to states and qualities normally associated with animate beings only (Mcintosh 1991: 298-99). Thus a person would be characterised as wis and a piece of advice as wislic (Mcintosh 1991: 299). However, -lie derivatives are occasionally used with animate nouns as well. Mcintosh (1991: 301) finds that in such cases the

90

Kristin Killie

-lie suffix indicates that the characteristics imputed to the agents are "merely inferred from an observation of their behaviour". He gives several examples of creatures who are judged from their appearance to be modiglic, grimlic, etc. (Mcintosh 1991: 301-302). A typical example is taken from Cynewulf s poem Christ, where he describes how on Judgement Day the Lord will appear to sinners grimlic to geseonne ('of an aspect terrifying to behold', [Mcintosh 1991: 302]). Mcintosh (1991: 301) claims that the -lie suffix in cases such as these carries the meaning '-like, -looking, having the appearance of being'. In the course of time English has stopped using -lie (or -ly) adjectives in the way outlined above. According to Mcintosh (1991: 300), most of the Old English -lie adjectives have passed out of the language, and the remaining -lie {-ly) partners "no longer serve any purpose except in those rather few cases where the meaning of one partner has deviated from that of the other in some special way", as in the cases of sick versus sickly and good versus goodly. To indicate that the characteristic imputed to the subject is only inferred from a visual impression, present-day English uses the base (and only) form of the adjective, or it chooses alternative constructions, e.g. compounds of the base form and the suffix -looking (Mcintosh 1991: 308, note 5). There are some factors that may be interpreted as evidence in favour of the adjective view. Mcintosh himself backs up his hypothesis by referring to the current use of the Icelandic -legur suffix. This suffix, he maintains, makes reference "not to an intrinsic quality or characteristic but to the impression, usually but not always visual, which has been created" (Mcintosh 1991: 306). The use of the -legur suffix seems to provide good support in favour of Mcintosh's theory. Both English and Icelandic are Germanic languages, but compared to English, which has changed rather dramatically, Icelandic has remained relatively stable. It seems plausible, then, that we can learn more about the use of the Old English -lie suffix by studying its present-day Icelandic cognate. Another interesting fact to consider is that the meaning attributed to the Old English -lie suffix by Mcintosh bears close resemblance to what is assumed to be the original meaning of the suffix, i.e. "having the appearance or form indicated by the first element or word" (Oxford English Dictionary: -ly, suffix 1). This phrasing of course also reveals a dependence on the sense of vision.

The case of the LOOK +

-ly

construction in English

91

Further evidence in favour of Mcintosh's analysis is found in several of the entries for -ly adjectives in the Oxford English Dictionary, where the adjectives in question are paraphrased by way of the suffix -looking or the noun appearance, examples being darkly, goodly and grimly. What I would like to suggest, then, is that the -ly elements found with copular LOOK at earlier stages of the language are in fact not illegitimate adverbial uses at all, but represent a continuation of the tradition of indicating by way of the adjectival -ly suffix that the characteristic imputed to the subject is inferred from a visual impression. Indeed, if the adjectival -ly suffix continued to have this meaning after the Old English period, there would be no environment where it would be more appropriately used than in collocation with copular LOOK. It has been suggested to me that a condition for accepting that the -ly forms under discussion are adjectives must be that they also occur in other, more prototypical adjective positions, i.e. attributively as well as predicatively after the verb BE. It has also been claimed that the -ly forms in question ought to occur after other verbs of "seeming", such as SEEM and APPEAR. In order to check how the -ly forms in question behave with respect to these linguistic environments, I did two things: (i) checked if the forms in question are listed as adjectives in the Oxford English Dictionary, and (ii) checked all instances of SEEM and APPEAR in the corpus, as well as in the Oxford English Dictionary, focusing on those uses which could be argued to be semantically more or less parallel to the LOOK + -ly pattern. Of the -ly forms which were found to occur with copular LOOK, the following are shown to have had other adjective uses: angrily, bloodily, freshly, gloriously, grimly, lightly, meekly, mildly, piteously, solemnly, sweetly, youngly,12 Some examples of these adjectives in use are provided in (14)—(17) below, all of which are from the Oxford English Dictionary. Interestingly, the adjectives in question almost without exception appear attributively.13 Only in a few cases do they appear predicatively with BE; however, to my mind this is no evidence against their adjectivehood, as even in present-day English there are a number of adjectives which are restricted to either attributive or predicative use, e.g. ill, well, chief, sheer, sole, etc. (14)

Mid mildliche worden (c.1205 Lay. 8832)

92

Kristin Killie

(15)

In came Margaretis grimly ghost, And stood at Williamis feet. (1611 Beaum. & Fl. Knt. Burn. Pestle ii. v)

(16)

Who ys he that wolde not foil gretly sorow to see so feire and so solemly a body to be caste under so grete iniuriis and sore peynys. (1482 Monk of Evesham [Arb.] 106)

(17)

Fayne maye thy frendes be in fere, To see thy sweetlye [v.r.frely]face(?a.l500 Chester PL [Shaks. Soc.] II. 2)

With respect to copular APPEAR and SEEM, these hardly ever occur with -ly forms. In fact, there was only a single example in all of the corpus material, i.e. appear nobly in the early Modern English section (HCE3) of the Helsinki Corpus. Presumably, the fact that the verbs in question used to take different kinds of complements is related to a difference in their lexical meanings. Thus, although the three verbs are commonly considered to belong to the same class ("verbs of perception", according to Quirk et al. 1985: 203-204), I would claim that they are in fact not as similar as has been assumed. Essentially, while copular LOOK typically expresses a fairly concrete state of affairs, APPEAR and SEEM in most cases denote more abstract relationships. To be more specific, LOOK has stronger links with the concrete external world, while APPEAR and SEEM depend more strongly on internal or mental reasoning, as in It appears/seems to me that he is going to win the election (as well as all uses of APPEAR and SEEM in this paper). Note that I am not claiming that LOOK cannot be used in such abstract contexts as well, or that APPEAR and SEEM are never used to denote concrete appearance; I only claim that these may not be their prototypical uses. In other words, LOOK has not become quite as abstract as the two other verbs in question. At least, this appears to be the case in present-day English and may also have been the situation in earlier English. There is some evidence to back up such a hypothesis. Firstly, the examples under APPEAR and SEEM in the Oxford English Dictionary almost exclusively contain the more abstract uses of these verbs, while there are a large number of concrete examples under LOOK. Secondly, I have come across only very few LOOK + -ly examples where LOOK expresses a more abstract meaning. Thirdly, consider the behaviour of the three verbs in the sentences in (18) and (19) below.

The case of the LOOK + -ly construction

in English

93

(18) a. She looks old/tired/charming/nice. b. She seems old/tired/charming/nice. (19) a. She looks blood-stained/tall/fat. b. IShe seems blood-stained/tall/fat. With regard to the first pair, it seems that in the (a) sentence the characteristics denoted by the adverbs are inferred on the basis of a visual impression, while in (b) they may be based on other kinds of information, such as the subject's behaviour. The difference between the two verbs is particularly clear in the case of nice, which may function as a description of external appearance in (a) but must refer to inner qualities in (b). As for the sentences in (19), the adjectives here denote external characteristics. It appears that LOOK is appropriate to denote such attributes. SEEM, by contrast, is pragmatically odd in this context, probably because its abstract and indirect nature is not compatible with properties which are so easily observable.14 In sum, LOOK, SEEM and APPEAR, although some of their uses overlap semantically, are not synonyms, but are characterised by different levels of abstraction and imply different cognitive processes. This may explain why their complements used to take different morphological shapes. On the basis of this argument, which unfortunately cannot be proved empirically, I would suggest that the fact that APPEAR and SEEM do not collocate with -ly forms is not a valid counter-argument against the adjective hypothesis.

4.2. The adverb hypothesis With respect to the adverbial hypothesis, there are many claims mentioned above that need to be addressed. First of all, Heuer's claim that the -ly forms found in collocation with copular LOOK are the result of a confusion with active LOOK may of course be correct, but unfortunately, this is a claim that can neither be confirmed nor rejected empirically. It would, however, be easier to believe in this hypothesis if we were talking about a few examples and authors. It is more difficult to believe that so many writers — i.e. 43 — were unable to keep the two uses of LOOK apart. In this connection, we may also ask why they did not confuse active and stative SEEM and APPEAR

94

Kristin Killie

as well. As for Mustanoja's hypothesis, the same kind of problems apply. Nevalainen's objections to the adjective hypothesis are definitely worth considering. To recapitulate, her arguments are (i) that the derivation of adjectives from other adjectives by way of -ly was not a very productive process in late Middle and early Modern English, and (ii) that most of the adjectives derived in this manner in the periods in question have specialised senses, typically metaphorical and moral ones. With respect to the former claim, Nevalainen is probably right in making this assumption (cf. Dalton-Puffer 1996: 175-176). However, it may still be the case that the process was more productive in certain linguistic contexts, for example in collocation with copular LOOK. Of course, these are only speculations that cannot be tested. As for Nevalainen's second claim, the truth of this claim cannot be tested on the basis of the evidence presented in that article. Surely, we need more evidence than five adverbs to make any categorical claims about the nature of deadjectival -ly derivation in the periods in question. It seems to me that the uses and meanings of adjectival -ly must have been many, cf. for example Middle English whitely and redly, where the -ly suffix seems to be equivalent to present-day English -ish, and warmly, where the Oxford English Dictionary is unable to explain the difference between this adjective and the plain form, but where there may well be one. I would also like to add that although the LOOK + -ly pattern is a feature of English from the late Middle English period onwards, it does not follow from this that the -ly elements used in the construction must date from the same period. Indeed, several of the -ly forms in the corpus data, for example angrily, grimly, lightly, mildly and sweetly, go back to Old English, and their senses are not any more metaphorical than those of their unsuffixed partners. Neither is the Middle English creation meekly. For this reason, I find Nevalainen's conclusion about deadjectival -ly derivation in late Middle English and early Modern English (and with it her claim that the forms in question are adverbs) rather too hasty. On the basis of the discussion in this section, it seems that neither the adjective nor the adverb hypothesis can be easily refuted or confirmed. However, in contrast to what scholars have claimed before me, I would suggest that the adjective hypothesis is the most likely one.

The case of the LOOK + -ly construction

5.

in English

95

Concluding remarks: implications for the field

As stated earlier in this paper, I believe the claims made by the linguists discussed in section 2 to be wrong. With respect to the hypothesis that the forms in question are adverbs, the basis of this claim is probably the knowledge that the -ly suffix was fairly well established as the adverbial suffix in the Middle English period, the productivity of adjectival -ly being much reduced. As for the conclusion that the forms in question are anomalies, this follows almost inevitably from the first assumption, since from a present-day point of view adverbs are disallowed in collocation with copular verbs. In the light of these premises, it is quite understandable that the linguists discussed in this paper have come to the conclusions they have, and of course, they may be right in their claims. If my analysis is the correct one, however, the case under discussion here demonstrates how careful we need to be in dealing with earlier stages of a language. Ideally, our analyses of older data should rest not on our intuitions about the present state of the language, but solely on detailed and unbiased studies of historical data. This goal is clearly difficult to obtain since our current intuitions are an apparatus we carry with us all the time, and which it is difficult to switch off; it is. Ideally, however, we should simply approach an earlier stage of a known language as we would approach a foreign language. If my analysis of the LOOK + -ly pattern is correct, this also shows that it may prove helpful to study even older varieties of the language than that which we wish to describe, as the origin of the structure under study — or as in this case, the semantic principle governing its use — may sometimes be traced backwards in time.

Appendix Primary sources (number of words in parenthesis after the title) Middle English texts: The texts below are all made available by the University of Michigan Humanities Text Initiative (Corpus of Middle English Prose and Verse) at: http://www.hti. umich.edu/english/mideng/bibl.html.

96

Kristin Killie

Anon., Anon., Anon., Anon., Anon.,

Blanchardyn and Eglantine. (77,300) Book of the Knight of La Tour-Landry. (81,300) Melusine. Part I. (130,600) Merlin: or the History of King Arthur: a Prose Romance. (289,000) Prose Life of Alexander. (47,700)

Anon., Ratis Raving, and Other Moral and Religious Pieces, in Prose and Verse. (10,700) Anon., Religious Pieces in Prose and Verse. (24,300) Anon., Sawles Warde. (4,900) Anon., Sir Gawain and the Green Knight. (21,300) Anon., The Three Kings' Sons. Part I. (108,400) Anon., Three Prose Versions of the Seereta Secretorum. (101,000) Julian of Norwich, A Revelation of Love. (55,100) Chaucer, Geoffrey, The Canterbury Tales. (207,200) Chaucer, Geoffrey, Troilus and Criseyde. (74,500) Chaucer, Geoffrey, A Treatise on the Astrolabe. (17,800) Helsinki Corpus of English texts: Diachronic part. In: International Computer Archive of Modern English. Bergen: Norwegian Computing Centre for the Humanities. (608,600)

Early Modern English texts: Ascham, Roger. The Scholemaster. (23,750) http://darkwing.uoregon.edu/~rbear/ascham 1 .htm and http://darkwing.uoregon.edu/~rbear/ascham2.htm: Judy Boss. Bacon, Francis. The New Atlantis. (15,400) http://www.hti.umich.edu/english/pd-modeng/bibl.html: Humanities Text Initiative. Behn, Aphra: The City Heiress. (31,500) http://etext.lib.virginia.edu/: Oxford Text Archive. Behn, Aphra. The Unfortunate Happy Lady. (10,500) http://curly.cc.utexas.edu/~churchh/bhnufhpl.html [no indication of editor /publisher]. Cavendish, George. The Life and Death of Cardinal Wolsey. (72,150) http://library.utoronto.ca/www/utel/ret/ret.html: Web Development Group.

The case of the LOOK + -ly construction in English

97

Cavendish, Margaret, Duchess of Newcastle. A True Relation of My Birth, Breeding, and Life. (9,000) http://www.acs.ucalgary.ca/~meedward/CTrue.html. [no indication of editor/publisher]. Chettle, Henry. Kind-Harts Dreame. (14,000) http://darkwing.uoregon.edu/~rbear/kind.html: Richard Bear. Elyot, Thomas. The Boke Named The Governour. (40,150) http://darkwing.uoregon.edu/~rbear/gov/gov 1 .htm: Ben Ross Schneider, Jr. Jonson, Ben. The Alchemist. (32,000) http://www.hti.umich.edu/bin/pd-idx?type=header&idno=JonsoAlche: Hugh Craig. Jonson, Ben. Volpone. (28,100) http://eserver.org/drama/volpone.txt: Hugh Craig. Middleton, Thomas. No Wit, no Help Like a Woman's. (29,000) http://www.tech.org/~cleary/nwnh.html.: Chris Cleary. Middleton, Thomas. The Witch. (17,550) http://www.tech.org/~cleary/witch.html.: Chris Cleary. More, Sir Thomas. The History of King Richard the Third. (30,850) http://www.r3.org/bookcase/more.moretext. html: Richard Bear. More, Sir Thomas. Utopia. (42,600) http://zippy/dct.ac.uk/www/books: Kirk Crady. Helsinki Corpus of English texts: diachronic part. In: International Computer Archive ofModern English. Bergen: Norwegian Computing Centre for the Humanities. (551,000) Shakespeare, William (1989) William Shakespeare the Complete Works. Macintosh version. Oxford: Oxford University Press

Eighteenth century: Brooke, Frances. The History of Julia Mandeville. (63,400) http://english-www.hss.cmu.edu/18th: University of Virginia Library. Burney, Fanny. Dr. Johnson and Fanny Burney. (57,200) http://pubweb.ucdavis.edu/documents/engac/craciun. womrom.html: Jamie L Spriggs.

98

Kristin Killie

Defoe, Daniel. Robinson Crusoe. (123,000) http://www. 1 i.net/~scharf/defoe.html: Robert Scharf. Fielding, Henry. The History of Tom Jones. (346,300) http://milton.mse.ihu.edu:8003/quixote/fielding. html [no indication of editor/publisher]. Gay, John. The Beggar's Opera. (22,400) http://darkwing.uoregon.eduTrbear/beggar.html: Richard Bear. Goldsmith, Oliver. The Vicar of Wakefield. (63,800) http://www.english.upenn.edU/-ilynch/18th/g.html: Judy Boss. Schaw, Janet. Journal of a Lady of Quality, Being the Narrative of a Journey from Scotland to the West Indies, North Carolina, and Portugal, in the Years 1774 to 1776. (32,850) http://www.arts.gla.ac.uk/COMET/starn/prose/schaw/iourney.htm· [no indication of editor/publisher]. Sheridan, Frances. Memoirs of Miss Sidney Bidulph. (200,000) http://www.hti.umich.edu/bin/pdidx?type=header & byte=57420960: Humanities Text Initiative. Wollstonecraft, Mary. Maria, or the Wrongs of Woman. (46,700) http://www.georgetown.edu./irvinemi/english016/franken/marial0.txt: Judy Boss. Woolman, John. Journal of John Woolman. (60,000) http://english-www.hss.cmu.edu/18th.html: Judy Boss. Also: The Oxford English Dictionary Second Edition on Compact Disc. 1993. Oxford: Oxford University Press.

Notes 1. The ideas put forth here were first suggested in Killie (1993: 70-74), but without much data to back them up empirically. 2. Eventually, the -ly suffix came to be conceived of as the adverbial suffix per se, but that is a story which is not to be discussed here. For a fuller account of the history of the adverbial -ly suffix, see Killie (forthcoming). 3. A similar functional indeterminacy is claimed to obtain in Old English (Mitchell 1985: 4 7 2 ^ 7 2 ) .

The case of the

LOOK + - l y

construction in English

99

4. The suffix label " - L Y 1 " is used here to indicate that we are dealing with an instance of adjectival -ly, as opposed to -LY2, which is adverbial -ly. 5. I have not scanned the Old English section of the Helsinki Corpus, as the verb LOOK, according to the Oxford English Dictionary as well as my own observations, was not used in the relevant sense until quite late in the Middle English period. 6. I would, of course, have preferred the number of words to be the same without counting the Helsinki material, thus providing a higher degree of comparability between the same set of texts with respect to both number of authors and number of words. However, due to the short length of the early Modern English texts available on the Internet, the number of authors and texts represented would have to be much higher than for the eighteenth-century corpus for this to be possible. Hence I had a choice between balancing the number of words and the number of authors represented, of which I gave priority to the latter alternative, as I find it interesting to study how individual grammars vary and to compare this variance. 7. Those examples which are found in two sources are of course counted only once. Thus one of Chaucer's LOOK + -ly clauses is found both in the Helsinki Corpus and in the Chaucer corpus; however, this instance has been counted only once in the table and placed under the rubric Chaucer. Similarly, some of Shakespeare's uses are rendered in the Oxford English Dictionary, but are counted with the rest of the Shakespeare data in the table. 8. Four of these are from his Canterbury Tales', the remaining two are found in The Legende of Good Women and The Book of the Duchess, cited by Heuer (1932: 108) and Mustanoja (1961: 650), respectively. 9. The fact that the number of -ly forms in Chaucer is higher in table 1 than in table 2 is because in the former the examples provided by Mustanoja and Heuer are included (cf. note 8). 10. The figure for unsuffixed modifiers in Shakespeare is inflated by the highfrequency items pale and sad, which occur at a rate of 43 and 13 occurrences, respectively, i.e. they make up more than half of the total figure. 11. Where an example is cited by someone else, this source is given after the page number. For fuller information on publication data, the reader is referred to the relevant source. 12. The sources used here are The Oxford English Dictionary, Stratmann (1940) and Kurath and Kuhn (1952-). 13. Hence Nevalainen's claim that these forms do not occur in front of nouns does not hold. 14.1 have chosen to use SEEM in these examples, to the exclusion of APPEAR, which seems to have an intermediate status between LOOK and SEEM as far as level of abstraction is concerned.

100

Kristin Killie

References Dalton-Puffer, Christiane 1996 The French Influence on Middle English Morphology: A Corpusbased Study of Derivation. Berlin/New York: Mouton de Gruyter. Guimier, Claude 1985 On the origin of the suffix -ly. In: Jacek Fisiak (ed.), Historical Semantics, Historical Word-formation, 155-70. (Trends in Linguistics 29.) Berlin/New York/Amsterdam: Mouton Publishers. Heuer, Hermann 1932 Studien zur syntaktischen und stilistischen funktion des Adverbs bei Chaucer im Rosenroman. Anglistische Forschungen 75: 1-157. Jespersen, Otto 1961 Α Modern English Grammar on Historical Principles. Part 6. London: Allen & Unwin. First edition [1946]. Killie, Kristin 1993 Early Modern English Subject Modifiers. (Tromse Studies in Linguistics 13.) Oslo: Novus. forthcoming Stativity, Productivity and Adverbial Derivation: A Synchronic and Diachronic Study. Kurath, Hans and Sherman M. Kuhn (eds.) 1952Middle English Dictionary. Ann Arbor, Michigan: Michigan University Press. Lass, Roger 1997 Historical Linguistics and Language Change. (Cambridge Studies in English 81.) Cambridge: Cambridge University Press. Mcintosh, Angus 1991 Old English adjectives with derivative -lie partners: some semantic problems. Neuphilologische Mitteilungen 92: 297-310. Mitchell, Bruce 1985 Old English Syntax. Volume 1. Oxford: Clarendon Press. Mustanoja, Tauno F. 1960 A Middle English Syntax. Part 1. Helsinki: Sociötd Niophilologique. Nevalainen, Terttu 1997 The processes of adverb derivation in late Middle and early Modern English. In: Matti Rissanen, Merja Kytö, and Kirsi Heikkonen (eds.), Grammaticalization at Work: Studies of Longterm Developments in English, 145-189. Berlin/New York: Mouton de Gruyter. Phillipps, K.C. 1970 Jane Austen's English. London: Andre Deutsch.

The case of the LOOK + -ly construction in English

101

Pyles, Thomas and John Algeo 1982 The Origins and Development of the English Language. San Diego: Harcourt Brace Jovanovich. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Stratmann, Francis Henry 1940 A Middle English Dictionary: Containing Words Used by English Writers from the Twelfth to the Fifteenth Century. London: Oxford University Press. Western, August 1906 Some remarks on the use of English adverbs. Englische Studien 36: 75-99.

The indefinite pronoun mam "nominal" or "pronominal"? Linda van Bergen

1.

Introduction

In generative studies on Old English word order a distinction is normally made between two types of noun phrases: "nominal" and "pronominal". This division is of particular importance in relation to what has been analysed as clitic (or clitic-like) behaviour of pronominals. So far, the classification of the indefinite pronoun man in relation to that division appears to have been assumed without argument. In most cases it is regarded as nominal, for example in van Kemenade (1987) and Koopman (1997a), but it is treated as pronominal in Haeberli and Haegeman (1995: 86). In this paper I will present evidence that man cannot be grouped with the nominale. This leads to problems of analysis, as will be shown in section 3.

2. Man and inversion One of the main arguments that has been used to support a clitic analysis for personal pronoun subjects is the fact that, unlike nominal subjects, they normally do not invert with the finite verb in clauses with a topicalised constituent, unless this initial constituent is a wA-word, the negative particle ne or a member of a limited set of adverbs (pa 'then' and ponne 'then' in particular).1 See for example van Kemenade (1987: chapter 4) and Pintzuk (1991: chapter 4). This phenomenon is illustrated in (1) and (2), with a non-inverted personal pronoun subject and an inverted nominal subject respectively. (1)

Midlare he sceal him tcecan with learning he must them teach 'He must instruct them with learning' OECHom I, 17, 240.4)

104

(2)

Linda van Bergen

on twam pingum hcefde god pees marines saule gegodod in two things had God the man's soul endowed 'God had endowed man's soul with two things' OECHomI, 1,20.1)

Although inversion is not found consistently with nominal subjects, as demonstrated by Koopman (1996, 1997b), there is still a clear difference between the behaviour of pronominal subjects on the one hand and nominal subjects on the other, since inversion with pronominal subjects is all but absent in this context. This difference, then, provides us with an important diagnostic for deciding whether a particular lexical item behaves as a pronominal subject or as a nominal one. So to determine whether a classification as nominal is justified, I have investigated the frequency of inversion found for man} The data have been taken from several sources. The first is the morphologically tagged version of the Old English part of the Helsinki Corpus} The Helsinki Corpus was chosen to get a general insight into the behaviour of man across texts. But given that the rate of inversion with nominal subjects varies by text, it seemed appropriate to collect data from more homogeneous corpora as well. For this I have chosen King Alfred's translation of Cura Pastoralis and the works of jElfric. To ensure that the results for Cura Pastoralis and iElfric could be compared with the frequency of inversion found by Koopman, I have followed his criteria as closely as possible. This means that the data have been restricted to uncoordinated main clauses with a fronted object or prepositional phrase. Only clauses with a single topicalised constituent have been included in the counts (ignoring any preceding subclauses), but I will mention any cases of interest which have more than one fronted constituent.

2.1. The Helsinki Corpus The data from the Helsinki Corpus are the most complex of the sets of data to be discussed here, which is due to interference from other factors. The numbers found for topicalised objects are given in table 1.

The indefinite pronoun man\ "nominal" or "pronominal"?

105

Table 1. Inversion of man after fronted objects in the Helsinki Corpus Not inverted Inverted Indicative, non-negated "77Γ773 19(5) Indicative, negated 6 Subjunctive 10 2(2) 21 Total 16 Although the totals show a fairly even division between inversion and noninversion, this is misleading. As the table shows, inversion of man is rare with fronted objects if the verb is neither negated nor in the subjunctive mood. Examples with indicative, negated and subjunctive verb forms are given in (3), (4) and (5) respectively. I will come back to the cases with negation and verbs in the subjunctive mood in section 2.2. (3)

a. Micele maran gyltas man mceg gebetan her on pisum life much greater sins one can atone-for here in this life 'Much greater sins can be atoned for here in this life' ( £ H o m 16, 157)5 b. Ciricsceat mon sceal agifan to pam healme & to pam heorde church-scot one must give to the roof and to the hearth 'Church-scot must be given according to the roof and hearth' (Lawine 61)

(4)

a. pa wergendan ne sceal mon na ongeanwerian ac ma the cursing not must one not revile-in-return but rather bletsian bless 'those who curse must not be reviled in return, but rather be blessed' (BenR 17.13) b. Dam dysegan ne mceg mon na mid wordum gestyran the foolish not can one not with words urge 'The foolish cannot be urged with words' (BenR 13.22)

106

(5)

Linda van Bergen

a. Sunnandceges freols healde man georne Sunday's feast hold one gladly 'The feast-day of Sunday should be observed gladly' (LawVIAtr 22.1) b. Wydewan & steopcild werie man & nerie widows and orphans protect one and save 'One should protect and save widows and orphans' (WHom 10c, 162) c. (Gif... gerihtlcecan,) hine manpreage mid teartran steore him one punish with severe discipline '... he should be punished with severe discipline' (BenR 52.6)

The only case of inversion with a non-negated indicative verb involves more than one fronted constituent. It is given in (6) below. But there are good reasons for rejecting it as a counter-example.6 (6)

yrsunge tidelice sceal mon gehealdan anger temporarily must one retain 'one must retain anger temporarily' (BenR 17.6)

A similar picture to that found for topicalised objects emerges when the same type of data is gathered for fronted prepositional phrases. The numbers are given in table 2. Because most verbs in this case are simple indicatives the results look much clearer, but they are not essentially different. Examples are given in (7)-{9).

Indicative, non-negated Indicative, negated Subjunctive Total

Not inverted 137 -

13

Inverted -

1 1 2

The indefinite pronoun man: "nominal" or "pronominal"?

(7)

107

Be disum lytlan man mceg understandan by this little one can understand 'By means of this little thing can be understood' 0EGenPref72)

(8)

On hcepenum peodum ne dear man forhealdan lytel ne micel in heathen peoples not dare one withhold little nor much 'Among the heathens one dare not withhold little or much' (WHom 20.3, 27)

(9)

JEt pam uhtsange rcede man pcere godcundan lare bee at the matins read one the divine teaching's books 'At matins, books of divine teaching should be read' (BenR 33.17)

Again, the only cases of inversion found with a non-negated indicative verb are in clauses which have more than one fronted constituent. In one of these, given in (10), the other constituent is the adverb pa 'then', which normally triggers inversion of all types of subject, and inversion in this example is therefore unsurprising. (10)

Da cefterpcere tide pa bead mann eft fyrde be ftillum wite then after that time then ordered one again army by full penalty 'Then after that time the army was ordered again on full penalty' (ChronE 1016.9)

The other counter-example, given in (11a), appears to be genuine, especially since there are more cases of inversion to be found in the parts of Bald's Leechbook which are not contained in the Helsinki Corpus. Non-inversion occurs as well under similar circumstances, as exemplified in ( l i b ) , so in this text inversion of man is really variable. The same variation is found with topicalised objects in this text; see example (12). (11) a. Onfruman mid onlegenum & sealfum sceal mon lacnian in beginning with poultices and ointments must one treat 'First one must treat it with poultices and ointments' (Lch II (2), 19.1.10)

108

Linda van Bergen b. On fruman mon sceal dceg odöe .IL togcedere gefcestan in beginning one must day or 2 together fast 'First one must fast a day or two' (Lch II (2), 32.1.2)

(12) a. Pas Icecedomas sceal mon don wip sidansare these medicaments must one do against side's sore 'These medicaments one must use against soreness of the side' (Lch II (2), 46.1.2) b. Pis mon sceal writan on husldisce this one must write on housel-dish 'This must be written on the housel-dish' (Lch 11(1), 62.3.1) I have not yet investigated this text in detail, but the cases I have found of personal pronoun subjects in a comparable context do not invert. Nevertheless, even if it could be demonstrated that the behaviour of man is different from that of personal pronouns in this text or dialect, this would not alter the conclusion that in most texts man seems to behave as a pronominal subject with respect to inversion, at least when the verb is neither negated nor in the subjunctive mood. This, then, leaves the question whether the separate treatment of negation and subjunctives is justified and whether the inversion found in these contexts should be taken as an indication that the behaviour of man does not entirely match that of personal pronouns.

2.2. Negation and subjunctive verbs Given the high incidence of clauses with a negated verb in initial position, and the consequent inversion found for all types of subject in this environment, it seems plausible that inversion after a topicalised constituent is more frequent with negated verbs, even for subject types that normally do not invert. Examples of this with personal pronoun subjects can indeed be found as well, as shown in (13). See also van Kemenade (1997a, 1997b). It is not clear, however, whether it is the dominant pattern for all or even most texts. Notice that in (13c) inversion is followed in the next clause by non-inversion.

The indefinite pronoun man: "nominal" or "pronominal"?

109

(13) a. witodlice pone cwyde pcere amcensumunge ne gesprcec truly the word of-the excommunication not spoke hena py pe he not because 'truly, he did not speak the word of excommunication because (GD2(C), 23.152.13) b. Drihten, pine rihtwisnysse ne behidde ic an minre heortan Lord your justice not concealed I in my heart 'Lord, I did not conceal your justice in my heart' (ChrodR 1, 79.49) c. His lichoman lustum ne sceal he fulgan; his agene wyllan his bodily desires not must he follow his own will he ne sceal fulgan he not must follow 'His bodily desires he must not follow; his own will he must not follow' (ThCap 1,21.331.80) It cannot be judged on the basis of this corpus whether the pattern found for negated verbs and man is representative, because, with just one exception, all examples are found in a single text, a prose translation of the Benedictine Rule, with most occurring in the same passage. It would be interesting to see for any texts in which there is a strong preference for either inversion or non-inversion of personal pronouns with a negated verb form, whether the same holds for man (and vice versa), but low frequency is likely to be a problem. I have found only three relevant clauses with a fronted object or prepositional phrase and a personal pronoun subject in the Benedictine Rule, two of which are coordinated. The uncoordinated clause has inversion, and one of the coordinated clauses does as well, but the other one does not. The examples are given in (14). This makes it impossible to be certain for this text what the behaviour of subject pronouns in the context under investigation is, both because of the low numbers and because of the potential influence of coordination on word order.

110

Linda van Bergen

(14) a. eornestlice, of damdcegenah he fordon his agenes indeed from that day not-has he therefore his own lichoman geweald body's ruling 'indeed, from that day he will therefore not have the ruling of his own body' (BenR 101.21) b. ne hleahter ne sceal he lufian nor laughter not must he love 'nor must he love laughter' (BenR 18.8) c. Minne gylt ic gecydde pe and mine unrihtwisnesse ic ne my sin I made-known you and my iniquities I not bediglode hid 'My sin I made known to you, and I did not hide my iniquities' (BenR 28.19) However, note that there is also one case of non-inversion with man in the sample of the Benedictine Rule contained in the Helsinki Corpus, given in (15), which has not been included in the counts because it has more than one fronted constituent. So it is also possible that inversion with a negated verb in this environment is not obligatory in this text. Alternatively, the clauses surrounding (15), which do not have inversion, may have had an impact on the word order, although that is perhaps made less likely by the fact that in the case of (4a) (repeated in [16] together with the surrounding clauses), occurring in the same chapter of the same text, the position of mon varies according to whether the finite verb is negated or not. (15)

estmettas no to grcediglice mon ne sceal lufian, ac dainty foods not too eagerly one not must love but fcesten mon sceal lufian fast one must love 'one must not love dainty foods too eagerly, but one ought to love fasting' (BenR 16.23)

The indefinite pronoun man: "nominal" or "pronominal"?

(16)

Ill

his fynd mon sceal lufian for Godes lufan; pa wergendan ne sceal mon na ongeanwerian, ac ma bletsian; ehtnesse for rihte mon sceal eadmodlice gepolian 'one must love one's enemy for the love of God; the accursed must not be reviled, but rather blessed; persecution for the sake of justice must be endured humbly' (=[4a] BenR 17.12)

In any case, the influence of negation on the placement of man in this particular text is clear. Table 3 shows the numbers for the sample of the Benedictine Rule included in the Helsinki Corpus (according to the criteria used so far). Whether other texts show as clear a difference and whether the pattern is the same for personal pronouns and man alike remains to be seen, but for now it is enough to know that the negated verb forms have to be kept separate from non-negated verbs in determining the rate of inversion. Table 3. Inversion of man in the Benedictine Rule Fronted prepositional phrase non-inverted inverted Indicative, 3 non-negated Indicative, negated

Fronted object non-inverted inverted 3 6

With respect to the subjunctive, the construction of subjunctive verb plus man is often used in a way which is similar to the imperative. To illustrate this, in example (17) it is used in a list of straightforward imperatives. It is probably best seen as parallel to the imperative construction with an expressed subject. See also Mitchell (1985: §§879-915) for a discussion of the imperative and what he calls the "jussive subjunctive". (17)

do

ponne sumne dcel para wyrta pcerto; clcem

do [IMP] then

arfeet;

donne on

some part of-the herbs thereto smear [IMP] then

leet

standan nygon niht, wende

in

man celce dcege

brazen-vessel let [IMP] stand nine night stir [SUBJ] one every day 'then add a part of the herbs to it, then smear it into a brazen vessel, let it stand for nine nights, one should stir it every day'

(Med 3, 37.4)

112

Linda van Bergen

The similarity in syntactic behaviour of these two categories includes word order, at least to the extent that they both tend to occur in clauseinitial position, with resulting inversion of the subject (if present). So as with negated verb forms, it would not be surprising if inversion of subjunctives after a topicalised constituent were found with all types of subject. Note that imperative clauses with inversion of an expressed subject pronoun after a topicalised constituent can be found, as shown in (18). The same holds for subjunctives with a directive function and a personal pronoun subject, as in (19). However, I do not know how regular inversion is in this context, nor is it clear how widespread the use of topicalisation in these contructions is. (18) a. Pine teodan sceattas & pine frumripan gongendes & your tenth properties and your first-fruits of-going and weaxendes agif% pu Gode of-growing give you to-God 'Give your tithes and your firstlings of the going and growing to God' (LawAfEl 38) b. Na us, drihten, na us, ac pinum naman sele pu wulder not us Lord not us but your name give you glory 'Give glory not to us, Lord, but to your name' (BenR 4.3) c. Sodlice of dam treowe ingehydes godes & yfeles truly from the tree of understanding of-good and of-evil ne et du not eat you 'Truly, do not eat from the tree of knowledge of good and evil' (Gen 2.17 [attr. to ^lfric]) (19) a. upahefednysse fleo he; ealde men arige he arrogance flee he old people honour he 'he should flee arrogance; he should honour elderly people' (ThCap 1,21.331.94)

The indefinite pronoun man: "nominal" or "pronominal"?

113

b .Ilhida cet Eadmunddescotan hcebbe he & wel bruce 2 hides at Edmundesscota have he and well use 'Let him have two hides at Edmundesscota and use them well' (Ch 1394, 5) c. (Gif... witan,) LXscill. gebete he & oper LXgeselle to 60 shillings pay he and other 60 give as wite fine ' ... let him pay 60 shillings and give another 60 as a fine' (Lawine 6.2) The data on man shed only a very limited light on these questions: the cases found in the Helsinki Corpus do not come from a single text this time, but they do nearly all come from the same author. Nearly all cases are found in the extracts from Wulfstan's homilies (nine instances in total, all occurring in the same homily, 10c, and most in the one particular passage). Two others, essentially the same clause in both cases, are found in the law codes VjEthelred and VI ALthelred respectively, again written by Wulfstan. The comparatively frequent use of this construction with man, then, may be an idiosyncrasy of Wulfstan's writing. For my purposes it is sufficient for the moment to check whether Wulfstan makes a consistent difference in the position of man according to the mood of the verb. There are only a few relevant instances with indicative verb forms in the Helsinki Corpus, so I have checked the whole of Wulfstan's homilies, 9 the law code VI Aithelred and the text Institutes of Polity for further instances. What evidence there is supports the claim that man does not invert with indicative verbs. The numbers are given in table 4, and examples with indicative verbs in (20). The only case of inversion with an indicative verb, given in (21), involves negation. I have also checked the coordinated clauses, and the same pattern is found here: inversion with negated and subjunctive verb forms and no inversion with non-negated indicatives. The examples of coordinated clauses involving a modal verb are given in (22). Unfortunately, I have not found any instance of a personal pronoun subject and a subjunctive verb in a clause with topicalisation in reading through these texts, so it cannot be determined whether these would also invert.10

114

Linda van Bergen

Table 4. Inversion of man in Wulfstan's writing Fronted prepositional phrase non-inverted inverted -co/+co

Indicative

WHom WPol LawVIAtr Subjunctive WHom WPol LawVIAtr Ind. negated WHom WPol LawVIAtr

1 /3 2 "/-/ -/ -/-/-/-/-/-

-co/+co - / -

Fronted object non-inverted

inverted

-C0/+C0

-C0/+C0

4(2)/6(l)

-/-

- / -

- / -

-/-

8/5

-/-/-

2/7 -/I

- / - / - / -

-/I 1 /- / -

- / -

- / -

- / -

- / -

- / -

(20) a. Mid pant pater nostre man sceal to Gode gebiddan with the pater noster one must to God pray 'One must pray to God with the paternoster' (WHom 8c, 17) b. Pis man behcet for celcne pcera pe fiilluht underfehd this one promises for each of-those who baptism receive O n e promises this for each of those who receive baptism' (WHom 13, 26) (21)

On hcepenum peodum ne dear man forhealdan lytel ne micel in heathen peoples not dare one withhold little nor much 'Among the heathens one dare not withhold little or much' (=[8] WHom 20.1,22)

(22) a. Ac pas tide man sceal mid mycelre forhcefednesse healdan but this time one must with great temperance hold 'But this time must be observed with great temperance' (WHom 14, 28) b. & purh post man sceal gewunian pcet man riht healde and through that one must habituate that one right keep 'and through that one must habituate oneself to preserve right' (WHom 21, 13)

The indefinite pronoun man: "nominal" or "pronominal"?

115

In conclusion, the data from the Helsinki Corpus strongly suggest that man normally does not invert after a topicalised constituent. Once the influence of subjunctive mood and negation is recognised, the behaviour of man with respect to inversion is very similar, and possibly identical, to that of personal pronouns. The question whether these two contexts are consistent exceptions in all texts is secondary, and will need to be left open for now. But it is clear that for the texts in which cases are found in this corpus, mood and negation are factors that need to be taken into account.

2.3. Cura Pastoralis and the works of JElfric The numbers found for Cura Pastoralis confirm the general pattern found in the Helsinki Corpus, as can be seen in table 5. This text shows a relatively high percentage of non-inversion with nominal subjects according to the investigation done by Koopman (1996, 1997b). Also, the numbers for man are low. Nevertheless, the complete absence of inversion would be unexpected if man behaved as a nominal subject. Moreover, since Koopman included man among the nominal subjects, the percentages for non-inversion of nominal subjects drop if man is excluded from this category. The difference is only slight for fronted prepositional phrases (from 31 per cent to 27 per cent, according to my calculations), but noticeable for fronted objects (from 49 per cent to 27 per cent),12 and the exclusion of man would also bring the two percentages in line with each other. Table 5. Inversion of man in Cura Pastoralis

Indicative, non-negated Negated/subjunctive

Fronted prepositional phrase non-inverted inverted 6

Fronted object non-inverted 11(5)

-

-

inverted

The writings of ^Elfric investigated by Koopman emerge as having a consistently low frequency of non-inversion of nominal subjects: 11 per cent or less. This maximises the difference between nominal and pronominal subjects. The pattern found for man in jElfric's works could therefore be particularly revealing. The data given in table 6 are based on a search with the help of the Toronto Corpus through nearly all texts attributed to /Elfric.

116

Linda van Bergen

All clauses included had non-negated indicative verbs. Table 6. Inversion of man in jElfric's works Fronted prepositional phrase non-inverted inverted jECHom i 2 jECHom ii 1 JELS 2 ,ΕΗοπι 16, 22,31 1 ,EHomM 7,8, 14 jCAbusMor 1 1 jELet 3 1 jELet 4 (SigeweardZ) ^Gram 1 ^EGenPref 10 13 Total -

-

Fronted object non-inverted inverted 2 -

3(3) 3(1) 3 1(1) -

1 3 -

16

The resulting numbers are unfortunately low, but what evidence is available is unambiguous and again supports the conclusion reached so far: man normally does not invert with the finite verb after a topicalised constituent. There is only one case of inversion after a fronted prepositional phrase, but the adverb pa also occurs in a fronted position in this clause; it has been excluded from the data summarised in table 6, together with all other examples with more than one fronted constituent. Some other cases have also been excluded for various reasons, but none of these have inversion.

3. A paradoxical conclusion The data discussed in section 2 have shown that, as far as inversion is concerned, the behaviour of man is similar to that of personal pronoun subjects (at least in most texts). This clearly indicates that man should not be grouped with nominal subjects. However, there are difficulties with classifying it as a pronominal subject: there are two contexts in which man appears to follow the behaviour of nominal subjects instead. In subordinate clauses, object pronouns may intervene between the subordinator and the subject. Whereas this construction does not occur with personal pronoun subjects, it is found with both nominal subjects and man. See examples (23)-(25).

The indefinite pronoun man: "nominal" or "pronominal"?

117

(23) a. dcet hie mon mid nanre swingellan gebetan ne mceg that them one with no flogging reform not can 'that they cannot be reformed with any flogging' (CP 37.263.8) b. And we gelyfad poet hine man on rode ahenge and we believe that him one on cross hung 'And we believe that he was hung on the cross' (WHom 7a, 27) (24) a. dcette hine sio gewilnung dcere gifernesse of his modes that him the desire of-the gluttony from his mind's fcesdrcednesse ne gebrienge fortitude not bring 'that the desire of gluttony does not bring him from his mind's fortitude' (CP 43.317.6) b. And we gelyfad pcet hine clcene mceden gebcere, Sancta Maria and we believe that him pure virgin bore Sancta Maria 'And we believe that a pure virgin bore him, Holy Mary' (WHom 7a, 24) (25) a. dcet is, deah he hine mid ryhte tcelcm mcege, dcet he hit ne doo that is although he him with justice blame can that he it not do 'that is, although he could with justice blame him, that he does not do so' (CP 28.199.21) b. geleornige huru poet he hine dus cunne learn at least that he him thus know 'let him learn at least so that he knows it thus' (WHom 7a, 43) Van Kemenade (1987: 127) analyses this as cliticisation of the object pronoun onto the complementiser. In her analysis subject pronouns are clitics themselves, which explains why they normally do not invert in main

118

Linda van Bergen

clauses. In subordinate clauses they cliticise onto the complementiser. Hence the object pronoun cannot intervene. However, if non-inversion of man is accounted for by analysing it as a clitic as well, then the frequent placement of a clitic object pronoun immediately following the complementiser is unexpected in clauses which have man as their subject. The same problem occurs with clauses which have inversion of all subject types. Object pronouns are found intervening between the finite verb and both nominal subjects and man, but again this does not happen with personal pronoun subjects. Examples are given in (26)-(28) below. (26)

Pa scede him man poet hi of engla lande wceron then said him one that they from Angles' land were 'then he was told that they were from England' OECHom II, 9, 74.60)

(27)

Da ge-bced hine se bisceop bealdlice to gode then prayed himself the bishop boldly to God 'then the bishop prayed boldly to God' (jELS [Apollonaris] 157)

(28)

da betealde he hine swide geaplice then defended he himself very cunningly 'then he defended himself very cunningly' OECHom I, 5, 80.8)

In this respect, then, man is strikingly unlike personal pronoun subjects, and appears to behave like a nominal subject, and this almost certainly explains why it has usually been assumed that it should be classifed as nominal. But if man is grouped with nominal subjects, then the lack of inversion after topicalised constituents becomes very difficult to account for. Variation in the status of man between nominal and pronominal does not really seem to be an option, given that the data on inversion in clauses with topicalisation do not provide any evidence for such variation. If there were variation of status, the frequency of inversion for man would be expected to lie somewhere in between that of nominal subjects and personal pronoun subjects. The absence of inversion would lead to a claim that man can only be a nominal subject in subordinate clauses and in main clauses with inversion of all subject types, but

The indefinite pronoun man: "nominal" or "pronominal"?

119

not in other main clauses. Such an approach would be highly suspect from a theoretical point of view.14 A more promising way of dealing with the problems concerning the placement of object pronouns is likely to be available in a clitic analysis, given that the ordering of clitics in clitic groups can be subject to essentially idiosyncratic and language-specific constraints (see for example Halpern 1995: chapter 6). This will be pursued in future research. But even without a precise answer to the question concerning the status of man it is safe to conclude on the basis of the data presented here that, in spite of the resulting problems, it should not be treated as nominal.

Notes 1. I have included ne here because some analyses do so. However, it should almost certainly be analysed as procliticised onto the finite verb, and clauses of this type are better regarded as verb-initial. 2. Note that Roth (1914: 12-32 passim) suggests on the basis of his work on main clauses in parts of the Parker Chronicle (entries for 800-900) and the Peterborough Chronicle (1066-1154) that man's behaviour is similar to personal pronouns in this context, but his data for the particular construction needed are far too limited for any safe conclusions as well as suffering from problems, such as the lack of distinction between coordinated and uncoordinated main clauses (with & routinely being omitted in the examples cited) and the fact that the larger portion of text used is of such a late date. Fourquet (1938: 66-67) likewise claims that man behaves as a pronominal subject and does not invert under such circumstances, but his data largely overlap with Roth's, and are thus still very limited. Although he usually cites coordinating conjunctions, he does not keep coordinated clauses separate from non-coordinated clauses (except when subjectless coordinated clauses are involved), and most of his examples with non-inversion of man are in fact coordinated clauses. No conclusions can be based on such a limited number of examples, especially when in most cases non-inversion could be due to the tendency of finite verbs to occur late in a coordinated clause. The suggestion does not appear to have been followed up in any subsequent work. Although Bacquet (1962: 38) refers to this part of Roth's work approvingly, he does not give any relevant information on the subject in the discussion of his own data on Alfredian prose. (Notice also his remarks on man in relation to Fourquet's claims [Bacquet 1962: 264].)

120

Linda van Bergen

3. A preliminary version was used. I am grateful to Ans van Kemenade and Frank Beths for making this available. The data taken from this corpus, as well as those taken from the Toronto Corpus, have been checked against printed editions where possible. 4. The numbers between round brackets in the tables specify the number of fronted personal pronoun objects involved. Although they occur in what is usually seen as the topic position, I kept track of how many instances there were because of the different behaviour of personal pronouns in general. 5 For explanation of the abbreviations used for the names of texts, see Healey and Venezky ([1980] 1985). 6. The clause is a translation of iracundice tempus non reservare. There is only one witness for the version in (6). The four other extant manuscripts containing this part of the text have tide ne instead of tidelice (this includes the manuscripts containing a fragment of the text, which had not been collated in Schröer's [1885-1888] edition, i.e. Cotton Tiberius A. iii and Gloucester Cathedral 35; see D'Aronco [1983] for a supplementary edition). As Schröer (1885-1888: 207) points out, these manuscripts follow the Latin source more closely. Tidelice could be a later alteration, in which case the word order may have been preserved in copying from an original with the negated verb form. To this may be added that the surrounding clauses, which have inversion of man with negated verb forms, could have influenced the word order. Moreover, in the only manuscript which has the version given in (6), ne has been added above the line after tidelice in a later hand. 7. Three clauses in which the initial constituent is a preposition plus a form of the demonstrative pronoun se have been excluded because of their potential ambiguity (main or relative clause). 8. Editorial emendation. The manuscript reads (/ is found in two other manuscripts). 9. Duplicates from different versions of the same homily have been excluded. 10.Editions used: Bethurum (1957), Liebermann (1903: 246-259). For Institutes of Polity I have had to use the text as given in the Toronto Corpus. 11. One clause in which the initial constituent is a preposition plus a form of the demonstrative pronoun se has been excluded because of its potential ambiguity (main or relative clause). 12. Based on the numbers given in Koopman (1996), which differ marginally from those in Koopman (1997b). If Koopman (1997b) is used, the frequency of noninversion with fronted objects changes from 51 per cent to 31 per cent. 13. Four clauses in which the initial constituent is a preposition plus a form of the demonstrative pronoun se have been excluded because of their potential ambiguity (main or relative clause). 14. My current research indicates that treating man as nominal in subclauses and clauses with inversion would also fail to account adequately for the data. For details, see van Bergen (forthcoming).

The indefinite pronoun man: "nominal" or "pronominal"?

121

References Bacquet, Paul 1962

La structure de la phrase verbale a I 'epoque alfredienne. (Publications de la Facult6 des Lettres de l'Universitö de Strasbourg 145.) Paris: Les Belles Lettres. van Bergen, Linda forthcoming Pronouns and word order in Old English, with particular reference to the indefinite pronoun man. [Working title.] PhD dissertation, Department of English and American Studies, University of Manchester. Bethurum, Dorothy (ed.) 1957 The Homilies ofWulfstan. Oxford: Clarendon Press. D'Aronco, Maria A. 1983 II IV capitolo della Regula Sancti Benedicti del ms. Londra, B.M. Cotton Tiberius A. iii. In: Patrizia Lendinara and Lucio Melazzo (eds.), Feor ond Neah: Scritti di Filologia Germanica in Memoria di Augusto Scaffidi Abbate, 105-28. Palermo: University of Palermo. Fourquet, J. 1938 L 'ordre des elements de la phrase en germanique ancien: etudes de syntaxe de position. (Publications de la Facultö des Lettres de l'Universit6 de Strasbourg 86.) Paris: Les Belles Lettres. Haeberli, Eric and Liliane Haegeman 1995 Clause structure in Old English: Evidence from negative concord. Journal of Linguistics 31:81-108. Halpern, Aaron 1995 On the Placement and Morphology of Clitics. Stanford, CA: CSLI Publications. Healey, Antonette DiPaolo and Richard L. Venezky 1985 A Microfiche Concordance to Old English: The List of Texts and Index of Editions. Toronto: The Dictionary of Old English Project. First edition Toronto [1980], van Kemenade, Ans 1987 Syntactic Case and Morphological Case in the History of EnglishDordrecht: Foris. 1997a Topics in Old and Middle English negative sentences. In: Stanislaw Puppel and Raymond Hickey (eds.), Language History and Language Modelling, 293-306. Berlin/New York: Mouton de Gruyter. 1997b Negative-initial sentences in Old and Middle English. In: Jacek Fisiak (ed.), Festschrift for Roger Lass on his Sixtieth Birthday, 91-104. (Studia Anglica Posnaniensia 31.) Poznan: Uniwersytet im. Adama Mickiewicza W Poznaniu.

122

Linda van Bergen

Koopman, Willem F. 1996 Topicalization and the V2 rule in Old English. Paper given at the Linguistic Seminar, University of Manchester, 15 October 1996. 1997a Another look at clitics in Old English. Transactions of the Philological Society 95: 73-93. 1997b Topicalisation in Old English and its effects: Some remarks. In: Stanislaw Puppel and Raymond Hickey (eds.), Language History and Language Modelling, 307-21. Berlin/New York: Mouton de Gruyter. Liebermann, F. (ed.) 1903 Die Gesetze der Angelsachsen. Halle: Max Niemeyer. Mitchell, Bruce 1985 Old English Syntax, 2 volumes. Oxford: Clarendon Press. Pintzuk, Susan 1991 Phrase structures in competition: Variation and change in Old English word order. PhD dissertation, Department of Linguistics, University of Pennsylvania. Roth, Wilhelm 1914 Die Wortstellung im Aussage-Hauptsatz angelsächsischer Orginalprosa (Annalen 800-900, 1066-1154.) (Kapitel A-E.). PhD dissertation, Friedrich-Wilhelms-Universität, Berlin. Schröer, Arnold (ed.) 1885—1888 Die angelsächsischen Prosa Bearbeitungen der Benediktinerregel. (Bibliothek der angelsächsichen Prosa 2.) Kassel: Georg H. Wigand.

1.2. Form andfunction

Coordinate deletion, directionality and underlying structure in Old English Rodrigo Perez Lorido

1.

Introduction

Though it can be said that the topic of Old English word order has been tackled from almost every conceivable linguistic perspective in the last few years, it is surprising that so far virtually no attempt has been made at analysing it from the point of view of coordination and coordination reduction. The earliest attempts in this direction can be traced back to Ross's (1970) Directionality Constraint, a very interesting and insightful principle which analysed the interplay between underlying order and the direction in which elements are deleted in coordinate structures. This principle, despite having been formulated so long ago, is in my view still a valid one and may certainly provide new evidence to help elucidate some of the hitherto unsolved questions about the grammatical nature of Old English, such as its underlying syntactic structure (verb-final, verb-medial or both), the extent to which word order in Old English obeys restrictions of a pragmatic nature, and the extent to which it depends on psycholinguistic restrictions on parsing and language processing. The present paper aims, therefore, at offering an analysis of word order in Old English inspired by Ross's Directionality Constraint, presenting statistical data for the incidence of forward and backward deletion in coordinate structures in Old English, and comparing the Old English facts with those from other West-Germanic languages like Modern German and Modern Dutch. The study focuses on verbal deletion (Gapping) and its database comprises eight complete texts from the early and late Old English periods. The theoretical approach adopted here is a deletion-based one, according to which any coordinate construction with an incomplete conjunct, like (la) below, is considered as a derived structure, which starts from two underlying complete sentences (lb), in which the identical elements have undergone a process of deletion under identity.1

126 (1)

Rodrigo Perez Lorido a. John lives in Canada and Mary 0 in Florida. b. John lives in Canada and Mary lives in Florida.

For reasons of time and space, I have restricted the scope of my analysis to the type of construction exemplified in (1) (known as Gapping), in which it is assumed that the rule deletes the verb, and potentially some of its complements, in a clause conjoined to another clause containing the same verb in the underlying structure. Another important feature of Gapping which distinguishes it from other deletion rules is that the subject in the second conjunct is always overt.

2. Theoretical background: Ross's Directionality Constraint Gapping is a single rule which operates both forward and backward. The direction of Gapping depends on the input phrase structure configuration: forward if the identical elements are on left branches, backward if they are on right branches (Ross 1970: 251) As we can see in the quotation above, Ross's formulation of the Directionality Constraint attaches special importance to syntactic configurations and to the position of the different elements in the sentence, since in his opinion they determine the direction in which deletion in coordination takes place. According to him, the direction in which deletion occurs is opposite to the branching direction of the elements involved in coordination: if the identical elements are on left branches, deletion operates forward (that is, the first conjunct is complete and the second incomplete); if the identical elements are on right branches, deletion operates backward (first conjunct incomplete, second complete). Therefore, if we consider the tree diagram in figure 1, which corresponds to a language with Subject-Verb-Object word order, like Modern English or Modern Spanish, in which both subject and verb occupy a position in a left branch, and the object occupies a position in a right branch, the only possible deletion outputs, according to Ross (1970), would be the ones indicated in (2a-c). In (2a) deletion of the verb can only take place in the second conjunct (forward), as it is on a left branch. The subject in (2b) is also deleted forward as it also occupies a place in a left branch. Finally, the

Coordinate deletion, directionality and underlying structure

127

object, which occupies a position in a right branch, can only delete backwards, leaving the first conjunct incomplete. This final operation is also known as Right Node Raising, and for some time it was considered a different type of rule (of movement rather than deletion), constituting the major objection to Ross's two-way deletion hypothesis.2

NP (Subject) NP (Object) Figure 1. Subject-Verb-Object order (2)

a. SVO+SVO - » SVO + SO *SO + SOV

b. SVO+SVO

SVO + VO *VO + SVO

c. SVO+SVO

SV + SVO *SVO + SV

John washed the walls and Susan the windows. *John the walls and Susan washed the windows. John some *Ate some

ate a sandwich and drank milk. a sandwich and John drank milk.

John washed and Susan wiped the dishes. *John washed the dishes and Susan wiped.

In spite of its many merits, however, Ross's formulation of the Directionality Constraint contained a number of notable flaws and erroneous assumptions. One of them was that deletion was unordered with respect to movement rules (it was, in his words, an anywhere rule that could apply after or before movement rules), which considerably limits the possibility of making any sort of inference about underlying structure, and, on the other hand,

128

Rodrigo Pirez Lorido

diminishes the explanatory power of the principle. In this work, I will assume, however, following van Oirsouw (1987: 252), that the rule of coordinate deletion is a "late" rule, which applies post-cyclically to surface structures, i.e. after all movement rules have applied, as can be observed in the two examples in (3) in which w/j-movement will precede deletion: (3)

a. Who did John hug and Bill kiss ? b. *John hug and who did Bill kiss?

In addition to this, another major problem for the purpose of this paper concerns the assumed universality of the Directionality Constraint. Ross's work was typologically inspired, and therefore aimed at such universality, but the selection of languages was restricted almost exclusively to IndoEuropean ones. If we extend the database to non-Indo-European languages we find that the Directionality Constraint may exceptionally fail to apply.3 In spite of this fact, I think that, though not a language universal, the Directionality Constraint cannot be considered as absolutely languageparticular. Specifically, western European languages show a striking regularity in the application of the Directionality Constraint (as shown in [2] above), and therefore I consider that a cross-linguistic analysis of directionality features in coordinated constructions among genetically related languages may turn into a valid test for the grammatical status of such languages. This constitutes the major premise in the organisation of this paper, as I base my conclusions on a comparison between the verbal deletion patterns of West-Germanic languages, like Dutch or German, and those found in my corpus of Old English texts, assuming the working hypothesis that — being genetically related — they should present similar behaviour with regard to the application of the Directionality Constraint.

3.

Gapping and directionality in German, Dutch and Old English

3.1. Main clauses As far as Gapping is concerned, the apparent simplicity in the application of the Directionality Constaint shown in (2) for Modern English turns into

Coordinate deletion, directionality and underlying structure

129

something slightly more complex when dealing with languages which allow for more than one surface position for the verb, like Modern German, Modern Dutch or Old English. In these cases one of the verbal positions has to be considered as underlying, and the others as derived by means of movement rules. As a working hypothesis, I assume the most widely accepted view that West-Germanic languages (except Modern English) are underlyingly verb-final, this pattern occurring basically in subordinate clauses, and that other verbal positions (basically in main clauses) are derived via the application of V2 rules. As far as Old English is concerned, I take as a starting point the position in van Kemenade (1987), which assumes an underlying verb-final status for Old English, and the application of V2 rules to account for the great majority of structures with the verb in other positions. Assuming that deletion applies after movement, V2 rules should place the verb in Dutch, German and presumably Old English in positions comparable to those for the Subject and Verb in figure 1 for Modern English, and therefore make it accessible for forward deletion, according to the Directionality Constraint, as shown in (4) and the corresponding examples in (5) and (6). In (5) we have forward deletion of the verb in second position, preceded by the subject, an extremely common pattern, both in German and Dutch (5a) and Old English (5b and c). In (6) we also have forward deletion of the verb in second position, after the topicalisation of another constituent: the adverbs Dann/Dan in the German and Dutch examples (6a), and the adverb Da, the prepositional phrase On pcem gefeohte and the specifiers of the determiner phrase in the genitive pcera worda respectively in the Old English examples (6b-d). (4)

Gapping outputs (main clauses) SVO+SO (X)VSO+SO

(5)

SVO+SO a. Johann isst Fleisch, und Peter 0 Reis. Jan eet vlees, en Peter 0 rijs. 'John eats meat, and Peter rice.'

13 0

Rodrigo Perez Lorido

b. Hire feeder hatte Ioachim & hire moder Anna. 'Her father was called Joachim and her mother Anna.' OECHomii 271.4)4 c. pcet flcesc sodlice gewind ongean pone gast and se gast ongean poet flesc. 'The flesh truly fights against the spirit and the spirit against the flesh.' (iE LS 1.364.6) (6)

(X)VSO+SO a. Dann hat Johann ein Auto, und Peter ein Fahrrad. Dan he eft Jan een auto, en 0 Peter een fiets. then has John a car and Peter a bike b. Da wearp Eustatius upon his horse & his geferan uppon then was Eustatius upon his horse and his followers upon heora. theirs 'Then Eustatius got upon his horse and his followers upon theirs.' (ChronE 172.24 [1048]) c. On pcem gefeohte wees Antigenes ofslagen, & his sunu of In the battle was Antigones slain and his son of pcem rice adrefed. the kingdom driven-out 'Antigones was slain in the battle, and his son driven out of the kingdom.' (Or 81.25) d. ... peera worda wceron dreo on anre tabelan awritene & ... of-the words were three on one tablet written and seofon on dcere odre. seven on the other 'Three of the words were written on one tablet and seven on the other.' OECHom ii 114.137)

Coordinate deletion, directionality and underlying structure

131

All the previous examples belong to main clauses, in which the verb is in non-final position after the application of movement rules, and therefore in an accessible site for forward deletion according to the Directionality Constraint. On the other hand, subordinate clauses in German and Dutch quite systematically present the verb in final position, a position which is considered underlying in the standard analyses, and which, according to the Directionality Constraint, should permit only backward deletion. That this position is an accessible site for backward deletion in Germanic languages, regardless of which element stands in it, is demonstrated by the fact that post-verbal objects in main sentences in German and Dutch delete backward, as shown in (7).5 This process is extremely rare in Old English, and in fact example (8) is the only genuine occurrence of backward object deletion in the whole of the corpus of Old English texts examined: (7)

SV+SVO Johann kauft 0, und Peter verkauft Käse. Jan koopt 0, en Peter verkoopt kaas. 'John buys, and Peter sells cheese.'

(8)

SV+SVO He onfeng & se cyning him sealde feower mcessepreostas. He obtained and the king him gave four mass-priests 'He obtained and the king gave him four mass-priests.' (Bede 222.6)

3.2. Subordinate

clauses

Assuming that the right-peripheral position is an accessible site for backward deletion, verbs should delete backwards in German and Dutch subordinate clauses in agreement with the Directionality Constraint, as indeed happens (example [9]):

132 (9)

Rodrigo Pirez Lorido SO+SOV

Ich glaube, dass Johann Fleisch 0, und Peter Reis isst. Ik geloof dat Jan vlees 0, en Peter rijs eet. I

believe that John

meat

and Peter rice eats

Nevertheless, (9) is not the only possible deletion output for verbs in subordinate clauses in German and Dutch. It is a well known fact that verbs may delete both forward and backward in conjoined subordinate sentences in these languages, as shown in (11) and (12) below, and that there is no difference in acceptability or naturalness between forward-deleted or backward-deleted verbs in such contexts, as van Oirsouw (1987: 260) points out. The reason why the verb may delete forward as well as backward in Germanic languages may be found in some restrictions on natural language processing of the type observed by Frazier (1979, 1985), Reinhart (1983) and Hawkins (1986, 1994) among others: due to limitations on the human capacity of short-term memory, the processing of a linguistic input involves a mechanism of establishing closure of units as soon as possible ("closure" understood as the correct interpretation of a full clause or perception unit). Since the parsing process proceeds linearly (from left to right), perceptual complexity of sentences may arise if the closure is delayed (as in the case of backward deletion or centre-embedding structures — see Kuno 1974), or if a closed unit must be reopened (as in the case of "garden path" sentences [Reinhardt 1983: 208]). 6 As far as verbal deletion is concerned, forward Gapping can be considered as less complex than backward Gapping from the point of view of language processing and perception, as closure of the whole coordinate construction is achieved sooner in the former than in the latter. In forward-gapped structures the verb is present in the first conjunct, from where it is retrieved to induce the correct semantic interpretation of the second conjunct, following the linear left-to-right direction of parsing and, hence, demanding very little of short term memory to provide closure of the whole coordinate structure. In backward-gapped structures, however, the verb is processed last of all, so it is not until the whole coordinate structure is parsed that the correct subject-predicate relationship can be established in the first conjunct. This requires a certain amount of back-tracking on the part of the parser and delays the closure of the whole coordinate

Coordinate deletion, directionality and underlying structure

133

construction, which increases perceptual difficulty, and which explains the usual preference for forward over backward deletion on the part of most speakers, when both syntactic structures are available to express the same idea.7 (10)

Gapping outputs (subordinate clauses) SO+SOV SOV+SO

(11) a. Ich glaube, dass Johann Fleisch 0, und Peter Reis isst. I believe that John meat and Peter rice eats b. Ich glaube, dass Johann Fleisch isst, und Peter Reis 0. I believe that John meat eats and Peter rice (12) a. Ik geloof dat Jan vlees 0, en Peter rijst eet. I believe that John meat and Peter rice eats b. Ik geloof dat Jan vlees eet, en Peter rijs 0 I believe that John meat eats and Peter rice These processing strategies clearly overlap (and sometimes are in clear opposition) with the purely syntactic and structural conditions for deletion mentioned in the previous section (the Directionality Constraint), which depend on syntactic configurations and ultimately on the typological status of the language at stake. We might say, therefore, that German and Dutch display backward deletion of the verb in subordinate sentences (in agreement with the Directionality Constraint) simply as a result of being typologically Subject-Object-Verb, and forward deletion due to the restrictions on processing mentioned above. Bearing all those assumptions in mind, the basic strategy I have pursued in this paper consists of examining all the instances of verbal deletion in Old English conjoined subordinate sentences in the corpus (in theory the only ones capable of undergoing backward deletion according to the Directionality Constraint if Old English is considered a Subject-ObjectVerb language), checking the incidence of both forward and backward deletion, and comparing the results with those for German and Dutch. If

134

Rodrigo Pirez Lorido

Old English behaved exactly like those languages, we should expect a more or less even incidence of forward and backward deleted verbs in conjoined subordinate sentences. The more it leaned towards backward deletion the more it would reflect the typological Subject-Object-Verb status of Germanic languages, and the more it would lean towards forward deletion the more it would reflect the importance of processing and parsing mechanisms in Old English word order.

4. The data 4.1. Verbal deletion The corpus I have examined consists of eight complete texts from both the early and the late Old English periods (arranged more or less chronologically in [13]): (13)

early OE: late OE:

Bede (Bede); Orosius (Or); Boethius (Boeth); Pastoral Care (CP); Parker Chronicle (ChronA); Peterborough Chronicle (ChronE); ^ilfric Catholic Homilies-Second Series (^ECHom ii); ^ l f r i c Lives of Saints (JELS 1 & 2); Wulfstan Homilies (WHom).

The selection of relevant examples comprises all the instances of Gapping in conjoined subordinate clauses found in them. Examples of this type are really scarce,8 to the extent that in all the previous studies on directionality in coordinate deletion in Old English that I know of, they have been overlooked. This renders the conclusions in works like Smith (1971), Mitchell (1972) or Taylor (1973) irrelevant, as the authors hastily conclude that Old English is a Subject-Verb-Object language from the simple verification of forward Gapping in main sentences of the type in (5) and (6) above.9 A total of 51 conjoined subordinate sentences with deleted verbs have been found in the corpus,10 out of a total of 265 Gapping instances, which amounts to 18 per cent of the total. Of these, all the examples display forward deletion, as illustrated in examples (15) to (25):

Coordinate deletion, directionality and underlying structure

(14)

135

Total conjoined clauses: 265 Conjoined subordinate clauses: 51 (18.8%) Forward deletion in subordinate conjoined clauses = 100% Backward deletion in subordinate conjoined clauses = 0%

First conjunct verb-final: (15)

hit is nieddearf öcet mon his hlaford ondrcede, it is necessary that man his lord

& se cneoht

fear [SUBJ] and the servant

his magister. his master 'It is necessary for a man to fear his lord, and the servant his master.' (CP 109.12) (16)

...for py da godcm p[cet\ god on riht secap & pa yfelan on woh ... because the good that good in right seek and the wicked in wrong '... because the good seek good rightly, and the wicked wrongly' (Boeth 178.6)

(17)

Da pet gafol gelest wees & pa frid adas gesworene, when the tribute paid was and the peace oaths sworn pa toferde se here... then marched the army ... 'When the tribute was paid and the peace oaths sworn, then the army marched...' (ChronE 143.4 [1012])

(18)

Langsum is to reccenne ... hu cristen se cyning weard & hu tedious is to narrate ... how Christian the king was and how eawfeest seo cedele cwen & hu estfull ealpees cyninges folc pious the noble queen and how devout all of-the-king people 'It is tedious to narrate ... how Christian the king was, and how pious the noble queen, and how devout all the people of the king.' OECHom ii 277.43)

13 6

Rodrigo Perez Lorido

(19)

... peer peer hine ...where

nan manne can

ne he

him [ACC] no-one not-know nor he

ncenne mon. no-one man '... no man knows him nor he any man.' (Boeth 98.21) First conjunct verb-medial: (20)

... od poet his feax weox swa swa wimmanna & his neglas ... until his hair grew like of-women and his nails swa swa earnes clawa. like of-eagle claws '... until his hair grew like women's, and his nails like an eagle's claws.' OECHom ii 253.116)

(21)

...forpam pepara leoda peawas sind swipe ungelica & ... because of-the nations manners are very unlike and heora gesetnessa swide mislica. their institutions very various '... because the manners of the nations are very unlike, and their institutions manifold.' (Boeth 64.21)

(22)

God bid mannum dcet celc hcebbe

his agen wif, & celc

good is for-men that each have [SUBJ] his own wife and each

wif hire ceorl, wife her husband 'It is good for men that each have his own wife, and each wife her husband.' (CP 397.18) (23)

...for dan de ealle his weorc sind sode & his wegas rihtwise. '...because all his deeds are true, and his actions righteous.' OECHomii 253.130)

Coordinate deletion, directionality and underlying structure

137

(24)

Fordi synd laga gescette ... pcet pa riht-wisan beon therefore are laws fixed ... that the righteous be [SUBJ] ge-herode and pa unriht-wisan gehynde. praised and the unrighteous shamed 'Therefore laws are fixed that the righteous may be praised and the unrighteous shamed.' 04ELS 1.134.282)

(25)

Hie ne wendon dcette cefre menn sceolden swce reccelease they not think that ever men should so careless weordan ond sio lar swce odfeallan. become and the learning so decay 'They didn't think that men would ever become so careless, and learning decay so (much).' (CP 5.22)

A first conclusion that can be drawn from this overwhelming statistical fact is that word order in Old English depends quite considerably on the restrictions on language processing mentioned in section 3.2, and that those restrictions interact quite heavily with syntactic, structure-based constraints, certainly more than they do in German and Dutch.11 On the other hand we must bear in mind that the absence of backward deletion represents in theory no formal counter-argument to the Subject-Object-Verb hypothesis for Old English from a comparative point of view, since German and Dutch display both possibilities for deletion. If we turn now to word order in the first conjunct of the structures we are dealing with, we will observe (cf. tables 1, 2 and 3) that the tendency is for the finite verb to appear in medial position rather than in final position: we find in the selection 40 instances of verb-medial structures in the first conjunct (78 per cent of the total) against 11 instances of verb-final ones (21.5 per cent of the total). This is relevant since, given the assumption that deletion follows movement, patterns showing verb-mediality in the first conjunct due to postposition rules within the verb phrase such as extraposition, verb-raising or verb-projection raising12 should not be treated differently from other main sentences as regards the applicability of the Directionality Constraint, which would reduce the number of relevant examples to 11 out of the whole corpus.

13 8

Rodrigo Pirez Lorido

Table /.Verbal deletion patterns in conjoined subordinate sentences Number of instances Word Order SOV + SO 8 36 SVO + SO 4 OVS + OS 2 OSV + OS 1 OSV +SO 51 Total Table 2. Distribution of verbal deletion patterns in the different works in the corpus SOV + SO SVO + SO OVS + OS OSV + OS OSV + SO 2 Bede 2 1 2 Or 14 2 1 Boeth 4 1 2 CP 1 ChronA -

-

-

-

ChronE jECHom ii ^LS WHom Total

1 -

8

_ 6 6 4 36

-

-

_ 1

_ 1

-

-

-

-

4

2

1

Table 3. Verbal position in the first conjunct of coordinate elliptical clauses in early and late Old English Verb-final Verb-medial Total 11 (21.5%) 40 (78.4%) 9(28.1%) 23 (71.8%) Early Old English 2 (10.5%) 17 (89.4%) Late Old English

From a diachronic perspective, the percentages of verb-finality decrease as we move from early Old English (Bede, the Orosius, Boethius, the Pastoral Care and the Parker Chronicle) with 28.1 per cent to late Old English {Peterborough Chronicle, ^Elfric's works, Wulfstan's Homilies) with 10.5 per cent, moving in the opposite direction to verb-mediality, which increases as we advance in time: 71.8 per cent in early Old English versus 89.4 per cent in late Old English, as we can see in figure 2.

Coordinate deletion, directionality and underlying structure

100%

139

r

! 80% L 60% i-

40% h

20% 0%

-

Early Old English

Late Old English

Figure 2. Incidence of V-finality vs. V-mediality in the first conjunct of conjoined elliptical subordinate clauses over time I am unsure about the implications of this for the chronology of the change Subject-Object-Verb —> Subject-Verb-Object. Apparently the figures seem to suggest that there is a steady movement in the direction of the development Subject-Object-Verb —> Subject-Verb-Object well within the Old English period, but the figures diverge quite considerably if we compare them with other studies in which different stages of Old English are analysed with regard to the position of the finite verb, like Stockwell and Minkova (1990), so for the time being I confine myself to stating the evidence, and leave the discussion of the problems it may pose for further research.

4.2. Non-deletion of identical verbs In this section I would like to present some facts very relevant to the discussion at stake, which have — in my view — been persistently overlooked in the studies of coordinate deletion in Old English in particular: I am referring to the possibility that identical verbs, instead of being deleted — whether forward or backward — are repeated in the contexts under analysis, as in the examples (26) to (32) below: 13

140

Rodrigo Perez Lorido

Both conjuncts verb-final: (26)

Ealle stcerwriteras secgad pcet Asiria rice cet Ninuse all historians say that of-Assyrians kingdom at Ninus begunne & Romana rice cet Procose begunne. began and of-Romans kingdom at Procas began 'All historians say that the kingdom of the Assyrians began at Ninus, and the kingdom of the Romans at Procas.' (Or 37.5)

(27)

Fordon pa hio ofslcegne weran, & hiora lichaman in for when they slain were and their bodies into da ea worpene wceran, da gelomp hit pcet... the river cast were then happened it that... 'For when they were slain, and their bodies cast into the river, then it happened that...' (Bede 416.31)

(28)

Nu pu donne wast hwcet da leasan gescelpa sint & now you therefore know what the false goods are and hwcet pa sopan gescelpa sint. what the true goods are 'Therefore, you know now what the false goods and what the true goods are.' (Boeth 126.27)

Both conjuncts verb-medial: (29)

Pa he pa wees piderwardes, & sio operu fierd wees when he then was towards-there and the other army was hamweardes,... homewards ... 'When he was on his way there, and the other army on their way home ...' (ChronA 86.3 [894])

Coordinate deletion, directionality and underlying structure (30)

141

...forpam de done welan lyst anwealdes & done ... because the wealth [ACC] desires of-power and the anwealde lyst weorpscipes & pone weorpscipe lyst power [ACC] desires of-dignity and the dignity [ACC] desires mcerpa. of-glory

'... because wealth desires power, and power dignity, and dignity glory.' (Boeth 124.7) (31)

... dcette on his swidran ... that

winestran wcere left

wcere

lang lif,

& on his

in his right [hand] were [SUBJ] long life and in his

wela

&

wyrdmynt.

were [SUBJ] wealth and honour

'... that in his right hand was long life, and in his left wealth and honour.' (CP 389-16) Asymmetry: Subject-Object-Verb + Subject-Verb-Object: (32)

... pa pa dunstan iung man wees andse swurd-bora wees ... when Dunstan young man was and the swordbearer was forealdod man. very-old man '... when Dunstan was a young man and the swordbearer was a very old man.' 0ELS2.314.6) 14

Such an unnatural and uneconomical retention of a redundant element in communication is a highly relevant indicator of the different strategies in language processing and the mechanisms involved in the derivations from underlying to surface structure. Therefore it is worth examining in detail. Consequently, in addition to the data about coordinate reduction presented in the previous section, I would like to analyse here the statistical facts corresponding to the repetition of identical verbs in conjoined subordinate clauses, taking into account — as I did in the previous section — different word orders (cf. table 4) as well as diachronic variation (cf. table 5):

142

Rodrigo Perez Lorido

Table 4. Verbal repetition in conjoined subordinate sentences according to word order in both conjuncts Number of instances Word Order 3 SOV + SOV 4 SVO + SVO 1 OVS + OVS 3 OSV + OSV 1 SOV + SVO Total: 12 Table 5. Distribution of word order patterns in the different works in the corpus SOV + SVO + OVS + OSV + SOV + SOV SVO OVS OSV SVO Bede 1 Or 1 2 2 Boeth 1 1 1 CP 1 ChronA ChronE jECHom ii /ELS WHom Total

1

3

4

1

3

1

If we now look closely at the data in table 6 below, we will notice that the percentage of repeated verbs in conjoined subordinate sentences (19 per cent) is far too high to be simply attributed to slips of the pen or scribal error. Table 6. Incidence of verbal deletion versus verbal repetition in conjoined subordinate clauses in early and late Old English. Verbal Total Verbal % repeated deletion repetition verb clauses 12 19 Total 63 51 10 32 23.8 Early Old English 42 2 21 19 9.5 Late Old English If we concentrate on the distribution of the examples according to the position of the verb in the sentence (cf. table 7) we will observe that — leaving aside example (32) in which the position of the repeated verbs in the

Coordinate deletion, directionality and underlying structure

143

coordinated construction is asymmetrical — the identical verbs are repeated in subordinate sentences almost four times as much in constructions in which the verb is in final position (38.8 per cent) as in constructions with Subject-Verb-Object order (11.1 per cent). Table 7. Distribution of verbal deletion versus verbal repetition according to verbal position Total Deleted verb Repeated verb % repeated verb 18 11 Verb-final 7 38.8 Verb-medial 45 40 5 11.1 From a diachronic point of view, repetition of the verb decreases as we advance in time (23.8 per cent of repeated verbs in early Old English versus 9.5 per cent in late Old English), and in very close proportion to the movement from verb-finality to verb-mediality in the first conjunct of examples with deleted verbs (compare the figures in table 6 and table 3). Summarising, identical verbs in conjoined subordinate sentences are repeated predominantly in clauses with verb-final order in the first conjunct, and as the change verb-final —> verb-medial sets in, the number of repeated verbs in such constructions decreases proportionally.

5.

Conclusion

The analysis of the data about coordinate verbal deletion in Old English presented so far has shown that identical verbs in conjoined subordinate clauses delete quite systematically forward. This fact seems to run counter to a strict application of Ross's (1970) Directionality Constraint, which states that verbs in final position (the place where they normally occur in subordinate clauses in Old English) should always delete backward. From a strictly formal point of view, however, I don't think that the absence of backward-gapped instances in the corpus should constitute any sort of counter-evidence to the Subject-Object-Verb hypothesis for Old English, since other West-Germanic languages like German or Dutch allow for both forward and backward deletion of verbs in conjoined subordinate clauses. The explanation for the zero incidence of backward Gapping in the corpus might lie in the fact that, besides syntactic, structurally-based restrictions,

144

Rodrigo Perez Lorido

Old English word order is also sensitive to constraints of a psycholinguistic nature, such as the restrictions on processing mentioned in section 3.2 (which can be summarised in the preference for syntactic structures in which closure of the processing unit can be established as soon as possible), or at least that Old English is more sensitive to these restrictions than German and Dutch are. On the other hand, a number of instances have been found in the corpus in which identical verbs are lexically realised in both conjuncts in conjoined subordinate clauses instead of having undergone deletion. In my view, the repetition of identical verbs in the contexts under analysis can be seen as a compromise solution between backward deletion and forward deletion, which is eventually resolved by non-deletion, thus reflecting the conflict between the strictly syntactic constraints on deletion imposed by the Directionality Constraint assuming an underlying Subject-Object-Verb status for Old English (backward deletion) and the preference for forward deletion imposed by the restrictions on processing mentioned before. Finally, since both forward and backward verbal deletion are possible patterns in most Subject-Object-Verb languages, but backward verbal deletion is not possible in Subject-Verb-Object ones, the presence of the verb in the second conjunct (though present in the first too) in a considerable number of examples in the corpus can be considered as an indicator of verb-finality rather than of verb-medial ity, thus constituting an argument (subtle though it may be), in favour of the Subject-Object-Verb hypothesis for Old English. This is confirmed, in my view, by the proportional decline of verbal repetition as Subject-Verb-Object order displaces Subject-Object-Verb order in subordinate sentences over time.15

Appendix /ECHom ii

JELS

JElfric's Catholic Homilies. The Second Series. Edited by Malcolm Godden. (Early English Text Society, Supplementary Series 5.) London: Oxford University Press [1979], JElfric's Lives of Saints, Two Volumes. Edited by Walter W. Skeat. London: Oxford University Press [1966]. First published in: Early English Text Society, Original Series 76, 82, 94 , 114. London: N. Trübner [1881-1900],

Coordinate deletion, directionality and underlying structure Bede

Boeth

Chron A/E

ChronC

CP

Or

WHom

145

The Old English Version of Bede 's Ecclesiastical History of the English People. Two Volumes. Edited by Thomas Miller. Millwood, New York: Kraus Reprint [1978]. First published in: Early English Text Society Original Series 95, 96, 110, 111. London: N. Trübner [1890-1898]. King Alfred's Anglo-Saxon Version of Boethius' De Consolatione Philosophiae. Edited by Samuel Fox. New York: AMS Press [1970]. First published London: G. Bell [1864], Two of the Saxon Chronicles Parallel: Based on an Edition by John Earle. Edited by Charles Plummer. Oxford: Clarendon Press [1892], The Anglo-Saxon Chronicle, According to Several Original Authorities. Edited by Benjamin Thorpe. London: Longman, Green, Longman and Roberts [1861]. King Alfred's West-Saxon Version of Gregory's Pastoral Care Edited by Henry Sweet. Millwood, New York: Kraus Reprint [1988]. First published in: Early English Text Society Original Series 45, 50. London: N. Trübner [1871-1872], The Old English Orosius. Edited by Janet Bately. (Early English Text Society, Supplementary Series 6.) London: Oxford University Press [1980], The Homilies ofWulfstan. Edited by Dorothy Bethurum. Oxford: Clarendon Press [1957],

Notes 1. For a very good discussion of the notion "identity" within the deletion-based approach see van Oirsouw (1987: 178-220), and for an interesting analysis of the conditions for "identity" in Old English subject deletion see Allen (1986,1995). 2. Two of the classic analyses in this respect are Maling (1972) and Hankamer (1979: 47-54). In the last few years, however, very solid arguments have been provided by authors like McCawley (1982), Levine (1985) and more recently Kayne (1994: 67, 146-147 footnote) against a non-deletion analysis of structures like (2c) above. 3. Consider, for example, the examples in (ia) and (ib), from Sanders (1977: 246), from Zapotec and Tojolabal Maya: (i) a. Zapotec (SO + SVO) xwain 0 jumE, ne makU been yuu. Juan 0 a-basket and Marcos made a-house 'Juan made a basket, and Marcos a house'

146

Rodrigo Perez Lorido b. Tojolabal Maya (SO +VSO) 0 b 'ak 'et Hwan, sok yi 'a tek 'ul Manwel. 0 meat Juan and took fruit Manuel 'Juan took (some) meat and Manuel (some) fruit'

4. References to Old English texts are made by page and line number in the authors' numberings in the editions mentioned in the Appendix. Additional figures in square brackets in the Anglo-Saxon Chronicle refer to year entry. 5. The same holds for any other languages with Subject-Verb-Object order, like Modern English or Spanish, in which the only possible backward deletion output is object deletion: (i)

Juan Ιανό 0 y Maria seco los platos. John washed 0 and Mary wiped the dishes

6. For example, if we consider the famous example the horse raced past the barn fell, the tendency to establish closure as soon as possible will dictate processing the horse raced past the barn as a clause, but, as fell is reached, the closed clause must be reopened and reanalysed. 7. A clear example of this is the fact that in some languages exhibiting backward object deletion, such as Spanish, there is a preference for repetition of the identical object in the second conjunct — as a clitic —, as in (ib), rather than deleting the object backwards, as illustrated in (ia) below: (i) a. Juan Ιανό 0 y Maria seco los platos. John washed 0 and Mary wiped the dishes 'John washed and Mary wiped the dishes.' b. Juan Ιανό los platos y Maria los seco. John washed the dishes and Mary them wiped 'John washed the dishes and Mary wiped them.' 8. It must be admitted that instances of conjoined subordinate sentences with identical verbs are not among the commonest in everyday speech, so it is not surprising that examples of those constructions in the Old English literature are scarce. This is the reason why the texts have been examined in their entirety, and not just a mere share of them. On the other hand, given that the absolute incidence of the construction under analysis is so low, variation in one or two examples in statistics may render considerable differences in total percentages. 9. These works are all early transformational ones carried out in the early seventies, when there was probably too little emphasis on data, but it is also true that two of them (Mitchell 1972 and Taylor 1973) use as databases the

Coordinate deletion, directionality and underlying structure

147

Anglo-Saxon Chronicle, which — due to its narrative style — contains fewer instances of subordination than the other texts. In the case of Smith (1971), the author's database comprises ten manuscripts copied in the first half of the eleventh century, and the following Old English gloss of Latin is the only instance of Gapping quoted by the author from her corpus: (i)

/Er is lyft. Ignis fyr. Terra eorde. Aqua water. '"iEr" is air, "ignis" fire, "terra" earth, "aqua" water.' (^lfric. De Temporibus Anni)

10. In the selection and presentation of the data the following conventions have been adopted: • "Object" stands not just for "direct object" but for any element of the verbal predicate. • Only coordinated constructions with and and ne have been considered. • Auxiliary verbs have been put together with non-auxiliary ones, the order "Object-Verb" representing the order non-finite/finite verb. • Examples of Gapping consisting of repetitions of a model, enumerations, or lists of items, like the protracted example in (i) below from the Cura Pastoralis, have been counted as just one. (i) XXIIII Dcette on odre wisan sint to manianne weras, on odre wif; XXV Doette on odre wisan sint to manianne da iungan, on odre da ealdan ; XXVI Dcette on odre wisan sint to manianne da welegan, on odre da wcedlan; 'XXIIII that in one way are to be admonished men, in another women; XXV that in one way are to be admonished the young, in another the old; XXVI that in one way are to be admonished the rich, in another the poor.' (CP 179.14) 11. Another way of looking at the striking fact that verbs are never deleted backwards in subordinated clauses in Old English would be to analyse it from the point of view of pragmatics and pragmatic word order, if we assume that, by deleting identical verbs only forward, the order of the elements in the sentence would conform to the Old English tendency to present linguistic material with regard to their degree of informativeness in a given-new perspective, as shown in Kohonen (1978: 197-202) and Perez Lorido (1996), among others. The line of inference would be more or less as follows: Functional Sentence Perspective principles require constituents that are deleted to be contextually known, while constituents left behind necessarily represent new information. This obviously favours forward deletion, since in backward deletion the material erased can never be contextually known, because it is the

14 8

Rodrigo Pirez Lorido

antecedent occurrence of the identical material, and not the subsequent occurrence. Such an argument seems to me at best a bit convoluted, and I find the explanation in terms of processing mechanisms much more natural. There is nothing, on the other hand, that prevents different factors from simultaneously playing a role in the generation of Old English sentences, and I fully agree with Denison (1986: 293) when he says that Old English word order must be seen as the interaction of those different factors, and that perhaps the best way to tackle it is a multi-factorial approach which takes into account not only syntactic and structural aspects, but also discourse-based, and even psycholinguistic ones. 12. See van Kemenade (1987: 39-41, 55-63) and Pintzuk (1991: 157-177) for a discussion of movement rules within the verb phrase in Old English. 13. Examples from the corpus have been filtered, disregarding any instances in which the repetition of the verb can be justified on the basis of stylistic or rhetorical needs, even though I am aware that this is a highly subjective criterion. 14. The asymmetry in the position of the two identical verbs in this example is explained by Stockwell and Minkova (1990: 503) on account of a similar possibility in colloquial German, but I would like to point out that, strictly from the point of view of the application of the Directionality Constraint, it might well indicate the state of competing underlying phrase structures (infl-medial and infl-final) available synchronically in Old English in the model proposed by Pintzuk (1991). Nevertheless, given the very low incidence of asymmetrical structures like (32) in the corpus, I have to reject the idea that coordinate reduced patterns in Old English should constitute an argument for Pintzuk's Double Base hypothesis. 15. In (i) below I propose a possible scenario for the different ways in which verbfinality may be reflected in Gapping patterns, from more to less marked, namely: (a) just backward deletion of the verb, (b) coexistence of backward and forward deletion, (c) non-deletion of the identical verbs, and finally (d) just forward deletion. (i) a. SO + SOV b. SO + SOV / SOV + SO c. SOV + SOV d. SOV + SO Out of these four options, Old English displays (c) and (d). An interesting example illustrating this situation is (ii) below from the Anglo-Saxon Chronicle, in which one entry from MS Ε (example [17] above), which presents forward deletion of the verb, appears in MS C as a coordinated construction in which both identical verbs are repeated at the end of each conjunct:

Coordinate deletion, directionality and underlying structure

149

Da fret gafol gelest wees & pa frid adas gesworene, pa When the tribute paid was and the peace oaths sworn then toferde se here ... dispersed the army ... 'When the tribute was paid and the oaths of peace sworn, the army dispersed ...' (ChronE 143.4 [1012]) Da p[cet] gafol gelcest wees & pa frid apas asworene When the tribute paid was and the peace oaths sworn weeronpa toferde se here... were then dispersed the army ... (ChronC 268.38 [1012])

References Allen, Cynthia 1986

Reconsidering the history of like. Journal of Linguistics 22: 375-409. Case Marking and Reanalysis. Oxford: Clarendon Press.

1995 Denison, David 1986 On word order in Old English. Dutch Quarterly Review 16: 277-295. Frazier, L. 1979 Parsing and constraints on word order. University of Massachusetts Occasional Papers in Linguistics 5: 177-198. Syntactic complexity. In: D. Dowty, L. Karttunen and A. Zwicky 1985 (eds.), Natural Language Parsing: Psychological, Computational, and Theoretical Perspectives, 320-357. Cambridge: Cambridge University Press. Hankamer, Jorge 1979 Deletion in Coordinate Structures. New York: Garland. Hawkins, John A. 1986 A Comparative Typology of English and German. London: Croom Helm. 1994 A Performance Theory of Order and Constituency. Cambridge: Cambridge University Press. Kayne, Richard 1994 The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press, van Kemenade, Ans 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris.

150

Rodrigo Perez Lorido

Kohonen, Viljo 1978 Kuno, Susumu 1976 1974

On the Development of English Word Order in Religious Prose around 1000 and 1200 A.D. Abo: Abo Akademi Foundation. Gapping: a functional analysis. Linguistic Inquiry 7: 300-318. The position of relative clauses and conjunctions. Linguistic Inquiry 5: 117-136.

Levine, R.D. 1985 Right node (non)-raising. Linguistic Inquiry 16: 492-497. Maling, Joan M. 1972 On Gapping and the order of constituents. Linguistic Inquiry 3: 101-108.

McCawley, J. 1982

Parentheticals and discontinuous constituent structure. Linguistic Inquiry 13: 91-106. Mitchell, Lawrence 1972 Old English as an SVO language: Evidence from the auxiliary. Papers in Linguistics 5: 183-201. van Oirsouw, Robert 1987 The Syntax of Coordination. London: Croom Helm. Pirez Lorido, Rodrigo 1996 On the grammatical domain of Gapping in Old English: Syntax and pragmatics. Diachronica 13: 319-346. Pintzuk, Susan 1991 Phrase structures in competition: Variation and change in Old English word order. PhD dissertation, Department of Linguistics, University of Pennsylvania, Philadelphia. Reinhart, Tanya 1983 Anaphora and Semantic Interpretation. London: Croom Helm. Ross, John 1970 Gapping and the order of constituents. In: Manfred Bierwisch and Karl Heidolph (eds.), Progress in Linguistics, 249-259. The Hague: Mouton. Sanders, S.A. 1977 A functional typology of elliptical coordinations. In: F.R. Eckman (ed.), Current Themes in Linguistics: Bilingualism, Experimental Linguistics, and Language Typologies, 237-270. Washington: Hemisphere.

Coordinate deletion, directionality and underlying structure

151

Smith, Robin D. 1971 Coordination of linguistic units in selected Old English prose of the early eleventh century. PhD dissertation, University College, London. Stockwell, Robert P. and Donka Minkova 1990 Verb phrase conjunction in Old English. In: Henning Andersen and Konrad Koerner (eds.), Historical Linguistics 1987, 498515. Amsterdam/Philadelphia: John Benjamins. Taylor, Mary Vaiana 1973 Another look at word order in Old English. Edinburgh Working Papers in Linguistics 3: 93-99.

The position of the adjective in Old English Olga Fischer

1. Introduction This study is part of a larger project in which I would like to find out what changes have occurred in the position of the adjective within the noun phrase in the history of English.1 The questions I would like to see answered are: Were there any differences in usage/meaning between prenominal and postnominal adjectives in Old English; were such differences in any way tied up with the traditional distinction between strong and weak adjectives; and, finally, were there any radical changes in this area, or is the present-day position of the adjective (which usually occurs before the noun but occasionally also after) a natural and gradual development of its position in Old English, involving only minor shifts? Concerning the last question, opinions seem to differ on this at present. Consider, for instance, Lightfoot (1979: 168-186), who writes (when discussing the position of quantifiers in Old English, which in his view are indistinguishable from adjectives at this stage) that Old English adjectives were free to occur prenominally, in predeterminer position, postnominally and "floating". 2 Lightfoot concentrates here on the quantifiers, and the radical change he proposes concerns especially these, but it implies a (radical) change for adjectives too, since it was the fact that adjectives could no longer occur in all of the above-mentioned positions that caused the quantifiers (which did remain in those positions) to change into a separate category. However, apart from this, Lightfoot notes in a later chapter (1979: 205-208) that the adjectives, too, underwent a change: OE adjectival modifiers appeared characteristically in prenominal position, particularly for unco-ordinated adjectives and participles. However, in ME postnominal adjectives become increasingly common. In this period most adjectives can occur as pre- and postnominals and the latter appear to represent a productive position, because newly borrowed adjectives are often introduced as postnominals.

154

Olga

Fischer

Lightfoot (1979: 206) believes that this is not due to French influence (as is the common suggestion), first of all because "in contemporary French adjectives normally occurred prenominally", but mostly because he "resist[s] such foreign influence interpretations on ideological grounds: such 'explanations' ... are usually unilluminating". I also do not believe that French played any crucial role in the position of adjectives in Middle English, but it clearly played some role, considering the fact that of the postnominal adjectives used in present-day English, many are French (and often used in set phrases); see also below.3 In spite of that, Lightfoot's suggestion that postnominal adjectives became productive and much more frequent in Middle English due to the basic word order change from Subject-Object-Verb to SubjectVerb-Object seems to me unlikely too. First of all, Greenberg's typology, in which Subject-Verb-Object order would favour noun-adjective order, is based on statistics and does not necessarily entail a causal relation between Subject-Verb-Object and noun-adjective order. More importantly, however, I do not believe that postnominal adjectives did become more productive in Middle English. Of course, in order to be able to compare and to make any factual pronouncements on this, we need to know more about adjective position in both Old and Middle English. It is clear from the above that Lightfoot favours a number of radical changes in this area. His view is strongly contrasted with that of some other linguists, such as, for instance, Karl Brunner, who favours a more "gradual" story, a story with little change: Attributive Adjektive stehen in der ae. Prosa, me. und auch noch heute in der Regel vor dem Substantiv, das sie ergänzen. [Attributive adjectives in OE prose, in ME, and also still today, stand as a rule before the noun which they complement], (Brunner 1962: 69 [translation mine])

And in what follows, Brunner continues to describe under what circumstances postposition still occurs in present-day English. First, postposition is used in order to highlight the adjective. Here he remarks that the Old English postnominal adjective already had this function. Second, it is found when the adjective is accompanied by complements that give extra information about the adjective (as in all the implements necessary for salmon fishing). At this point he does not explicitly refer to Old English, but it will be clear from the discussion below that this is also a common structure in Old

The position of the adjective in Old English

155

English. Thirdly, postposition occurs with a number of Anglo-French legal expressions such as heir apparent, princess royal, and a number of nachgebildete expressions such as art military, love eternal, etc. In other words, this scenario suggests that there may not have been an awful lot of change between Old and Middle English as far as the adjective position is concerned, apart from a very limited area of fossilised expressions. Still, most Old English grammar books have a separate section that deals with the special position of the adjective in Old English, in particular when more than one adjective is used (cf. Mosse 1946: 196, Davis 1953: 59, Brook 1955: 82, Quirk and Wrenn 1955: 88-89). The reason given for this variety in word order is usually that "[t]he order of words is less rigid in Old English than in Modern English because the Old English inflectional system, much fuller than that of Modern English, made it possible for a writer to make clear the relation of a word to the rest of the sentence without making use of word order for this purpose" (Brook 1955: 82). Although Quirk and Wrenn especially give more details about the kind and number of adjectives that appear postnominally, an explanation for the position itself in syntactic, semantic or any other terms is not given in any of these grammars. When I looked into the bible of Old English syntax (Mitchell 1985), I felt satisfied and reassured that indeed there is "room for more work on the arrangements when two attributive adjectives qualify the same noun" (Mitchell 1985: §173). This is what I intend to do in this study, starting with Old English. On the basis of a tagged Old English corpus,4 I will try and begin to find out what differences there are in adjective position in Old English and what semantic, pragmatic and/or syntactic factors lie at the basis of these differences. I hope that a closer study of the order of elements within the noun phrase will lead to more insight into the rules that may govern it, just as a more detailed study of Old English clausal word order has proven to be very fruitful, leading to the conclusion that Old English word order is not as free as was for a long time supposed (see above).

2. Adjective position: some general comments Before turning to Old English, I would like to have a look at what may determine the order of adjective and noun in a general way and what

156

Olga Fischer

meaning differences may be attached to differences in position of the adjective. In Dutch, as in English, adjectives do not normally follow the noun; prenominal position is the rule. Still, we do find occasional examples where the adjective follows. A fairly recent one is the expression een broodje gezond 'a roll (of bread) healthy'. Since this is such a new and still fresh expression, it may be fruitful to ask what the difference is between this and the normal order, een gezond broodje 'a healthy roll', because this difference may still be alive (in the sense of not yet conventionalised) to the native speaker. For most Dutch speakers that I questioned, there are two differences. First of all, gezond could be somehow adverbial; it can be replaced by the prepositional phrase met gezonde ingredienten 'with healthy ingredients' (lettuce, tomato, etc.). So gezond may have started off as a kind of abbreviated form of such a prepositional phrase. At the same time, however, it was felt that the postnominal position also gave the word gezond a kind of emphasis; it was much more striking than een gezond broodje. In fact, it expresses a contrast between rolls that are the norm and not so good for you (in these days of slimming and health-awareness) and the one that is different because it is good for you. It was also suggested that een broodje gezond is patterned on een broodje kaas 'a roll cheese', but here the postnominal word is itself a noun. However, it has to be noticed that the opposite order, een kaasbroodje 'a cheese roll', is indeed a totally different matter. The first is a roll, cut open, with slices of cheese put in between, the second is a kind of pastry which has been baked with cheese in it, usually served hot. In other words, the postposed noun again suggests a prepositional phrase ('a roll with cheese'), which is not true for the preposed noun ('a cheesy roll'). So we can conclude from this single example that the order makes a difference, that somehow the adjective before the noun changes the noun in question (the 'roll' is no longer an ordinary 'roll'), while the adjective after the noun provides extra information about the noun, and possibly gives it extra emphasis. Whether the extra emphasis is due to the postposition itself or to the unnaturalness of this position in Dutch remains to be seen. Melita Stavrou (1996) has written an article on adjective position in Modern Greek, in which she argues that the postnominal and prenominal adjective positions in Greek convey subtle semantic and pragmatic differences. She follows ideas expressed by Bolinger on attributive and

The position of the adjective in Old English

157

predicative adjectives in English (Bolinger 1967) and on linear modification (Bolinger [1952] 1972). Bolinger's ideas in these two articles are clearly perceptually (I would say iconically) based. Stavrou does not follow up or go deeply into these iconic insights, but she shows that the semantic/pragmatic differences that follow from what Bolinger sees as an iconic or perceptual order also apply to Modern Greek. Bolinger (1972: 31) writes: "the linear geometry of the sentence imposes certain relationships upon the elements that compose it". The principle that he uncovers is perceptual in that whatever comes first in a linear sequence determines to some extent how the next element is to be interpreted (Bolinger 1972: 32). Such a contrast is of course only possible if the elements concerned can occupy more than one position. This is the case with adjectives in Modern Greek and also in Spanish (Bolinger's examples are from Spanish), but much less so in Modern English because adjectives are on the whole confined to prenominal position. Bolinger schematises his idea as follows:

Ν

Ν

Figure 1. Linear modification in noun phrases (after Bolinger 1972) The diagram indicates that the element that comes first (adjective or noun) modifies the rest of the phrase, so in Spanish un hermoso edificio (adjective-noun), refers to 'a building that has beauty as an inbuilt characteristic'; in other words, the topic of the sentence is 'a beautiful building'. When the adjective follows — un edificio hermoso — the adjective, as it were, splits up the noun, the topic: 'building' is now contrasted with other buildings that are not beautiful. Stavrou (1996: 83-84) describes this difference for Modern Greek as follows: "the prehead AP denotes a pre-existing (...) or defining property, whereas the posthead one asserts the (perhaps temporary) possession of a property". In addition, she notes that in Modern Greek postposition of the adjective is really only possible with indefinite noun phrases (see [1]). Postposition with a definite noun phrase only occurs when the postposed adjective functions as a so-

158

Olga Fischer

called object complement (small clause); i.e. the adjective is then governed by the verbal predicate and not by the noun phrase head, as in (2). (1)

a. Katharise enamilo kokino He/she-peeled an apple red 'He/she peeled a red apple' b. *Katharise to milo kokino He/she-peeled the apple red 'He/she peeled the red apple' (Stavrou 1996: 80)

(2)

Theli ti bira pagomeni He/she-wants the beer cold 'He/she wants the beer cold' (Stavrou 1996: 86)

Because of the differences expressed in (1), Stavrou further links the semantic differences expressed by adjective position with basic semantic differences between definiteness and indefiniteness. Vincent (1986) has considered the position of the adjective in Italian.5 His article provides a very useful overview of the most important studies that have appeared on the position of the modern Romance adjective, and how this applies to Italian. Although a great variety of descriptive terms can be found to differentiate between the two possible adjective positions, Vincent (1986: 192 [translation mine]) shows that a common denominator can be found for each position in terms of theme/rheme: L'aggettivo preposto, essendo parte inseparabile della testa, non puö avere un valore indipendente, mentre l'aggettivo postposto e sempre rematico rispetto al nome che modifica, anche se la sua posizione sintattica gli conferisce il ruolo di rema secondario. [The preposed adjective, being an inseparable part of the head, cannot have an independent value, while the postposed adjective is always rhematic with respect to the noun that it modifies, even though its syntactic position confers upon it the role of a secondary rheme.]

The position of the adjective in Old English

159

He further shows how the structure of the adjective-noun phrase resembles in linear terms the structure of existential clauses, which have only a rheme, while that of the noun-adjective phrase resembles a predicative clause, which has a theme as subject and a rheme as complement, suggesting the similar order theme/rheme for the noun and adjective respectively and their "independent value" with respect to one another. His suggestion then resembles the suggestion given for the difference in adjective position in Modern Greek, namely that the postposed adjective functions as a type of secondary predicate. In addition, the notion of theme/rheme corresponds closely to the difference between definiteness and indefiniteness that Stavrou suggested. Confronted with these ideas of linear (iconic) order, the expression of contrast and the role played by definite/indefiniteness or given/new information, the Old English noun phrase, with its possibility of a regular contrast between postnominal and prenominal adjectives, becomes highly interesting. Table 1. Strong and weak adjectives in Old English (a) declension of (b) declension of strong adjectives weak adjectives masc neut fem masc neut -a -e nom. sg. -/-u -e -an acc. sg. -ne -e -es -es -re -an -an gen. sg. -um -re -an -an dat. sg. -um -

-

nom. pi. acc. pi. gen. pi. dat. pi.

-e -e -ra -um

-/-u -/-u -ra -um

-a/-e -a/-e -ra -um

-an -an -ra/-ena -um

-an -an -ra/-ena -um

fem -e -an -an -an -an -an -ra/-ena -um

Since Old English has both strong and weak adjectives, which seem to be linked somehow with (in)definiteness, it is possible that similar semantic and/or pragmatic distinctions may have played a role there, and that there is a further link between position and weak/strong inflection of the adjectives. We must, therefore, briefly turn to the question of what the strong and weak forms stand for in Old English. The differences in form between the weak and strong category are given in table 1, which, incidentally, indicates quite clearly that the strong adjective is linguistically more highly marked than its weak equivalent. 6

160

Olga Fischer

Concerning the relation with (in)defmiteness, it is well known that strong adjectives are usually found when there is no determiner present, or with an occasional indefinite article {an or sum, which is not yet fully an article in Old English), while weak adjectives typically occur after a demonstrative pronoun functioning as a budding definite article. For most linguists, this distinction serves a function in Old English because it exercises a "principle of economy", as Barbara Strang (1970: 301) put it: "so long as a preceding word carried the full differentiae the adjective could appear in a less highly differentiated form". In other words, when there is no other defining element, the strong adjective ending is useful because it is distinctive for case and gender (unlike the weak adjective), while such a distinctive ending in a weak adjective is not useful because case and gender are already clear from the preceding demonstrative. This principle, or the functional interdependence of article and adjective ending, also provides an explanation for the rise of the article system in Middle English, or, if one wishes, for the disappearance of the weak/strong distinction in Middle English. Because the two are interpreted as clearly linked, the increasing presence of the one (the article) obviates the need for the other (the strong/weak adjectival distinction). 7 However, apart from being motivated by economy, the distinction between weak and strong adjectives may also be more intimately related to what the presence or absence of an article stands for, i.e. they may have a direct relation to (in)defmiteness. This is indeed suggested by Brunner (1962: 53-54, and a number of other linguists there referred to), who first of all remarks that the distinction drawn above by Strang and indeed by most Old English grammar books is not the original one. In Old English poetry we still find both types of adjectives without any determiner. So we need a further (that is, deeper) explanation for their use. Brunner (1962: 53-54 [translation mine]) writes: Die schwachen Formen sind daher individualisierend, gegenüber den allgemeinen starken. Sie werden zuerst wiederaufnehmend verwendet ... Die schwache Form steht daher attributiv wenn eine bestimmte Einzelperson beschrieben wird ... Als der bestimmte Artikel zur Kennzeichnung von Einzelpersonen oder Dingen in Aufnahme kam, wurde nach ihm die schwache Form des Adj. verwendet. Daraus erklärt sich auch, daß die ebenfalls individualisierenden Komparative und Superlative meist schwach flektiert werden...

The position of the adjective in Old English

161

[The weak forms are therefore individuating, in contrast to the generalising strong forms. They are at first used to refer back to an already mentioned entity ... The weak form is therefore used attributively when a certain individual is described ... When the definite article was on the increase to characterise individual persons or things, then the weak form of the adjective continued to be used after it. This also explains why the likewise individuating comparative and superlative forms usually carry a weak inflection ...] It is clear that for Brunner, the weak adjective is connected with definiteness and is, in that sense, individuating. It is also connected with already given information. He has less to say about the function of the strong forms. I do not find the term "allgemein", with which he describes them, very informative, but we will see below that the strong adjectives can indeed be linked with new information, and they are thus contrasted with weak ones. With the above ideas in mind it is time now to have a look at the adjectives in Old English.

3.

Adjective position in Old English

As I mentioned in the introduction, not much detailed work has been done on the position of the adjective in Old English. Mitchell (1985) is a fairly reliable guide to what had been done up to 1985. His own remarks on the topic are vague. In (1985: §172) he notes that the reason for postposition may be Latin influence, a desire for emphasis, rhythmic and stylistic variation, metre, etc. In (1985: §160) he writes that it is not always clear whether with an "attributive adjective in post-position ... we have to do with an attributive, predicative or appositional, use." It seems that Mitchell himself believes that these adjectives are still attributive (witness his remark that adjectives after the noun "may seem predicative to some readers" [1985: §168 (emphasis mine)]). Mitchell, as usual, describes all the various possibilities but does not go into any semantic and/or syntactic differences that may account for these possibilities. He is a taxonomist (a very good one) and not really interested in the why of variation. So clearly, for a linguist interested in an explanation for the variation, there is indeed room for more work.

162

Olga

Fischer

What other possible explanations have been offered? Mitchell (1985: §171) refers to S0rensen (1956: 262-263), who writes: "Anyhow it is tempting to assume that the widespread use of the construction adjective + substantive + adjective in OE. was supported by, if not a direct outcome of, [a] general tendency towards looseness in construction". This "looseness in construction" — which is also visible in other "split constructions", e.g. split subjects, objects, genitive phrases, etc. (cf. Reszkiewicz 1966 for a list of them) — is, according to S0rensen, probably due to a certain immaturity in the writing style, which is also manifested in the more frequent use of anacolutha in Old English. This leads him to the conclusion that postnominal adjectives are possibly cases of "afterthought" (Sorensen 1956: 262). Although I think it is too easy to explain these cases simply as immaturity and/or "afterthought", there is a remark in Sorensen's study that is certainly worth considering, but which he does not follow up himself. S0rensen (1956: 262) refers to Bogholm's (1939) English Speech from an Historical Point of View [which I have not been able to see], who has pointed out that in Old English "the rule is for parallel words to be kept apart". This links up with a remark made by Spamer (1979: 244) that in Old English adjectives were not recursive. I will come back to this below. Some further interesting and very pertinent observations made by Sorensen (1956: 264) with reference to modern postnominal adjectives preceded by and are: • •

•

That this "adjective may be interpreted as a predicative in its own right, as it were not very closely connected with the preceding substantive". That the postnominal and + adjective construction "may be caused by the need for linking up the second adjective with what follows" (he offers this example from Somerset Maugham: It was a pleasant sight and grateful to the sensibility ... ). That the postnominal and + adjective construction may be used to express contrast where the qualities expressed by the first and second adjective do not refer to "one primary"; i.e. in black spirits and grey, grey may refer to another category of spirits.

I will show later that all these factors may well be at work in Old English, and that they are all linked. Spamer (1979) proposes that there are two kinds of adjectives in Old English, which correspond with the morphological weak/strong distinction.

The position of the adjective in Old English

163

The weak adjectives, according to him, are "adjuncts" (Quirk and Greenbaum [1973: 123] term these "denominal adjectives" and give examples of the type a stone wall, a Shakespearian critic), while the strong ones function as determiners and are on a par with Old English se, seo, pest. In his study, Spamer uses this distinction in order to explain the development of the article system.8 It would take too long to go into all his arguments here, but he clearly suggests that there may be a link between type of adjective and (in)definiteness, as suggested also by Stavrou for Modern Greek. He notes that Old English adjunctival (i.e. weak) adjectives are recursive, while strong adjectives, being determiners, are not. He then notes that this explains why one may have two weak adjectives following one another prenominally and a strong adjective followed by a weak adjective prenominally, but not two strong adjectives in a row (Spamer 1979: 245). When two strong adjectives modify the same noun, the word order in Old English is "significantly different" from that of Modem English (Spamer 1979: 244). As an example of this different order he quotes the following, from Mitchell (1968: 65): (3)

pcet hi ncefre cer swa clcene gold, ne swa read ne gesawon that they never before such pure gold nor such red not saw 'that they have never seen such pure gold nor such red gold'

whereby he reads read as a modifier of an understood noun gold (which is deleted because it is co-referential). Further, he remarks that "[i]t is a striking feature of Old English syntax that a series of adjectives [strong ones, presumably] almost invariably contains and between every single adjective and the one following"; this is necessary according to Spamer because the "adjectives which they conjoin are non-recursive" (Spamer 1979: 244). So, in fact, he suggests that the use of and and the use of postnominal position are, as it were, forced by the non-recursiveness of strong adjectives in Old English. This hypothesis indeed explains quite nicely why quantifiers (which are almost always strong) do not normally co-occur with the demonstrative pronoun (determiner) in Old English. In my prose data I did not find a single example.9 It also explains why, when there are two strong adjectives and no determiner, front placement of both is rare (I found only eight examples in my data: see table 2, row 6); the usual stratagem being to place and before the second adjective, or to place (and +)

164

Olga Fischer

adjective, or both adjectives connected by and, after the noun. There are, however, quite a few problems with the data as presented by Spamer, and more specifically with the data ignored by Spamer. Table 2. All instances of adjective-noun order (without the connector and) involving more than one adjective in the Old English prose section of the tagged Helsinki Corpus. Dt A A A Prt Ν Total Q 9 X X X 220 X X X 15 X X X 77 X X X X 11 X X X X 1 X X X 8 X X X X 2 X X X X 1 + X X X 0 + X X X X 0 + X X X X 0 + X X X 0 + X X X 10 + X X X X 1(1?) + agen X X 3 ylca or sylf X X 13 X mycel X 11 Key: Dt = Determiner, Q = Quantifier, A = Adjective, Prt = Participle, Ν = Noun -

-

-

-

First of all, the type consisting of two quantifiers followed by a noun, where both quantifiers are typically strong, should not be possible (because, according to Spamer [1979: 243], the strong adjective is nonrecursive). However, this type is extremely frequent in my data (see table 2, row 3: 77 examples). In addition, I have found no examples of a weak quantifier followed by a strong one. Some examples: (4)

a. JEnig oder sceat [nom. sg. masc.; both strong] any other tribute (/El. Horn. Supplem. II 1.39)10

The position of the adjective in Old English

165

b. Sumere oöre stowe [dat. sg. fern.; both strong] in-some other place (JEI. Horn. Supplem II 1.197) c. Sume feawe dagas [acc. pi.; both strong] some few days (DOE, LS9 (Giles)822) Secondly, when we have a strong quantifier (which therefore must function as a determiner) followed by an adjective, we would expect the adjective to be weak according to Spamer's theory. However, both quantifiers and adjectives are usually strong in my data (Mitchell [1985: §125] confirms this): (5)

a. JElc haligfeeder [nom. sg. masc.; both strong] each holy father/patriarch (^Elet4 1.203) b. On sumne blindne sead [acc. sg. masc.; both strong] Into some blind pit (>Elet4 1.167)

This is also true when the first adjective is not a quantifier but in fact a true adjective. In this case both adjectives are also strong: (6)

a. mid ofermcete with excessive

unclcene luste [instr. sg. masc; both strong] unclean lust (Marti. 1.150)

b. of surre rigenre grut [dat. sg. fem.; both strong] from sour rye-made groats (Laeceboc 3. 1.59.1.1) Note that in the last example the typical adjunctival (adnominal) adjective rigen is in fact declined like a strong adjective, whereas Spamer (1979: 245) asserts that only the weak adjectives are adjuncts.

166

Olga Fischer

Table 3. All instances of noun-adjective order (except where the first postnominal adjective is preceded by and: those figures are in brackets) involving more than one adjective in the Old English prose section of the tagged Helsinki Corpus. Dt Q Nm Α Α Ν and Α Α Total Q Q Α Χ χ 102 -(+) Χ celmihtig 45 leaf 3 + Χ χ 74 + an χ 58 χ genoh 8 χ χ χ -(1 exc) 11 χ χ χ 1 χ χ χ 2 X χ χ χ 16 χ χ χ χ 3 χ χ χ χ (1) χ ping genitive 29 wiht adjective etc χ χ Χ 9 χ χ χ Χ 2 + χ Χ χ χ 5 + χ Χ 1 χ χ Χ 35 -(+, 3x) Χ χ χ χ 5 + Χ χ (wk!) χ (wk!) 2 χ + χ (wk!) χ Χ χ (wk!) 1 χ χ Χ 1 Χ χ χ 7) (χ χ χ Χ 1 Key: Dt = Determiner, Q = Quantifier, Nm = Numeral, A = Adjective, Ν = Noun, wk = weak -

-

-

-

-

-

Spamer's theory also does not explain why there are cases of postnominal adjectives when the need does not arise (cf. the data in table 3, above), as in:

The position of the adjective in Old English (7)

167

Pone ilcan ceaddan iungne [acc. sg. masc.; strong] the same Chad young (Chad. 1.184)

Here iung could easily have preceded the noun, syntactically speaking (semantically it is a different matter altogether), because it would have been weak after the determiner (like the other adjective, ilcan), and two weak adjectives are allowed in his theory. Also, the hypothesis does not explain why and is not always inserted (see [8] and also table 3, above) before the first postnominal strong adjective, as in the example he quotes from Mitchell (see [3] above). (8)

Gyldenne wingeard trumlicne andfcestlicne [acc. sg. masc.; strong] golden vineyard durable and firm (Alex. 1.107)

Nor is it clear why we should get postnominal adjectives when there are no prenominal ones at all. After all, in that case, there is no reason to opt for a "significantly different" word order (Spamer 1979: 244). Such instances are, however, highly frequent in my data (see table 3, row 1), especially when the postnominal adjective is a quantifier, or an adjective in -weard:u (9)

a. poet hi hyra cehta ealle beceapedon that they their possessions all sold (/Eletl 1.40) b. pa wilnode ic indeum innanwearde to geseonne then wanted I India inner to see 'then I wanted to see inner India' (Alex. 1.129) c. Petrus hcefde wif ful sod Peter had wife full true OEletw. 1.17) d. Gersuman unateallendlice treasures uncountable (ChronE 2, 1086.59)

168

Olga Fischer

A question that arises in this connection is: Why are postnominal adjectives practically always strong (there are only two exceptions in the data in table 3)? This is so even when a determiner precedes the noun, as Mitchell indicates: (10)

in pissum life ondwardum in this life present 'in this present life' (Mitchell 1985: §126, Ch 1508)

The postnominal adjective is only weak when the determiner is repeated too. Most cases of this recorded in Mitchell (1985: §126) are superlatives, which tend to be weak adjectives even when predicative (for an explanation of this see Brunner 1962: 54, quoted above). A last point that Spamer's hypothesis does not satisfactorily account for is the frequent occurrence of two prenominal weak adjectives connected by and, where there should be no need for and (because the weak adjective is recursive). A quick glance through the Dictionary of Old English revealed many of these. Set against the few examples of weak adjectives in a series (all discussed below), this is, to say the least, remarkable. (11) a. Se ceresda stream pcere clcenan and hlutran burnan The first stream of-the clean and clear spring (DOE, GDPref 1 2,18) b. For pcere micclan and stiöan drohtnunge Because-of the intense and severe way-of-life (DOE, LS7(Euphr)233) There is something, however, that pleads very strongly in favour of Spamer's analysis of weak adjectives as adjuncts, but which he does not discuss himself, and this is the complete absence of constructions such as: (12)

*se/his swipe ealda man the/his very old man12

The position of the adjective in Old English

169

Adjunctival adjectives (as Spamer qualifies the weak adjectives, see note 8), being of a nominal character cannot normally be modified by adverbs, cf. *a very stone wall. I have checked the Dictionary of Old English, and have not been able to find a single example with a weak adjective, neither with swipe nor with Jul. What I did find quite regularly were strong adjectives modified by swipe, used predicatively, as in he wees swipe eald, 'he was very old'; in indefinite noun phrases, such as swipe eald man, 'a very old man', and postnominally as in: (13) a. and peer of slogan anne brittiscne man swipe cepelne and there killed a British man very noble 'and there killed a very noble Briton' (DOE, ChronC [Rositzke] 501.1) b. nu cweöaö oft preostas poet petrus hcefde wif Jul sod now say often priests that Peter had wife full true 'now priests often say that Peter had a very true wife' (cf. [9c.] above) The only exception is: (14)

His Jul leof feeder His very dear father (BlHom3 1.14, p. 101-102)

Where ful may originally have been an adjective like the micel examples which I will discuss below. What does all this suggest? Spamer seems to be on to something when he associates weak adjectives with adjuncts (which makes the adjective + noun a kind of compound; cf. note 8), but he is wrong about strong adjectives being determiners. He is also on to something when he mentions non-recursiveness of adjectives, but this seems to be true for both strong and weak ones (quantifiers excepted).13 However, he has missed the significance of why adjectives may be positioned after the headword. Postnominal position is not an automatic consequence of the impossibility of serial strong adjectives, rather it seems to involve a choice and it seems to make a semantic distinction. This distinction may well be linked to

170

Olga Fischer

strength and weakness of adjectives, or to indefiniteness and definiteness, with which these adjectives were traditionally linked. I now want to go back to the ideas I mentioned at the beginning: the link between adjective position and definiteness, between adjective position and contrast, and between adjective position and linear iconicity. What I would tentatively like to suggest is that all strong adjectives in Old English are, or started off as, predicative adjectives, and that weak adjectives tend to be adjunctive (in the sense of Spanier). The natural position for predicative adjectives would be after a noun phrase followed by a copula, but they also came to be used straight after nouns in so-called secondary predicates (traditionally subject or object complements), conveying in both cases additional and new information. The natural position for weak or adjunctive adjectives, on the other hand, would be prehead, conveying already known (and therefore definite) information (it would explain at the very least why postnominal adjectives are always strong).14 This would link up with Bolinger's idea that when there is a choice in position, the prenominal adjective qualifies the noun and turns it into a different category. The postnominal adjective, on the other hand, gives extra information about the noun and the quality it expresses is not interpreted as inherent in the noun itself. The postnominal adjective in this sense has more of a verbal character (as said above, it functions like a secondary predicate), and can therefore also be modified by elements that can modify verbs, such as adverbs and prepositional phrases. This would then explain why in Old English degree adverbs like swipe can only modify verbs and strong adjectives, as we have seen. It also explains why only strong adjectives can be followed by a prepositional phrase (as far as I could tell from my data; and compare the remark made by Sorensen above), as in: (15) a. God da ford ateah of dcere moldan celces cynnes God then forth out-pulled from the earth of-each kind treow fceger on gesyhde tree beautiful in sight 'God then pulled forth from the earth every kind of tree beautiful to see' OEI.Old Test. 1 1.2.9)

The position of the adjective in Old English

b.

171

swide geleafull wer welig on cehtum (a) very faithful man wealthy in goods (/Elet4,1.737)

c. bollan fulne hates wines cup full of-hot wine (Lasceboc 2.14.2) Similarly, it would explain why in some cases the strong postnominal adjective itself is comparable to a prepositional phrase, as in: (16)

and peer sint swide micle meras fersce and there are very many lakes fresh 'and there are very many fresh-water lakes' (Mitchell 1985: §126; Oros. 19.5)

The adjective fresc here does not qualify the lake itself but adds information about what is in the lake, and is thus comparable to the Dutch example een broodje gezond, discussed in section 2. Compare this also to (17b) below, where fresc modifies 'water' and does stand before the headword.15 The nature of the Old English adjectives, i.e. predicative and adjunctive, also explains why adjectives cannot really modify each other. Predicative adjectives are always parallel and not hierarchically ordered. Thus, one can say a beautiful young woman, but not *the woman was beautiful young. This would explain why the strong adjective is not recursive (cf. Spamer's claim, but made here on different grounds) and why adjectives used postnominally must be strung together by means of and in Old English. Adjunctive adjectives are also more difficult to string together, because of their compound characteristics. We saw above that we sometimes do find a series of strong adjectives without and prenominally in Old English (see table 3 — I have found very few examples of a postnominal series of adjectives without and). How are these then to be explained? First of all, the most usual string is that of two quantifiers following each other. A possible explanation for these strings is that the first quantifier acts like an indefinite determiner (which accounts for the fact that the adjective or quantifier following always has the strong form), while the second may be more adjectival in nature.

172

Olga Fischer

Concerning two true adjectives in a series prenominally, I have found eight examples in my data of adjective-adjective-noun where both adjectives are strong; one example of three strong adjectives and a noun (but these adjectives are all the same, so a different case altogether: halig halig halig Drihten, [iElet4, 1.1154]; which, moreover, is a translation from the Latin sanctus sanctus sanctus) and two examples of a quantifier followed by two strong adjectives (see table 2, rows 6, 8 and 7 respectively). What is noteworthy about these examples is that the second adjective refers in six cases either to a material (cescenne, 'of ashwood' [Lascebocl 13.10.1]; beren, 'of barley' [Lascebocl 19.4.1]; rigenre, 'of rye' [Laeceboc3 159.1.1]; sylfrenum 'of silver' [Documents 4, 2a, 1.21]) or to a nation (brettisc, 'British' [ChronA Early 1.501.1]; englisc, 'English' [Documents 4, 2a 1.41]); in other words they are denominal adjectives. These adjectives (cf. Quirk and Greenbaum 1973: §13.40) always stand close to the noun in Modern English because they are the least adjective-like. It may be for this reason that this series is possible because the second adjective is so nounlike. If we are witnessing a change from purely parallel use of adjectives to serial use, this is where we might expect it to start. It is interesting to note, by the way, that adnominal adjectives are characterised by Spamer as typical examples of weak adjectives (adjuncts), but they are strong here. The other examples are more difficult to account for: (17) a. Swa beorht scinende steorra Such (a) bright shining star (Bede3.1.13.428.18) b. Swetum fer scum wceterum With-sweet fresh water (Laxeboc2, 1.16.1.4) c. God hluttor eala Good clear ale (Lasceboc3 1.30.1.1) d. Mid ofermcete With excessive

unclcene luste unclean lust (=[6a] Marti 1.150)

The position of the adjective in Old English

173

The example from Bede may have been influenced by the Latin text, and the fact that one adjective is participial may play a role too. Possibly the other three involve adjective-noun combinations which may well be idiomatic and can therefore be seen almost as a compound, thus allowing another adjective in front of it. It could also be said that god and ofermcete express degree and could begin to function like adverbs modifying the next adjective, just like the micel examples discussed below. Concerning a series of adjectives between a determiner and a headword (thus weak adjectives), I have found ten instances of adjective-adjectivenoun (table 2, row 13), and one of three adjectives plus a noun (table 2, row 14). The examples are similar to the previous group in that in five cases a locality/nation is present (sudwesterna 'southwestern' [Apollonius 1.11.11]; frencisce 'French' [ChronE2.1.1087.1]; romaniscan 'Roman' [Greg. Dial.2, MS Η, B.5.8.2]; macedonisc 'Macedonian' [Marvels 12.10, 2x]). Four are similar again in that the adjective may form an idiomatic unit with the noun (godan willan 'goodwill' [Cura Past.3 1.53.417.12]; synfulla man, 'sinful man [sinner]' [WHoml.1.43]; soda God, 'true God' [WHom8.1.25]; eadige fcemne, 'blessed woman [Mary]' [Marg. 1.350]) and there are two I cannot immediately account for: (18) a. poet ofstandene picce slipige horh the remaining thick slimy mucus (Lasceboc2 1.16.1.14) b. on pam ceftemestan mceran freolsdcege on the after-most well-known festival-day (West Saxon G. 1.7.37) It could be that 'slimy mucus' is an idiomatic unit too (note that the adjective and noun express more or less the same thing); that participial adjectives behave somewhat differently, and that in the second example ceftemestan functions like a kind of ordinal numeral. A rather different group of the determiner-adjective-adjective-noun type concerns cases where the first adjective is agen, 'own' (table 2, row 15: three instances) or ylca/self'self (table 2, row 16: thirteen instances). To my mind these adjectives are different because they seem to form a unit with the determiner rather than with the noun. Note indeed that in Middle

174

Olga Fischer

English the ilke often comes to be written as thilke. Yet another special group (table 2, row 17) concerns the following type: (19)

On pcere micclan niwan gecydnysse In that big/very new testimony (;Elet4 1.925)

Here it is not entirely clear whether micclan functions as an adjective to the noun or as an adverb to niwan. Both interpretations would make sense. Note that the ending in -an in micclan functions both as a weak ending and as an adverb according to the traditional Old English grammars, but the adverbial usage of micclan may well have started in these constructions and have developed from an original adjective (because -an is a highly unusual ending for an adverb). In example (20), it is much more likely that micle must be interpreted adverbially because of the comparative adjective, and the -e ending in micle (if it was an adjective, it would presumably have been micel): (20)

Sua micle hefigre wite So great/much heavier punishment 'So much heavier punishment' (Cura Past.2 136.247.6, p.157)

I found eleven instances of micel + adjective + noun. These examples may well show the beginning of a different system: a system where adjectives are no longer used strictly in parallel, but are beginning to modify each other, and where, due to the inability to distinguish clearly between strong and weak adjectives in the later period, a determiner system begins to develop, which takes over the expression of definiteness/indefiniteness formerly expressed by the weak/strong distinction. The loss of that distinction may also have led to the loss of the adjunctive/predicative distinction in adjectives, which in turn may have led to the loss of a regular position for postnominal adjectives. The predicative or temporary character of a modifier may now have to be expressed differently; for instance by means of a relative clause after the headword or by special stress on prenominal adjectives. Of course, occasionally, we still find noun-adjective order in Modern English, as in the people present,

The position of the adjective in Old English

175

where the adjective is clearly predicative, and this order is used to contrast it with the attributive the present people. Also, we find it when an adjective is followed by a prepositional phrase, as noted above by Brunner (section 1) and Sorensen (section 2). But indeed these postnominal adjectives are still predicative and can usually be replaced by a relative clause. The third possibility mentioned by Sorensen was the use of and + adjective as in, black spirits and grey, where the second adjective refers to another category. This indeed may also have been a reason for an adjective to follow and postnominally in Old English: (21) a. gesceop god celmihtig ealle gesceafta gesewenlice and ungesewenlice created god almighty all creatures visible and invisible 'God almighty created all creatures, visible and invisible' (Byrht.M.2 1.80.15) b. ne scyle nan mon siocne monnan and gesargodne swencan not must no man (a) sick man and wounded oppress (Boeth.4 1.38.123.32) c. to dan blidan wunenessum dara hwittra gasta andfcegra to the joyful dwelling of-the white spirits and beautiful (Bede3 1.13.430.37) d. ... pe wynsumlicor pa myclan byrpenne and pa hefian ... the more-joyfully the big burden and the heavy aberan mihton endure might 'the more joyfully might endure the big burden' (BlHom3 1.65) e. pees swetan wcetres and pees ferscan of-the sweet water and the fresh (Alex. 1.338) It is clear from the examples in (21) that it is not always evident whether the adjectives refer to two different categories or to one and the same. The entities may be different ones in (b) to (d); they must be different in (a)

176

Olga Fischer

(but here both adjectives follow the noun) and they are unlikely to be different in (e). Examples (d) and (e) make clear that the second adjective may be used substantively (because of the repeated determiner) and in the same way the second adjective of (b) and (c) could be so interpreted. Substantive adjectives were a regular feature of Old English because adjectives could still be declined for number, case, and gender. The loss of these endings led to the use of the propword one in Modern English, and, indeed, in Modern colloquial English the expression black spirits and grey would normally be followed by the (plural) propword ones. So then this change would not be one in adjective position, but rather would involve the loss of substantival adjectives after the Old/Middle English period.

4. A brief conclusion Although more extensive work is clearly necessary, both on Old and, especially, on Middle English data, a few tendencies have come to the fore. It seems that there is indeed a relation between the position of the adjective and the feature of (in)definiteness in Old English: weak adjectives are related, as Brunner and others already suggested, with definiteness, and as given information they tend to occur before the noun. Strong adjectives are predicative rather than attributive in nature, which makes it difficult for them to occur hierarchically in a series. They are associated with indefiniteness and with new information. For this reason, the postnominal position is still frequent. With the rise of a new determiner system and the loss of the strong/weak distinction, the two types of adjectives may have lost some of their distinctive features, with the result that they came to be used more or less indiscriminately prenominally (with some relic postnominal positions left over), but how exactly their position was affected in the period immediately following the Old English period, is a topic for further study.

Notes 1. I would like to thank Willem Koopman and Bettelou Los for their careful reading of a second version of this paper and for their useful suggestions and comments.

The position of the adjective in Old English

177

2. Not everyone agrees with the idea that quantifiers were part of the category "adjective" in Old English. Although it is true that adjectives could occur in all the positions quantifiers occurred in, there were severe restrictions for adjectives, both on the type of adjective as well as the textual genre (poetry versus prose). For more details see Bennett (1979), and Fischer and van der Leek (1981: 311-337). For some further differences between adjectives and quantifiers see also notes 9 and 11 below. Note also that in other languages, such as Modern Italian, where adjectives can take up more than one position, the quantifiers behave differently from the adjectives. For this reason Vincent (1986: 187) does not include them in his discussion of the position of adjectives. 3. Lightfoot (1979: 206-209) illustrates, by means of a study based on a small corpus of Chaucerian texts, how productive postnominals are with native Germanic words. I believe that these examples need to be carefully looked at before any suggestion can be made about this postnominal position becoming more productive in Middle English. It seems to me that many of the examples he gives are examples of predicatively used adjectives (below we will see that these are typically postnominal in Old English), and his illustrations make clear that he also includes examples involving more than one adjective, where postnominal position of one or two adjectives was very common in Old English (Lightfoot 1979: 208). Moreover most of his examples come from poetic texts, and it seems highly likely that postnominal position is more frequent there than in prose, as it was in Old English and as it still is in Modern English poetry. 4. The corpus I have used is the prose part of the Old English section of the Helsinki Corpus (the glosses excepted), which is currently being tagged at the Free University of Amsterdam. I would like to thank Frank Beths, who is the main "tagger", for his expertise and freely given help. The tags used in this text refer to the tags used in the Helsinki Corpus (for which see Kytö 1991), but some abbreviations have been enlarged for easier reference. Some further examples come from the Dictionary of Old English Corpus. 5. I would like to thank Nigel Vincent for drawing my attention to this article during the presentation of this paper in Manchester. 6. In a forthcoming paper (Fischer 1999) I will discuss why the more pronounced marking of strong adjectival endings, compared to weak ones, is significant from a functional point of view. 7. For a different point of view concerning this link, see McCoIl Millar (forthcoming). 8. He argues that strong adjectives were determiners, and that weak adjectives were "adjuncts", and that, with the loss of inflectional endings, these adjectives fell together; as a result of which the initial demonstrative, which was originally adjectival, became realised as the definite article. His solution for the rise of the article is very simple and elegant, but it may not quite work, since, as I will show below, the distinction between adjective/determiners and adjective/adjuncts

178

Olga Fischer

does not cover all the facts of Old English. It should also be noted that Spamer's use of the term "adjunct" is rather idiosyncratic, and is not really described clearly in his article. As far as I could make out from his description ("The chief defining characteristic of the adjunct is that it behaves in many respects as if it were the first part of a compound noun" [Spamer 1979: 242]), he uses the term to refer to what Quirk and Greenbaum (1973: 123) have called a "denominal adjective". 9. Mitchell (1985: §145) does give some examples (e.g., ealnepisne middangeard) but they are rare, and, when they do occur together, the quantifier, if it follows the determiner, is indeed a weak adjective (whereas normally quantifiers are strong): e.g. pa monigan cyningas, pas feawan dagas. 10. In my treatment of the Old English examples I have provided glosses, and generally no translation (unless the word order of the clause as a whole was such that it might raise confusion). The reason for this is twofold. First of all most glosses speak for themselves; adding a translation would be superfluous because it would be exactly the same as the gloss. Secondly, and more importantly, it is difficult to translate the adjective in the discussed noun phrases adequately in Modern English because their meaning is influenced by the position they take in the noun phrase. How this meaning is influenced by position is the topic of this paper, and I do not wish to pre-empt the issue by translating all the adjectives in a prenominal position, when perhaps a different translation (i.e. by means of a relative or adverbial clause or indeed postnominal position) may do more justice to the meaning of the phrase. 11. This is another indication that the quantifiers in Old English did not quite behave like adjectives, as was suggested by Lightfoot (1979) (cf. section 1 above). 12. Bettelou Los (personal communication) suggested that the absence of such constructions may not have a syntactic cause, that in fact such constructions may be rare in present-day English too; or that, when they occur in present-day English, they are the result of the development of a literary style. It would be difficult to prove the influence of style. However, I have consulted the recent newspaper corpus collected by a colleague, Tom van Brederode, which is about the same size as the Old English corpus I have used here (4,682,625 words as against the Old English corpus of 5,894,365 words), and have found 100 occurrences of phrases with the determiner the and very followed by an adjective and a noun (e.g. the very intelligent suggestion, the very limited powers, the very strong-willed attitudes). I have not checked any other forms of determiner nor any other adverbs, but this at least suggests that the construction is quite common in present-day English. I have also briefly checked the website of the Middle English Dttionary {Middle English Compendium http://www.hti.umich.edu/mec/) and found that these constructions only begin to occur in the fifteenth century; only constructions such as the moste + adjective + noun are somewhat earlier, e.g. I found pe most kyd kmjjtes in Sir Gawain and the Green Knight and the

The position of the adjective in Old English

179

moste reale place in The Alliterative Morte Arthur. But again these too only become common in the fifteenth century 13. The adjectives in -weard are a special case, in the sense that they are the only adjectives which, at least in prose, regularly occur prenominally (i), postnominally (ii), and even before the determiner (iii). When they are postnominal, they can often be interpreted adverbially as in (iib); the dictionaries even give the -um ending as an adverbial ending (this case could be similar to the one mentioned below: i.e. the adverbial status of miclan and also miclum). The fact that they can also be used as particles no doubt accounts for their "adverbial" usage. It is also quite clear from the examples given below that the meaning of the adjective depends quite crucially on position in these cases. Thus, inneweard usually means 'on the inside' postnominally, but 'sincere' prenominally as in (ia). Likewise toweard used postnominally expresses the fact that the event lies in the future (see iia), whereas prenominally it refers to a kind of future life, i.e. the Christian heaven. (i) a. we hine biddad mid inneweardre heortan we him pray with inward heart 'we pray to him with a sincere heart' OEHom 1.39) b. pcet he ne purfe prowian on dam toweardan life so-that he not need suffer in the future life 'so that he need not suffer in the future life' OEHom 1.157) (ii)a. and him sige towardne geheht and to-him victory impending promised 'and promised impending victory to him' [towardne is acc. sg. strong; hence adjectival] (BlHom5 1.95) b. se leoma gehran pcem treowum ufonweardum the light touched the trees highest/above 'the light touched the highest trees/touched the trees above' [here ufonweardum can be both adjectival and adverbial] (Alex. 1.849) (iii) and pa heafda ealle wurdon gesette on ufeweardum pam geate and the heads all were placed on above the gate 'and the heads were all placed on top of the gate' (Apol. Tyre 1.6.1)

180

Olga Fischer

14. For some exceptions see Fischer (1999). 15. Fischer (1999) goes deeper into the nominal and verbal character of the weak and strong adjectives respectively.

References Bennett, Paul A. 1979 Observations on the transparency principle. Linguistics 17: 843-861. Bolinger, Dwight L. 1967 Adjectives in English: Attribution and predication. Lingua 18: 1-34. 1972 Linear modification. In: F. Housholder (ed.), Syntactic Theory, 31-51. Harmondsworth: Penguin. First published in Language 67: 1117-1144 [1952], Brook, G.L. An Introduction to Old English. Manchester: Manchester University 1955 Press. Brunner, Karl 1962 Die Englische Sprache. Ihre Geschichtliche Entwicklung, Volume 2. Tübingen: Niemeyer. Davis, Norman 1953 Sweet 's Anglo-Saxon Primer. Oxford: Clarendon Press. Dictionary of Old English Dictionary of Old English Project, The Complete Corpus of Old 1998 English in Machine-Readable Form. Healey, Antoinette DiPaolo and Richard Venezky (eds.) University of Michigan. [Web site: http://www.hti.umich.edu/english/oecl. Ferris, Connor The Meaning of Syntax. A Study in the Adjectives of English. 1993 Harlow: Longman. Fischer, Olga 1999 The position of the adjective in Old English from an iconic perspective. Paper given at the Second Symposium on Iconicity in Language and Literature, University of Amsterdam [March 1999]. Fischer, Olga and Frederike C. van der Leek 1981 Optional versus radical re-analysis: Mechanisms of syntactic change. Lingua 55: 301-50. Kytö, Merja Manual to the Diachronic Part of the Helsinki Corpus of English 1991 Texts. Coding Conventions and Lists of Source Texts. Helsinki: University of Helsinki.

The position of the adjective in Old English

181

Lightfoot, David W. 1979 Principles of Diachronic Syntax. Cambridge: Cambridge University Press. McColl Millar, Robert forthcoming Some suggestions for explaining the origin and development of the definite article in English. In: Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), Pathways of change. Grammaticalisation in English. Amsterdam: Benjamins. Mitchell, Bruce 1968 A Guide to Old English, Second edition. Oxford: Blackwell. First edition [1964]. 1985 Old English Syntax, Volume 1. Oxford: Clarendon Press Mosse, Fernand 1946 Manuel de I'anglais du Moyen Agedes origines au XI Ve siecle, Volume I, Second edition. Paris: Aubier. First edition [1945]. Quirk, Randolph and C.L. Wrenn 1955 An Old English Grammar. London: Methuen. Quirk, Randolph and Sidney Greenbaum 1973 A University Grammar of English. London: Longman. Reszkiewicz, Alfred 1966 Split constructions in Old English. In Mieczyslaw Brahmer, Stanislaw Helsztynski and Julian Krzyzanowski (eds.), Studies in Language and Literature in Honour of Margaret Schlauch, 313-326. Warsaw: Polish Scientific Publishers. Serensen, Knud 1956 Substantive with two epithets. English Studies 37: 261-264. Spamer, James B. 1971 The development of the definite article in English: A case study of syntactic change. Glossa 13: 241-250. Stavrou, Melita 1996 Adjectives in Modern Greek: An instance of predication, or an old issue revisited. Journal of Linguistics 32: 79-112. Strang, Barbara M.H. 1970 A History of English. London: Methuen Vincent, Nigel 1986 La posizione dell'aggettivo in italiano. In: Harro Stammerjohann (ed.), Tema-Rema in Italiano. Symposium, Frankfurt am Main, 26/27-4-1985, 181-195. Tübingen: Gunter Narr.

On the history of the s-genitive Anette Rosenbach, Dieter Stein and Letizia Vezzosi

1.

Introduction

The genitive and its history have recently received renewed interest, as witnessed amongst other things by the recent discussion in the journal English Language and Linguistics (Allen 1997, 1998; Klemola 1997) and in other publications (Seppänen 1997). The traditional view is that the English inflected genitive went the same way as other inflections in the course of the typological development of the language from a synthetic to a more analytic character. Indeed, in Middle English the inflected genitive had largely been replaced by the (^construction. However, the process has not been completed to total loss of the inflected genitive, which still occurs in today's English. In this paper we do not intend to re-write the history of the inflected genitive. Rather, we would like to present new data concerning its development or, more precisely, the development of the major morphosyntactic devices for expressing possessive relationships. We will focus on the distribution and development of the two dominant nominal devices that express possessive relationship in English: the inflected genitive, henceforth referred to as the "i-genitive" {Simon's father), and the periphrastic genitive, referred to as the "o/-genitive" (the father of Simon)} These two constructions also differ in the relative order of the two elements involved, the possessum and the possessor. In the s-genitive the possessor comes first; in the o/-genitive the possessor follows the possessum (see table 1). The terms of "possessum", "possessor" and "possessive relationships" are used in the general framework of possession (see section 2.3 for further definition). Table 1. Definition of "possessive constructions" Inflected genitive

Periphrastic genitive

"j-genitive"

"o/· genitive"

Simon's father

the father of Simon

possessor-possessum

possessum-possessor

184

Anette Rosenbach, Dieter Stein andLetizia

Vezzosi

1.1. Historical background As is well known, during the Middle English period English changed from a synthetic character, in which syntactic relations were mainly expressed paradigmatically by inflection, to an analytic character, in which syntactic relations are encoded syntagmatically by word order and prepositional constructions. In accordance with this overall typological drift in English, the inflected genitive too was replaced to a large extent by a prepositional construction, namely the o/-genitive, during Middle English. For Middle English, the relative distributions of the s-genitive and the o/-genitive are generally depicted as in figure 1 (the figures in the graph are based on Mustanoja [1960: 75], who quotes a corpus analysis by Thomas [1931]). 120 η

100 -prototypical possession > agent possessors > object possessors. If the s-genitive occurs at all with locally given possessors, the possessive relationship seems to be the determining factor: it occurs first (interval I) only in +prototypical relation, and extends in the second interval f r o m +prototypicaI to -prototypical possession to agent relations. 2

+ proto

-proto agent 1500-1559

object

Figure 9c: Interaction of possession/valency and topicality/givenness (1500-1559): relative frequency of s-genitive (in percentages) (cf. table VIc in Appendix 1)

Possession

+ proto

Valency

-proto 1560-1630

agent

object

Figure 9d: Interaction of possession/valency and topicality/givenness (1560-1630): relative frequency of j-genitive (in percentages) (cf. table VId in Appendix 1)

196

Anette Rosenbach, Dieter Stein andLetizia

Vezzosi

In the last two intervals we find a large extension of the ί-genitive with locally given possessors. While the type of possessive relationship with globally given possessors does not seem to play a role any longer, it still seems to determine the frequency of the s-genitive with locally given possessors. There is, however, a notable difference in the preferred context for locally given possessors. While in the first two intervals the ^-genitive with both globally and locally given possessors extended along the same continuum (+prototypical > -prototypical > agent > object), from the third interval onwards agent possessors outrank -prototypical possessive relations as a preferred context for the occurrence of the s-genitive

3. Summary and conclusion To sum up, we would like to suggest the following answers to our initial research questions (see section 1.2). The s-genitive cannot be considered as an accidental leftover in the English language. There is a clear rise in frequency of occurrence from 1400 to 1630. In particular, three contexts seem to be crucial in triggering the s-genitive. The strongest criterion for the occurrence of the s-genitive in our corpus is animacy. If the s-genitive can occur at all, then it is almost exclusively with [+animate, +human] possessors. The two other factors, topicality and possession/valency, do not have the same impact on the increase of the s-genitive. Within the animate context, the further restricting criterion for the occurrence of the s-genitive shifts from topicality in the first two intervals to the type of possessive/valency relationship in the last two intervals, with agent possessors becoming more important than -prototypical possession. The syntactic restriction of the s-genitive to the order possessorpossessum can be accounted for if we take a cross-linguistic perspective and look at the general word order constraints imposed by the factors animacy, topicality and agency. It is widely known that these factors are the determining factors in various areas of grammar. Among the various hierarchies that have been proposed in the literature, Givon's (1984: 364) topic hierarchy probably shows best how the different factors may interact with each other:

On the history of the s-genitive Semantic case-role: Pragmatic case-role: Humanity / animacy: Deflniteness:

197

AGT > DAT/BEN > PAT > OTHERS SUBJ > D-OBJ > OTHERS HUMAN/ANIMATE > NONHUMAN/INANIMATE DEF > INDEF

In this hierarchy it is always the leftmost element that is more likely to undergo a certain grammatical operation than the elements on the righthand side. For example, Givon (1984: 365-372) has shown — among other things — that obligatory grammatical agreement is more likely to occur with subject noun phrases than with object noun phrases. If object agreement can occur at all, it often depends on the animacy — and sometimes also on the deflniteness — status of the object noun phrase. While in Swahili, for example, object agreement is obligatory with human objects, with inanimate objects it becomes obligatory only with definite objects. 3 With respect to word order, there is cross-linguistic evidence for animate noun phrases occuring before inanimate noun phrases (e.g. Siewierska 1988: 56-60; Ortmann 1998: 75-76). Moreover, according to theories of information structure (e.g. Firbas 1964; Gundel 1988a) given information (i.e. the theme/topic) usually precedes new information within a sentence. Transferring this principle to the phrasal level, Rosenbach and Vezzosi (1997, forthcoming) have argued that the preference for the s-genitive to occur predominantly with topical possessors is pragmatically motivated. If the possessor is known it is much easier to identify the intended referent of the head. This anchoring function is particularly important in a structure where the possessor comes first, i.e. the s-genitive. Furthermore, for English as a Subject-Verb-Object language we would generally expect to find properties strongly connected with the subject, such as agency, animacy and deflniteness, to occur first — before object properties respectively — in any given linear sequence. To sum up, looking from a broader cross-linguistic and typological perspective there appears to be a good reason why the possessor comes first only — as it does in the s-genitive — if it is [+animate]/[+human], [+topical] and +prototypical possession/agent. Which brings us to a couple of $64,000 questions. Why did these factors not become operative earlier, or later, for that matter? Were there external or internal-structural factors that made these contexts produce more s-genitives not present in earlier Middle English?

198

Anette Rosenbach,

Dieter Stein and Letizia

Vezzosi

One embedding, if not actuating, factor that cannot possibly be overlooked is that apparently the most preferred surviving inflectional ending both in the verbal and nominal paradigm is -s, which also tends to generalise within the respective paradigms in non-standard dialects, as it did in Northern Old English genitive and present tense verbal inflections. Diachronically, what we observe in the development of the s-genitive, is, amongst other things, a pragmatisation of its use, or a functionalisation for functional sentence-perspective purposes, much in the same way as has been shown for the use of DO and cognates in pre-standard and modern dialectal uses (cf. Stein 1990). The shift from globally given possessors to locally given possessors as the preferred context for the s-genitive that was observed in the last interval (1560-1630) reflects a shift from a lexicallybased topicality to a more discourse-based topicality. The inflected genitive has acquired a textual function in the Hallidayan sense. On top of that, it can today have a personalisation function, as shown by Dabrowska (1998). She argues that it can be used for a metaphorising and perspectivising effect. In other words, it looks as if the post-Middle English new-fangled inflectional genitive has risen to undergo subjectivisation in much the same way as other expressions. Further theoretical analysis will show whether it is possible to treat the historical development of the genitive in terms of a historical re-ranking or addition of constraints, as is being done in Optimality Theory.

Appendix 1 Table I. Distribution of genitive constructions according to time intervals (our

corpus)

o/-genitive s-genitive

1400-1449(1)

1450-1499(11)

1500-1559(111)

1560-1630 (IV)

η

η

%

η

%

η

%

81.2 14.6

2257 320

81.3 11.5

1826 449

76.0 18.7

1341 119

% 89.6 8.0

2059 371

Table IIa. Distribution of the 5-genitive and o/genitive with [-animate] possessor 1400-1449 1450-1499 1500-1559 1560-1630 5-genitive o/~-genitive

η

%

η

%

η

%

η

%

8 732

1.1 98.9

25 1082

2.3 97.7

17 1614

1.1 98.9

42 1287

3.2 96.8

On the history of the s-genitive

199

Table lib. Distribution of s-genitive and o/genitive with [+animate] 1400-1449 1450-1499 1500-1559 η % η % η % s-genitive 110 21.6 344 31.9 298 36.8 of-genitive 398 78.4 735 68.1 512 63.2

possessor 1560-1630 η % 402 50.3 398 49.7

Table III. Topicality of possessor: distribution of given versus new 1400-1449 1450-1499 1500-1559 η % η % η % Given 396 95.4 790 96.7 509 93.6 New 19 4.6 27 3.3 35 6.4

possessors 1560-1630 η % 531 94.5 31 5.5

Table IVa. Topicality (givenness) 1400-1449 s-genitive η % Globally given 80 24.0 Locally given 2 3.5 Total 82

of possessor (1400-1449 and 1450-1499) 1450-1499 o/-genitive s-genitive o/-genitive η % η % η % 259 76.0 280 43.0 371 57.0 55 96.5 15 10.8 124 89.2 314 295 494

Table IVb. Topicality (givenness) 1500-1559 s-genitive η % 45.7 Globally given 191 33 36.3 Locally given 224 Total

of the possessor (1500-1559 and 1560-1630 o/^genitive s-genitive η % η % 227 54.3 223 58.4 58 63.7 102 68.5 285 325

Table Va. Possession and valency 1400--1449 s-genitive η % 22 31.4 +prototype -prototype 37 26.1 Agent 16 14.8 9.2 Object 7 Total 82

(1400-1449 and 1450-1499) 1450--1499 o/-genitive s-genitive η % η % 92 48 68.6 50.5 105 73.9 97 31.6 92 85.2 59 30.1 69 90.8 18 23.7 314 266

1560-1630) o/-genitive η % 159 41.6 47 31.5 206

o/-genitive η % 90 49.5 210 68.4 137 69.9 58 76.3 495

200

Anette Rosenbach, Dieter Stein and Letizia Vezzosi

Table Vb. Possession and valency 1500-1559 s-genitive η % 51.2 +prototype 87 114 -prototype 47.1 29.4 Agent 20 Object 10.3 3 Total 224

(1500-1559 and 1560-1630) 1560-1630 o^genitive s-genitive η % η % 83 48.8 143 69.1 128 52.9 122 61.0 70.6 52 48 57.1 26 89.7 8 24.2 285 325

o/-genitive η % 64 30.9 78 39.0 39 42.9 25 75.8 206

Table Via. Interaction: topicality (givenness) and possession/valency (1400-1449) LOCALLY GIVEN GLOBALLY GIVEN o/genitive s-genitive o/-genitive s-genitive η % η η % η % % 33.9 66.1 2 18.2 9 81.8 +prototype 20 39 14 -prototype 37 28.9 91 71.1 100 Agent 70 81.4 22 100 16 18.6 89.4 10 100 Object 7 10.6 59 2 55 Total 80 259 Table VIb. Interaction: topicality (givenness) and possession/valency (1450-1499) GLOBALLY GIVEN LOCALLY GIVEN s-genitive o/-genitive s-genitive o/-genitive η % η % η % η % 47.2 65.2 +prototype 84 52.8 75 8 34.8 15 41 92 35.2 169 64.8 5 10.9 89.1 -prototype 2 42 37.5 95 62.5 4.5 95.5 Agent 57 32 64.0 26 100 Object 18 36.0 124 371 15 Total 251 Table Vic. Interaction: topicality (givenness) and possession/valency (1500-1559) GLOBALLY GIVEN LOCALLY GIVEN s-genitive o/-genitive s-genitive o/genitive η % η % η % η % 70 49.3 13 46.4 +prototype 72 50.7 15 53.6 12 72.1 -prototype 102 51.3 97 48.7 27.9 31 40 72.7 38.5 61.5 Agent 15 27.3 5 8 1 Object 2 9.1 20 90.9 14.3 6 85.7 33 58 Total 191 227

On the history of the s-genitive Table VId. Interaction: topicality (givenness) and GLOBALLY GIVEN j-genitive o/-genitive η % η % 36.4 +prototype 91 63.6 52 36.4 63.6 60 -prototype 105 55.1 Agent 22 44.9 27 80.0 Object 5 20.0 20 223 159 Total

201

possession/valency (1560-1630) LOCALLY GIVEN s-genitive o^genitive η % η % 52 12 81.3 18.7 17 48.6 18 51.4 30 12 71.4 28.6 3 5 37.5 62.5 102 47

Appendix 2 1400-1449:

portions

readfrom

Helsinki Corpus

Chambers, R.W. and M. Daunt (eds.) 1967 A Book of London English 1384-1425. Oxford: Clarendon Press. First published [1931]. Davis, N. (ed.) 1971 Paston Letters and Papers of the Fifteenth Century, Part 1. Oxford: Clarendon Press. Dickins, B., and R.M. Wilson (eds.) 1956 Early Middle English Texts. London: Bowes & Bowes. First published [1951]. Fisher, J.H., M. Richardson and J.L. Fisher (eds.) 1984 An Anthology of Chancery English. Knoxville: The University of Tennessee Press. Francis, W. N. (ed.) 1942 The Book of Vices and Virtues. A Fourteenth Century English Translation of The Somme Le Roi of Lorens D Orleans. (Early English Text Society 217.) London: Oxford University Press. Gaytryge, Dan Jon 1969 Dan Jon Gaytryge's Sermon. Religious Pieces in Prose & Verse. Edited by G.G. Perry. (Early English Text Society Original Series 26.) London: Oxford University Press. First edition [1914]. Kempe, Margery 1940 The Book of Margery Kempe, Volume 1. Edited by S. B. Meech and H.E. Allen. (Early English Text Society 212.) London: Oxford University Press.

202

Anette Rosenbach, Dieter Stein andLetizia Vezzosi

Mandeville, Sir John 1919 Mandeville's Travels Translated from the French of Jean D Outremeuse, Volume 1. Edited by P. Hamelius. (Early English Text Society Original Series 153.) London: Oxford University Press. Rolle, Richard The Bee and the Stork. A Handbook of Middle English. Edited by 1952 F. Mosse. Translated by J.A. Walker. Baltimore: The Johns Hopkins Press. Rolle, Richard 1921 English Prose Treatises of Richard Rolle de Hampole. Edited by G. Perry. (Early English Text Society Original Series 20.) London: Oxford University Press. First published London: N. Trübner & Co. [1866].

1450-1499: portions readfrom Anonymous 1963

Helsinki Corpus

The Statutes of the Realm. Printed by Command of his Majesty King George the Third in Pursuance of an Address of the House of Commons of Great Britain, Volume 2. London: Dawson of Pall Mall. First published [1816],

Capgrave, John 1983 John Capgrave's Abbreviacion of Chronicles. Edited by P. J. Lucas. (Early English Text Society 285.) London: Oxford University Press. Capgrave, John 1971 John Capgrave's lives of St. Augustine and St. Gilbert of Sempringham and a sermon. Edited by J. J. Munro. (Early English Text Society Original Series 140.) London: Oxford University Press. First edition [1910]. Caxton, William 1956 The Prologues and Epilogues. Edited by W. J. B. Crotch. (Early English Text Society 176.) London: Oxford University Press. First edition [1928], Caxton, William 1970 The History of Reynard the Fox. Translated from the Dutch Original by William Caxton. Edited by N. F. Blake. (Early English Text Society 263.) London: Oxford University Press. Cely, George 1975 The Cely Letters 1472-1488. Edited by A. Hanham. (Early English Text Society 273.) London: Oxford University Press.

On the history of the s-genitive Davis, N. (ed.) 1971

203

Paston Letters and Papers of the Fifteenth Century, Part 1. Oxford: Clarendon Press. Fitzjames, Richard 1907 Sermo Die Lüne in Ebdomada Pasche. Westminster, Wynkyn de worde (1495?). Edited by F. Jenkinson. Cambridge: Cambridge University Press. Gregory, William 1876 The Historical Collections of a Citizen of London in the Fifteenth Century. Edited by J. Gairdner. (Camden Society, N. S. 17.) Westminster. Hilton, Walter 1967 Walter Hilton's Eight Chapters on Perfection. Edited by F. Kuriyagawa. Tokyo: The Keyo Institute of Cultural and Linguistic Studies. Horstmann, C. (ed.) 1887 The Early South-English Legendary or Lives of Saints. [The life of St. Edmund]. (Early English Text Society Original Series 87.) London: N. Trübner & Co, Julian of Norwich 1978 Julian of Norwich's Revelations of Divine Love. The Shorter Version. Edited from B.L. ADD. MS 37790. Middle English Texts. Edited by F. Beer. Heidelberg: Carl Winter Universitaetsverlag. Malory, Thomas 1954 The Works of Sir Thomas Malory. Edited by E. Vinaver. London: Oxford University Press. Methan, John 1916 The Works of John Methan Including the Romance of Amonyus and Cleopes. Edited by H. Craig. (Early English Text Society Original Series 132.) London: Oxford University Press and Kegan Paul, Trench, Trübner & Co., Ltd. Mirk, John 1905 Mirk's Festial: A Collection of Homilies by Johannes Mirkus (John Mirk), Part 1. Edited by T. Erbe. (Early English Text Society Extra Series 96.) London: Kegan Paul, Trench, Trübner & Co., Ltd. Nichols, J. G. (ed.) 1875 Two Sermons Preached by the Boy Bishop, at St. Paul's Temp. Henry VII, and at Gloucester, Temp. Mary. (Camden Society Miscellany 7. Camden Society N. S. 14.) London. Reynes, Robert 1980 The Commonplace Book of Robert Reynes of Acle. An Edition of Tanner MS 407. Edited by C. Louis. (Garland Medieval Texts 1.) New York and London: Garland.

204

Anette Rosenbach, Dieter Stein and Letizia Vezzosi

Ross, W. O. (ed.) 1940 Middle English Sermons, Edited from British Museum MS. Royal 18 Β. XXIII. (Early English Text Society 209.) London: Oxford University Press. Shillingford, John 1965 Letters and Papers of John Shillingford, Mayor of Exeter 1447-1450. Edited by S. A. Moore. (Camden Society N. S. 2.) New York. First published [1871]. Stonor, Elizabeth 1919 Stonor Letters and Papers, 1290-1483, Volumes 1-2. Edited by C. L. Kingsford. (Camden Society Third Series 29-30.) London.

Other texts Davis, N. (ed.) 1971

Paston Letters and Papers of the Fifteenth Century, 1-20. Oxford: Clarendon Press.

1500-1560: portions read from Helsinki Corpus Dickens, A. G. (ed.) 1962 Clifford Letters of the Sixteenth Century. (Surtees Society 172.) Durham and London. King Edward VI The Diary of Edward VI. Literary Remains of King Edward the 1963 Sixth, Volume 2. Edited by J. G. Nichols. (Burt Franklin Research and Source Works Series 51.) New York. First published [1857]. Ellis, H. (ed.) 1846 Original Letters, Illustrative of English History; Including Numerous Royal Letters, Third Series, Volume 1. London: Richard Bentley. Elyot, Thomas The Boke Named the Gouernour, (Everyman's Library.) Edited 1907 by E. Rhys. London and New York: J.M. Dent & Co. and E.P. Dutton & Co. First published [1531] Leland, John 1964 The Itinerary of John Leland in or about the Years 1535-1543, Volume 1, Parts 1-3. Edited by L.T. Smith, London: Centaur Press Ltd.

On the history of the s -genitive

205

Mowntayne, Thomas 1859 The Autobiography of Thomas Mowntayne. Narratives of the Days of the Reformation, Chiefly from the Manuscripts of John Foxe the Martyrologist. Edited by J.G. Nichols. (Camden Society 77.) London. Rogers, E. F. (ed.) 1947 The Correspondence of Sir Thomas More. Princeton: Princeton University Press. Roper, William. 1958 The Lyfe of Sir Thomas Moore, Knighte, Written by William Roper, Esquire, whiche Maried Margreat, Daughter of the Sayed Thomas Moore. Edited by E.V. Hitchcock. (Early English Text Society 197.) London: Oxford University Press. First edition [1935]. Torkington, Richard. 1884 Ye Oldest Diarie of Englysshe Travell: Being the Hitherto Unpublished Narrative of the Pilgrimage of Sir Richard Torkington to Jerusalem in 1517. Edited by W. J. Loftie. (The VellumParchment Shilling Series of Miscellaneous Literature 6.) London: Field & Tuer, Ye Leadenhalle Presse, E.C., ETC.

Other texts Machyn, Henry 1968 The Diary of Henry Machyn, 1-40. Edited by John Gough Nichols. London/New York: AMS Press. First published London: J. B. Nichols and Son [1848], Medwall, Henry. 1970 Fulgens and Lucrece. In: Frederick S. Boas (edj, Five PreShakespearean Comedies. London: Oxford University Press. First published [1934],

1560-1630: portions readfrom Helsinki Corpus Ascham, Roger. 1870 The Scholemaster. Written between 1563-1568. Posthumously Published. English reprints. Edited by E. Arber. London. First Edition [1570].

206

Anette Rosenbach, Dieter Stein and Letizia Vezzosi

Bacon, Francis [1970]

Forman, Simon 1849

The Twoo Bookes of the Proficience and Advancement of Learning (1605). Facsimile reprint. (English Experience 218.) Amsterdam: Theatrvm Orbis Terrarvm Ltd. and New York: Da Capo Press. The Autobiography and Personal Diary of Dr. Simon Forman, the Celebrated Astrologer, from A.D. 1552, to A.D. 1602. Edited by J.O. Halliwell. London: Privately printed.

Gifford, George [1931] A Handbook on Witches and Witchcraft. A Dialogue Concerning Witches and Witchcraftes, 1593. Introduction by B. White. (Shakespeare Association Facsimiles 1.) London: Humphrey Milford and Oxford University Press. Harman, Thomas 1937 A Caveat of Warning for Commen Cursetors Vulgarely Called Vagabones. Collated with the second edition of 1567 in the Bodleian Library, Oxford, and with the reprint of the fourth edition of 1573. Edited by E. Viles and F. Furnivall. (Early English Text Society Extra Series 9.) London: N. Trübner & Co. First edition [1869, 1898]. Hoby, Margaret 1930 Diary of Lady Margaret Hoby, 1599-1605. Edited by D.M. Meads. London: George Routledge & Sons, Ltd.. Hughey R. (ed.) 1941 The Correspondence of Lady Katherine Ρas ton, 1603—1627. (Norfolk Record Society 14.) Norwich: Norfolk Record Society. Jeayes, I. H. (ed.) 1906 Letters of Philip Gawdy of West Harling, Norfolk, and of London to Various Members of his Family, 1579-1616. London: J. B. Nichols and Sons. Madox, Richard 1976 An Elizabethan in 1582: The Diary of Richard Madox. Fellow of All Souls. Edited by E. S. Donno. London: Hakluyt Society. Markham, Gervase [1973] Countrey Contentments, 1615. Facsimile reprint. (The English Experience 613.) Amsterdam: Theatrvm Orbis Terrarvm Ltd. And New York: Da Capo Press Inc. Searle, A. (ed.) 1983 Barrington Family Letters, 1628-1632. (Camden Fourth Series 28.) London.

On the history of the s-genitive

207

Stapleton, T. (ed.) 1839 Plumpton Correspondence. A Series of Letters, Chiefly Domestick, Written in the Reigns of Edward IV. Richard III. Henry VII. And Henry VIII. (Camden Society 4.) London. Stow, John. 1580 The Chronicles of England from Brute unto this Present Yeare of Christ. London: Ralphe Newberie.

Other texts Shakespeare, William 1989 The Merry Wives of Windsor. Edited by T.W. Craik. Oxford: Clarendon Press. Shakespeare, William 1917 The Tragedy of Hamlet, Prince of Denmark. Edited by Tucker Brook and Jack Randal Crawford. New Haven: Yale University Press; London: Geoffrey Cumberlege, Oxford University Press.

Notes 1. The examples used in this paper are all taken from our corpus. 2. Unfortunately, the total number of locally-given possessors in the first interval (1400-1449) is relatively small as compared to the other intervals (cf. tables Vla-d in Appendix 1). However, the only two examples of s-genitives with locally-given possessors found in the 1400-1449 interval do occur in •prototypical relation. Only in the second interval (1450-1499) does the s-genitive with locally-given possessors extend also to -prototypical relations. 3. We also recommend Ortmann (1998) for an excellent overview of the role of animacy in inflection.

References Allen, Cynthia 1997 1998

Middle English case loss and the 'creolization' hypothesis. English Language and Linguistics 1: 63-89. Genitives and the creolization question. English Language and Linguistics2: 129-134.

208

Anette Rosenbach, Dieter Stein and Letizia Vezzosi

Chafe, Wallace 1987

Cognitive constraints on information flow. In: Rüssel Tomlin (ed.), Coherence and Grounding in Discourse, 21-51. Amsterdam/Philadelphia: John Benjamins.

Dabrowska, Eva 1998 How metaphor affects grammatical coding: The Saxon genitive in computer manuals. English Language and Linguistics 2: 121-127. den Breejen, Bastiaan 1937 The Genitive Case and its 'Of-Equivalent in the Latter Half of the Sixteenth Century. Amsterdam: H. J. Paris. Dryer, Matthew S. 1996 Focus, pragmatic presupposition and activated propositions. Journal of Pragmatics 26: 475-523. Firbas, Jan 1964 On defining the theme and functional sentence analysis. Travaux Linguistiques de Prague 1: 267-280. Fischer, Olga 1992 Syntax. In: N. Blake (ed.), Cambridge History of the English Language, Volume 2: 1066-1476, 207^108. Cambridge: Cambridge University Press. Givön, Talmy 1984 Syntax, Volume 1. Amsterdam/Philadelphia: John Benjamins. 1993 English Syntax. Amsterdam/Philadelphia: John Benjamins. Gundel, Jeanette Κ. 1988a Universals of topic-comment structure. In: Michael Hammond, Edith Moravcsik and Jessica Wirth (eds.), Studies in Syntactic Typology, 209-239. Amsterdam/Philadelphia: John Benjamins. 1988b The Role of Topic and Comment in Linguistic Theory. New York: Garland. Gundel, Jeanette Κ., Nancy Hedberg and Ron Zacharski 1993 Cognitive status and the form of referring expressions in discourse. Language 69: 274-307. Halliday, Michael A.K. 1967 Notes on transitivity and theme in English, Part 2. Journal of Linguistics 3: 199-244. Heine, Bernd 1997 Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge: Cambridge University Press. Klemola, Juhani 1997 Dialect evidence for the loss of genitive inflection in English. English Language and Linguistics 1: 350-353.

On the history of the s-genitive

209

Lambrecht, Knud 1994 Information Structure and Sentence Form: A Theory of Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Mustanoja, Tauno 1960 A Middle English Syntax, Part I. (Memoires de la Societe N^ophilologique 23.) Helsinki: Societö Niophilologique. Nichols, Johanna 1988 On alienable and inalienable possession. In: William Shipley (ed.), In Honor of Mary Haas: From the Haas Festival Conference on Native American Linguistics, 557-609. Berlin/New Yoric: Mouton de Gruyter. Ortmann, Albert 1998 The role of [± animate] in inflection. In: Ray Fabri, Albert Ortmann and Teresa Parodi (eds.), Models of Inflection, 60-84, Tübingen: Niemeyer. Prince, Ellen F. 1981 Toward a taxonomy of given-new information. In: Peter Cole (ed.), Radical Pragmatics, 223-255. New York: Academic Press. Rosenbach, Anette and Letizia Vezzosi 1997 Genitive constructions in early Modern English: New evidence from a corpus analysis. Talk presented at the workshop "Stability and variation in word-order patterns over time" at the International Conference on Historical Linguistics 1997, Düsseldorf. forthcoming Genitive constructions in early Modern English: New evidence from a corpus analysis. In: Erich Poppe, Ariel Shisha-Halevy and Rosanna Sornicola (eds.), Stability and Variation in Word-Order Patterns over Time. Amsterdam/Philadelphia: John Benjamins. Seiler, Hans Jakob 1983 Possession as an Operational Dimension of Language. Tübingen: Gunter Narr Verlag. Seppänen, Aimo 1997 The genitive and the category of case in the history of English. In: Raymond Hickey and Stanislaw Puppel (eds.), Language History and Linguistic Modelling. A Festschrift for Jacek Fisiak on his Sixtieth Birthday, Volume 1, 193-214. Berlin/New York: Mouton de Gruyter. Siewierska, Anna 1988 Word Order Rules. London: Croom Helm. Stein, Dieter 1990 The Semantics of Syntactic Change: Aspects of the Evolution of 'do' in English. Berlin/New York: Mouton de Gruyter.

210

Anette Rosenbach, Dieter Stein andLetizia Vezzosi

Taylor, John R. 1995

Linguistic Categorization: Prototypes in Linguistic Theory. Oxford: Clarendon Press. First edition [1989]. Possessives in English. Oxford: Clarendon Press.

1996 Thomas, Russell 1931 Syntactical processes involved in the development of the adnominal periphrastic genitive in the English language. PhD dissertation, University of Michigan.

The passive as an object foregrounding device in early Modern English1 Elena Seoane Posse

1.

Introduction

In this paper I examine the object foregrounding, or object topicalisation, function of the passive in the early Modern English period, that is, the assignment of subject/topic or merely topic function to a non-agent. In the case of the passive, this process involves the fact of the active object becoming passive subject, termed by Givon (1981: 168; cf. also Givon 1994: 9; Vezzosi 1996) clausal topic assignment: "the subject/agent of the active clause ceases to be the topic, and a non-actor argument of the active then assumes, by whatever means, the clausal-topic function". Other authors label this phenomenon foregrounding of the active object (Langacker and Munro 1975; Frajzyngier 1982; Foley and van Valin 1985; Keenan 1985), and interpret it as a consequence of the search for subject backgrounding (cf. Seoane Posse forthcoming).2 More specifically, the aim of this research is to ascertain the extent to which the passive voice is used in early Modern English as a device whereby a patient noun phrase endowed with inherent and discourse topicality, in Siewierska's (1984) terminology, is highlighted through its promotion to subject/topic position. As is well known, several factors determine the eligibility of a noun phrase as passive subject, namely its conveying given information, its degree of definiteness, its human and animacy features and its semantic role. Though the relevance of these aspects for present-day English passives has been examined in a number of studies (cf. Duskova 1971; Givon 1979: 57-66, 1981, 1990: 563-580; Doherty 1996; among others), little is known about their incidence in early Modern English or, for that matter, in earlier stages of English. Yet given the substantial developments in the syntax of the passive (e.g. the consolidation of the indirect and prepositional passives, the establishment of the by-agent) that were still under way in early Modern English (cf. in

212

Elena Seoane Posse

this connection Seoane Posse 1993: 199-212; Moessner 1994: 218-224), it is tempting to hypothesise that such developments may have semantic and discoursal correlates. In what follows, therefore, I will explore the role of inherent and discourse topicality in the passive constructions recorded in the early Modern English section of the Helsinki Corpus of English Texts. A brief account of the sample used is given in section 2 below, while the overall findings are expounded in sections 3 and 4.

2. The corpus The sample used is not the whole early Modern English section of the corpus, which comprises 500,000 words, but a selection from it, comprising about 153,000 words. A preliminary overview of these 153,000 words showed that the size of the sample was sufficient for my purposes. In other words, it became clear that the analysis of a larger sample would not result in the addition of novel or significant instances of the passive, nor would it modify the proportional distribution of such instances in any relevant way. Table 1 displays the number of words studied in each chronological subperiod. Table 1. Number of words examined per chronological subperiod. Number of words Subperiod 50,000 El (1500-1570) E2 (1570-1640) 48,000 E3(1640-1710) 55,000 153,000 Total

As regards text types, I thought it necessary to select text types which could provide data representative of both formal and informal settings. This decision was informed by the fact that the passive is, like any other thematic device, closely interrelated with style. Among the text types classified as formal by the compilers of the Helsinki Corpus, I singled out Statutes, Science and Sermons. Among the informal text types, I selected Private Letters, Drama and Fiction (cf. Nevalainen and Raumolin-Brunberg 1993). Table 2 displays the number of words examined by text-type. The first step consisted of identifying the passive constructions in the corpus. As can be observed in table 3, 2,256 passives were found, which constitute 22.7 per cent of all transitive clauses in the sample. The count of

The passive as an object foregrounding device

213

active constructions was restricted to those for which a passive counterpart would be available, that is, to those active transitive constructions with an overt object eligible to become passive subject. Table 2. Number of words per text-type. Text type Statutes Science Sermons Private letters Drama Fiction Total

Number of words 37,000 25,000 15,000 37,000 19,000 20,000 153,000

Table 3. Number of words examined, with indication of actives and passives and of relative frequency of passives with respect to actives. Words Actives %Actives Passives %Passives 50,000 2,236 78.5 612 21.4 El 48,000 2,550 77.9 722 E2 22.0 55,000 2,893 75.8 922 24.1 E3 Total 153,000 7,679 112 2,256 22.1 Before proceeding any further, it seems in order to point out that, though passivisation is a multifunctional thematic device, and, consequently, several factors can often determine its occurrence at any given moment, I decided to restrict the study of the object foregrounding function of the passive to passives with an overt agent, i.e. so-called agent passives. One justification for this decision lies in the fact that the primary function of agentless passives is clearly to avoid mentioning the agent, since they demote it completely by omitting it. Though object foregrounding factors may also intervene in the choice of agentless passives, such factors are secondary and are most commonly derived from the former; as Givon (1981: 189) points out, "if the identity of the agent — the most likely/unmarked clausal-topic — is suppressed, by some means the language user will confer the topic function upon the most likely remaining argument in the clause". On the contrary, agent passives lack, by definition, this agent-suppressing function and, therefore, object foregrounding factors seem more relevant. As a result, a second step in this research involved isolating the passives with an overt agent (table 4); as in other quantitative studies (cf. Svartvik

214

Elena Seoane Posse

1966 and Duskovä 1971 on present-day English passives; Baskken 1998 on early Modern English) the percentage of agent passives (15.4 per cent) was found to be much lower than that of agentless passives (84.5 per cent). Table 4. Number of agent and agentless passives. Agentless El 526 (85.9%) E2 615 (85.1%) E3 766 (83.1%) Total 1,907(84.5%)

3.

Agent 86(14.1%) 107(14.8%) 156(16.9%) 349(15.4%)

The object foregrounding function of the passive in early Modern English

3.1. Discourse

topicality

Once the agent passives in the corpus had been identified, I examined the degree of communicative dynamism (Firbas 1966: 270) of the subject and agent noun phrases in each example, that is, I checked whether they conveyed given or new information. 3 As is well known, the unmarked order of information in a clause is given/new. In contexts where the distribution of information does not comply with this principle, the subject being new and the object given, passivisation allows the unmarked given/new order to be restored due to the rearrangement of clausal elements it involves (Werth 1984: 236; Downing and Noonan 1995: 35; Baekken 1998: 375). In other words, there is a correlation between topic position and given information, which favours the choice as topics of those noun phrases conveying given information. This correlation is termed "discourse topicality" by Siewierska (1984: 222), since it is contextual, discourse factors that determine the topicality of a noun phrase.4 Baekken (1998: 307) has shown that this crosslinguistic tendency to assign the syntactic role of subject to given noun phrases is already present in early Modern English. In his corpus (a selection of prose texts from 1480 to 1730) he found an average of 84 per cent given as against 16 per cent new subjects. The type of information of the noun phrases involved was established by analysing both the extralinguistic and linguistic contexts. Deictic pronouns which were found to point at entities in the situational context, for example,

The passive as an object foregrounding device

215

were considered given information, since they are within the perception and consciousness of both interlocutors. Other noun phrases classified as conveying given information were those whose referents belong to the "generically shared context or cultural knowledge" (Givon 1992: 12), and noun phrases whose referents were found to be mentioned in the prior context. Table 5 provides the results: Table 5. Information conveyed by the subject/agent in passive clauses. Given/new 211 (60.4%) New/new 97 (27.7%) Given/given 27 (7.7%) New /given 14 (4.0%) Total 349(100%) As shown in table 5, most agent passive clauses, namely 60.4 per cent, contain given patients and new agents, while the reverse order, new/given, is observed in only 4.0 per cent of passives. These data indicate that, in the corpus, contexts in which the agent is new and the patient is given information favour the use of the passive construction, since only by rendering the clause in the passive voice can the unmarked given/new order be achieved. Therefore the discourse topicality of the noun phrases involved (i.e. the type of information they convey) is confirmed as an important triggering factor for the use of the passive as an information rearranging device in early Modern English. The significance of this parameter for early Modern English passives is further corroborated by some of Baskken's (1998: 320) findings. He examined the distribution of information in Object-Verb-Subject clauses, that is, active transitive clauses with fronted objects (such as all this saw the boy), and found that only 9.6 per cent of the total had the unmarked given/new distribution of information. This indicates that fronting strategies in early Modern English, as in present-day English (Siewierska 1984: 231-237), do not generally serve as order rearranging devices alternative to the passive in those cases where the agent is new and the patient given, a finding which helps explain the high incidence of passives with this function in the corpus. However, examples such as (1) show that, at times, the patients in passive constructions are topicalised for reasons other than their conveying given information. In (1), the patient-subject (the /Äa/-clause) is new and the agent, thadvyse and auctoritie aforesaid, is given; the distribution of

216

Elena Seoane Posse

information is nevertheless unmarked because the order of subject and agent has been reversed, the agent appearing before the subject. In other words, the example above has the order given agent + new patient, an order which could also have been attained by means of an active clause. (1)

And be it furthermore ordeyned and enacted by thadvyse and auctoritie aforesaid that the Kyng our Sov~ayn Lord or eny other p~sones take not any advantage or p~fuyt of any penalties of forfaitures by an Act made in the p~liament holden at Westm~ the xxiij day of Januare in the first yere of the Reign of Richard the third late in dede and not in right Kyng of Englond conc-nyng the makyng and drapyng of wollen Clothes geven lymyted or assigned for eny cloth made or hereafter to be made but oonly accordyng to the seid ordenanc~ and statut~ nowe made in this p~sent p~liament, Eny acte statute ordenance or p~vysion to the contPry hertofore made notwithstondyng. (QE1 S T A L A WSTAT3: 9).

Table 6 analyses the distribution of information in passive clauses taking into account whether the agent precedes or follows the patient-subject. Table 6. Distribution of information in passive clauses taking into account the order of constituents. Subject precedes agent Agent precedes subject 39(11.0%) Given + given 3 (0.7%) New + new 65(18.6%) 32(9.1%) 122 (34.9%) Given + new 71 (20.3%) New + given 6(1.7%) 11 (3.0%)

The data in table 6 confirm the findings shown in table 5, namely that the most common distribution of information in passive clauses in the corpus is given/new, and that, consequently, the passive is predominantly used with the aim of restoring the given/new structure of information, independently of whether the agent follows or precedes the subject. However, table 6 also shows that, in 71 passive examples (20.3 per cent of the total), the order of the semantic functions coincides with that of active clauses (agent + patient) and the distribution of information is unmarked (given + new), as illustrated in example (1) above. In other words, the same

The passive as an object foregrounding device

217

informational structure could have been achieved by resorting to the active voice. The choice of the passive voice in such sentences, therefore, is not determined by the discourse topicality of the noun phrases involved, and must be accounted for in some other way, an issue that will be addressed in the next sections.

3.2. Inherent

topicality

There are other factors that, irrespective of the given/new status of a noun phrase, can also determine its eligibility as topic. Thus, despite the fact that any noun phrase may function as topic of a clause, 5 some noun phrases are more likely to carry one of these functions than others. Statistical data gleaned from cross-language investigations show that the eligibility of nominal topics also depends on (i) the degree of animacy of the referent of the noun phrase; (ii) its semantic role, and (iii) its degree of definiteness. These properties constitute the inherent topicality of noun phrases (in Siewierska's 1984 terminology) because they render the noun phrase topical with independence of any contextual and discourse features. 6

3.2.1. Human [H] and animacy [A] features As regards the human and animacy features (Silverstein 1976; Siewierska 1984: 221; Croft 1990: 112), noun phrases which are [+H, +A] tend to occupy topic position. Situations arise in which the patient of an active clause is human and animate, while the agent is minus human, minus animate ([-H, -A]). Passivisation, by promoting the patient to topic position, has the effect of topicalising the more eligible noun phrase to become topic. In order to ascertain the extent to which the passives in the corpus serve this purpose, the human and animacy features of the subject and agent noun phrases were examined and compared. Table 7 displays all the combinations found in the corpus. However, only combinations (i) and (ii) are significant. Thus, combinations (iii) and (iv) correspond to passives in which the subject and agent share the same human and animacy features, so this parameter proves irrelevant as regards the use of the passive. Combinations (v)-(vii) correspond to passives in which either the passive agent, or the subject, or both have an

218

Elena Seoane Posse

animal as referent. Only one instance was found for each case, so these types are poorly represented in the corpus and do not warrant any conclusions. Table 7. Human and animacy features of the subject (S) and agent (A). Ν % 33 9.4 S=[+H, +A] / A+[-H, -A] (0 S=[-H, -A] / A=[+H, +A] 84 24.1 (ii) (iii) 34 9.7 S=[+H, +A] / A=[+H, +A] S=[-H, -A] / A=[-H, -A] 69 19.7 (iv) S=[+H, +A] / A=[-H, +A] 1 0.2 (v) S=[-H, +A] / A=[+H, +A] 1 0.2 (vi) S=[-H, +A] / A=[-H, +A] 1 0.2 (νϋ) (viii) S=relative as / A=[+H, +A] 8 2.2 (ix) S=relative as / A=[-H, -A] 15 4.2 A=[-H, -A] / S=clause 14 4.0 (x) 89 (xi) Α=Γ+Η, +A] / S=clause 25.5 Total 349 100 Combinations (viii)—(ix) in table 7 correspond to passives in as-comment clauses, such as the one in (2) below, where the subject is the relative as and hence acts as structural and textual connector with respect to the preceding context. Lastly, in combinations (x)-(xi), as in (3), the subject is a clause. Consequently, comparison between the subject and agent as regards the animacy and human features is out of the question. (2)

And that ev~y Capteigne ... shall uppon the payne afore seid pay... the Wag~ ratably as is allowed unto theym by the Kyng oure Sov~aigne Lorde or the Tresourer of his Warres without lessyng or withdrawyng of any parte therof. (I QE1 STA LAW STAT3: 6)

(3)

Provided alwaies and be it enacted by the authoritie of this p~sent Parliament, That the Correcc-on and Punishment of such as shall offend againste this Acte, or any parte therof within either of the two Univ-sities of this Realme, or the p~cinct~ or Liberties of the same, shall be done upon the Offenders, and Justice shall be ministred in this behalfe accordinge to the intent and true meaninge of this Lawe, ... (|QE2_STA_LA WSTAT4: 15)

The passive as an object foregrounding device

219

Coming back to combinations (i) and (ii) in table 7, these patterns show that only 9.4 per cent of passives topicalise a patient noun phrase whose referent is higher than the agent in the animacy hierarchy, a case illustrated in (4). The reverse case, that in which a [-H, -A] patient is promoted to topic position despite the fact that the agent is [+H, +A] and should therefore have become topic, occurs in more than 24 per cent of the passives, as in (5). (4)

a Civil man is never bitter against a Friend or a Stranger, much less to him that enters under his Roof, and is secured by the Laws of Hospitality. (|QE3_IR_SERM_ JETAYLO: 18).

(5)

But as the Earth, the Mother of all Creatures here below, sends up all its Vapours and proper emissions at the command of the Sun ... So are the proprieties of a Wife to be dispos 'd of by her Lord; ... (|QE3_IR_SERM_JETAYLO: 12).

In the vast majority of cases, therefore, agents outrank subjects on the animacy hierarchy. We must then conclude that, unlike in present-day English (Givon 1979: 58; Siewierska 1984: 222), the animacy and human features of noun phrases are not a determining factor for the use of the passive in early Modern English, at least to the same extent as today. This finding seems to go against the cross-linguistic tendency to talk about human subjects (see, among others, Siewierska 1984: 221; Givon 1979: 58), and appears to confirm the claims of Strang (1970: 151), Söderlind (1951-58: 24) and Moessner (1994: 226) as regards earlier periods of the language. Strang (1970: 151) argues, in relation to indirect passives, that, though they are still rare in early Modern English, their increasing frequency in this period is "one aspect of a yet wider tendency to prefer human, specially first person subjects, where possible". In a similar line, Moessner (1994: 226) points out that "the indirect passive has the advantage that it allows the foregrounding of the person-denoting object; this may be one of the reasons why it developed at all". The data in the corpus, as shown in table 7, lead to the same conclusion, namely that the tendency to topicalise noun phrases with human referents is only in its inception in early Modern English.

220

Elena Seoane Posse

The inherent topicality of a noun phrase is also said to correlate with rank in the empathy hierarchy (Kuno and Kaburaki 1977; DeLancey 1981). According to this, speech act participants, that is, noun phrases with first and second person referents, outrank human third person referents, which in turn outrank non-human animate and inanimate referents. Among the 349 agent passives in the corpus, only 14 involve speech act participants, specifically, eight first person singular pronouns, three first person plural pronouns and three second person singular pronouns. As can be observed in table 8, most of the speech act participants (12 out of 14) are patient noun phrases which have been topicalised via passivisation, and thus conform to the tendency for speech act participants to occupy subject/topic position. Table 8. Speech act participants (SAP) as subject (S) and agent (A). S=SAP / A=[-H, -A] 7 (50.0%) 4 (28.5%) S=SAP / A=[+H, +A] S=SAP / A=[-H, +A] 1 (7.1%) 2 (14.2%) S=r-H, -Al / A=SAP 14 (100%) Total There are, however, two instances, illustrated in (6) and (7), in which this tendency is reversed, since an agent speech act participant is relegated to final position and to the status of oblique adjunct. (6)

I may hope to see you at Easter, which time will be much longed for by me. (|QE2_XX_CORP_HARLEY: 90)

(7)

.... which seeming to contradict what has been observed by others and by us also, that Amber... (|QE3_EX_SCIO_BOYLE: 59)

These examples would hardly be possible today (in this connection, see further section 3.2.3 below). Present-day English would definitely prefer an active construction with the speech act participant in subject position (e.g. a time I will much long for/what we and others have observed). From the occurrence of these two passives, in which a [-H, -A] noun phrase is singled out as topic at the expense of a speech act participant, we might tentatively infer that the present-day English tendency reflected in the

The passive as an object foregrounding

device

221

empathy hierarchy was not yet firmly established by early Modern English times. This hypothesis, however, needs to be confirmed by further research.

3.2.2. Semantic role The eligibility of a noun phrase as topic is also related to its semantic role. According to the semantic hierarchy, the most likely candidates to become topics are noun phrases with the role of agent, followed, in this order, by recipients/benefactives, patients and obliques (Siewierska 1984: 221; Croft 1990: 112). This hierarchy accounts, inter alia, for the higher frequency of actives over passives in English, since the former have agents as subjects/topics; secondly, it also accounts for the preference for indirect over direct passives in present-day English (cf. Huddleston 1984: 440-441), since indirect passives topicalise recipient/benefactive noun phrases, whereas direct passives topicalise patients. 7 In order to ascertain whether the semantic role of noun phrases has a bearing on the passive examples in the corpus, I analysed the semantic role of the topicalised noun phrase in the passives of ditransitive clauses. As is obvious, passives from monotransitive clauses were not examined, since they only have the possibility of topicalising the patient noun phrase. As for passives from ditransitive actives, the noun phrase subject of the passive can be a patient, if it is the direct object that is promoted to subject, or a recipient/benefactive in the case of indirect objects promoted to passive subjects (Denison 1993: 103). When analysing the animacy features and degree of definiteness of the noun phrases involved in passivisation, it is necessary to compare the subject and agent noun phrases (cf. sections 3.2.1 and 3.2.3); for the study of the semantic role, however, only the topicalised noun phrase is relevant, regardless of whether there is an overt agent or not. Therefore to examine this particular variable I have included both agent and agentless passives from ditransitive actives, as can be seen in table 9. Table 9. Passives from ditransitive actives. Indirect object as subject 5 (25%) Agent passives 27 (22.5%) Agentless passives 32 (22.8%) Total

Direct object as subject 15(75%) 93 (77.5%) 108 (77.1%)

222

Elena Seoane

Posse

As an illustration, consider the behaviour of TELL as a passive ditransitive verb. There are only two instances: the first, example (8), has the direct object as subject; the second is the kind of passive favoured in present-day English, with the indirect object as subject (9). As might have been expected, the first dates back to El, the second to E3. (8)

Come on fellow it is tolde me thou art a shrew iwysse, Thy neighbours hens y=u= takest, and pi ay es the two leggedfaxe (|QEl_XX_COME_STEVENS: 54)

(9)

I was told this day that the heralds had yet a quarter of their work to do: (|QE3_XX_CORP_ANHATTO: 98)

To judge from the findings of the present study, the semantic role of the noun phrases involved in passivisation is less important in early Modern English than in present-day English: unlike in present-day English, topicalised benefactives (i.e. indirect objects) show lower percentages in the corpus, namely 22.8 per cent, than topicalised patients (i.e. direct objects), which average 77.1 per cent. In other words, the present-day English tendency to topicalise noun phrases with roles high up on the semantic hierarchy is not observed in my early Modern English corpus. However, as is well known, passivisation of indirect objects is greatly restricted in early Modern English not really because of semantic factors, but rather because the indirect passive was only recently available in the language and was still undergoing a process of consolidation. As Strang (1970: 151) notes, the tendency to prefer personal indirect objects as subjects can already be seen at work in early Modern English, as is demonstrated by the fact that indirect passives become gradually consolidated in this period. However, she also admits that this trend is only in its inception, as seems to be corroborated by the preponderance of direct passives over indirect passives in my corpus (cf. also Denison 1993: 110-112).

3.2.3. Degree of definiteness Thus far, I have dealt with two of the three factors that can make a noun phrase inherently topical and hence eligible to become passive subject,

The passive as an object foregrounding device

223

namely its position in the animacy hierarchy and its semantic role. A third factor contributing to the inherent topicality of the patient noun phrase is its degree of definiteness. The relation between definiteness and topicality lies in that definite noun phrases are topics "which the speaker assumes the hearer can identify uniquely, is familiar with, are within his file (or register) and thus available for quick retrieval" (Givon 1983: 10 [emphasis his]). Definite noun phrases are used when the speaker believes that his interlocutor will be able to identify, from the possible range of referents, the one he has in mind (Chafe 1976: 39). On the contrary, indefinite noun phrases are normally introduced for the first time, and are consequently not familiar and more difficult to identify. Hence, the choice of indefinite noun phrases as clausal topics is rarer; topic noun phrases are most commonly definite, since they convey given information. To check whether passivisation in the corpus was or was not determined by degree of definiteness I examined the subject and agent noun phrases in each agent passive example. Passives in which either the agent or the subject was realised by means of a clause (126 examples in all) were excluded, since no comparison regarding degree of definiteness is possible in such circumstances. The noun phrases involved in the remaining 223 agent passives were found to fall into the following four classes, ordered from most to least definite: •

• • •

Pronouns: both personal pronouns and relative pronouns, which involve prior mention in the discourse and whose referents can be unambiguously identified Proper N, a category including both proper nouns and proper names (cf. Huddleston 1984: 229-30), which label specific referents directly 8 Definite noun phrases, which refer either to unique referents, such as the moon, or to referents that have been mentioned before Indefinite noun phrases, which introduce referents that the hearer/reader cannot identify and are new information 9

Tables 10-12 show the results of the analysis. Table 10 provides the figures for those patterns in which subject and agent are on a par as regards definiteness.

224

Elena Seoane Posse

Table 10. Subject and agent noun phrases in terms of degree of definiteness (same degree of definiteness). Definite noun phrase/definite noun phrase 63 (28.2%) Indefinite noun phrase/indefinite noun phrase 14 (6.2%) Pronoun/pronoun 7(3.1%) Proper N/proper Ν 4(1.7%) Total 88 (39.4%) Table 11 provides the figures for the subject/agent combinations in which the subject is more definite than the agent; the percentages for each of the four combinations recorded (cf. examples [10]-[13]) are relative to the total number of clauses examined, namely 223. As can be seen in table 11, 42.6 per cent of the agent passives in the corpus have a subject more definite than the agent. Furthermore, in the majority of the passives within this group, namely in 6310 out of 95, the subject is a pronoun, i.e. the most definite type of constituent. Table 11. Subject and agent noun phrases in terms of degree of definiteness (subject more definite than the agent). 39 (17.4%) Pronoun/definite noun phrase 32 (14.3%) Definite noun phrase/indefinite noun phrase Pronoun/indefinite noun phrase 14 (6.2%) 10(4.4%) Pronoun/proper Ν 95 (42.6%) Total (10)

And where as I am enformed by my sone Heron of the losse of our barns and our neighbours also with all the come that was therin,... (|QE l_XX_CORP_MORELET: 96)

(11)

Forasmuche as the science and connyng of Physyke and Surg~ie ... ys daily within this Royalme exc~cised by a grete multitude of ignoraunt p~sones of whom the grete partie have no man~ of insight in the same nor in any other kynde of lernyng... (|QE 1 _STA_LAW_STAT3: 11)

(12)

for stubble though it be quickly kindled, yet it is as soon extinguished, if it be not blown by a pertinacious breath, or fed with new materials. (|QE3_IR_SERM_JETAYLO: 10)

The passive as an object foregrounding device (13)

225

but for as much as those tables be not altogether truely Printed, and for that they haue beene lately corrected, and made perfect by (AClauiusA), who doth set downe the saide Tables in quarto and not in folio, ... (|QE2_EX_SCIO_BLUNDEV: 14)

Table 12 shows the figures for passives in which the subject is less definite than the agent; the percentages, as before, are relative to the total number of passives analysed (223). Most of the agent passives of this type have an indefinite noun phrase as subject and a definite noun phrase as agenζ as in (14). Table 12. Subject and agent noun phrases in terms of degree of definiteness (subject less definite than the agent). 25(11.2%) Indefinite noun phrase/definite noun phrase Definite noun phrase/pronoun 13 (5.8%) 2 (0.8 %) Indefinite noun phrase/pronoun 40 (17.9%) Total (14)

To the end that an Accompt may bee taken by the said Wardens or theire Deputy or Deputies thereof upon every Person and Persons offending herein shall forfeite (|QE3_STA_LAW

Master and Paine that and pay... STAT7: 7).

Particularly remarkable are the other two patterns represented in table 12; in both, the subject is more indefinite than the agent, since this latter is realised by means of a pronoun, the category of nominal that comes highest on the definiteness hierarchy. According to Siewierska (1984: 224), passive clauses with pronominal agents are rare in present-day English, and "for many speakers of English are acceptable only under a contrastive interpretation". Among pronominal agents, first person human agents are least likely to be demoted to passive agent position, since they are at the top of the animacy, definiteness and empathy hierarchies, and are, therefore, the most suitable candidates to be topics (cf. Kuno and Kaburaki 1977: 650). Yet the corpus provides two instances of this type, illustrated in examples (6) and (7) above. Finally, example (15) illustrates the third pattern in table 12, namely indefinite noun phrase/pronoun.

226

(15)

Elena Seoane Posse

Nor secondly, is this last more difficult to be explicated, then that a Body, as Silver for Instance, put into a weak (AMenstruumA), as unrectified Aqua fortis should, when it is put in a great heat, be there dissolved by it, and not before; (|QE3_EX_SCI0_H00KE: 9).

In sum, my findings reveal that 135 agent passives (cf. the totals in tables 11 and 12) exhibit a discrepancy between the degree of definiteness of the subject and the agent, and that, in the majority of these cases, namely 95, the patient is more definite and has been promoted to subject position. Degree of definiteness, therefore, is an important determining factor for passivisation in my data. The remaining 40 cases, in which a patient is topicalised despite being less definite than the agent, must be accounted for by reference to some of the other linguistic parameters examined in this paper.

4.

Summary and conclusions

The analysis of the object foregrounding function of the passive in early Modern English suggests that, of the two variables explored in this paper, namely discourse topicality and inherent topicality, only the first was well established as an important factor behind the use of the passive as a topicalisation device. As shown in section 3.1.1, which analysed the discourse topicality of the subject and agent noun phrases, 211 passives out of the 349 with an overt agent have given patients and new agents, which indicates that in early Modern English, as in present-day English, one of the functions of the passive voice is to restore the informationally unmarked given/new order of information by foregrounding given patients. With respect to the inherent topicality variable, in contrast with presentday English, the animacy and human features of the subject and agent do not trigger passivisation in the corpus, for only 9.4 per cent of the passives exhibit the unmarked combination [+H, +A] subject and [-H, -A] agent (cf. table 7). Similarly, the universal tendency to foreground speech act participants fails to apply in 14.2 per cent of the cases (cf. table 8). These data run counter to the "strong egocentric bias" exhibited by human discourse crosslinguistically (Siewierska 1984: 221) and confirm the claims of earlier

The passive as an object foregrounding

device

227

research (Strang 1970: 151; Moessner 1994: 226) that the current preference for human, animate subjects was only in its inception in early Modern English. Another remarkable difference between early Modern English and present-day English passives concerns the semantic role of the promoted noun phrase: the present-day English tendency to topicalise noun phrases with a benefactive role (i.e. indirect objects) over those with a patient role (i.e. direct objects) does not apply to early Modern English, since only 32 out of the 140 passives from ditransitive structures topicalise benefactives. As benefactives are usually human and/or animate, this finding ties in with the one mentioned in the previous paragraph, where I drew attention to the relatively unimportant role of the human/animacy parameter for early Modern English passives. What seems clear, at any rate, is that the passivisation of indirect objects was still greatly restricted in the period under discussion, thus confirming that the indirect passive itself, though available in the language since at least the fifteenth century, was not yet well established. Lastly, of the factors that render patient noun phrases inherently topical and that, therefore, might determine their promotion to topic position through passivisation, only the degree of definiteness of subject and agent was found to play a significant role in early Modern English passives. More particularly, 42.6 per cent of the patients ranked higher than the agent in the definiteness hierarchy (cf. table 11). Examination of this variable thus yielded similar results to those obtained from the study of the type of information conveyed by subject and agent noun phrases, which seems to indicate that the strong connection existing in present-day English between given information and definiteness is already at work in early Modern English. To conclude, one word on object foregrounding in relation to text type. The various registers represented in my corpus were found to behave very much alike with respect to the linguistic parameters examined in this paper. This suggests that the factors controlling object foregrounding in passives are not sensitive to style and apply generally across all kinds of texts. Further research based on a larger sample is however needed in order to confirm this hypothesis.

228

Elena Seoane Posse

Notes 1. Research for this paper was funded by the Spanish Ministry of Education and Science (DGICYT grant numbers PB 94-0619 and PB97-0507). I am also grateful to Professor Teresa Fanego for her valuable comments at various stages and to an anonymous reviewer of this volume for a number of helpful criticisms and suggestions. 2. As employed in this paper, the terms "foregrounding" and "backgrounding" are only distantly related to their use in text linguistics (cf. Enkvist and Warvik 1987; Warvik 1990), where they refer to the distinction between foregrounded clauses (i.e. those carrying the most essential, main line narrative material) and secondary, supportive, or backgrounded, clauses. 3. Chafe (1987: 36, 1994: 72) proposes a scale from given to new, applying the labels "already active" to given information, "previously inactive" to new information, and "previously semi-active" to an intermediate category. The crucial question in debating the passive as an information-rearranging device is the difference between subject and agent noun phrases as regards communicative dynamism, and not so much the degrees of newness or givenness of each noun phrase individually. For this reason, the givenness scale is disregarded in this paper and only the binary distinction given versus new is taken into account. 4. A different position is taken by Givön (1992: 16-17). According to him, two discourse-pragmatic aspects turn out to be relevant for discourse topicality, namely (i) referential accessibility, which may be assessed basically by means of "referential distance" (number of clauses from the last occurrence of the noun phrase in the preceding discourse) and (ii) thematic importance, which is measured in terms of "topic persistence" or occurrence of the participant in the ten subsequent clauses, and "overall persistence" or total number of times the same referent appears as clausal argument in the discourse. However, Givön's parameters prove difficult to apply to the agent passives in the corpus, since these occur in stretches of discourse varying widely in size, from ten lines in the case of Private Letters to 150 lines in Law. In this connection see also Baekken (1998: 301), who aptly points out that there is no agreement as to the extent of the span of discourse during which an element will remain present in the consciousness of the interlocutor. 5. Constituents other than noun phrases (e.g. verbs, adverbs) can also function as topics, but these non-nominal topics are not relevant for a discussion of the passive and hence are disregarded in this paper. 6. This is somewhat of an oversimplification, for it must be admitted that the inherent topicality of a noun phrase is directly related to the type of information it conveys. Concerning degree of definiteness and semantic hierarchy, for instance, definite agents are the most eligible constituents to be topic, since they prototypically convey old information. We are obviously moving among interrelated functions which are not always easy to distinguish.

The passive as an object foregrounding device

229

7. The relationship between semantic role and choice of voice has also been accounted for in terms of attention flow and viewpoint (DeLancey 1981; Marin Arrese 1997). Attention flow is the order in which the speaker wishes the hearer to attend to the participants involved in the event, and its basis is the temporal order of the phases of the event. Therefore the unmarked linguistic attention flow in a transitive sentence is from agent to patient, since the agent is "the first mover in a transitive event" (DeLancey 1981: 650). According to DeLancey, the passive is a mechanism for reversing the natural attention flow in those cases where the patient is a speech act participant (i.e. a first or second person referent; see section 3.2.1 above) and thus inherently eligible for viewpoint status, that is, topic/initial position. 8. Despite the inherent definiteness of proper nouns and proper names, I follow Croft (1990: 127) in ranking them after pronouns in the definiteness hierarchy. 9. Indefinite topics are possible whenever they are used generically, as in a chair is a piece of furniture to sit on, where the entire class of chairs can be assumed to be a concept familiar to any speaker. 10. The figure 63 is obtained by adding up 39 passives with the combination pronoun/definite noun phrase, 14 where it is pronoun/indefinite noun phrase and ten of the pattern pronoun/proper Ν (cf. table 11).

References Baskken, Bjerg 1998

Word Order Patterns in Early Modern English with Special Reference to the Position of the Subject and the Finite Verb. (Studia Anglistica Norvegica 9.) Oslo: Novus Press. Chafe, Wallace L. 1976 Givenness, contrastiveness, definiteness, subjects, topics and points of view. In: Charles N. Li (ed.), Subject and Topic, 25-55. New York: Academic Press 1987 Cognitive constraints on information flow. In: Russell S. Tomlin (ed.), Coherence and Grounding in Discourse, 21-51. (Typological Studies in Language 11.) Amsterdam/Philadelphia: John Benjamins. 1994 Discourse, Consciousness and Time. Chicago/London: The University of Chicago Press. Croft, William 1990 Typology and Universals. Cambridge: Cambridge University Press. DeLancey, Scott 1981 An interpretation of split ergativity and related patterns. Language 57: 627-657.

230

Elena Seoane Posse

Denison, David 1993

English Historical Syntax: Verbal Constructions. London/New York: Longman

Doherty, Monika 1996 Passive perspectives; different preferences in English and German: A result of parameterized processing. Linguistics 34: 591-643. Downing, Pamela and Michael Noonan 1995 Word-Order in Discourse. Amsterdam/Philadelphia: John Benjamins. Duskovä, Libuse 1971 On some functional and stylistic aspects of the passive voice in present-day English. Philologica Pragensia 14: 117-143. Enkvist, Nils Erik and Brita Warvik 1987 Old English pa, temporal chains and narrative structure. In: Anna Giacalone Ramat, Onofirio Carruba and Giuliano Bernini (eds.), Papers from the Seventh International Conference on Historical Linguistics, 221-237. (Current Issues in Linguistic Theory 49.) Amsterdam/Philadelphia: John Benjamins. Firbas, Jan 1966 On defining the theme in functional sentence analysis. Travaux Linguistiques de Prague 1: 267-280. Foley, William A. and Robert D. van Valin, 1985 Information packaging in the clause. In: Timothy Shopen (ed.), Language Typology and Syntactic Description, Volume 1, 299-347. Cambridge: Cambridge University Press. Frajzyngier, Zygmunt 1982 Indefinite agent, passive and impersonal passive: A functional study. Lingua 58: 267-290. Givon, Talmy 1979 On Understanding Grammar. New York: Academic Press. 1981 Typology and functional domains. Studies in Language 5: 163-193. 1983 Topic continuity in discourse: An introduction. In: Talmy Givon (ed.), Topic Continuity in Discourse. A Quantitive Cross-language Study, 1-41. (Typological Studies in Language 3.) Amsterdam/Philadelphia: John Benjamins. 1990 Syntax. A Functional Typological Introduction, Volume 2. Amsterdam/Philadelphia: John Benjamins. 1992 The grammar of referential coherence as mental processing instructions. Linguistics 30: 5-55. 1994 The pragmatics of de-transitive voice: Functional and typological aspects of inversion. In: Talmy Givön (ed.), Voice and Inversion, 3-44. (Typological Studies in Language 28.) Amsterdam/Philadelphia: John Benjamins.

The passive as an object foregrounding device

231

Huddleston, Rodney 1984 Introduction to the Grammar of English. Cambridge: Cambridge University Press. Keenan, Edward L. 1985 Passive in the world's language. In: Timothy Shopen (ed.), Language Typology and Syntactic Description, Volume I, 243-281. Cambridge: Cambridge University Press. Kuno, S and E. Kaburaki 1977 Empathy and syntax. Linguistic Inquiry 8: 627-673. Langacker, Ronald W. and Pamela Munro 1975 Passives and their meaning. Language 51: 789-830. Marin Arrese, Juana I. 1997 Cognitive and discourse-pragmatic factors in passivisation. Atlantis 19: 203-218. Moessner, Lilo 1994 Early Modern English passive constructions. In: Dieter Kastovsky (ed.), Studies in Early Modern English, 217-31. Berlin/New York: Mouton de Gruyter. Nevalainen, Terttu and Helena Raumolin-Brunberg 1993 Early Modern British English. In: Matti Rissanen, Merja Kytö and Minna Palander-Collin (eds.), Early English in the Computer Age: Explorations through the Helsinki Corpus, 53—73. Berlin/New York: Mouton de Gruyter. Rissanen, Matti, Ossi Ihalainen and Merja Kytö (compilers), 1991 Helsinki Corpus of English Texts: Diachronic and Dialectal. Helsinki: University of Helsinki. Seoane Posse, Elena 1993 The passive in early Modem English. Atlantis 15: 191-213. forthcoming Impersonalising strategies in early Modern English. English Studies. Siewierska, Anna 1984 The Passive: A Comparative Linguistic Analysis. London: Croom Helm. Silverstein, Michael 1976 Hierarchies of features and ergativity. In: Robert M.W. Dixon (ed.), Grammatical Categories in Australian Languages, 112—171. Canberra: Australian Institute of Aboriginal Studies. Söderlind, Johannes 1951-58 Verb Syntax in John Dryden's Prose, Volumes 1 and 2. Uppsala: Lundequistska Bokhandeln. Strang, Barbara Μ. H. 1970 A History of English. London/New York: Routledge.

232

Elena Seoane Posse

Svartvik, Jan 1966 On Voice in the English Verb. The Hague/Paris: Mouton Vezzosi, Letizia 1996 Passive constructions: The case of Anglo-Saxon. Paper delivered at the Ninth International Conference on English Historical Linguistics, Poznan, 26-31 August 1996. Warvik, Brita On grounding in English narratives: A diachronic perspective. In: 1990 Sylvia Adamson, Vivien Law, Nigel Vincent and Susan Wright (eds.), Papers from the Fifth International Conference on English Historical Linguistics, 559-575. Amsterdam/Philadelphia: John Benjamins. Werth, Paul 1984 Focus, Coherence and Emphasis. London: Croom Helm.

Reinforcing adjectives: A cognitive semantic perspective on grammaticalisation1 Carita Paradis

1.

Introduction With the birth of cognitive semantics, new ideas from the field of theoretical semantics have found their way to the study of meaning changes, and that should not come as a surprise: since one of the major things cognitive semantics is interested in is polysemy — and polysemy is, roughly, the synchronic reflection of diachronic semantic change. (Geeraerts 1997: 6)

This passage sparked off the idea of a study where I could combine my general interest in both cognitive semantics and grammaticalisation theory with a more specific interest in the lexemes in the domain of degree (Paradis 1997; forthcoming). I was intrigued by what happens semantically when lexemes are recruited to the domain of degree and hence grammaticalised. In other words, what is grammaticalisation in lexical semantic terms? This paper is a case study of ten lexemes with a reinforcing reading. They are: absolute bliss a complete bitch a perfect idiot total crap utter nonsense

an awful mess a dreadful coward a horrible muddle a terrible bore extreme pleasure

In these phrases the adjectives are specifiers of degree at the same time as they convey an evaluation of the reliability of the proposition, i.e. they are epistemic markers. They are expressive in nature rather than descriptive of the nouns they apply to. However, both historically and in contemporary language they have applications at both ends of the developmental cline, in

234

Carita Paradis

that they have propositional readings (an absolute measure, an awful sight) or they can be applied as markers of degree, expressiveness and subjectivity (absolute bliss, an awful mess)? This paper is devoted to the polysemy of the above adjectives, i.e. to the constraints associated with their propositional and reinforcing readings. 3 More precisely, my aim is to account for their well-formedness and their interpretability in lexical semantic terms, both as descriptive adjectives and as grammaticalised reinforcers. Three questions are at the heart of this study: • • •

What are the semantic properties that these adjectives have in common which make it possible for them to develop a reinforcing reading? What are the semantic properties of the nouns they combine with? What are the structural constraints for a reinforcing reading?

I argue that the polysemes map on to different onomasiological domains, i.e. to the PROPOSITIONAL Χ and the EPISTEMIC/DEGREE domains. Different domains require different semantic bindings, and the interpretation and wellformedness of the actual expressions rely on different valence structures (Langacker 1988: 91-125). The noun and the adjective exert semantic pressure on one another. The properties of the noun constrain the interpretation of the adjective, and the properties of the adjective assign a perspective in which the noun is to be viewed. For instance, measure in the phrase an absolute measure profiles the expression, and the adjective binds a certain property of the noun, i.e. in this case the possibility of measurement being 'absolute' rather than 'relative'. In the expression absolute bliss, bliss profiles the expression and absolute binds the gradable property of the noun and thus assigns the perspective. The motivation for semantic shifts across domains is to be found in two powerful processes. These are subjectiflcation in grammaticalisation (Traugott 1995), which is a speakeroriented process, and relevance, which is a hearer-oriented process (Deane 1988; Blakemore 1992; Sperber and Wilson [1986] 1995;Nicolle 1998).4 The procedure of the argument is as follows: First I shall set the scene by giving a brief outline of the diachronic development of the adjectives in order to trace their semasiological path of grammaticalisation and discuss various contextual constraints on the reinforcing readings. Secondly, the development of these adjectives into grammaticalised reinforcers will be outlined in terms of their mapping on to different onomasiological domains

A cognitive semantic perspective on grammaticalisation

235

and their forming of paradigms. Finally, I shall analyse the constraints on their reinforcing readings in contrast to their propositional readings within a cognitive semantic framework and conclude by explicating their grammatical ised status semantical ly.

2. Historical development According to the Oxford English Dictionary, the reinforcing readings of these adjectives are relatively recent developments in the history of the English language. Most of them developed this reading during the early Modern period. Table 1. The approximate date of appearance of reinforcing adjectives in chronological order Reinforcing adjectives Approximate date of appearance utter 1430 1460 horrible extreme 1460 1574 absolute 1600 terrible perfect 1611 complete 1645 1647 total dreadful 1700 awful 1809 Complete, perfect, total and absolute were originally descriptive of 'completeness' and subsequently recruited into the domain of degree and reinforcement. Complete comes from 'having all its parts', 'entire', 'full', perfect from 'completed', 'accomplished', total from 'relating to the whole of something'. Absolute was originally a participle meaning 'disengaged from' (1), or 'free from imperfection' (2), which later came to be used in the sense of'complete degree' (3): (1)

Men sen it vtterly fre and absolut from alle necessite. (1374, Chaucer Boethius 175; OED s.v. absolute, I 1 adj. Obs.)

236

Carita Paradis

(2)

A young man so absolute, as yat nothing may be added to his further perfection. (1579, Lyly Euphues 123, OED s.v. absolute II 4)

(3)

The honour of its absolute sufficiency (1641, Milton Ch Discip. (1851) I 32, OED s.v. absolute, II 5a. adj.)

In both their descriptive and reinforcing readings complete, perfect, total and absolute are complementary adjectives, i.e. they are associated with a bounded, "either-or" conceptualisation (Paradis 1997: 57-58), and they can be limited by, for instance, almost. That is, either something is complete or almost complete, whole or almost whole, accomplished or almost accomplished. In their capacity as reinforcing adjectives, a generalised property of the degree of totality has taken over completely, and they have been recruited as markers of reinforcement by implication.5 Utter and extreme were originally descriptive of outermost locations in space (4), and for extreme also outermost location in time, e.g. the extreme unction. This reading then lent itself to a degree reading (5): (4)

Chichester is in the extream part of the shire. (1503, Act 19 Hen VII, c 24, OED s.v. extreme 1 adj)

(5)

Theire adherents made extreme resistens. (1512, Act 4 Hen VIII, c 20 Pream, OED s.v. extreme 4 adj)

The original meanings of awful, dreadful, horrible, terrible were 'awe/dread/horror/terror-causing' (6). These senses suggest an extreme point on a scale. They are what we might call inherent superlatives (Paradis 1997: 54-57). The negative superlative property that characterises these adjectives is recruited for reinforcement. (6)

They reared thence vnto the Saxons such awefulle armies. (1602, Warner Alb. Eng. Epit. (1612) 360, OED s.v. awful 1 adj)

(7)

To what an awful extent the Spanish peasant will consume garlic. (1845, Ford Handbk. Spain i.28, OED s.v. awful 4a. adj.)

A cognitive semantic perspective on grammaticalisation

237

Obviously, our ten lexemes are all predisposed to gradability in their propositional readings, albeit in different ways. Some of them are construed according to an either-or reading of totality (absolute, total), while others are conceptualised according to an extreme point on a scale {extreme, terrible)!" Gradability emerges as the semantic feature that all the adjectives have in common. When used descriptively, the adjectives are also configured according to gradable modes of construal. When the adjectives are employed as reinforcers, these gradability modes are made prominent and dominate their interpretation.7

3. Grammaticalisation: Semasiology and onomasiology There are two opposite approaches to the study of lexemes in historical linguistics: the semasiological approach and the onomasiological approach (Traugott 1997; Geeraerts 1997; Traugott forthcoming a and b). The semasiological approach to the study of the development of lexemes in historical linguistics has the lexeme as its starting-point. It typically asks the question: Given lexeme L, what meanings does it express? In the case of absolute, we may give the following answer: absolute 'disengaged from' > 'free from imperfection' > 'total'. A typical subject of semasiology is polysemy. The onomasiological approach starts from the content side. It typically asks the question: Given concept C, what lexical items can it be expressed with? Here the focus is on near synonyms in conceptual domains, e.g. the domain of EPISTEMIC/DEGREE with members such as total, complete, perfect, absolute being functional synonyms. A typical subject of onomasiology is synonymy.8 Having traced the semasiological path of our ten adjectives (section 2), we can state that their original application is propositional. In their development into markers of reinforcement they have undergone a weakening in their propositional meaning. This weakening, however, is compensated for by pragmatic strengthening. In Traugott's (1995) terminology this is called subjectification? Subjectification follows from a cognitive-communicative motivation on the part of the speaker to be more informative (Traugott 1995: 49). What is strengthened in absolute mess as compared to an absolute measure is the subjective stance of the speaker.

238

Carita Paradis

Another sign of grammaticalisation is that these adjectives have formed two paradigms whose members are functional synonyms. These paradigms are formed by analogy.10 TOTALITY absolute bliss a complete bitch a perfect idiot total crap utter nonsense

SCALARITY an awful mess a dreadful coward a horrible muddle a terrible bore extreme pleasure

The development of the adjectives into reinforcers represents a typical case of grammaticalisation when defined as the process whereby lexical items in constrained contexts undergo reanalysis and come to serve increasingly pragmatic functions. The well-formedness and interpretability of the new meanings arise through implication induced by contextual constraints and a search for relevance in the communicative situation.

4. The cognitive semantic approach The general theoretical framework of the present study is cognitive. I am following scholars such as Lakoff (1987), Langacker (1987a, 1988, 1991), Taylor (1992, 1995), Cruse (1995a, 1995b) and Cruse and Togia (1996). But I also incorporate ideas and views from the generative approach to lexical semantic theory (Pustejovsky 1995) and from the positivistic approach (Warren 1984, 1988). I assume that the meanings of linguistic expressions arise by the activation of conceptual patterns in the cognitive system. Lexical items map onto certain concepts in a cognitive network. In each case it is the linguistic and pragmatic contexts that evoke the relevant conceptual pattern and determine the interpretation. Meanings are perspectival in nature and polysemy emerges as a natural consequence of the human ability to think flexibly (Deane 1988: 325). Concepts are built up by domains that are of two kinds (Cruse and Togia 1996: 113-117; Paradis 1997: 48-51, 64-66). There is a content domain and a schematic domain (or mode of construal). Content domains involve propositional meanings that mirror our perception of the world,

A cognitive semantic perspective

on grammaticalisation

239

while the schematic domain imposes a specific configurative frame on the content. Gradability belongs in the schematic domain. The interpretation of reinforcing adjectives is dominated by the schematic domain and the content domain is backgrounded. In their capacity as descriptive adjectives, on the other hand, the content domain is in the foreground and their mode of construal is backgrounded.11 As was demonstrated in the previous section, reinforcing adjectives map on to two different modes of construal, one of totality, i.e. the property in question is conceived as bounded and one of scalarity, i.e. the property in question is conceived as unbounded. Langacker (1987b) makes a broad distinction between two kinds of linguistic expressions ("predications"), those that designate entities (i.e. nouns) and those that designate relations (i.e. verbs, adjectives, adverbs, and prepositions). The interpretation of polysemous adjectives is largely determined by the nouns they relate to. Nouns and adjectives combine and form more complex units. The mechanism which combines two elements and makes them well-formed and possible to interpret is valence. According to Langacker (1988: 102) "a valence relation between two predications is possible just in case these predications overlap, in the sense that some substructure within the other one is construed as identical to it". Within a compositional generative framework, Pustejovsky (1995) accounts for the well-formedness and interpretability in terms of three different modes of predication, i.e. argument structure, event structure and qualia structure. Lexical items are strongly typed and the elements involved in an expression have to conform to make well-formed expressions. This part of the model can be described as the stabiliser of the model. Yet the model is also equipped with generative mechanisms by means of which words can assume a potentially infinite number of senses in context. These mechanisms are responsible for polysemy in language. They allow for the flexibility of language. The various meanings of a polysemous adjective in context depend on the semantics of the nouns it combines with, in that the adjective is able to make available a selective interpretation of a noun through a mechanism called selective binding. The adjective selects and binds a certain quale of the noun. The qualia structure of a noun encodes information about particular properties and activities associated with them, such as their constituent parts, purpose, function and mode of creation. Quales are aspects of meaning. For instance, the noun novel has the following quales:

240

Carito Paradis

the constitutive role is NARRATIVE, the formal role is BOOK, the telic role is FOR READING and the agentive role is WRITTEN. It is to the matter of the underpinnings of well-formedness and interpretability that the rest of this paper will be devoted.

5. The problem of polysemy Adjectives are intrinsically prone to ambiguity and vagueness in that they are semantically underspecified. They require the presence of a noun for a fully-fledged interpretation. 12 Imagine being asked to give a succinct rendering of the adjective old. Out of context the answer would probably be 'something that has lasted long'. But in order to fully understand old we need to know what the combining noun is. Let us therefore consider old car, old friend and old boyfriend. They are all ambiguous, but they may have default interpretations out of context. The default interpretation of old in an old car is likely to be 'that the car has been around for a long time, whatever a long time is in the context of cars'. Similarly, the default interpretation of old in an old friend is also 'somebody who has been around for a long time as a friend', but this friend may very well be quite young. It is the duration of the friendship that is perspectivised. Finally, an old boyfriend may neither have been around for a long time nor be old. What old means in this phrase is that 'this man is no longer my boyfriend'. 13 Our first rendering 'having long duration of time' applies to old in an old car. It applies to some degree to an old friend, but it does not apply at all to an old boyfriend. A central problem for the study of polysemous adjectives is whether (i) a given lexeme is interpreted as a single inherently vague concept, or (ii) whether the various shades of meaning of a given lexeme are to be attributed to the context. In discussing the goals of lexical semantic theory, Pustejovsky (1996) describes the two views as (i) the monomorphic model, which treats ambiguity as multiple listing. Our old example would then be treated as having three different meanings. The polysemy is situated in the inherent meaning of the adjective and senses are fixed. This view has been predominant in the Montague school (Montague 1970). The inverse of the monomorphic model in Pustejovsky's terminology is (ii) the restricted polymorphic model which denies the role of fixed senses.

A cognitive semantic perspective on grammaticalisation

241

Polysemy is pragmatically determined and there are no fixed senses inherent in the language that constrain the meaning of a word in context. This is a view held by Searle (1979: 117-136). Obviously, both these accounts of polysemy are uneconomical. Apart from being unable to make generalisations about the senses in a polysemous relation, both models are descriptively and explanatorily inadequate. The main argument against the monomorphic view is that it fails to recognise the sense relations between polysemies, and the main argument against the restricted polymorphic language is that it fails to recognise fixed senses. Following Pustejovsky, I subscribe to a middle position which seems better suited to solving the polysemy problems. The middle position is (iii) the weak polymorphic model.14 Adherents of weak polymorphism treat polysemy as in part lexically determined and in part structurally and pragmatically determined. Consequently, I argue that the ten adjectives have inherently determined lexical properties which relate the various polysemous readings (section 2). Secondly, the semantic properties of the modified nouns and the way they relate to the adjectives are important for the analysis (sections 6 and 7). Finally, there ought to be structural constraints on their interpretation (section 8).

6. Reinforcing adjectives and their nouns In the introduction I stated that the properties of the noun constrain the interpretation of the adjective, and the properties of the adjective assign a perspective from which the noun is to be viewed. In the light of this statement, it seems reasonable to assume that nouns which take reinforcers are degree nouns. The reasoning behind this is that the combining elements have to harmonise for a successful match (Paradis 1997: 158-165). Pretheoretically, the assumption seems to be correct as far as the occurrences in British National Corpus (BNC) are concerned. Nouns which combine with clear-cut reinforcing adjectives are nouns such as bargain, bastard, crap, contempt, darkness, despair, disaster, disgrace, failure, fool, heat, horror, idiot, mess, nonsense, poverty, purity, rubbish, shame, shit, wanker. An important observation is that nouns that are capable of taking reinforcers correspond to gradable property concepts. They are, in other words, rather more adjective-like than typical nouns.15 The relationship

242

Carito

Paradis

between a reinforcing adjective and its head is much like the relationship between a degree modifier and its adjectival head {absolutely pure, awfully messy, totally contemptuous, extremely dark, completely disastrous, utterly disgraceful and terribly boring). This claim provokes the question: What is gradability in nouns and how can it be tested? Firstly, gradability has traditionally been associated with adjectives, and the concept of degree has been dealt with in terms of whether an adjective can undergo comparison or not. This view was challenged by Sapir (1949), Bolinger (1967, 1972) and Gnutzmann (1975), who claimed that gradability is not only a feature of adjectives but also of nouns and verbs. What gradable words have in common is a feature which we perceive as variable in intensity and which therefore can be reinforced. Some criteria for a valid characterisation of gradability in nouns are thus needed. One such criterion is the possibility for certain words to be the focus of exclamatory utterances. For nouns then, gradability is drawn out in exclamatory "Such x!" or "What x!" utterances, e.g. What a bastard!, Such rubbish!, What a bargain!, Such poverty!. Bolinger (1972: 60) points out that we make a difference between identifying such and intensifying such. He gives the following examples to illustrate this (emphasis and additions are mine): (8)

Such a person always frightens me.

[Identifying such]

(9)

Such a blunder always frightens me.

[Intensifying such]

(10)

Such behaviour always frightens me.

[Identifying such]

(Π)

Such misbehaviour always frightens me.

[Intensifying such]

In (8) and (10) such refers to the identification of the nouns (jc identity) and the nouns are not inherently gradable. In (9) and (11) such refers to the intensity of the gradable feature of the noun (jt intensity) and the nouns are degree nouns.16 By the same token degree nouns are capable of invoking reinforcing readings in gradable adjectives such as absolute, complete, utter and terrible {an absolute blunder and terrible misbehaviour). The critical reader may suggest expressions such as Such a person! and Such behaviour! as counter-evidence to the above criterion for degree

A cognitive semantic perspective on grammaticalisation

243

nouns. It is true that person and behaviour can serve intensifying functions when they are in the focus of exclamatory such. There is a difference, however, between person and behaviour on the one hand, and blunder and misbehaviour on the other, in that a gradable adjective has to be inferred and the polarity of that adjective has to be contextually induced (Such a [wonderful/boring] person! and Such [good/bad] behaviour!). The same phenomenon is of course true of adjectives + non-degree nouns (a terrible person, referring to 'a person who does things that I find terrible'), where terrible is evaluative and descriptive, and adjectives + degree nouns (a terrible fool) where terrible is purely intensifying. The gradable element is inherent in fool)1 This line of thought will be developed further in section 7. Another observation is that degree nouns tend to be hyperbolic. They come with an inherent evaluation of either a positive or a negative character {absolute bliss, extreme pleasure, complete mess, terrible bastard). Also, reinforcers differ in their preferences regarding the polarity of the degree nouns they combine with. Awful, dreadful, horrible and terrible are restricted to negative nouns. An awful muddle is natural, while it would be unnatural to talk about Tan awful pleasure}* In the vast majority of cases, the other reinforcers also occur with negative nouns in the examples from BNC, but they are not restricted to negativity. For instance, absolute bargain, complete and utter joy, extreme bravery all match perfectly well.19 Neutral nouns are either not gradable or gradable only by the addition of an implied evaluative adjective, and hence they are not true degree nouns. Moreover, degree nouns are non-typical nouns in that they do not conceptualise entities or phenomena. Rather they are property concepts and as such adjective-like. It is important to note that the main function of degree nouns is not to act as identifying elements, but to describe entities and phenomena. Degree nouns are thus relational and underspecified just like descriptive adjectives. Jespersen (1968: 75) puts this very neatly: on the whole, substantives are more special than adjectives, in the parlance of logicians, the extension of a substantive is less, and its intension is greater than that of an adjective. The adjective indicates and singles out one quality, one distinguishing mark, but each substantive suggests, to whoever understands it, many distinguishing features by which he recognises the person or the thing in question.

244

Carito Paradis

The distinction between nouns and adjectives has to do with how the properties are conceptualised. Nouns designate complex entities or 'kinds of things', while adjectives designate properties. As Jespersen points out, nouns tend to involve a large number of properties, whereas adjectives normally designate one property. The prerequisite for well-formedness in phrases containing a reinforcing adjective is that there has to be some kind of overlap first and foremost in the schematic domain, i.e. in terms of gradability. This allows us to conclude that nouns that take reinforcers are degree nouns.

7.

Valence

Apart from the lexical content that is inherent both in the adjective and the noun, well-formedness is a result of the way adjectives attach to nouns. Warren (1984: 21-28) discovered that there is a system of relations that underlies the interpretations of classifying and descriptive adjectives. This system is based on a restricted set of covert relators that trigger different types of attachments between adjectives and their nouns. Warren's relator is on a par with Pustejovsky's generative mechanism of selective binding, whose function is to bind the appropriate quale of the noun. Warren states that in order to uncover the nature of the adjective, we must be familiar with both the denotation of the adjective and with its relation to the noun it qualifies. Her semantic analysis involves two components, referential content and relator. Consider figure 1:

a sad girl

referential content 'sadness'

relator 'experiencing'

Figure 1. Warren's (1984) model of the semantic structure of an adjective

A cognitive semantic perspective on grammaticalisation

245

The meaning of sad in a sad girl, is analysed in two components and glossed 'x experiences sadness'. There is thus an opaque relation, which the language user at some level has to be aware of in order to fully interpret the phrase. 20 Warren's model is thus weakly polymorphic in that it is partly lexically determined through the overt content part and is contextually defined by means of a set of connecting relations providing for the compositional interpretation of adjectives and nouns. Warren's relators selectively bind the appropriate quale of the noun. Consider the polysemy of nervous in a nervous breakdown and a nervous man in figure 2. a nervous breakdown

'nerves/nervousness'

'caused by'

a nervous man

'nervousness'

'experiencing'

Figure 2. The semantic elements of nervous A nervous breakdown can be glossed 'x is caused by nerves', and a nervous man 'x experiences nervousness'. Nervous in nervous breakdown is one type of breakdown. The quale that nervous binds in breakdown has to do with mental phenomena. Nervous in a nervous man binds the quale of man as an experiencer. Polysemy in an adjective can be said to be due to differences in at least one of the two main semantic parts of adjectives, i.e. in the referential content and/or in the relator. I shall now apply Warren's model to three of the adjectives under discussion, both in their propositional capacity and as reinforcers:

PROPOSITIONAL READING a complete sentence a perfect body a terrible nightmare

REINFORCING READING complete nonsense a perfect idiot a terrible bore

246

Carito Paradis

As has already been pointed out, Warrren's study was aimed at propositional readings of adjectives in the first place. By applying her model on grammaticalised readings of adjectives we may be able to display what the systematic semantic differences between these two onomasiological domains are. Figures 3, 4 and 5 show the decomposition of the overt and covert parts of the three lexemes: a complete sentence

'completeness'

'constituting'

complete nonsense

TOTALITY

?

Figure 3. The semantic elements of complete

The meaning of a complete sentence is decomposed into its two parts and can be glossed 'x constitutes completeness'. There is the overt referential content of the adjective, 'completeness', a concept which maps on to a content proper domain and a schematic domain. Complete in a complete sentence is dominated by the content proper domain. It is the content proper that makes the basis for the valence relation. The relator is 'constituting'. The quale bound by the adjective is the propensity for a sentence to be complete or not. The foregrounded content part of 'completeness' maps on to the formal and constitutive quales of sentence as something consisting of parts and rules for their arrangement. This is the common ground where the two predications meet in a well-formed valence relation. The meaning of complete in complete nonsense is not decomposable in the way complete in a complete sentence is. First of all, there is loss of content proper in the referential content part of its semantic structure. The application of complete is abstracted away from its foundation in the content domain and has assumed a function as a marker of degree and epistemic modality. The semantic loss is compensated for by pragmatic enrichment. What has happened in the referential content part is that the schematic domain is foregrounded, while the content proper is hovering in the background. The propositional loss is the reason for the inability of

A cognitive semantic perspective on grammaticalisation

247

complete to bind a particular quale in the noun. What complete can identify and bind is gradability. Furthermore, complete nonsense can be glossed 'x that is characterised by being nonsensical to the degree of totality'. This glossing makes it clear that a referent is described by nonsense. In the paraphrase of a complete sentence, i.e. 'x constitutes completeness', the referent is specified by the modified noun. Now consider perfect in figure 4. a perfect body

'perfection'

a perfect idiot

'constituting'

TOTALITY

?

Figure 4. The semantic elements of perfect

The same pattern is repeated for perfect. A perfect body can be decomposed into a referential content part 'perfection' and a relator 'constituting', and the phrase can be glossed 'x constituting perfection'. In a perfect idiot the content component is weakened and perfect is employed at an abstract level of reinforcement, and consequently there is a breakdown in the relator. The only link that can be picked up by perfect and made compatible with idiot in an overlapping substructure is the gradability of the property of'idiocy'.

a terrible nightmare

'something very bad'

'constituting'

Figure 5. The semantic elements of terrible

a terrible bore

HIGH DEGREE

9

248

Carita

Paradis

In a terrible nightmare the content component is weakened from 'terror' to 'something very bad', and the relation is one of 'constituting'. The phrase can be glossed as 'x constitutes something very bad'. 21 The decomposition of a terrible bore is analogous with complete nonsense and perfect idiot in that the content component is generalised into HIGH DEGREE. Not only has terrible become more abstract, but related to this is the loss of the conceptualisation of terrible as schematically representing the extreme boundary on a scale in the schematic domain.22 Again there is no propositional link to the noun bore. The interpretability and well-formedness lie in the grading potential in both of them and that is where the language user can find relevant overlap and identity of substructures. This goes to show that semantic bleaching, as in α terrible nightmare, is not enough for a breakdown in the relator part, since it is still dominated by the content proper domain and the propositional function, and not by the mode of construal and the expressive function. Finally, from the point of view of grammaticalisation, the reinforcing readings have undergone a process of semantic bleaching and abstraction. This process is not unique to reinforcers. It is a well-established fact that a lexical element can go through a process of semantic bleaching without being grammaticalised, which is shown in a terrible nightmare. This analysis explains why bleaching is a necessary but not sufficient condition for an item to undergo grammaticalisation. Thus, grammaticalisation in semantic terms means a shift in the content component that is not only a weakening of the content, but also an increase in generality coupled with pragmatic enrichment. Secondly, the covert relators in propositional adjective + noun combinations are no longer to be found. The failure to establish a relator between the reinforcer and its noun is a natural consequence of the fact that their valence relation is not to be found in the propositional element. The reinforcers make the mode of construal prominent and the reading is automatically linked up with scopal properties and speaker attitude. Driven by a unidirectional urge towards subjectification on the part of the speaker and relevance on the part of the hearer, our reinforcers have moved into another onomasiological domain. They have moved along the scale from semantic to more pragmatic meanings, i.e. from the PROPOSITIONAL X domain to the EPISTEMIC/DEGREE domain.

A cognitive semantic perspective on grammaticalisation

249

8. Structural constraints The interpretation of adjectives and their nouns cannot be separated from the structure they are embedded in, a view held by both cognitivists and generativists. One restriction on the use of these reinforcers is that they occur in premodifier position only. The validity of this statement can be demonstrated by entailment, where the semantic interaction between the adjective and the noun and the positional constraints are made manifest at the same time. Consider first the set of propositional readings of complete, perfect and terrible, in example (12): (12)

It is a complete sentence It is a sentence which is complete It is a perfect body MANNER > DEGREE (Peters 1993:282; Traugott 1995:44). 9. Langacker (1990) also uses the term "subjectivity". Both Langacker and Traugott view subjectivity as a ubiquitous phenomenon, but they use the term somewhat

254

Carito Paradis

differently. Langacker focuses primarily on subjectivity as degrees of grounding in the situation construed by speakers. I am only going to use subjectification in the sense proposed by Traugott (1999: 187-188), i.e. the diachronic shift from the physical world to the mental world. 10. As pointed out in section 2 (note 5), all the adjectives were originally associated with a bounded conceptualisation as their mode of gradability, either in terms of 'completeness' or 'an extreme point'. The 'completeness' adjectives, absolute, complete, perfect and total represent one type of analogy formation. Extreme and utter are extreme point adjectives and represent another type of analogy. Extreme has lost its extreme application and means 'very high degree' 0Oxford English Dictionary, s.v.; cf. extremely; Paradis 1997: 83). Finally, awful, dreadful, horrible and terrible as reinforcers come from the same sort of strongly negative, extreme-point adjectives. All the members of the group of scaling adjectives have lost their inherent superlativity and assumed an upward scaling function. It should be pointed out that not only have the scalar reinforcers gone through a process of subjectification, but they have also become unbounded in their gradability. That is, they have lost their association with the extreme point of the scale to high degree. 11. This accounts for the traditional classification of words into function words and content words. 12. Polysemy is a phenomenon which is closely linked to ambiguity and vagueness. For a discussion of these terms, see Warren (1988, forthcoming), Geeraerts (1993), Tuggy (1993). 13. Notice that if you say my old car instead of an old car the interpretation can be, but is not necessarily, similar to the interpretation of an old boyfriend, i.e. 'no longer my car'. Contextual effects of this kind will be discussed in section 8. See Taylor (1992) for a semantic analysis of old, and Healey (1997) for the historical development of old. 14. This is the view held by most semanticists, e.g. Warren (1984, 1988), Taylor (1992), Cruse (1995a, 1995b), Kamp and Partee (1995), Jackendoff (1997). 15. See Dixon (1982), Hopper and Thompson (1984), Wierzbicka (1986), Langacker (1987b), Thompson (1988), Wetzer (1996) for discussions on "adjectivehood" and "nounhood". 16. This criterion is comparable to the "how x!" criterion for gradable adjectives (How excellent!, How sad!) (Paradis 1997: 50). Another criterion for gradable adjectives is "how χ is it?" (How good is it?, How old is it?). Applied to gradable nouns this criterion would be: "how much of a(n) χ is it?" (How much of a blunder is it?, but not *How much of a person is he?). There is also an element that can undergo comparison in degree nouns (This was much more of a blunder than that, but not * He is much more of a person than you are). 17. It deserves to be pointed out that degree nouns can be identifying as well as intensifying (cf. section 8). Identifying nouns, however, cannot assume an intensifying reading without the addition of an implicit gradable evaluative adjective.

A cognitive semantic perspective on grammaticalisation

255

18. This is not the case for degree adverbs in combination with adjectives, which is probably a sign of their being even more grammaticalised. The related adverbs are not restricted to negative contexts, e.g. terribly nice and awfully good (Paradis 1997: 76-91). 19. There are 25 tokens of utter in my material, all of which are reinforcers. Interestingly, 19 of these 25 tokens are coordinations with other adjectives. There are 13 occurrences of complete and utter and one occurrence of the following combinations: utter and total, utter and unbelievable, utter and absolute, total and utter, sheer and utter, absolute and utter. 20. Warren's model of the meaning of adjectives involves a paraphrase of the noun phrase in question, whereby two semantic elements are revealed, one overt referential ('sadness') and one covert relational ('experiencing'). She finds that there is a limited number of recurring covert relations, all of which may occur with classifying adjectives, but a more limited number of them with descriptive adjectives. In cognitive semantic terms, Warren's referential content involves mappings on to both the content proper domain and the gradability domain discussed in section 4. The relator is the exponent of the valence relation, i.e. the type of match between the adjective and the noun. 21. There is not always an established noun representing the referential content of an adjective. Another interpretation, which is possible but favoured by fewer of my informants, is that the content component actually is 'terror' and the relator is 'causing'. The gloss would then be 'x causes terror'. The fact that there are these two readings is however irrelevant to my argument. 22. This path of development is not uncommon in shifts in the category of degree words (Paradis forthcoming).

References Aston, Guy and Lou Burnard 1998 The BNC Handbook. Edinburgh: Edinburgh University Press. Bolinger, Dwight 1967 Adjectives in English: Attribution and predication. Lingua 18: 1-34. 1972 Degree Words. The Hague: Mouton. Blakemore, Diane 1992 Understanding Utterances. Oxford: Blackwell. Crowdy, Steve 1995 The BNC spoken corpus. In: Geoffrey Leech, Greg Myers and Jenny Thomas (eds.), Spoken English on Computer, 224-242. Harlow: Longman.

256

Carita Paradis

Cruse, Alan 1995a

Polysemy and related phenomena from a cognitive linguistic viewpoint. In: P. St. Dizier and E. Viegas (eds.), Computational Lexical Semantics, 33-49. Cambridge: Cambridge University Press. Between polysemy and monosemy. In: H. Kardela and G. Persson 1995b (eds.), New Trends in Semantics and Lexicography, 25-34. Umeä: Swedish Science Press. Cruse, Alan and Pagona Togia 1996 Towards a cognitive model of antonymy. Journal of Lexicology 1: 113-141. Deane, Paul 1988 Polysemy and cognition. Lingua 75. 325-361. Dixon, Robert M. W. 1982 Where Have all the Adjectives Gone? Berlin/New York/Amsterdam: Mouton. Geeraerts, Dirk 1993 Vagueness's puzzles, polysemy's vagaries. Cognitive Linguistics 4: 223-272. 1997 Diachronic Prototype Semantics. Oxford: Clarendon Press. Gnutzmann, Claus 1975 Some aspects of grading. English Studies 56: 421 -433. Healey, Antonette Di Paolo 1997 Words, story, history: The mapping of meaning and Toronto's Dictionary of Old English In: Terttu Nevalainen and Leena KahlasTarkka (eds.), To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, 35-49. (Mömoires de la Soci6t6 N6ophilologique de Helsinki 52.) Helsinki: Soci6t6 Neophilologique. Hopper, Paul and Sandra Thompson 1984 The discourse basis for lexical categories in universal grammar. Language 60: 703-752. Jackendoff, Ray 1997 The Architecture of the Language Faculty. Cambridge, Massachusetts/London, England: MIT Press. Jespersen, Otto 1968 The Philosophy of Grammar. London: Allen and Unwin. Kamp, Hans and Barbara Partee 1995 Prototype theory and compositionality. Cognition 57: 129-191. Lakoff, George 1987 Women, Fire, and Dangerous Things. Chicago: The University of Chicago Press.

A cognitive semantic perspective on grammaticalisation

257

Langacker, Ronald 1987a Foundations of Cognitive Grammar. Stanford: Stanford University Press. 1987b Nouns and verbs. Language 63: 53-94. 1988 The nature of grammatical valence. In: Brygida Rudzka-Ostyn (ed.), Topics in Cognitive Linguistics, 91-125. Amsterdam/Philadelphia: John Benjamins. 1990 Subjectification. Cognitive Linguistics 1: 2-38. 1991 Concept, Image and Symbol. Berlin/New York: Mouton de Gruyter. Montague, Richard 1970 Universal grammar. Theoria 36: 373-398. Nicolle, Steve 1998 A relevance theory perspective on grammaticalisation. Cognitive Lingusitics 9: 1-35. Oxford English Dictionary 1992 Second edition, CD-ROM. London: Oxford University Press. Paradis, Carita 1997 Degree Modifiers of Adjectives in Spoken British English. Lund: Lund University Press. forthcoming It's well weird. Degree modifiers of adjectives revisited: The nineties. In: John Kirk (ed.), Corpora Galore: Analyses and Techniques in Describing English. Amsterdam/Atlanta: Rodopi. Peters, Hans 1993 Die englischen Gradadverbien der Kategorie Booster. Tübingen: Gunter Narr Verlag. Pustejovsky, James 1995 The Generative Lexicon. Cambridge, Massachusetts/London, England: MIT Press. Sapir, Edward 1949 Grading: Α study in semantics. Selected writings of Edward Sapir. In: D. G. Mandelbaum (ed.), Language, Culture and Personality, 122-149. Berkeley: University of California. Searle, John 1979 Expression and Meaning. Cambridge: Cambridge University Press. Sperber, Dan and Deirdre Wilson 1995 Relevance: Communication and Cognition, Second edition. Oxford: Blackwell. First edition [1986], Taylor, John 1992 Old problems: Adjectives in cognitive grammar. Cognitive Linguistics 3: 1-35. 1995 Linguistic Categorization. Oxford: Clarendon Press.

258

Carito Paradis

Thompson, Sandra 1988 A discourse approach to the cross-linguistic category 'adjective'. In: J. A. Hawkins (ed.), Explaining language universal, 167-185. London: Blackwell. Traugott, Elizabeth Closs 1995 Subjectification in grammaticalisation. In: Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectivisation, 31-54. Cambridge: Cambridge University Press. 1997 Semantic change: an overview. Glot International 2: 3-6. 1998 Historical pragmatics and the development of scalar focus particles. Paper given at the Tenth International Conference on English Historical Linguistics, University of Manchester, 21st-26th August 1998. 1999 The rhetoric of counter-expectation in semantic change: A study in subjectification. In: Peter Koch and Andreas Blank (eds.), Historical Semantics and Cognition, 177-196. Berlin/New York: Mouton de Gruyter. forthc. a. Lexicalization and grammaticalization. In: Alan Cruse, Franz Hundsnurscher, Dieter M. Job and P.R. Lutzeier (eds.), Lexikologie — Lexicology. Berlin/New York: Mouton de Gruyter. forthc. b. The role of pragmatics in semantic change. In: The Proceedings of the Pragmatics Conference in Reims 1998. Tuggy, David 1993 Ambiguity, polysemy, and vagueness. Cognitive Linguistics 4 (3): 273-290. Ungerer, Friedrich 1988 Syntax der Englischen Adverbialen. (Linguistische Arbeiten 215.) Tübingen: Max Niemeyer Verlag. Warren, Beatrice 1984 Classifying Adjectives. (Gothenburg Studies in English 56.) University of Gothenburg, Sweden: Department of English. 1988 Ambiguity and vagueness in adjectives. Studia Linguistica 42: 122-172. forthcoming The role of links and/or qualia in head-modifier construction. In: D. Clarke, V. Herman, B. Nerlich and Z. Todd (eds.), Polysemy: Patterns of Meaning in Mind and Language. Amsterdam/Philadelphia: John Benjamins. Wetzer, Harrie 1996 The Typology of Adjectival Predication. Berlin/New York: Mouton de Gruyter. Wierzbicka, Anna 1986 What's in a noun? (Or: How do nouns differ in meaning from adjectives?). Studies in language 10: 353-389.

2. Text types

Variation and change: Text types and the modelling of syntactic change1 Wim van der Wurff

1. Introduction This paper will address a general issue in the study of historical syntax. I will take two approaches to historical syntax that at first sight seem entirely different in methods adopted and results aimed at, and then present some reflections on the possibilities of integrating them. Such an attempt at integration, even if only partial, may help us understand better the causes and pathways of syntactic change, which is the ultimate goal of the remarks that follow. To begin with, I shall describe what the two approaches are, and in what ways they differ from each other, using two case studies, one from my own current work and one from a recent article using the Helsinki and ARCHER corpora, as concrete examples. After that, I shall put forth some general considerations tending to suggest that it may be possible and profitable to combine insights from the two approaches, and I will try to show what direction this sort of enquiry could take in the two concrete examples. Although the suggestions in this paper are at some points programmatic and at no point exhaustive, they may be helpful in stimulating others to apply the methods described in more detailed data-based studies or to contemplate further possible connections between these two types of approach to historical syntax.

2. Two approaches to historical syntax A first very common approach to historical syntax consists in the study of the structural factors conditioning syntactic losses and innovations. These structural factors can take various forms, such as semantic generalisation, phonetic weakening or reduction, symmetry of patterning, analogy, phonetic

262

Wim van der Wurff

splits and mergers, or all kinds of consequences of prior syntactic changes. In all of these cases, the investigator is typically focused on trying to figure out what there is in the structure or system of the language at a certain point in time that made a new construction possible, or even natural, or that allowed an existing construction to become more general, or — the opposite development — that brought about the decline of an existing construction. This is the first approach to historical syntax that I would like to distinguish; I will label it the "structural" approach. A concrete example will follow in a moment. Another common approach in historical syntax may be called the "variationist" approach. It is concerned with documenting the distribution and patterning of forms and structures in terms of text types, styles, regions, and social and other characteristics of speakers/writers, and determining whether this distribution and patterning changes or stays stable over time. The leading idea behind this approach is the "orderly heterogeneity" of Weinreich, Labov and Herzog (1968), and historical syntactic studies of this type can in fact be viewed as examples of sociolinguistic inquiry, which are somewhat special in that many of them necessarily rely on written sources only. Since information about writers of historical texts is often spotty and scanty, comparisons along social and other speaker-related dimensions can be difficult to make. Text-related characteristics are usually somewhat easier to establish, and there are many variationist studies in which these receive a great deal of attention. In what follows, I will indeed also focus on the study of the syntactic characteristics of text types, to the virtual exclusion of other dimensions of syntactic variation, in the hope that the results of this rather restricted (perhaps even reductive) inquiry will be detailed and precise enough to make meaningful predictions about other domains of variation as well.

3.

The structural approach: an example

Let us now turn to a concrete example of the structural approach, which can serve to illustrate the fundamental assumptions and methods of this way of doing historical syntax. The example that follows concerns several syntactic properties of late Middle English. I will try to show how it can be argued that there is a structural connection between these properties (in

Text types and the modelling of syntactic change

263

doing so, I will draw on a recent article by Bobaljik and Thräinsson 1998, who have suggested the existence of such a connection in the modern Germanic languages; see also Thräinsson 1996). The example, I think, illustrates well how historical linguists usually go about the search for structural factors determining syntactic change. In English texts written around 1400, we find five phenomena that are absent from present-day English. Examples are given in (l)-(5), all taken from prose writings. (1)

And ther seyde oones a clerk in two vers, "What is bettre than gold?..." 'And once a clerk said in two verses, "What is better than gold?'" (Chaucer, Melibee 1437)

(2)

a. We wol that ye see that theer be taaken dewe accomptes 'We want you to see to it that there is taken due account' (1418 Proc.PrivyC. 2.240) b. per was a newe kyng chosen 'There was chosen a new king' (al500 [?al450] GRom [Glo] 727/16)

(3)

a. yf it so be pat pyn answere scholde turne eny oper persone to schame or desclaundre 'if it should be the case that your answer might bring shame or disrepute to someone else' (Five Wyttes 22.34; ca. 1400) b. pou scalt noujt pe persone discouere ne accuse 'you must not betray or accuse that person' (Five Wyttes 23.3; ca. 1400)

(4)

Syn thou feelist thus thise thinges 'Since you feel thus about these things' (Chaucer, Bo 3.pr. 12.48)

264

(5)

Wim van der Wurff

The riche folk, that embraceden and oneden al hire herte to tresor of this world 'the rich people, who embraced the treasure of this world and joined their hearts to it' (Chaucer, CT.PT. 192)

The sentence in (1) illustrates what, in comparative Germanic syntax, is known as the transitive expletive construction: there is an active transitive verb (seyde) with an expletive subject (ther). This is no longer possible in Modern English, which only allows expletive subjects with passives and intransitives (as in There followed a long list of grievances). The existence of the construction in (1) in late Middle English has not — as far as I am aware — been noted in the literature, even though a great deal of work has been done on the construction in the modern Germanic languages (see, for example, Jonas 1996). The examples in (2) illustrate word order variability in other expletive constructions: (2a) has the logical subject, dewe accomptes, in a position following the lexical verb, while the logical subject in (2b), a newe kyng, immediately follows the auxiliary. In Modern English, only the first of these two options remains. The sentences in (3) show variability in the position of the object: (3a) has a postverbal object, but (3b) a preverbal one. Again, Modern English only allows the former possibility. Sentence (4) has a postverbal object (thise thinges) separated from the verb (feelist) by an adverbial (thus), which is only possible in Modern English when the object is long or heavy. The example in (5), finally, has two verbs (embraceden, oneden) with separate inflectional marking for tense (-ed-) and number (-en). In Modern English, of course, there is no separate marker for number anymore. At first sight, the set of examples in (l)-(5) may look like a random collection of facts, but there has been some work attempting to establish connections between several of these late Middle English patterns. Making use of ideas about verb positions developed on the basis of cross-linguistic data, Roberts (1993: 246-273) has argued that there is a link between the facts illustrated in (4) and (5), and van der Wurff (1997) has suggested that (3) may also be part of the underlying pattern. Here, I would like to suggest that (l)-(5) are all part of the same pattern. In a recent contribution to the area of comparative Germanic syntax, Bobaljik and Thräinsson (1998) propose that the Germanic languages can

Text types and the modelling of syntactic change

265

be divided into two groups, on the basis of their setting of the Split-IP Parameter. The effects of the parameter setting can be seen in the following five properties (Bobaljik and Thrainsson 1998: 67). (I) (II) (III) (IV) (V)

the possibility of transitive expletive constructions (i.e. there + transitive verb phrase) the availability of two subject positions (as shown in the position of the logical subject in intransitive expletive constructions) the availability of two object positions (i.e. preverbal and postverbal position) raising of the verb out of the verb phrase (yielding Vj + X + [ V P t; OBJ], where V ends up to the left of an adverb X) the possibility of multiple inflectional morphemes on the verb stem, specifically the co-occurrence of discrete tense and agreement morphemes.

Bobaljik and Thrainsson (1998: 67) note that there has been a great deal of previous research on these five properties and their clustering in the Germanic languages, and that there have been proposals to subsume some of them under a single parameter; however, as they say, "we believe ours to be the first proposal that attempts to unify ... all five". Their specific proposal is that languages can differ in having either a clausal structure as in figure 1, in which there is just one functional projection, the IP, dominating the verb phrase, or a structure as in figure 2, in which the IP is split, and there are an object agreement phrase AgrOP, tense phrase TP, and subject agreement phrase AgrSP, and possibly others as well, dominating the verb phrase.

Spec

Γ I

VP Spec

SUBJ Figure 1. Non-split-IP

V' V

OBJ

266

Wim van der Wurff AgrSP AgrSP'

Spec

.TP

AgrS Spec

Τ' Τ

AgrOP Spec

AgrO' AgrO

VP Spec

SUBJ

V V

OBJ

Figure 2. Split-IP The tree structures in figures 1 and 2 perhaps look complicated, but the essential point for our purposes is simple. In both language types, we can think of the subject and the object as starting out inside their natural home and habitat, the verb phrase, which of course has the verb V as its head. At some point in the derivation, subject (SUBJ) and object (OBJ) may leave the verb phrase, to move up in the tree structure of the clause. Since languages of the type in figure 1 have relatively little structure in the clause, they may be expected to have relatively few possibilities in terms of word order and constructions with additional formatives, such as expletive elements like dummy there (i.e. they lack the properties listed under [I][IV]). In addition, there is no separate structural position for tense and agreement in figure 1, so it is not surprising to find that these languages do not have layering of discrete tense and agreement morphemes (they also lack property [V]). By contrast, language of type in figure 2 have several structural positions in the clause that SUBJ and OBJ can move into, entailing the existence of more constructional possibilities and also the occurrence of separate tense and agreement morphemes on the verb, since there are separate structural positions for such formatives, in the form of

Text types and the modelling of syntactic change

267

the tense phrase TP and the subject agreement phrase AgrSP. This, in a nutshell (and leaving many technicalities aside), is how the account of Bobaljik and Thrdinsson (1998) works. What would be the implications of all this for the history of English? In English texts from around 1400, all five properties listed in (I)-(V) are attested, as we saw in examples (1) to (5). By 1500, however, they have more or less disappeared. If we accept that there is a Split-IP Parameter, the reason for the simultaneity of the decline of these five superficially quite different phenomena can be said to lie in the fact that there is actually only one structural difference involved, i.e. the difference between figures 1 and 2, which reflects the setting of the Split-IP Parameter. As for the reason why this set of properties started to decline in the first place, this could plausibly be attributed to the reduction of inflectional marking on verbs in late Middle English. As data given by Lass (1992: 97-98) make clear, the verbal plural marker -en disappeared rapidly in the course of the fifteenth century (for detailed discussion, see also Roberts 1993: 246-273). Once there were no longer any forms marked for past tense and plural, i.e. when the forms in (5) had declined, there would no longer be separate structural positions for tense and agreement as in figure 2; instead, the clause came to look like figure 1. According to this analysis, the shapes and meanings of verbal forms would therefore serve to determine the basic architecture of the clause, and in the course of the fifteenth century that would become the option in figure 1. When this happened, the rather minimal structure in figure 1 would entail that the possibilities for accommodating expletives and logical subjects, extra object positions, and extra verb positions would also be rather minimal — specifically, none of the sentence types in (1), (2b), (3b) and (4) would be possible any more. This, then, is an example of what I am calling the structural approach to syntactic change. The main assumption underlying this approach is that specific linguistic features and constructions hang together structurally, and that a change in one form or construction may have consequences for — in principle — many other forms and constructions. The types of structure assumed in studies of this sort of course differ with the general theory embraced by the researcher. The example above is based on recent generative ideas about language structure, but similar examples exist in the dozens or even hundreds which are based on other theories of syntax but also try to identify structural factors pushing and pulling one or more

268

Wim van der

Wurff

syntactic phenomena this way or the other. Just like the example given above, these studies usually remain silent about the question of what kinds of concrete situations, utterances or texts these abstract structural factors are supposed to have done their pushing and pulling in. This must certainly be considered a serious weakness in work of this type. However, I shall argue that it may be overcome by using insights from the variationist approach, to which we now turn.

4. The variationist approach: an example A good example of the variationist approach to historical syntax can be found in Kytö (1997), a detailed corpus-based study of the diachrony of the use of BE and HAVE as auxiliaries of the perfect. As is well known, earlier English showed variation in the choice of BE and HAVE with intransitive verbs in the perfect, as illustrated for the verb COME in (6) and (7). (6)

My godfathyr has be syke byt he ja? whell mendyd, thankyd be God. [Thys same] day my Loord ys comyn to London to aske the Kyng leue to go to the Rodys for he ys sent for. 'My godfather was ill, but he has recovered, thank God. This same day my Lord has come to London to ask the King permission to go to Rhodes, because he has been sent for'. (Richard Cely, The Cely Letters 107 [=Kytö 1997, (3a)])

(7)

... and when mattens vhos done thay whente to a kynnyswhoman off the jewnge gentyllwhomane; and I sent to them a pottell of whyte romnay, and thay toke hyt thankefully, for thay had cwm a myle a fote that mornyng 'and when matins was done, they went to a kinswoman of the young lady; and I sent them a pottle of rumney, and they accepted it gratefully because they had come a distance of a mile on foot that morning'. (Richard Cely, The Cely Letters 152 [=Kytö 1997, (3b)])

Eventually, HAVE wins out in all except a few expressions, such as they are gone or the sun is set, and Kytö (1997) traces the course of this development

Text types and the modelling of syntactic change

269

from the fourteenth century up to the present. The texts that she examined come from the Helsinki Corpus (for the period 1350 to 1710), the Century of Prose Corpus (1680-1780), and the ARCHER Corpus (1650-1990); together they contain nearly three million words. Among the linguistic factors considered by Kytö are the distinction between stative verbs (e.g. STAND, LIVE) and mutative verbs (e.g. COME, GROW), with stative verbs favouring the choice of HAVE from the earliest period onwards; the distinction between action verbs (such as ARRIVE, RETURN) and process verbs (such as GROW, BECOME, WAX), with action verbs shifting over to HAVE somewhat earlier than process verbs; the influence of the nature of the verbal cluster (the past perfect, for example, has higher proportions of HAVE than the present perfect, and the perfect infinitive has even higher proportions); the presence or absence of an object-like element following the main verb (cases with such an element, as in (7), tend to favour the choice of HAVE); the effect of durative, iterative, conditional and optative contexts (all of these promote the use of HAVE); and the frequency of the individual main verbs (for example, the highly frequent verbs COME and GO continue to favour the auxiliary BE longer than other verbs). These and similar findings reported in Kytö (1997) raise several questions about structure and causation, but for present purposes I would like to focus on her results concerning the influence of text type or register. Here, the question is: Is the variation between BE and HAVE and its development different in different text types? Although the corpora examined by Kytö do not have exactly the same composition, there are six text types that she is able to follow across each of the corpora: Fiction, Science, Sermons, Drama, Diaries/Journals, and (private) Letters. In table 1, I summarise the figures for these six text types given in Kytö (1997: 44), which are based on the eighteenth- and nineteenth-century texts in the ARCHER Corpus. Table 1. Percentage of HAVE in six Corpus (Kyto 1997: 44) Fiction Journals 1700-1750 38% 41% 80% 1750-1800 58% 1800-1850 89% 71% 88% 1850-1900 85%

text types from four subperiods of Letters 45% 63% 38% 88%

Drama 18% 33% 63% 85%

Science 53% 33% 91% 91%

ARCHER

Sermon 50% 88%

270

Wim van der Wurff

The figures show that the use of HAVE was predominant by the second half of the nineteenth century in all text types, but that its rise to prominence occurred a century earlier in the category of Journals and, to some extent, in Letters than in other texts. Kytö's data show some other differences between text types (for example, the category of Science outstrips the other categories in the use of HAVE in the nineteenth century) but these seem relatively minor compared with the early ascendance of HAVE in Letters and Journals. Since these two text types are usually taken to closely approximate ordinary speech in at least some of their linguistic characteristics, it might be thought that the early rise of HAVE in these categories is a reflection of its rise in speech, or in speech-related genres in general. Kytö therefore tested the data for possible correlations between the use of HAVE versus BE and level of formality and closeness to spoken language, by combining the figures for the relevant text types. However, the evidence did not show that there was a significant effect of this type. It therefore appears that letters and in particular journals are the two specific text types that were early to favour the use of HAVE. Kytö's study thus allows us to determine very precisely in what textual categories HAVE and BE played out their battle for ascendance. In this, it is a good representative of the variationist approach to syntactic change, which Rissanen et al. (1997: 1) characterise as follows: In diachronic studies, the variationist approach provides us with a good opportunity to observe the actual process of change. We can trace the birth and death of variant expressions, but perhaps more interestingly, their changing frequencies and distributions within a variant field at subsequent periods of time and in various genres, and we can analyse changes in the intricate mesh of linguistic and extralinguistic factors conditioning the occurrence of these variants. Fittingly, Helsinki has been the leading centre of historical syntactic work of this type, as the contributions and references in volumes such as Rissanen et al. (1997) amply testify. Although most variationist studies pay close attention to linguistic as well as distributional factors, their focus is usually not on the question of what structural factors are responsible for the observable changes. Their strength rather lies in the exploitation of historical corpora of ever

Text types and the modelling of syntactic change

271

increasing size and scope to identify the routes and pathways along which the observable changes proceed in concrete situations and texts. As we saw in section 3, this is exactly the type of concern that is usually neglected in the structural approach to syntactic change. It therefore appears that, in a sense, the two approaches complement each other, and in what follows I shall suggest that there may indeed be ways of turning this complementation to good use in our quest for a better understanding of syntactic change.

5. Integrating the two approaches: general considerations Viewed within the overall framework set out in Weinreich, Labov and Herzog (1968) and Labov (1982), the two approaches to historical syntax distinguished above can be said to differ according to which particular aspect of syntactic change they are primarily focused on. The structural approach is mainly concerned with the actuation of change, while the variationist approach studies its transition or diffusion (further restricted in this paper to diffusion through text types). The question that we turn to now is: Can these two approaches be integrated? In earlier work, I have grappled with the question whether it is possible to integrate the actuation of syntactic change and its diffusion through the speech community, and to develop a coherent model of change in which both factors are included (see van der Wurff 1990). In what follows, I would like to present some reflections on whether this can also be done with actuation and diffusion through text types or registers. It has to be recognised right at the outset that there are certain obstacles to the construction of such an integrated model. At the practical level, it cannot really be said that the study of historical registers is very advanced; in the study of present-day registers, a wide range of methods has been employed (see for example Ghadessy 1988, and especially the very detailed work in Biber 1988, 1995, and Biber and Finegan 1994), but much less work has been done on historical registers. There have been some studies exploring the historical continuities and discontinuities of several registers (see especially work by the Helsinki school and by the group around Douglas Biber and Edward Finegan; also some of the papers in Jucker 1995) but most studies have tended to view historical registers as accidental repositories of old or new forms and structures, rather than as coherent wholes.

272

Wim van der Wurff

At a more theoretical level, there is some lack of agreement on what exactly a text type or register is. The most influential view of the concept is probably that developed in Hallidayan systemic functional grammar (e.g. Eggins 1994, Martin 1992), but even that suffers from a certain lack of clarity, as well as variable use of terminology by systemic linguists themselves (see Martin 1992: chapter 7). And much more variation is found when we move to the work of other linguists, where we encounter a true kaleidoscope of terms such as register, style, speech style, genre, way of speaking, text type and others (for discussion, see the review of the term "register" in Cheshire 1992). However, in all these different models and approaches it is possible to discover a certain measure of consensus about the types of variation that need to be incorporated in the theory of register; the differences mainly concern the relative importance attached to specific types of variation and the level of detail that investigators are willing to pursue. Since the aim of the present paper is to explore possible connections between existing approaches to historical syntax, rather than set either of the approaches themselves on a different footing, we will simply adopt the well known systemic functional view of text type or register and try to determine what relations there may be between structural factors and register viewed in this way. Adoption of other views may necessitate some reformulation of the results, but these seem relatively minor (for example, the suggestions made below can be transferred fairly easily to the framework put forward in Biber 1994, a conceptually elegant synthesis of earlier models of register). In the systemic functional scheme of things, register is determined by the three variables of field (the topic, or subject, of a text), tenor (the relationship between the people involved in the production and reception of a text) and mode (the role being played by language — compare for example a shopping list, an e-mail message, face-to-face conversation, language written to be spoken as if it is face-to-face conversation, as in film scripts, etc.). If we also include the overall function of the text (which in systemic functional grammar is sometimes regarded as a separate variable, that of genre), the language used in the following situations would all be examples of registers: • •

a transaction between a post office clerk and a customer buying stamps across the counter a caption to a photo accompanying a news article in a national daily

Text types and the modelling of syntactic change

• •

273

an Old English primer for use by first year university students a telephone conversation between two mothers making arrangements for baby-sitting

As writers about register never tire of pointing out, we can think of these and other registers as the actual locus of language use: this is where communication takes place, where meanings are made and structural options are put to use. It is not difficult to see that the purpose and context of language use has an effect on the linguistic choices that are made by speakers and writers: just consider the vast differences between the types of language found in the four registers given above. A plausible case can be made for the idea that these linguistic choices, which differ systematically from register to register in accordance with differences in field, tenor, mode and purpose, create different syntactic environments, and therefore different structural pressures promoting or retarding syntactic change. That is, if we take some common pathways of syntactic change, like the development from resultatives into perfects, or from nouns into prepositions, or from deontic modals into epistemic modals, or from word orders marked by topicalisation into canonical word orders, or from twoclause structures into monoclausal structures, we might expect these changes initially to manifest themselves most clearly in those registers where the relevant input conditions (the source construction plus the conditioning factors) are most strongly present. Taking seriously the notion of register variation would therefore result in a model of change shown in the diagram in figure 3: register 1

—construction A -> construction Β —conditioning factors

register 2

—construction A —

register 3

/

— / —conditioning factors

Figure 3. Model of the connections between structural and textual factors in syntactic change.

274

Wim van der Wurff

What is schematically represented here is a construction A, which may undergo change (i.e. develop into construction B) because of certain structural pressures (here called "conditioning factors"). There is one register in which construction A and the conditioning factors are indeed frequent, and there the change can take place. In registers 2 and 3, however, either the construction in question or the constructions creating the conditioning factors are absent or infrequent, so that there is no change in these registers. This scenario is of course very simple, but it supplies us with a possible framework for thinking about connections between structural and textual factors in syntactic change. For one thing, it may help us understand why a certain change advances most rapidly in a specific register. At the same time, it may enable us to identify more precisely the conditioning factors primarily responsible for the change, if we assume that registers which exhibit high frequencies of the new phenomenon also contain high frequencies of the conditioning factors. A problem in implementing these ideas is that in some structurally oriented theories of syntax, the notion of register does not seem to play any significant role. It is well known, for example, that in mainstream generative work, very little attention is paid to issues related to register variation. From reading an introductory textbook to generative grammar such Culicover (1997), one might even get the impression that basically there is no language-internal variation of this type. Only two individual registers, if that is the word, are singled out for attention by Culicover (1997: 2-3), in the following passage in his introductory chapter: ... people make mistakes or speak in a peculiar manner when they are drunk or tired. This is something that we (some of us) are familiar with. The mistakes that people make should not be included as part of'knowledge of language'. I am not aware of the existence of any study of the syntax of tired or drunken speech (unlike the phonology of drunken speech, some features of which Aitchison [1981] 1991: 174-175 very entertainingly compares with the types of sound change that are historically attested). But even apart from this, it is clear that Culicover's remarks are not very helpful for anyone interested in identifying the connections between structural factors and text types in syntactic change.

Text types and the modelling of syntactic change

275

There are, however, some generative studies of specific constructions in which register variation is taken seriously. Thus, Haegeman (1988) writes on empty objects in what she calls "instructional" texts, such as recipes, giving examples like (8). (8)

Skin and bone chicken, and cut 0 into thin slices. Place 0 in bowl with mushrooms. Purie remaining ingredients in blender and pour 0 over chicken and mushrooms. Combine 0 and chill 0 well before serving 0. (Haegeman 1988: 236-237)

Definite empty objects in English are quite restricted in occurrence, but Haegeman points out, citing Huang (1984), that in many languages, such as Portuguese, Chinese, Japanese and Korean, they are found quite freely. Huang (1984) analyses the difference between English and these other languages as one involving a syntactic parameter, by which languages are either subject-prominent or topic-prominent, as shown in figure 4: topic-prominent (Portuguese, Chinese, Japanese, etc.) versus subject-prominent (English, French) Figure 4. Parameter responsible for empty objects (Huang 1984) Haegeman (1988) then suggests that the phenomenon in (8) may be viewed as a case of language-internal variation in parameter setting: in the specific English register of instructional texts, the parameter setting is [+topic prominent], but in other registers, English is [+subject prominent]. In later work, Haegeman (1990, 1997) has looked at register variation involving empty subjects in English. Some examples of this phenomenon given by Haegeman (1997: 237) are: (9)

a. 0 walked there - feeling light and airy, (diary) b. 0 wish you were here, (informal note) c. Brilliant 0 could have stayed all day. (visitors' book)

276

Wim van der Wurff

Very prominent use of such empty subjects is made in Bridget Diary in the Daily Telegraph, from which (10) is an example. (10)

Jones's

Unfortunately, however, when 0 came to pour out tea 0 realised 0 did not have any milk or sugar. (Helen Fielding, Bridget Jones's Diary, Daily Telegraph, 2/5/1998: 24)

Note the two empty subjects in the embedded clauses in (10), a possibility that Haegeman (1997: 239) explicitly says does not exist. What is happening in (10) indeed seems to stretch the language quite far. It is interesting to see that in the passages of direct speech included in Bridget Jones's Diary, characters are not made to speak like this; the empty subjects in nonsentence-initial position seem to fit better in written language than in representations of spoken language. Haegeman (1997: 233) characterises the text types allowing empty subjects as abbreviated written and informal spoken registers. For this phenomenon, she proposes a special process of clausal reduction which is operative in these registers but not others. Although Haegeman does not use the term for this case, it certainly looks like another case of languageinternal register-based parameter variation. Since, in the theory of generative grammar, variation between languages is viewed as involving different parameter settings, it is indeed natural to invoke this notion also for variation internal to one language. It is often done for dialectal variation (see, for example, many of the papers in Black and Motapanyane 1996), and it seems plausible for register variation as well. Once we admit the possibility that different registers can instantiate different parameter settings, we are led to recognise the possibility that a parameter might undergo a diachronic shift on a register-by-register basis. The rather general remarks and reflections presented in this section are intended to suggest that it may indeed be possible to identify connections between structural factors and textual factors operative in syntactic change. The following section will contain a sketch of the questions and possible answers that all this would imply for the two specific cases dealt with in sections 3 and 4, i.e. the changes in word order possibilities around 1500 and the diachronic shift in the use of HAVE and BE to form perfects of intransitive verbs.

Text types and the modelling of syntactic change

277

6. Integrating the two approaches: specific cases We saw in section 3 that the loss of the plural marker in the past tense, as in (5), may have been responsible for the disappearance of the constructions illustrated in (1) to (4), i.e. for the resetting of the Split-IP Parameter. It must be assumed that this resetting of the parameter took place gradually. What registers, then, are likely to have been in the vanguard of the change? Those, we can now say, may well have been the registers in which there was most evidence for the new past tenses, i.e. those lacking the plural marker -en. This would mean that such registers would at least have to have a substantial number of past tenses in the first place. One such register might be narratives, in which Biber (1988) has found a high frequency of past tenses in present-day English. Moreover, if we assume that processes of phonetic reduction, of which the loss of plural -en is an example, are more likely to start in relatively informal spoken registers, then we can hypothesise that one of the registers first showing the effects of the resetting of the Split-IP Parameter may have been the register of informal oral narrative. This, of course, is a rather bold prediction, and detailed empirical research would be needed to substantiate this idea. The type of reasoning we have engaged in perhaps allows us somewhat more confidently to rule out certain registers as the initial locus of the new parameter setting. These would be written texts, and also spoken registers with low frequencies of use of the past tense. In terms of the diagram in figure 3, the situation for some specific registers in late Middle English could therefore be represented as in table 2: Table 2. Resetting of the Split-IP Parameter by register Register Properties Informal oral narrative AgrSP-TP-AgrOP-VP (i.e. fig. 2) past tense frequent zero plural marker frequent Formal oral narrative AgrSP-TP-AgrOP-VP past tense frequent -en plural marker frequent Transactional encounter AgrSP-TP-AgrOP-VP past tense infrequent zero plural marker frequent

Change to: IP-VP (i.e. flg. 1)

/

/

278

Wim van der Wurff

In this way, we could envisage the change in the Split-IP Parameter as having been gradual, because of the differential input conditions in different text types. Let us now try to see what structural issues arise for the variationist approach to HAVE versus BE as auxiliaries of the perfect. In the first section of her article, Kytö (1997: 18) mentions some of the factors that earlier research has suggested as causes of the shift towards HAVE (see also Denison 1993: 340-370). These include the heavy functional load of BE as auxiliary of the passive and the rapidly evolving progressive, the coalescence of is and has in contracted forms like it's and s/he's, and normative grammatical prescriptions. As far as the early prominence of HAVE in the text types of Journals and Letters is concerned, the influence of normative grammarians does not seem to be a plausible explanation, since it is difficult to see why it would affect these text types more strongly than others. The other two factors, however, suggest promising lines of further inquiry when the situation is viewed in the light of the diagram in figure 3. Thus, if it can be shown that the passive and progressive were particularly frequent in Journals and Letters, this means that these text types carried in them a conditioning factor for the shift from BE to HAVE. If it can be shown that contracted forms were also frequent in these texts, another factor could be added. Of course, other text types must also have contained instances of these conditioning environments, but they may have been less frequent and therefore have exerted less pressure towards change in the choice of HAVE versus BE. Some of the linguistic factors studied by Kytö (1997) may also be relevant to this further inquiry. As we saw in section 4, she found that HAVE was favoured in the following contexts: stative rather than mutative verbs, action rather than process verbs, the past perfect and the perfect infinitive rather than the present perfect, clauses containing an object-like element, and durative, iterative, conditional and optative contexts. If the registers of Journals and Letters contained high proportions of these particular environments, this would contribute to the higher frequency of HAVE, not only directly in these environments but also indirectly through stronger pressure on other environments to change by analogical levelling. Needless to say, all of this would require detailed further empirical work on the syntactic characteristics of historical registers through time. But there are some interesting pointers in existing work. Thus, the data in

Text types and the modelling of syntactic change

279

Appendix III of Biber (1988) show that the register of Personal Letters in Modern English has a high frequency of contractions, a category which includes the form 's for is/has, and judging by the description in Biber and Finegan (1989) this also appears to have been the case in earlier centuries, though at a slightly lower level. Biber and Finegan interpret this characteristic as being part of an underlying dimension along which texts can range from being highly informational to being highly involved, with personal letters coming towards the involved end of the continuum. The data in Biber (1988) also show that contractions in modern English are most frequent in face-to-face and telephone conversation, but much less so in some other text types that one might intuitively take to be fairly close to conversation. Projecting such differences into the past, we might conclude that historical registers, such as those in table 1, need to be studied on an individual basis, and that registers with certain shared situational characteristics do not necessarily have shared linguistic characteristics.

7.

Summing up

It may be useful to provide a brief summing-up of the main points made in this article. First of all, we noted that it is by now well established that there exists register variation of many kinds, including syntactic variation. As Biber, Conrad and Reppen (1998: 82) put it, "grammatical features are distributed in systematic ways across registers". As far as syntactic change is concerned, this means that registers will show systematic differences in the structural input conditions for syntactic losses and innovations to take place. Hence, we are led to expect systematic differences in the rate and nature of change in different registers. We examined two examples of rather different types to illustrate the kind of light that these general considerations can shed on the distribution and diachronic development of specific syntactic phenomena. To readers that are structurally inclined, this article presents a plea for a greater emphasis on empirical matters, to be precise, an emphasis on register variation and its possible structural consequences. I think this greater emphasis is necessary, since some structural studies have tended to neglect empirical matters, and have done so to their detriment. Sometimes, what gets explained in such studies is a very broad development, in which

280

Wim van der Wurff

matters of transition and diffusion are completely neglected (for a forceful recent statement of this point, see Currie 1995). As I hope to have shown, patterns of transition and diffusion may be amenable to interesting structural interpretation too. For researchers interested in studying the stylistic factors governing the distribution of variants and the spread of changes, this article can be regarded as a plea for greater emphasis on the structural significance of register variation, i.e. for a greater focus on the issue of actuation alongside those of transition and diffusion. I hope the reflections presented here, modest in scope though they are, constitute a step towards our common goal of locating language variation and change both in structures and in texts.

Notes 1. I would like to thank the audience at the Tenth International Conference on English Historical Linguistics and an anonymous referee for helpful suggestions and remarks.

References Aitchison, Jean 1991

Language Change: Progress or Decay? Second edition. Cambridge:

Cambridge University Press. First edition [1981], Biber, Douglas 1988 1994

1995

Variation across Speech and Writing. Cambridge: Cambridge

University Press. An analytical framework for register studies. In Douglas Biber and Edward Finegan (eds.), Sociolinguistic Perspectives on Register, 31-56. London: Oxford University Press. Dimensions of Register Variation: A Cross-Linguistic

Comparison.

Cambridge: Cambridge University Press. Biber, Douglas, Susan Conrad and Randi Reppen 1998

Corpus Linguistics: Investigating Language Structure and Use.

Cambridge: Cambridge University Press. Biber, Douglas and Edward Finegan 1989 Drift and the evolution of English style: A history of three genres. Language 65: 487-517.

Text types and the modelling of syntactic change

281

Biber, Douglas and Edward Finegan (eds.) 1994 Sociolinguistic Perspectives on Register. London: Oxford University Press. Black, James R. and Virginia Motapanyane (eds.) 1996 Microparametric Syntax and Dialect Variation. Amsterdam/Philadelphia: John Benjamins. Bobaljik, Jonathan David and Höskuldur Thräinsson 1998 Two heads aren't always better than one. Syntax 1: 37-71. Cheshire, Jenny 1992 Register and style. In: William Bright (ed.), International Encyclopedia of Linguistics, Volume 3,324-327. London: Oxford University Press. Culicover, Peter W. 1997 Principles and Parameters: An Introduction to Syntactic Theory. London: Oxford University Press. Currie, Oliver Written texts that speak out against theories of language change: 1995 The evidence for word order change in early Modern Welsh. Paper presented at the Twelfth International Conference on Historical Linguistics, University of Manchester, August 1995. Denison, David English Historical Syntax. London/New York: Longman 1993 Eggins, Suzanne 1994 An Introduction to Systemic Functional Linguistics. London: Pinter. Ghadessy, Mohsen (ed.) 1988 Registers of Written English: Situational Factors and Linguistic Features. London: Pinter. Haegeman, Liliane Register variation in English: Some theoretical observations. 1988 Journal of English Linguistics 20: 230-248. Non-overt subjects in diary contexts. In Joan Mascaro and 1990 Marina Nespor (eds.), Grammar in Progress, 165-174. Dordrecht: Foris. 1997 Register variation, truncation, and subject omission in English and in French. English Language and Linguistics 1: 233-270. Huang, James On the distribution and reference of empty pronouns. Linguistic 1984 Inquiry 15: 531-574. Jonas, Dianne Clause structure, expletives and verb movement. In: Werner 1996 Abraham, S.D. Epstein, Höskuldur Thräinsson and C. Jan-Wouter Zwart (eds.,), Minimal Ideas: Syntactic Studies in the Minimalist Framework, 167-188. Amsterdam/New York: John Benjamins.

282

Wim van der Wurff

Jucker, Andreas (ed.) 1995 Historical Pragmatics. Amsterdam/New York: John Benjamins. Kytö, Merja 1997 Be/have + past participle: The choice of the auxiliary with intransitives from late Middle to Modern English. In: Matti Rissanen, Merja Kytö and Kirsi Heikkonen (eds.), English in Transition: Corpus-Based Studies in Linguistic Variation and Genre Styles, 17-85. Berlin/New York: Mouton de Gruyter. Labov, William 1982 Building on empirical foundations. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 17-92. Amsterdam: John Benjamins. Lass, Roger 1992 Phonology and morphology. In: Norman F. Blake (edj, The Cambridge History of the English Language, Volume 2: 1066-1476, 23-155. Cambridge: Cambridge University Press. Martin, J.R. 1992 English Text: System and Structure. Amsterdam/Philadelphia: John Benjamins. Rissanen, Matti, Matti Kilpiö, Anneli Meurman-Solin, Saara Nevanlinna, Päivi Pahta, and Irma Taavitsainen 1997 Introduction. In: Matti Rissanen, Merja Kytö and Kirsi Heikkonen (eds.), English in Transition: Corpus-Based Studies in Linguistic Variation and Genre Styles, 1-15. Berlin/New York: Mouton de Gruyter. Roberts, Ian 1993 Verbs and Diachronic Syntax: A Comparative History of English and French. Dordrecht: Kluwer. Thräinsson, Höskuldur 1996 On the (non-)universality of functional projections. In: Werner Abraham, S.D. Epstein, Höskuldur Thräinsson and C. Jan-Wouter Zwart (eds.), Minimal Ideas: Syntactic Studies in the Minimalist Framework, 253-281. Amsterdam/New York: John Benjamins. Weinreich, Uriel, William Labov and Marvin I. Herzog 1968 Empirical foundations for a theory of language change. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Directions for Historical Linguistics, 95-115. Austin: University of Texas Press. van der Wurff, Wim 1990 Diffusion and reanalysis in syntax. PhD dissertation, Department of English, University of Amsterdam. 1997 Deriving object-verb order in late Middle English. Journal of Linguistics 33: 485-509.

The progressive form and genre variation during the nineteenth century Erik Smitterberg

1.

Introduction

The progressive form (or, simply, the progressive) has been increasing in frequency throughout Modern English. However, there is evidence to suggest that the construction has developed differently in different genres. A progressive form is defined here as a verb phrase containing a form of the verb BE and a present participle (or -ing form); the term "genre" will be used throughout the study to refer to texts grouped together on extralinguistic grounds. The aim of this study is to shed further light on the development of the progressive form over the nineteenth century by comparing five genres: Drama, Fiction, History, Letters and Science. Genres such as History and Science have not always been included in previous analyses, since the progressive form is assumed not to be frequent in these genres, and since many scholars have focused on genres which are assumed to be closer to speech. It is to be hoped that their inclusion here will yield some interesting results. Moreover, this study can complement earlier studies of the use of the progressive in late Modern English, such as Arnaud (1973, 1983) and Strang (1982). In what follows, I will first introduce the corpus and some models for the description of the progressive. I will then turn to the features under scrutiny in this paper, i.e. the analyses of normalised frequencies and clausal distribution. My reasons for selecting these two features will be presented when the analyses are dealt with. Throughout the study, I will focus on variation across genres.

2.

Material

The texts analysed in the present study make up the Corpus of NineteenthCentury English (henceforth CONCE). The CONCE project was started

284

Erik

Smitterberg

with the aim of providing some nineteenth-century British English followup material to the Helsinki Corpus. At present, CONCE comprises about 800,000 words, and includes five genres: Drama, Fiction, History, Letters and Science. These genres were selected for several reasons: they could be expected to yield a sufficient number of representative texts; they are among the core genres of the Helsinki Corpus·, and they can be said to align along a formal/informal continuum. CONCE is still being expanded to include some speech-based genres. The texts in CONCE are divided into three subperiods. Period 1 includes texts produced between 1800 and 1830, period 2 between 1850 and 1870, and period 3 between 1870 and 1900. The periods thus represent, broadly speaking, the beginning, the middle and the end of the century. The Letters genre comprises ten texts from each period/genre subsample; the other genres comprise three texts from each such subsample. However, one Drama text from period 1 and one Letters text from period 3 have yet to be included in the corpus. Each text comprises roughly 10,000 words. Since frequencies will be normalised throughout this study, slight variations in text length will not skew the results.

3.

Method

In previous research the progressive form has received a good deal of attention in recent decades. Some scholars have tried to discern a basic meaning applicable to all uses of the progressive. An interesting recent model in this respect is that presented in Ryden (1997). Ryden (1997: 421) sees the basic meaning of the progressive as expressing a "dynamic process" (physical, mental or textual). The uses of the progressive form are divided into two facets. Facet 1 is action-focused and includes aspectual functions. Facet 2 is attitude-focused and includes subjective, evaluative and interpretative functions. Other scholars have focused on certain functions of the progressive. Here can be mentioned for instance Wright's (1986, 1994, 1995) work on the progressive form as a vehicle for encoding subjectivity and foreground/background distinctions into texts. Where appropriate, I will return to these models when presenting the results of my study. Something should also be said about the collection of data for the study.

The progressive form and genre variation

285

First, a computer program (WordSmith Tools) was used to search for combinations of forms of the verb BE and words ending in -ing (including spelling variants). These combinations were then gone through in order to include only progressive forms proper in the analyses. Instances of BE going to functioning as a future auxiliary were not included. Finally, instances which shared part of their verb phrase with a previous progressive, such as was ... dancing in He was singing and dancing, were excluded from the study.1 After these selection processes, the total number of progressive forms analysed was 1,799.2

4. Results 4.1. Normalised frequencies An analysis of normalised frequencies, which now follows, makes it possible to relate developments within a genre to those attested for other genres, for the whole corpus, and in other studies. The analysis is therefore of interest within a framework of diachronic genre variation. Throughout the study, the term "frequency" is taken to mean "normalised frequency"; for the actual number of instances, the term "raw frequency" will be used. The commonest unit when measuring the frequency of the progressive is the M-coefficient (M after Mosse 1938). The M-coefficient equals the number of progressive forms normalised to a text length of 100,000 words. Since this figure is easily calculated and widely used, it was chosen for the present study.3 It should be pointed out that normalised frequencies can sometimes mask low underlying raw frequencies. For this reason, the raw frequencies for each period/genre subsample are given in table 1. Table 1. Raw frequencies for all period/genre subsamples in CONCE Fiction History Period Drama Letters 241 1 34 85 10 111 124 36 421 2 67 65 388 3 131

Science 28 24 34

It can be seen from the table that for History, period 1, and for Science, periods 1 and 2, the raw frequencies are below 30. This makes conclusions

286

Erik Sm itterberg

drawn from these period/genre subsamples less reliable. However, they can still give valuable indications of developments. The frequency developments for individual genres are given in table 2 and figure 1; table 2 also includes figures for the whole corpus. (For the sake of accuracy, the information given in figures is supplemented by tables throughout.) The figures for the entire corpus show that the progressive form roughly doubled its frequency in written British English over the nineteenth century — from 155.3 to 307.1, as measured in M-coefficients. It might be mentioned again that, at present, the Letters genre constitutes almost half of the corpus, which means that it influences the overall results more than the other genres do. Table 2. Normalised frequencies per 100,000 words for all period/genre subsamples in CONCE and for the whole corpus Fiction History All Period Drama Letters Science genres 202.1 32.3 193.5 73.5 1 165.5 155.3 75.2 2 317.7 317.3 117.9 317.9 266.2 212.4 414.4 345.7 222.1 110.7 307.1 3

. Drama . Fiction . History . Letters . Science

Period

Figure 1. Normalised frequencies per 100,000 words: genre developments

Concerning the individual genres, all five show consistent frequency increases over the century, with one exception: Fiction shows a decrease

The progressive form and genre variation

287

between periods 2 and 3. At this stage of the study, I surmise that the proportions of dialogue and narrative passages in the various Fiction texts affect the frequency of the progressive here. (I will pursue this hypothesis in my forthcoming work on the topic.) The figures for individual genres can be related to Biber and Finegan's (1997) work on diachronic relations among genres. In their study a factor analysis was made of a large number of linguistic features in texts from different genres. The study is based on the ARCHER Corpus, which includes American and British texts from the seventeenth century to the present day. The factor analysis resulted in a number of dimensions on which the linguistic features loaded.4 Three of these dimensions were given the labels "Involved vs. Informational Production", "Situation-Dependent vs. Elaborated Reference" and "Non-Impersonal vs. Impersonal Style". On these three dimensions, the genres Drama, Fiction and Letters approached Involved Production, Situation-Dependent Reference and Non-Impersonal Style over the centuries. Science, on the other hand, approached Informational Production, Elaborated Reference and Impersonal Style. This development resembles the frequency development of the progressive form. Drama, Letters and (apart from period 3) Fiction show high and clearly increasing figures. Science, however, shows lower and less clearly increasing figures.5 The result is that these two groups of genres become increasingly separated over the century with respect to the frequency of the progressive form. The progressive is not included in Biber and Finegan's factor analysis; however, the results of the frequency analysis presented here suggest that the construction may co-occur with other features indicating Involved Production and Situation-Dependent Reference.6 It is notable that the labels which Biber and Finegan assign to their dimensions lend support to a connection to the progressive. Genres with high frequencies of progressive forms tend towards the labels "Involved Production", "Situation-Dependent Reference" and "Non-Impersonal Style". Such terms are common in descriptions of the progressive, and Ryden (1997: 421, 424 respectively) specifically mentions situational immediacy and personal involvement as characteristic of the progressive form. Moreover, Biber and Finegan (1997: 260) say that the three dimensions mentioned above "can be considered 'oraP/'literate', in that they distinguish between stereotypically oral (conversational) registers at one

288

Erik

Smitterberg

pole, and stereotypically literate (written expository) registers at the other pole"; and many scholars have discussed the conversational nature of the progressive form. The History genre is not included in Biber and Finegan's study. In the present analysis, however, History seems to change over the nineteenth century with respect to the frequency of the progressive form: History is close to Science in period 1, but increases clearly in frequency over the century and is closer to Fiction in period 3.

4.2. Clausal analysis I will now turn to an analysis of the clausal distribution of the progressive form. Strang (1982) and Wright (1994) both discuss the distribution of the progressive over different types of clause. However, Strang focuses on Fiction and Wright on Drama. I decided to extend their work within a framework of genre variation. In the present study, the 1,799 progressive forms were classified according to the type of clause in which they occurred: main, adverbial, comparative, nominal, or relative. Quirk et al. (1985: 1045-1146, 1244-1260) was followed when distinguishing between clause types. (A small number of ambiguous instances were left unclassified.) I then calculated percentages for each period/genre subsample. In this way I obtained a unit independent of the frequency of the progressive form. At this point it can be claimed that this unit depends not only on the tendency of the progressive to occur in the five clause types, but also on distributional changes among these clause types in themselves. For instance, a general increase in main clause usage might be interpreted as a change from subclauses to main clauses in the distribution of the progressive form. 7 However, Strang's (1982: 442) investigation into the clausal distribution of the progressive in narrative prose in late Modern English indicates that diachronic changes in the clausal distribution of the progressive form cannot be accounted for by this factor alone. The results for the whole corpus are given in table 3 and figure 2. As can be seen from the table and the figure, there is a clear percentage increase in the use of the progressive in main clauses during the century. Over the same period, the percentage of adverbial clauses decreases. This

The progressive form and genre variation

289

development is interesting in several ways. It might be linked to theories of connections between aspect, clausal distribution and grounding (see, for instance, Hopper 1979). Main clause occurrence is listed in Wright (1994: 472) as a diagnostic for what she calls the "modal" progressive, which is foregrounding and non-aspectual, as opposed to progressives that are backgrounding and aspectual. This division has much in common with Ryden's (1997) division between facet 1 and facet 2. The results presented in table 3 and figure 2 seem to indicate that important developments concerning the distribution of these two facets took place in the nineteenth century. Table 3. Percentages of Period Adverbial 1 17.8 2 12.6 3 12.3

clausal distribution for the whole CONCE Corpus Comparative Main Nominal Relative Unclass. 0.0 49.2 19.1 12.6 1.3 0.4 53.8 17.5 13.5 2.2 56.2 17.1 0.1 13.0 1.3

. Adverbial . Comparative .Main .Nominal . Relative -Unclassified

Figure 2. Developments in clausal distribution for the whole CONCE Corpus

If we proceed to look at the individual genres, a division into three groups can be made with respect to clausal distribution. As can be seen from tables 4 and 5 and figures 3 and 4, Drama and Letters both have a main clause percentage of 50 per cent or more already in period 1, and

290

Erik Smitterberg

show consistently high percentages throughout the century. Moreover, the corresponding decreases in percentage come chiefly in adverbial clauses for both genres. The findings for the Drama genre can be compared to those presented in Wright (1994: 474) for prose comedy between 1670 and 1710. Wright found no significant change in clausal distribution overtime. Table 4. Percentages of clausal distribution for the Drama genre Main Nominal Relative Period Adverbial 1 50.0 11.8 17.6 17.6 8.1 10.8 66.7 2 10.8 17.6 16.0 61.1 3 5.3 Table 5. Percentages of Period Adverbial 16.2 1 2 10.9 3 10.01

Unclassified 2.9 3.6 0.0

clausal distribution for the Letters genre Comparative Main Nominal Relative 0.0 53.5 19.5 9.1 56.3 20.0 10.9 0.2 0.3 60.6 18.3 9.5

Unclass. 1.7 1.7 1.3

• Adverbial • Main .Nominal • Relative • Unclassified

Figure 3. Developments in clausal distribution for the Drama genre

The progressive form and genre variation

291

Adverbial Comparative Main Nominal Relative Unclassified

Figure 4. Developments in clausal distribution for the Letters genre The second group consists of Fiction and History. The results for these two genres are shown in tables 6 and 7 and figures 5 and 6. Fiction and History have lower but increasing percentages of main clause progressives. Moreover, the corresponding decreases in percentage come in nominal and relative clauses rather than in adverbial clauses. 8 The findings for the Fiction genre are in line with those presented by Strang (1982) for narrative prose. Strang (1982: 441-443) found that the clausal distribution of main and subordinate clauses shifted over the nineteenth century in favour of main clauses. Table 6. Percentages of clausal distribution for the Fiction genre Period Adverbial Comparative Main Nominal Relative 1 15.3 0.0 44.7 21.2 18.8 0.8 43.5 17.7 16.9 2 17.7 16.4 0.0 53.7 16.4 9.0 3 Table 7. Percentages of clausal distribution for the Main Period Adverbial 30.0 1 20.0 33.3 2 19.4 46.2 3 16.9

History genre Nominal 30.0 13.9 10.8

Unclass. 0.0 3.2 4.5

Relative 20.0 33.3 26.2

292

Erik Smitterberg

.Adverbial . Comparative .Main .Nominal .Relative -Unclassified

1

2

3

Period Figure 5. Developments in clausal distribution for the Fiction genre

•Adverbial •Main •Nominal •Relative

Figure 6. Developments in clausal distribution for the History genre The third group consists only of Science. As can be seen from table 8 and figure 7, Science is the only genre to show a decrease in the percentage of main clauses over the nineteenth century. Moreover, it seems that at least between periods 2 and 3 there is a corresponding increase in the percentage of adverbial clauses. This is a development opposite to that of the genres in the first group, namely Drama and Letters.

The progressive form and genre variation Table 8. Percentages of clausal distribution for the Science genre Period Adverbial Comparative Main Nominal Relative 39.3 0.0 32.1 14.3 1 14.3 2 12.5 4.2 33.3 20.8 25.0 47.1 0.0 11.8 14.7 23.5 3

293

Unclass. 0.0 4.2 2.9

. Adverbial . Comparative .Main .Nominal . Relative -Unclassified

Figure 7. Developments in clausal distribution for the Science genre It is also possible, however, to merge the first two groups. This larger group, including Drama, Letters, Fiction and History, would have the common denominator of high and/or increasing main clause percentages, as opposed to Science only. The group would then consist of two subgroups, one including Drama and Letters and one including Fiction and History.

5.

Summary and conclusion

In this study, two analyses of the progressive form during the nineteenth century were carried out. First, a frequency analysis showed that all the genres in the corpus except Science show high and/or clearly increasing frequencies over the century. These results were related to the factor analysis carried out by Biber and Finegan (1997). My hypothesis was that the progressive would be an interesting candidate for inclusion in the factor analysis. I also suggested that the History genre seems to change over the nineteenth century with respect to the frequency of the progressive form,

294

Erik Smitterberg

from being closer to Science to being closer to Drama, Fiction and Letters. A clausal analysis supported the division of the genres into two groups. Letters, Drama, Fiction and History all show high and/or increasing percentages of main clauses, whereas Science shows decreasing percentages. It is possible to make a further division between Drama and Letters on the one hand and Fiction and History on the other. I also related my clausal analysis to those carried out by Strang (1982) for prose narrative and Wright (1994) for prose comedy. A few caveats are called for. First, some period/genre subsamples yielded too few examples of progressive forms for frequencies and percentages based on them to be wholly reliable. Second, CONCE as yet includes too few speech-based genres to allow us to generalise the results to both speech and writing. However, this is being remedied by the inclusion of more speech-based genres in the corpus (keeping in mind the editorial process nineteenth-century spoken texts went through when being transferred into written form). Third, overall changes in the relative distribution of clause types will affect the results attained from the clausal analysis. Despite these caveats, the present study has shown that the progressive form seems to become an increasingly genre-specific feature over the nineteenth century. The clear trends in the data show that one century may be a sufficient time frame for carrying out diachronic analyses of the progressive form. However, a prerequisite for this is that the researcher has access to a relatively large, stratified corpus. The present study has also shown that important insights into the nature of the English progressive can be obtained from studying genres which yield comparatively few instances. If Science had not been included in the study, for instance, the parallels between the results of the present study and Biber and Finegan's (1997) factor analysis would have been less clear. Needless to say, the genre parameter alone does not account for the use of the progressive form. In my future work, I intend to look into further linguistic and sociolinguistic factors.

Notes 1. As is well known, the presence of the progressive auxiliary in a previous instance would make a following progressive form more likely than would otherwise be

The progressive form and genre variation

2.

3.

4.

5.

6. 7. 8.

295

the case. For this reason, it was decided that such instances should be excluded from quantitative analyses in this study. To ensure that the electronic collection of instances of progressives was adequate, I checked one period/genre subsample manually. Period 3 of the Fiction genre was selected, and the manual checking revealed a further two instances. This means that the accuracy rate of the collection procedure was (67/69 χ 100)% » 97.1%. This was considered sufficient. (The two additional instances found were not included in the present study, since doing so might skew the genre comparison.) An alternative to using the M-coefficient would be to use a normalisation procedure which takes into account only those verb phrases where there is actually competition between simple and progressive forms, rather than the total number of words, which may have little to do with how many progressive forms would be possible in a given text. An example is the K-coefficient used by Nickel (1966), and later by Scheffer (1975). However, there are several drawbacks associated with such a procedure. First, it presupposes an agreedupon theory of the progressive form which clearly states in what environments the progressive is grammatical. (This is a big problem especially in diachronic studies of genre variation, where the possibility of inserting a progressive form may change over time — and, possibly, between genres.) Second, it is considerably more time-consuming, since the researcher has to go through all verb phrases in his/her corpus in order to exclude those where there is no competition between simple and progressive forms. In this factor analysis, "each factor represents ... a grouping of linguistic features that co-occur with a high frequency" (Biber 1988: 79). As for factor loadings, they "indicate the degree to which one can generalize from a given factor to an individual linguistic feature ... Thus, features with higher [i.e. further from 0.0] loadings on a factor are better representatives of the dimension underlying the factor, and when interpreting the nature of a factor, the features with large loadings are given priority" (Biber 1988: 81 [addition mine]). As shown by Hundt and Mair (forthcoming) in a study carried out on the LOB, Brown, FLOB and Frown corpora, academic texts still lag behind other genres in this respect. There are no positive features for the dimension called "Non-Impersonal vs. Impersonal Style". I am grateful to Sali Tagliamonte for drawing my attention further to this issue. For History, period 1, only ten instances of progressive forms were found. For this reason, both percentages calculated for period 1 and percentage changes between periods 1 and 2 are rather unreliable for this genre. The similarity to the development of the Fiction genre becomes clearer if only changes between periods 2 and 3 are taken into account as regards History.

296

Erik Smitterberg

References Arnaud, Rend 1973 1983

Biber, Douglas 1988

La forme progressive en anglais du XIXe siecle. Lille: Service de reproduction des thöses. On the progress of the progressive in the private correspondence of famous British people (1800-1880). In: Sven Jacobson (ed.), Papers from the Second Scandinavian Symposium on Syntactic Variation, 83-94. (Stockholm Studies in English 57.) Stockholm: Almqvist & Wiksell.

Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas and Edward Finegan 1997 Diachronic relations among speech-based and written registers in English. In: Terttu Nevalainen and Leena Kahlas-Tarkka (eds.), To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, 253—275. (Mömoires de la Sociötö Nöophilologique de Helsinki 52.) Helsinki: Soci£t£ Nöophilologique. CONCE = A Corpus of Nineteenth-Century English forthcoming Compilers: Merja Kytö (Uppsala University) and Juhani Rudanko (University of Tampere). Hopper, Paul J. 1979 Some observations on the typology of focus and aspect in narrative language. Studies in Language 3 (1): 37-64. Hundt, Marianne and Christian Mair forthcoming "Agile" and "uptight" genres: The corpus-based approach to language change in progress. International Journal of Corpus Linguistics. Mossi, Fernand 1938 Histoire de la forme periphrastique etre + participe prisent en germanique, Volumes 1-2. Paris: C. Klincksieck. Nehls, Dietrich 1988 On the development of the grammatical category of verbal aspect in English. In: Josef Klegraf and Dietrich Nehls (eds.), Essays on the English Language and Applied Linguistics on the Occasion of Gerhard Nickel's Sixtieth Birthday, 173-198. (Studies in Descriptive Linguistics 18.) Heidelberg: Julius Groos Verlag. Nickel, Gerhard 1966 Die Expanded Form im Altenglischen. Neumünster: Karl Wachholtz.

The progressive form and genre variation

297

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Rydin, Mats 1997 On the panchronic core meaning of the English progressive. In: Terttu Nevalainen and Leena Kahlas-Tarkka (eds.), To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, 419-429. (Mömoires de la Societe Niophilologique de Helsinki 52.) Helsinki: Sociötö Nöophilologique. Scheffer, Johannes 1975 The Progressive in English. Amsterdam: North-Holland. Strang, Barbara Μ. H. 1982 Some aspects of the history of the be + ing construction. In: John Anderson (ed.), Language Form and Linguistic Variation: Papers Dedicated to Angus Mcintosh, 427-474. (Current Issues in Linguistic Theory 15.) Amsterdam: John Benjamins. Wright, Susan 1986 Tense, aspect and text: Processes of grammaticalisation in the history of the English auxiliary. PhD dissertation, Cambridge University. 1994 The mystery of the modal progressive. In: Dieter Kastovsky (ed.), Studies in Early Modern English, 467-485. (Topics in English Linguistics 13.) Berlin/New York: Mouton de Gruyter. 1995 Subjectivity and experiential syntax. In: Dieter Stein and Susan Wright (eds.), Subjectivity and Subjectivisation: Linguistic Perspectives, 151-172. Cambridge: Cambridge University Press.

The conjunction and in early Modern English: Frequencies and uses in speech-related writing and other texts Jonathan Culpeper and Merja Kytö

1.

Introduction

The general aim in this and in our other recent papers is to improve our understanding of the language of spoken face-to-face interaction in the early Modern English period. To this end, we have been working on the construction of a corpus of speech-related dialogue texts (the Dialogues Corpus referred to below).1 Though of course even so-called verbatim transcripts, such as trial proceedings, are not accurate recordings of a speech event, these kinds of text are the best evidence we have.2 The more specific aim of this study is to explore the usage of and in two text types in that corpus — trial proceedings and drama. To facilitate a comparison we will also explore two typical written text types, science and history, both drawn from the Helsinki Corpus. Our reason for deciding to investigate and is that scholars have shown that and plays an important role in spoken interaction.3 This point, however, seems to have escaped the attention of many historical studies and currently popular theoretical approaches. Amongst historical studies, Traugott (1986) presents an insightful account of the semantic-pragmatic development of and. However, the remit of her paper does not extend to a consideration of and as part of different text types and language media, and the implications this may have had for its development. Amongst currently popular theoretical approaches, the Relevance Theory approach is noteworthy. For example, Carston (1994) argues for a Relevance-based account of the sequential and consequential relations in the interpretation of conjunctive utterances. However, Carston (1994) only considers the role of and in coordinating two clauses in a sentence. No attention is given to context — the role of multiple usages of and, the textual context, and used across exchanges, and so on.

300

Jonathan Culpeper and Merja Kytö

There are three main parts to this essay. In the first, we will briefly introduce three studies which have considered and in present-day speech and writing. In the second, we will consider the frequency distribution of two types of and in both speech-related and non-speech-related early Modern English text types. In the third, we will discuss a number of examples, in order to show how and is being used in our various text types.

2. And in present-day English speech and writing 2.1. Chafe ([1982] 1984), and Chafe and Danielewicz (1987) In his often-cited study, Chafe points out that the so-called "idea units" of speech are often held together by and:4 (1)

A—nd it's ... very well.. urn equipped ....You know the kitchen, ... and and it's got a dishwasher, and it's got... uh .. all kinds of... you know ... mixers and ... uh ... plates and ... you know every kind of equipment you need. And .. a—nd uh— .. staple .. things. (Chafe 1984: 38 [our lineation])

Chafe also gives frequencies for and in informal spoken language and formal written language, basing his study on two types of language — dinner-table conversations and academic writing — which he took to be "maximally differentiated" examples of speech and writing (Chafe 1984: 36). And occurs four times as often in spoken conversation and much more frequently than the other conjunctions examined (see table 1). Table 1. Frequency of conjunctions in spoken and written data (occurrence per 1,000) (Informal spoken = 9,911 words; Formal written = 12,368) (from Chafe 1984:39) Spoken Written 44.2 10.1 and 4.1 9.8 but 7.4 0.0 so 4.7 1.5 because 66.0 15.8 "combined"

The conjunction and in early Modern English

301

A more detailed picture emerged in a follow-up study, which comprised conversations, lectures, letters and academic papers (for details, see Chafe and Danielewicz 1987). In the light of the results obtained in this study, conversations contained by far the largest amount of clause-level coordination (even lectures contained twice as much as letters), whilst academic writers tended to avoid this use more than others (Chafe and Danielewicz 1987: 103). Conversely, phrase-level coordination was more frequent in writing (letters and academic papers) than in speech (conversations and lectures) and, in fact, three times more frequent in academic writing than in conversation (Chafe and Danielewicz 1987: 101).

2.2. Longman Grammar (Biber et al. 1999) Interesting results concerning the occurrence and distribution of and in various text types on a larger scale are presented in the Longman Grammar of Spoken and Written English (1999).5 Essentially, the grammar is based on a corpus of some 40 million words, containing spoken conversation, fiction, news and academic prose. In this corpus the incidence figures for and (when counted per 1,000 words for each text type) are around 20 (spoken conversation), 28 (fiction), 20 (news) and 27 (academic prose) (Biber et al. 1999: 81). Moreover, clause-level and is the most frequent in spoken conversation (in nearly 80 per cent of the instances as against phrase-level and) and phrase-level and in academic prose (in nearly 70 per cent of the instances as against clause-level and). These results move beyond the findings obtained by Chafe and Danielewicz (1987). Notice, firstly, that — contrary to what one might have supposed from Chafe's work — conversation does not have the highest frequency of and: both fiction and academic writing contain more. However, in accord with Chafe and Danielewicz (1987), it is conversation that contains the highest proportion of clausal and, whereas academic writing contains the highest proportion of phrasal and. We will follow up these aspects in our own data.

302

Jonathan Culpeper and Merja Kytö

3. And in early Modern English With the present-day studies introduced in section 2 in mind, we turn to consider the frequencies and uses of and in early Modern English texts. Clearly, the Longman Grammar of Spoken and Written English has revealed the importance of distinguishing between clause-level and phraselevel and in the examination of different registers. In fact, this can be a difficult distinction to make, particularly when distinguishing between the coordination of clauses versus the coordination of verb phrases. Section 3.1 addresses some of the classificational issues. However, there is no claim here that the functions of and are restricted to syntactic coordination. It rapidly becomes apparent when studying the uses of and in speech-related texts that and can coordinate speech acts and serve other discoursal functions. We will consider these at the beginning of section 3.3.

3.1. Classificational criteria In order to classify our data reliably, we needed to consider how coordination works. As pointed out by Quirk et al. (1985: 46-47), coordination is a type of construction where "two or more units of the same status on the grammatical hierarchy may constitute a single unit of the same kind". The number of coordinated constituents is open-ended, but commonly a conjoint consists of two coordinated units (or conjoins). These are mostly clauses, phrases, or words. As coordination has many variations and complications, an exhaustive discussion of all aspects of the issue is beyond the scope of the present study; instead, we will focus on those issues that turned out to be of major relevance to our classification of examples. To investigate what it is that and holds together in early Modern English, the distinction made between clause-level and phrase-level coordination in studies carried out on present-day English was a useful starting point. To enable comparisons with our data and the results presented in the Longman Grammar, it was advisable to apply, as far as possible, the classificational principles put forth in that work and in Quirk et al. (1985). The instances below are given in the latter as prototypical examples of the two categories (1985: 46):

The conjunction and in early Modern English

303

Clause-level coordination: (2)

[[It was Christmas Day,] and [the snow lay thick on the ground]].

Phrase-level coordination: • (3) • (4)

coordination of prepositional phrases: You can go [[by air] or [by rail]]. coordination of nouns: His [[son] and [daughter]] live in Buenos Aires.

Example (4) calls for a comment on the role played by ellipsis. Examples such as this could be analysed as instances of clausal coordination, i.e. (4')

[[His son lives in Buenos Aires] and [his daughter lives in Buenos Aires]].

However, as the conjoins here expand a simple noun phrase, these examples can more conveniently be considered as instances of phrasal coordination (as explained in the Longman Grammar). Another classificational issue of particular importance and difficulty concerns the verb phrase. While noun phrases (heaven and earth, falshoods and untruths, Pride, Wit, and Quality), adjective phrases (readiest and clearest, sincere and hearty) and adverb phrases {up and down, now and then, with Love and with Despair)6 were relatively easy to identify, the question of how to pin down a verb phrase needed further consideration. A helpful discussion can be found in Quirk et al. (1985: 943-944, 967). In the sentence they give, (5), two functionally comparable verb elements are conjoined to express what can be regarded as a single combined activity (two separate actions being the less felicitous interpretation). (5)

Peter washed and dried the dishes.

In the case of this example, the question-answer pair What are you doing to the dishes? I'm washing and drying them can be applied to test the

304

Jonathan Culpeper and Merja Kytö

semantico-pragmatic status of the action involved. Applying the above strategy to our corpus examples was not all that easy to start with.7 The danger of arbitrary labelling could be prevented only by setting up an analytical apparatus that could make the claim of replicability. Two rounds of analysis were needed, the first based on the application of syntactic criteria and the second on the application of semantico-pragmatic criteria. Over the first round instances with functionally comparable verbs which would make potential candidates for the verb phrase group were set aside, and over the second round an attempt was made to identify the sameness or unity of action that would warrant a distinction between phrasal and clausal readings. The first round was carried out relatively smoothly. As syntax and semantics work together in that complementation tends to separate actions, only instances of "VP + (one element) + and + VP + (one element)" were included in the group to be screened over the second round. By "one element" is meant here extensional elements accompanying the verb in a close relationship, e.g. reflexive pronouns and prepositions. Over the second round, a number of data-generated contextual tests were resorted to, among them scrutiny of the semantic properties of the verbs and the degree of seriality involved in the action. The criteria supporting a phrasal interpretation included semantic parallelism between the verbs and the applicability of the "(in order) to + infinitive" construction indicating purpose (with verbs of locomotion, in particular). By way of illustration, examples of semantic parallelism are given in (6a) to (6e). In (6a) to (6c) the conjoins are used to denote the same action ([6a] contains a rhyming word pair at that). In (6d) we have semantically parallel and contrasting actions. In some examples, one of the conjoins could be dropped and the verb phrase still makes perfect sense (for example [6b]). (6) a. He told me, none but naughty Women sate there, whom they tous'd and mous'd; but I wou'd have ventur 'dfor all that. (DC/Drama/Wycherley 273) b. Sir (ATim.A)

Dear Cosin, be in good humour, I could wish my self well beaten for mistaking one that loves me so; I would I might ne 're stir, if I did not think you had

The conjunction and in early Modem English

305

been in earnest; well, but I vow and swear I am mightily beholden to you, that you think me so fine a person, and love me so dearly; ... (DC/Drama/Shad we 11156) c. ... That 'tis some time before the Cork that covers the young and tender sprouts comes to be discernable; That it cracks, flaws, and cleaves into many great chaps, the bark underneath remaining entire; ... (HC/Science/Hooke 13.5,115) d. Mrs. (ARawlinsA) and I were in a great consternation, wondring what they would do with us, I took hold of her Arm and told her I would live and dye with her. (DC/Trial/Swendsen 5) Examples are given in (7a) and (7b) supporting the sameness of action criterion and the phrasal interpretation with verbs of locomotion where paraphrase by the "(in order) to + infinitive" construction applies (i.e. [7a] is equivalent to go to get and [7b] to come to give). (7) a. When she took my Ring away, I askt her what she would do with it, she said she would go and get a Wedding Ring made by it. (DC/Trial/Swendsen 8) b. I know my Duty to my King better, and have always exercised it, I defy any body in the world that ever knew the contrary, to come and give Testimony. (DC/Trials/Lisle IV.122.C2) A test that seemed to help distinguish instances of clausal coordination from the verb phrases over the second round of analysis, was the 'and then' gloss. When seriality of action emerges as a powerful feature in the context, we have two separate actions at hand, as in (8). (8)

At his return therfore he built and dedicated a Church to St. (AEdmundA) at (ABuryA), whom his Ancestors had slain, threw out

306

Jonathan Culpeper and Merja Kytö

the secular Preists who had intruded there, and plac'd Monks in thir stead; ... (HC/History/Milton X,278) Clearly, natural language data also contains "indeterminate" instances, where it is not possible to draw the line between a clausal and phrasal interpretation. When no test or combination of tests determined whether the instance in question was clausal or phrasal, we classified the instance as indeterminate and excluded it from our counts. Such instances are given in (9a) to (9c); in these cases it is not clear whether the verbs denote one and the same action or whether they represent two clearly separate actions. (9)

a. Then Mrs. (ASabina BusbyΛ) was called and Sworn, as also Mrs. (^Nightingale*). (DC/Trials/Swendsen 3) b. I went into the House, and she vanisht, and there was nothing but the Cat in the middle, who spit and star'd at me, and I was frighted away. (DC/Drama/Shadwell 145) c. My Mother recovered, and came and hung about my Neck, so that they could not get me into the Coach, and (AMr. Page*) went to call Company to rescue me. (DC/Trials/Mohun 16)

As some subjectivity is unavoidable when weighing the semanticopragmatic shades of contextual features, we are aware of the problems involved in the application of the above tests. What looks phrasal from one point of view is clausal from another. However, by using context-anchored screening over two rounds of analysis, we hope to have arrived at a classification that can be defended in terms of consistency and replicability. We might also point out that as a whole, the verb phrase group represents a relatively small section of the data: out of the 1503 instances of phrase coordination, no more than nine per cent (or 135 examples) belong to the category of verb phrase coordination. Various types of indeterminate instances (between clause and any type of phrase coordination) remain an even smaller group, all in all 43 instances.

The conjunction and in early Modern English

307

3.2. Frequency distribution in early Modern texts Excluding a small number of indeterminate examples (43, or one per cent of the grand total of 4150 instances), table 2 presents the distribution of all instances of and in our four text types (for graphic illustration, see figures 1 and 2). Table 2. The distribution of and as clause-level versus phrase-level connector. (Figures in brackets represent incidence per 1,000 words.) (Science and History are drawn from the Helsinki Corpus·, Drama and Trials are drawn from the Dialogues Corpus.) Text type Clausal Phrasal Total 54.7% SCIENCE (1640-1710) 192 159 45.3% 351 Words = 11,280 (17.0) (14.1) (31.1) HISTORY (1640-1710) 235 56.6% 180 43.4% 415 (20.2) (15.5) Words = 11,640 (35.7) DRAMA (1640-1720) 938 60.5% 612 39.5% 1550 (15.5) (10.1) Words = 60,639 (25.6) 69.2% TRIALS (1640-1720) 1239 552 30.8% 1791 Words = 60, 531 (20.5) (29.6) (9-1) 2604 63.4% Total 1503 36.6% 4107 (10.4) Words = 144,090 (28.5) (18.1)

• Phrasal Clausal

Science

History

Drama

Trials

Figure 1. The relative proportions of and as a clause-level versus phrase-level connector (according to the percentages given in table 2)

308

Jonathan Culpeper and Merja Kytö

40 35 30 25 20 0 Phrasal

15

•Clausal

10 5

0 Science

Hstory

Qama

Trials

Total

Figure 2. The relative densities of and as clause-level versus phrase-level connector (according to the normalised figures given in table 2) On the basis of the results obtained for present-day data reported in section 2 above, we might expect a clear difference between the speech-related text types and the non-speech-related text types, and variation in the figures obtained for the individual text types. The table shows that the proportion of clausal usages increases steadily from Science through History and Drama to Trials (Science having some 55 per cent and trials some 70 per cent; the figures in table 2 are statistically significant: chi-square = 50.99, 3 d.f.; ρ < 0.001). In some respects, this is in accord with the Longman Grammar (and Chafe and Danielewicz 1987) in that spoken conversation contains the largest proportion of clausal coordination, whereas academic writing contains the least. However, one might note that Drama — which some have claimed to be the closest to spoken conversation8 — lags behind Trials. This, we will suggest, is because trials contain narratives delivered by witnesses, and these narratives are dense with clausal and, as we will show below. In narrative, and is used as a means of sequencing the events that constitute the narrative, and typically has the sense 'and then'. Narrative also explains the relatively high density — 20.2 — of clausal and in History, which of course also contains narrative of past events. The

The conjunction and in early Modern English

309

relatively high proportion of phrasal and in both Science and History, we would suggest, relates to both informational density and rhetorical effects (we will show examples of this below). That History is relatively dense in both phrasal and clausal coordination accounts for the fact that it has the highest overall density of and— 35.7. This text type in particular contains narrative, informational density and rhetorical effects. These observations are simplifications to some extent. We will cast further light on the use of and in different text types by looking at some examples in section 3.3. It is intriguing to compare our overall figures for the incidence of and with the trends described in the Longman Grammar. Of course, there is no easy one-to-one correspondence between the text types (and theirs was a much bigger corpus and they had a classification system different from ours in some respects). Nevertheless, the general trend is clearly that the incidence was higher for the early Modern English texts. For our texts, the overall incidence is about 29, while for the Longman Grammar it is about 24. This is also the case when one compares our figures with Chafe's, and Chafe and Danielewicz's. Moreover, the difference is striking when one focuses specifically on clause-level and. For our texts, the incidence of clause-level and is about 18.1, while the approximate figure for the Longman Grammar is 13.9. This finding for clause-level and seems to tally with what Blake (1996: 226) observes when discussing the use of conjunctions in the period 1400 to 1660: In the formation of complex sentences, one can detect a growing use of conjunctions to make clear the relationship between the various clauses which form a sentence. The growth in the concept of a sentence accelerated this development. In previous periods of English and still today in colloquial English, syndetic co-ordination and subordination are common.

The final sentence hints that the incidence of clause-level and has decreased in some varieties of English, since the early Modern English period. Our evidence would support this. As to why this shrinkage might have taken place, Blake (1996: 226) hints at a possible reason: "The increasing regularisation of clauses meant that a large number of new conjunctions was needed and the function of those already in the language had to be more clearly distinguished". In other words, other conjunctions might have been used instead of and for particular functions. This is clearly the case

310

Jonathan Culpeper and Merja Kytö

for the conditional usage of and, which has been taken over by i f . In our data spanning 1640 to 1720 there are only 19 instances of the conditional and and its variant an, all occurring in Drama (see example [10a]); the 19 examples amount to 0.31 instances per 1,000 words in Drama texts); of these nine occur in more or less fixed politeness formulae, and four in dialectal speech (see example [10b]). (10) a. (ALod.A)

b. La. (ASha.A) (^onstA) (ATho. ο G.A)

Well (AVallentiusA), and you be caught ith' purlues: and you be not stung for't lie forsweare privacie, and all that belongs too't, I have a Girle, the very spirit of what she was made for, and she were honest, she might crave supremacie of (AHellenA), and make her ride behinde. (DC/Drama/J.D. B4V) Come, has any one else any thing to inform? Yes, an't please your worship, here is a Neighbour, (AThomas ο Georges A) Why, a n ' t please your Worships, I was at (AMal. Spencer 'sA) House, where she wons i 'th' Lone, and whoo has a meeghty great Cat, a black one by'r Lady, and whoo kist and whoo clipt Cat, and ay sent me dawn a bit (meet a bit) and believe Cat went under her Coats. (DC/Drama/Shadwel 1 146)

However, it is beyond the scope of this study to provide conclusive evidence for Blake's point that other conjunctions might have been used instead of and for particular functions. To do this, one would need to look at all conjunctions. Clearly, there is further empirical work to be done here. We suspect that the developments in punctuation played a key role in the shrinkage of clause-level and in written registers. Consider that in spoken English the concept of the sentence is of little use in comprehension: there are no graphological features — capitals and fullstops — to signal the beginnings and ends of sentences. Instead, to identify units of speech, we rely on:

The conjunction and in early Modern English

• • •

311

grammatical features (notably, clause boundaries) prosodic features (notably, tone units) lexical features (notably, conjunctions)

And, as we have already shown, is the linker par excellence in speech. In the history of written texts, as Lennard (1995: 67) puts it aptly, with reference to an early pre-publication draft of Robinson (1998), "a conceptual change from the 'period', defined aurally and rhetorically, to the 'sentence', defined visually and syntactically, should be dated to the mid-to-late seventeenth century". This view is also supported by Treip (1970: 49-50). Consequently, until the syntactic basis of the sentence was established, the concept of the sentence was not particularly helpful in comprehension, owing to the fact that there was no broad consensus about what constituted a sentence and that typically the sentence constituted a substantial stretch of text. Thus, one relies on: • • •

grammatical features (notably, clause boundaries) a variety of punctuation marks lexical features (notably, conjunctions)

Our research has suggested that and is the most frequent conjunction in the historical written texts we examined. Our line of reasoning here might generate the following hypothesis: "In terms of the frequency and function of clause-level and, historical texts used to have a closer resemblance to modern speech than modern writing, before developments around the midto-late seventeenth century". The implication of our hypothesis is that as far as the use of and is concerned, there has been a move away from an oral style to a visual or literate style, due to the evolving concept of the sentence and the role played by and in relation to it. This is thus a development in the opposite direction to the general pattern of drift towards more oral styles described by Biber and Finegan (1989, 1992). Specifically, while they document development towards greater use of features associated with colloquial language, the development shown here for and points to increasing use of patterns associated with modern-day written registers. This hypothesis has been supported in some preliminary work we have undertaken on words collocating with and. We selected five words that collocate immediately to the right of and: then, therefore, when, thus, and

312

Jonathan Culpeper and Merja Kytö

because. We calculated what proportion of these words collocate with and (e.g. how many of the total number of instances of then follow and). Our assumption was that and was less likely to occur in these collocations in written registers after the mid-to-late seventeenth century, because of: •

•

redundancy (a full-stop could signal the separation of clauses, as could the following adverbial or subordinator, which could also specify a particular sense relationship) the prescriptive grammatical rule prohibiting the use of sentence-initial and

Furthermore, there is some empirical evidence that and then is more likely in speech than writing. Regarding and then, Beaman (1984: 77) gives a frequency figure of 8.8 (per 1000 words) for her spoken data and 0.9 for her written; against 2.1 (per 1000 words) for then in her spoken data and 5.8 for her written. The data we searched were: (i) the Science and History sections from the early Modern English section of the Helsinki Corpus (70,020 words from the period 1500-1710), (ii) the Dialogues Corpus (340,915 words from trials, depositions, drama, and prose from the period 1600-1720), (iii) the written section of the British National Corpus

Sampler

(1,109,718 words from the late 1980s and early 1990s), and (iv) the spoken section of the British National Corpus Sampler (1,874,580 words from the late 1980s and early 1990s). Table 3 presents the cumulative totals for all five words. 9 Table 3. Words collocating with and. (The first figure represents the number of instances that follow and. The figure in brackets represents the total number of instances. The percentage represents the number of instances that follow and expressed as a percentage of the total number of instances of that word.) Science and Dialogues Written section Spoken section History from Corpus of the British of the British the Helsinki National National Corpus Corpus Sampler Corpus Sampler 102 (608) 440 (2734) 454 (4784) Total for 5 1456 (8824) words 16.8% 16.1% 9.5% 16.5%

What table 3 shows is a strong similarity between the historical texts — regardless of whether they are speech-related or not — and modern

The conjunction and in early Modem English

313

speech. Modern written texts stand apart with a lower proportion of instances preceded by and. Some may object that a proportion of data from the Helsinki Corpus and the Dialogues Corpus overlaps with the crucial transitional period for punctuation we mentioned above — the mid-to-late seventeenth century. In fact, more detailed analysis reveals further support for our hypothesis. For Science and History from the final period of the Helsinki Corpus early Modern English section (i.e. 1640-1710), the percentage drops to 12.9 per cent; in other words, it begins to move towards the figure for modern written texts. Of course, whilst our findings are consistent with our points about punctuation, they may not be fully explained by it. For this, more research is needed into the semantic relationship between and and the particular collocating word, and also the wider context.10

3.3. Uses of and We now move on to discuss some specific examples. Our emphasis will be on and in "our speech-related text types. We have organised our examples into three groups. One concerns clause coordination and another phrase coordination. To these we have added a third group, which we call exchange coordination and which concerns the coordination of adjacency pairs, such as questions and answers, across turns. This group is designed to capture examples which are not adequately described by either a clauselevel or phrase-level analysis, but which are captured by an analysis at the level of the exchange. Instances in this group overlap with the other types in our classification (and we do not aim at quantifying all possible instances). We start with this group.

3.3.1. Exchange coordination Our understanding of exchange coordination is partly inspired by Sweetser (1990: chapter 4). Sweetser (1990: 89) repeatedly points out that "[a]nd conjoins speech acts as well as content items or logical premises". She states that "precisely when conjunction of clauses is intended to conjoin speech acts, the conjoined clauses behave as syntactically independent.

314

Jonathan Culpeper and Merja Kytö

This is because they are linked not at the syntactic but at the pragmatic level" (Sweetser 1990: 156, note 7). Four of her examples are given in (11). (11) a. Glad to meet you, sir; and what makes you think I can be of assistance to your work? b. Thank you, Mr Lloyd, and please just close the door as you go out. c. Darling, you 're wonderful, and how about dinner at Chez Panisse tonight? d. The Vietnam War WAS morally wrong, and I'll gladly discuss the reasons why I think so. (Sweetser 1990: 89) It is clearly the case in our more interactive speech-related data that speech acts are frequently conjoined across exchanges (see, for example, Sinclair and Coulthard 1975: 21-53 for the notion of the exchange). In other words, the interpretation of the conjunction can only be fully understood in relation to the exchange of which it is a part. This, of course, brings us into contact with some of the discoursal uses of and discussed in Schiffrin ([1987] 1994: 128-152). In example (12) the function of and is to coordinate the first parts of adjacency pairs in the exchange. Specifically, it is used to create a series of questions or, as Schiffrin (1994: 146) puts it, to "link questions in a question agenda". (12)

( A L. C. J.A) ( A Mary Tilden.A) (AL. C. J A) ( A Mary Tilden A) ( A L. C. J A) (AMary Tilden.A)

Can you speak positively as to this night, the Saturday that he was kill 'd? He was at home that night. And where was he the Sunday? He was at home. And you are sure he was at home every night? Yes, while we were in Town. (DC/Trial/Green, Berry and Hill 53)

In the next four examples and is used in the second part of the questionanswer adjacency pair. And is used to signal that the answer is to be extended; it gives thinking time to the speaker. Example (13) also

The conjunction and in early Modern English

315

illustrates the sometimes quite fragmented nature of impromptu speech. For example, the use of I am sure after the elliptical clause William Drake is what Leech (1998) in his work on the grammar of speech would call a "retrospective tag" (see also Biber et al. 1999: 1080-1082). (13)

L. (APres.A) (AIaquel.A)

... I ask you then before the Magistrate here, who was present at that meeting? (AWilliam Drake,A) I am sure, and Major (AAlfordA) was there, and Captain (APotter,A) and severalI others, as I remember, Master (AIenkins.A) (DC/Trial/Love 41)

In examples (14a) to (14c), and is strategically used to go beyond what is required to answer the question (cf. Grice 1975). In (14a), the prisoner successfully feeds extra information which the judge then takes up in his attempts to pin down the truth about Mrs. Rawlins' actions. (14) a. (AL. C. J A) (APrisoner.A) ( A L. C. J.A) (APrisoner,A) ( A L. C. J.A)

Did she kiss you? Yes, my Lord, and squeez'd me often. Did you not think her very coming? Yes, I did, and when we talk 'd of Marriage, she seemed to be very well pleased. Mrs. (ARawlins,A) you hear what he says. Did you Squeeze him by the Hand, and Kiss him? Is it true? (DC/Trial/Swendsen/10)

In (14b) we have a similar instance drawn from a drama text. (14) b. ( A Aim. A ) (AGib.A)

Pray, Sir, han't I seen your Face at (AWill'sA) Coffee-house? Yes, Sir, and at (AWhite 'sA) too. (DC/Drama/Farquhar 27)

In example (14c) we have an instance where the straightforward answer yes is ellipted, but the extension is there.

316

Jonathan Culpeper and Merja Kytö

(14) c. (AArch.A) (ACher.A)

Scruple! no, no, but — two thousand Pound you say? And better. (DC/Drama/Farquhar 20)

In examples (15) and (16) we now turn to adjacency pairs other than the question/answer type. In example (16) there is an instance of what might be characterised as a "comment/comment" pair. Here and links the two comments and makes possible the ellipsis in the second part. (15)

(ACynt.A) (AL. Froth.A)

He does not indeed affect either pertness, or formality; for which I like him: Here he comes. And my Lord with him: pray observe the difference. (DC/Drama/Congreve 15)

Finally in our discussion of Exchanges, the passage given in example (16) is a kind of parody of the marriage vow. What we have here looks like an instance of phrasal coordination split between two speakers, Harriet ("Har.") and Young Bellair ("Y. Bell."). The latter seems to use and as a means of interrupting and dominating the conversation. (AHar.A) (AY. Bell.A) (AHar.A) (AY. Bell A) (AHar.A) (AY. Bell.A) ( A Both A ) (AHar.A) (AY. Bell A) (AHar.A)

With all my heart, I never thought I should have given you mine so willingly. Here I (AHarrietA) — And I (AHarryA) — Do solemnly protest — And vow — That I with you — And I with you — Will never marry — A match! And no match! How do you like this indifference now? You expect I should take it ill, I see! (DC/Drama/Etherege 221)

The conjunction and in early Modern English

317

3.3.2. Clause coordination We now move to take a look at instances of clausal and, and specifically its uses in speech narrative. Mrs. Bracegirdle uses and to link the elements of her story. (17)

Mrs. (ABracegirdle.A) My Lord, I was in (APrince's Street) at Supper at Mr. (APageAj's, and at ten a Clock at Night, Mr. (APageA) went Home with me; and coming down (ADrury-Lane,A) there stood a Coach by my Lord (ACravenA) 's Door, and the Boot of the Coach was down, and a great many men stood by it; and just as I came to the Place where the Coach stood, two Soldiers came and pulled me from (AMr. Page,A) and four or five more came up to them, and they knocked my Mother down almost, for my Mother and my Brother were with me. My Mother recovered, and came and hung about my Neck, so that they could not get me into the Coach, and (AMr. PageA) went to call Company to rescue me. Then (AMr. HillA) came with his Sword drawn, and struck at (AMr. PageA) and my Mother; and when they could not get me into the Coach, because Company came in, he said he would see me Home, and he led me by one Hand, and my Mother by the other. And when we came Home, he pulled (AMr. PageA) by the Sleeve, and said, Sir, I would speak with you. (DC/Trial/Mohun 16)

Interestingly, an and precedes every clause except when Bracegirdle wishes to mark a particularly significant event. In this passage she does this twice, first when pointing out that her mother recovered, and then when reporting on the dramatic arrival of Mr. Hill on the scene. The fact that narrative is fundamental also to history writing can been seen in example (18). Here we can see the same type of linking of events and states making up a narrative.

318

(18)

Jonathan Culpeper and Merja Kytö

His shoes were clean, his money was in his pocket: but nothing was about his neck, and a mark was all round it, an inch broad, which shewed he was strangled. His breast was likewise all over marked with bruises, and his neck was broken. All this I saw; for Lloyd and I went to view his body. There were many drops of white waxlights on his breeches; which he never used himself; and since only persons of quality or priests use those lights, this made all people conclude in whose hands he must have been. And it was visible he was first strangled, and then carried to that place, where his sword was run into his dead body. For a while it was given out that he was a hypochondriacal man, and had killed himself. Of this the king was possessed, till Lloyd went and told him what he had seen. The body lay two days exposed, many going to see it, who went away much moved with the sight. And indeed men's spirits were so sharpened upon it, that we all looked on it as a very great happiness that the people did not vent their fury upon the papists about the town. (HC/History/Burnet 1,11,164-165)

Furthermore, the final instance in example (18) (And indeed men's spirits were so sharpened upon it ...) illustrates what Halliday and Hasan ([1976] 1990: 240-244, 321) would describe as a shift from internal to external and, that is to say, a shift from the narrative of the story to the author's comment on it. In example (19), and signals a shift from dialogue to narrative in the form of a scribal comment. (19)

(ALord Great Chamberlain.A)

The Witnesses must be brought to this place. And accordingly, way was made for the Witnesses, as called to stand at the Bar between the Prisoner and the Kings Councel. (DC/Trial/Mohun 12)

Finally, in the clausal coordination group, a brief comment on the use of and in science. We were struck by the frequency with which and was used to indicate the steps in the argument, in particular, that what follows and is a conclusion based on what precedes.

The conjunction and in early Modern English (20)

319

It does not seem to have any eye-lids, and therefore perhaps its eyes were so placed, that it might the better cleanse them with its fore-legs; and perhaps this may be the reason, why they so much avoid and run from the light behind them, for being made to live in the shady and dark recesses of the hair, and thence probably their eye having a great aperture, the open and clear light, especially that of the Sun, must needs very much offend them; (HC/Science/Hooke 13.5,211)

3.3.3. Phrase coordination Moving on to phrase coordination, and does occur in simple lists — packages of information — as in example (21), drawn from a science text. (21)

Having made this Distillation, I say, and continued it till it had afforded a good proportion of phlegm, Spirit, Volatile Salt, and Oyl, the Retort was warily broken, and the remaining matter was taken out in a lump, ... (HC/Science/Boyle 24, 21)

These packages of information do not contain random information but items that belong to a particular domain. The contents of the list in (21) are all components of the distillation. In fact, in our examples phrasal coordination seems to be more about coordinating common semantic elements. For example, the first instance of and in example (22a) coordinates the two professions. (22) a. He was a gardener and a chemist, and was full of projects and notions. He had got some credit in Cromwell's time, and that kept him poor. He was a very mean divine, and seemed credulous and simple, but I had always looked on him as a sincere man. (HC/History/Burnet 1,11,156) The instances of and in (22a) and also in (22b) and (22c) illustrate the fact that far more frequent than lists are coordinated word-pairs. Moreover, these word-pairs very often consist of synonymous or near-synonymous expressions.

320

Jonathan Culpeper and Merja Kytö

(22) b. ... and perhaps this may be the reason, why they so much avoid and run from the light behind them, for being made to live in the shady and dark recesses of the hair, and thence probably their eye having a great aperture, the open and clear light, especially that of the Sun, must needs very much offend them; ... (=[20] HC/Science/Hooke 13.5,211) Clearly, this kind of rhetorical device in not a feature of today's history or science writing. Interestingly, the density of phrasal coordination gives away the fact that the passage in the final example (22c), drawn from a trial proceeding, has to be to a great extent planned discourse, not the impromptu speech of the courtroom. As we said, phrasal coordination is typical of non-speechrelated texts, which have a greater degree of planning. (22) c. (AL.C.J.A)

I hope, Gentlemen of the Jury, you take notice of the strange and horrible Carriage of this Fellow; and withal, you cannot but observe the Spirit of that sort of People, what a villanous and devilish one it is: Good God! that ever the thing called Religion (a Word that People have so much abused) should ever wind up Persons to much a height of Impiety, that it should make them lose the Belief that there is a God of Truth in Heaven; that sees and knows, observes and registers, and will punish and take vengeance of Falshood and Perjury. It may well make the rest of Mankind, that have any sort of Faith in a Deity and a future Life, to abhor and detest both the Men and their Religion, if such abominable Principles may be called so. (DC/Trial/Lisle IV,117.C2)

4. Conclusion Our work on and is a reminder that speech and writing are in no way uniform varieties of language. Speech and writing are, as Biber's (1988)

The conjunction and in early Modern English

321

work has already shown, multi-dimensional. Furthermore, our work is a reminder that and is not simply a marker of speech. In brief, the main points of this paper are: •

•

•

Like the work done on present-day English, we have shown that it is important to distinguish between and as a clause-level coordinator and and as a phrase-level coordinator in early English. Clause-level and is relatively frequent in our speech-related text types compared with phrase-level and. Our figures compared with those of the Longman Grammar and Chafe and Danielewicz (1987) suggest that over time there may have been a general shrinkage in the incidence of and in written texts. We argued that the development of the sentence would, at least in part, explain this finding as far as clause-level and is concerned. Our preliminary work on collocations supports the hypothesis that, in terms of frequency and function, clause-level and in texts before approximately the mid-to-late seventeenth century has a closer resemblance to modern speech than modern writing. Our examples show the multifunctionality of and. Particular usages are associated with certain text types, and, moreover, with certain dimensions, such as narrative, which cut across both speech-related and non-speech-related text types. Furthermore, we have demonstrated how for dialogic texts an analysis must consider the operation of and at the level of the exchange.

Notes 1. The overall design and the principles of compilation of the corpus are described in Culpeper and Kytö (1997). 2. We elaborate on the problems related to the study of "spoken texts" of the past in a recent paper of ours, where we examine four seventeenth-century speechrelated text types (witness depositions, trial proceedings, prose fiction and drama) for a number of features assumed to be characteristic of spoken face-toface interaction (see Culpeper and Kytö forthcoming). A discussion of the findings presented in research carried out so far on differences in present-day spoken and written language can be found in Miller and Weinert (1998).

322

Jonathan Culpeper and Merja Kytö

3. For some well known studies, see, e.g., Chafe (1982: 38-39, 1985: 111), Chafe and Danielewicz (1987: 101-105), Schiffrin (1987: chapter 6), Biber (1988: 106, 245) and Biber et al. (1999: 79-84). 4. In our examples all instances of and are in plain type. Those relevant to the point we are making are also in bold. The corpus coding (Λ...Λ) stands for italics in the original. 5. We are grateful to the authors of the Longman Grammar for kindly allowing us to consult the sections on and prior to their publication. 6. Examples drawn from the Dialogues Corpus. 7. It would, of course, have been possible to give up distinguishing verb phrases from the rest of the data altogether and opt for a clausal interpretation even with these instances of functionally and semantically analogous verb-words. However, this strategy was deemed to be a less fruitful starting point, given that prototypical clauses do differ from these instances to a great extent and that the distribution figures might have been skewed as a consequence of including only non-verb constructions in the category of phrasal coordination. 8. For discussion, see, e.g., Brown and Gilman (1989); Kopytko (1995: 516); Salmon (1987: 265); Biber and Finegan (1989; 1992: 693); Taavitsainen (1995: 460). 9. We are aware that frequency is not a sophisticated measure in the identification of collocations. (In fact, it may be more accurate to dispense with the term "collocation" altogether.) Moreover, the reason that we refrain from presenting figures for individual collocations in the body of the paper is that further coding of our concordance examples is necessary. For example, we need to identify when when is a conjunction and when it is an adverb, and we need to identify when thus is clause final and when it is not. This may slightly alter the picture for individual figures, though we do not think that it will change the overall picture for the cumulative totals, which is why we present these in the body of the paper. Nevertheless, readers may be interested in our current results, and so we present them in table I below, whilst urging that they be treated with caution. 10. Our work does not preclude the possibility that the shrinkage of clause-level and started earlier than the key period we suggest, the mid-to-late seventeenth century. Looking at the frequencies — raw and unscreened — of the instances of and in the Helsinki Corpus, one can detect a steady decline from the Middle English period through to the early Modern English period. However, such a decline may be due or may be partly due to changes in the frequency of phraselevel and. Clearly, we need to extend our work on identifying clause-level and phrase-level and to cover a much broader period, in order to establish exactly what is happening.

The conjunction and in early Modern English

323

Table I. Words collocating with and. (The first figure represents the number of instances that follow and. The figure in brackets represents the total number of instances. The percentage represents the number of instances that follow and expressed as a percentage of the total number of instances of that word.) Science and Written section Dialogues Spoken section History from Corpus of the British of the British National the Helsinki National Corpus Corpus Sampler Corpus Sampler 194 (1234) 285 (1599) 1219(3614) Then 43(310) 13.9% 15.7% 17.8% 33.7% Therefore 27(61) 123 (300) 47 (226) 75 (173) 44.3% 41.0% 20.8% 43.4% 14(138) 94 (812) 65 (1924) 142 (2957) When 3.4% 10.1% 11.6% 4.8% 9(40) 15 (179) 44 (310) 1(3) Thus 8.4% 14.2% 22.5% 33.3% 9(59) 14 (209) 13 (725) 19(2077) Because 1.8% 15.3% 6.7% 0.9% 102 (608) 440 (2734) 454 (4784) 1456 (8824) Total 9.5% 16.8% 16.1% 16.5%

References Beaman, Karen 1984

Biber, Douglas 1988

Coordination and subordination revisited: Complexity in spoken and written narrative discourse. In: Deborah Tannen (ed.), Coherence in Spoken and Written Discourse, 45-80. Norwood, New Jersey: Ablex Publishing Corporation.

Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas and Edward Finegan 1989 Drift and the evolution of English style: A history of three genres. Language 65: 487-517. 1992 The linguistic evolution of five written and speech-based English genres from the 17th to the 20th centuries. In: Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.), History of Englishes: New Methods and Interpretations in Historical Linguistics, 688-704. Berlin/New York: Mouton de Gruyter.

324

Jonathan Culpeper and Merja Kytö

Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 Longman Grammar of Spoken and Written English. Harlow: Longman. Blake, Norman F. 1996 A History of the English Language. London: Macmillan. British National Corpus Sampler 1999 The BNC Sampler. Oxford: Oxford University Computing Services. Brown, Roger and Albert Gilman 1989 Politeness theory and Shakespeare's four major tragedies. Language in Society 18: 159-212. Carston, Robyn 1994 Conjunction and pragmatic effects. In: R.E. Asher (ed.), The Encyclopedia of Language and Linguistics, 692-698. Oxford: Pergamon. Chafe, Wallace 1984 Integration and involvement in speaking, writing, and oral literature. In: Deborah Tannen (ed.), Spoken and Written Language: Exploring Orality and Literacy, 35-53. Norwood, N.J.: Ablex. First edition [1982]. 1985 Linguistic differences produced by differences between speaking and writing. In: David R. Olson, Nancy Torrance and Angela Hildyard (eds.), Literature, Language, and Learning: The Nature and Consequences of Reading and Writing, 105-123. Cambridge: Cambridge University Press. Chafe, Wallace and Jane Danielewicz 1987 Properties of spoken and written language. In: Rosalind Horowitz and S. Jay Samuels (eds.), Comprehending Oral and Written Language, 83-113. San Diego, New York, etc.: Academic Press. Culpeper, Jonathan and Merja Kytö 1997 Towards a corpus of dialogues, 1550-1750. In: Heinrich Ramisch and Kenneth Wynne (eds.), Language in Time and Space. Studies in Honour of Wolfgang Viereck on the Occasion of his Sixtieth Birthday, 60-73. (Zeitschrift für Dialektologie und Linguistik, Beiheft 97.) Stuttgart: Franz Steiner Verlag. forthcoming Data in historical pragmatics: Spoken interaction (re)cast as writing. Journal of Historical Pragmatics. Dialogues Corpus In prep. A Corpus of English Dialogues 1560-1760. Compilers: Jonathan Culpeper (Lancaster University) and Merja Kytö (Uppsala University).

The conjunction and in early Modern English Grice, H. Paul 1975

325

Logic and conversation. In: Peter Cole and Jerry L. Morgan (eds.), Syntax and Semantics 3: Speech Acts, 41-58. New York: Academic Press. Halliday, Michael A.K. and Ruqaiya Hasan 1990 Cohesion in English. London and New York: Longman. First edition [1976], Helsinki Corpus 1991 The Helsinki Corpus of English Texts. Helsinki: Department of English, University of Helsinki. Kopytko, Roman 1995 Linguistic politeness strategies in Shakespeare's plays. In: Andreas H. Jucker (ed.), Historical Pragmatics: Pragmatic Developments in the History of English, 515-540. Amsterdam/Philadelphia: John Benjamins. Leech, Geoffrey 1998 Grammar and spoken English: Examples of conversation from the British National Corpus. A lecture given at Lancaster University. Lennard, John 1995 Punctuation: And — 'Pragmatics'. In: Andreas H. Jucker (ed.), Historical Pragmatic: Pragmatic Developments in the History of English, 65-98. Amsterdam/Philadelphia: John Benjamins. Miller, Jim and Regina Weinert 1998 Spontaneous Spoken Language: Syntax and Discourse. Oxford: Clarendon Press. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Robinson, Ian 1998 The Establishment of Modern English Prose in the Reformation and the Enlightenment. Cambridge: Cambridge University Press. Salmon, Vivian 1987 Sentence structures in colloquial Shakespearean English. In: Vivian Salmon and Edwina Burness (eds.), A Reader in the Language of Shakespearean Drama, 265-300. (Benjamins Paperbacks 7.) Amsterdam: John Benjamins. Schiffrin, Deborah 1994 Discourse Markers. Cambridge: Cambridge University Press. First edition [1987], Sinclair, John McH. and R. Malcolm Coulthard 1975 Towards an Analysis of Discourse: The English Used by Teachers and Pupils. London: Oxford University Press.

326

Jonathan Culpeper and Merja Kytö

Sweetser, Eve Ε. 1990 From Etymology to Pragmatics: Methaphorical and Cultural Aspects of Semantic Structure. (Cambridge Studies in Linguistics 54.) Cambridge: Cambridge University Press. Taavitsainen, Irma 1995 Interjections in early Modern English: From imitation of spoken to conventions of written language. In: Andreas H. Jucker (ed.), Historical Pragmatics: Pragmatic Developments in the History of English, 439-465. Amsterdam/Philadelphia: John Benjamins. Traugott, Closs Elizabeth 1986 On the origins of 'and' and 'but' connectives in English. Studies in Language 10: 137-50. Treip, Mindele 1970 Milton's Punctuation and Changing English Usage 1582-1676. London: Methuen.

3. Sociolinguistics and dialectology

Processes of supralocalisation and the rise of Standard English in the early Modern period Terttu Nevalainen

1. Introduction The purpose of this paper is to advocate a localised and sociolinguistically driven framework for describing the diffusion of Standard English between the late fifteenth and the mid-seventeenth centuries. I shall suggest that the concept of supralocalisation used by sociolinguists and dialectologists describes this era before normative grammar more adequately than the notions of overt or deliberate standardisation. If the level of abstraction is lowered, and the language of individual speaker-writers from different parts of the country is examined in detail, it is possible to arrive at a realistic picture of how supralocal usages were diffused in early Modern England. The approach shows that what has been called a "single-minded march towards Standard English" (Lass 1976: xi) was in fact divided into a variety of processes of supralocalisation which involved linguistic features of diverse regional and social origins. This investigation was sparked off by the divergence/convergence model of language change suggested by James Milroy (1992: 51). Milroy blames traditional accounts of the history of English for their bias towards the standard language instead of offering a balanced account of the natural divergence that is characteristic of linguistic evolution. What amounts to a counter-argument to Milroy's views can be found in many empirical studies on late Middle and early Modern English. They show that the degree of (spelling) standardisation in extant materials is so high as to make most of them impossible to localise. Both viewpoints are discussed in section 2. My own mediating position is that while it may not be possible to localise most individual texts in the later periods, it is possible to follow the processes of change that take place in particular localities and regions at a given time. This mainly applies to developments other than spelling. My argument is based on the notion that, viewed as supralocalisation, language standardisation becomes regionally observable: it is possible to investigate

330

Terttu Nevalainen

where the components of national norms originated and at what rate they were diffused. This processual view of standardisation can, for instance, be tested against contemporary early modern writers such as Puttenham (1589 [1968]), who maintained that linguistic usages worth following nationwide were found in the London area, particularly in the Royal Court. In order to provide a sociohistorical context for the empirical part of my study, I shall present in section 3 an overview of the research on demographic developments and the processes of urbanisation in early Modern England with a particular focus on the role of London. This information will be drawn upon in section 4, where my research questions are outlined. They include the following: first, is the direction of linguistic innovations invariably from London to the rest of the country? Second, does the capital region accept features from other varieties? Finally, does the language of literate Londoners differ from the linguistic usage of the Court? In section 5, the empirical part of my study, I shall investigate the diffusion of four morphosyntactic features in late Middle and early Modern English: • • • •

generalisation of the present indicative form are of the verb BE generalisation of the suffix -(e)s in the third person singular present indicative generalisation of the oblique form you in the subject function disappearance of multiple negation and generalisation of non-assertive indefinites.

The material used in this study comes from the 2.7 million word computerreadable Corpus of Early English Correspondence (CEEC), compiled by the Sociolinguistics and Language History team at Helsinki University in 1993-1998. 1 My results show that what we at a higher level of abstraction call the rise of standard language consists of sets of individual processes moulded by a number of social and demographic factors.

2.

Modelling the rise of Standard English

2.1. Divergence and convergence In recent years, English historical linguists have often been taken to task for their predominant concern with linguistic developments that either lead

Supralocalisation and the rise of Standard English

331

to, or pave the way for, Standard English. In his (1992) book, Language Variation and Change, James Milroy critically contrasts what he calls the divergence model of the Indo-European family tree approach and the convergence model found in language histories for recent centuries. He argues that, while Indo-European scholarship has been able to show that the long-term development in related languages tends towards greater divergence, histories of English would suggest just the opposite. The blame does not fall on Middle English scholarship as heavily as on later periods, but signs of the ideology of convergence are already detected: At the Middle English stage, the description of divergence is still very salient (partly because the states attested in writing are unquestionably divergent states), but we also begin to notice attempts to launder the data retrospectively in such a way as to focus on those features that lead to modern 'standard' English and to ignore, reject and explain away those features that deviate from it. (Milroy 1992: 50-51) Milroy goes on to argue that research on the following period, early Modern English, decisively narrows the base of the funnel, and proceeds, from about 1550 on, to tell only the story of Standard English. This divergence/convergence model is shown in figure 1.

it ft 1550

Present English

=>

Convergence

ft

Divergence

ft

U li Middle English JJ Old English •U Germanic Indo-European

Figure 1. Divergence/convergence model (after Milroy 1992: 51)

332

Terttu Nevalainen

Similar complaints have been expressed by a number of historical linguists in the 1980s and 90s (e.g. Crowley 1989, Hogg 1998, and Leith 1983: 153-212). Traditional scholarship, notably the work of Wyld ([1920] 1936) and Dobson [1957] 1968), is charged with historicising Standard English in general and Received Pronunciation in particular. It is argued that their model has put blinkers on later research, which has assumed only the canonical forms of the standard language as a proper object of study. Leaving their possible ideological pursuits aside, which have been amply discussed in the sources referred to, due attention should, however, also be paid to problems faced by late Middle and early Modern English scholars. Not surprisingly, these practising language historians, too, have their complaints to make, and they to a large extent centre on what might be called the bad data problem. By the "bad data" problem, I refer to the label given by the sociolinguist William Labov (1994: 11) to one of the most obvious obstacles to historical sociolinguistics, the lack of access to the spoken language of the past. In principle, this point has been reiterated in the literature on early Modern English from Wyld (1936) and Jespersen ([1909] 1949) to Rissanen (1986) and Görlach (1990). A well known formulation is given in the introduction to A Linguistic Atlas of Late Mediaeval English (LALME): In the course of the fifteenth century ... regional diversity gives way increasingly to Chancery Standard, the official language of the London administrators and the direct ancestor of modern Standard English. By the end of the same century, moreover, the establishment of printing was instrumental in the redevelopment of a national literary standard. The dialects of the spoken language did not die out, but those of the written language did; and although there are some late survivals, they are no sufficient basis for a dialect atlas. (Mcintosh, Samuels and Benskin 1986: 3)

We might, however, ask whether the thin end of the funnel could be broadened by looking at linguistic developments other than spelling and the way in which pronunciation is reflected in spelling. Spelling belongs to the areas in language that readily lend themselves to standardisation. More variation seems to be tolerated in many areas of morphology and syntax. The social and regional profiles of these domains have largely remained unexplored in early Modern English, and in syntax even in Middle English

Supralocalisation and the rise of Standard English

333

(Blake 1992: 14-15). In other words, although our access to phonological variation in the post-Middle English period may be limited, there are other areas of language that are clearly more open-ended than the orthographic and phonological systems. One of the empirical challenges of future early Modern English scholarship is to explore the extent to which they, too, were regularised in the data sources that have come down to us. Finally, although it may sometimes be useful to view a historical period of a language as a product rather than as a process, this textbook approach is not justified if one wishes to trace linguistic variation. As pointed out by Milroy (1992: 127-131), the idea of a fixed early Modern variety is partly created at this high level of abstraction. One might add that a more realistic picture will certainly emerge if early Modern English is approached either in terms of subperiods shorter than the two hundred years from 1500 to 1700, or in real time without fixed time limits.2

2.2. A processual view of standardisation The product versus process perspective can also be extended to the study of standardisation. 3 There is a relative consensus among researchers that standardisation, too, may be construed as a process. Haugen ([1966] 1997) makes a basic distinction between form and function in the standardisation process. Simplifying matters, one can say that the formal side involves (i) the selection of a particular language variety to become the standard and (ii) its subsequent codification. The functional side of the process consists of the acceptance by the community, and the subsequent elaboration, of the chosen variety. According to this model, selection and acceptance are social issues, while codification and elaboration are linguistic (see figure 2). I shall next focus on the two social issues: selection and acceptance. We have grounds for arguing that there was one particular variety that was selected to serve as a national standard in the fifteenth century, viz. Chancery English. At the end of the century, the printing press became another standardising agency exercising its power to promote a language variety for public use. In both cases the variety has been identified as that of the South-East Midlands, in particular, that of London (see e.g. Samuels 1963, Fisher, Richardson and Fisher 1984, Fisher 1996). Despite its acceptance as a printed medium, this fifteenth-century variety did not, however, have a

334

Terttu Nevalainen

linguistically stable existence outside the language community that promoted it. Even competing norms were developing in certain areas of language use, such as early scientific writing, which reflected the Central Midland variety (see Taavitsainen forthcoming). Neither had the London-based variety reached internal consistency either in spelling or in grammar; for this reason Smith (1996: 70) prefers to call it a "focused" usage rather than a "fixed" standard.

SOCIETY

FORM Selection

FUNCTION Acceptance

LANGUAGE

Codification

Elaboration

Figure 2. Haugen's (1997: 351) model of language standardisation. The language of the London area was also explicitly accepted as a model for literary composition in the sixteenth century. In his Arte of English Poesie, George Puttenham (1589: 120) warned his readership, aspiring poets, against the ill effects of using most regional and social varieties of English: undesirable language mixing took place in port towns, at frontiers and in the two universities, and dialectal usage was associated with rural areas, including the north. The language of the lower ranks, "the inferior sort", was similarly to be avoided, regardless of region, as uneducated. For the aspiring poet, Puttenham recommended, not the language of official documents nor the printed books of the time, but the language of the "better-brought-up sort" of London, and of the Royal Court in particular. He was not alone in promoting the language of London. Similar comments can be found, for instance, in the writings of John Hart, the London orthoepist, some 20 years earlier (Dobson 1969: 64). The process of functional elaboration of the selected variety was then clearly in progress. However, as Haugen (1997: 343) observes, when Puttenham talks about "London speech" he is not referring to a national standard but to a regional variety. Puttenham's (1589: 120) definition of "language" includes nationwide acceptance of the variety when he writes: "after a speach is fully fashioned to the common vnderstanding & accepted by consent of a whole country & nation, it is called a language". In order to see how this process took place, the terms selection and acceptance will be applied to individual linguistic processes of supralocalisation in section 3.3.

Supralocalisation and the rise of Standard English

3.

335

Agencies of supralocalisation

It is true that we may see a convergence in the history of English at the juncture of late Middle and early Modern English. In fact, this would be predicted by cultural historians, who associate language standardisation with other phenomena compounded under "modernity", including urban as opposed to rural residence, geographical mobility, and contact with mass media (see e.g. Barton and Hamilton 1996). In all these respects, London emerges as the centre of activity in early Modern England.

3.1. Print culture and literacy As pointed out above, the printing press was one of the agencies in the selection of the language variety to be employed in mass communication. Eisenstein (1983: 50-90) stresses the role of print in cultural diffusion and written language standardisation. The availability of multiple copies of the same text makes them accessible to a number of people simultaneously, spreading certain messages and conventions, and ignoring or suppressing others. The significance of print is, however, inextricably connected with what Eisenstein (1983: 90) calls the "social penetration of literacy". The number of English people who could both read and write was not very large around 1500. Cressy (1980: 175-177) estimates that it was roughly ten per cent of the male population at the time, when the whole population was about two million. Full literacy was, however, much higher in London than elsewhere. By 1640 it had reached an estimated average level of 30 per cent of the male population in the entire country, and some 60 per cent of the male population of London. Cressy (1980: 129) notes that even servants and apprentices in London were extraordinarily literate. The major social dividing line between the gentry and the non-gentry also surfaces in literacy figures: while the overall literacy of women in the seventeenth century lagged considerably behind that of men, we may assume that the gentry, women and men, could by this time both read and write (Cressy 1980: 141-174, Heal and Holmes 1994: 252-254). Book production similarly grew at a fast rate: the number of titles published in 1500 was only 54; in 50 years it as much as quadrupled, amounting to 214 in 1550. For 1640, the corresponding figure in the Short

336

Terttu Nevalainen

Title Catalogue was 577. London was the capital of the book trade: 98 per cent of the books published in England in the early Modern period were printed in London (Bennett 1969, Görlach 1991: 6-7, 13). These figures bear on Samuels's (1963) requirement for the quantity written in written language standardisation, and suggest the growing importance of London.

3.2. Urbanisation and mobility Testimonies like Puttenham's suggest that it is the process of urbanisation in the sixteenth century that we should turn to in order to understand how the "usual speech" of London might have come about. Even a cursory look at the demographic evidence available will show that early Modern England was predominantly a rural society. According to Clay (1984: 165-166), no more than 10-12 per cent of the English population lived in towns around 1500. There were, however, significant population movements in the fifteenth and sixteenth centuries. People moved both from densely populated farming regions to undeveloped land in the forests and the fens, and from the countryside to London and other urban centres (see Stone 1966, Corfield 1990). The importance of migration to the capital cannot be underestimated at a time when the population of London quintupled from about 100,000 in 1550 to 500,000 in 1700, accounting for over ten per cent of the population of England around 1700 (Finlay and Shearer 1986: 42-51). The role of migration to London becomes even more important if we take account of the fact that London's death rate often exceeded its birth rate because of epidemic and endemic diseases. Rappaport (1989: 77), for instance, estimates that five out of every six men who became citizens and companymen in the early 1550s had emigrated to the capital. On the basis of research on court records, Coleman and Salt (1992: 27) conclude that a mere 15 per cent of Londoners had been born in the capital in the sixteenth and seventeenth centuries, and that more than two thirds came from at least 50 miles away. Moreover, London was a highly cosmopolitan city with a sizeable number of immigrants from the continent (Yungblut 1996). The apprenticeship, citizenship and court records that are used in demographic studies suggest that a very high proportion of the late fifteenthand sixteenth-century immigrants to London came from the northern counties, but that there were relatively few migrants from regions like East

Supralocalisation and the rise of Standard English

337

Anglia (Stone 1966, Wareing 1980, Finlay and Shearer 1986, Rappaport 1989: 78-9). A large number of the immigrants consisted of young single males, who were employed in low-paid, transient jobs (Beier 1990). Another large group consisted of apprentices, who had quite bright prospects to advance in life. Social historians generally point out that the London scene in the sixteenth and seventeenth centuries gave rise to what Stone (1966: 33) calls a phase of unprecedented individual social mobility, both upwards and downwards. Moreover, the population of early Modern London was boosted not only by subsistence migrants in search of livelihood, and upwardly mobile "betterment migrants" (Wareing 1980: 249), but also by members of the landed gentry and the nobility. London's social season attracted the nobility and gentry from all over England. Wrigley (1967) estimates that large sections of the early Modern English population came into contact with the capital at some time during their lives. Longer or shorter visits to London were paid by one out of eight English adults in 1550-1650, and by one out of six in 1650-1750. Large-scale immigration like this could lead to segregation and the rise of closed immigrant communities. Historical topographers like Power (1986), however, suggest that this was not the case in early Modern London, arguing that people from different walks of life did, to a large extent, intermingle in the same parishes and even in the same streets. Even in the Restoration, the City of London shows little evidence of the rigid class segregation so common in many present-day cities. John Stow, the famous author of the Survey of London (1598), provides a similar testimony when he says London's population was: "by birth for the most part a mixture of all counties, by blood gentlemen, yeomen, and of the basest sort without distinction" (cited in Rappaport" 1989: 86). Population movements of all kinds typically give rise to changes in the communities' social network structures. We may therefore assume that the rapid urbanisation and phenomenal growth that took place in early Modern London increased loose-knit and single-function social networks among the London population. As suggested by James and Lesley Milroy (e.g. 1985a, [1992] 1997), circumstances like this are particularly apt to promote language variation and change. Linguistic homogenisation also commonly takes place under similar circumstances today (Chambers 1995: 58-59). Demographic developments could therefore have helped to make London speech more

338

Terttu Nevalainen

usual throughout the country. On the other hand, if London English had already been selected and accepted as the model for nationwide use, the demographic mobility of the period would lead us to suspect that the variety was not as stable as writers like Puttenham, and Caxton before him, had wished it to be.

3.3. Modelling supralocalisation of linguistic variables I shall use the term supralocalisation to refer to the spread of a linguistic feature from its region of origin to neighbouring areas; when the process involves reduction of marked, minority variants, it is often called dialect levelling or koineisation (Trudgill 1986, Milroy and Milroy 1993, Holmes 1997, Milroy 1999).4 It is this stage in the lives of linguistic variables that is frequently missing in histories of early Modern English. As I wish to relate early Modern English research to what can be detected in the actual linguistic communities, I shall make use of the terms selection and acceptance with reference to individual linguistic features. "Selection" may be related to the work done both by innovators and by early adopters of new linguistic features within the Milroys' framework of social networks. In early Modern London a potential innovator could be an immigrant who retained a particular feature of his or her local dialect. An innovator for a group of the "better-brought-up sort" might also be a speaker of another sociolect, one of the "inferior sort", for instance. An early adopter within a network would be the one to do the actual selecting by picking up a new feature from an innovator. By "acceptance" I refer to the diffusion of new features within the network or community at large. The aim of this model is to further operationalise Haugen's standardisation scheme by focusing on interaction between different social groups and their sociolects, as is suggested by the demographic research on early Modern London. It is typically the acceptance phase, the diffusion of changes in a linguistic community, that we have access to in historical materials. As far as possible, I shall, however, also try to locate those social groups whose members must have been active in the selection process, thus referring to the distinction between change from below and change from above with reference to the social hierarchy (one of the two definitions of the term in

Supralocalisation and the rise of Standard English

339

Labov 1994: 78; see Nevalainen and Raumolin-Brunberg 1996). As I am looking at features that later became part of Standard English, I need to be able to measure the degree of supralocalisation and subsequent standardisation of a linguistic variable. The measure I shall use will be the completion of an S-curve by the feature in question in the data examined.

4.

Four empirical studies of supralocalisation

4.1. Hypotheses There are a number of specific hypotheses to do with supralocalisation that arise from the above discussion. They include, first of all, the directionality of change. A strong version of the standardisation model sketched in section 2.2 would imply that the direction of linguistic processes in the early Modern period should have been from London to the rest of the country, because the selection and acceptance of the standard would have taken place at the level of a particular linguistic variety. As a corollary, it would be predicted that literate Londoners would have been reluctant to promote features from other regional varieties. However, as suggested in section 3.2, the sociolinguistic model based on the demographics of London and the increase in weak ties would lead one to expect the opposite, that contact influence would in fact be detected, both regionally and socially. Another, more specific question focuses on London itself, and has to do with the possible difference between the Court and London proper. Puttenham particularly singles out the language of the Court as the model to be imitated, but this, we may think, was for pragmatic reasons only. He was, after all, writing for the aspiring poet who would be courting favour at the Queen's Court. The alternative hypothesis is provided by some textbooks on the history of English suggesting that it was the new urban social group, wealthy London merchants, and not the Court, who were the most influential group in standardisation (Milroy and Milroy 1993: 69). These two contending views are based on a social distinction: the Court in Westminster was the centre of administration where government documents were produced. Caxton set up his printing press not in the City but in Westminster "under the shadow of the government offices" (Fisher 1977: 899). Merchants, by contrast, were a growing commercial influence

340

Terttu Nevalainen

both in the capital and nationwide. This "mobile group of transients", as Stone (1966: 19) calls them, did not belong to the land-owning gentry, but were frequently assimilated into it on the basis of their income and lifestyle (Everitt 1966). The view of the merchants' role in language change may be supported by findings of modern sociolinguists. It is the intermediate social groups which typically promote linguistic variation and change in modern speech communities, while the upper and lower classes show the greatest resistance to change (Romaine 1988: 1464; Labov 1990: 226-227).

4.2. The material My data are drawn from the Corpus of Early English Correspondence (CEEC), which was compiled by the research team on Sociolinguistics and Language History at the University of Helsinki between 1993 and 1998. The research reported in this paper forms part of the work carried out by Helena Raumolin-Brunberg and myself within the project (Nevalainen and Raumolin-Brunberg forthcoming a and b). The corpus covers the period 1417-1681 and comprises 2.7 million running words. The material consists of the personal correspondence of 778 writers. Some pilot studies on it were published in Nevalainen and Raumolin-Brunberg (eds.) (1996). Table 1 shows the range of data the CEEC offers for studying early Modern English on a social and regional basis. The speaker-writers, as we may call the informants, are here classified according to rank and domicile. Four regions receive particular attention in the Corpus: London, the Court, East Anglia, and the North. Those people who lived in London (basically in the City, though Southwark is also included) are entered in our records as Londoners. The Court refers to people, mostly resident in Westminster, who were courtiers or otherwise belonged to the royal household, and to high-ranking government officials, including diplomats, directly reporting to the Queen and/or the King, or the Lord Chancellor. There is an obvious difference in the social status of the writers who represent the Court and those who represent London. The London social spectrum includes only a few members of the nobility, but the lower gentry, professional people and merchants are well represented. The Court, by contrast, is typically represented by royalty, other nobility, and the upper gentry.

Supralocalisation and the rise ofStandard English

341

Table 1. The Corpus of Early English Correspondence (1998 version). Time span: 1417-1681 (2,700,000 words; 778 writers; circa 6000 letters) WRITERS BY SOCIAL RANK WRITERS BY DOMICILE 2.4% Court: Royalty: 7.8% 14.7% London: Nobility: 13.9% East Anglia: Gentry: 39.3% 17.1% Clergy: 13.6% North: 12.5% 11.2% Other: Professionals: 48.6% Merchants: 8.4% Other non-gentry: 9.4% No similar social divide can be found in the other areas studied but, as far as possible, all literate ranks are represented in the corpus. Writers resident in Norfolk and Suffolk count as East Anglians, and those living in the counties north of the Chester-Humber line as Northerners. Our figures do not include people who had emigrated from their native area and settled permanently somewhere else. It is also worth noting that no basic distinction is made between those inhabitants of a locality who had migrated there from elsewhere, and those who were born there. Looking at those people whose place of birth has been identified, we can see that over 80 per cent of our Northerners were also born in the North, and almost 80 per cent of the East Anglians were native East Anglians. London, including the Court, turns out to be very different. Of all the writers who lived and worked in London and in the Court on a permanent basis, and whose place of birth we know, only about one third are recorded as having been born in London. However, this is still higher than the average estimate of 15 per cent cited by Coleman and Salt (1992: 27) for native Londoners in this period.5

4.3. The processes

examined

Using the CEEC as my database, I propose to examine the diffusion of four morphosyntactic features in late Middle and early Modern English. None of them was a major variant in Chancery English in the fifteenth century (Fisher, Richardson, and Fischer 1984). Some did not appear in Chancery English at all, while others were attested only as rare variants. The processes are: (i) the generalisation of the present indicative form are of the verb BE, (ii) the

342

Terttu Nevalainen

generalisation of the suffix -(e)s in the third person singular present indicative, (iii) the generalisation of the oblique form you in the subject function, and (iv) the simultaneous disappearance of multiple negation and generalisation of non-assertive indefinites. As shown below, the four processes are complex and would also merit a closer linguistic examination. My attention in this paper will, however, be directed to their extralinguistic profiles under the assumption that, whatever their linguistic paths of diffusion, all these processes were also subject to external conditioning.6

4.3.1. Generalisation of are The main regional variants of the present indicative plural forms of the verb BE in late Middle English are given in LALME. The LALME maps (Mcintosh, Samuels and Benskin 1986: 118-128) show that, between 1350 and 1450, we can still find a regional preference in the distribution of the two variants be and are. While both forms can be found throughout the country, be predominates in the south and west, and are in the north, where it goes back to old Anglian dialects. Norfolk has a high incidence of are and am, while the typical Scandinavian er is confined to Lincolnshire and further north (Samuels 1985: 274). One more point should be made about the be/are variable: the two forms do not exhaust the paradigm of the present indicative plural of BE. In a more complete study of the variable, the form is would count as one of the variants, as is suggested by example (lc), below. Bailey (1989) shows that is was an important variant in the fifteenth-century Cely correspondence. In the letters of the younger Celys, for instance, it represented some 20 per cent of the cases, while be covered 63 per cent, and are only 17 per cent. Interestingly, in the first early Modern English period of the Helsinki Corpus (1500-1570), is is used in the plural only in about three per cent of the cases (Kilpiö 1997: 101). We know from earlier studies that are mostly replaced be in large parts of the south in the sixteenth century, and that the two forms were used by many writers interchangeably in this period (Forsström 1948: 103, Barber 1976: 246, Kilpiö 1997: 101). Are dominated in the CEEC in the early seventeenth century (Nevalainen 1996). It was implemented by the 1662 revisers of the Book of Common Prayer, who replaced the vast majority of

Supralocalisation

and the rise of Standard English

343

earlier be forms with are (Nevalainen 1987). Be does, however, still occur in the Survey of English Dialects materials from the south and southwest, as shown in Upton and Widdowson (1996: 64). This form was elicited in an emphatic context ("Ohyes, we ...!"). Some examples of the be versus are variable from the CEEC are given under (1). The writers are identified by domicile, year of writing and name. Traditional regional forms continue to occur in the fifteenth and early sixteenth centuries. The typical Scandinavian form er is found in the North (la) and am in East Anglia (lb), while be is a typical London form (lc-d). (1)

a and Sir Roger Fenwike with ccc. men burnt the Town of Langton and distroyed all the comes therein: which Townes er in the hert of the countre two myle beyond Jedworth opon the watere of Chevyot. (North, 1513, Thomas Dacre; 94) b. I haue be there with youre tenauntys for here areragys, and they am better plessid to pay you pe mony thanne Master Will P. (East Anglia, 1470s, William Pecock; 11,412) c. Syr, owr father and modyr ys, and whe aull be, in good hey11, thankyd be God, and ar ryght glad that Ze pwrpos yow to be wyth them thys Kyrstemes. (London, 1478, Richard Cely Jr.; 40) d. All your menservauntes have bene of counsaill with hym, for they be of no les opynion, declaring that your breid is not good ynoghe for dogges, and drincke so evill that they cannot drinck it, but ar fayn when they go into the towne to drincke to their dynnars. (London, 1545, John Johnson; 250)

4.3.2. Generalisation of -(e)s The second process to be discussed is the diffusion of the third person singular present indicative suffix -(e)s throughout the country. Following the research tradition, I shall concentrate on -(e)s and -(e)th. In a more comprehensive study, the suffixless alternative should also be considered (see [2d]). The organisation

344

Terttu Nevalainen

of the whole present indicative paradigm would similarly be of interest, notably the Northern Subject Rule and its reflections in the Midlands and the South (Mcintosh 1983, Schendl 1996, Nevalainen and Raumolin-Brunberg forthcoming).7 LALME maps 645 and 646 indicate that there was a surprisingly clear regional divide in the distribution of the two variants -(e)s and -(e)th in the late Middle English period. North of the Chester-Wash line -(e)s dominated, south of it, -(e)th. Compared with the diffuse data on be and are, the situation in the third person singular present is clearly focused. Norfolk, for instance, shows only a few instances of the northern -(e)s in the singular. The southern -(e)th continued in the Chancery texts, and was later selected by Caxton as the form to appear in print (see Sandved 1968, Fisher 1996). We know, however, from the wealth of earlier research that -(e)s was generalised in most kinds of writing in the course of the seventeenth century. The dental fricative was preserved longest in two verbs, HAVE and DO. A particular characteristic in the history of the two suffixes is that they also became associated with register differences, -(e)th being connected with formal, literate styles, and -(e)s with informal and oral (Holmqvist 1922, Stein 1987, Kytö 1993). So the dental fricative was retained in the 1611 Bible and 1662 Prayer Book, for instance (Brook 1965). It continued as a regional form in southwestern dialects, and was recorded in Wright's (1905) dialect dictionary in Somerset and Devon.8 Some illustrations of the two forms in the CEEC are presented in (2). (2)

a. The Kyngpurposeth as to morow to be at Wendesor, andfrom thens to Notyngham. My Lord Chamberleyn rides to morue hame to Leycestre. (London, 1478?, Richard Page; 11:59) b. Your yong jentleman, Mr. Prat, hathe complayned by his lettre to his mother that tie lackythe bothe meat and drycke, as well his brekefastes, as also at meles not sufficient. (London, 1545, John Johnson 250) c. ... that I se not that he told you who bade him talke with Morgan of the price of my bloude, wiche he knowes, I am assured, right wel; nor yet hathe named the man that shuld be the murtherar of my life. (Court, 1585, Elizabeth I; 12)

Supralocalisation and the rise of Standard English

345

d. thy father remembers his loue to the and take thy wrightinge to him very kindly: thy brother remember his louingest loue to the ... I had thought to haue written to mr Roberts this time, but this sudene Iornye of this mesinger affordethe me not so much time (East Anglia, 1626, Katherine Paston; 90)

4.3.3. Generalisation of you The replacement of the second person pronoun ye by the oblique form you in the subject function has received less scholarly attention than the two verbal changes discussed above. Although some occasional instances of you can be found in the subject position from the fourteenth century onwards, the breakthrough of the oblique form largely took place in the sixteenth century (see Nevalainen and Raumolin-Brunberg 1996). It is sometimes attributed to phonological confusion between the weak forms of the plural pronoun, or analogy with the second person singular thou (Mustanoja 1960: 125, 129-130), sometimes to the general tendency of oblique forms to assume the subject function — one of Sapir's (1921) famous drift phenomena (for further discussion, see Lutz 1998). No specific regional origin is suggested for the process in LALME. Chancery texts still systematically use ye in the subject function, but the 1552 Prayer Book, for instance, mixes the two forms. The distinction between ye and you was, however, consistently restored in the 1662 Book of Common Prayer (Nevalainen 1987: 298). The examples in (3) illustrate the variable use of the two forms in the first half of the sixteenth century. (3)

a. Right wyrchypfull brodere, I hartyly recowmawnd me unto yow, besechyng yow to be gud brodere to me, and haffyng gret marvell of yowr unkyndnes, that ye wold not be here yowr selfe at thys tyme, nor nowne for yow, for here hays bene all my husbandes hu[{n{]kkylles & broders and I hayd nobody to speke in all my cawse bot my selfe. (North, 1518, Anne Clifton; 93) b. but that I wolde yow were in myne armes or I in yours, for I thynk it long, syns I kyst yow. (Court, 1528, King Henry VIII; 144)

346

Terttu Nevalainen c. Yf ye knowe they complayn with cawse, Ipraie you se it amendyd: (London, 1545, John Johnson; 250) ά. I do send this berar Fewren your servaunt hierwith unto you, and eftsones putt you owte of doubt that you may boldely remayne wher you ar, or ellis wher you lyst to be at your owne pleasur, ... (London, 1546, Otwell Johnson; 573)

4.3.4. Loss of multiple negation and rise of non-assertive indefinites My last case deals with the loss of multiple negation and the rise of nonassertive indefinites. In the written language the disappearance of multiple negation was well under way in the fifteenth century (Mustanoja 1960: 340, Tieken-Boon van Ostade 1994, Iyeiri 1999), and the process was nearly completed in the course of the seventeenth (Barber 1976: 283, Singh [1973] 1987, Nevalainen 1998). The sentential negator ne was practically lost by the end of the fifteenth century, and the use of more than one negative form in one clause was typically reserved for purposes of emphasis. Multiple negation started to lose ground supralocally in this function, too, in the early sixteenth century, giving way to a single negator followed by non-assertive indefinites. My study focuses on the replacement of multiple negatives by a single negative accompanied by non-assertive indefinites of the awy-type. The linguistic variable includes clauses with multiple negation and those with single negation where multiple would have been possible, but where nonassertive indefinite pronouns are used instead (Nevalainen 1998: 267-272). The examples in (4) illustrate the use of the variable. (4)

a. that they hafe your wryt[{y{]ng and a writyng vnder Ser John Paston ys seale promyttyng and warauntyng that neyther the seyd Ser John Paston ne none yn hys behalf, nother that none othyr havyng a rule late therebefore, shall not distreyn her catell and godes. (East Anglia, 1470, William Worcester; II, 583)

Supralocalisation and the rise of Standard English

347

b. yff they can to gett a savegard vnder the seall off Flounders that noo merchaunt off the Stappell schall be trubullyd yn Flaunders ffor any maleffett doon be any oder Englyscheman to any person off Flaunders, wythowte hytt be ffor tresspasse be hym don. (Calais, 1484, William Cely; 208) c. and that the dewke of Gelder send me no vord vat I sale do, nor heelpes me nat with notheng, as Petter sale chove yov, (Court, 1505, Edmund de la Pole; I, 254) d. there shall no poore neghebore of myne berre no losse by eny chaunce hapned in my howse. (Court, 1529, Thomas More; 423) e. and wher as I accordinglye haue not in lyke wise remembrid and rescribid it hath bene for that I haue not hade anything to wryt of to your aduauncement. (Court, 1523, Thomas Cromwell; I, 313) Many people used both multiple and single negation with non-assertive forms side by side in the sixteenth century, as in (4d). Supralocally, multiple negation was preserved later than elsewhere in co-ordinate structures with neither and nor, as in (4a) and (4c) (Iyeiri 1999; Nevalainen 1998: 278-280).

5. Results 5.1. Generalisation of are The overall distribution of are out of the variable total of be plus are in the indicative in the four regions is shown in figure 3. The corresponding numerical information is presented in the Appendix.9 As it is notoriously difficult to distinguish between the present indicative and subjunctive uses of BE, especially in the early part of the period, I decided to omit from the figures all instances of BE in clauses where the indicative mood could in principle alternate with the subjunctive, that is, clauses complementing a verb of wishing, commanding, doubting and hoping, as well as in optative, conditional,

348

Terttu Nevalainen

concessive, temporal, and exceptive clauses and generalised relative clauses (Nevalainen and Raumolin-Brunberg 1996: 323-324). The instances of are may therefore be somewhat over-represented in the results for the earliest periods.

Figure 3. The frequency of are in the present indicative (CEEC 1998; means of individual scores). These results are, however, sufficient to provide some answers to the questions posed in section 4.1. First of all, it is clear that features that do not originate in the South are acceptable in both London and the Court; what is more, many people are happy to use these two variants side by side for a long period of time. London seems to have accepted are earlier than the Court, and East Anglia earlier than London. The North, not surprisingly, manifests the highest proportion of are throughout the period. I interpret these results as a case of ordinary dialect diffusion: as are was percolating southwards in the fourteenth and fifteenth centuries, it had a stronger impact on East Anglia than on London. Its progress seems to have had little to do with the processes of standardisation discussed in section 2.2. An interesting fact that further appears from figure 3 is that be was also

Supralocalisation and the rise of Standard English

349

employed by northern speaker-writers, albeit as a minority variant. It has been suggested by Traugott (1972: 116) that be was used in northern dialects in predictions, for instance. As I tried to exclude any such modal uses of BE from my analysis, the result may be taken to suggest that dialect diffusion was a two-way process, at least for the literate section of the population that is focused on here.

5.2. Generalisation

of -(e)s

The variation between the third person present indicative singular forms -(e)s and -(e)th in the CEEC is shown in figure 4.

Figure 4. The frequency -(e)s in the third person singular present indicative (CEEC 1998; excluding HAVE and DO; means of individual scores). HAVE and DO, both auxiliary and main verb, have been excluded from these figures because of their very slow adoption of the incoming suffix -(e)s (see

350

Terttu Nevalainen

Nevalainen, Raumolin-Brunberg and Trudgill forthcoming). Figure 4 reflects the information contained in LALME that -(e)s was consistently retained in the North. It also shows that -(e)th made some inroads there in the sixteenth century. Just like the other northern form are, -(e)s was present in London in the late fifteenth century, especially in the correspondence of merchants, but also in the letters written by the lower gentry and members of the category of professionals given in table 1. The Court is not very well represented in the first period, but what data we have would indicate that the use of -(e)s was avoided in the Court. This would suggest that London and the Court again patterned differently with respect to a northern variant. East Anglia, by contrast, is very well represented in the fifteenthcentury materials, and shows next to no instances of the northern form. This state of affairs was also discovered by Wyld (1936: 336), who was puzzled not to find a case of regular dialect diffusion: We are placed in this dilemma, that the only apparent possible intermediary between the North and London and the South, by which a dialectal peculiarity could pass, is the E. Midland area, whereas this peculiar characteristic does not appear to be especially widespread in the E. Midland dialects, or among such writers as might be expected to show direct influence from these dialects in the fifteenth and sixteenth centuries. A plausible answer to Wyld's dilemma is offered by London demographics: the flow of migrants from the northern regions in the fifteenth and sixteenth centuries. According to social historians, East Anglia was quite self-contained at the time, to the extent that Phythian-Adams (1993) refers to it as a cultural province of its own. East Anglian apprentices, for instance, preferred to go to Norwich rather than to London (Patten 1976). Norwich was, after all, the largest provincial town in England at the time. The London situation was therefore more likely to have been brought about by dialect hopping by means of interurban links than steady dialect diffusion (see Trudgill 1986 and Labov 1994 on modern parallels, and Ringrose 1998 on urban links in the early Modern period). But migration does not explain everything. Despite the continuing flow of migrants to the capital from the North in the sixteenth century, the first half of the century shows a drop in the frequency of ~(e)s in London, and confirms that it was not used in the Court. Obviously, the diffusion of ~(e)s

Supralocalisation and the rise of Standard English

3 51

was later than that of are. While are was already to some extent used by Caxton in his printing press at Westminster, -(e)s was extremely rare (Sandved 1968). It was not the form to be selected at the time. Table 2. The distribution of -(e)s versus -(e)th in the Johnson circle (1542-1553), who were mostly active in London and on the continent, according to recipient ("intimates" = nuclear family). Most writers have data in only Total/all -(e)s + -(e)th

Intimates Ν of -(e)s

Distants Ν of-fete

(%)

(%)

LONDON, CALAIS

Cave, Anthony (London, Calais) Johnson, John (London, Calais) Johnson, Otwell (London) Johnson, Richard (London, Calais) Sandell, Richard (London, Essex)

128 165 90 76 24

Saunders, Blase (London) Pinder, Ralph (London) Total frequency of s Variable totals (-s + -th)

3 2 54 488

4 (5%) 5 (6%) 1 (1%)

29 (23%) 10(12%)

0 (0%) 3 13 (5%) 248

2 41 (17%) 240

CALAIS, ANTWERP, ETC.

Saunders, Ambrose (Calais, Antwerp) Southwick, Henry (Calais, Antwerp) Warner, Bartholomew (Calais) Andrew, Robert (Antwerp)

75 33 26 19

3 (4%)

Garbrand, Henry (Antwerp) Master, John (Kent, London) Master, Peter (Kent, London) Tupholme, John (Boston, Lines.) Tupholme, William (Boston, Lines.) Total frequency of - s Variable totals (-s + -th)

8 4 2 2 2 16 171

6

1 (3%) 0 (0%) 0 (0%)

1 1 2 2 11 (10%) 111

5 (8%) 60

It looks as though the position of ~(e)s had not been very securely established even in the merchant community. The form still occurs in the middle of the sixteenth century, but only as a minority variant in the section of the community that we have access to. The Johnson circle were relatively well-to-do wool merchants in London and Calais in the 1540s

352

Terttu Nevalainen

and 1550s. A look at their usage reveals their strong preference for -(e)th even when writing to their immediate family. The relevant figures are shown in table 2. The only person to use -(e)s to a higher degree, about 20 per cent of the time, is Anthony Cave. He was born in Northamptonshire and belonged to the older generation of the merchant circle. As shown in figure 4, the southern -(e)th also spread to the North in the course of the fifteenth and early sixteenth centuries. Interested to see how it had been received, we analysed the CEEC data in terms of their register implications. The results suggest that, in general, the southern form was less commonly used by northerners when addressing family and personal friends than when addressing more distant correspondents. Evidence for register differentiation in the use of the suffixes in the North may therefore partly explain why -(e)s receded in London, too, in the first half of the sixteenth century: it may have been regarded as dialectal, colloquial, or socially stigmatised (Nevalainen and Raumolin-Brunberg forthcoming a and b). However, as shown by figure 4, a change in the social evaluation of the two forms must have taken place in the second half of the sixteenth century, when -(e)s began to spread. As earlier, it was first accepted in London as opposed to the Court. The Court, however, soon caught up and surpassed London in the early seventeenth century. The situation from the latter half of the sixteenth century onwards repeats the pattern that we saw with are: a northern form becomes generalised in the dominant supraregional variety. Although both cases suggest a change from below in social terms, they differ in that there is more evidence for a register split with the third person singular.

5.3. Generalisation of you Figure 5 shows that the replacement of ye with you in the subject function was completed in less than a century in the CEEC. Once under way, the change was led jointly by London and the Court, with East Anglia and the North catching up by the end of the sixteenth century. At one point, between 1480 and 1520, the Court even seemed to be leading the process, but the difference is due to the heavier concentration of the London data in the first half of the period, and the Court data in the latter.10

Supralocalisation and the rise of Standard English

353

Figure 5. The frequency of you in the subject function (CEEC 1998; means of individual scores). The process could be interpreted as another change from below, in the traditional sociolinguistic sense of one taking place below the level of social awareness (Labov 1994: 78). However, in individual cases, some evidence can be gathered for register differentiation. In a sample of four writers, three used you more frequently in their family letters than in other contexts of writing. At the same time, the difference was negligible even in a style-conscious writer like Sir Thomas More: he used you 70 per cent of the time in his non-private correspondence (41/59 cases), and 76 per cent of the time in his private letters (28/37) (Raumolin-Brunberg and Nevalainen 1997). We may therefore assume that the supralocalisation of you was so rapid that there were contemporaries who readily adopted it and others who hardly caught up with it at all. Its selection and acceptance seem to have overlapped. Moreover, the diffusion of you in the subject function is the first process examined so far to show a clear southern advantage: the change appears to be initiated in the capital rather than in the North. What may be crucial to

354

Terttu Nevalainen

this change of direction in supralocalisation is its timing: the bulk of the change takes place in the sixteenth century.

5.4. Loss of multiple negation and rise of non-assertive

indefinites

As the multiple negation variable is much less frequent than the other three cases that I have discussed (a couple of thousand instances as opposed to 5,000 or 10,000: see the Appendix), I have contrasted London and the Court with the rest of the country. The results are presented in figure 6. It shows the gradual disappearance of multiple negation in male usage, and implies a simultaneous rise of non-assertive forms with single negation.

Figure 6. The frequency of multiple negation (CEEC 1998; means of period totals; male writers). It appears from figure 6 that there is some evidence of non-assertive forms in negative contexts in the latter half of the fifteenth century. The Court is

Supralocalisation

and the rise of Standard English

355

poorly represented in the first subperiod, but if we look at the situation in the second, a statistically significant (chi-square test) difference can be detected between London and the Court. The Court continues to lead in the third subperiod, and the margin is significant at the five per cent level, but the differences even out in the fourth, around 1600. The tail end of the process is seen in the last 20 years covered by the corpus, 1660-1681. I would suggest that the loss of multiple negation progressed from above the level of consciousness. One thing shared by those who promoted it was their familiarity with the legal language of the time. This suggestion is based on an analysis of the social distribution of multiple negation in the whole of the CEEC. β Nobility, C e n t n • Social Movers, Professionals, M e r c h a n t s • Non-gentry

,

,

.IHI.-ll

1460-1490 1520-1550 1580-1610

1660-1681

Figure 7. The frequency of multiple negation according to writer's rank in four periods: 1460-90, 1520-50, 1580-1610 and 1660-81 (CEEC 1998; means of period totals; male writers).

Figure 7 indicates that in the sixteenth century multiple negation was particularly avoided — and single negation with non-assertive forms therefore preferred — by men who were also accustomed to communicating in writing in their professional capacity. They include members of the legal

356

Terttu Nevalainen

profession and, in particular, upwardly mobile administrators in the Court, as well as some well-to-do London merchants.11 A similar preference for nonassertive forms can also be observed in the Statutes of the Realm at the time, as shown in Rissanen (forthcoming). The implication of these findings is that the change was diffused from above in the sense that it was closely associated with professional usage. In Nevalainen (1998: 275) I further show that the disappearance of multiple negation correlated with gender: it was favoured by men. The information in figure 7 also supports the notion of change from above in social terms by indicating that multiple negation remained frequent in non-professional ranks below the gentry. Although these non-gentry categories are numerically small in the corpus, the tendency is worth noting.

6.

Discussion

Haugen (1997: 349) probably comes close to the truth when he notes that where a new linguistic norm is to be established, the problem will be as complex as the sociolinguistic structure of the people involved. With my four case studies I hope to have established that, in the late Middle and early Modern English periods, the morphosyntactic features selected for supralocalisation were not the sole property of London English. It appears, in particular, that those future standard language features that had begun to supralocalise in the fifteenth century or earlier need not have originated in the south. Both are and -(e)s were of northern origin (and there is evidence for a number of other similar features, including the third person plural forms of personal pronouns; see e.g. Heikkonen 1996). Even these two had, however, different profiles: the supralocalisation of are can be identified as a case of regular dialect diffusion that progressed over the centuries, as is also suggested by LALME. The third person suffix -(e)s, by contrast, looks more like a case of "dialect hopping": it had not properly reached the Central Midlands in the LALME sources, but showed up in London earlier than in East Anglia in the CEEC. Moreover, when it supralocalised in the sixteenth century, the process was led by London, while the Court clearly lagged behind. In both cases, the processes of supralocalisation between the North and the South were two-way streets, not only of the two northern features

Supralocalisation and the rise of Standard English

357

percolating south but also of the southern features -(e)th and be spreading northwards. The mediating role of the Central Midlands in these processes cannot be assessed on the basis of the present data (cf. Samuels 1963). As are and -(e)s showed such vastly different profiles of diffusion, however, one may assume that no generalisation could be made at the variety level. There are a number of social and demographic factors to do with the time courses of individual processes that would have influenced their outcome, patterns of internal migration and register associations being among the most important of them. However, it appears that many, if not most, of the morphosyntactic features that were supralocalised in the sixteenth and seventeenth centuries diffused from the capital region to the rest of the country. In my survey based on four regions, this was the case with you in the subject function, and the replacement of multiple negation with the use of single negation and non-assertive forms (for a number of other, similar processes, see Nevalainen and Raumolin-Brunberg forthcoming b). This does not, however, mean that variety-specific generalisations could be made about London English in these cases either. As with are and -(e)s, a difference emerged between the Court and London proper in the disappearance of multiple negation when the process was nearing its 50 per cent mark, i.e. the mid-range of its S-curve. This evidence supports the general argument that, in the early Modern period, supralocalisation took place both from below and from above in terms of social status as well as social awareness. The case of you also brings home the fact that it is difficult to speak about standardisation at the level of varieties because the supralocalisation of individual features may vary from a few decades, as with you, to some hundreds of years, as in the case of are. Moreover, a number of real time studies carried out within the Sociolinguistics and Language History Project suggest that the social alignment of a given change in progress need not remain constant as the process advances. The diffusion of -(e)s is a case in point (see also Palander-Collin 1998). It is therefore interesting to ask whether changes in these processes might not have coincided with social upheavals and periods of increased social and geographical mobility. Present-day English studies suggesting this include Bailey et al. (1996). Some macro-level correlations have also emerged in the early Modern period. Raumolin-Brunberg (1998) shows the relevance of the Civil War as an accelerator of linguistic changes in the

358

Terttu Nevalainen

mid-seventeenth century, and Nurmi (this volume) the correlation of the arrival of the Scottish court in London with the (non-)use of periphrastic DO at the beginning of the seventeenth century. Similarly, it is plausible that the diffusion of you may also have been precipitated by the turmoils of the Reformation in London, coinciding with a population boom in the capital. Porter's ([1994] 1996: 37) summary of the social and economic impact of the Reformation in London vividly illustrates this period of rapid social change (see further Brigden 1989, Inwood 1998: 157-161): The Reformation revolutionized the face of the medieval city ... At a stroke, priories, hospitals, chapels and shrines changed hands. Church lands and properties were, formally speaking nationalized, but in effect most were privatized, being sold off or dished out by the Crown to its courtiers and cronies. London acquired a new breed of parvenu landlords, who felt free to do as they pleased with their windfall wealth. A hectic property market followed, encouraging opportunistic redevelopment comparable perhaps to the speculative fever following the Second World War. And this real-estate boom occurred just when population was soaring and the economy hotting up.

7.

Conclusion

To summarise, I have suggested that a processual model could profitably be adopted to study the linguistic convergence that took place in early Modern English. It may be implemented empirically by studying actual linguistic usages simultaneously in various parts of the country, including and contrasting the City of London and the Royal Court. A framework for interpreting the ongoing processes of supralocalisation is provided by the social and demographic developments in the country as a whole, especially by relations between the capital and the rest of the country. In a rapidly changing urban setting, the relevant sampling period for the linguistic data may be as short as 20 years or even shorter. Borrowing the terms "selection" and "acceptance" from standardisation research, I hope to have thrown some light on the evolution of supralocal usages propagated by the literate section of the population. Although the capital region typically led these processes from the sixteenth century onwards, features originating from regions other than the East Midlands

Supralocalisation

and the rise of Standard English

359

were also selected, and gained acceptance as part of a nationwide sociolect. Moreover, processes from both above and below in the social hierarchy were in evidence, as was socially unmarked change. These findings may be interpreted sociolinguistically by suggesting that the composition of the metropolitan usage, and consequently many aspects of future supralocal norms of English, were negotiable in the late Middle and early Modern English periods. Finally, thinking of processes of supralocalisation in more global terms, one should not forget that the narrowing of Milroy's funnel around 1550 took place at one level only, that of supralocal usage in England. At the same time, the funnel began to widen geographically owing to the rise of extraterritorial varieties of English. These new lines of development have gathered momentum ever since, extending the convergence versus divergence debate to English worldwide at the turn of the twenty-first century.12

Appendix: Numerical data

Table I. (Figure 3) The relative frequency of are versus be in the present indicative according to writer's domicile: variable totals (N), totals (are) and means of individual scores of are (Average) per period (CEEC 1998; all writers). 14601500154015801620Total 1499 1539 1579 1619 1639 318 64 London are 48 3 155 588 Average 25.9 33.3 56.9 88.8 95.6 (170) (322) (358) (69) (928) (N) (9) Court

are Average (N)

0 0.0 (10)

27 11.2 (217)

116 58.5 (215)

293 87.8 (350)

59 495 100.0 (59) (851)

North

are Average (N)

6 46.2 (13)

35 70.0 (50)

11 65.0 (15)

43 90.6 (49)

222 317 98.3 (227) (354)

East Anglia

are Average (N)

75 36.1 (226)

3 37.5 (8)

154 68.1 (221)

90 74.3 (126)

136 458 95.0 (148) (729)

360

Terttu Nevalainen

Table II. (Figure 4) The relative frequency of -s versus -th in the third person singular present indicative according to writer's domicile: variable totals (N), totals (-s) and means of individual scores of -s (Average) per period, excluding HAVE and DO (CEEC 1998; all writers). 1460- 1500- 1540- 1580- 1620- 1660- Total 1499 1539 1579 1619 1659 1681 London 2 224 -s 147 53 579 206 1211 Average 31.9 4.2 8.4 82.4 64.8 98.6 (51) (516) (759) (265) (508) (209) (2308) (N) Court

-s Average (N)

North

-s Average (N)

East Anglia

-s Average (N)

0 0.0 (12)

9 28 2.4 5.7 (735) (387)

21 59.6 (35)

30 21.4 (139)

10 0.9 (690)

0 0.0 (11)

270 38.4 (865)

129 91.4 (143)

231 667 98.5 (236) (2378)

7 32.0 (22)

38 41.2 (84)

265 75.0 (452)

204 91.0 (227)

19 4.8 (428)

20 5.2 (202)

291 59.8 (390)

90 430 84.1 (126) (1847)

565 (959)

Table III. (Figure 5) The relative frequency of you versus ye in the subject function according to writer's domicile: variable totals (N), totals (you) and means of individual scores of you (Average) per period (CEEC 1998; all writers). 14601480152015601600Total 1479 1519 1559 1599 1619 London 16 195 662 you 31 358 1262 9.0 Average 6.6 45.6 100.0 99.7 (402) (1516) (717) (195) (663) (3493) (N) Court

you Average (N)

North

you Average (N)

East Anglia

you Average (N)

0 0.0 (17)

109 431 15.4 40.7 (212) (1070)

878 96.8 (941)

294 1712 96.4 (296) (2536)

5 9.4 (57)

10 6.3 (170)

16 11.5 (124)

27 100.0 (27)

72 100.0 (72)

18 3.9 1283

0 0.0 75

2 12.5 16

935 99.1 944

99 100.0 99

130 (450) 1054 2417

Supralocalisation and the rise of Standard English

361

Table IV. (Figure 6) The relative frequency of multiple versus single negation with non-assertive forms according to writer's domicile: variable totals (MN + SN), totals (MN) and relative frequencies (%) of multiple negation per

London

MN

% (MN + SN) Court

MN

% (MN + SN) Rest of country

MN

% (MN + SN)

14601479 85 88 (97)

14801519 33 94 (35)

15201559 79 40 (197)

15601599 9 12 (73)

16601681 1 1 (76)

0 0 (0)

18 58 (31)

76 31 (249)

15 9 (162)

1 2 (61)

162 94 (173)

32 80 (40)

103 46 (222)

55 14 (403)

11 4 (278)

Total 207 (478) 110 (503) 363 (1116)

Table V. (Figure 7) The relative frequency of multiple versus single negation with non-assertive forms according to writer's social rank: variable totals (MN + SN), totals (MN) and relative frequencies (%) of multiple negation per period (CEEC 1998; male writers). 1460152015801660Total 1490 1550 1610 1681 MN 180 111 29 9 Gentry* 329 4 % 90 49 10 (MN + SN) (201) (227) (300) (232) (960) Middle ranks**

MN

% (MN + SN)

102 95 (107)

76 27 (282)

16 100 (16)

16 84 (19)

9 5 (183)

3 2 (178)

190 (750)

1 38 20 (MN + SN) (50) (5) (* consisting of the nobility, upper and lower gentry, and upper clergy) (** consisting of social movers, the lower clergy and other professionals, and Other non-gentry

MN

%

merchants)

5 50 (10)

362

Terttu Nevalainen

Notes 1. The Sociolinguistics and Language History team, which holds copyright on the CEEC, consists of Jukka Keränen, Minna Nevala, Arja Nurmi, Minna PalanderCollin, Helena Raumolin-Brunberg and myself. 2. See for example Nevalainen and Raumolin-Brunberg (1989: 75-81) and Görlach (1991: 9) for further discussion on time limits. They range from 1500-1700 to 1500-1660, 1500-1800, 1540-1750 and, in The Cambridge History of the English Language, from 1476 to 1776 (Lass 1999). 3. See further for example the monograph-length discussions by Milroy and Milroy (1985b) and Joseph (1987). 4. I would like to thank Lesley Milroy for providing me with some comparisons with supralocal developments in Present-day English. See also Hickey (1998). 5. The four regions discussed represent about half of the Corpus in terms of running words. The other half consists of writers whose domicile we may not have been able to identify or who permanently lived in other parts of the country. Most interesting from the supralocalisation point of view are the Central Midlands, on the one hand, and the West Country, on the other. Where possible, they will be considered in future work. 6. Keeping the linguistic and sociolinguistic profiles of linguistic changes separate is advocated by many sociolinguists, for instance, Wolfram (1991). 7. The widespread diffusion of this rule throughout the English-speaking world also attracts the attention of modern sociolinguists and social dialectologists; see Clarke (1997) and Godfrey and Tagliamonte (1999). 8. But no longer found in the Devon data analysed and discussed by Godfrey and Tagliamonte (1999). 9. The 20 year sampling intervals introduced by the Sociolinguistics and Language History team are here (and in most of my subsequent figures) conflated into 40 year periods, but preserved as the basis for calculating the means of the individual writers' scores. The relative frequency of are (-(e)s, and you) was calculated for each writer with more than ten instances of the study variable per period. The rest were scored in the aggregate. The calculation was performed in order to avoid any skewing of the period means of high frequency variables by individuals who contribute relatively large samples to the CEEC (see further Nevalainen and Raumolin-Brunberg, forthcoming a and b). The numerical details for figures 3-7 are given in the Appendix. 10. The difference between London and the Court is due to the scarcity of London data in the second half of the period: in the subperiod 1500-1519 the London variable total is four, while the corresponding Court total is 141. 11. An interesting picture emerges if we take a closer look at the people who contributed to the parallel results in London and the Court in period 2. In London we find that the largest contribution to the variable was made by wealthy merchants, notably the Johnson circle discussed in section 5.2. In the

Supralocalisation and the rise of Standard English

363

Court the contributors mostly consisted of high ranking upwardly mobile professionals, for example, Thomas Cromwell, Stephen Gardiner, and William Paget. These men are good examples of the upward social mobility that was characteristic of London at the time. All three of them were born to ranks below the gentry, but had highly successful public careers and were raised to the peerage, either temporal or spiritual (Nevalainen 1998). 12. For some recent discussions, see Crystal (1997), McAilhur (1998), and Trudgill (1998).

References Bailey, Guy 1989

Sociolinguistic constraints on language change and the evolution of are in early Modem English. In: John Trahern (ed.), Standardizing English: Essays in the History of Language Change, 158-171. Knoxville: University of Tennessee Press. Bailey, Guy, Tom Wikle, Jan Tillery and Lori Sand 1996 The linguistic consequences of catastrophic events: An example from the American Southwest. In: Jennifer Arnold, Renee Blake, Brad Davidson, Scott Schwenter and Julie Solomon (eds.), Sociolinguistic Variation: Data, Theory, and Analysis, 435-451. Stanford, CA: CSLI Publications. Barber, Charles 1976 Early Modern English. London: Andre Deutsch. Barton, David and Mary Hamilton. 1996 Social and cognitive factors in the historical elaboration of writing. In: Andrew Lock and Charles R. Peters (eds.), Handbook of Human Symbolic Evolution, 793-858. Oxford: Clarendon Press. Beier, A. L. 1990 Social problems in Elizabethan London. In: Jonathan Barry (ed.), The Tudor and Stuart Town, A Reader in English Urban History 1530-1688, 121-138. London/New York: Longman. Beier, A. L. and Roger Finlay 1986 The significance of the metropolis. In: A. L. Beier and Roger Finlay (eds.), The Making of the Metropolis: London 1500-1700, 1-33. London/New York: Longman. Bennett, H. S. 1969 English Books and Readers: 1475-1557, Volume 1. Cambridge: Cambridge University Press. Blake, Norman F. 1992 Introduction. In: N. F. Blake (ed.), The Cambridge History of the English Language, Volume 2: 1066-1476, 1-22. Cambridge: Cambridge University Press.

364

Terttu Nevalainen

Bridgen, Susan 1989 Brook, Stella 1965 Chambers, J.K. 1995 Clarke, Sandra 1997 Clay, C. G. A. 1984

London and the Reformation. Oxford: Clarendon Press. The Language of the Book of Common Prayer. London: Deutsch. Sociolinguistic Theory: Linguistic Variation and its Social Significance. (Language in Society 22.) Oxford: Blackwell. English verbal -s revisited: The evidence from Newfoundland. American Speech 72: 227-259.

Economic Expansion and Social Change: England 1500-1700: People, Land and Towns, Volume 1. Cambridge: Cambridge University Press. Coleman, David and John Salt 1992 The British Population: Patterns, Trends, Processes. Oxford: Oxford University Press. Corfield, Penelope 1990 Urban development in England and Wales in the sixteenth and seventeenth centuries. In: Jonathan Barry (ed.), The Tudor and Stuart Town, A Reader in English Urban History 1530-1688, 35-62. London/New York: Longman. Cressy, David 1980 Literacy and Social Order. Cambridge: Cambridge University Press. Crowley, Tony The Politics of Discourse: The Standard Language Question in 1989 British Cultural Debates. London: Macmillan. Crystal, David English as a Global Language. Cambridge: Cambridge University 1997 Press. Dobson, E. J. 1968 English Pronunciation 1500-1700, Volume 1. Oxford: Clarendon Press. First edition [1957]. Eisenstein, Elizabeth 1983 The Printing Revolution in Early Modern Europe. Cambridge: Cambridge University Press. Everitt, Alan 1966 Social mobility in Early Modern England. Past and Present 33: 56-73. Finlay, Roger and B. Shearer 1986 Population growth and urban expansion. In: A. L. Beier and Roger Finlay (eds.), The Making of the Metropolis: London 1500-1700, 37-59. London/New York: Longman.

Supralocalisation and the rise of Standard English Fisher, John H. 1977

365

Chancery and the emergence of standard written English in the fifteenth century. Speculum 52: 870-899. 1996 The Emergence of Standard English. Lexington, KY: The University Press of Kentucky. Fisher, John H., Malcolm Richardson and Jane L. Fisher 1984 An Anthology of Chancery English. Knoxville: University of Tennessee Press. Forsström, Gösta 1948 The Verb 'To Be' in Middle English: A Survey of the Forms. (Lund Studies in English 15.) Lund: Gleerup. Godfrey, Elizabeth and Sali Tagliamonte 1999 Another piece for the verbal -s story: Evidence from Devon in Southwest England. Language Variation and Change 11: 87-121. Görlach, Manfred 1990 The study of Early Modern English variation — the Cinderella of English historical linguistics? In: Manfred Görlach (ed.), Studies in the History of the English Language, 108-122. Heidelberg: Carl Winter. 1991 Introduction to Early Modern English. Cambridge: Cambridge University Press. Haugen, Einar 1997 Language standardization. In: Nikolas Coupland and Adam Jaworski (eds.), Sociolinguistics, 341-352. London: Macmillan. First published as: Dialect, language and nation. American Anthropologist 68: 922-935 [1966]. Heal, Felicity and Clive Holmes 1994 The Gentry in England and Wales 1500-1700. London: Macmillan. Heikkonen, Kirsi 1996 Regional variation in standardization: A case study of Henry V's Signet Office. In: Terttu Nevalainen and Helena Raumolin-Brunberg (eds.), Sociolinguistics and Language History: Studies Based on the Corpus ofEarly English Correspondence, 111-127. Amsterdam/Atlanta, GA: Rodopi. Hickey, Raymond 1998 Supraregionalisation as a type of language change. Paper presented at the Tenth International Conference on English Historical Linguistics, Manchester, 21st-26th August 1998. Hogg, Richard M. 1998 On the ideological boundaries of Old English dialects. In: Jacek Fisiak and Marcin Krygier (eds.), Advances in English Historical Linguistics, 107-118. (Trends in Linguistics, Studies and Monographs 112.) Berlin/New York: Mouton de Gruyter.

366

Terttu Nevalainen

Holmes, Janet 1997

Setting new standards: Sound changes and gender in New Zealand English. English World-Wide 18: 107-142.

Holmqvist, Erik 1922 On the History of the English Present Inflections, Particularly -th and-s. Heidelberg: Carl Winter's Universitätsbuchhandlung. Inwood, Stephen 1998 A History of London. London: Macmillan. Iyeiri, Yoko 1999 Multiple negation in Middle English verse. In: Ingrid TiekenBoon van Ostade, Gunnel Tottie and Wim van der Wurff (eds.), Negation in the History of English, 121-146. (Topics in English Linguistics 26.) Berlin/New York: Mouton de Gruyter. Jespersen, Otto 1949 A Modern English Grammar on Historical Principles, Volume 1. London: Allen and Unwin. First published Heidelberg: Carl Winter's Universitätsbuchhandlung [1909], Joseph, John Earl 1987 Eloquence and Power: The Rise of Language Standards and Standard Languages. London: Pinter. Kilpiö, Matti 1997 On the forms and functions of the verb be from Old to Modern English. In: Matti Rissanen, Merja Kytö and Kirsi Heikkonen (eds.), English in Transition: Corpus-based Studies in Linguistic Variation and Genre Styles, 87-120. (Topics in English Linguistics 23.) Berlin/New York: Mouton de Gruyter. Kytö, Merja 1993 Third-person present singular verb inflection in early British and American English. Language Variation and Change 5: 113-139. Labov, William 1990 The intersection of sex and social class in the course of linguistic change. Language Variation and Change 2: 205-254. 1994 Principles of Linguistic Change. Oxford: Blackwell. Lass, Roger 1976 English Phonology and Phonological Theory. Cambridge: Cambridge University Press. Lass, Roger (ed.) 1999 The Cambridge History of the English Language, Volume 3: 1476-1776. Cambridge: Cambridge University Press. Leith, Dick 1983 A Social History of English. London: Routledge and Kegan Paul.

Supralocalisation Lutz, Angelika 1998

McArthur, Tom 1998

and the rise of Standard English

367

The interplay of external and internal factors in morphological restructuring: the case of you. In: Jacek Fisiak and Marcin Krygier (eds.), Advances in English Historical Linguistics, 189-210. (Trends in Linguistics, Studies and Monographs 112.) Berlin/New York: Mouton de Gruyter. The English Press.

Languages.

Cambridge: Cambridge

University

Mcintosh, Angus 1983 Present indicative plural forms in the Later Middle English of North Midlands. In: Douglas Gray and Eric G. Stanley (eds.), Middle English Studies Presented to Norman Davis, 235-244. Oxford: Oxford University Press. Mcintosh, Angus, Michael L. Samuels and Michael Benskin 1986 A Linguistic Atlas of Late Mediaeval English [LALME], Volume 1. Aberdeen: Aberdeen University Press. Milroy, James 1992 Linguistic Variation and Change. Oxford: Blackwell. Milroy, James and Lesley Milroy 1985a Linguistic change, social network and speaker innovation. Journal of Linguistics 21: 339-384. 1985b Authority in Language: Investigating Language Prescription and Standardization. London/Boston/Henley: Routledge and Kegan Paul. 1993 Mechanisms of change in urban dialects: The role of class, social network and gender. International Journal of Applied Linguistics 3 (1): 57-77. 1997 Network structure and linguistic change. In: Nikolas Coupland and Adam Jaworski (eds.), Sociolinguistics, 199-211. London: Macmillan. First published as: Speaker innovation and linguistic change. In: James Milroy, Linguistic Variation and Change: On the Historical Sociolinguistics of English, 176-191. Oxford: Blackwell [1992], Milroy, Lesley Women as innovators and norm-creators: The sociolinguistics of 1999 dialect leveling in a northern English city. In Suzanne Wertheim, Ashlee C. Bailey and Monica Corston-Oliver (eds.), Engendering Communication: Proceedings of the Fifth Berkeley Women and Language Conference, 361-376. Berkeley, CA: BWLG.

368

Terttu Nevalainen

Mustanoja, Tauno F. 1960 A Middle English Syntax. Part I: Parts of Speech (Mömoires de la Sociötö Nöophilologique de Helsinki 23.) Helsinki: Sociötö Niophilologique. Nevalainen, Terttu 1987 Change from above: A morphosyntactic comparison of two early Modern English editions of The Book of Common Prayer. In: Leena Kahlas-Tarkka (ed.), Neophilologica Fennica, 295-315. (M&noires de la Socidtö Niophilologique de Helsinki 45.) Helsinki: Sociötö Neophilologique. 1996 Social stratification. In: Terttu Nevalainen and Helena RaumolinBrunberg (eds.), Sociolinguistics and Language History: Studies Based on the Corpus of Early English Correspondence, 57-76. Amsterdam/Atlanta, GA: Rodopi. 1998 Social mobility and the decline of multiple negation in early Modern English. In: Jacek Fisiak and Marcin Krygier (eds.), Advances in English Historical Linguistics, 263-291. (Trends in Linguistics, Studies and Monographs 112.) Berlin/New York: Mouton de Gruyter. Nevalainen, Terttu and Helena Raumolin-Brunberg 1989 A corpus of early Modern Standard English in a social perspective. Neuphilologische Mitteilungen 90 (1): 67-110. 1996 Social stratification in Tudor English? In: Derek Britton (ed.), English Historical Linguistics 1994, 303-326. (Current Issues in Linguistic Theory 135.) Amsterdam/Philadelphia: John Benjamins, forthc.a. The changing role of London on the linguistic map of Tudor and Stuart England. In: Dieter Kastovsky and Arthur Mettinger (eds.), The History of English in a Social Context. (Trends in Linguistics.) Berlin/New York: Mouton de Gruyter. forthc.b. Historical Sociolinguistics. London: Longman. Nevalainen, Terttu and Helena Raumolin-Brunberg (eds.) 1996 Sociolinguistics and Language History: Studies Based on the Corpus of Early English Correspondence. Amsterdam/Atlanta, GA: Rodopi. Nevalainen, Terttu, Helena Raumolin-Brunberg and Peter Trudgill forthcoming Chapters in the social history of East Anglian English: The case of the third-person singular. In: Jacek Fisiak and Peter Trudgill (eds.), East Anglian English. Cambridge: Boydell and Brewer. Nurmi, Arja this volume The rise and fall of periphrastic DO in early Modern English, or "Howe the Scotts will declare themselv 's".

Supralocalisation and the rise of Standard English

3 69

Palander-Collin, Minna 1998 Grammaticalization of I THINK and METHINKS in late Middle and early Modem English: A sociolinguistic perspective. Neuphilologische Mitteilungen 99 (4): 419-442. Patten, John 1976 Patterns of migration and movement of labour to three preindustrial East Anglian towns. Journal of Historical Geography 2 (2): 111-129. Phythian-Adams, Charles 1993 Introduction: An agenda for English local history. In: Charles Phythian-Adams (ed.), Societies, Cultures and Kinship, 1580-1850, 1-23. London/New York: Leicester University Press. Porter, Roy 1996 London: A Social History. Harmondsworth: Penguin Books. First edition [1994], Power, M. J. 1986 The social topography of Restoration London. In: A. L. Beier and Roger Finlay (eds.), The Making of the Metropolis: London 1500-1700, 199-223. London: Longman. Puttenham, George [1968] The Arte of English Poesie (1589). (English Linguistics 110.) Menston: Scolar Press. First published London: Richard Field. Rappaport, Steve 1989 Worlds within Worlds: Structures of Life in Sixteenth-Century London. Cambridge: Cambridge University Press. Raumolin-Brunberg, Helena 1998 Social factors and pronominal change in the seventeenth century: The Civil War effect? In: Jacek Fisiak and Marcin Krygier (eds.), Advances in English Historical Linguistics, 361-388. (Trends in Linguistics, Studies and Monographs 112.) Berlin/New York: Mouton de Gruyter. Raumolin-Brunberg, Helena and Terttu Nevalainen 1997 Social embedding of linguistic changes in Tudor English. In: Raymond Hickey and Stanislaw Puppel (eds.), Language History and Linguistic Modelling: A Festschrift for Jacek Fisiak on his Sixtieth Birthday, 701-717. Berlin/New York: Mouton de Gruyter. Ringrose, David R. 1998 Capital cities, urbanization, and modernization in early Modern Europe. Journal of Urban History 24 (2): 155-183.

370

Terttu Nevalainen

Rissanen, Matti 1986

Variation and the study of English historical syntax. In: David Sankoff (ed.), Diversity and Diachrony, 97-109. Amsterdam/ Philadelphia: John Benjamins. forthcoming Standardisation and the language of early statutes. In: Laura Wright (ed.), Standardisation. Cambridge: Cambridge University Press. Romaine, Suzanne 1988 Historical sociolinguistics: Problems and methodology. In: Ulrich Ammon, Norbert Dittmar and Klaus J. Mattheier (eds.), Sociolinguistics: An International Handbook of the Science of Language and Society, 1452-1469. Berlin/New York: Walter de Gruyter. Samuels, Michael L. 1963 Some applications of Middle English dialectology. English Studies 44: 81-94. 1985 The great Scandinavian belt. In: Roger Eaton, Olga Fischer, Willem Koopman and Frederike van der Leek (eds.), Papers from the Fourth International Conference on English Historical Linguistics, 269-281. Amsterdam/Philadelphia: John Benjamins. Sandved, Arthur O. 1968 Studies in the Language of Caxton's Malory and That of the Winchester Manuscript. Oslo: Norwegian Universities Press. Sapir, Edward 1921 Language: An Introduction to the Study of Speech. New York: Harcourt, Brace and Company. Schendl, Herbert 1996 The third plural present indicative in early Modern English — variation and linguistic contact. In: Derek Britton (ed.), English Historical Linguistics 1994, 143-160. (Current Issues in Linguistic Theory 135.) Amsterdam/Philadelphia: John Benjamins. Singh, Rajendra 1987 Multiple negation in Shakespeare. In: Vivian Salmon and Edwina Burness (eds.), A Reader in the Language of Shakespearian Drama, 339-345. Amsterdam: John Benjamins. First published in Journal of English Linguistics 7: 50-56 [1973]. Smith, Jeremy 1996 An Historical Study of English. London/New York: Routledge. Stein, Dieter 1987 At the crossroads of philology, linguistics and semiotics: Notes on the replacement of th by s in the third person singular in English. English Studies 68: 4 0 6 ^ 3 1 .

Supralocalisation and the rise of Standard English

371

Stone, Lawrence 1966 Social mobility in England 1500-1700. Past and Present 33: 16-55. Stow, John 1598 A Survay of London. London: J. Wolfe. Taavitsainen, Irma forthcoming Scientific language and standardisation 1375-1550. In: Laura Wright (ed.), Standardisation. Cambridge: Cambridge University Press. Tieken-Boon van Ostade, Ingrid 1994 "After a copie unto me delyverd": Multiple negation in Malory's Morte Darthur. In: Francisco Fernandez, Miguel Fuster and Juan Josö Calvo (eds.), English Historical Linguistics 1992, 353-364. Amsterdam/New York: John Benjamins. Traugott, Elizabeth Closs 1972 A History of English Syntax. New York: Holt, Rinehart and Winston. Trudgill, Peter 1986 Dialects in Contact. Oxford: Blackwell. 1998 World Englishes: Convergence or divergence? In: Hans Lindquist, Staffan Klintborg, Magnus Levin and Maria Estling (eds.), The Major Varieties of English, Papers from MAVEN 97, 29-34. (Acta Wexionensia, Humanities 1.) Växjö: Växjö University. Upton, Clive and John D. A. Widdowson 1996 An Atlas of English Dialects. Oxford: Oxford University Press. Wareing, John 1980 Changes in the geographical distribution of the recruitment of apprentices to the London companies 1486-1750. Journal of Historical Geography 6: 241-249. Wolfram, Walt 1991 The linguistic variable: Fact and fantasy. American Speech 66 (1): 22-32. Wright, Joseph 1905 The English Dialect Grammar. Oxford/London/Glasgow: Frowde. Wrigley, E.A. 1967 A simple model of London's importance in changing English society and economy 1650-1750. Past and Present 37: 44-70. Wyld, Henry Cecil 1936 A History of Modern Colloquial English. Oxford: Blackwell. First edition [1920], Yungblut, Laura Hunt 1996 Strangers Settled Here amongst Us: Policies, Perception and the Presence of Aliens in Elizabethan England. London/New York: Routledge.

The rise and fall of periphrastic DO in early Modern English, or "Howe the Scotts will declare themselves Arja Nurmi

1.

Introduction

This paper argues that Ellegärd's (1953) traditionally accepted timing of the spread and regulation of the periphrastic auxiliary DO is in some respects misleading. Unlike Ellegärd, I time the decline of DO in affirmative statements to the first decades of the seventeenth century, and the beginning of the regulation process of DO in negative statements to the following decades. Because of this new timing I am able to offer a new hypothesis on the reasons for the beginning of the disappearance of DO from affirmative statements in the evolving standard English, namely a Scottish influence. While it is unlikely that any single factor can explain this development, it seems highly probable that the sudden influential presence of a mostly DOless group of people at court and in London played a part in the decline of DO. It is only after this drop that the regulation of the NICE qualities (i.e. the use of DO in Negations, Inversions, Code and Emphasis) began. The new dating for this development is based on the Corpus of Early English Correspondence (CEEC).2 This corpus is superior to Ellegärd's materials in two ways. Firstly, it consists of one clearly defined genre, personal correspondence,3 while Ellegärd's text types were in no way stratified and could in fact differ considerably from decade to decade. Using the Helsinki Corpus, Rissanen (1991) has shown that the frequency and speed of change of DO varies according to genre: this supports my view of the importance of balanced material. Secondly, the accurate dating of correspondence is easier than the dating of most other genres: more often than not we know the exact date when a letter was written. For many other genres (plays, fiction, even sermons) it is difficult to ascertain the actual date of composition, which could differ considerably from the date of publication. Furthermore, the influence of various editors and publishers along the way is an unknown variable with most genres, and so knowing

374

Ar ja Nurmi

which decade these texts linguistically represent is difficult. That is why I claim that the data I am using offer a more reliable timing for the development of periphrastic DO.

2.

Constructions studied

This study concerns two contexts where periphrastic DO appears: affirmative and negative statements (only statements negated by not are included here). Questions would be an interesting addition to these two, but unfortunately there are too few instances in the CEEC to be meaningfully included. Examples (1) and (2) are drawn from the corpus and illustrate the two sentence types studied. (1)

The plan of all the incroachments about Paules is fully finished. I heear that the measons do begin to mak up that part of the East end y^h (fey fajve demolished, not well, but with uneven courses of stone. (ARUNDEL: 1620, Inigo Jones, 169)4

(2)

the Maid w^ you sentt vnto me Cam to Franakly vpon the 3{h of July she had had a Longe Iorny yet she did well endwere it for she did not shew any dislyke of it nor I yet of Hir (GAWDYL: 1608, Mary Littleton, F. 96)

As it would have been extremely time-consuming to retrieve the whole linguistic variable for affirmative statements in an untagged corpus I have given the results for them in normalised frequencies (instances of affirmative DO per 10,000 words of running text). I am satisfied (see Nurmi 1996: 155-160) that this gives reliable results. Defining the variable for periphrastic DO in affirmative statements is not simple, in any case. It can be argued that the variable consists of instances of DO-use and those affirmative statements where DO could have been used (i.e. simple present and past tenses excluding clauses where the main verb is BE or HAVE). This version of the variable is used by Ellegärd (1953: 156). The other alternative for the variable is to count the percentage of periphrastic constructions in all finite verb forms, as suggested by Frank (1985). See examples (3) and (4) for affirmative constructions applicable

The rise and fall ofperiphrastic

DO in early Modern English

375

for Ellegard's and Frank's variables. In the case of negative statements the whole variable (finite clauses negated by not and containing no other operator than DO) can easily be retrieved from our corpus and so the results are generally in percentages (see also table 1 below). Example (5) shows an instance of the non-use of DO in negative statements. (3 )

When I writt I wept and I marvell it was not perceived. (STUART: 1603, Arabella Stuart, 130)

(4)

then is Garret discharged, for as I take it, his dett was but 7001'. Martin, or William Mills can tell you. (HOLLES: 1625, John Holies, 11,315)

(5)

My Lord, they deale not charitably, Who are too witty in another mans Workes (JONSON: 1605, Ben Jonson, 195)

3.

The turn of the sixteenth and seventeenth centuries

Figure 1 shows the development of periphrastic DO in affirmative statements in the CEEC in the sixteenth and seventeenth centuries (the corresponding tables can be found in Nurmi 1998 and Raumolin-Brunberg and Nurmi 1997). The most remarkable feature of this graph is the very sudden drop in the frequency of DO at the turn of the sixteenth century. Figure 2 shows the corresponding drop in the use of DO in negative statements (based on tables in Nurmi forthcoming b). Unlike DO in affirmatives, the negative construction later recovers from this crisis and starts its rise towards present-day usage. When using the chi-square test, the overall development of DO from 1500-1681 as shown in figures 1 and 2 is statistically highly significant (p : < V >

a. hupä cedelingas / eilen

fremedon

b. op dcet him ceghwylc / dara ymbsittendra

(Beo 3) (Beo 9)

500

Donka Minkova c. is ig ond utfüs, / cedelinges fcer d. ece drihten, /pees pe he Abel slog

(Beo 33) (Beo 108)

Conceptually, this practice has puzzled scholars for 200 years and has produced some correspondingly animated rhetoric. Commenting on the asymmetry between consonant and vowel alliteration, Lass (1995: 143) writes: ... given the specificity of the system elsewhere, the restrictiveness of consonantal alliteration and the laxness of vowel alliteration suggests that the latter might be misinterpreted, swept under the rug, as an unmotivated 'elsewhere' ... This is weak, and may show a facile disrespect for the technical proficiency of the practitioners of a major verse tradition (poor guys, once they had alliteration, vowels were just a locus desperatus). Drawing a parallel between the rigid identity patterns of consonantal alliteration and the practice of vowel alliteration is, indeed, a challenge. If alliteration in verse was based on identity of one or more linguistic properties, the question is: What kind of identity could have licensed vowel alliteration? Formulating the question in terms of phonological sameness presupposes that the Anglo-Saxon scops and scribes did not follow some arbitrary versification rule which allowed them the freedom to mix vowel qualities. Since the latter proposition is not testable — there are no explicit instructions on how to write alliterative verse prior to the circa 1220 treatise by the Icelander Snorri Sturlason (see Möbius 1897), and in any case Snorri's instructions that non-identical vowels alliterate better can only be taken as a post-hoc summary of the Old English tradition — I will assume therefore that the linguistic question is the interesting one, and that a reasonable answer to it will also provide the solution to the otherwise puzzling practice of mixed vowel alliteration in Germanic. There have been four proposals concerning the nature of the identity in Germanic vowel alliteration. Identity in these proposals is described in terms of (I)-(IV) below. All four proposals are well rehearsed in the literature and need not be discussed in detail here.3

Syllable ONSET in the history of English (I)

501

a shared feature [+vocalic]

(II)

diachronic vowel identity

(III)

a null/zero onset

(IV)

a glottal stop preceding the vowel

A shared feature account, based on e.g. IDENT VOCALIC as in (2) below, is so general that it amounts to a restatement o f the problem.4 Candidates

IDENT VOCALIC

IDENT PLACE

© fa] : Tal © Tal : Γ®1

*

© ral : ΓΠ

*

© f y l : ΠΙ

*

Account (II), the reference to diachronic vowel identity, is based on a study by Classen (1913), who established that as many as 75 per cent o f the vowel alliterations in Beowulf

are "identical vowels" in some diachronic

sense (Classen 1913: 64).5 A s commented on repeatedly in the literature, most recently in Lass (1995: 144-145) and Suzuki (1996: 310), Classen's assumed "identity" is of questionable synchronic validity since it relies heavily on reconstructed values. The third approach, Jakobson's (1963), is an attempt to reconcile the structural requirement of onset identity with the troublesome lack of identity of alliterating vowels. To account for this apparent inconsistency, he proposed

a treatment

of

all

stressed

syllable

vowel

onsets

as

incorporating an unmarked (zero) glide, which is structurally akin to a consonant but has no phonetic substance, though it is supposed to be phonologically opposed to its tense (marked) counterpart /h/. His proposed system of alliteration rests on a three-way distinction: consonants, tense glides (h-), and lax glides ( 0 ) . He finds "the expression or suppression of these Anlaut signals in the delivery ... of no significance" (Jakobson 1963: 91). Kiparsky's (1978) proposal is a conceptual extension of the zero onset idea: he accepts a null onset as sufficient to fulfil the structural requirement of identical onsets in alliteration. Accounts ( I ) - ( I I I ) are all unsatisfactory in some way: account ( I ) is too general, ( I I ) is empirically suspect and ( I I I ) is rendered unnecessary by a stronger version of ( I V ) . N o critique of ( I I I ) will be offered here on the

502

Donka Minkova

assumption that if we can identify a sufficiently broad-ranging set of phenomena which weigh in favour of the glottal stop hypothesis, a recourse to a structural zero/null onset as the basis of alliterative identity is obviated. All our reconstructions are ultimately probabilistic, yet if the aggregate data favour, or are compatible with, a probability which rests on a positive, well defined and physically testable entity, surviving to this day, that positive phonetically-based hypothesis is preferable, even though some of the same data might be compatible with a negative/null entity. Accounts (III) and (IV) are mutually exclusive since both hypotheses rely on the same notion of syllable composition and seek to establish identity in the same structural position; an acceptable defence of (IV) makes (III) indeed moot. Moreover, the appeal of the structural zero onset in Old English is confined to the analysis of alliteration, it has not been shown to be independently needed in the phonology, and is prompted solely by the need to identify the basis of identity, while the glottal stop provides that basis and its presence can be traced independently in other phonological phenomena. The essence of account (IV), first proposed by Rapp (1836: 53), is the postulation of a "glottal catch" similar to the Modern German glottal stop occurring before stressed vowels and after a pause. Jakobson (1963: 88) rejected Rapp's hypothesis as ad hoc, and dubbed it a vicious circle, i.e. the glottal stop was never written, it was concocted to justify vowel alliteration, while the only evidence for it is the very assumption that it was necessary for vowel alliteration. Starting with Hammerich (1948), and culminating most recently with Lass (1995), the glottal stop theory for Germanic has been resurrected, the glottal stop being analysed as a continuation of an Indo-European laryngeal (Hammerich 1948: 33, 71), Lass (1995: 147-149). In what follows, I adduce new data and arguments showing that the reconstruction of an obligatory filled onset for Old English is not circular. First, however, a clarification of the notional overlap between three phenomena which will be much discussed in the next sections: the glottal stop, hiatus and elision. The position taken here is that in Old English the glottal stop was always realised in the onset of vowel-initial stressed syllables. The relevance of hiatus and elision in this context is that the insertion of a glottal stop before stressed vowels in (scribal) hiatus environments would prevent pre-vocalic elision. Thus there is a positive correlation between graphic hiatus and the glottal stop, and a negative

Syllable ONSET in the history of English

503

correlation between phonological elision and the glottal stop, as in (3):6 (3)

[+ η s (glottal stop congruent with graphic hiatus ) [+ ?] ~ - V # - # V -

0 -

#v-

(glottal stop incompatible with elision)

2.

Non-alliterative evidence for the glottal stop in Old English

If alliteration were all that we had to go by in the reconstruction of the syllable structure in Old English, the charge of circularity in positing a glottal stop realisation would not be unfounded. The alternatives do, indeed, look feeble, but that does not remove the circularity. The likelihood of an epenthetic glottal stop as a universal solution to preserving the integrity of ONSET helps some: Lombardi (1997: 159) cites an extensive list of languages with pre-vocalic glottal stop epenthesis: Malay, Selayarese, Ilokano, Czech, English, Arabic, Hebrew, German, Kisar, Sundanese, Tamil, Gokana, Tunica ... , adding that the list is by no means exhaustive. For Old English, in addition to the cross-linguistic naturalness of a pre-vocalic glottal stop in stressed syllables, there are three recoverable and fairly good indicators that the glottal catch idea may not have been a figment of the imagination of nineteenth-century philologists. The first argument in support of that idea comes from elision in Old English verse.

2.1. Elision in Old English verse Generally in Old English, the assumption of pre-vocalic elision does not affect the metricality of a verse line; the examples cited in (4) would be metrical with or without the final vowels preceding the alliterating vowelinitial syllables.

(4)

a. Nold(e) eorla hleo / cenige dinga

(Beo 791)

b. swylc(e) oncydpe / ealle gebette

(Beo 830)

504

Donka Minkova c. egeslfc for eorlum / onddcer(e) idese mid

(Beo 1649)

d. onsyn cenig, / ac hyn(e) ecgfornam

(Beo 2772)

e. ellensiocne, / doer he hin(e) cer forlet

(Beo 2787)

These attestations cannot be treated as either positive or negative evidence for the presence of a glottal stop. In contrast to such uninformative lines, however, there are instances in the verse corpus for which the assumption of elision will threaten seriously the acceptability of a verse, as in (5) below, where monosyllabic lange, eacne, dyre, rihte, urne would render the verses three-positional and therefore defective. Notice that the cases cited in (4) and (5) are quite distinct from the cases of elision of abutting unstressed vowels. The latter process occurs freely and is unrelated to the claim that ONSET was obligatory in stressed syllables.7 a. leof landfruma / lange ähte.

(Beo 31)

b. eacne eardas, / da se ellorgäst

(Beo 1621)

c. dyre Fren, / deer hyne Dene slogon

(Beo 2050)

d. recene geryno, / ond rihte ce

(Andreas 1511)

e. purh rihte ce / reccan cudon

(Elene 281)

f. on Urne eard / in becömon

(BM 58)

Elision of an unstressed vowel before another vowel in speech and especially in verse is a well attested cross-linguistic phenomenon. Assuming, as I have been doing throughout this paper, that in comparable syntactic structures a glottal stop is more likely to inhibit elision than a null onset, I believe that the existence of verses like the ones in (5) is a good argument in favour of the glottal stop: elision occurs freely before unstressed vowels, but before stressed vowels it is the glottal stop which safeguards the metricality of the verses because elision is suspended and effectively blocked by the realisation of a glottal stop before major class words such as

Syllable ONSET in the history of English

505

ähte, eard(as), ce, and Iren.8 A general assumption of a null onset in all vowelinitial syllables would not affect the metricality of these verses either, but it would leave the difference in the metrical behaviour of stressed and unstressed syllables with respect to elision unaccounted for. The significance of such examples is highlighted further by the fact that, unlike Old English, Middle English allows pre-vocalic elision freely, though in the syllable-counting verse this evidence can be somewhat oblique in the sense that elision often regulates the number of syllables per line, while the assumption of what the "regular" number is depends on elision. Still, the statistical predominance of identical syllabic count in nonelision environments, and the strict iambic alternation in some Middle English poetic works, support the assumption that the elisions in the examples in (6)—(7) are genuine:9 (6)

(7)

a. Pin blettsinng tunnderrganngenn

(Orrmulum 10661)

b. Forr tunnderrfanngenn Crisstenndom

(Orrmulum 11112)

a. Patt Godess Sune Allmahhtij Godd

(Orrmulum 11042)

b. Here upponn eorperiche

(Orrmulum 12132)

c. Iwis it was ure 03er broker,

(O&N 118)

d. & hold hire eje noperwafrjd,

(O&N 144)

e. vor rijte nipe & for fule onde

(O&N 1096)

f. Ayther were armed on a stede g. How hyt befelle owre eldurs olde

(Sir Tryamowr 1204) (Octavian 0011)

In the history of English verse elision does not occur provably before the end of the twelfth century; The Ormulum and The Owl and the Nightingale are the first compositions in which elision in hiatus is testable on the basis of a stable syllable-counting nature of the verse line.10 In addition to elision in verse, there is substantial evidence of scribal pre-vocalic elision of final

506

Donka Minkova

-e in Middle English prose, suggesting that elision was enabled by the absence of a glottal stop. It developed into a hiatus-avoidance strategy which became an important component of the phonological history of the unstressed vowels (Minkova 1991: 62-69). The parallels between Old and Middle English with respect to elision therefore support a hypothesis of progressive disappearance of the glottal stop in the post-Conquest period.

2.2. Hiatus at morpheme boundaries A second interesting body of evidence in favour of the glottal stop comes from the way in which vowel-final prefixes attach to vowel-initial roots in Old English derivation: the scribes respect the break between the two vowels.11 There are few vowel-final prefixes in Old English: α-Ice-, be-lbi-, ge-, so the number of forms which are potentially informative in this context is limited. The productivity and the length/stress status of a- is uncertain, and its semantic meaning is vague (Kastovsky 1992: 378). The be- and ge- prefixes, also vague, but somewhat more identifiable semantically, are well known for their prosodic invisibility: a correlate which would suggest susceptibility to phonological change — compare the allomorphic relation between the full form of and- and its unstressed form on-}2 Given the premise of low morphemic and prosodic salience, the scribal consistency of the non-elided prefixed forms can be interpreted to mean that the prefix and the root remained separated by, presumably, the glottal stop. The list in (8) illustrates the point; the hyphenation is from Bosworth and Toller (1898). (8) a-cedan a-eargian a-ebbian a-eode a-etan a-idlan a-idlian a-iepan

'lay waste' 'become slothful' 'ebb away' 'happened', pp. 'devour' 'become idele' 'annul' 'lay waste'

a-urnen

'run out' pp.

ge-cehtan ge-cernan ge-cetywan ge-cedelian ge-eahtle ge-edciegan ge-edcucoda ge-edleanend ge-efenlcecung

'subdue' 'gallop' 'show' 'ennoble' 'esteem' 'recall' 'one revived' 'requiter' 'imitation'

Syllable ONSET in the history of English be-ceftan be-cewnian be-ebbian be-eode ge-aned ge-acsian ge-adlian ge-cef(e)nian

'behind, after' 'marry, wed' 'leave aground' 'venerated', pp. 'united' 'discover' 'become ill' 'become evening'

ge-endung ge-openian ge-orwyrped to-amearcian to-cetiecan to-eacan to-efenes to-endebyrdnes

507

death' open' disgraced' assign' increase' in addition' on a level with' order'

The pattern of non-contraction is stable. Some of the items listed above were probably of lower frequency, but frequency cannot be a determining factor: a check in the Dictionary of Old English Corpus13 shows 1120 attestations of ge-end-, 332 of ge-open-, 158 of to-eacan. Since all of these non-contracted prefixes behave metrically as proclitics in Old English verse, i.e. they are subsumed under the strong metrical position and do not constitute part of the dip,14 they are commensurate with the proclitic contractions in hiatus which begin to appear much more frequently in Middle English. Some forms attested in the Ormulum and elsewhere in Middle English are given in (9) (see Jordan [1934] 1974: 151, Jespersen 1909: 276): (9) pallderrmann pangle parrke popre toffrenn

'the alderman' '(of) the angel' 'the arc' 'the other' 'to offer'

dof teken teve bceften toppe

'do way' 'in addition' 'yesterday evening' 'behind, after' 'above' (Kentish)

Jordan ([1934] 1974: 151) views such contractions as a consequence of the "acceleration of the speaking tempo", a phrase repeatedly used by Luick with reference to the causes of Middle English unstressed vowel reduction and schwa loss (Luick [1914-1940] 1964: §§451-452, 456, 602, etc.). Puzzling as this assertion by the great anglicists of the first half of this century may be — after all, we can't assume that post-Conquest speakers of English suddenly felt the urge to talk faster than their predecessors — , it appears that their intuition can be justified in linguistic terms: the contractions in question, licensed by the disappearance of an obligatory syllable ONSET constraint, would physically speed up the production of the

508

Donka Minkova

particular strings of function + lexical words as in (9). The difference between the earlier view and the one proposed here is that the change in syllable structure, the realignment of prosodic boundaries, is seen as the trigger, and not as the effect of the perceived "acceleration". Compounds whose second root begins with a vowel could provide another link to glottal stop insertion. The relevant evidence is limited since what we need to isolate in order to establish the probability of elision within the compound is based on resolution in Old English verse as in (10): (10)

healcerna mcest; / scöp him Heort naman

(Beo 78)

Resolution of *hea-lcer produces an unacceptable three-position verse; the metricality of (Beo 78a) is guaranteed by the fact that healcerna 'of hallhouses' is syllabified as heal.cer.na and has a heavy first syllable for this reason. As noted in Stockwell and Minkova (1997: 64 and footnote 6), this is contrary to Modern English intuitions and is possibly due to the presence of glottal stops as the initial sounds of roots that apparently begin with vowels. Unlike the maximal onset syllable division of Modern English, in compounds such as Old English ceteaca 'addition', cofincel 'little chamber', cwicceht 'livestock', fripap 'peace-oath', frumieldo 'first age', twi-ecge(de) 'two-edge(d)', wynele 'pleasant oil', the first syllables are heavy by virtue of the fact that the post-vocalic consonants in cet, cof, cwic, frip, etc. belong to a separate syllable and count in calculating syllable weight. This is contrary to the rules of syllabification in the classical languages, where word boundaries are ignored: ab oris is metrically a.bo.ris.15 The chances of such maximal syllabification with vowel-initial second roots in Old English would be thwarted by the glottal stop: heal• "fern, though admittedly the evidence is compatible also with an assumption of morphological independence of the compound elements. Ambivalent as the resyllabification evidence for Old English is, in contrast to the behaviour of heal.cerna 'of hall-houses', Middle English does show some instances where the morphological boundaries in compounds with a vowel-initial second root have been ignored, as attested by the history of obscured compounds of the type barn < bere cern 'barley-house', bridal < brydale 'feast, bridal ale-drinking', daisy < dceges eage 'day's eye', smallage < smal ache 'parsley'.16 One additional reason why evidence from hiatus at compound boundaries should be treated with caution but still allowed into the picture, is the left-heavy

Syllable ONSET in the history of English

509

stress pattern of compounds: unlike the roots of prefixed words, the second elements of compounds are prosodically subordinate to the first element, hence the diminished justification for [^-insertion compared to P]insertion in prefixed words. The weakness of the tendency towards blurring of morphological boundaries results from the conflict between the independent morphological status of the two parts of a compound, which appears to be the dominant factor in the history of compounds, and the lesser, but still present factor of rightward stress-subordination.17 Two more considerations are relevant in the context of morphological and syllabic realignment in Middle English. First, the contractions which are consistent with a weakening of the ONSET constraint seem to have spread at a different rate in the different dialects, with the northern and the east Midlands documents preserving uncontracted forms longer than the corresponding southern and west Midlands texts — this geographical distribution suggests that the loss of an obligatory ONSET might have been externally induced.18 Second, chronologically, there is a clear upward curve in the number of recorded "false junctures" for vowel-initial words from late Middle to early Modern English. The list of Middle English examples in (9) can be augmented later by well known instances of morphological reanalysis of the type newt, nickname, and the converse apron, adder, attested from the fourteenth and the fifteenth centuries (see also Salmon 1958: 234). From the evidence cited in Jespersen (1909: 276), it appears that elision across morphological boundaries reached its peak during the eighteenth centuiy, after which the tide turned — the effect of morphological faithfulness which correlates well with an external set of factors: education, standardisation, literacy, the non-linguistic forces that were so powerful in the late eighteenth and nineteenth centuries. After all, gout < go out, attested in 1600, and dup < do up, attested in 1685, are as good as doff and don: the survival of some such forms against the demise of others must be an issue of frequency, fashion, education.

2.3. The inorganic Another potential piece of evidence, which, though available in earlier scholarship, has not been interpreted explicitly in terms of syllable structure constraints for Old English, is supplied by the pre-vocalic insertions and omissions of the letter in the Old English manuscripts.

510

Donka Minkova

My data is based on Scragg (1970), who reports comprehensively on the use of initially before a vowel in both stressed and unstressed items, and in the onset of a stressed syllable after a prefix. 19 The study presents impressive scribal and other arguments against an assumption of "instability" of the glottal fricative in Old English, concluding that the vagaries of spelling cannot be equated with /h-/ loss in prevocalic position. Since the scribal evidence weighs so heavily on the side of continuous stability of the prevocalic /h/ in Old English, the forms identified as deviant make an instructive point with reference to syllable boundaries. The data shown in table 1 is drawn from Scragg's narrative (1970: 167-182) and the attestations cited in Schlemilch (1913: 51-52); since my focus is on stressed syllable onsets, I have excluded pronominal forms and conjunctions.20 Table 1. Inorganic in Old English Epinal Glossary Erfurt Glossary Corpus Glossary VP Gloss Bede Vesp. D Orosius CP (MS Hatton 20) Parker Chronicle Junius MS Vercelli MS 21 Exeter Book Beowulf Translation of Benedictine Rule22 WS Gospels vElfric Glosses to Aldhelm 24 Total

Omissions 2(-l?) 1 -

1 1 -

1 -

1(?) 5 8 1 -

1 22

Insertions 3 1 1(?) -

7 4 -

7 3 -

8 8 9 3 4 23 8 5 71

The discrepancy in the frequency of attested omissions versus insertions is striking; it cannot be logically correlated with phonological "weakness" in terms of consonantal strength, or to any other phonological properties of the /h-/. Moreover, if the unevenness of the scribal evidence was of phonological significance, one would expect to find a much higher proportion of the instability data to be associated with words of low prosodic prominence, which is,

Syllable

ONSET in the history of English

511

surprisingly, not the case in most of the texts, see Scragg (1970: 190). There must be some other reason for both the higher number of insertions and omissions before stressed syllables, and for the lopsidedness of the distribution of omissions versus insertions of stressed syllable-initial 's. The first type of imbalance was noticed and discussed by Sweet ([1876] 1908: 49) whose interpretation is that since prevocalic /h-/ was silent in Latin, "in the oldest texts h is sometimes wrongly prefixed to native words, and is also used to show hiatus". Pointing out that such insertions occur with considerable frequency in "sentence-stressed words" in non-Latinate texts, van Langenhove (1923: 16-17) states that "in these instances h was not a mere symbol for the eye, but represented an actual sound — a sound occasional of course inasmuch as the emphatic pronunciation of the word was itself but occasional". The second type of imbalance is observed by Scragg (1970: 186, 192), who suggests that this development in the written language is connected to the use of in late Latin as a diacritic signifying hiatus. My proposal shares with van Langenhove the idea that inserted 's are an attempt to represent an actual sandhi glottal stop, though I see the insertion as a predictable consequence of the structural requirement for a filled onset, and not as a correlate of "emphatic pronunciation" (van Langenhove 1923: 16). Sweet and Scragg, on the other hand, were right to invoke Latin scribal habits to account for the use of in these cases, though in my account the is more than "a diacritic signifying hiatus" (Scragg 1970: 186, 192): in the absence of a special symbol for it is likely that the scribes would occasionally insert as a convenient way of marking the obligatory sandhi in stressed vocalic onsets. By definition, a proof of what the scribes really intended is beyond reach. Some additional considerations point in the direction of scribal as a glottal stop marker: First, as is clear from the data in table 1, the number of relevant attestations of insertions in verse is relatively higher than that in the prose documents: a scribe recording an alliterative poem would be more sensitive to the presence of the phonetic signals associated with prominent stress. 25 Second, if insertion was just a hiatus marker anywhere, including hiatus between two unstressed vowels, and not an attempt to highlight a stressed syllable onset, one would expect a goodly proportion of unstressed vowel-initial words with an excrescent . However, as mentioned above, the number of insertions in words of low prosodic prominence is unimpressive: proportionally there is a much more

512

Donka Minkova

considerable number of unetymological 's in hiatus before a stressed syllable, as in the examples in (11):

(11) ahebbad geharn gehendod gehiht gehyppe gehywdest

p.part of 'to ebb' 'ran' 'ended' 'increased' 'disclose' 'thou showed'

gehydlect gihiodun gihalda pa hyfelan pa hilcan feonda handweald

'repeated' 'they went' 'hold' 'the evil' 26 'the same' 'enemy power'

To summarise section 2: The argument that Old English stressed syllables required a filled onset is not circular. The original idea linking vowel alliteration in verse with the realisation of a prevocalic glottal stop is supported by a good deal of solid empirical and inductive support from language typology, from elision in Old English verse, from the observance of morphological boundaries, and from irregular spellings. As for why it is the glottal stop and not some other segment that should be reconstructed, the arguments are (i), that it can be seen as the natural continuation of the Indo-European laryngeals and (ii) that it is the observed placeless epenthetic consonant cross-linguistically (see Lombardi 1997, Ortmann 1998). Finally, turning to phonetics, as has been suggested by Donca Steriade (personal communication), [?] is favored epenthetically and disfavored underlyingly for a single reason, low perceptual salience. 27 This choice is fully consistent with the fact that the epenthesis is only functional in the onsets of stressed syllables; it cannot be reconstructed reliably elsewhere, i.e. stressed syllable onsets by being highly perceptually salient compensate for the deficiency of the glottal stop and allow it to be functional.

3. The upset of

ONSET

Before we move on to a discussion of the phonological properties of ONSET in Modern English, here are some ways in which the intermediate history of syllable structure can be reconstructed.

Syllable ONSET in the history of English

513

3.1. Middle English vowel alliteration In a London dissertation written 106 years ago, Lawrence (1893) made an interesting remark concerning vowel alliteration in Middle English: he described it as being "evidently in a moribund condition" (Lawrence 1893: 109 [emphasis mine]). His inference was prompted by the fact that the Middle English poetic material he studied showed a sharp drop in the proportion of vowel alliteration to consonantal alliteration compared to Old English. Schumacher (1914: 44-56) expanded the scope of Lawrence's data base. The list in (12), showing the proportion of alliteration that is vocalic, combines information from both sources:

(12) Beowulf SGGK St. Erkenwalde Piers Plowman A Piers Plowman Β Piers Plowman C

15.5% 4.8% 1.1% 3.6% 3.8% 3.9%

Wynnere and Wastoure Alexander A Alexander Β William of Palerne The Destruction of Troy The Sege of Jerusalem

2.1% 3.7% 5.0% 2.3%28 5.0% 2.0%

The decrease is from 15.5 per cent in Beowulf to an overall three per cent in fourteenth-century verse. In addition to the general avoidance of vowel alliteration, there is also a manifest tendency to alliterate on identical vowels. The interpretation of these figures is riddled with questions that go beyond my focus: changing conventions, textual transmission, literacy, the disassociation of alliteration and stress in Middle English. Within the framework adopted here, however, which rests on the assumption that change of poetic form is linked to linguistic change, such data is strongly symptomatic of the poets' and scribes' decreased confidence in the appropriateness of alliteration on vowels of different quality. Within my proposal, the new practice emerged due to a typological change in the syllabic structure of the language under the influence of Anglo-Norman loan phonology: from being obligatory in Old English, glottal stop epenthesis had become optional in Middle English. With regard to the data in (12), the glottal stop hypothesis provides an account of the diachronic alliterative facts, or at least conforms to them. In comparison, its strongest theoretical rival, the hypothesis that empty onsets provided the common

514

Donka Minkova

denominator in alliteration in the history of English, cannot accommodate the same data by reference to the properties of the linguistic system, and would have to relegate the explanation for the statistically significant drop of vowel alliteration rates in Middle English to alternative extralinguistic factors. In anticipation of the formalisation in section 4, we can say that the evidence in ( 1 2 ) indicates a reranking of the ONSET constraint with respect to faithfulness to the base. Two factors facilitate the abandonment of an obligatory ONSET in Middle English: first, the glottal stop had never been part of the contrastive, underlying phonological system of the language — it was always realised as an output epenthetic segment. Second, the model for a lower-ranked ONSET was provided by scores of newly borrowed vowelinitial words: actor, aim, air, April, arch, eagle, ease, emperor, envy, err, odour, olive, optic, orange, order, etc.

3.2. Middle English elision alliteration: Stab der Liaison29 Further prosodic evidence for Middle English is provided by an intriguing alliterative practice, referred to as Stab der Liaison, illustrated in (13):30 (13) a. An oper noyse fulnewe

\ nejed biliue (SGGK 132)

Pat schal I telle pe trwly,' \ quop pat oper penne (SGGK 2444) And non eire ofmyn own \ neuer yet I hadde (Destr. of Troy 5315) Toax, a tide mon | of pat oper side (Destr. of Troy 6805) If this were nedles note \ anothir comes aftir (Wynnere and Wastoure 338) On the assumption that resyllabification occurs if the host stressed syllable has a surface zero onset, the practice illustrated in (13a) could not have

Syllable ONSET in the history of English

515

arisen if prevocalic glottal stops had been a feature of the language of the poets. There are no comparable instances in the Old English alliterative corpus; the negative evidence, as in (13b) below, is easy to trace in numerous examples which show that e.g. the definite articles, the demonstratives se (masculine), seo (feminine), poet (neuter) and pa (plural), and their inflexional forms pees, pcem, peer(e), could not have been resyllabified without garbling the alliteration: öcet se ecghete / äöumsweoran

(Beo 84)

Base ellengcest /earfopltce

(Beo 86)

in poet orlege / unryhte swealg. fsigfepera; /ful oft poet earn bigeal, pa he pees eorles / earm ämyrde.

(Crist 560) (Sea 23) (Maid 165)

On the other hand, the evidence in (13a) suggests strongly that though obligatory ONSET was no longer part of the phonology of Middle English, and there was no requirement for glottal stop epenthesis in vowel-initial stressed syllables, a filled onset was still a highly desirable option. Unlike the satisfaction of ONSET through glottal stop insertion, which optimises the syllable structure without compromising syllable and morpheme boundaries, an oper syllabified maximally as a.no per, is also a better phonological structure than att.o per, but it violates the morphology-prosody alignment. Not surprisingly, the non-alignment occurs in the context of potentially proclitic weakly stressed words: an, at, pat, myn, pyn, his, etc., where the damage done to the morphosyntactic bracketing is minimal. Again, the process illustrated in (13a) results in a filled onset, but the chances of resyllabification must have been enhanced by the absence of a prevocalic glottal stop; in other words (13a) testifies both to the violability of the ONSET constraint, and to its continuing relevance to the phonology of Middle English. Stab der Liaison is not pervasive in the alliterative corpus of the period, but it is a good indicator of the flexible status of ONSET with respect to prosodification of the group function word + host.

516

Donka Minkova

3.3. Non-verse evidence for ONSET in Middle English A special Middle English phonological development which has direct bearing on the issue of ONSET, is the narrowing of the first component of Old English ea, accompanied by loss of sonority and transfer of accent. The spellings in table 2 attest to this specifically Kentish development, which is assumed to have taken place after the eleventh century (Jordan 1974: 109). Table 2. Kentish prevocalic glide epenthesis Old English Middle English (Kentish) eadij

eare

east

eastre

eap

Ead (Devonshire) According to the philological accounts, the reconstructed /je:/ remained most stable in initial position; in medial position there is one single example of that type, the word dead 'dead', spelled . There is no evidence for the /j/ in other words where the ea is non-initial. Traditionally, all of these examples have been treated as instances of Akzentumsprung, but as the data and the analysis in Stockwell (1991: 317) show, the phenomena subsumed under this rather large umbrella are much better characterised for English with reference to assimilation. Since the assimilatory account cannot apply to the de-vocalisation in table 2, which looks more like a word-initial epenthesis, Stockwell (1991: 319 [addition mine]) writes: 1 have no explanation for the development of rising diphthongs in word-initial position either in these examples or, for that matter, all over Icelandic and Frisian. They seem to me to represent a genuinely different phonetic process which only superficially resembles the phonetic assimilations [(1) front vowel absorption by preceding palatals and (2), back glide nucleation by following back glides] exemplified in (1) and (2) above. It is not uncommon for grammars to scatter evidence of various constraints across different subsystems and varieties of language. If indeed the Kentish

Syllable ONSET in the history of English

517

developments cited in table 2 represent a genuinely different phonetic process, and if they can be attributed to the continuing preference for a filled onset, as I would like to argue, then the observed discrepancy between word-initial position and medial position is no longer a puzzle; it is a logical prediction of the account. The assumption of a persistent ONSET constraint in Middle English is further evidenced by the spellings in (14): 31 (14)







'us' 'oak' 'old' 'Ulster' 'only' 'once' oar 'oats' 'oath' 'other' C

ΐ







'egg' 'eel' 'end' 'any' 'ear' 'earth' 'evening' 'also' 'easy' 'even'

These forms have so far remained problematic, in the sense that they do not fit any general patterns describing the overall properties and directions in the phonological history of English. Factoring in ONSET as a continuing, though violable, constraint on the structure of the stressed syllable, accounts for such epenthetic phenomena in a natural way. Jones (1989: 177-181) comes closest to recognising the importance of the epenthetic segments in terms of a continuing structural constraint on the stressed syllable in English. He points out, rightly, that accent shifting is merely a non-explanatory label, and that "glide epenthesis" can occur in instances where the syllable-bearing peak remains intact. 32 The approach advanced here differs from his in two significant ways. Jones believes that the epenthesis is a manifestation of a stratagem to highlight and foreground the complex (bimoraic) vowel space in any environment. I focus on ONSET as a universally available constraint on syllable structure; its effect on the output can be seen both with monomoraic and bimoraic peaks, though, admittedly, the epenthesis effect is perceived as stronger with bimoraic peaks — an extraneous consequence of the higher statistical probability of the co-occurrence of stress and bimoraicness in Middle and Modern

518

Donka Minkova

English. Also, Jones relates the phenomenon of epenthesis to the stability, and even increase of prominence of complex peaks in the later language, whereas in the analysis advocated here, the probability of ONSET-induced epenthesis in English has been diminishing.

4.

ONSET

in Modern English

ONSET does not contribute to the calculation of syllable weight, it is therefore routinely ignored in describing the prosodic properties of English, diachronically and synchronically. The evidence for consonantal epenthesis in vowel-initial native lexical items in table 2 and (14) above, and the stress-related considerations put forth in Jones (1989), raise the question of the correlation between ONSET and stress in English. I have made no reference to the behaviour of ONSET in unstressed syllables; the focus on stressed syllables must be defended. In the remaining sections of this paper I want to show that incorporating ONSET into the prosodic history of English is revealing both for the historical phonologist and from the point of view of the Modern language, and that the link between a filled onset and stress, which is essential to the coherence of the proposed historical account, is not an incidental correlate of syllable geometry.

4.1. The correlation between ONSET and stress in Modern English. From the familiar observation that a filled ONSET is universally preferred, it does not necessarily follow that the presence of a prevocalic consonant will be more compelling in stressed syllables than in unstressed ones. Throughout this discussion, I have been focussing on the fate of ONSET in stressed syllables. None of the arguments supporting the idea of an obligatory onset in stressed syllables in Old English will go through for unstressed syllables. Similarly, the onset traces tracked in section 3 are phenomena confined to stressed syllables. In principle, this is not surprising: stressed and unstressed syllables exhibit quite distinct phonological properties in many other respects. Against the background of ONSET being treated as the poor relative of other syllable properties, however, it is worth looking at some Modern English data which support

Syllable ONSET in the history of English

519

the idea that positing co-occurrence of ONSET and stress historically is well justified. Here I will draw on data from Hammond (1995), who establishes, from a database of 20,000 English words, a significant correlation between the distribution of stress and word-initial onsets in disyllabic words. According to Hammond's calculations, zero-onset stressed syllables are only nine per cent of the 3,949 trochaic words counted, the able versus gable type, while zero-onset unstressed syllables are 37 per cent of the 1,416 iambic words counted, the about versus garage type, i.e. unstressed syllables tolerate zero onset much better than stressed syllables, or stressed syllables prefer onsets.33 (15)

V521 342

CV895 3607

Trochaic Iambic

% V37 9

Hammond comments that "This statistical skewing is repeated in words of all lengths". The chart in (16) shows the distribution of initial stress as a function of whether the word begins with a vowel or a consonant-vowel sequence. The column labeled "%" shows the percentage of stressless syllables for each length and segmental condition. In contrast to the marked preference for CV- structure for stressed syllables, initial stressless syllables consistently form a greater percentage of the total in the vowelinitial condition than in the consonant-initial condition: 48 per cent versus 17 per cent for disyllabic words, 42 per cent versus 28 per cent for trisyllabic words etc.

# of a ' s 2 3 4 5 6

CV (',σ) 3919 1945 346 28 0

V 790 761 663 67 10

% 17 28 66 71 100

CJ

Σ

556 599 116 10 0

)

(σ,-) 521 437 400 76 10

% 48 42 78 88 100

The results support the exclusion of unstressed syllables from the account; they are also a corollary of the hypothesis that only stressed vowel-initial syllables in Old English required [^-epenthesis. We will never know

520

Donka Minkova

syllables in Old English required P]-epenthesis. We will never know whether the insertion was sensitive to register and/or rate of delivery, but the odds are in favour of its surface presence in careful speech and especially in verse recitation.

4.2. Alignment: a historical

perspective

One of the important prosodic issues related to positing an obligatory ONSET in Old English and a subsequent demotion of that constraint is the issue of morphology-prosody alignment. A comparison of the Old English data in (8) (a-etan 'devour', be-cefian 'after', be-ebbian 'leave aground', etc.) and (11) (ahebbad past participle of 'to ebb', gehended 'ended', etc.) with the Middle English contractions in (9) (bceften 'after', toffren 'to offer', teken 'in addition', etc.) and the Stab der Liaison in (13a) makes it clear that similar or equivalent morphosyntactic sequences must have been prosodified differently in the history of English. At first sight it may appear strange to assume that English has allowed two types of alignments historically. However, ALIGN, one of the basic components of Universal Grammar, does not have to be a constraint by itself, but simply a schema for creating language or time-specific constraints; see Russell (1997: 119). The historical data presented in this study indicate that in Old English the edge of a grammatical category Lex (Lexical Word) is aligned with the prosodic category PrW (Prosodic Word). 34 As shown in figure 1, except for the negative proclitic ne, in Old English most affixes and function words are located outside the Prosodic Word — they are free clitics. Initial position in Prosodic Word is associated with certain phonetic phenomena such as voiceless stop aspiration in Modern English, or, in the Old English parallel, glottal stop epenthesis. By deterring resyllabification within the Prosodic Phrase, glottal stop epenthesis becomes also a diagnostic for the type of morphology-prosody alignment in the earliest stages of English. In figure 1 the Prosodic Phrase (PPh) dominates the Prosodic Word (PrW) directly; all arguments adduced in section 2 indicate that it is the appropriate representation for Old English.

Syllable ONSET in the history of English

521

PPh

/

σ I1

abetoon min

\ PrW I1 etan ebbian eacan Englisc ende

'devour' 'ebb away' 'in addition' 'in English' 'my end'

Figure 1. The prosodic phrase in Old English Selkirk (1996: 441) describes three different types of clitics corresponding to the first element in a syntactic phrase (Fnc Lex): free clitic, where the function word is sister to Prosodic Word and daughter to the Phonological Phrase; internal clitic, where the clitic is dominated by the same Prosodic Word that dominates its sister lexical word; and affixed clitic, where the function word is located in a nested Prosodic Word structure. The type of alignment involving free clitics as in figure 1 is argued by Selkirk (1996: 452) to be the optimal grammatical representation of non-final function words in Modern English. In Middle English, on the other hand, the evidence suggests that the prosodification of the same morphosyntactic sequences has changed as shown in figure 2. Here the boundaries of Prosodic Phrase and the higher level recursion of the Prosodic Word coincide. The coextensiveness of the two domains renders the prefixes and function words both PrW-initial and PPh-initial, allowing greater prosodic cohesion within the syntactic phrase (Fnc Lex). More importantly for the fate of ONSET, the fact that both sister nodes are internal to the Prosodic Word at a lower level licenses resyllabification, which satisfies ONSET, but the price of the satisfaction is more complex prosodic nesting, a violation of the independent one-to-one alignment between lexical and prosodic word at one prosodic level. Thus figure 2 below highlights the conflict between alignment and ONSET in Middle English: ONSET, though no longer inviolable, is sufficiently strong to trigger internal clitic prosodification; prosodically, the function words have moved away from acting like independent words, the state of "wordhood", towards the dependent state of "affixhood".

522

Donka Minkova

PPh PrW

/\

σ

t(o) t(o) a(n) the

PrW

eken eve other other ende



Figure 2. The prosodic phrase in Middle English

The alignment in figure 3, which has the same geometry as figure 1, is the appropriate structure for Modern English, as attested by the absence of resyllabification/aspiration: PPh

σ

PrW

that let eight

object enter answers

*tha t h object *le t h enter *eigh t h answers

Figure 3. The prosodic phrase in Modern English

Without going into further detail, we can note that the Middle English structure of nested Prosodic Word and internal cliticisation has not been completely abandoned; it persists in the prosodification of Modern English object pronouns, which present a special case of morphosyntactic cliticisation, as shown by Selkirk (1996: 358). Here belong pairs of the type:35 (17)

feed us see you will it

= Ξ =

fetus Mia billet

[fills] [mija] [bilit]

Syllable ONSET in the history of English

4.3. ONSET

versus Faithfulness

(DEP-IO)

523

in the history of English

The three stages in the development of ONSET outlined here lend themselves to a formalisation within the framework of Optimality Theory. The survey of the diagnostic ONSET facts in the history of English was conducted with reference to the constraints listed in (18): (18)

ONSET: Syllables must have onsets. 36 DEP i0 : Every segment of the output has a correspondent in the input. 37 ALIGN-L (LEX, PRW): Align the left edge of the lexical word with the left edge of the Prosodic Word. The alignment leaves the "free" clitic outside the domain of the Prosodic Word. 38 STRICT LAYERING: No PPH immediately dominates a syllable. 39

Old English ranked the constraints as in (19). (19) Old English Candidates

ONSET

ALIGN-L

STRICT

(LEX,

LAYERING

DEPJO

PRW)

© öing © ?eorl

*

Θ eorl © to - 9 eacan

I*

Θ to - eacan Θ *teacan

!*

® *o.nEnglisc © on ^Englisc

*

*

* ,

!* !* *

*

In Old English, ONSET for stressed syllables was undominated whereas input-output faithfulness was ranked lower: ONS » DEP io . The alignment type consonant with the Old English evidence is one of simple lexical word edge to Prosodic Word edge: ALIGN-L (LEX, PRW), which also appears to be very strong within this set of parameters. There is some difficulty in

524

Donka Minkova

ranking ONSET with respect to ALIGN-L (LEX, PRW): if ONSET

»

ALIGN-L (LEX, PRW), P]-epenthesis would occur first and would block resyllabification, which would then guarantee the desirable alignment. If, on the other hand, ALIGN-L (LEX, PRW) » ONSET, there is nothing to prevent epenthetic satisfaction of ONSET, so a question may be raised whether the two constraints need to be ranked with respect to each other.40 The ranking, as shown in (19), is chosen following the logic that if resyllabification could satisfy ONSET in Old English by filling it with any initial consonant, as in *o.nEnglisc, one should be able to establish some textual or metrical support for that process comparable to that found generously in Middle English. As shown in (13b), the broader version of alignment, allowing (re)syllabification within the domain of the Prosodic Phrase, though logically available in Old English, is not attested. The absence of fully contracted forms outside of the negative clitics (ncefre < ne + cefre 'never', ncenig < ne + cenig 'no(ne)', but *teacan < to-eacan 'in addition') helps with the ranking of ONSET with respect to the STRICT LAYERING constraint,

thus ONSET »

STRICT LAYERING.

Moreover,

assuming the blocking effect of the glottal stop on elision, the metrical preservation of hiatus in (5) and the non-contractions in (8), support the proposed top ranking of ONSET independently. Two powerful factors are thus at play in the prosodic behaviour of vowel initial stressed syllables in Old English: ONSET and ALIGN-L; the satisfaction of ONSET guarantees

satisfaction of ALIGN-L, but not vice versa. The two constraints are stronger than the requirement for function words to form "internal" clitic groups with the following lexical words.41 The situation with regard to the interaction of constraints is considerably less distinct in Middle English, where the data is more varied. The hierarchy of the constraints in Middle English is obscured by a number of factors, hence the introduction of a third possibility: the indifferent face ©. The data show ambivalence about initial epenthesis, which means that O N S E T has lost its dominant position among the other constraints. D E P I O has gathered strength, and possibly outstripped all other competitors, but forms like woth 'oath' and yende 'end' indicate that it continued to be violable. Contractions and resyllabiflcations of the type another, teken 'in addition', and the Middle English Stab der Liaison argue that what used to be a fatal violation, the misalignment of the lexical and Prosodic Word A L I G N - L ( L E X , PRW), has become acceptable. Both an.other and a.nother

Syllable

ONSET in the history

of English

525

are allowed, but an.other gathers more violations. Crucially, none of the Old English "losers" in (19) can be reconstructed for Middle English; uncomfortably, anything goes. (20) Middle English Candidates

DEPI0

ONSET

STRICT

ALIGN-L

LAYERING

(FNC

LEX,

PPH)

© © © © © ©

thinq oath ?oath(?) woth 'oath' an . other a. nother

* * *

*

* *

(21) Modern English Candidates

DEPJO

ALIGN-L

ONSET

(LEX,

STRICT LAYERING

PRW)

© © © © © Θ

thinq oath under oath unde. roath that object tha. t h object

* *

*

*

*

*

1*

The options in Modern English are more limited than in Middle English, at least with respect to ^syllabification not involving /r/.42 The tableau for Modern English ignores the possibility of glottal stop epenthesis since such epenthesis is limited to some varieties and registers. In any case, including 9 oath-type forms would not affect the main-line argument beyond providing local support for the perseverance of a more robust ONSET. The tableaux for Middle and Modern English show the interaction between constraints in the history of post-Conquest English: the violable

526

Donka Minkova

and low-ranked ALIGN-L of Middle English is elevated in the constraint hierarchy of Modern English. ONSET and STRICT LAYERING, on the other

hand, drop further down in the hierarchy. As in Middle English, the relative ranking of the latter two constraints remains indeterminate. Cliticisation occurs independently of the requirement for a filled onset, but arguably a violation of ALIGN-L (by "internal" cliticisation and Irl ^syllabification) is tolerated only if both lower-ranked ONSET and STRICT LAYERING are

jointly satisfied.43 In sum, the Optimality Theory analysis produces the following picture: The universal requirement for a filled onset changes historically in English, but this prosodic well-formedness constraint cannot be traced in isolation. ONSET interacts with DEPJO, but the morphosyntax-prosody alignment matters too, and so does the nature of the clitic formation in terms of prosodic domination. To account for the observed changes, we need to refer to two types of clitics: free clitics and internal clitics. Free clitics violate prosodic STRICT LAYERING, internal clitics violate ALIGN-L. The tolerance for constraint violations will vary depending on the phonetic nature of the segments at the juncture, on the degree of cliticisation, on rate of speech, on style of delivery, on dialect, and possibly on other factors. Why more cases of internal procliticisation such as thende 'the end', toffer 'to offer' have not survived in the standard language,44 and why the nuncle