Studies in the History of the English Language: A Millennial Perspective [Reprint 2011 ed.] 9783110197143, 9783110173680

Citation preview

Studies in the History of the English Language

≥

Topics in English Linguistics 39

Editors

Elizabeth Closs Traugott Bernd Kortmann

Mouton de Gruyter Berlin · New York

Studies in the History of the English Language A Millennial Perspective

Edited by

Donka Minkova Robert Stockwell

Mouton de Gruyter Berlin · New York 2002

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

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication-Data Studies in the history of the English language : a millennial perspective / edited by Donka Minkova, Robert Stockwell. p. cm. ⫺ (Topics in English linguistics ; 39) Includes bibliographical references and index. ISBN 3 11 017368 9 1. English language ⫺ History. 2. English language ⫺ Grammar, Historical. I. Minkova, Donka, 1944⫺ II. Stockwell, Robert P. III. Series. PE1075 .S88 2002 4201.9⫺dc21 2002067795

Die Deutsche Bibliothek ⫺ Cataloging-in-Publication-Data Studies in the history of the English language : a millennial perspective / ed. by Donka Minkova ; Robert Stockwell. ⫺ Berlin ; New York : Mouton de Gruyter, 2002 (Topics in English linguistics ; 39) ISBN 3-11-017368-9

” Copyright 2002 by Walter de Gruyter GmbH & Co. KG, 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. Printed in Germany.

Table of Contents

Foreword

1

Millennial perspectives From etymology to historical pragmatics Elizabeth Closs Traugott

19

Mixed-language texts as data and evidence in English historical linguistics Herbert Schendl

51

Dialectology and the history of the English language William A. Kretzschmar, Jr.

79

Origin unknown Anatoly Liberman

109

Issues for a new history of English prosody Thomas Cable

125

Chaucer: Folk poet or littérateur? Gilbert Youmans and Xingzhong Li

153

A rejoinder to Youmans and Li Thomas Cable

177

Phonology and metrics On the development of English r Blaine Erickson

183

Vowel variation in English rhyme Kristin Hanson

207

Lexical diffusion and competing analyses of sound change Betty S. Phillips

231

vi

Table of Contents

Dating criteria for Old English poems Geoffrey Russom

245

How much shifting actually occurred in the historical English vowel shift? Robert Stockwell

267

Restoration of /a/ revisited David White

283

Morphosyntax/Semantics Pragmatic uses of SHALL future constructions in Early Modern English Maurizio Gotti

301

Explaining the creation of reflexive pronouns in English Edward L. Keenan

325

Word order in Old English prose and poetry: The position of finite verb and adverbs Ans van Kemenade

355

The “have” perfect in Old English: How close was it to the Modern English perfect? Jeong-Hoon Lee

373

Reporting direct speech in Early Modern slander depositions Colette Moore

399

The emergence of the verb-verb compound in twentieth century English and twentieth century linguistics Benji Wald and Lawrence Besserman 417

Envoy A thousand years of the history of English Richard W. Bailey

449

Name index Subject index

473 483

Foreword This collection contains papers selected from those presented at a conference at UCLA in the Spring of the year 2000. The conference was called Studies in the History of the English Language, abbreviated SHEL-1. It was intended to be the first in what we hope will become a regular biennial series at various sites in North America; SHEL-2 is being organized at the University of Washington as we are preparing this volume. The intention of the series is to stimulate research and other scholarly activity in the field of historical English linguistics. Our emphasis was deliberately on the history of English as a discipline: how healthy was it at the end of the millennium and what if anything needed to be done to maintain its scholarly energy and relevance? A comparable series of meetings in Europe, the International Conferences on English Historical Linguistics (ICEHL), have proved to be very vigorous since the first conference, in 1979, at Durham. The scholars who work in our field in non-English speaking countries have the advantage that English is a popular choice for foreign language study there and often there is generous government and public support for this kind of enterprise. They have a steady student demand based on that fact. In Britain, the 19th and 20th century dedication to historical English studies created a robust research tradition and a new generation of outstanding scholars in whose hands the field is thriving. The tradition flourished in British and American universities in the 19th century and during the first half of the 20th century, but during the last two decades only in British universities has this kind of scholarship remained in the mainstream of academic life. It has been apparent for some years that in America comparable vigor did not exist in this field. In fact, the field has been declining as scholars unfamiliar with the linguistic history of English themselves fail to see the relevance of a traditional subject in a newly fashioned humanistic curriculum. We believe this to be a misapprehension of the field and its present-day involvement with other disciplines. Researching the cognitive and social conditions, causes, and consequences of language change should not be brushed aside as peripheral to the concerns of the new generation of humanists in this country. We organized the first SHEL with the conviction that the study of the history of English is central to the interpretation of our cultural and literary heritage, and both the meeting, and the contributions to this volume, reaffirmed our conviction.

2

Foreword

So the first motivation for us to convene this conference was to provide a forum for the presentation of research in English linguistics, specifically English historical linguistics, to serve as a stimulus for quality and a probe for what more needs to be done. Measured by the quality and variety of the presentations and discussions, 35 speakers and over 80 participants, we happily report that energy and vigor continue to characterize our field. A Workshop, organized by Anne Curzan (University of Washington), addressed various approaches to the teaching of the History of English. The North American representation at SHEL-1 was greatly augmented by transatlantic scholars whose participation was deeply appreciated. For the record, a list of paper presenters and session chairs appears at the end of the Foreword. The second motivation was to provide, in as many areas as possible, a sort of millennial stock-taking. The millennial focus is fully apparent in the envoy paper by Richard Bailey (University of Michigan); and in the field-survey papers by Elizabeth Traugott (Stanford University); Herbert Schendl (University of Vienna); William Kretzschmar (University of Georgia); Anatoly Liberman (University of Minnesota); Thomas Cable (University of Texas, Austin); and Gilbert Youmans and Xingzhong Li (University of Missouri and Central Washington University). The papers addressing individual issues fall into two natural groups: (1) Phonology and metrics – the papers by Blaine Erickson (Kanazawa Institute of Technology, Japan); Kristin Hanson (University of California, Berkeley); Betty Phillips (Indiana State University, Terre Haute); Geoffrey Russom (Brown University); Robert Stockwell (University of California, Los Angeles); and David White (University of Texas, Austin). (2) Morphosyntax/Semantics – the papers by Maurizio Gotti (University of Bergamo); Edward Keenan (University of California, Los Angeles); Ans van Kemenade (Katholieke Universiteit, Nijmegen); Jeong-Hoon Lee (University of Texas, Austin); Colette Moore (University of Michigan); and Benji Wald and Lawrence Besserman (Los Angeles and Hebrew University, Jerusalem).

Foreword

3

Millennial Perspective Elizabeth Traugott’s contribution is entitled “From etymology to historical pragmatics”. As is well known, she was one of the principal scholars in the initiation of research into the historical process of grammaticalization, by means of which new grammatical formatives come into languages. In introducing her survey of the field, she notes that many of the major themes of the early twentieth century, e.g. the arbitrariness of the sign, the non-predictability of change, and distinctions between structure and use, were developed against the background of historical work primarily on Romance and Germanic languages. Among many themes debated at the end of the century have been the extent to which change is predictable and non-arbitrary, this time against the background of synchronic work primarily on English, as well as many other languages. She points out that throughout the twentieth century diachronic theory has lagged behind synchronic. Though the seeds of many ideas about morphosyntactic change that dominated the end of the century are to be found in work at its beginning, the pragmatics, semantics and syntax were not sufficiently far advanced for those ideas to be developed in principled ways. New possibilities for understanding language change have opened up with advances in the study of the relation between language and use, especially from the perspective of work on grammaticalization and historical pragmatics. One area of intense debate arising from these new studies is what status in linguistic theory the widely attested unidirectionalities in semantic and morphosyntactic change should have. On the one hand there has been an active research program seeking to identify unidirectionalities (e.g. Traugott 1982, 1989, Heine, Claudi and Hünnemeyer 1991, Bybee, Pagliuca and Perkins 1994); this has recently been complemented by a research program seeking to formalize them (e.g. Roberts 1993, Kemenade 1999). 1 On the other it has been argued that since unidirectionalities are tendencies, not absolute universals, they are epiphenomena and not explanatory (e.g. Roberts 1993, Newmeyer 1998 on the unidirectionalities identified in grammaticalization), and even that such searches are a hold-over from nineteenth century historicism (Lightfoot 1999). Her paper suggests some ways in which historical pragmatics can shed light on the unidirectionalities observed in grammaticalization, as the field moves toward an explanatory theory that can account for them.

4

Foreword

Herbert Schendl is one of the world’s premier experts on the new and relatively unexplored subject of mixed-language texts. His essay here, “Mixed-language texts as data and evidence in English historical linguistics”, essentially lays out the whole field, what has been accomplished and what needs to be done. As he writes, the changing multilingual situation of medieval Britain as well as the foreign influence on the lexicon and structure of English over the centuries have been extensively studied. He highlights the surprising fact that most historical linguists still regard monolingual English texts as the only data base for research and have paid little attention to the numerous older texts showing alternation of languages, especially of Latin, French and English. While the central role of code-switching for sociolinguistics and general linguistics has been established for some time, older mixed-language texts have been seen as literary peculiarities or as the result of insufficient language competence. For the Middle English period it has even been claimed that bilingual texts are more frequent than monolingual ones. Their relative neglect by historical linguists is thus the more surprising. Only in the last few years, Schendl claims, has there emerged an awareness of the relevance of such texts, which explicitly document bilingualism and code-switching in action and thus provide invaluable evidence of the close contact between different languages in the history of English. His paper surveys research in this promising new field of historical linguistics. After a brief discussion of some general aspects of mixed-language texts, the author addresses the earlier literary and philological approaches to such macaronic writings. The main body of the paper is devoted to a discussion of recent linguistic research into these texts from a syntactic and a functional-pragmatic point of view with illustrations from a wide range of text types. Finally, two more general questions are addressed; first, the distinction between code-switching and mixed codes; second, the relationship between code-choice, code-switching, and code-shift. William A. Kretzschmar, Jr.’s paper is entitled “Dialectology and the history of the English language”. In it he argues persuasively that dialectology has much to offer to historical linguistics and to the history of the English language in particular. Survey research is the hallmark of the field, in line with the empirical emphasis of the Neogrammarians who were the initiators of this approach, yet dialectologists have been among the strongest critics of the central Neogrammarian position on mechanical sound

Foreword

5

change. Given this state of affairs, the central role of dialectology is to describe, for the particular time of a survey, the facts of the complicated language behavior in the survey area. These facts make up a body of raw material for historical linguistics to use, first as studies of individual words, and second for description of the situation of language variation today which can enlighten our attempts at historical reconstruction. Kretzschmar maintains that contemporary variation is indeed the trace of change for particular words, because the low-frequency responses often reflect historical precedents. The asymptotic curve (A-curve), the pattern of few common linguistic types and many infrequent types which emerges from survey research on a single linguistic target, appears to be a basic fact about the distribution of linguistic types and tokens. The A-curve gives historical linguistics a more realistic model than the all-or-nothing pattern predicted by mechanical sound change. The A-curve and the familiar S-curve are thus complementary quantitative descriptions of the distributional facts of variant linguistic forms upon which we can develop improved “realism” in accounts of linguistic change. Anatoly Liberman’s brilliantly entertaining, yet deeply scholarly essay, is entitled “Etymology unknown”. The author is engaged in the production of a new etymological dictionary of English that focuses entirely on the mysteries of English etymology – the words for which no adequate etymology exists in any of the standard works like the Oxford English Dictionary. He notes that many words are dismissed in etymological dictionaries with the verdict “origin unknown.” In English, some such words go back to Common Germanic and even Indo-European, while others are more recent. Sometimes, “origin unknown” means that there is no consensus on a word’s prehistory, but equally often scholars indeed have no clue to the sources and the semantic motivation of words current today. Relatively easy to etymologize are only onomatopoeias, words from names, some simplified compounds, and borrowings (if an agreement is reached not to pursue the origin of foreign words in the lending languages). In other cases, historical linguistics usually fails to achieve the avowed goal of etymological research and present an unmotivated sign as motivated. The most urgent task of English etymology as a science, Liberman believes, is to produce a dictionary comparable to those by Feist, Vasmer, Frisk, and others with a full critical survey of the existing literature on every word. If the idea of such a dictionary materializes, many words of allegedly unknown etymol-

6

Foreword

ogy will be upgraded to words of uncertain or even acceptable etymology, and scholars will be able to pick up where their numerous forgotten, disregarded, and misunderstood predecessors left off. Anatoly Liberman has been working on “an English Feist” for thirteen years, and his suggestions are the result of this work. Thomas Cable’s paper, “Issues for a new history of English prosody”, proceeds from the observation that an adequate new history of English prosody must resolve the main incompatibilities of three rich traditions of prosodic study: temporal, accentual, and generative. He argues that the problems and possible solutions can best be illustrated by focusing on generative metrics (a particular development of the stress tradition) because of the explicitness of its goals and rules, a virtue often missing in the other two approaches. In his critique of the generative tradition, he assigns special relevance to three contexts: (1) the diagnostic relevance of the W, or Weak, position, (2) the lack of attention to the metrical pause, and (3) the misleading attention to the caesura. He examines and seeks to correct generative analyses of the iambic pentameter as used by Chaucer and Shakespeare, and by the poets between them, in a way that he believes clarifies this pivotal period in the history of English prosody. He concludes that there is a significant difference in the internal structure of the line in Chaucer and Shakespeare, and that the most succinct formulation of this difference is to say that Shakespeare made use of the iambic foot and Chaucer did not. Cable’s original and controversial position on the difference between the Chaucerian and the Shakesperean iambic line is disputed in a counterpoint paper entitled “Chaucer: Folk Poet or Littérateur?” by Gilbert Youmans and his student Xingzhong Li. Youmans and Li were not participants in the conference but became involved in the exchange at our request. Thus we have a comparison of these two views on how the prosody of iambic pentameter can be best described, Cable’s own view and a version of the generative view represented, in this instance, by Youmans and Li. The differences are vast and extremely interesting, and having them aired this way we believe is greatly to the benefit of us all. In a framework consistent with Optimality Theory, Youmans and Li propose a central prototype for iambic pentameter verse and describe gradient “tension” rules (violable constraints) for measuring degrees of deviation from this prototype. They provide statistical evidence in support of the traditional view that Chaucer’s decasyllabic verse is iambic pentameter [WS][WS] rather than an unfooted sequence of alter-

Foreword

7

nating syllables WSWS. Chaucer’s verse, they acknowledge, is more regular than Shakepseare’s and Milton’s, but all three poets share the same central prototype, a 4//2/4 line structured [WS][WS]//[WS]/[WS][WS]; and variations from prototypical stress patterning are most common at the beginnings of their lines (and hemistichs) and least common at the ends of their lines (and hemistichs). Thus, Chaucer’s verse differs stylistically from Shakespeare’s and Milton’s, but it embodies strikingly similar metrical principles. Cable’s response summarizes the differences between his theory of pentameter and that of the generativists, as represented by Youmans and Li, under six categories. (1) A rigorous insistence on finding five beats to the line – no reading without five beats is acceptable, some beats no doubt weaker than others but still stronger than the adjacent “weak” beats. (2) “Tilting” some weaker syllables toward strong, and some stronger syllables toward weak, is both necessary and proper. (3) Non-lexical words, i.e. function words, not just lexical words, can bear ictus under the right contextual circumstances. (4) Distinctions between unstressed syllables that are cliticized as function words, and lexically unstressed syllables, are irrelevant for the purposes of assessing poetic rhythm. (5) Constituent bracketing, except with respect to caesura marking, is irrelevant to the determination of metricality. (6) Metricality is not the same as tension: What is relevant to measuring tension may not be relevant to assessing metricality. What may reasonably be considered a seventh category, though not so numbered by Cable, is his insightful discussion of trisyllabic feet in Chaucer, in particular pointing out that there are no trochaic reversals in which the reversal depends on a weak position two syllable consisting only of a final -e. Though there are epiphenomenal reasons which might explain this gap, the final -e in position two would be an incontrovertible and untiltable trochaic reversal: Cable argues that this gap requires explanation. As editors we would not be willing to take sides and say who won this exchange, even if we had a firm clear answer (which we don’t). We found the exchange extremely enlightening, and we believe our readers will also. We are deeply grateful to all three scholars for having the integrity and courtesy to present these vigorously conflicting views of the nature of the iambic pentameter as it is (mis)understood at the turn of the century. Finally, among the millennially-oriented papers, Richard W. Bailey’s paper, printed at the end of the book because of its envoy tone, “A thousand

8

Foreword

years of the history of English”, is a swashbuckling tour through English philology of the 19th and 20th centuries (“text-centered study of language”, in Bailey’s words) in support of the view that we need a “renewed philology”. The paper is a vigorous reminder that descriptions of linguistic facts are not by any means free from assumptions, and that our views of the history of English often express biases that are only dimly apparent to us. Bailey compares our own histories of English with those of our predecessors: members of the Philological Society squabbling over standards as its New English Dictionary began to appear in 1884; Thomas Percy, the eighteenth-century clergyman who nearly single-handedly invented a distinctive Middle English and demonstrated how he could revive feudalism and place himself within it as a cosseted minstrel even in the midst of the Age of Reason; John Free, writing the first book-length history of English, who managed to revive pride in Germanic languages by demonstrating that English was one of them; Byrhtferth, in the twelfth-century, who used English to demonstrate revealed truth through words. His argument concludes with the claim that we make better histories if we understand, and make public, our assumptions, an appropriate appeal to the scholars of the twenty-first century.

Phonology and Metrics Blaine Erickson’s paper on “The history of English r” is not a full-scale history of all varieties of /r/ found in the English language so much as it is an effort to explain how the phonetic properties of the common American English /r/ came about. This topic deserves such special treatment because it is the odd articulation of /r/ that is the most strikingly different feature of American as compared with any other variety of English. In American English, retroflex and central approximant articulations are the most common phones found for /r/. Cross-linguistically, both sounds are rare, and since they are almost unknown in other Germanic languages, the r phones of American English are more likely to be innovations than retentions. Many other Germanic languages have a coronal trill as their phone for /r/; evidence suggests that this was also the original phone for English /r/. Erickson assumes that during the development of pre-modern English, the

Foreword

9

place of articulation of the coronal consonants changed from dental to alveolar. This caused the tongue to take on a sulcal shape, which set the stage for the change from trill to retroflex. Retroflex articulation occurred first as a conditioned change: when /r/ appeared before other coronals. Later, retroflex articulation spread to other environments, and it supplanted the trill. After that, the central approximant appeared as an alternate pronunciation for /r/. This is motivated by the fact that the retroflex and central approximant articulations are nearly identical acoustically, even though the tongue configurations for the two are opposite: tongue tip up for the former, and down for latter. Kristin Hanson’s paper, entitled “Vowel variation in English full rhymes”, is an effort to account for imperfect rhymes in the history of English verse. Virtually all modern English poets include some rhymes in their practice in which the vowels differ in present day English, such as Shakespeare’s famous love/prove rhymes in the Sonnets. Versification handbooks and history of English textbooks tend to suggest that English sound changes explain such rhymes, in that the vowels would have been identical in the poets’ own dialects. But philological studies paint a more complex picture: Kökeritz (1953), for example, concludes that many of Shakespeare’s rhymes would have involved genuine differences in the vowels, and would have counted as rhymes only in virtue of literary precedents, the presence of dialect variants in the linguistic milieu, or mere phonetic similarity. Moreover, such differences in rhyme vowels often seem aesthetically significant. These considerations invite the question of whether for a given poet’s rhyming practice can be formally defined in such a way as to accommodate expressive variation while still distinguishing possible from impossible variations. For Shakespeare’s Sonnets at least, the explanations Kökeritz offers are subtly problematic for this purpose, in that they invoke a conception of language which is social and imitative rather than individual and creative. A closely related alternative which would reformulate the role of history in shaping these practices and locate them within a unified grammar rests on the observation that vowels which are paired in such rhymes are often those which participate in alternations such as reduce/reduction. Whatever the value of this observation for Shakespeare’s rhymes, it is clear that the general question of what principles govern the vowel variation in English poets’ full rhymes, and how these are transmitted and changed from poet to poet, constitute significant open research questions.

10

Foreword

Betty S. Phillips’ paper, entitled “Lexical diffusion and competing analyses of sound change”, seeks to show that the current theory of lexical diffusion can be used to help choose between competing analyses of a given sound change, specifically the shift of Early Modern English /u:/ to /U/ in such words as GOOD and STOOD. Ogura (1987) hypothesized that the impetus behind this change was an attempt to keep the duration of syllabic units relatively constant – following a suggestion by Dobson (1968) that the early stages of the change complemented lengthening in an open syllable. Görlach (1991: 71) suggested that “the short vowel in words like GOOD and STOOD was introduced on the pattern of words in which the occurrence of a short or a long vowel was determined by the type of syllable the vowel appeared in (GLAD vs. GLADE).” Yet since both explanations require lexical analysis, as defined in Phillips (1999), one would expect the least frequent words to have been affected first. Ogura (1987), however, finds that the change in question affected the most frequent words first. For that reason, Phillips offers an alternative, phonetically motivated explanation for this sound change. Geoffrey Russom’s paper “Purely metrical criteria for dating Old English poems”, seeks to provide a new set of dating criteria for early English texts, supplementing standard philological techniques, because improved dating would greatly facilitate research on English linguistic history. He proposes a method for dating Old English poems by reference to frequencies of verse types. Several strategies are employed to deal with problems arising in comparable studies. First, verses are selected for analysis in a way that avoids bias toward the native heroic sphere, which can affect frequencies for many variants of types C, D, and E. Further, nonstandard frequencies due to limited poetic ability are factored out by reference to expert judgments based in part on the number of verses for which normal scansion is impossible; and the expert judgments are then cross-checked against frequencies of type A3, the verse pattern that would be easiest for a poet of limited talent to construct. Appropriate relative frequencies are predicted by fundamental principles of the word-foot theory (Russom 1987, 1998), which yield a linguistically-based metrical norm and ways to compute deviation from the norm. Deviation from the norm restricts frequency, as expected; and the decline of the metrical tradition is accompanied by major shifts in the placement as well as the frequency of deviant verses. The results of this study confirm, in striking detail, the results achieved by Fulk (1992), who employs entirely independent dating criteria.

Foreword

11

Robert Stockwell’s paper entitled “Short shrift for the ‘great’ vowel shift” is an effort to provide an alternative and substantially simplified view of what might have happened to produce the appearance of vowel shifting without such a complex set of events having actually occurred. The English Vowel Shift is traditionally thought of as a chain in which the next vowel up the chain is displaced, in a continuous cycle. Stockwell argues that this picture is wrong in two ways: (1) Most stages of the English shift are not chained but actually result in mergers. The targets of these mergers are diphthongs, either inherited (like MAIDEN or GROW) or borrowed (like DAINTY or PRAY from French, or RAISE from Scandinavian). (2) Diphthongization of [i:] and [u:] toward [ai] and [au] is not a chain at all, since nothing is ever displaced: all members of these parts of the shift coexist to this day (e.g., the misnomer “Canadian raising”) and must be considered allophones or diaphones (Kurath’s rarely used but highly appropriate coinage). Combining (1) and (2), Stockwell gets a very different view of the vowel shift, reducing the “shift” to the raising of [e:] and [o:] to [i:] and [u:] (FEEL and FOOT) and the initial diphthongization, though not the continuing diaphonic development, noted in (2). This difference of interpretation raises difficult problems for the view (e.g. Lass 1999) that assumes leveling of the RAISE vowel to a long monophthong, and similarly the GROW vowel, with the general diphthongal quality found over the past three centuries taken as innovative. David White’s paper deals with a highly technical philological question about certain sound changes in Old English. The paper is entitled “Restoration of /a/ revisited”. According to the traditional historical phonology of Old English, all instances of the phoneme /a/ are supposed to have changed to /æ/, with some /æ/’s later changing back again into /a/ in certain environments where /a/ later appears. This last change is known as “restoration of /a/”. That it is inelegant is obvious, and we should only believe it for a very good reason. The traditional reason, for /h/ to have acted as a blocker of back-assimilation, as is implicitly suggested by Moore and Knott (1942), is phonetically implausible. But it is possible that the relevant phenomenon was not blocking but dissimilation. /h/ is such a weak consonant that /aho/ might as well be /ao/. If stressed back vowels were dissimilated before following unstressed back vowels where no “buccal” consonant intervened, then /aho/ would develop into /æho/, from which later /æo/ (or whatever we hold long “ea” to have been) is not problematic. Under such a scenario, dis-

12

Foreword

similation would also motivate the change of /au/ to /æo/. White’s conclusion is that the change of /a/ to /æ/ was not unconditioned, that words like ea ‘water’ (from /aho/) never had short diphthongs from breaking, and that restoration of /a/ in words like dagas ‘days’ never occurred. Such words always had /a/.

Morphosyntax and Semantics Maurizio Gotti’s paper on “Pragmatic uses of shall future constructions in Early Modern English” carries us forward in the philological tradition referred to in Richard Bailey’s paper – the “text-centered study of language”. The object of the paper is the analysis of the uses of the future tense with shall in an Early Modern English corpus; the texts examined are those included in the third section of the Early Modern English part of The Helsinki Corpus of English Texts. The results of this analysis are compared to the pragmatic values pointed out in a few grammars of the same period in order to assess their degree of faithfulness to the real usages in contemporary texts. The period taken into consideration is 1640–1710 and includes the grammars which first describe the pragmatic uses of this modal auxiliary for the formation of the future tense. The analysis of the shall-forms contained in the corpus first identifies all the quotations referring to futurity and then subdivides them into semantic/ pragmatic categories. This is done in order to confirm the basic differentiation between pragmatic and dynamic uses of this modal suggested by the 17th-century grammarians’ adoption of categories labelled “prediction”, “promising”, “threatening”, “declaration”, “command”. The survey also points out the qualitative difference between the deontic and dynamic aspects of the shall-forms found in the corpus. Another aspect which is investigated is the degree of correlation between these modality categories and the text types contained in the corpus; the results indicate that some genres show a very high correlation rate, while others are characterised by a much wider range of modal categories. Edward L. Keenan, a distinguished logician and semanticist, has recently gotten deeply involved in the explanatory aspects of historical philology. In this paper he takes on the subject of reflexive pronouns, under the title “Explaining the creation of reflexive pronouns in English”. He pro-

Foreword

13

vides an account for the creation and interpretation of the English reflexives (himself, herself, …) in terms of two general forces of change, two universal semantic constraints on language, and the start state of the Old English anaphora system. His account is based on the study of some 11,000 instances of locally bound objects of verbs and prepositions, drawn from over 100 texts dating from c750 to c1750, complementing Keenan (2000) which is a more extensive presentation of these data than we could accommodate here. He contrasts his account with two other accounts of the creation of grammatical formatives: grammaticalization and parameter resetting. The creation of reflexives in English falls into neither category though it shares one feature with grammaticalization. A comprehensive presentation including a list of Source Materials is to be found in Keenan (2001). Ans van Kemenade initiated very important work on clitic pronouns in Old English near the end of the 1980’s. Her contribution here, “The word order of Old English poetry compared with prose once more”, deals with some non-pronominal aspects of clitic word order in Old English, and has significant bearing on the issue of dating the earliest documents. The paper highlights a difference between verse and prose word order in negative root clauses: in particular, the existence of one main clause word order pattern that is unique to the early poetry, namely negative-initial clauses in which the finite verb does not appear adjacent to the negative word. Her study is both a contribution to the description and analysis of the earliest syntactic patterns in English and, by implication, a syntactician’s confirmation of the dating of poetic texts. The same theme, dating the poetic corpus, is addressed from a different perspective in this volume by Geoffrey Russom. The word order of Old English poetry has given rise to some considerable debate: some hold that the principles governing it represent a genuine departure from those ruling the prose; others argue that it is governed by the same principles as the prose, albeit at differential frequencies dictated, for instance, by metrical considerations. The special aspect of the word order that Kemenade focuses on here is that a negative element is the first constituent, but the finite verb is not attracted to it, as it is in later Old English. The diachronic picture for the history of English suggests that this pattern is an older one. This in turn supports the familiar view that the extant poetry is, in some respects at least, older than the extant prose. Jeong-Hoon Lee deals with the semantics of Perfect tense in Old English in his paper “The “have” perfect in Old English: How close was it to the

14

Foreword

Modern English perfect?” He examines the semantics of the have perfect tense in Old English and the perfect tense in Modern English. Lee argues that there is no significant semantic difference between these perfect tenses, contrary to the accepted view that the have perfect tense in Old English had at best only the so-called “resultative” meaning. Based on the three meanings of the Modern English perfect, which represents three subcategories by perfect tense forms, i.e., the existential, universal, and resultative meanings, he shows that, in addition to Old English have perfect constructions with the resultative meaning, there were also many Old English have perfect constructions with either an existential or a universal meaning. He also argues that there were many Old English have perfect constructions that could not have been interpreted as the resultative perfect, namely, have perfect constructions modified by manner adverbials, and have perfect constructions in negative sentences. Additionally, he suggests that many Old English have past perfect constructions had the same ambiguity as that of the Modern English past perfect; that is, it was not clear whether they had the “perfect” sense, or they expressed just an event/situation before a past time. Colette Moore’s paper is entitled “Reporting direct speech in Early Modern slander depositions”. Determining how speakers used language in the Early Modern period presents difficulties due to the deficiencies of colloquial source material. Court records, with their recording of witness testimony, provide one source of “speech-based” texts, and they are being more widely considered in this light after their inclusion in such computer corpora as the Helsinki corpus. Without dismissing the value of depositions as a potential source for spoken English, we must also be aware that their usefulness comes within a certain context. She investigates the context of these records and examines the depiction of direct speech in a sample of them. Defamation depositions reveal both code- and style-switching, and reported speech plays a central role in the records since it presents the alleged criminal action. She analyses the treatment of direct and indirect speech in the defamation court records in an attempt to understand the switching between spoken and written language in Early Modern English and the conventions that surround the reporting of speech within a written form. This, in turn, allows us, she believes, to perceive better the nature of the linguistic construction of these depositions and its relation to spoken English, so that we may make more considered use of court records on their own terms and within computer corpora.

Foreword

15

Benji Wald and Lawrence Bessermann’s paper is entitled “The verb-verb compound in twentieth century English and twentieth century linguistics”. It deals with the emergence of the endocentric VV (verb-verb) compound as a unified productive pattern. This pattern poses an interesting set of problems and challenges for twentieth century historical and synchronic linguistics. Although VV seems to have eluded discussion and even exemplification in most major treatments of English throughout the twentieth century, such studies anticipated a range of relevant historical and synchronic problems in the analysis of the type. The principal problem that they consider in detail is that VV compounds frequently seem ambiguous as between VV and NV analyses. How can we tell whether sleep in sleep-walk is N or V? They refer to this analytical problem as the ambiguous category problem and offer a historical solution to it. After discussing the analytical issues, and recent innovations from which they have arisen, including a general constraint on verb compounding which they label “the activity constraint”, they address the longer term question: why was the twentieth century a critical period in the development of the VV compound? Their answer is that the development of VV has depended on a number of preconditions that have successively accumulated since the Old English period. They briefly consider the historical evolution of VV, first through the development of its preconditions in late Old English and Early Middle English, and then through a quantitative study of VV and related items listed in the second edition of the Oxford English Dictionary and its later additional supplements.

Acknowledgements The idea of convening SHEL-1 could not have materialized without the financial backing of the Center for Medieval and Renaissance Studies, the Department of English, and the Department of Linguistics at UCLA. We thank the leaders of these units for their support. For encouragement and advice in preparing this volume we are grateful to the TiEL series editors, Elizabeth Traugott and Bernd Kortmann, and to Anke Beck, Birgit Sievert, and Wolfgang Konwitschny at Mouton de Gruyter. Above all, we extend our thanks to our authors who were prompt in submitting, patient in awaiting the comments on their papers, and conscientious in revising. They have been a terrific group to work with.

16

Foreword

Each paper in this volume was reviewed anonymously by two outside reviewers, and many of the submissions underwent revision several times. It is with very special gratitude that we list the names of our colleagues who took time away from their own pursuits to advise us and the authors, often writing critiques that could easily have become independent contributions. It is a heartening comment on the scholarly dedication to the field that so many people were ready to render first rate professional service with paper bags over their heads. Disclosing their names is the least we can do to acknowledge their help. We hereby express our very special thanks to them, with the important further acknowledgement that their criticism was never ignored, though it was not always accepted with agreement either by the authors or the editors, no doubt to our eventual discomfort, and certainly no one but the authors and editors can be blamed for errors that remain. Leslie Arnovick, University of British Columbia, Vancouver, Canada Laurel Brinton, University of British Columbia, Vancouver, Canada Fran Colman, University of Edinburgh, Scotland Claire Cowie, University of Sheffield, England Christanne Dalton-Puffer, University of Vienna, Austria Hans Jürgen Diller, University of Bochum, Germany Martin Duffell, Queen Mary and Westfield College, London, England Nigel Fabb, University of Strathclyde, Scotland Susan Fitzmaurice, Northern Arizona University, Flagstaff Rob Fulk, Indiana University, Bloomington Elly van Gelderen, Arizona State University, Tempe Michael Getty, University of Toronto, Canada Robert Howell, University of Wisconsin, Madison Andreas Jucker, University of Giessen, Germany Juhani Klemola, University of Helsinki, Finland Willem Koopman, University of Amsterdam, Holland Barbara Kryk, University of Poznan, Poland Merja Kytö, Uppsala University, Sweden Bettelou Los, Vrije Universiteit, Amsterdam, Holland Angelika Lutz, University of Erlangen, Germany David Matthews, University of Newcastle, New South Wales, Australia Chris McCully University of Manchester, England Frances McSparran, University of Michigan, Ann Arbor Michael Montgomery: South Carolina, Columbia, S.C. Robert Murray, University of Calgary, Canada Mieko Ogura, Tsurumi University, Japan Päivi Pahta University of Helsinki, Finland

Foreword Theo Vennemann, University of Munich, Germany Katie Wales, University of Leeds, England Tony Warner, University of York, England Gilbert Youmans, University of Missouri, Columbia, MO

Presenters and Session Chairs at SHEL-1 (UCLA, May 2000) Noriko Akatsuka, UCLA Henning Andersen, UCLA Richard Bailey, University of Michigan, Ann Arbor Lawrence Besserman, Hebrew University, Jerusalem Mary Blockley, University of Texas, Austin Laurel Brinton, University of British Columbia Thomas Cable, University of Texas at Austin Ruth Carroll, University of Turku, Finland Don W. Chapman, Brigham Young University Anne Curzan, University of Washington, Seattle Edwin Duncan, Towson University Nancy Elliott, Southern Oregon University Blaine Erickson, Kanazawa Institute of Technology, Japan Susan Fitzmaurice, Northern Arizona University Robert Fulk, Indiana University Elly van Gelderen, Arizona State University Michael Getty, University of Toronto Gwang-Yoon Goh, Ohio State University Chris Golston, Fresno State University Maurizio Gotti, University of Bergamo, Italy Kristin Hanson, UC Berkeley Yukio Haraguchi, Kumamoto Gakuen University, Japan Bruce Hayes, UCLA Richard Hogg, University of Manchester, England Richard Janda, Ohio State University Dieter Kastovsky, University of Vienna, Austria Edward L. Keenan, UCLA Henry Ansgar Kelly, UCLA Ans van Kemenade, Katholieke Universiteit, Nijmegen, Holland Willem Koopman, University of Amsterdam, Holland

17

18

Foreword

William A. Kretzschmar, Jr., University of Georgia, Athens Anthony Kroch, University of Pennsylvania Barbara Kryk, University of Poznan, Poland Ian Lancashire, University of Toronto Eva Delgado Lavin, University of the Basque Country, Spain Jeong-Hoon Lee, University of Texas, Austin Anatoly Liberman, University of Minnesota, Minneapolis Hans Lindquist, Växjö University, Sweden Soon-Ai Low, University of Maryland Joseph P. McGowan, University of San Diego Frances McSparran, University of Michigan Anna Meskhi, Isik University, Turkey Donka Minkova, UCLA Colette Moore, University of Michigan Betty Phillips, Indiana State University Terre Haute Thomas Riad, Stockholm University, Sweden Philip Rusche, University of Nevada, Las Vegas Geoffrey Russom, Brown University Herbert Schendl, University of Vienna, Austria Catherine Smith, Northern Arizona University William Spruiell, Central Michigan University Robert Stockwell, UCLA John Sundquist, Indiana University Akinobu Tani, Hyogo University, Japan Elizabeth Traugott, Stanford University Geoffrey Russom, Brown University Benji Wald, Los Angeles David L. White, University of Texas, Austin

Donka Minkova and Robert Stockwell Los Angeles, March 2002

Note 1.

For the bibliographical details on all short-form references embedded in the summaries in our Foreword, please see the reference lists attached to the appropriate paper.

From etymology to historical pragmatics Elizabeth Closs Traugott

0.

Introduction 1

One of the much-discussed themes of linguistic work in the last hundred years has been how to think about the relationships between synchrony and diachrony, between structure and use, between arbitrariness and motivatedness, between what Roberts has called “a random ‘walk’ through the space defined by the set of possible parameter values” (Roberts 1993: 252) and observed directionalities of change. Throughout the twentieth century diachronic theory lagged behind synchronic. While rapid advances have in recent years been made in the study of historical morphosyntax, no aspect of language change has lagged behind synchronic theory more than pragmatic-semantic change. Though the seeds of many ideas that dominated the end of the century are to be found in work at its beginning, neither the semantics nor the pragmatics were sufficiently far advanced for those ideas to be developed in principled ways. Recently, however, there has been the possibility of catch-up, and in that catch-up it has been the functionalist, pragmatic aspects of linguistics that have carved out a path that in diverse ways link up with older concerns. What began as work on the semantic correlates of grammaticalization, often using texts in ways that privilege discourse and genre, has among some practitioners now come to be known as “historical pragmatics” (see Jucker 1995). There are many paths that one could follow with such a topic, but I will highlight work on certain aspects of the regularities and directionalities in change that have been explored, with particular attention to work on the interplay between language and use (see most recently Croft 2000), as it is conceptualized in the study of grammaticalization and especially historical semantics and pragmatics. In recent years we have heard that the search for tendencies and directionalities is a hold-over from nineteenth century historicism, that change is an epiphenomenon, even that history itself is an epiphenomenon (Lightfoot 1999: 261). In other words, it falls out from things other than itself, from

20

Elizabeth Closs Traugott

what people do. There is no doubt that language change occurs because of forces outside of language, most particularly processes of production and interpretation. If language acquisition is a prime reason for change, if competing motivations such as “be clear and processable” vs. “be quick and easy” (Slobin 1977; Langacker 1977) or maxims such as have been proposed by Keller (1994) are valid reasons for language change, then it is obvious that language change is an epiphenomenon. We have a tradition of calling change within a relatively homogeneous community that is brought about by language acquisition “internal change”, as opposed to “external change” brought about by contact, but the first is actually no more “internal” than the latter – it does not happen “in the language”, or “in the grammar”, only “in” transmission. Expressions like “grammars change” are short-hands for differences between grammars over time brought about by system-external factors such as acquisition, and not to be taken literally. Likewise, “directionalities” are not deterministic tendencies that require some change, and definitively not tendencies that live some reified existence as cognitive paths, trajectories, or whatever other metaphor might be used, but they are nevertheless powerful tendencies that demand historical thinking. To couch debates about such issues in language referring to “formal” and “functional” theories is admittedly to reproduce old dichotomies many of which were fortunately falling by the wayside in the nineteen nineties (see e.g. Croft 1995; Newmeyer 1998). Nevertheless, it is useful in that it allows us to acknowledge that the questions different researchers pose may be fundamentally different (see also Kemenade 1999). To say on the “formal” (generative) side of the debate that all one needs in order to explain language change is: “(a) an account of how trigger experiences have shifted and (b) a theory of language acquisition that matches PLD (primary linguistic data) with grammars in a deterministic way” (Lightfoot 1999: 225) simply puts the explanation off. WHY does the trigger change? Has the alleged “determinism” of the search for directionality been replaced by a different determinism that predicts that if two children were to have the same trigger experience they would acquire the same E-language? On this view, language change is the result of innovation in the individual compared to some other, older, individual. Same input (PLD) into the individual (LAD) will produce same output (G). In this scenario, the individual is a processor of systems, largely passive, a logic

From etymology to historical pragmatics

21

machine, a “language acquisition device”, presumably devoid of personal differences or preferences. It is an implausible scenario not only because we know children negotiate and experiment with language, but above all because it does not allow for abduction in language acquisition (see Andersen 1973; Anttila [1972] 1989 on abduction as the key mechanism in language change). From a “functionalist” perspective, by contrast, change is the result of strategic interaction, specifically of choice-making on the part of speakers/writers in interactional negotiation with addressees/readers. This includes, but is not limited to, conveying of information. On this view, language change is the result of innovation in the individual and spread of the innovation to the community, as suggested by Weinreich, Labov and Herzog (1968). This individual is active, an abducer who is a producer as well as processor of language, and likely to have personal differences and preferences. Furthermore, innovation is not limited to early child language acquisition (see e.g. Andersen 1973; Labov 1994; Ravid 1995). On this view too explanations for observed directionalities (and failures of directionality) should be especially interesting and precisely what historical linguists need to seek to provide. Here I hope to suggest some ways in which to organize our thinking as we look for ways to account for directionality. Unidirectionality is a thread common to the study of grammaticalization and at least some branches of historical semantics and pragmatics. Some arguments against unidirectionality are primarily philosophical – dependent on what the researcher thinks a scientific argument is or should be – arguments of this type have been put forward by Lightfoot (1999), and Lass (2000). Other arguments have been primarily of an empirical sort, for example the work of Joseph and Janda (1988), Janda (1995), and Ramat (1992). Both kinds are presented in Newmeyer (1998) and Campbell (2001). 2 I take these arguments seriously. Nevertheless I do not find it useful to take a position such as “I take any example of upgrading as sufficient to refute unidirectionality” (Newmeyer 1998: 263). If change is a social product, the result of the interaction of language and use, how could there be no counterexamples? Humans are not machines, and do not use language mechanically. They use it for purposes of strategic interaction. Strategies must be flexible not rigid 100%s of anything! But likewise, since unidirectionality is not exceptionless, it is also not useful to adopt Leh-

22

Elizabeth Closs Traugott

mann’s ([1982] 1995) and Haspelmath’s (1999) claims that there are no, or at least no genuine, counterexamples to grammaticalization. Our road map is as follows. Section 1 provides some background on grammaticalization, with focus on English studies in the last hundred years, and discusses aspects of the “deconstruction” of grammaticalization in Newmeyer (1998: 5). Section 2 provides some background on semantic change and the role of pragmatics in semantic change. In section 3 I suggest a possible model of grammaticalization informed by historical pragmatics. Section 4 puts forward some challenges for the future.

1.

Grammaticalization

The term grammaticalization, as is well known, seems to have originated with Meillet ([1912] 1958). Since then it has been closely associated with reanalysis that involves a change from lexical material in constructions to functional category status, e.g. from nominal or verbal status to markers of case, tense, aspect, mood, conjunctions, etc. Although the term grammaticalization did not find its way into much work on English until the early nineteen seventies, Jespersen’s view of historical syntax was in several ways germane to work on it. We recoil today from Jespersen’s ideological claims that the modern languages are “better” than older ones, that they have made “progress” or that simplification is “beneficial” (Jespersen [1922] 1959: 363). We probably reject the idea that perfect languages would “express the same thing by the same, and similar things by similar means ... sound and sense would be in perfect harmony” (Jespersen 1959: 442). Yet his interest in “the transition from freedom in word position to greater strictness” (Jespersen 1959: 363), in shifts to more regular paradigms, and to shorter forms, coupled with his idea that language originated in “half-musical unanalyzed expressions ... words and quasi-sentences” (Jespersen 1959: 441) resonates with some of the thinking behind Givón’s famous “cycle” (as he calls it), reproduced in (1): (1)

discourse > syntax > morphology > morphophonemics > zero (Givón 1979: 209)

From etymology to historical pragmatics

23

We should note in passing that Jespersen rightly saw that it is incoherent to think of inflectional loss as preceding fixing of word order, i.e. in a cycle: “if this were true we should have to imagine an intervening period ... in which speech was unintelligible and consequently practically useless” (Jespersen 1959: 361). This is exactly what Lehman (1985) found himself having to repeat again sixty years later. Despite Jespersen’s major work, Modern English Grammar on Historical Principles, that spanned forty years (1909–49), and Visser’s English Historical Syntax that spanned ten (1963–73), the book that probably did most to jump-start historical English morphosyntax was Lightfoot’s Principles of Diachronic Syntax (1979). It put the spot-light on the history of modals, and of for-to constructions, both of which continue to be major areas of research. From this work in (1979) to The Development of Language twenty years later (1999), Lightfoot has focused on the question of what a scientific approach to language change might be, and has found an answer in language acquisition. Follow-up work in the formal generative tradition (broadly construed) includes work by Kroch (e.g. 1989), Fischer (1994, 2000), Kemenade (1987, 1999), Warner (1993), and several papers in Kemenade and Vincent (1997), to mention only a few. In all there is the basic assumption that the interesting question is how grammars change (not how people bring about changes in certain parts of grammars via strategic interaction), and how child language acquisition may explain grammar change. It is usually argued that because children bring about change, there cannot be unidirectionality. In perhaps the clearest position statement from this perspective, Lightfoot has said that if a child were to be thought to access unidirectionality one would have to falsely postulate a “racial memory of some kind” (1999: 209). “Racial memory” is obviously an incoherent idea. But Lightfoot’s hypothesis that the trigger experience is simply the data heard, and the frequency of it, is not a good substitute because it is unrealistic. The child is thought of as receiver and perceiver, apparently without any social purposes or awareness of the social purposes to which language is put, of differences in language-use correlated with age, or of construction of identity through language, and so forth, all of which can give the child access (or partial access) to layers of history. In his (1979) book Lightfoot discussed catastrophic change – innovation of new syntactic categories or orders (what more recently has been re-

24

Elizabeth Closs Traugott

thought as parametric change), while acknowledging that the challenge to researchers was to identify the local changes leading to them. It is precisely these local changes that have generated work in the more functional perspective. The same year that Lightfoot’s Principles of Diachronic Syntax appeared, Givón’s On Understanding Grammar was also published. This highlighted the importance of cross-linguistic and typological studies (e.g. Lehman 1995 (though only remotely “functional”); Bybee 1985; Heine and Reh 1984; Heine, Claudi, and Hünnemeyer 1991; Kemmer 1993; Svorou 1993; Bybee, Perkins and Pagliuca 1994; Haspelmath 1997). Sometimes historical data from English was used to illustrate in depth the kind of change that is sketched or even reconstructed for other languages. Just as Latin cantare habeo ‘to sing have I’ to Fr. chanterai has become a prototype change in discussions of Romance, so the development of be going to has dominated the pages of work on grammaticalization (see Hopper and Traugott 1993; Bybee, Perkins, and Pagliuca 1994, both strongly influenced by Pérez 1990). But a great deal has also been devoted to English as a subject in itself (e.g. Brinton 1988, 1996). Central in this line of work has been the research conducted on English in the context of the English Department at the University of Helsinki (e.g. Rissanen, Kytö, and Heikkonen 1997). The contributions to grammaticalization in English are far too many to mention here, but some (introduced in short form in Rissanen et al., but developed into full-length monographs) deserve special mention: Kytö (1991) on modals, Nevalainen (1991) on the development of the focusing adverbials but, only, just, and Palander-Collin (1999) on the grammaticalization of I think and methinks as epistemic parentheticals. In all these works, the pragmatic-semantic correlates, indeed putative precursors of, grammaticalization have played a major role. Change is conceptualized as motivated by discursive and social practices. Note that this is not the same as saying discourse > syntax – Givón was hypothesizing change in structure from loose to tighter syntax, etc. “Discursive practices” as I am using the term assumes structure and syntax, and treats speakers and addressees’ interaction in discourse as a motivator of change not a stage or phase in historical development. On this view, the local changes investigated are often so particular, and so deeply entrenched in the idiomatic, constructional, semantic and pragmatic nature of language structure and use that they cannot be usefully conceived of as the products of small children setting parameters, but rather of older teens and adults engaged in

From etymology to historical pragmatics

25

discourse practices for specific social purposes or of language-acquirers in general. Many of these issues are discussed in Newmeyer’s (1998) book on Language Form and Language Function. A whole chapter is devoted to “deconstructing grammaticalization”. He characterizes grammaticalization as describing “the loss of grammatical independence of a grammatical structure or element” (Newmeyer 1998: 227) and cites as “the standard definition”: “grammaticalization ... where a lexical unit or structure assumes a grammatical function, or where a grammatical unit assumes a more grammatical function (Heine, Claudi, and Hünnemeyer 1991: 2). In Newmeyer’s view, and indeed, that of most researchers on grammaticalization to date, the “or” is interpreted as inclusive – see also the following rather similar definition: “the process whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions, and, once grammaticalized, continue to develop a new grammatical function” (Hopper and Traugott 1993: xv). Newmeyer’s arguments are based on formal assumptions: change comes about through child language acquisition, only phenomena that are universal in the absolute sense that they are predicted to occur 100% of the time are considered to be explanatory. I am not the first to find that despite himself Newmeyer actually constructs grammaticalization to some extent (see Haspelmath 2000, Heine 2000), partly because he finds that in the end he has to admit that “unidirectionality is almost true” (277) because it is so prevalent. Let us consider how Newmeyer approaches what I would call the motivation and mechanism issue, and what he, Roberts, Lightfoot and others call the epiphenomenon issue. The move that Roberts and Newmeyer try to make is that grammaticalization is an epiphenomenon not only of the mechanisms that bring change about (e.g. reanalysis, analogy/rule extension) but of other types of change, specifically downgrading analysis (decategorialization), phonetic reduction (erosion), appropriate semantic change (bleaching). Newmeyer models this as in Figure 1. This model may be intended simply to illustrate and critique what Newmeyer interprets some of the more extreme claims in the grammaticalization literature to be. But it has some properties worth discussing. In one sense this model appears to have no directionality. However, it has an inherently inbuilt directionality in two of its dimensions (downgrading and phonetic reduction). But what about semantic change? Is it not directional?

26

Elizabeth Closs Traugott

I will argue that it is – and in this directionality lies the key to the whole set of changes we think of as grammaticalization. This model shows that there is phonetic reduction independent of grammaticalization, and semantic change independent of grammaticalization. This is indeed correct. But what downgrading is there outside of semantic change and phonological reduction that is not a case of grammaticalization?

DA

ASC

GR

PR

Figure 1.

Grammaticalization as an epiphenomenon (Newmeyer 1998: 260). Legend: GR = grammaticalization, ASC = appropriate semantic change, DA = downgrading analysis, PR = phonological reduction. Reprinted by kind permission of the MIT Press.

I propose that we think of grammaticalization differently. We need to tease apart two aspects of the theory of unidirectionality in grammaticalization. The possibility to do this is interestingly anticipated in the definitions above and the differences between them, provided we read “or” as exclusive in Heine, Claudi and Hünnemeyer’s definition above, and note that “new grammatical function” is not the same as “more grammatical function” in Hopper and Traugott’s, also cited above. One aspect of unidirectionality has to do with the grammatical function of categories: this I will call “primary grammaticalization”. Primary gram-

From etymology to historical pragmatics

27

maticalization is the development in specific morphosyntactic contexts of constructions and lexical categories into functional categories. For example, in English, lexical verbs become auxiliaries in certain of their uses (or, in a different metalanguage, only a tiny subset of verbs that once could “raise to I” can still do so). Primary grammaticalization is compatible with recent work of Kemenade (1999), and Roberts and Roussou (1999). Indeed, that is how they define it; Kemenade specifically identifies the final stage of grammaticalization as “base-generation as a functional head” (1999: 1001). There is almost no argument against this type of unidirectionality. On this interpretation of unidirectionality, it makes no sense to say that grammatical material becomes more grammatical, or that it necessarily erodes, bonds, reduces syntactic scope, etc. Indeed, it is feasible to think of grammaticalization as allowing increased scope/freedom in external syntax (for example, that as a complementizer has wider syntactic scope than that as a pronoun serving an argument function (Tabor and Traugott 1998)). The other aspect of unidirectionality, and the one that is most controversial, has to do with the form rather than the function of the categories under consideration. On this view grammaticalization is the development of morphophonemic “texture” associated with the categories in question; here the issue is the degree of morphological bonding/fusion, phonetic erosion, bleaching, etc. (see specially Bybee, Pagliuca, and Perkins 1991; Bybee, Perkins, and Pagliuca 1994). With respect to such changes, it does make sense to talk about shifts to a more grammatical status, although it would be more accurate to say that “expressions of functional categories become more bonded over time”. Thus auxiliaries can undergo reduction (will > ’ll, would > ’d, have > ’ve). This kind of change I call “secondary grammaticalization”. It is presumably what Kurylowicz has in mind when he wrote the definition that lies behind those cited above: “the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to more grammatical status, e.g. from a derivative formant to an inflectional one” (Kurylowicz [1965] 1976: 52). The crucial difference between this definition and the others is the “e.g.” This part is presumably usually left off because of the difficulty of determining exactly what role derivation has in grammaticalization. These two kinds of change are linked in ways still to be understood, but in general we can say that changes of type B are later than, or at least start at

28

Elizabeth Closs Traugott

the same time as changes of type A, and crucially not before them. I hypothesize that the shift from construction or from lexical to functional category is a subset of possible morphosyntactic changes (others, like word order are of a different kind), and licensed by the set of regular pragmaticsemantic changes (though not all pragmatic-semantic changes involve grammaticalization, for example the development of performative meanings for speech acts verbs).

2.

Semantic change

As traditionally understood, semantic change would seem to have very little to offer to grammaticalization beyond the well-known mechanisms, articulated by Bréal ([1900] 1964), Meillet ([1905–6] 1958) and others about a century ago, mostly in contrasting pairs, such as metaphor, metonymy, specialization and generalization, pejoration and amelioration. Studies almost exclusively in terms of nominal referents (e.g. master, mistress, sentiment, telephone, bead) have little to contribute to grammaticalization or to questions about unidirectionality. But there were some hints of what were to become predominant themes in the effort to conceptualize the interface of meaning and grammaticalization. Bréal speaks of subjectification, a concept I will return to, Meillet of the linguistic contexts for change (for example, how homme ‘man’, chose ‘thing’ were affected by negation, interrogatives, conditional contexts (Bréal 1964: 239)), an idea absolutely fundamental to work on recurrent patterns of semantic change. Though methodologically alien now, Stern’s account of meaning change was particularly suggestive. In his chapter on “permutation” he wrote of: “A shift in the apprehension of a complex referent, denoted by a phrase, ... [which] will, by repeated use, become associated to the word expressing the earlier apprehension of it” (Stern [1931] 1968: 353). His example of the transfer of meaning from beads ‘prayer’ to ‘beads of a rosary’ is wellknown. Perhaps less well known is how he treated it: not as a transfer (metonymy) in the world of non-linguistic action, but in the linguistic world of the phrase (language use). This example is one from the domain of nominals, with all their attendant problems attributable to changes in the real world referents (are rosary beads made of wood, obsidian, plastic?). But he also discussed some other meaning changes that have no real-world refer-

From etymology to historical pragmatics

29

ents such as while with its development “from temporal to adversative and concessive import” (Stern 1968: 379). With hindsight one can see emerging in Stern’s attempt to characterize “permutation” as what in Grice’s ([1975] 1989) and Levinson’s (1983) conceptions came to be known as the conventionalizing of conversational implicatures. 3 I have developed the Grice/Levinson insight into the Invited Inferencing Theory of Semantic Change or IITSC (Traugott 1999; Traugott and Dasher In press; for an earlier version see Traugott and König 1991). Here I will mention only two things about IITSC because they will be crucial later. I have extended Geis and Zwicky’s (1971) notion of “invited inference”, seeing in this term (rather than “implicature”) the possibility of alluding to both speaker’s strategic action (inviting) and the hearer’s response (inferencing). Furthermore, following Levinson (1995, 2000), I distinguish between invited inferences that arise “on the fly” or are not salient in the community (these are Invited Inferences or IINs), and those that are well established (Generalized Invited Inferences, or GIINs). An example of a GIIN is the causative inference from after in certain contexts; this has been available since Old English. It is not only stable but well known and usable, but has not yet been semanticized as a polysemy. Recent work on semantic change motivated by pragmatic factors has its roots in various places. Four are of particular importance: lexical field theory, cognitive linguistics, grammaticalization, and historical pragmatics/ semantics. These in some sense span the century, with lexical field theory emerging first, then grammaticalization and cognitive linguistics, and finally historical pragmatics. Working at a time when Trier’s famous (1931) work on terms of intellect in Middle High German around 1200–1300 was much discussed (it was published in the same year as Stern’s book), Stern attempted to get away from the contemporary focus on discrete components of meaning divided up as in a mosaic. He took a psychological view of meaning and argued that change within a lexical field could be independent of culture. His special interest was in changes evidenced by the field RAPIDLY (e.g. OE swifte, georne) → IMMEDIATELY in the context of perfective verbs “denoting the action as a unit” (Stern 1968: 185–191). He argued that this change affected all words that meant RAPIDLY before about 1300; words from other sources that came to mean RAPIDLY or were borrowed with that meaning

30

Elizabeth Closs Traugott

after 1400, e.g. fleetly (1598) and rapidly itself (1727) did not undergo this change. The theoretical interest of Stern’s hypothesis is several-fold. He aimed to provide evidence that related words could show parallel semantic changes, and that semantic changes can take place over a certain period, and then cease to be in effect. He interpreted the data as an example in semantics of regular change similar to that posited by the neogrammarians for phonological change. More importantly, he suggested that the changes were dependent on the lexical aspect of the verb, rapidly being associated with process (imperfective) events, immediately with punctual (or perfective) ones. In other words, the change was associative, arising out of the syntactic “phrasing”, in this case use in what we would now think of as non-canonical contexts. He also showed that it was implausible for the reverse change to take place: “it is evident that if a person rides rapidly up to another, the action is soon completed; but we cannot reverse the argument and say that if a person soon rides up to another, then the action is also rapidly performed” (Stern 1968: 186). This is an argument for unidirectionality, and for the conventionalizing of implicatures. While Stern’s work was often ignored because of problems with his attempt to equate semantic with phonological change, lexical field analysis gained momentum in the sixties and seventies, largely influenced by anthropological taxonomies (for example Berlin and Kay’s 1969 work on color terms). English historical lexical field studies include such works as Williams (1976) on synaesthesia, the metaphorical extension of terms of the five senses onto the domain of the other senses and onto social behaviors, language, and so forth (cf. acid remark, bitter irony); Dahlgren (1985) on the social construction of kingship; Lehrer (1985) on the metaphorical use on zoological terms (e.g. baboon, vulture) for pejorative purposes; Goossens (1985) on verbs of speaking, to name just a few. The most recent line of research in lexical fields has been developed within historical prototype theory (Geeraerts 1997), e.g. Koivisto-Alanko (2000) on the field of wit, Molina (2000) on the field of give sorrow. Some of these studies began to suggest patterns of directionality, whether increase in the set of color terms along predictable paths, or patterns of change in synaesthesia. Not, however, till the rise of cognitive linguistics and of grammaticalization as active fields of research did interest in unidirectionality in semantic change really come into its own, for in these

From etymology to historical pragmatics

31

contexts the domains explored became increasingly conceptual and abstract. Nouns were paid less attention than verbs, most especially functional categories like tense, aspect, modality (e.g. Bybee and Pagliuca 1985), adverbials with temporal properties (e.g. König and Traugott 1982), focus particles (e.g. König 1991), etc. Especially significant for the study of semantic change was Sweetser’s (1990) work on modals and her suggestion (largely based on synchronic data) that one way to conceptualize the relation between deontic and epistemic modality was to think of it in terms of mapping of image-schemata. Sweetser hypothesized that in the case of the modal may sociophysical obligation is metaphorically mapped onto obligation in the world of reasoning; and further mapped onto what she called speech act meaning in the world of speaking, e.g. Kim may go (permission) onto Kim may be tired (epistemic), and onto Kim may be a nice guy, but I don’t trust him. (“speech-act modal”). Sweetser argued that meaning change is from content to reasoning and to “speech act” meanings, not vice versa (to avoid confusion with speech acts and strictly performative meaning, the terms “metatextual” (Dancygier 1992) or “procedural” (see Blakemore 1987) are preferable). This kind of thinking was fundamental to work on grammaticalization in African languages associated with Heine and his colleagues starting with Claudi and Heine (1986). Meanwhile I was working on an alternative approach, one that it turns out was more in keeping with Stern’s ideas, since I proposed looking not so much at the discontinuities of metaphorical mappings, at the sources and targets of change, but at the inferential processes at work in the flow of speech, and the local continuities across times and texts. Here the focus was on associative meanings arising in the production and reproduction of discourse. Much of the work on unidirectionality in semantic change was done in the context of grammaticalization, for example Traugott (1982), Traugott and König (1991), Bybee and Pagliuca (1985), Bybee, Perkins, and Pagliuca (1994). But this was only an entry-point into the large territory of semantic change in general. Historical semantics and pragmatics is of course a very diverse field. Studies in semantics range from focus on the sociocultural macro-context of changes in discursive practices to focus on the micro-contexts of meaning development in communication, including locally constrained strategic interactions in conversation, narrative, and other genres. One thing that

32

Elizabeth Closs Traugott

these lines of research have made crystal clear is that we can go beyond the well-known contradictory-seeming (and therefore arbitrary-seeming) pairs of semantic change that we have all learned: amelioration vs. pejoration, generalization vs. specialization, euphemism vs. taboo, metaphor vs. metonymy. Likewise we can go beyond the apparent randomness of referencebased meaning changes (such as changes in the meaning of chalk or horse, cf. Eckardt 1999). When we move from thinking about reference change to sense change, and from concepts expressed by nouns to concepts expressed by other parts of speech, especially verbs or conjunctions, a new picture emerges in which unidirectionalities predominate: (2) a. SPACE > TIME not vice versa b. DEONTIC > EPISTEMIC not vice versa c. CONTENT > PROCEDURAL not vice versa 4 d. SOCIOPHYSICAL > EPISTEMIC > METATEXTUAL not vice versa An even larger picture emerges when we move from thinking about words out of context to constructions in context, and from the study of the OED or the MED to study of corpora. Now we find ourselves squarely in the domain of discourse and genre analysis; of rhetorical purposes such as information, explication, persuasion, strategic interaction, and rhetorical force (Fitzmaurice 2000). In this line of work, metaphor and metonymy, generalization and specialization, amelioration and pejoration still play an important role. However, though necessary, they are no longer sufficient. Rather, they are seen as outcomes (epiphenomena, yes) of two mechanisms that are well-known to lead to morphosyntactic change: analogy and reanalysis. Metaphor is closely allied to analogy because it is largely paradigmatic, and metonymy is closely allied to reanalysis because it is largely context-dependent and therefore associative (see Anttila 1989: 141–142). The hypothesis is that as speakers strategize discourse, these two mechanisms guide and constrain the conventionalizing of conversational implicatures. A third important mechanism is subjectification, a mechanism whereby meanings come over time to encode or externalize the speaker/ writer’s perspectives and attitudes as constrained by the communicative world of the speech event, rather than by the so-called “real-world” char-

From etymology to historical pragmatics

33

acteristics of the event or situation referred to (Traugott 1989; Stein and Wright 1995; Traugott and Dasher In press 5). Amelioration and pejoration are allied to subjectification at a relatively trivial level, and the development of metatextual meanings at a more substantive level. These three mechanisms are not only cognitive but also communicative and arise out of communicative practices in the (minimally dyadic) speech situation, and in writing. The overarching directionalities we find are shifts from content meanings (what I used to call “propositional meanings” and Sweetser calls “sociophysical meanings”), to non-truth-conditional, and procedural meanings. Examples include: (3) a. as long as ‘measure of spatial length’ > ‘measure of temporal length’ > ‘conditional’ = ‘provided that’ b. actually ‘in an affective manner’ > ‘epistemic adversative’ > ‘additive discourse marker’ The history of actually illustrates an increase in semantic scope (the manner adverbial is VP-internal, the discourse marker is sentence-external). 6 It, and many others like it, also illustrates a shift from meanings unrelated to discourse to meaning entirely associated with discourse (the content > procedural meaning change). Procedural meanings index metatextual relations between propositions or between propositions and the nonlinguistic context; they cue addressees to speakers’ and writers’ attitudes to the discourse and the participants in it. Therefore many pragmatic/semantic changes, including those illustrated in (2) and (3) can, I hypothesize, be seen as the outcome of subjectification (Traugott and Dasher In press). To return to grammaticalization, we can now better understand how to interpret and probe the hypothesis that constructional or morphosyntactic lexical > functional category change will not occur without prior semantic change. Given the IITSC model of semantic change in which GIINs play a major role, while semantic change often does precede, as in the case of the modals, it is not obvious that semantic as opposed to pragmatic change must precede, in the sense that a new coded polysemy must have arisen before the category shift occurs. What has to happen minimally before grammaticalization can take place is that there is a widely understood, and often exploited GIIN (which is a pragmatic polysemy, see Horn 1985; Sweetser

34

Elizabeth Closs Traugott

1990). This is drawn on and used when a lexical item comes to be grammaticalized. An example is provided by the development of even as a scalar particle from even(ly) ‘smoothly, uniformly’: (4) a. tak perfore a ... bord pætte be smothe shaue by leuel & euene polised (c. 1392 Equatorie of Planets, p. 18) b. trusted me, even att my very first cominge into his noble service (1556 Roper, Life of More, p. 90) c. Yet behold, even they that are wisest amongst us living (1614 Hooker, Sermons St. Jude, p. 5) Although many of the intermediate details are omitted here, one can readily see that an inference from uniformity at the VP adverbial stage (4a) to precise matching must have preceded the grammaticalization to a particle that expresses the speaker’s view of the appropriateness to the situation described of some PP as in (4b) or NP as in (4c). Otherwise even could not have been recruited to the already extant set of scalar particles like only.

3.

A possible model for grammaticalization

Bringing these two threads together, grammaticalization and pragmaticsemantic change, we can replace Newmeyer’s model with a rather different one, Figure 2. This still shows grammaticalization as involving a subset of pragmatic-semantic, morphosyntactic, and phonological changes, but it is not the default intersection of those changes. Grammaticalization carves out a special domain: precisely those changes that are lexical to functional category changes, i.e. instances of primary grammaticalization. None of the changes is specified as being directional (contrast “morphosyntactic change” with Newmeyer’s “downgrading reanalysis”, “phonological change” with his “phonetic reduction”). Secondary grammaticalization is assumed to be a possible by-product of rapid speech, which can lead to changes especially in morphosyntax and morphophonology.

From etymology to historical pragmatics

GR

35

PSC

Lex > FCat MSC

PC

Figure 2.

The relationship of grammaticalization to pragmatic-semantic, morphological, and phonological change. Legend: GR = grammaticalization, PSC = pragmatic-semantic change, MSC = morphosyntactic change, PC = phonological change, Lex = lexical item or construction, FCat = functional category.

Where is the directionality? Not in the changes themselves. Rather, it is conceptualized as enabled by and resulting from speakers using the lexical items or constructions, such as actually, as long as, any way, even, must, be going to in the flow of speech. If the morphosyntactic contexts in which the IINs arise are replicated often enough and become salient enough in the community, GIINs will arise. These widely used and understood GIINs enable the innovating speaker to recruit constructions and lexical items to functional categories.

4.

Challenges and tasks

Despite a century of vigorous research, much remains to be done. Most of my suggestions could and should be carried out in the context of research on any language that has a long textual history and so can give substance to claims about the order of developments across time. But English studies

36

Elizabeth Closs Traugott

have paved the way for research on many other languages, and should continue to be on the fore-front of theoretical work. To the extent that historical pragmatics can shed light on unidirectionality in grammaticalization, far more needs to be understood about pragmatics, especially of meanings that arise in language use. This is a relatively new problem in linguistics. The nature of explanation is, however, an old problem. Though we are probably on the right track, I do not think we have really progressed very far. To say that changes in the trigger experience explain differences between one individual and another is true, but it does not explain why the trigger experience is different. To say that there are competing motivations is also true. But have we really progressed very much beyond Jespersen’s insight that language change results from the demands of “a maximum of efficiency and a minimum of effort” (1959: 324)? He greatly valued the language “which is able to express the greatest amount of meaning with the simplest mechanism” (324). This sounds like Slobin’s “be clear” vs. “be quick” (1977), Langacker’s (1977) “perceptual optimality” vs. “signal simplicity”, Kiparsky’s “optimization” (1996). Preferring to seek explanation in one rather than two activities, Roberts has focused on the “ease theory” (Jespersen 1959: 263) side of the competition, a solution that best fits an account of grammaticalization after it is well in progress. More recently Haspelmath (1999) has postulated a Maxim of Extravagance instead, in an attempt to constrain Lehman’s (1985) claim that speakers’ attempts to be “expressive” account for grammaticalization. Both solutions fit best the phenomena at the onset of grammaticalization. But in what sense are early uses of e.g. be going to as the expression of future (rather than of motion) really examples of “extravagance”? We need simply remember the famous quotation that just might be a precursor of temporal be going to: (5)

Thys onhappy sowle ... was goyng to be broughte into helle for the synne and onleful (‘unlawful’) lustys of her body (1482 Monk of Evesham)

to see that this hardly appears to be an “extravagance”. Likewise early relatively unambiguous examples do not appear to be extravagant, though some might be intended as humorous, e.g.

From etymology to historical pragmatics

(6)

37

So, for want of a Cord, hee tooke his owne garters off; and as he was going to make a nooze, I watch’d my time and ranne away (1611 Tourneur, The Atheist’s Tragedie [Garrett 2000])

To get a better handle on the century-old insights about possible motivations for grammaticalization, and semantic change we need to understand better such cross-subdisciplinary questions as: a) How do children learn language and what exactly does that have to do with language change (understood as replication or spread of innovations from the individual to the group)? As Eckert has pointed out, an important distinction here is “between evidence of social variation among children that may reflect simple exposure, as in class and ethnic differences, and evidence of the social use of variation” (Eckert 1999: 11, italics added). We badly need well-coordinated long-term studies of language acquisition by children pre-puberty and by adults of all ages that pay attention to those areas of linguistic change of interest to historical researchers. b) What can we learn from combining the study of written text and spoken discourse with a fine-grained rhetorical analysis in which not only propositional but also persuasive, argumentative, and other procedural processes of language use are probed (Fitzmaurice 2000: 5)? What will this tell us about the extent to which children rather than young and older adults really are responsible for change? c) To what extent do changes in written languages reflect changes in spoken language, or are they relatively independent? Here the study of text types and written genres needs to meet up with the study of spoken discourse types and genres; major work in this area has been started, once again at Helsinki, for example by Nevalainen and Raumolin-Brunberg (1995), though the emphasis of this work is on social stratification rather than on genre and register. d) After a hundred years or so of recording, what new hypotheses can we develop about the role of the interplay of meaning and intonation in the history of a language? e) What is the nature of categories, such that we can reasonably identify Lexical > Functional Category? In particular, what is the grammatical status of such categories as “particle”, “marker”, or derivational morpheme such that they can be meaningfully used in positing clines of grammaticalization (Zwicky 2000)?

38

Elizabeth Closs Traugott

f) What is the nature of constructions, such that we can provide stronger theoretical underpinnings to the fact that grammaticalization often begins in a construction, and apparently always results in a new construction for the string in question (e.g. the string be going to did not start out as a construction in any well-understood sense of that term, but ended up as one)? 7 g) Our studies have tended to be far too monolingual, and too focused on the standard language. We need to find out more about non-standard varieties not only to complete the record but to test the worry expressed by Lass (2000) and others that many of the changes we observe might be artifacts of standardization and the pedagogical and discursive practices associated with it. Much has been done on dialects of ME and EMdE (e.g. Rissanen, Ihalainen, Nevalainen, and Taavitsainen 1992; several papers in Bermúdez-Otero, Denison, Hogg, and McCully 2000; Taavitsainen, Nevalainen, Pahti, and Rissanen 2000). But there is a large gap after that regarding historical materials. There is a lot of synchronic material. Görlach has opened up the field in ways unimaginable a decade ago through the series Varieties of English Around the World, and his own work (e.g. 1991, 1995, 1998). Others have concentrated on particular areas of the world, for example on India (e.g. Kachru 1983), or Australia and New Zealand (e.g. Burridge and Mulder 1998), or they have concentrated on types of language development such as creolization (see Baker and Syea 1996 for historical discussion of several creoles; Rickford 1987 and especially Bruyn 1995 provide historical discussion of particular creoles), but in general there is very little that has yet been published that reveals the linguistic history of these varieties of English rather than the history of their speakers (but see Krug 2000). The problems of using records of nonstandard varieties have been carefully documented in Bailey, Maynor, and Cukor-Avila’s (1991) book on evidence for African American Vernacular English. But we have little actual data to ground competing claims about creole or English origins of African American Vernacular English (see e.g. Rickford 1999 vs. Poplack 2000), not only because representations by literate whites of what slaves said is somewhat suspect, but also because we do not know enough yet about non-standard British, Scots-Irish and other spoken non-standard influences on American English (of any variety) prior to this century. h) Finally, if we are to succeed on these endeavors, we need to rethink how we institutionalize the study of the history of English. If for a moment we

From etymology to historical pragmatics

39

imagine multi-volume bookends for the history of English in the twentieth century that are relevant to my themes, we might think of Jespersen’s Modern English Grammar on Historical Principles (1909–49) on the one end and The Cambridge History of the English Language (1992–) on the other. It would appear that much work remains to be included in multi-volume histories of this kind on the linguistics of contact languages, dialectology, semantic change, or historical pragmatics.

Notes 1.

2.

3. 4.

5. 6. 7.

This paper is dedicated to Ekkehard König. The title is adapted from Eve Sweetser’s important book From Etymology to Pragmatics (1990). While the original title was deliberately ambiguous, allowing a diachronic as well as a “range of topics” interpretation, mine is intended diachronically. Members of my grammaticalization seminar, Spring 2000 and the Stanford working group on historical morphosyntax gave me the impetus to think again about grammaticalization, most especially Brady Clark, Luc Baronian, Ashwini Deo, Paul Kiparsky, and above all Arnold Zwicky, whose probing remarks and thoughtful insights are a constant inspiration. Fritz Newmeyer commented on an earlier draft of this paper. To all my thanks. To the extent that I have paid heed to the many suggestions and objections I have received, this paper is the better; to the extent that I have not, it is the poorer. Discussion of particular papers in Campbell (2001) does not appear in this paper, since they were not available to me at the time of the conference, and this paper is intended as a review of twentieth century English historical linguistics. Stern may, however, have laid greater emphasis on syntactic context than either Grice or Levinson would have required [Andrew Garrett p.c.]. For the distinction between “content” and “procedural” meaning see Blakemore (1987) and Wilson and Sperber (1993); briefly, the distinction is between meanings that are context-independent, and those that cue addressees to interpret meanings in context (e.g. deictics and other indexicals). Contrast Langacker (1990, 1999), who restricts subjectification to changes involving the semantics and pragmatics of the subject, as in raising. The best-known example is the scopal increase involved in deontic > epistemic change. For the increasing interest in constructions and grammaticalization, see Bybee, Perkins, and Pagliuca 1994; Traugott In press).

40

Elizabeth Closs Traugott

References Andersen, Henning 1973 Abductive and deductive change. Language 49: 765–93. Anttila, Raimo 1989 Historical and Comparative Linguistics. Amsterdam: Benjamins, 2nd ed. [1st ed. New York: Macmillan, 1972]. Bailey, Guy, Natalie Maynor and Patricia Cukor-Avila (eds.) 1991 The Emergence of Black English. Text and Commentary. Amsterdam: Benjamins. Baker, Philip and Anand Syea (eds.) 1996 Changing Meanings, Changing Functions. London: Westminster Creolistics Series, Volume 2. 1969 Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press. Berlin, Brent and Paul Kay 1969 Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press. Bermúdez-Otero, Ricardo, David Denison, Richard M. Hogg, and C.B. McCully (eds.) 2000 Generative Theory and Corpus Studies. A Dialogue from 10 ICEHL. Berlin: Mouton de Gruyter. Blakemore, Diane 1987 Semantic Constraints on Relevance. Oxford: Blackwell. Bréal, Michel 1964 Semantics: Studies in the Science of Meaning. Trans. by Mrs. Henry Cust. New York: Dover [1900]. Brinton, Laurel J. 1988 The Development of English Aspectual Systems. Cambridge, UK: Cambridge University Press. Brinton, Laurel J. 1996 Pragmatic Markers in English: Grammaticalization and Discourse Function. Berlin: Mouton de Gruyter.

From etymology to historical pragmatics

41

Bruyn, Adrienne 1995 Grammaticalization in Creoles: The Development of Determiners and Relative Clauses in Sranan. Amsterdam: Institute for Functional Research into Language and Language Use. Burridge, Kate and Jean Mulder 1998 English in Australia and New Zealand: An Introduction to its History, Structure, and Use. Oxford: Oxford University Press. Bybee, Joan L 1985 Morphology: A Study of the Relationship between Meaning and Form. Amsterdam: Benjamins. Bybee, Joan L. and William Pagliuca 1985 Cross-linguistic comparison and the development of grammatical meaning. In: Fisiak (ed.) 59–83. Bybee, Joan L., William Pagliuca and Revere D. Perkins 1991 Back to the future. In: Traugott and Heine (eds.), Volume 2: 17–58. Bybee, Joan, Revere D. Perkins and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. The Cambridge History of the English Language. 1992– Cambridge, UK: Cambridge University Press. Campbell, Lyle (ed.) 2001 Grammaticalization: A Critical Assessment. Special issue of Language Sciences 23, Nos. 2 and 3. Claudi, Ulrike and Bernd Heine 1986 On the metaphorical base of grammar. Studies in Language 10: 297–335. Croft, William 1995 Autonomy and functionalist linguistics. Language 71: 490–532. 2000 Explaining Language Change: An Evolutionary Approach. Harlow, Essex, UK: Pearson Education. Dahlgren, Kathleen 1985 Social terms and social reality. Folia Linguistica Historica 6: 107–125.

42

Elizabeth Closs Traugott

Dancygier, Barbara 1992 Two metatextual operators: Negation and conditionality in English and Polish. In: Laura A. Buszard-Welcher, Lionel Wee, and William Weigel (eds.), Proceedings of the Eighteenth Annual Meeting of the Berkeley Linguistics Society, 61–75. Berkeley: Berkeley Linguistics Society. Eckert, Penelope 1999 Linguistic Variation and Social Practice. Oxford: Blackwell. Eckardt, Regine 1999 On the Underlying Mechanics of Certain Types of Meaning Change. Fachbereich Sprachwissenschaft Universität Konstanz Arbeitspapier 104. Fischer, Olga 1994 The development of quasi-auxiliaries in English and changes in word order. Neophilologus 78: 137–64. 2000 Grammaticalisation: Unidirectional, non-reversible? The case of to before the infinitive. In: Fischer, Rosenbach and Stein (eds.), 149–169. Fischer, Olga, Anette Rosenbach and Dieter Stein (eds.) 2000 Pathways of Change: Grammaticalization in English. Amsterdam: Benjamins. Fisiak, Jacek (ed.) 1985 Historical Semantics: Historical Word-formation. Berlin: Mouton de Gruyter. Fitzmaurice, Susan 2000 Some remarks on the rhetoric of historical pragmatics. Journal of Historical Pragmatics 1: 1–6. Garrett, Andrew 2000 Syntactic vs. pragmatic contexts in semantic change; or, a Jacobean farrago. Paper presented at the Linguistics Colloquium, Stanford University, April. Geeraerts, Dirk 1997 Diachronic Prototype Semantics: A Contribution to Historical Lexicology. Oxford: Clarendon Press. Geis, Michael L. and Arnold M. Zwicky 1971 On invited inferences. Linguistic Inquiry 2: 561–566.

From etymology to historical pragmatics

43

Givón, Talmy 1979 On Understanding Grammar. New York: Academic Press. Goossens, Louis 1985 Framing the linguistic action scene in Old and Present-Day English: OE cweDan, secgan, sp(r)ecan and present-day English speak, talk, say, and tell compared. In: Jacek Fisiak (ed.), Papers from the 6th International Conference on Historical Linguistics, 149–170. Amsterdam: Benjamins. Görlach, Manfred 1991 English: Studies in Varieties of English 1984–1988. Amsterdam: Benjamins. 1995 More Englishes: New Studies in Varieties of English 1988–1994, Amsterdam: Benjamins. 1998 Even More Englishes: Studies 1996–1997, Amsterdam: Benjamins. Grice, Paul 1989 Logic and conversation. Studies in the Way of Words, 22–40. Cambridge, MA: Harvard University Press. (Orig. publ. in Peter Cole and Jerry L. Morgan (eds.), Speech Acts, 41–58. New York: Academic Press, 1975). Haspelmath, Martin 1997 From Space to Time: Temporal Adverbials in the World’s Languages. (LINCOM Studies in Theoretical Linguistics 3.) Munich: LINCOM EUROPA. 1999 Why is grammaticalization irreversible? Linguistics 37: 1043–1068. 2000 Why can’t we talk to each other? Lingua 110: 235–55. Heine, Bernd 2000 On the evolution of grammatical forms: The case of reflexive marking. Stanford University, May. Heine, Bernd and Mechthild Reh 1984 Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago: University of Chicago Press. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge, UK: Cambridge University Press. Horn, Laurence R. 1985 Metalinguistic negation and pragmatic ambiguity. Language 61: 121–174.

44

Elizabeth Closs Traugott

Janda, Richard D. 1995 From agreement affix to subject “clitic” – and bound root: -mos > -nos vs. (-)nos(-) and nos-otros in New Mexican and other regional Spanish dialects. CLS Parasession on Clitics 31: 118–139. Jespersen, Otto 1909– A Modern English Grammar on Historical Principles. Copenhagen: 1949 Munksgaard, 6 Volumes. 1959 Language: Its Nature, Development and Origin. London: Allen and Unwin [1922]. Joseph, Brian D. and Richard D. Janda 1988 The how and why of diachronic morphologization and demorphologization. In: Michael Hammond and Michael Noonan (eds.), Theoretical Morphology: Approaches in Modern Linguistics, 193–210. San Diego: Academic Press. Jucker, Andreas H. (ed.) 1995 Historical Pragmatics. Amsterdam: Benjamins. Kachru, Braj B. 1983 The Indianization of English. The English Language in India. Delhi: Oxford University Press. Keller, Rudi 1994 On Language Change: The Invisible Hand in Language. Trans. by Brigitte Nerlich. (Orig. publ. as Sprachwandel: Von der unsichtbaren Hand in der Sprache, Tübingen: Francke, 1990). Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. 1999 Functional categories, morphosyntactic change, grammaticalization. Linguistics 37: 997–1010. Kemenade, Ans van and Nigel Vincent (eds.) 1997 Parameters of Morphosyntactic Change. Cambridge, UK: Cambridge University Press. Kemmer, Suzanne 1993 The Middle Voice. Amsterdam: Benjamins.

From etymology to historical pragmatics

45

Kiparsky, Paul 1996 The shift to head-initial VP in Germanic. In: Höskuldur Thráinsson, Samuel David Epstein, and Steve Peter (eds.), Comparative Germanic Syntax, 140–179. Amsterdam: Kluwer. Koivisto-Alanko, Päivi 2000 Abstract Words in Abstract Worlds: Directionality and Prototypical Structure in the Semantic Change in English Nouns of Cognition. Doctoral dissertation, University of Helsinki. Helsinki: Société Néophilologique. König, Ekkehard 1991 The Meaning of Focus Particles: a Comparative Perspective. London: Routledge. König, Ekkehard and Elizabeth Closs Traugott 1982 Divergence and Apparent Convergence in the Development of ‘Yet’ and ‘Still’. In: M. Macaulay and R.D. Gensler, et al. (eds.), Proceedings of the Eighth Annual Meeting of the Berkeley Linguistic Society, p. 170–179. Berkeley: Berkeley Linguistic Society. Kroch, Anthony S. 1989 Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Krug, Manfred G. 2000 Emerging English Modals. Berlin: Mouton de Gruyter. Kurylowicz, Jerzy 1976 The evolution of grammatical categories. repr. in Jerzy Kurylowicz, Esquisses linguistiques, Volume 2: 38–54. Munich: Fink [1965]. Kytö, Merja 1991 Variation and Diachrony, with Early American English in Focus. Frankfurt am Main: Peter Lang. Labov, William 1994 Principles of Linguistic Change: Internal Factors. Oxford: Blackwell. Langacker, Ronald W. 1977 Syntactic reanalysis. In: Charles N. Li (ed.,), Mechanisms of Syntactic Change. 57–139. Austin: University of Texas Press. 1990 Subjectification. Cognitive Linguistics 1: 5–38. 1999 Losing control: Grammaticalization, subjectification, and transparency. In: Andreas Blank and Peter Koch (eds.), Historical Semantics and Cognition, 147–175. Berlin: Mouton de Gruyter.

46

Elizabeth Closs Traugott

Lass, Roger 2000 Remarks on (uni)directionality. In: Fischer, Rosenbach and Stein (eds.) 207–227. Lehman, Christian 1985 Grammaticalization: Synchronic variation and diachronic change. Lingua e Stile 20: 303–318. 1995 Thoughts on Grammaticalization. Munich: LINCOM EUROPA (originally published as Thoughts on Grammaticalization: A Programmatic Sketch 1. University of Cologne: Arbeiten des Kölner Universalienprojekts 48. Köln: Universität zu Köln. Institut für Sprachwissenschaft. 1982). Lehrer, Adrienne 1985 The influence of semantic fields on semantic change. In: Fisiak (ed.), 283–295. Levinson, Stephen C. 1995 Three levels of meaning. In: F.R. Palmer (ed.), Grammar and Meaning: Essays in Honor of Sir John Lyons, 90–115. Cambridge, UK: Cambridge University Press. 2000 Presumptive Meanings: The Theory of Generalized Conversational Implicature. Cambridge, MA: MIT Press, a Bradford Book. Lightfoot, David 1979 Principles of Diachronic Syntax. Cambridge, UK: Cambridge University Press. 1999 The Development of Language: Acquisition, Change, and Evolution. Malden, MA: Blackwell. Meillet, Antoine 1958 Comment les mots changent de sens. Linguistique historique et linguistique générale. Paris: Champion, 230–80. (Repr. from Année sociologique 1905–1906). 1958 L’évolution des formes grammaticales. Scientia (Rivista di Scienza) 12, No. 6, repr. in Meillet, Linguistique historique et linguistique générale, 130–48. Paris: Champion [1912]. Molina, Clara 2000 Give Sorrow Words. Doctoral thesis, University of Madrid. Nevalainen, Terttu 1991 BUT, ONLY, JUST: Focusing Adverbial Change in Modern English 1500–1900. Helsinki: Société Néophilologique.

From etymology to historical pragmatics

47

Nevalainen, Terttu and Helena Raumolin-Brunberg 1995 Constraints on politeness: The pragmatics of address formulae in Early English correspondence. In: Jucker (ed.), 541–601. Taavitsainen, Irma, Terttu Nevalainen, Päivi Pahta and Matti Rissanen (eds.) 2000 Placing Middle English in Context. Berlin: Mouton de Gruyter. Newmeyer, Frederick J. 1998 Language Form and Language Function. Cambridge, MA: MIT Press, Bradford Books. Pagliuca, William (ed.). 1994 Perspectives on Grammaticalization. Amsterdam: Benjamins. Palander-Collin, Minna 1999 Grammaticalization and Social Embedding: I THINK and METHINKS in Middle and Early Modern English. Helsinki: Société Néophilologique. Pérez, Aveline 1990 Time in motion: Grammaticalisation of the be going to construction in English. La Trobe University Working Papers in Linguistics 3: 49–64. Poplack, Shana (ed.) 2000 The English History of African American English. Malden, MA: Blackwell. Ramat, Paolo 1992 Thoughts on degrammaticalization. Linguistics 30: 549–60. Ravid, Dorit Diskin 1995 Language Change in Child and Adult Hebrew: A Psycholinguistic Perspective. New York: Oxford University Press. Rickford, John R. 1987 Dimensions of a Creole Continuum: History, Texts, and Linguistic Analysis of Guyanese Creole. Stanford: Stanford University Press. 1999 African American Vernacular English. Oxford: Blackwell. Rissanen, Matti, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.) 1992 History of Englishes: New Methods and Interpretations in Historical Linguistics. Berlin: Mouton de Gruyter.

48

Elizabeth Closs Traugott

Rissanen, Matti, Merja Kytö and Kirsi Heikkonen (eds.) 1997 Grammaticalization at Work: Studies in Long-term Developments in English. Berlin: Mouton de Gruyter. Roberts, Ian G. 1993 A formal account of grammaticalisation in the history of Romance futures. Folia Linguistica Historica 13: 219–58. Roberts, Ian and Anna Roussou 1999 A formal approach to “grammaticalization”. Linguistics 37: 1011–41. Slobin, Dan I. 1977 Language change in childhood and in history. In: John MacNamara (ed.), Language Learning and Thought, 185–214. New York: Academic Press. Stein, Dieter and Susan Wright (eds.) 1995 Subjectivity and Subjectivisation in Language. Cambridge, UK: Cambridge University Press. Stern, Gustav 1968 Meaning and Change of Meaning. Bloomington: University of Indiana Press [1931]. Svorou, Soteria 1993 The Grammar of Space. Amsterdam: Benjamins. Sweetser, Eve V. 1990 From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge, UK: Cambridge University Press. Tabor, Whitney and Elizabeth Closs Traugott 1998 Structural scope expansion and grammaticalization. In: Anna Giacalone Ramat and Paul J. Hopper (eds.), The Limits of Grammaticalization, 227–270. Amsterdam: Benjamins. Traugott, Elizabeth Closs 1982 From propositional to textual and expressive meanings; Some semanticpragmatic aspects of grammaticalization. In: Winfred P. Lehmann and Yakov Malkiel (eds.), Perspectives on Historical Linguistics, 245–271. Amsterdam: Benjamins. 1989 On the rise of epistemic meanings in English: an example of subjectification in semantic change. Language 57: 33–65.

From etymology to historical pragmatics

49

1999

The role of pragmatics in a theory of semantic change. In: Jef Verschueren (ed.), Pragmatics in 1998: Selected Papers from the 6th International Pragmatics Conference, Volume 2: 93–102. Antwerp: International Pragmatics Assoc. In press Constructions in Grammaticalization. In: Richard D. Janda and Brian D. Joseph (eds.), A Handbook of Historical Linguistics. Oxford: Blackwell. Traugott, Elizabeth Closs and Richard B. Dasher In press. Regularity in Semantic Change. Cambridge, UK: Cambridge University Press. Traugott, Elizabeth Closs and Bernd Heine (eds.) 1991 Approaches to Grammaticalization, Amsterdam: Benjamins, 2 Volumes. Traugott, Elizabeth Closs and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In: Traugott and Heine (eds.), Volume 1: 189–218. Trier, Jost 1931

Der deutsche Wortschatz im Sinnbezirk des Verstandes. Heidelberg: Winter.

Visser, F. Th. 1963–73 An Historical Syntax of the English Language. Leiden: Brill, 3 Volumes. Warner, Anthony R. 1993 English Auxiliaries: Structure and History. Cambridge, UK: Cambridge University Press. Weinreich, Uriel, William Labov, and Marvin I. Herzog 1968 Empirical foundations for a theory of language change. In: W.P. Lehmann and Yakov Malkiel (eds.), Directions for Historical Linguistics: A Symposium, 98–195. Austin: University of Texas Press. Williams, Joseph M. 1976 Synaesthetic adjectives: A possible law of semantic change. Language 52: 461–478. Wilson, Deirdre and Dan Sperber 1993 Linguistic form and relevance. Lingua 90: 1–25. Zwicky, Arnold 2000 Constructions and morphological categories. Stanford University, May.

Mixed-language texts as data and evidence in English historical linguistics Herbert Schendl

1.

Introduction

It is common knowledge that medieval England was a multilingual country, where a number of languages, especially English, Latin and French were used, partly side-by-side, partly in specific geographical, social or discourse areas. 1 Twentieth-century research has looked into the changing status and roles of these three languages, and equally, the importance of language contact for linguistic change has been investigated by historical linguists, with particular emphasis on lexical, less so on structural borrowing. This research has, however, been almost exclusively based on data from monolingual English-language texts, while mainstream historical linguistics has completely neglected the analysis of mixed-language texts, i.e. of texts which show alternation and mixing of languages in various forms. This is the more surprising since there is a large number of such texts both literary and, particularly non-literary, from most of the attested history of England. 2 While historical linguists are often unaware of the large number of such texts, philologists, medievalists, and literary scholars have tended to hold them in low esteem and not worthy of serious examination. These so-called “macaronic” literary texts were frequently regarded as instances of artificial, sometimes highly artistic language-play or as exercises of clerics and students, while the non-literary mixed-language texts were predominantly seen as reflecting the insufficient language competence of some medieval writer or scribe. 3 This negative attitude towards mixed-language texts mirrors nineteenthcentury views that languages are distinct, clearly separate and separable entities, and that any mixing of languages leads to corrupt texts. As late as 1932, R.W. Chambers expresses this view when he complains that after the Norman Conquest “the English people (...) began to conduct their legal

52

Herbert Schendl

business, first of all in the French language, and then in no language at all”. (Chambers 1932: lxxxii, emphasis mine). What Chambers calls “no language at all” is “this jargon of ‘Law French’”, which he illustrates by the mixed French-English sentence “Il jecte un graund brickbat que narrowly mist”. Modern linguistic theories which rely on an ideal speaker in a homogeneous speech community have not helped to change this prejudice against language mixing. Only slowly has mainstream linguistics begun to realise that research into bilingualism can provide valuable insights into general aspects of language structure and that code-switching is a relevant part of bilingual competence. The achievements of modern sociolinguistics have led to an increasing acceptance of sociolinguistic insights and methodologies in historical linguistics, and the corpus-based sociohistorical approach as well as research in historical pragmatics are well-known cases in point. Modern research in bilingualism has also enabled us to see medieval mixed-language texts in a new light, namely as instances of code-switching, which we will define following Gumperz as “the juxtaposition within the same speech exchange of passages of speech belonging to two different grammatical systems or subsystems” (1982: 59). 4 However, the advances in modern code-switching research have so far not had the impact on historical linguistics that they deserve. Equally surprising is the fact that modern linguistics has not been concerned with the historical dimension of code-switching, even though the debate on “universal” syntactic constraints has for long dominated grammatical research into code-switching. We would claim that historical data provides an important testing ground for competing theoretical models (see the discussion of Archan 2000 in section 4.3.1.). To sum up, both English historical linguistics and modern linguistic research on code-switching have so far either not been aware of or not much interested in code-switching in older texts. It is the aim of this paper to provide a brief state-of-the-art report of research into this neglected area and to emphasise the relevance of these mixed-language texts as data and evidence for English historical linguistics in the hope that this may stimulate further research in the field.

Mixed language texts as data and evidence in English historical linguistics

2.

53

Multilingualism in medieval Britain

Research into bilingualism in general, and into code-switching in particular, has to take into account the macro- and micro-sociolinguistic situation in which the various languages are used. As so often in historical linguistics, however, we are frequently faced with the fact that not all the relevant variables are recoverable for a particular period or, even more so, for a particular text. Among the extralinguistic variables more easily accessible are the status and roles of the individual languages, the extent and evaluation of bilingualism, and the nature of the medieval diglossic situation. In medieval England, as in most European countries, Latin predominated as the high language in many public domains (religion, scholarship, education, administration, law and literature), and it furthermore served as the international lingua franca. The status and functional ranges of English and French, on the other hand, changed more dramatically during the Middle English period. English slowly developed from a low variety mainly used in the private domain to a more and more prestigious one, a development completed with the wide acceptance of an English standard in the Early Modern period. French, on the other hand, changed from the prestige vernacular of the upper ranks of society to a prestigious taught language of culture, and a written language “of record and administration, law and education” (Trotter 1996: 21) from the 13th century onwards (cf. Crane 1995: 105–106, Rothwell 1994: 54, Rothwell 1998). As such it was widely used in literature and, alongside Latin, in public domains such as administration and law up to the 15th century and even beyond – though the linguistic situation had changed decisively from the second half of the 14th century onwards (see Rothwell 1994, Clanchy 1993). 5 The predominance of Latin and French in public domains, in which English was hardly used up to the 14th century, reflects a diglossic situation, though the functional ranges of the different languages were not rigidly separated during much of the time (cf. Machan 1994, Voigts 1996, etc.; for a general discussion of diglossia see Romaine 1995: 33–38). Multilingualism must have been widespread with the educated and literate part of society, i.e. with the producers of the surviving texts, and in particular with the clerks maintaining a multilingual administration (cf. Trotter 1996: 31, Clanchy 1993), though the overall percentage of bi- and multilin-

54

Herbert Schendl

gual speakers seems to have been relatively small. There are numerous medieval reports on bilingual speakers and their use of different languages according to communicative situation (cf. Clanchy 1993, Richter 1979, Short 1980, etc.), and the extant written manuscripts and texts are further proof of the bilingualism of an educated elite (cf. Hunt 1991, Voigts 1989a, Wenzel 1994, etc.). But even monolingual texts mirror the intensive contact between the three languages: French linguistic influence is, for example, visible in Latin and English administrative texts (cf. Rothwell 1994: 48), Latin influence is seen in English ones (Rodríguez Álvarez 1996), as is English influence in French ones (Rothwell 1994: 60), with translations being a special case. Thus it should not come as a surprise that language contact and bilingualism are also apparent in numerous medieval mixed-language texts. These provide direct evidence for code-switching in written texts, some of which are closer to speech than others, and it would be surprising if codeswitching had not existed as a particular discourse strategy in spoken discourse as well, the more so since code-switching is increasingly thought to be a linguistic universal in bilingual communication. While we have a certain amount of information about the overall bilingual situation of medieval England, we know far less about the relevant variables of particular communicative situations in which specific texts were produced and delivered. Similarly, we know little about their influence on code-choice and code-switching. The participants in discourse (writers and addressees) and their multiple relationships, the immediate sociocultural context in which they communicated, as well as the concrete setting of the communicative event, the topic, purpose and function of specific texts, and their level of formality, are frequently not clear, though detailed philological research may help to uncover some of these factors (cf. section 4.2. on “macaronic” sermons). The constant negotiation of codechoice observed in bilingual speech may not apply to the same extent to written texts; but a bilingual writer may equally choose between a monolingual and a mixed-language text, his choice having specific functions and being clearly influenced by the above-mentioned extralinguistic variables.

Mixed language texts as data and evidence in English historical linguistics

3.

55

Types of historical mixed-language texts

As stated above, language-mixing is a widespread phenomenon in medieval texts, though its frequency of occurrence may differ according to text type, genre and level of formality. It is particularly frequent in scientific and medical texts (Voigts 1989a: 95–96), in sermons (Wenzel 1994: 13), business accounts (Wright 1992: 762), and in certain types of administrative and legal texts such as Year Books (Davidson, forthcoming) and depositions. 6 It is not unusual in other prose texts (Edwards 1981, Voigts 1989a: 96), in poetry, drama and religious verse (see below 4.1.), while switching seems to be rarer in letters (cf. Schendl, forthcoming). Most of these texts are testimony for historical “bilingualism in action” (Wenzel 1994, Chapter 6), i.e. for the linguistic performance of individual bilinguals using more than one language within a single text or discourse. Their relevance is in no way diminished by the fact that they represent various levels of written register only – a fact which (socio)historical linguistics in general has to cope with. As recent corpus-based sociohistorical research has shown, features of older speech may be recovered from written sources by systematically comparing texts with different levels of formality including some that are closer to speech. There are evidently very different ways of mixing languages within a single text, and definitions of what constitutes a “mixed-language text” differ; the length of an insertion may vary from single words to phrases, clauses, sentences and extended stretches of text or subtext, and so may the syntactic integration of the material from the individual languages. Not all of these mixed texts are equally interesting from a linguistic point of view. Frequently, medieval manuscripts written by a single scribe contain a number of more or less discrete monolingual texts in different languages, while other texts change language between subtexts, such as recipes in a medical work, lists of property or the deposition of a witness in a legal text (cf. Voigts 1989a, 1996). Such “inter(sub)textual switching” often resembles the situational code-switching typical of diglossia, and may reflect the status of the languages and their domains of use at a given time. Different from such intertextual switches are those switches where the different languages are syntactically and semantically fully integrated (cf. Machan 1994, Schendl 2000b). Such code-switches may be intersentential, i.e. happen between sentences, or intrasentential, i.e. occur “within the

56

Herbert Schendl

clause or sentence boundary” (Romaine 1995: 123; cf. also Myers-Scotton 1993: 4); especially intrasentential switches have been claimed to presuppose a high degree of linguistic competence on the part of the text producer. More difficult to classify are texts that show a high incidence of non-integrated or partly integrated single words from another language, but no foreign language material above the level of lexemes. To some extent, many or even most medieval English texts are “mixed” in this sense due to their high amount of French and Latin loans. The decision whether to classify such texts as monolingual with a high percentage of loan words, as instances of code-switching, or as a specific kind of mixed code is not uncontroversial (see the discussion in section 4.4.2. below). Though taxonomies of mixed-language texts may provide a useful matrix for classification, individual texts often cannot be clearly assigned to one particular type, since different switching patterns may combine within a single text and thus rather form a continuum. (For a more functionally based taxonomy see below 4.2.) The relevant thing about all or at least the vast majority of mixed texts, however, is that they are not the result of insufficient language competence, but presuppose a high level of bilingualism on the part of author, scribe and reader alike. Furthermore, we have clear contemporary evidence of the widely accepted status of such texts for much of the attested history of English (see Schendl 1996, 2000a, forthcoming).

4.

Approaches to older mixed-language texts

Research on mixed texts has so far centred either on a specific text, such as Piers Plowman, or on specific text types or genres (poems, drama, sermons, business accounts, medical texts). With most of these studies the emphasis is on functional aspects of language-mixing in the widest sense of the word, though their theoretical bases have been very different and sometimes conspicuously absent. With no claim to completeness, this section will first look at some literary analyses of “macaronic” poetic and dramatic texts, which view language-mixing as a stylistic device with predominantly literary functions. More recent philological and medievalist research in non-literary texts will be discussed in 4.2. Section 4.3., finally, will survey linguistic studies of literary and non-literary texts based on theoretical in-

Mixed language texts as data and evidence in English historical linguistics

57

sights from modern code-switching research. Though historical research following these principles is still small, there is some promising work in progress. However, as the following discussion will show, there is often no neat separation between these three research traditions, which can fruitfully complement each other in various ways.

4.1 Literary approaches to “macaronic” texts Though there was some interest in “macaronic poetry” before the 20th century, serious research only began with Wehrle’s study on medieval macaronic hymns and lyrics (1933). Wehrle establishes a typology of macaronic poetry from the 13th to 15th centuries and classifies patterns of Latin insertions from a formal literary perspective, viewing them as “a genre of versification” (1933: vii). The second half of the 20th century saw quite a number of literary studies on the aesthetic and poetic functions of language-mixing in macaronic poems, such as Zumthor (1960, 1963) in a European perspective, Harvey (1978) for Anglo-Norman lyrics, or Archibald (1992) for the poems of Dunbar and Skelton. These more recent studies emphasise the often highly artistic stylistic functions of poetic language-mixing. The several hundred Latin insertions in Piers Plowman, a late 14th century religious verse piece, have for some time attracted the attention of literary scholars. As early as (1932), Sullivan looked at the stylistic significance and syntactic integration of these insertions, most of them biblical and religious quotations. Later studies investigated their specific stylistic and rhetorical functions, such as giving authority to the text, but have also uncovered their discourse function in providing the matrix of the whole poem (e.g. Alford 1977, 1992, Machan 1994, with further bibliographical references). For illustration, a passage from Piers Plowman (Text C, V, lines 84–88; Pearsall 1994: 101), in which the original quotation has been changed by Langland to fit the purpose of the passage, is provided under (1): 7 (1) Preyeres of a parfit man and penaunce discret Is the leuest labour pat oure lord pleseth. Non de solo,’ y sayde, ‘for sothe viuit homo,

58

Herbert Schendl

Nec in pane et in pabulo; the pater-noster wittenesseth; Fiat voluntas dei – pat fynt vs alle thynges.’ [‘Prayers of a perfect man and judicious penance Is the most precious work which pleaseth our Lord. Not from the soil’, I said, ‘in truth doth man live, Nor in bread and in food; the Lord’s Prayer witnesses, Let God’s will be done, who provides us with everything.’] The textual functions of code-switching in medieval drama are discussed in Diller (1997/98). While French, as the language of power, characterises speakers of the powerful elite, “Latin as the divine language (...) symbolises authority, not exclusiveness” (1997/98: 519) and serves as the language of communication between God and humanity. Diller’s analysis of the functions and structural use of code-switching as well as of the integration of Latin material into the English text has a predominantly literary perspective, but is clearly influenced by modern code-switching theory.

4.2 Philological and medievalist approaches The increased medievalist interest in non-literary texts in the last decades has led to some awareness of the high frequency of language mixing in such texts. Of the 178 manuscripts with scientific and medical writings from between 1375 and 1500 investigated by L.E. Voigts, almost half were bi- and trilingual (Voigts 1989a: 95–96). Similarly, S. Wenzel claims that a high percentage of sermons “produced in England during the century from 1350 to 1450 (...) show some degree of such mixture” of Latin and English (1994: 13). These facts must be seen as a reflection of the size of a multilingual audience and/or readership. The research carried out by these two authors into mixed medieval sermons (Wenzel 1978, 1994) and scientific and medical texts (Voigts 1989a, 1989b, 1996, etc.) deserves particular attention. In spite of their mainly medievalist focus (which also includes monolingual texts) both authors explicitly refer to modern research in bilingualism and code-switching, whose basic terminology they adopt. Both also provide (partly overlapping) taxonomies of the mixed-language texts of their respective text types. Wenzel’s rather detailed functional classification will be discussed here

Mixed language texts as data and evidence in English historical linguistics

59

briefly for illustration. According to the functions of the switched English language material within a Latin sermon, he distinguishes between a-elements, which “basically translate a part of the Latin discourse or are imported into it in the manner of quotations” (1994: 17), and b-elements, which are “English words, phrases, clauses, periods, or paragraphs that serve as divisions, subdivisions, or distinctions in the sermon in which they appear. These functional elements are essential for the structure of the scholastic sermons” (1994: 17). With the third category, so-called c-elements, switching between Latin and English occurs frequently within the sentence or clause, without any apparent motivation, so that “[t]he English material (...) forms syntactically integrated parts of bilingual prose sentences” (1994: 22). On the basis of the occurrence of these elements and the nature of the matrix, i.e. dominant language (Latin or English), Wenzel distinguishes five types of mixed-language sermons; only sermons having c-elements, which are predominantly intrasentential switches, are termed “macaronic” and form the corpus of his study. Both researchers basically agree that code-switching in these late-medieval texts is the result of a “bilingual culture [which] exploited the linguistic resources of two languages” (Voigts 1996: 823), frequently “as (...) effective discourse strategy” (Voigts 1996: 817). We have emphasised above the importance of extralinguistic variables for the study of code-switching, and it is here that a philological and medievalist analysis of texts can provide the basis for any linguistic study. As a result of his close examination of the sermon manuscripts as well as of the author, topic, occasion, and intended audience of the individual sermons, Wenzel (1994, Chapter 3) shows that there are favourite topics of and preferred dates for macaronic sermons and that they were partly addressed to the bilingual clergy, partly to a mixed audience of clergy and laity (1994: 71–73). The passage under (2) shows part of a macaronic sermon, which according to Wenzel may have been delivered in this macaronic form – a view which contradicts mainstream theories about language use in medieval preaching (cf. 1994, Chapter 6; see also Fletcher 1994: 240–241). (2) Sermo pro toto aduentu Domini (middle 15th century; source: Wenzel 1994: 125–126) Dixi eciam quod venit in hominem et doith omni die by grace of łe godhede. Karissimi, debetis intelligere quod it farith per graciam Dei

60

Herbert Schendl

as it doith per solem. Videtis bene ad oculum quod quando sol splendet bryZte, splendet ita prompte in loco qui est inmundus sicut in loco qui est mundus, in domo qui est clausus sicut in domo qui est apertus, in casu illi qui fuerint in domo clauso voluerint aperire hostia et fenestras et letyn it in. Veraciter recte sic it faryZth per graciam Dei: it is ita prompta malo sicut bono, peccatori sicut iusto, ita quod non est defectus in Deo qui tribuit graciam; set si aliquis defectus sit, it is of mennys wicked hertis qui possent recipere graciam Dei et nolunt. [I have further said that he comes to people and does [so] every day through the grace of his godhead. Beloved, you must understand that it goes with the grace of God as it does with the sun. You see well with your eyes that when the sun shines brightly, it shines as readily on a place that is unclean as on one that is clean, in a house that is closed as in one that is open, if those who are in the closed house will open the doors and windows and let it in. Truly, in just this way it goes with God’s grace: it is as ready for an evil person as for a good one, for a sinner as for a righteous person, so that there is no fault in God, who distributes his grace. But if there is any fault, it rests in people’s wicked hearts, who could receive God’s grace and do not wish to.] For reasons of space we cannot discuss some other relevant work on bilingualism and code-switching in the medievalist tradition, but the following two should be singled out: Clanchy (1993) is an excellent comprehensive treatment of the linguistic situation in the early Middle English period, while Fletcher (1994) not only discusses a macaronic sermon in some detail, but also the overall status of the macaronic style in medieval England.

4.3 Linguistic approaches to language-mixing Most linguistic research views mixed-language texts as written manifestation of code-switching and explicitly or implicitly adopts the “uniformitarian principle” that we can use the present to explain the past (see Lüdi 1985: 188, Machan 1994, note 5). In other words, code-switching in written older texts is assumed to follow the same basic principles as switching in modern speech and writing (cf. Lüdi 1985: 172; for the increasing

Mixed language texts as data and evidence in English historical linguistics

61

importance of modern written code-switching data see Jacobson 2000, Section 3). This does not imply, however, that the actual manifestations of code-switching in historical texts are identical to those in modern speech or writing, nor that modern theoretical models can be applied to historical data without modification. Modern code-switching research has traditionally been split up into grammatical-syntactic and functional-pragmatic approaches (including the discourse analytic approach). Grammatical-syntactic studies have especially focused on intrasentential switching (including its relation to intersentential switching), while inter- and extrasentential as well as intertextual switching are particularly amenable to functional-pragmatic analyses, which also take the overall sociolinguistic and communicative context into account. This division has some methodological justification, though recent research has increasingly emphasised the necessity of a multilevel approach for a full explanation of switching (cf. Halmari 1997: 2, 10–11, 67). Historical code-switching research has been successfully carried out both within a syntactic-grammatical and a pragmatic-functional framework, though – as often in historical linguistics – it faces a number of limitations due to lack of data and insufficient information on extralinguistic and textual variables. Furthermore, the language varieties involved in medieval code-switching, in particular those of medieval Latin and French are not always clearly separable (see below 4.4.) and less well described (for neglected differences in the lexicon of Anglo-Norman and continental French see Rothwell 1998).

4.3.1 Syntactic-grammatical studies As a rule-governed phenomenon, code-switching is subject to certain syntactic constraints, and the search for possibly universal constraints has instigated much research in the last two decades. 8 The frequent violations of the proposed categorical constraints in natural data have increasingly led to a redefinition of constraints as probabilistic tendencies which allow for variation and can be overridden by sociolinguistic and pragmatic factors (cf. Muysken 1995: 184, Halmari 1997: 3; for a recent proposal based on Optimality Theory see Bhatt 1997). Syntactic research into historical code-switching started in the 1990s with informal descriptive analyses of switched syntactic constituents and

62

Herbert Schendl

switching points within traditional, surface-based grammatical models, such as Machan (1994: 369–372) for Piers Plowman, Wenzel (1994, Chapter 5) for medieval sermons, Wright (1992, 1994, 1995, etc.) for business accounts, and Schendl (1996, 1997) for macaronic poems. Particularly detailed is Wenzel’s structuralist analysis, which illustrates the syntactic nature and length of the English switches in his corpus of Latin sermons, and provides statistics on the total frequency of switches per sermon, though not on the frequency of particular switched constituents. In Wenzel’s data Latin provides the syntactic matrix into which the English elements are fully integrated (1994: 82). However, the determination of a matrix language is not always as straightforward as in Wenzel’s data, and may present considerable difficulties in historical studies. 9 Both Machan (1994) and Schendl (1997) extensively illustrate switching sites between sentences and clauses as well as between and within major constituents in poetic texts and come to the conclusion that the syntactic switching patterns in their respective corpora basically correspond to those found in modern code-switching (cf. Machan 1994: 368–372; Schendl 1997: 56–60). Their informal comments on the relative frequencies of the main patterns are, however, not backed by any statistical information. This equally applies to the analyses of switched constituents in medieval and Early Modern English mixed-language business accounts analysed in a number of papers by Wright (cf. Section 4.4.2.). Wright’s studies also document diachronic changes in the switching patterns between her medieval and Early Modern data, such as a move from case inflections to prepositional particles, and a change in the distribution of word-classes of the Latin and English lexical material within the texts. These and similar changes form the basis for her classification into youthful, mature, and moribund macaronic business writing (Wright 1999), a diachronic development so far not observed in any of the other genres. 10 Detailed statistical information on switching patterns in a small corpus of 15th century poems and sermons is provided in Schendl (2000b) with the goal of establishing “frequency hierarchies of switched constituents and switching points for particular texts and text types or genres” (2000b: 70). The frequency hierarchies of these two Middle English sets of data, the first with English, the second with Latin as the matrix language, are then compared to statistics provided in three studies of modern codeswitching. This study has confirmed that the differences between the two

Mixed language texts as data and evidence in English historical linguistics

63

sets of medieval data on the one hand, and the modern data on the other hand, are in general not very conspicuous. There do not seem to be any fundamental differences on the abstract level of syntactic constituents between the two diachronic states. Where diachronic differences between Middle and Modern English do occur, “the main difference between switching strategies in the medieval written and the modern spoken data seems to lie less in the presence or absence of specific features or types, but rather in their relative frequencies” (Schendl 2000b: 82). Such differences are “at least partly due to pragmatic and stylistic factors (...) though the different structures of the languages involved may also play some role” (2000b: 82). A check of the attested switching patterns against some major syntactic constraints proposed in the literature has furthermore shown that violations of these constraints do occur, but only constitute a small percentage of the data, i.e. there are certain clearly preferred patterns and a small number of unusual ones – a result which supports the more recent view of constraints as probabilistic tendencies. A valuable attempt to systematically check the validity of syntactic code-switching models and constraints for historical data is made for three such models in an unpublished M.A. dissertation (Archan 2000). After a statistical comparative analysis of switched constituents in a medieval sermon and an Early Modern medical text, Archan checks some of her data against Myers-Scotton’s Matrix Language Frame Model, Poplack’s Equivalence Constraint, and the Government Model of Di Sciullo, Muysken and Singh. For Archan’s data, Myers-Scotton’s model has the greatest explanatory power, accounting for about 80% of the switches, as against 55% for the Equivalence Constraint and a mere 35% for the Government Model (Archan 2000: 136–137). While all the studies discussed so far provide surface-based descriptions of syntactic patterns, Halmari and Adams (2000) apply the Government Model to explaining “case-assignment of the Latin elements in Piers Plowman”, in particular why a Latin insertion entering a case-assignment relation (such as object of preposition or of transitive verb) “appears either with or without the overtly marked Latin case-ending” (quoted from the ESSE-5 abstract). Further syntactic research within specific theoretical models of codeswitching is highly desirable, ideally on a corpus of diachronically stratified texts from different text types and genres, to investigate possible

64

Herbert Schendl

genre-specific and diachronic differences and to test the validity of these models in more detail. 11 We should, however, be aware of the often widely differing linguistic competences of the authors of the extant texts, some of whom evidently mastered the languages used to a high degree, while others relied heavily on a set of fixed phrases, text portions and quotations in their second (and/or third) language(s).

4.3.2 Functional-pragmatic studies The pragmatic and discourse functions of code-switching have been recognised for a long time (cf. Gumperz 1982, Chapter 4), and functional-pragmatic approaches have been central areas of research in recent years. Particular switches may fulfil a variety of specific functions, but not every single switch can be explained in this way; often it is rather the fact of switching per se that carries communicative information and speakers may use code-switching to identify themselves as members of a specific social or ethnic group or to mark their multilingual identity. Frequently, the choice and switching of codes are the result of the negotiation of a speaker’s rights and obligations or influenced by conversational factors. 12 The interpretation of code-switching also depends on the linguistic competence of the participants of discourse and a mixed-language text will carry different connotations for monolingual and for bilingual speakers (cf. Lüdi 1985: 169–171). Research into the stylistic and structural functions of literary languagemixing has been briefly reviewed in 4.1. above and will not be taken up here again. There are, on the other hand, obvious similarities and overlaps between a more general philological or medievalist and a linguisticallyoriented functional approach, which justifies their joint treatment in this section. In general, however, linguistically-oriented functional-pragmatic models of code-switching provide a more consistent framework and terminology, which should be applicable to both non-literary and literary texts. Though a functional-pragmatic analysis of historical data meets a number of difficulties due to insufficient information on extralinguistic variables (see section 2), such an approach has yielded some encouraging results so far. As in modern speech, the functions of code-switching in historical texts are likely to differ according to the status of the switched language, the

Mixed language texts as data and evidence in English historical linguistics

65

period of text composition, and the type, purpose, or audience of a text. In a number of historical studies specific functions of individual, particularly intersentential switches, have been illustrated from different texts or text corpora, partly referring functions in modern speech, as listed in Gumperz (1982) or Romaine (1995). Some attested functions are providing a specific background to a text passage, or contributing to speaker characterisation – both particularly found in medieval poetic texts –, and re-iteration, attested e.g. in poetic, medical and scientific texts (Machan 1994, Schendl 1997, Voigts 1996). A prominent function of switching is marking quotations, which is met across a wide range of medieval genres and text types. However, switched quotations may also fulfil various more specific discourse functions, such as structuring or organising a text (Wenzel 1994 for medieval sermons, Pahta 2000 for medical texts), lending authority to an argument (Wenzel 1994, Machan 1994), or serving as a rhetorical device. Other frequently attested discourse functions of intertextual or intersentential switching are the differentiation of text categories or domains, the distinction of topic and discussion, and of text and commentary, as, for instance, between an English-language recipe and the Latin-language comments or discussions in the same hand (Voigts 1996, Pahta 2000, for medieval medical and scientific texts; Rodríguez Álvarez 1996 for deeds, i.e. administrative texts). The pragmatic force of code-switching in two FrenchEnglish semi-official medieval letters is investigated in Schendl (forthcoming), while more recent epistolary data, namely English-Scots 18th and 19th century private letters, are discussed in Dossena (2000). A contrastive discourse analytic study of switching in Piers Plowman, i.e. a literary English religious text with Latin insertions, and the non-literary Year Books, informal French legal texts with Latin switches, is undertaken in an unpublished Ph. D. dissertation by M.C. Davidson (forthcoming). In the Year Books, “Latin organises and clarifies the abridged French transcriptions of court proceedings” and thus has a “co-operative role” in encoding “oral pleading techniques for apprentices learning common law”, a function which is claimed to be found in literary texts as well (quotations from author’s abstract, personal communication). While the above functions are particularly exploited in intersentential or intertextual switching, such specific functions are less obvious with intrasentential switches. As observed for modern conversational switching, “[t]he very fact that a speaker makes alternate use of both codes, itself has

66

Herbert Schendl

interactional motivations and implications, beyond any particular effects of specific switches. (...) It is then the choice (or not) of this mode which is of significance to participants rather than the choice of switch points” (Poplack 1980: 614). This function as “an overall discourse MODE” (Poplack 1980: 614) seems equally present in medieval texts, such as Piers Plowman (see Machan 1994) and macaronic poems (Schendl 1997). 13 As with syntactic approaches, we need further functional-pragmatic analyses of a wide range of texts before we can be certain that code-switching in the various medieval genres and text types is a uniform linguistic and cultural phenomenon accessible to a common approach. But both earlier and more recent research have shown beyond any doubt that historical code-switching is to a large degree functionally motivated.

4.3.3 Multilevel approaches to code-switching As stated above, a satisfactory explanation of code-switching requires a multilevel approach, which takes syntactic, functional-pragmatic, discursive and sociolinguistic factors into account. Though such an approach is still lacking for most historical data, in some of the above-mentioned studies, particularly Machan (1994) and Wenzel (1994), a wide range of relevant aspects and factors have been taken into consideration. Machan explicitly places his text into the wider sociolinguistic context of its period while Wenzel’s detailed philological analysis of his manuscripts as well as of the topics, dates, authors and audience of the sermons provides much of the sociolinguistically relevant information on the extralinguistic variables in which code-switching occurs (see section 2). This information provides the background for their discussions of the various functions of switching in their respective corpora, but it is not related to their syntactic analyses; in particular, a possible correlation between sociolinguistic and pragmatic factors on the one hand, and violations of syntactic constraints on the other lies outside their theoretical framework, though it might be an interesting objective for further research. The extralinguistic contexts and the lexical-semantic and syntactic structures of the texts are also to some extent included in Wright’s analyses of mixed business accounts, the language of which is seen as a kind of lingua franca serving important communicative functions in business transactions (see also section 4.4.2 below).

Mixed language texts as data and evidence in English historical linguistics

67

4.4 From code-switching to mixed codes The definition of code-switching as the juxtaposition of two codes presents a number of problems, which are particularly apparent with inserted single words and short phrases from language A in a text which is otherwise largely in language B: firstly, the distinction between switching and borrowing; secondly, the related question of a possible distinction between code-switching and a new type of “mixed code” made up of elements of two different codes; and, thirdly, the discrepancy between the perception of different codes by the participants of discourse and the distinction of the two different codes by the analysing linguist. These are central topics of modern code-switching research, which have also been addressed to various degrees in the analyses of historical mixed-language texts.

4.4.1 Borrowing vs. switching The controversial issue of distinguishing between borrowing and switching has profound consequences both for theoretical models and empirical studies alike (for discussions of this problem see Myers-Scotton 1993: 163–207, Romaine 1995: 142–61, Halmari 1997: 16–17 and Chapter 7). All of the parameters proposed, such as the degree of linguistic integration into the recipient or base language, the frequency of occurrence in the speech of the community or the individual, or the presence of bilingual competence, are theoretically or empirically problematic, and even more controversial is Poplack’s concept of “nonce-borrowing” (see Romaine 1995: 62–63). There is also disagreement whether the two phenomena are basically different or rather represent a continuum. 14 In the analysis of older texts an unambiguous and theoretically sound decision seems hardly possible, especially if we take phonological integration and frequency of use as criteria for the distinction (see also Trotter 1996: 29). All of the above-mentioned parameters imply that we can clearly determine the etymology of a particular lexical item and thus ascribe it to a particular language. However, as lexicographers of medieval Latin, English and Anglo-Norman have long realised, such a clear etymological separation of distinct languages may be difficult to make (see Trotter 1996 for Anglo-Norman, Latham 1965, quoted in Trotter 1996: 25 for medieval Latin). According to Trotter, “[a]t best this is inconvenient: at worst, it calls

68

Herbert Schendl

into question what is meant by different languages in the Middle Ages” (1996: 25). This is made worse by the fact that even the direction of borrowing is not always unambiguous, since borrowing not only happened from the more prestigious languages Latin and Anglo-Norman into English, but frequently also from English into medieval Latin and Anglo-Norman (see Rothwell 1994: 48–49, 60; Trotter 1996: 22–23).

4.4.2 Code-switching, linguistic continuum, or new mixed code? Section 4.3. has treated the languages involved in code-switching as clearly separate and separable codes, and there is no doubt that Latin, French and English were distinct languages in the Middle English period. However, in specific texts the different languages may be so intimately mixed or even fused that the question arises whether we are still faced with intrasentential code-switching or rather with a – possibly institutionalised – new mixed code. This possibility is discussed by Wenzel (1994: 127) for macaronic sermons (for a definition of c-elements see section 4.2.): Could the sermons that mix their basic Latin with a substantial number of English c-elements represent an attempt to create a mixed-language made for special bilingual audiences (...) [which] was suited for delivery from the pulpit? If such a view is acceptable, these sermons are products of a linguistic aggiornamento and a stylistic trial that runs parallel to, though in the reverse direction of, the aureate diction used by some contemporary poets.

In spite of its apparent attractiveness, this view does not seem to be supported by the switching patterns of these mixed sermons, which can be described satisfactorily in terms of intrasentential code-switching, with the two languages generally being clearly separate. The situation is, however, fundamentally different in the medieval and Early Modern business accounts investigated by L. Wright, where a much more intimate mixture of Medieval Latin and English or Medieval Latin and Anglo-Norman is attested over a period of about 500 years (see Wright 1992, 1995, 1999, etc.). As mentioned in 4.3.1, the Medieval Latin/English business accounts show a high frequency of isolated single word switches in English in the Latin base text, especially noun phrases, deverbal –ing forms and adjectives, with a diachronic change in the respective use of Latin and English in the Early Modern period. However, an extensive sys-

Mixed language texts as data and evidence in English historical linguistics

69

tem of abbreviations and suspensions often makes it impossible to ascribe “all material in an entry to either Medieval Latin, or Anglo-Norman, or English” (Wright, forthcoming, p. 14). 15 This extensive use of abbreviations and suspensions and the ensuing lack of explicit morphological information results in or at least comes very close to a new mixed code, which “required considerable knowledge of Medieval Latin to write, although not a commensurately great knowledge of Latin to read” (forthcoming, p. 2). The morphological neutralisation of the two languages in these texts leads to a focus on the semantic content of the lexical roots, which are often identical in Latin, Anglo-Norman and English, and a distinction between borrowing and switching is even more problematic in these texts (cf. Wright 1994: 455–456). Wright gives convincing arguments for classifying the language of these texts as representing a special, supra-regional code for accounts-keeping. Their systematic syntactic and lexical mixing or rather “fusing” of languages and their rule-governed variation cannot easily be accounted for by most existing code-switching models, though Myers-Scotton’s Matrix Language Frame Model “comes closest (...) to capturing some of the characteristics of medieval accounts-keeping” (Wright, forthcoming, p. 13).

4.4.3 Linguistic or “interpretive” approach to code Recent code-switching research has emphasised that “the question of what counts as a code is not easily answered” (Auer 1998: 13), and that the linguist’s definition of a particular code is not necessarily identical with the speaker’s perception of code (see Auer 1998: 13–21, and the papers in Part I of Auer 1998). We also have evidence that the historical linguist’s analysis in this respect may differ from the medieval speaker’s perception (cf. also 4.4.2.). Referring to research by T. Hunt and W. Rothwell, Trotter (1996: 26) states that in medieval glosses, “[t]ime and again, the appellations anglice and (especially) gallice refer to the ‘wrong’ language (...), with the balance of inaccuracy (...) seemingly tilted towards gallice, ‘French’, as the generic term for non-Latin”. Even for literate speakers of medieval England the various languages at their disposal were obviously less clearly distinct than often claimed by scholars, and in general “[f]or medieval Englishmen (...) French was not someone else’s language, individual items of which might be ‘borrowed’ for the circumstance” (Rothwell 1994: 56).

70

Herbert Schendl

Frequently, “Middle English and Anglo-French are too closely linked for the simplistic notion of ‘borrowing’” (Rothwell 1994: 63; see also Trotter 1996: 21). Trotter, in discussing the difficulty of deciding which language a given word belongs to, nicely sums up the problem: Neither for the modern lexicographer nor for the user of the languages of Britain in the Middle Ages can clear-cut distinctions between vernaculars, or between vernacular and Latin, be considered an accurate representation of linguistic and social reality. Glossators and writers alike show scant regard for watertight divisions between languages. 16 (1996: 21)

4.5 Code-choice, code-switching and language shift For some centuries, medieval bi- and multilingual speakers in England had to make code-choices according to the nature of the communicative act, in an overall multilingual situation that became less and less stable in the late medieval period. It is of major interest how the attested code-switching fits into this general process which finally led to a shift to English. There is some evidence that the majority of mixed-language texts dates from the transitional period of language shift from French and Latin to English, with some chronological differences according to text types and domain (on the relation between code-switching and language shift see Myers-Scotton 1993: 223–224). Wenzel’s Latin-English sermons (1994) date from the middle of the 14th to the middle of the 15th century, and a similar date (1375–1475) is given for the majority of scientific and medical mixed texts (Voigts 1996: 814; cf. also Pahta 2000), the period of the vernacularization of medicine and science in England. Similarly, some code-switching in semi-official letters occurs at the beginning of the 15th century, only a few years after the first monolingual English letters are attested (see Schendl, forthcoming). Though both Latin and French continued to be used in most of these domains and text types for almost two centuries, the vernacularization process towards English with the development of a national medium of communication was evidently fast advancing in that period. 17 This chronological correlation between language shift and code-switching in late medieval England seems to be too obvious to be due to chance, though further systematic research is needed to support this hypothesis. 18

Mixed language texts as data and evidence in English historical linguistics

5.

71

Conclusion

This paper has tried to present a survey of recent research into historical code-switching, a fairly new field in historical linguistics, and to identify some areas for future research. The study of medieval mixed-language texts has clearly confirmed that code-switching in this period is a central linguistic and cultural phenomenon, which exploits the linguistic potential of a multilingual society. These texts provide invaluable data and evidence for English historical linguistics in a variety of fields, such as historical sociolinguistics, pragmatics, lexicology, language contact and bilingualism. But, as the present survey has also shown, we are still a long way from a comprehensive study of medieval code-switching. We not only lack a comprehensive list of mixed texts, but also still need detailed analyses of a wide range of texts from different genres, text types and domains. Modern research on code-switching has provided us with a suitable theoretical frame for a comprehensive, multilevel approach to these historical texts, inviting further investigation into this fascinating field.

Notes 1.

2. 3.

4.

5.

This paper will only deal with these three languages in the Middle English period and will disregard some less used languages as well as Early Modern mixed-language texts. For a brief survey of mixed texts from different text types and genres, with examples, see Schendl (1996). For the different uses of the term ‘macaronic’ see, e.g. Archibald (1992: 127–128). In the present paper, this term is used in its wider sense as relating to any text which mixes two or more different codes. In general, however, we will refer to such language alternation within one text as language-mixing or ‘code-switching’ to emphasise our linguistic approach. We will only deal here with switching between languages and disregard switching between language varieties. For a discussion of competing terms, such as codemixing, language alternation, etc. see Halmari (1997: 16). For the present purpose we neglect the highly controversial distinction between Anglo-French and Continental French, see Rothwell (1994, 1998). For the role of French in medieval England see e.g. Berndt (1965), Crane (1995), Kibbee (1991), Short (1980).

72 6.

7.

8. 9. 10.

11.

12. 13. 14.

15.

16.

17.

18.

Herbert Schendl As for late medieval administrative texts, Wright claims that “it is easier today to find macaronic documents from the late Medieval period in Record Offices than it is to find monolingual texts” (1992: 762). Some literary scholars have explicitly linked language-mixing in macaronic poems and in Piers Plowman to the same phenomenon in medieval non-literary texts, especially in “macaronic” sermons, guild records and court rolls, see e.g. Alford (1977: 81), Archibald (1992: 148), Machan (1994: 359). For research surveys on syntactic constraints see, e.g. Muysken (1995), Halmari (1997, Chapter 4). For a survey on different ways of distinguishing matrix language and embedded language and some of the difficulties involved see Halmari (1997: 19–21). While the accounts in Medieval Latin and English predominantly show isolated English noun phrases, normally single nouns, in an otherwise Latin text, in the Early Modern “moribund” form of these accounts Latin has become restricted to function words such as numbers, prepositions, conjunctions, etc. and English language material is no longer restricted to single nouns, see Wright (1994: 452), etc.). There is also an Anglo-Norman and English mixed variety with such texts, see Wright (1995). For a different view see Voigts (1996: 818), who rejects the usefulness of “theoretical linguistic models of code mixing (...) for a diachronic analysis of written texts”. For surveys of functions of code-switching see Romaine (1995: 161–165), Halmari (1997: 25–29), etc. For the special status of mixed business accounts see section 4.4.2. Picone (1994: 271) advocates a third category between borrowing and switching, a so-called “buffer code” in which the different codes are neutralised. This concept seems to be applicable for cases such as the medieval business accounts discussed below, see Wright (forthcoming, p. 15). The importance of sigils and “carecters” in late-medieval medical and scientific texts, including mixed-language texts, is also emphasised in Voigts (1989a: 95) and in Rothwell (1994: 51–52). The same problem is addressed for the mixed-language town protocols from multilingual medieval Fribourg (Switzerland) by Lüdi (1985: 180), who also raises the question whether the scribes were aware of their switching, or whether the analysing linguist imposes such a view on the texts. Mixed business accounts, on the other hand, were used for some centuries, but their disappearance by the middle of the 16th century is also linked to the establishment of standard English in this domain (see Wright 1995, and forthcoming). This correlation is also found in the multilingual town of Fribourg, where Latin as the language of local administration was replaced by French around 1360, which had in turn been largely superseded by German by the 1480s, with increasing

Mixed language texts as data and evidence in English historical linguistics

73

French-German code-switching from the late 1450s onwards (Lüdi 1985: 165–168).

References Alford, John 1977 The role of the quotations in Piers Plowman. Speculum 52: 80–99. 1992 Piers Plowman: A Guide to the Quotations. Binghamton: State University of New York, Center for Medieval and Early Renaissance Studies. Archan, Sabine 2000 Functional and structural aspects of code-switching in Middle English and Early Modern English texts. Unpublished M.A. dissertation, Vienna. Archibald, Elizabeth 1992 Tradition and innovation in the macaronic poetry of Dunbar and Skelton. Modern Language Quarterly 53: 126–149. Auer, Peter (ed.) 1998 Code-switching in Conversation. Language, Interaction and Identity. London and New York: Routledge. Bhatt, Rakesh M. 1997 Code-switching, constraints, and optimal grammars. Lingua 102: 223–251. Berndt, Rolf 1965 The linguistic situation in England from the Norman Conquest to the loss of Normandy (1066–1204). Philologica Pragensia 8: 145–163. Chambers, R.W. 1932 The continuity of English prose from Alfred to More and his school. In: Elsie V. Hitchcock (ed.), The Life and Death of Sir Thomas Moore, Knight, sometymes Lord High Chancellor of England, written in the tyme of Queene Marie by Nicholas Harpsfield. London: Early English Text Society, O.S. 186, xlv-clxxiv. Clanchy, Michael T. 1993 From Memory to Written Record. England 1066–1307. 2nd ed. Cambridge, Ma. and Oxford: Blackwell.

74

Herbert Schendl

Crane, Susan 1995 Social aspects of bilingualism in the thirteenth century. In: Michael Prestwich, R.H. Britnell and Robin Frame (eds.), Thirteenth Century England VI: Proceedings of the Durham Conference 1995. Woodbridge: Boydell & Brewer Ltd., 103–115. Davidson, Mary C. forthLanguage mixing and code-switching in England in the late medieval pecoming riod”. Ph.D. dissertation Toronto. Diller, Hans-Jürgen 1997/98 Code-switching in medieval English drama. Comparative Drama 506–537. Dossena, Marina 2000 “A good Scots accent of the mind”: the pragmatic value of code-switching between English and Scots in private correspondence. Paper delivered at ESSE 5, Helsinki, August 2000. Edwards, A.S.G. 1981 Towards an index of Middle English prose. In: A.S.G. Edwards and Derek Pearsall (eds.), Middle English Prose. Essays on Bibliographical Problems. New York & London: Garland, 23–41. Fletcher, Alan J. 1994 “Benedictus qui venit in nomine Domini”: A thirteenth-century sermon for Advent and the macaronic style in England. Mediaeval Studies 56: 217–245. Gumperz, John J. 1982 Discourse Strategies. Cambridge: Cambridge University Press. Halmari, Helena 1997 Government and Codeswitching: Explaining American Finnish. Amsterdam, Philadelphia: Benjamins. Halmari, Helena and Robert Adams 2000 Case-assignment of the Latin elements in Piers Plowman. Paper delivered at ESSE 5, Helsinki, August 2000. Harvey, Carol J. 1978 Macaronic techniques in Anglo-Norman verse. L’Esprit créateur 18: 70–81.

Mixed language texts as data and evidence in English historical linguistics

75

Hunt, Tony 1991 Teaching and Learning Latin in Thirteenth-century England. 3 vols. Cambridge: D.S. Brewer. 2000 Code-switching in medical texts. In: David Trotter (ed.), Multilingualism in Later Medieval Britain. Cambridge: D.S. Brewer, 131–148. Jacobson, Rodolfo 2000 Codeswitching Worldwide II. Berlin: Mouton de Gruyter. Kibbee, Douglas A. 1991 For to Speke Frenche Trewely. The French anguage in England, 1000–1600: Its Status, Description and Instruction. Amsterdam: Benjamins. Lüdi, Georges 1985 Mehrsprachige Rede in Freiburger Ratsmanualen des 15. Jahrhunderts. Vox Romanica 44: 163–188. Machan, Tim W. 1994 Language contact in Piers Plowman. Speculum 69(2): 359–385. Muysken, Pieter 1995 Code-switching and grammatical theory. In: Lesley Milroy and Pieter Muysken (eds.), One Speaker, Two Languages. Cross-disciplinary Perspectives on Code-switching. Cambridge: Cambridge University Press, 177–198. Myers-Scotton, Carol M. 1993 Duelling Languages. Grammatical Structure in Codeswitching. Oxford: Clarendon Press. Pahta, Päivi 2000 The multilingual context of Middle English medical writing. Paper delivered at ESSE 5, Helsinki, August 2000. Pearsall, Derek (ed.) 1994 William Langland, Piers Plowman. The C-Text (Exeter Medieval English Texts and Studies). Exeter: University of Exeter Press. Picone, Michael 1994 Lexicogenesis and language vitality. Word 45: 261–285.

76

Herbert Schendl

Richter, Michael 1979 Sprache und Gesellschaft im Mittelalter. Untersuchungen zur mündlichen Kommunikation in England von der Mitte des elften bis zum Beginn des vierzehnten Jahrhunderts. Stuttgart: Hiersemann. Poplack, S. 1980 Sometimes I’ll start a sentence in English y termino en espanol: toward a typology of code-switching. Linguistics 18: 581–618. Rodríguez Álvarez, Alicia 1996 Latin for specific purposes and latinized English in 15th-century vernacular deeds. Revista de lenguas para fines específicos 3: 255–272. Romaine, Suzanne 1995 Bilingualism. 2nd ed. Oxford: Blackwell. Rothwell, William 1994 The trilingual England of Geoffrey Chaucer. Studies in the Age of Chaucer 16: 45–67. 1998 Arrivals and departures: the adoption of French terminology into Middle English. English Studies 79: 144–165. 2000 Aspects of lexical and morphosyntactic mixing in the languages of medieval England. In: David Trotter (ed.), Multilingualism in Later Medieval Britain. Cambridge: D.S. Brewer, 213–233. Schendl, Herbert 1996 Text types and code-switching in medieval and Early Modern English. Vienna English Working Papers 5: 50–62. 1997 “To London fro Kent/Sunt predia depopulantes”: Code-switching and medieval English macaronic poems. Vienna English Working Papers 6: 52–66. 2000a Linguistic aspects of code-switching in medieval English texts. In: David Trotter (ed.), Multilingualism in Later Medieval Britain. Cambridge: D.S. Brewer, 77–92. 2000b Syntactic constraints on code-switching in medieval texts. In: Irma Taavitsainen, Terttu Nevalainen, Paivi Pahta, Matti Rissanen (eds.), Placing Middle English in Context, Berlin: Mouton de Gruyter, 67–86. forthCode-choice and code-switching in some early 15th century letters. In: coming Peter Lucas and Angela Lucas (eds.), Middle English in Dublin: Papers from the Third International Conference on Middle English. Bern: Lang.

Mixed language texts as data and evidence in English historical linguistics Short, Ian 1980

77

On bilingualism in Anglo-Norman England. Romance Philology 33: 467–479.

Sullivan, Carmeline 1932 The Latin Insertions and the Macaronic Verse in Piers Plowman. Washington, D.C.: The Catholic University of America. Trotter, David A. 1996 Language contact and lexicography: the case of Anglo-Norman. In: Hans F. Nielsen and Lene Schøsler (eds.), The Origins and Development of Emigrant Languages. Proceedings from the Second Rasmus Rask Colloquium, Odense University, November 1994. Odense: Odense University Press, 21–39. Voigts, Linda E. 1989a The character of the carecter: Ambiguous sigils in scientific and medical texts. In: A.J. Minnis (ed.), Latin and Vernacular. Studies in Late-medieval Texts and Manuscripts. Cambridge: D.S. Brewer, 91–109. 1989b Scientific and medical books. In: Jeremy Griffiths and Derek Pearsall (eds.), Book Production and Publishing in Britain, 1375–1475. Cambridge: Cambridge University Press, 345–402. 1996 What’s the word? Bilingualism in late-medieval England. Speculum 71: 813–826. Wehrle, William O. 1933 The Macaronic Hymn Tradition in Medieval English Literature. Washington, D.C.: The Catholic University of America. Wenzel, Siegfried 1978 Verses in Sermons. Fasciculus morum and its Middle English Poems. Cambridge, Ma.: Medieval Academy of America. 1994 Macaronic Sermons. Bilingualism and Preaching in Late-medieval England. Ann Arbor: University of Michigan Press. Wright, Laura 1992 Macaronic writing in a London archive, 1380–1480. In: Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.), History of Englishes. New Methods and Interpretations in Historical Linguistics. Berlin: Mouton de Gruyter, 762–770. 1994 Early Modern London business English. In: Dieter Kastovsky (ed.), Studies in Early Modern English. Berlin: Mouton de Gruyter, 449–465.

78

Herbert Schendl 1995

A hypothesis on the structure of macaronic business writing. In: Jacek Fisiak (ed.), Medieval Dialectology. Berlin: Mouton de Gruyter, 309–321. 1999 Macaronic business writing: five hundred years of code-switching. In: Ernst H. Jahr (ed.), Language Change. Advances in Historical Sociolinguistics. Berlin: Mouton de Gruyter, 99–118. forthLexical and grammatical convergence and language shift in London coming mixed-language business writing, 1066–1550. In: Raymond Hickey (ed.), Festschrift for Roger Lass. Elsevier. (quoted from manuscript) Zumthor, Paul 1960 Un problème d’esthétique médievale: l’utilisation poétique du bilinguisme. Le Moyen Age 66: 301–336, 561–594. 1963 Langues et techniques poétiques à l’époque romane (onzième-treizième siècles). Paris: Klincksieck.

Dialectology and the history of the English language William A. Kretzschmar, Jr.

A few years ago I delivered a paper at a meeting of the American Dialect Society entitled “Why Dialectology” (Kretzschmar 1996a). The paper proposed that, generally speaking, dialectology was not about the discovery and description of neat dialect areas each with its own linguistic system, but instead responded to language variation more comprehensively. Dialectology recapitulates the normal human interest we take in how the people we talk to talk differently than we do ourselves; we all have a lifetime’s experience in matching different ways of speaking with the places or groups from which different speakers come. Dialectologists just work more systematically to chart particular linguistic features, both in geographical space and in terms of other characteristics of human society, such as gender, education, age, and so on. The paper suggested that what most people call “dialects” are actually perceptual categories, based on our universal habit of matching different perceived ways of speaking with different places and groups – and thus there is no real object (such as a linguistic system) called a “dialect” for dialectologists to find and affirm. These are ideas to which I will return. The reason for mentioning this “Why Dialectology” paper is that Robert Stockwell, a fine English historical linguist, came up to me after the talk and asked why I had not included historical linguistics in my rationale for dialectology. Of course he was right: the field of dialectology began in connection with historical linguistics and language change, and it can make a contribution to historical study of language today. What I would like to do in this essay is to make up for what I left out before, to assert what it is that dialectology has to offer to historical linguistics and to the history of the English language. Let us begin, if not at the beginning then at least at a significant moment in the history of dialectology. It is generally agreed that the field of dialectology began with Georg Wenker’s survey of the German-speaking area in Europe beginning in 1876 (Wenker and Wrede 1895). There were of course

80

William A. Kretzschmar, Jr.

people who were interested in language variation before that. One might mention from our Western tradition the Greek and Latin playwrights who represented literary dialects (Aristophanes and Petronius are good examples), and the grammarians who have decried popular barbarisms and solecisms such as the author of the third or fourth-century AD Appendix Probi (Hall 1974: 73). These sources tell us not only that people were interested in dialects and variation, but also tell us a little about what kind of variation existed at the time. Sever Pop has offered an interesting timeline of notable events and publications related to language variation stretching from the twelfth century up to 1948 (Pop 1950: 1.xxiii–lv; see also 2.1179–97), including, for example, Dante’s fourteenth-century commentary on Italian dialects, Joachim du Bellay’s 1549 suggestion that dialect words be used along with Latin and Greek sources for the enrichment of the French language, the proclamation by King Gustav of Sweden in 1630 that advocated the study of dialects and place names, and finally, on the heels of the French Revolution, the 1794 proclamations by the new French government which required each commune to have a teacher of (standard) French and which attempted to bar the use of local patois, which it considered to be evidence of resistance to the government. Plus ca change, plus c’est la meme chose. Interest in dialects of a more academic sort arose in the nineteenth century in conjunction with the general rise of interest in linguistics, and in particular with historical linguistics among the Neogrammarians. Robins has written (concisely) that “the neogrammarians wished to make historical linguistics an exact science in line with those natural sciences which had made such striking advances in the nineteenth century” (1979: 184). He then quoted Osthoff and Brugmann as writing what Robins called “a rather brutally expressed paragraph” that “Only that comparative linguist who forsakes the hypothesis-laden atmosphere of the workshop in which Indogermanic root-forms are forged, and comes out into the clear light of tangible present-day actuality in order to obtain from this source information which vague theory cannot ever afford him, can arrive at a correct presentation of the life and the transformations of linguistic forms” (1979: 184–85). For the Neogrammarians present-day languages were the product of their history, and therefore any “correct” representation of linguistic facts must include their historical development. The “tangible present-day actuality” which stood in contrast to rationalist speculation referred to two

Dialectology and the history of the English language

81

kinds of facts: evidence from manuscripts or inscriptions about the historical features of languages, and also the testimony that contemporary dialects might offer. So, dialectology per se began with Wenker because he was the first to conduct a dialect survey, and thus scientifically and systematically to gather facts about dialects. Survey research began as the hallmark of the field because of Neogrammarian positivism, and it remains so to this day. As a result of the great influence of the Neogrammarian position, again to echo Robins, in some sense “we are all neogrammarians now” (1979: 182), not only the historical linguists and dialectologists among linguists. That being the case, I have always been struck by the irony that dialectologists, from the beginning, have been among the strongest critics of the central Neogrammarian position on the Ausnahmslosigkeit der Lautgesetze, that sound changes develop without exception in the same environments within a single dialect or language. This mechanical view of sound change arose in contrast to previous speculative work which was thought to be unscientific because it did not account systematically for the evidence. It was also a perfect fit for Schleicher’s Stammbaum theory, because a uniform mechanical change within one variety could differentiate that variety from others in which the change did not occur, and thus create the branching of the linguistic tree. Unfortunately dialectologists have failed to confirm the Neogrammarian position, even from the earliest studies. Walther Mitzka, the third director of the project initiated by Wenker, described the early results of his project: Im philologischen Streit um die Ausnahmslosigkeit der Lautgesetze sollten diese Zeugnisse der Mundarten rings um Düsseldorf den geographischen Beweis für die Anhänger dieser Lehre erbringen. Das Ergebnis führte zur gegenteiligen Meinung. 1885 berichtet er vor der Philologenversammlung in Gießen: “Ich lebte in der schönen und beruhigenden Überzeugung, diese Characteristika müßten ganz oder nahezu ganz zusammengehen. Jene Voraussetzung erwies sich bald genug als eine durchaus irrige, die Grenzen der vermeintlichen Characteristika liefen eigensinnig ihre eigenen Wege und kreutzen sich oft genug” (Mitzka 1943: 9). [As regards the philological controversy about the exceptionlessness of sound laws, the evidence of the dialects around Düsseldorf should have brought proof for the adherents of this teaching. The results supported the contrary view. [Wenker] reported in 1885 before the Giessen conference of philologists that, “I lived in the fair and calming conviction that these [linguistic] features must completely or nearly completely go together. That assumption turned out soon enough to be utterly mistaken: the boundaries of the contemplated features stubbornly took their own way and often crossed each other.” my trans.]

82

William A. Kretzschmar, Jr.

4. Ungleiche räumliche Verbreiterung gleicher Lautveränderungen: hs in sechs, Ochsen, wachsen im dt. Westen. (Nach Th. Frings, in: Kulturströmungen [s. § 41]) Figure 1.

Bach’s map of the sechs/sess sound change, with Hans Kurath’s annotations (1950: 57). Reprinted by kind permission of Indiana University Press.

Dialectology and the history of the English language

83

What a shock it must have been for Wenker, as he received the more than 44,000 surveys returned to him, that his work undercut the very theory that had motivated him. Adolf Bach has illustrated the problem well in his map of the sechs/sess sound change in three words (Figure 1; 1950: 57, with Hans Kurath’s annotations, see below). We see that this sound change has different geographical boundaries south of Düsseldorf in the three different words. Evidence like this might be taken to support lexical diffusion, as opposed to the Neogrammarian position of mechanical change within a dialect for the entire class of words affected. However, Neogrammarians ever since have labored to find ways to respond to dialect evidence. Perhaps the best recent example is William Labov’s Principles of Linguistic Change (1994), a frank and monumental defense of sound laws which I have reviewed elsewhere (Kretzschmar 1996b) and to which I will return a little later. To go back a little further in time than Labov, in the copy of the Adolf Bach text in the working library of the Linguistic Atlas (now at the University of Georgia), Hans Kurath, the founder of the Linguistic Atlas project in North America, annotated Bach’s commentary. In response to Bach’s statement that “Der Sprachatlas des Dt. Reiches ist geboren aus dem Kampf um die Ausnahmslosigkeit der Lautgesetze (s. §20). Er hat dieses Postulat der Junggrammatiker nicht bestätigt” [The Linguistic Atlas of Germany was born from the position of the exceptionlessness of sound laws. It has not confirmed this postulate of the Neogrammarians. my trans.], Kurath has simply written “Wrong,” underlined, in the margin. In response to Bach’s next sentence, that “Seine Karten lassen vielmehr deutlich erkennen, daß dieselben Laute in verschiedenen Wörtern sich in der Tat nicht in gleicher Weise im Raum entwickeln” [Its maps let us know much more clearly that the same sound in fact does not develop in the same way in space in different words. my trans.], Kurath has simply written in the margin, “Where?” In response to the evidence of Bach’s map, plus two other parallel examples, Kurath writes that “relics do not disprove the rule,” again with underlining (see below for the reference there to Bohnenberger). Finally, Kurath just says “no,” with a double underline and exclamation point, to Bach’s summary statement that “Der Sprachatlas zeigt im Gegensatz zu dieser Annahme, daß im Grunde jedes einzelne Wort und jede einzelne Wortform ihre eigenen Geltungsbereiche, ihre eigenen Grenzen im Sprachraum besitzen” [The Linguistic Atlas shows in contrast that, fundamentally, each separate word and each separate word

84

William A. Kretzschmar, Jr.

form have their own range, its own borders within the speech area. my trans]. More on what underlies such comments in a moment, but the comments are not cryptic; Kurath was a defender of the Neogrammarian position on sound laws, even though he was a dialectologist and was confronted daily with data that required serious accommodation. The chief alternative in dialectology to the Neogrammarians was the approach taken in France, particularly by Jules Gilliéron. Gilliéron wrote, with particular application to French but clearly in opposition to the Neogrammarians, that La réflexion et les faits s’accordent pour détruire cette fausse unité linguistique dénommé patois, cette conception d’une commune ou même d’un groupe qui serait resté le dépositaire fidèle d’un patrimonie latin ... Force nous est donc de repousser le patois comme base d’opération scientifique. Aucune recherche de dialectologie ne partira de cette unité artificielle, impure and suspecte: et à l’étude du patois nous opposerons l’étude du mot. (emphasis original, Gilliéron and Mongin 1905: 27). [Both reflection and the facts come together to destroy the false linguistic unity called “patois” [‘dialect’ or ‘vernacular’], this notion of a community or even of a group which remains the faithful recipient of an inheritance from Latin ... Accordingly we must reject “patois” as the basis for any scientific study. No inquiry of dialectology will take for its starting point this artificial unity, impure and suspect; but against the study of “patois” we will set the study of the word. my trans.]

This statement extended the line of thought expressed by Gilliéron’s sponsor, Gaston Paris, who wrote that there are no dialects but instead a linguistic continuum, that the various patois of France constituted “une vaste tapisserie dont les couleurs variées se fondent sur tous les points en nuances insensiblement dégradées” [a vast tapestry of which the varied colors flow out from locations [i.e. towns] in imperceptible gradations (cited in Jaberg 1936: 32)]. Gilliéron, as we have seen, went beyond Gaston Paris to reject even the local patois as an organizing element in order to pursue his concentration on single words and their history as the center of his approach to linguistics. Gilliéron is supposed by many to be the author of the famous credo, “Chaque mot a son histoire” ‘each word has its own history’ (although I have been unable as yet to recover that phrase from any of his writings). Some of his principal publications, besides the famous Atlas linguistique de France (1902–10), include a monograph on the word scier ‘saw’ (Gilliéron and Mongin 1905), and what he called a “genealogy” of the French words for ‘honeybee’ (1918). His map for the distribution of words

Dialectology and the history of the English language

85

for ‘honeybee’ reveals evidence for sound change, but also a wide array of lexical replacements which call into question a slavish obsession with mechanical sound change (see also Gilliéron 1921, 1922). The primary mechanism of argument in these works was Gilliéron’s ability to display regional distributions in France of different words for the same thing, which allowed him to present detailed arguments about how modern French words seem to have developed from Latin in complex patterns of interaction between words at different historical stages. He wrote (Gilliéron and Mongin 1905: 24): Et elle met en pleine lumière le danger couru par l’étymologiste appuyé sur la phonétique pure: celui d’interpréter comme le produit d’une évolution autochthone régulière et en partant du latin, une forme qui ne s’est implantée sur un point qu’au milieu de XIXe siècle peut-être, une forme d’une romanité tertiaire ou quaternaire, immatriculée par les patois et revêtue par eux d’un aspect auquel les variation de leur vitalité phonétique nous défendent d’accorder la moindre confiance. [And [the application of linguistic geography] puts in plain view the danger run by the etymologist who applies pure phonetic [laws]: the very thing that he interprets as the product of regular evolution in a particular place and as coming from Latin, may actually be a form which was only implanted at the place during the middle of the nineteenth century, a form with only tertiary or quaternary relation to the Romance languages, enrolled by the country speakers and reclothed by them with an appearance which prevents us from granting the least confidence to its phonetic vitality. my trans.]

The replacement of word classes and sound laws by individual words does not so much attack the Neogrammarian position as it substitutes a different set of questions for those addressed by the Neogrammarians. However, in the politics of linguistics, Gilliéron and the Neogrammarians were diametrically opposed and they acted the part. Let us return to American dialectology, and look more closely at Hans Kurath’s accommodation of the facts of language variation to his belief in the Neogrammarian position on sound laws. His clearest statement on the issue occurs in his Pronunciation of English in the Atlantic States (Kurath and McDavid 1961), in a section entitled “How Systematic is Speech?” Kurath put the essential problem this way: “The task of presenting a complicated and fluid linguistic situation in readily intelligible simplified form exposes one to the risk of making the dialects look more regular and systematized than the observed facts warrant. One should, of course, uncover

86

William A. Kretzschmar, Jr.

all the regularities; but one should not overlook recalcitrant data or play them down” (Kurath and McDavid 1961: 2–3). The implication here is that the facts are necessarily messy, and that making them intelligible requires a rhetorical act of arrangement that may, for the incautious reader, be mistaken for a genuine property of the facts themselves. Kurath has recognized that the act of description of certain linguistic facts as a dialect, something that I have discussed elsewhere as “modeling” (Kretzschmar 1992), creates something new, an abstraction which is never objectified in real language behavior. Kurath then made explicit the connection between historical linguistics and the facts of language variation, and it will be worth quoting the passage in extenso: For, though language is essentially systematic, it is never wholly without irregularities and oddities, whatever their origin. This is a simple matter of observation and should surprise no one who is unwilling to forget that all natural languages are historical products developed in the give-and-take between individuals and social groups of a speech community and between speech communities. In this complicated historical process, so different from the creation, once and for all, of an artificial code, features taken from other social and regional dialects are not always adapted to the native system, and innovations in the native system may as yet not be established with consistency, so that elements of an older system survive as relics. To treat such tangible irregularities, current in all natural languages and dialects, as if they were built into the system is to misjudge linguistic realities for the sake of a working theory ... In our handling of the data we single out the inconsistencies and the vacillations in usage after the regularities have been established. Such deviations from the system are indicative of change in progress, of trends in usage. In fact, changes in systematization, which all living dialects undergo from time to time, are inconceivable without temporary disorganization. (Kurath and McDavid 1961: 3).

We see here the old Neogrammarian strategy of talking about dialect borrowing as an explanation for the unsystematic appearance of dialect facts. More important, though, is Kurath’s assertion that innovations and relics, necessary objects in the unending historical process of language change, are not yet or no longer part of the system. Thus Kurath proposes what we can take to be a notion of core and periphery in the linguistic facts, core elements that are part of the linguistic system current at the time of study, and peripheral elements that are not integrated into the system, whether acquired through borrowing or as a consequence of the historical process, and this duality of core and peripheral elements will always characterize natural language. That which is “regular” is by definition part of the system, and

Dialectology and the history of the English language

87

whatever else left over is peripheral. The importance of the peripheral elements is that they may well be signs of change to come or evidence of change accomplished. In some ways Kurath’s pragmatic description flies in the face of the Neogrammarian position, because the linguistic system is accompanied by unsystematic elements for which a mechanical process of change cannot immediately account. At the same time, however, Kurath preserves the notion of a core linguistic system in which sound laws could operate without exception, moreover, a linguistic system which is explicitly the product of its historical development. The practical effect of Kurath’s theory is that the dialectologist (or the historical linguist) is always in the position of having to decide what is part of the core and what is part of the periphery. What makes that task difficult is that, without knowing the boundaries of a dialect in advance – that is, who speaks a particular dialect and who does not speak it – it is impossible to tell what linguistic features occur with regularity and which ones fall outside the system. For example, a dialectologist would have to define one set of regularly occurring features if Canada and the United States are both considered to be part of the same North American dialect, and a relatively larger set for each country if the two places are considered separately. Features that occur regularly only in the United States would be part of its system, but not part of the North American system, and vice versa for Canada. Kurath’s procedure thus runs the risk of circular reasoning, of having to know already what the dialect is before you can discover it. For Kurath, the solution to this dilemma can be found in the use of individual diagnostic features. Kurath recommends that the dialectologist first identify those items from a dialect survey which reveal “fairly clear-cut dissemination patterns” (Kurath 1972: 24; see below, Figure 3, for an example), so as to create a “skeletal structure of the area” with which to evaluate more complicated dissemination patterns of other items. The analytical tool of choice is the “heterogloss,” a line drawn on a map to separate the variants charted. Kurath expects that these heteroglosses will form bundles, claiming that “every systematic sampling survey has revealed this situation” (1972: 24). We see here that the recommended discovery procedure does not actually establish the unity or coherence of a dialect, but instead uses diagnostic features to discover where there is a change in usage – thus the term “heterogloss,” ‘to mark a difference’, instead of the term “isogloss,” which describes an area of consistent usage. Once dialect areas have been sketched

88

William A. Kretzschmar, Jr.

by the discovery of bundles of heteroglosses, the dialectologist can look for any regularities in usage inside of the area. I have pointed out the problems with this approach in other places (e.g. Kretzschmar 1992, 1996c), and suffice it to say here that, in the end, the selection of these diagnostic heteroglosses and the identification of bundles is quite a subjective process; it is not so cut and dried as Kurath would have us believe. So, to complete my thought about the choice of which features belong to the core and which ones to the periphery, if we cannot reliably reproduce the discovery process and identify the speakers of a dialect, we can never know what is regular and what is variable, and we can never get to any linguistic system. Kurath’s account of the procedures of several European dialectologists will clarify the real bases of his use of heteroglosses to discover dialects. Of the work of Ferdinand Wrede, the second director of the German Sprachatlas, Kurath wrote that he accepted traditional nineteenth-century groupings of the German dialects, and described Wrede’s methodological contribution as the notion of “basing each boundary, major or minor, upon a single heterogloss whose course had been established by the Sprachatlas” (Kurath 1972: 85). We see the effect of this procedure in the schematic map of Figure 2: one difference in a linguistic feature divides each of the categories, both Low from Central German ( ik from ich) and Central from Upper German (appel from apfel), and also the subdivisions within each category. Kurath said that this procedure “furnishes the investigator with guidelines for discovering and describing heteroglossic lines of similar trend on which dialect boundaries can be established. Moreover, it commits him to a consistent orientation, so essential in dealing with complicated linguistic situations” (Kurath 1972: 85). Kurath also highly approved of the work of Karl Bohnenberger, who for his work on the Alemannic area had specifically selected diagnostic heteroglosses that had historical linguistic importance (Kurath 1972: 82–83). Kurath in addition discussed similar work in the Netherlands, Italy, the Iberian peninsula, and even France. He cites the work of Karl Ettmayer and Robert Hall, who used the evidence provided in Gilliéron’s Atlas linguistique de France, as rebuttals of Gilliéron’s assertion of the failure of mechanical sound change. Figure 3 shows Hall’s French heteroglosses, constructed to try to determine the situation of the Franco-Provençal variety, each of which is based on unequal historical development of a sound from Latin (Kurath 1972: 97).

Dialectology and the history of the English language

89

Figure 2. Wrede’s map of German dialects (Kurath 1972: 85).

Again, whether or not dialects can actually be “discovered” by this process is radically unclear. Wrede and Bohnenberger in Germany, and Ettmayer and Hall in France, had already accepted a dialect structure for their regions of interest, a structure which originated in popular perception or in work not closely based on the dialect evidence. All that they needed to do was find some feature or bundle of features which agreed with the dialectal structure they already believed to exist. If all of the evidence did not really support their description of a dialect as a systematic, well-bounded entity, that is not necessarily a serious objection; if you believe that the dialect is really there, then the glimpses of the dialect that the evidence does provide are enough to convince you that what you believe to be true is actually true. Let me try to make this point more clearly by use of an analogy: if you believe that there is a white house in the woods, and you see some flashes of white through the leaves and brush, then you will find the white flashes to be convincing evidence for the house even if you never go and find the

90

William A. Kretzschmar, Jr.

Figure 3.

Hall’s French heteroglosses (Kurath 1972: 97). Reprinted by kind permission of Indiana University Press.

Dialectology and the history of the English language

91

house itself. My problem with such an argument is that the white flashes may in fact belong to birds or animals or flowers; it seems better to me to assume, until there is compelling evidence otherwise, that the white flashes belong to the woods and not to some house that may or may not actually be there. My own position is closer to what Gaston Paris and Adolf Bach were talking about; they were interested not in boundaries but in the entire set of features for a given place or area, and how one set of features might be similar or different when compared to the entire set of features from another place or area. The French school, along with Bach and those Marburgers of like mind working on the German Sprachatlas, were interested in seeing the whole white house, in the terms of my analogy. Gilliéron was not interested in sets of features but in separate semantic targets, and the list of variants that he could find from different places; he was interested in cataloguing the different flashes of white in the forest and trying to explain where each one came from. While we may not wish to abandon the notion of dialects entirely, we can certainly see that the accommodation proposed by Kurath does not solve the essential problem or address what the French were interested in; at the very most it suggests that a few historical or popular features might be salient for some dialect that has come to exist in subjective perception. Labov’s (1994) Principles, which greatly expands the ideas of his (1981) article “Resolving the Neogrammarian Controversy,” attempts a defense of the Neogrammarian position on much the same grounds as we have already seen. Since changes from below proceed first and at a faster rate among the interior social groups, a clear view of the change in apparent time then requires that we examine these innovating groups separately in order to get a clear view of the main direction and rate of progress of the apparent change. At the same time, we lose information on the behavior of the variable among the less advanced social groups. In the progressive refinements ... information is lost as well as gained. (1994: 54)

As I wrote in my review, this statement “clearly reflects the notion that progressive redefinition of groups is required for ‘a clear view’ of what is going on. Those not yet convinced of the validity of the procedure may well ask, as my students did, why it is reasonable to exclude black speakers, and later the upper- and lower-class white speakers, when by this act the notion of ‘NYC speech community’ that originally guided data collection is fragmented and lost; even if we grant that black speakers may form a separate

92

William A. Kretzschmar, Jr.

speech community, Labov’s progressive redefinition has removed nearly half of the white speakers, and so his notion of speech community for even the white speakers is sharply limited” (1996b: 200). The review subsequently documented evidence that Labov extracted from the Linguistic Atlas and other sources and sketched the particular view that Labov had to make of the evidence to defend his proposed sound changes, all in order to illustrate the point that “Labov’s drive toward abstraction, to make generalizations in keeping with the Neogrammarian program, must necessarily color his observation of real-time evidence and its difficulties of interpretation” (1996b: 203). It is perhaps ironic, given his detailed and innovative defense of the Neogrammarian position in 1994, that in 1981 Labov had written I would strike from our agenda the questions “Does every word have its own history?” “Is it phonemes that change?” “Are the Neogrammarians right or wrong?” and start a research program of a different sort. We begin with respect ... for the achievements of our predecessors; but that does not mean that we rest content with the data they have gathered. An appreciation of their work is shown, not by the remanipulation of the original observations, but by adding a wider and deeper set of inquiries that will display the value and limitations of these initial results (1981: 305).

This statement might well be read as advocating a radical departure from Neogrammarian doctrines, when in fact Labov strove mightily to add just those enhancements to the traditional Neogrammarian strategies for management of evidence and speech groups which could preserve the traditional central position of exceptionless sound change. The importance of all this for historical linguistics is that, to my mind, there has still never been any very adequate defense of the Neogrammarian position on sound change. Highly subjective and selective use of the dialect data can generate arguments for or against particular interpretations of the development of sound changes, but one has to wonder, given their subjectivity, whether such arguments are even worth making. The findings of dialectologists remain at odds with our traditional ways of thinking about sound change in historical linguistics. It is well to keep in mind, however, that besides a mechanical understanding of sound laws and language change that is open to serious question, the Neogrammarians also promoted the use of scientific methodology and close attention to actual data, two standards under which any of us (at least any of us historical linguists and

Dialectology and the history of the English language

93

dialectologists) could be happy to march. 1 They can also offer an alternative to today’s prevailing rationalist notions of linguistic systems, in their emphasis on individual speakers and their speech habits or Language Custom (Sprachusus; see, e.g., Herman Paul 1889, whose ideas were commented upon rather extensively in Weinreich, Labov, and Herzog 1968). Indeed, the theory of mechanical sound change might be thought to impose an impossible burden on this part of the Neogrammarian program, since mechanical sound change introduces the methodological requirement of a relatively fixed notion of speech communities, when emphasis on individual speakers suggests a more fluid sense of speech communities. We are not obligated to indulge in party-line thinking and accept every plank in a platform, just because we generally prefer the shape of a platform to those of its competitors. This brings me to a more positive side of the relationship between dialectology and historical linguistics. If we free ourselves from an unwarranted defense of mechanical sound change as the main target of either historical linguistics or dialectology, we can use systematic attention to the facts of language variation to help us understand the historical development of languages. In so doing, I would prefer to think along with both the Neogrammarians and the French, which is not only possible but profitable, once the problem of mechanical sound change is no longer in the way. If we are no longer interested in separating some region into dialect areas, we have no need for heteroglosses to mark boundaries. Instead, we can try to describe the distribution of individual features, in space and in time, on their own terms – this is really the heart of what Gilliéron proposed. And we can observe the speech behavior of individuals, and then make generalizations about how such habitual behavior by individuals might best be considered in terms of groups – which is really at the heart of the Neogrammarian program, and which is also quite in accord with the ideas of Gaston Paris. We will be free to accept what Kurath called “a complicated and fluid linguistic situation” for itself, as the normal state of natural languages. We can recognize that language variation is not a distortion of a linguistic system, neither merely borrowing nor just change in progress, but is itself the normal condition of a language. And we can assert that people, being the intelligent creatures that they are, manipulate and exploit the fact of language variation in the maintenance of their social structures.

94

William A. Kretzschmar, Jr.

Therefore, the central role of dialectology will thus be to describe, for the particular time of a survey, the facts of the complicated language behavior in the survey area. These facts make up a body of raw material for historical linguistics to use. There are two particular benefits that dialectology can offer to historical linguistics: first, studies of individual words, and second, description of the situation of language variation today which can enlighten our attempts at historical reconstruction. Studies of individual words may or may not lead to the insight that there are whole groups of words in which the same kind of historical development seems to have taken place, as the Neogrammarians predicted with the theory of mechanical sound change; that will always be a question to be answered, instead of a given. And descriptions of language variation will certainly move us away from old fixed notions of speech communities towards more fluid and flexible constructions, a direction suggested by the Neogrammarian position on idiolects and Language Custom (Sprachusus). As an example of a modern-day word study, I would like to share with you the work of my student, Allison Burkette (2001), on the Linguistic Atlas question designed to reveal terms for a piece of case furniture that you have in your bedroom, often called a bureau or a chest of drawers. Burkette is interested in antiques, and thus had enough specialized knowledge to imagine and then to carry out a thorough study of this item. Figure 4 shows the responses that were collected from 1162 speakers for the Linguistic Atlas of the Middle and South Atlantic States from the 1930s to 1970s (see Kretzschmar, McDavid, Lerud, and Johnson 1993), and also those collected by Ellen Johnson (1996) from 39 Southeastern speakers in her paired-sample survey conducted in 1990. Johnson’s findings match those of the Linguistic Atlas pretty well, except that dresser and chest of drawers have replaced bureau as the most frequent answer to the question. In order to verify that these findings really referred to the same piece of furniture as semantic referent, in 1999 Burkette showed pictures of related furniture items to sixty college students at Georgia, and found that between 30% and 40% of the students used the terms dresser and a similar number used chest of drawers, that the simplex forms chest and drawers were used by about 6% of the students, and that bureau was only used by 3% of the students. A total of 23 different words were used to describe the pictures, a result which also corroborates the finding in the Atlas and Johnson’s follow-up work that a few common terms were supplemented by a great many

Dialectology and the history of the English language

95

peripheral terms (see below). Burkette’s data describes a linguistic change, that dresser and chest of drawers have overtaken bureau as the most common term for this piece of case furniture; what was formerly the most common term now survives only at a low frequency of occurrence. LAMSAS Responses bureau dresser chest of drawers chest sideboard washstand highboy chiffonnier trunk drawers bureau drawers commode dressing table box stand lowboy chest on chest vanity desk Johnson data Responses dresser chest of drawers bureau chest washstand highboy dresser drawers

LAMSAS Occurrences 1104 382 227 44 34 30 27 22 22 19 19 17 9 8 7 5 4 4 3

Occurrences 18 17 5 5 2 2 1

Responses case of drawers dresser drawers stand of drawers set of drawers blanket chest cabinet checkrobes chest upon a chest clothes stand clothespress chifforobe drawers cupboard bookcases cabinet table kast vanity dresser wardrobe wash hands stand

Johnson data Responses bachelor’s chest chest on chest dressing table linen press press vanity dresser

Figure 4. Chest of Drawers data (after Burkette 2001: 140–141).

Occurences 3 3 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Occurrences 1 1 1 1 1 1

96

William A. Kretzschmar, Jr.

It is interesting enough that Burkette could show a change in the most common terms for this piece of furniture, but she also tracked down the origin of every one of the other terms elicited by Atlas field workers. She found, with assistance from the Museum of Early Southern Decorative Arts (MESDA) and the pattern books of colonial furniture makers, that all of the peripheral terms had historical precedent. Trunk, cupboard, and box predated the colonial period; the William and Mary period contributed blanket chest and stand; the Queen Anne period case of drawers and cabinet; the Chippendale period chiffonier, lowboy, chest on chest, chest upon chest, press, wardrobe, and kast; the Hepplewhite period dressing drawers, commode, and dressing table; and finally the Sheraton period, bookcase, as a name for a piece that has shelves above the drawer unit. A few terms were carried over from the set used to designate the larger piece of bedroom furniture, the wardrobe. And, finally, a few of the Atlas terms are compounds, like bureau drawers, or blends, like chifforobe from chiffonier and wardrobe, that were composed from the other available terms. Batchelor’s chest, from Johnson’s survey, is the only term that cannot be explained with reference to earlier furniture history; it appears to be a neologism from the 1960s. Burkette’s study shows us how variation is indeed the trace of change, because the low-frequency responses all reflect historical precedents. I believe that this is what we might expect of word studies, perhaps not that all low-frequency survey responses for all semantic targets will have historical explanations, as in this case, but that each semantic target will have its own particular history that we can document to the extent that the evidence will take us. A different side of this problem has been shown to us by Robert Stockwell and Donka Minkova, though they might well prefer that I not cite the evidence. If we inspect the pronunciation of a single word in an area, we are likely to find variants that either recapitulate different presumed stages in the historical sound changes that affect it, or show different reflexes than we might have predicted. For example, in Figure 5 (from Stockwell and Minkova 1988, following Boisson) we see, for each of several words in the Middle English long, close e class, with the Modern Standard English reflex [i:], the testimony of two individual speakers from the Survey of English Dialects. As one looks down the chart for speaker 39.1, there are various realizations of the vowel, from the familiar modern [i:] to apparently unshifted [e:], which Stockwell and Minkova describe as “mildly surpris-

Dialectology and the history of the English language

97

ing idiosyncracies in this individual’s speech” (Stockwell and Minkova 1988: 362). In the same way, for speaker 39.3, we see a mixture of [i:] and [e:], now along with two tokens of [ai], which Stockwell and Minkova interpreted as evidence that the speaker “was either very confused or very hard of hearing such that the responses cannot be taken as those of a coherent dialect” (Stockwell and Minkova 1988: 361). The evidence from other Hampshire and other Southern speakers, however, shows that [e:] and even [ai] do seem to occur in pronunciations of some words in the class. This is the kind of data that is most inconvenient for discussion of mechanical sound changes, because it shows the various stages of the proposed change within the pronunciation of the same individual as well as more widely in a region. Perhaps the best conclusion, though one that Stockwell and Minkova might not share, is that these speakers, with their associates in Hampshire and nearby counties, really do not have a “coherent dialect” in the frame of reference that we are used to using to describe sound change. other Hamp

other South

[i:]

[i:] [Iu]

[i:] [Iu] [i@]

[i:]

[e:]

[i:] [e:] [ai]

[i:] [Ii] [ I]

[i:] [Ii]

[I ]

[ I]

[I] [i:] [Ii]

[ I] [Ii] [i:]

close e¯

[i:]

[e:]

[e:]

[i:] [Ii]

[i:] [Ii]

yeast

close e¯

[i:] [Ii]

[i:]

[ai]

[i:] [Ii]

niece

close e¯

[i:] [Ii]

[e:]

[ai]

[i:] [Ii] [ei] [e:]

[i:] [Ii] [E] (x1) [i:] [Ii] [ei] [e:]

Figure 5.

Middle English long, close e class, with speakers from the Survey of English Dialects (after Stockwell and Minkova 1988).

Form

ME

ModStE

SED 39.1

fields

close e¯

[i:] [Ii]

[i:]

feed

close e¯

[i:] [Ii]

keep

close e¯

cheese

SED 39.3

98

William A. Kretzschmar, Jr.

However, these speakers do illustrate quite well the situation that we might expect from the Neogrammarian emphasis on the speech behavior of individuals and the creation, in Herman Paul’s term, of the Language Custom (Sprachusus) of the area. We see in Hampshire that individuals apparently have different behaviors (at least for reflexes of ME long close e), and that accordingly the Language Custom of the area is complex with regard to this feature. We can expect that some of the different behaviors might be historical residue (perhaps like the [e:] .pronunciations there) or the beginnings of change in progress, and that some of them (like the [ai] pronunciations) are just, well, different – although not necessarily the result of mental or physical impairment. To be fair to Stockwell and Minkova, their article (one of the best of its decade, and one that I regularly assign to my students) presents an argument against overgeneralization of sound changes into chain shifts, and so rather than criticizing them I am really just going one step further than they did. Their argument, that different English dialects develop their own sound changes in complex ways that make it difficult to hypothesize a language-internal general English vowel shift, is one that I certainly accept. I also appreciate their speculation that “In the socalled chain shift it is more likely that schoolmasters and language-conscious educators made choices in codifying the standard, choices that favored pronunciations that they believed, rightly or wrongly, resulted in greater clarity of expression. These choices resulted in a system which, viewed from the narrow perspective of any version of purely structural linguistics, appeared to form a neat cohesive sequence of events within a homogeneous dialect community which in fact surely never existed” (Stockwell and Minkova 1988: 378). All that I am asserting here is that we should go one step further and accept the complexity and fluidity of individual speech behavior within an area, in addition to talking about the complexity and fluidity of dialects across the areas of a region. There is no denying that individual people just “talk funny,” talk in ways that we cannot entirely predict, and that the unpredictable elements in their speech do not necessarily mean that they are in any way language-impaired. 2 Some of these unpredictable characteristics, whether the result of creativity or chance, whether above or below the level of personal or social consciousness, will get into the Language Custom of some group of people, and at that point we can then begin to track them and assign them to the category of linguistic change. I am not proposing anything new here, just reasserting

Dialectology and the history of the English language

99

what Hermann Paul and other Neogrammarians were saying before we became so preoccupied with ideal linguistic systems that we forgot about the rest of their theory. If we expect to find this sort of complexity in speech behavior, and not the artificial regularity imposed by methodological considerations, and our own daily experience and the dialect evidence consistently say that we should expect it, then we need only ask ourselves how best to describe it. My consistent preference has been to introduce quantitative measures to replace the qualitative decisions that necessarily accompany the notion of mechanical sound change in an ideal system. Since I have written pretty extensively about quantitative methods in other places (e.g. Kretzschmar and Schneider 1996; Kretzschmar 1996c, 1996d), I would like here just to illustrate one outcome of good counting in dialectology and historical linguistics, which I believe might well be elevated to the status of a general model for historical linguistic descriptions. First of all, to go back to Figure 4, we see a quantitative fact that repeats itself seemingly everywhere that we count linguistic types and tokens. There are a few types that are very common, but a great many more that are much less common. Indeed, “one” is the most frequent count of lexical tokens for any linguistic type. This appears to be true for work with linguistic corpora in addition to what we find in our linguistic atlas surveys. For the Atlas bureau question, for instance, only bureau was common enough to be used by more than half of the speakers in the older survey, and no words were used by more than half of the speakers in either Johnson’s or Burkette’s smaller surveys. On the other hand, there were fourteen terms used exactly once by the 1162 speakers from the older survey, and seven terms used exactly once by the 39 speakers from Johnson’s survey, and yet every one of these oncers could be explained by the historical development of the semantic target. While we might have expected some creativity or aberration from speakers in a survey, which we do not happen to see in this case, we cannot explain away the consistent appearance of single occurrences as mere creativity or aberration. It is not only the lexicon that shows this kind of distribution. We have recently been attempting a new kind of analysis of Linguistic Atlas phonetics in which we attempt to decompose the sounds of words into their component vectors. For example, in the pronunciation of the initial vowel of the state name Alabama, we identify every possible level of vowel height and

100

William A. Kretzschmar, Jr.

every possible step of frontness/backness along the front/back vector, as determined by the relative possible positions of the available symbols on the Linguistic Atlas vowel chart (a version of IPA; see Kretzschmar, McDavid, Lerud, and Johnson 1993: Chapter 5). For a word like Alabama, we make a separate listing for each phonic segment. What we find consistently is that there is a wide range of possible realizations – even though Linguistic Atlas field workers have sometimes been accused of oversystematization in their transcriptions (Labov: “Dialectologists in the field are often quite conservative in their notation, and they tend to limit their entries to variants of the forms they have encountered before, even when the sound changes have advanced across several levels of the sound system” (1994: 74). Within the range there tend to be one or two common values and a large number of infrequent values. For the initial vowel of Alabama, we get the a range of results for degrees of height and of frontness/backness from 1036 tokens (Figure 6). For this vowel, we see that [æ] is clearly the most frequent level of vowel height, followed by its raised and lowered variants. There are, however, no fewer than six other possible degrees of vowel height attested in the data. Similarly, for the front/back vector, [æ] is by far the most common position along the vector, followed by its backed variant, yet again in this case there are six other possible degrees of frontness/backness attested in the data. The second vowel of Alabama has mid-central schwa as the most common position for both height and frontness/backness, yet even though weakly stressed it still shows seven different gradations of height and six gradations along the front/back vector. The third vowel of Alabama, again stressed, shows twelve degrees of height and thirteen degrees of frontness/backness, while [æ] has over 800 tokens as the most common position for each vector. We can observe a range of possibilities for every sound, consonants as well as vowels, in every word that we analyze, and most often the range of values is distributed with one or a few frequent variants and a larger number of infrequent variants along each vector. If we were to try the same kind of analysis from acoustical phonetic data, which of course we do not have for the Atlas because most interviews were not recorded, we would observe a similar wide range of possible realizations. We can observe a continuum of values along the F1 and F2 dimensions, for example, in each of the numerous acoustical phonetic plots in Labov 1994. Linguistic Atlas data is not suitable for derivation of frequencies of syntactic variants. This is not the place for a full-scale demonstra-

Dialectology and the history of the English language

101

tion, but it certainly appears that both lexicon and pronunciation appear to conform to ... this distributional pattern. height

front/back

E E lowered

number 1 7 & lowered 1 æ raised 41 æ 928 æ lowered 51 a raised 3 a 3

E

Figure 6.

æ fronted æ æ backed a a backed & fronted A

number 8 4 872 144 4 2 1 1

symbols E = epsilon & = reverse epsilon æ = aesc a = typed a A = script a

Degrees of height and of frontness/backness from 1036 tokens of the initial vowel of Alabama.

bureau dresser chest of drawers chest of drawers sideboard washstand highboy chiffonnier trunk drawers

1104 382 227 44 34 30 27 22 22 19

bureau drawers commode dressing table box stand lowboy chest on chest vanity desk

19 17 9 8 7 5 4 4 3

102

William A. Kretzschmar, Jr. case of drawers dresser drawers stand of drawers set of drawers blanket chest cabinet checkrobes chest upon a chest clothes stand clothes press

3 3 2 2 1 1 1 1 1 1

chifforobe drawers cupboard bookcases cabinet table kast vanity dresser wardrobe wash hands stand

1 1 1 1 1 1 1 1

Figure 7. Example of the asymptotic curve (A-curve)

This state of affairs appears on a graph as an asymptotic curve (Figure 7; I will refer to it by the name “A-curve”), and the A-curve appears to be a basic fact about the distribution of linguistic types and tokens that we can use to our advantage. If we take the A-curve as a model for the distribution of features that we should expect in historical linguistics, we should expect, not the all-or-nothing pattern predicted by mechanical sound change, but instead that there should be one or a few common types for any linguistic feature of interest, and a larger group of types that do not occur very often. This will be true even amongst the rather paltry evidence on which we sometimes have to rely from surviving manuscripts, since the A-curve effect is scalable for data sets of different sizes. In the absence of a sufficient quantity of evidence, the difficulty will be the identification of which types are the common ones and which types are the uncommon ones. Of course good counting will not help if we have so little evidence that every linguistic type is a oncer. Even then, however, if we know about the existence of the asymptotic distribution, we can make historical arguments that make quantitative sense, and avoid arguments that mistake the paucity of evidence for monolithic qualitative proof. Of course the flip side of what I am saying is that a single counter example, or even a few counter examples, can no longer serve as proof that a linguistic change has not occurred for some population of speakers. We should not be interested just in the one or two forms at the top of the A-curve; if that were true, then we would have to say that bureau was the only term for bedroom case furniture that we cared about. Low-frequency forms are just as important a part of the comprehensive distribution of features for some linguistic target as the most common features. Our interest should be in the body of the curve, in as comprehensive a description as we

Dialectology and the history of the English language

103

can make of the linguistic features that correspond to a linguistic target. This does not mean that we can no longer talk about linguistic change, far from it. Rather, it means that we will have a much more flexible model for discussion of linguistic change than we had before, because we can account for frequency differences in the distribution of the components of the curve. We see this in the bureau question in the difference between the older Atlas survey and the more recent work by Johnson and Burkette: it is not true that the term bureau has been lost during the course of the century, but only that it has slipped down the curve while chest of drawers and dresser have climbed it. The notion of “movement along the curve” is shorthand for the logical relationship of different A-curve plots: 1) any feature will occupy a specific place on the A-curve prepared for any single moment in time; 2) the same feature is likely to occur at different locations along the A-curves plotted for several different moments in time; 3) we can imagine that the time sequence of locations for the feature might suggest that the feature is rising on the A-curve (i.e. becoming more frequent over time), or falling on the A-curve (i.e. becoming less frequent over time), or remaining stable in position (i.e. maintaining its frequency) with respect to the frequencies of other features. If we accept the A-curve as a model, we have a new framework for thinking about linguistic change. This argument about distributions which form an A-curve is in no way at odds with discussions of the “S-curve” that has been described for the progress of linguistic change (notably in Kroch 1989, Labov 1994: 65–67). The two curves are different expressions of the same basic distributional facts. The A-curve takes “frequency” for its Y axis, and the different “variant forms” for a given linguistic type as the X axis. The S-curve, on the other hand, takes “frequency” as the Y axis just like the A-curve, but its X axis measures the variable “time.” Thus the S-curve describes the successive frequencies of a single variant form at different moments in time. When I have said that bureau moves “down the curve” over time and chest of drawers and dresser “climb” it, if we were to draw different A-curves that corresponded to different moments in time, I was referring to the same process that is graphed directly for a variant on an S-curve chart. The A-curve and the S-curve are thus complementary quantitative descriptions of the distributional facts of variant linguistic forms. If we embrace quantitative analysis through the use of measures like these two curves, I believe that we can, as Hans Kurath recommended, “uncover all the regularities; but ... not

104

William A. Kretzschmar, Jr.

overlook recalcitrant data or play them down.” Perhaps quantitative description is more demanding in some ways than traditional qualitative analysis, but we now have new tools to help us – computers – and we have a right to expect better, more systematically processed results. Finally, quantitative description of the language variation in the evidence will allow for effective correlational studies of linguistic variables with regional and social variables. Stockwell and Minkova argued persuasively, in the article I have already cited, for “realism” in accounts of linguistic change, and I see correlational study of the language and aspects of the culture of the time as central to the pursuit of realism. We are fortunate in historical linguistics long to have recognized the mutual importance of what we have called “internal” and “external” histories of a language. Our traditional problem has been to forge some real connection between the two. Quantitative description of the kind I have suggested offers a better chance for that connection. As a fine example of what we can attempt, I can cite Nevalainen and Raumolin-Brunberg’s collection called Sociolinguistics and Language History (1996), a group of articles based on the Corpus of Early English Correspondence. Many of our research targets will not be amenable to the close attention to individuals that the Helsinki group has brought to their corpus, but we can keep before us Kurath’s truism that “all natural languages are historical products developed in the give-and-take between individuals and social groups of a speech community and between speech communities.” This point was in fact an important consideration for Kurath in the application of his method of using selective diagnostic heteroglosses: we know that he looked for linguistic features that corresponded to settlement and cultural patterns in order to mark his dialect areas. Just what constitutes evidence for a “cultural pattern” is the topic for another day. Suffice it to say, however, that a model for analysis which preserves the comprehensive distribution of features for a linguistic target, preserves as well the possibility for correlational study of any of the linguistic features with any regional or social variable that the investigator is motivated to define. I am proposing no more than the traditional “who says what where” of dialectology – but the special relevance of quantitative correlational study for historical linguistics is a field just waiting to be further explored as soon as we introduce a more flexible analytical model than what was permitted under the hypothesis of mechanical sound change.

Dialectology and the history of the English language

105

So, what does dialectology have to offer to the history of the English language? Just what we always thought it did. Dialectology records the facts of language behavior at particular moments in time. The massive evidence of dialectology, in England and America and other English-speaking places, is a storehouse for historical linguistic forms that have been preserved at low frequencies of use. Huge amounts of this data already exist online, on my own Linguistic Atlas web site (http://us.english.uga.edu) and elsewhere, ready for use by historical linguists. All we have to do is to document forms of interest, and then plot the linguistic change we see just as Johnson and Burkette have done. And the evidence of dialectology also offers the best available model for the linguistic behavior of individuals and the Language Custom of groups. If human language behavior has not essentially changed in recent millennia, as we must trust that it has not, then the model that we draw from dialectology is precisely the model that we should use for historical reconstruction. We might say, at the end of the day, no longer that variation is the trace of change, but that the new history of the English language could well be the trace of variation. All that we need to do to exploit the findings of dialectology for historical linguistics is to release our insistence on mechanical sound change, while we retain the scientific methods and other traditional aspects of historical linguistics. Labov was quite right when he wrote that we must “begin with respect for the achievements of our predecessors” and that we will advance “by adding a wider and deeper set of inquiries that will display the value and limitations of these initial results” (1981: 305). The whole point of standing on the shoulders of giants is that we are not ourselves the giants, but instead that we take advantage of their achievements and rise above them with our own new insights. Our predecessors were less able than we are to cope with the evident variation in the language that they studied, and so they relegated variation to a subordinate role in the comparative and historical method. We, however, now have the means to record and to analyze language variation, and so to use it for the purposes of historical study. This we should do, in order to get out of the long shadows cast by our predecessors and to bring new vision to our field.

106

William A. Kretzschmar, Jr.

Notes 1.

2.

Robert B. Lees took the opposite point of view in his highly approving review of Chomsky’s Syntactic Structures: It is a common misconception on the part of many a scientist, strange to say, that correct scientific theories are discovered by making many observations of nature, that somehow the right answers just leap up out of the laboratory notebook if only we have measured enough things accurately. … the theories by means of which we order our experiences, on the street or in the laboratory, are generated only by those flashes of insight, those perceptions of pattern, which mark off the brilliant scientist from the dull cataloguer of data (1957: 379–80). It would be hard to imagine a more romantic notion of science than this one, a notion of science more dependent upon intuition and insight. It has been suggested that an inability to manipulate the linguistic variation in a community is actually a pathological condition. Perhaps the best literary example of the point occurs in John Kennedy Toole’s A Confederacy of Dunces (1980), in which the main character is handicapped by his inability to shift out of his Standard English speech in order to deal effectively with the many different dialect speakers in the New Orleans setting of the novel.

References Bach, Adolf 1950 Deutsche Mundartforschung. 2nd ed. Heidelberg: Carl Winter. Burkette, Allison Forth- The Story of Chester Drawers. American Speech. coming Gilliéron, Jules 1902–10 Atlas linguistique de France. Paris: Champion. 1918 Généalogie des mots qui désignent l’abeille. Paris: Champion. 1921 Pathologie et thérapeutique verbales. Paris: Champion. 1922 Les Étymologies des étymologistes et celles du peuple. Paris: Champion. Gilliéron, Jules, and J. Mongin 1905 Scier dans la Gaule romance. Paris: Champion. Hall, Robert A., Jr. 1974 External History of the Romance Languages. New York: Elsevier.

Dialectology and the history of the English language

107

Jaberg, Karl 1936 Aspects géographique du langage. Paris: Droz. Johnson, Ellen 1996 Lexical Change and Variation in the Southeastern United States in the Twentieth Century. Tuscaloosa: University of Alabama Press. Kretzschmar, William A., Jr. 1992 Isoglosses and Predictive Modeling. American Speech 67: 227–49. [reprinted in M. Linn, ed., Handbook of Dialects and Language Variation (Orlando: Academic Press, 1999), 151–72.] 1996a Why Dialectology? RASK 4: 35–49. 1996b Review of William Labov, Principles of Linguistic Change. American Speech. 71: 198–205. 1996c Quantitative Areal Analysis of Dialect Features. Language Variation and Change 8: 13–39. 1996d Dimensions of Variation in American English Vocabulary. English WorldWide 17: 189–211. Kretzschmar, William A., Jr., Virginia McDavid, Theodore Lerud, and Ellen Johnson 1993 Handbook of the Linguistic Atlas of the Middle and South Atlantic States. University of Chicago Press. Kretzschmar, William A., Jr. and Edgar Schneider 1996 Introduction to Quantitative Analysis of Linguistic Survey Data. Thousand Oaks, CA: Sage. Kroch, Anthony 1989 Reflexes of Grammar in Patterns of Language Change. Language Variation and Change 1: 199–244. Kurath, Hans 1972 Studies in Area Linguistics. Bloomington: University of Indiana Press. Kurath, Hans and Raven I. McDavid, Jr. 1961 The Pronunciation of English in the Atlantic States. Ann Arbor: University of Michigan Press. Repr. 1982, University of Alabama Press. Labov, William 1981 Resolving the Neogrammarian controversy. Language 57: 267–309. 1994 Principles of Linguistic Change: Internal Factors. Oxford: Blackwell.

108

William A. Kretzschmar, Jr.

Lees, Robert B. 1957 Review of Noam Chomsky, Syntactic Structures, Language 33: 375–408. Mitzka, Walther 1943 Deutsche Mundarten. Heidelberg: Carl Winter. Nevalainen, Terttu, and Helena Raumolin-Brunberg 1996 Sociolinguistics and Language History. Atlanta: Rodopi. Paul, Hermann 1889 Principles of the History of Language. New York: McMillan. Pop, Sever 1950 La Dialectologie. 2 vols. Louvain: Duculot. Robins, Robert H. 1979 A Short History of Linguistics. 2nd ed. London: Longman. Stockwell, Robert and Donka Minkova 1988 The English Vowel Shift: Problems of Coherence and Explanation. In Dieter Kastovsky and Gero Bauer (eds.), Luick Revisited: Papers Read at the Luick Symposium at Schloss Lichtenstein (Tübingen: Gunter Narr), 355–394. Toole, John Kennedy 1980 A Confederacy of Dunces. New York: Grove Press. Weinreich, Uriel, William Labov, and Marvin Herzog 1968 Empirical Foundations for a Theory of Language Change. In W. Lehmann and Y. Malkiel, eds., Directions for Historical Linguistics, 95–188. Austin: University of Texas Press. Wenker, Georg, and Ferdinand Wrede 1895 Der Sprachatlas des deutschen Reichs. Marburg: Elwert

Origin unknown Anatoly Liberman

... what was her name again? I know it began with a B, and ended with a g, but whether it was Waters or – no it couldn’t have been that, either ... Charles Dickens, Nicholas Nickleby, Chapter 21.

1.

Words of unknown etymology as a problem of etymological lexicography

Every etymological dictionary contains numerous words said to be of unknown origin. They are included but left unexplained. This is also true of English dictionaries. The study of some such words has passed through several cycles. For example, in the 17th and 18th centuries the etymology of stubborn was “known”. At that time, European scholars tried to find look-alikes in Hebrew and Greek, from which most words of modern languages were allegedly derived, and, when they succeeded (and they succeeded amazingly often), their search was over. The Greek adjective στιβαρóς means ‘obstinate, stubborn’. The similarity of sound and meaning convinced Junius (1743) and Bailey (1721) that στιβαρóς is the source of stubborn. Later investigations showed that the two words are not related. Consequently, today we “know” less about the origin of stubborn than did scholars three centuries ago. According to the O.D.E.E. (s.v.), stubborn, from Late Middle English stibourne, is of unknown origin, for connection with OE stubb, stybb, *stobb ‘stub’ cannot be proved. The situation with stubborn should not be treated as an example typical only of the rift between “prescientific” etymology and the etymology based on the achievements of the comparative method. Disagreement among modern scholars is as common in word studies as in the areas of phonology and grammar. One example, chosen almost at random, will suffice. This is

110

Anatoly Liberman

what Skeat (1910, s.v.) says about the word toper in the last edition of his dictionary: TOPER, a great drinker. (F. or Ital.- Teut.) ‘Tope, to drink briskly or lustily;’ Phillips, ed. 1706. ‘The jolly members of a toping club;’ Butler, Epigram on a Club of Sots, I.I. ‘Tope! here pledge me! (drinks);’ Etheredge, The Comical Revenge, A. ii. sc. 3. Certainly connected, as Wedgwood shows, with F. tôper, to cover a stake, a term used in playing at dice; whence tôpe! interj. (short for je tôpe, lit. I accept your offer), used in the sense of good! agreed! well done! It came to be used as a term in drinking, though this only appears in Italian. ‘According to Florio [i.e. in ed. 1688] the same exclamation was used for the acceptance of a pledge in drinking. [He gives]: topa, a word among dicers, as much as to say, I hold it, done, throw! also by good fellows when they are drinking; I’ll pledge you;’ Wedgwood. b. Of Teut. origin; from the striking together of hands or glasses; cf. Picard toper, to strike hands in bargaining, Ital. in-toppare, to strike against an obstacle. Originally from the act of placing together the tops of the thumbs, at the same time crying topp! See topp in Ihre, Outzen, and the Bremen Wörterbuch. Cf. Top (I). Note the aggressive certainly at the beginning of the entry and compare the short entry on tope in the O.D.E.E. (toper is not featured in this dictionary), with its cautious perhaps: tope toup drink (heavily). XVII. perh[aps] alt[eration] of synon[ymous] † top XVI (of unkn[own] origin) by assoc[iation] with † tope (XVII) interjection used as a pledging formula in drinking – F[rench] tôpe. [The dagger sign means ‘obsolete’, the dash stands for ‘(borrowed) from’.]

The O.D.E.E. is a conservative dictionary. It depends almost entirely on the O.E.D. (which made it outdated even before it was published) and follows James Murray’s behest: when in doubt, say nothing, for no etymology is better than a wrong one. The rule adopted by the O.E.D. is hardly applicable to etymological dictionaries, whose goal is not only to reveal the truth but also to discuss controversial problems, but this is not an issue here. More important is the fact that the professed ignorance of the origin of some words turns out to be relative: from some people we hear ‘certainly’, from others ‘perhaps’; still others prefer to reserve judgment altogether. Users of etymological dictionaries will not fail to observe the multitude of verbal clichés employed by authors and editors. Thus the O.D.E.E. states that ship has cognates in all the Germanic languages and goes back to skipam, “of unkn[own] origin.” Shilling has a similar set of cognates, but its

Origin unknown

111

source *skilliŒaz is “of much disputed origin.” The equally ancient noun sword, with cognates all over the Germanic speaking world except Gothic, has been traced to *swerDam “of doubtful origin.” The disputes and doubts are not laid bare, but the art of dodging is manifest (incidentally, dodge is a word “of unknown origin” and has no cognates anywhere). Cob is a relatively recent addition to the vocabulary of Standard English, and what are now hesitantly considered to be its different meanings may go back to three or even four homonyms. The truth is hard to get at, and the O.D.E.E. quite correctly calls cob a word “of obscure origin.” See also spigot “of uncertain origin.” Besides words of much disputed, uncertain, doubtful, obscure, and unknown origin, there are compounds like butterfly and strawberry, apparently butter+fly and straw+berry, but in both cases “the reason for the name is unknown.” Since the discovery of such a reason is the ultimate goal of etymology, no difference of principle exists between ship, shilling, sword, cob, and dodge on one hand and butterfly and strawberry on the other. Nor are butterfly and strawberry in a much better situation than amusing monstrosities like plug-ugly and hot dog, whose origin is “unknown.” While leafing through the O.D.E.E., one finds a few more interesting things. Skimp ‘scanty’ is said to be “possibly related” to its synonym scrimp, but scrimp turns out to be of unknown origin, so being akin to it is of little advantage in the world of etymology. Slum is “of cant origin.” Some words are “of dial[ectal] origin” (see below). Reference to thieves’ cant and dialectal use can hardly be called an etymology. Donkey and its synonyms dicky, neddy, and cuddy have been derived from proper names, but why were Duncan, Dick, Ned, and Cuthbert honored so? Is this a legitimate question, or should we accept all of them, along with john ‘lavatory’, jenny ‘skeleton key’, etc. as being due to pure chance? It may be useful to define the concept “origin unknown,” stop hedging, and distinguish clearly among the many degrees of etymological ignorance. Some words are of obscure history, some are of obscure origin, some are “not basically” or “obscurely” related to other obscure words. Hesitations and discord beset etymologists, and being uncertain is their lot. Despite all their efforts, numerous words are still of “unknown origin.” What are they? To answer this question, we should look at words the etymology of which is known.

112

2.

Anatoly Liberman

Words of undisputed etymology

2.1. The easiest case seems to be onomatopoeia. However, the number of transparent words like cuckoo or moo is small (most verbs of imitative, or echoic, origin, are only presumed to be such), and even they have sometimes lost their ties with natural sounds. The Common Germanic name of the cuckoo had the root *gauk- , but in Old English the change *au >e¯a and palatalization turned *gauk- to ge¯ac, while in Old High German the Second Consonant Shift yielded gouh. With half of the stops gone in each of these forms, the sound complexes ge¯ac, gouh but dimly resemble k/guku(k) that one believes is heard from this bird. Although cuckoo is a perfect example of an onomatopoeic formation, the modern English word is not a native synonym of ge¯ac. It was apparently borrowed from Old French in the 13th century. In due time, the first [u] in cuccu became [ˆ], which is extant in cuckold, and only around 1800 [´kˆku:] was replaced with [´kuku:], to make it sound more like the cuckoo’s cry. If it is true that even miaow is a borrowing from French (it gained currency only in the 17th century, and so did bow-wow and yap; moo is a century older, as far as written records are concerned), it does not come as a surprise that the onomatopoeic origin of verbs is hard to prove. Thump, dump, bump, and jump must rhyme for a good reason (they imitate the sound of a heavy fall or of feet coming to the ground), but each of them has a history of its own, and the situation is further complicated by their late emergence: dump was first recorded in the 14th century, the other three in the 16th. Onomatopoeia is sometimes impossible to distinguish from sound symbolism. A parallel group of predominantly 16th century -mp words (or is it the same one?) – dump, glump, grump, hump, mump, and lump (in like lump) – denotes sulkiness. This makes the origin of lump ‘shapeless mass’, hump (on the back), hunch ‘compress into a hump’, hunk, and HumptyDumpty less obvious than expected. Natural sounds have been studied in great detail, partly in connection with the bow-wow, pooh-pooh, and ding-dong theories of the origin of language, partly for their own sake. See, for example, Curti (1885), Wackernagel (1867), Winteler (1892), Garcia de Diego [no date], and Diensberg (1988). Baby words like daddy, mammy, and biddy ‘birdie’ (cf. chickabiddy) can also be mentioned here. Not all of them are easy to recognize. For instance, the root of the Indo-European word for brother and Engl. boy may

Origin unknown

113

have originated in baby talk. Interjections seldom attract the attention of modern etymologists. They are supposed to reflect emotions directly. However, when people shriek (“holler”) with pain, they do not say “ouch”, and no one will produce the coquettish oops without learning it like any other word. They share common cause with cuckoo and its ilk in that they are often borrowed. Consult the O.E.D. for the recorded history of bah, haw, humph, pshaw, pish, tush, tut, fie, fe-fi-fo-fum, la, lo, pooh, poof, phew and for inspiration (re)read Kipling’s just so story “How the Camel Got His Hump.” Despite all the difficulties mentioned above, if a word can be traced to onomatopoeia, its origin has been established once for all. 2.2. Many common nouns and quite a few adjectives and verbs go back to proper names. Sometimes their etymology is a matter of stating the type of word formation: cf. mesmerism/mesmerize, Shakespearean, quixotic, Micawberish, and so forth. More complicated is the etymology of milliner, tawdry, and the like, but, once discovered, it becomes final: milliner is from Milan, and tawdry is from (Seyn)t Audrie(s lace). Words like spencer, sandwich, duffel, Stilton, cambric, Yperite, from personal and place-names, are transparent. Yet this etymological path is also rough. Not infrequently we only suspect that a common noun is from a name. Weekley (1921, s.v.) mentioned the possibility of trot ‘old woman, hag’ reflecting the memory of Dame Trot of Salerno, 11th century doctor and witch. ME trat(t)e is “obscurely related” to Gower’s Anglo-Norman trote “of unknown origin.” I offered my own etymology of trot (Liberman 1992: 91–92), for I believe that Weekley was mistaken, but if he was right, scholars attempting to recover this word’s antecedents elsewhere deserve nothing but pity. Sometimes we are fortunate enough to have reliable anecdotes about such words; a classic case is the history of bunkum. Partridge (1949) collected many examples of “names into words”, but his material, like everything else from Partridge’s pen, should be treated with caution. Very curious has been the search for Dr. Condom, Tom Blanket, and especially for Thomas Crapper (Rayburn 1971). 2.3. Some simplified words can be traced to their sources with absolute certainty. Such are, first and foremost, disguised compounds. Cupboard rhymes with Hubbard, but its spelling betrays its origin. Bonfire (from bone fire) is less obvious, daisy (day’s eye) and woman (w¯ıfman) still less so,

114

Anatoly Liberman

wassail (Old Norse ves heill = OE wæs ¯ ha¯l) and barn (bere+ern/ærn) are totally opaque, but their history is recoverable. Some back formations (sculpt from sculptor) and clippings (flu from influenza) pose no problems either. Ancient blending cannot be demonstrated, but there are blends (“portmanteau words”) whose origin has not yet been forgotten (jerrymander, chortle, motel, brunch). Finally, coinages by individuals exist. Sometimes we know their sources but are unable to account for their motivation: the authors are usually not around to ask (cf. theodolite, gas, Lilliputian, spoof, jeep, and so forth; a much more sizeable list than is usually believed). 2.4. Our evaluation of borrowings depends on how far we are ready to pursue their etymology. I will, for the sake of the argument, disregard the words whose status is in doubt: they may be borrowed or native. Quite a few well-known words found their way into Standard English from either northern dialects or Old Danish. Others seem to be of Low German/Dutch provenance, but perhaps their counterparts in English dialects remained unknown to etymologists. There are obscure words that only because of their obscurity were at one time declared Celtic, and words that look like loans from French but are probably English (e.g., flatter). However many such cases may be, incomparably more English words exist that undoubtedly came from French, Latin, Scandinavian, Low German, Dutch, and other languages all over the world. As far as their history in English is concerned, their origin (source) is known, but in their respective languages they may be etymological cruces. Since English etymology as a science must have its geographical and historical limitations, it is more reasonable to expect that someone else should decide why the colibri and the yak are called what they are. Yam is from Portuguese or Spanish, and buffet ‘sideboard’ is from French. Both are “of unknown origin.” May Romance scholars solve their puzzles (if they can).

3.

Some English words of unknown etymology

To find a word’s etymology means to break through the arbitrariness of the sign, to reach a stage at which the sound complex before us stops being absolutely conventional. In the categories listed above (onomatopoeia, words

Origin unknown

115

from names, disguised compounds, clippings, back formations, blends, and proven borrowings), this stage has been reached. The etymology of such words is “known,” but no etymology, except for the clear-cut cases of onomatopoeia, is ever complete. As stated in section 2, milliner is from Milan and breakfast is from break+fast while buffet is a borrowing from French, but we still have to answer the question about the origin of Milan, break, fast, and Fr. buffet. We are most fortunate if we succeed in making one step back. Consider some cases of internal and comparative reconstruction. Green < OE gre¯ne < *gro¯njaz can be shown to be akin to grow and grass, so green emerges as the color of growth. The non-Germanic cognates of comb mean ‘tooth’; the original meaning of the Germanic word was probably also ‘tooth’. The etymology of green and comb is “known,” but why Germanic *gra-/*gro¯- suggested growth and why IE *gombhos meant ‘tooth’ will hardly ever be explained. The same mist envelops the origin of primitive roots and the origin of language. Here is then the obvious concluding statement: words that do not allow modern investigators to “make one step back in history” are those whose etymology is “unknown.” In practical terms, they are words not mentioned in 2.1–2.4, above, and words that, unlike green and comb, do not yield to the procedures of internal and comparative reconstruction. We will now examine them briefly. Age has little to do with etymological obscurity. Impenetrable words can go back to Old, Middle, Early Modern English, and any of the last four centuries. Since the number of Old English words recorded in Old English and still current is relatively small, most obscure words are post-1066; especially many of them surfaced between 1500 and 1800. To give an idea of the relevant material, I will list some of the words cited in my article “The ‘Dregs’ of English Etymology” (Liberman 1992a). Only universally known words will be featured; the numbers in parentheses refer to the centuries in which they were first recorded: ache (OE), awning (17), baffle (16), bald (14), bask (14), berth (18), bicker (13), big (16), bird (13), blizzard (19), bludgeon (18), boast (13), boy (13), brag (13), breeze (OE), bunch (14), ceiling (14), chide (OE), cluster (OE), coarse (14), coax (17), croft (OE), cub (16), cudgel (OE), curse (OE), dagger (14), dandruff (16), dandy (17), dastard (15), dig (13), dude (19), dye (OE), elope (17), ever (OE), fieldfare (12), flash (16), flimsy (17), fun (18), gamble (16), garbage (16), gaunt (16), gingerly (16), girl (13), gravy (14), guilt (OE), guess (17), hassock (OE), heifer (OE), hoax (18), hobo (19),

116

Anatoly Liberman

hoodlum (19), horehound (OE), hurl (13), jazz (20), jeer (16), job (15), jug (16), keep (OE), lad (13), larrikin (19), lass (13), lettuce (13), lollipop (18), lorry (19), marigold (14), moan (13), mongrel (15), motley (14), muffin (18), nincompoop (13), no (OE), nook (13), nod (14), nudge (17), oat (OE), pang (16), peat (14), pedlar (14), peevish (14), penguin (16), pig (OE), pimp (17), plod (16), plot (OE), pony (17), pour (13), pout (14), prance (14), prank (16), prig (16), prim (17), prod (15), prowl (14), puzzle (16), qualm (16), queer (16), quirk (16), rabble (16), ragamuffin (16), rod (12), rogue (16), rowdy (19), rubbish (14), runt (16), scorch (15), scoundrel (16), scramble (16), scruff (18), scull (14), shallow (15), shark (16), shelter (16), shrug (14), shun (OE), slang (18), slant (17), slash (14), slender (15), slum (19), smell (12), sprawl (OE), stooge (20), swift (OE), tarpaulin (17), tarry (14), taut (14), throb (14), tire (OE), toad (OE), toss (16), trash (16), trickle (14), turmoil (17), wench (13), whim (17), wither (14), yeoman (14). These are isolated words. They are known to English speakers but lack established (and most often even putative) cognates in other languages. They are especially difficult to work with, for they exist in an etymological vacuum. Some words about whose origin nothing definite is known may have a cognate or two in Germanic or cognates in and outside the Germanic group. Dictionaries are prone to replace etymologies with strings of cognates, as though the fact that in addition to OE br¯yd we have Old Frisian bre¯d, breid, breyd, Old Saxon bru¯d, Old High German brut, Old Icelandic bru¯Dr, and Gothic bru¯łs, all meaning ‘bride’ and traceable to Common Germanic *bru¯Diz, makes the etymology of bride any clearer. The only use of cognates is that their form and meaning will display enough variety for making the desired step back (consider what is said about green and comb, above). If it is true that day (OE dæg, Gothic dags) is related to non-Germanic words designating heat (summer) and the products of combustion, we can conclude that *dagaz originally meant ‘hot time’. Likewise, if winter is related to wet, it follows with a rather high degree of probability that the people who gave such a name to this season inhabited the lands in which rain rather than frost prevailed in the months they called winter. But if all we learn about tree is that in most Germanic languages, Sanskrit, Greek, Balto-Slavic, and Celtic its cognates also mean ‘(some kind of) tree, wood’, we are not closer to the ulterior etymology than if tree were as isolated as

Origin unknown

117

willow, with cognates only in Old Saxon and (Middle) Low German, or hemlock, with no cognates at all. The reconstructed Germanic and IndoEuropean forms *trewam and *deru-/*doru- cannot be considered a sufficient reward for a long search, because the sign, though traced to a distant epoch, remains arbitrary. It is typical of the state of the art that the O.D.E.E. calls the origin of bride unknown but does not say the same about tree. (I constantly refer to the O.D.E.E., for of all the inadequate modern etymological dictionaries of English it is the best. Skeat’s dictionary was written by a great scholar, but the edition available to the public is a reprint of the 1910 work, which is in many respects outdated.) Producing roots like *trewan and *deru-/*doru- has been the goal of etymologists for nearly two centuries. We find them in the works of Jacob Grimm, Theodor Benfey, August Pott, August Schleicher, Karl Brugmann, and of course in the comparative dictionaries of Indo-European, Romance, and Slavic (curiously enough, a comparative Germanic dictionary has not been compiled). Except for cases of onomatopoeia, the origin of roots is hidden. To put it differently, every etymology sooner or later reaches a dead end. It is instructive to connect winter with wet, wet with water, and water with otter, but we cannot explain why water, a word whose cognates have been recorded from Norway to India, was, thousands of years ago, called *wod-. Did the complex wod suggest wetness to the early Indo-Europeans or the presence of fish, or the joy of quenching thirst, or did they say wodwod, wed-wed, ud-ud in every grade of ablaut as we say glug-glug? Thus wet, otter, and possibly winter have etymologies, while the origin of water is unknown. For a non-onomatopoeic native word to possess a recoverable etymology, it has to be derivative either at the level of word formation (daisy = dæges eage, quixotic = quixote+ic) or at the semantic level (bunkum from Bunkum, green from grow, otter from water). Why is it that some words, like green, are “derived”, while others, like grow, are “prime”? A few typical examples will answer this question. Cases like water will, however, be passed by below. Adz(e), from OE adesa, resembles some names of the hatchet from Hittite to German, but in the form in which this Old English word has come down to us it is isolated. Tools cross state borders easily, and so do their names. Adesa might be a corrupted version of a foreign word or a blend of some kind. Speculations on this score are not fruitless, but an absolutely convincing answer will never be obtained. Key is another tool with a

118

Anatoly Liberman

strange name, and yo-yo came to us from Canada; the toy is familiar, the word is obscure. Shakespeare used the adjective sprag ‘smart, clever’. Two centuries later its synonym sprack was recorded; it is probably a doublet of sprag. Also in the 18th century spry ‘active, brisk’, allegedly connected with the verb spray, emerged. Spruce is a 16th century word; at that time it meant ‘brisk, lively’. Spree surfaced in the 19th century. Evidently, the group spr- suggested (and still suggests?) sprightliness. Language is not only always in a state of flux; it is also continually at play. We will easily grant the status of ludic forms to popsy-wopsy, fuzzy-wuzzy, and miminy-piminy, but it is hard to tell whether sprack, sprag, and spree were coined in jest. Even if they were, the remainder of each word has to be accounted for. Funk (obsolete), punk (chiefly American), and spunk mean ‘touchwood’. One does not have to be a specialist in rhyming slang to suggest that they are connected. If we knew the origin of at least one of them, we would have relatively little trouble with the other two, but we do not. Funk ‘cowering fear’, Funk, last name, and punk in the sense known only too well (originally ‘prostitute’) provide no help. In the 17th century there was a certain Captain Fudge, nicknamed Lying Fudge. Is he the gentleman to whom we owe the verb fudge ‘fake, falsify’? After all, grog, a mixture of rum and water, is named after Admiral Vernon, nicknamed Old Grog; he wore a grogram cloak, and in 1740 he gave the order to use the mixture instead of neat spirit. And there is an anecdote about the origin of the word schooner: “The first schooner ever constructed is said to have been built in Gloucester, Mass., about the year 1713, by a Captain Andrew Robinson and to have received its name from the following trivial circumstance: When the vessel went off the stocks into the water, a bystander cried out, “O, how she scoons [= glides]!” Robinson instantly replied “A scooner let her be”; and, from that time, vessels thus masted and rigged have gone by this name” (Webster: 1864). Skeat (1910, s.v.) remarks: “As a rule derivations which require a story to be told turn out to be false; in the present case, there seems to be no doubt that the story is true.” The letter h crept in because the word was supposed to be of Dutch origin. However, fudge (v.) has probably nothing to do with the lying captain: it looks like many f-words denoting erratic motion and movement back and forth (cf. fiddle, fiddle-faddle, fickle, fitful, etc.). Of course, one cannot be certain, and it is an embarrassment that the inventors of fudge did not write

Origin unknown

119

a story about how they came by the name of their sweetmeat. Unless such a story happens to be unearthed by a patient scholar, the origin of the noun fudge will remain “unknown,” and that of the verb fudge “uncertain” or “doubtful.” Local words have always striven for national recognition, and many of them have made their way into our dictionaries, with or without the abbreviation dial. Those who will look through Wright’s (1898–1905) monumental work and DARE will discover a vocabulary so foreign that it can hardly be equated with the English they speak. Its origin is usually obscure. The Scandinavian element is not too difficult to recognize. The same holds for some obvious borrowings from Celtic. But most native words have no ascertainable Middle English, let alone Old English, sources, and the amount of philological labor expended on them falls short of what is needed for offering even preliminary hypotheses about their origin. The entries on freak, moither, skedaddle, and gumption in the O.D.E.E. are typical. Freak: “prob[ably] of dial[ectal] origin”; moither: “of dial[ectal] and obscure origin”; skedaddle: “first in U.S. mil[itary] sl[ang], unless the earlier Eng[lish] dial[ectal] use is taken into account”; gumption: “orig[inally] Sc[ottish]; ... of unkn[own] origin.” Equally uninformative are the entries dealing with social dialects and modern slang; cf. filch: “orig[inally] thieves’ slang, of unkn[own] origin.” Many plant and animal names pose insurmountable difficulties to language historians. Ivy has been compared to Latin ibex ‘wild goat’ (that is ‘climber’), but, unfortunately, the origin of ibex is obscure (“prob[ably] Alpine word like camox ‘chamois’”; cf. chamois: “prob[ably] ... from a preRom[ance] name current in Alpine areas” – O.D.E.E.), and, according to the golden rule of etymology, one should never explain obscurum per obscurius. When scholars violate this rule, they invariably get wrong results. The unexplained word ibex is unable to shed light on ivy by definition. Not only the distant etymology of ivy is a puzzle; its cognates also look odd. The Old English form is ¯ıfig, seemingly ¯ıf+ig, but the Old High German name of ivy was ebah (eb+ah?), while Dutch has eiloof = ei+loof ‘leaf’. How are ¯ıf-, eb-, and ei- related? Ivy is a poisonous plant, and taboo could be at work in its denomination. If this is so, we will probably never know how the original form (assuming that it was the same for English, German, and Dutch) was changed out of fear and superstition. In other languages, the etymology of the word for ivy poses equally difficult problems.

120

Anatoly Liberman

A similar case is the history of clover, a well-attested West Germanic plant name. I think that the old etymology (from cleave ‘stick fast, adhere’) is correct, but few people believe in it, and clover is now sometimes called a substrate word. Dutch scholars are especially active in promoting the idea of the pre-Germanic substrate. In the history of the Romance languages, the substrate has to be reckoned with (cf. what is said above about ibex and the source of chamois), and no one has done more for excavating the pre-Romance strata of the modern vocabulary than Italian and French philologists. But in Germanic the situation is different, and extreme caution is needed to prevent the substrate from becoming a respectable-looking dump for words of unascertained, and often unascertainable, origin. Apparently we will gain nothing by labeling the animal name seal an Arctic substrate word if we cannot pinpoint the language from which it spread to Germanic. How good is the chance that words “of unknown origin,” from adze to yo-yo will ever shake off the obscurity in which they are now shrouded? At present, only a preliminary and evasive answer can be given to this question. Many cases are undoubtedly hopeless, but, before it becomes clear which words should be given up at least for the time being, it is necessary to have a dictionary of English etymology comparable to those we have for Sanskrit, Ancient Greek, Latin, Gothic, French, Spanish, Russian, Lithuanian, and to a lesser extent for German, Dutch, and Old Icelandic, that is a dictionary that critically evaluates all the existing literature pertaining to the origin and history of English words. My work on such a dictionary (see Liberman 1994) has shown that many words believed to be of unknown origin have in fact been explained very well, but the notes, reviews, articles, and even books offering ingenious conjectures have been missed by our leading etymologists (which, in the absence of a cumulative bibliography, is not surprising). Also, luck plays an outstanding role in etymological work. A chance association may produce the spark needed for guessing the origin of some word. In 1889 Mayhew guessed the connection of Engl. clough ‘ravine’ and German Klinge. Ten years later, someone, who called himself S.J.A.F. (1899), explained oof ‘money’ (slang) as a shortening of Yiddish offtisch ‘on the table’, that is ‘cash on the nail’, said of money laid on the table in gambling. Correct solutions did not stop appearing a hundred years ago. There is no excuse for repeating Skeat’s, Murray’s, and Bradley’s opinions, as though they were biblical truths. None of these excellent

Origin unknown

121

scholars was omniscient or infallible, and, more importantly, since their time thousands of pages have been written on the origin of English words and their cognates. Once we have an up-to-date survey of the literature, the picture will become clearer. “An English Feist” or “an English von Wartburg” will take years to complete, but a bibliography of English etymology featuring about 15,000 titles and approximately 13,000 words will probably be published in 2002 by the University of Minnesota Press (Minneapolis) and Il Calamo (Rome). I hope that its appearance will be a step in the right direction. Since the epigraph to this paper contained one of Mrs. Nickleby’s “sublime inanities,” as a witty critic called them, it will be appropriate to finish our rambles through etymological thickets by quoting another episode from the same novel (Chapter 27). Mrs. Nickleby has been enticed to go to the theater, and two minor villains Puke and Pluck are there to wait on the silly woman. ... and so polite were they, that Mr. Pyke threatened with many oaths to ‘smifligate’ a very old man with a lantern who accidentally stumbled in her way – to the great terror of Mrs. Nickleby, who, conjecturing more from Mr. Pyke’s excitement than any previous acquaintance with the etymology of the word that smifligation and bloodshed must be in the main one and the same thing, was alarmed beyond expression, lest something should occur. Fortunately, however, Mr. Pyke confined himself to mere verbal smifligation, and they reached the box with no more serious interruption by the way, than a desire on the part of the same pugnacious gentleman to ‘smash’ the assistant box-keeper for happening to mistake the number.

Those with a taste for such problems are humbly requested to discover the exact meaning and origin of the verb smifligate.

References Bailey, Nathan 1721 An Universal Etymological English Dictionary. London: Printed for E. Bell. Reprinted as Anglistica and Americana 52. Hildesheim, New York: G. Olms, 1969. Curti, Theodor 1885 Die Entstehung der Sprache durch Nachahmung des Schalles. Stuttgart: E. Schweizerbart (E. Koch).

122

Anatoly Liberman

DARE 1985–

= Frederic G. Cassidy, chief editor, Joan H. Hall, associate editor, Dictionary of American Regional English. Cambridge, Massachusetts and London, England: The Bellknap Press of Harvard University Press, 1985– .

Diensberg, Bernhard 1988 A Phonological and Semantic Description of Animal Sounds as Found in Modern English Dialects. Studia Anglica Posnaniensia 20: 19–32. Garcia de Diego, Vincente [no date] Diccionario de voces naturales. Aguilar. Feist, Sigmund 1939 Vergleichendes Wörterbuch der gotischen Sprache. 3rd ed. Leiden: E.J. Brill. 4th ed. by W.P. Lehmann. Junius, Franciscus 1743 Etymologicum Anglicanum [= Francisci Junii Francisci filii Etymologicum Anglicanum ex Autographo descripsit & accesionibus permultis auctum edidit Edwardus Lye]. Oxonii: E Theatro Sheldoniano. Reprinted Los Angeles: Sherwin & Freutel, 1970. Liberman, Anatoly 1992a Etymological Studies IV. The “Dregs” of English Etymology. General Linguistics 32: 16–35. 1992b. Etymological Studies V. Some English Words Related to Sex and Prostitution. General Linguistics 32: 71–94. 1994 An Analytic Dictionary of English Etymology. Dictionaries 15: 1–29. Mayhew, Anthony L. 1889 The Etymology of ‘clough.’ The Academy 36: 137, 154, 188. O.D.E.E.

= C.T. Onions, editor, with the assistance of G.W.S. Friedrichsen and R.W. Burchfield, The Oxford Dictionary of English Etymology. Oxford: Clarendon Press, 1966.

O.E.D.

= James A.H. Murray et al., eds., The Oxford English Dictionary, 1884–1928. 2nd ed. by J.A. Simpson and E.S.C. Weiner, 1992.

Partridge, Eric 1949 Name into Word. Proper Names That Have Become Common Property. A Discursive Dictionary. London: Secker and Warburg.

Origin unknown

123

Rayburn, Wallace [1971] Flushed with Pride. The Story of Thomas Crapper. Englewood Cliffs, N.J.: Prentice Hall. S.J.A.F. 1899

Ooftisch. Notes and Queries IX/4: 166.

Skeat, Walter W. 1910 An Etymological Dictionary of the English Language. 4th ed. Oxford: Clarendon Press. Webster, Noah 1864 An American Dictionary of the English Language. Springfield, Mass.: G. & C. Merriam Company. Wackernagel, Wilhelm 1867 Voces variæ animalium. Programm für die Rectoratsfeier der Universität. Basel: C. Schultze. Wartburg, Walther von 1934– Französisches etymologisches Wörterbuch: eine Darstellung des galloro[1998] manischen Sprachschatzes. Leipzig, Berlin: B.G. Teubner. Weekley, Ernest 1921 An Etymological Dictionary of Modern English. London: John Murray. Reprinted, New York: Dover Publications, 1967. Winteler, Jost 1892 Naturlaute und Sprache. Ausführungen zu W. Wackernagels Voces variae animalium. Aarau: H.R. Sauerländer. Wright, Joseph 1898– The English Dialect Dictionary. London: H. Frowde. Reprinted, London: 1905 Oxford University Press, 1970.

Issues for a new history of English prosody Thomas Cable

It has been nearly a century since the first volume of George Saintsbury’s History of English Prosody appeared. Any new history of English prosody will have to sort through the voluminous twentieth-century writings on the subject, some leading to significant discoveries that appear to be holding up well, changing our understanding of the older poetry, others going down dead ends. The subject too, of course, has changed since Saintsbury’s time. Any new history should have quite a bit to say about the rise of free verse and its relation to the continuing tradition of writing in form. In a more subtle way the poetry that Saintsbury wrote about has changed too, or our understanding of it has changed, including our understanding of its rhythm. It is probably also true that familiar ways of reading poetry have become less familiar. At the metrical level the simple notion of counting out beats, for example, which had been shared by timers and stressers, despite the differences in theory constructed around those beats, is less of a staple than it once was. The focus of the present essay will be on the iambic pentameter between Chaucer and Donne, but the issues that are raised have implications for most other poetic forms in English, including Old and Middle English alliterative poetry, iambic tetrameter, and dipodic verse. In 1966 Halle and Keyser introduced generative metrics. Although their iambic pentameter template had five S, or Strong, positions, there was no invitation to match these abstractions to a tapping out of five beats, and there was certainly no indication that such a tapping should modify the most “neutral” stress patterns of ordinary phonology. Indeed under their working hypothesis “No violation of the linguistic givens for the sake of meter,” where “linguistic givens” included even phrasal stress, the implications for reading poetry became very peculiar. Without the idea of tapping out a beat, it is impossible to talk about the interaction of the beat and the words that make up the line of poetry. It takes words to set up a linguistic beat, but the beat in turn modulates the words. It is especially strange that this idea should be so foreign to generative metrics, because some of the most significant developments in generative pho-

126

Thomas Cable

nology since 1975 have included descriptions of the interaction between stressed syllables and beats where no syllable occurs. Some of the same linguists write on both subjects. Because of the effect of generative prosody on our perception of the poets’ craft, we shall consider three elements of that theory after we get to Shakespeare. The first occurs in all versions since Halle and Keyser 1966, and the last two occur in most current theories: (1) the diagnostic relevance of the W, or weak, position; (2) the lack of attention to metrical pauses; and (3) what is seen here as the misleading attention given to other kinds of syntactic breaks, usually concerning the “caesura.”

Chaucer and stress shift Reading through English poetry of the fourteenth and fifteenth centuries is an experience, as regards meter, of changing expectations. An Alliterative Revival poem such as Sir Gawain and the Green Knight unfolds rhythmically according to the natural stress patterns of the words as we understand them, so that the abstract pattern of metrical stress is known only after the fact – except, of course, for the short lines of the stanzaic “wheel,” where a basic iambic trimeter tune lets the reader know which way to tilt the stress, if necessary. In Chaucer and Gower this tune of alternating stress carries through eight or ten syllables with great regularity so that expectations of it are soon established. It is the pattern of the great majority of lines, and in the rest, the more ambiguous lines, it is almost unfailingly possible to the ear of a modern reader. The effect is of a clear tune in the mind somewhat like an afterimage on the retina, and it is quite different from the effect of reading Shakespeare, Milton, or Wordsworth, or even Pope’s couplets. Not all modern readers have this experience, to be sure (see Southworth 1954 and Robinson 1971), but it is implied by a tradition of probably the majority of Chaucerians who have written on the subject and of Middle English metrists, going back to Ten Brink 1901, and including Bischoff 1897, Baum 1922, Donaldson 1948, Wimsatt 1970, Mustanoja 1979, and Smithers 1983. The experience is rather different as one moves into the fifteenth century and takes up Lydgate and Hoccleve. Their lines often conform neither to the rhythms of the Alliterative Revival nor to those of Chaucer and Gower,

Issues for a new history of English prosody

127

though they are closer to the latter. Indeed, Lydgate and Hoccleve looked to Chaucer as their master. Their prosody has been severely criticized, and Lydgate’s “broken-backed” lines are notorious. Not all the criticism has been fair because of problems that had already been sorted through for Chaucer but that still vexed Lydgate studies – for example, spurious works and corrupt texts and less willingness among prosodists to smooth rough lines with the sounding of historical -e in contexts where it is generally assumed for Chaucer and Gower. Still, three or four lines on every page of Lydgate cannot be redeemed metrically, even after resort to all the familiar licenses of Chaucerian verse. For example, in iambic pentameter: (1) And of deth stood vnder arreste (Troy-book 3.4180) In addition to the ten percent or so that cannot be made to fit, even by a generous reading, the profusion of lines that depart from an easy alternating rhythm is such that the fabric of the whole is different from that of Chaucer or Gower; many lines must be read twice to get the stresses right; and the iambic pentameter tune is missing. For present purposes Lydgate can throw illumination back on Chaucer as a reminder that not all late medieval verse falls into such regular rhythms. Why is it that Chaucer and Gower seem so regular to the modern reader? One reason is that Weak and Strong positions alternate so regularly in stress that the pattern carries over to syllables where stress is less determinate. The metrical template seems to shape and tilt the ambiguous patterns of stress – those that occur among consecutive monosyllabic words and in the well-known instances of variable stress in words like honour. Furthermore, despite sporadic dissension, it is clear from more than a century of philological investigation that the metrical template affects (at the very least) the occurrence and non-occurrence of historical -e by processes of apocope and various forms of elision. Finally, headless lines – those of nine syllables beginning with a stressed syllable – add one more variable to Chaucer’s meter. With so many options – or uncertainties – in the reading of Chaucer’s line, it is no surprise that modern theories vary widely. The earliest statement of generative metrics was extremely strict in holding to the notion of “No violation of the linguistic givens for the sake of meter,” as those linguistic givens were defined within a specific phonological theory. Partly

128

Thomas Cable

because the linguistic rules were so strict, the metrical patterns described by the Halle-Keyser theory were quite loose. My own understanding of Chaucer’s meter is that it was extremely regular, and the flexibility was more in the phonology, especially in the stress shifts that it allowed. For example, in the 858 lines of the General Prologue to The Canterbury Tales I cannot find a single line that contains a clear, indisputable example of an “inverted first foot,” the kind of variation familiar in Shakespearean lines such as: (2) / x x / x / x /x /x Music to hear, why hear’st thou music sadly? (Sonnet 8) (Note: the markings “/” and “x” throughout this paper correspond to metrical ictus and non-ictus respectively, entities that usually correspond to relative linguistic stress – though not always, as will be discussed below. An “x” at the end of the line is extrametrical. Foot structure, for the moment, will be ignored.) There are numerous lines in the General Prologue that, out of context, a modern reader might scan with this pattern, but most of these yield to the regularly alternating pattern when read in context, and every single one of the rest is, from a twenty-first century point of view, problematic. For example, three lines begin with disyllabic words ending in -y: (3)

x/ x / x / x / x / x Redy to wenden on my pilgrimage (A GP 21) x / x / x / x / x / x Hardy he was and wys to undertake (A GP 405) x / x / x / x / x / Worthy to been stywardes of rente and lond (A GP 579)

We know from the occurrence of these words in rhyme that stress on the second syllable was possible in Chaucer’s English (redy: Tr. 1.988; hardy: B 3106; worthy: RR 2209, 5063, 7238, 7573, PF 635). In addition, hardy and worthy occur within the verse where stress on the second syllable would be expected: hardy: LGW 1528; worthy: Compl. Lady 73, and less

Issues for a new history of English prosody

129

certainly, F 555), though whether stress shifted here to conform to the meter is, of course, part of the question being asked. Clearly, the more common pronunciation was with stress on the first syllable, but a pronunciation with stress on the second syllable was acceptable in the language, and it would not be the kind of violation of normal phonology that it would be for any of these words in Present-day English. 1 Occasionally in Chaucer a verb in the present indicative third person singular occurs at the beginning of a line that seems to require a scansion with an “inverted first foot.” One of these occurs in the General Prologue: (4) “Cometh neer,” quod he, “my masteer and my lord” (A GP 839) Kökeritz 1961: 18, with this line in mind, observes: “Medial -e could be syncopated between two consonants of the same word: cometh > com’th [kumT] 839.” The distribution of the word in Chaucer’s poetry bears out Kökeritz’s observation. Of 65 occurrences, 48 are in contexts where syncopation would be expected for the meter. Twelve are in contexts where the disyllable is needed. And five are at the beginning of the line as in 839 above, where the ambiguity is between a disyllable forming an inverted first foot and syncopation forming a heavy foot. This kind of ambiguity attaches also to lines beginning with preterits and past participles of weak verbs, as in: (5) Lyned with taffata and with sendal (A GP 440) Tukked he was as is a frere aboute (A GP 621) In Modern English these have become monosyllabic, as has loved, the example in Kökeritz 1961: 19: loved > luv’d [luvd]. Kökeritz’s example is a preterit, but the past participle also occurs in contexts in which the meter suggests syncopation: (6)

x / x / x / x /x /x Hadde loved hir best of any creature (F Fran T 939)

Monosyllabic words are also problematic at the beginning of the verse. Youmans 1996: 208 cites as obvious examples of inverted feet:

130

Thomas Cable

(7) Short was his gowne, with sleves longe and wyde (A GP 93) Whit was his berd as is the dayesye (A GP 332) The idea of “schwebende Betonung,” “hovering stress,” “level stress,” or “fluctuating stress,” proposed by Bischoff 1897–98, Ten Brink 1901: §§274–75, and Dobson 1957: 446–49 and recently revived by Redford 2001, should be given consideration for both disyllabic words and monosyllabic phrases. At the very least a cautionary note should be struck about assuming that the Middle English pattern followed the most neutral pattern of Modern English. The regularities in a system of meter can serve a heuristic function, leading to a reexamination of assumptions about specific parts of the language that had seemed unquestionable. For example, Kiparsky 1977: 220–21 found that in Shakespeare the prepositions among, between, before, above, and upon, all usually having the stress pattern x/, occur in contexts in which one version of his theory of meter would have expected /x. Kiparsky suggests that a solution could be an extension of the well-known Rhythm Rule to allow stress on the first syllable of each of these words: (8)

x / x / x / x / x / Henceforth be never numbered among men (MND 3.2.67)

A problem with just the reverse pattern of stress occurs in Chaucer: the prepositions after and under, normally having the stress /x, occur in contexts in which x/ might be expected. These words are favorites among metrists who object to the traditional idea of stress shift in meter, as in the citation by Halle and Keyser 1971 of after in A 136 and under in RR 6451, BD 164, B Mk. 3741, and E Cl. 22. Their criticism was directed toward a traditional scansion by Wimsatt 1970. However, it is not at all clear that Middle English after and under had invariable stress on the first syllable in Middle English poetry. The composite prepositions into, unto, upon, and toward also occur in both /x and x/ metrical contexts, and above, among, before, biforn, and bitwix occur almost exclusively in Chaucer and Gower in /x contexts or at the beginning of a line (where we are suspending judgment). Because the overwhelmingly most frequent patterning for disyllabic prepositions is the sequence of x/ or W S positions, one possibility is that after and under occasionally conformed to normal prepositional stress.

Issues for a new history of English prosody

131

Of the 180 occurrences of after in the Canterbury Tales, 28 or 15.5 percent occur in clear W S contexts, as in (9): (9)

x / x/ x / x / x / x And heeld after the newe world the space (A GP 176)

Another 25 or 13.8 percent occur at the beginning of the line. Taking these together, then, about 30 percent of the occurrences of after in the Canterbury Tales are metrically uncertain. For under, the uncertain percentage is even higher. Of the 78 occurrences in the Canterbury Tales, 31 or 39.7 percent occur in clear W S positions, as in (10): (10) x / x / x / x / x / x The smylere with the knyf under the cloke (A KnT 1999) Another 23 or 29.5 percent occur at the beginning of the line to give a total of 54 or about 69 percent of the occurrences of under as uncertain in metrical stress. As often happens in historical metrics, both the phonology and the meter are uncertain. A degree of circularity, and of going back and forth, is always involved in reasoning through these matters, just as in a child’s learning a language, although it would be a vicious circle simply to assume either the phonological pattern or the metrical pattern and make the other pattern march accordingly. After and under are especially interesting, since no evidence has been adduced for stress shift on these words when used as prepositions in Old English poetry. It is also true that Old English meter cannot be diagnostic for these words, because they are always swept into the“strong dip” of that strong-stress poetry, a matter of key importance in the history of English prosody, to which we shall return below. However, there is other relevant historical evidence. Nearly two centuries before Chaucer, Orm composed his long poem with careful attention to meter and to syllable structure, as reflected in his idiosyncratic spelling. Mossé 1952: 357 says the following about meter and stress shift in that work:

132

Thomas Cable

The Ormulum is written in absolutely regular septenary verse of 15 syllables, with cæsura after the eighth syllable (for typographical convenience this is usually printed in two lines): x x´ | x x´ | x x´ | x x´ || x x´ | x x´ | x x´ | x At the beginning of the second half-line there is sometimes x´ x instead of x´x (goddspelles 14, affterr 2, 15, unnderr 9, Hælennde 3355, drihhtin 3375, summ łing 3363); it is useless in such cases to suppose any displacement of the word stress.

Mossé’s self-contradictory assertion is, of course, an invitation to take a second look, especially since he includes goddspelles and summ łing. In fact, affterr occurs at the beginning of the half-line 94 times out of 241 total occurrences, or 38.8 percent. Twelve of the total number of occurrences, or five percent, are within the line in W S contexts, in both the four-stress and the three-stress half-lines; for example: (11) x / x / x / x / Ne maZZ itt nohht affterr ¼att daZZ (6,486) x / x / x / x / AZZ stanndenn inn affterr hiss mahht (11,182) x / x / x /x Ne don affterr hiss lare (16,587) A tentative conclusion to draw is that these prepositions, which seem so obvious in their stress pattern to the modern reader, might very well have undergone stress shift in poetry of the thirteenth and fourteenth centuries. In trying to determine if the inverted first foot was normal for Chaucer, the most curious fact of all is that he made no use of the category of syllable that would be conclusive in establishing the pattern: the final -e that figures so prominently in his prosody. If the schwa were to occur as the second syllable in the pattern / x x / at the beginning of a line, not even the most regular reader would want to suggest that stress could be shifted to it. Yet the pattern does not occur in the 858 lines of the General Prologue. Given the numerous morphological categories that require or permit final -e, the pattern would be easy to construct. For example, line 440, was read above as a headless line with syncopation:

Issues for a new history of English prosody

133

(12) / x /x/ x / x / Lyned with taffata and with sendal, An adjective with a historical -e would make the pattern:

(13) / x x / x/ x / x / *Riche with taffata and with sendal Or a noun:

(14) / x x /x/ x / x / Lace with taffata and with sendal An adverb with a historical -e would suffice: (15)

/ x x /x / x / x / *Smothe with taffata and with sendal

An infinitive would work also: (16)

/ x x /x/ x / x / *Werke with taffata and with sendal

This one might require a series of verb phrases with the finite auxiliary verb in a previous line, but no complications of syntax would be required to construct a plentitude of missing patterns, including a plural adjective and a noun with stress on the second syllable: (17) / x x/ x / x / x / *Derke desires now so streyneth me (Cf. Tr 3.1482) “Derke desires” shows up in Chaucer’s prose but not in his poetry. Although the inverted first foot with final -e does not occur in the General Prologue, I am aware of three occurrences elsewhere in Chaucer’s poetry, once with blesse in the imperative mood, where an -e would be expected:

134 (18)

Thomas Cable

/x x / x / x / x / Blesse this hous from every wikked wight (A MilT 3484)

and once with herkne (B 1 ML Pro 113) and once with herke (B 1 MLT 425). There are doubtless other exceptions. However, the extreme paucity of these patterns requires an explanation, since final -e is such a common feature of Chaucer’s prosody. One story that makes sense is that traditional notions of stress promotion and demotion, syncopation, and headless lines apply to lines that might appear to the modern reader to have an inverted first foot, and the rest, those patterns with an -e in second position followed by a metrically unstressed syllable, were avoided because they required the proscribed pattern. A more subtle argument for very regular alternating stress in Chaucer’s poetry is the rarity of foot structures that would allow four rising levels of stress, such as occur everywhere in Shakespeare and Shelley (see Wimsatt 1970: 775). In all possible occurrences of this pattern in Chaucer, a reading with alternating stress is plausible, and – given the tune that is established in Chaucer’s poetry – many readers would say more plausible than not. A description of this pattern, however, requires a sketch of certain elements of Shakespeare’s meter.

Shakespeare, the foot, and the weak position Cable 1991 argued that Chaucer’s regularly alternating meter does not imply foot structure, although English poetry beginning tentatively with Sidney and certainly with Spenser does. If the basic principle of foot meter stated there is true to the facts, then it is in clear conflict with all theories of generative prosody. By that view and the view of the present essay, the key to foot meter in English is relative ictus within the foot, the ictus rising between W and S. Neither W nor S is more important than the other, because the conditions that give W less ictus than S are the same conditions that give S more ictus than W. All theories of generative prosody, including Halle and Keyser’s Stress Maximum Principle, Kiparsky’s Monosyllabic Word Constraint, and Bruce Hayes’s prosodic hierarchy, are asymmetrical in valuing, at the linguistic level, a stressed syllable over an unstressed syllable and, at the

Issues for a new history of English prosody

135

metrical level, W over S. Thus, only a stressed syllable in a W position can serve as the diagnostic for metricality. It is interesting that no poet or metrist before Halle and Keyser ever suggested that for English iambic meter stressed syllables were regulated in determining metricality but unstressed syllables were not. (In Halle and Keyser’s system the only way unstressed syllables can render a line unmetrical, as opposed to merely more complex, is for there to be too many of them to map onto ten positions.) Kiparsky’s theory is an improvement over Halle and Keyser’s in bringing the focus to bear on stress within words, where the pattern of relative stress is more fixed, but it too finds the relevant diagnostic in W positions through the constraint that “weak positions must not contain syllables that are strong within the prosodic word” (Hanson and Kiparsky 29). In Halle and Keyser 1991, which presented a revised theory of generative metrics, a stress maximum was defined as “a stressed syllable occurring between two unstressed syllables in the same syntactic constituent within a line of verse.” Thus, the stress maximum on the first syllable of Percy was the purported location of unmetricality in the unmetrical construct (19): (19) / x x / / x / x / / x Ode to the West Wind by Percy Bysshe Shelley /

However, one could as easily flip the whole theory and propose a stress minimum: “an unstressed syllable occurring between two stressed syllables in the same syntactic constituent within a line of verse.” Then the location of unmetricality would be the S position, there being, in fact, two such locations in this construct: (20) / x x / / x / x / / x Ode to the West Wind by Percy Bysshe Shelley /

/

Our two competing principles will usually identify the same lines as metrical or unmetrical, as here. They do not select exactly the same sets, but any unmetrical pattern that slips past the Stress Minimum Principle will have a corresponding unmetrical pattern that slips past the Stress Maximum Principle. (Cf. “Ode to the West Wind by Percy and Mary,” disallowed by the Stress Maximum but not by the Stress Minimum, and “Ode to the West

136

Thomas Cable

Wind sang Percy Bysshe Shelley,” disallowed by the Stress Minimum but not by the Stress Maximum.) Kiparsky’s Monosyllabic Word Constraint has largely superseded the Stress Maximum Principle, but the same point applies: there is a corollary to his principle of locating unmetricality in the W position. Simply flip the terms and locate unmetricality in the S position: “Syllables of any degree of stress can occur anywhere in the line except that a syllable that is less heavily stressed than an adjacent syllable within the word cannot occur in S.” More will be said about the Monosyllabic Word Constraint in the next section. By reversing the terms of the criteria the familiar generalizations of all generative metrics studies will also be reversed. Stressed syllables are unregulated in the iambic pentameter poetry and can occur freely in S and W positions. The only feature for determining metricality (as opposed to tension) is the placing of the unstressed syllables. These can occur freely in W positions, but in S positions they may not occur adjacent to a more heavily stressed syllable within a word. If they do, the sequence is unmetrical. This is a parody, of course. The point is that stressed and unstressed syllables must be considered together. It is misleading to say that either the stressed syllables or the unstressed syllables are unregulated with respect to W or S. They form a relationship of rising ictus, and it is exactly this relationship that the concept of the iambic foot captures. The general approach of the present essay, while it may seem eccentric, is quite traditional – or was traditional until about the middle of the twentieth century. It is much in the spirit of metrical writings by Ivor Winters and W.K. Wimsatt. It is also in the spirit of Jespersen 1913, an essay to which generative metrists often trace their heritage. For example, Jespersen analyzes the first four syllables of the line in (24) from The Merchant of Venice: (21)As his wise mother wrought in his behalf (1.3.74) Using a system for notating stress in which 4 is the strongest and 1 the weakest, he writes: “This figure is frequent in English verse, but not in other languages. I incline to read it with 1234 and thus to say that the ascent is normal between the first and second as well as between the third and the fourth syllable ...”

Issues for a new history of English prosody

137

This syntactic pattern has long been recognized as producing a metrically ambiguous structure. Stewart 1935 chose a similar line from Alexander Pope, containing a function word, a possessive pronoun, a monosyllabic adjective, and a noun to illustrate “hovering accent”: (22)

On her white breast a sparkling Cross she wore (Rape of the Lock2.7)

Stewart writes that “if we consider white a mere conventional epithet, our reading might be”: (23)

/ x \ | ^ On her white breast a sparkling Cross she wore

“But considering it really descriptive,” he continues, “we should read”: (24)

x x / | / On her white breast a sparkling Cross she wore

The monosyllabic word constraint Kiparsky 1977: 209 analyzes lines syntactically similar to these, including Venus and Adonis 982, the key word being the possessive pronoun in second position: (25)

Which her cheek melts, as scorning it should pass

Kiparsky invokes the syntactic hierarchical constituency of the whole clause, which, by the present view, is irrelevant for metrical purposes and, in its complications, misleading. However, in his syntactic analysis and discoveries, Kiparsky provides the key for disambiguating the “hovering accent” and for choosing between (26a) and (26b) (ignoring for the moment foot division): (26)

x x / / a. Which her cheek melts, as scorning it should pass

138

Thomas Cable

x \ ^ / b. Which her cheek melts, as scorning it should pass Following up on Magnuson and Ryder 1970, Kiparsky shows that Shakespeare avoided the syntactic, word division, and stress patterns of the first four syllables in Donne’s line: (27) x x / / Shall behold God, and never tast deaths woe (Holy Sonnet VII) The same rule that accounts for the avoidance of this pattern by Shakespeare, the Monosyllabic Word Constraint, also accounts for Shakespeare’s avoidance of patterns like: (28)

/ / x x *And peace has green olives of endless age

These observations are acute, original, and extremely valuable for understanding Shakespeare’s meter. However, there is more than one way of interpreting them. It will be argued that Kiparsky has folded into one constraint two patterns that have quite different metrical dynamics, as can be seen in Jespersen 1913. By an alternative analysis, the pattern in (27) was avoided because there is no possible rise in metrical stress between Shal and be-; i.e., the line cannot begin x/. That leaves two possibilities for metrical stress: x x and /x, the latter of which is frequent in Shakespeare in the “inverted first foot,” as for example in (1). However, the stress on behold requires rising ictus, and thus the inverted foot pattern with behold in the middle two positions of /x x/ is not a possibility: (29)

/ x / / *Shall behold God ...

The assumption here is that foot divisions are never made between positions of non-ictus. One way of putting it is that there is no possibility of a trochee and an iamb:

Issues for a new history of English prosody

139

(30) * / x | \ / Shall behold God Another way to say it is that there is no “reversed foot.” By this view, English as a stress-timed language will always have the grouping of non-ictus syllables as in (31): (31) / x

x /

and not as in (32), regardless of the syntactic structure or the occurrence of clitics: (32) / x

x /

This principle is essential for understanding the patterning of non-ictus syllables in Old, Middle, and Modern English, in both the strong-stress and the syllable-stress meters. To return to the line from Donne, we are left then with x x / /. Since Shakespeare avoids the clear syntactic patterns that would produce this metrical effect, a reasonable conclusion to draw is that the metrical pattern itself was proscribed in Shakespeare’s meter (though not in Donne’s). There are other syntactic patterns in Shakespeare that the modern reader might be inclined to scan that way, but to do so would be to impose on ambiguous verses a metrical pattern congenial to modern ears that has no realization in syntactic patterns that are unambiguous in their stress. The implication, then, is that Shakespeare’s avoidance of unambiguous x x / / tells us that he would opt, where ambiguity presents itself, for x \ ^ /; thus (26b). By contrast, the pattern in (28) was avoided, we may infer, for two reasons that work together: (1) the phrase green olives does not permit a shift of stress within the word olives, and (2) the phrase, a combination of adjective and noun with no syntactic break does not permit – in Shakespeare’s poetry – the regularizing effect of a metrical pause, to which we shall turn shortly.

140

Thomas Cable

Kiparsky’s Monosyllabic Word Constraint, which has become the most influential principle of generative metrics, focuses on not allowing the stressed syllable of disyllabic and polysyllabic words to occur in W positions, or on allowing them to occur only under certain conditions. However, in fitting these words into an iambic meter that goes x /, it matters immensely that olives is stressed / x, and behold x /. Thus, with monosyllabic words there can be a climbing of stress through four levels in strict accord with the iambic pentameter template, as in 26b, two feet with rising stress, the W position of the second foot being stronger than the S that precedes. The word olives does not fit into that pattern because its stress does not rise; behold does not fit, we can infer, because the unstressed be- cannot be given more stress than the preceding monosyllabic shal. The weak stress on be- is locked into its relationship to -hold. The obvious fact is that monosyllables are more tiltable in their stress than disyllables and polysyllables – hence the Monosyllabic Word Constraint, topsy-turvily named as it is. Of course, let it be repeated, the final unstressed syllable of a polysyllable can serve as the bottom rung for the climbing monosyllables that follow, famously in: (33)

/ x \ | ^ I summon up remembrance of things past

Because Kiparsky’s discoveries would surely have greater currency if separated from generative metrics’ puzzling thirty-five year fixation on W positions, I would restate the constraints as follows. (1) Shakespeare does not use the pattern x x / / of (26a). Lines that appear to have that pattern actually have the pattern x \ ^ / of (26b). (2) Shakespeare does not use the pattern / / x x unless the first two stresses are separated by a syntactic break, which is filled by a metrical pause.

Metrical pause The usual response from linguistic metrists and many traditional metrists is that the choice between (26a) and (26b) and certainly anything involving a pause are matters of the performance of the line and not the structure. However, such a response confuses the typographical representation of the line and the line itself. Two different lines are represented in (26), even though

Issues for a new history of English prosody

141

they have the same sequence of written letters and punctuation. This difference can be made clearer by indicating the spacing pause of (26a): (34)

x x / | (x) / Which her cheek melts, as scorning it should pass

This kind of pause – a metrical pause as opposed to a syntactic or rhetorical caesura – has been a part of the temporal tradition of English metrics since the nineteenth century. 3 An equivalent pause in ordinary discourse has been a part of a raft of writings in metrical phonology since Liberman 1975. These include writings by phonologists who discuss English verse, though the spacing pause has never been incorporated into their descriptions of English poetic meter (Jakobson 1963: 90, Halle and Keyser 1966: 203, Kiparsky 1972: 175, Hayes 1989: 222). At the interface of syntax and ordinary phonology, this metrical pause is sensitive to the degree of syntactic break as determined by constituent structure, though it is not completely determined by such breaks. At the interface of ordinary phonology and poetic meter, the syntactic break is even less determinative because of the overriding strength of the metrical template. In adjective phrases, which have little internal syntactic structure, stress shift in certain kinds of words is familiar: (35)

/ / \ / thirteen men → thirteen men

However, often stress shift is not possible: (36)

/ / / (x) / content cows → content cows 4

This is the structure in the line from Pope (by one interpretation of what constitutes the line), where a pause occurs between the adjective and noun, white and breast, with no syntactic motivation: (37) x x / | (x) / On her white breast a sparkling Cross she wore

142

Thomas Cable

It is also the structure in the line from Donne between verb and object, behold and God: (38) x x / | (x)/ Shall behold God, and never tast deaths woe Finally, it is the structure in (34) above, where a metrical pause occurs between cheek and melts. This line would be plausible in Shakespeare, though the evidence provided by Kiparsky indirectly supports the other possible line: (39)

/ x \ | ^ Which her cheek melts, as scorning it should pass

It appears that Shakespeare confined the metrical pause to major syntactic breaks, in contrast with later poets who inserted metrical pauses between an adjective and a noun. The use by Shakespeare of a metrical pause coinciding with a dramatic pause has long been recognized in lines that have fewer than ten syllables and major syntactic breaks, sometimes punctuated by an exclamation point, as in the first two of the three exclamation points in (39) (the third being followed by at unstressed syllable): (40) / (x) / (x) / x / x / Stay! speak, speak! I charge thee, speak! (Hamlet 1.1.51) It is worth the somewhat artificial effort to refer to (34) and (39) as different lines rather than different performances of the same line. This distinction makes the point that the metrical pause in (34) replaces a syllable in a position of nonictus, and if that pause were omitted, the line would be as incomplete as if a syllable had been dropped. The line in (34), incidentally, has a trisyllabic foot at the beginning, a substitutable unit that fits in the present theory as well as in Kiparsky’s. For discussions of unelidable trisyllabic feet in Shakespeare, involving words such as fortification, see Kiparsky 1977: 197 and Hanson and Kiparsky 1996: 295–300.

Issues for a new history of English prosody

143

Ictus Something should be said about metrical ictus as the tapping of five fingers. For the present author, the beats proceed from the small finger to the thumb. Usually, of course, the tap of a finger coincides with a stressed syllable, or with a syllable that could be promoted to stress without doing violence to the language. Occasionally but infrequently the tapped beat occurs on a syllable which, if given greater stress than those on either side, would seem, in fact, to wrench the language, especially if notated by the usual graphic devices, as with in in (41): (41) x / x / x / x / x / Resembling strong youth in his middle age Or with by in (42): (42) x / x / x / x / x / To eat the world’s due by the grave and thee (Sonnet 1) It would be a grossly mechanical reading to pronounce stress as marked. The problem is not in locating beats by the tapping of fingers but in a conception of metrical ictus as determined solely by linguistic stress. A normal way of thinking about stress and timing in phrases is to say that stress is assigned to lexical words and then function words march accordingly. Selkirk 1984 has rules for destressing function words and for cliticization. In the iambic pentameter there is an external frame of beats (the tapping of fingers) to which the words of the poetic line must be matched. One way of thinking about it is that the iambic pentameter frame occasionally causes de-cliticization. Instead of by-the-grave, we have by the-grave, a slower timing, allowing the middle finger to tap by without a tricky syncopation. The objection that one can expect from all this, and a reasonable objection, is that the middle finger doesn’t tap on by in any event. The taps, it might be objected, are as indicated by the stress marks: (43) x / x / / x x / x / To eat the world’s due by the grave and thee

144

Thomas Cable

However, it is exactly this reading that Kiparsky’s Monosyllabic Word Constraint shows to be implausible. Except at strong phrase boundaries /x x does not occur in unambiguous patterns – patterns with a disyllabic word that locks in the stress: olives of. A theory of suprasegmentals that adjusts timing and stress as sketched here, not to mention pitch, is obviously beyond the scope of the present paper, but no parts of it are incompatible with an approach such as Selkirk 1984. This kind of adjustment has been registered impressionistically by metrists for more than a century, with both more and less precision than finger tapping. Kiparsky 1975 makes a good point about the different styles of performance over the centuries – the eighteenth-century way of allowing a pause at the end of a line, and so on. However, the restricted and precisely defined use of silent beats in the present description is a different matter. It is as integral a part of the linguistic and metrical description as the syllable. It stands in place of a syllable. It is the kind of entity described in much phonology since 1975 – for example, the rules for Silent Demi-Beat Addition, and so, in Selkirk 1984. Following Jakobson 1960: 366, I have tended to leave lines like those in (34) and (39) ambiguous, as “hovering stress,” or chose one or the other, depending on the feel and the context. However, as just explained, one understanding of the implications of studies by Kiparsky leads to the conclusion that the pattern in (34) was not used by Shakespeare, although it was used by English poets before and after. Furthermore, the uncertainty between the two possible lines in (44) can be resolved in favor of (44b): (44)

x x / | (x)/ a. In the deep bosom of the ocean buried x \ | ^ / b. In the deep bosom of the ocean buried

Following Jespersen’s reading that adds enough stress to the to count for ictus, we can justify the pattern on the basis that Shakespeare clearly treated proclitics and initial unstressed syllables of lexical words differently (the Monosyllabic Word Constraint again). As with (41) and (42), adjustments of timing may figure in the individual performance more than

Issues for a new history of English prosody

145

stress. The crucial matter is where the beat falls, however manifested phonetically. Consider again: (45)

Which her cheek melts, as scorning it should pass

By most readings in generative metrics, there is a lack of stress on both which and her, and this lack of stress is irrelevant to determining the metricality of the line, although the occurrence of unstressed her in an S position does contribute, by whatever mechanism of measurement in the particular theory, to complexity or tension. By many traditional readings – of both timers and stressers – her has metrically significantly more ictus than which. By a generative metrics reading that superimposes syntactic structure on the whole line, the line conforms to the metrical template by having a syntactic break after the fourth syllable. By most traditional readings – of both timers and stressers – the caesura is widely recognized as occurring most often after the fourth syllable in a line, as in (45), then after the sixth, and so on, depending on the specific poet’s style. However, as C.S. Lewis put it, the caesura “can occur anywhere the poet chooses and need not occur at all, and is therefore no part of the pattern, though it may be a very important part of the poet’s handling of the pattern so as to move passion or delight. It is a rhetorical and syntactical fact, not a metrical fact” (Lewis 1938: 29). Note that the analysis in (46) inserts a metrical pause (not a caesura) where the syntactic break is only between an adjective and a noun: (46) x x / | (x) / On her white breast a sparkling Cross she wore This metrical pause corresponds to a rest in music, which must be counted in the measure. By contrast, the pause that coincides with a syntactic break but no missing syllable and is often called a “caesura” – as marked by the comma in, “Which her cheek melts, as scorning it should pass” – corresponds to a hold, or fermata. It is open to interpretation, longer or shorter, depending on the performer. In standard orthography, commas ambiguously indicate both caesuras and metrical pauses, and are often absent from

146

Thomas Cable

both. On this matter, a traditional “stresser” such as Wimsatt and all the generative metrists would define the metrical pause as not being a metrical structure. The “timers” in note 1 have a different view.

Summary Reference to the work of metrists such as the ones just mentioned can provide benchmarks for indicating the discoveries and consolidations in English metrical study since Saintsbury. The classic essay by Wimsatt and Beardsley 1959 is the best starting point for an overall sketch of English prosody, and a map to a new history. To that account could be added one idea that Wimsatt and Beardsley never accepted and three ideas that have developed since the publication of their essay. The idea to which they remained recalcitrant was the spacing pause within stress clash, which had been a part of Timers’ theory for decades and which has been incorporated into phonological studies since Liberman 1975. The three main developments in English metrics since their essay have been the discovery or fuller understanding of the following elements: (1) syllable quantity in Old English verse (Fulk 1992, Suzuki 1996); (2) syllable counting in fourteenthcentury alliterative verse (Duggan 1986, Cable 1988); and (3) the Monosyllabic Word Constraint (Kiparsky 1975, 1977). The last of these assumes an interpretation such as the one proposed in the present essay. This interpretation, in effect, jettisons most of the syntactic structure, retaining only the terminal syntactic break at clause ends, and it translates the findings into superficial phonological and metrical elements without hierarchical structure. The bias behind this point of view goes back to one of the main points of the Wimsatt-Beardsley essay and to other writings by Wimsatt, with which the present author is in full accord (save the matter of the metrical pause). There are two main metrical traditions in English poetry: strong-stress meter (Old and Middle English alliterative verse) and syllable-stress foot meter (especially the iambic pentameter from Chaucer to Frost). Essential to the first tradition is the fact that the pattern of metrical ictus for a given line is known only in retrospect, after the line has unfolded according to certain grammatical principles of stress, with very little distortion of those patterns by the metrical template. Essential to the second tradition is that the metrical template can “tilt” the

Issues for a new history of English prosody

147

patterns of stress so as to arrive in the pentameter at five feet, each with rising metrical ictus. (Note that ictus is largely determined by stress in English iambic verse but is not isomorphic with it.) The idea of five rising feet was completely lost in the original version of generative metrics. It is perhaps implicit in Kiparsky 1977 and later, where five feet of a W S configuration occur at the abstract level, but if so, its mapping onto an actual line of verse is not – to reveal the limitations of the present metrist – perspicuous. Finally, to the three developments named in the preceding paragraph, there is the idea that Chaucer wrote in an alternating meter and not in iambic pentameter, which began to develop only with Sidney and Spenser and the poets who followed. However, because the present metrist is the only person who holds this idea, it is best mentioned as an eccentric possibility.

Notes 1.

2.

3.

4.

On variable stress in Chaucer, see Nakao 1977 and Minkova 1997. Though neither deals specifically with the lines considered here, Minkova’s assumptions about the effect of meter on phonological rules is compatible with the scansions given here, at least for syllables in rhyme. See also Minkova 1997: 162: “While the two systems [verse meter and language prosody] are by definition homologous, complete isomorphy on every level need not be assumed.” For analyses that are similar to that in (34) but that are ignored by generative metrists, see, for example, Alden 1903: 19–21 on “pauses filling the time of syllables”; Baum 1922: 62–65 on “metrical pause”; Smith 1923: 42–54 on “compensatory pause”; Schwartz 1962 on Wimsatt and Beardsley’s main failing in not recognizing such pauses; and Attridge 1982: 172–75 on “implied offbeats,” though without the structure of the foot. These are but a sample of a line of prosodists influenced by temporal considerations, stretching back to Steele 1775. The mechanisms vary from one theory to the next. What is described here is essentially Beat Addition in Selkirk 1984: 185–86.

References Alden, Raymond M. 1903 English verse. New York: Holt.

148

Thomas Cable

Attridge, Derek 1982 The Rhythms of English poetry. London: Longman. Baum, Paull F. 1923 The Principles of English versification. Cambridge, MA: Harvard University Press. 1961 Chaucer’s verse. Durham, NC: Duke University Press. Bischoff, Otto 1897 Über zweisilbige Senkung und epische Caesur bei Chaucer. Englische Studien 24: 353–392 and 25: 339–398. Brink, Bernhard ten 1901 The Language and metre of Chaucer. 2nd ed. rev. Friedrich Kluge. Trans. M. Bentinck Smith. New York: Macmillan. Brogan, T.V.F. 1981 English versification, 1570–1980. Baltimore: Johns Hopkins University Press. Cable, Thomas 1991 The English alliterative tradition. Philadelphia: University of Pennsylvania Press. 1988 Middle English meter and its theoretical implications. Yearbook of Langland studies 2: 47–69. Dobson, E.J. 1957 English pronunciation, 1500–1700. 2 vols. Oxford: Clarendon Press. Donaldson, E. Talbot 1948 Chaucer’s final -e. PMLA 63: 1101–1124. Duggan, Hoyt N. 1986 The Shape of the B-Verse in Middle English alliterative poetry. Speculum 61: 564–592. Halle, Morris, and Samuel Jay Keyser 1966 Chaucer and the study of prosody. College English 28: 187–219. 1971 Illustration and defense of a theory of the iambic pentameter. College English. 33: 154–176. Hanson, Kristin, and Paul Kiparsky 1996 A Parametric theory of poetic meter. Language 72: 287–335.

Issues for a new history of English prosody 1997

149

The Nature of verse and its consequences for the mixed form. In: Prosimetrum: Cross-cultural perspectives on narrative in prose and verse, ed. Joseph Harris and Karl Reichi. Cambridge: Brewer, 17–44.

Hayes, Bruce 1989 The Prosodic hierarchy in meter. In: Phonetics and phonology, vol. 1: Rhythm and meter, ed. Paul Kiparsky and Gilbert Youmans. San Diego: Academic Press, 201–60. Jakobson, Roman 1960 Concluding statement: linguistics and poetics. In: Style in language, ed. Thomas A. Sebeok. Cambridge, MA: MIT Press, 350–77. 1963 On the so-called vowel alliteration in Germanic verse. Zeitschrift für Phonetik, Sprachwissenschaft und Komunikationsforschung 16: 85–92. Jespersen, Otto 1933 Notes on metre (1913). In: Linguistica: Selected papers in English, French and German. Copenhagen: Levin & Munksgaard, 249–74. Kiparsky, Paul 1972 Metrics and morphophonemics in the Rigveda. In: Contributions to generative phonology, ed. Michael K. Brame. Austin: U of Texas Press, 171–200. 1975 Stress, syntax, and meter. Language 51: 576–616. 1977 The rhythmic structure of English verse. Linguistic Inquiry 8: 189–247. Kökeritz, Helge 1961 A guide to Chaucer’s pronunciation. New York: Holt, Rinehart and Winston. Lewis, C.S. 1938 The Fifteenth-century heroic line. Essays and Studies 24: 28–41. Liberman, Mark 1975 The Intonational system of English. Diss., MIT. Magnuson, Karl, and Frank G. Ryder 1970 The Study of English prosody: an alternative proposal. College English 31: 789–820. Minkova, Donka 1997 Constraint ranking in Middle English stress-shifting. English Language and Linguistics 1: 135–175.

150

Thomas Cable

Mossé, Fernand 1952 A Handbook of Middle English, trans. James A. Walker. Baltimore, MD: Johns Hopkins Press. Mustanoja, Tauno F. 1979 Chaucer’s prosody. In: Companion to Chaucer studies, ed. Beryl Rowland. 2nd ed. New York: Oxford University Press. Nakao, Toshio 1977 The Prosodic phonology of late Middle English. Tokyo: Shinozaki Shorin. Redford, Michael 2001 Chaucer and Middle English stress. In: Development in prosodic systems, ed. Paula Fikkert and Haike Jacobs. Berlin/New York: Mouton de Gruyter, forthcoming. Robinson, Ian 1971 Chaucer’s prosody. Cambridge: Cambridge University Press. Saintsbury, George 1906–10 A History of English prosody. 3 vols. London: Macmillan. 1910 Historical manual of English prosody. London: Macmillan. Schwartz, Elias K 1962 Rhythm and “exercises in abstraction.” PMLA 77: 668–669, 671–674. Selkirk, Elisabeth O. 1984 Phonology and syntax: the relation between sound and structure. Cambridge, MA: MIT Press. Smith, Egerton 1923 The Principles of English metre. London: Oxford University Press. Smithers, G.V. 1983 The Scansion of Havelok and the use of ME -en and -e in Havelok and by Chaucer. In: Middle English Studies presented to Norman Davis in honour of his seventieth birthday, ed. Douglas Gray and E.G. Stanley. Oxford: Clarendon Press, 195–234. Southworth, James G. 1954 Verses of cadence: an introduction to the prosody of Chaucer and his followers. Oxford: Blackwell.

Issues for a new history of English prosody

151

Steele, Joshua 1775 An Essay towards establishing the melody and measure of speech. London: J. Almon. Stewart, George R., Jr. 1930 The Technique of English verse. New York: Holt. Wimsatt, W.K. 1970 The rule and the norm: Halle and Keyser on Chaucer’s meter. College English 31: 774–788. Wimsatt, W. K., Jr., and Monroe C. Beardsley 1959 The Concept of meter: an exercise in abstraction. PMLA 74: 585–598. Youmans, Gilbert 1996 Reconsidering Chaucer’s prosody. In: English historical metrics, ed. C.B. McCully and J.J. Anderson. Cambridge: Cambridge University Press, 185–209.

Chaucer: Folk poet or littérateur? Gilbert Youmans and Xingzhong Li

It was the schooner Hesperus, That sailed the wintry sea; And the skipper had taken his little daughtèr, To bear him company. Henry Wadsworth Longfellow “The Wreck of the Hesperus”

1.

Verse prototypes

Adopting the notation of Liberman and Prince (1977), Youmans (1989: 346–347) proposes (1) as the prototype for Milton’s iambic pentameter verse line, with binary branching nodes labeled S for more prominent verse constituents and W for less prominent ones:

(1)

(a) Prototypical iambic pentameter line

S W W W

S S

S

–s +s 1 2

W

W

W

S

S

W

S

W

S

W

S

–s +s 3 4

–s 5

+s 6

–s +s 7 8

–s 9

+s 10

(b) Condensed version [W 2S] [W 3S] // [W 2S] / [W 2S] [W 5S]

154

Gilbert Youmans and Xingzhong Li

Relative phonological prominence in English typically corresponds with relative degrees of stress; hence, all odd-numbered positions in (1) are labeled [–stress] and all even-numbered positions are labeled [+stress]. Alternatively, these positions could be labeled [±long], or [±heavy], or [±strong], depending upon the category that defines phonological prominence for a particular poet composing verse in a particular language (Hanson and Kiparsky 1996). The hierarchy of relative prominences implied in (1) can be represented as a metrical grid (2), with the height of each column representing relative degrees of phonological prominence (Liberman and Prince 1977): (2)

x x x x x x x x x x x x x x x x x x

Hence, (1) embodies a claim that, in the prototypical iambic pentameter line, the tenth syllable is the most prominent one, all the more so in rhymed verse; the fourth syllable is the next most prominent one; and the odd-numbered syllables are least prominent. In addition to this pattern of relative prominences, (1) also embodies a hierarchy of metrical boundaries, prototypically: (a) a syllable boundary between metrical positions, (b) a word boundary between metrical feet, (c) a subsidiary phrase boundary after the third foot, (d) a major phrase boundary after the second foot, and (e) a clause or sentence boundary at the end of the line. As a corollary, deviations from the prototypical pattern in (1) are most disruptive at the end of the line, next most disruptive at the end of the first hemistich, least disruptive at the beginning of the line, and next least disruptive at the beginning of the second hemistich. Youmans (1996) and Li (1995) conclude that (1) is the prototype for the iambic pentameter verse of Chaucer as well as Milton and Shakespeare. In addition, Piera (1980) proposes a similar prototype for Spanish eleven-syllable verse, as does Biggs (1999) for French ten-syllable verse. Hence, a prototype similar to (1) seems to be implicit in the verse of a wide range of poets writing in different languages and in different centuries. The iambic pentameter prototype (1) parallels the typical Subject//Verb/ Complement pattern of English clauses and sentences, as in (3):

Chaucer: Folk poet or littérateur?

155

(3) Contrite designs // invite / divine delight. (Constructed line) Example (3) is nearly an ideal realization of the prototype in (1). Nevertheless, its iambic feet, though very similar, are not identical. Conversely, all sequences of identical iambic feet, as in (4), lack the higher-level hierarchical structure of sentences: (4) Repent! Repent! Repent! Repent! Repent! (Constructed line) Hence, the prototype in (1) is a Platonic abstraction rather than a statistical norm or an aesthetic ideal. All actual lines of verse (and prose) deviate from this prototype to a lesser or greater degree. Metrical theorists who adopt a Platonic approach such as this have two distinct tasks: (a) to define central prototypes such as (1) for iambic pentameter, trochaic tetrameter, ballad meter, and so on, and (b) to formulate statistically normative “tension” rules for measuring degrees of deviation from these prototypes. By contrast, Aristotelian metrists are more concerned with defining an exact boundary between metrical and unmetrical lines, through supposedly obligatory rules such as the Stress Maximum Principle (Halle and Keyser 1966 and 1971) and the Monosyllabic Word Constraint (Kiparsky 1975 and 1977). Recent work in this tradition, such as Hanson and Kiparsky (1996 and 1997), is innovative and suggestive, although Hanson and Kiparsky continue to treat rules such as the MWC as categorical constraints rather than as strongly normative rules – despite counterexamples to the MWC cited in Youmans (1983), Li (1995), and elsewhere. Recent linguistic approaches to metrical theory within the framework of Optimality Theory (Prince and Smolensky 1993) tend to place more emphasis upon gradient metrical rules than upon categorical ones; that is, upon ranked, violable constraints. (For example, see Hayes and MacEachern 1998.) By such accounts, the set of permissible lines of verse is inherently “fuzzy” rather than well-defined. Of course, some principles of versification such as rhyme schemes do include a categorical component. But even in this case it turns out that an adequate definition of rhyme requires gradient rules rather than purely categorical ones – rules that accommodate atypical variants such as rhyme riche, near rhyme, slant rhyme, off rhyme, eye rhyme, and so on. Hence, we shall focus on gradient, statistically normative rules hereafter.

156

2.

Gilbert Youmans and Xingzhong Li

Statistical evidence for verse prototypes

2.1 Stress patterning The hierarchical pattern of stress prominences proposed in (1) and (2) is consistent with the statistics compiled in Tarlinskaja (1976). Tarlinskaja scanned 625-line samples from the nondramatic iambic pentameter verse of thirty-two English poets from Chaucer to Swinburne and then constructed stress profiles for each poet (Tarlinskaja 1976: 279–80). For every poet in this sample, the highest incidence of stressed syllables is in the tenth position, and the next highest is in the fourth position. Conversely, in W metrical positions, the highest incidence of stressed syllables is in the first position. Table (1) shows Tarlinskaja’s statistics for Chaucer, Shakespeare, and Milton. Table (1). Percentages of stessed syllables in each metrical position (nondramatic iambic pentameter verse)

Position Chaucer

1

2

18.0 80.3

3

4

5.6 89.2

5

6

8.3 78.5

7

8

5.2 76.8

9

10

3.2 99.5

Shakespeare 27.3 65.8 11.6 90.2 13.1 71.1 11.6 76.0 10.4 94.3 Milton

31.6 72.2

9.6 79.3

9.9 75.2

8.6 75.7

4.6 96.2

Table (1) supports Cable’s observation (1991, 1996, and this volume) that Chaucer’s verse corresponds more closely than Shakespeare’s does with the alternating pattern of stresses that Cable takes to be the defining characteristic of Chaucer’s decasyllabic verse, WSWSWSWSWS. By contrast, Cable classifies the iambic pentameter of poets such as Shakespeare and Milton as typologically distinct, with metrically significant foot boundaries: [WS] [WS] [WS] [WS] [WS]. Unlike Cable, we adopt the traditional view that Chaucer wrote iambic pentameter verse, with the foot as a metrically significant constituent. Table (1) illustrates that deviations from the prototypical WSWS pattern occur significantly less often in Chaucer than in Shakespeare and in Milton.

Chaucer: Folk poet or littérateur?

157

However, when deviations from the prototype in (1) do occur, they tend to occur in the same metrical positions for all three poets. In this respect their verse is alike. The highest incidence of stresses is in position ten, and the next highest is in position four. The lowest incidence of stresses is in position nine. Compared with other W positions, position one has the highest percentage of stresses, and position five has the next highest. Hence, all three poets constrain stress patterning most strictly at the ends of their lines (and hemistichs), and least strictly at the beginnings of their lines (and hemistichs). Such facts tend to support the hierarchical pattern of stresses represented in (1) and (2) – for Chaucer as well as for Shakespeare and Milton – rather than an undifferentiated sequence of WSWS metrical positions. Chaucer deviates less often from a WSWS pattern than Shakespeare and Milton do, but when he does, he deviates in the same ways and in the same places that later iambic poets do. Li (1995) provides substantial additional evidence for (1) as a prototype for Chaucer’s iambic pentameter. Li’s statistics are based on his scansion of a systematic sample (every tenth line) of Chaucer’s verse, including 3020 lines of iambic pentameter. In contrast with table (1), which records statistics for single syllables, table (2) records the percentages of trochees in Chaucer’s pentameter verse (that is, SW sequences of syllables occupying putatively WS metrical positions).

Table (2). Incidence of trochees in 3020 lines of Chaucer’s iambic pentameter

Foot Trochees

1 241 (8%)

2

3

4

50 (1.7%) 111 (3.7%) 108 (3.6%)

5 6 (0.2%)

Again, deviations from prototype (1) are most frequent at the beginning of the line (in the first foot) and next most frequent at the beginning of the second hemistich (the third foot in 4//6 lines and the fourth foot in 6//4 lines). Conversely, deviations are least common at the end of the line (the fifth foot) and next least at the end of the first hemistich (the second foot in 4//6 lines). Hence, Li’s statistics parallel Tarlinskaja’s closely, and both studies lend strong support to a hierarchically structured prototype for Chaucer’s iambic pentameter such as (1).

158

Gilbert Youmans and Xingzhong Li

2.2 Subdivisions within lines (caesuras) Li (1995) compiles statistics for linguistic boundaries as well as stress patterning in Chaucer’s verse, and these statistics, too, tend to confirm prototype (1). As part of his scansion, Li identifies the most prominent syntactic subdivision within each line of his 3020-line sample. Table (3) records the percentages of primary subdivisions that occur after each metrical position (Li 1995: 223).

Table (3). Percentages of primary subdivisions after each metrical position in Chaucer’s pentameter verse

Position Primary subdivision

1

2

3

4

5

6

0.17

5.4

4.9 48.9 13.2 18.6

7

8

9

6.7

2.0 0.03

As table (3) indicates, by far the most common location for primary subdivisions is after the fourth position, resulting in prototypical 4//6 lines, exactly as depicted in (1). The next most common subdivision is after the sixth position, as also depicted in (1). The third most common subdivision is after the fifth position. The resulting 5//5 lines deviate from the prototype in (1) in that they divide the line within a metrical foot rather than between feet; however, they divide the line roughly in half – roughly, because 5//5 lines typically have two stresses in the first hemistich and three in the second. In this sense, they can be considered variants of the prototypical 4//6 line. Lines divided 3//7, 7//3, 2//8, and 8//2 are less common, and lines divided 1//9 and 9//1 are very rare. Most of these lines have secondary subdivisions after the fourth, fifth, or sixth syllables, making them 4/6, 5/5, 6/4 variants of 4//6, 5//5, 6//4 lines. All together, 75% of Chaucer’s primary subdivisions occur after evennumbered metrical positions; which is to say, they coincide with foot boundaries. Conversely, 25% of Chaucer’s primary subdivisions occur within a metrical foot. Statistically speaking, then, Chaucer’s syntactic boundaries tend to reinforce foot boundaries far more often than not, and they lend further support to the claim that a hierarchically structured prototype such as (1) is required to represent Chaucer’s decasyllabic verse accurately.

Chaucer: Folk poet or littérateur?

159

2.3 Variations in the number of syllables per metrical position The prototype in (1) implies that, in the unmarked case, each metrical position is occupied by one and only one syllable. Of course, Chaucer occasionally omits a W syllable at the beginning of his lines, or (less often) at the beginning of the second half-line. In his 3020-line sample, Li (1995: 238) identifies only thirty-four (1.1%) headless lines; four lines (0.13%) with syllables omitted from position five, and five lines (0.17%) with omissions in position seven. These statistics, too, tend to confirm the prototype in (1), because omissions are most common at the beginning of the line and next most common (though very rare) at the beginning of the second hemistich. Chaucer omits syllables more often in his tetrameter than in his pentameter verse; for example, fifty (9.4%) of the lines in Li’s tetrameter sample are headless. Midline omissions, resulting in so-called “broken-backed lines,” are much more common in Lydgate than in Chaucer: Tarlinskaja (1976: 278) identifies omissions from position five in 5.4% of the lines in her sample from Lydgate. Addition of unstressed syllables is far more common than omission. Both Chaucer and Shakespeare frequently include an extrametrical W syllable at the end of a line or (far less often) at the end of the first hemistich. Both poets, too, adopt various ways of introducing unstressed syllables in their verse; for example, Chaucer’s archaic final -e’s and Shakespeare’s pronunciation of –ed as an extra unstressed syllable when needed. All such metrical aids reinforce the prototypical pattern of alternating stresses WSWS in iambic verse. Like many other iambic poets, Chaucer occasionally includes two unstressed (or weakly stressed) syllables in metrical W positions: (5) Of a solempne and a greet fraternitee A.GP.364 1 23 4 5 6 7 8 9 10 (6) As wel of this as of othere thynges moore D.WB.584 1 2 3 4 5 6 7 8 9 10 Li (1995: 236) identifies just thirty-four anapests such as these in his sample from Chaucer’s iambic pentameter verse. This sort of disyllabic oc-

160

Gilbert Youmans and Xingzhong Li

cupancy of W positions is much more common in ballads and other folk verse forms than it is in iambic or trochaic verse. For example, consider (7): (7) “Last night, the moon had a golden ring, w s w s w w s w s And to-night no moon we see!” w w s w s w s The skipper, he blew a whiff from his pipe, w s w w s w s w w s And a scornful laugh laughed he. (“The Wreck of the Hesperus,” 17–20) w w s w s w s Categorically speaking, the disyllables in W positions in (5) and (6) are similar to the ones in (7); but the frequency of such disyllables is much higher in Longfellow’s ballad than in Chaucer’s verse, giving the two a very different flavor. Hence, statistical frequencies, more than categorical rules, differentiate the anapests in ballads such as (7) from those in iambic verse. The perceptual effect of ballad verse, with one or two unstressed syllables between major stresses, is to accentuate rhythmic beats, which also tend to be “temporally rigid” (Oehrle 1989) – occurring at perceptually similar intervals. This rhythmic regularity is especially obvious in verse forms that can be set to music, as well as in nursery rhymes, limericks, chants, and comic verse. Hence, readers of (7) are likely to “tilt” Longfellow’s ballad stanza toward rhythmic regularity – by speeding up unstressed disyllables, slowing down unstressed monosyllables, and by giving extra stress to rhythmic beats and less stress to offbeats. Cable (1996 and this volume) describes his own reading of Chaucer in similar terms. However, Chaucer’s iambic pentameter verse, like that of Renaissance poets, arose out of a continental literary tradition (the French decasyllable and/or the Italian hendecasyllable) rather than from native English folk verse. Consequently, Chaucer’s iambic pentameter verse, like its continental forebears, shows greater consistency in its syllable-count than English folk verse does, combined with greater variety (and less insistency) in its rhythmic patterning, and therefore less tendency toward tilting

Chaucer: Folk poet or littérateur?

161

than native folk verse. (See Stockwell and Minkova 2001 for discussion of continental influences on Chaucer.)

2.4 The syntactic inversion test Gascoigne ([1575] 1868) includes the following advice in “Certain notes of instruction,” his manual for aspiring Renaissance poets: You shall do very well to use your verse after the English phrase, and not after the manner of other languages ... if we should say in English a woman fair, a house high, etc. it would have but small grace: for we say a good man, and not a man good, etc. ... Therefore even as I have advised you to place all words in their natural or most common and usual pronunciation, so would I wish you to frame all sentences in their mother phrase and proper Idióma, and yet sometimes the contrary may be borne, but that is rather where rhyme enforceth, or per licientiam Poëticam, than it is otherwise lawful or commendable. (Gascoigne 1868: 37)

Thus, Gascoigne explicitly advises poets against altering the usual pronunciation of English words to suit a poem’s meter, and against altering idiomatic English word order, except where rhyme or meter “enforceth.” Insofar as other English poets have shared these views, we might expect their verse to avoid altered pronuncations (such as Longfellow’s daughtèr in the epigraph to this article), and we might expect syntactic inversions such as a woman fair to show a strong correlation with metrical rules governing rhyme and/or stress patterning, as in (8) and (9): (8) Whan that Aprill with his shoures soote ...*soote shoures

A.GP.1

(9) The droghte of March hath perced to the roote A.GP.2 w s w s w s w s w s (x) *Hath perced the droghte of March … w s w w s w s In (8), “rhyme enforceth” the inversion from an adjective – noun to a noun – adjective sequence. In (9) the inversion from verb – object to object – verb has no effect on rhyme, but it does regularize the placement of

162

Gilbert Youmans and Xingzhong Li

stressed and unstressed syllables in the line. We shall refer to all such syntactic movement transformations that reinforce rhyme schemes and/or increase metrical regularity as “Gascoigne transformations.” Gascoigne transformations comprise the overwhelming majority of syntactic inversions in Chaucer. Youmans (1996) found that some 77% of the inversions in the General Prologue to the Canterbury Tales resemble the one in (8); that is, they shift a rhyme to the end of a line. The statistics for Shakespeare’s Sonnets are similar: about 79% of the syntactic inversions in the Sonnets shift a rhyme to the end of a line (Youmans 1983). Conversely, no syntactic inversion shifts a rhyme away from the end of the line. This fact will surprise no one, because end-line rhyme is a categorical requirement in Chaucer’s verse and in Shakespeare’s Sonnets. More surprisingly, Youmans (1983, 1989, 1996) and Li (1995) also find a nearly perfect correlation between syntactic inversions and strongly normative metrical rules such as the Stress Maximum Principle and the Monosyllabic Word Constraint, as table (4) illustrates:

Table (4). Syntactic inversions with no effect on rhyme in Chaucer, Shakespeare, and Milton

Verse Sample Chaucer (Youmans) Chaucer (Li) Shakespeare Milton Totals

Total Lines of verse

Syntactic Inversions with no effect on rhyme

Inversions Inversions Preventing viola- Causing violations of the tions of the SMP or MWC SMP or MWC

5200

538

287 (53%)

0

3020 4906 7339 20465

366 503 1990 3397

219 (60%) 291 (58%) 1193 (60%) 1990 (59%)

1 (?) 0 0 1 (?)

(The single inversion that may induce rather than prevent a violation of the SMP and the MWC is in D.ClT.600, if aperceyve is scanned SWSW, although this word is also aligned in WSW position in TC.4.656, so it may have been stressed WSW in both lines.) All together, then, Youmans and Li have analyzed 3397 syntactic inversions having no effect on rhyme in 20465 lines of Chaucer, Shakespeare,

Chaucer: Folk poet or littérateur?

163

and Milton; 1990 (59%) of these inversions prevent violations of the SMP and/or the MWC, and one may have the opposite effect. Hence, like the correlation between syntactic inversions and rhyme schemes, the correlation between inversions and strongly normative metrical rules is almost perfect, and the statistical probability that this correlation occurs by chance is nearly zero. Gascoigne would be pleased. Syntactic inversions also show a strong positive correlation with other normative metrical principles. For example, Youmans (1996) found thirtyeight additional Gascoigne transformations such as the one in (10), which does not prevent a violation of either the SMP or the MWC, but which brings the pattern of stresses closer to the WSWS prototype: (10)

That lene he wex and drye as is a shaft w s w s

A.KT.1362

That he wex lene … w w s s All together, more than 70% of the syntactic inversions with no effect on rhyme have positive metrical effects, as summarized in table (5): Table (5). Metrical effects of syntactic inversions in Chaucer, Shakespeare, and Milton

Verse Sample Chaucer (Youmans) Chaucer (Li) Shakespeare Milton

Inversions with no effect on rhyme

Positive Metrical Effect

538

70%

25%

5%

366 503 1990

85% 77% 85%

13% 20% 13%

3% 3% 2%

Neutral or Minor Metrical Effect

Negative Metrical Effect

Hence, there is a strong correlation between syntactic inversions and positive metrical effects: more than three-fourths of Chaucer’s inversions preserve his rhyme schemes and about the same percentage of his remaining inversions have positive metrical effects, which is to say that some 95% or

164

Gilbert Youmans and Xingzhong Li

more of Chaucer’s inversions are Gascoigne transformations, serving to regularize his rhyme schemes and/or his meter. The small number of inversions with negative metrical effects are especially informative, because they highlight permissible variations in Chaucer’s verse. Youmans (1996) found just twenty-seven inversions with negative metrical effects, and Li (1995) found just ten. Of these, twelve (32%) transform prototypical WSWS sequences into initial trochees SWWS: (11) Short was his gowne ... [His gowne was short ...] (12) Whit was his berd ... [His berd was whit ...]

A.GP.93

A.GP.332

(13) Swelleth the breast of Arcite ... [The breast of Arcite swelleth] A.KT.2743 Cable (1996 and this volume) points out, correctly, that Chaucer’s verse has significantly fewer initial trochees than Shakespeare, but inversions such as the ones in (11) – (13) demonstrate that initial trochees are not only permissible variations in Chaucer but also (occasionally) desirable rhythmic alternatives. Initial trochees are by far the most common and least disruptive trochaic substitutions in iambic verse. Next most common and next least disruptive are trochees in the third or fourth foot; a smaller number of Chaucer’s inversions introduce trochaic cadences in these feet: (14) That Calkas, traitour fals, fled was and allyed ... was fled …

TC.1.87

(15) Soul as the turtle that lost hath hire make. ... hath lost …

E.MerT 2080

(16) How dorst I mo tellen of this matere … tellen mo …

TC.3.370

Such lines are rare, but, when they are combined with other lines containing trochaic cadences in foot three and four, they suggest that Chaucer

Chaucer: Folk poet or littérateur?

165

treats these deviations, too, as (occasionally) desirable variations from his rhythmic norm, rather than as desperate remedies.

3.

Contesting the statistical evidence for verse prototypes

The syntactic inversion test is criticized occasionally because there is no way to know in individual cases whether the motivation for a specific inversion is purely rhetorical rather than metrical; and, in fact, a minority of inversions do have neutral or even negative metrical effects. Statistically speaking, though, the motivation for individual inversions is less significant than the fact that all lines of verse must meet simultaneous demands of meter, meaning, and syntax, whereas their rephrased versions are free from any metrical constraints. On the average, then, verse lines are bound to be more metrical than their rephrased counterparts. Unlike the usual set of unattested lines cited by metrical theorists, these rephrased lines are synonymous with, and syntactically more regular than, their verse counterparts. Hence, they are more likely to isolate purely metrical variables in verse. Tarlinskaja’s and Li’s statistics on stress patterning in §2.1 might be contested, too. For example, Cable (this volume) would count most of the trochees identified by Tarlinskaja and Li as iambs, radically reducing the frequency of rhythmic variations in Chaucer’s verse. In the case of phrases composed of monosyllabic words, Cable claims that Short was his gowne in (11) and Whit was his berd in (12) begin with spondees (“hovering stress” or “level stress”) rather than trochees. But by this account there would be little motivation for the syntactic inversions in these phrases. Topicalization transformations typically preserve underlying stress prominence relations:

.

x

.

x x

x x

.

.

x

(17) His gowne was short → Short was his gowne This is the primary function of topicalization transformations – to give special emphasis to “moved” constituents by situating primary stresses in syntactically and phonologically marked positions. But this initial emphasis disappears if we accept Cable’s analysis:

166

Gilbert Youmans and Xingzhong Li

x x . x . x x x . x (18) His gowne was short → Short was his gowne From Gascoigne’s point of view, this scansion makes Chaucer’s line (and dozens more like it) triply inept, because abnormal word order supposedly induces abnormal pronunciation – at the expense of metrical regularity. We do not attribute any such ineptitude to Chaucer. Hence, we contend that (17) is the correct scansion for this line and that all such lines provide additional evidence that SW onsets are desirable variations in Chaucer’s verse rather than prohibited cadences. In the case of polysyllabic words, Cable contends that Chaucer’s verse includes more stress doublets such as hónour and honóur than commentators such as Tarlinskaja, Li, and Youmans recognize. For example, Cable reads disyllabic words ending in –y, such as redy, hardy, and worthy, as iambs rather than trochees in lines such as the following: (19) Redy to wenden on my pilgrymage (20) Hardy he was and wys to undertake

A.GP.21 A.GP.405

He was hardy … (21) Worthy to been stywardes of rente and lond

A.GP.579

By contrast, we scan these italicized words as initial trochees, taking them as further evidence of the permissiveness of the first foot in Chaucer’s verse. Line (20) is especially significant, because it includes a Gascoigne transformation that prevents a violation of the SMP and the MWC, and it moves a trochaic word from the second (more strictly constrained) foot into the first (least strictly constrained) foot. Cable acknowledges that words such as redy, hardy, and worthy occur more often in SW than in WS positions, indicating that Chaucer normally treats these words as trochees. However, Cable also points out that such words occasionally occur in rhymes, indicating that they may be intended as spondees or even iambs in final position. Cable concludes, therefore, that such words are also intended as iambs in other WS positions, as

Chaucer: Folk poet or littérateur?

167

in (19)–(21). However, Youmans (1996) raises a different possibility in his discussion of (22): (22) Arsenyk, sal armonyak, and brymstoon G.CY.798 … there are many ... disyllabic words in Chaucer’s verse that are aligned exclusively with SW positions except when these words occur at the end of a line – that is, when they are used as rhymes. This distribution suggests two alternatives: either Chaucer relaxed the SMP in position 9 (which would be unusual, because verse tends to become more, not less, regular at the end of lines), or he followed a stress-promotion convention for rhyming syllables, changing a line-final word such as brimstoon from falling stress to level or rising stress. (Youmans 1996: 188)

That is, Youmans (1996) acknowledges the possibility of Cabelian tilting for words such as redy, hardy, and worthy in rhyming position, but he does not take this as evidence of tilting elsewhere in the line. In fact, the statistical distribution of these words tends to support the opposite conclusion. Paradoxically, words such as redy, hardy, and worthy occur in rhyming position more often than in any other WS position in Chaucer’s line; that is, they are most common in Chaucer’s most strictly constrained foot. Outside this final rhyming position, such words normally occupy SW positions, and when they occupy WS positions, they occur most often in Chaucer’s least strictly constrained feet (especially in the first foot). That is, except in rhyming position, words such as redy, hardy, and worthy tend to occur in exactly the same metrical positions as other trochees. (Minkova 1997 and 2000 reaches the same conclusion about other possible stress “doublets” such as author, purpose, labor.) The distribution of the word squier illustrates this principle. There are twenty-seven occurrences of squier, squyer, squier(e)s, and squiereth in Canterbury Tales (Benson 1993). Chaucer treats squier as disyllabic, although he also treats squier(e)s and squiereth as disyllables, implying that either elision or syncopation occurs in these two words. Squier and its variants occur eighteen times (67% of the time) in SW metrical positions, as in (23): (23) This lusty squier, servant to Venus,

F.FranT.937

However, squier(es) also occurs in WS rhyming positions seven times (26% of the time):

168

Gilbert Youmans and Xingzhong Li

(24) With hym ther was his sone, a yong SQUIER, A lovyere and a lusty bacheler, A.GP.79–80 This is an atypical rhyme for Chaucer. It requires readers to do one of two things: either alter their usual pronunciation of the rhyming words, or suspend their strong expectation of a WS closure to Chaucer’s lines. We know of no conclusive evidence for choosing between these two alternatives, but, given the rarity of final trochees in Chaucer’s verse, we are inclined to agree with Cable that Chaucer intended his rhyming syllables to be stressed, regardless of their usual pronunciation. Outside of rhyming position, squier occurs just twice in WS positions, once at the beginning of a line: (25) Squier, com neer, if it youre wille be,

F.IntSqT.1

And once as an interpolated noun of address in the second foot: (26) In feith, Squier, thow hast thee wel yquit

F.SqT.673

Interpolations constitute separate intonation phrases. They rank among the most frequent sites for trochaic substitutions in iambic verse. Hence, we conclude that squier is intended to be pronounced in its usual way, SW, in both (25) and (26), as it is everywhere else in Chaucer’s verse outside of rhyming position. The statistics for genuine stress doublets such as hónour and honóur are markedly different. Out of seventy occurrences in Canterbury Tales, honour (noun) appears in rhyming position thirty-two times. Elsewhere, honour occurs nineteen times in SW positions and nineteen times in WS positions – twelve times in the second foot (at the end of the first hemistich) and only once in the first foot. Hence, there is strong statistical evidence that Chaucer’s pronunciation of honour vacillated between WS and SW. This systematic contrast between words such as honour and squier illustrates an important flaw in Cable’s tilting argument: it fails to explain the statistical distribution of variations in stress patterning described by analysts such as Tarlinskaja and Li. Words with vacillating stress, such as honour, are distributed more or less equally throughout Chaucer’s lines, whereas more radical candidates for tilting, such as squièr, are most frequent at

Chaucer: Folk poet or littérateur?

169

the end of Chaucer’s lines, where stress patterning is most strictly constrained. On the whole, however, tables (1) and (2) demonstrate that Chaucer’s verse includes very few candidates for tilting at the ends of his lines (and hemistichs). To the contrary, most of the deviations in Chaucer’s stress patterning occur in the same metrical positions that are constrained least strictly by Shakespeare, Milton, and other iambic poets; that is, at the beginnings of lines and hemistichs, as in (27): (27) Trouthe and honour, fredom and curteisie. A.GP.46 Hence, we conclude that both half-lines in (27) begin with trochees, and that such trochaic onsets are permissible variations in Chaucer’s verse rather than prohibited cadences. Among the least likely candidates for metrically induced tilting are trochaic words formed from a root plus an inflectional suffix, as in lines such as the following cited by Cable (this volume): (28) Lyned with taffata and with sendal

A.GP.440

(29) Tukked he was as is a frere aboute

A.GP.621

In these cases, Cable reads (28) and (29) as headless lines with syncopated vowels in lyn’d and tukk’d. From the point of view of Optimality Theory, Cable’s analysis reverses the order of Chaucer’s metrical constraints, because headless lines are far less common than initial trochees in his verse. But from Cable’s point of view, his analysis explains why apparent trochees such as lyned and tukked occur in initial positions more often than elsewhere – because there they can be treated as the onsets to headless lines. However, similar trochaic words also occur occasionally in the third foot: (30) Ther may no man clepen it cowardye. A.KntT.2730 (31) Of worldly folk holden the siker weye; (32) How that a knyght, called Virginius, And in the fourth foot:

E.MerT.1390 C.PhyT.180

170

Gilbert Youmans and Xingzhong Li

(33) But hood, for jolitee, wered he noon, ... he wered …

A.GP.680

(34) His felawe hadde a staf tipped with horn,

D.SumT.1740

If syncopation were invoked to explain these examples, then they would have to be classed as “broken-backed” lines (which are very rare in Chaucer). In one case (33) the trochee even results from a syntactic inversion, which is further evidence that such trochees are permissible variations in Chaucer’s verse. We scan the italicized words in all these examples (28)–(34) as trochees and consider them further evidence that Chaucer’s verse permits variations in stress patterning remarkably similar to those in Shakespeare and Milton.

4.

A brief comment on foot-based metrics

Many metrists since Jespersen (1913) have abandoned the foot as a significant metrical unit. Major phonological boundaries, including even sentence boundaries, are free to occur in mid-foot. How, then, can feet be metrically significant? Our answer is to treat foot boundaries as statistically normative rather than as categorical constraints. For example, in both Chaucer and Shakespeare, 4//6 lines and 6//4 lines are much more common than 5//5 lines; analysis of Gascoigne inversions in Milton shows that he prefers iambic bracketing at the ends of lines and hemistichs (Youmans, 1989); and Tarlinskaja (1976) and Kiparksy (1977) point out that trochaic words such as holden occasionally occur in WS positions in iambic verse, whereas iambic words such as behold very rarely occur in SW positions, suggesting that iambic verse permits variations in either the stress patterning or the metrical bracketing of polysyllabic words, but very rarely in both at the same time. Hence, foot boundaries seem to be metrically significant, a result that is consistent with recent theories of metrical phonology. So far as we know, Cable is unique in proposing that feet are metrically normative in Shakespeare but not in Chaucer. For Cable, the supposed absence of foot boundaries in Chaucer’s verse makes it more, rather than less, strictly constrained, because stress contrasts are metrically significant in two directions rather than one. That is, the second syllable in Chaucer’s line

Chaucer: Folk poet or littérateur?

171

should be stronger than both the first and third syllables; the third syllable should be weaker than both the second and fourth syllables, and so on throughout the line, whereas Shakespeare need worry about relative stress in only one direction, between syllables within a single foot. In support of his view, Cable cites examples from Shakespeare such as the following: (35) When to the sessions of sweet silent thought I summon up remembrance of things past (Sonnet 30.1–2) Traditional metrists might classify the italicized sequences in this passage as ionic feet (WWSS) – double iambs, so to speak. Such syncopated rhythms are surprisingly euphonious in iambic verse. Cable scans such syllables as sequences of rising stresses [12][34], which supposedly makes them perfectly regular in iambic meter but irregular in alternating meter. Cable contends that such sequences are relatively common in Shakespeare, but rare in Chaucer, although Chaucer does include similar patterns occasionally: (36) For with good hope he gan fully assente (37) Love hath byset the wel; be of good cheere! For of good name and wysdom and manere

TC.1.391

TC.1.879–80

Reviewing his sample from Chaucer, Li identifies such WWSS sequences in about 6.8% of Chaucer’s lines – uncommon, yes, but clearly permissible. Unlike Cable, we scan the stresses in the italicized sequences in (35) as [1134], with no significant contrast between of and the preceding unstressed syllables –sions and –brance. Of is the weakest of all English prepositions. It is frequently reduced to schwa in speech, with no final consonant [v], and it is commonly cliticized with either the preceding or the following word. In lines such as (35) no crucial metrical issue depends on whether of is pronounced with the same stress as the preceding syllable, or with slightly more stress or slightly less. Readers may experiment with these different pronunciations to see whether such minor variations have any significant metrical effect. Even more damaging to Cable’s theory, however, is the fact that both Chaucer and Shakespeare permit fully stressed trochees [41] such as clepen and tellen in WS positions, so it is un-

172

Gilbert Youmans and Xingzhong Li

likely that strict constraints are placed on pyrrhic feet [11] or weak trochees [21] in these positions. We are also skeptical about Cable’s claim that there is an obligatory metrical pause between the adjacent stressed syllables in WWSS sequences. Readers may experiment with this claim by placing an actual pause between sweet and silent in (35) ( ... sweet, silent thought). To our ears, any such pause makes the line less, not more, regular. Conversely, if we artificially tilt the line toward its iambic norm, we are more inclined to rush over the words sweet and things than to linger over them. Consequently, we believe that Cable’s analysis of metrical pauses is incorrect. Finally, Cable criticizes generative metrists for focusing on W metrical positions and ignoring S positions. He points out that, because stress is relative, prohibitions such as the SMP and the MWC necessarily place simultaneous constraints on both W and S positions. However, metrical phonologists since Liberman and Prince (1977) have treated phonological S and W positions asymmetrically. Syllables within words and phrases are grouped into feet, with unstressed syllables subordinated to stressed syllables. For example, the first and last syllables of a word such as remembrance are subordinated to the middle S syllable, but they are not compared directly with each other or with any other syllables in the line, even with the adjacent unstressed word of, which can be subordinated either to the preceding stressed syllable –mem–, or to the following one, things. Within metrical phonology, it is impossible to measure whether one unstressed syllable is more salient than another even if the two syllables are adjacent. By contrast, the middle S syllable of remembrance is compared directly with the unstressed syllables before and after it, and it is also subordinated to the final stressed syllable in the same phrase, past. Theoretically, then, all unstressed syllables are compared directly with one (and only one) stressed syllable; whereas stressed syllables are compared directly with one, two, or more syllables. From this point of view, an asymmetry between S and W positions in generative metrical theory makes sense. Basically, the SMP and the MWC single out especially salient stressed syllables (stress maxima and stressed syllables in polysyllabic words) and prohibit such syllables in W positions in verse. These rules can be restated as positive prescriptions: all especially salient syllables must be aligned with S positions. As for less salient syllables, they are freer to fall where they may.

Chaucer: Folk poet or littérateur?

5.

173

Conclusion

An alternating stress pattern, WSWS, is statistically normative in all iambic pentameter verse. Chaucer conforms more closely with this WSWS prototype than Shakespeare does, but less closely than Cable contends. When Chaucer and Shakespeare deviate from prototypical patterns, they do so in remarkably similar ways. Therefore, we consider the difference between them to be a matter of metrical style rather than metrical type. In short, we accept the traditional view that Chaucer deserves full credit for writing the first iambic pentameter verse in English. Chaucer – folk poet or littérateur? A court poet, polyglot, translator, European ambassador, a father who felt it necessary to explain to his tenyear-old son why he wrote Treatise on the Astrolabe in English rather than Latin – our answer is littérateur.

References Benson, Larry D. 1993 A Glosorial Concorrdance to the Riverside Chaucer. New York: Garland. Biggs, Henry 1999 The Classic French decasyllable of DuBellay (16th Century): A generative metrics perspective. Conference paper presented at: Formal approaches to poetry and recent developments in generative metrics. University of Toronto. Cable, Thomas 1991 The English alliterative tradition. Philadelphia: University of Pennsylvania Press. Cable, Thomas 1996 Clashing stress in the metres of Old, Middle, and Renaissance English. In: McCully and Anderson (eds.), 7–29. Cable, Thomas (this Issues for a new history of English prosody. volume) Chaucer, Geoffrey 1989 The Complete Poetry and Prose of Geoffrey Chaucer, 2nd edition. Edited by John H. Fisher. New York: Holt, Rinehart and Winston.

174

Gilbert Youmans and Xingzhong Li

Gascoigne, George 1868 Certain notes of instruction. Reprinted in English Reprints 10–12. Edited by Edward Arber. Pp. 31–40. First published [1575]. Halle, Morris and Samuel Jay Keyser 1966 Chaucer and the study of prosody. College English 28: 187–219. Halle, Morris and Samuel Jay Keyser 1971 English Stress: Its Form, Its Growth, and Its Role in Verse. New York: Harper & Row. Hanson, Kristin and Paul Kiparsky 1996 A parametric theory of poetic meter. Language 72: 287–335. Hanson, Kristin and Paul Kiparsky 1997 The Nature of verse and its consequences for the mixed form. In: Prosimetrum: Cross-cultural perspectives on narrative in prose and verse. Edited by Joseph Harris and Karl Reichi. Cambridge: Brewer, 17–44. Hayes, Bruce and Margaret MacEachern 1998 Quatrain form in English folk verse. Language 74: 473–507. Jespersen, Otto 1913 Notes on metre. In: Linguistica: Selected Papers in English, French and German. Copenhagen: Levin & Munksgaard, 249–274. Kiparsky, Paul 1975 Stress, syntax, and meter. Language 51: 576–616. Kiparsky, Paul 1977 The rhythmic structure of English verse. Linguistic Inquiry 8: 189–247. Kiparsky, Paul and Gilbert Youmans (eds.) 1989 Phonetics and Phonology I: Rhythm and Meter. San Diego: Academic Press. Li, Xingzhong 1995 Chaucer’s meters. Ph.D. dissertation, Department of English, University of Missouri, Columbia. Liberman, Mark and Alan Prince 1977 On stress and linguistic rhythm. Linguistic Inquiry 8: 249–336. Longfellow, Henry Wadsworth 1975 The Poetical Works of Longfellow. Boston: Houghton Mifflin.

Chaucer: Folk poet or littérateur?

175

McCully, C.B. and J.J. Anderson (eds.) 1996 English Historical Metrics. Cambridge: Cambridge University Press. Minkova, Donka 1997 Constraint ranking in Middle English stress-shifting. English Language and Linguistics 1(1): 135–175. Minkova, Donka 2000 Middle English prosodic innovations and their testability in verse. In: Irma Taavitsainen, Terttu Nevalainen, Päivi Pahta, and Matti Rissanen (eds.), Placing Middle English in Context, Berlin/New York: Mouton de Gruyter, 431–461. Oehrle, Richard 1989 Temporal structures in verse design. In: Kiparsky and Youmans. Pp. 87–119. Piera, Carlos. 1980 Spanish verse and the theory of meter. Ph.D. dissertation, University of California, Los Angeles. Prince, Alan and Paul Smolensky 1993 Optimality theory: constraint interaction in generative grammar. Unpublished ms. Rutgers University and University of Colorado, Boulder. Stockwell, Robert and Donka Minkova 2001 On the partial-contact origins of English pentameter verse. In: Dieter Kastovsky and Arthur Mettinger (eds.), Language Contact in the History of English, Frankfurt am Main: Peter Lang Verlag, 337–363. Tarlinskaja, Marina 1976 English Verse: Theory and History. The Hague: Mouton. Youmans, Gilbert 1983 Generative tests for generative meter. Language 59: 67–92. Youmans, Gilbert 1989 Milton’s meter. In: Kiparsky and Youmans. Pp. 341–79. Youmans, Gilbert 1996 Reconsidering Chaucer’s prosody. In: McCully and Anderson. Pp. 185–209.

A rejoinder to Youmans and Li Thomas Cable

The paper by Youmans and Li is useful in its own right for the new information that it presents and as a commentary on my paper. Our full exchange may clarify our similarities on Chaucer’s meter and our differences on Shakespeare’s meter, along with differences in the general approach to metrics that are implied. Some of these differences are suggested by Youmans and Li, who call their approach Platonic, as against that of the Aristotelians. However, since Aristotle puts me in the company of the rest of the generative metrists, further distinctions will need to be drawn.

Summary of differences There is a clear difference in the way that Shakespeare is read by all generative metrists and by a large number of traditional metrists, cutting across the usual divisions of stressers (for example, W.K. Wimsatt) and timers (for example, George R. Stewart). It will help to state those differences at the beginning, most of which have been discussed by Wimsatt with specific reference to generative metrical theory of the time. (1) Five beats to the line. It is necessary to locate the syllables on which taps occur. There may be different “performances,” but only one of those is the poet’s performance. (2) The tilting effect of meter. To get the five beats, it is sometimes necessary to rearrange the supposedly “normal” timing and stress pattern of phrases, if there indeed is such a thing as a normal pattern. Dwight Bolinger argued that there is no normal pattern but only contextual factors, a view with much to recommend it. Whatever normal tendencies there may be are often overridden in poetry by the contextual factor of “the running disposition of ictus and non-ictus” (Chatman 1965: 121). (3) The significance of stress on non-lexical words. Except for words that had variable stress at certain points in the history of the language, lexical stress cannot be tilted: there is only one way to stress apricot. Generative

178

Thomas Cable

metrists stick with the four classes of “lexical” words as bearers of metrical ictus. However, by the traditional view, “non-lexical” or function words can bear ictus too. In the twenty-first century, there is only one way to stress under, and that pattern figures in the meter. (Kiparsky 1977 is noncommittal about stress on non-lexical words: having no status in determining metricality, it is relegated to the varying styles of performance.) (4) The isochronous foot. There may be reasons in lexical phonology for splitting a sequence of unstressed syllables in forming feet, but for purposes of poetic rhythm, the undivided sequence of unstressed syllables as described by Kenneth Pike, M.A.K. Halliday, and David Abercrombie makes for a more adequate theory. Examples such as “This is the house that Jack built” (from Abercrombie 1965: 18) are familiar in discussions of English as a stress-timed language. (5) The irrelevance of constituent bracketing. Except for a sharp phrasal break, where a medial “inverted foot” may occur, divisions between words and between syntactic constituents, or even between intonational constituents, are irrelevant for determining metricality, however useful they may be for discussing the more subtle matter of metrical tension. (6) The relevance of metricality. There is a clear distinction between metricality and tension. Youmans and Li are unusual among all metrists in opposing this idea with gradient rules, and with this we turn to their commentary.

Chaucer’s trisyllabic feet As Youmans and Li argue, I have doubtlessly overstated the frequency in Chaucer of tilted stress to avoid the inverted first foot. One reason for stating the proscription as strongly as possible is to force a new look at the apparent exceptions and, in a Neogrammarian spirit, look for explanations to the exceptions. The new information about after and under in my paper could be considered a result that might lead to further insights. Even more telling is the discovery that Chaucer does not use the inverted first foot with final -e in second position, a context that would be clear, unambiguous, and incontrovertible. It would be good to have an explanation for this missing pattern, which Youmans and Li do not discuss. Still, I am happy to remain agnostic on the matter until we know more.

A rejoinder to Youmans and Li

179

Youmans 1996 and Youmans-Li in this volume provide evidence that reinforces my general observation about Chaucer’s alternating meter. Perhaps a gradient rule of some kind is needed for “inverted first feet,” as they suggest and as Cable 1998 acknowledges. Even if the proscription on “inverted first feet” has to be relaxed, it is no more disruptive to take three of those syllables as an anapestic foot in an iambic meter than to take two of them as a trochee. Part of the traditional understanding of Chaucer’s prosody is “disyllabic position occupancy” (in generative terms).

Shakespeare’s trisyllabic feet and isochrony This point takes us to Shakespeare and our divergent treatments of “inverted first feet” in Modern English. For a line like Sonnet 5, line 8:

Beauty’s effect with beauty were bereft

The top branching shows how generative metrists would group the unstressed syllables. The bottom branching shows how an assortment of metrists from various traditions would group them, leaving aside the question of which stress the unstressed syllables are construed with (see Stewart 1930, Weismiller 1989 and references cited on p. 266, and Stockwell and Minkova 2001). These different points of view have their parallels in phonology, in which unstressed syllables are grouped into feet differently in two traditions, one mainly American (the studies following Liberman and Prince 1977, though these vary in their construction of feet) and one mainly British (Halliday, Abercrombie, Giegerich, but also Pike), where the idea of stress-timed languages in contrast with syllable-timed languages figures prominently. Hogg and McCully 1987: 220–27 succinctly summarize these two conceptions of the foot. My own view is closer to the British tradition, possibly because of a bias in coming to Modern English metrics from the isochronous structures of Old and Middle English poetry, especially the

180

Thomas Cable

“strong dip” (with the usual disclaimer that isochrony is perceptual and not of the clock).

Metrical pause There are other parallels between issues in phonology and in literary metrics. I am puzzled by both readings that Youmans and Li give to the lines from Shakespeare (37). The reading that they prefer, [1134], without a pause between the two stressed syllables, is contradicted by both the implications of Kiparsky’s Monosyllabic Word Constraint and by Liberman 1975 and many studies in metrical phonology that have followed Liberman. Because I thought that my teasing out the implications of Kiparsky’s MWC was one of the most useful parts of my paper, I am disappointed that Youmans and Li did not address that argument. Essentially it holds that since there are no unambiguous patterns of [1134] in Shakespeare’s verse (of the “Shal behold God” type), the implication is that ambiguous sequences follow some other pattern. I found my own reading to be in accord with Jespersen 1913. Reading the two stressed syllables without a pause ignores the original basis of metrical phonology in describing the avoidance of stress clash in sequences like “thirteen men” (by stress shift or beat insertion). Part of the problem may be in my informal use of “metrical pause,” by which I mean the kind of beat that occurs between stress clashes, whether or not an instrument shows that phonation actually ceases (which is not an essential or even very interesting matter).

Asymmetry in generative metrics In responding to the point about the asymmetrical focus on the W position, Youmans and Li say that that is the way it is in the theory of phonology they are following. This idea succinctly crystallizes the difference in our approaches. By my view, it is a pervasive flaw of generative metrics that it repeats syntactic and phonological analyses in the theory of meter. The literary metrist would like the grammarians and phonologists to get their work done and then pass the results along, keeping the sausage-making out of the metrics. Nothing fancier than a stress-marked lexicon is needed. This is

A rejoinder to Youmans and Li

181

simply the enduring concern in linguistics for the separation of levels of description. When the literary metrist receives the linguist’s description of normal stress, the considerations that led to a particular phonological constituent structure – stress shift, sandhi, etc. – are invisible. The metrist isn’t satisfied to be told that there was an asymmetry in the phonological theory that produced these patterns if there is no apparent reason in the verse. Even if we disregard the theoretical objection to importing phonological analyses into the meter, there are empirical problems. Every theory of generative metrics claims that a deviation in a W position counts for more than a deviation in an S position. This evaluation is very clear in Halle and Keyser, as in (19) of my paper, but it is also the case for Kiparsky, more obliquely, who would find two “mismatches” in “The curfew tolls the knell of parting day” (curfew and parting). For Kiparsky, mismatches are of two kinds – stress mismatches and bracketing mismatches – and both count in determining metricality. Curfew and parting are bracketing mismatches, because they don’t form iambic feet; however, the mismatch becomes relevant for considering metricality only when the stressed syllable occurs in a W position. It doesn’t matter where the unstressed syllable falls. I would eliminate the mix of syntactic and phonological information and translate Kiparsky’s findings into phonological terms. (Hayes 1989’s attempt to do this is still tied too closely to syntax.) The mismatches of word division then would contribute very slightly to tension, as would other contributing factors such as syllable weight, consonant clusters, and the caesura, but they would be irrelevant for determining metricality.

References Abercrombie, David 1965 A Phonetician’s view of verse structure. In: Abercrombie, David, Studies in phonetics and linguistics, London: Oxford University Press, 16–25. Cable, Thomas 1998 Metrical similarities between Gower and certain sixteenth-century poets. In: R.F. Yeager (ed.), Revisioning Gower 39–48. Asheville, NC: Pegasus Press. Chatman, Seymour 1965 A Theory of meter. The Hague: Mouton.

182

Thomas Cable

Hayes, Bruce 1989 The Prosodic hierarchy in meter. In: Paul Kiparsky and Gilbert Youmans (eds.). Phonetics and phonology I: Rhythm and meter, 201–60. San Diego: Academic Press. Hogg, Richard, and C.B. McCully 1987 Metrical phonology: A coursebook. Cambridge: Cambridge University Press. Jespersen, Otto 1913 Notes on Meter. In: 1933, Linguistica: Selected papers in English, French and German, 249–74. Copenhagen: Levin & Munksgaard. Kiparsky, Paul 1977 The Rhythmic structure of English verse. Linguistic Inquiry 8: 189–247. Liberman, Mark 1974 The Intonational system of English. Dissertation, Massachusetts Institute of Technology. Liberman, Mark, and Alan Prince 1977 On stress and linguistic rhythm. Linguistic Inquiry 8: 249–336. Stewart, George R., Jr. 1930 The Technique of English verse. New York: Holt. Stockwell, Robert P., and Donka Minkova 2001 The Partial-contact origins of English pentameter verse: the Anglicization of an Italian model. In: Dieter Kastovsky and Arthur Mettinger (eds), Language contact in the history of English, 337–63. Frankfurt am Main: Peter Lang. Weismiller, Edward R. 1989 Triple threats to duple rhythm. In: Paul Kiparsky and Gilbert Youmans (eds), Phonetics and phonology I: Rhythm and meter, 261–90. San Diego: Academic Press. Wimsatt, W.K. 1970 The rule and the norm: Halle and Keyser on Chaucer’s meter. College English 31: 774–88. Youmans, Gilbert 1996 Reconsidering Chaucer’s prosody. In: C.B. McCully and J.J. Anderson (eds.), English historical metrics, 185–209. Cambridge: Cambridge University Press.

On the development of English r Blaine Erickson

1.

Introduction1

In American English, the retroflex approximant [– ] and the non-retroflex central continuant [Ú] are the most common phones found for /r/. Cross-linguistically, both sounds are rare, and since they are not found as the dominant phones for /r/ in the other Germanic languages, the r phones of English are more likely to be innovations than retentions. Here, I investigate the history of r in English, and by employing data from English and other languages, propose a series of phonological changes to account for the modern English phones. I also explain other selected phenomena related to r. My purpose is to present a plausible hypothesis to account for the phonetic realizations of modern English r.

2.

The phonetic values of Germanic rhotics

In the majority of Germanic languages – Afrikaans (Lass 1997: 205), Faroese (Barnes and Weyhe 1994), Frisian (Hoekstra and Tiersma 1994), Icelandic (Thráinsson 1994: 147), Norwegian (Askedal 1994: 222), and Swedish (Andersson 1994: 273) – the dominant r-sound is a coronal trill, best represented by [r]. Many of these languages also have the coronal tap [Q] as an allophone of /r/. For many if not all of these languages, in addition to trills and taps, an approximant realization of /r/ is found in syllable codas. Some Germanic languages, notably Danish (Haberland 1994: 320) and German (Eisenberg 1994: 355), have a uvular r instead; depending on dialect, this may be either the uvular trill [R] or the uvular fricative [“]. Some languages, like Dutch, German, Norwegian, Swedish (Chambers and Trudgill 1998: 170–175), and Yiddish (Jacobs, Prince, and van der Auwera 1994: 396) have, depending on dialect, both coronal and uvular realizations of /r/. English stands out for having a different kind of r as its primary

184

Blaine Erickson

phone: either the voiced alveolar frictionless continuant [Ú] or the voiced retroflex approximant [– ] (Ladefoged 1993: 84 et passim; Pullum and Ladusaw 1996: 164–165). Hagiwara (1994) reported a third type of American English r; see §4.1. The question, then, is what realization of /r/ is original in Germanic in general and English in particular? One explanation for the presence of uvular r in those Germanic languages where it is found is that the uvular pronunciation was an innovation in Parisian French, and spread into German by contact (Chambers and Trudgill 1998: 170). From there, the uvular pronunciation diffused into Danish and Dutch, and now, is even found in some Norwegian and Swedish dialects. However, this explanation of the origins of uvular r is not generally accepted by Germanicists, and other explanations, which refute the influence of French, have been proposed by Howell (1987), Runge (1974), and others. Based on the phonological effect of /r/ on adjacent vowels, as well as philological evidence, they argue that German had a uvular r before French did. Regardless of the origins of uvular r, it is considered to be an innovation and not the original realization of /r/. Further evidence suggesting that uvular r is an innovation is found in other Germanic languages. In Norwegian, for example, the uvular pronunciation is spreading at the expense of coronal articulation (Chambers and Trudgill 1998: 170, 173–174). Given the continued spread of uvular r, it seems reasonable to conclude that it is an innovation and not original. Since the r that is declining in the face of this northwestern European change is the coronal trill, and since the coronal trill is found in all varieties of Germanic, even if only dialectally in some, it seems safe to conclude that the original value of Germanic /r/ was a coronal trill. This is further supported by the dominance of coronal trill /r/ in other Indo-European languages (Maddieson 1984). The original value of /r/ in English is also in dispute. Based on its effect on neighboring segments, Lass (1983) has argued that earlier English /r/ had coronal, velar, and pharyngeal components. However, three factors favor the hypothesis that the original value for /r/ in English was a coronal trill: 1.

The original Germanic /r/ was, in all likelihood, a coronal trill.

2.

Some modern English dialects have a coronal trill for /r/ (Wells 1982: 367–370, 410–411).

On the development of English r

3.

185

Rhotacism changed some instances of Proto-Germanic /z/ to English /r/ (Hock 1991: 40–47). This same rhotacism was active in Latin, which is also presumed to have had a coronal trill for its /r/ (Hock 1991: 81–82, 130–131).

In light of these facts, I find no difficulty in taking the value of earlier English /r/ as a coronal trill.

3.

The development of r in English

Going on the assumption that the original pronunciation of /r/ in English was as a coronal trill, the next task is determining how it changed from trill to approximant or retroflex articulation. First, I will outline the relevant changes. Then, I will go into detailed descriptions of the phonological processes active in the sequence of changes, thereby accounting for the modern pronunciation.

3.1 Overview The change from coronal trill to coronal approximant followed these steps: 0.

Origin: coronal trill

1.

Trill lenites to tap before homorganic consonants

2.

Coronals change from dental to alveolar; tongue simultaneously acquires sulcal shape

3.

Tap lenites to retroflex approximant before homorganics

4.

Approximant articulation generalized to syllable coda

5.

Approximant articulation generalized to all positions

A number of other developments, attested in English and other languages, will also be explained.

186

Blaine Erickson

3.2 Step one: lenition of the trill Except in hyperarticulate speech, coronal trills are not found before coronal consonants in syllable codas in the Germanic languages. This suggests that there is some intrinsic articulatory difficulty in coda -rT sequences (where T stands for any coronal). This is not to suggest that such sequences are unpronounceable, but are, for reasons given immediately below (§3.2.1), not favored, and therefore susceptible to the application of phonological processes which overcome the articulatory difficulty (Stampe 1979). Here, I explain this phenomenon and propose phonological processes to account for the substitutions found.

3.2.1 From multiple to single trill The nature of the articulatory difficulty of -rT sequences is as follows. The articulator for a coronal trill is the tongue tip, so if trill [r] is followed by a segment with a different primary articulator, such as the lips or the back of the tongue, then there is no articulatory difficulty and there is no particular effect on the [r]. However, if the trill is followed by another segment that is also articulated by the anterior of the tongue, and if that segment requires occlusion, either full (e.g., [t], [d], and [n]) or partial (e.g., [s], [z], and [l]), then the continuancy of the trill and the closure of the following segment are in conflict. This is the first change: from a regular trill with multiple contacts of the tongue-tip to the roof of the mouth to a single trill with only one contact. A single trill is articulatorily distinct from a tap. “In a trill, the tip of the tongue is set in motion by the current of air. A tap or flap is caused by a single contraction of the muscles so that one articulator is thrown against another” (Ladefoged 1993: 168). Accordingly, for a trill, regardless of how many beats it has, the tongue is relatively loose, but for a tap, the muscles of the tongue are tense.In languages with trills, “often both [trills and taps] appear as allophones of the same phoneme” (Maddieson 1984: 79). The mechanism for this allophony lies in the substitution of a single trill for a multiple trill, and then the substitution of a tap for the single trill. The single trill is the result of a lenitive phonological process; that is, it is articulatorily motivated. However, the substitution of a tap is due to reinterpretation of the single trill as a tap; this is acoustically motivated, because a single trill and a tap are acoustically identical, as both involve the same ac-

On the development of English r

187

tion: the tongue-tip briefly touching the roof of the mouth once.The environment where allophony between trills and taps first occurred is before other coronals; this is formalized in figures 1 and 2. Because there is no IPA symbol for a single trill, I use an asterisk to represent it. [r]→[*]/_ C +cor

Figure 1.

Trill Lenition: From Multiple to Single Trill

The change in figure 1 can be summarized as follows: an apical trill is realized as a single trill before a coronal consonant. I call this process Trill Lenition. It is unclear to me how this should be represented in terms of features, and so I will omit a feature-based analysis here. This is a conditioned change, and is more likely to occur in hypoarticulate speech. As outlined above, the motivation for the change is the conflict between making a continuant coronal trill followed by a partially – or fully – occluded coronal consonant.

3.2.2 From single trill to tap The next change is the reinterpretation of the single trill as a tap; this is shown in figure 2. I call it REINTERPRETATION AS TAP. Unlike the change from multiple to single trill, which was articulatorily motivated, this change was acoustically motivated. [*] > [Q] Figure 2.

Reinterpretation as Tap: From Single Trill to Tap

This is an unconditioned change, made by at least some speakers, and I assume that this reinterpretation was cross-generational: an older generation of speakers had the process represented by figure 1 active in their speech, and a younger generation reinterpreted the older generation’s single trill as a tap. Once variation between trill and tap articulation had been established, which, in essence, conflated the processes shown in figures 1 and 2, it then became possible for the variation to expand to other environments, as the

188

Blaine Erickson

substitution was no longer motivated solely by articulatory factors. This is what appears to have happened in those English dialects, such as those in Scotland and northern England, that have both trills and taps as allophones of /r/ (Wells 1982: 367–370, 410–411). (Some of these dialects also have approximant [Ú] and retroflex [– ]; this is discussed in §3.5 and §4.1.) Trill Lenition and Reinterpretation as Tap (or their conflation) are active processes in other languages with allophony between trill [r] and tap [Q]. For the change of articulation of /r/ from a multiple to a single trill, and from single trill to tap, it is not necessary for the place of articulation to be alveolar; it can just as easily be dental. What is important is that both the rhotic and the following consonant are coronal, and therefore articulated with the same primary articulator: the tongue tip. Accordingly, the processes described above could have occurred at any time in the history of English – or any other language with the appropriate allophony of trills and taps. Note that “nearly every phonological process has an opposite” (Erickson 1998: 24), and the lenition of trill r to a tap is no exception. In Japanese, the trill [r] exists not only as a socially-marked variant of /r/ (Vance 1987: 28), but also as the result of the application of a fortition in emphatic speech. Wells observed that a trilled realization of /r/ in Scots English is “virtually restricted to formal or declamatory-styles” (1982: 411) – i.e., the Scots trill, which is a substitution for [Q], is found as the result of a fortition applying in hyperarticulate speech.

3.3 Step two: from dental to alveolar place Although premodern descriptions of language are often difficult to interpret, we are fortunate to have some remarkably accessible accounts of earlier English available to us. Here, I use the writings of John Hart and William Holder, both referred to in Lass (1997: 79). Writing in the sixteenth century, Hart reported that t and d were made with the tongue touching the teeth; I infer that s, z, r, l, and n were also dentals. In the seventeenth century, Holder reported that t and d were made with the tongue against the gums; “[t]his suggests a dental-to-alveolar shift somewhere between the mid-sixteenth and mid-seventeenth centuries” (Lass 1997: 79). Presumably, this innovated alveolar articulation also held

On the development of English r

189

for s, z, r, l, and n, although in some conservative dialects of Scotland, the apicals are still dental (Wells 1982: 409). The unconditioned change from the dental to the alveolar place of articulation is crucial to the change from trill to approximant manner of articulation for /r/, as will be demonstrated. The change from dental to alveolar for the apical English coronals (which excludes laminal /S/, /Z/, /tS/, and /dZ/, of course) caused a change in the overall shape of the tongue. For dental articulation, the tongue was held relatively flat; however, for apicals at the alveolar point of articulation, the tongue became sulcal, that is, cup-shaped. This sulcal tongue configuration is crucial to further change in the articulation of English r.

3.4 Step three: from tap to retroflex The change from tap to retroflex is motivated by three factors: one is the change of place of articulation, and another is the change in tongue shape, as noted above. The third factor, also mentioned above, is the inherent articulatory difficulty of making multiple consecutive closures at the same place of articulation. There is a natural tendency, seen in languages around the world, to resolve this difficulty through the application of appropriate phonological processes. One option is fortition of the rhotic, as found in Japanese. For example, Old Japanese törite ‘take continuative’ has become modern Japanese totte ‘id.’ by the application of a series of phonological processes which eliminated the high vowel, stopped the tap, and assimilated the tap to the following /t/ (Erickson 1998: 41, 47, ch. 4); the gemination of -rVt- sequences to -tt- is also found in the modern language (Saito 1986). In essence, the first closure – the closure of the tap – is made and held until the release of the second closure, which, for Japanese, results in geminate voiceless stops. Another resolution to this articulatory difficulty is lenition by the substitution of another phone without occlusion, and this is what happened in English, and also, as will be shown, certain other Germanic languages. In this case, the first closure (i.e., that for [Q] or [*]) is not made at all; instead, during the articulation of /r/, the tongue approaches the roof of the mouth without contacting it, and does not contact the roof of the mouth until the following coronal consonant is articulated. An apical trill or tap that approaches but does not contact the roof of the mouth is a retroflex [– ];

190

Blaine Erickson

retroflexion is enhanced by the sulcal tongue body. This change is illustrated in figure 3; I call it Tap Retroflexion. [Q]→[–]/_ C +cor

Figure 3.

Tap Retroflexion: From Tap to Retroflex

This can be stated as follows: tap r is realized as retroflex r before coronals. It could apply to a single trill, but is shown applying to a tap here. To analyze this change in features, it is necessary to determine what features taps have. I believe they are characterized, in part, by the features [–continuant, +sonorant] (Erickson 1998: 79), and in the change from tap to retroflex, [–continuant] becomes [+continuant]. 2 In other words, the non-continuant portion of [Q] – the contact of the tongue-tip to the roof of the mouth – is lost. The retroflexion of this r comes from two factors: the sulcal tongue shape induced by the apical alveolar articulation of English coronals; and the tongue tip’s approach to, but lack of contact with, the alveolar ridge. Although retroflex may be seen as a manner of articulation, Ladefoged (1993: 159) claimed that it is a distinct place of articulation. Accordingly, there must be some set of features that differentiates retroflex from alveolar and palato-alveolar, the adjacent places of articulation in English. I propose that the features [coronal], [high], [anterior], and [sulcal] are relevant here. The three places are contrasted in figure 4 below. (Although the alveolars of English are [+sulcal], this is not necessarily true for other languages, such as French and Japanese.)

Figure 4.

Alveolar

Retroflex

Palato-alveolar

+cor –hi +ant +slcl

+cor –hi –ant +slcl

+cor +hi –ant –slcl

Coronal Places of Articulation in English

On the development of English r

191

Since the tap [Q] and the retroflex [– ] are distinguished by two features, [continuant] and [anterior], Tap Retroflexion must also change the tap’s [+anterior] specification to [–anterior]. This process is shown in features in figure 5 below. Although “a process normally changes only one feature” (Donegan and Stampe 1979: 137), it is clear that some changes, such as the unconditioned substitution of [l] for [n] (or vice-versa), require the simultaneous change of two features, so proposing a two-feature process here is neither unprecedented nor unnatural. +son –cont +ant +slcl Figure 5.

→

+cont /_ –ant

–ant +slcl

Tap Retroflexion in Features

This can be stated as follows: a non-continuant sonorant becomes continuant and non-anterior before a non-anterior sulcal segment. Although the lenition (and even loss) of /r/ could have occurred at any time, the change to retroflex does require alveolar articulation, and therefore, the change to retroflex started sometime between the late sixteenth and mid-seventeenth centuries, when English coronals underwent an unconditioned change from dental to alveolar (Lass 1997: 79). As it turns out, another early scholar of English, Ben Jonson, writing in the mid-seventeenth century, noted that r was “sounded firme in the beginning of the words, and more liquid in the middle, and ends” (quoted in Lass 1997: 286). This is “the first description of a possible stronger onset/weaker coda articulation for /r/” (Roger Lass: personal communication). What exactly Jonson meant by this is open to interpretation; I believe it may be a description of the change to retroflex approximant articulation.

3.4.1 On the feature [sulcal] If I am correct in calling [sulcal] a feature, then it should appear in other languages and other phonological processes. Although it is not my purpose here to give a complete defense of the feature [sulcal], I can provide some evidence to support its existence.

192

Blaine Erickson

Hindi, like many languages of India, has retroflex consonants, distinct from dentals. It may seem more natural for English coronals like t and d to be borrowed into Hindi as dentals; however, they are borrowed as retroflexes (Pandit 1964). Why should this be? I propose that the specification [+sulcal] in English coronals causes Hindi speakers to perceive them as more similar to Hindi retroflexes, which are also [+sulcal], than to Hindi dentals, which are [–sulcal]. (Another explanation, proposed by David Stampe (personal communication), holds that the apical articulation of both English alveolars and Hindi retroflexes accounts for how the English alveolars are borrowed into Hindi.)

3.4.2 Evidence for retroflexion Returning to the change of earlier English r to a retroflex, primary evidence for the mechanism of this change is found in modern Swedish, where the phoneme /r/, normally realized as a trill [r] or a tap [Q], has another articulation before coronals. In Swedish, the sequence of an /r/ followed by a coronal consonant is normally realized as the retroflex equivalent of the coronal, as shown in the examples below (from Andersson 1994: 274). This mutual assimilation also applies in coronal clusters and across word boundaries. (1)

arla [A:ÏA] ‘early’

(2)

barn [bA:Ó] ‘child’

(3)

bort [bo‘:] ‘away’

(4)

hård [ho:ƒ] ‘hard’

(5)

mars [maÛ:] ‘March’

The pronunciation of /r/-coronal consonant sequences as retroflex consonants clearly goes back to sequences of /r/ followed by regular coronals. In fact, Faroese has alternation in its realization of /r/-coronal sequences: either as a rhotic followed by a coronal, as in English, or as a single retroflex consonant, as in Swedish (Barnes and Weyhe 1994: 194–195). The same variation is also found in some varieties of Norwegian (Askedal 1994: 222), and at least one variety of English shows a Swedish-like pattern (Wells 1982: 369–370).

On the development of English r

193

I explain the mutual assimilation of /r/ and coronals to retroflexes as follows: after retroflex [– ], the following coronal assimilates to the place of articulation of the preceding segment. I call this change Retroflexion. Thereafter, the r either vocalizes (see §4.2) or assimilates to the following consonant. The latter two changes are discussed below; first, I examine Retroflexion, shown in figure 6. [+cor] Figure 6.

–ant

→ +slcl

/

–ant +slcl

Retroflexion

Stated in prose, coronal consonants become retroflex after a retroflex segment. Like all phonological processes, this one “operate[s] over natural prosodic constituents” (Donegan and Stampe 1979: 136), so its effect is not limited to the segment adjacent to [– ]. This process need not necessarily operate across syllable boundaries. However, in Swedish, it normally affects all following coronals, tautosyllabic or not, as shown in the example below. (6) barnstuga [bA:ÓÛ‘ıgA] ‘children’s cottage’ (Andersson 1994: 274)

3.4.3 Retroflexion in English Retroflexion may apply in those varieties of English with [– ] as the phone for /r/, as shown in the following examples. It applies in normal, connected, moderately hypoarticulate speech. It affects not only the alveolars, but also the palato-alveolars. However, it does not normally apply to /s/ or /z/ in syllable codas. The data in (7)–(18) are based on my own pronunciation. (7)

bard [bA–ƒ]

(8)

barn [bA–Ó]

(9)

Bart [bA–‘]

(10)

Bardy [bA–—i] ([—] = retroflex flap)

(11)

Barney [bA–~—~~I ]

(12)

party [p}A–—i]

194

Blaine Erickson

and (13)

barge [bA–ƒ˙]

(14)

parch [p}A–‘Û]

(15)

harsh [hA–Û]

(16)

rouge [–u˙]

but (17)

bars [bA–z], not *[bA–˙]

(18)

parse [pA–s], not *[pA–Û]

In example 16, the affected segment is not adjacent to /r/, but is within the same prosodic constituent: the same syllable. I believe that the failure of Retroflexion to apply to /s/ and /z/ (examples 17 and 18) is due to two factors. One is that this process is optional, and therefore speakers have the choice of whether or not they will allow it to apply. The other, more important one, is that [Û] and [˙] are allophones of /S/ and /Z/, respectively, and that speakers choose to maintain the phonemic distinction between the apical and laminal coronal fricatives, rather than merge them, even though /s/ and /S/, and /z/ and /Z/, do not contrast in this position (i.e., after /r/). For many American English speakers, Retroflexion also fails to apply to a following /l/, because other processes affect /l/ in that position, as shown in example 19. However, for some speakers, it does apply to an /l/ that is in the onset of a following syllable, as shown in example 20. Retroflexion of onset /l/ is also found in some varieties of Irish English. (19)

Carl [k}A–‡" ] ~ [k}A–Ê], but not *[k}A–Ï]

(20)

Charlie [tS}A–.li] ~ [tS}A–.Ïi]

3.5 Step four: generalization of approximant articulation to codas The next change in the realization of /r/ was the generalization of the approximant articulation to all syllable codas, and not just before coronals. This is illustrated in figure 7 below.

On the development of English r

195

[Q]→[–]/_C0# Figure 7.

Generalization of Approximant Realization I

The mechanism for this change is analogy. It started with word-final /r/, which was realized as retroflex when followed by a coronal in a following word. Word-final /r/ came to be realized as retroflex even without a following coronal, and after that, all coda instances of /r/ changed to approximant realization, regardless of whether a consonant followed, and when a consonant was present, regardless of its place of articulation. Evidence for this change is found in some English dialects of Scotland, where medial /r/ is realized as a tap, and syllable-coda /r/ is realized as an approximant, either [– ] or [Ú] (Wells 1982: 411). Scots English is conservative in many aspects of its phonology, particularly the consonants (ibid., 408), and I believe this distribution is an example of this tendency. Furthermore, similar variation is found in Fulfulde, in which trill r is realized as an approximant before a consonant (Lindau 1980: 114). When English /r/ was undergoing this change, I believe that the tap, not the trill, was the dominant non-approximant realization. This is in keeping with the overall trend of lenition, and matches the overall pattern in modern Scotland, where the tap is more common than the trill (Wells 1982: 411).

3.6 Step five: generalization of approximant articulation to all positions. The next step in the change of the realization of /r/ was to approximant articulation in all positions. This unconditioned change is shown in figure 8. [Q ] > [– ] Figure 8.

Generalization of Approximant Realization II

I believe this is a cross-generational change: one generation of speakers had a tap in some positions; a following generation had variable realization as tap or approximant; and the next generation had nothing but approximant r. This pattern of implementation across generations for sound change

196

Blaine Erickson

matches that observed by Labov (1994) for many different changes, and also matches the pattern found for the loss of /Œ/ in Japanese (Vance 1987: ch. 9). This could also be the mechanism for other changes I have called cross-generational.

4.

Other developments

In addition to the changes discussed above, other changes have affected /r/ in English. The most salient ones are discussed here.

4.1 Variation in realization: non-retroflex approximants In many varieties of English, /r/ is not realized as the retroflex approximant [– ]. Instead, it is realized as a central approximant [Ú] (Ladefoged 1993: 84). Unlike the retroflex articulation, for which the tongue tip points up, the central approximant is realized with the tongue tip pointed down, and the tongue bunched in the center of the mouth (ibid.). There is a third realization of /r/, in which the tongue tip is neither pointed up nor down. Instead, the tongue tip points forward in a conguration Hagiwara called blade up (1994). Dellatre and Freeman (1968, referred to in Hagiwara 1994) found six different tongue configurations for American English /r/, and two more in British English. However many ways there are to make this sound, they all have one thing in common: a very similar auditory effect (Ladefoged 1993: 84). Here, the mechanism of change from [– ] to [Ú] is purely acoustic: the majority of children learning English learn to make an acceptable r-sound one with the proper “auditory effect,” regardless of tongue position. Whether the tongue tip is up, down, or elsewhere does not matter, just so long as the resultant phone is sufficiently sonorant and rhotic. (Of course, some children fail to master [– ] or [Ú], and there are speakers and dialects that substitute labiopalatals for rhotic /r/. See §4.3.3.) What constitutes rhoticity, however, is still an open question. Two relevant factors for English /r/ are pharyngeal constriction and the lowering of the third formant (Ladefoged 1993: 84, 227; Lindau 1980). As seen by the large number of poss-

On the development of English r

197

ible articulations that pass for /r/ in English, it does not seem to matter much how these are achieved, just so long as they are. Of course, this leads to the question of when this articulatory variation started to appear. I believe that as soon as continuant r started appearing outside the phonologically-conditioned environment – i.e., once [– ] started to appear in codas regardless of whether or not the following consonant was coronal – then other, non-retroflex realizations of /r/ could appear.

4.2 R-Vocalization In Swedish and the non-rhotic varieties of English, another process has applied to coda /r/. This change is R-Vocalization, which changes coda /r/ to a kind of schwa. I believe this change may affect both retroflex [– ] and central approximant [Ú], as well as other variant articulations. It is shown in figure 9. – Ú Figure 9.

→[@]/(V)_(C)#

R-Vocalization

Stated in prose, a rhotic loses its rhoticity in syllable codas. Note that syllabic r may undergo this process as well. This change is more difficult to state in features, because, as noted above (§4.1), there is more than one way to articulate the r in question. Regardless of its articulation, post-vocalic r induces a specfiic acoustic effect on the preceding vowel: it lowers the third formant (Ladefoged 1993: 227–228). As a first approximation, I will use the ad hoc feature [rhotic] here, and therefore one of the changes is from [+rhotic] to [–rhotic]. It may be that the primary change is not the loss of rhoticity but some other feature shared by the rhotics of English, and that rhoticity loss is incidental. However, the exact nature of this feature and of this particular change must await further investigation. Although it loses its rhoticity, the r retains its other features. The resultant sound is often treated as distinct from a regular schwa, and is often written as [&], a turned epsilon (Ladefoged 1993: 82, 84–85; Pullum and

198

Blaine Erickson

Ladusaw 1996: 55–56; Wells 1982: 137, 303–304). I believe that this sound is a sulcal schwa – that is, a schwa articulated with a sulcal tongue shape. Accordingly, the change shown in figure 9 must be modified. This is shown in figure 10 below. [–]→[&]/(V)_(C)# Figure 10. R-Vocalization Revised

Examples of R-Vocalization are shown below. (21)

Received Pronunciation bird [b&:d] (Wells 1982: 202)

(22)

RP term [t}&:m] (Wells 1982: 202)

(23)

RP turn [t}&:n] (Wells 1982: 202)

(24)

Welsh English year [j&:] (Wells 1982: 381)

The vowel I call a sulcal schwa, [&], may lose its sulcality to become a plain schwa [@]; in features, this entails the change from [+sulcal] to [–sulcal]. I call this change Sulcality Loss; it is shown in figure 11 below. It seems likely that [Ú], which is not retroflex, may also change directly to schwa without an intermediary [&] stage. [&]→[@] Figure 11. Sulcality Loss

Below are some examples of the output of Sulcality Loss. (25)

RP beard [bI@d] (Wells 1982: 218)

(26)

RP scarce [skE@s] (Wells 1982: 218)

(27)

RP cure [k}jU@] (Wells 1982: 219)

(28)

Hawaii English/Hawaii Creole English fair [fE@]

(29)

Southern U.S. door [do@] (Wells 1982: 549)

(30)

New York City nearby [nI@bAi7] (Wells 1982: 507)

On the development of English r

(31)

New England shore [SO@] (Wells 1982: 525)

(32)

Welsh English poor [pu:@] (Wells 1982: 381)

(33)

Australian English beer [bI@] (Wells 1982: 600)

199

The schwa that results from R-Vocalization and Sulcality Loss may undergo other changes. One possibility is Lowering (Donegan 1978: 122–131), where schwa changes to [A]; the feature that changes is APITÄLVsimply [–low] to [+low]. Lowering of schwa to [A] has occurred in many varieties of English; examples are shown below. (34)

RP far [fA:] (= [fAA7]) (Wells 1982: 219)

(35)

Welsh English start [stA:t] (Wells 1982: 378)

(36)

HE/HCE here [hi@] ~ [hiA]

(37)

HE/HCE more [moA]

(38)

HE/HCE party [pA:Qi]

(39)

Southern U.S. star [stA:] (Wells 1982: 544)

(40)

South African English part [pA:t] (Wells 1982: 615)

Another possible change is the assimilation of schwa (or [&]) to the preceding vowel. Not only is this attested in many different varieties of English, but it has also occurred in Swedish.3 Examples from English are shown below. (41)

RP more [mO:] (Wells 1982: 213)

(42)

Welsh English north [nO:T] (Wells 1982: 378)

(43)

Southern U.S. lord [lO:d] (Wells 1982: 544)

(44)

Southern U.S. poor [p}ou7] (where the offglide in earlier [poo7] has undergone Raising (Donegan 1978: 139–143, 188; Stampe 1972: 583))

(45)

Australian English cure [k}jO:] ~ [k}ju:] (Wells 1982: 596)

200

Blaine Erickson

4.3 Other effects of retroflex r Retroflex r may have other effects upon adjacent segments, and may itself be affected by other segments. Some of these changes are noted below, but I will forgo full analyses of these phenomena here.

4.3.1 Anticipatory retroflexion In addition to the perseverative process noted above (§§3.4.2–3.4.3), Retroflexion may affect preceding segments as well. In the majority of English dialects, a related process, Anticipatory Retroflexion, obligatorily applies to [s] before /r/, causing this preceding [s] to be realized as [Û]. Since [Û] is an allophone of /S/, this process has the effect of prohibiting */sr/- sequences in English, and only /Sr/- occurs. Accordingly, foreign sris borrowed as /Sr/-, as in Sri Lanka /SrilAŒk@/ [Û–ilAŒk@]. This process is active even for speakers with non-retroflex realizations of /r/, but it may cause retraction of the tongue instead of retroflexion per se. In a few dialects, Anticipatory Retroflexion does not apply, and these dialects allow /sr/- sequences, as in shrimp [sÚImp];4 these dialects must also have a non-retroflex value for /r/. In other dialects, Anticipatory Retroflexion affects not just /s/ but all preceding coronals. In these dialects, stops preceding /r/ may also be affricated. A few examples are shown below. (46)

˜ ƒ˙–u] Andrew [æ ˜ Óƒ˙–u] ~ [æ

(47)

dry [ƒ˙–Ai7]

(48)

his truck [hI˙‘Û–@k]

(49)

street [Û‘–it]

(50)

try [‘Û}–Ai7] (in writing Hawaii Creole English this is )

As seen in examples 46 and 48, this process operates across syllable- and word-boundaries. Anticipatory Retroflexion is obligatory in Hawaii Creole English, is very common in Hawaii English, and is found in many other American dialects.

On the development of English r

201

Even in those dialects in which the application of Anticipatory Retroflexion is limited to onset /s/, it may affect any coronal consonant in the onset of a syllable with syllabic /r/, as shown below. (51)

azure [æ˙–" ]

(52)

dirt [ƒ–" ‘]

(53)

nerd [Ó–" ƒ]

(54)

sure [Û–" ]

(55)

turn [‘–" Ó]

4.3.2 Tapping after interdentals Another effect of retroflexion is the substitution, found after /T/ and /D/, of a tap for the approximant.5 In the articulation of interdental fricatives, the front of the tongue is necessarily raised. If [– ] follows an interdental, then the tongue will be retracted to articulate the [– ]. As the tongue goes back, the tip curls up for the retroflex target. However, because the tongue body is raised, the curling tongue tip strikes the roof of the mouth and produces [Q9], a dental tap. Examples are shown below. (For orthographic convenience, the devoicing of the dental tap after [T] is not shown here.) (56)

throb [TQ9Ab]

(57)

through [TQ9u]

(58)

3-D [TQ9iQi]

(59)

Bathe right now! [bEi7DQ9Ai7t?nã˜u7 ]

This applies across word boundaries and after both /T/ and /D/. This substitution is more likely to occur in hypoarticulate speech. Of course, this substitution is not found for those speakers with non-retroflex realizations of /r/.

4.3.3 Labiopalatality of English /r/ In addition to its rhotic and pharyngeal components, English /r/ has labial and palatal components as well. For example, it is well known that onset r

202

Blaine Erickson

is accompanied by lip rounding; hyperarticulate pronunciation of syllabic r may also have concurrent rounding. Further evidence for labial and palatal components in r is found in language acquisition and language change phenomena. For example, English speakers learning languages with [Ô] and/or [˚] may substitute syllabic r for these foreign phones. Conversely, speakers of languages with mid labiopalatal vowels may substitute them for English /r/. In some dialects of English, syllable-initial /r/ is realized as a labiopalatal glide, and this substitution is also found for some children learning English. Syllabic /r/ is realized as labiopalatal in some dialects, and nonsyllabic /r/ may be realized as a labial, palatal, or labiopalatal glide. Examples are below. (Examples with syllabic r should be understood to represent either [– ] or [Ú], though only the latter is shown.) (60)

L1 English learning Cantonese: [gÚ" k] for Cantonese geuk /k˚k£/ [kÔ’7k£] ‘leg’

(61)

Loan phonology: English [gÚ" t}@] for German Goethe [g˚t}’]

(62)

Loan phonology: Cantonese sëutsàam /s˚t°saam°/ [s’t?°sa:m°] ‘shirt (lit., shirt clothes)’, from English shirt [ÛÚ" ‘] ~ [SwÚ" t] ~ [Sw&t]

(63)

Some English dialects: [Ëili] for really

(64)

Child SLA: [ËEst˚Ë] for rooster (Erickson 1996)

(65)

South African English: [fÔ:st] for first (Wells 1982: 615)

Perhaps it is this complex of so many features that make English rhotics so difficult for both children and second-language learners to acquire. The difficulty of simultaneously articulating these many different elements may also play a role in the lenition, substitution, and loss of r as well.

5.

Conclusion

Based on both internal and comparative evidence, earlier English /r/ was most likely realized as a coronal trill. Due to the action of the phonological processes described here, this trill came to be produced as a retroflex approximant, first before coronals, then more generally in syllable codas, and finally, everywhere. In some varieties of English, the retroflex articulation

On the development of English r

203

has been replaced by other realizations, and in some dialects, coda rhotics have been lost entirely and replaced by vowels. English r has labial, palatal, and pharyngeal elements in its articulation. The retroflex realization of English /r/ has other phonological effects as well, although not all dialects have a retroflex articulation of /r/.

Notes 1.

2. 3.

4. 5.

This paper has benefitted from comments I received at SHEL-1, as well as the comments of Donka Minkova, Robert Stockwell, and an anonymous reviewer. Any errors or deficiencies that remain are my responsibility alone, of course. Lass (1984: 107–111), however, proposed a different analysis of taps, which incorporates a scalar feature called rate. Swedish phonotactics require that all stressed syllables be long, but disallow overlong syllables. Common Scandinavian (C)V:C: shortened to Swedish (C)VC: (Haugen 1982: 240). Since this constraint continues to apply, forms like *CV:C: do not appear in examples 1–5. My thanks to Patricia Donegan for pointing this example out to me. My thanks to Patricia Donegan for pointing this phenomenon out to me.

References Andersson, Erik 1994 Swedish. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 271–312. London: Routledge. Askedal, John Ole 1994 Norwegian. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 219–270. London: Routledge. Baldi, Philip, and Ronald N. Werth (eds.) 1978 Readings in Historical Phonology: Chapters in the Theory of Sound Change. University Park: The Pennsylvania State University Press. Barnes, Michael P. and Eivind Weyhe 1994 Faroese. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 190–218. London: Routledge.

204

Blaine Erickson

Chambers, J.K. and Peter Trudgill 1998 Dialectology, 2nd ed. Cambridge: Cambridge University Press. Donegan, Patricia J. 1985 [1978] On the Natural Phonology of Vowels. New York: Garland. Donegan, Patricia J. and David Stampe 1979 The study of Natural Phonology. In: Daniel A. Dinnsen (ed.), Current Approaches to Phonological Theory. Bloomington: Indiana University Press. Eisenberg, Peter 1994 German. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 349–387. London: Routledge. Erickson, Blaine 1996 Child second language acquisition: Three Japanese brothers learning English. Preliminary report: Phonology. Presentation at Acquisition Group meeting. Honolulu: University of Hawai’i at Mänoa. 1998 The origins and development of Japanese mora nasals. Ph.D. dissertation, Department of Linguistics, University of Hawai’i at Mänoa. Haberland, Hartmut 1994 Danish. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 313–348. London: Routledge. Hagiwara, Robert 1994 Three types of American /r/. UCLA Working Papers in Phonetics (88): 55–62. Haugen, Einar 1982 Scandinavian Language Structures: A Comparative Historical Survey. Minneapolis: University of Minnesota Press. Hock, Hans Henrich 1991 Principles of Historical Linguistics. 2nd revised ed. Berlin: Mouton de Gruyter. Hoekstra, Jarich and Peter Meijes Tiersma 1994 Frisian. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 505–531. London: Routledge. Howell, Robert B. 1987 Tracing the origin of uvular r in the Germanic languages. Folia Linguistica Historica 7 (2): 317–349.

On the development of English r

205

Jacobs, Neil G., Ellen F. Prince and Johan van der Auwera 1994 Yiddish. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Language, 388–419. London: Routledge. König, Ekkehard and Johan van der Auwera (eds.) 1994 The Germanic Languages. London: Routledge. Labov, William 1994 Principles of Linguistic Change: Internal Factors. Oxford: Blackwell. Ladefoged, Peter 1993 A Course In: Phonetics. 3rd ed. Fort Worth: Harcourt Brace Jovanovich. Lass, Roger 1983 Velar /r/ and the history of English. In Davenport, Michael and Hans Friede Nielsen (eds.), Current Topics in English Historical Linguistics, 67–94, Odense: Odense University Press. 1984 Phonology: An Introduction to Basic Concepts. Cambridge: Cambridge University Press. 1997 Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Lindau, Mona 1980 The Story of /r/. UCLA Working Papers in Phonetics (51): 114–119. Maddieson, Ian 1984 Patterns of Sounds. Cambridge: Cambridge University Press. Pandit, P.B. 1964 Indian readjustments in the English consonant system. Indian Linguistics (25): 202–205. Pullum, Geoffrey K. and William A. Ladusaw 1996 Phonetic Symbol Guide. 2nd ed. Chicago: University of Chicago Press. Runge, Richard M. 1974 Proto-Germanic / r / : The Pronunciation of / r / Throughout the History of the Germanic Languages. Göppingen: Alfred Kümmerle. Saito, Sumio l986 Hanashikotoba ni okeru ragyöon oyobi nagyöon no möra onsoka (The moraic phonemicization of r- and n-sounds in the spoken language). Nihongo Kyöiku (60): 205–220.

206

Blaine Erickson

Stampe, David 1972 On the natural history of diphthongs. In: Paul M. Peranteau, Judith N. Levi and Gloria C. Phares (eds.), Papers from the Eighth Regional Meeting, 578–590. Chicago: Chicago Linguistic Society. 1979 A Dissertation on Natural Phonology. New York: Garland. [1973] Thráinsson, Höskuldur 1994 Icelandic. In: Ekkehard König and Johan van der Auwera (eds.), The Germanic Languages, 142–189. London: Routledge. Vance, Timothy J. 1987 An Introduction to Japanese Phonology. Albany: State University of New York Press. Wells, J.C. 1982 Accents of English. 3 vols. Cambridge: Cambridge University Press.

Vowel variation in English rhyme Kristin Hanson

“As with other aspects of the sound of a poem, rhymed and unrhymed are not only matters of degree, infinitely varied; they also vary, expressively, in the context of all the other aspects of the poem.” Robert Pinsky, The Sounds of Poetry, p. 95

1.

Introduction

One of the themes of this conference is the question of what new avenues of research and teaching present themselves at this start of a new millennium for studies in the history of the English language. 1 In this paper, I suggest that one such avenue lies in the study of the patterns of vowel variation in English rhyme throughout its history, as a problem not of historical phonology, but of poetic form. The issue has also been raised recently by Malone (1993), as it was, curiously, just at the turn of the last century, by Newcomer (1899a, b) and Ritchie (1901). It is a well known fact that in many poems from earlier periods in which rhymes normally have identical vowels, there also occur some rhymes in which for most 20th century English speakers, the vowels differ. For example, in the sonnet by Shakespeare (1564–1616) in (1), alongside the rhymes of minds/finds, mark/bark, cheeks/weeks and shaken/taken, there are the notorious rhymes of love/(re)move, come/doom, and proved/loved: (1) Let me not to the marriage of true minds Admit impediments. Love is not love Which alters when it alteration finds, 4 Or bends with the remover to remove. O no, it is an ever-fixèd mark That looks on tempests and is never shaken;

208 8

12

Kristin Hanson

It is the star to every wand’ring bark, Whose worth’s unknown, although his height be taken. Love’s not time’s fool, though rosy lips and cheeks Within his bending sickle’s compass come. Love alters not with his brief hours and weeks, But bears it out ev’n to the edge of doom. If this be error and upon me proved, I never writ, nor no man ever loved. Shakespeare, Sonnets 116

Similarly, in the passage from Pope’s (1688–1744) “The Rape of the Lock” in (2), the succession of rhymes with identical vowels gives way at the end to a rhyme with different vowels in give/(be)lieve: (2) Fairest of Mortals, thou distinguish’d Care Of thousand bright Inhabitants of Air! If e’er one Vision touch’d thy infant Thought, Of all the Nurse and all the Priest have taught, Of airy Elves by Moonlight Shadows seen, The silver Token, and the circled Green, Or Virgins visited by Angel-Pow’rs, With Golden Crowns and Wreaths of heav’nly Flow’rs, Hear and believe! thy own Importance know, Nor bound thy narrow Views to Things below. Some secret Truths from Learned Pride conceal’d, To Maids alone and Children are reveal’d: What tho’ no Credit doubting Wits may give? The Fair and Innocent shall still believe. Pope, “The Rape of the Lock” I 27–40 Especially for Shakespeare and Pope, such differences between rhymed vowels have been scrutinized time and again by philologists seeking to ascertain the actual phonology of these poets and their times – Ellis (1869–89), Gabrielson (1909), Wyld (1923), Kökeritz (1953), Dobson (1968) and Cercignani (1981), to name only some. Although these scholars differ in their conclusions, and consequently in their views as to how many and which of these vowel differences remain facts of these texts in their historical lin-

Vowel variation in English rhyme

209

guistic contexts, all concede that some vowel differences do remain. Most discussions of these differences seem to share two assumptions: first, that they are departures from the form, rather than an intrinsic part of it; and second, that the language with reference to which they can be explained is a broad social phenomenon, extending beyond the poet’s own idiolect to encompass literary predecessors’ practices, pronunciations of speakers of other dialects within the poet’s community, and so forth. What I would like to suggest here is that literary considerations invite reconsideration of both assumptions, and point to the interest of a different question: For a given poet, can rhyme be formalized in a way which incorporates variation in the vowels entirely as a matter of the poet’s individual linguistic competence? In section 2 I discuss some rhetoric about rhyme which has deflected interest from this question, confining even generative studies of rhyme largely to modern poets and other issues (e.g. Zwicky 1976, Holtman 1994, Hanson 1998 and others). In section 3 I turn to the specific vowel differences found in the rhymes of Shakespeare’s Sonnets, and the discussion of them in Kökeritz (1953); and suggest that an alternative explanation for these particular rhymes’ viability could be the presence in Shakespeare’s own grammar (as in ours) of vowel alternations resulting from historical changes, an idea which Brogan’s (1981) summary suggests may also have been put forth in Stryjewski (1940). Finally in section 4 I briefly note some ways in which the vowel differences found in Pope’s rhymes in “The Rape of the Lock” diverge from those in Shakespeare, in order to show that the historical issues raised by variation in rhyme are not only synchronic ones about literary language of the past, but also true diachronic ones of how poetic forms are transmitted within a tradition.

2.

The rhetoric of partial rhymes

2.1 Defining “partial rhyme” The array of conflicting terminology used in relation to rhyme necessitates a comment on how terms are used here. Loosely speaking, rhyme involves the occurrence for poetic purposes of the same sounds at the ends of discrete stretches of language. Describing it more formally involves specify-

210

Kristin Hanson

ing where those stretches of language are located in a text, which parts of them are repeated, and how complete the repetition is. In English, the canonical form of rhyme as a structural device in poetry – indeed its definition for many – involves the syllables in the final strong metrical position of discrete lines, and requires the nuclei of those syllables and all segments following to the end of the line to be identical. The issue of location, interesting though it is across the tradition, will not be of concern here; all the rhymes considered are end-rhymes in the canonical location, as in (1) and (2). What is trickier is the interrelationship of the other two issues, for the fundamental subject here is whether and if so how the nuclei of those syllables may be implicated in some relationship of repetition, and yet not be identical. The necessary distinction is carefully expressed by Rickert (1978), who reserves the term “rhyme” for the form in which requirements of repetition are imposed on the nuclei of rhyming syllables as well as what follows, as opposed to “slant rhyme”, in which requirements of repetition are imposed on the final consonant only. In either form, these requirements may be met through mere similarity rather than complete identity, so that either a rhyme or a slant rhyme can be termed “partial” as opposed to “full”. The subject here, therefore, is partial rhymes in which the partiality inheres in the nucleus.

2.2 The historical reality of partial rhymes Abetted by the use of the terms “perfect” and “imperfect” as opposed to the more neutral “partial” and “full”, the dominant pedagogical rhetoric accords partial rhymes no real status as a poetic technique in cases where phonological change makes the poet’s own language at all opaque. In poetry handbooks, when vowel variation in rhyme is acknowledged as a matter of craft rather than a lapse, it tends to be treated as a 20th century innovation, with examples suggesting that what is being referred to is the development of slant rhyme as a form in its own right: Until recently almost all English writers of serious poems have limited themselves to perfect rhymes, except for an occasional poetic license such as eye rhymes ... Many modern poets, however, deliberately supplement perfect rhyme with imperfect rhyme (Abrams 1999: 274).

Vowel variation in English rhyme

211

For poetry from earlier periods, in contrast, vowel differences in rhyme tend to be explained away as only apparent, artefacts of the historical development of the language having rendered what would have been full rhymes in the poet’s own language only partial rhymes in ours: A rhyme is stronger when the final words Seem less alike than pairs of mated birds. ... or when a common word like “love” Which rhymed in Shakespeare’s time with “move” and “prove” Ends up today a sight-rhyme, as above (Hollander 1981: 14).

In linguistics textbooks, vowel differences in rhyme tend to be presented as fairly straightforward evidence for how differently some words must have been pronounced at earlier stages in the language: If we go back to the eighteenth century we find Alexander Pope writing Good-nature and good-sense must ever join; To err is human, to forgive, divine ... where it is apparent that he pronounced join with the centralized diphthong [@I]. Again he writes Here thou, great Anna! whom three realms obey, Dost sometimes counsel take – and sometimes Tea. It is demonstrable that he pronounced tea as tay. Elsewhere he rhymes full – rule; give – believe; glass – place; ear – repair; lost – boast; thought – fault; obliged – besieged; reserve – starve. Since Pope’s time the pronunciation of at least one in each of these pairs has changed so that they are no longer considered good rhymes (Baugh and Cable 1993: 17).

But the relationship between language and verse forms is not so straightforward. In meter, for example, although the stressed syllables of polysyllabic words are generally excluded from weak metrical positions except after major breaks, it would be incorrect to conclude from lines like Milton’s (3a) that image retained the possibility of French final stress into the 17th century. Rather, (3a) reflects a variation characteristic of Milton’s metrical practice also found in (3b–c) and other lines which is entirely systematic and explictly describable as part of his metrical form (Bridges 1921, Kiparsky 1977):

212

Kristin Hanson

(3) a. Created thee, in the Image of God wsw s ws w s w s b. To the Garden of bliss, thy seat prepar’d ws w s w s w s w s c. In the Visions of God: It was a Hill w s w s w s w s w s

(Paradise Lost 7, 527)

(Paradise Lost 8, 299)

(Paradise Lost 11, 377)

And in detailed philological studies, which take a broader range of evidence than verse form into account precisely because of awareness of the special hazards the latter poses, many apparent partial rhymes in older poets’ practice are not in fact explained away as historical illusions. It is true that many of the phonological characteristics with respect to which vowels in rhymes differ are the same ones with respect to which change took place. In rhymes like love(d)/(re)move/prove(d) or come/doom equating [ˆ] and [u], for example, it is certainly true that in some contexts [u]’s shortened to [U]’s and [U]’s lowered and centered to [ˆ]’s, producing relationships like that of the [u] of do to the [ˆ] of done, to which we return below. But neither the dates of such changes nor the specific words involved in them add up to a system in which all rhymes seem to have actually been full rhymes for any poet from Early Modern English on. Kökeritz (1953), for example, argues that for Shakespeare, a surprising number of rhymes which are partial for us must have already been so for him. Kökeritz is unusually radical in his early dating of many vowel changes; Dobson (1968) dates many of the same changes much later. But that difference does not always undermine the historical reality of the individual partial rhymes found in Shakespeare, and even if it did, it would not eliminate the more general poetic problem. For if Kökeritz’s dating is too early, it only means that the point at which vowel variation becomes characteristic of English rhymes is later, and the poets in whom it merits study are different: the kinds of partial rhymes considered here continue in the practice of not just Pope, over a century after Shakespeare, but also, as Malone (1993) notes, in Blake (1757–1827) and Byron (1788–1824), nearly another century after Pope, and right on up to contemporary poets.

Vowel variation in English rhyme

213

2.3 The expressive value of partial rhymes At the same time that historical arguments suggest that the partiality of some rhymes may be real rather than only apparent, critical considerations suggest that such partiality may often – though as with any variation in poetic form, not always – have expressive significance. For example, Wright (1988: 85) suggests that in sonnet 116 in (1), the grand claims for love’s constancy are gently undermined by the metrical structure: The metrical irony of “brief” is inescapable; “brief hours” takes much longer to say than “and weeks” and seems to describe longer moments of time. The line in which the poet says “Love alters not” is full of metrical accelerandos and retards.

The partial rhymes can be understood as contributing similar effects. In line 4, just as true love’s unremovability is asserted, the expected full rhyme for love fails to appear. In line 12, not only does the thwarted expectation of full rhyme create uneasiness, but more specifically the increase in the length of the vowel of doom over that of its rhyming partner come emphasizes the length of time which true love is claimed to endure – “ev’n to the edge of doom”, which as Wright (1988: 86) notes is hardly a lyrical description. And in the final couplet in lines 13 and 14, the partial rhyme’s compromising of closure is consistent with Wright’s (1988: 86) suggestion that the “extravagant and illogical claim” of the couplet may ultimately “cast doubt upon the reasonableness of its proponent”. Similarly, in the passage from “The Rape of the Lock” in (2), the partiality of the rhyme in lines 39 and 40 contrasts sharply with the effect more commonly attributed to Pope’s couplets in which the apparent inevitable perfection of the rhymes lends a sense of incontrovertibility to assertions of universal truths (Wimsatt 1954). In these lines, the partiality of the rhyme suggests instead the incompatibilty of the beliefs ascribed to Belinda and her world with those held by those of a more philosophical bent. And it does so in a way subtly compatible with critics’ claims that part of the poem’s charm derives from the understanding of Belinda which tempers the satire (Tillotson et al. 1969: 567), in literally giving her beliefs not just the last word in the rhyme pair, but greater weight, through that word’s longer vowel. In his classic study of rhyme based especially on “The Rape of the Lock”, Wimsatt (1954) argues that it is the interplay of rhyme words’ phonological similarity with their semantic, syntactic and morphological

214

Kristin Hanson

differences which is crucial to rhyme’s creation of meaning. If the partiality of rhymes like those in (1) and (2) is significant, and part of the poets’ compositions rather than distortion caused by the lenses through which 20th century readers see these rhymes, then phonological differences too may play a role in creating such effects.

2.4 A note on the history of the rhetoric of rhymes The rhetoric of Early Modern poets themselves may provide further indirect support for the historical reality of partial rhyme as a poetic technique. According to Newcomer (1899a), the rhymes of Chaucer tend to be full rhymes which like the French rhymes they are modeled on derive their aesthetic interest from differences and similarities in the stretch of language which precedes the nucleus. Beginning with Spenser (though see also Oras 1955), the locus of aesthetic interest begins to narrow to the nucleus itself and the language which follows it, and the use of similar rather than identical sounds there is introduced. At this same time, as Lanz (1931: 145–199) documents, Puttenham and Sidney characterize rhyme as adding elements of melody and harmony to poetry, which depend upon the inherent qualities of individual vowel sounds. But later prosodists like Guest and Saintsbury take an interest in rhyme primarily as a rhythmic device, focusing on its role in demarcating poetic units like the line. In relation to this latter role, the specific quality of a rhyme may be less important than the fact of it, with the result that interest in the former issue is diminished in the very period when interest in the phonological history of English is developing, and the vowel differences in rhymes are first being studied systematically, whence perhaps the dominance of assumptions like Wyld’s (1923) that a “good” rhyme is a full rhyme. In an interesting twist, however, such assumptions may not have prevailed in American scholarly discourse on the subject as readily as they did in British. Of course the poetic context itself is different, given the presence in the American tradition of the innovative rhyme practices of Emerson and Dickinson. But the interest of Newcomer (1899a, 1899b), for example, in the subject is also clearly rooted in the question of whether Americans can have transparent access to the British poetic tradition, or whether “the Londoner’s poetry must be kept sacred to London ears”; and forms part of

Vowel variation in English rhyme

215

larger debates about linguistic diversity and correctness (Matthews 1898, Finegan 1980). Thus, the rhetoric surrounding partial rhyme seems a matter of cultural history deserving of research in its own right.

3.

Describing and explaining partial rhymes

3.1 Vowel variation in the rhymes of Shakespeare’s sonnets If it is true that rhyme allows for expressive partiality in its formal realization, then a question presents itself, analogous to that Halle and Keyser (1966) first posed for meter, and Zwicky (1976) pursued for the rhymes of rock lyrics. If identity of the vowels is not required for rhyme, what is? If similarity suffices, is it possible to define it in such a way as to distinguish variations which are actually used by a given poet from ones that would not be, as opposed to simply acknowledging the possibility of “poetic license”? If so, is it possible to express that as a grammatical principle, as opposed to a lexical list of poetic conventions? I do not have the expertise in the phonological history of English to answer this question here. Even for Shakespeare, for whom we already have Kökeritz’s (1953) painstakingly detailed study of Shakespeare’s own pronunciation, as well as subsequent studies such as Dobson (1968) or Cercignani (1981), the facts are too intricate and the conclusions too tentative and controversial for me to summarize here, let alone evaluate disputed points and argue for a grammatical principle based on the results. But it is nonetheless instructive to consider Shakespeare’s actual rhymes more fully, not in order to assess any specific claims about which differences in vowels would have been phonologically real for Shakespeare, nor any methods for ascertaining that, but rather in order to illustrate the kinds of explanations which have been offered for differences in vowels when they are concluded to be historically real, and to clarify how an answer to the question posed here would be subtly different. For this purpose, we turn to the rhymes of Shakespeare’s Sonnets and Kökeritz’s description of the vowels in them. Kökeritz (1953) does conclude that some of Shakespeare’s apparent partial rhymes would probably have been full in Shakespeare’s dialect: these include wind (n.) and find in sonnets 14 and 51 (218), deserts and parts in

216

Kristin Hanson

sonnet 17 (169, 250), or tongue and song in sonnet 17 (243). Some other rhymes are so complicated that I will set them aside for now: these include ear and bear in sonnet 8, war and bar in sonnet 46, and none and stone in sonnet 94. But in (4) are underlined some examples of rhymes in Shakespeare’s Sonnets that Kökeritz concludes with varying degrees of certainty would likely have involved actual phonetic differences in the rhyming vowels’ qualities in at least one register of Shakespeare’s own dialect. In each case the entire quatrain is given, in order to suggest something of the potential expressiveness of the difference; the phonetic notation is Kökeritz’s own: 2 (4) a. [@I] ~ [i:]: From fairest creatures we desire increase, That thereby beauty’s rose might never die, But as the riper should by time decease His tender heir might bear his memory: b. [E] ~ [i:]: Farewell, thou art too dear for my possessing, And like enough thou know’st thy estimate. The charter of thy worth gives thee releasing; My bonds in thee are all determinate.

(Sonnets 1, 1–4)

(Sonnets 87, 1–4)

c. [æ:] ~ [e:]: When to the sessions of sweet silent thought I summon up remembrance of things past, I sigh the lack of many a thing I sought, And with old woes new wail my dear time’s waste. (Sonnets 30, 1–4) d. [@u] ~ [o:] Time doth transfix the flourish set on youth, And delves the parallels in beauty’s brow, Feeds on the rarities of nature’s truth, And nothing stands but for his scythe to mow. (Sonnets 60, 9–12)

Vowel variation in English rhyme

217

e. [o:] ~ [¡]: When I consider everything that grows Holds in perfection but a little moment, That this huge stage presenteth naught but shows Whereon the stars in secret influence comment; (Sonnets 15, 1–4) f. [ˆ] ~ [u:]: Who will believe my verse in time to come If it were filled with your most high deserts? Though yet heav’n knows it is but as a tomb Which hides your life, and shows not half your parts. (Sonnets 17, 1–4)

3.2 Kökeritz’s explanations Kökeritz invokes a variety of overlapping explanations for such differences in rhymes’ vowels, including the availability of “doublets”, or variant pronunciations also current in Shakespeare’s linguistic milieu; the existence of literary precedents which make the rhymes “traditional”, often equated with their being “eye rhymes”; and the possibility of phonetic similarity rather than identity sufficing to justify a rhyme. For example, the rhyme in (4a) Kökeritz attributes to the existence of doublets. The vowel in die, he says, was already diphthongized at least to [@I] (1953: 216); for the final y in memory, however, two pronunciations must have been in use, an archaic and conservative [@I] as well as a more colloquial [i:] (219–220). Similarly, he suggests that although the vowel of releasing in (4b) could have been [i:] colloquially, it would have been [e:] in polite speech, and might have fluctuated in length as well, in these forms thus becoming a perfect rhyme with possessing, which would have had its current vowel [E] (194–209). Waste in (4c) may have had a conservative variant pronunciation with [æ:] alongside Shakespeare’s own more common, colloquial pronunciations with [æ:]; here, however, Kökeritz suggests that it is more likely that any such former pronunciation would have disappeared by Shakespeare’s time, leaving only variant spellings of these words without the final e, such that these “may already have become eye rhymes” (1953: 176; see also 34).

218

Kristin Hanson

The rhyme in (4d), used likewise by Shakespeare’s contemporaries but “phonologically ... difficult to explain satisfactorily” is “doubtless traditional” and “consequently phonetically inexact” (245). As for the rhyme in (4e), “we cannot dismiss the possibility ... of imperfect linking of [o:] and [¡]” (233): although for some words involving these vowels, such as both, there is evidence of variation between [o:] and [¡] in Shakespeare’s day, and for others such as boast, which is rhymed with cost, there is evidence that the long vowel might have had at one time a shortened variant which though no longer in use by Shakespeare’s day might have left a legacy of a traditional rhyme with a word containing a short vowel such as cost; for others such as those in (4e), while a parallel story might be possible, there is no direct evidence of one. Clear evidence that some phonetic inexactness is sometimes possible comes from the fact that “occasionally a long vowel is linked with its short equivalent as in dream:then, etc.” (34). Finally, the rhyme in (4f) of come and tomb returns to the sonnet in (1) and the rhymes of come/doom or love(d)/(re)move/prove(d). Kökeritz thinks it most likely that Shakespeare’s pronunciations of all these words were basically the modern ones: though “the existence of phonetic doublets must always be reckoned with” and “colloquial usage at that time vacillated between [u:] and its shortened products [U] and [ˆ]” (237), it seems most likely that “the rhymes in question were already traditional” (238) and examples of “eye rhyme” (31). There is a curious heterogeneity to these explanations. Given Kökeritz’s aim of ascertaining what Shakespeare’s pronunciation would have been, the heterogeneity makes sense: it arises from considering every possibility which could affect the validity of the evidence from rhyme. But from the point of view of trying to ascertain what Shakespeare’s principles for rhyming might have been, the variety of explanations suggests that what goes on in the poet’s mind is a kind of eclectic and post hoc justification of useful rhymes, rather than a tacit exploration of the properties of his artistic medium – his language – to develop a coherent, if also ever-changing, poetic form. Moreover, locating that justification not only in phonetic properties but also in spelling, others’ dialects, and others’ poetic practices defines that language as the linguistic practice the poet sees and hears around him, rather than a coherent linguistic system within his own mind. It thus reflects precisely the conception of language to which Chomsky’s (1959) generative proposal offered an alternative. While external linguistic prac-

Vowel variation in English rhyme

219

tice is obviously relevant to whatever the poet’s internal linguistic system is, and is in other ways especially important for a dramatist, as a conceptualization of the medium of the Sonnets it raises as many questions as it answers. What, for example, does it really mean to say that variant pronunciations in the linguistic milieu explain a rhyme? Is it plausible that variants would be freely interchangeable without any social or regional significance being carried by the choice? Kökeritz claims that it would, noting that the existence of many doublets in 16th century London was “probably less noticeable than would be today a vacillation between, say, [i:D@] and [aiD@] for either” (1953: 8). But even without specific associations, is it plausible that all variants could have been so equally accessible to the poet and his readers that a rhyme context could simply call up whichever variant would make a full rhyme, without any sense of correcting for expectations that are not met? Even the use of the rarer pronunciation of either can have destabilizing effects on the rhymes of poems today, given the right semantic context. Such a context is certainly provided in, say, the rhyme in (4a), in which both the stress on the final syllable that the meter causes to be expected, and also the quality of the vowel that the rhyme causes to be expected, never come to be quite fully embodied in the language, but hover ghostlike, precisely like the memory they denote. Finally, wouldn’t the use of variants in general be at odds with the construction of a lyric voice? Whatever freedom a single speaker might have had in choosing a pronunciation for a given word, it seems likely that as with spelling variants in the period, there would have been consistent characteristic practices, such that variation between forms would have compromised the coherent sense of a single speaking subject. And what exactly is an eye rhyme? The idea is summarized by Kökeritz (1953: 31) as follows: … already in the 16th century, and probably earlier, there had apparently developed a rhyming tradition that was to some extent independent of the contemporary pronunciation. Each new generation of poets preferred to use more or less the same rhymes as the preceding one and continued to do so long after some of the syllables they coupled in rhyme had ceased to be pronounced alike; in such cases the conventional spelling preserved an illusory, purely graphic or visual identity which had no counterpart in the pronunciation of the rhyming syllables. This is the genesis of the eye rhyme ...

220

Kristin Hanson

Thus, an eye rhyme is just a traditional rhyme in which the past identity on which the rhyme is based is preserved in the spelling. The spelling, however, seems to in fact be of very little consequence; it is certainly not a creative principle of rhyme in its own right in the Sonnets. Although Shakespeare would in any case have had little or no control over the spellings in any printed text of his work, the spellings of the apparent partial rhymes in the sonnets discussed above in the 1609 Quarto (Booth 1977) are nonetheless worth considering: (5) a. die : memory b. paSt : waSte c. moment : comment d. come : tombe e. come : doome f. loue : (re)moue g. proued : loued h. brow: mow Although all of (5b and d–h) are called “eye rhymes”, only (5f–h) actually have identical spellings from the first vowel on. Indeed, Kökeritz (1953: 33) says explicitly that while Shakespeare undoubtedly used eye rhymes, he did not have what Wrenn calls “traditional spelling rhymes,” that is, rhymes which were, “in the Early Modern period, based simply on spelling and were therefore linguistically incorrect.”

That is, identity of spelling is neither necessary nor sufficient even for those rhymes identified as eye rhymes; “eye rhyme” seems not to be a basic principle at all, but at best a supporting characteristic of some traditional rhymes. What then exactly are traditional rhymes, the rhymes described above as “more or less the same rhymes” as a preceding generation used? Are they pairs of words? Certainly that rhyme pairs in one poet’s practice can be repeated in another’s as a literary allusion seems true enough: love and prove will always rhyme after Shakespeare’s Sonnets. But is it plausible that poets also remember each others’ practice as word pairs when the semantic relationship is not so developed and meaningful? And if traditional rhymes do consist in word pairs, how are the infrequent but nonetheless real inno-

Vowel variation in English rhyme

221

vations in traditional rhymes to be accounted for, such as Shakespeare’s rhyming not only come with tomb, as did Spenser (Gabrielson 1909: 92), but also, for example, dumb in sonnet 83? At some point, the traditional rhyme must be construed as a relationship not between words, but between vowels, and hence a matter not of discrete lexical facts but of grammar. The one explanation which Kökeritz actually bases on the characteristics of the vowels themselves is that of phonetic similarity. But in what, exactly, does the relevant phonetic similarity consist? The classic study of the sufficiency of similarity in rhyme is Zwicky’s (1976) on rock rhyme, which finds that the allowable vowel differences in that genre, summarized in (6), can be characterized as differences in only one feature: (6) I-E i-e E-æ E-e

ˆ- O ˆ- A

u-o A- O

But Shakespeare’s rhymes do not seem susceptible of the same explanation: the rhyme in (4e), for example, which Kökeritz attributes to phonetic similarity, involves differences in both tenseness (or length) and height. Thus, while all the explanations are plausible in themselves, they also have limitations; individually they fail to generalize, and together they don’t form a coherent system.

3.3 Toward a grammatical explanation Is there any coherent system to be found? I would like to suggest that the poetry handbooks’ and linguistics textbooks’ claims that the history of English vowels is the explanation for their variation may be correct, but in a somewhat more subtle way than they imply. That history is what unites the relevance of dialect variants, traditional rhymes and in some cases their spelling, and phonetic similarity which transcends differences especially in length and height. But that history is not an extrinsic fact about the language which only a philologist knows; it is intrinsic to the language, inscribed in vowel alternations such as the following, available to any poet and reader regardless of their education:

222

Kristin Hanson

(7) a. b. c. d. e. f.

Ai-I i-E

e-æ Au-ˆ u-ˆ o-¡

wise ~ wisdom keep ~ kept vain ~ vanity profound ~ profundity produce ~ production cone ~ conic

There is a striking overlap between the pairs of vowels in the rhymes in (4) and in these alternations. 3 My suggestion, therefore, is that Shakespeare’s rhyme might incorporate the principle that vowels which alternate in the grammar in some contexts may be treated alike in rhyme, a principle similar to one I have proposed elsewhere for consonant variation in the very different partial slant rhyme practice of Robert Pinsky (Hanson 1998). This hypothesis does not offer an obviously better empirical account of Shakespeare’s rhyming practice. Some rhymes, like that in (4d), would not be predicted by it; at the same time, some rhymes like ones linking [@u] and [ˆ] (see (7d)) would be predicted, but don’t seem to occur; and many cases have not even been discussed here (see paragraph 3 of section 3.1). But perhaps it should not be construed sharply as an alternative to the kinds of explanations which Kökeritz adduces, so much as a principle which might coexist with them, emerging under their influence to shape the poet’s core practice, yet leaving a few peripheral rhymes unassimilated to it. As Newcomer (1899a) puts it for the tradition more generally, It may be admitted that certain rhymes, once good, have now and then, after a change of pronunciation, continued to lead a conventional existence for a time on the supposed suffrage of earlier poets. But that will not account for the steady growth in favor of certain rhymes, nor for the continuation of some to the exclusion of others. If blood and mood and have and grave live on, why have take and speak, desert and part, find and joined also at certain times or in certain localities perfect rhymes, been abandoned? There is but one answer: certain dissonances have approved themselves to the poet’s ear, while others have not.

What makes the hypothesis advanced here worth developing and exploring is more its literary value. It opens up the possibility of recognizing the expressive variation in the sonnets’ rhymes from a defensible philological foundation, and doing so within a strict definition of rhyme as a set of linguistic constraints which is individual rather than social, and creative rather

Vowel variation in English rhyme

223

than imitative, qualities more compatible with the lyricism of the Sonnets, and also with generative conceptions of language.

4.

Transmission of partial rhymes

Whatever the success of a hypothesis like the foregoing might have in accounting for Shakespeare’s practice, it seems unlikely that it can be generalized to the entire tradition; rather, every poet’s system of rhyme is their own. It has already been mentioned in connection with (6) that the partial rhymes occurring in Shakespeare are not the same as those in English rock lyrics; but more to the point, they are not even the same as in Pope. Pope’s “The Rape of the Lock” does include rhymes pairing many of the vowels in (4): (8) a. Belinda now, whom Thirst of Fame invites, Burns to encounter two adventurous Knights, At Ombre singly to decide their Doom; And swells her Breast with Conquests yet to come. (“The Rape of the Lock” III.25–28) b. Methinks already I your Tears survey, Already hear the horrid things they say, Already see you a degraded Toast, And all your Honour in a Whisper lost! (“The Rape of the Lock” IV 107–110) c. The busy Sylphs surround their darling Care; These set the Head, and those divide the Hair, Some fold the Sleeve, while others plait the Gown; And Betty’s prais’d for Labours not her own. (“The Rape of the Lock” I, 145–8) But it also shows subtle innovations: (9) a. And now, unveil’d, the Toilet stands display’d, Each Silver Vase in mystic Order laid.

224

Kristin Hanson

First, rob’d in White, the Nymph intent adores With Head uncover’d, the Cosmetic Pow’rs. (“The Rape of the Lock” I.121–4) b. Twas then Belinda! if Report say true, Thy eyes first open’d on a Billet-doux,... Here Files of Pins extend their shining Rows, Puffs, Powders, Patches, Bibles, Billet-doux. (“The Rape of the Lock” I.117–8; 137–8) c. He said; when Shock, who thought she slept too long, Leapt up, and wak’d his Mistress with his Tongue. (“The Rape of the Lock” I.115–6) Clearly the words involved are sometimes new: in (9a), for example, the vowel pairing of (4d) is extended to words not paired in Shakespeare. In (9b) the vowels differ in height without the concomitant difference in length characteristic of the pairings in (4) and (7); Kökeritz lists one traditional rhyme of these vowels in Shakespeare ((re)move:Jove (1953: 237)) but it does not appear to be representative of any systematic practice. The rhyme in (9c) is a pairing which does occur regularly in Shakespeare, but which Kökeritz argues may have been a full rhyme for him. If it is no longer one for Pope, it represents the establishment of a new traditional rhyme, and one which because it also involves no length difference, would elude any principle that vowel differences in rhyme parallel vowel alternations in the grammar as in (7). If Pope’s rhyming practice is to be accounted for by a grammatical principle rather than the kinds of explanations discussed in section 3.3, then, it will have to be a principle at least slightly different from any accounting for Shakespeare’s. In fact, when Ritchie (1901) poses just the question we have urged here of whether “the license poets allow themselves is governed and limited by some sort of scientific principle,” he describes phonological similarity in and of itself as playing a much more significant role for Pope than I have suggested here it does for Shakespeare. As with Shakespeare, however, the facts are far from settled. The value of the example in (9b) as phonological evidence of Pope’s rhyme practice is complicated by the fact that, as pointed out by Katy Breen (p.c.) in a class discussion, it seems to be a joke about the dubious relationship of English

Vowel variation in English rhyme

225

society to French pronunciation, possibly also relevant to the famous rhymes involving tea, such as that with away (I.60–61). Similarly, an incentive not to take the rhyme of (9c) as full comes from the very fact that as also pointed out by Breen (p.c.), its partiality adds an amusing phonological shock to the social one of Shock’s tongue providing the conclusion to Belinda’s erotic dream. All that seems certain is that far from being the repository of “bad rhymes” that for centuries a small cottage industry has been occupied with counting up (Brogan 1981), the “Rape of the Lock” may be a masterwork of jokes driven by the phonological as well as semantic expectations of rhyme, which should be taken account of in any philological study. Tradition and the individual talent may thus turn out to interact in the transmission of poetic forms in ways not so different from the transmission of language in general. It could be the case that on the basis of exposure to their predecessor’s rhymes, a poet infers a grammar; and, especially because of complications posed by special cases, may err in its construction, in the sense of inadvertently constructing a different grammar from what their predecessor had, or may deliberately innovate on it. All the major questions of the actual mechanisms of linguistic change thus come up in microcosm with respect to rhyme.

5.

Conclusion

I have discussed this issue in general and simplified terms. But my only purpose has been to suggest that the questions of whether various older poets’ rhyming practices can be characterized by a grammar, and if so, what the relationships of transmission between various such grammars might be, are fascinating questions for the new millennium, ones that follow directly from conceptions of language proposed at the end of the last one. In Kökeritz’s (1953: vi) own words, “once we know Shakespearean phonology we are equipped to deal with all sorts of textual and prosodic problems in the English literature of the Renaissance and following decades which have so far been treated only superficially. Here is an almost untilled field of research for the competent philologist.”

226

Kristin Hanson

Notes 1.

2. 3.

I would like to thank Katy Breen and Karen Huynh for help in researching this paper, Donka Minkova and Robert Stockwell for providing an occasion for writing it, and John Clark, Sharon Inkelas, Paul Kiparsky, two anonymous reviewers and participants in SHEL-1 for helpful discussion. In (4e) I have omitted an accent Booth (1977) prints on commént. The resemblance between (4a) and (7a) is stronger if the lengthening of [I] to [i] word-finally is borne in mind (Chomsky and Halle 1968).

References Abrams, M.H. 1999 A glossary of literary terms, 7th ed. New York: Harcourt Brace. Baugh, Albert C. and Thomas Cable 1993 A history of the English language, 4th ed. Englewood Cliffs: Prentice Hall. Booth, Stephen (ed.) 1977 Shakespeare’s sonnets. New Haven: Yale University Press. Bridges, Robert 1921 Milton’s prosody, with a chapter on accentual verse and notes, rev. final ed. Oxford: The Clarendon Press. Brogan, Terry V.F. 1981 English versification 1570–1980: a reference guide with a global appendix. Baltimore: Johns Hopkins University Press. Cercignani, Fausto 1981 Shakespeare’s works and Elizabethan pronunciation. Oxford: The Clarendon Press. Chomsky, Noam 1959 A review of B.F. Skinner’s Verbal Behavior. Language 35: 26–58. Chomsky, Noam and Morris Halle 1968 The sound pattern of English. Cambridge, Mass.: New York: Harper and Row.

Vowel variation in English rhyme

227

Clark, Arthur Melville and Harold Whitehall 1974 Rhyme. In: Preminger, Alex (ed.), Princeton encyclopedia of poetry and poetics, enlarged ed., 705–710. Princeton: Princeton University Press. Dobson, Eric J. 1968 English pronunciation 1500–1700, 2nd ed. (1st ed. 1957), 2vols. Oxford: The Clarendon Press. Ellis, Alexander J. 1869–89. On early English pronunciation, with special reference to Shakespeare and Chaucer. (Early English Text Society, Extra Series 2, 7, 14.) London: Trübner. Finegan, Edward 1980 Attitudes toward English usage: the history of a war of words. New York/ London: Teachers College Press. Gabrielson, Arvid 1909 Rime as a criterion of the pronunciation of Spenser, Pope, Byron and Swinburne. Uppsala: Almqvist & Wiksells Boktryckeri-A.B. Halle, Morris and Samuel Jay Keyser 1966 Chaucer and the study of prosody. College English 28: 187–219. Hanson, Kristin 1998 The linguistic structure of rhyme in Robert Pinsky’s The Inferno of Dante. Paper presented at the annual meeting of the Linguistic Society of America, New York, January 8–11. Hollander, John 1981 Rhyme’s reason, new, enlarged ed. New Haven: Yale University Press. Holtman, Astrid 1994 A constraint-based approach to rhyme. Linguistics in the Netherlands 19: 49–60. Hughes, Merritt. 1957 John Milton: Complete poems and major prose. New York: MacMillan. Kiparsky, Paul 1977 The rhythmic structure of English verse. Linguistic Inquiry 8: 189–247. Kökeritz, Helge 1953 Shakespeare’s pronunciation. New Haven: Yale University Press.

228

Kristin Hanson

Lanz, Henry 1931 The physical basis of rime: an essay in the aesthetics of sound. Stanford: Stanford University Press. Matthews, J. Brander 1898 An inquiry as to rhyme. The Bookman 8: 32–38. Pinsky, Robert 1998 The sounds of poetry: a brief guide. New York: Farrar, Straus and Giroux. Malone, Joseph L. 1993 Partial rhyme. Hellas 1: 107–122. Newcomer, Alphonso G. 1899a License in English rhyme. – I. The Nation: 68: 63–65. Newcomer, Alphonso G. 1899b License in English rhyme. – II. The Nation: 68: 83–85. Oras, Ants 1955 Intensified rhyme links in The Faerie Queene: an aspect of Elizabethan rhymecraft. The journal of English and Germanic philology 55: 39–60. Rickert, William E. 1978 Rhyme terms. Style 12: 35–46. Ritchie, Frank 1901 Rhyme. Longman’s magazine 37: 114–124. Stryjewski, Kurt 1940 Reimform und Reimfunktion: Untersuchungen zum Problem des reinen und unreinen Reims in der englischen Dichtung des 20. Jahrhunderts. Breslau: Prietbatsch. Tillotson, Geoffrey et al. 1969 Eighteenth-century English literature. New York: Harcourt, Brace & World. Wimsatt, W.K. 1954 One relation of rhyme to reason. In: W.K. Wimsatt, The verbal icon: studies in the meaning of poetry, 153–168. Lexington, Kentucky: Kentucky University Press.

Vowel variation in English rhyme

229

Wright, George T. 1988 Shakespeare’s metrical art. Berkeley: University of California Press. Wyld, Henry Cecil 1923 Studies in English rhymes from Surrey to Pope: a chapter in the history of English. London: John Murray. Zwicky, Arnold. 1976 Well, this rock and roll has got to stop. Junior’s head is hard as a rock. CLS 12: 676–697.

Lexical diffusion and competing analyses of sound change Betty S. Phillips

1.

Introduction

A little over 100 years ago, in 1885, Schuchardt planted the seeds of modern lexical diffusion theory when he noted that during a sound change “rarely used words drag behind; very frequently used ones hurry ahead” ([1972]: 58). In 1969, Wang picked up Schuchardt’s idea and expanded it, introducing the term lexical diffusion to describe the gradual effect of a phonological rule as it spreads to more and more words in the lexicon. Since then, a number of articles have found that word frequency does indeed influence the lexical diffusion of a sound change, many finding that the most frequent words have changed first (e.g., Leslau 1969; Fidelholz 1975; Hooper 1976; Phillips 1980, 1983, 1998b; Rhodes 1996). Phillips (1984), however, also documented several sound changes which affected the least frequent words first and has subsequently refined the theory of word frequency and lexical diffusion, most recently in Phillips (2001). The purpose of this paper is to demonstrate how an understanding of the relationship between word frequency and sound change can be used to help choose between competing analyses of a given sound change, specifically the shift of Early Modern English /u:/ to /U/ in such words as good, stood, book, took, and foot. 1

2.

Description

The change of Early Modern English /u:/ to /U/ noticeably involves both a change in vowel quantity (i.e., shortening) and vowel quality (i.e., laxing 2), not surprising since vowel length and quality were apparently mutually dependent (Stockwell and Minkova 1990). Yet was the motivation for this process primarily to make the vowel shorter? Or was the motivation a

232

Betty S. Phillips

change in vowel quality? That is, was it primarily a quantitative adjustment or a qualitative one? Dobson (1968: 497), Wells (1982: 197), and Lass (1999: 90) all describe this change as a shortening. In contrast, Pyles and Algeo (1993: 171–172) treat it as a laxing of the vowel: “Middle English [o¯], as in ro(o)te ‘root,’ became [u]. Laxing of this [u] to [U] has occurred in book, foot, good, look, took, and other words; in blood and flood there has been unrounding in addition to laxing, resulting in [@] in these two words.” The choice seems arbitrary until one looks at the lexical diffusion of the shift, which has been documented by Ogura (1987) and which allows us to compare her findings with two theories about the origin of the shift – Ogura’s own and Görlach’s (1991) – both of which depend upon its characterization as a quantitative adjustment motivated either by higher order syllable structure or by analogy.

3.

Evidence

Using the orthoepical evidence given by Dobson (1968), Ogura (1987: 147) investigates the lexical diffusion of the “shortening” of ME /o:/ from the 16th century through the 17th century, within each of eight environments: “1 before [d], 2 before [t], 3 before [T], 4 before [v], 5 before [f], 6 before [k], 7 before other consonants, 8 after [Cl]” (Cl = cluster). Frequency counts were based on The Harvard Concordance to Shakespeare (Spevack 1973), which contains 884,547 tokens (Ogura 1987: 145). Dobson’s 16th-century data were based on ten orthoepists, including Cheke, Smith, Hart, Bullokar, Mulcaster, and Coote; the 1600–1650 data were based on works by Tonkis, Robinson, Gil, Hayward, Sherwood, Butler, Daines, Johnson, The English Scholemaster, and Hodges; and the 1650–1700 data were based on works by fourteen orthoepists of that period (Ogura 1987: 189–190). Ogura discovered that phonetic conditioning played an important role, that is, that the shortening “took place first before [d], then spread before [v], and lastly before [t, T, k], and that within a given consonant, the vowel in the more frequent words and the vowel preceded by [Cl] changed first. That is, shortening was very common before [d], and fairly common before [v] in the 16th century, and before [t, T, k] shortening became operative in the former half of the 17th century” (145). Ogura describes a rich interplay between phonetic conditioning and word frequency:

Lexical diffusion and competing analyses of sound change

233

Before [d] the shortening affected first blood and flood where the vowel is preceded by [Cl], and the more frequent words good, (-)hood, then stood, and lastly the less frequent word food. Before [v], glove when [Cl] precedes the vowel, and the more frequent words (re)move and (ap)prove changed earlier than the less frequent words behove and hove. Before [t, T, k], the change corresponded roughly to an increase in word frequency. (1987: 145)

Ogura’s evidence is, admittedly, not air-tight. There are not a large number of words susceptible to this change, and they range over different word classes. The examples given above involving the same word class are especially more convincing – (re)move and (ap)prove vs. behove and hove. Ogura’s example involving the words ending in –ood, on the other hand, may more appropriately be exemplifying differences between word classes. A better comparison would probably be the evidence that food according to one orthoepist could take the innovative vowel, 3 whereas there is no evidence the word mood has ever had that option. Another strong example in her corpus is the word foot (frequency 0.0200) showing the innovative vowel by 61.9% of the late 17th-century orthoepists versus no evidence of innovation in root (frequency 0.0053) or boot (frequency 0.0035). In current English, of course, foot generally contains [U], boot generally contains [u], and root may vary. 4 Therefore, the general trend within phonetic environments and within word classes does nevertheless still support the more frequent words changing first.

4.

Explanations

Within lexical diffusion theory, the implication of (within phonetic categories) the most frequent words changing first is clear: such sound changes always seem to be physiologically motivated. 5 For instance, sound changes which have been shown to affect the most frequent words first include unstressed vowel reduction and deletion, as in mistook with initial [I] vs. mistake with initial [@] (Fidelholz 1975) and two-syllable nursery vs. three-syllable cursory (Hooper 1976); t/d-deletion in American English (Bybee 1997); Old English vowel raising before nasals (Phillips 1980); Middle English diphthongization of the type [ej] becoming [ei] (Phillips 1983); and various assimilations and reductions in Ethiopian languages

234

Betty S. Phillips

(Leslau 1969) – all of which can be attributed to the physical characteristics of the vocal mechanism and speech production. In contrast, three changes have been documented (Phillips 1984) as affecting the least frequent words first: Southern American English glide deletion (e.g., /tjun/ > /tun/, /djuk/ > /duk/, etc.), which has as its impetus the reorganization of phonotactic constraints in English (as suggested by Cooley 1978); the unrounding of Middle English [ö(:)] to [e(:)], which has as its impetus the vowel system universal (or near-universal) that languages without high front rounded vowels do not have mid-front rounded vowels; and the stress shift in such words as cónvict (noun) vs. convíct (verb) from earlier convíct (noun or verb), which requires access of the lexical entry to determine the part of speech. None of these has as its impetus the physiological features of the vocal tract. Rather, they have their bases in syllable structure, language universals, and analogy based on word class. As detailed in Phillips (2001), all of these changes affecting the least frequent words first require lexical analysis; that is, they require accessing of information from the lexical entry before they can take effect. In contrast, changes affecting the most frequent words first ignore information in the lexical entry, allowing features of segments to blur, producing assimilations and reductions, or allowing stress rules to reset to default positions. Yet Ogura (1987) and Görlach (1991) both posit explanations for what they call Early Modern English shortening that have to do with syllable structure and analogy, respectively – where one would expect the least frequent words to be changing first, despite Ogura’s data exhibiting the most frequent words changing first.

4.1 Ogura’s hypothesis Ogura (1987: 150–151) hypothesizes that the impetus behind this change was an attempt to keep the duration of syllabic units relatively constant: Dobson (1968: §24) suggests that shortening is ‘essentially due to a general tendency to shorten the vowels of closed syllables,’ though ‘the consonants differ in their power to cause it.’ The assumption underlying this explanation is that the duration per syllabic unit is relatively constant, and there is a negative correlation or temporal compensation between the vowel and the following consonant. We may assume that

Lexical diffusion and competing analyses of sound change

235

the first factor that motivated the change is this temporal compensation within the monosyllabic word.

She hypothesizes that since vowels tend to be longer before voiced consonants, shortening took place first before [d] (151). Note that this is precisely opposite the direction that one would expect articulatorily (Laver 1994: 446). By seeing this change as a shortening, she is led to an explanation whereby “ModE shortening was motivated by the higher-level articulatory unit, i.e., by the articulatory plan based on units larger than single segments” (151). But not only would such a motivation most likely result in the least frequent words being changed first, there is also no supporting evidence that English in the 16th century was moving toward a system of equal weight per syllable. The movement in Middle English that Dobson (1968) mentions and that has been investigated more fully by Murray and Vennemann (1983) never attained fulfillment, largely because of the counter-movement toward schwa loss, which created closed syllables where open ones had existed before. And so by the 16th century, a different motivation must be sought.

4.2 Görlach’s hypothesis Görlach (1991: 71) notes that “short vowels reflecting ME long vowel quantities are most frequent where ME has /E:, o:/ before /d, t, T, v, f/ in monosyllabic words, but even here they occur only in a minority of possible words.” 6 He suggests: It is likely that the short vowel was introduced on the pattern of words in which the occurrence of a short or a long vowel was determined by the type of syllable the vowel appeared in (glad vs. glade). When these words became monosyllabic in all their forms, the conditioning factor was lost and the apparently free variation of short/ long spread to cases like (dead). That such processes must have continued for some time is shown by words ending in -ood: early shortened forms (flood) are found side by side with later short forms (good) and those with the long vowel preserved (mood). (1991: 71)

Similarly, Strang (1970: 116) suggests, “There are changes of quantity, but their consequences in recent English are so confused that we must believe conflicting analogies have been at work.” And Ritt’s (1997: 69) appeal to

236

Betty S. Phillips

semiotics to explain a preference for short /u/ before dentals also relies on speakers “check[ing] the mental dictionary” to resolve the ambiguity of the /d/ in /blu:d/, for example, being a past tense marker or not. Yet an explanation that requires comparison with other words in the lexicon is unconvincing since analogical changes affect the least frequent words first; that is, “infrequent words tend to regularize before frequent ones” (Bybee 1995: 236). Hooper’s (1976: 99–100) study, for instance, found that more frequent verbs like kept, left, and slept resist being re-shaped with an –ed suffix, whereas less frequent verbs like crept, leapt, and wept have developed the new forms creeped, leaped, and weeped. Although the environment of Hooper’s example is morphological, it is similar to Görlach’s, Strang’s, and Ritt’s proposals in that the shape of an infrequent morpheme is chosen through comparison with other morphemes. Frequent forms, in contrast, can even lose their connection to their own lexical content, as in awful no longer being connected to awe, desperate to despair, or terrific to terrified (Bybee 1985: 88).

4.3 An alternative explanation So we are driven to search for a physiological motivation for this change. But are we looking for a motivation for shortening or for laxing? If the Early Modern English change were a true shortening, one would expect it to affect vowels before /t/ sooner than vowels before /d/, since languages generally have longer vowels before voiced segments, although not all languages have as great a difference as does Modern English (Laver 1994: 446). Similarly, one would least expect shortening of /u:/ before /d/, as Lehiste (1970: 20) explains in summarizing the work of Fischer-Jørgensen (1964) on the effect of a consonant’s place of articulation on the length of the preceding vowel: The duration of a vowel depends on the extent of the movement of the speech organs required in order to come from the vowel position to the position of the following consonant. The greater the extent of the movement, the longer the vowel. This explains the fact that all vowels were shorter before /b/ than before /d/ and /g/: since two different articulators are involved in the sequence vowel + labial, there is no time delay in moving the articulator (i.e., the tongue) from the vowel target to the consonant target. On the other hand, /u/ was particularly long before /d/.

Lexical diffusion and competing analyses of sound change

237

Judging from the phonetic evidence, alone, then, one would not expect this to be a vowel shortening. Therefore, the change is more likely to be a qualitative assimilation – which leads us to question what about the environment following a consonant-plus-[l] cluster and preceding [d] would lead to a qualitative change to [U]. The diagrams in Figure 1, taken from Ladefoged and Maddieson (1996: 184, 303), demonstrate how the tongue position of [l] could encourage the shift in tongue position from [u] to [U].

Lateral German

[u/U] (German, after Bolla and Valaczkai 1986)

Figure 1. Tongue Position of Laterals and High Back Vowels (Ladefoged and Maddieson 1996: 184, 303). Reprinted by kind permission of Blackwell Publishers.

As for why the change would occur before dentals earlier than before other consonants, the articulatory basis might be that, as Ohala (1974: 265) notes, “although it is not necessarily the case that the back of the tongue is lower than usual for an [u] pronounced in the environment of dental consonants, it is frequently the case.” Ohala (1974: 265) continues: “... velars have little effect on vowel quality, presumably because the lip configuration and the place of articulation of the ‘velar’ can vary rather freely as a function of the contiguous vowel, second, ... labials have the greatest effect on front vowels (though not on [i]), mainly tending to centralize them, and third, ... dentals have the greatest effect on back vowels, notably high back vowels, and the effect again is to centralize them.” All that remains to explain is why the change would occur before [d] rather than [t], and there is some evidence that [d] can be seen as relatively lax compared to [t]. That is, while some phonologists reject the identifica-

238

Betty S. Phillips

tion of voiceless consonants as tense/fortis and voiced consonants as lax/ lenis, Ladefoged and Maddieson (1996: 96), in their discussion of the issue, find that “rates of articulator movement and muscular activation levels often do differentiate between phonological voicing categories.” Although they warn that “as little is known about the articulatory dynamics of most languages, we would caution against making the assumption that phonological voicing differences are associated with articulatory strength differences in any particular case,” they state: Voiceless stops have a greater mean oral pressure than voiced stops, and also often have a greater peak oral pressure. Accordingly, the greater degree of articulatory activity in the formation of the closure may be an anticipation of this need to make a firmer seal. In principle, the parameters of voicing and gestural stiffness could vary independently. When they co-vary either one might be regarded as primary. (1996: 96)

So to suggest that less muscular energy is required before [d] than before [t] is not unreasonable. Therefore, I suggest that physiologically this sound change is a simple assimilation to the neighboring consonants. Only if we see it as primarily a qualitative change can we account for its behavior: its occurrence first after consonant + /l/ and before /d/, and its direction of diffusion from more to less frequent words.

5.

Conclusion

Our investigation of the laxing of Early Modern English /u:/ to /U/ has served two purposes: to clarify the phonetic motivation behind this sound change, and to illustrate how lexical diffusion theory may guide historical linguists in evaluating competing analyses of a given sound change. The theory of lexical diffusion is still under construction, it is true, but it promises to be a powerful tool for future studies in historical phonology.

Notes 1.

Many thanks go to Robert Stockwell, Donka Minkova, and two anonymous reviewers, all of whom had comments and suggestions that made this paper immeasurably better. Shortcomings, of course, remain strictly my own.

Lexical diffusion and competing analyses of sound change 2.

3.

4.

5.

6.

239

I will use the term laxing as a cover term to refer to the change in vowel quality involved in this shift. I realize that the appropriateness of the term lax has long been questioned, as in Lindau’s (1978: 557–558) discussion of the difficulty of defining the phonetic correlates of tenseness. Lindau concludes that a feature distinguishing tense versus lax vowels “is probably not required for classificatory purposes, because it never seems to occur without concomitant differences in length” and that “the difference between tense and lax vowels is best labeled by a feature Peripheral,” which is an acoustically defined feature first used by Stockwell, as in his (1973) interpretation of the Great Vowel Shift. The question posed in the current paper, however, is not what feature is best for “classificatory purposes” or which feature best describes the acoustic properties of the sounds in question, but whether the motivation behind the change is a quantitative adjustment or a qualitative assimilation. Ogura (1987: 200–201) accounts for the re-lengthened vowel by appealing to homonymic clash with the word fud ‘buttocks.’ Similarly, Barber (1997: 124) claims there is evidence “for the existence of all three pronunciations of the word foot, namely [fu:t], [fUt], and [fˆt], although the third was possibly vulgar.” Carroll et al. (1971) lists the frequency per million tokens of food as 467.35, root as 58.060, and boot as 5.6262. The word soot is described as short in half of the late 17th-century sources, but its frequency is not listed in Spevack’s (1973) Shakespeare concordance (Ogura 1987: 146, 200); Carroll et al. (1971) give the frequency per million tokens as 1.8513. One might speculate that it must have been a frequent word in Shakespeare’s time, despite its absence in his works. Although Labov (1994: 543) and Kiparsky (1995) both recognize lexically diffused sound changes, neither addresses the difference between lexically diffused changes which affect the most frequent words first and those which affect the least frequent words first. See Phillips (1984, 1998a) and Ogura (1995) for further discussion. For a Gilliéronian’s perspective on Labov (1994), see Kretzschmar’s (1996) review. For more on the parallel development of ME /E:/, see Labov (1994: 304–305; 528–529). The similarity of the changes may be gauged by Labov’s description of how members in this class in the 17th century “were either shortened to /e/ or rose to high position and merged with the e class, except those beginning with initial clusters of consonant plus /r/, which merged with the a¯ class” (306); “shortening” also occurred before dentals, as in “head, dead, breath, and sweat” (528). Ogura (1987: 173) states that “before dentals [d], [T], and [s], it seems that shortening [of ME /E:/] took place earlier in the more frequent words than in the less frequent ones, but before [t] no such tendency can be found.”

240

Betty S. Phillips

References Barber, Charles 1997 Early Modern English. Edinburgh: Edinburgh University Press. First published André Deutsch [1976]. Bybee, Joan 1985 Morphology: A Study of the Relation between Meaning and Form. Amsterdam: John Benjamins. 1995 Regular morphology and the lexicon. Language and Cognitive Processes 10: 425–455. 1997 The phonology of the lexicon: evidence from lexical diffusion. In: Michael Barlow and Suzanne Kemmer (eds.), Usage-Based Models of Language, 65–85. Stanford: CSLI (Center for the Study of Language and Information) Publications. Carroll, John, Peter Davies, and Barry Richman (eds.) 1971 The American Heritage Word Frequency Book. Boston: Houghton Mifflin. Cooley, Marianne 1978 Phonological constraints and sound changes. Glossa 12: 125–136. Dobson, E.J. 1968 English Pronunciation 1500–1700. 2 volumes. Oxford: Clarendon. Fidelholz, James 1975 Word frequency and vowel reduction in English. In: Robin E. Grossman, L. James San, and Timothy J. Vance (eds.), Papers from the Eleventh Regional Meeting of the Chicago Linguistic Society, 200–213. Chicago: Chicago Linguistic Society. Fischer-Jørgensen, Eli 1964 Sound duration and place of articulation. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 17: 175–207. Görlach, Manfred 1991 Introduction to Early Modern English. Cambridge: Cambridge University Press. Hooper, Joan 1976 Word frequency in lexical diffusion and the source of morphophonological change. In: William M. Christie, Jr. (ed.) Current Progress in Historical Linguistics, 95–105. Amsterdam: North-Holland.

Lexical diffusion and competing analyses of sound change

241

Kiparsky, Paul 1995 The phonological basis of sound change. In: J.A. Goldsmith, ed. The Handbook of Phonological Theory, 640–670. Oxford: Blackwell. Kretzschmar, William 1996 A glass half empty, a glass half full. American Speech 71: 198–205. Labov, William 1994 Principles of Linguistic Change, Volume 1: Internal Factors. Oxford: Blackwell. Ladefoged, Peter and Ian Maddieson 1996 The Sounds of the World’s Languages. Oxford: Blackwell. Lass, Roger 1999 Phonology and Morphology. In: Roger Lass (ed.) The Cambridge History of the English Language, Volume 3: 1476–1776. 56–186. Cambridge: Cambridge University Press. Laver, John 1994 Principles of Phonetics. Cambridge: Cambridge University Press. Leslau, Wolf 1969 Frequency as a determinant of linguistic change in the Ethiopian languages. Word 25: 180–189. Lindau, Mona 1978 Vowel features. Language 54: 541–563. Murray, Robert and Theo Vennemann 1983 Sound change and syllable structure in Germanic phonology. Language 59: 514–528. Ogura, Mieko 1987 Historical English Phonology: A Lexical Perspective. Tokyo: Kenkyusha. 1995 The development of Middle English ¯ı and u¯: a reply to Labov (1992, 1994). Diachronica 12: 31–53. Ohala, John 1974 Phonetic explanation in phonology. In: Anthony Bruck, R.A. Fox, and M.W. LaGaly (eds.) Papers from the Parasession on Natural Phonology. 251–275. Chicago: Chicago Linguistic Society.

242

Betty S. Phillips

Phillips, Betty 1980 Old English an ~ on: a new appraisal. Journal of English Linguistics 14 (1980): 20–23. 1983 ME diphthongization, phonetic analogy, and lexical diffusion. Word 34: 11–24. 1984 Word frequency and the actuation of sound change. Language 60: 320–342. 1998a Lexical diffusion is NOT lexical analogy. Word 49: 369–381. 1998b Word frequency and lexical diffusion in English stress shifts. In: Richard Hogg and Linda van Bergen (eds.), Historical Linguistics 1995, Volume 2: Germanic Linguistics. 223–232. Amsterdam: John Benjamins. 2001 Lexical diffusion, lexical frequency, and lexical analysis. In: Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structure. 123–136. Amsterdam: John Benjamins. Pyles, Thomas and John Algeo 1993 The Origins and Development of the English Language, 4th ed. Fort Worth: Harcourt Brace Jovanovich. Rhodes, Richard 1996 English reduced vowels and the nature of natural processes. In: Bernhard Hurch and Richard A. Rhodes (eds.), Natural Phonology: The State of the Art, 239–259. Berlin: Mouton de Gruyter. Ritt, Nikolaus 1997 Early Modern English vowel shortenings in monosyllables before dentals: a morphologically conditioned sound change. In: Raymond Hickey and Stanislaw Puppel (eds.), Language History and Linguistic Modelling: A Festschrift for Jacek Fisiak on his 60th Birthday, Volume 1: Language History. 65–71. Berlin: Mouton de Gruyter. Schuchardt, Hugo [1972] Über die Lautgesetze: gegen die Junggrammatiker/ On sound laws: against the Neogrammarians. In: Theo Vennemann and Ronnie Wilber (eds.), Schuchardt, the Neogrammarians, and the Transformational Theory of Phonological Change, 39–72. Frankfurt, Athenäum. First published Berlin: Oppenheim, 1885. Spevack, Marvin 1973 The Harvard Concordance to Shakespeare. Cambridge, MA: Harvard University Press.

Lexical diffusion and competing analyses of sound change

243

Stockwell, Robert 1973 Problems in the interpretation of the Great English Vowel Shift. In: M. Estelle Smith (ed.), Studies in Linguistics: Papers in Honor of George L. Trager, 344–362. The Hague: Mouton. Stockwell, Robert and Donka Minkova 1990 The early modern English vowels, more o’ Lass. Diachronica 7: 199–214. Strang, Barbara 1970 A History of English. London: Methuen. Wang, William S-Y 1969 Competing changes as a cause of residue. Language 45: 9–25. Wells, J.C. 1982 Accents of English, Volume 1: An Introduction. Cambridge: Cambridge University Press.

Dating criteria for Old English poems Geoffrey Russom

1.

Use of Sievers’s five-types system in attempts to date Old English poems

In the system of Sievers (1893), Old English meter is represented as a set of five verse types labeled A, B, C, D, and E. Researchers interested in the development of this meter have observed some apparent changes in type frequencies over historical time. Cable (1981), for example, claims that the frequencies of types C, D, and E declined toward the end of the Old English period. One important goal of Cable’s research was to test dating criteria for Old English poems against the evidence of type frequency. He concluded that previous attempts to date Beowulf might have placed the poem a hundred years too early. 1

1.1 Problems caused by varying employment of compounds Calculation of type frequencies is beset with methodological problems. Fulk (1992: 256) observes that the statistical variations observed by Cable are rather small and attributes them to non-metrical constraints on employment of compound words. Consider the following examples from Beowulf: (1) to¯ łæ¯ m to the-DAT.SG ‘to the splendid hall’

goldsele goldhall-DAT.SG

Beo 1639a 2

(2) swa¯ wæs so be-3SG.PRET ‘so it was for Beowulf’

B¯ıowulfe Beewolf-DAT.SG

Beo 3066a

(3) go¯d good-NOM.SG ‘the good king’

gu¯Dcyning battleking-NOM.SG

Beo 2563a

246

Geoffrey Russom

(4) cwe¯n woman-NOM.SG ‘Hrothgar’s wife’

Hro¯Dga¯res Gloryspear-GEN.SG

Beo 613a

(5) be¯orscealca beerdrinker-GEN.PL ‘One of the retainers’

sum one-NOM.SG

Beo 1240b

(6) Ecgla¯fes Swordsurvivor-GEN.SG ‘Ecglaf’s child’

bearn child-NOM.SG

Beo 499b

Examples 1, 3, and 5 illustrate employment of traditional poetic compounds in types C, D, and E, respectively. Forms like gold-sele, guD-cyning, and be¯or-scealc express a native heroic ethos and could not be exploited straightforwardly in the poems on Christian subjects that make up the bulk of the corpus. Examples 2, 4, and 6, respectively, are C, D, and E verses with compound proper names of the Germanic type. In Beowulf and The Battle of Maldon, verses like these appear with considerable frequency. In the long Old English poems based on imported Christian narratives, there are no Germanic characters, hence no verses like 2, 4, and 6.

1.2 Inherent limitations of the five-types system In any metrical history, the researcher needs to identify nonstandard metrical practices that arise when traditional rules have lost their authority. Sievers offers no reliable procedures for carrying out this essential task. His five-types system has no metrical rules in the usual sense – no general principles of well-formedness that apply in the same way to all verses. Within Sievers’s system, the metrical status of a given verse is assessed by calculating the frequency of other verses that are similar in certain respects assumed to be significant, such as number and arrangement of stresses. Such calculations cannot readily distinguish complex but acceptable variants of a legitimate type from unmetrical verses. In the variant of Sievers’s theory formulated by Bliss (1967), this problem arises even more acutely (Russom 1998: 194–196). If assessment of metrical complexity is based entirely on

Dating criteria for Old English poems

247

verse counts, moreover, changes in type frequencies cannot be assumed to represent the decay of a verse tradition. Independent criteria for metrical complexity would make analysis of frequency changes much more straightforward.

2.

The word-foot theory as an alternative to the five-types system

In this article I reevaluate the metrical evidence from the perspective of the word-foot theory (Russom 1987, 1998). This theory consists of two general principles requiring correspondence between metrical form and linguistic form: (7) Principle 1: Foot Pattern = Word Pattern Corollary: Since feet corresponding to common word patterns are easiest to identify during real-time scansion, the ideal foot has the paradigmatic word pattern (trochaic). 3 Feet corresponding to other word patterns add to metrical complexity. Principle 2: Alliterative Binding Rule = Compound Stress Rule Corollary: Since compounding is a binary operation joining two smaller words into one larger word, binding of feet into verses and of verses into lines by alliteration will proceed in a binary fashion, yielding verses of two feet and lines of two verses.

2.1 Deviation from verse norms Within the word-foot framework, the most basic realization of each verse type is a pair of words. From principle 1 it follows that the metrically simplest or ideal verse consists of two trochaic words, as with example 8. (8) longe long-ACC.SG ‘for a long time’

łra¯ge time-ACC.SG

Beo 54b

Verses with the stress pattern of 8 are subsumed under an ideal type A1 by Sievers on purely statistical grounds. Within the word-foot framework, the

248

Geoffrey Russom

metrical simplicity of type A1 follows from general principles of versecraft, quite independently of verse counts. As predicted, two-word realizations of type A1 have a remarkably high frequency that cannot be attributed to purely linguistic factors. Although the trochaic word pattern is clearly paradigmatic in Old English, complete syntactic constituents like 8, with two trochaic words and no unstressed function words, are not particularly common in OE prose or in the nontraditional alliterative verse composed by Ælfric. 4 Ælfric’s 287-line poem on the death of St. Oswold, for example, contains no instances of two-word type A1. The first 287 lines of Beowulf contain more than eighty instances. Within the word-foot framework, verse counts serve to validate independent assessments of complexity.

2.2 Management of complexity within the line Old English meter is subject to a universal principle of closure: “beginnings free, endings strict” (Hayes 1983: 373). Verses that deviate significantly from the norm have an elevated frequency in the first half of the line, called the a-verse. Some types of deviation occur only in the a-verse. Verses close to the norm have an elevated frequency in the second half of the line, called the b-verse (Bliss 1967; Russom 1987: 49–50). From the perspective of the word-foot theory, it is possible to identify – and to explain – a dramatic change toward the end of the Old English period in placement of two-word A1 variants like example 8. In Beowulf, 65% of such variants appear in the b-verse. The figures for most OE poems lie between 65% and 75%. In two poems known or believed to be very late, however, the percentage jumps quite remarkably. The Battle of Maldon has just three a-verses and 79 b-verses of two-word type A1 (96% b-verses). The longest psalm of the Paris Psalter, number 118, has 14 a-verses and 204 b-verses (94% b-verses). Wholesale displacement of the simplest verses to the second half of the line is necessary in these late poems because they contain so many variants of extreme complexity. In metrical psalm 118, nearly half of the type D and E verses are preceded by extrametrical unstressed words that complicate scansion, what specialists call anacrusis. Example 9 is a representative type E variant with the anacrusis between brackets.

Dating criteria for Old English poems

(9) [and ic] unrihte and I-NOM.SG unrighteous-ACC.PL ‘and I (despise) unrighteous ways’

249

wegas Ps118 128.2a 5 way-ACC.PL

Variants like 9 do not occur in poems of the early period, when a more robust inflectional system made it easier to omit excess function words at the beginning of the verse. During this early period, the function of the personal pronoun ic in 9 could be performed by a verbal inflection, while the conjunction and could be avoided through paratactic juxtaposition of clauses. 6 In psalm 118, the simpler variants of types D and E have significant frequency in the b-verse, but the 23 examples with peculiar anacrusis are situated without exception in the a-verse. The poet knows he is in trouble with these variants and mitigates their awkwardness by putting them in their proper place. The poet also realizes that the best way to free up room in the a-verse for additional complex variants is to shift metrically simpler variants to the b-verse. 7

2.3 Selection of appropriate variants for study Maldon and the Paris Psalter provide fascinating insight into metrical death but little insight into earlier periods of Old English metrical history. Certain variants of types C, D, and E provide more broadly useful dating criteria. These variants have a long simplex word with a heavy medial syllable rather than a compound word with a secondary root. I will refer to them as simplex C, simplex D and simplex E. Representative examples are given in items 10–12. (10) cwæD, he¯ on mergenne Beo 2939a say-3SG.PRET he-NOM.SG in morning-DAT.SG ‘he said that in the morning he (would kill his enemies)’ (11) ræd ¯ ænigne ¯ counsel-ACC.SG any-ACC.SG ‘any plan’

Beo 3080b

250

Geoffrey Russom

(12) slæpendne ¯ rinc sleeping-ACC.SG warrior-ACC.SG ‘the sleeping warrior’

Beo 741a

In item 10, an example of simplex C, the high-frequency finite verb cwæÍ is treated as unstressed and alliteration falls on mergenne, the first word with metrically significant stress. Like other type C variants, simplex C allows for rather free addition of extrametrical unstressed words at the beginning of the verse. Items 11 and 12 represent simplex D and E, respectively. The problems identified by Fulk do not arise for simplex variants because the long words they contain can be employed straightforwardly in Christian as well as native heroic narratives.

3.

Date of composition and strictness of versecraft

Diachronic studies of meter begin with the hypothesis that language change makes it increasingly difficult to compose in a traditional form. On this hypothesis, all other things being equal, the earlier of two poems should exhibit the stricter versecraft. Since every era hosts a variety of poetic talent, however, we need to consider not only the date but the quality of composition.

3.1 Expert assessments of individual poems Table 1 provides recent expert assessments of date and quality for the Old English poems to be analyzed. The first column in Table 1 (“work”) gives the abbreviated title for each verse selection. Selections in the upper portion of the table are ordered as in Cable’s test chronology, which is based on consensus dates of previous scholarship. This is the chronology used as a starting point in Fulk’s rehabilitation of linguistic dating criteria (Fulk 1992). No traditional consensus is available for the five selections in the lower portion of the table. These are dated according to best guesses in Fulk (1992) and Hutcheson (1995). Note that estimates in column 2 (“date”) allow for reordering of some adjacent items. For simplicity of exposition, I reduce these estimates in sub-

Dating criteria for Old English poems

251

Table 1

1 work

2 date

3 merit

4 size

5 A3%

GenA

early

so-so

Dan Beo Ex Cyn And Met Brun Jud Chron Mald

early early > 825 > 850 9th c. in 897 < 937 900–50 < 942 < 991

so-so best best best good bad best best good bad

4636 1092 6240 1100 4934 3294 3500 145 698 202 648

9.3 4.3 4.1 5.9 5.9 8.6 2.1 2.6 1.9 9.2

GuthA Chr & Sat GuthB Phoen Ps 118

early < Beo < Cyn < Cyn late

bad so-so best best bad

1544 1398 1088 1354 1071

9.5 9.4 1.4 3.6 13.8

4.0

sequent tables to three period categories: early, middle, and late. In column 3, I assign poems to four merit categories: best, good, so-so, and bad. These categories summarize expert judgments based on the number of verses for which there is no normal scansion. 8 Column 4 gives the size of the selection, expressed as a number of verses.

3.2 Distribution of type A3 as independent evidence for merit Given the limitations of the five-types system, scholarly assessments of versecraft tend to be cautious. Merit rankings based on the number of

252

Geoffrey Russom

anomalous verses receive independent support, however, from the distribution of type A3 verses like 13. (13) nu¯ now

ge¯ mo¯ton gangan you-2PL may-2PL go-INF ‘now you may go’

Beo 395a

Such verses are very easy to construct, since they may contain any number of unstressed words along with a verse-final stressed word that has the high-frequency trochaic pattern (e.g. gangan). In other types, the poet must adhere to special constraints on extrametrical words, placement of the caesura, and placement of alliteration, but no such constraints are imposed on variants like 13. With its single stressed root, this type of variant deviates significantly from the metrical ideal, which has two stressed roots. 9 In “good” and “best” poems, type A3 is restricted to the a-verse, the proper place for deviant metrical patterns. A certain minimal frequency of type A3 would be necessary to accommodate strings of function words in some indispensable syntactic constructions, but a good poet would not overuse this type. The results for the poems in our corpus are presented in the last column of table 1 (“A3%”). 10 All but one of the poets independently ranked as “best” restrict type A3 to less than 5% of total verses. The “good” Chronicle poetry (Chron) and one “so-so” poem (GenA) also stay below 5%. The “best” poetry of Cynewulf (Cyn) and the remaining “good” poem (And) stay below 6%. All the “bad” poems have more than 8% type A3. The average “bad” poet exploits this type about twice as frequently as the average “best” poet.

3.3 Anomalous faults and refinements Given our hypothesis that language change made it increasingly difficult to use the traditional meter, we would expect certain relations to hold among date, talent, and conformity to metrical standards. In the “best” and “good” categories, we would expect anomalous metrical faults primarily during the late period, when recalcitrant linguistic material would challenge even the most gifted poet. Anomalous metrical refinements in the “bad” and “so-so” categories, on the other hand, should appear primarily during the early peri-

Dating criteria for Old English poems

253

od, when strict traditional versecraft would be easiest to achieve. Further, the frequency of metrical faults should increase earlier in “good” poems than in “best” poems and earlier in “bad” poems than in “so-so” poems. These expectations are summarized in 14. (14)

Relation of metrical anomaly to date and merit: a. Anomalous metrical faults should occur in late “best” poems and in “good” poems that are not early. b. Anomalous metrical refinements should occur in early “bad” poems and in “so-so” poems that are not late.

Observe that the only anomaly worth mentioning in the distribution of type A3, a remarkably low frequency of 4% in an early “so-so” poem (GenA), is consistent with 14b.

4.

Deviation from verse norms in simplex C, D, and E

With preliminaries in place, we now turn to the simplex C, D, and E variants (examples 10, 11, and 12, respectively). Simplex D is normative in that it has two stressed roots. Simplex E also has two stressed roots, but this type is deviant in that the longer word comes first. In most Old English verses containing long simplexes or compounds of similar length, the long word comes last, and types with the long word first are subject to special constraints that testify to their metrical complexity (Russom 1987: 29–31). Simplex C, with only one stressed root, stands still farther away from the metrical ideal. To construct a simplex D verse, the poet often has to use archaic word order, placing a modifier with an overt inflection after the constituent it modifies, as in 11. Simplex E variants like 12 require no such efforts. Simplex C is even easier to construct. Like type A3, it imposes no special constraints on alliterative patterning, extrametrical words or placement of the caesura. All other things being equal, then, we would expect a “best” or “good” poet to have a higher percentage of simplex D in relation to simplex E and simplex C. “Best” and “good” poets should also have a higher percentage of simplex D in relation to total D verses, since the other variants of type D, represented by 3 and 4, deviate from the norm in having three stressed roots. The same higher percentage for simplex E would be expected, since the other variants of this type, represented by examples 5

254

Geoffrey Russom

and 6, also have three stressed roots. Employment of archaic word order may reduce the percentage of simplex E variants, however, by transforming them into simplex D variants with an even higher metrical value. Unlike simplex D and E, simplex C is the deviant realization of its type. With just one stressed root, a simplex C variant like 10 lies farther from the metrical norm than C variants like 1 and 2, which have the canonical number of stressed roots. The percentage of simplex C in relation to total type C should accordingly be lower in the work of “good” or “best” poets.

4.1 Statistical evidence Table 2 illustrates relations between relative frequency and metrical complexity in simplex C, D, and E. Table 2 7 8 6 sim- % E D/E +D % plex Ea:b 26 : 24 21.9 21.8 4:5 23.1 40 16 : 24 9 62.9 5:5 8 50 20 : 19 18.3 36.1 16 : 20 22.5 40 5 : 16 13 4.5 0:0 0 100 4:1 26.3 37.5 1:3 40 20 5:2 20.6 12.5

10 11 9 sim- % C D/C +D % plex C a:b 35 : 20 7.5 20.3 11 : 11 9.1 21.4 40 : 13 4.8 56.2 8:0 5.3 55.5 59 : 30 9.6 19.8 28 : 8 5.9 40.0 36 : 15 9.9 1.9 0:0 0 100 4:7 10.9 21.4 1:0 4.3 50 4:1 6.1 16.6

5 %D

so-so so-so best best best good bad best best good bad

4 simplex Da:b 8:6 1:5 23 : 45 6:4 8 : 14 9 : 15 0:1 0:1 2:1 0:1 0:1

bad so-so

2:1 5:9

7.9 36.8

9:4 7:8

32.5 31.9

18.8 19 : 24 48.2 11 : 1

14.2 7.7

6.5 53.8

best best bad

4:6 0:4 0:0

19.6 11.4 0

4:2 2:1 1:0

9.4 6 16.7

62.5 57 0

11.6 6.4 11.8

40 23.5 0

1 work

2 date

3 merit

GenA Dan Beo Ex Cyn And Met Brun Jud Chron Mald

early early early early middle middle late late late late late

GuthA early Chr & middle Sat GuthB late Phoen late Ps 118 late

9.4 15.4 17.9 17 10.7 9.2 1.3 33.3 10 14.3 8.3

15 : 5 10 : 3 10 : 1

Dating criteria for Old English poems

255

4.1.1 Simplex D in relation to simplex E Column 4 in Table 2 gives the number of simplex D variants for each selection, with separate counts for a-verses and b-verses. Column 5 expresses the frequency of simplex D verses as a percentage of total D verses. Columns 6 and 7 give corresponding figures for simplex E. 11 With two exceptions, the figures for “best” poems are larger in column 5 than in column 7, indicating a higher frequency for simplex D. All “bad” poems have the larger figure in column 7, indicating a higher frequency for simplex E. The “so-so” poem with a higher frequency for simplex D (Chr & Sat) is not late. The “good” poems with a higher frequency for simplex E (And, Chron) are not early. One “best” poem with a higher frequency for simplex E (Jud) is late, as expected. The only unexpected result is for Cynewulf, a “best” poet placed in the middle period of the traditional chronology who shows a higher frequency for simplex E. Column 8 in Table 2 expresses the frequency of simplex D as a percentage of the totals for simplex D and simplex E, added together (column 4 + column 6). All the “best” poems have 35% or more simplex D. If this figure seems a little low, recall that construction of simplex D often requires employment of archaic syntax. The “good” entry showing less than 35% (Chron) is not early. The two “so-so” poems with more than 35% (Dan, Chr & Sat) are not late.

4.1.2 Simplex D in relation to simplex C Column 9 in Table 2 gives the number of simplex C verses for each selection, with separate counts for a-verses and b-verses as before. Column 10 expresses the frequency of simplex C verses as a percentage of total C verses. For most poems, the figure in column 10 is smaller than the figure in column 5, indicating that simplex C has a lower frequency than simplex D. The one “best” poem showing a higher frequency for simplex C (Jud) is late, as expected. The other poems showing a higher frequency for simplex C are all “bad.” The principle of closure can help us gauge the poet’s sensitivity to metrical deviance in simplex C. Most poets have a lower frequency of this deviant type in the b-verse, as indicated in column 9. One “bad” poem (GuthA) has a higher frequency in the b-verse. One “so-so” poem (Dan)

256

Geoffrey Russom

has the same frequency in the a-verse and b-verse. The “best” poem with a higher frequency in the b-verse (Jud) is late as expected. Column 11 in Table 2 expresses the frequency of simplex D as a percentage of the totals for simplex D and simplex C, added together (column 4 + column 9). All the “bad” poems have frequencies under 20% for the metrically superior simplex D. Most “best” poems have frequencies over 40%. They are joined by the two “good” poems and a “so-so” poem of the middle period (Chr & Sat). Two “best” poems with less than 40% (Jud, Phoen) are late, as expected. The remaining “so-so” poems have frequencies from 20% to 30%. The unexpected result is again for Cynewulf, a “best” poet of the middle period in the traditional chronology with a frequency of 19.8%. By this measurement, Cynewulf falls well below the standard of Andreas (And), a “good” rather than “best” poem of the middle period with a frequency of 40%.

4.2 Frequency of two-word type A1 a-verses as an independent dating criterion Earlier we observed a spectacular shifting of two-word type A1 verses to the closing half of the line in a pair of late “bad” poems (Mald and Ps118). As it turns out, another aspect of type A1 distribution is more broadly informative. The need to make room for complex variants in the a-verse can be measured independently by calculating the frequency of two-word a-verses relative to total a-verses in type A1. The results are summarized in table 3. In this table, column 4 gives the total number of two-word A1 variants in the a-verse. Column 5 gives the total number of a-verses for this type, and the last column expresses the frequency of two-word type A1 a-verses as a percentage of total type A1 a-verses. All the early poets have 34% or more of their two-word type A1 variants in the a-verse. Most have about 40%; the low score of 34% is for an early “bad” poem. A “good” poem of the middle period (And) has 41%. A late “best” poem (Jud) has a frequency of 33%. With two exceptions, frequencies below 30% in column 6 correspond to late datings in column 2. One exceptional entry is for a “so-so” poem of the middle period (Chr & Sat), which has a frequency of 27%. The other is for Cynewulf, a “best” poet in the middle period of the traditional chronology

Dating criteria for Old English poems

257

Table 3

1 work

2 date

3 quality

4 two-word a-verses

5 total a-verses

6 %

GenA Dan Beo Ex Cyn And Met Brun Jud Chron Mald

early early early early middle middle late late late late late

so-so so-so best best best good bad best best good bad

286 74 390 52 200 154 133 6 28 7 3

730 173 1027 134 679 375 479 23 84 29 66

39 43 38 39 29 41 28 26 33 24 5

GuthA Chr & Sat GuthB Phoen Ps 118

early middle late late late

bad so-so best best bad

95 68 44 64 14

277 251 157 219 60

34 27 28 29 23

who is now placed with “best” poets of the late period by three independent measurements.

4.3 Revisions to the traditional chronology Unexpected results for Cynewulf impose an urgent explanatory demand, since this poet is a superb artist by any number of criteria and few verses in the manuscripts of his signed poems present significant problems of scansion. As it turns out, the results might be squared with the traditional consensus in its more cautious formulation (Table 1, column 2). The Andreas poet is placed below Cynewulf because the center of the 9th-century

258

Geoffrey Russom

range for Andreas is 850 and Cynewulf’s poems are dated prior to 850. If we place Andreas in the earlier part of its range, closer to Exodus, Cynewulf will be chronologically adjacent to the late “best” poets. In this case as in others, purely metrical evidence converges with Fulk’s metrical-linguistic evidence. Fulk suggested only two major revisions to the traditional chronology, and one of these was to place Cynewulf’s poems after Andreas (Fulk 1992: 348). Conner (1996) has recently argued that a source text used by Cynewulf cannot be earlier than the mid-ninth century, and Fulk (1996: 16) concurs. Cynewulf must apparently be placed even later than 850. Fulk’s second major revision was to place Beowulf first among the early poems. The Beowulf poet’s splendidly high percentage of simplex D in relation to simplex E, which required extensive employment of archaic syntax, might be taken as support for this revision (see Table 2, column 8).

5.

Conclusions and envoi

The variety of independent measurements supporting the traditional chronology, as amended by Fulk, would suffice even if these measurements were unmotatived. Recall that expert merit rankings based on the number of anomalous verses have been tested against the independent distribution of type A3 before being used for analysis of simplex C, D, and E, each of which has its own, independent distribution in sixteen verse selections. These measurements are motivated, however, by straightforward expectations about metrical history (14) and by the two fundamental principles of the word-foot theory (7), which provide independent assessments of metrical complexity. A crucial aspect of complexity for the analysis, deviation from the metrical norm in the number of stressed roots per verse, cannot even be described within Sievers’s system, which notates subordinate stress but does not distinguish stressed secondary roots from the stressed medial syllables of long words in simplex C, D, and E verses. 14 The success of the analysis thus provides additional support for the hypothesis that metrical feet employed by Germanic poets are projections from the domain of morphology into the domain of verse form. A final question: why do even small poems behave as expected? Consider for example the striking contrasts in Table 2 that distinguish small

Dating criteria for Old English poems

259

“best” poems of the late period (Brun, Jud, GuthB) from small “bad” poems of this period (Met, Mald, Ps118). The extreme case is Brunanburh (Brun), a “best” poem with a single relevant verse that has the most highly valued pattern. The principle of metrical variety (Russom 1987: 25), which falls under the heading of interest (Hanson and Kiparsky 1996: 295), provides a straightforward answer. It has been clear for some time that Old English poets avoid repetition of types and subtypes in close proximity (Lehmann 1975), returning frequently to metrical norms and distributing complex variants evenly throughout their works to keep the effort of intuitive scansion within reasonable limits (Russom 1987: 146–149). Such metrical orchestration makes Old English poems especially valuable as sources of information about the history of English.

Appendix: Verse lists for Table 2 A. List of simplex D a-verses and b-verses (fourth column): GenA 88a, 133a, 874a, 1419a, 1653a, 2106a, 2123a, 2610a, 211b, 926b, 1393b, 1453b, 1691b, 2501b; Dan 354a, 129b, 148b, 163b, 165b, 395b; Beo 148a, 291a, 351a, 412a, 428a, 546a, 581a, 663a, 947a, 1021a, 1083a, 1183a, 1203a, 1418a, 1601a, 1653a, 1690a, 1759a, 2062a, 2159a, 2350a, 2424a, 3142a, 30b, 47b, 58b, 170b, 216b, 229b, 270b, 294b, 312b, 345b, 371b, 456b, 500b, 521b, 620b, 652b, 778b, 847b, 1069b, 1096b, 1210b, 1321b, 1498b, 1563b, 1769b, 1851b, 1860b, 1871b, 2026b, 2052b, 2101b, 2259b, 2381b, 2484b, 2537b, 2603b, 2620b, 2809b, 2889b, 2912b, 2985b, 3002b, 3005b, 3080b, 3107b; Ex 45a, 62a, 392a, 453a, 478a, 519a, 49b, 499b, 184b; CYN: ChristB 536a; El 523a, 540a, 1062a, 1096a, 1169a; Jul 266a, 306a; ChristB 500b, 635b, 808b; El 242b, 391b, 538b, 850b, 891b, 1087b, 1114b, 1220b, 1257b; Jul 261b, 689b; And 393a, 437a, 443a, 491a, 591a, 860a, 1300a, 1477a, 1508a, 307b, 340b, 397b, 570b, 595b, 629b, 657b, 776b, 811b, 1062b, 1163b, 1409b, 1431b, 1574b, 1709b; Met 21.30b; Brun 18b; Jud 142a, 334a, 85b; Chron 6.3b; Mald 281b; GuthA 162a, 724a, 267b; Chr & Sat 86a, 157a, 400a, 526a, 573a, 152b, 172b, 194b, 299b, 394b, 424b, 576b, 646b, 720b; GuthB 856a, 1038a, 1056a, 1063a, 819b, 1209b, 1217b, 1241b, 1336b, 1379b; Phoen 59b, 477b, 620b. B. List of simplex E a-verses and b-verses (sixth column): GenA 180a, 890a, 956a, 1071a, 1081a, 1278a, 1413a, 1654a, 1698a, 1737a, 1858a,

260

Geoffrey Russom

2002a, 2091a, 2093a, 2146a, 2232a, 2409a, 2445a, 2558a, 2559a, 2620a, 2657a, 2670a, 2720a, 2831a, 2933a, 89b, 182b, 191b, 918b, 981b, 1161b, 1285b, 1447b, 1497b, 1660b, 1662b, 1676b, 1762b, 1970b, 2062b, 2120b, 2131b, 2347b, 2464b, 2544b, 2598b, 2662b, 2664b, 2722b; Dan 373a, 475a, 478a, 616a, 484b, 333b, 583b, 700b, 734b; Beo 50a, 179a, 547a, 673a, 741a, 815a, 888a, 1091a, 1268a, 1536a, 1697a, 2447a, 2464a, 2597a, 3075a, 3170a, 33b, 159b, 429b, 802b, 899b, 1199b, 1408b, 1565b, 1581b, 1690b, 1767b, 1944b, 1953b, 1961b, 1972b, 1973b, 2337b, 2789b, 2817b, 2832b, 2979b, 3028b, 3127b, 3145b; Ex 101a, 111a, 277a, 300a, 490a, 56b, 73b, 213b, 321b, 488b; CYN: ChristB 515a, 529a, 599a, 741a, 835a, 845a; El 393a, 465a, 580a, 599a, 927a, 994a, 1002a, 1068a, 1110a, 1259a, 1293a; Jul 339a, 642a, 688a; ChristB 481b, 704b, 743b, 755b, 765b; El 12b, 99b, 484b, 486b, 885b, 899b, 1197b, 1273b; Jul 7b, 10b, 61b, 386b, 626b, 730b; And 144a, 395a, 481a, 506a, 616a, 725a, 956a, 1095a, 1174a, 1389a, 1543a, 1572a, 1590a, 1608a, 1610a, 1667a, 44b, 277b, 459b, 493b, 568b, 574b, 576b, 623b, 655b, 735b, 771b, 887b, 896b, 949b, 957b, 1223b, 1450b, 1557b, 1672b, 1713b; Met 6.14a, 26.62a, 26.84a, 5.7b, 7.16b, 8.45b, 9.6b, 10.33b, 18.6a, 20.64b, 10.35b, 10.42b, 13.33b, 20.214b, 20.216b, 22.33b, 20.221b, 26.24b, 26.91a, 26.31b, 29.72b; Jud 78a, 86a, 216a, 313a; 45b; Chron 4.34a, 2.12b, 4.17b, 4.20b; Mald 26a, 47a, 73a, 97a, 146a, 134b, 154b; GuthA 110a, 181a, 189a, 204a, 207a, 254a, 615a, 646a, 799a, 65b, 491b, 525b, 555b; Chr & Sat 71a, 133a, 212a, 376a, 413a, 561a, 677a, 105b, 118b, 267b, 284b, 292b, 327b, 353b, 387b; GuthB 1061a, 1138a, 1139a, 1363a, 1048b, 1316b; Phoen 264a, 275a, 327b; Ps118 146.3a. C. List of simplex C a-verses and b-verses (ninth column): GenA 125a, 153a, 203a, 227a, 927a, 997a, 1023a, 1214a, 1237a, 1241a, 1288a, 1324a, 1338a, 1437a, 1452a, 1461a, 1670a, 1694a, 1704a, 1706a, 1826a, 1853a, 1884a, 2066a, 2071a, 2110a, 2277a, 2351a, 2381a, 2444a, 2530a, 2536a, 2637a, 2740a, 2780a, 2848a, 2883a, 890b, 1059b, 1088b, 1234b, 1582b, 1616b, 1647b, 1761b, 1775b, 1791b, 2029b, 2144b, 2333b, 2435b, 2592b, 2609b, 2804b, 2842b, 2864b, 2873b; Dan 31a, 141a, 212a, 393a, 489a, 520a, 543a, 572a, 586a, 617a, 715a, 295b, 304b, 306b, 380b, 413b, 519b, 521b, 550b, 637b, 689b, 763b; Beo 3a, 113a, 221a, 258a, 274a, 363a, 416a, 461a, 471a, 565a, 627a, 634a, 649a, 708a, 916a, 982a, 1026a, 1076a, 1135a, 1244a, 1294a, 1309a, 1406a, 1414a, 1596a, 1640a, 1685a, 1804a, 1828a, 1829a, 1956a, 1988a, 2272a, 2298a, 2374a, 2382a, 2435a, 2569a,

Dating criteria for Old English poems

261

2617a, 2683a, 2823a, 2828a, 2841a, 2850a, 2857a, 2871a, 2939a, 3004a, 3013a, 3109a, 206b, 644b, 1490b, 1659b, 1772b, 2292b, 2329b, 2665b, 2807b, 2888b, 2949b, 2964b, 3122b; Ex 8a, 112a, 123a, 264a, 324a, 364a, 393a, 504a; CYN: ChristB 470a, 491a, 505a, 705a, 729a, 786a, 823a, 824a; El 18a, 138a, 169a, 177a, 186a, 245a, 277a, 299a, 323a, 389a, 484a, 732a, 738a, 744a, 820a, 931a, 942a, 983a, 1024a, 1109a, 1122a, 1124a, 1173a, 1233a, 1274a; Fates 3a, 71a, 85a; Jul 37a, 94a, 115a, 133a, 152a, 166a, 204a, 206a, 207a, 233a, 248a, 250a, 338a, 340a, 373a, 382a, 394a, 456a, 477a, 572a, 663a, 702a, 727a; ChristB 520b, 761b; El 25b, 130b, 153b, 536b, 611b, 652b, 673b, 689b, 930b, 990b, 1018b, 1064b, 1106b, 1201b, 1234b; Jul 149b, 160b, 179b, 205b, 242b, 247b, 339b, 449b, 484b, 543b, 614b, 653b, 662b; And 59a, 186a, 215a, 218a, 220a, 230a, 378a, 386a, 424a, 464a, 704a, 763a, 820a, 857a, 876a, 992a, 1032a, 1034a, 1141a, 1349a, 1354a, 1370a, 1392a, 1394a, 1544a, 1571a, 1614a, 1620a, 1685a, 86b, 312b, 649b, 699b, 805b, 1068b, 1177b, 1592b; Met 0.6a, 1.21a, 1.40a, 1.73a, 1.79a, 9.25a, 9.26a, 11.8a, 11.9a, 11.34a, 12.20a, 13.54a, 15.8a, 18.7a, 20.143a, 20.168a, 20.191a, 20.192a, 20.199a, 20.203a, 20.212a, 20.220a, 22.41a, 22.63a, 24.49a, 25.28a, 25.65a, 25.66a, 26.36a, 27.23a, 2.2b, 3.4b, 7.50b, 11.26b, 13.28b, 16.22b, 21.24b, 25.32b, 28.42b, 29.82b; Jud 40a, 172a, 242a, 73b, 108b, 178b, 231b, 283b, 295b, 314b; Chron 2.10a; Mald 65a, 116a, 122a, 143a, 322b; GuthA 77a, 94a, 109a, 149a, 166a, 194a, 219a, 342a, 385a, 415a, 437a, 458a, 471a, 484a, 544a, 589a, 621a, 628a, 664a, 758a, 818a, 31b, 38b, 81b, 121b, 128b, 150b, 168b, 208b, 209b, 404b, 427b, 496b, 505b, 578b, 604b, 616b, 645b, 649b, 655b, 658b, 671b, 672b, 679b, 689b, 720b, 778b, 802b; Chr & Sat 54a, 62a, 171a, 328a, 363a, 381a, 519a, 594a, 620a, 624a, 643a, 220b; GuthB 821a, 822a, 845a, 899a, 907a, 935a, 975a, 1004a, 1099a, 1122a, 1167a, 1180a, 1296a, 1321a, 1359a, 836b, 919b, 939b, 1147b, 1252b; Phoen 8a, 193a, 207a, 244a, 254a, 368a, 431a, 453a, 531a, 631a, 201b, 395b, 441b; Ps118 28.2a, 44.3a, 89.1a, 111.3a, 142.3a, 143.1a, 143.2a, 152.3a, 160.3a, 161.1a, 93.1b.

Notes 1.

For detailed criticism of this and other late datings of Beowulf, see Fulk (1992), which responds to influential criticisms of linguistic dating criteria in Amos (1980).

262 2.

3.

4.

5.

6.

7.

8.

Geoffrey Russom Poem titles are abbreviated as follows in tables and brief references: Andreas (And), Beowulf (Beo), The Battle of Brunanburh (Brun), Christ and Satan (Chr & Sat), Daniel (Dan), Exodus (Ex), Genesis A (GenA), Guthlac A (GuthA), Guthlac B (GuthB), Judith (Jud), The Battle of Maldon (Mald), The Meters of Boethius (Met), The Phoenix (Phoen), and Psalm 118 of the Paris Psalter (Ps118). The signed poems of Cynewulf, analyzed as a group, are abbreviated as “Cyn.” Poems 2, 3, 4, and 6 from the Anglo-Saxon Chronicle, analyzed as a group, are abbreviated as “Chron.” In word-for-word translations, the following abbreviations are employed: inf(initive), masc(uline), pres(ent), pret(erite), refl(exive), nom(inative), gen(itive), dat(ive), acc(usative), s(in)g(ular), pl(ural). Persons are abbreviated as 1, 2, 3. When a verse cannot be translated comprehensibly in isolation, I add translated material from adjacent verses, placing such material between parentheses (e.g. in example 9). Examples from Beowulf are cited from Klaeber (1950), with systematic suppression of sentence-initial capitalization and line-final punctuation. Line-number references to other poems follow Krapp and Dobbie (1931–1953), from which example 9 is cited. See Dresher and Lahiri (1991), where the trochaic pattern is represented as the phonological ideal for Germanic words. For our purposes it does not matter whether Dresher and Lahiri’s Germanic Foot is basic, as they claim in this article, or derived, as suggested in Hanson and Kiparsky (1996: 305). To qualify as a plausible analogue of a two-word A1 verse, a prose string must constitute a complete syntactic constituent. Otherwise, it would fail to meet the metrical requirement of syntactic integrity imposed on the verse (Russom 1998: 30). With regard to Ælfric’s so-called rhythmical prose, see Russom (1987: 134). In this type E variant, the light disyllable wegas, which has a short root syllable followed by an unstressed syllable, occupies a single strong metrical position by resolution, becoming metrically equivalent to sum and bearn in the examples of type E cited as 5 and 6 (Russom 1998: 103–105). Frequent employment of these strategies in Beowulf provides further evidence that the poem is in fact early, as argued in Fulk (1992). Within the brief compass of Beo 4a–11b, punctuated as a single complex sentence in Klaeber (1950), there is paratactic juxtaposition of clauses with unexpressed pronouns at 6a, 8a, and 8b, for example. This type of knowledge or realization was probably unconscious, surprising though that may seem. With respect to unconscious control of metrical constraints, see Jakobson (1979). Observe that the poet’s own sense of metrical complexity attests an intention to conform to traditional standards. Hutcheson (1995: 32–38) gives rankings for GenA, Beo, Ex, Cyn, And, Brun, Jud, Chron, Mald, and Phoen. For Met, Dan, Chr & Sat, and Ps118 see Fulk (1992: 35, 65, 396, 410). Rankings for GuthA and GuthB are based on observations of Roberts (1979: 61–62), though she herself is largely non-judgmental. Since the merit rankings provide only one measurement of poetic talent (conformity to traditional

Dating criteria for Old English poems

9.

10.

11.

12.

13.

263

standards of versecraft), I keep them in quotes to distinguish them from rankings based on overall literary merit. It seems clear, however, that all poets discussed in this article tried to conform to traditional standards. Deviation from these standards cannot be attributed to a Romantic-style agenda of innovation (Fulk 1992: 33–35). Sievers also classifies a low-frequency variant with a bisyllabic compound in final position as type A3, but I exclude this variant from consideration. The counts include A3 variants in which a finite main verb with subordinate phrasal stress is treated as equivalent to an unstressed function word, appearing without alliteration to the left of the verse-final stressed word (Russom 1998: 128–135). Statistics cited in this paper are drawn from a database incorporating an electronic transcript of Krapp and Dobbie (1931–1953) by Patricia Bethel and O.D. MacraeGibson. Thanks are due to Oxford Text Archive for making the transcript available to me. To obtain comparable frequencies for type A3, I have systematically excluded from verse counts the so-called hypermetrical verses and verses with no determinate scansion. Most of the latter contain foreign proper names whose linguistic significance for Old English poets is unknown (Fulk 1992: 84). The “size” column of Table 1 provides the number of normal (nonhypermetrical) verses with determinate scansion. Statistics in Table 2 are restricted by definition to verses with determinate scansion. The counts for simplex D and E in Table 2 exclude low-frequency variants with extrametrical unstressed words that increase metrical complexity in these types, making it more difficult to assess the kinds of metrical complexity at issue here. Variants with extrametrical words are included in the count for simplex C, however, since in this type additional unstressed words do not increase complexity and most instances contain at least one such word (Russom 1987: 38). Frequencies for simplex C may accordingly be inflated in relation to frequencies for simplex D and E. Systematically excluded from the count are verses with compounded or prefixed simplexes, including inflected infinitives, since the poet’s placement of such constituents is severely limited by the irrelevant constraints on extrametrical syllables and by purely grammatical norms for ordering of constituents in compounds. The rather high relative frequency of 60% for simplex C in “fine” poems can be attributed to the usefulness of this pattern and perhaps to necessarily selective treatment of extrametrical syllables in verse counts (see footnote 11). Other measurements, notably the proportion of a-verses to b-verses, show that simplex C is more deviant than simplex D. Among the poems classified here as late, Fulk (1996: 17) posits a Cynewulfian group that includes GuthB and Phoen, a later Alfredian group including Met and Jud, and a still later group including Mald, Ps118, and the eleventh-century Chronicle poems. The tenth-century Chronicle poems are placed in the Alfredian group. Given these groupings, the metrical faults of Maldon can be reconciled more easily with its rhetorical merit (Russom 1978); and the high frequency of

264

14.

Geoffrey Russom verses with normal scansion in Judith, the last “best” poem, seems a truly heroic achievement. Medial syllables of long words in our simplex variants receive subordinate stress because they are heavy, with the medial vowel followed by two consonants (Dresher and Lahiri 1991: 260). The special behavior of simplex variants cannot be attributed to stress differences captured by Sievers’s notation. Simplex C and D variants discussed in this article should be distinguished from light C and D variants in which the medial syllable of the long word is light and (I think) unstressed (Russom 1998: 114–116, 167–169). Light C and D variants also contain long simplexes, but are assigned to distinct metrical patterns within the word-foot framework (x/Sxx and S/Sxx, respectively, as distinct from x/Ssx and S/Ssx). There are no light E variants.

References Amos, Ashley Crandall 1980 Linguistic Means of Determining the Dates of Old English Literary Texts. Cambridge, MA: Medieval Academy of America. Bethel, Patricia, and O.D. Macrae-Gibson No date Electronic transcription of Krapp and Dobbie (1931–1953). Oxford Text Archive. Bjork, Robert E., (ed.) 1996 Cynewulf: Basic Readings. New York: Garland. Bliss, Alan J. 1967 The Metre of Beowulf, revised edition. Oxford: Blackwell. Cable, Thomas 1981 Metrical Style as Evidence for the Date of Beowulf. In: Colin Chase (ed.), The Dating of Beowulf, 77–82. Toronto: University of Toronto Press. Conner, Patrick W. On Dating Cynewulf. In: Bjork (1996: 23–53). Dresher, B. Elan and Aditi Lahiri 1991 The Germanic Foot: Metrical Coherence in Old English. Linguistic Inquiry 22: 251–286. Fulk, Robert D. 1992 A History of Old English Meter. Philadelphia: University of Pennsylvania Press.

Dating criteria for Old English poems 1998

265

Cynewulf: Canon, Dialect, and Date. In: Bjork (1996: 3–21).

Hanson, Kristin, and Paul Kiparsky 1996 A Parametric Theory of Poetic Meter. Language 72: 287–335. Hayes, Bruce 1983 A Grid-based Theory of English Meter. Linguistic Inquiry 14: 357–393. Hutcheson, B. Rand 1995 Old English Poetic Metre. Cambridge: D.S. Brewer. Jakobson, Roman 1979 On the So-Called Vowel Alliteration in Germanic Verse. Selected Writings V: On Verse, Its Masters and Explorers, 433–485. The Hague: Mouton. Kiparsky, Paul 1977 The Rhythmic Structure of English Verse. Linguistic Inquiry 8: 189–247. Klaeber, Frederick (ed.) 1950 Beowulf and the Fight at Finnsburg. Third edition. Lexington: D.C. Heath. Krapp, George Philip, and Elliot Van Kirk Dobbie 1931– The Anglo-Saxon Poetic Records. Six volumes. New York: Columbia 1953 University Press. Lehmann, Ruth P.M. 1975 Broken Cadences in Beowulf. English Studies 56: 1–13. Roberts, Jane 1979 The Guthlac Poems of the Exeter Book. Oxford: Clarendon Press. Russom, Geoffrey 1978 Artful Avoidance of the Useful Phrase in: Beowulf, The Battle of Maldon, and Fates of the Apostles. Studies in Philology 75: 371–390. 1987 Old English Meter and Linguistic Theory. Cambridge: Cambridge University Press. 1998 Beowulf and Old Germanic Metre. (Cambridge Studies in Anglo-Saxon England 23.) Cambridge: Cambridge University Press. Sievers, Eduard 1893 Altgermanische Metrik. Halle: Niemeyer.

How much shifting actually occurred in the historical English vowel shift? Robert Stockwell

The traditional picture of the English vowel shift that took place in the south of England between 1200 and 1600 was represented as in Figure 1 by Otto Jespersen (1909: 232), who named it the “Great Vowel Shift”: ai ←

i: ↑ e: ↑ E: ↑ a:

u:→ au

↑

o:

↑

O:

Figure 1

This says, each lower vowel moves up to the next higher position, and the highest vowel becomes a diphthong. Jespersen dates the change to around 1450. A quite non-traditional picture, the one for which I shall argue, is shown in Figure 2: i: → ↑ e: ej← E: ↑ a:

u:

@j ↓ ↓ ˆj ↓ aj

@w ↓ ↓ ˆw ↓ aw

Center Drift Figure 2

←↑

Upper Half

o: O:→ow ↑ ¡(:)

Lower Half

268

Robert Stockwell

In Figure 2, mergers are shaded, diaphones are inside the box, and true chain shifts are shown in the upper left and right peripheries. (I shall return to the term “diaphone” shortly: for now, think of them as allophones which are non-contrastive by virtue of their distribution across dialects.) The claim embodied in Figure 2 is an elaboration of a claim that Donka Minkova and I made (Stockwell and Minkova 1988), namely that the lower half of the traditional Southern British English vowel shift is a series of mergers among the available contrastive units, ultimately favoring [ej] and [ow] as the only occupants of that space. I shall call this “implosion”, whereby a set of two or three or more contrastive entities within a substantially overcrowded vowel space merge to a single unit. Obviously “implosion” is just a mnemonic, not a characterization of a process: it is the opposite of a chain shift because nothing is displaced and everything is merged. I claim that the fundamental condition for a chain, namely the no-merger condition, is only supported among the vowels with the feature [–LO]. The most conspicuous part of the vowel shift, which I shall call the “center drift”, is not a chain at all. “Center drift” is the diphthongization, centralizing, and lowering (in either order, chronologically) of [i:] and [u:] to some variant of [aj] and [aw]. The defining feature of sound changes which are arranged in a chain is this: Any two adjacent contrastive entities – any two links in the chain – must move in lockstep, without merger. They always maintain their distance and their functional contrast. But in these famous vowel-shift instances of drifting down the center, on the contrary, nothing is ever displaced, no merger occurs, and no phonological change has occurred. All the variants, top to bottom, co-exist. They are non-contrastive in all dialects and all accents. The only thing that suggests that each step downward calls for analysis as a new phoneme or phoneme cluster is distinctive feature theory. This is a kind of hangover from structuralist days and the notion “once a phoneme always a phoneme”. That is, once the diphthong moves from [+HI] to [–HI], or from [–LO] to [+LO], features that are distinctive elsewhere in the vowel system, one might argue that they must be taken as distinctive here too. But I see no reason to accept that consequence of the theory. We know that in childhood all speakers build their own grammars. When it comes to learning language, everyone is an island. From that point of view, when people are still immature and haven’t traveled much or mixed in wider social circles, [ˆj] and [ˆw] (Canadian or Virginia pronunciations) may not

How much shifting actually occurred in the historical English vowel shift?

269

immediately be perceived as allophonically related to [aj] and [aw]. But the island joins the mainland as it matures. As we expand our linguistic horizons and talk with Australians and Cockneys and Canadians and Virginians and Philadelphians, and as we go to school and learn that the front diphthongs are called “long I” (for good non-phonetic reasons, like regularities in morphophonemic alternations), and we grow up using dictionaries where the front diphthongs are marked as “long I”, these dialect variants become functionally single phonemes. (The back diphthongs are not marked or referred to as “long U”, because of the “OU” spelling that French scribes introduced in the 12th and 13th centuries, leaving “long U” as the name for the earlier IW vowel, the vowel of cute, beauty, feud). If we prefer to keep “phoneme” for a narrower sense, we can use Hans Kurath’s term “diaphone” (Kurath 1961) – i.e., dialect variants that function like allophones. The two most important criteria of phonemic identity, namely psychological unity and functional equivalence, are found in these diaphones at least among sophisticated adult speakers. “Canadian raisings” are hardly noticed by the rest of us because they are increasingly familiar diaphones of /aj/ and /aw/. It has been claimed – I am only one of many – that the center drift through the sequence [ij] to [@j] (pace Lass) to [ˆj] to [aj] to [Oj] (Cockney) or [a:] (Southern American) – and similarly for the corresponding back diphthong – is plausibly motivated in the dissimilatory stages (i.e., as it gets further from [ij] and closer to [Oj]) by perceptual optimization of the diphthong followed at the assimilatory stage by articulatory optimization (producing the Southern long monophthong [a:] for [aj] in words like mine, tide). Across a much wider range of phenomena, it is the main theme of Boersma (2000), who attributes the notions of “minimization of articulatory effort” and “minimization of perceptual confusion” to Passy (1891). These notions are made much more explicit in Boersma’s study than in the numerous previous studies where they have been invoked to account for the tendencies of high vowels to diphthongize and of wide diphthongs to become narrow diphthongs or monophthongs. I claim that it is a phonetically conditioned low-level drift, not in principle different from the development of intervocalic flapping of /t/ or /d/ in latter-ladder, or of nasalization in can’t and don’t. This drift could of course end up being rephonologized, but it hasn’t been rephonologized yet, at least not in mainstream varieties of English.

270

Robert Stockwell

With regard to the formulation of center drift, I shall put aside as irrelevant to my focus here one aspect of the formulation of the details of the drift, namely: did the new diphthongs first drop down one step and then move to center, or first move toward center and then drop down? The argument has recently been vigorously revived in Lass 1999. His account fails to explain why the putative [ej], derived from [ij] in Lass’s first stage of the vowel shift, did not merge with the diphthong [ej], both inherited and borrowed. This diphthong was extremely salient in that area of the vowel space. A corresponding argument, point for point, applies to the putative [ow] from [uw], except we never can say anything about it on the basis of early phonetic descriptions because of the French spelling . Let me also make clear that in raising the notion of diaphones I am not turning the clock back to the early structuralist notion of “overall patterns”. The overall pattern allowed all accent differences to be represented with a selection of symbols from a single very rich and reasonably coherent set of benchmarks which were attributed to everyone’s phonological competence equally, whether they made a given specific contrast or not. Overall pattern notation is still in use, in an only slightly modified form, by Labov (1994), whose work is often misunderstood because the Trager-Smith system is now unfamiliar and not generally taught. It nevertheless has an honorable history that includes such names as Batchelor (1809), Sweet (1891), Bloomfield (1933), Trager and Smith (1953), and Hockett (1955). Modern scholars commonly forget that it was the surface phonemic notation system, with minor modifications, used by Chomsky and Halle (1968) as the output of their morphophonemic (i.e., abstract phonological) rules. With respect to the vowel shift, my view favors the reality of received orthography and the good sense of lexicographers. I take it, following this line of thought, that the most famous part of the vowel shift, the only part – center drift – that it shares with other West Germanic languages, is diaphonic drift, not a chain shift. “Long I” and “Long U” are the right designations for these entities, and they cover a range of instantiations from Edinburgh to Baton Rouge. This is not only true as a diachronic claim, I think it is true also as a synchronic claim about the reality of these entities in the perception of sophisticated speakers: but for my argument against “chaining” to go through, it only has to be true as a diachronic claim. Now I would like to turn to my second main point, an anti-chaining view of the bottom half of the vowel shift that Minkova and I argued for in 1985.

How much shifting actually occurred in the historical English vowel shift?

271

We apparently did not make it clear, or if clear then not convincing, since half-a-dozen full-scale histories of English, or more, have been published subsequently and none of them have had their vision altered by the new light we offered. The metaphor of a set of links in a chain, as noted above, entails lockstep without merger. The metaphor proposed here, to explicate what Minkova and I were suggesting, entails multiple mergers to a single focal point. I think it can legitimately be viewed as a form of optimization. As implied earlier, it is motivated by a universal principle which says that over-crowding an area of vowel space is to be avoided, i.e., too many similar-sounding nuclei in the same vowel space create a non-optimal situation. The overcrowding in this instance resulted from the introduction of highly-similar [ej]-type and [ow]-type nuclei from a variety of sources: 1. 2. 3. 4. 5.

From Old English with vocalization or epenthesis in the sequence V + [x/ç], resulting in [Vw] or [Vj] From Scandinavian loans From French loans From Middle English Open syllable lengthening of the three [–HI] short vowels. From the words like great, break, steak, yea. The rest of the words in this set are a special problem, really for everybody, the famous meet-meat-mate problem.

These were powerful forces for merger. We are talking about very large numbers of words containing diphthongs similar enough to [ej] and [ow] such that overcrowding could destabilize the inherited contrasts. In Figure 3 I have tried to represent the implosion picture for low and mid front vowels, and for low and mid back vowels. Below are lists of representative examples. Some explanation of notation is needed. Where the modern English form is shown in parentheses, it means that the example was at one point in its history a proper Middle English example which underwent further (and for this argument irrelevant) development or dialect interference that removed it from the merger class to which it earlier belonged. The dotted g of OE editors is here written after front vowels. The ambiguous symbol is used after back vowels when it is responsible for later lowering of the vowel, as in daughter, which

272

Robert Stockwell

Front (low to mid) ej

← eC@

↑ eg

OE brecan

Back (low to mid) OE nosu

oC@

ow

↑

OE bregdan eçC

OE ehta

OE boga

og

OE mægden

OE growan

ow

o:

OE knawan

↑

↑

OE ba:t

Ow

O:

↑ ei(g) ← æi(g)

↑ æ@

OE great (PGmc *grauto-)

↑ OE a:gan

O:g

Figure 3

otherwise should be [do:tr]. Middle English Open Syllable Lengthening (MEOSL) added many more examples: all of the and words like bake and nose, discussed below. Length of vowels with glides is not indicated except where absolutely necessary to identify the items in question. Length was always assimilated to the coda glide of the stressed syllable – i.e., -V(:).wV → -Vw.V, and –V(:).jV → -Vj.V. A putative contrast between long vowel plus glide, and short vowel plus glide, is completely spurious at least in the history of English, because the two types, if they once existed before resyllabification took place, always develop identically afterwards. The story in Scandinavian is quite different but clearly irrelevant here. In addition to the direct reflex relations shown above, there are, as noted above, the results of MEOSL; these new long vowels merged immediately with the phonetically most similar diphthongs. Examples: ache, ale, bake, bale, bane, bathe, blade, blaze, cake, and dozens more; and similarly nose. The precise nature of the intermediate stage(s) of this merger is not clear. It cannot have been simple raising because that would merge with the inherited OE ‘long æ’, which is the only vowel in this area that did NOT merge to [ej] – cf. OE hælan, nædl, MnE heal, needle. The best guess would appear to be that since the lengthened vowel would have started out

How much shifting actually occurred in the historical English vowel shift?

273

as [æ:] or even more probably [æ@] and would have been an orphan in the system (there being, by this time, no in-gliding diphthongs remaining, though they were common in OE), it fell in with the out-gliding [æj] and merged with [ej] accordingly. From Old English Mergers to [ej]

Mergers to [ow]

bow < OE boga (vocalization) break < OE (= eC@ > [ejC]) MEOSL braid < OE ( = [ej]) close < OF closen brain < OE (= [ej]) dough < OE , ME [dO:x] (glide epenthesis) clay < OE (= [æj] > [ej]) (daughter) < douZter < OE dohtor) (glide epenthesis) day < OE (= [æj] > [ej]) flown < OE flogen (vocalization) eight < OE (= [eç > ejç > ej]) grow < OE (resyllabification of /w/ from onset to coda) fey < OE (= [æj] > [ej]) know < O:w < OE (glide assimilation and resyllabification) (fight) < OE nose < O: < OE MEOSL (= [eç > ejç > ej]) gray < OE (=[æj] > [ej]) (nought) < ME , OE (glide epenthesis) (high) < hejZ < OE he:h (glide ep- own week depended on the assumption that “short i” was at approximately the same height as “long e”, and that “short e” was at the same height as “long open e”. That is the clear and necessary implication of Lass’s own account of OSL as it applies to high and mid vowels (CHEL II: 48):

278

Robert Stockwell

i:

i

u

u:

e:

e

o

o:

E:

O:

Figure 7

He simply creates a rule of “lowering” applying to the high and mid short vowels. It is completely unmotivated, unless it has the phonetic content of Figure 8: i:

u: I

U

e:

o: E

E:

O O:

Figure 8

Lass rejects the phonetic implications of Figure 8 in favor of an arbitrary rule of lowering when lengthened. The only reason for this rejection is to preserve his assessment of John Hart’s excellence as a phonetician. The price he pays for rejecting the phonetic motivation seen in Figure 8 is to call all of these lengthenings and shortenings – examples like week, sick, wood, door, evil, etc. (see Trnka 1959 or Stockwell 1961 for complete lists) – “sporadic”, which is surely a court of last resort, since when lengthened, sporadically or regularly makes no difference, they merge with the lower, not the higher, long vowel. This fact is not sporadic, even if the lengthenings and shortenings themselves are.

How much shifting actually occurred in the historical English vowel shift?

279

Finally, by way of summary, note especially that my claim is not that in the vowel space from low to mid, both in the front and in the back, there was no raising to [ej] and [ow], – there is no denying that implosion in some instances involved raising – but rather that such raising resulted in merger with [ej] and [ow], which were nuclei that already existed in substantial numbers. They did not merge with the next higher vowel of an on-going chain. There was a complete destabilization and ultimately collapse of all contrasts in the vowel space that might be called long low and long mid front; and similarly a collapse in the corresponding back vowel space. Furthermore it is unlikely to be accidental that the diphthong which was the target of the merger, conspicuously in the front and to a lesser extent in the back, had already been borrowed in large numbers from Scandinavian and French, though there are significant numbers of native examples also. Nonetheless, the imports, rather like Honda and Toyota for American auto manufacturers, provided a new standard of excellence for [–HI] vowels When [ej] and [ow] became so common, it is not surprising that the only similar-sounding vowels ([e:] and [o:]) in their respective areas moved higher, becoming [i:] and [u:] as in beet and boot. In these instances the push kind of motivation favored by Karl Luick (1896), Richard Carter (1975), and Roger Lass (see Lass 1987: 226–227 for a succinct presentation of the evidence) makes some sense. Whether this development justifies positing a push chain (with only one link) depends on whether the highest vowels had already become diphthongs. Stenbrenden (2001) reports evidence from the Linguistic Atlas of Early Middle English (in progress at Edinburgh University) for diphthongization of an earlier date (numerous spellings for Middle English reflexes of Old English ‘long i’ and of other Old English spellings that result in Middle English ‘long i’ forms). These apparently antedate the evidence for raising (seemed written symed). If this is so, it is damaging to any push-chain theory. In Stockwell-Minkova (1988), we pointed to other serious data problems in the Luick-Lass-Carter argument for a push chain, that argument being based on some Scots and Northern English dialects. Frankis (1986) has pointed to other Germanic data which, at the very least, argue for a more conservative position than the push-chain theory allows. It seems doubtful whether this question can ever be resolved satisfactorily, but either resolution of it – push or drag – fails to impact the main argument about mergers made above.

280

Robert Stockwell

References Batchelor, T. 1809 An Orthoepical Analysis of the English Language. London: Didier and Tebbett. Bloomfield, Leonard 1933 Language. New York: Holt Boersma, Paul 2000 Functional Phonology. Formalizing the interactions between articulatory and perceptual drives. Netherlands Graduate School of Linguistics. The Hague: Holland Academic Graphics. Carter, Richard 1975 Some theoretical implications of the Great Vowel Shift. In: Didier L. Goyvaerts and Geoffrey K. Pullum (eds.), 369–376. Essays on The Sound Pattern of English. Ghent: Story Scientia. Chomsky, Noam, and Morris Halle 1968 The Sound Pattern of English. New York: Harper and Row. Donka Minkova and Robert P. Stockwell 1991 The Early Modern English vowels, more o’ Lass. Diachronica 3.1–18. Frankis, John 1988 The Great Vowel-Shift and Other Vowel-Shifts. In: G. Nixon and J. Honey (eds.), An Historic Tongue: Studies in English Linguistics in Memory of Barbara Strang, 133–137. London: Routledge. Hockett, Charles 1955 A Course in Modern Linguistics. New York: MacMillan. Jespersen, Otto 1909. A Modern English Grammar on Historical Principles, Vol. 1, Sounds and Spellings. London: Allen and Unwin. Kurath, Hans, and Raven I. McDavid, Jr. 1961 The Pronunciation of English in the Atlantic States. Ann Arbor: The University of Michigan Press. Lass, Roger 1987 The Shape of English. Structure and History. London & Melbourne: J.M. Dent and Sons.

How much shifting actually occurred in the historical English vowel shift?

281

Lass, Roger 1999 Phonology and Morphology. In: Lass, Roger (ed). The Cambridge History of the English Language. Vol. III, 1476–1776, 56–186. Cambridge: University Press. Luick, Karl 1896 Untersuchungen zur englischen Lautlehre. Strassburg: Truebner. Luick, Karl 1914/41 Historische Grammatik der englishen Sprache. Leipzig: Tauchnitz. Passy, Paul 1891 Etude sur les changements phonétique et leurs caractères généraux. Paris: Librairie Firmin-Didot. Stenbrenden, Gjertrud F. 2001 On the Interpretation of Early Evidence for ME Vowel Change. Unpublished paper presented at the 15th International Conference on Historical Linguistics, Melbourne, August 2001. Stockwell, Robert P. 1986 Assessments of alternative explanations of the Middle English Phenomenon of High Vowel Lowering when lengthened in the Open Syllable. In: Roger Eaton, Olga Fischer, Willem Koopman, and Frederike van der Leek (eds). Papers from the Fourth International Congress on English Historical Linguistics, The Hague: Mouton de Gruyter, 125–34. Stockwell, Robert P., and Minkova, Donka 1988 “The English Vowel Shift: Problems of Coherence and Explanation.” In: Dieter Kastovsky and Gero Bauer (eds), 1988, Luick Revisited, Tübingen: Gunter Narr. Sweet, Henry 1891 A Handbook of Phonetics. Oxford: Henry Frowde. Trager, George L., and Henry Lee Smith, Jr. 1953 An Outline of English Structure. Studies in Linguistics, Occasional Papers 3. Norman, Oklahoma: Battenburg Press. Trnka, B. 1959

A phonemic aspect of the Great Vowel Shift. Mélanges de linguistique et de philologie. Paris: Didier. 440–443.

Restoration of /a/ revisited David White

1.

Introduction

The purpose of the present paper is to examine the evidence and arguments concerning the prehistoric Old English sound change known as restoration of /a/, plus some other issues related to breaking and supposed lengthening of short diphthongs. An attempt is made to show that forms such as “slean” that are traditionally derived by positing 1) unconditioned first fronting, 2) breaking, 3) restoration of /a/, 4) loss of former /x/, and 5) compensatory lengthening of short diphthongs (attendant upon loss of unstressed vowels in hiatus) can be explained by positing simply 1) conditioned first fronting, 2) loss of former /x/, and 3) hiatus resolution involving dissimilation. If first fronting is conditioned, there is no need for restoration of /a/, so that the somewhat awkward developments seen in the posited development of /dagas/ > /dægas/ > /dagas/ can be done away with. It will also be seen that hiatus resolution rather than breaking followed by lengthening can explain the development of forms like “feos” < /fexes/ with two original front vowels. Various conventions must be explicated before proceeding. As the purpose is to provide an alternative explanation for those forms traditionally regarded as regular in West Saxon, irregular or dialectal variants, however interesting in their own right, will not be considered. Cases where pre-Old English /a/ before a following nasal evidently changed to something like /o/, e.g. “monn” and “hond”, are not relevant, and for the sake of avoiding often having to intrude “except in cases with a following nasal”, will in general be ignored. The second element of diphthongs (except “ie”, which will not come up) will be taken as a mid central vowel, which will be rendered as /-@/, though nothing hinges on it, and this paper is not intended to argue for or against this or any other interpretation. It will be also be assumed (pace Colman 1983), as is traditional, that Old English had a phonemic distinction between short /a/ and /æ/ from the historical moment the phonetic difference first developed. The phoneme that is spelled “g” will be

284

David White

taken as /g/ throughout, notwithstanding the fact that for pre-Old English it might well be better taken as /Â/. For the phoneme spelled “h”, most often /x-h/ will be used as a shorthand for “/x/ or /h/”, where there is no clear reason to prefer one over the other. In tables, where /x-h/ would be awkward, “x” indicates /x-h/ pronounced as [x] and “h” indicates /x-h/ pronounced as [h]. Similarly “/a-o/” will be used to mean “/a/” or “/o/” before nasals, where these must be mentioned. These decisions are for purposes of presentation only, and are not intended to be taken as assertions, novel or otherwise, about the actual truth (such as it can be recovered) of Old English historical phonology. As for typographic conventions, things like “/stapol/”, in both quotes and slashes, will be used where a spelling and a phonemic representation may be taken as the same. Likewise where a phoneme and phone are the same, things like /[h]/ may be used, where use of either [h] or /h/ alone might be misleading or inappropriate. Short diphthongs will be indicated by underlining, as /pæ@rrok/, and long vowels will be indicated by doubling, as /aa/. “y” is used for IPA “j”.

2.

Reasons for the traditional interpretation

It is advisable to first review the evidence and arguments that are generally held to justify both 1) an unconditioned first fronting and 2) restoration of /a/. (Henceforth mention of the second of these will be considered to do duty for both, as the two go together logically.) All things being equal, a change of /a/ to /æ/ then back to /a/ again has nothing to recommend it. But of course not all things are equal, and restoration of /a/ would never have been posited without good reason. The case for it may be made as follows. From the distribution of Old English “/æ/” vs. “/a/” in most cases, we would think that pre-Old English /a/ changed to /æ/ except when followed by both 1) a back vowel in the next syllable and 2) a single consonant, a geminate, or an /s/ plus plosive cluster. Some examples of /a/ being found in such circumstances, taken from the treatment in Hogg (1992: 96–99), are “/stapol/”, “mattuc”-/mattuk/, and /flaskan/. (This is the plural of “flasce”, where “/a/” is to be taken as analogical rather than phonologically regular.) But, contrary to what would be expected according to this first attempt at a formulation, /æ/ rather than /a/ evidently occurred 1) when an in-

Restoration of /a/ revisited

285

tervening single consonant was /x-h/, 2) when an intervening geminate is /rr/, /ll/, or (strangely enough) /pp/, and 3) when an /s/ plus plosive cluster is /sp/. Some examples, again taken from Hogg (1996: 96–99), of /æ/ occurring where /a/ might, according to the first formulation (before exceptions), be expected are: 1) */slæxan/, later “slean”-/slæ@n/, 2) /pæ@rrok/, “/tæppa/”, and 3) “/æspan/” (which is the plural of “/æspe/”, where “/æ/” is to be taken as analogical.) There is some doubt about whether there is any such rule as has been given for /sp/ and /pp/, which is to say whether “/æ/” should be regarded as regular before these. Thus Hogg (1992: 96–97) takes “/æ/” before “/sp/” as regular, but asserts that “/æ/” before “/pp/” is no more than an accident. Instances are so rare that accident is indeed hard to rule out. Nonetheless, as there are some possible phonetic explanations to be suggested if “/æ/” is regular before both “/sp/” and “/pp/”, it will be assumed in what follows that it is. The matter, which will not be treated till section 5, is of fairly minor importance either way. The effects of /rr/ and /ll/ are not terribly problematic, as will be seen below (section 4), so that it is the development of forms like “slean”-/slæ@n/, with lost /x-h/, that is the heart of the matter. Most observers would agree that this must somehow be from /slæxan/, from earlier /slaxan/. The problem is that according to the general distributional rule stated above (before exceptions), /slaxan/ should have stayed as it was, and should then have developed into /slaan/, spelled “slan”. (This does occur, but in Northumbrian.) Thus it seems that /x/ has somehow acted as a blocker against the influence of a following back vowel. But this does not make sense phonetically, as no other single consonant, and certainly no other velar consonant, has this effect. The evidence of breaking indicates that pre-Old English /x/ was always back, or it presumably would have turned into “ich-laut” rather than alter a preceding front vowel, and there is nothing in such a resolutely velar /x/ that should have blocked the influence of a following back vowel. Far from it, we might expect that /a/ rather than /æ/ would always have occurred before /x/. When /x-h/ in such forms was later lost, it is reasonable to presume that this was by way of [h], for two reasons: 1) [h], which has only a glottal component, is a weaker consonant than [x], which has both a glottal and a buccal component, and 2) in the two best-known cases where pre-vocalic former /x/ is known to have been lost, Cockney English and Romance

286

David White

(Palmer [1954] 1988: 227–229), it is certain that former /x/ had become /[h]/ before being lost. Surely a progression from [x] to [h] to [] (i.e. “zero”) is what we would expect must have happened in this case, in the absence of any evidence to the contrary. The idea occurs then that the problem of phonetic plausibility might somehow be solved by positing that [h] rather than [x] was the sound of /x-h/ at the time in question. But this does no good, since there is nothing about [h] that should have blocked anticipation of a following vowel. As Laver says (1994: 305): … the quality of the resonance that any laryngeal friction excites in the vocal tract will be that appropriate to the following vocoid, for which the vocal tract will already be in position. In English, the resonant quality of [h] in a word such as “he” [hi] is that of a whispered or breathed version of [i], and in “hoop”, the quality anticipates that of [u].

In other words, [h] is ordinarily a voiceless version of the following vowel. Therefore /a/ should have remained in [slahan], as [h] if anything would have been unusually open to transmitting the influence of a following vowel. But this is the opposite of what seems to have happened. To sum up, regardless of whether [x] or [h] is posited as the relevant realization of /x-h/, /a/ rather than /æ/ would still be expected, if first fronting had been conditioned. Thus if /a/ did change, in a conditioned manner, to /æ/ in such cases there is a serious problem with phonetic plausibility. If on the other hand the change was unconditioned, so that /slæxan/ existed early, then a development from this to /slæ@xan/ through /slæ@.an/ to /slæ@n/ would positively be expected by independently motivated changes. As Campbell (1959: 61) says: … breaking can hardly be regarded otherwise than as a change affecting front vowels, and from this it follows that Old English slean is developed from slæxan. Now it is hardly conceivable that when Prim. Germanic slaxan was at the stage slæxan, Prim. Germanic draGan would not be at the stage dræGan; from which it follows that Old English dragan has restored a in its first syllable by later change.

Thus restoration of /a/ restores /a/ in the environments where we might think it had always been. This sort of argument has been traditional since Luick (1921: 155 ), though due to its very indirect nature Luick seems notably less certain than some of his more recent followers. Luick’s view of the restoration of /a/ is adopted not only by Campbell (1959: 61), but also

Restoration of /a/ revisited

287

by Brunner (1965: §50), and (save for cases with lost /w/) by Hogg (1992: 99–100). On the other hand, Moore and Knott (1942: 39), Cassidy and Ringler (1971: 22) and Bülbring (1902: §§132–134) view first fronting as conditioned. Sievers ([1903] 1968: 31–33) also seems to be of this view. The position of the Wrights (1925: 40–41, 46–47) is seemingly contradictory, as they first assert that first fronting was conditioned, but later imply /æ/ in /slæhan/, where /a/ should be expected according to their previous formulation, without explanation. Quirk and Wrenn (1994 [1957]: 139–40) simply leave the matter open. A breaking solution works technically, and has some additional appeal in that /rr/ and /ll/, where /æ/ also appears to have occurred, are breaking environments as well. In words of this sort we may posit a development from /parrok/ to /pærrok/ to /pæ@rrok/, spelled “pearroc”, which conveniently enables the vowel in question, once broken, to escape restoration of /a/, however formulated. Thus it is that a breaking solution can seemingly kill two birds with one stone, but only with restoration of /a/. The main price we pay, obviously, is that the restoration of /a/ is, in the vast majority of forms where it is held to have applied, completely redundant. But to sum up, the position of traditionalists is essentially this: first fronting must have been unconditioned, not because a breaking solution works, but rather because no conditioning by which /a/ could change to /æ/ before /x-h/ and a back vowel is phonetically plausible.

3.

Another possibility: hiatus resolution

But /æ/ need not necessarily have been created by first fronting. It could have been created after loss of /x-h/ by dissimilation. Under this view, the form just before loss of /x-h/ would have been [slahan] phonetically, and as [h] was lost /sla.an/ was first dissimilated to /[slæ.an]/, then re-analyzed as /slæ@n/ (through /slæ.@n/). The last part is in fact regular when front vowels are followed by back vowels in hiatus (Hogg1992: 5.135). Such dissimilation may be seen as in part motivated by the desire to avoid seeming to say /slaan/, with long /aa/ and one syllable, which would have been a different possible word. But there is another possible motivation, which probably played some role, for Brosnahan (1953) notes what may be called “the Brosnahan Effect”: for some reason stress in Old English is associated with

288

David White

frontness and lack of stress with backness. (As this is the conclusion of his entire book, no more specific reference is appropriate.) In accordance with this, creating a front plus back sequence might have been one way of signaling a stressed vowel followed by an unstressed vowel, as opposed to a single long vowel /aa/. Changing /a.a/ to /æ.@/, later reanalyzed as /æ@/, does this. The proposed rule then is that short stressed /a/ becomes /æ/ before an immediately following back vowel. From there, development to “slean”-/slæ@n/ is regular under any interpretation. Under this scenario, there is no need to posit an unconditioned change of /a/ to /æ/, and thus there is no need to posit restoration of /a/. The question is whether there are any unwanted side effects: does short stressed /a/ immediately followed by a back vowel ever not change to /æ@/? As far as I have been able to determine, the answer is no, and therefore the theory is not falsified. Though Hogg (1992: 182) asserts that “When the sequence [in hiatus] consisted of back vowel + back vowel ... in WS the usual development was that the unstressed vowel was lost”, there appear to be no examples (in West Saxon) where an original short /a/ plus back vowel develops into either /a/ or (if compensatory lengthening is held to be applicable, as it probably would be) /aa/. There are two types of cases: those where hiatus is original or from lost /x-h/, and those from lost /w/. Of the first sort, all the examples I have been able to find by checking Hogg (1992: 182), Campbell (1959: 9–100), Wright (1925: 73–74), Sievers (1968: 74–80), Luick (1921: 222), and Quirk and Wrenn (1957: 138) of stems where such developments are posited are the following: /fooh-/ ‘take’, /hooh-/ ‘hang’, /doo-/ ‘do’, /gaa-/ ‘go’, /Sooh-/ ‘shoe’, /faah-/ ‘hostile’, /flaah-/ ‘treason’, /slaah-/ ‘slay’, /taah-/ ‘toe’, /Tooh-/ ‘clay’, /raah-/ ‘roe’, /maah-/ ‘shameless’, /tooh-/ ‘tough’, /wooh-/ ‘crooked’ , /hooh-/ ‘heel’, /Truuh-/ ‘trough’. In none of these do we see /a.a./ developing into /a/ or /aa/, or into anything, because in no case do we see short stressed short /a/ at all. All the forms in question have original long vowels. Thus the traditional formulation is arguably over-general. But in any event, no development of such forms can provide evidence against the posited change of short stressed /a/ to /æ/. Forms with lost /w/ do, however, provide some examples of stressed short /a/ developing in hiatus with a following back vowel. The development of “clea”-/klæ@/ is traditionally supposed to have been /klawu/ > /kla.u/ > /klæu/ > /klæ@/. There is nothing in this that is problem-

Restoration of /a/ revisited

289

atic for the alternative theory developed above, henceforth “hiatus solution”, for what is posited under this is exactly what is traditionally supposed to have happened in the development of “clea”. This, it may be noted, is not the same as the traditional interpretation as given above. The traditional interpretation in fact asserts that the loss of /w/ in such cases was significantly earlier than the loss of /x-h/, and connects “ea”-/æ@/ here with “ea”-/æ@/ from original /au/, whereas under the hiatus solution “ea”-/æ@/ in words like “clea”-/klæ@/ could be regarded as more to be connected with the same sort of thing in “slean”. Dissimilation of /a/ to /æ/ before back vowels might even be connected with the change of /au/ to /æ@/, though for the present purpose original /au/ is not relevant. Regardless of how we construe these matters, there is nothing in the development of such forms to contradict the proposition that the sequence of short stressed /a/ immediately followed by a back vowel changed into /æ@/ in Old English. Another area of Old English phonology where the hiatus solution might possibly be falsified involves palatalization. If all pre-Old English /a/ changed to /æ/ as in the traditional interpretation, then we would expect that preceding /k/ and /g/ should have been palatalized; whereas if /a/ in cases like “slean” did not change to /æ/ until the loss of intervocalic /x/, palatalization would not occur, since its active period was presumably earlier than loss of /x-h/. Thus the facts of palatalization might enable us to determine which theory is correct. But as is happens, none of the Old English words with “cea-” and “gea-” (where “ea” is long) go back to forms which have lost /x-h/, so there simply is no evidence on this point: all would have had initial palatals under either theory. The words, or stems, in question, taken from Clark-Hall and Merritt ([1894] 1960), with etymologies from Holthausen (1934) and/or Watkins (1981), are as follows: “ceac” ‘basin’ < Latin /kauk-/, “ceap” ‘cattle’ < Latin /kaup-/, “ceas” ‘chose’ < Germanic /kaus/, “ceas(t)” ‘strife’ < Latin /kaus-/, “ceaw” ‘chewed’ < Germanic /kauw/, “geac” ‘cuckoo’ < Germanic /gauk/, “geaf-” ‘gave’ < Germanic /gææf-/, “geagl” ‘throat’ < Germanic /gaugl/, “geap” ‘open’ < Germanic /gaup/, “gear” ‘year’ < Germanic /yæær-/, “geat” ‘pour’ < Germanic /gaut/, “gea(t)-” ‘yea, to say yea’ < Germanic /yææ(t)/ (? of uncertain development, but not from /gax-/), “geaD” ‘foolishness’ < Germanic /gauD/, “gean” < Germanic /gagin/. Though the origin of “ceace” ‘cheek’ is obscure, it is not from any Germanic /kaxu-/, and therefore does not represent a difference of prediction. How “gean”- /yæ@n/ de-

290

David White

veloped from /gagin/ is not clear, but in any event both solutions under consideration here would predict that first fronting should have turned /gagin/ into /gægin/, so there is not a difference of prediction. To review, what we have at this point is two possible solutions (this term is intended in the loose sense of “technically possible solution”), which may be laid out in tabular form as follows. “O” means any back vowel. In the first table, “Restoration” is short for “Restoration of /a/”, and “LL” stands for “Loss with Lengthening”, by which when a short diphthong is immediately followed by an unstressed vowel, this unstressed vowel is (qualitatively) lost and the short diphthong is compensatorily lengthened. In the second table, “Diss.” means “dissimilation”, and “Red.” means “reduction”, by which unstressed back vowels in hiatus after stressed front vowels are reduced to /@/. Note that this change is part of the traditional interpretation, for other forms, so the only new change being posited here is dissimilation. The loss of the syllable boundary is silently passed over in each case. Fronting is short for “First Fronting”, and is to be regarded as unconditioned (at least for relevant forms) in the first table, but in the second table fronting is to be regarded as conditioned in the way suggested by various non-believers in restoration of /a/. What the exact formulation would be need not concern us: for the present purpose it is sufficient to say that a conditioned first fronting, under any formulation, would not apply to forms with /a/ followed by a single consonant and a back vowel. Thus in the second table first fronting has no application. Nor does breaking have any application in the second table, as there are no front vowels for it to operate on. Table 1. The breaking solution

Start

Fronting Breaking

axO aCO

æxO æCO

æ@xO

/x/-Loss Restoration LL

Spelled

æ@.O

“ea”

æ@ aCO

“aCO”

Under either solution, /axO/ must wind up as /æ@/, spelled “ea”. Arguments for the phonetic plausibility of the two critical changes, Dissimilation and Reduction, have been given above. By contrast, the exact phonetic nature of breaking remains an unsolved problem (Stockwell 1994). The main difference between the two solutions is the development of “/dagas/”, and here

Restoration of /a/ revisited

291

the breaking solution is at a disadvantage since it posits a change that must be undone. It should be noted as well that over the two types of cases, the breaking solution must posit six changes, whereas the hiatus solution posits only three changes in arriving at exactly the same results. Table 2. The hiatus solution

Start axO

Fronting Breaking

/x/-Loss

Dissim.

Red.

Spelled

a.O

æ.O

æe

“ea”

aCO

4.

“aCO”

Hiatus with a following front vowel

The traditional interpretation of how forms of the “slean” type developed seems to be independently supported by other cases where a stressed front vowel followed by an unstressed front vowel has led, after loss of /x-h/, to a diphthong. A representative case is “feos”, where the following development is traditionally posited: /fexes/ > /fe@xes/ > /[email protected]/ > /fe@s/. It is therefore advisable to investigate whether a suitably expanded hiatus solution might work for cases of the “feos” type as well. What is hereby posited is simply that /e/ in hiatus after front vowels changed to /@/, so that the development of “feos” was as follows: /fexes/ > /fe.es/ > /fe.@s/ > /fe@s/-“feos”. Under this scenario, breaking does not occur (in cases of lost /x-h/) because intervocalic /x-h/, on its way to loss, is assumed to have been pronounced as [h], which would not have triggered breaking. As has been seen, a value of [h] in such cases, though not certain, is probable. More technically, it would be said that non-moraic /x-h/ does not condition breaking. As for phonetic motivation, the posited change of /e/ to /@/ may be seen as an example both of dissimilation and of the Brosnahan Effect. By the time /x-h/ is lost, there are only two relevant unstressed front vowels, /i/ and /e/ (Hogg 1992: 177), other unstressed front vowels having become /e/. In cases where breaking is traditionally supposed to have led to short diphthongs, there are only three relevant stressed front vowels, /æ/, /e/, and /i/. Therefore there should be six possibilities: 1) /æxe/, 2) /æxi/,

292

David White

3) /exe/, 4) /exi/, 5) /ixe/, and 6) /ixi/. However, /exi/ does not occur, all previous instances having been changed to /ixi/, and original (or semi-original) /ixe/ is rare, the only example apparently being various forms of the verb /twixe-/ (EWS “twion”), which after umlaut probably had /e/ from older /oy/. How these five possibilities should develop is laid out in the tables below. Table 3. The expanded breaking solution

Start ixi ixe exe axe axi axO

Fronting

æxe æxi æxO

Breaking

Umlaut /x/-Loss

LL

Spelled

i@xi

iexi

ie.i

ie

“ie”

[email protected] [email protected] æ@.e ie.i æ@ O

i@ e@ æ@ ie æ@

“io” “eo” “ea” “ie” “ea”

V-Change

Spelled

i.i

ie

“ie”

i.e e.e æ.e i.i a.O

i@ e@ æ@ ie æ@

“io” “eo” “ea” “ie” “ea”

i@xe e@xe æ@xe æ@xi æ@xO

iexi

Table 4. The expanded hiatus solution 2

Start ixi ixe exe axe axi axO

Fronting

æxe æxi

Breaking

Umlaut /x/-Loss

ixi

Again there are two possible solutions: the breaking solution and the hiatus solution. That the forms that are predicted by the breaking solution are also predicted by the hiatus solution is easily seen. “Vowel Change” in the hiatus solution includes three changes: 1) stressed /a/ (in hiatus) > /æ/, 2) unstressed /e/ (in hiatus after a short front vowel) > /@/, and 3) unstressed /i/ (after a short syllable and not before a palatal) > /e/ 1. As none of these can apply to the same word, and all occurred (by hypothesis) at about the same

Restoration of /a/ revisited

293

time, putting them in the same column seems justifiable as a matter of presentational convenience. Of these, the case for the first has been made above in dealing with cases of the “slean” type, and the third, though it has not before now been held to apply to forms of this type, is part of the traditional interpretation (Moore and Knott 1942: 65). Restoration of /a/, which if included would have made the table too large to fit across the page, has been left out. Note that with the inclusion of /axe/, the hiatus solution now makes exactly the same predictions as the breaking solution, even in cases (not treated above in dealing with “slean”) where the following unstressed vowel is not back. One difference between the two solutions is that under the hiatus solution, i-umlaut creates more cases of /ixi/ but does not phonemically affect old /ixi/, which is assumed to have been pronounced [ihi] up until /x-h/ was lost. Another difference between the two is that in the hiatus solution the change of unstressed /i/ to /e/ (which is not part of the breaking solution for such forms) is considered to be after loss of /x-h/, but as both happened around 700 (Moore and Knott 1942: 131, 134), this is a minor change. The two do not interact in any event. Both solutions work technically, without any serious problems of phonetic plausibility, except again the problems with breaking noted by Stockwell (1994). Thus it might seem then that there is no basis for a choice between them. But there is the development of words like “/dagas/” to consider. In this case, the hiatus solution has the advantage of simplicity, at least in the sense of avoiding otherwise unmotivated reversals. The obvious question is whether there are any cases of hiatus from sources other than loss of /x-h/ that develop differently. A possible problem with positing that /i.e/ changed into /i@/ is that there are more cases where it did not change than where it (by hypothesis) did: “/sie/” ‘be, subj.’, “/hie/” ‘they’, and “/Drie/” ‘three’. But these are also a problem for the breaking solution, at least as presented by Hogg, for as he says (1992: 184) in discussing these forms, “… the hiatus sequence should have been resolved by loss of the unstressed vowel.” Hogg speaks (1992: 184) of unstressed /e/ in hiatus after /i/ (in the Vespasian Psalter) being “protected by inflexional juncture”, a scenario which is hardly unreasonable, and could apply to the three forms in question. Alternatively, /-e/, if changed, may have quickly been restored by analogy, since it would have been regular in all cases where it is later found. The Wrights state (1925: 237) that “Drie had its e

294

David White

from the adjectives”, and “hie” too is described (Wright 1925: 242) as having “e from the adjectives”. Likewise /-e/ is of course regular in a subjunctive like “/sie/”. It must be admitted, however, that for short vowels independent evidence for the proposition that (in West Saxon) front vowels in hiatus after front vowels were lost is (apparently) non-existent, as the only example of a (non-round) front vowel in such circumstances appears to be various forms of /gææ.e-/ (OE “gan”), where the vowel is long. Note that under the hiatus solution the posited change of /e/ to /@/, being restricted to short vowels, would not apply, so that wrong “gean” and so on are not to be expected. Cases of “/ie/” before “/w/”, as “/-triewe/” and “/niewe/”, are not relevant because there is no syllable boundary, which again means that the posited change would not apply. Alternatively, /e/ could have been maintained by pointed dissimilation before /w/, in which case the rule given above would have to exempt the position before /w/. Campbell’s (1959: 100) “nicenned” ‘new-born’, with long “/i/”, is in accordance with the hiatus solution, as before a palatal consonant /i/, and therefore (after loss of /w/) /ii/ rather than /ie/, is to be expected. Even if it was not, it is far from clear that the two moras here could not be original, since “niew” ‘new’, a possible alternative source for the first element, had two moras. To sum up, if what has been said above is true, it is not necessary to posit breaking followed by lengthening in order to account for cases like “feos”. This would mean that the parallel scenario for cases like “slean” does not derive independent support from cases like “feos”. In other words, it is not true that we must posit breaking followed by lengthening anyway, in order to account for the facts in cases of lost /x-h/.

5.

Sonorant geminates

It remains to be seen whether the somewhat anomalous conditioning behavior of various geminates and clusters can be explained. Cases of /s-/ plus plosive and /-p/ will be treated in the next section. It must be stressed that all suggestions made here and in the next section are to be regarded as highly tentative. But with this understood, it still seems better to attempt an explanation than not. To review, the general rule for the distribution of /a/ versus /æ/ is that /a/ occurs before a single consonant followed by a back

Restoration of /a/ revisited

295

vowel, though for this purpose geminates and /s-/ plus plosive clusters tend to act like single consonants, with the exception of /rr/, /ll/, /pp/ and /sp/, which follow the general rule. Note that with the nasal geminates /nn/ and /mm/, where “/a-o/” appears before a following back vowel, this could just as well be attributed to the general rule whereby “/a-o/” appears before nasals, geminate or single, and therefore has no necessary bearing on the question of whether nasal geminates acted as single consonants. If we want to express the distributional facts in terms of the broadest possible natural classes, which is generally considered good practice, we might just as well state that /a/ appears before a following obstruent geminate followed by a back vowel, with either /æ/ or /a/ appearing before a sonorant geminate, depending on the usual rule for following nasals. The question is whether there is anything phonetically implausible in the idea that such a rule might represent the original distribution of /a/ versus /æ/ during a conditioned first fronting. It is reasonable to presume that if sonorant geminates acted like two consonants, it is because they received their expected full value in phonetic output, whereas if obstruent geminates acted like one consonant, it is because they were somewhat rushed in phonetic output, so that in such cases anticipation of a following back vowel would naturally have had a greater effect than might otherwise be expected over two consonants. There is some supporting evidence for this sort of thing (rushing of obstruent geminates) in the operation of /i/-umlaut, where geminates often act like single consonants, while consonant clusters can act as blockers (Moore and Knott 1942: 63), though the matter is difficult to interpret. A possible phonetic motivation for rushing the pronunciation of geminates might be that obstruents, being less vowel-like, are in general less easily sustained than sonorants. It is of course possible to sustain an /s/ indefinitely, but this does not exclude the possibility that the Anglo-Saxons may have felt it somehow inappropriate to sustain any sound that was not voiced, since as a general rule speech is voiced. 3 By contrast, sonorant geminates, being more vowel-like, are as a class more easily sustainable. Thus there is no reason to expect that they should have been rushed, and accordingly there is no reason to think that a following back vowel would have affected the preceding vowel, any more than was usual in cases of two consonants. Therefore it would seem that there is nothing phonetically implausible in finding /æ/ (or its descendant) before /rr/ and /ll/, and such cases do not provide evidence for restora-

296

David White

tion of /a/. 4 In terms of the sequence of forms, the development of a word like “pearroc” would be as under the traditional interpretation, the only difference being that the change of /a/ to /æ/ would be seen as fairly naturally conditioned by the following fully realized sonorant geminate blocking the influence of a following back vowel.

6.

Cases of /s-/ plus plosive and of /–p/

The evidently exceptional behavior (in some cases) of /s-/ and /-p/ remains to be explained. Clusters of /s-/ plus sonorant, like /m/, /n/, or /l/, seem to have received normal or full value as two consonants in general do, which is hardly surprising. The problem is /sk/ and /st/, both of which should act like two consonants but in fact act like one, in contrast to /sp/. With regard to /sk/, there are two possibilities. The first is that the newer conventional wisdom is wrong, and that /sk/ had become /S/ in all positions at the time in question, even before unstressed back vowels, so that of course it would have counted as one consonant. Generally, the more recent authorities such as Luick, Campbell, and Hogg support the view that /sk/ remained before unstressed back vowels, but as recently as Quirk and Wrenn (1994 [1957]: 16) the older view that the change was unconditioned is maintained, and it is possible that this is correct. If the change was unconditioned, it would of course have to have been very early in order to affect the distribution of /a/ and /æ/ during a conditioned first fronting, but there does not seem to be anything against this. The second possibility, probably less controversial, is that anticipatory backing of /k/ before back vowels helped in spreading the leftward influence of these, especially after a phonemic distinction of some sort had developed, probably from West Germanic gemination after loss of conditioning /y/, between back and front /k/. Again, this perhaps puts a prehistoric sound change a bit earlier than is generally assumed, but there seems to be nothing against it. Note that mere phonetic backing of /k/ should have been no stronger than backing of /p/, or rather of the tongue position for /p/, so that only stronger backing intended to clearly convey a phonemic distinction will do. But a phonemically back /k/ in something like /flaskan/ would probably have been enough to, in effect, transmit the influence of the following unstressed /a/ farther leftward than was usual.

Restoration of /a/ revisited

297

With regard to /st/, this is more easily rushed than is /sp/. Thus it is that the /t/ of “writer” can easily be speeded up or rushed through, by flapping as in American English, whereas no such option is available for the /p/ of “riper”. Accordingly the second (in American English) takes longer to say. This is not to say that Old English /t/ in such circumstances was necessarily flapped after /s/ (though I myself find no difficulty in pronouncing a flapped or at least very rapid version of /t/ in such circumstances), but rather that the tip of the tongue is more agile and easily moved than are the lips, which is what makes flapping available as a reduction device for dentals in the first place. Thus /st/ would be more easily shortened or rushed than /sp/, and, to judge by results, probably was. As for /pp/, it is the most easily sustained of plosive geminates. It is more easily sustained than /bb/ because there is no phonation to cause build up of pressure behind the closure, and more easily sustained than /kk/ or /tt/ because the exterior of the oral cavity (cheeks and lips) can expand somewhat to receive air. It is possible therefore that /pp/, being more easily given its full value, was not rushed, whereas /tt/ and /kk/ were. This would explain the difference of result. 5 For conceptual clarity it should be noted that the suggestions made above all relate to phonetic implementation, not to abstract phonemic representations. Therefore it is not posited that /st/, /tt/, and /kk/ “occupied one slot” or any such thing, but rather that they were shortened somewhat in phonetic output, just as the /t/ of American English “writer” is phonetically, rather than phonemically, shorter than the /p/ or “riper”. In sum, there does not seem to be anything phonetically implausible in the idea that /pp/, /sp/, and cases of /s-/ plus sonorant could have conditioned /æ/ in the manner expected of two consonant sequences, while /tt/, /kk/, /st/, and either /S/ or /sk/ (with phonemically back /k/) conditioned /a/, before a following back vowel. 6 In any event, it should be noted that whatever problems arise in motivating /æ/ or /a/ as part of a conditioned first fronting also arise in motivating /æ/ or /a/ during restoration of /a/, so that regardless of whether the possibilities suggested above are right or wrong, the traditional interpretation cannot be said to have an advantage in this regard.

298

7.

David White

Conclusion

Restoration of /a/ is traditionally posited because it is thought that for /æ/ to have been conditioned before /x-h/ and a following back vowel is phonetically implausible. This, as far as it goes, is true. But /x-h/ in such cases was lost, so that it is not necessarily relevant. By contrast it is not phonetically implausible for /æ/ to have developed before a following back vowel in hiatus, in part by dissimilation and in part by the Brosnahan Effect. There is no negative evidence: relevant cases of hiatus other than from loss of /x-h/ do not develop in a different manner, and the usual assertions made about back vowels followed by other back vowels in hiatus in fact apply, upon examination, only to cases with long vowels. Thus it is simply not true that restoration of /a/ is necessary because no other scenario is phonetically plausible. Dissimilation after loss of /x-h/ has not previously been considered, but once considered, does not seem to be problematic. Though it is perhaps not entirely clear what lies behind the mildly anomalous behavior of /st/, /tt/, and /kk/ in conditioning /a/, such cases in no way justify restoration of /a/. Since restoration of /a/ involves an otherwise unmotivated reversal of development in many words like “/dagas/”, it should not be regarded as a proven sound change of prehistoric Old English. Upon critical examination cases of the “feos” type do not necessarily involve breaking, as they too can fairly easily be reconceptualized as cases of hiatus resolution. The scenario suggested above for cases like “feos” requires only one new sound change: unstressed /e/ > /@/ after front vowels. This change would require inflectional juncture or analogy to explain “sie”, “hie”, and “Drie”, but this is to some extent part of the traditional interpretation anyway, and does not seem problematic. Cases of the “feos” type do not provide independent support for breaking followed by lengthening, and therefore indirectly for restoration of /a/, in cases of the “slean” type. All in all, the case for restoration of /a/, and for breaking followed by lengthening in cases of lost /x-h/, is notably weaker than is generally assumed.

Restoration of /a/ revisited

299

Notes 1. 2. 3. 4. 5.

6.

The loss of final /i/ after long syllables cannot be relevant to cases of /x-h/ loss. Vowel-change and loss of the syllable boundary are collapsed for presentational convenience. To start speaking is to start phonating. Nor has it ever been suggested, to my knowledge, that such cases do motivate restoration of /a/. It should be remembered, however, that Hogg (1992: 97) denies that the apparent anomalous conditioning behavior of /pp/ is anything more than an accident, so scanty is the evidence. As with the case of geminate /rr/ and /ll/, no one has ever (to my knowledge) claimed that such cases motivate restoration of /a/.

References Brosnahan, Leonard F. 1953 Some Old English Sound Changes: an analysis in the light of modern phonetics. Cambridge: W. Heffer. Brunner, Karl 1965 Altenglische Grammatik (nach der angelsächsischen Grammatik von Eduard Sievers neubearbeitet, (3rd edn). Tübingen: Max Niemeyer. Bülbring, Karl D. 1902 Altenglisches Elementarbuch, I: Lautlehre. Heidelberg: Carl Winter. Campbell, Alistair 1959 Old English Grammar. Oxford: Clarendon Press. Cassidy, Frederick G. and Richard Ringler 1971 Bright’s Old English Grammar and Reader. 3rd edn., second corrected printing. Orlando: Harcourt Brace Jovanovich College Publishers. Clark-Hall, J.R. and Herbert D. Merritt 1960 A Concise Anglo-Saxon Dictionary. Fourth Edition. Toronto: University of [1894] Toronto Press.

300

David White

Colman, Fran 1983 Old English /a/ /æ/ or [a] ~ [æ]? Folia Linguistica Historica 4: 265–285. Hogg, Richard. 1992 A Grammar of Old English. Oxford: Blackwell Publishers. Holthausen, Ferdinand 1934 Altenglisches Etymologisches Wörterbuch. Heidelberg: Carl Winters Universitätsbuchhandlung. Laver, John 1994 Principles of Phonetics. Cambridge: Cambridge University Press. Luick, Karl 1921 Historische Grammatik der englischen Sprache. Stuttgart/Oxford: Bernhard Tauchnitz/Basil Blackwell. Moore, Samuel and Thomas Knott 1942 The Elements of Old English. Ann Arbor: George Wahr. Palmer, Leonard 1968 The Latin Language. Norman: University of Oklahoma Press. [1954] Quirk, Randolph and C.L. Wrenn 1957 An Old English Grammar. 2nd edn. London: Methuen. Sievers, Eduard 1968 An Old English Grammar. Translated and edited by Albert S. Cook. New [1903] York: Ginn and Company. Stockwell, Robert P. 1994 On Old English Short Diphthongs and the Theory of Glide Emergence. In: Britton, Derek (ed), English Historical Linguistics 1994, 57–72. Philadelphia: John Benjamins. Watkins, Calvert 1981 Indo-European Roots. In: The American Heritage Dictionary of the English Language, 1505–1550. Boston: Houghton Mifflin Company. Wright, Joseph, and Elizabeth M. Wright 1934 Old English Grammar. 3rd edn, Oxford: Oxford University Press. [1925]

Pragmatic uses of SHALL future constructions in Early Modern English Maurizio Gotti

0.

Introduction

The object of this paper is the analysis of the uses of the future tense with SHALL 1 in an Early Modern English corpus; the texts examined are those included in the third section of the Early Modern English part of The Helsinki Corpus of English Texts. The results of this analysis will be compared to the pragmatic values pointed out in a few grammars of the same period 2 in order to assess their degree of faithfulness to the real usages found in contemporary texts. The Helsinki Corpus has been chosen as it includes a large selection of texts compiled in Early Modern English, as well as in Old and Middle English. It offers a good diatypic coverage as it contains a wide range of texttypes, genres and registers. Both literary and non-literary texts are included, with preference for the non-literary ones; most texts are in prose. Some, such as sermons, trial records, and official correspondence, are classified as “formal”; others, such as private correspondence, light fiction and comedy, are classified as “informal”. Spoken English is reflected in speechbased text-types such as drama, sermons and trial records, and therefore interesting comparisons can be made with the other, non-speech-based texts. The pragmatic aspects are of particular interest for our analysis, as they concern the speech acts (cf. Austin 1962) which the modal verb SHALL expresses. As such, therefore, they refer to specific contextual elements of the speech event (Dressler and Merlini Barbaresi 1994: 5), with particular reference to the locutor’s goals and intentions, and are meant to point out the illocutionary act potential (Kiefer 1997: 243) of the modal verbs examined, and not merely their typical formal behaviour or conventional semantic meaning. The description of the pragmatic uses of the future tense formed with SHALL has been the subject of previous analyses and heated debate: some

302

Maurizio Gotti

scholars have seen it as the result of the prescriptive intent of some of the early grammars, with the conclusion that “[t]he complete conventional rules for shall […] appearing first from such a source cannot safely be assumed to represent the practice of the language” (Fries 1925: 983). Other scholars, instead, believe that these criticisms are unjustified and are part of a negative attitude towards early grammarians; in their opinion, the considerations found in these grammars reflect the actual usage of the common language; cf., for example, Taglicht’s words: A careful reconsideration of the question leads to the conclusion that the rules given by Ward, like those given by Wallis before him, are inductive generalisations based on actual usage, and […] though perhaps they are not perfect in every detail and undoubtedly are in need of further elaboration, they represent a definite advance and a respectable achievement in the grammatical description of the English language. (Taglicht 1970: 197)

The period taken into consideration is 1640–1710, as this stretch of time includes the grammars which first describe the pragmatic uses of this modal auxiliary for the formation of the future tense. In this period, Wallis is the first grammarian to introduce a specification of the values of the auxiliaries used to indicate the future tense. 3 This specification is based on pragmatic criteria, linking the appearance of each modal to specific speech acts to be performed: In the first person shall simply indicates a prediction, whereas will is used for promising or threatening. In the second and third persons shall is used for promising or threatening, and will of a straightforward prediction. I shall burn, you will (thou wilt), he will; we shall, ye will, they will burn all simply predict what will happen; whereas I will, you shall (thou shalt), he shall, we will, ye shall, they shall burn are used for guarantees or pledges of what will happen. (Wallis: 94–5/339)

The great innovation of Wallis’ pragmatic approach has been particularly appreciated in recent studies. For an example of this appreciation cf. Arnovick: A careful examination of the Wallis Rules reveals the way in which they tell us “how to do things with words”. The formulations indicate whether a speaker commands, promises, or predicts; they indicate whether the speaker is the agent of volition or prediction or whether he or she is the questioner of such agency. By distinguishing speaker attitude and speaker involvement the Wallis paradigm makes formal distinctions of modality which are central to the utterance of a speech act. (Arnovick 1997: 140–141)

Pragmatic uses of SHALL future constructions in Early Modern English

303

Indeed, in this specification Wallis was preceded by George Mason’s Grammaire Angloise, London, 1633 (EL 261): Le signe du futur est, shall or will, mais il n’en faut pas vser indifferemment: car si vous vsez de ce signe, shall, quand il faut dire, will, il a mauvaise grace, outre qu’il semblera que vous parliez d’audace: example, vous pouvez dire elegamment, If I doe eate that, I shall be sicke, si je mange cela, je seray malade: au lieu que si vous disiez, I will be sicke, il sembleroit que volontairement vous voulussiez estre malade: ainsi vous pouvez dire: I hope you will be my good friend, j’espere que vous me serez amy: If you doe that, you shall bee beaten or chidden, Si vous faites cela, vous serez battu ou tancé: but I shall not, mais non seray: but you shall not chuse, mais vous ne choisirez pas, cest a sçauoir, ce ne sera pas à vostre chois: Pour le faire court, il est mal-aisé d’en bailler reigle certaine, parquoy je vous renvoye a l’vsage, auquel, afin de mieux y parvenir, nous vous proposerons la variation de certains verbes. (Mason: 25–26) [The signs of the future are shall or will, but they shouldn’t be used indiscriminately: since if you use the sign shall when will should be used, it will sound improper and it will seem as if you are speaking too audaciously: for example, you could say appropriately, If I doe eate that, I shall be sicke, if I eat that, I shall be sick: instead if you said, I will be sicke, it will seem as if you willingly wish to be sick: similarly you could say: I hope you will be my good friend, I hope you will be my good friend: If you doe that, you shall bee beaten or chidden, if you do that, you shall be beaten or chidden: but I shall not, but I shall not: but you shall not chuse, but you shall not choose, that is to say, it will not be your choice: To be short, it is not easy to give a fixed rule, therefore I refer you to usage, and in order to understand it better, we shall propose the conjugation of certain verbs.]

A few decades later Cooper amplifies Wallis’ specifications stating that SHALL used with first persons indicates declaration, while with second and third persons it stands for an order: Shall in primis personis innuit declarationem, in secundis & tertiis, mandatum; ut I shall prepare, we shall prepare. You shall prepare, ye shall prepare, he shall prepare, they shall prepare. (Cooper: 143) [In the first persons shall indicates a declaration, in the second and third persons an order; such as I shall prepare, we shall prepare. You shall prepare, ye shall prepare, he shall prepare, they shall prepare]

Cooper then specifies that the pragmatic function of the form expressing the future tense also depends either on its specific meaning, or on the locutor’s aim or particular emphasis: Sive declaratio, promissum, mandatum vel resolutio significatur, a sensu & loquentis proposito & emphasi, apparet. (Cooper: 143)

304

Maurizio Gotti

[Whether a declaration, a promise, an order or a resolution is meant becomes apparent from the meaning, the locutor’s intention and the emphasis]

Aickin derives his description from Cooper’s text, from which he also borrows the exemplification: Shall and will declare the future, as, when I shall love. Note that, shall in the first persons signifies a declaration of ones mind, in the second and third a command. I shall prepare, thou shalt prepare. (Aickin 2nd part: 11)

As can be seen from the analysis of these grammars, the main uses of the future tense expressed by SHALL have been grouped in two categories: i) prediction or declaration, ii) commands, promises or threats. Using Palmer’s terms (1986), the former category refers to the ‘dynamic’ 4 future uses of SHALL, commonly described with the label ‘prediction’; the latter, instead, refers to ‘deontic’ future usage, which indeed comprises several other pragmatic values apart from those pointed out by the early grammarians (such as ‘intention’, ‘proposal’, ‘advice’, ‘permission’, etc.). It is the purpose of this paper to identify the various uses found in the corpus and group them in appropriate categories.

1.

The analysis of the corpus

The corpus analysed consists of extracts whose length varies from 2,000 to 20,000 words from a number of texts published in the period 1640–1710, for a total of 171,040 words. The various texts have been grouped into different textual categories, each representing a main text genre. The textual categories included are the following: law, handbooks, science, educational treatises, philosophy, sermons, trial proceedings, history, travelogues, diaries, biographies/autobiographies, fiction, drama/comedies, private/ non-private letters. 5 The instances of SHALL contained in our corpus are 541; the corpus analysed does not only include the occurrences of the basic form shall (529 occurrences) but also of its graphic variants shalt (6 occurrences), shan’t and shant (3 occurrences each; no other orthographic variants were found). Most of these have future time reference (527 instances, i.e. 97%); the second largest group concerning time aspect pertains to present time (8 cases,

Pragmatic uses of SHALL future constructions in Early Modern English

305

i.e. 1%) and mainly includes phrases introducing an exposition or concept; six, instead, have timeless reference and occur in the description of habitual actions and permanent characteristics, or in sentences referring to moral obligation and permanent rules of behaviour, or in the context of instructions. 6 Of all the occurrences found in the corpus, only the 527 having future time reference will be taken into consideration in the rest of the paper. The various quotations have been labelled according to the prevalent pragmatic function in the context of the text analysed. However, in a few cases different values overlap and it is not always easy to see which can be considered the most relevant; this overlapping is at times made explicit by the author of the text himself, as can be seen in the following quotation: (1) To which (^Austin^) thus presaging, some say menacing, replies, since ye refuse to accept of peace with your brethren, ye shall have War from your enemies; and since ye will not with us preach the word of life, to whom ye ought, from their hands ye shall receive death. (MILTON X: 149) All cases in which more pragmatic values are present in the same quotation have been grouped in square brackets in our tables; normalised data – showing the frequency of the modal per ten thousand words – have instead been placed in round brackets. Table 1 shows the figures for the various types of modality found; 7 as can be seen, the quantitative difference between deontic and dynamic meanings is not great (239 vs 269 instances). As can be seen from the table, certain types of modality seem to correlate strongly with specific text types; this is the case of obligation and prediction which characterise statutes. 8 The other genres, instead, offer a much wider range of modal categories; handbooks, sermons and comedies, in particular, present instances of several deontic and dynamic uses. A variety of modal functions has also been found in educational treatises, philosophical works, history books, fictional works and letters.

1.1 Volition The SHALL-forms found in the corpus give rise to a variety of pragmatic forms, as can be seen in Table 2.

28 (25)

2 (2)

1 (1)

12 [2] (13)

12 [2] (13) 13 [2] (11)

13 [2] (11)

12 (11) 16 (18) 15 (13)

9 [1] (8)

9 [1] (8)

1 (1)

1 (1)

2 (2)

2 (1)

2 (1)

2 (1)

6 (5)

1 (1)

1 (1)

5 (4)

5 (4)

2 (2)

1 (1)

1 (1)

1 (1)

1 (1)

DIAR

4 (3)

3 (2)

3 (2)

1 (1)

1 (1)

BIO

34 [3] (29)

34 [3] (29)

15 [2] (12)

16 [2] (13)

31 [4] (25)

FICT

30 [3] (24)

31 [3] (25)

2 (1)

7 [1] (5)

17 [1] (14)

26 [2] (20)

DRAMA

33 [3] (17)

33 [3] (17)

5 (3)

15 [2] (8)

20 [2] (11)

LETT

65 (54) 57 (45) 53 (28)

246 (187)

13 [2] (12)

1 (1)

1 (1)

3 [1] (4)

4 [1] (5)

HIST

Total

116 [6] (88)

Prediction

13 [3] (12)

1 (1)

2 (2)

3 (3)

SERM TRIAL

1 (1)

116 [6] (88)

Dynamic

1 (1)

1 (1)

PHIL

Necessity

6 (5)

Permission

4 (3)

Obliga- 124 [2] tion (94)

15 [1] (13)

11 [1] (10)

130 [2] (99)

HAND SCIEN EDUC

Volition

Deontic

LAW

Table 1. Occurrences of SHALL future constructions according to modality type

508 (30)

269 [23] (16) 268 [22] (15.9) 1 [1] (0.1)

8 (0.5)

158 [5] (9)

73 [7] (4.5)

239 [12] (14)

TOT

306 Maurizio Gotti

Threat Promise

Volition Intention

LAW

1

5 [1]

6

1

11 [1]

1

1

2

HAND SCIEN EDUC

3 [1]

3 [1]

PHIL

1

1

SERM TRIAL

Table 2. Occurrences of SHALL-forms expressing volition

4 1

5

HIST

1

1

DIAR

1

1

BIO

1 6

9 [2]

16 [2]

FICT

1 11 [1]

5

17 [1]

DRAMA

5

10 [2]

15 [2]

LETT

6 30 [1]

37 [6]

73 [7]

TOT

Pragmatic uses of SHALL future constructions in Early Modern English

307

308

Maurizio Gotti

1.1.1 Intention The most frequent deontic usage of SHALL-forms expressing volition refers to speech acts of intention; these are most widespread in letters, fictional works, handbooks and educational treatises, and normally correlate with first-person pronouns (see quotations 2–3), although sentences containing a third person subject have also been found (see quotation 4): (2) I am offerd an Advowsen of what is sayd to be worth seavenskore pound by the yeare for 40 = l = to be paid in hand: the incumbent is said to be 80 yeare old etc, but shall not make any agreement till I speake with thee. (HOXINDEN: 292) (3) But I shan’t put you to the trouble of farther Excuses, if you please this Business shall rest here. (VANBR I: 34) (4) The old man dotes on her: she must not set neare a doore when she was hot with dancinge and tell her he shall give her somethinge; it is not his way to brag but he will not take five hundred pound for what he is worth; and many expretions of kindness to her and her mother. (HOXINDEN: 333)

1.1.2 Threat In the corpus analysed threats have only been found in history books, fictional works and comedies. As regards their forms, the second-person interlocutor is at times placed in the subject position, as in the following instance: (5) angry to be so upbraided, therfore said he with a chang’d countnance, Traytor to God and to me, thou shalt die; (MILTON X: 279) Threats sometimes concern a third-person referent; this is commonly placed in the subject position, but the threatening agent is often mentioned in some other part of the sentence, as can be seen in the following quotation: (6) therefore they first, saith he, shall feel our Swords; for they who pray against us, fight heaviest against us by thir prayers, and are our dangerous enemies. (MILTON X: 149)

Pragmatic uses of SHALL future constructions in Early Modern English

309

1.1.3 Promise Promises appear in various text types, such as comedies, handbooks, fictional works, letters, history books and educational treatises. Promissory statements are usually expressed by a first person pronoun, reinforced by specific expressions, such as the epistemically marked phrase ye may be sure: (7) Nevertheless because ye are strangers, and have endur’d so long a journey, to impart us the knowlegde of things, which I perswade me you believe to be the truest and the best, ye may be sure we shall not recompense you with any molestation, but shall provide rather how we may friendliest entertain ye; (MILTON X: 144)

In dialogues and correspondence, promises having a second person pronoun are often used. Although the subject is not a first person pronoun, the locutor remains responsible for the action promised, as the following instance shows: (8) (^Tom.^) but if any should offer any such thing to you, you must be sure to let me know what they say or do to you. (^Ione.^) You shall be sure to know all Sir. (PENNY: 268) Promises are also expressed by means of statements having a third person subject, as in the following example: (9) Dear Nurse, this goodness of yours shan’t go unrewarded, (VANBR I. 64) In some cases the deontic value of SHALL is reinforced by the use of some harmonic verb such as engage, which underlines the subject’s commitment in a very effective way, as can be seen in the following example: (10) on condition they be sent constantly to the Schoole, and that their Parents do engage they shall keep good order, and be cleanly and neat in their apparel; (HOOLE: 228)

2

1

Advice/Request for advice

1

4

1

1

HAND SCIEN EDUC

Proposal

28

96 [2]

Order

Prohibition

124 [2]

Obligation

LAW

1

1

PHIL

1

1

SERM TRIAL HIST DIAR

Table 3. Occurrences of SHALL-forms expressing obligation

BIO

1 [1]

3

2

9 [1]

15 [2]

FICT

6 [1]

1

7 [1]

DRAMA

5

5

LETT

5 [1]

11 [1]

31

111 [3]

158 [5]

TOT

310 Maurizio Gotti

Pragmatic uses of SHALL future constructions in Early Modern English

311

1.2 Obligation Obligation is also expressed in a variety of pragmatic forms (as can be seen in Table 3), although the great majority refer to order and prohibition (90% of all cases).

1.2.1 Order Order is the most common pragmatic function performed by the deontic aspect of the modal analysed; with its 111 occurrences it corresponds to one fifth of all SHALL-forms found in the corpus. In particular, these appear in statutes, where they represent the most frequent type of modal meaning; for SHALL this is not surprising, as the main function of statutes is to provide prescriptive statements. Here is an example: (11) Provided alwayes and bee itt enacted, that the severall Rates and Impositions hereby imposed upon the respective Goods and Merchandizes aforesaid shall bee collected and paid according to the respective Rates and Proportions herein expressed and bee raised levyed collected and paid unto His Majesty dureing the respective Tymes (STAT7 VII: 97) The prescriptive speech act performed by means of SHALL is at times made more explicit by the use of the nouns order and direction in the same sentence: (12) and the Money thereby raised shall be imployed and accounted for according to the Order & Direction of the said Justices for and toward~ the purchasing of the Land to enlarge the said Highways and for the making the said Ditches and Fences and the said Assessment shall by order of the said Justices be levied by the Overseers of the High-ways by Distresse and Sale of the Good~ of Persons so assessed not paying the same within Ten Days after Demand (STAT7 VII: 210) As can be seen from the quotations above, in the legal texts analysed SHALL only occurs in third person verbal forms, as the obligations to

which the documents refer usually concern agents who are presented in a general and impersonal way; hence the subject is often “he or she”, “any person or persons”, “any Justice or Justices of the Peace”, “hee shee or

312

Maurizio Gotti

they” etc. Also in official correspondence – where the tone is formal and prescriptive acts are often performed – the subject is frequently a third person singular or plural noun or pronoun: (13) For which these Our Letters shall bee your Warrant. (CHARLES I: 25) (14) And for soe doing this shall be your Warrant. (CHARLES I: 113) Non-legal text types, instead, commonly show instances of orders given to a second person. This is due to the fact that such forms occur either in quoted speech or in dialogues, where the recipient of the directive act is addressed in a straightforward way: (15) (^Tom.^) Why then in brief these they are. First, you shall kiss my hand and swear that you will acknowledge me to be your Lord and Master. (^Ione.^) I will Sir. (^Tom.^) Secondly, when I come home drunk a nights, you shall be diligent to make me unready and get me to bed, and if I chance to befoul my self, you are to make me clean without chiding me. (PENNY: 268)

1.2.2 Prohibition Another frequent deontic category is that of prohibitions; these are particularly common in statutes (90% of all cases concerning prohibition), where they provide rules to prevent a certain type of behaviour. Sentences containing prohibitions usually rely on the negative form of SHALL, as can be seen in the following examples: (16) that such Malt shall be landed in the said Islands of Jersey and Guernsey or one of them the Danger of the Seas only excepted for the Use of the Inhabitants there and shall not be landed or sold in any other Parts whatsoever (STAT7 VIII: 455) (17) O fie! Hast not thee declared among the Brethren that it shall not be lawful for a Sister to define her self? (PENNY: 148) In other cases the negative command is conveyed by sentences in which the subject is preceded by no, as we can see in the following example:

Pragmatic uses of SHALL future constructions in Early Modern English

313

(18) Provided alsoe That noe Person shall bee discharged out of Prison or have any Benefitt or Advantage by force or virtue of this Act who shall bee really and (\bona fide\) indebted in more than the Su~m of One hundred (STAT7 VII: 76)

1.2.3 Proposal Proposals commonly appear in fictional works, comedies and handbooks, and always occur in the context of a dialogue, as in the following examples: (19) let me invite you along with us; I’ll bear you charges this night, and you shall bear mine to morrow; for my intention is to accompany you a day or two in Fishing. (WALTON: 212) (20) (^Lov.^) Let me see his Wound. (^Ser.^) Then you shall dress it, Sir; for if any body looks upon it, I won’t. (VANBR I: 40) Proposals may be reinforced by appropriate harmonic expressions, as can be seen in the following quotations: (21) Then if it please thee to begin, thy hand-maid shall bear a part, but be sure it be such pure Language as is used among our Friends when assembled at Bull-and-Mouth. (PENNY: 149) (22) But I shan’t put you to the trouble of farther Excuses, if you please this Business shall rest here. (VANBR I: 34) (23) And it shall be to morrow Morning still, if you’ll consent? (VANBR I: 62)

1.2.4 Advice/Request for advice Instances of advice have been found in educational treatises and sermons; whenever advice is given to people in general, a third person subject is used: (24) And as amongst men and women humility is the way to be preferr’d; so it is in Husbands, they shall prevail by cession, by sweetness and counsel, and charity and compliance. (JETAYLOR: 23)

314

Maurizio Gotti

Requests for advice are particularly frequent in fictional works and comedies, although they also appear in handbooks, sermons, educational books and philosophical treatises. The reason for their popularity in the former text types is the strict correlation of this speech act with the dialogic form. Most questions have a first person singular subject and the addressee of the request for advice is at times explicitly mentioned, as can be seen in the following quotation: (25) O Master Master, by and by, nay, ‘tis my Dames voice, whether shall I depart? where shall I run to hide myself from Sarah? (PENNY: 149) In some sentences the subject is first person plural and the request concerns the interlocutor’s advice or opinion on the action to be taken jointly, as in the following instance: (26) Well, well honest Jone, I know thy meaning, come give me thy hand, let us to Church and be married with speed but now I think on’t, what Church shall we go to Ione? (PENNY: 269) The two patterns are very similar in use and are often adopted within the same context, as can be seen in this series of questions that a bride-to-be asks her fiancé: (27) This will be brave indeed John, but what shall we do with our Ale? … What shall we do for Clothes? … Who shall we bid to our Wedding? … What Musick shall we have? … I laugh to think how the young men will turn the Lasses about in dancing, and how they will buss them, methinks I see them already, but good Jack how shall I do to behave my self at that time amongst so many; I shall be so ashamed I shant know what to do. … What Favours shall we give? (PENNY: 117–8) Also the instances found in handbooks and philosophical works are commonly inserted in dialogues:

Pragmatic uses of SHALL future constructions in Early Modern English

315

(28) Oh me, all the Horses are got over the River, what shall we do now? Shall we follow them over the water? (WALTON: 211) (29) (saith she) see how well and irrefragably thou canst prove what I have said, to wit, that god is wholly replenished with the sovereign Good. How shall I do that? (replied I). (BOETHPR: 135)

1.3 Permission Another deontic use of SHALL is that of permission; the cases found in the corpus only occur in statutes and in comedies. In the former text type, when permission is granted to judges, the SHALL-form is usually accompanied by the expression have power, as can be seen in the following quotation: (30) That the Justices of the Peace […] shall have power to enlarge or widen any Highways in their respective Counties Riding~ Divisions Liberties or Places so that the Ground to be taken into the said Highways (STAT7 VII: 210) Instead, when permission is granted to common citizens, harmonic words such as at liberty or free are generally employed; here are a few examples: (31) the Party injured shall be at liberty to bring his Action for the Damages by him sustained (STAT7 VIII: 457) (32) to which all Persons concerned (upon reasonable Request) shall have free accesse without Fee or Charge Any thing herein contained to the contrary notwithstanding (STAT7 VIII: 459) In comedies permission occurs whenever some sort of authorization is implied in the locutor’s words, as can be seen in the following quotations: (33) O my dear (^Nurse^), forgive me this once, and I’ll never misuse you again; no, if I do, you shall give me three thumps on the Back, and a great pinch by the Cheek. (VANBR I: 60) (34) Well, I’m such a tender-hearted Fool, I find I can refuse nothing; so you shall e’en follow your own Inventions. (VANBR I: 64)

316

Maurizio Gotti

1.4 Prediction The highest number of dynamic occurrences concerns prediction; with its 268 occurrences, it corresponds to approximately one half of all the cases of SHALL future constructions taken into consideration and covers all the text types included in the corpus. Prediction is expressed in different pragmatic forms, as can be seen in Table 4.

1.4.1 Neutral prediction Neutral prediction (or “pure future”, as this category is often referred to in the literature) is very widespread in the corpus analysed; indeed, there are examples of usage with all persons, and in all kinds of sentences: affirmative, negative and interrogative. Here below are examples of the various possibilities found in the data, starting from first-person subjects: (35) because I shall not be able, at this rate, to know whether I have a Guide, and whether I follow him or not. (TILLOTS II, ii: 448) (36) I’le now lead you to an honest Ale-house where we shall find a cleanly room, (^Lavender^) in the windows and twenty (^Ballads^) stuck about the wall; (WALTON: 216) Here, instead, is an instance with a SHALL-form depending on a secondperson subject: (37) Well, you shall have your Choice when you come there. (VANBR I: 63) The following represent cases containing third-person subjects:

(38) I would then say, that he hath a mind to arrive at the sovereign Good; but can it be thought that it shall ever be found in these Acquisitions, which I have shewed already, not to be able to perform any thing they promise? (BOETHPR: 127) (39) The boys, and the pedlers, and the fruiteres shall tell of this man, when he is carried to his grave, that he lived and died a poor wretched person. (JETAYLOR: 10) The main verbs that follow SHALL are usually of the stative type; indeed, if we examine the cases occurring in non-legal texts (in legal texts neutral

Pragmatic uses of SHALL future constructions in Early Modern English

317

prediction is commonly expressed by SHALL; WILL-forms appear only in 5 cases out of 116), the three most frequent verbs are be (23 occurrences), have (13 instances) and find (6 cases); these findings confirm the results of similar analyses carried out by other scholars (cf. Kytö 1991). When the prediction is the result of some kind of hope, fear or expectation, it is commonly preceded by an explicit harmonic expression, as can be seen in the following quotations: (40) In this part I hope I shall finde easy beleive and noe body can have a pretence to doubt (LOCKE: 49) (41) My lord, I hope I shall not be condemned without being heard. (LISLE IV: 120C1) (42) I’m afraid we shall lose that Character, Madam, whenever you happen to change your Condition. (VANBR I: 35) (43) You shall be hang’d first, but if ever you expect that I shall be friends with you, there must be two things granted. (PENNY: 271) In philosophical books and sermons this predictive usage of SHALL often expresses a logical or inevitable consequence, as can be seen in the following quotations: (44) Thou hast now then the Form and causes of that adulterate Felicity: now turn again the Eyes of thy Consideration upon the contrary Prospect, and thou shalt soon comprehend that true and genuine happiness which I so long have promised thee. (BOETHPR: 127) (45) it is as if a King should beat his Viceroy and use him like a Dog; from whom most of that Reverence and Majesty must needs depart, which he first put upon him, and the subjects shall pay him less duty, how much his Prince hath treated him with less civility; and the government of the whole family shall be disordered, if blows be laid upon that shoulder which together with the other ought to bear nothing but the cares and the issues of a prudent government. (JETAYLOR: 25) Results expressed by SHALL-forms have also been found in fictional dialogue, as can be seen in the following quotation: (46) I shall observe my own the more, but you must not look towards me, for then I shall laugh and that will shame me quite. (PENNY: 118)

Prediction Neutral prediction Prophecy Assurance

116 [6] 116 [6]

LAW

1

11 [1]

2 [1]

1

13 [2]

HAND SCIEN

9 [1]

9 [1]

EDUC

SERM

1

3

11 [2] 10 [2]

12 [2] 13 [2]

PHIL

2

2

TRIAL

Table 4. Occurrences of pragmatic forms expressing prediction

1

1

HIST

1

1

DIAR

3

3

BIO

2

32 [3]

34 [3]

FICT

3 [1]

27 [2]

30 [3]

DRAMA

1

1

31 [3]

33 [3]

LETT

9 [2]

4

268 [22] 255 [20]

TOT

318 Maurizio Gotti

Pragmatic uses of SHALL future constructions in Early Modern English

319

1.4.2 Prophecy SHALL is also used – particularly in religious contexts – to express proph-

ecy; indeed, of the four occurrences found in the corpus, three appear in sermons, and one in correspondence. However, the quotation found in the letter also concerns religion: (47) and that I am at such adistance from my famaly, but the great god hath saide, that all thinges shall worke together for good, to him that truly fare him, the lord to rais up all yo=r= spirrits to trost in him, and sarve him, and live to him, which is the daily prayere (JPINNEY: 19) Prophetic quotations are commonly taken from biblical texts and their explicit mention is justified by the fact that their interpretation is often the very object of the sermons themselves, as can be seen in the following examples: (48) I. (^First^), we will consider the nature of the sin here mentioned, which is (^scoffing^) at Religion, (^There shall come scoffers^): These it seems were a sort of people that derided our Saviour’s prediction (TILLOTS II, ii: 418) (49) and reasons of his coming to judgement, (^infidelity^), and (^prophane scoffing^) at Religion. (^When the Son of man comes, shall he find faith on the earth?) And St (^Jude^) out of an ancient prophecy of (^Enoch^) (TILLOTS II, ii: 430) At times the source of the prophetic quotations is explicitly specified, as is shown in the following occurrence: (50) St. (^Jude^) in his (^Epistle^) gives much the same character of them that St. (^Peter^) here does, ver. 18, 19. (^There shall come in the last days mockers, walking after their own ungodly lusts, sensual, (TILLOTS ii, II: 419)

1.4.3 Assurance Various cases of SHALL-forms expressing assurance have been found in the E3 part of the Helsinki Corpus; they are not concentrated in any specific text type, but occur in five different genres. In some cases utterances are

320

Maurizio Gotti

made stronger by means of locutor-oriented boosters, intended to highlight the utterer’s truthfulness and reliability. Here are a few examples: (51) I’le to my own art, and I doubt not but at yonder tree I shall catch a (^Chub^), and then we’l turn to an honest cleanly Hostess, that I know right well; (WALTON: 215) (52) and I shall wait upon thee too that same day, as it is my duty. (PENNY: 118) (53) Prevail with him – or he shall never prevail with me, I can tell him that. (VANBR I: 64) In other cases, instead, the emphasis is laid on the certainty of the interlocutor’s acceptance of the validity of the proposition, as can be seen in the following quotations: (54) You shall not doubt it long, for you shall see me do it presently: (WALTON: 215) (55) Be sure that nothing shall be squandered away can be prevented. (HOXINDEN: 276) [sic] (56) They shall be dispoil’d of it sooner than perhaps thou wouldst have them, (BOETHPR: 178) In further cases, the instances of assurance rely on content-oriented boosters, which are meant to increase the illocutionary force of the speech act by underlining the certainty of the proposition asserted and thus emphasize its validity. The following is an example of this usage: (57) and then we shall have good handsel indeed, (PENNY: 120) Emphasis is at times obtained by modifying the conventional word order inside the sentence, as can be seen in the following quotations: (58) I’ll give my Wench a Wedding-Dinner, tho’ I go to Grass with the King of (^Assyria^) for’t, and such a Dinner it shall be, as is not to be Cook’d in the Poaching of an Egg. (VANBR I: 61) (59) Now on my Knees, my Dear, let me ask your pardon for my Indiscretion, my own I never shall obtain. (VANBR I: 40)

Pragmatic uses of SHALL future constructions in Early Modern English

321

1.5 Necessity Necessity concerns those cases of obligation deriving from external factors, in which the performing of a certain action is to be considered an essential requisite for the successful achievement of another. Here is an instance of this usage: (60) (^Arch.^) If the Sun rides fast, and disappoints not Mortals of to Morrow’s dawn, this Night shall crown my Joys. Mrs. (^Sull.^) My Sex’s Pride assist me. (^Arch.^) My Sex’s Strength help me. Mrs. (^Sull.^) You shall kill me first. (FARQUHAR: 60)

2.

Conclusions

As can be seen from the analysis of SHALL-forms carried out so far, examples of the various pragmatic uses identified by 17th-century grammarians have been found in the corpus taken into consideration. Also the differentiation between deontic and dynamic uses of this modal verb – suggested by the adoption of categories labelled “prediction”, “promising”, “threatening”, “declaration”, command” – has been confirmed by the present survey, which has also pointed out that the quantitative difference between its deontic and dynamic aspects is minimal. The most frequent deontic usage of SHALL-forms expressing volition refers to speech acts of intention; these are most widespread in letters, fictional works, handbooks and comedies. Obligation is also expressed in a variety of pragmatic forms, although the great majority refer to order and prohibition. Another deontic use of SHALL is that of permission; the cases found in the corpus only occur in statutes and in comedies. The highest number of dynamic occurrences concerns prediction; indeed, this pragmatic value is conveyed by approximately one half of all the cases of SHALL future constructions taken into consideration and covers all the text types included in the corpus. In particular, neutral prediction has proved to be very widespread in the corpus analysed. The present analysis has also highlighted the degree of correlation between the modality categories pointed out by early grammars and some

322

Maurizio Gotti

specific text types; the results indicate that some genres show a very high correlation rate. This is the case of obligation and prediction which characterise statutes, while proposals commonly appear in fictional works and comedies and always occur in the context of a dialogue. Instances of advice have been found prevalently in educational treatises and sermons, while requests for advice are particularly frequent in fictional works and comedies. A pragmatic function which is strongly correlated with topic is prophecy, which has proved to be exclusively linked with religious contexts.

Notes 1. 2.

3. 4.

5. 6. 7. 8.

Throughout the text, capitals will be used to denote the lexeme SHALL, while italics will be used for its graphic variants such as shall, shalt, shan’t, etc. The grammars taken into consideration are the following: – John Wallis, Grammatica Linguae Anglicanae, London, 1653 (EL 142); – Christopher Cooper, Grammatica Linguae Anglicanae, London, 1685 (EL 86); – Joseph Aickin, The English Grammar, London, 1693 (EL 21). The acronym EL stands for “English Linguistics 1500–1800”, the series of texts on microfiches, selected and edited by R.C. Alston for The Scolar Press. Wallis’ grammar has been innovative in many other aspects. For the analysis of its most relevant innovations cf., among others, Vorlat (1975) and Robins (1986). The term ‘dynamic’ was originally suggested by von Wright (1951: 1–2) to refer to such cases that cannot be categorized as deontic modality (expressing an order, permission or promise) or epistemic modality (expressing an inference by the speaker). More detailed information about the single texts may be found in the manual to the Helsinki Corpus (Kytö 1996). For a detailed analysis of all SHALL-forms found in the corpus see Gotti (forth.) No SHALL-forms have been found in travelogues. For a detailed analysis of the use of SHALL (and WILL) in EModE statutes cf. Gotti (2001).

Pragmatic uses of SHALL future constructions in Early Modern English

323

References Arnovick, Leslie K. 1997 Proscribed collocations with shall and will: the eighteenth-century (non-)standard reassessed. In: Jenny Cheshire and Dieter Stein (eds.), Taming the Vernacular, 135–151. London: Longman. Austin, John L. 1962 How to Do Things with Words. Oxford: Oxford University Press. Dressler, Wolfgang U. and Lavinia Merlini Barbaresi 1994 Morphopragmatics: Diminutives and Intensifiers in Italian, German, and Other Languages. Berlin: Mouton de Gruyter. Fries, Charles C. 1925 The periphrastic future with shall and will in Modern English. Publications of the Modern Language Association of America XL: 963–1024. Gotti, Maurizio 2001 Semantic and pragmatic values of shall and will in Early Modern English statutes. In: Maurizio Gotti and Marina Dossena (eds.), Modality in Specialized Texts, 89–111. Bern: Peter Lang. Forthcoming Shall. In: Maurizio Gotti, Marina Dossena, Richard Dury, Roberta Facchinetti and Maria Lima, Variation in Central Modals: Analysis of a Corpus of Middle English and Early Modern English Texts. Kiefer, Ferenc 1997 Modality and Pragmatics. Folia Linguistica XXXI/3–4: 241–253. Kytö, Merja 1991 Variation and Diachrony, with Early American English in Focus. Frankfurt am Main: Peter Lang. 1996 Manual to the Diachronic Part of the Helsinki Corpus of English Texts. Helsinki: Department of English, University of Helsinki. Palmer, Frank Robert 1986 Mood and Modality. Cambridge: Cambridge University Press. Robins, Robert H. 1986 The evolution of English grammar books since the Renaissance. In: Gerhard Leitner (ed.), The English Reference Grammar, 292–306. Tübingen: Max Niemeyer Verlag.

324

Maurizio Gotti

Taglicht, J. 1970 The genesis of the conventional rules for the use of shall and will. English Studies 51/3: 193–213. Von Wright, Georg Henrik 1951 An Essay in Modal Logic. Amsterdam: North Holland. Vorlat, Emma 1975 The Development of English Grammatical Theory 1586–1737. Leuven: Leuven University Press.

Explaining the creation of reflexive pronouns in English* Edward L. Keenan

We provide an account of the historical creation and interpretation of the English reflexives (himself, herself, …) in terms of (1) two general, not specifically linguistic, forces of change: Decay and Inertia; (2) two universal semantic constraints on language: Constituency Interpretation and Anti-Synonymy, and (3) the initial state of the Old English (OE, c750–c1150) anaphora system. In comparison with other approaches the changes as we present them do not rely on parameter resetting and are not primarily instances of grammaticalization, though this plays a role in one change. We also, standardly, invoke Pattern Generalization, whereby a rule or paradigm that applies to a limited range of cases extends to new ones. Our study is based on some 11,000 instances of locally bound objects of verbs and prepositions, drawn from over 100 texts dating from c750 to c1750. Keenan (2000, 2001) present these data extensively, as well as the list of source materials.

1.

Forces of change: DECAY and INERTIA Decay Things wear out

Decay is familiar from cases where phonological reduction obscures morpheme boundaries. An example is the negative n’t in Don’t blink, where we may distinguish at least four steps reading back from Modern Standard English (MSE) 1 to Old English: n’t < not < ne+aught < ne+a/an+wiht ‘not one whit (thing)’. More systematic cases throughout Middle English (ME, c1150–1500) are vowel reduction in unstressed syllables and the loss of word final unstressed vowels, often the last remaining traces of case inflection. Thus OE disyllabic nama ‘name’, naman ‘names’ with word initial stress, becomes

326

Edward L. Keenan

disyllablic name, namen in Early Middle English and then monosyllabic name, names in Modern Standard English. A commonly noted case of semantic Decay is the historical weakening of demonstratives to become definite articles. Thus Old French nom. masc. sg. li ‘the’ derives from Latin illi ‘that’, and the English definite article the traces its origin to se, sio, łæt ‘the, that’ (masc., fem., neut. sg.), which had a demonstrative use absent from MSE the. In the present work an important case of Decay in Old English and Middle English is: (1) Function Word Proclisis (FWP) A short closed class item ω forms a phonological word with a following word W. ω loses obligatory stress (if it had it) and may phonologically reduce. on+weg → a way on+slep → asleep on+life → alive on+gegn → again bi+cause → because be+sidan → beside be+geondan → beyond eal(l)+ane → (a)lone all+theigh → although all+to+gædere → altogether

FWP predicts that the main stress in ω+W is that of W, despite being noninitial, contra the core Germanic pattern. In general, a cliticized item continues to exist as an independent word, a phenomenon called Split in Heine and Reh (1984), Divergence in Hopper (1991). Modern Standard English retains all, on, and by, sometimes in doublets: All aboard! vs Everyone on board; beside her vs by her side; He sent me away vs on my way. We even say all alone. Similarly Fr. avoir encliticized to infinitives: parlerai, parleras, … ‘I/you(sg) … will speak’, but remained as a main verb. FWP is also active in the creation of paradigms, as with what(so)ever (Allen 1980). so (< swa) procliticizes to ever within the relative clause, and what, how and later who procliticize to the result. Note that in (2a), the three items are not adjacent and do not form a syntactic or a semantic unit. But by the late 1300’s we see forms like (2b, c), cited in Allen (1980: 211, 212), with separation possible as late as 1607. V & V 1.67 (2) a. Luue Dine nexte al swa De seluen, hwat manne swo he aeure bie! Love thy neighbor as thyself, whatsoever man he be! c1200

Explaining the creation of reflexive pronouns in English

327

b. But what-som-ever woo they fele, they wole not pleyne Chaucer Rom 5041 But whatsoever woe they feel, they will not complain c1385 c. And what as evere that ye seie, riht as ye wole so wol I Gower CA i.1830 And whatsoever that you say, just as you will so will I c1390 d. hee would ... patronize his cause against all opponants of what sort soeuer, ... Horsman 1607 Inertia Things stay as they are unless acted upon by an outside force or Decay Vocabulary borrowing during language contact is a standard “outside force” case. The Norman French conquered England in 1066 establishing Anglo-Norman as the language of administration; by the 1200s there is extensive influence of Central French, with massive French borrowings in the period 1250–1350. Equally French rhyming poetry replaced the native alliterative verse (despite the occasional alliterative revival). A less obvious case is the replacement of the indigenous 3rd pl pronouns by their Scandinavian counterparts. In 1250, London English had 3rd pl forms with shapes like he/hy (nom), hem (dat/acc), and here (gen), with much spelling variation in texts. By 1450 we find the Scandinavian forms łai, łaim, and łayr (with much spelling variation). The Northeast quadrant of England was subjected to extensive Scandinavian influence beginning with raids in the 8th century, then settlement and imposition of Danelaw until the mid 11th century. And this pronominal replacement takes place first in the North and spreads South, a clear case of outside influence. Many morphological changes followed this same geographical spread: the loss of case endings proceeds from North to South (Allen 1995: Ch 5), as does the spread of the predicate and postpositive genitive forms in -s: This is yours; a friend of yours, as opposed to your friend (Mossé 1952: 59). It is Inertia which justifies our assumption in historical studies that what needs to be explained are the changes. Inertia explains the nonchanges. Ironically, Inertia is the most widely used of the principles we invoke. Reason: a small change coupled with the maintenance of other properties (Inertia) can lead to a noticeable change in overall pattern.

328

Edward L. Keenan

Here is one such case, one that helps us apprehend the sense in which synchronic data, (3a, b), can be historically explained. (3) a. In Modern Standard English single word modifiers of Ns precede them: a sleepy lion, no living athlete, every criminal lawyer, your plastic spoon. b. But contrary to (3a) many nominal modifiers in a- follow their N: a lion asleep/*an asleep lion, no man alive/*no alive man, every passenger aboard/*every aboard passenger To explain (3b) observe first that in Middle English (as in Modern Standard English) PP modifiers of Ns follow them, as the examples below from Chaucer c1385 illustrate: (4)

lordship over youre persone, folk in sorwe, recours to the juge, recchelesnesse in spekynge, servauntz and thralles to synne, another remedie agayns Leccherie, humylitee in confession

We explain (3b) on the grounds that the a- modifiers, alive, asleep, ... used to be PPs headed by the preposition on, which procliticized to the following N by FWP reducing to a-, see (1). The resulting expression didn’t change position or interpretation (Inertia).

2.

Two semantic universals: constituency interpretation and anti-synonymy Constituency Interpretation (CI) The constituents of an expression are semantically interpreted

CI follows from Compositionality - the semantic interpretation of a complex expression is a function of the interpretations of the expressions it is built from. Here we only need the weaker Words have meanings. Independent support for Compositionality and thus CI is that it explains how we understand novel utterances: we understand what their ultimate parts (the lexical items) mean and how things constructed in that way take their meaning as a function of the meanings of their parts.

Explaining the creation of reflexive pronouns in English

329

Anti-synonymy (AS) Different words mean different things Direct support for AS is the relative rarity of pure synonyms and the tendency for near synonyms to semantically differentiate: e.g. pork, mutton, beef are reserved for the cooked foodstuff, but their original French sources could refer to the live animal as well. Kuno (1987: 64–67) discusses subtle semantic differences between the local bindings in Johni pulled the blanket over himselfi /himi. Clark (1993), from whom AS is taken, and CarstairsMcCarthy (1999) cite psycholinguistic support for AS. I am inclined to think that AS derives from a learning dynamic Gainful Learning: New words mean new things. Language learners are motivated to expand what they can say and this goal is furthered by assigning new meanings to new words. So Gainful Learning can itself be thought of as a particular case of Inertia.

3.

The anaphora profiles of Modern Standard English and Old English

We present first the anaphora properties of Modern Standard English that will be historically derived.

3.1 Properties of self-forms in Modern Standard English 3.1.1

Locality. In (5) himself is understood to be (semantically) bound to its co-argument the king; it cannot be bound to John or refer to a third party.

(5) John denied that the king defended himself By contrast the pronoun him replacing himself cannot be bound by the king but can be bound by John or refer to a third party. So, as is well known, selfforms in Modern Standard English must be locally bound when occurring as the complete (henceforth bare) argument of a predicate, and the personal pronouns (him, her, them, …) cannot be locally bound (= must be locally free) there.

330

Edward L. Keenan

3.1.2

self-forms do not occur as bare subjects: (He says that) *himself/he is clever Pronoun subjects cannot be referentially dependent on co-arguments (Keenan 1976, Lee 1993). E.g. in She criticized herself/the queen the pronoun she must refer independently of herself or the queen.

3.1.3

self-forms occur non-locally bound in ICEs (Inherently Contrastive Expressions):

a. Exception NPs: Each studenti thought that no one but himselfi would get an A b. Coordinations: Johni tackled a problem that neither himselfi nor the teacher could solve c. Comparatives: No onei likes to work with anyonej smarter than himselfi,j 3.1.4

self-forms occur as non-arguments: adverbially and in apposition to definite NPs, always with a contrastive value:

a. John washed the floor himself He is himself a vegetarian b. John himself washed the floor John interviewed the queen herself 3.1.5

self-forms are not possessors; NP’s + own expresses emphatic/contrastive possession:

a. He is using his own/*himself’s pen b. John’s own father didn’t vote for him

Explaining the creation of reflexive pronouns in English

331

3.2 The Anaphora Profile of Old English 3.2.1 Local (and non-local) binding is done with the ordinary personal pronouns. (6) a. syDłan he hine to guDe gegyred hæfde. once hei (nom) himi (acc) for battle girded had

Beowulf 1473 c750

b. wit unc wiD hron-fixas werian łohton Beowulf 540 wei (1.nom.dual) usi (1.dat/acc, du) whale-tusks protect c750 thought ‘We thought to protect ourselves from whale-tusks’ c. ła gegyrede heo hy mid hærenre tunecan ond ... then dressed shei heri (acc) in a tunic of hair and ...

Mart 190 c875

d. ... Dæt hie sylf hie Deowen nemde BlHom 5.I.13 c971 ... that shei (nom) self (nom) heri (acc) handmaiden calls ‘... (that was undoubtedly modesty) that she herself calls herself handmaiden’ e. ... hweDer he hine gefreclsian wolde ... whether hei (nom) himi (acc) set-free would

BlHom III.29 c971

f. Sibyrhtes broDor and swiDe mænig ołer ... cene hi Mald 282 weredon c1000 [Sibyrht’s brother and very many others]i ... bravely themi(acc) defended

3.2.2 Non-theta (pleonastic) pronouns in Late Old English Late Old English makes extensive use of non-theta (pleonastic) pronouns. These are uses of dative, sometimes accusative, never nominative or genitive pronouns which are not required by the verb and do not satisfy its semantic role requirements (whence “non-theta”). They always take the local subject as antecedent, agreeing with it in person, number and gender; semantically they may heighten the involvement of its referent: e.g. subject acted intentionally or was affected by the action. They may suggest a telic interpretation of the action. In the Saxon Chronicles from 999 to 1067 we find 58 locally bound dative pronouns: 57 are non-theta, one is theta. Pas-

332

Edward L. Keenan

sages may be peppered with them (7f). They may take quantified NP antecedents, not just definite ones (7b, d). (7) a. for Dæm hi him ondrædaD Da frecenesse De hi ne gesioD because they fear them the danger that they do not see

CP 433 c880

b. ... on łæm magon nigon men standende him gebiddan, ... Mart 52 ... on which nine men may stand and pray them ... c875 c. ... & namon him wif of eallum Dam De hi gecuron. ... and (they) took them wives from all those that they chose ... acceperunt sibi uxores ex omnibus quas elegerant (the Latin original)

ÆGen 6.2 c1000

Chron(E) d. ... ge swicon ła łære fyrding. & færde ælc mann him ham. 1016 ... (they) abandoned then the expedition and went every man him home Dream e. Aledon hie Dær linwerigne, gestodon him æt his lices heafdum indicate words or letters that were lost because of damage and are supplied through editorial conjecture.

References Biber, Douglas and Edward Finegan 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. (Topics in English Linguistics 10.) Berlin: Mouton de Gruyter. Brinton, Laurel 1990 The development of discourse markers in English. In: Jacek Fisiak (ed.), Historical Linguistics and Philology, 45–71. (Trends in Linguistics, Studies and Monographs 46.) Berlin: Mouton de Gruyter. Culpeper, Jonathan and Merja Kytö 2000 Data in historical pragmatics: Spoken interaction (re)cast as writing. Journal of Historical Pragmatics 1/2: 175–199. Cusack, Bridget (ed.) 1998 Everyday English 1500–1700. Edinburgh: Edinburgh University Press. Fludernik, Monika 1995 Middle English po and other narrative discourse markers. In: Andreas H. Jucker (ed.), Historical Pragmatics: Pragmatic Developments in the History of English, 359–392. (Pragmatics & Beyond New Series 35.) Amsterdam: John Benjamins Publishing Company.

Reporting direct speech in Early Modern slander depositions

415

Görlach, Manfred 1991 Introduction to Early Modern English. Cambridge: Cambridge University Press. Hope, Jonathan 1993 Second person singular pronouns in records of Early Modern ‘spoken’ English. Neuphilologische Mitteilungen 94: 83–100. Helmholz, Richard H., ed. 1985 Select Cases on Defamation to 1600. London: Selden Society 101. Ingram, Martin 1987 Church Courts, Sex and Marriage in England, 1570–1640. Cambridge: Cambridge University Press. Jucker, Andreas H. 1997 The discourse marker well in the history of English. English Language and Linguistics 1/1: 91–110. Jucker, Andreas H. and Irma Taavitainen 2000 Diachronic speech act analysis: Insults from flyting to flaming. Journal of Historical Pragmatics 1/1: 67–95. Kryk-Kastovsky, Barbara 2000 Representations of orality in Early Modern English trial records. Journal of Historical Pragmatics 1/1: 201–230. Labov, William 1994 Principles of Linguistic Change: Internal Factors. Oxford: Blackwell. March, John 1647 Actions for Slaunder, or a Methodicall Collection under certain Grounds and Heads, of what words are actionable in the Law, and what not? London: M. Walbank and R. Best. Nevalainen, Terttu 1999 Making the best use of ‘bad’ data. Neuphilologische Mitteilungen 100: 499–533. Partee, Barbara Hall 1973 The syntax and semantics of quotation. In: P. Kiparsky and S. Anderson (eds.), A Festschrift for Morris Halle, 410–418. New York: Holt, Rinehart and Winston.

416

Colette Moore

Raine, James, ed. 1845 Depositions and other Ecclesiastical Proceedings from the Courts of Durham. London: J.B. Nichols and Son. Richman, Gerald 1986 Artful slipping in Old English. Neophilologus, 70/2: 279–290. Sharpe, James A. 1980 Defamation and Sexual Slander in Early Modern England: the Church Courts at York. Borthwick Papers 58. York. Trotter, David A. 2000 Multilingualism in Later Medieval Britain. Cambridge: D.S. Brewer. Visser, Fredericus Theodorus 1963– A Historical Syntax of the English Language. Leiden: Brill. 1969

The emergence of the verb-verb compound in twentieth century English and twentieth century linguistics* Benji Wald and Lawrence Besserman

1.

Introduction

The twentieth century emergence of the endocentric VV (verb-verb) compound as a unified and productive construction posed an interesting set of problems and challenges for twentieth century and later historical and synchronic linguistics. Examples are: crash-land, freeze-dry, slam-dunk, stirfry, strip-search, among an increasingly large number of items of the same type. The historical problems center around the question: why is it that this addition to the ways of forming verbs among English word formation processes has shown signs of becoming highly productive only since the middle of the twentieth century? Why didn’t it happen sooner? In approaching an answer to such questions we first need to consider the nature of the VV compound as an object which has evolved, and then the evidence for its path of evolution over time. In turning to previous studies, we do not find anything specifically addressing the VV compound until late in the twentieth century. However, earlier more general discussions of the word formation processes are relevant to views of the nature of VV. Most often these discussions specifically involve NV, e.g., apartment-hunt, baby-sit, whistle-blow, but imply the same analysis of VV, had the latter pattern been distinguished. Concerns about NV are most intimately related to concerns about VV in the very frequent apparent ambiguity of category of the first constituent of the compound, e.g., sleep – noun (N1) or verb (V1)? – in sleep-walk. We will expand on this issue in section 2.2 below. VV seems to have eluded discussion and even exemplification in most major treatments of English throughout the twentieth century, from Sweet (1881/1905) through Quirk at al. (1972) and its successor Quirk et al.

418

Benji Wald and Lawrence Besserman

(1985). Marchand’s classic study of English word formation processes (1960, rev. 1969) includes a possible example, type-write, as noted by Bauer (1983: 208), but Marchand argues that the appearance of compounding in this and similar items is parasitic on its conversion or backformation, as opposed to a productive directly formed compound type, usually NN (e.g., type-writer). In his view, the crucial difference between direct formation, on the one hand, and back-formation or conversion, on the other, is in the semantic interpretation of the construction. Only direct formation necessarily refers to the internal structure of the compound for its semantic interpretation. He argues that the verb type-write is not interpreted in reference to a semantic relationship between its internal constituents, but in reference to the synthetic compound noun type-writer, as embedded in the paraphrase “use a type-writer”. Marchand considers all compounds except Particle + Verb to be of this syntactically and semantically parasitic type, and hence labels them “pseudo-compounds”. Thus, he does not overtly recognise the endocentric VV type that we are concerned with. Our subsequent discussion will show that we accept Marchand’s view on direct formation, but not his implicit view of how the VV compound is currently constructed and understood. Pennanen (1966) is exceptional for his time in including many examples of possible VVs, e.g., dive-bomb, kick-start, sleep-walk. He dissented from Marchand by viewing the great increase in the NV type from the late nineteenth through the early twentieth century as evidence for a change in the status of back-formation as a synchronic process from sporadic to highly productive (1966: 150). In an even more radical departure from Marchand’s position, he argued that he had detected some movement toward direct formation of NV (e.g., chain-read modeled on the probable back-formation chain-smoke) (1966: 115). However, in focusing on the issue of backformation, he did not distinguish VV from NV, e.g., ice-skate, skin-dive, spoon-feed, etc. The issue of the productivity of NV, and implicitly VV, remained controversial throughout the twentieth century (cf. Kastovsky 1986: 419, Sˇtekauer 1998: 155–60). Faced with a continually rising number of VVs in the last quarter of the twentieth century, a younger generation of synchronic linguists began to explicitly recognise the VV compound, and note problems in its analysis. Thus, Adams (1973: 109–10) devotes some attention to drip-dry. Among strictly synchronic generative studies, Selkirk (1982: 16–18) is exceptional

The emergence of the verb-verb compound

419

for her time in overlooking examples of the VV compound and explicitly denying its existence. In contrast, Lieber (1983: 265) takes the type for granted, but notes problems in how to analyse it. In response to problems noted by Lieber, Boase-Beier (1987: 75–80) discusses the grammatical analysis of VV at greater length. We will return to Boase-Beier’s proposal in the next section. To proceed with the evolution of the VV compound, we organise discussion as follows. In section 2 we survey the major analytical problems involved in synchronically characterising the VV compound. From these we select the problem of distinguishing VV from NV as an issue for historical development. In section 3 we begin to examine the evolving distinction between NV and VV through the development of grammatical preconditions in late Old English (OE) and early Middle English (eME). In section 4 we provide empirical evidence for the evolution of the distinction between NV and VV from late Middle English to the end of the twentieth century. In section 5 we summarise our findings and draw conclusions about the nature of VV and why it has emerged so recently.

2.

Problems in VV analysis

Problems in VV analysis involve the various aspects of its internal structure. Underlying the dependence of the semantic interpretation of VV on the relationship between its two verb constituents are the distinct and independent issues of headedness (V2) and category analysis of V1. 2.1 Headedness. As an endocentric type, VV compounds are head-final. V1 modifies and restricts semantic interpretation of the head V2. Thus, to crash-land is to land in a specific way, “by” crashing. VVs are generally paraphrasable as V2 + Preposition + V1 -ing, but the semantic relationship between the modifier and the head can vary, e.g., to drop-kick is not to kick “by” dropping, but to kick “after” dropping. The analysis of VV as a headed compound conflicts with Marchand’s analysis of “pseudo-compounds” as synchronically transparent backformations or 0-derivations, cf. the possible conversion of slam-dunk from a nominal to a verb compound, without reference to its internal structure. Adams’ (1973, 110) attestation of written drip-drys rather than drip-dries, as an inflected verb, was intended to indicate a lack of internal analysis of

420

Benji Wald and Lawrence Besserman

this VV compound. Treatment of V2 as a strong verb provides clearer evidence for its status as an independent constituent. This is always the case for appropriate verb compounds we recognise as VV. However, there are examples of NV compounds that reveal that historical conversions and back-formations are subject to further analysis which restores their internal structure. Thus, the Oxford English Dictionary, second edition (OED2), first attests the NV joyride in the inflected form joy-rided [1915], rather than joy-rode. The regularisation of the past tense of ride reflects a lack of analysis of the internal constituents, suggesting semantic interpretation by reference to 0-derivation from the NN joy-ride [1909] or back-formation from the synthetic NN joy-rid#er [1910]. Similarly, Quirk et al. (1972: 1029) expressed discomfort with a past tense for the NV baby-sit, as a back-formation from baby-sitt#er, but resolved that discomfort in favour of the strong past baby-sat in Quirk et al. (1985: 1579). In such cases, Marchand’s pseudo-compound verbs are historically subject to further reanalysis which restores their internal composition, including headedness, and erases the historical priority relationship between the nominal and verb forms of such compounds for later speakers. Adams’ example of dripdry+s > drip-dries suggests that VV may similarly emerge through loss and subsequent reanalysis of internal structure. The essential feature of headedness is that the VV compound takes on the independent semantic and syntactic features of V2, regardless of the autonomous features of V1. This observation contradicts Boase-Beier’s (1987: 75–80) judgment that in grammatically well-formed VVs both constituents must be coordinated for transitivity, such that the object of a transitive VV must also be the object of V1 taken individually. Attested counter-examples include switch-hit, dive-bomb, sleep-walk, drop-forge, among many others. For example, “to switch-hit the ball” is “to hit the ball”, but not “to **switch the ball”, and similarly for all other examples; “to drop-forge a metal object” is to forge the object by dropping a weight on it as part of the forging process, not “to **drop the object”. The head alone determines transitivity relations. The headedness of VV compounds deserves more synchronic and historical discussion than we can devote to it in this paper. Our primary purpose in mentioning it here is to demonstrate that VV compounds have internal structure. Our primary concern from this point on will be the first constituent, V1.

The emergence of the verb-verb compound

421

2.2 The ambiguous category problem. The ambiguous category problem is the problem of whether the first constituent of a compound should be analysed as verb or noun, V1 or N1, or indeed whether it has a relevant categorial analysis at all. In some cases, the evidence for V1 hinges on the lack of a plausible N1 alternative. For example, in the VV stir-fry, the first constituent stir is highly syntactically restricted as an independent nominal with an appropriate meaning. Thus, the analysis of stir as a nominal in the stereotyped frames “give it a stir” and possibly “have a stir” can be questioned, despite appearances. Nevertheless, in the majority of the cases that we have observed, the first constituent has a semantically appropriate N1 analogue without further syntactic restrictions, e.g., crash-land, divebomb, kick-start, sleep-walk. Speakers themselves can indicate distinctive analyses by paraphrase, and may disagree. For example, among naive speakers we have questioned about paraphrases for tap-dance, some paraphrase as “dance by tapping (feet, shoes)”, evidence of a V1 analysis, while others paraphrase as “dance with taps (on shoes)”, associating tap with a lexicalised concrete noun referring to an instrument enabling the activity. 1 Among documented examples of possible reanalysis of NV as VV are play-act and press-pack. The example play-act cautions us to recognise that semantic shift may allow the possibility of a V1 analysis without insuring that analysis, providing room for disagreement among contemporary speakers, as in the case of tap-dance, mentioned above. The first OED2 citation of play-act [1882] is unproblematically NV, where play is a reference to a theatrical performance. However, while the head verb act alone has continued to dominate this sense, play-act was later deflected to a sense in which the suspension of disbelief, required by theatrical performance, gave way to a focus on disbelief modifying the activity expressed by the head, i.e., to the sense of “act insincerely, pretend, feign”. Analysis of the V1 play- with this function is reinforced for play-act by numerous more recent examples, e.g., play-fight, -kill, -bite (none listed in OED2 or Additional Supplements [OEDAS] as compound verbs). These examples do not resolve the ambiguous category problem, but they easily allow the V1 analysis of play- alongside the N1 analysis of “act, fight, kill, etc. in play”. In press-pack, V1 is currently preferable to N1 as an analysis of press-. The OED2 analyses press-pack [1864] historically as NV, citing Webster’s definition: “to pack ... by means of a press”. However, it analyses the more recent compound verb press-fit [1970] as VV. In this case, the obsolescence

422

Benji Wald and Lawrence Besserman

of the concrete object referred to in press-pack before the coinage of pressfit eliminated an appropriate nominal analysis of press-. The compound verb press-pack itself may be currently obsolete, but it remains readily interpretable to speakers under a V1 analysis as “pack by pressing”. In sum, reanalysis of NV as VV is often a possibility rather than a certainty, allowing for disagreement among contemporary speakers. We will allow for the ambiguous category problem in the analysis of VV by considering such items possible VVs, rather than indisputable VVs. Ultimately we will not suggest that reanalysis of what may be historically N1 as V1 is a cause of the emergence of the VV compound as an independent type, but that it is a consequence of the more general establishment of a word formation process producing VV compounds. We will propose that the more general existence of the VV compound has been established by the unification of historical NVs with VVs of other historical sources. The following subsection completes discussion of the current nature of the possible VV compound by considering a constraint which provides further support for a V1 analysis at the expense of an N1 analysis for additional verb compounds. 2.3 The activity constraint. The activity constraint imposes a formal and semantic economy on the internal structure of a compound verb by limiting it to what is necessary to semantic interpretation of that compound verb, and suppressing what is unnecessary in that context. Thus, when nominal or adjectival marking alters the syntactic properties of the first constituent of a compound, but does not otherwise alter its semantic properties, it is suppressed in favour of the root verb alone in the verb form of that compound. In all the VVs we have found, the essential semantic content of the root verb as V1 is an activity of some type. Activity nominals (N1) and participles (A1) are suppressed as semantically unnecessary first constituents of compound verbs in favour of the activity semantics inherent in the root verb (V1) itself. The activity constraint reveals a historical source of possible VVs in addition to those already discussed. We will consider the activity constraint first in the context of the adjectival, i.e., participial, marking of the first constituent, because it is relatively straightforward, and then in the context of nominal marking, where additional complications further elucidate the nature of the constraint.

The emergence of the verb-verb compound

423

2.3.1 The activity constraint with participial first constituents (A1). Participles are adjectives derived from verbs by overt marking. They commonly figure as first constituents in adjectival (A A) and nominal compounds (A N) without constraint, but not in compound verbs. Participles distinguish active/passive voice and have syntactic effects when used in phrase level constructions, but they add no semantic information to the verb from which they are formed when they serve as the first constituent of a compound verb. Hence, they are avoided by the activity constraint in favour of their root verb, resulting in a VV compound. The logic of the constraint is, as discussed in 2.1 above, that the head (V2) alone determines the syntactic characteristics of the verb compound, including voice, rendering the participial marking superfluous. A model example is the VV spin-dye [1961] (referring to dyeing fibres during the spinning process), but not the AV **spun-dye, attested alongside a corresponding AA spun-dyed [1955] (cf. red-hot). Similarly, the possible VV force-land [1921], but not the AV **forced-land, is attested alongside the AN nominal forced landing [1917]. For the possible VV spin-bowl (a cricket term), the earliest attested nominal is spinning-bowling [1862], analysed by the OED as AN, i.e., spinning is a present participle. The activity constraint blocks the AV **spinningbowl. Among previous scholars, Jespersen came closest to recognising the activity constraint in his suggestion that the now rare VV mix-bathe [1906], referring to men and women swimming together, is “probably an abbreviated form of mixed-bathe” [1965a: 168]. Jespersen assumed this (AV) compound verb on the basis of the obvious source nominal mixed bathing, but the compound verb violates the activity constraint and is unattested. 2.3.2 The activity constraint with derived nominal first constituents (N1). In principle, the activity constraints operates in the same way with derived-nominal first constituents as with participial ones – by avoiding them in favour of their root verb when the derived nominal refers strictly to an activity and thus adds nothing essential to the semantics of the root verb. This is commonly the case when the derived nominal is a gerund (an -ing form discussed in further detail later). However, in contrast to participles, derived nominals frequently develop more specialised lexical meanings than mere activity, in which case elimination of the nominal marking could result in the loss of material relevant to semantic interpretation. As the ac-

424

Benji Wald and Lawrence Besserman

tivity constraint predicts, we do not find gerunds as first constituents of headed compound verbs. 2 But we do find examples of aggressive use of the activity constraint to de-lexicalise a lexically specialised derived nominal in favour of the root verb where this still does not result in loss of information necessary to interpretation of the compound. An example is the VV spell-check [1983], foreshadowed by the nominal spelling-checker [1980]. There is no attested 0-nominal spell in the appropriate sense to create the ambiguous category problem for the VV analysis. Presumably, spell-check arose historically from the back-formed verb spelling-check (from the N spelling-check#er), where spelling is lexicalised as a product nominal of the root verb activity, as in “that’s the wrong spelling”. Reanalysis of spelling as a gerund, as in “spelling it that way is wrong”, enabled the activity constraint to operate by eliminating superfluous morphology. VV interpretation of spell- check is paraphrasable as “check (words) after spelling (them)”. The VV blow-dry [1968], foreshadowed by the NN blower-dryer [1948], has a similar history. The N1 blower is lexicalised as a reference to a concrete object, a component of a blower-dryer. The NV blower-dry (not attested to our knowledge) does not conflict with the activity constraint. However, recognising the NN blower-dryer by its head as a kind of dryer, rather than as a kind of blower, enabled speakers to eliminate the nominal morphology as unnecessary information in favour of the semantics of the root verb, paraphrasable in the verb compound as “dry (hair) by blowing (on it)”. Examples like spell-check and blow-dry indicate a path of formation of VVs from NVs which does not depend on reanalysis of a 0-nominal N1 as a verb. It is further available to unattested cases, such as reforming the NV roller-skate [1928] as the VV **roll-skate, paraphrasable as “skate by rolling” rather than “skate on rollers”. However, the constraint depends crucially on not eliminating information necessary to semantic interpretation. Thus, the NV sucker-punch [1844] cannot be reformed as **suck-punch, because the semantic relation of sucker to the verb suck is obscure, and would result in loss of information necessary to semantic interpretation. Similarly, we expect -ing to be preserved in cases like build **(-ing)-inspect (back-formed from building inspector) and wed**(-ing)-crash (modeled on the NV gate/party-crash). These -ing nominals are not direct references to the activity expressed by the root verb, but further specialised

The emergence of the verb-verb compound

425

lexicalisations which cannot be recovered from the root alone, e.g., building is a concrete product nominal, wedding is a reference to the cultural institution of a wedding as a type of ceremony, for which the root verb wed is currently disprefered in favour of marry. By distinguishing gerund uses of -ing from lexicalised -ing nominals, the activity constraint indicates an internal analysis of the VV compound that recognises the independent semantic contribution of the first constituent to the semantics of the compound verb. In sum, the activity constraint reduces non-lexicalised synthetic forms to their constituent V1, but only for verb compounds, not for nominal or adjectival compounds. This behaviour distinguishes verb compounds from nominal/adjectival compounds, and provides a source of VV in addition to the paths of reanalysis discussed earlier. It is worth noting that in most cases the historical record does not indicate an intermediate stage of backformation for VVs formed by this process, in which case their internal analysis would have to be subsequently restored in order for the activity constraint to operate. Instead, it seems that the VV model with its internal analysis already existed for coiners of new VV compounds, who took the elements of the nominal or adjectival compound sources and fashioned them directly into that VV model, eliminating the unnecessary marking in accordance with the activity constraint. The activity constraint suggests that V1 is a preferable analysis to N1, just in case that N1 is a 0-nominal restricted to interpretation as an activity nominal. In this case, N1 analysis adds nothing to the semantic interpretation of the first constituent, and thus is superfluous in the context where nominal syntactic properties are irrelevant. If the first constituent of a verb compound is interpretable as a more specialised lexicalisation of a 0-nominal, the N1 analysis persists. Frequently the ambiguous category problem persists because either an activity or a further lexicalised interpretation are both possible. For example, the first element race in the compound verb race-walk [1983] (Barnhart 1991) allows either a VV or NV analysis. If race refers directly to the activity denoted by the V race, it is V1. However, if it refers more directly to the cultural institution of a race as a type of contest, then it is N1. Again, we expect variant analyses, even individual hesitation, in such cases.

426

3.

Benji Wald and Lawrence Besserman

The history of grammatical preconditions for VV.

In approaching the longer term history of the VV compound, and why it has emerged so recently as a productive process, we turn to the historical preconditions for its development. We will not be able to discuss all the preconditions in the detail that they deserve. However, we will discuss in considerable historical detail the major role that the ambiguous category problem has played in the evolution of VV, and how many evolving preconditions have given momentum to the subsequent rise of V1 in both nominal and verb compounds, including factors facilitating the reanalysis of N1 as V1. 3.1 Compounding in Old English. Scholars agree that PV, e.g., download, upgrade, among many current examples, is the oldest productive English compound verb type (Sweet 1905, Lass 1994), already abundantly attested in Old English. Marchand (1960/1969) considered it still the only genuine compound verb type in the mid-twentieth century. The age of the NV type is much less certain, though most studies of Old English list a number of examples in which V2 is also attested as an independent verb, e.g., morgen-wacian (morning-wake), ellen-campian (zeal-fight ), geDancmetian, (thought-measure, i.e., deliberate/think over) [Kastovsky 1992: 374, Lass 1994: 198]. Lass considers the possibility that the type may predate English, on the basis of a few examples in other old Germanic languages, e.g., Gothic feihu-geironjan (wealth-lust = be greedy, Skeat 1868: 62). However, like other scholars, he notes that NV is sporadic; it may be the product of back-formation from nominal compounds, operating independently but sporadically in Old English, Gothic and other ancient and modern Indo-European languages. Whatever the grammatical status of NV in Old English and later, it contrasts with VV in that there are no examples of VV in Old English. V1 does not occur in any compound type until late Old English. At that point the ambiguous category problem arises, not in the context of compound verbs but in the context of compound nouns. 3.2 VN in late Old English. Scholars agree that VN arose late in the Old English period (Koziol 1972, Lass 1994, Marchand 1969, Visser 1967). Old English had previously followed the ancient Germanic practice of nominalising the verb root by addition of a nominal stem forming suffix, re-

The emergence of the verb-verb compound

427

sulting in an NN compound, cf. Old Saxon skek-i-fat (pour-*ja/o: -vat), Old High German trag-a-betti (carry-*a-bed = portable bed), where the starred element is the Proto-Germanic nominaliser (cf. Voyles 1992: 278). In some cases nominalisation was accomplished by ablaut, e.g., OE ra:d-hors (ride]v ]n -horse, cf. Sauer 1992: 190). By the late Old English period the integrity of the noun/verb class systems had been greatly eroded, largely due to the merger and variable loss of unstressed vowels. The consequence was the loss of overt marking of many verb derivatives as nominals. This created the conditions in which the ambiguous category problem could emerge, allowing various N1s to be reanalysed as V1s (cf. Sauer 1992: 725). Here we will suggest that while there is no reason to doubt that V1 did establish a beachhead among compound nouns by the late Old English period, there was widespread and prolonged resistance to it, continuing well into the twentieth century in various contexts and by various means, so that VV has only recently gathered enough momentum from its historical struggles to emerge as an increasingly productive compound type. Our major evidence for resistance to V1 in late Old English is the connective -e – mediating the two elements of a possible VN compound. We suggest that this “link-vowel” is an attempt to retain the older Germanic nominalising pattern, and to resist the reanalysis of N1 as V1. This linkvowel is reminiscent of Campbell’s (1959: 153) “parasitic connecting vowel”, associated with a nominal extension, usually the earlier OE -a nominal stem, e.g., wer-e-wulf. Thus, it is the sign of a final nominal boundary, -e-]n. Campbell (1959: 149ff) notes that in some compounds it remains where it should have been deleted by regular phonological processes, and that in other cases it is inserted unetymologically. Though Campbell characterises hierst -e-panne (fry-e -pan) as a VN compound, he accounts for the link-vowel as a reflex of an -ia (rather than the expected -i) nominal stem formative. Strictly speaking, then, hierste-panne is not a VN compound, but an NN compound in which N1 is a derived form of V1. VN seems to be a retrospective analysis anticipating the later loss of the linkvowel as a nominaliser. We have found various other cases where a link-vowel mediates between the two elements of a possible VN compound, most interestingly in cases where there is no independently attested nominalised form of V1, e.g., hwete-sta:n (whet-e -stone) alongside hwet-sta:n, et-e -lond (eat-e -land = pas-

428

Benji Wald and Lawrence Besserman

ture), grind-e -to:D (grind-e-tooth = molar), all listed in Clark-Hall (1960). Whether or not -e- indicates a contemporary pronunciation, we suggest that it reflects resistance to underived V1 in favour of N1, possibly even when N1 does not occur as an independent word. Thus, the connective reflects ambiguity for at least some speakers of the period as to whether possible VN should be analysed as V1-N2 or as V1(-e-)]n -N2. More conservative speakers would be resisting V1 by perceiving the much more firmly established NN pattern. In essence, we are suggesting that -e- functions as an overt marker of what would later become the 0-nominal. We hasten to add that we are not suggesting this as the solution to the ambiguous category problem, i.e., that all later apparent V1s are descendants of (or otherwise to be analysed as) 0-nominals. The evidence we are about to discuss indicates a more complex evolutionary path, by which V1 succeeds in establishing itself amid prolonged resistance. A final comment on the late Old English period is that we have found a few examples where a linking -e- is not attested between V1 and N2, e.g., ceorf-seax (carve-knife = scalpel), and yet no independent N1 **ceorf is attested, favouring an unambiguous V1 analysis. 3.3 VN in early Middle English. The link-vowel continues into early Middle English. Sauer (1992: 187ff) calls it a “Fugenelement”. He suggests that in the possible VN context it may derive from the OE -i-an infinitive, for which it would be a regular reflex, e.g., her-e-word (< *her-ian-word; praise-e-word). He compares it with the -e(n) reflex of all Old English infinitives in the early Middle English period, without suggesting that this is an appropriate analysis of -e- in early Middle English. The following table compares his 52 early Middle English cases of possible VN (1992: 188–194) where the root verb is extended by -e- or not extended at all (-0-).

Table 1.

Sauer’s corpus of early Middle English VNs in which V is extended by -eor -0-

continued from OE eME coinages Total eME corpus

-e-

-0-

Total

6 14 20

13 19 32

19 33 52

The emergence of the verb-verb compound

429

Excluded from Table 1 are the two additional forms fæst-en- dæg (fast-enday), where -en- may be an orthographic variant of -e-, and grind-el- stanes (grind-el- stones), where -el- unproblematically nominalises the root verb. The increase in the proportion of the link-vowel among early Middle English coinages may suggest that the threat of a V1 analysis of-0- was increasingly perceived from late Old English to early Middle English times, so that -e- was inserted, at least orthographically, as resistance to this analysis. With the continuation of phonological erosion, categorial ambiguity increased in early Middle English, providing further impetus to the development of VN. The majority of Sauer’s possible VNs (about 36/54) are ambiguous with NN, whether or not extended. Probably for this reason, he relies on semantic criteria for recognising possible VNs. But in so doing he sometimes ignores formal criteria. For example, he includes the early Middle English coinage lod-cniht (lead-servant = travel-leader, cf. OE la:d-mann ‘lead-man’) as VN, while acknowledging that lod is unambiguously a nominal form of “lead”. Accepting Sauer’s semantic insight, such examples continue the earliest Old English practice of deriving nouns from verbs by any available means to occupy first position in a nominal compound (cf. OE ra:d-hors > eME rode-horse, 1992: 191). Thus, such examples represent another form of resistance to V1. As Middle English evolved further, the variety of means for nominalising verbs of Old English origin diminished for various reasons. For example, lexically determined nominalisations like lod developed specialised meanings so that they were no longer available as activity nominals. For phonological (and related orthographic) reasons, -e-, already variable in late Old English, became increasingly unreliable as an indicator of nominalisation. Such developments interact with the Middle English generalisation of the -ing form of nominalisation. We turn now to -ing as an increasing form of resistance to V1 during Middle English and later periods. 3.4 The rise of competition between V1 and the -ing gerund. Already in Old English there are numerous NNs in which N1 is a deverbal -ing form. Visser (1967: 1112) lists 90 examples of Old English V-ing-N compounds. In addition, we find examples of incipient variation between V1 and V1-ing [= N1] during this period, e.g., ceorf-seax (carve-knife = scalpel) and ceorf-ing isen (carving iron = carving knife), both listed in Clark-Hall (1960). The use of the gerund instead of the bare root verb is a portent of

430

Benji Wald and Lawrence Besserman

competition which increases dramatically by the late Middle English period. It suggests that as the linking -e- and other means of nominalisation grew increasingly unreliable for purposes of resisting the appearance of V1, the gerund became a major resource to continue that resistance. Scholars quite commonly characterise such variation from the Middle English period on as a form of competition (e.g., for early Middle English, Sauer 1992, from Middle English on, Visser 1967, Marchand 1969). For early Middle English, Sauer (1992: 197ff) provides such alternations as swi (-ing)-wike (be.silent-(ing)-week), where V1 is at issue, and fæst-yng -day (fast-ing-day) vs. fæst-en-dæg (fast-en-day), where variation between -ing and the “Fugenelement” (i.e., link-vowel) is overt (1992: 190). Visser (1967: 1113) observes continued competition into the mid-twentieth century, citing Galinsky’s (1952) suggestion that American English seems to prefer V1 forms like wish-bone, cook-stove, while British English prefers V1-ing (= N1) forms like wishing-bone, cooking-stove. Facilitating the use of -ing as an overt form of resistance to V1 is the well known generalisation of -ing as an activity nominal from a restricted Old English class of verbs to all verbs during the Middle English period (e.g., Sweet 1905: 461, Mustanoja 1960: 566–581). Thus, as the link-vowel became increasingly unreliable as a form of resistance to V1, the gerund had become available to continue the resistance. Sauer’s considerable corpus of 85 early Middle English texts, representing the language between 1066 and 1300, yielded only 28 V-ing-N items, as opposed to the approximately 55 items he analysed as possible VNs (1992: 735). It is evident that -ing took a considerable amount of time to develop during the Middle English period as a form of resistance to V1. We have already noted that the majority of the possible VNs are ambiguous by Sauer’s criteria (36/55), that 20 have the link-vowel, and that he includes as VNs a few overt NNs, e.g., lod-cniht, because first position is occupied by an activity nominal. Nevertheless, accepting Sauer’s counts, 34/55 possible VNs are early Middle English coinages, rather than continuations of Old English coinages, an increase of 62%, while 21/28 V-ing-Ns are early Middle English coinages, an increase of 75%. Thus, even in early English we can see V-ing leading possible V1 in productivity as the first constituent of a nominal compound. In our view, there are no grounds for dissociating the later Middle English surge in V-ing-N compounds from late Old English resistance to V1.

The emergence of the verb-verb compound

431

By the late Middle English period V1-ing compounding was a fully productive process for all verbs, such that Raab (1936: 26) declined to count examples. Among examples are compounds that competed with earlier possible VNs but did not widely survive into later periods, e.g., whettingstone [1432], alongside OE hwet (-e)-sta:n reflected in still current whet stone. Following the generalisation of -ing to all verbs, the progressive development of verb syntax by the gerund, beginning in the Middle English period, increasingly strengthened the semantic identity of V and V-ing as competing first elements of a compound (cf. Mustanoja 1960: 566–81, Jespersen 1965: 88–143, Visser 1967: 1183ff). As a gerund, -ing acquired the option of adverbial rather than adjectival modification and direct rather than genitival objects, e.g., [nominal syntax] “their prompt recount (ing) of the ballots” vs. [verb syntax] their /them promptly recounting the ballots”. The competition between VN and V-ing-N during the Middle English period helps us resolve the ambiguous category problem. If VN and V-ing-N are in competition, then we can examine the growth of VN as the result of a conflated option, schematised in (1) below where lower case n stands for “nominal” and is an identification of the category to which the material to its left belongs: (1)

V(-ing]n) –N

The brackets of the optionality minimise the amount of internal analysis of the compound to that which is motivated by overt morphology. The logic is that if -ing initially gathered momentum as resistance to the possibility of VN analysis of a compound, then eventual increasing productivity of VN indicates the overcoming of this resistance. This perspective facilitates reanalysis of the absence of -ing as indicating a shift to V1 rather than to a 0-nominal form of N1. The progressive development of verb syntax by the gerund reinforces this analysis by undermining the analysis of V-ing as unambiguously N1. Thus, it might be more precise to view the variation as V(-ing )(]n )-N. In section 4 below, we will indeed distinguish the history of competition between V1 and V1-ing in nominal compounds, according to the formula in (1) above, from the history of competition between V1 and 0-derived N1 in both nominal and verb compounds.

432

4.

Benji Wald and Lawrence Besserman

Quantitative study of the rise of V1 from late Middle English to the twentieth century

In order to perform a quantitative study of the evolution of V1 and VV from late Middle English to the end of the twentieth century reported below, we constructed a controlled data base of nominal and verb compounds that were coined during that time span. The controlled data base was constructed as follows. Our initial collection of possible VV items began with personally familiar items and further observation. Using the V1s of these items we alphabetically searched various dictionaries for other possible VVs (e.g., Webster’s Ninth 1988, Partridge 1984, Barnhart 1991). Eventually, we extended our dictionary searches to other frequently used verbs for further examples of VV. The most important dictionaries used were the OED2 and OEDAS (1993, 1997). We quickly became aware that the OED often did not recognise our analyses but analysed the first constituent of a possible VV as N1, as discussed further below. Considering the trend in the linguistic literature toward increasing recognition of VV during the course of the twentieth century, we undertook to investigate whether these changes were also reflected in the OED analyses. From this investigation we developed an OED data base as an external control on the historical pattern evident from our own analyses. We are well aware of problems with the OED, both in terms of completeness and accuracy of dating. Quite a few forms which we analyse as VVs do not occur at all in the OED, including many which are attested in other sources. Nevertheless, completeness not being a practical option, we use the OED as a control, under the practical assumption that it is representative, despite its limitations (cf. Willinsky 1994: 92ff). A further advantage to using the OED is that it tends to give historically motivated analyses of listed compounds or “combinations”. While in some cases its analyses seem arbitrary to us (as exemplified in our ensuing discussion), its general use of historical criteria in its analyses independently helps us reveal a change over time in the relative frequency of competing patterns, and increasing recognition of V1 first in nominal and then in verb compounds. The criteria we use for inclusion and exclusion from our control data base are determined by the ambiguous category problem of V1. As a point

The emergence of the verb-verb compound

433

of departure we listed all the possible VVs that had come to our attention by any means, and selected all their V1s in order to count the number of different endocentric nominal and verb compounds in which the V1s occur in the OED. In total, 89 distinct possible V1s emerged from our original data base, and all were matched at least by a nominal compound in the OED. Examples include: (2)

blow, copy, drip, fly, jump, practice, spin, trickle

In all cases they are relatively frequent and semantically general action verbs, mostly monosyllabic. They are generally the same kinds of verbs that occur in VP nominals, e.g., blow-up, spin-off, etc. This seems not to be a coincidence, but rather an indication of the susceptibility of verbs of such semantic generality to further semantic restriction by means of word formation processes less available to the semantically more complex polysyllabic Latinate stratum of the English lexicon. Perhaps the most complex V1 in our original list was practice, as in the possible VV practice-teach [1952]. This VV was not listed in the OED, but in Webster (1988). Nevertheless, our procedure obligated us to count all items listed in the OED in which practice figured as a possible V1, e.g., practice-room [1921]. Virtually all the V1s are also attested as 0-nominals. This is the essence of the ambiguous category problem. Indeed, the OED analyses practiceroom as NN, not VN. Accordingly, we accepted the OED analysis in constructing our control data base. In following the OED analyses, we recognise three distinct categories based on our original possible V1s; V1, N/V and N1. In our control data base, V1 refers strictly to those compounds which the OED analysed unequivocally as V, e.g., jump-start, shrink-wrap. All OED examples of V1 confirm the activity constraint. The category N/V refers to cases in which the OED noted ambiguity, and hesitated to choose between an N1 and V1 analysis, e.g., copy-book (cf. copying-book), scratch-paper. In hesitating over the analysis of the nominal force-pump [1659], hence N/V, the OED2 notes the variant forcingpump [1727]. However, it analyses force- in the verb force-feed [1934] only as N1, despite the nominal entry forced feeding [1901], synonymous with force-feeding [1909]. Our control data base must accept this classification, although the OED analysis seems arbitrary to us. Included as N/V are cases such as fly-wheel, where the OED2 listed all compounds under the noun fly, not the verb, but added, “... In many of these

434

Benji Wald and Lawrence Besserman

the first element may be really the verb-stem ...”. In the absence of more specific item identifications, we classified them all as N/V in the OED control data base. Pursuant to our focus of concern, we excluded from the control data base all items in which N1 could not be construed as an activity nominal, e.g., fly paper, ram head, press release, where the reference is to objects and institutions, not activities. Only activity nominals figure in the ambiguous category problem, and can compete with V1-ing as possible V1s. Included were examples potentially ambiguous for activity, e.g., ram-vessel (= vessel with a ram]n). This example might be construed as VN, i.e., a vessel for ramming, but our control data base classifies it only as NN, according to the OED analysis. The result of our procedure is that all included NNs are reanalysable as VNs, regardless of their historical origins. In addition to NNs construable as VNs, we included as a distinct category V-ing -Ns listed by the OED in all cases where they are gerunds of our original V1s, partially listed in (2) above. The exclusion of participles appropriately restricts our interest to competition between V1 and the V1-ing gerund. Thus, we included such items as flying school, spinning wheel (i.e., V1-ing]n -N2), but excluded such items as flying saucer, spinning bowling (i.e., V1-ing]a -N2). Finally, we excluded from our control data base all synthetic compounds involving productive morphological marking of V2 as a noun. Synthetic compounds are of great importance to the study of the historical sources of NV and VV, but they would detract from the emerging pattern for our present purposes. Thus, we exclude such items as blow-dryer, drip-feeding, trickle irrigation, regardless of whether they are attested earlier or later than the corresponding VV, or whether a corresponding VV is attested at all. In sum, the OED data base acts as a control on our original analyses in two essential ways. First, it takes category analysis of the first constituent out of our control; for us, all unmarked first constituents are simply possible V1s, with no further differentiation and no reliance on historical considerations. Second, it provides empirical control on first constituents by limiting them to those for which (a) competition between verbs and nominals is attested, and (b) possible VVs are attested. Table 2 below summarises our resulting data base for nominal compounds.

The emergence of the verb-verb compound Table 2.

N NVN/VV-ingTotal

435

OED control data base for studying the ambiguous category problem in competition between VN and V-ing-N (sources: OED2 and OEDAS).

OED example

types

tokens

possible VV

test-flight hangman force pump spelling bee

59 34 13 65 89

1028 396 219 615 2258

test-fly hang-glide force-feed spell-check

In Table 2, the “possible VV” column on the extreme right exemplifies the original data base from which V1s were selected for building the OED data base. The extreme right column “–N” presents the categories reflected in the OED analyses discussed above. These are exemplified in the “OED example” column. The “types” column counts the number of distinct verbs for each category, e.g., 59 distinct first constituents were analysed by the OED exclusively as N1, although we consider them possible V1. The “tokens” column reports the total number of distinct nominal compounds listed for each category. Thus, the 59 distinct N1s by OED analysis occur in 1028 distinct nominal compounds, indicating that most of these N1s are very productively used in compounding processes. Due to recurrence of the same possible V1s, by our original analysis, in different OED categories according to the particular compound at issue, the first three rows add up to more than the 89 distinct possible V1s which we collected in the original data base and have controlled with the OED data base. Recall from discussion in section 2.3 that the V-ing- example with V1 spell does not have a semantically appropriate 0-nominal. Therefore, strictly speaking, there is no competition between spelling- and spell- as a first constituent in nominal compounds. The same is true for look-, scan- and stir-. The 61 other V1-ing types recur without -ing in one or more of the other categories, and thus can be considered in competition with them. Since we are interested in the more abstract grammatical trends that have given rise to the particular lexical items that constitute the data base, rather than in the particular items themselves, there is no reason to impose further restrictions on our subsequent counts. We are about to show that the procedural controls we have already imposed are sufficient to reveal the evolution of V1 and VV.

436

Benji Wald and Lawrence Besserman

4.1 Quantitative analysis of the historical evolution of V1 in VN contexts The historical patterns we examine below rely on the first citation dates of the items collected in our OED data base. The first historical pattern we look at in Figure 1 is the relative frequency of VN and V-ing-N.

Figure 1.

Relative proportions of VN and V-ing-N from 1300–1992

Accepting the OED analyses does not detract from recognising the steady growth of the VN pattern at the expense of the V-ing-N pattern over time. VN is disfavoured from the late Middle English period, but steadily gains strength from the 18th century onward. It only becomes favoured in the late 20th century. The historical pattern of increase is clear.

The emergence of the verb-verb compound

437

To provide a contrast we also examined a pattern now rare outside of attributive position, Verb + Object]n (VO), excluded as non-endocentric from our main count, e.g., break fast, scare crow, pick pocket. When we combine VO with the distinct VNs and V-ing-Ns, we see a precipitous decline in the proportion of VO from a peak in the 16th century. VO]n is unrelated to the continuous rise of VN. Figure 2 introduces a new consideration.

Figure 2.

Proportions of VN and VP]n to their respective -ing nominals: 1500–1992

Here we compare the rise in VN with VP]n, e.g., blast-off , blowout, hangover, etc. Although VP]n may often be best analysed currently as a 0-nominalised phrasal verb, its well studied rate of increase in the last few centuries (cf. Kennedy 1920: 47–9, Lindelöf 1937) contrasts with a parallel nominal V-ing-P pattern, strongly favoured at the same time as the V-ing-N pattern in Middle English (cf. Denison 1993: 404). V-ing-Ps in the sample include only syntactically nominal examples of such items, as in (3)b below.

438

Benji Wald and Lawrence Besserman

(3)a. V P]n: hang over/s, blast-off/s, blow out/s, etc. b. V-ing]n -P: breaking/s-off, striking/s-out, passing/s-over The OED is usually accommodating in listing only nominalised examples of V-ing-P, and not gerund uses that properly belong to sentence grammar. The two are distinguishable in that only nominalised examples can pluralise; they do not have the verbal properties of gerunds, such as taking direct objects, adverbial modification, or allowing a constituent to intervene between V-ing and P, as exemplified in (4) below. (4)gerundive:

their (/them)suddenly breaking off relations/breaking relations off

nominal:

their (/**them) sudden breaking off of relations/ (**suddenly ) breaking of relations **off

Figure 2 confirms, in a more controlled way than in previous studies, that VP]n has become a highly productive nominalisation pattern since the 18th century. The figure shows that its rise is parallel to that of VN, at the expense of competing -ing nominals. In contrast to VO]n, the parallelism between the VN and VP]n patterns suggests that as a word formation process in which first position is occupied by a verb, VP]n may have assisted in overcoming residual historical resistance to V1 analysis of appropriate nominal compounds.

4.2 Quantitative analysis of the historical evolution of V1 in VV contexts. At this point we consider VV. Figure 3 adds only the compound verbs listed in the OED determined by our procedure. It omits any examples not entered in the OED, and a probably small number of examples listed by the OED that we were not aware of at the time, and thus failed to capture by our procedure, e.g., gape-gaze [1863]. The figure distinguishes the OED’s three categories of first position for both compound nouns and verbs. It supports two important points for us. The first point is that before the late 20th century none of the compound verb patterns are anywhere near as common as the parallel compound noun

The emergence of the verb-verb compound

Figure 3.

439

Proportion of compound verbs to compound nouns for each catagorial status: 1300–1992

patterns. Until the twentieth century the difference in relative frequency of all three compound verb patterns is negligible, because verb compounding as a whole is operating at a very low frequency. That changes dramatically in the late 20th century, when verb-compounding as a whole increases substantially. New VNs still outnumber VVs, but the gap is narrowing. In interpreting the above trend, recall that we have excluded synthetic VNs. In principle, if synthetic VNs are included, the total number of VVs will always be smaller than the total number of VNs, because there are at least two synthetic VNs that can be derived from any VV by -ing and -er suffixation, e.g., drop-forge [VV 1925], drop-forging [VN 1884], dropforger [VN 1957]. Our procedure emphasises the pattern of increase of VV at the expense of the size of the gap between attested VNs and VVs.

440

Benji Wald and Lawrence Besserman

The other point emerges from the persistent bunching of the three (OED) first position types from the 14th century period of marginality through the most recent period of substantially increased productivity. The historical patterns derived from the OED fail to show a distinction in their separate developments, suggesting that they are all examples of the same possible V1 position. Thus, the OED control data base provides independent confirmation of our initial analyses. The examples in (5) below are typical. (5)

OED NV = our possible VV: crash-land, force-feed, kick-start, etc.

Our initial selection of possible V1s presupposed that the sample of compound verbs in the data base could all be reanalysed as VVs, regardless of how they may have been historically differentiated in earlier periods. The lack of differentiation in the behaviour of the three first position types throughout the period of development of compound verbs supports our initial assumption that the examples of NV included in the sample are all reanalysable as VV, at least by some speakers, and probably by many.

4.3 Quantitative analysis of the historical resolution of the ambiguous category problem. Our final Figure 4 combines all compound verbs in the sample to provide further insight into the relationship between the OED analyses and our own initial analyses of all these verbs as possible VVs. The actual number of compound verbs in Figure 4 shows that as late as the late 19th century, our control sample is sparse for any compound Vs that we consider reanalysable as VVs. That situation changes dramatically in the 20th century, when the total number of compound Vs increases to respectable levels. As the data increase, we see a cross-over between V-initial analyses and ambiguous N/V-initial analyses by the OED’s own criteria. As new coinages have increased during the 20th century, the OED analysts show an increasing willingness to analyse first position as V1, rather than as N1 or ambiguous between N1 and V1 analyses. We suggest that for pretwentieth century coinages, the OED analyses are influenced by the same kind of historical resistance to the admission of V1 in first position of a compound that we have proposed as operant throughout the history of English, first for VN and then for VV. For late nineteenth and early twentieth

The emergence of the verb-verb compound

441

century coinages, the OED analyses may accurately reflect a more widespread conservative impressionistic analysis of that period, and are understandable in historical context as resistance to V1. We have already noted that for some entries the conservative analyses seem arbitrary with respect to linguistic justification, e.g., the OED analysis of force-feeding [1909] as N V-ing (probably through paraphrase as “feeding by force”), where the listing of synonymous forced-feeding [1901] (= V-en -V-ing) could have caused hesitation. Until the second half of the twentieth century it seems that the weight of history favoured an N1 analysis when possible. However, Figure 4 suggests that V1 had overcome the weight of history by the late twentieth century, and that OED analysts reacted accordingly, by showing an increasing tendency to resolve potential N/V ambiguity in favour of a V1 analysis. We read this as providing independent evidence not only for a future increase in VV coinages, but also for increasing reanalyses of other appropriate compound verbs as VVs by English speakers unacquainted with the historical sources of such compounds.

Figure 4.

Proportion of each first position catagory for total number of compound verbs: 1850–1992

442

5.

Benji Wald and Lawrence Besserman

Concluding remarks

On the basis of our current observations and twentieth century linguistic studies of English word formation processes, we conclude that the VV compound emerged in the twentieth century as an increasingly productive pattern with a bright future ahead of it. Its relatively recent emergence has been conditioned by the historical resistance it had to overcome. English has been historically inhospitable to the direct formation (or base derivation) of compound verbs, except for PV. It has also historically resisted V1 in all compounds. Even at present VA compounds are marginal, e.g., fail-safe, tamper (ed)-evident/resistant. V1 has been assisted in its development by a number of successive changes beginning in the late Old English period. Among them was the breakdown of the Old English system of noun/verb classes. This breakdown gave rise to both the ambiguous category problem and the generalisation of the -ing gerund to all verbs in the Middle English period. Together these two developments gave rise to a prolonged competition between the gerund and the ambiguous N/V stem in first position in nominal compounds. The progressive development of verb syntax by the gerund, beginning in the early Middle English period, reinforced the potential for the competition to be perceived as competition between two verb forms, at the expense of the historically nominal interpretation of first position. The activity constraint has prevented the same competition from spreading to first position of verb compounds. Its age is difficult to assess, because it operates only in the context of compound verb types that were rare until the twentieth century. The activity constraint seems to have started as a by-product of the various ways that compound verbs have been formed in all periods. However, in the twentieth century it seems to have become more actively involved in the production of VVs by favouring delexicalisation of derived N1 nominals to their root V1s when this is does not interfere with semantic interpretation. Despite the varied historical sources for VV, the growing recognition of the pattern in the late twentieth century literature on English word formation reflects its increasing productivity and the perception that it is a unified construction. We have suggested that the recent change in attitude in favour of unequivocal V1 analysis, especially in VV, reflected in OED analyses of twentieth century coinages, betokens the coming of age of the VV pattern for even more conservative English speakers. It seems that a

The emergence of the verb-verb compound

443

critical mass of familiar VVs has accumulated which are recognised as such regardless of their historical origins, and that, in addition, that critical mass establishes both the possibility of further direct generation of new VVs, and reanalysis of older compound verbs as VVs. 3 To conclude, the likely lines of development for VV compounds in the twenty-first century seem clear, although we shall have to wait and see. Among the items that we have noted in sources other than the OED are drip-feed, drop-ship, force-draft (a law), force-fit (a curve), freeze-etch, freeze-fracture, jerk-bump, jump-serve, play-bite, play-fight, play-kill, play-test, practice-teach, spit-shine, skim-read, skip-bomb and test-play. These and other items that continue to come to our attention suggest that we will not have to wait long.

Notes *

1.

2.

This paper is a revision of a lecture delivered at the SHEL-1 conference (first meeting of the Studies in the History of the English Language) at UCLA on May 20, 2000. We wish to thank the organisers Donka Minkova and Robert Stockwell for valuable additional information and editorial comment, and to the audience of the original lecture for a stimulating discussion. We are particularly indebted to the chair of our session, Dieter Kastovsky, for generously sharing his extensive knowledge of English word formation processes with us, and for providing us with a copy of Pennanen’s (1966) important monograph. We also gratefully acknowledge the helpful comments of two anonymous referees of a pre-publication version of the manuscript. The paraphrase method of uncovering variant analyses suggests a psycholinguistic research project which can be further systematised along the lines recommended more generally for empirical investigations of problematic grammatical constructions by Schütze (1996). Such research would supplement the evidence we will present in this paper for the reanalysis of N1 as V1, as VV has evolved as an independent and unified compound type. In earlier versions of this paper we also related the activity constraint to the avoidance of N > V conversion of V-ing-N nominals, e.g., the preference for “machinewash [the clothes]” over “washing-machine [the clothes]”. Though rare and not yet listed in dictionaries, we have found some examples of the conversion of V-ing-N to verbs, e.g., “to washing-machine [a man to death]”, “to shaving-cream [somebody’s car/doorknob]”. Unlike cases where the activity constraint operates, the semantics of these verbs is dependent on the semantics of the source nominal compound, e.g., “to use a washing-machine (to kill somebody)”, “to use shaving-

444

3.

Benji Wald and Lawrence Besserman cream (to play a prank)”. There is no head verb (V2) in such converted verbs, nor any other evidence of internal analysis. The semantics is quite different from machine-wash and the unattested but conceivable cream-shave, which are headed and interpretable on the basis of internal analysis alone. In the OED we have so far found only one V-ing-N conversion shoeing-horn [1650], converted from the now obsolete noun [1440]. The currently used conversion shoe-horn [1926] from the noun [1589] does not require an internal analysis for its semantic interpretation, though one may be possible by identification of V2 with “horn in (on)”. It is worth noting in passing that we are not alone in our assessment that VV has a promising future. Dieter Kastovsky (p.c.) has called to our attention a futuristic novel in which many examples of VVs, some apparently coined by the authors, are used for futuristic effect, e.g., “She glide-walked quickly toward the big holo tank ...” (David Brin and Gregory Benford, Heart of the Comet . New York, Bantham Books 1986: 167)

References Adams, Valerie 1973 Introduction to English Word Formation. London: Longman. Barnhart, David K. 1991 Neo-Words. New York: Collier. Bauer, Laurie 1983 English Word-Formation. Cambridge UP. Boase-Beier, Jean 1987 Poetic Compounds. Tübingen: Max Niemeyer Verlag. Campbell, Alistair 1959 Old English Grammar. Oxford UP. Clark-Hall, John Richard 1960 A Concise Anglo-Saxon Dictionary. Cambridge University Press. Denison, David 1993 English Historical Syntax: Verbal Constructions. London & New York: Longman. Galinsky, Hans 1952 Die Sprache des Amerikaners II. Heidelberg.

The emergence of the verb-verb compound

445

Jespersen, Otto 1965 A Modern English Grammar on Historical Principles. Part 5. London: George Allen & Unwin Co. 1965a A Modern English Grammar on Historical Principles. Part 6. London: George Allen & Unwin Co. Kastovsky, Dieter 1986 Diachronic word-formation in a functional perspective. In: Dieter Kastovsky & Aleksander Szwedek (eds.), Linguistics across Historical and Geographical Boundaries: In Honour of Jacek Fisiak on the Occasion of his fiftieth Birthday, 409–21. Berlin: Mouton de Gruyter. 1992 Semantics and vocabulary. In: Richard M. Hogg (ed.), The Cambridge History of the English Language, Vol. I: The Beginnings to 1066, 209–408. Cambridge UP. Kennedy, Arthur G. 1920 The Modern English Verb-Adverb Combination. Stanford University Press. Koziol, Herbert 1972 Handbuch der englischen Wortbildungslehre. Heidelberg: Carl Winter. Lass, Roger 1994 Old English: A historical linguistic companion. Cambridge University Press. Lieber, Rochelle 1983 Argument linking and compounds in English. Linguistic Inquiry 14.2: 251–286. Lindelöf, U. 1937 English Verb-Adverb Groups converted into Nouns. Societas Scientiarum Fennica: Commentiones Humanum Litterarum IX.5 Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation. München: Beck. [first edition 1960]. Mustanoja, Tauno F. 1960 A Middle English Syntax. Helsinki: Mémoires de la Société Néophilologique de Helsinki 23. Oxford English Dictionary. Second Edition. [OED2] 1989 Prepared by J.A. Simpson & E.S.C. Weiner. Oxford: Clarendon Press.

446

Benji Wald and Lawrence Besserman

Oxford English Dictionary Additional Series. Vol.1–2. [OEDAS1–2] 1993 Edited by J.A. Simpson & E.S.C. Weiner. Oxford: Clarendon Press. Oxford English Dictionary Additional Series. Vol.3. [OEDAS3] 1997 Edited by Michael Proffitt. Oxford: Clarendon Press. Partridge, Eric (ed.) 1984 A Dictionary of Slang and Unconventional English. (8th edition.) New York: Macmillan Publishing Co. Pennanen, Esko V. 1966 Contributions to the Study of Back-Formation in English. Tampere: Acta Academiae Socialis. A.4. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik 1972 A Grammar of Contemporary English. London: Longman. 1985 A Comprehensive Grammar of the English Language. London: Longman. Raab, Eduard 1936 Mittelenglische Nominalbildung. Inaugural-Dissertation. A. Rossteutscher: Coburg. Sauer, Hans 1992 Nominalkomposita im Frühmittelenglischen: Mit Ausblicken auf die Geschichte der englischen Nominalkomposition. Max Niemeyer Verlag: Tübingen. Schütze, Carson T. 1996 The Empirical Base of Linguistics: Grammaticality Judgment and Linguistic Methodology. Chicago/London: University of Chicago Press. Selkirk, Elisabeth O. 1982 The Syntax of Words. Cambridge & London: MIT Press. Sˇtekauer, Pavol 1998 An Onomasiological Theory of English Word Formation. Amsterdam & New York: Benjamins. Skeat, Walter W. 1868 A Moeso-Gothic Glossary. London: Asher. Sweet, Henry 1905 A New English Grammar. Oxford: Clarendon Press. [first edition 1881].

The emergence of the verb-verb compound

447

Voyles, Joseph B. 1992 Early Germanic Grammar: pre-, proto-, and post-Germanic languages. San Diego & New York: Academic Press. Visser, F. Th. 1967 A Historical Syntax of the English Language. Part 2. Leiden: EJ Brill. Webster’s Ninth New Collegiate Dictionary. 1988 Frederick C. Mish, chief editor. Springfield, Mass.: Merriam-Webster Inc. Willinsky, John 1994 Empire of Words: The Reign of the OED. Princeton University Press.

A thousand years of the history of English Richard W. Bailey

Insofar as critical theory has conferred any benefit on literary and linguistic studies, it has made practitioners self-conscious about their practices. If, as American linguists have believed at least since William Dwight Whitney, language is a socially-constructed artifact, then it is self-evident that discourse about language is similarly both arbitrary and conventional. Language itself is sidelong and evasive; commentary about language is slippery and elusive. Any text-centered study of language – any philology – is constantly being distracted by unexamined assumptions. In The Search for AngloSaxon Paganism, Eric Stanley shows just how persistent the idea was that Old English literature was either pre-Christian or had the signs of preChristian cosmology about it. Here was an idea connecting the Germanic myths of Wagnerian romanticism with Beowulf, the exuberant and natural worship of the Aryan gods in contrast to the gloomy and artificial life of the early medieval Christian church. In his conclusion to this book, Stanley wrote that “the history of scholarship is a history of error” (1975: 122). Scholars have stopped searching for paganism, or at least have ceased to see the pagan gods where they are not certainly found. Perhaps the best that can be said is that we are now making still other unexamined errors that merely await an Eric Stanley to unmask them. In the essay that follows, I argue for a renewed philology, one far more comprehensive than any definition of this inquiry usually offered in the twentieth century. For Whitney and his contemporaries, philology was a comprehensive and historical discipline. For his successors, it became narrowly antiquarian and mired in curricular arguments about whether early Germanic languages should be required of all who teach English in colleges and universities. At the beginning of this new century, scholars are liberated from those burdens. Philologists no longer limit themselves to dead languages; doctoral requirements no longer force all students to this kind of arid knowledge that was so productive of bitterness and resentment.

450

Richard W. Bailey

We can now do as we choose and, I believe, we should choose to renew the philological enterprise.

1.

The idea of a standard English

Searching for a standard in English bears an uncomfortable resemblance to that quest for paganism unmasked by Eric Stanley. While the idea of “common” and “courtly” dialects within English is old, the notion that everyone should speak one kind of English is not. It arises in the middle of the eighteenth century with the explosive growth of schools and schoolbooks – highly profitable to authors and publishers – produced for use in them. And as variety in English increased (as it did through the geographical dispersal of its speakers in the nineteenth century), the notion that there was (or should be) a single standard flourished. Like the grail quest, the search for standard accumulated various mythologies. Daniel Jones thought it could be found in the ancient fee-paying schools of southern England; H.C. Wyld regarded the officer’s mess of a regular army regiment as the place where it might be found; Americans in the west, a century ago, imagined it to be spoken in Boston. More recently, linguists have discovered standard English in their own mouths. In a wonderful essay, my colleague A.L. Becker gets right to the heart of my discomfort: “Notice how ethnocentric that is: are there stages of literacy? I recall Jean-Paul Sartre: ‘progress, that long steep path which leads to me’” (1983: 45). This sort of ethnocentrism is endemic to the study of language change. We hold the mirror of linguistic history in front of our own faces. In talking to my classes, I define myself as a speaker of “standard English” and explain that whatever I say goes. This is increasingly difficult as the standard erodes before my very eyes – as students say calm, palm, psalm, and almond (for instance) with a consonant long silent in standard English and a back-shifted vowel. Mostly they get the idea that I’m kidding. (Or so I hope.) And they are supposed to get the idea that their own language bigotries are no more founded in the immutable standard than mine are. This pedagogy is supposed to make them tolerant both of their parents (and other old people) who can’t help being severe about innovation and of people younger than

A thousand years of the history of English

451

they are. And nicer to their younger siblings – to my wonderment they are willing to criticize, at age eighteen or nineteen, the illiterate babblings of brothers and sisters still in middle school. These pitiful siblings are far less standard than they themselves were five or six years earlier – not to mention much less clever. My difficulty with this fascinating ethnology is that we have very inadequate ways of representing it. We are still locked in Newtonian thinking in an Einsteinian world. I recently came across a clear acknowledgement of this situation in a fine essay by Roy Porter about medical quackery in eighteenth-century England. Porter has wonderful examples that sound a great deal like the quacks in Ben Jonson’s Alchemist. Ned Ward (of facetious memory) concocted the following preposterous example. Gentlemen, you that have a Mind to be Mindful of preserving a Sound Mind in a Sound Body, that is, as the Learned Physician Doctor Honorificabinitudinatibusque has it, Manus Sanague in Cobile Sanaquorum, may here at the expense of twopence, furnish himself with a parcel, which tho’ it is but small containeth mighty things of great Use and Wonderful Operation in the Bodies of Mankind, against all Distempers, whether Homogeneral or Complicated; whether deriv’d from your Parents, got by Infection, or proceeding from an ill habit of your own Body. (Quoted by Porter, 1987: 81–82)

It would be natural enough to contrast the quacks with the credentialed practitioners – the “faculty physicians” – in their citation of ancient authority, their use of specialized vocabulary, and their syntax of efficacy and promise. Porter, however, does not follow this procedure of setting the standard against the dialect, the unmarked vs. the jargon. Instead, he points to the mistake of seeing evolutionary progress as ascending up to me. Here lies an important lesson. Historians of medicine have traditionally sided with regular medicine, deriding quacks as evil or absurd. In linguistic terms, this would be the equivalent of implying that faculty physic spoke ‘standard English’ – that normal, perfect, translucent, neutral, objective medium of communication – whereas quacks used some degenerate argot or jargon (one hesitates to say quacks had their own pidgin). But this begs many questions. Faculty-talk and quackspeak need to be treated symmetrically. Both orthodox and irregular medicine generated their own linguistic sub-cultures, whose vocabulary, tones and speech mannerisms, though somewhat distinct, played similar socially and even therapeutically active roles. (1987: 76)

452

Richard W. Bailey

In other words, the placebo may effect the cure if the accompanying language is powerful enough. In the history of English, we have sided with “faculty linguistics” and treated the “other” with condescending sympathy. Sometimes this strategy involves adroitness in drawing boundaries. “International English,” I was once told, includes me, Lord Bloomsbury (i.e., Randolph Quirk), Braj Kachru, and quite a number of others, no doubt. I won’t mention the names of people who are left out. They are just as various in their English as we are. The difference is that we are the namers and they are the named. How can we have adopted a political strategy of circling the linguistic wagon around ourselves and defending out position with repeating rifles while the vast majority of speakers of English whoop and holler around us in a combination of derision and dread? While the metaphor may seem a little forced, it strikes me that there is a discomforting truth captured by it. After all, we let people’s lives depend upon linguistic nuance. Or perhaps I exaggerate. As lexicographers and grammarians we are the underlings of the test-makers who provide the “objective measurement” upon which those with the real power will decide who will be selectively admitted (and who not), lucratively employed (and who not), decisively powerful (and who not).

2.

Philologists and sempstresses

Most of the testing and sorting of people is a twentieth-century development, though there were certainly usages criticized before that and there are shibboleths even more ancient. Reading Lynda Mugglestone’s collection, Lexicography and the OED, brought again to my attention the presidential address to the Philological Society presented by J.A.H. Murray on May 16, 1884. Murray was basking in the celebrity of his dictionary, the first part of which had appeared at the end of February. Quick to resent criticisms, he surveyed the reviews that had thus far appeared – most of them thoroughly gratifying to the lexicographical ego – and was happy to tell his audience about the risible and incompetent responses.

A thousand years of the history of English

453

The method employed to indicate pronunciation has received the general approval of foreign reviewers as simple, intelligible, and practical... It is only in the United States therefore that any attempt has been made to criticize the actual pronunciation exhibited; always excepting the Church Times, which in a notice generally appreciative, says: The system adopted for fixing the pronunciation ... is extremely complicated, and involves the employment of the eccentric alphabet invented by the fanatics of the Fonetic Nuz. ... But we object to Dr. Murray’s pronunciation on other ground; namely, that, in certain words the sound he assigns is not the true one, but that in use amongst lower middle-class Londoners. For example, alone, as he marks it, is given the drawling “ow” sound of the sempstress voice, as heard in Kentish Town or Peckham. This whole passage is so delicious, substituting, as it does, for my pronunciation the writer’s perverse interpretation of the symbols (there is a ‘Key’ to them, too), that I think it deserves preservation, and I venture to enshrine it in the amber of this Address for the admiration of distant readers and future philologists. The joke will be appreciated most of all by those who know the discussions we have had here upon the method employed in the Dictionary to indicate the Pronunciation, in which the main charge has been that we have not adequately recognized current colloquial tendencies. (Mugglestone 2000: 527–528)

We are, of course, the future philologists Murray addressed, and so I hold up this globule of amber for your inspection. Many issues spring to our attention. Let us begin with Murray’s remark that: “It is only in the United States therefore that any attempt has been made to criticize the actual pronunciation exhibited.” Here he apparently alludes to a review, in the American Journal of Philology, by James M. Garnett. Garnett would become a regular reviewer of the parts of the dictionary as they appeared – eventually contributing sixteen thoughtful, judicious, and congratulatory notices – and in this, his very first, review he ranked Murray’s dictionary with those of Grimm and Littré and urged the backers to supply “all the editorial and clerical assistance necessary” to hasten completion of the work (1884: 366). Garnett described the pronunciation scheme enthusiastically, and he praised Murray for making “a more exact analysis of pronunciation, particularly of the vowel sounds, than we are accustomed to in the ordinary dictionaries” (1884: 365). So far, so good. But then Garnett makes the following criticism: “Some might take exception to certain statements, but with respect to pronunciation, more than anything else, it is true that ‘no one man’s English is all English’” (1884: 365). Only the most up-to-date of Murray’s hearers would

454

Richard W. Bailey

have seen the issue of the American Journal of Philology in which Garnett’s review appeared. 1 It would not have seemed implausible to them that some American might take a perverse view of English pronunciations. Only very recently, in 1880, had the American journalist Franc B. Wilkie declared that “the original and pure English of America” was the true descendent of the Saxons and Normans and what he found spoken in Britain was a debased jargon in which peers in the House of Lords, for instance, inserted or deleted h’s in truly disgusting ways (see Bailey 1991: 154–156). Americans might “criticize the actual pronunciation exhibited” in Murray’s dictionary from some such wild and perverse failure to grasp the source of “true English.” (This is hardly to exaggerate the degree of anti-American feeling among British philologists of Murray’s day. In 1890, Murray himself would declare that Americans recognized dictionaries as authoritative since they had no other access to right speaking: “the absence of living English usage” in America Murray called it [see Bailey 2000: 219].) 2 Having swept aside, without any detailed indictment, American objections, Murray turned his attention to the well-intentioned though wrongheaded reviewer in the Church Times. The first laugh at this writer came at the mention of “the eccentric alphabet invented by the fanatics of the Fonetic Nuz,” particularly amusing because two of these fanatics were members of the Philological Society – A.J. Ellis, who had been writing on spelling reform for nearly forty years, and Henry Sweet, celebrated by Murray in his address as “the ringleader of these ‘fanatics’ is the learned Father of English Phonology, who has five times addressed us from this chair, as our honoured President” (Murray 1882–84: 528). By criticizing these eminences, the reviewer had virtually disqualified himself from having any valuable opinion whatsoever about the pronunciations in the dictionary. Nowadays, it is difficult to recollect the mania for spelling reform that animated the great personalities of English philology in the nineteenth century. Noah Webster was, of course, such a reformer, though we now mainly recollect him for his successes in promoting as “American” variant spellings that had already existed in British English – the -or in color, honor, and so forth, and the -er in center, meter, and the like. (Webster’s advocacy of iland ‘island’, ake ‘ache’, and crum ‘crumb’, are mostly forgotten.) Ellis and Sweet were far more ardent and ambitious than Webster had been, and Ellis published his reports of meetings of the Philological Society in a thor-

A thousand years of the history of English

455

oughgoing “rational” spelling. While Murray embraced the cause, he was less exuberant in advancing it, though he was not reluctant to distinguish variant spellings – like sempstress for seamstress – from “bad” or “erroneous” ones in his dictionary. For our purposes, it is important to understand that the motive for spelling reform went far beyond the benefits eternally adduced for children learning to read and for foreigners acquiring the language. A reformed spelling, its proponents argued, would create greater uniformity in spoken English because the spelling would be an unambiguous cue to pronunciation. Webster hoped that a reformed spelling would maintain the “purity” and “uniformity” of English in America, traits he and visitors from England found to be remarkable in the vast expanse of the United States. Sweet had a far more ambitious version of this vision: to make English uniform worldwide. His famous Handbook of Phonetics, published in 1877, had a sub-title often omitted in citations of it: “Including a Popular Exposition of the Principles of Spelling Reform.” When I first came upon the appendix – through a reference in an essay by Randolph Quirk – I was puzzled by it. (Quirk was attentive to the words I am about to quote because, in the 1980s, R.W. Burchfield had predicted a separation of the languages of Britain and America into descendants as different as French and Italian within a century [Quirk 1985: 3].) Sweet’s prediction was this: “It must also be remembered that by that time” – a century from his day, or, in other words, the 1970s – “England, America, and Australia will be speaking mutually unintelligible languages, owing to their independent changes of pronunciation” (1877: 196). Quirk, of course, pounced on this quotation: “Sweet’s forecast (which, given the circumstances and received knowledge of his time, had a greater plausibility than Burchfield’s) proved dramatically wrong because he overestimated the rate of sound change” (1985: 3). Eventually it struck me that Sweet – who was no fool about the rate of sound change – was not merely responding to the “circumstances and received knowledge of his time.” In fact, most people of that day were, if anything, overly enthusiastic about the rate of increase in English as a world language. Prime Minister Gladstone, in 1884, would quote an authority who thought that there would be a billion English-speakers in the 1980s, and a Swiss savant, writing at the same time as Sweet, estimated the 1970 population of England as 124 million, and the United States, Canada, and Australia as having another 736 million (see Bailey 1991: 111–113).

456

Richard W. Bailey

English, in these estimators’ eyes, would sweep up the Amazon basin and pour over the Andes to occupy all of South America, eventually reaching across the Pacific to connect with English speakers who had migrated eastward to New Zealand and Australia. But how could this global sweep result in mutual unintelligibility? Even Abel Magwitch (in Great Expectations) found his way from Australia back to England, and Dickens had posited no issues of intelligibility in representing his conversations with Pip. So how could Sweet be so wrong in predicting mutual intelligibility in the future varieties of (former) English? In fact, Sweet was trying to throw a scare into the opponents of spelling reform. “The only way to meet these evils” – that is, the fissiparous tendencies in English both locally and globally – “is strictly to subordinate spelling to pronunciation” (1877: 196). Putting it in another way, he connected his argument to the long-tradition of hand-wringing about the very fact of language change itself: “In fact,” he wrote, “one of the worst features of a fixed orthography is that it loses all control of pronunciation, and thus indirectly proves the cause of such changes as have completely changed the character of English in the last few centuries” (1877: 194). Close inspection of the grammar of this sentence exposes Sweet’s argument. (In memory of George Orwell, we might name this technique the outing of undercover agents.) What exactly is it that “loses all control of pronunciation”? The grammar says it’s the “worst features of a fixed orthography.” No more definite culprit can be named. Who should have been in charge? Perhaps if we were to assign blame we might look at the dossiers of the Wycliffites, or maybe the scribes who used to dwell in Chancery Lane, or maybe William Caxton. Sweet argues that someone should be put back in the control room. And guess who’s ready to take on the job? Well – to return to Murray’s presidential address – how about the ringleader of these “fanatics of the Fonetic Nuz”? By adopting an orthography founded on their own speech, Sweet and his fellow fanatics could wrest control of English and direct it in the “right way.” 3 Now we return to that unfortunate clergyman reviewing the first part of the OED in the Church Times. Aside from his objection to the phonetic key employed in the dictionary, he complained of the model Murray had chosen: “in certain words the sound he assigns is not the true one, but that in use amongst lower middle-class Londoners.” To illustrate this claim, he cites the representation of alone which, he says, has the “drawl-

A thousand years of the history of English

457

ing ‘ow’ sound of the sempstress voice, as heard in Kentish Town or Peckham.” Murray sweeps this complaint aside without notice, alleging that the reviewer cannot interpret properly the key to the pronunciation. Yet it seems to me that the reviewer knew exactly what he was about (and Murray knew it too). And, should we uncover the identity of the reviewer, we should not be surprised to learn that he, like Murray, was a Northerner. The pronunciation of alone found in the dictionary is the diphthongal one with a back glide, though with the [o] solidly located as a back vowel rather than the more fronted versions found in modern RP. In 1869, A.J. Ellis had observed speakers who distinguished between the simple vowel in no and the diphthong in know (see MacMahon 1998: 411–412, and Wells 1982: 312–323). 4 Certainly the subsequent history of this vowel sound invites us to think that our clerical reviewer was onto something when he assigned the diphthongal vowel in alone to the “sempstress voice” of Kentish Town or Peckham. According to J.C. Wells, the monophthongal realization still occurs (though with a central vowel), but there is a richness in London unknown in America including a rounded front terminus for the diphthong said by Wells (following Beaken 1971) to be used only by Cockney girls (1982: 309). Things were not much simpler in the nineteenth century. In Murray’s own meticulous account of the dialect of the southern counties of Scotland, published in 1872 under the sponsorship of the Philological Society, he wrote of the vowel in alone: “It is also a uniform simple sound, and not a diphthong or quasi-diphthong, like the o of the South of England, which begins with o, but tapers off into oo, thus no¯ > oo, ro¯ > ood (Pal[eotype] noou, rooud), while the Scotch sound is no¯ - o, ro¯ - od (noo, rood)” (1872: 120). From this account, we may presume that the reviewer for the Church Times was a northerner, who, unlike Murray, had made little attempt to adapt his speech to the corrupt speechways of the metropolis. In Joseph Wright’s complicated account of the phonology of English dialects – published in 1905 but based on earlier inquiry – the picture is indistinct but, in general, the diphthongs are southern and the monophthongs northern (1905: 88–92). But more than merely a regional distinction was at work here. Writing in 1890, Sweet would note that the diphthongal pronunciation tended to vanish in unstressed syllables.

458

Richard W. Bailey

This change often takes place in stressed syllables as well, where it is often used in polite or conciliatory address; thus [oU] is often used in polite or conciliatory address; thus [oU noU] and [ˆU nˆU]= oh no may be heard from the same speaker, the former in more decided and dogmatic statements. The constant use of [ˆ] gives a character of effeminacy or affectation to the pronunciation. In Cockney English the first element is unrounded and the second lowered and partially unrounded – [ˆo:], [sˆo:]. Scotch keeps the old monophthong [o:], [so:] (1890: 76)5

Of course in “the original and pure English of America” we have had no truck with this stuff but are merely amused bystanders – just as were the northerners and Scots who were wise enough to avoid this peculiarly southeastern imbroglio. Thus an American observer, writing for publication in 1919, noted: “This sound is much less diphthongal in American than in British speech. In the latter a great variety of diphthongal shadings occur, some of them familiar in the exaggerated representations of Englishmen and their speech on the American stage. In the speech of many, perhaps of most, Americans there is scarcely any trace of diphthongal quality in the sound ...”(Krapp 1919: 82). Remember that we began with this example by contrasting the “true sound” of the word alone with the “sempstress voice of Kentish Town or Peckham.” Now that we have come to understand better what the issues were that animated the reviewer for the Church Times, we see that, for him, the “true sound” was the monophthong used by Americans, northerners, and those Scots who had not abandoned their language for the refined accents of the south. The “sempstress voice” was the one coming to the awareness of purists, and effeminate affectation, exaggerated representations of Englishmen, and RP were all about to merge in fastening on the diphthong as the “true sound.” Murray was also right in preserving this issue in amber for our inspection. Discussions in the Philological Society had resulted in selecting for the dictionary the most advanced specimen of a recent sound change. He could allay fears that “we have not adequately recognized current colloquial tendencies” in the pronunciation scheme. The larger issue for us, however, is that our historical accounts of English preserve the notion of upward progress leading to one destination – truth versus the sempstress of Kentish Town, the Scot, the American, the northerner, and the other unfortunates. So we teach about the Great English Vowel Shift even in classrooms where the Vowel Shift has not taken place.

A thousand years of the history of English

459

Canadians deliberate “raising” in the onsets of writer and bout but “lowering” in rider and bowed (see Chambers 1979). Historically speaking, we ought to be intrigued by “non-lowering” in the former pair, but so strong is the notion that the Great Vowel Shift is normative that we talk about Canadian exceptionalism as “raising” even though there is no historical evidence for such an event. Sweet’s argument for spelling reform returns here in an interesting way. Our conventional spelling – and he uses italics to emphasize his point – is far from “preserving etymology” as the anti-reformers claim. “In short, historical spelling destroys the materials on which alone history itself can be based” (1877: 200). If his reforms had been in place earlier, we would know when seventeenth-century speakers ceased to pronounce the w in wretch or the k in knight. What a democratic vista for spelling! And what a boon for the historian of English! His proposals, had they been adopted, would at once proclaim to the world that I have different vowels in Don and dawn and most Californians do not. Regional and social distinctions would be immediately apparent as soon as someone put pen to paper. Of course Sweet had no such prospect in view, despite this natural extension of his argument that I have presented. Here is what he says: The history of h and r in modern times is an instructive instance of how pronunciation may be controlled by a changing spelling. It is certain that if English had been left to itself the sound h would have been as completely lost in the standard language as it has been in most of the dialects. But the distinction between house and ’ouse, although in itself a comparatively slight one, being easily marked in writing, such spellings as ’ouse came to be used in novels, & c. as an easy way of suggesting a vulgar speaker. The result was to produce a purely artificial reaction against the natural tendency to drop the h, its retention test of education and refinement. (1877: 194–195)

At the time Sweet wrote, my paternal grandparents were acquiring English on subsistence farms in a rural backwater of southern Michigan, and I can assure you that nobody in their community had the slightest “natural” tendency to misplace their aitches. Joseph Wright’s later investigation of dialects in Britain revealed that aitch was not lost in Ireland, insular or mainland Scotland, Northumberland, “and perhaps also in portions of n[orthern] Dur[ham] and n[orthern] Cum[berland]” either (1905: 254). In other parts of England, however, Wright affirmed Sweet’s claim that initial h had “disappeared.”

460

Richard W. Bailey

It can now be revealed that Sweet had no interest whatsoever in letting Scots, Northerners, inhabitants of Ireland, or rural Michiganders spell as they chose. Instead, they all should follow influential tastemakers meeting in London and calling themselves, no doubt, members of the Philological Society. Thus it is that the word control figures so largely in Sweet’s argument, for his proposals give control to a centralized authority. Perhaps it will be apparent how close to the mark Shaw was in transforming, in Pygmalion, the real Henry Sweet into the fictional Henry Higgins – rather cruelly giving to the latter the title “professor” that had always eluded the former. Sweet’s vision was “the restoration of the original harmony of the English with the Continental values of the letters” (1870: 197), which would create a kind of Euro-spelling in which English, French, and German would become far more transparent in classrooms (and elsewhere) than was possible with their non-phonetic spelling systems. Notice, however, his appeal to a golden age in which harmony prevailed over discord. Like the reviewer in the Church Times, he divided the linguistic world into “true” pronunciations and the pronunciation of the sempstresses, the evolution of a standard and the deviance of the dialects, the descent of a once unified English into divergent descendents that needed to be reunited.

3.

Radically studying the Saxon and English languages

Going back one more century puts us among a distinctly different set of unexamined assumptions. In 1798, Samuel Henshall – Fellow of BrazenNose College, Oxford – published an anthology of Old English with an impressive title: The Saxon and English Languages Reciprocally Illustrative of Each other; The Impracticability of Acquiring an Accurate Knowledge of Saxon Literature, Through the Medium of Latin Phraseology, Exemplified in the Errors of Hickes, Wilkins, Gibson, and other Scholars, and a New Mode Suggested of Radically Studying the Saxon and English Languages. Henshall asserted the virtues of the method we would use today – the use of interlinear glosses to tease out the meaning of the Old English texts. Taking Caedmon’s hymn, for instance, Barbara Strang provides the Old English: Nu scylon hergan/hefaen-rices uard. She then gives us the word-by-word version: ‘Now (we) must praise/(of) heaven-kingdom

A thousand years of the history of English

461

(the) guardian’. She then gives a modern English translation: “It is now our duty to praise the Guardian of the Heavenly Kingdom” (1970 [1994]: 369). Henshall follows the same procedure, giving first one of the many source texts of the hymn: Nu we sceolan herigean heofon Rices Weard. And then his interlinear gloss: “Now we shall hearen heaven’s Reach-word” (1798: 47). And so he proceeds by what we might call, the false-friend method of extracting the ancient meaning through like-sounding modern words. As Hans Aarsleff explains in his account of Henshall’s ignorance, uard ‘guardian’ is rendered by him as “world” with the explanatory footnote: “The omission of a letter here, l, is sometimes not to be much regarded” (1798: 47n). In 1807, as Aarsleff also points out, Henshall published his Etymological Organic Reasoner to illustrate the idea that “throughout all languages there is a resemblance in the sound, and an affinity of ideas, attached to the tones produced by the exertions of the same organic powers of human speech” (quoted by Aarsleff, 1967: 77). John Mitchell Kemble regarded Henshall as an “irrecoverable madman,” but his method of approaching the history of English was well within the mainstream of scholarship in his day. Tales almost as fantastic as Henshall’s are found in abundance in David Matthews’ recent book, The Making of Middle English, 1765–1910. Of the characters in these stories, the best remembered is Thomas Piercy who, in 1765, published The Reliques of Ancient English Poetry. Piercy was an ambitious scholar who parlayed his Oxford education into modest living in a country parish. Discovering a manuscript of ballads and romances – allegedly being used by a neighbor’s parlor-maid to light fires – he changed his name to Percy and enhanced the manuscript to emphasize “several ballads about the medieval Perceys” (1999: 6). This much improved version of Middle English was the foundation for the publication of The Reliques and the means to Thomas Percy’s eventual preferment as a Bishop, since the powerful and wealthy family of Percy was pleased to have its ancient lineage, already celebrated by Shakespeare, put on public display. It involved the eighteenth-century clergyman becoming a medieval minstrel. Percy was by no means the only person to seek the patronage or good opinion of the Northumberlands through poetry, but he alone was able to remake himself as a minstrel so successfully as to reinvent, through his ‘barbarous productions,’ that ‘spirit of chivalry’ that would ensure that he was ‘protected and caressed’ (Matthews, 1999: 22). 6

462

Richard W. Bailey

Matthews explains that Thomas Percy’s interest in the history of English and his publication of The Reliques “was a brilliant career maneuver, one bringing rapid material and social improvement” (1999: 12). Samuel Johnson likewise made good through the English language, and, though the preface to his Dictionary of 1755 is well known, the history of the English Language that follows it is far less frequently read. It consists mainly of specimens, but Johnson’s introductory remarks illuminate his view of the subject. What was the form of the Saxon language, when, about the year 450, they first entred [sic] Britain, cannot now be known. They seem to have been a people without learning, and very probably without an alphabet; their speech therefore, having been always cursory and extemporaneous, must have been artless and unconnected, without any modes of transition or involution of clauses; which abruptness and inconnection may be observed even in their later writings. This barbarity may be supposed to have continued during their wars with the Britains, which for a time left them no leisure for softer studies; nor is there any reason for supposing it abated, till the year 570, when Augustine came from Rome to convert them to Christianity. The Christian religion always implies or produces a certain degree of civility and learning; they then became by degrees acquainted with the Roman language, and so gained, from time to time, some knowledge and elegance, till in three centuries they had formed a language capable of expressing all the sentiments of a civilized people, as appears by King Alfred’s paraphrase or imitation of Boethius, and his short preface, which I have selected as the first specimen of ancient English. [D1, recto]

If “involution of clauses” be a true measure of civility and learning, then Johnson is nearly at the summit of that “long steep path” of linguistic evolution. Of course Johnson’s view was widely shared in his day: that the early stages of English reflected a culture artless, unconnected, and barbarous.

4.

The politics of linguistic history

In 1749, John Free published the first book-length history of English, a work remarkably lacking in linguistic detail. 7 Free, however, met head-on the usual view that the Latin-Romance part of English was reflective of “knowledge and elegance” and the Germanic part of tumultuous barbarity. In so doing, he claimed the authority of Edmund Gibson who had welcomed the Hanoverian succession by writing, in 1722, that English people were

A thousand years of the history of English

463

all Teutonic and had suffered the misfortune of being ruled by unworthy French people since 1066. 8 Free warmed to this theme in considering the history of English, drawing on patriotic zeal lately exercised in the thwarting of the Jacobite uprising of 1745 that nearly toppled the Hanoverian monarchy. I know that this is a Truth disagreeable enough to some People, who are sorry for the strict Relation that England bears to Hanover; and therefore, tho’ it be published aloud by all our antient Histories, we find nothing more industriously concealed and evaded by the faction of Renegado English, whom we call the Jacobites. Creatures so unnatural! as to disavow their own Original, and to brand with the Name of Foreigners those very People, who not only gave them Being, but procured them also, at the Expence of their Blood, all the Advantages, which from their Conquest we now enjoy, English Ground and English Liberty; the Wealth and Commerce, in which we glory, the busy Thames and Severn, the English Rivers and the English Seas, by which our Name and Power have been extended to both the Indies, and dreaded even in the remotest and most powerful Empires of the Globe. And yet our English-Saxon Fathers, the authors of these inestimable Blessings, by an Ingratitude not to be paralleled in any nation under Heaven, are considered as Foreigners, and the country which sent us hither in their Loins, hardly ever mentioned but with Contempt! (1788: v–vi)

Once we have stopped laughing at Free, we need to remind ourselves of the deadly proto-Aryanism that suffuses his bluster. What his view has in common with the opinions of the Philological Society, of course, is a sort of identity politics in which an argument for cultural superiority masquerades as linguistic history. Free was welcoming the prospect of George III coming to the throne – this dedication was composed to him as Prince of Wales. Like Thomas Piercy, he perhaps anticipated “rapid material and social improvement.” 9 Pressing our search ever backward, we find that the linguistic historians of the mid-sixteenth century had yet another variant of identity politics. Protestants used Old English texts to justify the use of the vernacular in translating the Bible. In 1571, the scholars employed by Matthew Parker, Archbishop of Canterbury, made this point vividly clear. Some againe haue iudged our natiue tounge vnmeete to expresse Gods high secret mysteries, being so barbarous & imprefecte a language as they say it is. As though the holy spirite of truth mente by his appearing in clouen tounges, to debarre any nation,

464

Richard W. Bailey

or any tounge, from vttering forth the magnificent maiestie of Gods miraculous workes. (Gospels A.ii verso–A.iii recto)

Even though few could read these “Saxon” gospels, their publication made impressive testimony for the Protestant cause. It was the English church, through the continuity of vernacular tradition, that could claim direct descent from the primitive rites of early Christians – not, to use the words of an English divine of the 1560s, the “continuall chaine” of textual descent claimed by the “Romanistes” (Testimonie: 18).

5.

Etymology and truth

As we inquire about the work of even earlier linguistic historians, the idea of “high secret mysteries” becomes ever more important while class-superiority, national pride, and religious legitimacy diminish as metaphors for explanation. For medieval intellectuals, etymology was a technique for revealing truth and part of the armamentarium of every textual scholar. 10 In these deeds of interpretation, the Bible provided inviting examples of the “high secret mysteries” hidden within words. Before linguistic science drove out the bad old ways, all that was needed to advance an etymology was phonetic resemblance, semantic approximation, and a memorable anecdote. So Samuel Henshaw was firmly in the center of respectable etymologizing when he rendered Old English “heofon Rices Weard” as “Heaven’s Reach-Word” – a proto-deconstructionist interpretative act if ever there was one. My example of this line of linguistic speculation comes from the important survey of English Language Scholarship published in English by Helmut Gneuss in 1996. Gneuss draws attention to Byrhtferth’s Handboc, an English work based largely on the Latin writings of Bede on astronomy and rhetoric. Writing in 1011, Byrhtferth explained the origin of the Latin word autumn: “The third season is called autumnus in Latin, and in English autumn. Scholars explain the name as derived from ‘ripening’ or ‘gathering.’ They say: ‘Harvest (derives its name) from harvesting or ripening.[’]” (Crawford: 1929: 93). 11 There is some complexity to this example, particularly because the modern editor of the Handboc translates “in English autumn” even though the

A thousand years of the history of English

465

English word autumn would not exist in English until several centuries after Byrhtferth. For Byrhtferth, the third season was harvest. But in explaining Latin autumnus Byrhtferth explicates a hidden set of associations. James Murray, writing in the OED, described autumn as “of doubtful etymology.” A far less conservative etymologist, Eric Partridge, used the abbreviation o.o.o. for ‘of obscure origin’, and in writing about autumn he declared: “... L[atin] autumnus is o.o.o., perh[aps] Etruscan, although the old L[atin] explanation by auge¯re, to increase, may, after all, be correct, autumn being the season when the crops’ increase is yielded at the harvest.” Here is linguistic history founded on a deep and widespread human conviction that words contain mysteries. In 1851, Richard Chenevix Trench (who would influence the shape of the OED) declared in his lectures On the Study of Words: “... words often contain a witness for great moral truths – God having impressed such a seal of truth upon language, that men are continually uttering deeper things than they know” (1851 [1927]: 13–14). While we may not nowadays so readily embrace this claim, it is certainly true that psychoanalysis (and other versions of the “talking cure”) depends upon precisely the assumption that there is more to words than meets the ear. By close inspection of language, we can reveal truths. When I wrote that last sentence, I thought immediately that I should revise it to: “By close inspection of language communities, we can reveal truths.” It seems to me of paramount importance that we see people in the past as like ourselves and yet different. James Murray, Henry Sweet, Samuel Henshall, Byrhtferth the Monk – all are characters in the history of English.

6.

All the lonely people: Where do they all belong?

Leo Spitzer, one of those great émigré scholars who were ejected from Europe to enrich America, wrote notes toward a biography of his linguistic training, and I have often thought of them during my own career. [W]hen I attended the classes of French linguistics of my great teacher Meyer-Lübke no picture was offered us of the French people, or of the Frenchness of their language; in these classes we saw Latin a moving, according to relentless phonetic laws, toward French e (pater > pére); there we saw a new system of declension spring up from nothingness, a system in which the six Latin cases came to be reduced to two, and later

466

Richard W. Bailey

to one – while we learned that similar violence had been done to other Romance languages and, in fact, to many modern languages ... In reference to a given French form Meyer-Lübke would quote Old Portuguese, Modern Bergamesque and Macedo-rumanian, German, Celtic, and paleo-Latin forms; but where was reflected in this teaching my sensuous, witty, disciplined Frenchman, in his presumably 1000 years of existence? He was left out in the cold while we discussed his language; indeed, French was not the language of the Frenchman, but an agglomeration of unconnected, separate, anecdotic, senseless evolutions ... (1962: 2–3)

In our histories of English, where are those people who spoke the language? As indistinct, I fear, as the communities of French speakers in the profoundly learned classes of Meyer-Lübke. The great comparative-historical enterprise of the nineteenth century was sidetracked into arid pedantry and minute particularizing. Though Saussure never intended his view of diachrony to lead scholars away from historical study, that was very much the effect of structuralism. His idea of comparing two synchronic, coherent, and variable communities was never realized, but it remains worthy of our praise and dedication. By looking at trajectories of change, we miss the sense that speakers have of the stability of their own language. (Such stability may be illusory, but it is still compelling – just as the most ardent Einsteinian relativists live in an apparently stable Newtonian world.) Our discontent with the unexamined assumptions of our own inquiry can lead us to new (or renewed) theoretical and practical directions. Perhaps it is best to end with the words of that most intellectual and most American of Presidents, Thomas Jefferson. What he declared two hundred years ago is still true: “We want an elaborate history of the English language” (Foley 1900: 472).

Notes Acknowledgements: I am grateful to the organizers of this conference for many things: Anne Curzan, Donka Minkova, and Robert Stockwell. Two anonymous reviewers made useful suggestions that have influenced this published version of my plenary address there. One of them recommended, to my great benefit, Allen J. Frantzen’s book treating the history of scholarship and teaching in the field of Old English.

A thousand years of the history of English 1.

2.

3.

4. 5.

6.

7. 8. 9.

467

It seems implausible that the complete part issued early in 1884 could have reached the United States in time for Garnett’s review to appear and for a copy of it arrive back in England in time for Murray’s May 16th address. Two reviews appeared in 1883, and these must have been based upon preliminary and perhaps partial copies. Perhaps Garnett had access to one of these. According to the Monthly Reports of meetings of the Philological Society, it was on January 18, 1884, that “three copies of Part I. of the Societies new English Dictionary, edited by Dr. Murray, were laid on the table” (v–vi). The brief account of Murray’s address published in the Monthly Reports just after the May 16th meeting mentions nothing about pronunciation, and he may have added the quoted passage later. It is also possible that Murray alludes to some review other than Garnett’s since, at the May meeting, he reported, without enumerating them, that “at least 12” reviews had appeared in America by that date. C.P.G. Scott, chief etymologist for the Century Dictionary, wrote a spirited response in which he derided ‘true-born’ Englishmen for loss of r and h (see Bailey 2000: 220n). In the Huntington Library, the Furnivall papers contain the final page of a letter written by Sweet to Furnivall: “The advantage of using new alphabets only in scientific works at first, is that their merits & defects get discussed, & they are perfected & fixed before the people at large begin to use them. Of course every one would be taught to write exactly as he pronounces. If the pron. were made uniform the writing would be uniform. If, however, this were not the case there would be no more difficulty in understanding a different piece of writing than a differentlypronounced sentence. If this were not not [sic] done the alphabet would not be phonetic” (FU 838). The word exactly was inserted as Sweet reviewed the letter before sending it. The date of this letter is not known. MacMahon adduces that this distinction is a relict form reflecting the differing Middle English vowels in these two words. I have changed Sweet’s Broad Romic to IPA. In 1892, Sweet asserted that the back round vowel in English is “Close (oo) in Scotch no, know, where Standard E. has the diphthong (ou)” (19–20). Mistakenly, MacMahon writes: “No other nineteenth-century writer” (other than Ellis) “draws attention to this distinction, and it is not commented on later” (1998: 412). The internally quoted words are from Percy’s dedication of the Reliques to the Countess Percy who would, in the year following, become the Duchess of Northumberland. I base my claim that Free’s is the first on A.G. Kennedy’s excellent bibliography (1961: 14). Elizabeth Elstob was similarly enthusiastic about the Hanoverian succession in her Grammar for the English-Saxon Tongue (1715). The biographical details for Free are elusive but he seems not to have advanced beyond being Vicar of East Croker in Somersetshire.

468 10.

11.

Richard W. Bailey In his excellent survey of etymological study, Frank L. Borchardt states: “The charged significance of names and their truth was so firmly established in the Western mind as not even to require an individual defense and, much less, formidable literary authority” (1995: 5). He later explains how speculation about native words connected the Germanic north of Europe with classical antiquity. In the Renaissance, this line of thinking was devoted to “the elucidation of the past as a national political expression” (1995: 15). Crawford transcribes the ms. thus: “Se pridda tima ys autumnus on Lyden gecweden, & on Englisc harfest: boceras getrahtniaD pæne naman for pære ripunge oDDe for pære gaderunge. Hig cweDaD: Autumnus propter autumationem uel propter maturitatem” (Crawford: 1929: 92). The authority here cited is Bede who himself credits Isodore of Seville.

References Aarsleff, Hans 1967 The Study of Language in England, 1780–1860. Princeton: Princeton University Press. Bailey, Richard W. 1991 Images of English: A Cultural History of the Language. Ann Arbor: The University of Michigan Press. 2000 ‘This unique and peerless specimen’: The reputation of the OED. In Mugglestone, Linda (ed.), 206–227. Beaken, M.A. 1971 A study of phonological development in a primary school population in east London. Ph.D. dissertation, University of London. Becker, A.L. 1983 Literacy and cultural change: Some experiences. In: Richard W. Bailey and Robin Melanie Fosheim (eds.), Literacy for Life: The Demand for Reading and Writing, 45–51. New York: The Modern Language Association. Borchardt, Frank L. 1995 Etymology in tradition and in the Northern Renaissance (1968). In: Nancy Struever (ed.), Language and the History of Thought, 1–15. Rochester: University of Rochester Press.

A thousand years of the history of English

469

Chambers, J.K. 1979 Canadian English. In: The Languages of Canada, 168–104. Montreal: Didier. Crawford, S.J. 1929 Byrhtferth’s Manual (A.D. 1011). E.E.T.S., o.s. 177. London: Oxford University Press. Elstob, Elizabeth 1968 The Rudiments of Grammar, for the English-Saxon Tongue (1715). London: J. Bowyer and C. King. Facsimile. (English Linguistics, 1500–1800, no. 57.) Menston: Scolar Press. Foley, John P., ed. 1900 The Jeffersonian Cyclopedia. New York: Funk & Wagnalls. Frantzen, Allen J. 1990 Desire for Origins: New Language, Old English, and Teaching the Tradition. New Brunswick, NJ: Rutgers University Press. Free, John 1788 An Essay towards an History of the English Tongue [1749]. London: Printed for the Author. Garnett, James M. 1884 Review of the New English Dictionary, pt. 1, A-Ant. American Journal of Philology 5: 358–366. Gneuss, Helmut 1990 The study of language in Anglo-Saxon England. Bulletin of the John Rylands University Library of Manchester 72.1: 3–32. 1996 English Language Scholarship: A Survey and Bibliography from the Beginnings to the End of the Nineteenth Century. (Medieval & Renaissance Texts & Studies, no. 125.) Binghamton, NY: State University of New York at Binghamton. The Gospels of the Fower Euangelistes translated in the Olde Saxons Tyme. London: 1571 John Day. Henshall, Samuel 1798 The Saxon and English Languages Reciprocally Illustrative of Each Other. London: Printed for the Author.

470

Richard W. Bailey

Johnson, Samuel 1979 A Dictionary of the English Language [1755]. Facsimile ed. London: Times Books. Kennedy, Arthur G. 1961 A Bibliography of Writings on the English Language (1927). New York: Hafner Publishing Co. Krapp, George Philip 1919 The Pronunciation of Standard English in America. New York: Oxford University Press. MacMahon, Michael K.C. 1998 Phonology. In: Suzanne Romaine (ed.), The Cambridge History of the English Language: 1776–1997, 373–535. Cambridge: Cambridge University Press. Matthews, David 1999 The Making of Middle English, 1765–1910. Minneapolis: The University of Minnesota Press. Mugglestone, Lynda, ed. 2000 Lexicography and the OED: Pioneers in the Untrodden Forest. Oxford: Oxford University Press. Murray, James A.H. 1872 The Dialect of the Southern Counties of Scotland. London: Asher & Co. 1882– Thirteenth address of the president, to the Philological Society, delivered at 1884 the anniversary meeting, Friday, 16 May, 1884. Transactions of the Philological Society, 501–531. Partridge, Eric 1959 Origins: A Short Etymological Dictionary of Modern English. Second ed. New York: The Macmillan Company. Porter, Roy 1987 The Language of Quackery in England, 1660–1800. In: Peter Burke and Roy Porter (eds.), The Social History of Language, 73–103. Cambridge: Cambridge University Press.

A thousand years of the history of English

471

Quirk, Randolph 1985 The English language in a global context. In: Randolph Quirk and H.G. Widdowson (eds.), English in the World: Teaching and Learning the Language and Literatures, 1–6. Cambridge: Cambridge University Press. Spitzer, Leo 1962 Linguistics and Literary History: Essays in Stylistics (1948). New York: Russell & Russell. Stanley, Eric G. 1975 The Search for Anglo-Saxon Paganism. Cambridge: D.S. Brewer. Strang, Barbara M.H. 1994 A History of English (1970). London: Routledge. Sweet, Henry 1877 A Handbook of Phonetics, Including a Popular Exposition of the Principles of Spelling Reform. Oxford: Clarendon Press. 1890 A Primer of Phonetics. Oxford: Clarendon Press. 1892 A Short Historical English Grammar. Oxford: Clarendon Press. A Testimonie of Antiqvitie Shewing the Auncient Fayth in the Church of England. [1566?] London: John Day. Trench, Richard Chenevix 1851 On the Study of Words; English Past and Present. London: J.M. Dent & [1927] Sons. Wells, J.C. 1982 Accents of English. Cambridge: Cambridge University Press. Wright, Joseph 1905 The English Dialect Grammar. Oxford: Henry Frowde.

Name index

Aarsleff, Hans, 461, 468 Abrams, M.H., 210, 226 Adams, Jim, 359, 370 Adams, Robert, 63, 74 Adams, Valerie, 418–420, 444 Akatsuka, Noriko, 17 Alden, Raymond, 147 Alford, John, 57, 72–73 Algeo, John, 232, 242 Allen, Cynthia, 44, 280, 326–327, 350, 445, 469 Amos, Ashley Crandall, 261, 264 Andersen, Henning, 17, 21, 40 Anderson, C. Anthony, 352 Anderson, J.J. 151, 173, 175, 182 Anderson, Stephen, 415 Andersson, Erik, 183, 192–193, 203 Anttila, Raimo, 21, 32, 40 Arber, Edward, 174 Archan, Sabine, 52, 63, 73 Archibald, Elizabeth, 71–73 Arnovick, Leslie, 16, 302, 323 Askedal, John, 192, 203 Attridge, Derek, 147–148 Auer, Peter, 69, 73 Austin, John, 301, 323 Auwera, Johan van der, 183, 203–206

Bailey, Guy, 38, 40 Bailey, Nathan, 109, 121 Bailey, Richard, 2, 7-8, 12, 17, 449, 454–455, 467-468 Baker, Carl, 333, 351 Baker, Philip, 38, 40, Baldi, Philip, 203 Barbaresi, Lavinia, 301, 323

Barber, Charles, 239–240 Barlow, Michael, 240 Barnes, Michael, 183, 192, 203 Barnhart, David, 425, 432, 444 Batchelor, T., 270, 280 Bauer, Gero, 108, 281 Bauer, Laurie, 418, 444 Baugh, Albert, 211, 226 Baum, Paull, 126, 147–148 Beaken, M.A., 457, 468 Beardsley, Monroe, 146–147, 151 Becker, A.L., 450, 468 Benson, Larry, 167, 173 Bergen, Linda van, 242 Berlin, Brent, 30, 40 Berndt, Rolf, 71, 73 Berwick, Robert, 349, 351, 353 Besserman, Lawrence, 2, 17, 417 Bethel, Patricia, 263–264 Bhatt, Rakesh, 61, 73 Biber, Douglas, 413–414 Biggs, Henry, 154, 173 Bischoff, Otto, 126, 130, 148 Bjork, Robert, 264–265, 355 Blakemore, Diane, 31, 39–40 Bliss, Alan J., 212, 246, 248, 264 Blockley, Mary, 17, 393 Bloomfield, Leonard, 270, 280 Boase-Beier, Jean, 419–420, 444 Boersma, Paul, 269, 280 Booth, Stephen, 220, 226 Borchardt, Frank L., 468 Bréal, Michel, 28, 40 Bridges, Robert, 211, 226 Brink, Bernhard ten, 126, 130, 148 Brinton, Laurel, 16–17, 24, 40, 394, 396, 413–414

474

Name index

Briscoe, Ted, 349, 351 Britton, Derek, 300 Brogan, Terry, 148, 209, 225–226 Brosnahan, Leonard F., 287, 291, 298–299 Bruck, Anthony, 241 Brunner, Karl, 287, 299 Bruyn, Adrienne, 38, 40 Bülbring, Karl D., 287, 299 Burchfield, R.W., 122, 455 Burridge, Kate, 38, 41 Bybee, Joan, 3, 24, 27, 31, 39, 41, 233, 236, 240, 242, 376–377, 396

Cable, Thomas, 2, 6–7, 17, 125, 134, 146, 148, 156, 160, 164–166, 168–173, 177, 179, 181, 211, 226, 245, 250, 264 Campbell, Alistair, 286, 288, 294, 296, 299, 427, 444 Campbell, Lyle, 21, 39, 41 Carey, Kathleen, 375–377, 396 Carroll, John, 239–24 Carroll, Ruth, 17 Carstairs-McCarthy, Andrew, 329, 351 Carter, Richard, 279–280 Cassidy, Frederic, 122, 287, 299, 395 Cercignani, Fausto, 208, 226 Chambers, J.K, 183–184, 204, 459, 469 Chambers, R.W., 51–52, 73 Chapman, Don, 17 Chomsky, Noam, 106, 108, 218, 226, 270, 280 Christie, William, 240 Clanchy, Michael T., 53–54, 60, 73, Clark, Arthur Melville, 227 Clark, Eve V., 39, 329, 351 Clark, Robin, 349, 351 Clark-Hall, J.R., 289, 299, 428-429, 444 Claudi, Ulrike, 3, 24–26, 31, 41, 43 Colin, Chase, 264

Colman, Fran, 16, 283, 300 Comrie, Bernard, 375, 396 Conner, Patrick W., 258, 264 Cooley, Marianne, 234, 240 Cowie, Claire, 16 Crane, Susan, 53, 71, 74 Crawford, S.J., 464, 468–469 Croft, William, 19–20, 41, 115 Cukor-Avila, Patricia, 38, 40 Culpeper, Jonathan, 399–400, 414 Curti, Theodor, 112, 121 Curzan, Anne, 2, 17, 466 Cusack, Bridget, 402–404, 406, 408–409, 411–414

Dahlgren, Kathleen, 30, 41 Dalton-Puffer, Christanne, 16 Dancygier, Barbara, 31, 41 Dasher, Richard B, 29, 33, 49 Davenport, Michael, 205 Davidson, Mary, 55, 65, 74 Davies, Peter, 240 Declerck, Renaat, 375, 396 Denison, David, 38, 40, 437, 444 Diensberg, Bernhard, 112, 122 Diller, Hans-Jürgen, 16, 58, 74 Dinnsen, Daniel, 204 DiSciullo, Anna-Maria, 352 Dobbie, Elliot Van Kirk, 262–265 Dobson, E.J., 10, 130, 148, 208, 212, 215, 227, 232, 234–235, 240 Donaldson, E. Talbot, 126, 148 Donegan, Patricia, 191, 193, 199, 203–204 Dossena, Marina, 65, 74, 323 Dresher, B. Elan, 262, 264 Dressler, Wolfgang U., 301, 323 Duffell, Martin, 16 Duggan, Hoyt N., 146, 148 Duncan, Edwin, 17 Duncan, Erin, 353 Dury, Richard, 323

Name index Eckardt, Regine, 32, 42, 348, 351 Eckert, Penelope, 37, 42 Edwards, A.S.G., 55, 74, 408 Einenkel, Eugen, 368, 370 Eisenberg, Peter, 183, 204 Elliott, Nancy, 17 Ellis, Alexander J., 208, 227, 454, 457, 467 Elstob, Elizabeth, 467, 469 Erickson, Blaine, 2, 8, 17, 183, 188-190, 202, 204

Fabb, Nigel, 16 Facchinetti, Roberta, 323 Farkas, Donka, 353 Farr, James Marion, 333, 342, 351 Fasold, Ralph W., 352 Feist, Sigmund, 5–6, 121–122 Ferraresi, Gisella, 359, 370 Fidelholz, James, 231, 233, 240 Fikkert, Paula, 150 Finegan, Edward, 215, 227, 413–414 Fintel, Kai, von, 348, 351 Fischer, Olga, 23, 42, 46, 281, 362–363, 370 Fischer-Jørgensen, Eli, 236, 240 Fisher, John, 173 Fisiak, Jacek, 41–43, 46, 242, 414, 445 Fitzmaurice, Susan, see also Wright, Susan, 16–17, 32, 37, 42 Fletcher, Alan, 59–60, 74 Fludernik, Monika, 413–414 Foley, John P., 466, 469 Fosheim, Robin Melanie, 468 Fox, R.A., 241 Frankis, John, 279–280 Frantzen, Allen J., 466, 469 Free, John, 8, 462–463, 467, 469 Friedrichsen, G.W.S., 122 Fries, Charles C., 302, 323 Fulk, Robert, 10, 16–17, 146, 245, 250, 258, 261–264, 355, 370

475

Gabrielson, Arvid, 208, 221, 227 Galinsky, Hans, 430, 444 Garcia de Diego, Vincente, 112, 122 Garnett, James M., 453, 467, 469 Garrett, Andrew, 37, 39, 42 Gascoigne, George, 161–164, 166, 170, 174 Geeraerts, Dirk, 30, 42 Geis, Michael L., 29, 42 Gelderen, Elly, van, 16–17, 343, 351 Getty, Michael, 16–17, 355–356, 360–361, 366–367, 370 Givón, Talmy, 22, 24, 42 Gneuss, Helmut, 464, 469 Goh, Gwang-Yoon, 17 Golston, Chris, 17 Goossens, Louis, 30, 42 Görlach, Manfred, 10, 38, 43, 232, 234–236, 240, 400, 415 Gotti, Maurizio, 2, 12, 17, 301, 322–323 Gould, Stephen Jay, 350–351 Goyvaerts, Didier L., 280 Gray, Douglas, 150, 273, 387 Greenbaum, Sidney, 446 Grice, Paul, 29, 39, 43 Grossman, Robin E., 240 Gumperz, John, 52, 64–65, 74

Haberland, Hartmut, 183, 204 Hagiwara, Robert, 184, 196, 204 Hall, Joan, 122, Hall, Robert A., 80, 88–90, 106 Halle, Morris, 125–126, 130, 134–135, 141, 148, 151, 155, 174, 181–182, 215, 226–227, 270, 280 Halmari, Helena, 61, 63, 67, 71–72, 74 Hanson, Kristin, 2, 9, 17, 135, 142, 148, 155, 174, 207, 209, 222, 227, 259, 262, 265 Haraguchi, Yukio, 17 Harder, Peter, 375, 396

476

Name index

Harris, Joseph, 149, 174 Harvey, Carol, 57, 74 Haspelmath, Martin, 22, 24–25, 36, 43 Haugen, Einar, 203–204 Hayes, Bruce, 17, 134, 141, 149, 155, 174, 181–182, 248, 265 Heikkonen, Kirsi, 24, 48, 353 Heine, Bernd, 3, 24–26, 31, 41, 43, 49, 326, 351–352, 354 Helmholz, Richard H., 402, 405, 408, 413–415 Henshall, Samuel, 460–461, 465, 469 Herzog, Marvin I., 21, 49, 93, 108 Hickey, Raymond, 78, 242 Hock, Hans Henrich, 185, 204 Hockett, Charles, 270, 280 Hoekstra, Jarich, 183, 204 Hoffmann, Gerhard, 376–377, 396 Hogg, Richard, 17, 38, 40, 179, 182, 242, 284–285, 287–288, 291, 293, 296, 299–300, 354, 445 Hollander, John, 211, 227 Holthausen, Ferdinand, 289, 300 Holtman, Astrid, 209, 227 Honey, J., 280 Hooper, Joan, 231, 233, 236, 240 Hope, Jonathan, 171, 303, 317 Hopper, Paul, 24–26, 43, 48, 242, 326, 348, 352 Horn, Laurence R., 33, 43, 170, 444 Höskuldur, Thráinsson, 45, 183, 206 Howell, Robert, 16, 184, 204 Hughes, Merritt, 227 Hünnemeyer, Frederike, 3, 24–26, 43 Hunt, Tony, 54, 69, 75 Hurch, Bernhard, 242 Hutcheson, B. Rand, 250, 262, 265

Ihalainen, Ossi, 38, 47, 77, 414 Ingram, Martin, 411, 415 Inoue, Kyoko, 375, 396

Jacobs, Haike, 150 Jacobs, Neil, 183, 205 Jacobson, Rodolfo, 61, 75 Jakobson, Roman, 141, 144, 149, 262, 265 Janda, Richard, 17, 21, 44, 49 Jespersen, Otto, 22–23, 36, 39, 44, 136, 138, 144, 149, 170, 174, 180, 182, 267, 280, 371, 423, 431, 445 Johnson, Samuel, 94–96, 99–100, 103, 105, 107, 232, 462, 470 Joseph, Brian, 18, 21, 44, 49, 123, 149, 174, 228, 300, 322, 447, 457, 459, 471 Jucker, Andreas, 16, 19, 44, 47, 399, 413–415 Junius, Franciscus, 109, 122

Kachru, Braj B., 38, 44, 452 Kastovsky, Dieter, 17, 77, 108, 182, 281, 418, 426, 443–445 Kato, Kozo, 364, 370 Kay, Paul, 30, 40 Keenan, Edward, 2, 12–13, 17, 325, 330, 336, 341–342, 344, 346, 348, 352 Keller, Rudi, 20, 44 Kelly, Henry Ansgar, 17 Kemenade, Ans van, 2–3, 13, 17, 20, 23, 27, 44, 333, 352, 355–356, 359, 361–363, 367–371 Kemmer, Suzanne, 24, 44, 240 Kennedy, Arthur G., 437, 445, 467, 470 Keyser, Samuel Jay, 125–126, 130, 134–135, 141, 148, 151, 155, 174, 181–182, 215, 227 Kibbee, Douglas, 71, 75 Kiefer, Ferenc, 301, 323 Kiernan, Kevin, 355, 368, 371 Kiparsky, Paul, 36, 39, 45, 130, 134–138, 140–142, 144, 146–149, 154–155, 174–175, 178, 180–182,

Name index 211, 226–227, 239, 241, 259, 262, 265, 415 Klaeber, Frederick, 262, 265, 368–369 Klein, Wolfgang, 375, 396 Klemola, Juhani, 16 Kluge, Friedrich, 148 Knott, Thomas, 11, 287, 293, 295, 300 Koivisto-Alanko, Päivi, 30, 45 Kökeritz, Helge, 9, 129, 149, 208–209, 212, 215–222, 224–225, 227 König, Ekkehard, 29, 31, 39, 45, 49, 203–206, 352 Koopman, Willem, 16–17, 281, 362–363, 370 Koziol, Herbert, 426, 445 Krapp, George Philip, 262–265, 458, 470 Kretzschmar, William, 2, 4–5, 18, 79, 83, 86, 88, 94, 99–100, 107, 239, 241 Kroch, Anthony, 18, 23, 45, 103, 107, 346, 352 Krug, Manfred, 38, 45 Kryk-Kastovsky, Barbara, 16, 18, 413, 415 Kuno, Susumo, 329, 352 Kurath, Hans, 11, 82–91, 93, 103–104, 107, 269, 280 Kurylowicz, Jerzy, 27, 45 Kytö, Merja, 16, 24, 45, 48, 317, 322–323, 353, 399–400, 414 Labov, William, 21, 45, 49, 83, 91–93, 100, 103, 105, 107–108, 196, 205, 239, 241, 270, 399, 415 Ladefoged, Peter, 184, 186, 190, 196–197, 205, 237–238, 241 Ladusaw, William, 198, 205 LaGaly, M.W., 241 Lahiri, Aditi, 262, 264 Lancashire, Ian, 18 Langacker, Ronald, 20, 36, 39, 45 Lanz, Henry, 214, 228

477

Lass, Roger, 11, 21, 38, 46, 78, 116, 183–184, 188, 191, 203, 205, 232, 241, 243, 269–270, 275–281, 344, 353, 404, 426, 445 Laver, John, 235–236, 241, 286, 300 Lavin, Eva Delgado, 18 Lee, Hyunoo, 330, 336, 353 Lee, Jeong-Hoon, 2, 13-14, 18, 373, 393-394 Leech, Geoffrey, 446 Lees, Robert B., 106, 108 Lehman, Christian, 23-24, 36, 46, 348, 353 Lehmann, Ruth, 259, 265 Lehmann, Winfred, 48-49, 108, 122 Lehrer, Adrienne, 30, 46 Leslau, Wolf, 231, 234, 241 Levi, Judith, 206 Levinson, Stephen, 29, 39, 46 Lewis, C.S., 145, 149 Li, Charles, 45 Li, Xingzhong, 2, 6–7, 153–155, 157-159, 162, 164-165, 167-168, 171, 174, 177–180 Liberman, Anatoly, 2, 5–6, 18, 109, 113, 115, 120, 122 Liberman, Mark, 141, 146, 149, 153–154, 172, 174, 179-180, 182 Lieber, Rochelle, 419, 445 Lightfoot, David, 3, 19–21, 23–25, 46 Lima, Maria, 323 Lindau, Mona, 195–196, 205, 239, 241 Lindelöf, U., 437, 445 Lindquist, Hans, 18 Longfellow, Henry Wadsworth, 153, 160–161, 174 Los, Bettelou, 16 Low, Soon-Ai, 18 Lüdi, Georges, 60, 64, 72–73, 75 Luick, Karl, 108, 279, 281, 286, 288, 296, 300 Lutz, Angelika, 16

478

Name index

MacEachern, Margaret, 155, 174 Machan, Tim, 53, 55, 57, 60, 62, 65–66, 72, 75 MacMahon, Michael K.C., 457, 467, 470 Macrae-Gibson, O.D., 263–264 Maddieson, Ian, 184, 186, 205, 237–238, 241 Magnuson, Karl, 138, 149 Malkiel, Yakov, 48–49, 108 Malone, Joseph, 207, 212, 228 March, John, 407, 415 Marchand, Hans, 418–419, 426, 430, 445 Matthews, David, 16, 461-462, 470 Matthews, J. Brander, 215, 228 Mayhew, Anthony, 120, 122 Maynor, Natalie, 38, 40 McCoard, Robert, 375, 396 McCully, Christopher, 16, 38, 40, 151, 173, 175, 179, 182 McDavid, Raven I., Jr., 85–86, 94, 107, 280 McGowan, Joseph P., 18 McSparran, Frances, 16, 18 Meillet, Antoine, 22, 28, 46 Merritt, Herbert D., 227, 289, 299 Meskhi, Anna, 18 Mettinger, Arthur, 175, 182 Michaelis, Laura, 375–376, 379, 382, 388, 396 Minkova, Donka, 18, 96–98, 104, 108, 147, 149, 161, 167, 175, 179, 182, 203, 226, 231, 238, 243, 268, 270–271, 280–281, 443, 466 Mitchell, Bruce, 333, 336–337, 353, 355, 360, 366–368, 371, 377, 397, 461 Mittwoch, Anita, 375, 397 Molina, Clara, 30, 46 Montgomery, Michael, 16 Moore, Colette, 2, 14, 18, 399 Moore, Samuel, 11, 287, 293, 295, 300

Mossé, Fernand, 131–132, 150, 327, 353 Mourek, V.E., 368, 371 Mugglestone, Linda, 452–453, 470 Mulder, Jean, 38, 41 Murray, James, 110, 120, 122, 452–458, 465, 470 Murray, Robert, 16, 235, 241 Mustanoja, Tauno, 126, 150, 344, 353, 376, 397, 430–431, 445 Muysken, Pieter, 61, 63, 72, 75 Myers-Scotton, Carol, 56, 63, 67, 69–70, 75

Nakao, Toshio, 147, 150 Nevalainen, Terttu, 24, 37–38, 46–47, 76–77, 104, 108, 175, 399, 414–415 Newcomer, Alphonso, 207, 214, 228 Newmeyer, Frederick, 3, 20–22, 25–26, 34, 39, 47 Nielsen, Hans Friede, 77, 205 Niles, John, 370 Nixon, G., 280 Niyogi, Partha, 349, 353

Obermeier, Anita, 355, 370 Oehrle, Richard, 160, 175 Ogura, Michiko, 333, 353 Ogura, Mieko, 10, 16, 232–234, 239, 241 Ohala, John, 237, 241 Oras, Ants, 214, 228

Pagliuca, William, 3, 24, 27, 31, 39, 41, 47, 376, 396 Pahta, Päivi, 16, 47, 65, 70, 75–76, 175 Palander-Collin, Minna, 24, 47 Palmer, Frank Robert, 46, 304, 323 Palmer, Leonard, 285-286, 300 Pandit,P.B., 192, 205

Name index Partee, Barbara Hall, 406, 415 Partridge, Eric, 113, 122, 432, 446, 465, 470 Passy, Paul, 269, 281 Pearsall, Derek, 57, 74–75, 77 Peitsara, Kirsti, 342, 353 Pennanen, Esko V., 418, 443, 446 Peranteau, Paul M., 206 Percy, Thomas, 8, 461–462, 467 Pérez, Aveline, 24, 47 Perkins, Revere, 3, 24, 27, 31, 39, 41, 376, 396 Phares, Gloria, 206 Phillips, Betty, 2, 10, 18, 231, 233–234, 239, 242 Picone, Michael, 72, 75 Piera, Carlos, 154, 175 Pinsky, Robert, 207, 222, 227–228 Pintzuk, Susan, 355, 360, 371 Poplack, Shana, 38, 47, 63, 66–67, 76 Porter, Roy, 451, 470 Preminger, Alex, 227 Prince, Alan, 153–155, 172, 174–175, 179, 182 Prince, Ellen, 183, 205 Pullum,Geoffrey, 184, 197, 205, 280 Puppel, Stanislaw, 242 Pyles, Thomas, 232, 242

Quirk, Randolph, 287–288, 296, 300, 417, 420, 446, 452, 455, 471

Raab, Eduard, 431, 446 Raine, James, 402, 416 Ramat, Anna Giacalone, 48 Ramat, Paolo, 21, 47 Raumolin-Brunberg, Helena, 37, 47, 104, 108 Ravid, Dorit Diskin, 21, 47 Rayburn, Wallace, 113, 123 Redford, Michael, 130, 150

479

Reh, Mechthild, 24, 43, 326, 351 Reichl, Karl, 149 Rhodes, Richard, 231, 242 Riad, Thomas, 18 Richman, Barry, 240 Richman, Gerald, 401, 416 Richter, Michael, 54, 76 Rickert, William, 210, 228 Rickford, John, 38, 47 Ringler, Richard, 287, 299, 395 Rissanen, Matti, 24, 38, 47–48, 76–77, 175, 353, 414 Ritchie, Frank, 207, 224, 228 Ritt, Nikolaus, 235–236, 242 Roberts, Ian, 3, 19, 25, 27, 36, 48, 348–349, 351, 353 Roberts, Jane, 262, 265, Robins, Robert H., 80–81, 108, 322–323 Robinson, Ian, 118, 150, 232 Rodríguez Álvarez, Alicia, 54, 65, 76 Romaine, Suzanne, 53, 56, 65, 67, 72, 76, 470 Rosenbach, Anette, 42, 46 Rothwell, William, 53–54, 61, 68–72, 76 Roussou, Anna, 27, 48, 348, 353 Rowland, Beryl, 150 Runge, Richard, 184, 205 Rusche, Philip, 18 Russom, Geoffrey, 2, 10, 13, 18, 245–248, 253, 259, 262–265 Ryder, Frank G., 138, 149

Saintsbury, George, 125, 146, 150, 214 Saito, Sumio, 189, 205 Sauer, Hans, 427–430, 446 Schendl, Herbert, 2, 4, 18, 51, 55–56, 62–63, 65–66, 70–71, 76 Schiffrin, Deborah, 352 Schuchardt, Hugo, 231, 242 Schütze, Carson T., 443, 446

480

Name index

Schwartz, Elias, 147, 150 Schwenter, Scott, 375, 397 Selkirk, Elisabeth, 143–144, 147, 150, 418, 446 Sharpe, James A., 416 Short, Ian, 54, 71, 77 Siemund, Peter, 333, 352 Sievers, Eduard, 245–247, 258, 263–265, 287–288, 299–300 Simpson, J.A., 122, 445–446 Skeat, Walter, 110, 117–118, 120, 123, 369, 426, 446 Slobin, Dan, 20, 36, 48 Smith, Catherine, 18 Smith, Carlota S., 375, 385, 393, 397 Smith, Egerton, 147, 150 Smith, Henry L., 270, 281 Smith, Estellie, 243 Smithers, G.V., 126, 150 Smolensky, Paul, 155, 175 Southworth, James, 126, 150 Spaelti, Philip, 353 Sperber, Dan, 39, 49 Spevack, Marvin, 232, 239, 242 Spitzer, Leo, 465, 471 Spruiell, William, 18 Stabler, Edward P., 348, 352 Stampe,David, 186, 191–193, 199, 204, 206 Stanley, Eric G., 150, 449–450, 471 Steele, Joshua, 147, 151 Stein, Dieter, 33, 42, 45–46, 48, 323 Stenbrenden, Gjertrud, 279, 281 Stern, Gustav, 28–31, 39, 48 Stewart, George R., Jr., 137, 151, 177, 179, 182 Stockwell, Robert, 2, 11, 18, 79, 96–98, 104, 108, 161, 175, 179, 182, 203, 226, 231, 238–239, 243, 267–268, 278–281, 290, 293, 300, 443, 466 Strang, Barbara, 235–236, 243, 280, 460, 471

Struever, Nancy, 468 Stryjewski, Kurt, 209, 228 Sullivan, Carmeline, 57, 77 Sundquist, John, 18 Svartvik, Jan, 446 Svorou, Soteria, 24, 48 Sweet, Henry, 270, 281, 369, 417, 426, 430, 446, 454–457, 459–460, 465, 467, 471 Sweetser, Eve, 31, 33, 39, 48 Syea, Anand, 38, 40 Szwedek, Aleksander, 445

Taavitsainen, Irma, 38, 47, 77, 175, 399, 414 Tabor, Whitney, 27, 48 Taglicht, J., 302, 324 Tani, Akinobu, 18 Tarlinskaja, Marina, 156–157, 159, 165–166, 168, 170, 175 Tieken-Boon van Ostade, Ingrid, 370 Tiersma, Peter Meijes, 183, 204 Tillotson, Geoffrey, 213, 228 Tottie, Gunnel, 370 Trager, George L., 243, 270, 281 Traugott, Elizabeth Closs, 2–3, 15, 18–19, 24–27, 29, 31, 33, 41, 43, 45, 48–49, 333, 336, 348, 352, 354, 377, 393–394, 397 Trench, Richard Chenevix, 465, 471 Trier, Jost, 29, 49 Trnka, B., 278, 281 Trotter, David A., 53, 67–70, 75–77, 413, 416 Trudgill, Peter, 184, 204 Tsoulas, George, 371

Vance,Timothy, 188, 196, 206, 240 Vasmer, Max, 5 Vennemann, Theo, 17, 235, 241–242 Vincent, Nigel, 23, 44

Name index Visser, Fredericus Theodorus, 23, 49, 375–377, 397, 401, 416, 426, 429–431, 447 Voigts, Linda, 53–55, 58–59, 65, 70, 72, 77 Vorlat, Emma, 322, 324 Voyles, Joseph B., 427, 447 Vrba, Elizabeth, 350–351

Wackernagel, Wilhelm, 123, 370 Wald, Benji, 2, 15, 18, 417 Wales, Katie, 17 Wang, William, 231, 243 Warner, Anthony, 17, 23, 49, 371 Wartburg, Walther von, 121, 123 Watkins, Calvert, 289, 300 Webster, Noah, 118, 123, 421, 454–455 Weekley, Ernest, 113, 123 Wehrle, William, 57, 77 Weiner, E.S.C., 122, 445–446 Weinreich, Uriel,, 21, 49, 93, 108 Wells, J.C., 184, 188–189, 192, 195, 198–199, 202, 206, 232, 243, 457, 471 Wenzel, Siegfried, 54–55, 58–59, 62, 65–66, 68, 70, 77 Werth, Ronald, 203 Weyhe, Eivind, 183, 192, 203 White, David, 2, 11–12, 18, 283 Whitehall, Harold, 227

481

Widdowson, H.G., 471 Wilber, Ronnie, 242 Williams, Joseph, 30, 49 Willinsky, John, 432, 447 Wilson, Deirdre, 39, 49 Wimsatt, W.K., 126, 130, 134, 136, 146–147, 151, 177, 182, 213, 228 Winteler, Jost, 112, 123 Wrenn, C.L., 287–288, 296, 300 Wright, Elizabeth, 287-288, 293-294, 300 Wright, Georg, Henrik von, 322, 324 Wright, George T., 213, 229, Wright, Joseph, 119, 123, 287-288, 293-294, 300, 457, 459, 471 Wright, Laura, 55, 62, 66, 68-69, 72, 77-78 Wright, Susan, 33, 48, see also Fitzmaurice Wurff, Wim van der, 362–363, 370 Wyld, Henry Cecil, 208, 214, 229, 450

Youmans, Gilbert, 2, 6–7, 17, 129, 149, 151, 153–155, 162–164, 166–167, 170, 172, 174–175, 177–180, 182

Zelëny, Michael, 352 Zumthor, Paul, 57, 78 Zwicky, Arnold, 29, 37, 39, 42, 49, 209, 215, 221, 229

Subject index

abduction, 21 activity constraint, see constraint acquisition, see language acquisition A-curve, see asymptotic curve Advice see pragmatic value adverb(ial)(s), 14, 24, 31, 33–34, 43, 46, 133, 330, 333, 335, 338–339, 344, 355–357, 359–366, 369, 379–383, 385–386, 391–393, 403, 431, 438, 445 Alfred(ian), 73, 263, 369, 462 Ælfric, 248, 262, 369 African American Vernacular English, 38, 47 Afrikaans, 183 Aktionsart, see aspect alliteration, 125–126, 146, 148–149, 247–248, 250, 252–253, 263, 327, 333, 335, 355 alliterative binding rule, 247 Alliterative Revival, 126 allophone, 11, 186–188, 194, 268–269 alternating meter, see meter alternating stress, see stress alveolar, 9, 184–185, 188–191 ambiguity, 14, 43, 129, 139, 236, 384, 396, 409, 417, 428–429, 433, 441 ambiguous category, 15, 421–422, 424–428, 431–435, 440, 442 amelioration, see semantic change American English, 8, 38, 45, 47, 107, 183–184, 194, 196, 233–234, 297, 323, 430, 458 anacrusis, 248–249 analogy, 25, 32, 89, 91, 195, 232, 234, 242, 293, 298 anapest, see foot

anaphora, 13, 325, 329, 337, 345–346, 349–350 Anaphora Profile, 331, 344, 347 Anglo-French, 70–71 Anglo-Irish, 277, see also Irish English Anglo-Norman, 57, 61, 67–69, 72, 74, 77, 113, 327 Anglo-Saxon, 262, 265, 295, 299, 351, 369, 444, 449, 469, 471 ANTI-SYNONYMY, 325, 328–329, 345, 347–348 apical, 187, 189–190, 192, 194 apocope, 127 apposition(al), 330, 341, 338–339, 344 approximant, 8–9, 183–185, 188–189, 191, 194–197, 201–202 argot, 451 articulation, 8–9, 184–198, 201–203, 236–238, 276, 280 – articulator, 186, 188, 236, 238 – articulatory effort, 269 – articulatory factors, 186–189 – articulatory optimization, 269 – articulatory unit, 235 – hyperarticulate, 186, 188, 202, 276 – hypoarticulate speech, 187, 193, 201 – place of articulation, 9, 188–190, 193, 195, 236–237, 240 AS, see anti-synonymy aspect, 12, 22, 30–31, 304, 311, 377, 396–397 – aktionsart, 394 assimilation, 11, 192–193, 199, 233–234, 237, 238–239, 273 Assurance, see pragmatic value asymptotic curve, 5, 102–103

484

Subject index

auditory effect, 196 Australia, 16, 38, 41, 455–456 Australian English, 199 auxiliary (verb) 27, 42, 133, 302, 358

back vowel, see vowel back-formation, see word-formation backness, see vowel base derivation, see word-formation Beat Addition, 144, 147 Bede, 336, 464, 468 Beowulf, 245–246, 248, 258, 261–262, 264–265, 331, 335–336, 355–361, 364–371, 390, 395, 449 bilingual(ism) 4, 52–56, 58–60, 64, 67–68, 71, 74, 77 Binding (Theory), 63, 331, 336–337, 342, 345- 346, 352 – local binding, 337, 345, 346, 349–350 – locally bound, 13, 325, 329–331, 347 Blake, William, 212 bleaching/bleached, see semantic change borrowing/borrowed, 5, 11, 29, 51, 67–70, 72, 86, 93, 110, 112–115, 119, 192, 200, 270, 275, 279, 327, 404 boundary, 154, 158, 170, 192–193, 200, 325, 404 – clause, 154 – dialect, 81, 87–88, 91, 93 – foot, 158, 170 – metrical, 154 – morpheme/morphological, 325, 427–428 – phonological, 170 – phrase, 144, 154–155 – sentence, 56, 154, 170 – syllable, 154, 193, 200, 290, 294, 299 – syntactic boundary, 158

– word, 154–155, 192, 200–201 breaking, 12, 283, 285–287, 290–294, 298, 438 British English, 196, 268, 351, 430, 454 Brosnahan Effect, 287, 291, 298 buccal, 11, 285 buffer code, 72 Byrhtferth, 8, 464–465, 469 Byron, 212, 227

caesura, 6, 7, 126, 141, 145, 181, 252–253 Canadian English, 87, 117, 469, 455, 459 Canadian raising, 11, 268–269, 459 Canterbury Tales, see Chaucer Caxton, William, 456 Celtic, 114, 116, 119, 466 center drift, 268–270 central vowel, see vowel centralizing/centralized, 211, 237, 268 chain shift, 98, 268, 270 Chancery, 402, 413, 456 Chaucer, 6–7, 76, 125–134, 146–151, 153–154, 156–171, 173–175, 177–179, 182, 214, 227, 327–328, 341, 350 – Canterbury Tales, 128, 131, 162, 167–168 – General Prologue, 128–129, 132–133, 162 Christian(s), 46, 246, 250, 342, 368, 449, 462, 464 – pre-Christian, 449 CI, see Constituency Interpretation clitic(ization), 7, 13, 44, 143, 171, 326, 333 – de-cliticization, 143 – proclisis, 326, 338 – proclitic, 360 – procliticization, 350, 369

Subject index – procliticized, 326, 328, 344, 358–359 – procliticizing, 348 closure, see Principle of Closure Cockney, 269, 285, 457–458 code-(switching), 4, 52–76, 78, 400–407, 412–413, 420 – supra-regional code, 69 cognates, 110–111, 115–117, 119, 121 cognitive linguistics, 29–30 coinage(s), 114, 428–430, 440–442 Common Germanic, see Germanic comparative method/reconstruction, 109, 115 compensatory lengthening, see lengthening Compositionality, 328 compound(s), 5, 96, 111, 113, 115, 245–246, 249, 253, 263, 417–418, 420, 422–423, 425–427, 429–434, 441, 442, 445 – bisyllabic compound(s), 263 – Compound Stress Rule, 247 – compound verbs, 15, 420–424, 426, 438, 440–443 – disguised compound(s), 113, 115 – nominal compound(s), 429- 430, 433, 435, 438, 442–443 – pseudo-compound(s), 418–419 – synthetic compound(s), 418 – verb-verb, also VV compound(s), 15, 417, 419–420, 422, 425, 443 – VN(s) 426–431, 433–440 – V1-ing compounding, 431, 434–435 connotative meaning, 64, 410 consonant, 11, 171, 181, 187–189, 192–193, 195, 197, 201, 205, 210, 222, 232, 234, 236–239, 284–285, 290, 295, 297, 450 constituent, 13, 62, 135, 141, 156, 178, 181, 194, 334, 338, 345, 361, 364, 368–369, 417–425, 430, 432, 434–435, 438

485

– Constituency Interpretation, 325, 328, 338, 347 – constituent bracketing, 7, 178 – constituent structure, 141, 359, 368 – morphological constituent, 417–425, 430, 432, 434–435, 438 – negative constituent, 13, 359, 361, 368 – prosodic constituent, 193 – switching of constituents, 62–63, 165 – syntactic constituent, 61, 63, 135, 137, 178, 248, 253, 262–263 – verse constituent, 153, 156, constraint(s), 6, 13, 61, 63, 73, 76, 135, 140, 155, 165, 169–170, 172, 203, 222, 240, 245, 252–253, 262–263, 379, 423 – activity constraint, 15, 422–425, 433, 442–443 – constraint on verb compounding, 15, 422 – Equivalence Constraint, 63 – Monosyllabic Word Constraint, 134, 136–138, 140, 144, 146, 155, 162–163, 172, 180 – phonotactic constraints, 203, 234, 240 – semantic constraints, 13, 325 – syntactic constraints, 52, 61, 63, 66, 72, 336 contrast interpretation, 337, 346–348 core, 86–88, 222, 326, see also periphery coronal, 8–9, 183–195, 197, 200–202 corpus/corpora, 14, 32, 40, 52, 55, 63, 65, 99 – Corpus of Early English Correspondence, 104 – Corpus of Early Middle English, 12, 428, 430 – Corpus of (Latin) sermons, 62 – Helsinki Corpus, 12, 14, 301, 319, 322–323

486

Subject index

– Early Modern English corpus, 301, 304–8, 315–316, 319 – Old English poetic corpus, 13, 246 – Study of corpora, 14, 32, 40, 52, 55, 63, 65, 99 – Toronto text corpus, 364 court records, 14, 65, 72, 399–402, 410–416 Creole (English), 38, 198, 200 Cynewulf, 252, 255–258, 262, 264–265

Danelaw, 327 Danish, 114, 183–184, 204, 368 Dante, 80, 227 DARE, 119, 122 dating (criteria), 10, 13, 212, 245, 249–250, 256, 261, 264, 325, 355, 342, 361, 366–368, 432 decasyllabic (verse), 6, 156, 158, 160, 173 DECAY, 325–327, 346, 348 de-cliticization, see clitic defamation/defamatory, 14, 399–400, 402–4, 407, 410–413 denotative (meaning), 378, 383, 385, 392, 410 dental(s), 9, 185, 188–189, 191–192, 197, 201, 236–237, 239, 242, 297 deontic, see pragmatic value deposition, see text type dialect(s), 9, 38, 44, 79–81, 83–89, 91–93, 97–99, 104, 106, 114, 119, 183–184, 188–189, 195–196, 200–203, 209, 215–216, 218, 221, 268–269, 402, 450–451, 457, 459–460 – coherent dialect, 97 – dialect interference, 271 – dialectology, 4–5, 39, 78–79, 81, 84–85, 87–88, 92–94, 99–100, 104–105, 107–108, 204

dialogic form, 314 diaphones, 11, 268–270 Dickinson, 214 diglossia, 53, 55 diphthong(s), 11–12, 206, 211, 269–273, 275–276, 279, 283, 291, 457–458 – short diphthongs, 283–284, 290 diphthongization, 11, 233, 242, 267–273, 275–276, 279 dipodic verse, 125 direct speech, 14, 399–400, 402–406, 408–409, 411–413 directionality of change, 19–21, 25–26, 30, 35, 46 discourse, 19, 22, 24–25, 31–33, 37, 51, 54–55, 57, 59, 61, 64–67, 141, 214, 351, 384, 396, 400–401, 403, 409, 413–415, 449 – levels of discourse, 401, 409, 411 dissemination patterns, 87 dissimilation, 11–12, 283, 287, 289–291, 294, 298 disyllabic, 128, 130, 140, 144, 159–160, 166–167, 179, 325, 344 Donne, 125, 138–139, 142 doublets, 118, 166–168, 217–219, 326 Dunbar, 57, 73 Dutch, 114, 118–120, 183–184 dynamic, see pragmatic value

-e, 7, 127, 129, 132–134, 148, 150, 159, 178, 293–294, 427–431 Early Modern English, 10, 12, 14, 47, 53, 62–63, 68, 71–73, 76–77, 115, 212, 214, 220, 231, 234, 236, 238, 240, 242–243, 280, 301, 323, 399, 400, 402, 406, 413, 415–416 elision, 127, 167 Emerson, 214 endocentric, see word-formation English Vowel Shift, see vowel shift

Subject index epenthesis, 271, 273 epiphenomenon, 3, 7, 19–20, 25–26, 32 epistemic, 24, 31, 33, 39, 48, 322 epistolary, 65 Equivalence Constraint, see constraint ethnocentrism, 450 etymology, 3, 5–6, 19, 39, 48, 67, 85, 109–117, 119–123, 289, 300, 427, 459, 461, 464–465, 468, 470 Euro-spelling, 460 exaptive change, 350–351 existential, see pragmatic value Existential/experiential perfect, 374 extralinguistic, 53–54, 59, 61, 64, 66, 384 extrametrical, see meter

Faroese, 183, 192, 203 First Fronting, 283–284, 286–287, 290, 295–297 flapping, 269, 297 foot/feet, 6–7, 10–11, 112, 129, 134, 136–140, 142, 146–147, 154–158, 164, 166–172, 178–179, 181, 231–233, 239, 247–248, 258, 264, 367, 421 – anapestic foot/anapests, 159–160, 179 – foot boundaries, see boundary – iamb(ic), 6, 136, 138, 155, 166 – inverted first foot, 128–129, 132–134, 138, 178–179 – ionic, 171 – isochronous foot, 178 – reversed foot, 139 – trisyllabic, 7, 142, 178–179 – trochee(s), 138, 157, 164–172, 179 – word-foot theory, 248, 258 French, 4, 11, 51–54, 56, 58, 61, 65, 68–73, 75, 76, 80, 84–85, 88, 90–91, 93, 112, 114–115, 120, 149, 154, 160, 173–174, 182, 184, 190, 211, 214,

487

225, 269–271, 274–275, 279, 326, 329, 401–402, 413, 455, 460, 463, 465–466 frequency, 5, 10, 13, 23, 55, 58, 62–63, 67–68, 95–96, 100, 102–103, 105, 160, 165, 178, 233, 239–240, 242, 245–250, 252–256, 263, 305, 355–356, 359, 363, 432, 436, 439 frequency hierarchy, see hierarchy front vowel(s), see vowel frontness, see vowel function words, 7, 72, 143, 178, 248–249, 252, 326, 350 functional category, 22, 27–28, 31, 33–35, 37 functional head, 27, 362, 364 functionalist, 19, 21, 41 functional-pragmatic, 61, 64 future tense, 12, 301–302, 304–305 FWP, see clitic, proclisis

GAINFUL LEARNING, 329 geminate(s), 189, 284–285, 294–297, 299 General Prologue, see Chaucer Generalized Invited Inferences, see Invited Inferences generative metrics/prosody, 6–7, 125–127, 134–136, 140, 145–147, 172–173, 177–181 generative tradition/study, 20, 23, 40, 125, 149, 175, 179, 209, 218, 223, 418, genre(s), 12, 19, 31–32, 37, 55–57, 62–66, 71, 221, 301, 304–305, 319, 322, 413–414 Georgian, 350 German, 29, 72–73, 79, 88–89, 91, 112, 114, 116–117, 119–120, 149, 174, 182–184, 202, 204, 323, 348, 427, 460, 466

488

Subject index

Germanic, 3, 8, 45, 110–111, 115–117, 149, 183–186, 189, 203–205, 228, 241–242, 246, 258, 262, 264–265, 270, 279, 286, 289, 296, 326, 361, 426–427, 447, 449, 462, 468 – Common Germanic, 5, 112, 116 – Indogermanic, 80 – Proto-Germanic, 185, 205, 427, 447 – West Germanic, 120 gerund, 423–425, 429–431, 434, 438, 442 GIINs, see Generalized Invited Inferences glide, 202, 234, 272–273, 435, 444, 457 glottal, 285 Gothic, 111, 116, 120, 359, 426, 446 Government, see binding Gower, 113, 126–127, 130, 181, 327, 341 grammaticalization, 3, 13, 19, 21–22, 24–31, 33–49, 325, 348, 351–354, 370–371, 396–397 – secondary grammaticalization, 27 graphic variants, see variant Great Vowel Shift, see vowel shift Greek, 80, 109, 116, 120

h, 11, 118, 284–289, 291, 293–294, 298–299, 454, 459 h-deletion, 454 Hawaii English, 198, 200 headedness, 419–420 headless line(s), 127, 129, 132, 134, 159, 169 Hebrew, 2, 17, 47, 109, 350 Helsinki Corpus, see corpus hemistich(s), 7, 154, 157–159, 168–170 hendecasyllable, 160 heterogloss(es), 87–88, 90, 93, 104 hiatus, 283, 287–294, 298 hiatus resolution, 283, 287

hierarchy – case hierarchy, 349 – frequency hierarchy, 62 – prosodic hierarchy, 134, 146, 149, 154–155, 156–158, 182 – syntactic hierarchy, 137 high vowel, see vowel Hindi, 192 historical pragmatics, see pragmatics Hoccleve, 126–127 hovering accent, 137

iamb, see foot Icelandic, 183, 206 ICEs, see Inherently Contrastive Expressions ictus, 7, 128, 134, 136, 138–139, 143–147, 177–178 non-ictus, 128, 142, 177 idiolect(s), 94, 209 IINs, see invited inference illocutionary, 301, 320 implication/implicature(s), 29, 32, 379, 384, 394 implosion, 268, 271, 279 indirect speech, 14, 401, 403, 406–407, 409 Indo-European, 5, 80, 112, 117, 184, 300, 426 Indogermanic, see Indo-European, also Germanic INERTIA, 325, 327–329, 336–340, 344–345 -ing(form), 68, 419, 423–425, 429–431, 431, 434–435, 438, 439, 441–442 Inherently Contrastive Expressions, 330, 334, 341–342, 345, 347, 350 innovation(s), 8, 15, 20–21, 23, 37, 73, 86, 175, 183–184, 210, 220, 223, 233, 263, 302, 322, 373, 450 Intention, see pragmatic value

Subject index Interest, see metrical (variety) intersentential, 55, 61, 65 intrasentential, 55, 56, 59, 61, 65, 68 inverted first foot, see foot invited inference(s), 29, 35, 42 – Generalized Invited Inferences, 29, 33, 35 – Invited Inferencing Theory, 29 Ionic, see foot Irish English, 38, 60, 194, 459, see also Anglo-Irish isochrony, 178–180 isogloss, 87, 107 isomorphy, 147 Italian, 80, 110, 120, 160, 182, 323, 455 Japanese, 188–190, 196, 204, 206, 350 Jonson, Ben, 191, 451 King Alfred, see Alfredian labial, 201–203, 236 LAD, see language acquisition laminal, 189, 194 Langland, 57, 75, 148 language acquisition, 20–21, 23, 25, 37, 202, 204, 351 language change, 1, 3, 19–21, 23, 36–37, 45, 49, 79, 86, 92, 202, 250, 252, 351, 450, 456 et passim language contact, 51, 54, 71, 327 Language Custom/Sprachusus, 93–94, 98, 105 language shift, 70, 78 laterals, 237 Latin, 4, 24, 51, 53–54, 56–59, 61–63, 65, 67–70, 72, 74–77, 80, 84–85, 88, 114, 119–120, 173, 185, 275, 277, 289, 300, 326, 332, 359, 370, 400–407, 409, 411, 413, 460, 462, 464–466

489

Latinate, 433 laxing, 232, 236, 238–239 lengthening, 10, 226, 239, 271–272, 277–278, 281, 283, 288, 290, 294, 298 – compensatory lengthening, 283 – Open Syllable Lengthening, 10, 272, 277 lenition, 186, 188–189, 191, 195, 202 levels of discourse, see discourse lexical diffusion, 10, 83, 231–233, 238, 240, 242 lexical field, 29–30 lexical words, 7, 143, 144, 177–178 lexicalisation(s), 424–425 lexicography/ lexicographer(s), 67, 77, 109, 270, 452, 470 Linguistic Atlas, 83, 92, 94, 99–100, 105, 107, 279 link-vowel, see vowel loan words, 56, 404, see also borrowing local antecedent, 333, 336–337, 347 – non-local antecedence, 348 – non-local antecedent(s), 345, 347 local binding, see binding London English, 277, 327, 341 long vowel(s), see vowel Longfellow, 153, 160–161, 174 loss of unstressed vowels, see vowel Low German, 114 lowering, see vowel Lydgate, 126–127, 159

macaronic, 4, 51, 54, 56–57, 59–60, 62, 66, 68, 71–74, 76, 78 main verb(s), 263, 316, 326 manner adverbials, 14, 33, 382–383, 391, 392–393 Matrix Language Frame Model, 63, 69 Maxim of Extravagance, 36 merger(s), 11, 268, 271–272, 274–275, 277, 279, 427

490

Subject index

meter/metrical, 6–8, 10, 13, 125–128, 130–131, 134–147, 149, 151, 154–159, 161–167, 169–173, 175, 177–182, 210–211, 213, 229, 245–249, 252–253, 255, 258–259, 262–264, 356, 360, 367, 370 – alternating meter, 134, 147, 171, 179 – extrametrical, 128, 159, 248, 250, 252–253, 263 – metrical anomaly, 253 – metrical boundary, see boundary – metrical complexity, 246–247, 253–254, 258, 262–263 – metrical considerations, 13, 355, 360, 367 – metrical evidence, 247 – metrical faults, 252–253, 263 – metrical norm, 10, 254, 258–259 – metrical pause, 126, 139–142, 145–146, 172, 180 – metrical refinements, 252–253 – metrical resolution, 262, 283, 287, 298 – metrical template, 127, 141, 146 – metrical variant, 355 – metricality, 7, 135–136, 145, 178, 181 metonymy, 28, 32 Middle English, 4, 8, 15, 47, 53, 60, 62, 68, 70–71, 73–77, 96, 109, 119, 125–126, 130, 146, 148–150, 175, 179, 232–235, 241, 271–272, 274, 277, 279, 281, 301, 323, 325–326, 328, 341, 342, 344–346, 348, 351, 353, 370, 376–377, 397, 414, 419, 428–432, 436–437, 442, 445, 461, 467, 470 Milton, 7, 126, 153–154, 156–157, 162–163, 169–170, 175, 211, 226–227 mixed-language texts, see text types

modal verb(s), 31, 301, 321, 360, 366–367 – modal auxiliary, 12, 302 – modal categories, 12, 305 – modal functions, 305 modality, 12, 31, 302, 305–306, 321–322, 377, 396 Modern English, 12, 14, 23, 39, 44, 46, 63, 122–123, 129–130, 139, 179, 236, 275, 280–301, 323, 336–337, 343, 345, 373–378, 383–385, 387, 392–393, 445, 470 monolingual, 4, 38, 51, 54–56, 58, 64, 70, 72 monophthong(al), 269, 276, 457 monosyllabic, 127, 129–130, 137, 140, 165, 235, 326, 366–367, 433 Monosyllabic Word Constraint, see constraint Mopsey(s), 276–277 mora(ic), 204, 291, 294 morpheme, 27, 37, 236 morpheme boundaries, see boundary morphosyntax, 2–3, 12, 19, 23, 27–28, 32–35, 39, 44, 76 multilingual(ism), 4, 51, 53, 58, 64, 70–72, 75–76, 416 MWC, see Monosyllabic Word Constraint, constraint

nasal(s), 204, 233, 283–284, 295 Necessity, see pragmatic value negation, 28, 43, 356–360, 362, 367–369 – negative constituents, see constituent – negative-initial (element), 13, 355–356, 358, 361, 366–369 – negator (sentential), 359, 362–363, 368 Neogrammarian, 4, 30, 80- 81, 83–87, 91–94, 98–99, 107, 178, 242

Subject index neutralization, 69, 72, 384–385 New Zealand, 38, 41, 456 nominal compound, see compound nominalisation, 426–427, 429–430, 437–438 non-ictus, see ictus non-local antecedent(s), see local antecedent non-standard, 10, 38, 246 non-theta, see theta, 331, 333, 337 Norman French, 327, see also AngloNorman Norman(s), 51, 73, 454 Northumbrian, 285 Norwegian, 183–184, 192, 203

Obligation, see pragmatic value ODEE, 109–111, 117, 119, 122 OED, 5, 15, 32, 110, 113, 122, 344, 404, 423, 432–436, 438, 440–447, 452, 456, 465, 468, 470 OED2, 420–421, 432–433, 435, 445 OEDAS, 421, 432, 435 Old English, 10–11, 13–14, 29, 112, 115, 117, 119, 131, 233, 242, 245–246, 248–250, 259, 264–265, 271, 273–275, 279, 283–289, 297–300, 325–326, 329, 331, 333, 336, 339, 342, 344–346, 348, 353, 355–356, 358–362, 364, 366–371, 373–374, 376–378, 382–383, 385, 387, 392–395, 397, 401, 416, 419, 426–430, 442, 444–445, 449, 460, 463, 466, 468 Old Icelandic, 116, 120 Old Norse, 114 Old Saxon, 116–117, 395, 427 onomatopoeia, 5, 112–115, 117 Open Syllable Lengthening, see lengthening Optimality Theory, 6, 61, 155, 169 optimization, 36, 271

491

Order, see pragmatic value Orm/Ormulum, 131–132 338–339 orthoepist(s), 232–233, 276 orthography, 145, 201, 270, 304, 429, 456 Orwell, George, 456 Oxford English Dictionary, see OED

palatal(ization), 112, 201–203, 289, 292, 294 parameter(s), 19, 24, 44, 67, 238, 349, 397 – parameter resetting, 13, 325, 349–350 – parametric change, 24 participles, 129, 422–434 past perfect, see perfect tense PATTERN GENERALIZATION, 325, 339, 344 pejoration, 28, 30, 32–33 pentameter, 6–7, 125–127, 135–136, 140, 143, 146–148, 153–157, 159–160, 164, 168–173, 179, 181 perfect tense, 13–14, 373, 382–383, 394 – past perfect, 14, 373–374, 383–388, 391–395 – present-inclusive, 379–380, 382 – present perfect, 373–374, 376–378, 382–384, 392, 394, 396 – resultative perfect, 375, 392 – universal perfect(s), 375, 377–378, 380–383, 387, 389–390, 392–393 performance, 140, 144, 177–178 performative, 28, 31 periphery, 86–88, 95–96, 222, see also core Permission, see pragmatic value Personal pronoun subject(s), 361, 363–365 philology, 8, 12, 228, 449, 452, 454 Philological Society, 8, 370, 452, 454, 457–458, 460, 463, 467, 470

492

Subject index

phonology, 2, 8, 150, 174, 182, 203–206, 241–242, 280–281, 295, 353, 454, 470 – phonological boundaries, see boundary – phonological/phonetic reduction, 25–27, 34, 233–234, 240, 290, 297, 325, 344, 359, 368–369 phonotactic constraints, see constraint phrase boundary, see boundary Piers Plowman, 56–57, 62–63, 65–66, 72–75, 77 Pinsky, Robert, 207, 222, 227–228 pleonastic (pronouns), 331, 347 polysemy, 29, 33, 394 pragmatic polysemy, 33 polysyllabic, 140, 166, 170, 172, 211, 433 Pope, 126, 137, 141, 208–209, 211–213, 223–224, 227, 229, 369 possessors, 330, 337, 340, 345 pragmatic(s), 3–4, 12, 19, 21–22, 24, 28–29, 31, 33–36, 39–40, 42–43, 47–49, 52, 61, 63–66, 71, 74, 87, 301–305, 311, 316, 318, 321–323, 375, 396, 399–400, 413–414 – historical pragmatics, 3, 19, 22, 29 – pragmatic-semantic, 19, 24, 28, 34–35 pragmatic value, 304 – Advice, 310, 313 – Assurance, 318–319 – deontic, 12, 31, 39, 304–305 308–309, 311–312, 315, 321–322 – dynamic, 12, 304–305, 316, 321–322 – existential, 14, 374–379, 380, 382–383, 387–388, 392–393 – Intention, 307–308 – Necessity, 306, 321 – Obligation, 306, 310- 311, 321 – Order, see pragmatic value, 310–311

– Permission, 306, 315 – Prediction, 306, 316, 318 – Prohibition, 310, 312 – Promise, 307, 309 – Prophecy, 318–319 – Proposal, 310, 313 – Threat, 307–308 pre-Christian, see Christian Prediction, see pragmatic value preposition(s), 13, 62–63, 72, 130–132, 171, 325, 328, 345–346, 419 present perfect, see perfect tense present-inclusive, see perfect tense Principle of Closure, 248, 255 proclisis, see clitic Pro-Drop, 341, 349 Prohibition, see pragmatic value Promise, see pragmatic value pronominal (subject/object), 341, 345, 347–348, 350, 361–365, 370, 403, 406–407, 409, 414 Prophecy, see pragmatic value Proposal, see pragmatic value prosodic/prosody, 6, 125, 127, 131–132, 134–135, 146–151, 173–175, 179, 182, 193–194, 225–227, 360, 366 – prosodic constituent, see constituent – prosodic word, 135 pseudo-compound(s), see compound punctuation, 141, 262, 402, 408–409

quatrain, 174, 216

reanalysis, 22, 25, 32, 34, 45, 420–422, 424–427, 431, 443 Received Pronunciation/RP, 198–199, 457–458 reflexive(s) pronouns, 12–13, 325, 337, 348, 350–352

Subject index register(s), 37, 55, 216, 301, 399, 401–402, 404, 413 reported speech, 14, 400–401, 405, 408–409, 411 resolution, see metrical resolution resultative, 14, 373–380, 382–383, 385, 387–388, 391–393 resultative perfect, see perfect tense retro?ex(ion), 8–9, 183–185, 188–198, 200–203 reversed foot, see foot rhotic/rhoticity, 183, 185, 188–189, 192, 196–197, 201–203 rhyme(s), 9, 113, 154, 160, 166–167, 207–225, 228–229 – eye rhyme, 155, 210, 217–220 – full rhyme, 9, 210, 212–214, 219, 224 – near rhyme, 155 – off rhyme, 155, see also slant rhyme – partial rhyme, 209–215, 220, 223, 228 – perfect rhymes, 222 – rhyme riche, 155 – rock rhyme, 221 – sight-rhyme, see eye-rhyme – slant rhyme, 155, 210, 222 – traditional rhyme, 218, 220–221, 224 rhythm, 7, 125, 127, 174, 178, 182 Rhythm Rule, 130 rhythmical prose, 262 Romance languages, 85, 106, 120, 466 root clauses, 13, 356–359, 363–364, 367–368 RP, see Received Pronunciation R-Vocalization, 197–199

Saxon(s), 331, 369, 395, 454, 460, 462–464, 467, 469 Scandinavian, 11, 114, 119, 203–204, 271–272, 274, 279, 327, 368

493

scansion, 10, 129–130, 157–158, 166, 247–248, 251, 257, 259, 263–264 schwa, 100, 132, 171, 197–199, 235 schwa loss, 235, see also -e scientific texts, see text types Scots (English), 38, 65, 74, 188, 195, 279, 458, 460 scribe/sribal, 51, 55–56, 404–405, 407–412 S-curve, 5, 103, 346 Second Consonant Shift, 112 secondary grammaticalization, see grammaticalization Semantic(s), 2, 12, 40, 42, 45, 396, 445 – semantic change, 22, 25–26, 28–33, 37, 39, 42, 45–46, 48–49 – amelioration, 28, 32 – bleaching/bleached, 25, 27, 348 – specialization, 28, 32 – semantic constraints, see constraint – semantic meanings, 375–376 – semantic shift, 421 – semantic targets, 91, 96 – semantic universals, 328 sentence boundaries, see boundary sermon(s), see text types Shakespeare, 6–7, 9, 118, 126, 130, 134, 138–140, 142, 144, 154, 156–157, 159, 162–164, 169–171, 173, 177, 179–180, 207–209, 211–212, 215–218, 220–224, 226–227, 229, 232, 239, 242, 349, 461 – Sonnets, 9, 162, 208–209, 215–217, 219–220, 223 – The Merchant of Venice, 136 Shaw, 460 Shelley, 134–136 short diphthongs, see diphthong short vowel, see vowel shortening, 120, 231–237, 239, 242, 278 Sidney, 134, 147, 214, 446

494

Subject index

Sir Gawain and the Green Knight, 126 Skelton, 57, 73, 349 slander depositions, see text types SMP, see stress, Stress Maximum Principle sociolinguistic(s), 4, 37, 52–53, 61, 66, 71, 78, 104, 108, 399 sociolinguistic factors, 66 sociolinguistic variables, 104, 399 sociophysical meanings, 33 Sonnets, see Shakespeare sonorant, 190–191, 196, 294–297 sound change, 4–5, 9–11, 81, 83, 85, 88, 92–94, 96–100, 102, 104–105, 195, 231, 233, 238–242, 268, 274, 283, 296, 298, 455, 458 sound symbolism, 112 South African English, 199, 202 Specialization, see semantic change speech act/event, 31–32, 301–302, 311, 314, 320, 415 speech communities, 86, 93–94, 104, 400 speech-based texts, see text types spelling reform, 454–456, 459 spelling variation, 327, see also variant spellings Spenser, 134, 147, 214, 221, 227 Standard English, 72, 96, 106, 111, 114, 277, 325–326, 328–329, 336–338, 340, 343–347, 350, 450–451, 470 standardization, 38 stress – alternating stress, 126, 134, 173 – Compound Stress Rule, 247 – stress "doublets", 167 – Stress Maximum Principle, 134–136, 155, 162–163, 166–167, 172 – stress maximum/minimum, 134–136, 155, 162

– stress shift, 126, 128, 130–132, 141, 180–181, 234, 242 – stressed syllable, see syllable – stressed vowel, see vowel – stressers, 125, 145, 177, see also timers – stress-promotion convention, 167 – subordinate stress, 258, 264 – variable stress, 127, 147, 177 subjectification, 28, 32–33, 39, 45, 48 subordinate stress, see stress sulcal(ity), 9, 185, 189–192, 198–199 supra-regional code, see code Swedish, 183–184, 192–193, 197, 199, 203 Swinburne, 156, 227 syllable, 7, 10, 126–136, 139–140, 142–146, 154, 159–160, 170–172, 179, 181, 183, 193–195, 197, 200–202, 219, 232–235, 241, 249, 264, 271, 287, 290, 294, 299, 344, 369 – stressed syllable, 156, 172, 180, 272 – syllable coda, 183, 185–186, 191, 193, 195, 197, 202–203, 272–273 – unstressed syllable(s) 7, 135–136, 144, 159–160, 162, 171–172, 178–179, 325, 457 synaesthesia, 30 syncopation, 129, 132, 134, 143, 167, 170 synonyms, 111, 112, 118, 329, 347 synonymy, see ANTI-SYNONYMY syntax, 3, 22, 24, 27, 133, 141, 149–150, 165, 174, 181, 255, 258, 352, 361, 397, 407, 415, 431, 442, 451 – archaic syntax, 255, 258 – syntactic boundary, see boundary – syntactic break, 139–141, 145–146 – syntactic constraints, see constraint

Subject index

495

– syntactic inversion, 161–162, 165, 170, 361 synthetic compound, see compound

trochaic, 7, 155, 160, 164, 166, 168–170, 247–248, 252, 262 trochee(s), see foot/feet

tap(ping), 143, 177, 183, 185–192, 195, 201, 203 – dental tap, 201 – Tap Retroflexion, 190–191 telic interpretation, 331, 394 tension, 6–7, 136, 145, 155, 178, 181 tetrameter, 125, 155, 159 text types, 4, 12, 37, 55–56, 58, 62, 63, 65–66, 70–71, 301, 304–305, 309, 312, 314–316, 319, 321–322, 399 – deposition, 55, 400, 402, 406, 409 – legal English, 404 – mixed-language texts, 4, 51–52, 54–56, 58, 60, 67, 70–72 – scientific texts, 65, 72 – sermon(s), 54–56, 58–60, 62–63, 65–66, 68, 70, 72, 74, 301, 304–305, 313–314, 317, 319, 322, 413 – slander depositions, 14, 399, 402, 412 – speech-based texts, 301, 399, 401 theta (role), 331–334, 337–338, 340–342, 346–348 – non-theta, 331, 333, 337 Threat, see pragmatic value tilt(ing), 7, 126–127, 146, 160, 167–169, 172, 177 timers, 125, 145–146, 177, see also stress, stressers Topic Drop, 341 topicalization transformation, 165 Toronto text corpus, see corpus trigger (experience), 20, 23, 36, 366 trilingual, 58, 76 trill, 9, 183–190, 192, 195, 202 Trill Lenition, 187–188 trimeter, 126

umlaut, 292–293, 295 unidirectionality, 3, 21, 23, 25–28, 30–32, 36 uniformity, 34 universal perfect(s), see perfect tense unstressed syllables, see syllable unstressed vowel, see vowel uvular r, 183–184, 204

V1-ing compounding, see compound variable stress, see stress variables, 53–54, 59, 61, 64, 66, 104, 165 sociolinguistic variables, 37–38, 104, 399, 466 variable vowel loss, 427, 429 variant(s), 9, 87, 91, 96, 100, 155, 158, 167, 219, 221, 246, 248–250, 252–256, 263–264, 268–269, 283, 322 – graphic variants, 304 – variant spellings, 217, 454–455 variation, 5, 7, 9, 37, 46, 61, 69, 79–80, 85–86, 93–94, 96, 104–106, 128, 187, 192, 195, 197, 209–211, 213, 215, 218–219, 221–222, 235, 303, 353, 429–431 verb placement, 13, 142, 161, 355, 360–361, 366 – sentence-internal, 362–366 – variable, 363–364, 366 – verb-final, 355, 360, 366–367 – verb-fronting, 356, 360, 366–367 – verb-movement, 358–359 – verb-second, 361 verb-verb compound, see compound

496

Subject index

vernacular(ization), 53, 70, 76–77, 84, 323, 463 verse norms, 247, 253 VN(s), see compound vocalization, 197–199, 271, 273 vowel, 9–11, 96, 98–100, 149, 189, 197–199, 208–214, 217–221, 224, 231–237, 239–240, 242, 264, 267, 269–272, 274, 279, 284–285, 287, 291, 293–294, 297, 358–359, 368–369, 428, 430, 450, 453, 457, 467 – back vowel, 11, 237, 271, 279, 285, 287–290, 294–296, 298 – backness, 100–101, 288 – central vowel, 237, 283, 457 – front vowel(s), 237, 271, 283, 285–287, 290–292, 294, 298 – frontness, 100–101, 288 – high vowel, 267 – link-vowel, 427, 429–430 – long vowel(s), 10, 218, 235, 272, 275, 278, 284, 288–298 – loss of (unstressed) vowels, 283, 427 – lowering, 199, 267–268, 270–271, 278, 459 – short vowel, 218, 235, 271–272, 275, 278, 294 – stressed vowel, 288, 291 – unstressed vowel, 233, 288, 290, 293 – vowel alternations, 209, 221, 224 – vowel height, 99, 100 – vowel quality, 231–232

– vowel reduction, 233, 325, 369 – vowel variation, 207, 212 vowel shift, 11, 108, 243, 267–268, 270, 281, 458 – Great Vowel Shift, 239, 267, 280–281, 459 VV, see compound Welsh English, 198–199 West Germanic, see Germanic West Saxon, 283, 288, 294, 395 word boundary, see boundary word formation, 113, 117, 417–418, 422, 433, 438, 442–443, see also compound – back-formation, 418, 420, 425–426 – base derivation, 442 – direct formation, 418, 442 – endocentric, 15, 417–419, 433, 437 word frequency, 231–233 word order, 13, 23, 28, 42, 161, 166, 253–254, 320, 355–356, 358–361, 363–364, 366–367 word-foot theory, see foot Wordsworth, 126 Wycliff(ites), 343, 456 Yiddish, 120, 183, 205 Zero, 22 – Zero-derivation, 419–420, 431 – Zero-nominal, 424–425, 428, 431, 433, 435

Topics in English Linguistics Edited by Bernd Kortmann and Elizabeth Closs Traugott Mouton de Gruyter · Berlin · New York 1 Niels Davidsen-Nielsen, Tense and Mood in English. A Comparison with Danish. 1990. 2 Historical English Syntax. Edited by Dieter Kastovsky. 1991. 3 English Computer Corpora. Selected Papers and Research Guide. Edited by Stig Johansson and Anna-Brita Stenström. 1991. 4 Donka Minkova, The History of Final Vowels in English. The Sound of Muting. 1991. 5 Lia Korrel, Duration in English. A Basic Choice, Illustrated in Comparison with Dutch. 1991. 6 Andreas H. Jucker, Social Stylistics. Syntactic Variation in British Newspapers. 1992. 7 Ken-ichi Takami, Preposition Stranding. From Syntactic to Functional Analyses. 1992. 8 Bas Aarts, Small Clauses in English. The Nonverbal Types. 1992. 9 New Directions in English Language Corpora. Methodology, Results, Software Developments. Edited by Gerhard Leitner. 1992. 10 History of Englishes. New Methods and Interpretations in Historical Linguistics. Edited by Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen. 1992. 11 Early English in the Computer Age. Explorations through the Helsinki Corpus. Edited by Matti Rissanen, Merja Kytö and Minna Palander-Collin. 1993. 12 Towards a Standard English: 1600–1800. Edited by Dieter Stein and Ingrid Tieken-Boon van Ostade. 1993. 13 Studies in Early Modern English. Edited by Dieter Kastovsky. 1994. 14 Ronald Geluykens, The Pragmatics of Discourse Anaphora in English. Evidence from Conversational Repair. 1994. 15 Traute Ewers, The Origin of American Black English. Be-Forms in the HOODOO Texts. 1996. 16 Ilse Depraetere, The Tense System in English Relative Clauses. A CorpusBased Analysis. 1996. 17 Michiko Ogura, Verbs in Medieval English. Differences in Verb Choice in Verse and Prose. 1996. 18 Spanish Loanwords in the English Language. A Tendency towards Hegemony Reversal. Edited by Félix Rodríguez Gonzáles. 1996. 19 Laurel J. Brinton, Pragmatic Markers in English. Grammaticalization and Discourse Functions. 1996.

20 Christiane Dalton-Puffer, The French Influence on Middle English Morphology. A Corpus-Based Study on Derivation. 1996. 21 Johan Elsness, The Perfect and the Preterite in Contemporary and Earlier English. 1997. 22 Carl Bache and Niels Davidsen-Nielsen, Mastering English. An Advanced Grammar for Non-native and Native Speakers. 1997. 23 English in Transition. Corpus-based Studies in Linguistic Variation and Genre Styles. Edited by Matti Rissanen, Merja Kytö and Kirsi Heikkonen. 1997. 24 Grammaticalization at Work. Studies of Long-term Developments in English. Edited by Matti Rissanen, Merja Kytö and Kirsi Heikkonen. 1997. 25 Axel Hübler, The Expressivity of Grammar. Grammatical Devices Expressing Emotion across Time. 1998. 26 Negation in the History of English. Edited by Ingrid Tieken-Boon van Ostade, Gunnel Tottie and Wim van der Wurff. 1998. 27 Martina Häcker, Adverbial Clauses in Scots: A Semantic-Syntactic Study. 1998. 28 Ingo Plag, Morphological Productivity. Structural Constraints in English Derivation. 1999. 29 Gustav Muthmann, Reverse English Dictionary. Based on Phonological and Morphological Principles. 1999. 30 Metaphor and Metonymy at the Crossroads. A Cognitive Perspective. Edited by Antonio Barcelona. 2000. 31 Generative Theory and Corpus Studies. A Dialogue from 10 ICEHL. Edited by Ricardo Bermúdez-Otero, David Denison, Richard M. Hogg and C. B. McCully. 2000. 32 Manfred G. Krug, Emerging English Modals. A Corpus-Based Study of Grammaticalization. 2000. 33 Cause – Condition – Concession – Contrast. Cognitive and Discourse Perspectives. Edited by Elizabeth Couper-Kuhlen and Bernd Kortmann. 2000. 34 Hans-Jörg Schmid, English Abstract Nouns as Conceptual Shells. From Corpus to Cognition. 2000. 35 Placing Middle English in Context. Edited by Irma Taavitsainen, Terttu Nevalainen, Päivi Pahta and Matti Rissanen. 2000. 36 Michael G. Getty, A Constraint-based Approach to the Metre of Beowulf. Forthcoming. 37 Renaat Declerck and Susan Reed, Conditionals. A Comprehensive Empirical Analysis. 2001. 38 Alexander Kautzsch, The Historical Evolution of Earlier African American English. An Empirical Comparison of Early Sources. 2002.