Studies in the History of the English Language II: Unfolding Conversations 9783110897661, 9783110180978

Studies in the History of the English Language II: Unfolding Conversations contains selected papers from the SHEL-2 conf

381 17 16MB

English Pages 512 Year 2004

Report DMCA / Copyright


Recommend Papers

Studies in the History of the English Language II: Unfolding Conversations
 9783110897661, 9783110180978

  • 0 0 1
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Studies in the History of the English Language II


Topics in English Linguistics 45


Elizabeth Closs Traugott Bernd Kortmann

Mouton de Gruyter Berlin · New York

Studies in the History of the English Language II Unfolding Conversations

Edited by

Anne Curzan Kimberly Emmons

Mouton de Gruyter Berlin · New York

M o u t o n de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter G m b H & Co. K G , Berlin.

® Printed on acid-free paper which falls within the guidelines of the A N S I to ensure permanence and durability.

ISBN 3-11-018097-9 Bibliographic information published by Die Deutsche


Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at < h t t p : / / d n b . d d b . d e > .

© Copyright 2004 by Walter de Gruyter G m b H & Co. K G , 10785 Berlin All rights reserved, including those of translation into foreign languages. N o part of this book may be reproduced or transmitted in any f o r m 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 Section 1: Linguistics and philology Introduction: Linguistics and philology Anne Curzan and Kimberly Emmons Philology, linguistics, and the history of [hw]~[w] Donka Minkova An essay in historical sociolinguistics?: On Donka Minkova's "Philology, linguistics, and the history of [hw]~[w]" Lesley Milroy A brief response Donka Minkova Why we should not believe in short diphthongs David L. White Extended forms {Streckformen) in English Anatoly Liberman Linguistic change in words one owns: How trademarks become "generic" Ronald R. Butters and Jennifer Westerhaus Section 2: Corpus- and text-based studies Introduction: Corpus- and text-based studies Anne Curzan and Kimberly Emmons The meanings and uses of the progressive construction in an early eighteenth-century English network Susan M. Fitzmaurice


Table of contents

Investigating the expressive progressive: On Susan M. Fitzmaurice's "The meanings and uses of the progressive construction in an early eighteenth-century English network" Erik Smitterberg


A brief response Susan M. Fitzmaurice


Modal use across registers and time Douglas Biber The need for good texts: The case of Henry Machyn's Day Book, 1550-1563 Richard W. Bailey



The perils of firsts: Dating Rawlinson MS Poet. 108 and tracing the development of monolingual English lexicons Ian Lancashire


Section 3: Constraint-based studies Introduction: Constraint-based studies Anne Curzan and Kimberly Emmons


The evolution of Middle English alliterative meter Geoffrey Russom Old English poetry and the alliterative revival: On Geoffrey Russom's "The evolution of Middle English alliterative meter" Robert D. Fulk



A brief response Geoffrey Russom


A central metrical prototype for English iambic tetrameter verse: Evidence from Chaucer's octosyllabic lines Xingzhong Li


Table of contents vii Early English clause structure change in a stochastic optimality theory setting Brady Z. Clark


The role of perceptual contrast in Verner's Law OlgaPetrova


Section 4: Dialectology Introduction: Dialectology Anne Curzan and Kimberly Emmons


Historical perspectives on the pen/pin merger in Southern American English Michael Montgomery and Connie Eble


Digging up the roots of Southern American English: On Michael Montgomery and Connie Eble's "Historical perspectives on the pen/pin merger in Southern American English" Guy Bailey


A brief response Michael Montgomery and Connie Eble


Vowel merger in west central Indiana: A naughty, knotty project Betty S. Phillips


The spread of negative contraction in early English Richard M. Hogg


Name index


Subject index



The second biennial meeting of the Studies in the History of the English Language conference (abbreviated SHEL-2) was held at the University of Washington in Seattle in March of 2002. The conference series, which began at UCLA in 2000, originated in a desire to provide focus to and stimulate research in the field of historical English linguistics in North America. The papers in this volume, selected from the thirty papers presented at the conference, are a testament to the exciting and innovative research on the history of the English language happening in North America as well as the fascinating and productive conversations taking place among scholars in North America and in Europe. This volume is structured, in fact, around the theme of conversation. As the history of English unfolds all around us in the dialects of English in North America and in Britain, as well as in the distinctive varieties of World English around the globe, the tradition of scholarly conversation about these linguistic developments continues among scholars past and present. New resources such as electronic corpora and recent theoretical models such as optimality theory change some of the terms of the discussion and open rich new domains for historical research and critical analysis. At the same time, the goals at the core of historical English linguistics remain constant, and modern scholars revisit long-standing questions about the development of the language with new data and fresh perspectives. This volume witnesses conversations between new theories/methods and traditional fields such as phonology and syntax. It is also a conversation between the present and the past. As William Labov's uniformitarian principle articulates, our understanding of the mechanisms of current language change can critically inform our analysis of past language changes, on the assumption that historical forces of language change are the same or operate in similar ways - to present forces of change. Donka Minkova and Lesley Milroy propose explanations for historical variability in initial [h] within this framework; Richard M. Hogg examines patterns of dialect variation for negative contraction in medieval English; Susan M. Fitzmaurice and Erik Smitterberg reconstruct possible social networks affecting the spread of progressive constructions. Betty Phillips' study of current dialect variation confirms a long-standing hypothesis about females being innovators - a finding similarly suggested in Fitzmaurice's study of language



change some two centuries earlier. Moving from the past toward the present, Geoffrey Russom and Olga Petrova demonstrate the ways in which an understanding of earlier periods - of Old English alliterative meter or Proto-Germanic Verner's Law - can inform our understanding of later developments. Throughout the volume, scholars are negotiating the relationship between philology and linguistics, the complications of which are the more explicit focus of the first section of this volume. This volume is arranged into four sections: Philology and linguistics, Text- and corpus-based studies, Constraint-based studies, and Dialectology. In the spirit of conversation, we have identified key articles to lead off each section and invited preeminent scholars in each subfield to respond to these articles. The lead authors then agreed to provide brief remarks as a means of pointing toward future inquiry. These conversations are, we believe, a productive feature of this volume; and, although they could have continued through several more exchanges, the limitations of our print media necessarily leave them unfinished. Each section of this volume includes a separate introduction, in which we have identified points of intersection among the articles contained in the section. Throughout this volume, we see an ongoing conversation at the heart of historical English linguistics: the question of evidence and historical reconstruction. Robert Fulk puts it eloquently in his discussion of the oral nature of early English vernacular texts and the possibility, if not necessity, of creating linguistic arguments based on unavailable evidence; "it raises," he concludes, "profound questions about explanation in linguistics, most particularly whether the aim of historical linguistics should be to explain the data available or to analyze texts of earlier periods from a realistic historical perspective - that is, whether the primary allegiance of historical linguistics should be to linguistics or to history." Richard Hogg, after describing the paucity of data for contracted forms, notes that "[s]ome locations may accidentally not furnish the necessary material." Donka Minkova, recognizing the limits of what the written record can tell modern linguists about the spoken language, focuses specifically on alliteration, as a kind of textual evidence that may be able to speak beyond the written. Richard W. Bailey and Ian Lancashire speak directly to the question of how texts - the "data" from which historians of the language work - are made publicly accessible and analyzed within the historical linguistic tradition. Other contributors discuss newly available records for analysis, from the nineteenth-century letters described by Michael Montgomery and Connie Eble to the corpus of eighteenth-century documents described by Fitz-



maurice. In addition to these new resources, some of the same central historical texts resurface in multiple articles, most obviously resources such as the Oxford English Dictionary and the Middle English Dictionary, in addition to the Linguistic Atlas of Late Mediaeval English, and central literary texts such as Chaucer's, Sir Gawain, and, as it happens, Henry Machyn's sixteenth-century Day Book. The metaphor of folding involves bringing together the farthest points as part of a coherent whole. This volume spans topics and time periods from Proto-Germanic sound change to twenty-first century dialect variation, and methodologies from painstaking philological work with written texts to high-speed data gathering in computerized corpora. It is the richness of the intersections among these studies and approaches that makes history of English such an exciting field of study. The SHEL-2 conference would not have been possible without the generous support of several sponsors at the University of Washington, in particular the Walter Chapin Simpson Center for the Humanities, the Department of English, the Department of Linguistics, the Graduate School, and the Office of the Provost. We would like to thank all of the more than 80 participants in the conference for the lively conversation they generated at the panels. We have many people to thank for their support of this volume. In particular, we would like to thank the series editors, Bernd Kormann and Elizabeth Traugott, for their ongoing interest in publishing these volumes of selected papers from the SHEL conferences; we are particularly grateful to Elizabeth Traugott for her detailed feedback and advice at every step of the process. Our thanks also to Birgit Sievert, our editor at Mouton de Gruyter, for all her energy and work in making this volume happen, and to Rizwan Ahmad, for his work on the index. The papers in this volume all benefited enormously from the comments by outside anonymous reviewers, and we can only begin to repay them by listing them here. We would like to thank: Guy Bailey, University of Texas at San Antonio Rusty Barrett, University of Michigan Laurel Brinton, University of British Columbia Derek Britton, University of Edinburgh Gerald Cohen, University of Missouri-Rolla Robert Fulk, University of Indiana Matthew Gordon, University of Missouri Michael Hammond, University of Arizona

xii Foreword Kirk Hazen, West Virginia University Yoko Iyeiri, Kyoto University Robert Kirchner, University of Alberta Manfred Krug, University of Freiburg Robert Kyes, University of Michigan Robert Lewis, University of Michigan Christopher McCully, University of Manchester Erin McKean, Verbatim: The Language Quarterly Frances McSparran, University of Michigan Barbra Meek, University of Michigan Charles Meyer, University of Massachusetts-Boston Lesley Milroy, University of Michigan John Myhill, University of Haifa Chris Palmer, University of Michigan Joseph Pickett, Houghton Mifflin Robin Queen, University of Michigan Erik Smitterberg, Stockholm University Gail Stygall, University of Washington Sali Tagliamonte, University of Toronto Erik Thomas, North Carolina State University Thomas Toon, University of Michigan Laura Wright, Cambridge University Richard Wright, University of Washington All of the contributing authors to this volume have made it a pleasure to edit. We appreciate the intellectual energy that has gone into the writing and revising of these papers for and since the conference, and we have enjoyed participating in the conversations between authors and reviewers, as well as between authors and respondents. We look forward to continuing the conversations captured here at future SHEL conferences. Anne Curzan Kimberly Emmons

Section 1 Linguistics and philology

Introduction: Linguistics and philology Anne Curzan and Kimberly Emmons

The note below the modern definition of philology in the Oxford English Dictionary captures much of the tension between "philology" and "linguistics" that the authors in this section address in various ways. The third definition of philology in the OED reads: "The study of the structure and development of language; the science of language; linguistics. Now usu. restricted to the study of the development of specific languages or language families, esp. research into phonological and morphological history based on written documents." The editors then note: "This sense has never been current in the U.S. Linguistics is now the more usual term for the study of the structure of language, and, with qualifying adjective or adjective phrase, is replacing philology even in the restricted sense." Philology has often been marginalized as the close study of the language of texts for the purpose of etymological, comparative, or stylistic research, isolated from current linguistic theory. Donka Minkova, in "Philology, linguistics, and the history of [hw]~[w]," argues that combining the methodologies, theories, and insights of philology and linguistics is "both possible and desirable." Recognizing that the definitions of both terms vary, Minkova notes the consistent association of philology with the study of written texts and modern linguistics with the study of speech. Any rigid distinction between gathering and recording data (philology) and rigorous theoretical work (linguistics), she asserts, is not a productive one. Philology without theory provides little illumination, and linguistic theory must account for all the available data. Any such rigid distinction also often breaks down in practice. English historical linguistics employs rigorous analysis of carefully collected and categorized data. And as Lesley Milroy points out in response to Minkova's article, sociolinguistics, including the relatively new field of historical sociolinguistics, employs rigorous data collection. Minkova also calls into question the clear distinction between the study of written and oral texts when working with medieval texts. Given the orality of medieval literary culture, written texts can provide modern language scholars with insights about the intuitions and speech of "native


Anne Curzan and Kimberly Emmons

speakers" of early varieties of the language. Specifically, alliteration can provide critical evidence about phonological developments. As a case study, Minkova draws on her expertise in medieval English alliteration, phonological theory, and historical sociolinguistics to examine the development of the [hw]~[w] merger in English. The evidence she provides about early spelling confusion dates the beginning of the merger back into the Old English period. The progression of the merger, however, was not linear, as the distinction was reintroduced in subsequent centuries, probably for social reasons. Minkova's complex and original analysis of the available textual data draws on theoretical phonology and the effects of sonority hierarchies as well as on sociolinguistic theories for social triggers that could have influenced the reappearance of the innovative /m/. In response, Lesley Milroy draws striking parallels between the development of [hw]~[w] as described by Minkova and data on other reported mergers from sociolinguistic research. Importantly, she notes that Minkova's data, which suggests that the merger was variable, could be interpreted to mean that not all speakers underwent the merger in a given dialect area: commentators' reports of the merger may not capture the casual pronunciation of many speakers. Milroy also correlates Minkova's suggestion of the adoption of a northern form in the sixteenth and seventeenth centuries to morphosyntactic changes that followed a similar patterns, all of which may be related to demographic shifts in the period. Alliteration may provide some of the more easily interpretable phonological data for historical sociolinguists. In "Why we should not believe in short diphthongs," David L. White advises scholarly caution in assuming clear correspondences between written forms (specifically word spellings) and phonetic and phonological forms. White reopens a longstanding scholarly conversation about the existence of short diphthongs. Bringing together phonetic and phonological theory, White presents theoretical reasons to believe short diphthongs do not exist. White then uses a re-examination of the published data on short diphthongs, ranging over languages from Old Irish through Icelandic and Afrikaans to Ancient Greek, to argue that without convincing evidence of short diphthongs from other languages, living or dead, there is little justification for assuming short diphthongs in Old English. The graphic short diphthongs in Old English represent, instead, velarization. Without doubt, the conversation will continue with White's call for evidence of short diphthongs in living languages and the questions he raises about how best to interpret spelling evidence.

Section 1: Introduction


The subsequent two articles focus on etymology and semantic change, traditional domains of philological research. Both articles demonstrate how a focus on spoken language and extralinguistic factors adds depth to the analysis of philological data. In "Extended forms (Streckformen) in English," Anatoly Liberman's detailed analysis of extended forms in English supports his introductory assertion that philologists know much about phonetics, morphology, and syntax. Infixation, a minor word formation process in English, blurs traditional lines between morphology and syntax, etymology and word formation. In pulling together evidence that has typically been marginalized as "arbitrary" and creating connections across centuries among similar types of words, Liberman finds patterns to the form of compounds with -a- and -de-. These extended forms are typically playful and slangy, exactly the kind of spoken and casual language that philologists and linguists cannot afford to ignore. Ronald R. Butters and Jennifer Westerhaus, in "Linguistic change in words one owns," examine an often overlooked yet rich field of semantic change, that of trademarks becoming generic. This process, sometimes known as genericide, can involve modern advertising and the courts directly in semantic change. When the courts decide what the public understands a word to mean, they affect the marketing that helps shape the public's understanding of a word's meaning. Speakers themselves become central in this process as the test for genericness must be what speakers believe a given word to mean. Lexicographic and philological evidence cannot be abstracted away from actual speakers, and social factors such as advertising and governmental intervention should be seen as natural and important influences on the semantic changes that this subset of words continues to undergo in the history of English. As all of these articles demonstrate, "philology" and "linguistics," by any traditional definition, are richly intertwined in English historical linguistics, as researchers interpret the rich array of available textual evidence for what it reveals about linguistic theory and about the development of the spoken and written language.

Philology, linguistics, and the history of [ h w ^ w ] 1 Donka Minkova

Boswell: 'Did you find, Sir, his conversation to be of a superiour style?' Johnson: 'Sir, in the conversation which I had with him I had the best right to superiority, for it was upon philology and literature.' James Boswell, Life of Johnson

1. Overview The integration of philology and linguistics is a central topic in the continuously "unfolding conversations" in English historical studies. The first part of my paper addresses the tension between the disciplines of philology and linguistics and argues against a strict division of labor as advocated by some researchers. A focus on the "orality" of medieval literary culture provides a new angle on the debate; for the historical linguist, "orality" is the valuable philological link to the "native speaker's" intuitions of authors, scribes, and audiences. Phonological reconstruction should therefore be equally well informed by linguistic theory and by the largest available set of textual data. The second part of the study explores the consequences of recognizing the speaker's "voice" in reconstructing the linguistic properties and the regional, social, and more recently, age- and gender-based opposition between aspirated and non-aspirated reflexes of Old English . The topic reinforces the need for a renewed defense of the inseparability of philology and linguistics.

2. The mongrel linguistic philologist, or how to hold with the hare and hunt with the hound Were we to draw a metaphoric animal life map of our profession, the collective philologist would be of the Lepus family, a hare, a creature familiar with the secret warrens and burrows of medieval textual territory. The hare/philologist knows the old caves, coves, nooks and crannies of the


Donka Minkova

English language, has keen hearing, makes noise only when frightened or injured, and provides food and fur/texts and data for "higher order" mammals. The linguist is more like a member of the Canis family, a hound, fleet-footed, intent on new scents and sights, but not at home in the rough terrain and the arcana hiding in the labyrinths of philology. Each group has bred specimens of admirable purity and sterling quality and each group has distrusted the other and proclaimed its own achievements. In 1987 the late Cecily Clark, editor of the Peterborough Chronicle and a most distinguished onomastician, opened her presentation at ICEHL 5 with a feisty "I speak as a philologist." Her defensiveness was not unusual: in the latter half of the twentieth century the prestige of philology declined rapidly, triggering reappraisals of the state of the discipline. The clever self-deprecatory title of Winters and Nathan's (1992) paper - "First He Called Her a Philologist, and Then She Insulted Him" - reflects that same anxiety. The emphasis on linguistic modeling created a sense that philology was an old-fashioned enterprise and that its best days might be over. In 1982, surveying the philology/linguistics controversy, Koerner (1982: 404) wrote, sarcastically, that "The battle had been won in favor of 'linguistics' as the truly scientific discipline of the two, and only weaker minds could engage in the other field." Cecily Clark wanted to distance herself from linguistics. She described her kind of work as "linguistic archeology," "a branch of history." She wrote: (1) If one sees life as a ... seamless fabric in which language is woven together with politics, religion, economic developments and sociocultural relationships, then all linguistic manifestations are ... capable of illuminating these other spheres, in the same measure as language is enriched, impoverished, reshaped by the contexts in which it is used. (Clark 1987: 65) Nobody would then or now object to Clark's appeal to position the study of place names in its appropriate historical context, yet her implicit polarization of philology vs. (formal) linguistics rankles: a brilliant practitioner of the former, she apparently felt that the latter was of no interest. This attitude may be easier to understand within onomastics, but when it covers areas of intense linguistic concern, such as syntax, the perceived gap is more puzzling. Many will remember the flurry of uncomplimentary exchanges following the publication of Bruce Mitchell's monumental Old English Syntax in 1985. In response to the accusations by

Philology, linguistics, and the history of [hw]~[w] the reviewers that his approach was "taxonomic," Mitchell (1992: 97) wrote: "It will come as little surprise to at least some of my audience if I say that in my opinion modern linguistic techniques have so far done little to advance our knowledge of Old English syntax." Inattention to the full range of available data in the early generative work did result in abstract linguistic accounts unacceptable to the well-informed philologist. Yet vigorous and illuminating theoretical work revealing the properties of English historical syntax has repeatedly refuted Mitchell's condemnation. Still, if a scholar of Mitchell's stature could be so negative about the discovery potential of linguistics, the rift between the two disciplines needs to be addressed. In the United States it was the birth of the Linguistic Society of America in 1925 and the rise of structuralism that precipitated the confrontation between philologists and linguists.2 The priority of linguistics appeared self-evident to some, as is clear from Whorf s statement that "at the base of philology we must have linguistics."3 With the explosion of generative linguistics, the theoretical study of language drifted away from "the surface." Rules ruled, and modeling language in an elegant, economic, and coherent manner became a lofty intellectual goal while the parallel activity of data gathering and recording took a back seat as a more pedestrian enterprise. The breach was noticed early, and many outstanding scholars on both sides of the Atlantic set out to "heal" it - as Roger Lass described his goal in the preface to his famous (1969) Anthology. The relationship between philology and linguistics became the theme of important studies and whole volumes (see Koerner 1982; Fisiak 1990; Hogg 1994). All of them recommended a close partnership between philology and linguistics, and argued against the absurdity of philology without theory or theory without data, the senselessness of trying to separate the chicken from the egg, to use Hogg's metaphor. With the advent of "organized" historical linguistics on the international scene in the 1970s and the very successful conference series such as ICHL, and, for us, ICEHL and SHEL, the "partnership" appeared well established. The conference proceedings, the impressive Cambridge History of the English Language, dedicated publishers' series and specialized journals bear witness to the vitality of a world-wide research program combining philological knowledge and linguistic thinking. The divide, though not chasmic, perseveres, however. As recently as 1998, Werner (1998: 164— 165, 175-176) states that philology and linguistics "are not even as close as physics and chemistry but rather diverge like physics and history," arguing that they pursue different objectives, have different contents, apply



Donka Minkova

different methods, and employ different procedures of concept formation. He further asserts that:4 (2) Linguistics ... is a fully autonomous discipline, and its research is neither influenced nor controlled by philological considerations. (Werner 1998: 164-165) I will take this statement as a rhetorical target that the paper argues against.

3. The many faces and phases of philology As a brief look at the history of the branch of knowledge known as philology reveals, the discipline hasn't always been the hunted party, nor was it always defined as it is today. The first person on record who was honored as philologus was Eratosthenes of Cyrene, c. 3rd century B.C., head of the Hellenistic Library at Alexandria, known mostly as a great geographer and mathematician. His band of scholars, called "fatted fowls in a coop" by a contemporary, has been described as follows: (3) They [the philologists] had a carefree life: free meals, high salaries, no taxes to pay, very pleasant surroundings, good lodgings and servants. There was plenty of opportunity for quarreling with each other, (cited in Frank 1997: 486) From covering every conceivable field of learning in ancient Greece, including geography, history, mathematics, and philosophy, philology in the Middle Ages became more narrowly humanistic. In English, the first recorded appearance of the word is in Chaucer's Merchant's Tale: (4) Hoold thou thy pees, thou poete Marcian, That writest vs that ilke weddyng murie Of hire Philologie and hym Mercurie, And of the songes that the Muses songe

(MerT 11. 1732-1735)

Philologie here is the "personification" of linguistic and literary knowledge (MED), the lady matched in matrimony to the god Mercury.5 To seventeenth-century men of letters, philology was a focus on human liberal studies.6 By the middle of the eighteenth century, the term, as used in England, was specialized to mean 'the study of the structure and


linguistics, and the history of [hw]~[w]


development of language; the science of language; linguistics' {OED). In the post-Bloomfieldian scene in the United States, philology is richly polysemous in its use in academic circles. Nevertheless, as a survey reported in Winters and Nathan shows, there is a common denominator, and it is "the study of written texts" (1992: 363). Linking linguistics to philology, Encyclopedia Britannica also puts the written text as the primary target of philological study and describes the difference as follows:7 (5) The philologist is concerned primarily with the historical development of languages as it is manifest in written texts and in the context of the associated literature and culture. The linguist, though he may be interested in written texts and in the development of languages through time, tends to give priority to spoken languages and to the problems of analyzing them as they operate at a given point in time. ("Philology" in Encyclopedia Britannica Online, emphases DM)

4. The speaking hare: the text as a "native speaker" As is clear from the previous sections, the autonomy of linguistics versus philology has been defined on the basis of methodology (procedures of concept formation), cultural contextualization of the data, and the priority of the spoken language. If these differences between philological and linguistic work were really intractable, one would perhaps be wise to accept the division of labor and just plod on. Such a pessimistic view is unjustified, however. In this section I want to argue that none of the differences should stand in the way of integrating the two disciplines. Methodologically, philological research does not have to be characterized only as "attentiveness to minutiae ... a tolerance for pedantry, for the obscure, esoteric, and devious ..." (Frank 1997: 486). The kind of philology pursued by the community of English historical linguists, linguistic philology, selects its targets of study in a theoretically informed way. It employs the rigorous procedures of data classification and analysis mandated by other academic disciplines. Even if linguistic philology does not always draw on the latest formal devices that linguists use to model the properties of language in general, philological spadework is never blind prospecting. Both disciplines work towards recording and explaining language, and it is senseless to try to keep the empiricism of the philologist separate from the analyticity of the linguist.



Contextualization, Cecily Clark's concern, has not only been recognized as a necessary component of our work, but it has given us many remarkable insights into the causes and mechanisms of language change. All linguists involved in the historical study of English are philologists by her definition. For sociolinguists and the vigorous new area of historical pragmatics, contextualization is the methodology. With respect to conceptualization and contextualization, the breach between philologists and linguists has ceased to be an issue. This still leaves an important aspect of the presumed disciplinary incompatibility unaccounted for: the priority of the spoken language as a hallmark of linguistic research. If philology is defined by its concern with the past stages of languages, preserved only in written form, how do we mend that particular fence? The point calls for elaboration: it touches on the presence of the "native speaker" in the texts that we are working with.8 Clearly, all diachronic language study up to the invention of sound recording has to depend on texts. Texts reach us after scribal and editorial gestation; they are "secondary" speech products. I have argued elsewhere (Minkova 2003: Ch. 1), however, that although the physical immediacy of speech is irretrievably lost for the historical linguist, many textual features can inform us about the properties of the spoken language as closely as any native speaker would. One of the areas where the properties of speech are most reliably reflected is alliteration. In spite of the word's etymology, alliteration is a profoundly oral process; the selection of paired onsets can be argued to match closely the intuitions of speakers and scribes regarding phonological identity. Discussions of the "orality" of medieval literary culture have been quite divisive, usually along the lines of what constitutes formulaic language, a topic which cannot be addressed here.9 Instead, I want to highlight one particular aspect of poetic composition in early English which supports the idea of orality without reference to the formula. I have in mind the direct vocal, auditory nature of medieval verse production and transmission and the more general notion of "fit" between verse and the language in which verse is composed. Very briefly, the argument has to do with the decidedly aural intent and effect of alliterative verse. Along with Fleischman (1990: 20) and the references she cites, I assume that throughout the Middle Ages "writing was dictated and reading was carried out viva voce." As Fleishman writes:

Philology, linguistics, and the history of [hw]~[w]


(6) The term for writing as a method of composition [in the European Middle Ages] was dictare, whereas scribere generally referred only to the physical act of putting pen to parchment: these were different activities, carried out by different individuals. Legere, as late as the fifteenth century, normally entailed an oral articulation of the sounds being decoded. (Fleischman 1990: 20) This view of writing entails an intervening orality even in instances of mechanical copying. While the origins of silent reading go back to the eleventh century, the recording of verse represents a special case.10 In verse, author and reader have to rely on the shared esthetic properties of rhythm and sonority, which would require an aural trace even in overtly "silent" reading. Alliteration is something that not just the poet and the audience hear, it is also the structural and mnemonic glue that anybody involved in the preservation of a piece of verse would have been aware of. Copyists are also readers, and in reading the text prior to recording it, they must have drawn on their intuitions - this is what is meant by saying that alliteration is a first-hand reflection of the native intuitions about linguistic similarity and identity. Matching words which begin with the same sound is a straightforward and simple task which both children and adults can engage in without special instruction. Kith and kin, time and tide, people and places, gaggle of geese, sticks and stones are collocations whose survival owes much to alliteration. Psycholinguistically and cognitively, alliteration is a natural byproduct of human language, an ideal and immediate link between speaker and text. Thus, in spite of the fact that we work with "text" language, there are good reasons to assume that the evidence found in some forms of that language is a dependable source and target of linguistic study." The "orality" of the composition and transmission of texts, the reliance on dictation and reading aloud, and the natural acoustic basis of alliteration justify its elevation to the status of primary source of information for reconstructing the features of the spoken language. Acknowledging that there are other problems ensuing from the distance between manuscript forms and the actual spoken forms, I will assume here that alliteration emulates the native speaker's phonology accurately and that the implications of this "orality" deserve to be carried forward into the linguistic analysis. In alliterative verse the native speaker speaks to us, and we will do well to listen.12


Donka Minkova

5. Exploring the acoustic basis of alliteration: The story of /hw-/~/w-/ Thus much may serve by way of proem; Proceed we therefore to our poem. Jonathan Swift, On his Death 71 (1731) The rhetoric in defense of the bond between philology and linguistics will profit from a concrete example of how the voice can be extracted from the text, and what one can "do" with that voice.13 An issue which provides the kind of bridge between the past and the present that a linguistic philologist would want to build is the history of OE , whose realization in Modern English varies between a simple /w-/ and an aspirated /hw-/ or /w/(/λλ/).14 In trying to show how the two disciplines bear on one another, I will address the status of the contrast between, e.g., whet~wet, whine~wine in present-day English, and the history of the OE /hw-/ cluster, and I will offer a linguistic interpretation of the new philological findings. 5.1. The schizophrenic in Modern English Even within one single variety of English, the one described as "Received Pronunciation," or RP, the maintenance of the /hw-/~/w-/contrast varies depending on register and gender.15 Comparing the negative associations of eighteenth century [w-] pronunciation by Londoners to the attitude at the beginning of the twentieth century, Jespersen characterized the situation for RP as follows: (7) It [use of [w-]] is not, however, nowadays regarded as nearly so "bad" or "vulgar" as the omission of [h], and is, indeed, scarcely noticed by most people. In fact, a great many "good speakers" always pronounce [w] and look upon [hw] as harsh, or dialectal. In some schools, however, especially girls' schools, [hw] is latterly insisted on. (Jespersen 1909: 374) Gimson (1973: 217) refers to the voiceless labio-velar fricative /w/ as a variant found "amongst careful RP speakers and regularly in several regional types of speech, e.g. in Scottish English.... Among RP speakers, however - especially males - the use of /vy/ as a phoneme has declined rapidly (though it is often taught as the correct form in verse-speaking)." Wells (1982: 228-229) describes the realization of the historical /hw-/ cluster in RP as "schizophrenic." Within England, only Northumberland


linguistics, and the history of [hw]~[w]


preserves the contrast systematically to this day (Trudgill 1998: 40).16 Giner and Montgomery (2001: 350), investigating a late eighteenth-century Daybook, a document produced in Yorkshire, question the received view that the /hw-/ cluster was still used in northern England in the midnineteenth century.17 They found that "Variant spellings such as whenl and wich suggest its [the Daybook's] writer lacked initial aspiration and used what is the modern-day standard pronunciation in England /w/." In Scotland /w/ is still often described as contrastive, but in Ireland there are signs that the /w/~/w-/ contrast is a recessive feature, gradually confined to more rural areas, with simple /w-/ usual in Belfast and other urban parts, as described in Wells (1982: 408-409, 446). As for North American English, an interesting picture emerges from the dialect recordings of the last 100 years. Surveying the pronunciation of American English, Grandgent (1893: 277; 1895: 448) concluded that the reduction of the /hw-/ cluster was "comparatively rare."18 One generation later, Kurath and McDavid's Pronunciation of English in the Atlantic States (1961) recorded a wide-spread distinction between /hw/ and /w/ in whale and wail, which and witch in the North and the South, but not the Midland. The southern limit of the /hw/~/w/ distinction coincided almost completely with the lexical isogloss separating the North and North Midland through Pennsylvania. Another four decades later the situation has changed once again, making the simple voiced labio-velar approximant the dominant pronunciation. The 1997 map of the /hw/~/w/ distinction in the Phonological Atlas of North America makes this very clear:19

Figure 1. Contrast of /hw/ and /w/ from the Phonological

Atlas of North



Donka Minkova

The comment that accompanies the map is worth citing in full: (8) Since the LAMSAS data was gathered, the distinction has rapidly eroded. Map 8 shows only 71 of 587 speakers who maintain it. In this case, "Distinct" includes all those who were heard by the analyst as pronouncing the voiceless bilabial clearly (62 cases) or not quite clearly (9) cases. There were 3 individuals who thought that the pairs were different, but made no distinction in production; they were considered to be merged. (Cited from ) This is the picture at the beginning of the twenty-first century. There is no clear regional pattern, and only 10-12% of the speakers preserve the contrast. "The extent to which Ihl has been preserved (or perhaps restored) as a spelling pronunciation remains to be established" (Montgomery 2001: 143). On the one hand, there are some remaining /w/ areas in Britain: partly Scotland, Wales, and Ireland; on the other, in the US, the survival of /w/ is in question. The general North American shift from /hw-/ to /w-/ is shared by Canadian English (Brinton and Fee 2001: 430). This kind of instability of the /w/~/w/ contrast goes back to Old English, though the original proportions are reversed.

5.2. Old English/hw-/ In Old English, the sequence /hw-/ was unquestionably a bi-phonemic cluster, pairing with other initial clusters: , , . The evidence for that is both the comparative stability of the spelling and the way in which initial words were treated in the poetry: in fully stressed lexical items alliterates regularly on the initial consonants in the cluster, presumably [h-]20: (9) Ac se hwita helm / hafelan werede21 hea hornscipe, / ofer hwaeles eöel22 Hwalas öec herigaö, / and heofonfugolas 23

Beo 1448 Andreas 274 Dan 386

This is the predominant pattern of alliteration in Old English. There are, however, some early signs of occasional loss of the aspiration in the


linguistics, and the history of [hw]~[w]


cluster: some -less spellings and unetymological insertions are also found in Old English: (10) Manuscript form

Normalized form hwcelweg hwistle nahwar ahwcer ohwanan ceghwceper bilewit

Source The Seafarer 1. 63 Bosworth & Toller 1243 Vercelli Homilies xxii Meters of BoethiuslA, 33 Leiden Riddle 8 Riddle 88 27 24 Vercelli Homilies xvi. 113, 117

More revealing than the spellings of compounds in which, admittedly, the unaspirated form appears most often in the prosodically weak right-hand part, are alliterative matchings in (mostly) late Old English verse: (11) hwearfiim wrascmaecgas. / Woö up astag25 t>a hwile J)e hi waspna / wealdan moston. 26 weras ferhöe / hwearfum Jjringan27 wiö öy hwitan attre / wiö öy wedenan attre28 and he Jjar wunode / öa hwile ]?e he lyfode 29

Guthlac 263 Maldon 83 Judith 249 Charms 248 Death of Alfred 21

The examples in (11) suggest reduction and identification with /w-/. These are some of the earliest symptoms of the cluster's "schizophrenic" behavior.

5.3. Early Middle English: Lajamon's


In Middle English the evidence for loss of contrastive aspiration and merger with the pre-existing /w-/ becomes increasingly solid. When alliterative verse composition re-emerges at the end of the twelfth century, the etymological/xw-/ or/hw-/ was already eligible for alliteration with /w-/ i η both texts of Lajamon's Brut. The citations in (12) are from the more archaizing Caligula text; "O" means that the same alliterating pair is found also in Otho: 30


Donka Minkova

(12) buten while £>at Jier at-wond; }>urh wode bure 6e walles of stone; t>e duren of whales bone whar ich mihte on wildeme; wurchen aenne castel. Wiö him warfte Brien; al his iweden.

1084 (C, O) 182 (C, O) 7697 (C, O) 15343 (C)31

The hypothesis that for Lajamon the simplification of the cluster was an option is reinforced by additional scribal evidence in the two surviving copies. First, numerous inverse spellings of etymologically /w/- initial words in the Caligula version support an assumption of merger: (13) Etymological spelled in La^amon's Brut: iwhat 'went' (12784) iwhiten 'known' (7890) wharö 'became' (2467) what 'knew' (8572) whraöe 'wrathful' (9260)

whingen 'wings' (14604) whit 'wight, man' (5757, 7974, 12911, etc.) whit 'with' (2550, 2581, 2641, 12911, etc.) whitere 'brave' (10658) whreken 'revenge' (5392)

The examples in (13) suggest that the scribe was aware of the tradition of representing /w-/ with , but did not know which words merited that spelling. Compare the spellings in (13) to spellings elsewhere in the MED: (14) Early unetymological for in the MED:


'wet' + 'hall'


'wash'+'man' 'waiter'



Lay Subsidy Rolls, Sussex (12961332) The Place-Names of Cheshire (1308) EPNSoc.52 (Dor.) (1340) Feudal Aids 5 (1346) Thuresson ME Occup. Terms (1349) Reaney Diet. Br. Surnames (1374) The Place-Names of Essex (1376)

Philology, linguistics, and the history of [hw]~[w]


Conversely, etymological -initial words are commonly rendered as a simple : wcer 'where,' wat, wcet 'what,' wile 'while, time,' wanene, wonene 'whence,' wenne 'when,' wife 'white,' wulc 'which,' wuder, woder 'whether' are spellings in Caligula. It should be noted that while and spellings alternate in the Caligula manuscript, all etymological /hw-/ items are spelled in the Otho copy. In view of the research history of this issue, I want to emphasize specially the variety of lexical items in which the reduction is attested. The spelling evidence includes not only the interrogatives, which would be predictable "leaders" in this development, but also fully stressed historical /hw-/ words: (15) warf1 turning' (C) wcerf'crowd'(C) wate 'swift' (O) wuruen 'attacked' (C) wile 'time, while' (C, O) wite 'white' (C, O) iwet 'sharpened' (O)

OE hwearf (1036) OEhwearf (8727) OE hwaet (7137) OE hweorfan (9139) OE hwile (115, 174, 336, 456, etc.) OE hwit (594) OEhwettan (15263)

Again, the practice of the two scribes is confirmed independently by spellings found elsewhere in the twelfth- through early thirteenth-century records:32 (16) Early spellings for etymological /wh-/ in the MED:

'wheat-' 'wharf-' 'whorl-'

'white-' 'wheel-'

1177, Reaney Dict.Br. Surnames 378 1202, Ekwall Dict.EPN 4%Ί 1189-1199, EPNSoc.5 (N Riding Yks.) 1166, EPNSoc.8 (Dev.) 247 cl 192, EPNSoc.3\ (West Riding Yks.) 67

The type of alliterative evidence referred to in (12), the spellings that the two Lasamon's Brut scribes used in copying the original, and the independent early evidence provided by place and personal names reinforce each other to strengthen the case for an early reduction of the etymological cluster /hw-/, at least for some varieties of English spoken in the South West. The presence of reduced variants of /hw-/ continued throughout the Middle English period.


Donka Minkova

5.4. Fourteenth-century evidence of /hw-/ reduction The reduction of /hw-/ started in the south at about the same time that the other /h-/ initial clusters began to be simplified. Judging both from alliteration and from the present-day state of the opposition, this particular change did not occur or was delayed in the northern dialects. In verse, the more northern compositions, as, e.g., The Wars of Alexander, allow /hw-/ to be matched both to /h-/ and to /w-/, and even to etymological /kw-/, as in (17a)-(17c): (17) a. 3e behald me sa hogely quareon is 3our mynd Of }je quilke he hopid in his hert sumquat to knawe

Wars 269 Wars 679

b. For now vs wantis in a qwirre as quele turnes Wars\9%Q And sone be wacchemen without quen bai him bare sawe Wars 5290 c. Quirris furth all in quite of qualite as aungels Wars 1679 For h[i]m was quartirs of qwete vmqwile out of nombre Wars 4640 The poet's language, as reconstructed on the basis of the alliterative practice, reflects both familiarity with the southern dialects and the survival of the initial segment "in parts of the North and North Midlands ... [as] a very strongly aspirated /xw-/" (Duggan and Turville-Petre 1989: xxxvii). The Parlement of the Thre Ages is another poem which illustrates the uncertainty of the realization of the etymological in dialectally mixed texts. The original dialect of the poem is "the central or southern part of the West Riding of Yorkshire" 33 but influences from the East and South Midlands are recognizable in the way the etymological cluster is treated in alliteration: (18) The Parlement of the Thre Ages: And quopes 34 thaym to the querrye that quelled hym to the dethe (233) And he ne wiste in alle this werlde where he was bycomen (507) And his techynges will bene trowede whills the werlde standes, (604) For comparison, the Gawain

poet, located topographically between

La3amon and the Wars poet, also allows alliteration between reflexes of Old English and /w/:35

Philology, linguistics, and the history of [hw]~[w]

(19) l>e fyrst word jjat he warp,' Wher is', he sayd, SGGK Whettez his whyte tuschez; with hym }jen irked SGGK What! hit wharred and whette, as water at a mulne; SGGK And wyth quettyng awharf, er he wolde ly3t; SGGK


224 1573 2203 2220

The matching of etymological /hw-/ to /kw-/ () in northern texts reflects an important cluster development in the history of English. In Northumbrian the etymological cluster /kw-/ was unstable and its first element was spirantized to [χ-]; in the same dialect area the realization of the Germanic */χνν-/ was probably [xw-]. The two clusters merged, most likely as the result of substratum-induced sound-substitution in the speech of the indigenous British Celts (Laker 2002). 36 That northern pairing is therefore quite separate from the despirantization of the cluster in other varieties of Middle English. Moving further to the south, we find that like La3amon, Langland, whose dialect is also that of the South West Midlands, regularly pairs fully stressed words with /w-/: while 'time,' whiten 'whiten,' the why es 'the whys, causes,' etc.37 (20) Piers Plowman (B-Text): "Whit wyn of Oseye and wyn of Gascoigne (P.229) For to werche thi wille the while thow myght laste (3.028) Wowes do whiten and wyndowes glazen (3.061) Now awaketh Wrathe, with two white eighen (5.133) "For to werche by thi wordes the while my lif dureth (6.056) And wepen whan I sholde slepe, though whete breed me faille (7.121) For alle that wilneth to wite the whyes of God almighty (10.124) And thus thorugh wiles of his wit and a whit dowve (15.407) Similarly, in William of Palerne , commonly spelled also , alliterates on /w-/: (21) William of Palerne:™ Sehe awayted wel }>e white bere-skinnes (Will 1710) Whilum J)ei went on alle four, as doj) wilde bestes (Will 1788) Whanne f>e wite beres wist, Jjat were in }>e quarrer (Will 2401)


Donka Minkova

Notice, in the same text, the inverse spelling for wait, comparable to the spelling of warm in Langland's Piers Plowman: (22) He went to an hei3 weie to whayte sum happes (Will 1885) And wij) wharme 39 water at his eyghen wasshen hem after (PPl.B 15.192) Leaving the realm of tightly structured alliterative verse makes it harder to establish the continuity of the merger, but some fifteenth-century spelling evidence certainly exists. The forms / for 'which' are frequent alternatives in the 1384-1425 records of London English; there are 12 such instances listed in Chambers and Daunt (1931: 379). One of the texts where the merger is well attested, The Brewers' First Book, includes an inverse spelling of with as at 1. 816 for the year 1423. Other fifteenth-century spellings indicative of merger found in the MED are cited in (23) and (24): (23) 15th century spellings for in the MED: , 'whale' 40 'whale oil' 42 'wharf 4 4 'whey' 46 'whirl' 48 'wheat-flour' 50

'whether' 41 'whetstone' 43 'while'(n.) 45 'white' 47 'whin' 49 'bleach' 51

(24) 15th century spellings for in the MED"50 'vanish' cl450 53 'want' 147555 (feile-) 'ware' c 147557 'wash' al500 59 'without' cl400 61 'willow' al475 63

'water' c. 147554 'waves' al500 56 'wax-maker' al500 58 'weighs' cl469 60 'wight' a 147562 'wa\V(MED)

That takes us to the end of the fifteenth century. At that point the identification of the etymological /hw-/ with /w-/ is a fact in large areas of the South and the Southwest. Unlike other regional features, such as the clerk's vowels and the third person plural pronouns in Chaucer's Reeve's

Philology, linguistics, and the history of [hw]~[w]


Tale, there seems to be no association, negative or otherwise, between "provincialism" and either /w-/ or /hw-/; those kinds of associations developed later. Although the alliterative data I have presented are new in this context, the recognition of the merger is not. The assumption regarding the history of the cluster is that , re-spelled in Middle English, was lost first in the South and the SE Midlands and popularly certainly also in London (Jordan-Crook 1934/1974: 178-179). This dating is based on spelling evidence in The Peterborough Chronicle (1132 ff.), The Ayenbite of Inwyt, Poema Morale, Vices and Virtues, Trinity College Homilies, etc. Turning to the profiles in the Linguistic Atlas of Late Medieval English (LALME, Mcintosh et al. 1986), we find confirmation of the merger too. Volume 4 lists 97 spellings for what (278), shown on Map 1091.

K J ,


< φ »

. ,





»· * J « ^ · " Λ«, ' Ά >*« m* »


*« *% , ·






,* SgK - »»Λ · V v · Γ •*. : ' * it* »« , «** •»• *» ί · ' V

r-f i *

f ' _• V* $1





* *

* .



• « * V»^. » ·* * .- -" ,-Ρί-'Χ -·^ f V—1 -

ηι ν5

Figure 2. WHAT: wa- forms (LALME Map 1091)

LALME lists also 62 instances of used as a reverse spelling for as in 'was,' 'we,' 'with.' These spellings are distributed all over the south, with somewhat higher density in Essex (x9), Somerset (x6), Suffolk and Gloucestershire (x5), Berkshire and


Donka Minkova

Warwickshire (x4), etc. Similarly convincing is the distribution of while forms with (Map 253 below), and the overall w + V spellings for (Map 274): m

WW tß y

Λ - Λ>




χ*Υ* * Ι " * *··* . Bf*!·.· Ν ^ Λ ; . *; ί : · '·

ModE such, OE twa > ModE two, the ongoing post-coronal /ju/ > /u/ simplification which started in EMod English and keeps expanding, e.g rude, chew, suit, lute, and American southern pronunciation of Tuesday, news, etc. 76. See also the relationship of hurlpool, hurlwind, hurtleberry to whirlpool, whirlwind, whortleberry, etc., and of thwack, thwang to whack, whang, in the OED. The in whip is unetymological - according to the OED the word was borrowed from the Dutch wippe, wip, in the 14th century. Similarly, the dialectal whap 'a blow, an instant' is from ME wappen v. (MED). 77. A line of analysis which cannot be pursued here is suggested by the unmarkedness of [w] as compared to its voiced counterpart. This prompts the possibility that actual realization of the in the South that was introduced by the educators was a monophonemic [w], and not a bi-phonemic, or doubly-gestured aspirated /hw/. 78. Although "New Philology" is not what this paper is about, it should be pointed out that literary scholarship is also indebted to this discipline. Here the words of C. S. Lewis are still valid: "Those who ignore the relation of [Modem] English to Anglo-Saxon as a 'merely philological fact' irrelevant to

Philology, linguistics, and the history of [hw]~[w]


the literature betray a shocking insensibility to the very mode in which literature exists." Interesting discussions on the intersection of philological and linguistic scholarship are found in Lerer (1996).

References Anderson, John 2001 A major restructuring in the English consonant system: the delinearization of [h] and the de-consonantization of [w] and [j], English Language and Linguistics 5 (2): 199-213. Bailey, Richard 2002 A thousand years of the history of English. In Studies in the History of the English Language: A Millennial Perspective, Donka Minkova, and Robert Stockwell (eds.), 451-473. Berlin/New York: Mouton de Gruyter. Boersma, Paul. 1998 Functional Phonology. Formalizing the Interactions between Articulatory and Perceptual Drives. Netherlands Graduate School of Linguistics: LOT. Brinton, Laurel, and Margery Fee 2001 Canadian English. In The Cambridge History of the English Language Vol. VI, John Algeo (ed.), 422—440. Cambridge: Cambridge University Press. Brook G. L., and R. F. Leslie (eds.) 1963, 1978 La3amon's Brut, EETS 250, 277. London: Oxford University Press. Bunt, G.H.V. (ed.) 1985 William of Palerne. An Alliterative Romance. Groningen: Bouma's Boekhuis bv. Cerquiglini, Bernard 1999 In Praise of the Variant: A Critical History of Philology. Betsy Wing (trans, and intro.). Baltimore, MD: Johns Hopkins University Press. Chambers, R. W., and Maijorie Daunt 1931 A Book of London English 1384-1425. Oxford: At the Clarendon Press. Clark, Cecily 1987 Historical linguistics-linguistic archaeology. In Papers from the 5,h ICEHL, Sylvia Adamson, Vivien Law et al. (eds.), 55-69. Amsterdam: John Benjamins. Desmet, Piet, Peter Lauwers, and Pierre Swiggers 1999 Dialectology, philology and linguistics in the romance field: Methodological developments and interactions. Belgian Journal of Linguistics 13: 177-203.


Donka Minkova

Dobson, Eric 1957 English Pronunciation, 1500-1700. Oxford: Clarendon Press. Donoghue, Daniel 1997 Language Matters. In Reading Old English Texts, Katherine O'Brien O'Keeffe (ed.), 101-123. Cambridge, U.K./New York: Cambridge University Press. Duggan, Hoyt, and Thorlac Turville-Petre (eds.) 1989 The Wars of Alexander. EETS SS10. Oxford: Oxford University Press. Ellis, Alexander J. 1968 Reprint. The existing phonology of English dialects, part 5 of On Early English Pronunciation. Early English Text Society, ES 56. New York: Garland, 1889. Fisiak, Jacek (ed.) 1990 Historical linguistics and philology. Berlin/New York: Mouton de Gruyter. Fleischman, Suzanne 1990 Philology, linguistics, and the discourse of the Medieval text. Speculum 65 (1): 19-37. Frank, Roberta 1997 The unbearable lightness of being a philologist. JEGP 96.4, 486513. Fulk, Robert 1992 A History of Old English Meter. Philadelphia: The University of Pennsylvania Press. Gimson, A.C. 1973 An introduction to the pronunciation of English. 2d ed. London: Edward Arnold. Giner, Maria F. Garcia-Bermejo, and Michael Montgomery 2001 Yorkshire English two hundred years ago. Journal of English Linguistics 29 (4): 346-362. Grandgent, C. H. 1893 American pronunciation again. Modern Language Notes 8: 273-282. 1895 English in America. Die Neueren Sprachen 2: 443-467. Harris, David Payne 1954 The phonemic patterning of the initial and final consonant clusters of english from late old english to the present: A structural approach to their historical development Ph.D. diss., University of Michigan. Hickey, Raymond 1984 Syllable onsets in Irish English, Word 35: 67-74. Hogg, Richard 1992 A Grammar of Old English. Oxford: Blackwell Publishers.

Philology, linguistics, and the history of [hw]~[w] 1994


Linguistics, philology, chickens and eggs. In English Historical Linguistics, Francisco Fernandez, Miguel Fuster, and Juan-Jose, Calvo (eds.), 3-16. Amsterdam: John Benjamins. Jespersen, Otto 1909 A Modern English Grammar on Historical Principles. Vol. I Heidelberg: Carl Winter Universitätsbuchhandlung. Jordan, Richard 1934/1974 Handbook of Middle English Grammar: Phonology, Eugene Crook (trans, and rev.), Berlin: Mouton. Koerner, Konrad 1982 On the historical roots of the philology/linguistics controversy. In Papers from the 5th International Conference on Historical Linguistics, Anders Ahlqvist (ed.), 404-413. Amsterdam: Benjamins, Kristensson, Gillis 1967 A Survey of Middle English Dialects J290-1350: The Six Northern Counties and Lincolnshire. (Lund Studies in English 35.) Lund: Gleerup. Kurath, H., and R. I. McDavid Jr. 1961 The Pronunciation of English in the Atlantic States. Ann Arbor: U. of Michigan Press. Ladefoged, Peter, and Ian Maddieson 1996 The Sounds of the World's Languages. Oxford: Blackwell Publishers Ltd. La$amon's Brut, see Brook G. L. and R. F. Leslie Laker, Stephen 2002 An explanation for the changes kw-, hw-, > ?w- in the English dialects. In The Celtic Roots of English. Markku Filippula, Juhani Klemola, and Heli Pitkänen, (eds.), 183-198. (Studies in Languages 37.) University of Joensuu, Faculty of Humanities: Joensuu. Lass, Roger 1999 Phonology and morphology. In The Cambridge History of the English Language Vol. Ill, 1476-1776, Roger Lass (ed.). 56-187. Cambridge: Cambridge University Press. Lass, Roger, (ed.) 1969 Approaches to English Historical Linguistics. An Anthology. New York: Holt, Rinehart and Winston, Inc. Lerer, Seth (ed.) 1996 Literary History and the Challenge of Philology: The Legacy of Erich Auerbach. Stanford, California: Stanford University Press. Luick, Karl 1964 Reprint. Historische Grammatik der englischen Sprache. Stuttgart/ Oxford: Bernhard Tauchnitz/Basil Blackwell. 1914-1940.


Donka Minkova

Lutz, Angelika 1989 Phonotaktisch gesteuerte Konsonantenveränderungen in der Geschichte des Englischen. Tübingen: Max Niemeyer Verlag. MacMahon, Michael 1998 Phonology. In The Cambridge History of the English Language Vol. IV, 1776-1997. Suzanne Romaine (ed.), 373-535. Cambridge: Cambridge University Press. Madden, Sir Frederic 1847 [1970] Layamons Brut, or Chronicle of Britain; a poetical semi-Saxon paraphrase of the Brut of Wace. Now first published from the Cottonian manuscripts in the British Museum, accompanied by a literal translation, notes, and a grammatical glossary. London: Society of Antiquaries of London, 3 volumes. Matthews, William 1959 Review of Dobson (1957). Modern Language Notes lxxiv, April 1959, 356-363. Mcintosh, Angus, M. L. Samuels, and Michael Benskin, with the assistance of M. Laing, and K.Williamson. 1986 A Linguistic Atlas of Late Medieval English, 4 vols. Aberdeen: Aberdeen University Press. McSparran, Frances (ed.) 1986 Octovian. EETS 289. London: Oxford University Press. MED: Middle English Dictionary 1952-2001 Hans Kurath, Sherman M. Kuhn, Robert E. Lewis (eds.) Ann Arbor: University of Michigan Press. Milroy, James 1983 On the sociolinguistic history of /h/-dropping in English." In Current Topics in English Historical Linguistics, Michael Davenport (ed.), 37-54. (Odense University Studies in English, Vol. 4.) 1992 Middle English dialectology. In The Cambridge History of the English Language Volume II1066-1476. Norman Blake (ed.), 156— 204. Cambridge: Cambridge University Press. Minkova, Donka 2003 Alliteration and Sound Change in Early English. Cambridge: Cambridge University Press. Mitchell, Brace 1992 How to study Old English syntax? In History of Englishes: New Methods and Interpretations in Historical Linguistics, Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen, Irma Taavitsainen (eds.), 92-100. Berlin/New York: Mouton de Gruyter. Montgomery, Marshall 1910 Types of Standard Spoken English and Its Chief Local Variants. Twenty-Four Phonetic Transcripts from "British Classical authors " of the XIXth Century (Herrig-Foerster, vol. II) Strassburg: Trubner.

Philology, linguistics, and the history of [hw]~[w]


Montgomery, Michael 2001 British and Irish antecedents. In The Cambridge History of the English Language Vol. VI, English in North America, John Algeo (ed.), 86-153. Cambridge: Cambridge University Press. Oakden, J. P. 1930, 1935/1968 Alliterative Poetry in Middle English. The Dialectal and Metrical Survey, Archon Books. [Originally published 1930, 1935, reprint 1968] Offord, Μ. Y. (ed.) 1967 Reprint. The Parlement of the Thre Ages. EETS No. 246. London: Oxford University Press, 1959. Orchard, Andy 1997 Tradition. In Reading Old English Texts, Katherine O'Brien O'Keeffe (ed.), 101-123. Cambridge, U.K./New York: Cambridge University Press. The Oxford English Dictionary 1989 2d ed. J. A. Simpson and E. S. C. Weiner (eds.). 20 vols. Oxford: Oxford University Press. (Online version, 1992.) Pullum, Geoffrey, and William Ladusaw 1986 Phonetic Symbol Guide. Chicago: The University of Chicago Press. Saenger, Paul 1997 Space Between Words. The Origins of Silent Reading. Stanford, California: Stanford University Press. Schaefer, Ursula 1997 Ceteris Imparibus: Orality/Literacy and the Establishment of AngloSaxon Literate Culture. In The Preservation and Transmission of Anglo-Saxon Culture: Selected Papers from the 1991 Meeting of the International Society of Anglo-Saxonists, Paul Szarmach and Joel Rosenthal (eds.), 287-311. Kalamazoo: Western Michigan University. Sievers, Eduard 1901 Grundzüge der Phonetik zur Einfuhrung in das Studium der Lautlehre der indogermanischen Sprachen, fünfte verbesserte Auflage. Leipzig: Druck und Verlag von Breitkopf & Härtel. Sievers, Eduard, and Karl Brunner 1942 Altenglische Grammatik nach der Angelsächsischen Grammatik von Eduard Sievers, neubearbeitet von Karl Brunner. Halle/Saale, M. Niemeyer. Steriade, Donca 2001 The phonology of perceptibility effects: The P-map and its consequences for constraint organization. Manuscript. Unviersity of California, Los Angeles, 3/29/01.


Donka Minkova

Suzuki, Seiichi 1996 The Metrical Organization of Beowulf. Prototype and Isomorphism. Berlin/New York: Mouton de Gruyter. Trudgill, Peter 1998 The Dialects of England. 2d ed. Oxford: Blackwell Publishers. Vennemann, Theo 1988 Preference Laws for Syllable Structure and the Explanation of Sound Change. Berlin/New York: Mouton de Gruyter. Walker John 1791/1831 A Critical Pronouncing Dictionary and Expositor of the English Language ... Edinburgh: Thomas Nelson and Peter Brown. Wells, J. C. 1982 Accents of English. 3 vols. Cambridge: Cambridge University Press. Werner, Heinz 1998 Do linguistics and literature study still have something to say to each other? On the relationship between philology, language science, and linguistics. Abhandlungen zur Sprache und Literatur, 1998, 100: 161-189. Winters, Margaret E., and Geoffrey S. Nathan. 1992 First he called her a philologist and then she insulted him. In The Joy of Grammar: A Festschrift in Honor of James D. McCawley, Diane Brentari, Gary N. Larson, and Lynn A. MacLeod (eds.), 351-367. Amsterdam: Benjamins. Wright, Joseph (ed.) 1905 English Dialect Grammar. Oxford: Oxford University Press. Wyld, H. C. 1936 A History of Modern Colloquial English. Oxford: Blackwell.

An essay in historical sociolinguistics?: On Donka Minkova's "Philology, linguistics, and the history of [hw]~[w]" Lesley Milroy

I am delighted to have been invited by the editors to comment from the perspective of a sociolinguist interested in phonological variation and change on Donka Minkova's stimulating article "Philology, linguistics and the history of [hw]~[w]."' Adopting procedural and interpretative frameworks from fields which she distinguishes as "philology" and (fairly formal) "linguistics" - but not sociolinguistics - the author's chief goal is to argue for the long term instability of the distinction in English in words of the wine/whine sets. To this end, she presents and interprets exceptionally rich linguistic and other data drawn from a wide range of sources over a very long period. Dr. Minkova has recently published a book which explores the use of alliterative evidence from the Old English period and beyond to support accounts of sound change, and this article exemplifies the kind of argumentation she has developed. Its most striking and innovatory contribution is to place the early stages of the of [hw]~[w] merger much earlier than is customary, and to show that the two word classes were not systematically distinguished even in the Old English period. The argumentation is first rate, the author's starting point being the status of /hw-/ in Old English as one of a member of a set of /h-/ initial clusters. The other clusters (/hi-/, /hr-/, /hn-/) did not survive beyond 1300. The loss of initial /h/ in these sets is associated with more general changes affecting /h/. But the [hw]~[w] distinction survived much longer than the others in the quartet, and seems to be only now in the process of disappearing from contemporary speech communities in North America and northern parts of the British Isles (see Chambers 2002; Stuart-Smith 1999; Chirrey 1999). Dr. Minkova's discussion of the gradual attrition of this phonological distinction over more than a millennium is of great interest to sociolinguists working on patterns of variation and trajectories of change in con-


Lesley Milroy

temporary speech communities. In fact, many changes currently in progress are historically very deep-rooted; for example James Milroy (1992) provides evidence of /h/ dropping as early as the twelfth century. This process was foregrounded by Victorian writers on language and etiquette as a particularly deplorable vulgarism (see Mugglestone 1995) and is sociolinguistically salient in British English to this day. Quite recently, Kerswill and Williams (2002: 95) report that Α-less variants are receding in southern England, while remaining as robust as ever in the north. So it is no surprise to find that another ancient pattern of alternation displays interesting sociolinguistic dynamics. My remarks here are confined to issues salient in variationist sociolinguistics, and from this perspective Minkova's article is a substantial contribution to the developing subfield of historical sociolinguistics. Scholars working in this dynamic research area will appreciate the amount of primary data she has marshaled relevant to the issue of the [hw]~[w] merger, as well as her account of scholarly comments on its status from the sixteenth century onwards. She also presents information on a situation quite well known to sociolinguists, where a distinction is (apparently) reintroduced - in the sixteenth century in this case - a considerable time after a merger appears in the available sources to have progressed to completion. A brief summary of the field of historical sociolinguistics is in order at this point. In fact, the research of many variationist sociolinguists is quite heavily historically oriented, as those working primarily in contemporary communities often exploit the availability of written texts to add historical depth to analyses of variation and change. Minkova's contribution is particularly appreciated, since, with the notable exception of James Milroy's work (see again Milroy 1992) much sociohistorical research focuses on morphosyntax rather than phonology. For example, Tottie and Harvie (2000) include information on relative markers at earlier stages of English to support an argument for the close relationship between early African American English and dialects of English. And Trudgill (1996) provides details of a language contact situation in sixteenth-century Norwich to account for a contemporary pattern of alternation between zero and -s third-person singular present tense verb forms in contemporary Norwich vernacular. Historical sociolinguists characteristically do rather more, however, than using historical data to support arguments about essentially contemporary sociolinguistic issues, their chief goal being to understand trajectories of changes completed at earlier stages of the language. Variationist meth-

An essay in historical sociolinguistics?


ods are employed to illuminate the dynamics of these changes, and scholars work with the uniformitarian principle, which holds that structurally speaking the dynamics of variation and change in the past are similar to those observed in contemporary speech communities (see Lass 1997: 2632). The methodological challenges faced in applying this principle have led William Labov to describe historical linguistics in general as "the art of making the best use of bad data" (1994: 11). The "bad-data" problem has several dimensions: data are often patchy as a consequence of the random preservation of some texts and the equally random loss of others; the relationship between data derived from various kinds of written sources and the data of spoken interaction which forms the basis of much contemporary sociolinguistic work is unclear; reconstructing the social information needed to interpret patterns of variation in written texts is not always straightforward. So data might more accurately be described as incomplete rather than necessarily "bad." Labov's pessimistic comment was articulated in the course of an extended account of principles of phonological change, but he certainly does not contemplate the extensive use of alliterative material as a resource for such work, demonstrated by Minkova's analysis. Nevalainen (1999) argues that in any event his verdict is hardly relevant to accounts of morphological or grammatical change, which usually draw on substantial computerized corpora. She and her colleagues show also that earlier social worlds can to some extent be reconstructed from the detailed findings of contemporary social historians, some of which are of considerable sociolinguistic relevance (see, for example, Keene 2000). The Helsinki researchers are able to present well-motivated accounts of the social trajectories of particular grammatical changes associated with the sixteenth and seventeenth century, a period of particularly rapid social change dominated by massive migration into London from other parts of the country. One of the issues discussed by Minkova is the apparent readoption during the sixteenth and seventeenth century of the [hw]~[w] distinction, which was at that time still robust in northern English and Scottish dialects. Interestingly, some well-known morphosyntactic changes which took place during this period also involve the spread of northern features into London and the south - such as the variable use of the older third-person singular present tense verb form -ΕΤΗ as opposed to the innovatory northern dialect form -(E)S. These variants were in competition for over two hundred years, before -(E)s, which had emerged as an alternating variant in the fifteenth century, took over from -ΕΤΗ, becoming the norm by about


Lesley Milroy

1600 in all but "high registers" (Lass 1999: 162-165). Another northern morphosyntactic feature which became established in southern varieties over the same period is ARE rather than BE. Milroy (2002) discusses briefly some of the demographic changes in the sixteenth and seventeenth centuries likely to be of sociolinguistic relevance - such as the very high proportion of apprentices of northern origin migrating to London. It is possible that a systematic consideration of the sociolinguistic landscape in the early modern period could shed further light on the nature of the demographic dynamics underlying changes in the status and distribution of the [hw]~[w] distinction reported by Minkova. Indeed, she has already offered some necessarily tentative suggestions. Although Minkova's evidence places the [hw]~[w] merger very much earlier than is generally assumed, it is plain that it affected only some speakers in some communities, and that other speakers sometimes merged the relevant word classes and sometimes did not. In other words, the merger was variable. This point is important, for if a merger is assumed to be variable rather than categorical, the puzzle of how "unmerging" takes place dissolves. What is happening is not unmerging at all, but a sizable shift in frequency and distribution of unmerged and merged variants. A good deal of modern sociolinguistic research shows that mergers proceed slowly and ambiguously. Since some reported mergers have never taken place in the systems of some speakers and some communities, the unmerged pattern is present in the wider speech community and so has the potential to spread and reestablish itself more strongly as the sociolinguistic landscape changes. Interestingly, some reported mergers affect formal speech and citation forms, while distinctions remain in the informal spontaneous speech of at least some speakers. Labov (1994) discusses issues of this kind, while Milroy and Harris (1980) demonstrate not only that words of the MEAT class were variably realized by high mid-front variants overlapping with realizations of the MATE class, but that a three-way MATE/MEAT/MEET distinction was still present in the Belfast vernacular system. However, evidence of this distinction is not available in citation forms (word list readings), in self report and in rhymes of dialect poets. All these sources suggest that the three-way distinction had disappeared, as standardly reported in histories of English. This point is relevant to an interpretation of the orthoepic evidence presented by Minkova, which suggests a reinstatement, at least in southern dialects, of a merger which had gone to completion some centuries before. In view of the sociolinguistic evidence provided by Milroy and

An essay in historical sociolinguistics?


Harris, we are entitled to wonder how far commentators are describing current spoken language as opposed to ideal canonical or citation forms. The orthoepists' focus on the writing system as a guide to correctness, a point emphasized by Minkova, even means that they sometimes may not be commenting on spoken use at all. What is clear from the comments of much later writers such as Jespersen and Gimson is that the [hw]~[w] contrast was taught prescriptively, probably primarily in the nineteenthcentury British public school system (a network of high prestige private schools). And of course stage actors are traditionally trained to realize the [hw]~[w] distinction. Certainly it looks from the evidence presented by Minkova as if many English speech communities might have merged the distinction only variably, with instability persisting over centuries. And that instability continues. It is likely that the [hw]~[w] distinction is still robust in rural areas of Scotland and Ireland; but in reports published in 1999, Stuart-Smith in Glasgow and Chirrey in Edinburgh report its disappearance in the speech of younger people. Until quite recently the distinction was characteristic of these urban dialects; having been raised in Glasgow myself, I can vouch for its general health in that city 40 years ago and for its persistent presence in my own phonological system. Chambers (1996, 2001) reports a similar sociolinguistic pattern in Canada, where the merger has progressed to near completion over the last 40 years. It is likely that the robustness of the distinction he finds in the speech of elderly Canadians in the Golden Horseshoe area of central Canada arises from the large proportion of Scottish immigrants to that region. So Dr. Minkova has used historical evidence exceptionally innovatively to provide a detailed demonstration of the time depth of a particular pattern of sociolinguistic variation. She has also demonstrated that the trajectory of attrition followed by a disappearing phonological distinction is not linear. The [hw] variant, which seemed already to be in poor health several hundred years ago, did not simply face an obscure demise as a feature associated with remote northern British dialects. Rather, it was subsequently seized upon as a prescriptively correct variant to be used in realizations of words such as whine, which, whale. Such changes in social evaluation of variants are not uncommon; consider the case of non-prevocalic IT/ in New York City, documented by Labov (1972). But we seldom have available such an historically detailed account of a phonological change as this one. Minkova's article provides a fine resource for historical sociolinguists to


Lesley Milroy

investigate further the phonological and social dynamics of a very longterm change.

Note 1.

Thanks to James Milroy for reading and commenting on an earlier draft of this short review.

References Chambers, J. K. 2001 Patterns of variation including change. In Handbook of Variation and Change, J. Chambers, P. Trudgill, and N. Schilling-Estes (eds.), 349-372. Oxford: Blackwell. Chirrey, Deborah 1999 Edinburgh: descriptive material. In Urban Voices, P. Foulkes and G. Docherty (eds.), 223-229. London: Arnold. Keene, Derek 2000 Metropolitan values: Migration, mobility and cultural norms, London 1100-1700. In The Development of Standard English, 1300-1800, L. Wright (ed.), 93-114. Cambridge: Cambridge University Press. Kerswill, Paul, and Ann Williams 2002 "Salience" as an explanatory factor in language change. In Language change: The Interplay of Internal, External and Extra-Linguistic Factors, Mari C. Jones and Edith Esch (eds.), 81-110. Berlin/New York: Mouton de Gruyter. Labov, William 1972 Sociolinguistic Patterns. Philadelphia: Pennsylvania University Press. 1994 Principles of Linguistic Change, Volume 1: Internal Factors. Oxford: Blackwell. Lass, Roger 1997 Historical Linguistics and Language Change. Cambridge: Cambridge University Press. 1999 Phonology and morphology. In The Cambridge History of the English Language, Volume III 1476-1776, R. Lass (ed.), 56-186. Cambridge: Cambridge University Press. Milroy, James 1992 Linguistic Variation and Change. Oxford: Blackwell.

An essay in historical sociolinguistics?


Milroy, James, and John Harris 1980 When is a merger not a merger? The MEAT/MATE problem in a present-day English vernacular. English World-Wide 1 (2): 199-210. Milroy, Lesley 2002 Mobility, contact and language change: Working with contemporary speech communities. Journal of Sociolinguistics 6 (1): 3-15. Mugglestone, Lynda 1995 Talking Proper: The Rise of Accent as a Social Symbol. Oxford: Clarendon Press. Nevalainen, Terttu 1999 Making the best use of "bad" data: Evidence for sociolinguistic variation in Early Modern English. Neophilologische Mitteilungen 4 (C): 499-533. Stuart-Smith, Jane 1999 Glasgow: Accent and voice quality. In Urban Voices, P. Foulkes and G. Docherty (eds.), 203-222. London: Arnold. Tottie, Gunnel, and Dawn Harvie 2000 It's all relative: Relativization strategies in early African American English. In The English History of African American English, S. Poplack (ed.), 198-230. Oxford: Blackwell. Trudgill, Peter 1996 Language contact and inherent variability: The absence of hypercorrection in East Anglian present tense forms. In Speech Past and Present: Studies in English Dialectology in Memory of Ossi Ihalainen, J. Klemola, M. Kyto and M. Rissanen (eds), 412-445. Frankfurt am Main/New York: P. Lang.

A brief response Donka Minkova

Lesley Milroy's comments are not just kind (for which I am very grateful), but also very elegant and enriching. A variability approach to the history of [hw]~[w] provides a more realistic account of the change than an all-ornothing structural merger and unmerger. The suggested distinction between citation forms and rhymes versus casual speech forms is entirely apposite. Placing the change in the wider context of northern forms infiltrating the south in the sixteenth and the seventeenth century (-es vs. -eth, ARE vs. BE) is an excellent addition to the sociolinguistic characterization of the period. The parallel between the change under consideration and observed modern patterns of social re-evaluation (Irl in New York City) adds another valuable theoretical dimension to the account.

Why we should not believe in short diphthongs David L. White

1. Introduction The word philology seems to mean different things to different people. To some, the difference between philology and linguistics is a difference between fact and theory, or even (to resort to extremist caricatures) between a presumably a-theoretical empiricism and a dangerously a-factual rationalism, between having one's feet on the ground (and nothing more), or one's head in the clouds (and nothing more). Closely related to this is a tendency to associate philology with specialization and linguistics with generalization. To others, the difference between philology and linguistics is in large part a matter of diachrony versus synchrony. But in practice, we may say that philologists value familiarity (however indirect) with texts, and that much of the difference between philology and linguistics is between "text-based" and "speech-based" approaches to language. Minkova (this volume) effectively shows how texts can (in certain circumstances) serve almost as transcriptions of speech, so that the theoretical gap between text-based and speech-based approaches, or perhaps even between philology and linguistics in a larger sense, can be somewhat narrowed. This is in some sense a (wisely) limited movement back toward an older view. Around 1917, James Bright wrote ([1917] 1953: viii) that "The spelling of Anglo-Saxon before Alfred's reign, and to some extent after it, approached a phonetic transcription of the actual speech of the times." In other words, to put words to a view that has been widely assumed at one time or another, especially by older philologists, "in those days people spelled as they spoke." That would certainly be nice, not least by freeing us from the difficult and doubtful labor of having to actually interpret spelling. But it is important not to get carried away, for wishful thinking, the desire for greater certainty than can actually (so easily) be achieved, or even a sort of romanticism, may lead us to regard spelling as closer to speech than it actually was.


David L. White

How would the idea that "in those days, people spelled as they spoke" be applied to ecg, for example? Yet is only the most obvious of the problematic cases encountered in Old English (OE) spelling. It is generally acknowledged these days that is composed not of /e/ and /a/, but rather of /as/ and a somewhat obscure "something else." The distribution of "thorn" versus "edh" is neither phonemically nor phonetically rational, and upon examination is not in contrast with , which is a little disturbing if contrast was meant to be conveyed. My concern here, however, is not with these cases, however interesting they may be, but rather with the very vexed matter of the short diphthongs. When we ask why it has always been traditional to believe that Old English has short diphthongs, the answer is basically that Old English looks like it has short diphthongs. Or rather, that Old English looks like it has short diphthongs to an observer who assumes that vowels must "naturally" be grouped with vowels, so that Old English beorn must represent [beorn] rather than /beorn/. (Underlining will be used throughout to group sounds that somehow manage to count as one sound.) But though this analysis might seem quite obviously valid to a traditional English philologist, to a traditional Irish philologist it is just as obviously valid that Old Irish cuid must represent not /kuid/ but rather [kuid], with: 1) no short diphthong; and 2) a palatalized /d/ that is spelled as , in part because it is realized phonetically (in this context) as something like [id]. Here we have a problem, for two diametrically different interpretations, dependent ultimately on arbitrary preconceptions instilled by the educational experience of any given observer, cannot both be "obviously" valid. So far, it may seem that the "spelling as transcription" model is perfectly adequate to handle matters, for regardless of implicit groupings and subtleties of timing, is a respectable phonetic spelling of [kuid]. But though such spellings may be phonetically accurate, they violate nativespeaker intuitions, for Irish grammatical tradition has always recognized that the graphic short diphthongs (to English eyes) of Irish are intended to convey short vowels followed by palatalized or velarized consonants. A phonemic transcription of cuid would in fact have no "i" in it, for there simply is no Iii in Old Irish cuid, only a /u/ and a palatalized Id/ that necessarily has an [i] as part of its phonetic realization. Therefore is not accurate phonemically, and though in some cases the Irish may have spelled as they spoke, i.e., phonetically, having no better solution, they certainly did not spell as they thought, i.e., phonemically, which is what really matters. Short diphthongs have never been part of the traditional

Why we should not believe in short diphthongs


interpretation of Irish spelling, and the Irish interpretation of spellings like is pretty clearly correct for Irish. This does not in itself mean that the Irish interpretation must be correct for Old English, but it is a fair question, given that the two spelling systems were connected in space and time. The Irish were quite probably, despite the somewhat propagandistic impression created by Bede, the primary converters and "literacizers" of the Anglo-Saxons, and thus also the original esta-blishers of OE writing and presumably therefore of OE spelling. There is, after all, a reason why Old English is written in the Irish hand, not the Ro-man hand, a fact which would seem to put the burden of proof on any who would claim that OE writing and spelling were not established under the tutelage of Irish missionary linguists. Should we not wonder then whether the Old English spelling system was originally meant to be interpreted as Irish eyes would see it? To suggest this is of course to revisit the Daunt Hypothesis (Daunt 1939), generally (but wrongly, in my view) regarded as discredited. The issue as a whole is far too vast to be treated adequately within the compass of a single article, but one sub-issue that various un-believers (including Stockwell and Barrit 1951 as well as Daunt) have brought up from time to time can be treated here: the issue of whether short diphthongs are, as far as we can judge from other cases not in dispute, possible phonemes. We have just seen that how English or Irish philologists would answer this question might depend on whether they were English or Irish, which is surely less than ideal. But how would a linguist, operating from a more general and theoretical perspective, answer this question? It is of course difficult to prove a negative, but that does not mean that "anything goes," or that we should believe everything we read. The conclusion of this particular linguist (who is no despiser of philologists) is that short diphthongs, which even believers would have to admit are rare, are beyond rare, for upon critical examination short diphthongs appear to be non-existent. It must be stressed at the outset that the present essay is necessarily restricted to the issue of whether short diphthongs are known to exist, or should from general theoretical considerations be expected to exist, in other languages, living or dead, apart from the disputed case of Old English. The present article is therefore most definitely not an attempt to provide a definitive resolution of the digraph controversy, which would necessarily include an alternative intepretation of what spellings like "beam" vs. "baern" should be taken to mean, if the first does not have a short diphthong. For an attempt at a book-length treatment readers may see the au-


David L. White

thor's dissertation (White 2000), though the author warns that he has changed his mind about a thing or two since then. The present article, given its conclusion, is a step toward overthrowing the conventional wisdom, but it is only a step. Before we can know whether short diphthongs are possible phonemes, we must know what they would be if they were. First it must be understood that only phonemic short diphthongs are of interest. Thus in what follows "short diphthongs" will mean phonemic short diphthongs unless otherwise noted. (To those who do not accept the distinction between the phonemic and the phonetic, I have nothing to say, except that they should.) Phonetic short diphthongs, or sequences that might plausibly be called that, are another matter, and there can be no doubt that such things exist, as has just been seen in the case of Irish. Nearly all the vowels of modern English, as it happens, are phonetically diphthongized, and all English vowels are shorter before voiceless consonants. Putting these two together, we might say that the vowel of at, for example, is phonetically a short diphthong compared to the vowel of add. Yet there can be little doubt that the vowels of at and add are phonemically monophthongs. Turning now to true phonemic diphthongs, the vowel of right is short phonetically as compared to the vowel of ride. Again, we might think that the vowel of right qualifies as a short diphthong. Phonetically, perhaps it does, but not phonemically. The shortening in question is predictable and not conceivably contrastive, coming from the following consonant rather than being an inherent property of the vowel. Therefore the vowel, though it is surely a diphthong phonemically and is in a sense short phonetically, is not a short diphthong phonemically. For better or worse, phonemic short diphthongs are not something we can simply read off a spectrogram. More indirect methods must be used: we must know whether there is a possibility of contrast. If all is in order, phonemically short diphthongs should be contrastively short diphthongs, and it is the existence of these that, on examination, appears to be problematic. Before we can know what a short or monomoraic diphthong would be, we must know what a normal "long" or bimoraic diphthong is. It is a vowel composed of two (non-identical) vowels, not divided by a syllable boundary, each being counted as a mora. Modern English /ai/ is typical, and may be symbolized like this:

μ μ I I a i

Why we should not believe in short diphthongs


This sort of thing, though it may perhaps look exotic to those who are not used to it, is entirely normal, for under ordinary circumstances vowels are always moras. By contrast, a short diphthong would be two consecutive vowels that somehow manage to count as one mora. Short /ai/1 might be symbolized like this: μ /\ a i This is the sort of symbolism or conceptualization proposed by Lass (1994: 47). There are cases where two consecutive phones non-controversially count as one mora and as what will be called a unitary phoneme. For example, the final phoneme of the word church might be symbolized as /t|/: μ Λ t s because there is of course good reason to believe that it is a unitary phoneme, not a sequence of Ν and /§/, which would be two moras. But this mode of representing unitary phonemes has a serious liability: it does not allow us to capture the fact that such unitary phonemes can also exist in the onset of a syllable, where moras cannot provide us with any way to capture the difference between a unitary phoneme and a cluster. In other words, using moras alone as our representational tool, the initial phoneme of church, which is not moraic, would have to be represented as a sequence of Ν plus /s/. But to natives the first sequence in this word is just as much a unitary phoneme as the last, so such an analysis would be pretty clearly wrong. Therefore, it seems we have little choice but to regard Ν and /s/ in such cases as somehow linked without the aid of moras. The representational device that will be used here is underlining, so that "ch" would be /ts/, in both the initial and final cases, doubly-linked moras having nothing to do with anything. Following this sort of convention, short /ai/ would then be symbolized simply as /ai/, so that explicit moras and tiered representations can be dispensed with, as long as we recognize that /ai/ would still be, or rather be


David L. White

attached to, a mora. This mode of representation gets us out of the anomaly of having to posit that moras can sometimes be doubly linked, at the price of merely pushing the anomaly down one level, for now instead of wondering under what circumstances moras can be doubly linked we must wonder un-der what circumstances two consecutive phones can be linked as a unitary phoneme. Does anything go? Is /gs/ a possible unitary phoneme? What about /ap/? Why not just link any two consecutive segments whenever the result happens to be arguably elegant? But if no one believes that /ap/ is a possible unitary phoneme (as far as I know no one does), then everyone believes that there are some limitations. The question is exactly what these limitations are.

2. The theoretical case Strangely it seems that this matter, at least as framed above, has not attracted any attention, perhaps because it falls between the two stools of phonetics and phonology. Though the question might be considered one of phonology rather than of phonetics, most discussion of what types of unitary phonemes occur is in fact by phoneticians. There is a considerable difference of opinion between the two groups, for the general rule is that phoneticians regard sequences as unitary or linked only when they are physically linked, which is to say homorganic. Of course this is reasonable from their point of view, as phoneticians have no reason to regard sounds as linked unless they are physically linked. It will be argued below that this is not only reasonable from a narrow phonetic point of view but objectively correct. According to the phoneticians, unitary phonemes are not at all uncommon in the world's languages. Probably about half have at least one unitary phoneme, most of these being affricates, as in English. One may see, for example, the treatments in Ladefoged and Maddieson (1996: 118-131, 328-369), Laver (1994: 227-235, 256, 358-376), and Catford (1988: 1 ΙΟΙ 14). But upon examination, only a few well-defined types recur in the phonetic literature, and short diphthongs are conspicuously not among them. Apart from affricates, which are plosives with homorganic fricative egress, we find pre-affricates (plosives with homorganic fricative ingress), 2 pre-na-salized plosives (with homorganic nasal or egress), post-nasalized plosives (with homorganic nasal ingress), pre-occluded nasals (with homorganic plosive ingress), post-occluded nasals (with homorganic plosive

Why we should not believe in short diphthongs


egress), plosives with homorganic lateral egress, and plosives with homorganic rhotic egress. As if these were not enough, the different types of ingress and egress can freely co-occur, so that we also find things like prenasalized affricates, 3 and even more exotic entities. That the word "homorganic" recurs here is not a coincidence, for that it is what all the types recognized by phoneticians have in common: they all involve homorganic ingress to or egress from a closure of some sort.4 It is in a sense just a coincidence that any given homorganic ingress or egress may happen to be identifiable with a possible phoneme like /§/ or /n/. Whether the closure condition is important or not is unclear, but the phoneticians agree in repeatedly noting what might be called the homorganic condition, even if this might be considered tautological. But if this is the condition that limits the formation of unitary phonemes, then short diphthongs do not meet it, and therefore should not be possible phonemes. Phonologists for their part tend to regard sequences as unitary or linked whenever this results in a more elegant analysis. Of course it is possible to take the attitude that might be expressed as "Phonetics is phonetics and phonology is phonology, and never the twain shall meet," but it seems that few phoneticians or phonologists would actively recommend such a view these days. If phonetics and phonology do not lead entirely separate lives, then it is quite possible that there is or can be, at least in some cases, some actual empirical phonetic evidence on what is primarily a phonological question: when consecutive phones can be linked to form a unitary phoneme. In fact there appears to be some, and it supports the phoneticians' view. Ladefoged and Maddieson (1996: 331-332, 354-355) discuss the case of a language called SePedi (Northern Sotho), which has been said (Lombard 1985) to have various unitary phonemes consisting of a labial obstruents followed by sibilants, e.g. /gs/, on the grounds that such an analysis makes the phonology come out more elegant by making all syllables open. But reality is not necessarily elegant. Prospective unitary /p|/ does not meet the homorganic condition, and therefore if the phoneticians are right it should not occur. So what is the evidence? It is in fact quite clear from spectrographic evidence that SePedi [ps] takes as much time as would be expected for two fully independent phonemes in this language, not one.5 Speakers make no effort to compress it into the space that a single unitary phoneme should occupy, especially given that a prevocalic unitary phoneme would have to be non-moraic, and therefore could not be too long. The [p] in question is in fact notably prolonged, compared to the true


David L. White

non-moraic SePedi [ρ] illustrated right beside it (Ladefoged and Maddieson 1996: 354). Therefore it must be moraic, and the preceding syllable is closed, regardless of elegance. Furthermore, the fricative portions of the supposed unitary phonemes are suspiciously long in duration, given that in true unitary phonemes with fricative egress the fricative portion, being merely a mode of egress and not an independent phoneme, tends to be notably short (Laver 1994: 365). 6 Therefore SePedi [ps] is in fact a /p/ followed by an /§/, not a unitary /ps/, if facts are relevant. If facts are not relevant, then the hypothesis that it is a unitary phoneme is unfalsifiable. What is interesting is that SePedi [ps] and similar sequences are evidently not analyzed by native-users as unitary phonemes, despite the evidence of patterning which would have supported such an analysis, evidence that was enough to mislead at least one observer. The evident conclusion is that /gs/ is not a possible unitary phoneme. The reason why has just been seen: prospective unitary /ps/ does not involve homorganic ingress to or egress from a closure. In more mundane terms, the [s] cannot be taken as a phonetically natural way of getting out of a [p], so it cannot be phonemically linked with [p] in unitary /ps/, and must be taken as an independent phoneme. In other words, since the two cannot be phonetically linked, they cannot be phonemically linked. In consquence, the [p] and [s] must both be taken as independent phonemes /p/ and Is/, and must therefore (in this language) be timed out accordingly. That is what we see in the spectrographic evidence. Thus it seems that the phoneticians are right: it is not possible for just any two consecutive phones to be a single unitary phoneme. The problem in all this for the traditional interpretation of Old English spelling should be clear: the components of short diphthongs can by no means be seen as homorganic. In the supposed short diphthong /eo/, [o] is not homorganic with [e], and not a phonetically natural egress from [e], (Since [e] does not involve closure, the concept of egress is inapplicable.) If /ts/ and /eo/ are possible unitary phonemes, then why is it that /ps/ evidently is not? Surely /ps/ is more like /ts/ than like /eo/. What non-arbitrary standard will predict that /ps/ should be out if /eo/ is in? Above all, if the limit for possible unitary phonemes is not what has been posited above, then what is it instead? If there is not a natural limit, then believers in short diphthongs would seem to have little choice but to fall back on the idea that things like short /eo/ are arbitrarily included in some "Universal" list of All Possible Unitary Phonemes (the APUP, no doubt) that "just happens to" exclude things like /ps/. Maybe, but such very abstract propositions

Why we should not believe in short diphthongs


should be a last resort, forced upon us by the evidence, not a first one, and it will be seen below (section 3) that such evidence appears to be lacking. There are some other phoneme types that might be called unitary, but which involve phones that are much more nearly simultaneous than consecutive in timing. The first type is what Ladefoged and Maddieson call "doubly articulated" sounds, as in the unitary /kp/ and so on of many African languages (Ladefoged and Maddieson 1996: 332-354). Phonemes of this sort involve articulations that are (as much as is possible) truly simultaneous, as is made clear by spectrographic evidence. The articulation of the two segments to a great extent overlaps in time, as it must for the two to come out about being about as long as a normal single segment, which is what happens. It is not as if in some African languages independent consecutive /k/ and /p/ are opportunistically reanalyzed as unitary /kp/, just because some abstract consideration in the grammar would make this more elegant. Unitary /kp/ is truly a different sound, and phoneme, from a sequence /kp/.7 But returning now to putative short I col, an [e] and [o] with an overlapping articulation as nearly simultaneous as possible would be /Ö/, not /eo/, so short diphthongs do not qualify as possible unitary phonemes by this stan-dard either. The second type is secondary articulations, such as palatalization and velarization (Ladefoged and Maddieson 1996: 354-368; Laver 1994: 320333; Catford 1988: 105-110), which are considered just below in another connection. These are consonants in which the position of the tongue body and to some extent of the lips, is fixed, rather than being allowed to float freely in accordance with the demands of coarticulation, as is the usual case in languages without secondary articulations. Thus it may be possible, depending on the language in question, to make a back /k/ before a following front vowel, etc. In more technical terms, such phonemes involve a primary consonantal articulation and a secondary vocalic articulation, which is in general made to be roughly simultaneous with its associated consonant.8 In abstract terms the consonants so produced are like doubly articulated consonants, in that the two articulations involved are more nearly simultaneous than consecutive, so again secondarily articulated consonants do not qualify as analogues to the posited short diphthongs of Old English, as the nearest analogue would be not /eo/ but /ö/. To sum up, where consecutive phones are linked to form a true unitary phoneme, they are, as far as we can tell, always homorganic, and therefore are not like short diphthongs. In both double articulations and secondary articulations the two sounds are not a true sequence, and therefore are not


David L. White

like short diphthongs. Neither type can be presented as analogous to the posited short diphthongs of Old English, as these would be nonhomorganic phones in sequence. From theoretical considerations, short diphthongs do not seem to qualify as possible unitary phonemes.

3. Short diphthongs versus secondary articulations Yet another reason we should be reluctant to believe in short diphthongs is that it is probably not possible for short diphthongs and secondary articulations to exist in the same linguistic world, as the two would in many situations be phonetically indistinguishable. This has much to do with why Irish spells secondary articulations using vowels, presenting the impression of diphthongs, which is not phonetically inappropriate. Yet it is not controversial that secondary articulations do exist. Palatalization is typically realized as a roughly simultaneous high front sound, but such a sound cannot (generally) be heard during the consonant, so practically speaking the high-front sound must be heard before or after a palatalized consonant. In the before case, notably with final palatalized consonants, the primary vowel and the transitional glide together might reasonably be called a short diphthong phonetically, though this is to some extent an arbitary and wrong division. How this works can perhaps best be seen by considering a certain historical development in Irish. In what follows, palatalized consonants and front vowels will be indicated by italics. Old Irish Raid/, the "cuid" used as an example word in the introduction, had velarized ikJ and palatalized Idl. These days it has become /kid/ (MacEoin 1993: 108). The consonants have stayed the same, but the vowel has changed, taking on something of the nature of the following palatalized consonant. (Something like an inverse of this process may be seen in the common pronuncation of "porcupine" with /-kip-/.) For this to have happened, the original pronunciation, as was noted at the outset, must have been something like [kuid],9 with palatalization of moraic Idl being realized (in combination with the preceding vowel) as a preceding phonetically short diphthong with a rather short [-i]. But as MacEoin says (1993: 106), "... the balance between vowel and glide has become upset, so that the former glide is now the principal vowel." Using capitals to mark prominence, [kUid] became something more like [kuld], and this triggered reanalysis of /kud! as /kid/.

Why we should not believe in short diphthongs


The question for those who would have us believe in short diphthongs is this: why was [kuid] not simply taken all along as /kuid/, if short diphthongs are possible? Old Irish even had normal long /ui/ for putative short /ui/ to contrast with (Thurneysen 1946: 43), and so might be said to have had diphthongs differing only in length, phonetically. This is, as exactly as could be desired, the situation that Hogg says (1992: 17) led to the phonemicization of short diphthongs in Old English. Why then did it not lead to a parallel phonemicization of short diphthongs in Old Irish? Are we to believe that matters could have gone either way, with the equivalent of a linguistic coin toss determing the result? Essentially the same argument, using a living language, can be made from Estonian. This language has both palatalized consonants, which when moraic are implemented in more or less the Irish manner, through what might be called preceding short diphthongs (Robert Harms, personal communication), and diphthongs in /-i/ of the normal bi-moraic type. An example is short [ai] in [pain], phonemically /pa«/, versus long /ai/ in [painin], phonemically /painin/. As with earlier Irish, Estonian might be said to have diphthongs that contrast only in length phonetically, and again we must wonder why phonemicization of short diphthongs has not resulted, if Hogg's view of the parallel case in Old English is correct. The arguments made above apply only to moraic consonants, and it might be objected that perhaps in such cases what we have is not (for example) palatalized consonants but rather short diphthongs in short /-i/. This would be in effect to assert that the traditional analysis of Irish and Estoninan as having palatalized consonants is wrong, which would surely be news to specialists in these languages. Such an analysis does not work well with palatalized consonants that are final in some forms but syllable-initial in others, in such cases typically carrying palatalization over into the vowel of following syllable, where it can hardly be regarded as a short /-i/ in the preceding syllable. We could perhaps save such an analysis, in languages without initial palatalized consonants, by positing rather complex metathetic alternations with in effect a "jumping" /i/, or perhaps two /i/s, one of which may be deleted when not needed, but since the existence of palatalized consonants must be admitted in other languages, such a move would verge upon perverse. To sum up, if both secondary articulations and short diphthongs existed, it would in many cases be difficult for learners to tell the difference between them, as secondary articulations can often wind up in effect implemented as short diphthongs phonetically. This should create quite a


David L. White

learning problem, and we should expect to find differing opinions and divergent developments. For example, we might expect to find that some dialects, or idiolects, of Irish and of Estonian have short diphthongs in /-i/ and others have palatalization (assuming we could even tell the difference), or perhaps that Irish developed one way and Estonian another. Yet the diachronic and synchronic evidence indicates that real people, as opposed to some linguists, have no trouble at all with this issue: the theoretically ambiguous cases have been taken as phonemic secondary articulations, not as phonemic short diphthongs. Thus the situation that probably would be seen if both short diphthongs and secondary articulations were possible is not seen, yet it is not controversial that secondary articulations are possible. The obvious conclusion is that short diphthongs are not possible.

4. The empirical case That would certainly explain why it is that, so far as I have been able to determine, short diphthongs are not known to exist in any language, living or dead. Several languages, many of them conveniently dead, are sometimes trotted out as languages "known" to have short diphthongs, but upon critical examination, not one of these claims appears to be valid, from the evidence that has been produced at this point.

4.1. Icelandic Icelandic is often said to have short diphthongs, but the short diphthongs of Icelandic are phonetically short, not phonemically short. In North Germanic generally (save Danish) vowels are phonetically lengthened in stressed open syllables, or, looking at it the other way, shortened in stressed closed syllables.10 In words of more than one syllable, there are two syllable types phonetically: [VV] and [VC], which practically speaking means that mere [V] is not a possible stressed syllable. (Nor is [VVC], Words of one syllable permit one more consonantal mora in either case.) This is abstractly similar to what happens in English with shortening before voiceless consonants, the major difference being that what causes shortening in North Germanic is long consonants. Since Icelandic has phonemic diphthongs, we accordingly find phonetically short diphthongs in Icelandic. Shortness, either of diphthongs or monophthongs, is predictable

Why we should not believe in short diphthongs


in both cases, and therefore is merely phonetic, not phonemic. It is not even conceivably contrastive. This is most certainly not the state of affairs that is asserted for Old English, so Icelandic cannot rightly be presented as an analogue for Old English as traditionally interpreted. But the situation in Icelandic deserves more consideration, because there is another indirect argument that Icelandic has phonemic short diphthongs, though all observers admit that there is no possible contrast. For monophthongs it is fairly clear that we should posit a phonetic lengthening rule for short vowels in open syllables rather than a phonetic shortening rule for long vowels in closed syllables, as it would be odd to posit a language with long vowels but no short vowels. If Icelandic diphthongs are to be treated as parallel with Icelandic monophthongs, then they too must be regarded as phonemically short. But that is a big if. It is quite possible that Icelandic simply has two relevant rules of phonetic implementation: a lengthening rule for monophthongs in open syllables and a shortening rule for diphthongs in closed syllables. This give us one more rule, but though it is true that "all things being equal" rules should be minimized, in this case the "all things being equal" condition, as so often happens, is not met. Since in practical terms "phonemic" and "contrastive" mean much the same thing, being in a sense two sides of one coin, it surely verges on incoherent to assert that Icelandic has short diphthongs that are phonemic but not even conceivably contrastive. Refusing to posit even one more rule for one language would make nonsense of the phonemic principle. Languagespecific considerations, such as getting a maximally minimalist interpretation of Icelandic, must yield to language-general considerations, such as our understanding of the phonemic principle. For Old Icelandic, Stanley (1952: 29) argues for short diphthongs of a different sort, asserting that attested /ja/ from original /e/ was short /ea/ at some intermediate point. Stanley asserts that this diphthong must have been falling rather than rising. But rising /ea/ and /oa/ with non-moraic or in effect consonantal /e/ and /o/ (in effect more open versions of /j/ and /w/), which Stanley seems to have regarded as impossible, are in fact known from Rumanian (Mallinson 1988: 392), and as the Icelandic sequence must have become rising at some point in order to have become /ja/, there is no reason to think that Icelandic /ea/ could not have been rising with non-moraic consonantal /e/ as in Rumanian. Such an /ea/ with consonantal /e/ would be no more a short diphthong than was later /ja/, which is in effect /ia/ with consonantal Iii. Furthermore, Stanley's evidence that Icelandic had /e-/ rather than /y-/ in such cases is insecure. The early


David L. White

spellings that he points to could have been influenced by the spelling of English, where is used though or are not, and furthermore the testimony of "the First Grammarian," which is his source, could have been influenced either by these earlier spellings or by the fact that Icelandic /ja/ is typically related synchronically to Id.

4.2. Overlongs versus longs In Stockwell and Barritt's original article (1951), which was in part intended to question whether short dipthongs are possible, the question was framed as being whether contrast of length in diphthongs is possible. Unfortunately this is not the same question, and a great deal of confusion has resulted from the casual assumption that it is. Diphthongs that are shorter than some other diphthongs are not necessarily short in the sense of having only one mora, and contrast of length in diphthongs is certainly and quite non-controversially possible: between normal long diphthongs of two moras and "over-long" diphthongs of three moras. Homeric Greek (at least as traditionally interpreted) is an example of a language that does this. Afrikaans and Northern Welsh are presented by Lass (1988: 225; 1994: 46) as being languages that, because they show contrast of length in diphthongs, by implication are "known" to have short diphthongs, presumably rendering these totally unproblematic in Old English. But shorter diphthongs are not necessarily short, and upon examination absolutely no mono-moraic or short diphthongs are alledged by specialists in these languages. Rather, the situation in each appears to be much as in Homeric Greek. For Afrikaans, Ponelis (1993: 131) gives its diphthongs as /iu/, /ui/, /o:i/, /a:i/, /öi/, /ei/, and /ou/. Not a word about short or monomoraic diphthongs is said, and on any normal reading /o:i/ and /a:i/ are intended as trimoraic or over-long. Likewise for Northern Welsh, Thorne (1993: 6) says nothing about short diphthongs, but speaks only of the first elements of certain diphthongs being long, which again implies that the diphthongs as a whole are overlong. What these languages show is something that (outside of the small world of English philology) has never been in doubt: that contrast between long and over-long diphthongs is quite possible, if perhaps a bit rare. But contrast between diphthongs of three moras and diphthongs of two moras does not and cannot possibly prove the existence of contrastive diphthongs of one mora.

Why we should not believe in short diphthongs


4.3. Scots Certain dialects of PDE, notably Scots, are sometimes presented as having phonemically short diphthongs, just like Old English (Lass 1988: 225). The Scots evidence appears at first to show a situation somewhat similar to what occurs in Icelandic, the main difference being that the environments for shortening/lengthening are differently (and less elegantly) defined, in a way that need not detain us. But in Scots (as opposed to Icelandic) there are some minimal pairs, or possible minimal pairs, which raises the possibility that Scots truly does have phonemically short diphthongs. The Scots evidence involves things like tide [teid] vs. tied [taed]. Here the evidence given by Lass has been modified somewhat in order to make his case seem stronger than it does in his versions, for his evidence (in my symbology) is [teid] vs. [taaed], seemingly showing an over-long situation as in Afrikaans and Northern Welsh, which would prove nothing. Now if a Scot can tell the difference between tide and tied in isolation, without the benefit of any contextual or grammatical information about the presence of morpheme boundaries, then (to my mind) there must be a phonemic difference of some sort, probably at least roughly parallel to the difference in monophthongal cases like brood [brud] vs. brewed [bruud]. (The centralization of Scots /u/ will be ignored here.) It would seem reasonable to posit that [brud] is from /brud/ and [bruud] from /bruud/, this last derived somehow from analogically expected /bru-ed/. If the situation with tide [teid] and tied [taed] is similar, then [teid] is from /taid/, with predictable shortening much as in Icelandic (and some predictable qualitative changes as well), and [taed] is from /tai-ed/, with three moras phonemically but only two phonetically, the result again of a merely phonetic shortening quite parallel to what happens in the case of tide. (That phonetic shortenings of this sort can occur without phonemic significance has been seen above in connection with Icelandic.) That the vowel of five, from /faiv/, unshortened before /v/, which is not a shortening phoneme in Scots, comes out as [ae], like the vowel of tied from /tai-ed/, is just a garden-variety neutralization, and does not indicate that in other environments there is no contrast between /ai/ and /ai-e/. In any event, Lass cannot have it both ways, leaving the impresssion that, because of the morpheme boundary brewed, brood and brewed are not minimal pairs (Lass 1973: 319; 1984: 32-33), but that tide and tied are minimal pairs, thus supposedly proving the existence of contrastive and


David L. White

therefore phonemic short diphthongs, despite the morpheme boundary in tied. The Scots evidence appears to show a situation that combines aspects of what is seen in Icelandic (phonetic shortening) with what is seen in Afrikaans and Northern Welsh (over-long diphthongs). There is nothing in either of these (or their combination) to force the conclusion that Scots has phonemically short diphthongs. New York English is another non-standard dialect of English that is perhaps alledged (Lass 1988: 225) to have short diphthongs. It is quite unclear what Lass meant here, as the article on New York English he refers to (Lass 1981) does not in fact directly claim, or give evidence, that New York English has phonemically short diphthongs, though that is one interpretation a reader might perhaps come up with. But just to be on the safe side, the case of New York English will be briefly treated here. New York English has two distinct kinds of /ai/ (Lass 1981: 526), so that (among other things) I and eye are not homophones. No attempt will be made to reproduce Lass's symbology here, but it is roughly as if New York I has [ei], somewhat comparable the vowel of Scots tide, while New York eye has [ai], somewhat comparable to the vowel Scots tied. The evidence appears to show diphthongs distinguished primarily by quality and only secondarily by quantity. There is nothing in this to make us think that the phonemes could not simply be /ei/ and /ai/ (or something not far from it) rather than /ai/ and /ai/. It is true that the [e] of the first diphthong is shorter than than the [a] of the second diphthong, but it is normal in English for /e/ to be shorter than /a/. There is therefore no reason to think that New York English has been shown to have phonemically short diphthongs. Overall, the casual confidence of Lass's various assertions about the supposed existence of short diphthongs in non-standard English dialects seems to come fundamentally from a belief that short diphthongs are "known" to exist in Afrikaans and Northern Welsh, and therefore that their existence in other languages would be entirely unproblematic. But this is a linguistic or logical error: the existence of over-long diphthongs certainly does not prove the existence of short diphthongs. Once this is realized, his evidence supposedly proving the existence of short diphthongs in nonstandard Englishes does not stand on its own merits, and cannot be accepted as valid at this point, though perhaps it could be significantly elaborated and clarified at some future point. Lass claims (1988: 225) that "[w]e can probably now take universalist arguments against the possibility of two diphthong types as untenable." We certainly can, but that is a rhetorical straw man, however innocently and accidentally created. In context, Lass's

Why we should not believe in short diphthongs


assertion is clearly meant to convey that "universalist" arguments against short diphthongs are untenable. But this is a different proposition entirely, and I do hereby officially dispute it.

4.4. Old Irish Short diphthongs in /-u/ have been posited (Greene 1962; McCone 1996: 27) for Old Irish, apparently in part from the example of such things being "known" from Old English. But the analogizing here is going in the wrong direction. The starting point for Greene's theory is an objection to the idea that Old Irish had three series of secondarily articulated consonants, labiovelarized, "neutral," and palatalized, plus a full series of vowels and an unusually large number of normal bi-moraic diphthongs. This objection is in my view quite valid, as such a system indeed appears to be unattested in living languages, and would surely be difficult or even impossible to implement phonetically. Greene's idea is to replace the third series of labiovelarized consonants with short diphthongs in 1-uJ. Thus "fiur" would represent not /fir/ with labiovelarized Irl, as is traditionally posited, but rather /fiur/ with short /iu/. Unfortunately such a sound-system, with two secondary articulations and a series of short diphthongs, would be no more attested or easy to implement phonetically than the one it is intended to replace." The suggested improvement is not an improvement. The real problem, and real solution, are pointed to in McCone's observation (1996: 27) that "... there is no good empirical evidence for either of these two alleged extra phonemic contrasts [labiovelarized versus neutral and labiovelarized versus patatalized] in the synchrony of Old Irish." In other words, there is no good evidence that the problematic spellings in indicated any contrast at all in Old Irish. Upon examination, the supposed labiovelarized quality, or rather its appearance in spelling, turns out to be morphologically predictable. This is a problem for the traditional interpretation of Old Irish spelling, but quite a different problem than the one Greene and McCone set out to solve, for there is obviously no point in worrying about how a contrast was implemented if upon critical examination the contrast in question did not exist. Thus it seems best to accept that spellings with like "fiur" were an alternative (however mysterious in origin) for spelling velarized consonants in certain morphological contexts. In the end, there is and can be no good evidence that Old Irish had contras-


David L. White

tively short diphthongs, since there is no good evidence that spellings with "-u" indicated a contrast of any sort.

4.5. Ancient Greek Another language that might be thought to have had short diphthongs is Ancient Greek. The claim that Ancient Greek had short diphthongs is reluctantly accepted by Stockwell (1994: 62), apparently on the basis of some personal communications, but he had it right the first time. Evidence superficially supporting the existence of short diphthongs may be seen in any traditional Greek grammar. Pharr ([1959] 1985: 182) will serve: "Final /-ai/ and I-oil are counted [as] short when determining the accent, except in the optative..." From this it might seem that a maximally elegant explanation would be that the diphthongs in question really were short. The claim is almost (it is hard to tell) made by Kiparsky (1969: 124), though in fairness it is (wisely) retracted somewhat in a later article (Kiparsky 1973: 805). But in fact Pharr, like every other traditional authority going back to Hellenistic times, says not that these diphthongs were short but that they were counted as short for purposes of determining the accent. This is a critical distinction. The reason for it becomes evident in Pharr's next sentence: "These diphthongs are regularly long in metrical quantity..." 12 It is not probable that short diphthongs would count as long in poetry. If we are to believe that Greek had short diphthongs, we would have to posit that Greek metrical conventions were archaizing, so that the diphthongs in question, though they had been long in Homeric Greek (or at whatever point the metrical conventions of Greek were established), had become short by the time of Classical Greek. But it is improbable that Classical Greek verse would have gotten rid of many or most of the archaisms of traditional Homeric verse, as it did, but still have counted final short diphthongs as long. Though final diphthongs are sometimes counted as short before a following initial vowel, the same is true of final long vowels like omega (Smyth 1956: 36), which nobody claims were really short. Thus it is not so easy to dismiss the evident length of these diphthongs in metrics as a misleading archaism. Far from it, full length of diphthongs seems to have been, as far as we can tell, part of fully living poetic tradition. But this argument alone is not necessarily decisive. Further reasons to believe that the diphthongs in question were long can be seen in considering the issue of contrast and minimal pairs. Capitals will

Why we should not believe in short diphthongs


be used to indicate stress in what follows. Kiparsky (1969: 124) in effect presents as minimal pairs aorist infinitive [lUusai] vs. aorist optative [luUsai] and nominative plural [Oikoi] versus (fossilized) locative singular [olkoi].13 Kiparsky does not use the word "phonemic," and it is quite unclear whether at that time he accepted the phonemic principle (which was of course rejected by generative phonologists), but the innocent reader could surely be forgiven for thinking that these pairs are intended to be seen as distinguished by phonemic length of final diphthongs, with predictable (and therefore not phonemic) accent on the third mora from the end. But there are problems. First, the putative shortness of the final diphthongs, as opposed to the placement of the accent, is not a fact but part of Kiparsky's theory, as Greek thradition in later days marks accent but not shortness (if any) of diphthongs, so the evidence presented is not really evidence. Second, even if the evidence is valid, we could just as well assert that length of final diphthongs was predictable from the accent, as either one of these two is (in the limited cases seen at this point) predictable from the other. So the question then is which one of these two features is the real phonemic feature, as there is no point in regarding both as phonemic. The answer to this question is quite clear. The accent in Greek as a whole is not predictable from phonological considerations alone, and no amount of phonological sleight of hand, even positing short diphthongs, can make it so, as Kiparsky must have known.14 Abbott and Mansfield (1977: 132-134) list over sixty instances of minimal pairs distinguished by accent alone, with only two involving final diphthongs of disputable length. If that is not enough, Greek has many accent rules that refer to morphological rather than phonlogical information. Some examples are: 1. "monosyllabic nouns of the third declension accent the ultima of the genitive and dative" (Chase and Phillips 1941: 19); 2. "oxytones of the first and second declensions circumflex the genitive and dative of both numbers" (Chase and Phillips 1941: 8); and 3. "the ultima of the genitive plural of the first declension is always circumflexed" (Chase and Phillips 1941: 8). What exactly the various technical terms of Greek grammar mean need not detain us. The point is that Greek accent cannot be predicted without reference to non-phonological considerations, or there would be no need for such rules. Therefore it is better to regard length of diphthongs as predictable from accent than to regard accent as predictable from length of diph-


David L. White

thongs. We have independent evidence that the accent in Greek was phonemic, so there is no reason to believe that shortness of final diphthongs, even if existent, was also phonemic.15 Another problem, worth noting for strict propriety, is Kiparsky's implication that both nominative plurals and aorist infinitives have a phonologically predictable accent of the type that in Greek grammar is called recessive. This is not true, as aorist infinitives and a few other stray verbal forms have not recessive accent but an anomalous accent of their own (Pharr 1985: 251), as can be seen in the accent of aorist infinitive /paidEusai/ versus the accent of truly recessive nominative plural /mElainai/. If these two forms had the same type of phonologically predictable accent, they would be accented the same. Positing that Ancient Greek had short diphthongs is, among other things, part of a program to deny that Greek had any cases of four-mora accent, in order to save the generalization that the accent "cannot" go back more than three moras. A major problem with this program is that, upon critical examination, the generalization which is to be saved, thereby presumably vindicating our undying faith in the Beauty of Truth, turns out to be false. Under any interpretation, at least two compound adjectives, /dUseroos/ 'un-lucky in love' and /hupslkeroos/ 'high-horned', show a four-mora accent (Smyth 1956: 39).16 Furthermore, a word like /mElainai/ is undeniably a case of four- or even five-mora accent, regardless of whether it is seen as in some sense derived from the unacceptable (as a recessive form) */melAinai/. It cannot possibly be presented as a truism that the accent in Greek never goes back more than three moras,17 and there is no point in pretending otherwise. But if four-mora accents existed in Greek, this ruins the entire program of positing short diphthongs in order to deny that they did. For example, /Oikoi/, which might be thought to have a three-mora accent through the miracle of short diphthongs, could just as well have been /Oikoi/, with a mildly anomalous four-mora accent. Practically speaking, the effect of the traditional formulation for recessive accent would be achieved if we restated this as something like "Recessive accent generally goes back three moras, but goes back four moras in nominative plurals and present middles ending in I-oil or /-ai/." (In either case, it goes back one more mora in order to avoid a circumflexed penultima, as in /mElainai/ to avoid */melAinai/.) Kiparsky himself later concedes the existence of four-mora accents in middle forms (1973: 805), and there is no reason that this analysis could not

Why we should not believe in short diphthongs


just as well be extended to nominative plurals. There is absolutely no descriptive need to posit short diphthongs. In the end it is fairly easy to see what is going on: sometimes when a two-mora ending is morphologically associated with a one-mora ending, the two-mora ending has a four-mora accent and the one-mora ending has a three-mora accent. Obviously the effect, and quite probably the indirect intent, of these paired three-mora and four-mora accents is to put the accent in the same place in closely associated forms, often simultaneously rationalizing retention of inherited stress. The two major pairings are: 1) nominative plural /-oi/ and /-ai/18 with nominative singular /-os/ (and sometimes short /-a/); and 2) present middle /-ai/ with imperfect middle /-o/. That is just about all there is to it, and there is nothing in any of it to make us believe that Ancient Greek had phonemically (or even phonetically) short diphthongs. There is another reason to believe that nominative plural /-oi/ and /-ai/ might have been short, for when accented they always bear the simple acute/grave19 accent, and never the circumflex, which only occurs on long vowels and diphthongs. We might theorize then that nominative plural /-oi/ and /-ai/ cannot bear the circumflex because they are short. But the accusative plurals /-ous/ and /-aas/, which all observers accept as long, also cannot bear the circumflex, so there must be something else going on. In general, the nominative and accusative pattern together in Greek accent rules. Practically speaking, the traditional rule that "oxytones [words with final acute/grave accent] of the first and second declensions circumflex the genitive and dative..." might just as well be re-stated as "finally accented words of the first and second declensions have the acute/grave in the nominative and accusative but the circumflex in the genitive and dative." There is no objectively valid reason to regard such acute/graves as morphologically conditioned in the accusative plural (and singular) but as phonologically conditioned in the nominative plural, when they can just as well be seen as morphologically conditioned in both cases. To sum up, Greek had a morphologically conditioned phonemic accent and four-mora accents. It is senseless then to regard what appears to be /Oikoi/, with a morphologically con-ditioned phonemic four-mora accent, as really being /Oikoi/, with a predictable phonetic three-mora accent from final short -/oi/, especially given that the supposed shortness of this is not a fact but a mere guess seemingly contradicted by the evidence of Greek metrics.


David L. White

In the end, it can hardly be stressed too strongly that the traditional formulation of Greek grammarians was primarily intended as an aid to students in marking accent, not as an assertion that Greek really had short diphthongs, for the traditional grammarians were well aware that the diphthongs in question were long in metrics. As it happens, a rule counting some final diphthongs as short for the purposes of accent only, and excluding optatives from the general rule, is marginally simpler for students than a rule giving such diphthongs their true metrical value and excepting nominative plurals and present middles, and also permits the comforting but ultimately illusory generalization that Greek accent "never" goes back more than three moras (except when it does), to be presented as true. These and only these are the reasons for the traditional formulation, which was never meant to be taken seriously beyond the limited purposes it was designed to serve.

4.6. Conclusion Overall, not one of these examples of phonemically short diphthongs being supposedly "well-known" from other languages inspires confidence. To review, in Icelandic the same rule that creates phonetic short diphthongs precludes contrastive phonemic short diphthongs; in Afrikaans, Northern Welsh, and Scots the longer diphthongs involved in a contrast of length are overlong; in Old Irish there is no evidence for contrast of any sort; and in Ancient Greek it is the accent, not the putative shortness of some diphthongs, that is phonemic. As matters now stand, not even one living language with phonemic short diphthongs has ever been found. Dead languages are more difficult to assess, but there is no convincing case for short diphthongs in any dead language either. We must wonder why it is that short diphthongs are so often posited in conveniently dead languages, like Greek and Old English, where mis-analysis must be suspected, rather than simply being found in living languages. If short diphthongs really are possible, should it not be a fairly simple matter to find them in living languages? All sorts of other things that might seem less probable (e.g. pre-nasalized affricates) have in fact been found. Why then should it be so different with short diphthongs? Surely it is reasonable to ask, if contrasts of the sort /taid/ versus /taid/ are really out there somewhere in a living language, that someone should be

Why we should not believe in short diphthongs


able to find them present them to us. Perhaps someone will some day, but (as far as I have been able to determine) no one has yet.



There are theoretical reasons to believe that short diphthongs should not exist. Short diphthongs do not meet the homorganic condition, which, judging by the SePedi evidence and other considerations, seems to serve as a limit on when two consecutive phones can be analyzed as a unitary phoneme. Furthermore, short diphthongs and secondary articulations quite probably could not both exist without there being considerable confusion between the two, but there appears to be no evidence that this occurs. In purely empirical terms, no language, living or dead, has at this point been conclusively shown to have short diphthongs. The obvious question, if Old English did not have short diphthongs, is what it had instead. Old English is (to some extent) spelled as if it had phonemic velarization, like Old Irish (which is not a coincidence), but there is little reason to think that it did. The answer, in my view, was provided long ago by Daunt (1939): the graphic short diphthongs of Old English were used to spell velarization, as perceived by Irish missionary linguists, in a manner that would have seemed appropriate to them. But though velarization was phonemic in Old Irish, it was not phonemic in Old English. Three things are not controversial: 1. that Irish had phonemic velarization, spelled in short cases with what might appear to be short diphthongs; 2. that English had phonetic velarization at least; and 3. that people tend to hear what is phonemic in their language as phonemic in a foreign language, regardless of whether it is. From these considerations alone it is predictable that Irish missionary linguists would have perceived merely phonetic ve-larization in Old English as phonemic and would then have spelled this using graphic short diphthongs much as in their own language. The result would have been a situation where Old English graphic short diphthongs were predictable in distribution. That is, with very few doubtful exceptions, the situation that we see, so we have little reason to think that anything else lies behind it.


David L. White

It would lead us far afield indeed to say much more here, but it may be noted that the same methodology that within Old English studies has traditionally been assumed to "prove" that spellings like vs. must represent a contrast could equally well be used to "prove" that spellings like vs. must represent a contrast in modern English. Of course such spellings do indicate that someone perceived a contrast at some time, but in the case of Old English that original perception was not necessarily native, and therefore not necessarily accurate. The appearance of contrast could have been maintained, once established, by spelling rules. We tend to simply assume that Anglo-Saxon spellers would have been culturally incapable of applying spelling rules. But were people and the nature of spelling really so different "in those days"? The entire issue is overdue for re-examination, as the case for short diphthongs being an unproblematic type of unitary phoneme is no-tably weaker than has generally been assumed. In particular the somewhat casual assertions of Lass and others implying that short diphthongs are somehow "well-known" from Afrikaans, northern Welsh, and Scots, cannot be accepted as valid on the basis of the evidence that has been produced at this point. It seems that those who believe that Old English had short diphthongs would have us believe not just one or two but all four of the following propositions: 1. that there is an abitrary "Universal" list that just happens to exclude things like /ps/ as unitary phonemes, while including short diphthongs, but that no natural "homorganic" rule excludes short diphthongs; 2. that Old English is the only language in the world, or in the history of the world, that is "known" to have had short diphthongs; 3. that short diphthongs and secondary articulations co-exist quite happily in our world, though they would often produce essentially the same phonetic output; 4. that the fact that Old English is spelled just as we would expect it to be spelled by Irish missionary linguists, i.e. as having phonemic short diphthongs regardless of whether it did, is just a misleading coincidence without relevance for our interpretation of Old English spelling. It is necessarily left to individual readers to judge how probable or improbable such a world would be.

Why we should not believe in short diphthongs


Returning now, once more before closing, to the matter of philology and linguistics, it was noted at the outset that the two are often associated respectively with specialization and generalization. We tend to assume when we hear the term "English philology" that such a thing can exist successfully in a vacuum. But can it? It would seem that English and Irish philologists have assumed different conclusions (the implication of this wording is fully intended) about how graphic short diphthongs are to be interpreted. Should they not at least compare notes? No language is an island, and I submit that a person who does only English philology, untroubled by other philology or by general linguistics, is not likely to do it very well. It is not of course possible to be a specialist in, or a philologist of, everything, or perhaps even more than one thing. But it is not necessary to be a specialist in Irish to have a little knowledge of Irish, enough to be useful. A little knowledge can be a dangerous thing, perhaps. But even less knowledge can be an even more dangerous thing. There was no firewall of separation between the English and the Irish in early Anglo-Saxon England. There should be no firewall of separation between English and Irish philology and linguistics. Old Irish is relevant to Old English.

Notes 1.

2. 3.


5. 6. 7. 8.

Underlining will be used to indicate phonemic linkage and/or notable shortness (which do tend to co-oocur) in what follows; context should make clear which is primarily intended. Apparently this occurs in Scots Gaelic only, and is a variant of pre-aspiration, not of phonemic significance. Pre-nasalized fricatives are also mentioned, but these appear to be more properly pre-nasalized affricates, as a full closure is made, followed by fricative egress. It should also be noted that when two consecutive phones are linked as one, one of the two, the one that is not an ingress or egress, is always clearly primary. There is a significant and clear durational difference. Note that the [s] portion of English /ts/, for example, cannot be prolonged. It is even fairly common for there to be a possible contrast between the two types. The secondary articulation is more to the right in Russian. But Russian has a contrast between palatalized consonants and consonants with following /j/, so Russian palatalization cannot be seen as following /j/.


David L. White


Of course, [kuid] was used above, but the point of the underlining in either case is to indicate that the three segments shown alogether take the time expected for two, i.e. that there are not three moras, and to suggest the phonemic grouping that English or Irish observers would assume from the spellings. The two [kuidsjs are roughly equivalent, and no subtle difference of timing or grouping between the two is meant to be suggested. For strict propriety it should be noted that in the impementation of this rule final consonants in monosyllables are counted as not present, or rather not morale. Such rules, involving "extra-syllabic finals," are in fact fairly common across languages. The problems of phonetic implementations would be almost exactly the same, as was seen above. Kiparsky simply ignores this inconvenient fact. Kiparsky cites these forms as spellings, which of course means that there is no actual indication of putative shortness, since Greek spelling makes no provision for any such thing. This is one of the reasons I am reluctant to believe that Kiparsky truly meant to assert that Ancient Greek had phonemic short diphthongs. Even Aeolic, which had recessive accent for the most part, did not have recessive accent in prepositions and adverbs, and therefore had phonemic accent. Since in each case the /oo/ in question is omega, no claim that this was really omicron can be plausible. If /k/ and /p/ in final /ks/ and /ps/ count as moras, which they must if tradition is correct in regarding things like /oikophUlaks/ as having recessive threemora accent, then things like /kEeruks/ provide yet another example of fourmora accent. Four-mora accent in thematic feminine nouns probably spread (or was retained) by analogy from thematic feminine adjectives which had /-os/ in the masculine. Phonemically there is no point in distinguishing between the acute and the grave, as graves are simply final acutes.


11. 12. 13.

14. 15.

16. 17.



References Abbott, Evelyn, and E.D. Mansfield 1977 A Primer of Greek Grammar. London: Duckworth. Bright, James 1953 Bright's Anglo-Saxon Reader. Revised by James Hulbert. New York: Henry Holt and Company. Catford, John C. 1988 A Practical Introduction to Phonetics. Oxford: Clarendon Press.

Why we should not believe in short diphthongs


Chase, Alston H., and Henry Phillips, Jr. 1941 A New Introduction to Greek. Cambridge: Harvard University Press. Daunt, Marjorie 1939 Old English sound changes reconsidered in relation to scribal tradition and practice. Transactions of the Philological Society 1939: 108-137. Greene, David 1962 The colouring of consonants in Old Irish. Proceedings of the 4th International Congress of Phonetic Sciences, Helsinki 1961, A. Sofijärvi and P. Aalto (eds.), 622-624. The Hague: Mouton. Hogg, Richard M. 1992 A Grammar of Old English. Vol. 1. Oxford: Blackwell. Kiparsky, Paul 1969 A phonological rule of Greek. Glotta 44: 109-134. 1973 The inflectional accent in Indo-European. Language 49 (4): 794849. Ladefoged, Peter, and Ian Maddieson. 1996 The Sounds of the World's Languages. Oxford: Blackwell. Lass, Roger 1973 Linguistic orthogenesis? Scots vowel quantity and the English length conspiracy. In Historical Linguistics II, J. M. Anderson and C. Jones (eds.), 311-352. Amsterdam: North Holland. 1981 Undigested history and synchronic "structure." In Phonology in the 1980s, D. Goyvaerts (ed.), 525-544. Ghent: E. Story-Scientia. 1984 Phonology. Cambridge: University Press. 1988 The Akzentumsprung of Old English 'eo'. In Rhetorica, Phonoligica, Syntactica: A Festschrift for Robert P. Stockwell from his Friends and Colleagues, C. Duncan-Rose and T. Venneman (eds.), 221-232. London: Routledge. 1994 Old English: A Historical Linguistic Companion. Cambridge: University Press. Laver, John 1994 Principles of Phonetics. Cambridge: Cambridge University Press. Lombard, Daan 1985 An Introduction to the Grammar of Northern Sotho. Pretoria: Van Schaik. MacEoin, Gearid 1993 Irish. In The Celtic Languages, M. J. Ball.and J. Fife (eds.), 101125. New York: Routledge. Mallinson, Graham 1988 Rumanian. In The Romance Languages, M. Harris and N. Vincent (eds.), 391-419. Oxford: University Press.


David L. White

McCone, Kim 1996 Towards a Relative Chronology of Ancient and Medieval Celtic Sound Change. Maynooth: St. Patrick's College. Pharr, Clyde 1985 Homeric Greek. Revised by John Wright. Norman: University of Oklahoma Press. Original edition, 1959. Ponelis, Fritz 1993 The Development of Arikaans. Frankfurt am Main: Peter Lang. Smyth, Herber W. 1956 Greek Grammar. Revised by Gordon R. Messing. Cambridge: Harvard University Press. Original edition, 1920. Stanley, Eric G. 1952 The chronology of r-metathesis in Old English. English and Germanic Studies 5: 103-115. Stockwell, Robert P., and Barritt, Clyde W. 1951 Some Old English graphemic-phonemic correspondences - x, ea, and a. Studies in Linguistics:Occasional Papers No. 4. Stockwell, Robert P. 1994 On Old English short diphthongs and the theory of glide emergence. In English Historical Linguistics, D. Brittton (ed.), 57-72. Amsterdam/Philadelphia: J. Benjamins. Thorne, David 1993 A Comprehensive Welsh Grammar. Oxford: Blackwell. Thurneysen, Rudolf 1946 A Grammar of Old Irish. Dublin: The Dublin Institute for Advanced Studies. White, David L. 2000 Irish Influence and the Interpretation of Old English Spelling. Ph.D. diss., University of Texas at Austin.

Extended forms {Streckformen) in English Anatoly Liberman

This is the way the gentlemen ride; Gallop-a-trot, Gallop-a-trot! This is the way the gentlemen ride; Gallop-a-trot-a-trot. This is the way the farmers ride; Hobbledy-hoy, Hobbledy-hoy! This is the way the farmers ride, Hobbledy-hobbledy-hoy! (Nursery rhyme)

1. Introduction Almost all participants in the Studies in the History of the English Language 2 (SHEL-2) conference spoke on historical phonetics or historical grammar, whereas my subject then, as here, was etymology and word formation. Since the rise of structuralism early in the twentieth century, linguistics has stopped being part of philology. Our departments of linguistics do not look on text editing, paleography, lexicography, etymology, and even stylistics, unless it is tied to social norms, as an integral part of their curricula. Someone versed in Government and Binding, Optimality Theory, or the latest brand of semantics is a respectable member of such a department. But a young Ph.D. who has studied the history of Webster's dictionary, variation in medieval manuscripts, or glosses in their relation to the vocabulary of an old language, has no chance of employment as a linguist. Only a few language departments may take pity on such a "marginal" scholar. This is sad, because specialists in the subjects I have listed know a great deal about phonetics, morphology, and syntax, even if their knowledge may sometimes lack the sharpness of the proverbial cutting


Anatoly Liberman

edge. My experience of the last sixteen years has strengthened my belief that only a professional linguist can write persuasive word histories. Perhaps here I am preaching to the converted. Yet I would like to make my point clear: in giving a talk and publishing a paper on the origin of two obscure words, I intend not only to offer the solutions I find reasonable and call attention to a type of word formation that has passed almost unnoticed. My other goal is to highlight some benefits of the union between linguistics and traditional philology.

2. Infixation as a problem of historical linguistics This paper is devoted to a neglected aspect of infixation in the history of English. Infixation is a vague concept because it subsumes several heterogeneous phonomena: infixes go all the way from single phonemes to polysyllabic words. Their functions also differ considerably, and the only feature that unites them pertains to form, not to content: every infix, by definition, cuts a word into two parts and appears in it as a foreign body. The oldest recorded infixes of Indo-European played a noticeable role in both morphology and word formation. The best-known of them is /n/, as in Lat.1 iungo—junx—iunctum—iungere 'connect, bind, tie together', as opposed to jügum 'yoke', Go. standan 'stand'—stop 'stood' and fraihnan 'ask'—/ra/a 'asked'. So many other infixes exist, and there are so many of them that Karstien (1971) was able to write a thick book about them. Etymological dictionaries say unanimously that Lat. friäre 'crumble' and fricäre 'rub' are related, but the origin of/k/ in fricäre remains unexplained, for calling /k/ emphatic leaves all the questions open, the more so as other equally persuasive examples are absent. Engl .fillip is probably akin to flip; the German counterpart of the verb flip is fipsen. Intrusive /l/ poses the same problem as does Dd in fricäre. From the point of view of both modern and old languages such infixes are desemanticized: just /η 1 k/. On the other hand, what-so-bloody-ever contains a meaningful element (bloody) that has been undoubtedly inserted in whatsoever for emphasis and humorous effect. The nasal infix goes back to Indo-European, but l\l in flip and the like must be comparatively recent. One can imagine the loss of /l/ in fipsen or its addition in flip, or both words can be sound symbolic formations, their similarity being due to chance. I am not aware of a single work on such words. It is only (American) English words like abso-bloody-lutely and phrases like born bloody survivor that have attracted considerable atten-

Extendedforms (Streckformen,) in English


tion. McMillan (1980) discussed different types of infixing and appended a sizable glossary of extended forms to his article. In recent years, Michael Adams has been especially active in researching infixation and interposing in American English (see Adams 1999, 2001, 2002). Absolutely fell victim to infixation at the beginning of the twentieth century; abso-bally-lutely, Macmillan's first example, is dated 1914. Forms of this type were so common that Henry Cecil Wyld cites abso-blooming-lately as an example of tmesis (see the entry tmesis). Even then it must have been a euphemism for abso-frigging/fucking-lutely. The earliest recorded extended form is probably daffadowndilly or daffydowndilly for daffodil (1573, OED). It may be worthwhile to point out that despite the recent age (or recent attestation) of the abso-bloody-lutely model, it continues a very old trend. Tmesis, or diacope, occurs in Ancient Greek and in Old Icelandic skaldic poetry. It occasionally appears in Latin: cf. se-que-gregari, with que between se- and -gregari (segregari 'separate'). In the Authorized Version of the Bible, we find "of whom be thou ware also" (2 Tim. IV: 15) for "of whom beware thou also" (quoted in the Century Dictionary, tmesis). Gothic tolerated particles and pronouns between the prefix and the root, as in ga-p-pan-mip-sandidedum 'and we then sent along' (2 Cor. VIII: 18): p < up, assimilated from uh ' a n d \ p a n 'then', and mip 'with' stand between ga- and -sandidedum (3rd p. pi., preterit of gasandjan 'send, dispatch'). In the so-called incorporating languages, morphemes are inserted into a word in such a way that the line between the resultant pita bread-like word and a sentence disappears. We cannot know whether Gothic, with its packaging of adverbs and particles between prefixes and roots, reflects an archaic stage the other Germanic languages had left behind by the time their first written monuments were recorded or whether it developed its own morphological pattern. In Modern Germanic, incorporation as a grammatical means occurs in German verbs with separable prefixes, for example, aufgehen 'go up' versus auf-ge-gangen, the past participle of the same verb. In later Germanic, tmesis is rare. According to Duden (tmesis), in northern German dialects, but not only in them, one can hear "da hast du kein Recht zu" = "dazu hast du kein Recht" ('you have no right to do so': da...zu). Sta-n-d, fraih-n-an, and so forth are studied by grammarians. Fri-c-äre and f-l-ip come within the purview of etymology. Daffy-down-dilly and fan-damn-tastic should find a place in books on word formation. Neither Koziol (1937) nor Marchand (1969) mentions them, probably because both missed the pattern or because such words are "occasional" and invariably ludic. With born bloody survivior we are almost in the realm of syntax.


Anatoly Liberman

These spheres (morphology, etymology, word formation, syntax) are not delimited clearly when it comes to infixation and tmesis. Still another type of infixation seems to have never been discussed if one disregards sporadic mentions of it in etymological dictionaries. Sometimes syllables rather than separate phrases or whole words are inserted into words. From popular culture Modern English has the facetious eduma-cation (the early eighties?), and now b-az-itch (bitch) has made its entry. Edu-ma-cation, far from violating the spirit of the language, as they used to say in the nineteenth century, continues, though in a slightly modified form, an old trend. Words like razzmatazz are known in both urban slang and dialects.

3. Schroder's Streckformen In 1906 Heinrich Schröder, a German linguist full of original ideas even in such traditional areas of study as ablaut and word formation, brought out a book entitled Streckformen. Since Germ, strecken means the same as Engl. stretch, the term, coined by Schröder, can perhaps be translated as 'augmented/extended forms'. Schröder collected about 350 trisyllabic, mostly expressive, words of Modern German with stress falling in the middle. In noncompounded German words without a prefix stress is regularly assigned to the first syllable unless they are borrowings like Florett 'foil' (a weapon), Terzerol 'small revolver', Pistole 'pistol' or formations with Romance suffixes (for instance, Spielerei 'foooling around', süffisänt 'smug, complacent', organisieren 'organize'). Native words with the structure _ ^ _ are few in Standard German (cf. Holunder 'elder', a plant, Wacholder 'juniper', Forille 'trout'), and their history, as well as the etymology of some of them, such as Schmarotzer 'parasite', Schlaraffenland, the German counterpart of the land of Cockaigne, have been the object of intense research. Schröder suggested that German makes use of inserted syllables and cited dialectal verbs like b(aj)äckern, j(ad)ackern, sl(ad)acken, sch(aw)uppen, kl(ad)astern, kl(ab)astern (they mean 'run fast, run about aimlessly'), allegedly traceable to and being reinforced variants (extended forms) of bäckern, jackern, slacken, schuppen, klastern, all of which have been recorded in the same area; without "maternal" forms, the existence of Streckformen can obviously not be demonstrated. If Schröder had confined himself to obscure dialectal forms like those listed in his (1903) article, the earliest work in which he introduced his

Extended forms (Streckformen,) in English


ideas, his contribution to the study of German word formation would probably have been overlooked or used on a moderate scale in dialectology. The article contains 53 examples, of which only two or three belong to the Standard. But in his (1906) book he broadened his scope of investigation and analyzed numerous words, few of which are as convincing as m(ar)üshel 'virago' or kra(wä)ulen 'work hard but inefficiently' (Schröder 1903: Nos. 12 and 34). It should be repeated that extension arouses no doubts only when the presumed derivative with stress in the middle coexists with a close synonym which is shorter by a whole syllable, but when, for example, schmarotzen 'live like a parasite, sponge on one' is referred to Middle High German smutzen 'besmear oneself (Schröder 1903: No. 39, 1906: 83-87), the case looks weak, for the semantic link between 'parasite' and 'dirt' is far from obvious. Schroder's ideas were received coldly, and Kluge, the most important critic, flatly rejected the concept of the Streckform (Kluge 1906). He cited the least convincing examples and exclaimed: "It is a piece of rare luck when a single difficult word reveals its origin, and here 350 new etymologies are offered!" (393). His dictionary enjoyed great popularity, and he apparently disapproved of a strong competitor. Kluge's review was published in the prestigious Literaturblatt fur germanische und romanische Philologie. Otto Behaghel, the editor of the journal, wrote a short postscript to Kluge's diatribe (Behaghel 1906), in which he suggested that Streckformen should not be dismissed out of hand, because sometimes their existence is evident. Years later, he published an article on language at play and language humor and mentioned extended forms as a case in point (Behaghel 1923: 183). Independent of Schröder, practitioners of a trend Leo Spitzer called idealistic etymology (Voßler, Bally, Spitzer himself, and other contributors to Jahrbuch fur Philologie), that is, people interested in the aesthetics of language, pointed to similar examples. Spitzer (1925: 151, note 1) compared Spanish tartaruga with Christan Morgenstern's Schild-krö-kröte from Schildkröte (both mean 'tortoise') and thus allowed "anonymous" speakers to resort to the same jocular means of expression one finds in a work of a modern German poet. Nowadays, when research into the iconicity of language is commonplace, such conclusions cause no objections: turtles (tortoises) move slowly, and we understand that it is natural for a word designating a sluggish animal to last as long as possible. German and Dutch scholars have isolated words with initial ka-, ker- (Bause 1919; De Bont 1948: 28; cf. Coetzee 1995 for Afrikaans). Reinforcement in them


Anatoly Liberman

(the first syllable) differs from what we find in German (the middle syllable), but the principle is the same. Seebold, in Kluge-Seebold (1995), refers to Streckformen even when the putative base cannot be established (see schmarotzen: a reinforcement of an undiscovered base). Kluge would have disagreed with this etymology in the strongest terms.

4. Streckformen in English 4.1. The quasi-German type It appears that words like Germ, klabastern rarely occur in English. When they do, they are trisyllablic frequentative verbs of dialectal origin with stress in the middle. The examples below, except shillaber, are from Wright (1898-1905). Fineney 'mince, simper' is a doublet o f f i n e y , which suggests that it is fi(ne)ney. Wright gives fandangle 'ornaments, trinkets; capers' with stress on the first syllable; yet it looks like an extended form of fangle, as in newfangled. Fangle is glossed 'a conceit, whim; to trim showily, entangle; hang about, trifle, waste time'. Another word with stress allegedly falling on the first syllable is shillaber (American slang), of which the OED, The Second Supplement, has only one 1913 example. The lexicographers on the present staff of the dictionary could not have heard it, for it has been dead for almost a century, and only its doublet shill 'decoy or accomplice, especially one posing as an enthusiastic or successful customer to encourage other buyers, gamblers, etc.' is in use. Shill is probably an abbreviation of shillaber (so in the OED), whose origin is said to be unknown. The development from shillaber to shill does not prove that the longer word had initial stress: cf. prof from professor and, conversely, 'burbs from suburbs. Could shillaber be an extended form of Germ. Schieber 'black marketeer' (*shi-la-ber)l Fundawdle 'caress' and gamawedled 'tipsy' look like extensions of fondle and gaddle 'drink greedily and hastily'. Perhaps finagle also belongs with such verbs. Like the now forgotten shillaber, it first surfaced in the United States. The earliest citation o f f i n a g l e in the OED goes back to 1926. Webster2 compares finagle with fainaigue [fa'ni:g] 'revoke at cards, renege, play truant, cheat ...' from Wright (1898-1905). Griffith (1939: 292), who remembered growing up with the meaning of finagle 'fuss and feather over a small matter with fakery in it, a lackadaisical effort to sell a bargain, a small bargain', suggested that it is a respelling of the name of

Extended forms

(Streckformen,) in English


Gregor von Feinagle (71765-1819), a German mesmerist and whist expert, often ridiculed in Europe. He even considered the possibility of finagle originating from Byron's usage in Don Juan 1/11 ("For her Feinagle's were an useless art, / And he himself obliged to shup up shop — he / Could never make a memory so fine as / That which adorn'd the brain of Donna Innez."), "but more likely," he says, "it is the off-spring of the lampooning, local hit slang-inventiveness of music-hall taste; in England 1805-1820, the 'popular' pronunciation of the foreign lecturer's name would have been Fee-nä-gle." Most dictionaries prefer the fainaigue etymology, and it is indeed difficult to reconcile Feinagle's fame with the apparently dialectal provenance of finagle and its late attestation. The groups -ddle and -ggle often alternate in dialects. One can assume that jiggle is another pronunciation of fiddle. The OED gives only one example of figgle 'fidget about' (1652), but this verb exists in modern dialects (Wright 1898-1905). Finagle may then be fi(na)gle. One of the most disputed words of this type is skedaddle. It was first recorded in 1861 (DARE) in an American newspaper and immediately became known on both sides of the Atlantic. It has been traced to Ancient Greek, Irish, Welsh, Swedish, and Danish, as well as to various blends. (See Cohen 1976, 1979, 1985: 29-63; Cohen et al. 1979: 20-24; and my discussion in Liberman 1994: 173-175.) Most of the early hypotheses are without merit, though skedaddle being understood as a blend (such a suggestion exists) is more credible than as a borrowing or reshaping of some Homeric, Celtic, or Scandinavian word. The only reliable clue to its origin is furnished by the verb skedaddle 'spill milk' current in Dumfriesshire (Mackay 1877, s.v.). One can also skedaddle coals, potatoes, apples, and other substances falling from a cart. Skedaddle 'spill' and skedaddle 'retreat hastily' may be parallel formations. Engl. dial, scaddle means 'scare, frighten; run off in a fright, dare one to do something', while skiddle is glossed 'spill'. The phonemes /ae/ and /e/ tend to alternate in dialects; *skeddle is a probable doublet of scaddle. American skedaddle is, most likely, an extended form of scaddle or *skeddle: ske(da)ddlle or sca(da)ddle. There would be no difference between pretonic ske- and scaanyway. Northern English and Scottish skedaddle appears to be an extended form of skiddle: ski(da)ddle. Incidentally, in the 1861 citation the American verb was spelled skidaddle.


Anatoly Liberman

4.2. The augments -a- and -de4.2.1. Parasitic -aEnglish has a sizable number of words with "parasitic" (the German term is unorganisch) inserted -a- in the middle. One of them is Cockney < ME cockeney. It surfaced in the second half of the fourteenth century in two meanings, 'bad egg' and 'milksop, effeminate man; Londoner'. The etymology of cockenay 'milksop' is debatable (from French?) despite the considerable effort expended on it, but cockenay 'bad egg' has been shown by Murray (1890 and in the OED) to go back to the idea of 'cock's egg'. However, cockenay is not cocken-ay, as Murray believed, but cock-e-nay (so explained by Charles P. G. Scott: see cockney in the second edition of the Century Dictionary), -nay 'egg' arose by metanalysis (misdivision) from an ay (cf. nuncle < an uncle and the like); as is known, the Old English for 'egg' was ceg (> ME ey). The only problem with cock-e-(n)ey is e- in the middle. In words of French origin, the connecting schwa is a preposition: cf. vis-ä-vis and cap-a-pie. In the native vocabulary, -a- is a reduced form of on or of, as in twice a day, cat-o '-nine-tails, man-o '-war, Tam ο' Shanter. But when a model establishes itself, new formations arise, and it is futile to look for their prehistory. Tam o' Shanter was simply Tam Shanter in Burns's poems and acquired its ο' on the analogy of John ο' Groats and so forth. Fustianapes is an allegro form of fustian of Naples, but jackanapes developed from Jac(k) Napes, not Jack on or of Napes, and, surely, Jack-adandy never was *Jack of/on dandy. Will with the wisp forfeited its with the (o' substituted for them), and in a similar way the older form of lack-aday, the basis of lackadaisical, was alack the day (see these words in the OED and ODEE). As already pointed out, books on English word formation do not mention extended forms of any type. Etymological dictionaries deal with them on an individual basis and fail to solve the origin of -a-. The ODEE states that a in Blackamoor < black More is unexplained. The comment in the OED is much longer: Of the connecting a no satisfactory explanation has been offered. The suggestion that it was a retention of the final -e of ME. black-e (obs[olete] in prose before 1400) is, in the present state of the evidence, at variance with

Extended forms (Streckformen^ in English


the phonetic history of the language, and the analogy of other black- compounds. Cf. black-avised.

In the entry black-a-vised 'dark-complexioned' (first recorded in 1758, over two centuries later than Black-a-moor), we read: "... perh[aps] originally black-a-vis, or black ο' vis; but this is uncertain." Black-a-top 'blackheaded' (one 1733 example) is left without an etymology. The first element of caterwaul is said in the ODEE to be perhaps related to or borrowed from Low German/Dutch kater 'male cat', unless -er- "is merely an arbitrary connective syllable]"; this is a paraphrase of "some kind of suffix or connective merely" (the OED). Neither Murray nor Onions seems to have realized that cat-er-waul (= cat-a-waul?) is not an isolated example. A historian of English is hardly allowed to dismiss such puzzling insertions as merely arbitrary connective syllables/some kind of suffix. The Century Dictionary calls -a- in black-a-moor and jackadandy a meaningless syllable. This is true enough, but how did it originate? The OED invites us to compare grinagog (dial.) 'one who is always grinning' and its variant grinagod (the latest example is dated 1765, but this word occurs in Grose 1796) with stareagog and turlygog, apparently, three synonyms. Unless grinagog and stareagog are akin to agog (the OED does not suggest the connection), the origin of -a- in them remains a riddle. Returning to cock-a-ney, we notice a profusion of words beginning with cock and having a similar prosodic structure: cockabully 'a blunt-nosed fish in New Zealand', cockagee 'a kind of cider apple' (Wright 1898— 1905), cockagrice (obsolete) 'a cock and a pig cooked together', Scots cock-a-bondy 'fly for angling' and its near homonym cock-a-bendy 'instrument for twisting ropes', cock-a-leekie 'soup made from a fowl boiled with leeks' (-a- < of?), cock-a-hoop 'in a state of elation', and the universally known cock-a-doodle-doo. Cockalorum 'whippersnapper' and some words from Dutch and French, for instance, cockatiel 'a kind of parakeet', cockatoo 'a kind of parrot', the humorous cockamamie 'ludicrous', cockarouse (dial.) 'a person of distinction', and cockatrice also have -aafter cock. Scott was aware that cock-e-ney shared its insertion with some other words but cited only black-a-moor, pink-a-ney 'small or narrow eye', and Early Modern English mol(d)-e-warp 'mole' (animal). Borrowings are especially prone to acquiring -a- in the middle. Hackamore 'halter with a loop' is the rendering of Spanish jäquima 'halter, headstall of a halter'. The incomprehensible Dutch ter kaap varen 'go privateering' became capabarre 'misappropriate government stores'


Anatoly Liberman

(Anonymous 1912). Cockamaroo 'Russian bagatelle' (a ball game), first recorded in 1850, appears in the OED, Second Supplement, without any etymology. Whatever its origin, it sounds like cockamami and its kin. Cock-e-ney is the earliest recorded extended form with -a- (there was, of course, no difference between unstressed a and e in Chaucer's days), and the fourteenth century must have been approximately the time such words arose. Their emergence falls into the period of free apocope. The sound shape of hundreds of words became fluid and depended on their place in a sentence and syntactic function, and this is when English acquired its present-day rhythmic patterns. The existence of such patterns is seldom realized, but they determine the intuitive assessment of certain speech units as flowing or stilted. Brendum-Nielsen (1924) noted that humorous words in Germanic gravitate toward an anapestic structure. If cocke-ney had stress on the last syllable, it fits Brandum-Nielsen' s rule, and so does muck-a-muck 'a person in charge'. Hullaballoo, though it contains four syllables, is stressed on the ultima too. Note also the words of the peek-a-boo type in nursery rhymes: rick-a-ting-tang, rub-a-dub-dub, ringa-ring-a-roses; rock-a-by (hush-a-bye), baby; hick-a-more, hack-a-more. Unstressed i was also drawn into the process of coining extended -aforms. Cock-a-leekie has a doublet cockie-leekie, but ie in cockie is a spurious suffix. A similar case is piggyback 'carry on one's shoulders', from pickaback. According to Skeat (1882, and the same in all the editions of his dictionary), huckaback 'coarse durable linen' (earlier hugaback and hag-a-bag) is the English pronunciation of Low German huckebak 'pick-aback'; at one time, it presumably designated a pedler's ware, but the evidence is lacking and the OED says "origin unknown." If, however, Skeat's etymology is correct (and Kück [1905:14-15] seems to give it some support), huckaback is a doublet of pickaback/piggyback. Those who regard -a- to be a syllable devoid of meaning (though not an arbitrary connective syllable or a kind of suffix) are right, but reference to a type of word formation is needed. Pigeonholing extended forms will not absolve etymologists from doing their work because every word has an individual history. The presence of -e- in ME cock-e-ney and of -a- in chick-a-biddy should be taken in stride (it is an element that never meant anything), but -a- in dial, chickaleary with its amusing gloss 'aged pedestrians on winter mornings' has a recoverable past: cf. one-two-three-alairy: aged pedestrians were apparently at risk of looking like chickens with their legs wide apart or crossed over on a winter morning; cf. Scots hilliegeleerie or hilliegalair 'topsy-turvy'. Obsolete dial, hotagoe 'move

Extended forms (Streckformen) in English


nimbly' (said of the tongue: "You hotagoe your tongue," Ray [1874]; Wright [1898-1905] also has this word) looks like hot-a-go. But is it hot + extension + go? Even more impenetrable is dial, pistaquint 'pretty well'. Ease of articulation (phonetic economy) is the least important factor in the creation of such words. In cock-e-ney, -e- was syncopated, and hob-anob (hob and nob) became hobnob. Half of those whom I polled stated that they either preferred rigamarole to rigmarole or did not mind the extended form. It is unlikely that -a- appears in this word to make the pronunciation of -gm- easier. There is a universal tendency toward open syllables, but English is not Japanese and does not turn Berlin into Berulini. Rigmarole, because of its meaning and playful connotations, lends itself ideally to the α-extension. People who say rigamarole would never say dogamatic or stygamatize. The model with inserted -a- is productive, with transparent coinages and catchwords testifying to their vitality: cf. rope-a-dope (Muhammad Ali's phrase for roping a poor opponent; dope = idiot) and ads inviting us to dial-a-wish, dial-a-dream, call-a-bike, and buy a couch called hide-abed. Slangy phrases like bud-a-bing, bud-a-boom 'quickly and easily' (only in Brooklyn, New York?) belong here too; see also the phrases from nursery rhymes, above.

4.2.2. Parasitic -de- and -teThe number of words with parasitic -de- is not great. Some humorous coinages are especially interesting because they bring out the productivity of de-: cf. simper-de-cocket (the OED gives citations from 1524 to 1707), a term of relatively mild abuse for a woman (the original meaning must have been 'simpering coquette') and gobbledegook (gobbledygook). In the study of such compounds, French models from Coeur de Lion to dent-de-lion > dandelion spring to mind at once, but -de- (and in this respect it shares some common ground with -a-) probably has more than one source. In musterdevilliars, the name of woolen cloth well-known between the fourteenth and the sixteenth century (the latest citation in the OED is dated 1564), -de- is from French. But dandiprat 'small 16th-century coin' and 'worthless fellow' (no recorded examples before 1520) is obscure. Weekley (1921) wonders: "?Of the same family as Jack Sprat — 'This Jack Prat will go boast / And say he hath cowed me' — Misogonus, ii, i, c. 1550." Dandy, itself of dubious origin, surfaced only in the eighteenth


Anatoly Liberman

century. Prat could be a nickname (cf. prat 'trick', let alone prat 'buttock'), or it could refer to prate and prattle. Neither of the putative components is of French provenance, but the whole looks like dan-di-prat. Parasitic -di- occurs in several words denoting hubbub, ruckus, that is, noisy commotion and disturbance, and in the names of disreputable people and demons, in one of which it varies with -te-. The form -te- also speaks against the French origin of this augment. Here are some of the -de-/-te- words: hubble-te-shives 'confusion', flipper-de-flapper 'noise and confusion'; slabberdegullion 'lout' (a seventeenth-century word), tatterdemalian 'ragged person', grizzle-de-mundy 'stupid person who is always grinning' (cf. grinagogl); Flibber-ti-gibbet 'demon' (a word memorable from King Lear), with numerous variants, including Flibberdigibbet. The syllables -de- and -te- are common in Dutch, and some English words with this augment are probably of Low German/Dutch rather than Romance origin. In German we find holterdi(e)-polter 'upside down'. Its Dutch counterpart is holder-de-bolder 'helter-skelter', while Low German has helter-(de-)fulter, hullerdebuller, and hulterpulter, the last one without an augment. Needless to say, all of them are of obscure origin, but hardly any is a borrowing from French. Schröder (1906) regarded Schlaraffen- in Schlaraffenland to be a Streckform: schl(ar)aff(e), from schlaff 'loose, lax'. Kluge (1906: 401) found a mistake in Schroder's quotation and hauled him over the coals for it. However regrettable that mistake might be, it did not in principle invalidate Schroder's conclusion. Seebold (1995) also considers Schlaraffen to be an extended form even though its basis remains unclear (see above). The curious thing is that Schlaraffen developed from slü-de-raffe 'prosperous idler'. Consequently, one way or the other, Schlaraffen is an extended form. French and Low German must have contributed in equal measure to the spread of -dein English. The augment -a- has no parallels in the German-speaking area. Once this type of word formation came into being, hybrids like hobble-depoise 'easily balanced' (half-Germanic, half French), modeled on avoirdupois, and coinages like gobbledigook met with no resistance. The goal of the case study that follows is to show how an etymologist can use the concept of the Streckform in dealing with the origin of difficult words. Only -a- and -de- will be discussed. As always in works of this genre, solutions in the samples below are preceded by a survey of earlier conjectures.

Extendedforms (Streckformen,) in English


5. A case study: ragamuffin and hobbledehoy 5.1. The case of ragamuffin It has been known for a long time that in Langland's Piers Plowman, dated 1393 (c, XXI: 183, Skeat's edition, 1886, vol. 1) a devil called Ragamoffin is mentioned. The OED quotes the relevant passage. According to the MED, the name Isabella Ragamoffyn occurred as early as 1344. For two centuries ragamuffin (with any spelling) did not appear in written documents. Its uninterrupted history, now as a common name, goes back to 1581. The OED says the following about its origin: "[P]rob[ably] from Rag sb.' (cf. Ragged lc), with fanciful ending." The second part of ragabush 'worthless person' (now chiefly dialectal) is also said to contain a fanciful ending added to rag. Reference to fanciful endings does not make much sense when applied to sound strings like -amuffin and -abush. Whatever the origin of ragamuffin, its present day sense was influenced by rag, but it does not follow that the first ragamuffin was ragged or wore rags. The entry Ragman 'devil' in the OED contains a passing remark: "cf. Ragamuffin, Ragged, Sw[edish] ragg-en ['devil']." In the entry ragged, several examples make it clear that the Devil was often portrayed as having a ragged appearance. Sw. raggen can be understood as 'the shaggy (hairy) one', a tempting interpretation in light of the material from Middle English, or as 'the evil one' (rag is also a metathesized form of arg 'evil, wicked'). Hellquist (1948) preferred the second alternative, while the OED took the first one for granted. Spitzer (1947: 91) derived rageman (this is Langland's spelling) from French, and it is true that the idea of Ragemon (le bon) and Rogomant folk etymologized into rageman/Rageman carries more conviction than raggen borrowed from Swedish, because Sw. raggen is a neologism unrecorded in the other Scandinavian languages. The French source of ragman and ragamuffin was suggested long ago (Anonymous 1822: 618), but neither Spitzer nor his predecessors succeeded in discovering the ultimate origin of the French name. Its source may have been Germanic, especially if an old attempt to connect Engl, rag and Ital. ragazzo 'boy, youth' has any weight (then ragazzo = 'little devil', not 'person in rags'). Probably no other word of Italian has been discussed so often with such meager results; its etymology is still "unknown." The Germanic root rag- meaning 'fury' (cf. Dutch dial, ragen 'run around in wild excitement') probably existed, and a pagan divinity called Rageman,


Anatoly Liberman

someone like the Old English Herla cyning (the protoform of Harlequin) is not unthinkable (cf. Woden from *wöd- 'fury', Germ. Wut). Conjectures on the etymology of -muffin have been inconclusive: from Spanish mofar 'mock' or Ital. muffo 'musty' (Webster 1828; so in all the editions until 1864), from Germ. dial, muffen 'smell musty' (Webster 1864; the same until 1890), from Gaelic maoidh 'threaten' (Mackay 1877; Mackay, who derived most words of European languages from Gaelic, combined Gaelic ragair 'thief, villain' with maoidh, so that ragamuffin turned out to be 'dangerous scoundrel'), and from Engl, muff 'stupid, clumsy person' (UED: -muffin is only "compared" with muff). In Spitzer's opinion (1947: 93), ragamuffin goes back to Fr. *Rogom-ouf[l]e or *Ragam-ouf[l] e, which must be a blend of Ragemon 'devil', and such words as OF ruffien of the fourteenth century ... or maroufle ['scoundrel']; again, it could even be a coinage from the ragemon stem formed with the OF suffix -ouf[le], like maroufle itself. ... The idea of 'ragged' appears in ragamuffin only as late as 1440, and is consequently quite secondary.

He adds that ragamuffin still means a (ragged) street urchin and that perhaps 'street urchin' was the original meaning, which is associated naturally with 'devil, demon, imp, heathen'. In the (1890) edition of Webster's dictionary, ragamuffin is left without any etymology: only the name of Langland's demon is mentioned. For more than a century all reference works have followed this example; only Wyld (UED) risked a tentative comparison of -muffin with muff. Skeat did not even include the word in his etymological dictionary, but in Piers Plowman (1886,11:257, note on line 283) he wrote: Mr. Halliwell, in his Dictionary, remarks that Ragamofin is a name of a demon in some of the old mysteries. It has since passed into a sort of familiar slang term for any one poorly clad. The demons, it may be observed, took the comic parts in the old mysteries, and were therefore sometimes fitted with odd names.

Against this background all the more surprising is the entry in AHD3,4 in which -muffin is traced to Middle Dutch moffel, muff 'mitten' (is a bahuvrihi of the Redcap type meant: Ragamuffin = ragmitten or ragged mitten?). The entry is supplemented by a word history in which the discovery of the name Isabella Ragamoffyn is said to disprove the current derivation

Extended forms (Streckformen) in English


of ragamuffin from a devil's name. But ragamuffin has always been understood as a vague continuation rather than a reflex of ragamoffyn in Piers Plowman. Apparently, the woman in question had the character that earned her her unusual soubriquet. Some of the conjectures listed above can be ruled out by definition. Why should an English compound have an element straight from Spanish, Italian, German, Gaelic, or Middle Dutch? At best one could look for English cognates of those words. But -muffin has not been recorded ( m u f f i n 'cake' became known in the eighteenth century and has always meant what it means now). Spitzer's etymology is learned but too speculative. Engl. muff cited by Wyld was first used by Dickens in 1837, and this must have been the time when it gained currency in the streets of London. It has no ancestors, except muff' depreciatory term of a German or Swiss, sometimes loosely applied to other foreigners', which does not occur in extant texts after 1697. Modern Dutch muff'\ouV < mof (originally the same meaning as in Engl, muff) and Modern Germ. Muffel were recorded much later than ragamuffin. Even if their history were less opaque, their late attestation and the absence of their cognates in Middle English excludes their connection with ragamuffin (such is also the opinion of the OED). However, it is not entirely improbable that Engl, muff is an import from the continent. A seemingly correct etymology of -muffin can be deduced from the information published by Smythe Palmer in 1910. His article appeared in a popular journal, and I would never have discovered it if I had not been for many years engaged in compiling a bibliography for a new etymological dictionary of Modern English. An all-enclosing sweepnet lets relatively little slip through its meshes. That is how The Nineteenth Century happened to be screened among dozens of other serials. Smythe Palmer read the Supplementary List of Cumberland Words by E. W. Provost and noted the phrase Auld Muffy used by the older dalesmen for the Devil. As he observes: "The expression is now but seldom heard, and in a few years, probably it will be as extinct as the dodo." Muffy is Anglo-French maufe 'ugly, ill-featured', "which was once synonymous with the Evil One," a creature "notoriously hideous and deformed"; cf. Satan le maufi, etc. (Smythe Palmer 1910: 545-546; see additional details on p. 546). In all likelihood, both components of ragamuffin mean 'devil'. Only the origin of final -n is not quite clear, but so many nouns ended in -an, -en (cf. guardian, warden, and formations of the slabberdegullion and tatterdemal(l)ian type) that *ragamauffi could easily become *ragamauffi(an). The earliest spelling is ragamoffyn (with ο for Fr. au?) and Shakespeare

100 Anatoly Liberman

has rag of Muffin or rag of Muffian in 1 Henry IV, IV, iii: 272. If the reconstruction proposed here is correct, ragamuffin is a tautological extended form with the initial sense 'devil-a-devil', a coinage not unlike muck-amuck. Some confirmation of my hypothesis comes from the history of hobbledehoy, for hobbledehoy also seems to be connected with demons, fiends, and other malevolent spirits.

5.2. The case of hobbledehoy The recorded variants of the noun hobbledehoy are unusually numerous. It was spelled with two hyphens or without any and also as two or three separate words. The first element appeared in the forms hob(b)le-, hob(b)a-, hobbe-, hobby-, hobo-, hobbi-, hobbard-, and hab(b)er- (the OED; a similar array of forms can be found in Wright). Hobble-, hobbe-, and hobber(d)- reflect different pronunciations rather than the instability of spelling; this is especially true of hobble- and hobber-. The middle syllable could be -de-, -di-, -dy-, -da-, and -ty-. According to the OED, hobbledehoy is [a] colloquial word of unsettled form and uncertain origin. One instance in hobble- occurs in 1540; otherwise hober-, hobber-, are the prevailing forms before 1700; these with the forms in höbe-, hobby-, suggest that the word is analogous in structure to Hoberdidance, Hobbididance, and hobidy-booby, q.v.: cf. also HOBERD. Some of the variants are evidently due to the effort of popular etymology to put some sense into an odd and absurd-looking word. It is now perh[aps] most frequently associated with hobble, and taken to have ludicrous reference to an awkward and clumsy gait.

This summary is followed by a brief mention of the fact that Ray, Jamieson, Forby, and Skeat, too, tried to explain hobbledehoy. The derivation of hobbledehoy from Fr. hob(e)rau 'hobby' (a kind of hawk) is then called into question, but the Dictionary offers no etymology of its own. Hobbididance is a fiend like flibbertigibbet; it also occurs in King Lear. The OED has two citations (1603 and 1605). Hoberd (one 1450 example) is a term of reproach. Hobidy-booby (1720, another hapax) possibly means 'scarecrow'. Webster's dictionary of 1864 cites Engl. dial, hobbledygee (from Halliwell) 'with a limping movement' and suggests that we "compare" it with hobbledehoy. The main question is whether the tie between hobbledehoy

Extendedforms (Streckformen,) in English 101 and hobble is original. Chance (1887) answered it in the affirmative. A lad from fourteen upwards, he explains, "is uncertain, physically and morally, whether he will turn out ill or well. And besides this he frequently has an awkward and shambling gait, to which the term may more especially have been applied" (524). But the OED is probably right in stating that hobbledehoy is only now most frequently associated with hobble. Before 1700 the prevailing forms were those with hober- and hobber. Hobbledygee and hobble-de-poise may allude to unsteady movement, but it does not follow that hobbledehoy belongs with them. The picture becomes even more blurry if hubble-te-shives (Halliwell), a synonym of hubbleshow or hubbleshoo 'commotion, hubbub' (1515; OED), is taken into account. There is no need to reduce all hobble- and bubble- words to a single etymon (and no one has tried to connect hubble-te-shives and hobbledehoy). The conjectures on the origin of hobbledehoy, mentioned but not summarized in the OED, are as follows. Ray (1874) derived hobbledehoy from Spanish hombre de hoy 'a man of today'; it is hard to understand what his gloss means. Forby (1830: 161-162) combined OFr. hubi, the past participle of hubir 'cause to thrive by wholesome diet', and hui 'today' (as in Fr. aujourd'hui) and obtained 'one well thriven [sic] now', that is, 'wellgrown lad'. "The change of vowels," he says, "is absolutely nothing. It may have been made after the word became ours, for the rhyme's sake ('Hobi-de-hoy, / Neither man nor boy')." Roger Wilbraham suggested to Forby that Hobby is Robin, and hoy is hoyt, or hoyden. According to this etymology, hobbledehoy is Robin the hoyden, or hoyt. Forby cites Willbraham's suggestion as probable: "... by a metathesis in the last word, we come immediately to hobbite-hoy." Jamieson (1879-1882) also thought that hobbledehoy had come into English from French and cited Fr. hobreau (Cotgrave) 'country squire' ("hobledehoy has been undoubtedly borrowed from the French Hobereau"). Skeat (1885-1887: 302-303 = 1901: 731-732) traced hobbledehoy to Fr. hobel (= hoberel, hob(e)ran), originally 'country squire', which he understood or rather misunderstood as 'villain', and de hoy 'today', that is, he partly repeated the etymologies of Ray and Forby. His translation of the compound from French was 'vile fellow of today'. Neither his gloss nor the contrived etymology has much appeal (see the critique by Chance 1887: 523). Hobel, as Skeat explained, "is a diminutive of OFr. höbe, a hobby, and is allied to the E. hobby, a sparrow-hawk, a hawk of small size and inferior kind, whence it passed into a term of contempt." In the last edition of Skeat's concise dictionary (1910) not a trace of his early ety-

102 Anatoly Liberman mology is left (hobbledehoy is said to be of unknown origin). He calls hoy an unmeaning suffix and mentions Scots hoy 'shout' (noun and verb). But in the full edition he again cites (for comparison) Fr. hober 'remove from place to place' (Cotgrave's gloss) that he mentioned in his early article and adds hoppetihopp 'a giddy, flighty, eccentric man' (from Alsace) and hupperling 'boy who jumps about and cannot be still' (from Low German). However, he ends his entry with reference to hobby 'pet name for Robert', as did Wilbraham, and herein lies the most important clue to the origin of hobbledehoy. Hob 'sprite, elf is short for Rob. This name also happens to be the first element of hobgoblin. Robin Goodfellow and Knecht Ruprecht are medieval names of evil spirits. There is good reason to assume that the spelling hoberdehoy renders hoberd-de-hoy and that hobbledehoy is a folk etymological reshaping of that form. In 1557 Thomas Tusser published the book Fiue Hundred Pointes of Good Husbandries in one of its chapters a human life is divided into twelve periods, each lasting seven years. The chapter begins so (see it in Tusser 1878: 138, 60/3): The first seuen years, bring vp a childe; The next, to learning, for waxing too wilde; The next, keepe under sir hobbard-de-hoy.

This verse has often been discussed in connection with hobbledehoy (Johnson-Todd; Halliwell 1855; W. W. Ε. T. 1852; Η. B. F. 1872; Skeat 18851887: 302-303). Skeat looked on under as an adverb (keepe under / sir hobbard-de-hoy, not keep [the child] / under sir hobbard-de-hoy). No other reading makes sense. The meaning then is: until the age of seven bring up (take care of) your child; between seven and fourteen teach him, lest he get out of hand; between fourteen and twenty-one suppress Sir Hobbard de Hoy. It can hardly be doubted that Sir Hobbard de Hoy is the Devil, the call of the sex, for Tusser had a clear notion of when lust should be satisfied, when it is too late to start, and when it loses its appeal. Evidently, before a young man turns twenty-one, it should be kept in check. Thus Hobbard is a side-form of Robert. (Perhaps a trace of the ancient devilry can be discerned in Engl. dial, hobblety-hoy 'large, unmanageable hilltop': Brocket 1846.) The second half of the puzzle is -dehoy. Hoy, as pointed out by Wilbraham, is reminiscent of hoyden, originally 'rude person of either sex'. Both Richardson and Ogilvie compared them. Skinner took hoyden for an An-

Extendedforms (Streckformen) in English 103 glicized version of MDu. heiden 'heathen, gipsy'. Even the OED accepts this etymology for want of a better one. But in the entry on hoit 'indulge in riotous and noisy mirth; move clumsily', the word, which is the basis of hoity-toity 'riotous behavior, romping', hoyden, as in Brandreth (1885: vii), turns up again in an inconclusive context. Hoyden = hoitin' with intervocalic d < t, is more likely than hoyden = MDu. heiden. Hoyden hardly has any connections with either MDu heiden or Engl, hobbledehoy. Chance (1887: 524) compared hobbledygee ~ hobbledehoy with carters' and waggoners' cry gee-haw. (Is Skeat's hoy, a shout used in Scotland, an attempt to improve on Chance's hypothesis?) According to him, the first word meant 'turning right' and the other, 'turning left'. He visualized balances as in hobberdepoise, with the needle wobbling about the middle point. Then, he says, these terms began to be applied to people, but he was at a loss to explain why a hobbledehoy should have been someone inclining to the left. Chance, an undeservedly forgotten researcher into the history of words, never offered a silly suggestion, but, whatever the origin of hobbledygee may be, it is impossible to accept his etymology of hobbledehoy. The alliteration binding the elements of hobbledehoy is obvious. The RHD (both editions) states that hobbledehoy consists of hoberd (which seems to be correct) +y + hoy from boy for alliteration. The -y + boy part of this etymology (cf. Forby 1830) can be ignored. In contrast, Hughes's conjecture (1954: 606) is persuasive. He posits the initial form *Robert le Roy, that is, King Robert. After Robert became Hobar(d), Roy followed suit, and its r also changed to h. Spitzer (1947: 88) discusses at some length King Rageman, the putative ancestor of ragamuffin. The existence of King Robert is equally probable, and there must have been an element of humor in giving Robert Bruce the affectionate name Kyng Hobbe. Hobard-de-Hoy is a sibling of three fiends mentioned in King Lear, namely Hobberdidance, Obidicut (also known as Haberdicut), and Flibbertigibbet. One of them danced, another may have been fond of cutting capers (Chance 1881: 524, note), but hobber in their names, contrary to Chance's suggestion, did not refer to hopping. At a distance of more than half a millennium, the development of hobbledehoy can be reconstructed so. 1) One of the many names of an evil spirit, or the devil, was *Robert le Roy. 2) In popular speech, Robert was replaced by Hob, Hobard, *Hobert, etc. 3) Roy adjusted to the new pronunciation and became hoy; the result was a piece of alliterative gibberish. 4) *Hobert le Hoy degraded further into Hobert-de-hoy, for fiends' names

104 Anatoly Liberman typically had -de- in the middle. Although a desemanticized word in Tusser's days, it was still remembered as the Devil's name. 5) Evil sprites and all kinds of hobgoblins are occasionally represented as diminutive creatures, and, conversely, children tend to be associated with devils of small stature. A classic example is the history of imp: from 'offshoot' in Old English to 'offspring, child' in Middle English, and to 'child of Devil, little demon' in Early Modern English. 6) When hobartdehoy became the designation of an unwieldy adolescent, hobart turned into hobble, and the word acquired its present day form. It cannot be decided whether at some time -dehoy began to lead a semiindependent existence. Jamieson cited ride cockerdehoy 'sit on one, or on both the shoulders of another, in imitation of riding on horseback' and traced -dehoy in it to Fr. de haut 'from on high'. Chance (1887: 524-525) explained the phrase as meaning originally 'sit on the left shoulder', which is hardly credible. Doesn't it simply mean 'ride in the position of a "cocker" (= fighter, winner) and shout hoy'l Chance concludes his article by saying, "...the dehoy [in cockerdehoy] ought to have the same meaning as in hobbledehoy." His conclusion is not self-evident. The question remains open. Like *Hobard le Roy that changed to Hobard de hoy, to provide the phrase with alliteration, cockerdehoy produced a doublet cockerdecosie. How "cosy" it is to ride on someone's shoulders is a matter of opinion. Both ragamuffin and hobbledhoy were coined as the names of fiends (devils, sprites). Their original meanings are no longer remembered, but the negative connotations they once had have survived. Ragamuffin is a word that can be applied to a person of any age, though perhaps more often to a youngster (see Spitzer's remarks above), as in the title of James Greenwood's novel The True History of a Little Ragamuffin. King Rag(e)man, "Auld Maufl," and King Robert were full-grown devils, but the loss of their status resulted in the loss of stature. For comparison consider the history of Engl. boy. Boy probably has two sources: a baby word for 'brother' and a word for 'devil' (Mandel 1975; Liberman 2000: 204210). In Middle English it may have meant 'executioner'; ragman 'hangman's assistant' has been recorded too. The proper name Boi(e) was also current several centuries before the common name turned up in texts for the first time. Rag-a-muffin and hobble-de-hoy have not only had a similar semantic history: both are extended forms, though with different augments.

Extended forms (Streckformen) in English


6. Conclusion In English, three types of Schroder's Streckformen should be distinguished: 1. words with an inserted syllable, as presumably in ske(da)ddle and fi(na)gle; 2. compounds with -a- (schwa) in the middle, as in cock-a-doodle-doo and ragamuffin; Fr. a and the reflexes of the weakened prepositions of and on have merged in it; 3. compounds with -de- (rarely -te-) in the middle, such as fiddle-de-dee, and hobbledehoy; -de- is traceable to French and Low German (or Dutch). After the -a- and the -de- types had established themselves, native words of the same structure were allowed to appear. Chance (1887: 524, note) remarks that Engl, -de- "is not used as the French de is. It seems rather to mark some loose, often ill-definable relation between the two words which it connects, and may apparently be translated by with regard to, as or like, in, about, on, or towards." His formulation will not hold for -a-, which serves rhythm rather than semantics, whence the productivity of words like jack-a-dandy. However, -a- and -de- have sometimes been used interchangingly: cf. cater-a-fran and cater-de-flamp (Wright; both mean 'askew'); the OED cites raggedemuffin (1612). Streckformen in English belong to the low style: they are (or were at their inception) playful and slangy. Yet none of them should be dismissed by etymologists as fanciful, unmeaning, or meaningless. Language, like sprites, is fond of games, but, unlike them, it plays according to rules.

Note 1.

The following abbreviations are used below: dial. - dialectal, Engl. - English, Fr. - French, Germ. - German, Go. - Gothic, Ital. - Italian, Lat. - Latin, MDu Middle Dutch, ME - Middle English, OE - Old English, OFr. - Old French, Sw. - Swedish.

References Adams, Michael 1999 Another effing euphemism. American Speech 74: 110-112.

106 Anatoly Liberman 2001 2002 AHD3 1992 AHD4 2000 Anonymous 1822

Infixing and interposing in English: A new direction. American Speech 76: 326-331. Meaningful interposing. An accidental form. American Speech 77: 440-441. = The American Heritage Dictionary of the English Language. 3rd ed. Boston: Houghton Mifflin. = The American Heritage Dictionary of the English Language. 4th ed. Boston: Houghton Mifflin. Review of: Robert Nares, A Glossary; or. Collection of Words, Phrases, Names, and Allusions to Customs, Proverbs, etc. Which have been Thought to Require Illustration, in the Works of English Authors, Particularly Shakespeare, and His Contemporaries. The Academy 28: 328-329. Capabarre. The Mariner's Mirror 2: 164.

1912 Bause, Josef 1919 Kabuss. Korrespondenzblatt des Vereins für niederdeutsche Sprachforschung 37: 46-47. Behaghel, Otto 1906 Review of: Schröder, Heinrich, 1906. Literaturblatt für germanische und romanische Philologie 27: 401-402. 1923 Humor und Spieltrieb in der deutschen Sprache. Neophilologus 8: 180-193. Brandreth, E. L. 1885 Untitled. Transactions of the Philological Society 21 (1885-1887): vi-viii. Brocket, John T. 1846 A Glossary of North Country Words. 3rd ed. Newcastle upon Tyne: Emerson Charnley; London: Simpkin, Marshall. Brendum-Nielsen, Johannes 1924 Om ordet kissenlinke letfasrdigt pigebarn. In: Festskrift tillägnad Hugo Pipping pä hans sextioarsdag den 5 november 1924: 51-55. Helsingfors: Mercator. The Century Dictionary 1911 = The Century Dictionary.... Benjamin E. Smith (ed.). New York: The Century Co. Chance, F. 1887 Hobbledehoy. Notes and Queries VII (4): 523-525.

Extended forms (Streckformen,) in English


Coetzee, Anna E. 1995 Kaboems, kabolder, kerjakker, karbonkel, karfoefel: Vanwaar de hele kaboedel. Suid-Afrikaanse Tydskrifvir Taalkunde/South African Journal of Linguistics, Supplement 28: 27-44. Cohen, Gerald L. 1976 Skedaddle. Comments on Etymology V (12-13): 1-8. 1979 Skedaddle Revisited. Comments on Etymology VIII (10-11): 1-42. 1985 Etymology of skedaddle and related forms. In his Studies in Slang, Part I. Forum Anglicum 14/1, 26-63. Frankfurt am Main: Peter Lang. Cohen, Gerald L. et al. 1979 Concerning skedaddle. Comments on Etymology VIII (15): 20-24. Cotgrave, Randle 1611 A Dictionarie of the French and English Tongues. London: Printed by Adam Islip. DARE 1985= Dictionary of American Regional English. Frederick G. Cassidy, (chief ed.), and Joan H. Hall, (associate ed.). Cambridge, Massachusetts/London, England: The Belknap Press of Harvard University Press. De Bont, A.P. 1948 Over beduit(je) en wat dit meer zij: Tijdschrift voor Nederlandse Taal- en Letterkunde 66: 23—42. Duden 1983 = Duden Deutsches Universalwörterbuch. Bibliographisches Institut Mannheim - Wien - Zürich: Dudenverlag. Forby, Robert 1830 The Vocabulary ofEastAnglia.... London: J-B. Nichols and Son. Griffith, R. H. 1954 Phenagling. Modern Language Notes 39: 291-292. Grose, Francis 1796 A Classical Dictionary of the Vulgar Tongue. London: S. Hogser. Η. B. F. 1872 Hobbledehoy. Notes and Queries IV (9): 147-148. Halliwell, James O. 1855 A Dictionary of Archaic and Provincial Words.... London: Thomas and William Boone. Hellquist, Elof 1948 Svensk etymologisk ordbok. 3rd ed. Lund: C. W. K. Gleerup. Hughes, John P. 1954 On "h" for "r" in English proper names. The Journal of English and Germanic Philology 53: 601-612.

108 Anatoly Liberman Jamieson, John 1879-82 An Etymological Dictionary of the Scottish Language, edited by John Longmuir. Paisley: Alexander Gardner. Johnson-Todd 1827 = Samuel Johnson, A Dictionary of the English Language. Revised by H. J. Todd. London: Longman. Junius, Franciscus 1743 Etymologicum Anglicanum.... Edited by Edward Lye. Oxonii: Ε Theatro Sheldoniano. Kluge, Friedrich 1906 Review of: Heinrich Schröder, 1906. Literaturblatt für germanische und romanische Philologie 27: 393-401. Kluge-Seebold 1995 = Friedrich Kluge. Etymologisches Wörterbuch der deutschen Sprache 1/21. 23rd ed. Edited by Elmar Seebold. Berlin/New York: Walter de Gruyter. Koziol, Herbert 1937 Handbuch der englischen Wortbildung. Sammlung germanischer Lehr- und Handbücher 1/21. Heidelberg: Carl Winter. Kück, Eduard 1905 Niederdeutsche Beiträge zum Deutschen Wörterbuch. Beilage zum Jahresbericht des Gymnasiums zu Friedenau. Ostern 1905. Friedenau: Leo Schultz. Liberman, Anatoly 1994 Etymological studies VI: Some obscure English words.... General Linguistics 33:164-178. 2000 The Etymology of English boy, beacon, and buoy. American Journal of Germanic Languages and Literatures 12: 201-234. MED 1956-2000 = Hans Kurath et al., editors, Middle English Dictionary. Ann Arbor: University of Michigan Press. Mackay, Charles 1877 The Gaelic Etymology of the Languages of Western Europe.... London: Published for the author by N. Trübner and Co. Mandel, Jerome 1975 "Boy" as devil in Chaucer. Papers on Language and Literature 11: 407-411. Marchand, Hans 1969 The Categories and Types of Present-Day English Word-Formation: A Synchronic-Diachronic Approach. 2nd ed. Munich: C. H. Beck. McMillan, James B. 1980 Infixing and interposing. American Speech 55: 163-183.

Extended forms (Streckformen,) in English


Murray, lames A. H. 1890 Cockney. The Acadamy 37: 320-321. ODEE 1966 = The Oxford Dictionary of English Etymology. Edited by C. T. Onions, with the assistance of G. W. S. Friedrichsen and R. W. Burchfield. Oxford: Clarendon Press. OED 1992 = The Oxford English Dictionary. James A. H. Murray et al., (eds.), Original edition, 1884-1928. 2nd ed., by J. A. Simpson and E. S. C. Weiner. Oxford: Clarendon Press. Ogilvie, John, (ed.) 1882 The Imperial Dictionary of the English Language. London: Blackie & Son. RHD 1966 = The Random House Dictionary of the English Language. Jess M. Stein, (ed.). 2nd ed. by Stuart B. Flexner. New York: Random House. Ray, John 1874 A Collection of English Words, Not Generally Used.... Edited by Walter W. Skeat. English Dialect Society Publications, vol. 2. London: Published for the English Dialect Society by N. Trübner & Co. Original edition, 1674. Richardson, Charles 1858 A New Dictionary of the English Language, Combining Explanation with Etymology. London: Bell and Daldy. Schröder, Heinrich 1903 Streckformen. Beiträge zur Geschichte der deutschen Sprache und Literatur 29: 346-354. 1906 Beiträge zur germanischen Sprach- und Kulturgeschichte. I. Streckformen. Ein Beitrag zur Lehre von der Wortentstehung und der germanischen Wortbetonung. Germanische Bibliothek II/l: Untersuchungen und Texte. Heidelberg: Carl Winter. Seebold, Elmar See Kluge-Seebold Skeat, Walter W. 1882 An Etymological Dictionary of the English Language. Oxford: Clarendon Press. 2nd ed., 1884. 3rd ed., 1897. 4th ed., 1910. 1885-87 Notes on English Etymology. Transactions of the Philologial Society 21: 1-12, 283-333,690-722. 1886 (ed.) The Vision of William Concerning Piers the Plowman in Three Parallel Texts..., 2 volumes. Oxford: Clarendon Press. 1901 Notes on English Etymology. Oxford: Clarendon Press. 1910 A Concise Etymological Dictionary of the English Language. Oxford: Clarendon Press.

110 Anatoly Liberman Skinner, Stephen 1671 Etymologicum Lirtguce Anglicance.... London: Typis T. Roycroft. Smythe Palmer, A. 1910 Folk-lore in word-lore. The Nineteenth Century and After 68: 545557. Spitzer, Leo 1925 Aus der Werkstatt des Etymologen. Jahrbuch für Philologie 1: 129159. 1947 Ragamuffin, ragman, rigmarole and rogue. Modern Language Notes 62: 85-93. Tusser, Thomas 1878 Fiue Hundred Pointes of Good Husbandrie, edited by W. Payne and Sydney J. Heritage. English Dialect Society Publications, vol. 8. London: Published for the English Dialect Society by N. Trübner & Co. UED 1932 = The Universal Dictionary of the English Language. Hernry C. Wyld, (ed.). London: Routledge & Kegan Paul. WNWD 1970 = Webster's New World Dictionary of the American Language. David B. Guralnik, (ed.). 2nd ed., Cleveland/New York: The World Publishing Company. 3rd ed., by Victoria Neufeldt and David Guralnik. New York: Webster's New World, 1986. W. W. Ε. T. 1852 Sir Hobbard de Hoy. Notes and Queries I (5): 468. Webster, Noah 1828 An American Dictionary of the English Language. New York: S. Converse. 1864 An American Dictionary of the English Languag. Edited by Chauncey A. Goodrich and Noah Porter. Springfield, Massachusetts: G. & C. Merriam Company. 1890 Webster's International Dictionary of the English Language. Edited. by Noah Porter. Springfield, Massachusetts: G. & C. Merriam Company. Webster2 1934 = Webster's New International Dictionary of the English Language. William Allen Neilson and Thomas A. Knott, (eds.). 2nd ed., Springfield, Massachusetts: G. & C. Merriam Company. Weekley, Ernest 1921 An Etymological Dictionary of Modern English. London: John Murray. Wright, Joseph 1898-1905 The English Dialect Dictionary. London: H. Frowde.

Linguistic change in words one owns: How trademarks become "generic" Ronald R. Butters and Jennifer Westerhaus

1. Introduction Although trademarks abound in the lexicon of modern and postmodern life, they are treated by linguists (among whom we count lexicographers and other philologists) as something of an embarrassment. In particular, dictionaries - even unabridged ones - have entries for only the very most frequent of trademarks; trademarks are seldom discussed in the methodological and theoretical literature of lexicography and lexicology; and linguists have until quite recently ignored the contribution of this vast word stock to the history of twentieth-century English (but see Adams 1997; Baron 1989; Butters 2001; Clankie 2001, 2002; and Shuy 2002).1 Trademarks present a number of complex and interesting synchronic and diachronic problems, however, one of which is genericness. As a term of art in lexicography, the meanings of both generic (as applied to individual words) and trademark are essentially borrowed from the law. Sidney Landau writes that a trademark is "a symbol or name used by a maker of a product to distinguish the product from others of its kind" (2001: 405). Landau distinguishes trademarks from generic words, which are "ordinary words ... commonly used ... not for a brand of a kind ofthing but for the kind of the thing itself' (406). Thus Pepsi-Cola, Pepsi, Coca-Cola, and Coke are famous trademarks for commercial products that are referred to generically by such terms as soft drinks, pop, soda, tonic, and cola drinks. It is important to note that this concept of genericness is somewhat different from that found most commonly in the linguistic subdisciplines of syntax, semantics, and pragmatics, wherein any noun can be said to be in generic usage if, in a given context, it is to be construed as referring to all members of the class to which the noun in question intrinsically refers. For example, in the sentence, A giraffe is a mammal with a long neck, the word giraffe is said to be a "generic" reference because the predication of the sentence applies to all giraffes. In a sentence such as A giraffe bit my uncle, however, the word giraffe refers to a particular giraffe, not all giraffes, and

112 Ronald R. Butters and Jennifer Westerhaus is therefore not a generic reference under this definition. In lexicography (and the law), on the other hand, genericness is an intrinsic property of some words (normally nouns), but is not intrinsically a property of other words.

2. Trademark genericness and historical linguistics The relationship between generic words and trademarks is of interest to historical linguistics in a number of ways, central among which is the important fact that the status of a word with respect to its genericness can be open to question and can even change through time. Lexicographers and law-school professors cite such words as aspirin, shredded wheat, thermos, and escalator as words that once were trademarks but now are generics; lawyers term this process of historical linguistic change "genericide."2 And at least one term, Singer as applied to sewing machines, began as a nontrademark surname, became a trademark, underwent genericide, and then later "recaptured" its trademark status (Singer Mfg. Co. v. June Mfg. Co. 163 U.S. 169 (1896)). Compared to genericide, however, such resurrection is a rare occurrence. Genericide can be viewed as a subcategory of broadening, similar therefore to the process that has affected scores of English words - for example, dog, which at one time referred to a specific kind of canis familiaris rather than to dogs in general.3 Broadening, however, does not affect the genericness of ordinary words: dog is still generic, whether it refers only to longhaired canines that are used to herd sheep or to all canines. A trademark thus is really a kind of proper noun, but one that still refers to a class of things (e.g., Mercedes automobiles) rather than to a unique entity (e.g., Mercedes Norton, the first author's third-grade music teacher) or place (e.g., Mercedes, Texas). Furthermore, the sociolinguistic processes that bring about genericide are surely atypical of lexicosemantic change. Whatever powers caused hound and dog to change places in modern English (much less what mysterious medieval sociolinguistic forces created dog at all), they were in all likelihood forces unrelated to the legal system. Moreover, they were surely unrelated to the forces of modern marketing and advertising that have come increasingly into play in the creation of trademarks, the evolution of potential genericness, and what we believe to be the increasing inoculation of trademarks against potential genericide. At the same time, however,

Linguistic change in words one owns


traditionally identified linguistic processes also seem to be at work, in particular the tendency of ordinary speakers to make metaphorical and synecdochical (i.e., shorthand) utterances (see §4.1 below).

3. The social basis of trademarks in the law Owning a trademark is not a right to a word, but rather a right to a particular, distinctive use of a word. Unlike copyrights, which protect original works of authorship from misuse and misattribution, and patents, which grant inventors exclusive rights to their original inventions, trademarks are symbols and, in many cases, already were a part of the lexicon of the language before they were appropriated for use in commerce in connection with the sale of goods or services (e.g., Tide soap, Dove ice cream bar). American rights to trademarks are protected under the common law and via various state and federal statutes, most particularly through registration under the federal law known as the Lanham Act (15 U.S.C. § § 1501 et seq., available at ). Legally, at least in the United States, one comes to own a trademark simply by - under the proper conditions - using it. Chief among the proper conditions is that no one else owns the right to use the word to indentify goods or services for which the would-be owner wants to use it. In the earliest days of modem trademark usage (the later nineteenth and earlier twentieth centuries), some novel trademarks (e.g., cellophane, zipper, aspirin) evolved as names for new inventions for which the owners acquired patents. So long as only one company could manufacture the product because the ownership of a patent gave it a legal monopoly, the issue of what it was called was moot. But when the patents ran out, the trademarks were simply appropriated by the new manufacturers. As time progressed, however, the original patent owners began to perceive that the names themselves were of considerable financial value, and, they argued, it was not only unfair for them to lose the proprietary interest in their good name, it was also a disservice to a public who had come to rely upon the quality of the product indicated by the name. So long as the public relied on the brand name as a way of distinguishing between the old reliable manufacturer and the upstart new manufacturers, basic fairness to the consumer required that the brand name be protected from appropriation. Accordingly, the federal government - and the governments of the fifty states as well - created trademark boards to protect brand names

114 Ronald R. Butters and Jennifer Westerhaus through a registration process. In so doing, however, they were merely formalizing a process that the courts had already begun on the basis of common law. The basic test, then, for the legitimacy of a trademark thus became essentially the same as the linguistic test for word meanings: What is it that ordinary speakers of the language believe the word refers to? As the famous judge, Learned Hand, expressed it in the 1921 trademark case, Bayer Co. v. United Drug Company, the basic test is, "What do the buyers understand by the word for whose use the parties are contending?" (Bayer Co. v. United Drug Co., 272 F. 505, 509 (S.D.N.Y. 1921)). Because, it was decided in that case, the "consuming public had learned to know 'aspirin' as the accepted name for the drug," the word had indeed become generic. Similarly, with respect to cellophane, in 1920 a court ruled that "Customers knew no other name than 'cellophane' which they could use if they wished to buy a particular article" (DuPont Cellophane Co. v. Waxed Products Co., 85 F.2d 75 (2d Cir. 1936)). On the other hand, in Coca-Cola v. Koke (1920), Coca-Cola was deemed not to be generic because the name "means to most persons the plaintiffs familiar product." Even though coca is itself a word associated with cocaine, once the essential ingredient of Coca-Cola, and hence (in Landau's description) an "ordinary word commonly used not for a brand of a kind of thing but for the kind of thing itself," it nonetheless was not considered to be generic because it had acquired what is legally known as "secondary meaning" in the public mind as the particular product of a particular company (Coca-Cola Co. v. Koke Co. of Am., 254 U.S. 143,146 (1920)). In short, what is of historical linguistic interest here is that, in a somewhat circular process, significant aspects of the meaning of particular words have been determined and in effect established by governmental acts. The courts first decided - based upon what they deemed to be the best available linguistic evidence - what constituted the public's understanding of the meaning of a term such as Coca-Cola or aspirin·, and the judicial implementation of this decision, in turn, actually prohibited (in the case of Coca Cola) or allowed (in the case of aspirin) the uninhibited public commercial usage of that same term. As a result, the public was in turn presented with data that itself further reified the foundation of the court's decisions.

Linguistic change in words one owns


4. Sociolinguistic bases of the concept of trademark genericness Two other related sociolinguistic issues are of interest to the history of trademarks in American English during the past hundred or so years. First is the nature of the "public's understanding" of the putative genericness of trademarks and how the legal concept of genericness may or may not intersect with a lexicographer's view. Second is the fact that genericide as a productive linguistic process apparently is dying out.

4.1. Lexicography and the public's understanding The perspectives of lexicographers and lawyers are sometimes thought to be at odds on the issue of putative genericness (Landau 2001; see also Clankie 2001, 2002; Shuy 2002: 13). Thus Landau, invoking the traditional inductive methodology of dictionary makers, concludes that a number of trademarks are erroneously left unlabeled in dictionaries, even though, in his view, they ought to be labeled "generic." As he sees it, editors frequently succumb to pressures from their publishers, who, fearful that they will receive threatening letters from trademark lawyers, excise legitimate "generic" labelings from dictionaries. Specifically, he argues that BandAid, a registered trademark owned for decades by Johnson&Johnson, Inc., should be labeled "generic" in dictionaries of contemporary American English (408ff.). As a practicing lexicographer, Landau certainly speaks with authority about the pressures lexicographers may feel from publishers and lawyers. A closer look at his argument, however, suggests that his criticism of the absence of the label "generic" in dictionary entries (echoed in Clankie 2001, 2002; Shuy 2002) may miss some key social and linguistic considerations. The primary meaning of Band-Aid is surely well known to native speakers of American English: the product is a relatively small adhesive plastic strip with a small absorbent patch attached to the underside, used to cover small wounds. In England, they are generically called plasters. In the United States, companies other than Johnson&Johnson manufacture virtually identical products, and while admittedly it is a bit difficult to come up with a generic term for these things in American English, "plastic bandage strips" works for most people. (One of the legal tests for the lack of genericness historically has been the ready availability of an alternative

116 Ronald R. Butters and Jennifer Westerham generic term.) No manufacturers except Johnson&Johnson advertise their products as "band-aids," however. (Another of the legal tests for the lack of genericness historically has been how little the putative trademark is actually used generically in commerce; or, conversely, how well it has been "policed" by the company that claims ownership.) Furthermore, Johnson&Johnson uses the term Band-Aid to label products other than plastic bandage strips, for example, liquid bandages, tape, and gauze rolls. Since Band-Aid refers to products other than plastic bandage strips, the putative genericness of the term is further diminished. Landau bases his assertion that Band-Aid has undergone genericide entirely upon such data as the following, which are actual sentences collected from various media (408-409): 1.

And while the old pedagogy has failed for many reasons, clearly one of them is that its fundamental principles are wrong. It has mistaken a band-aid for the science of medicine. 2. Arofsky continues: "Everything the Knicks try is a Band-Aid remedy. There's no real vision of the team concept." 3. From my point of view, I've seen no change in how we're doing business to get qualified controllers. I don't see any light at the end of the tunnel. All we've had is a helter-skelter, Band-Aid, stop-gap type of approach. There's seemingly no plan to deal with the situation. 4. I see this as a Band-Aid, not a long-term program, says Mr. Pratt. 5. The notion of production control, as it is usually called, is indeed an appealing one: Enough of the band-aids of interdiction and domestic law-enforcement. ... 6. A minor near-term event like a Band-Aid on the budget deficit doesn't erase the five years of overconsumption and under-savings that contributed to the problem. 7. According to published accounts, Bush apparently intends to use the money freed up by defense spending cuts as an election-year BandAid for the health-care crisis. 8. Some policymakers would rather use a year-to-year approach, providing only annual subsidies or rent vouchers for tenants. This kind of band-aid strategy has the advantage of masking the budgetary effect of any expenditures. Clearly, the authors of these sentences are not using the word Band-Aid to refer specifically and exclusively to the products of Johnson&Johnson.

Linguistic change in words one owns 117 Such examples, however, simply do not in themselves demonstrate that, in the minds of the public, Band-Aid has become generic. Rather, Landau's data prove only that the term Band-Aid is so well known that it is in common use as a metaphor for a minor (and usually inadequate) remedy. For one thing, a majority of the examples (5 of 8) capitalize the word, thus treating it as a proper noun - another one of the tests that courts have employed as evidence that the users of the term think of it as a trademark, not a generic term. More important, all of Landau's examples are figurative; none of them merely uses the term in its ordinary usage to refer to small plastic bandages per se. The mere existence of metaphorical citations has relatively little bearing on the genericness question, since it does not address the crucial lexicographical issue of what the public at large actually believes about the reference of the term Band-Aid. That is to say, the fact that the term is sometimes used metaphorically does not tell us whether the public knows that the term is a trademark upon which consumers presumably can rely when making purchases of small plastic self-adhesive bandages. 4 Still, it is not at all unlikely that a search of newspaper and magazine articles from recent years would turn up some nonmetaphorical, genericseeming uses of bandaid or band-aid (though Landau presents none), as in the following hypothetical examples: 1. 2. 3.

[from a short story]: "Go get me some neosporin and a band-aid from the bathroom," Mother said. "Your little brother has cut his finger." [from a gossip column]: Why was Elton John wearing a Band-Aid? [from a sports story]: The floor of the locker-room was filthy. Dirty socks, abandoned razors, used bandaids - you name it, it was there.

No native speaker of contemporary American English would be surprised at such uses. Again, however, one cannot necessarily conclude from such examples that the writers would have been unaware that Band-Aid is a trademark. Synecdoche is a technical term in linguistics and literary studies that is defined as "semantic change shifting the meaning of a word because of a metonymic association of part and whole, species and genus, or material and product" (Pyles and Algeo 1993: 357); or, as the definition in Harmon and Holman (2000, s.v. synecdoche) reads: A trope in which a part signifies the whole or the whole signifies the part. To be clear, a good synecdoche ought to be based on an important part of the whole and, usually, the part standing for the whole ought to be

118 Ronald R. Butters and Jennifer Westerhaus directly associated with the subject under discussion. Thus under the first restriction we say 'threads' and 'wheels' for 'clothes' and 'car,' and under the second we speak of infantry on the march as foot rather than as hands just as we use hands rather than foot for people who work at manual labor. Both the linguists and the literary scholars here define a highly productive synchronic and diachronic process in language, sometimes used for its striking effect as a figure of speech (as in All hands on deck!, meaning 'All sailors on deck!'), sometimes as a kind of linguistic shorthand: rather than uttering an entire generic term, speakers and writers may use a briefer, nongeneric synecdochical term that is associated with the generic term. Thus speakers will sometimes say Kleenex to refer to a paper tissue and Xerox to refer to a photocopying machine, even though the speakers know full well that the terms they have employed are in fact brand names which are technically inexact labels for the items to which the speakers are referring. Such a speaker is not truly mistaking the part (Kleenex, Xerox) for the whole (paper tissues, photocopying machines), any more than the speaker who says threads to refer to clothes seriously believes that threads refers by definition to the entire universe of garments; and hearers are not confused precisely because they know that synecdoche is a normal linguistic process (and that threads 'clothes' is a commonplace synecdochical usage). Landau offers no philological evidence that, when Americans say or write Band-Aid to refer to small plastic bandages without apparent reference to the specific products of Johnson&Johnson, they are not merely making a synecdochical reference to the product itself. It seems clear that Landau's conclusions are based also on a somewhat faulty notion of what genericness means, at least in the legal context (which is the source, after all, of the linguistic sense of lexical genericness). The fundamental test for genericness must be what speakers themselves know and believe about the words in question. Any written evidence that might seem to support a conclusion that a Band-Aid is generic must be weighed against the vast amount of empirical philological data in which Band-Aid (and Kleenex and Xerox) are clearly used as brand names, as they are in immense numbers of cases (e.g., in advertising) to which the public is exposed. The appearance of Band-Aid in a relatively few environments wherein it is not clear whether the speakers intended synecdochical reference or generic reference is obviously not determinative. Moreover, although traditional lexicography, a philological inquiry, relies heavily upon inductive conclusions based primarily upon print-source

Linguistic change in words one owns


data, the use of introspective judgments is also a widely practiced methodology in American linguistics, and, in the case of Band-Aid and many other famous trademarks, linguists must also weigh their own linguistic knowledge as native speakers against the empirical data. As Wierzbicka notes (1985: 70): "... the observation of language in use under experimental conditions can produce artificial results unless the 'experiment' is prepared in such a manner as to engage the full knowledge and semantic intuition of the subjects. For this to happen, the intuition of the theorist must in turn be fully engaged." Such introspection obviously also argues against Landau's inductive conclusions based upon the figurative uses of Band-Aid that he reports in his book. The intuitions of some linguists (and other speakers of the language) who may not have known that Band-Aid is a trademark surely are outweighed by the introspective judgments of those who do have such knowledge (a group that, ironically, includes anyone who reads Landau's book). Of course, if one wants to know what the public thinks about the meanings of words in controversial cases, one might also consider asking them directly. While questionnaire methodology is not generally used by lexicographers (The Dictionary of American Regional English is something of a major exception to this), the use of market surveys in trademark cases where genericness is an issue is a common procedure. Courts generally give weight to the results of market surveys conducted by sociolinguists, psychologists, and advertising specialists - as well as to inductive lexicographical conclusions based on the philological record. This is at least in part because dictionaries generally are relatively conservative about lexical innovations and change. Such surveys, properly done, reify the results of traditional philological methods.

4.2. The putative death of genericide Finally, it is also the case that genericide in itself seems to be something of a dying historical linguistic process. The last major cases took place in the 1960s, over 40 years ago, when thermos and trampoline were declared by courts to be generic (King-Seeley Thermos Co. v. Aladdin Indus., 321 F.2d 577, 579 (2d Cir. 1963); Nissen Trampoline Co. v. American Trampoline Co., 193 F. Supp. 745, 749 (S.D. Iowa 1961)). The 1982 declaration by the Ninth Circuit court that Monopoly as a board game was generic (AntiMonopoly, Inc. v. General Mills Fun Group, 611 F.2d 296, 306 (9th Cir.

120 Ronald R. Butters and Jennifer Westerhaus 1979)) was expressly repudiated by a 1984 Act of Congress. By contrast, such widely different trademarks as Teflon®, the Mc of McDonald's® and McNuggets®, and the Beanie of Beanie Babies® have been adjudicated to be valid trademarks and not generic (E.I. DuPont de Nemours & Co., Inc. v. Yoshida Int'l, 393 F. Supp. 502, 528 (E.D.N.Y. 1975); Quality Inns Infi v. McDonald's Corp., 695 F. Supp. 198 (D. Md. 1988); McDonald's Corp. v. Druck & Gerner, P.C., 814 F. Supp. 1127 (N.D.N.Y. 1988); Ty Inc. v. The Jones Group, Inc. Case No. 99 C 2057; Ty Inc. v. Softbellys Inc. and Positive Products Consultants, Case No. 00 C 5230; Ty Inc. v. Perryman, 306 F.3d 509 (7th Cir. 2002)). Thus Landau's belief that dictionary makers are "weasels" in their refusal to declare numerous trademarks to be generic is apparently not shared by the courts. Linguists sometimes have met the seeming conflict between courts and lexicographers with a certain amount of cynicism. Clearly, numerous trademarks are so famous that they are often used, especially in speech, in ways that (at first glance at least) might seem to be generic: Jello®, KleenexXerox®, Velcro®, BarbieRoller Blade®, and Jeep®, to name only a few, are certainly terms that most people have used in informal situations to references not specifically the "brand of a kind of thing but the kind of the thing itself." Doubtless it is the case that such terms rarely are defended in court against genericness claims simply because no manufacturer has dared risk a lawsuit by attempting to use these trademarks for their competing products. Lawsuits are expensive, and in general, it is probably much more practical for businesses to use undisputed generic terms {gelatin dessert, photocopy, etc.) than to go to court with all the expense that that entails. As the Beanie, Mc, and Teflon cases make clear, it is not easy to win genericness cases, at least when the trademarks are not simply long-standing descriptive terms for the products they describe (e.g., the use of light and lite to indicate reduced-calorie content in foods and beverages). Such forces are sometimes seen as a distressingly unnatural interference with the natural linguistic processes that govern diachronic change, as well as a kind of restraint on the legal rights of freedom of speech and of the press (e.g., Clankie 2001, 2002; Shuy 2002). In our view, there is nothing less natural about the historical linguistic influence of modern advertising and the courts than there is about any other contemporary sociolinguistic process; lexicographical labeling ought not be based on resentment of the realities of everyday life, but on the nature of that reality. If genericide is a dying process, it is precisely because the courts in general - and most often these are juries, not judges - are

Linguistic change in words one owns 121 persuaded (often on the basis of market surveys and the lexicographical reports, but undoubtedly also on the basis of their own intuitions as native speakers) that the public at large is in fact highly aware that brand names are indeed brand names, even though a limited look at the empirical record might suggest otherwise. If there has been a change in the rate of genericide since the middle of the twentieth century, it indeed may well be because of increased vigilance on the part of trademark owners, who continue to strive to inform the public, especially through advertising, that their lexical property is in fact not generic, suggesting alternative generic terms.5 Lexicographical practice ought to be based on the fact that, while people do use trademarks as verbal shorthand, they are generally aware that they are, in reality, speaking synecdochically: that a Band-Aid® is a special kind of plastic bandage, that Jello® is a special kind of gelatin dessert, and that a reference to a "Disney" Beanie® is just a convenient synecdoche and not compelling evidence of genericide.

Notes 1.



John Simpson, Chief Editor, Oxford English Dictionary, notes (2003): "There are a number of myths about the Oxford English Dictionary, one of the most prevalent of which is that it includes every word, and every meaning of every word, which has ever formed part of the English language. Such an objective could never be fully achieved." The origin of the term genericide is uncertain, but its use in case law dates at least as far back as 1987, in G. Heileman Brewing Co. v. Anheuser-Busch, Inc., where the court used the term to describe a situation in which a trade name comes to be known by the consuming public as the name of a product itself; the court found that this was not the case for the term LA as applied to low-alcohol beer, ruling that LA is merely descriptive and not generic (676 F. Supp. 1436, 1488 (E.D. Wis. 1987)). See also Lefkowitz and Graham (1983), who use the term to describe Parker Brothers' loss of the trademark Monopoly (Anti-Monopoly Inc. v. General Mills Fun Group, 611 F.2d 296, 306 (9th Cir. 1979)). Genericide is used by McCarthy (2003: §12:1) to describe the situation in which "one seller develops trademark rights in a term which a majority of the relevant public then appropriates as the name of a product." According to McCarthy, in that situation, "trademark rights may cease. ... Once declared to be a generic name, the designation enters the 'linguistic commons' and is free for all to use." Two examples given in the online Oxford English Dictionary indicate that dog originally referred only to sheep dogs and other long-haired dogs: "1398 TRE-

122 Ronald R. Butters and Jennifer Westerhaus



TREVISA Barth. De P.R. XVIII. xxvi. (1495) 786 A gentyll hounde...hath lesse flesshe than a dogge and shorter heere and more thynne. cl440 Promp. Parv. 125/1 Dogge, shyppe-herdys hownde, gregarius." The subject of which figurative uses belong in a desk-top dictionary is itself a question that Landau elsewhere (2001: 201-2) notes is vexing. But even if the use of Band-Aid in this metaphorical sense were so commonplace as to warrant a separate meaning-entry, the issue of whether or not Band-Aid is in the public mind a brand name or a generic would not be determinable on this basis alone. Ty Inc., the manufacturer of Beanie Babies, does not advertise at all, however.

References Adams, Michael 1997 Lexical property rights: Trademarks in American dictionaries. Unpublished paper presented at the American Dialect Society Meeting, Chicago, IL, 4 January 1997. Baron, Dennis 1989 Word law. Verbatim 16 (1): l ^ t . Butters, Ronald R. 2001 Genericness in lexicography, general linguistics, and American trademark law. Unpublished paper presented at the Fifth Biennial Conference, International Association of Forensic Linguists, University of Malta, July 2001. Clankie, Shawn M. 2001 Why Bud Weiser can sell cars (but not beer). Verbatim 26 (3): 3. 2002 A Theory of Genericization and Brand Name Change. Edwin Mellon Press. DARE 1985 = Dictionary of American Regional English. Frederick G. Cassidy, (chief ed.), and Joan H. Hall, (associate ed.). Cambridge, Massachusetts/London, England: The Belknap Press of Harvard University Press. Harmon, William, and C. Hugh Holman 2000 A Handbook to Literature. 8th ed. Prentice Hall. Landau, Sidney 2001 Dictionaries: The Art and Craft of Lexicography. 2nd ed. Cambridge: Cambridge University Press. Lefkowitz, Saul, and Bary W. Graham 1983 Court Rules that "Monopoly" Has Suffered Genericide. Legal Times, 7 March 1983.

Linguistic change in words one owns


McCarthy, J. Thomas 2003 McCarthy on Trademarks. 4th ed. Minneapolis: The West Group, a Thomson business. Oxford English Dictionary Online 2003 . Pyles, Thomas, and John Algeo 1993 The Origins and Development of the English Language. 4th ed. Fort Worth: Harcourt Brace Jovanovich. Shuy, Roger 2002 Linguistic Battles in Trademark Disputes. Houndmills, UK/New York: Palgrave. Simpson, John 2003 Preface to the 3rd ed.: Distractions. Oxford English Dictionary Online. . Wierzbicka, Anna 1985 Lexicography and Conceptual Analysis. Ann Arbor: Karoma.

Section 2 Corpus- and text-based studies

Introduction: Corpus- and text-based studies Anne Curzan and Kimberly Emmons

Some of the very exciting new research in English historical linguistics has employed corpus linguistic methodologies and exploited other new electronic capabilities. Recent development in corpus linguistics have made it possible to collect and analyze large databases of texts - for example, ARCHER, the corpus used by Douglas Biber in this section, encompasses approximately 1.7 million words. These large samples are rendered manageable by the process of tagging (the addition of classification labels to words and other forms of interest) and computerized concordancing software. Since their early implementation in the 1960s, electronic corpora have been used by linguists to supplement traditional native-speaker intuitions about language structures. The first two essays, as well as the response, in this section engage with the methodologies and technologies of such text collections. These three studies demonstrate the ways in which searchable corpora allow broader generalizatios about the mechanisms of linguistic change, as well as demand accountability to the evidence of actual language use. All three authors also work to navigate the relationship of quantitative and qualitative analysis - a central forcus in the development of modern corpus linguistic research. All three studies highlight the ways in which quantitative findings about frequency and distribution are enhanced by close, qualitative work with specific textual examples, examining local and more subjective factors in language use and change. The last two articles in this section focus specifically on one historical text, demonstrating how critical close analysis of any one text in a collection is for responsible historical linguistic research. Both authors discuss texts in the process of being made available electronically, in hypertext editions, yet address the limits of research that does not work closely with the original manuscript - or does not exploit resources that are closely and faithfully linked to original manuscripts. Working both with complex corpora and with editions of single texts, all of the authors in this section raise methodological and procedural concerns for future scholarship. In the end, their essays not only provide care-

128 Anne Curzan and Kimberly Emmons ful textual analyses, but they also identify important concerns about the composition of corpora, examine linguistic variation across register and genre, and encourage more careful scrutiny of scholarly editions. Susan M. Fitzmaurice, in the lead article "The meanings and uses of the progressive construction in an early eighteenth-century English network," extends her own previous scholarship On what she terms the "expressive or subjective progressive construction." To explore the frequency of these constructions, which she argues participate "in the realization of speaker self-expression in grammar," Fitzmaurice uses the Network of Eighteenthcentury English Texts (NEET), a corpus containing texts in four registers written by individuals comprising a social network centered around essayist Joseph Addison. In addition to providing data on the frequency and locations of progressive constructions, Fitzmaurice argues, the corpus offers a chance to consider the effects of peer group on the usage of such constructions. Indeed, as Erik Smitterberg notes in his response to her article, Fitzmaurice's findings hint at potential variables (including membership in the network, gender, and socioeconomic status) that might condition experiential/subjective progressive constructions. Both Fitzmaurice and Smitterberg acknowledge the potential unreliability of intuitive categorization of instances of the progressive, and they suggest systematic criteria for locating potential tokens. Fitzmaurice develops a set of lexicogrammatical contexts in which she expects the experiential progressive constructions to occur; Smitterberg concludes that "experiential/subjective readings occur chiefly in genres that provide opportunities for people ... to express themselves in a spoken or speech-like manner." In her response to Smitterberg, Fitzmaurice notes that each token identified must still be read by the analyst, and she argues that such detailed work provides an important analytical tool to the corpus linguistic methodology. Together, Fitzmaurice and Smitterberg extend in fruitful and dynamic ways current conversations about the historical development of expressive/subjective progressive constructions and about the usefulness of social networks as a framework for corpora research. Douglas Biber, employs corpus linguistic methodologies to address a different historical change within the English verb prhase. In "Modal use across register and time," Biber works with the ARCHER corpus (A Representative Corpus of Historical English Registers), an invaluable corpus Biber and others at Northern Arizona University designed "to analyze historical change in the range of written and speech-based registers of English from 1650 to the present." As such, ARCHER represents a broad spectrum

Section 2: Introduction 129 of authors and registers and presents scholars with a diverse collection of diachronic texts. In the present study, Biber considers three categories of modals: permission/possibility modals (can, could, may, might), obligation/logical necessity modals (must, should), and volition/prediction modals (will, would, shall). In addition, he examines three semi-modals (have to, got to, be going to). Adopting a "comparative register perspective," Biber concludes that, while few global generalizations are possible, "modals have been decreasing in frequency in all registers, but semi-modals have shown noteworthy increases only in drama and personal letters." In his analysis of specific examples from the corpus, Biber draws on his innovative recent work on "stance" and its linguistic expression and adds a critical historical dimension here. This study provides extensive raw material for future study, and Biber concludes by suggesting that his own findings must be considered "relative to the full range of linguistic devices used to express stance in English." Thus, like Fitzmaurice and Smitterberg, Biber offers scholars the opportunity to use data derived from corpora like ARCHER to complement and instigate additional studies. The final two essays in this section turn from broad corpora to single works, highlighting the necessity for careful scholarship and editorial practices. Richard W. Bailey, in "The need for good texts: The case of Henry Machyn's Day Book, 1550-1563," argues that scholarly attention must be directed toward "the reliability of the material that makes up the data for our history writing." Considering the case of Henry Machyn's Day Book, Bailey outlines an editorial cautionary tale for future scholars of linguistic history. Noting that eighteenth-century historian John Strype "did not regard it as useful to give a literal transcript," Bailey chronicles various scholarly (mis)treatments of Machyn's work from that time to the present. Announcing the present effort underway at the University of Michigan to produce an accurate scholarly edition of Machyn's work, Bailey describes how such projects are now possible through electronic editions that incorporate manuscript images, literal translations, and modernizations. Such editions, he asserts, will be of much greater benefit to future linguistic and historical scholars. Ian Lancashire, an expert on Early Modern English lexicography, provides a parallel instructional example of how close work with a specific historical manuscript can correct previous misdatings and open up intriguing questions about the development of English-only lexicons. In "The perils of firsts: Dating Rawlinson MS Poet. 108 and tracing the development of monolingual English lexicons," Lancashire offers an examination

130 Anne Curzan and Kimberly Emmons of the scholarly editions available of selected Early Modem English lexicographic works. Focusing specifically on the difficulties of analyzing the material found in the commonplace book written by MS Rawlinson Poet. 108, Lancashire identifies the problems associated with dating and assessing lexicographic works without clear generic boundaries and without full scholarly editions of relevant works. In a move of interest to all historians of standardization in English. Lancashire argues for an early sixteenthcentury date for the first English-only lexicon. Lancashire includes his own scholarly edition of "The Rawlinson Dictionary" as an appendix to his article. As he argues here, it is only through the opportunity to engage with the full text of a carefully transcribed manuscript that scholars can evaluate any argument about dating or other linguistic features of a text. The essays collected in this section engage both with current linguistic scholarship and with future methodological concerns, as they examine historical texts and central questions in the development of the English language. Innovative scholarly tools like corpora and electronic editions are providing a rich source of future textual evidence. Combining careful attention to textual nuances with broad, cross-register comparisons, the scholars in this section suggest that these methodologies can produce important insights into the historical and present uses of English.

The meanings and uses of the progressive construction in an early eighteenth-century English network Susan M.




In the language of seventeen men and women writing in late seventeenthand early eighteenth-century England, we encounter the following instances of the progressive construction: (1) a.





She replied simply that he would be obeyed, and went to her closet to give way to the tears which she had suppressed with difficulty while he was speaking. (Lady Mary Wortley Montagu, Princess Docile, 1756 [mmficOOl]) We are destroying many thousand Lives, exhausting all our Substance, not for our own interest, which would be but common Prudence; not for a Thing indifferent, which would be sufficient Folly, but perhaps to our own Destruction, which is perfect Madness. (Jonathan Swift, Conduct of the Allies, 1711 [asess005]) But now I am talking of our poor, and indeed we are filled with the clamours of our people about the poor-with the least reason, and the least project of rectification that ever was seen in any like case. (Daniel Defoe [ddess012.txt]) But the compassionate case of very many of them is, that they are taken into such hands, without any the least Suspicion, previous Temptation, or Admonition to what place they are going. (Richard Steele, The Spectator 266, Friday, 4 January 1712 [rsess013.txt]) But I am not in the least pain upon that matter, because it is very well-known that they are every day dying and rotting by cold and famine, and filth and vermin, as fast as can be reasonably expected. (Jonathan Swift, A Modest Proposal, 1729 [asess001.txt])

132 Susan Μ. Fitzmaurice

These quotations exemplify some ordinary uses of the progressive construction in English. For example, the novelist Lady Mary Wortley Montagu deploys the past tense progressive verb phrase he was speaking in (la) with a subordinating conjunction, while, to indicate what was going on - the nature of the context - when the character was forced to suppress her tears. In (lb), the plural first person in combination with the present progressive verb are destroying allows Swift to present a view of a situation in progress - the continuing forfeit of English soldiers' lives in order to support a war being waged by an ally against Spain - the endpoint of which is considerable harm to English forces without advantage. Daniel Defoe uses the present progressive construction in (lc) to present a state of affairs in which the moment of utterance is coterminous with the act of talking about the poor. In other words, the progressive is used to present the simultaneity of a single occurrence of an activity (talking about the poor) and the moment of utterance (now). In (Id) Steele uses the present progressive with the activity verb go in a relative clause. The function of indicating the progress of an ongoing activity is one of the most common functions of the construction. In (le), the present progressive combines with the temporal adverb every day in a ίΑαί-complement clause to express the iterative, continual nature of the situation. Of course this is Swift talking about the Irish in his satirical essay, A Modest Proposal. Each of these instances straightforwardly illustrates normal, familiar functions of the English progressive construction to indicate the durative, iterative, temporary, and progressive nature of a situation (Huddleston and Pullum 2002). The uses in (2) are somewhat different. (2) a. b.



She is always seeing Apparitions, and hearing Death-Watches (1711. Spectator 7: 1. 34) In the mean time, betwixt my intervalls of physique and other remedies which I am useing for my gravell, I am still drudging on: always a Poet, and never a good one. (John Dryden to Mrs. Stewart, Candlemass Day, 1698/9 [jdlet047]) But I am gott into a Maze & am running in a Circle, wch is uneasy to me for ye present, & a very disagreable prospect for ye future. (George Stepney to Charles Montagu, November 12/22, 1698, Berlin [gslet002]) I am always troubling you, but shall endeavour, on any occasion you will give me, to approve my self, (Steele to James Anderson, 29 July 1718 [rslet071])

Meanings and uses of the progressive construction 133 e.

You formerly observed to me, that nothing made a more ridiculous Figure in a Man's Life, than the Disparity we often find in him Sick and Well. Thus one of an unfortunate Constitution is perpetually exhibiting a miserable Example of the Weakness of his Mind, or of his Body, in their Turns. (Pope, On Sickness, The Guardian #132, 1714)

In each of the utterances in (2), in contrast with those in (1), the progressive construction appears to be a grammatical choice driven by the attitude of the speaker to what he is talking about. The choice of the progressive in each case has the effect of expressing the writer's evaluation of or attitude towards what is being talked about, regardless of the situation. Hence in (2a), Addison invites the inference that he considers that the subject has an over-active imagination. By claiming that the subject's encounters with ghostly phenomena amount to a persistent habit, he makes a stronger claim than might be warranted, thereby allowing the reader to infer that he regards her as crazy. Specifically, the progressive construction "activates" the normally stative verbs see and hear, enabling them to be used to refer to a series of punctual events. To compound this, the temporal adverb always provides a resource for hyperbole, allowing the speaker to invite the reader to infer that he overstates the case to ensure that the reader gets the point. The root of the inference is the attitude of the speaker towards the subject, and his choice of the expression represents his choice to exaggerate in order to convey his point of view. In example (2b), John Dryden chooses a prepositional activity verb drudge on together with the durative adverb still to convey vividly the sense of the never-ending laboriousness with which he attends to the work of writing poetry which he constructs figuratively as drudgery. In example (2c) George Stepney deploys two metaphors to convey his frustration at not being permitted to leave Berlin for London. The pair of metaphors is constructed using the beperfect with the non-volitional verb get, and the progressive with the activity verb run to invite the reader to infer that he has no control of his own destiny. Of course, the resulting figurative force of the progressive chimes with the lexical prepositional phrase, in a circle, which arguably conveys something of the same sense of never-ending, pointless momentum that an adverb like always or perpetually would supply to such a construction. In (2d), Richard Steele ruefully acknowledges the weight of the imposition that he presents to his addressee by adopting the stative verb, trouble, with the pragmatic strengthened always. This is an interesting pragmatic strategy - by admitting the scale of the bother that he

134 Susan Μ. Fitzmaurice presents, he effectively takes that assertion away from the very person affected, thus disarming the addressee. He also manages to pay attention to his own face wants; by claiming responsibility for the imposition, he prevents other, harder, and unsympathetic responses. Finally, in (2e), Pope presents a sorry, rather savage portrait of the invalid. His choice of the progressive changes the stative verb exhibit into an accomplishment verb, and his use of the temporal adverb perpetually exacerbates this rhetorical picture of the permanently ailing person (which he himself was). In each of these cases, the speaker invites the addressee to attach to his utterance the sense that he means and thinks more than he is saying. Addison hints that the behaviour of the woman is extraordinary, Dryden suggests that his labour is endless and demeaning, Stepney fears that he has no control over his future, Steele admits that he is a continual source of bother, and Pope ridicules the figure of the invalid. What they share is a sense of the speaker's attitude or experiential perspective towards their topic - none of these examples exhibits neutral stance. Lexical features verb choice and adverb choice - may distinguish the subjective effects that the progressive conveys on any given occasion (such as those in (2)) from the standard functions of the progressive in (1). In particular, figurative or derived meanings of verbs may be recruited for non-aspectual, modal force with the progressive. For example, activity verbs that cannot be construed as referring to physical activity but may be construed figuratively are ideal recruits for modal meaning in the progressive. Thus, selected lexical collocates supply the required intensity to trigger the inference that the speaker means more, or indeed, other than what is conveyed by the standard construction. In this study, I assess the extent to which the markedly expressive meanings illustrated in (2) inhabit the uses and functions of the progressive construction in eighteenth-century British English. I draw upon the methods of corpus linguistics in order to present a general account of the meanings and uses of the progressive construction in a range of written prose registers of late seventeenth- and early eighteenth-century English.1 I adopt the corpus linguistics context of register analysis in order to situate the particular study of meaning change, thus drawing upon recent work in this area (Ryden 1997; Smitterberg 2000). I also draw upon pragmatic analysis in order to provide close readings of contexts in which the construction occurs in different registers in the Network of Eighteenthcentury English Texts Corpus (NEET) (see Section 2 below). In previous work (Wright 1994, 1995), I have argued that the progressive construction

Meanings and uses of the progressive construction


participates in the realization of speaker self-expression in grammar, what I have called experiential syntax. The thrust of those arguments was that the everyday subjective/expressive potential of the construction may be appropriated for use in literary discourse. In this paper, I return to basics in order to test this claim empirically. Thus I examine the contexts and frequencies with which progressive constructions such as those in (1) occur as crucial context for the study of the experiential progressive as illustrated in (2). First I explore a set of genres or registers, including letters, fiction, essays, and drama, produced by a network of writers in eighteenth-century London as well by a few people who are unconnected with the network. By "network" I mean a group of people connected by personal interests and affinities, perhaps also by common professional interests, over a period of time. For extended treatments of the nature of the ties forged among the people whose writing is represented in the Network of Eighteenth-century English Texts Corpus (NEET), see Fitzmaurice (2000a, 2000b, 2003a, 2003b). The writing represented in the NEET Corpus, including letters, essays, memoirs, fiction, prose comedy, and dialogues, was produced between 1670 and 1760, thus representing English writing from the Restoration period to the second half of the eighteenth century. I am interested in the extent to which genre and register differences might condition both the use of the progressive construction and its meanings in different prose styles in the period. I attend to syntactic and collocational contexts (including clause context, tense choice, verb class, and adverbial collocation) for the progressive as well as textual contexts, thus yielding a complex picture of the range and frequency of the construction's uses and functions. I then turn to a brief analysis of frequency and distribution of the construction in the context of the social network in which the writing that provides the dataset is situated. Finally, I turn to a closer examination of the distribution of the construction in the lexico-grammatical contexts that I have argued are most likely to trigger a subjective or experiential function. This analysis indicates that these functions, while not common in the corpus as a whole, appear in registers that seem to exhibit a greater frequency of the construction overall. Additionally, these registers are arguably less literary than private and they are less literate than speech-based. The outcome of the study is a detailed account of the relation of meaning and use of the progressive construction in a set of literary and nonliterary registers produced over a period of about 100 years.

136 Susan Μ. Fitzmaurice

2. The Network of Eighteenth-century English Texts Table 1 provides a numerical and historical profile of the data used in this analysis. The Network of Eighteenth-century English Texts (NEET) corpus consists of four prose registers: letters, essays, fiction and drama produced by a group of seventeen individuals writing between 1653 and 1762.2 The data sample is therefore not representative of the registers of eighteenthcentury English writing or indeed of the demographic range of British English writers of the period. Instead, the data sample is a product of the process of selecting a group of individuals who are connected with one another in various ways in a period that extends from 1695 to 1745. Table 1. The NEET corpus Dates of birth & death 1631-1700 1640-1688

Letters: Word #

Essays: Word #

23,068 N/A

57,299 N/A

Daniel Defoe Sarah Churchill

1660-1731 1660-1744

45,144 50,728

George Stepney Matthew Prior Mary Astell

1663-1707 1664-1721

19,690 20,848

Jonathan Swift Susannah Wesley Delariviere Manley William Congreve Edward Wortley

1666-1731 1667-1745 1669-1742 1671-1724 1672-1729 1672-1761

37,445 48,143 40,767 N/A 26,382 25,396

Joseph Addison Richard Steele

1672-1719 1672-1729

Alexander Pope Mary Montagu Eliza Haywood Range:

Writer John Dryden Aphra Behn

Fiction: Word # N/A

Drama: Word # 28,266



42,458 16,213

121,559 N/A





43,043 41,945 N/A 20,479 N/A

N/A N/A N/A 41,562 21,235 N/A

N/A N/A N/A N/A 28,604

50,791 40,951





1688-1744 1689-1762 1693-1756

41,919 40,767 N/A

41,284 24,171 N/A

N/A 23,064 13,698

N/A 14,107 19,426






14,520 40,407

N/A N/A 18,094

At the center of the group is the essayist Joseph Addison (born 1672) and his changing social network from 1700 until his death in 1719. The relationships that make up Addison's social network include professional collaborations, friendships, patron-client connections, acquaintanceships

Meanings and uses of the progressive construction


and enmities. For example, Richard Steele and Edward Wortley were lifelong friends of Addison's; Steele also collaborated with Addison on the Tatler and Spectator periodical projects. The two friends solicited essay contributions to their periodicals between 1709 and 1715 from the likes of Jonathan Swift, and the younger writers, Alexander Pope and Lady Mary Wortley Montagu, Wortley's wife. Addison regarded the oldest figure in the network, the poet laureate John Dryden, as a mentor, model and literary patron. He wrote a poem in praise of Dryden's translation of Ovid as early as 1693, and then spent time with the poet laureate in London later. Dryden published Addison's early poetry, as well as the poetry of Addison's friends and colleagues, George Stepney and Matthew Prior, and he was one of the young Congreve's most enthusiastic supporters. These men were fellow Whig writers and clients of the Whig grandee, Charles Montagu, Earl of Halifax (Fitzmaurice 2002). In addition, the group whose writing is collected in the corpus includes people whose connections with Addison are no more than indirect. These are Delariviere Manley, a correspondent of Steele's, and Mary Astell, a friend and confidante of Lady Mary Wortley Montagu's. Sarah Churchill, the Duchess of Marlborough, was well-known to all of the people in Addison's network, though she did not necessarily know them. The playwright and poet, Aphra Behn, the political journalist and novelist, Daniel Defoe, and mother of John Wesley, Susanna Wesley, are figures with no known connection with Addison. I included them in order to gain a perspective of the broader social and linguistic context that situates Addison's cohort.3 Because the corpus design is grounded in and informed by social networks analysis and the specific social-historical ties that characterize the group, the texts collected in the corpus represent only the products of the writing practices of a group of peer figures (Fitzmaurice 2002). The corpus was designed to facilitate the investigation of the role of author identity as well as register in linguistic variation in eighteenth-century English. To this end, it was important to include the kinds of writing produced by the figures in the group regardless of the fact that not all of the writers produce the same range or kinds of writing. Accordingly, fourteen of the seventeen figures are letter-writers, twelve are essaywriters, six are fiction-writers, and six (not the same six) are dramatists. Table 1 is organized chronologically, according to people's birthdates. There are three notional generations of people whose work is collected in the corpus. John Dryden, the earliest of the writers whose work is collected, and his contemporary, Aphra Behn, make up the first generation.

138 Susan Μ. Fitzmaurice The second generation consists of twelve of the seventeen people, all born between 1660 and 1672. This generation includes the people with whom the central figure in NEET, Joseph Addison, contracts the strongest and most long-lived ties. Born in 1672, Addison became close to his exact peers, Edward Wortley, Richard Steele, and William Congreve, as a college student or as a young man, and held those friendships for life. The youngest writers whose work is collected in NEET are Alexander Pope, Lady Mary Wortley Montagu, and Eliza Haywood, all born after 1687. Table 2 summarizes the periods occupied by each generation; it gives the life span of the generation first, and then gives the dates of the earliest and latest attested writings produced by each generation. The generational periods are thus 1630-1700: John Dryden's dates of birth and death; 16601745: the birth dates of Daniel Defoe and Sarah Churchill and the date of Swift's death; and 1688-1762: Alexander Pope's date of birth, and the date of Mary Wortley Montagu's death. The total period spanned by the writing in the corpus is over one hundred years - from 1653, represented by a letter from John Dryden to his wife, Honor, through to 1762, represented by a letter from Lady Mary Wortley Montagu to Lady Frances Steuart, written just before Lady Mary's death. The earliest drama text (out of a total of six) is John Dryden's Marriage A la Mode, produced in 1673, and the latest is Lady Mary Wortley Montagu's Simplicity (1734). The earliest piece of fiction in the corpus is Congreve's (1695) novella, Incognita, and the latest is Lady Mary Wortley Montagu's Princess Docile (1756). The essays range from Dryden's Essay on Dramatic Poesie (1668) to Sarah Churchill's (1745) political memoir. Appendix A presents the periods spanned by each individual's writings. Table 2. Writing periods Generation Generation 1 Generation 2 Generation 3

Dates of birth and death 1630-1700 1660-1745 1688-1762

Period of writing 1653-1700 1693-1731 1710-1762

William Congreve and Lady Mary Wortley Montagu alone among the seventeen writers are exponents of all four registers. Congreve (16701729) is known principally as a dramatist, but he was also a member of the Kit Cat Club, and a close friend and exact contemporary of fellow Whigs, Joseph Addison and Richard Steele. He regarded John Dryden as a mentor and he first met Lady Mary Wortley Montagu (1689-1762) as a little girl

Meanings and uses of the progressive construction


when her father, Lord Kingston, introduced her to his fellow Kit-Cats as the Toast of the Club. As the second youngest member of the set, and the last of the group to die, Lady Mary Wortley Montagu represents the language of the mid eighteenth century. In contrast, Congreve's repertoire is emblematic of the language of the early part of the century (see Fitzmaurice 2000a for an assessment of Congreve's language status among his peers). Two of the most senior members of the set, John Dryden and Daniel Defoe, produce three registers each - Dryden produces letters, essays and plays, while Defoe produces letters, essays and fiction. These writers represent two very different social strata. Although neither man seems to represent the mainstream - Dryden was a Catholic or Catholic sympathizer and Defoe was a nonconformist Tory sympathizer - they occupy opposite ends of the social milieu. Dryden had been the darling of James II and the premier poet of the English court before the Glorious Revolution of 1688. In contrast, Defoe made a living as a journalist and spy for Robert Harley, but never attracted the attention or the patronage of the grandees who sponsored literary projects and political careers of men like Addison or Steele. A few comments about the contents of the registers apart from the letters are in order. In general, there are fewer texts in the fiction, drama and essay collections for each writer than in the letters collections. One obvious reason for this is that these people are less prolific in writing novels or plays than they are in writing letters, both professional and private. Thus the fiction collected for individual writers may consist of a single piece rather than a set of pieces or, indeed, a single piece selected on a principled basis. For example, Congreve's fiction is represented by his (1695) novella, Incognita, the only piece of fiction he produced. Another obvious reason is that novels, unlike familiar letters, are sizeable documents, and therefore represent a substantial number of words in the register. Lady Mary Wortley Montagu's fiction is represented thus far in this corpus by Princess Docile, an autobiographical romance she wrote in retirement in Italy around 1756 (Grundy 1996). In contrast, although Steele wrote a number of comedies, his drama is represented in NEET only by the (1705) comedy, The Tender Husband. Lady Mary Wortley Montagu's drama is represented by her sole surviving attempt at playwriting, the (1734) comedy, Simplicity. It is striking how many of the writers are essayists - twelve out of the seventeen. Many of the eighteenth-century essays collected in NEET would most likely be regarded as journalism today. Defoe, Prior, Swift and

140 Susan Μ. Fitzmaurice Steele were political essayists, often producing propaganda and satire on behalf of the Whig or Tory parties. Defoe was editor of The Review (17111713), a political journal published at the same time as Addison and Steele produced The Spectator (1710-1712), followed by The Guardian (17131714), and The Freeholder (1715-1716). Lady Mary Wortley Montagu and Alexander Pope were occasional contributors to these journals. 4 The essays written by Congreve and Dryden are different in focus and audience from those of the polemicists. Their essays are literary criticism, often written in response to reviews of their own work. The women writers in NEET are different again in terms of their essays' audience and concerns. Susanna Wesley, mother of John and Charles Wesley, founders of the Methodist movement, wrote instructional epistles for her children, and conducted written theological debates with her sons. Mary Astell, a feminist and polemicist, wrote essays on women and education on the one hand, and conservative political essays on the other. These educated women writers, both reformers, differ markedly from Sarah Churchill, Duchess of Marlborough. Churchill's prose consists of political memoir - her own commentary on her experiences as a key political player in the court of Queen Anne by virtue of her role as the queen's confidante.

3. Meanings of the Early Modern English progressive There are three main treatments that provide the immediate research context for the corpus-linguistic survey of the historical construction's meanings and uses in seventeenth- and eighteenth-century English. Mats Ryden (1997) argues that the progressive controls core and non-core meanings respectively. He argues that the progressive's functional base of dynamic process is realized in two ways: the first is action-focused and aspectual; the second is attitude-focused, and concentrates on subjective, affective, and interpretative meanings. Rather than focus on the dynamic senses of the construction, Ziegeler (1999) attends to the function of the progressive in capturing duration. She argues (1999: 54) that the agentivity of the modern progressive is a "consequence of the diachronic generalization of the participle over an increasing range of dynamic verb classes, and as an inference from the developing aspectual senses of durativity." Further, she allows the inference of a subjective function of the progressive, for instance, in the construction's "capacity to 'activate' a verb which is normally defined by the inherent lexical semantics of stativity" (Ziegeler 1999: 55). This capacity is what triggers what I labeled as an experiential

Meanings and uses of the progressive construction


meaning. However, I have argued that the experiential meaning of the progressive is "non-aspectual; it is not always or directly apprehendable in truth-conditional terms, and is not even temporally salient. It focuses ... on the (speaking) subject's consciousness of being inside an event, state, activity, looking out" (Wright 1995: 156). Thus for many speakers, "the choice of a progressive draws attention to the stance or positioning of the emotional experience of the speaker with respect to what is being talked about" (Wright 1995: 156). Corpus linguistic surveys have also been conducted on the use of the progressive in historical corpora. Strang (1982) and Arnaud (1982, 1998) are studies that used databases of literary and epistolary texts respectively in diachronic surveys of the changing frequency and distribution of the construction. They were interested in tracing the use of the progressive over time as a framing device in narrative and comparing this with its use as a foregrounding device in nineteenth-century narrative prose. More recently, Erik Smitterberg (2000) surveyed the occurrence and distribution of the progressive in CONCE, the Uppsala-based Corpus of NineteenthCentury English. He found that for the three sub-periods covered by the corpus, the progressive occurred most frequently in letters, and in fact increased in this register over the three sub-periods. The distribution shifted across sub-periods for two other registers; it increased over time in drama, but after showing an increase in period 2 in fiction, decreased in period 3. And Nicholas Smith has recently conducted a survey of the use of the progressive in recent British English using the London-Oslo-Bergen (LOB) and Freiburg-London-Oslo-Bergen (FLOB) corpora. He observes that the progressive has spread from collocating with activity verbs to construction with mental and communication verbs, and that the principal contexts in which the construction has increased its frequency are main clause, present tense contexts. He speculates that these developments might have been influenced by the "spread of colloquial speech habits into written language. ... and an increase in subjective/interpretative uses" (2002: 330). The present study reaches back into the late seventeenth and eighteenth centuries to ascertain patterns of occurrence and meaning that might be understood as a basis for the trends noted by Smitterberg and Smith. A major challenge is to ascertain the contexts in which the experiential or subjective function of the progressive may be instantiated, compared with those in which the aspectual functions of the construction predominate. The appeal to pragmatic competence or intuition in ascribing

142 Susan Μ. Fitzmaurice to different occurrences of the progressive an experiential or subjective function is hard to account for at best and it is unreliable at worst. Consequently, I situate the search for the progressive construction in the registers represented in NEET in the following set of lexico-grammatical contexts (following Wright 1994, 1995): syntactic environment (main vs. subordinate clause); tense (present vs. non-present tense); identity of subject (first, second, third); verb type (e.g., private vs. activity); lexical support (e.g., adverbial modification). In line with earlier work, I hypothesize that first person main clause present tense progressives are more likely to convey experiential meanings than third person subordinate clause past tense progressives, and that the former context with a stätive verb such as understand or love is a good candidate for construal as experiential. However, in this study I draw upon the additional criterion of verb choice as a means of diagnosing a modal or experiential progressive. In other words, when a progressive construction is not aspectual or truthconditional, it is because the verb selected is figurative, and its colligation in the progressive changes its semantic category. It is reasonable to expect, however, that the majority of instances of the construction encountered in this study will convey aspectual meanings that have to do with the internal structure of an event or activity.

4. Analytical procedures and limitations Several search procedures were carried out in order to gather the data for analysis. Using MonoConc, a commercial concordancing package, I conducted searches of the progressive construction in the lexico-grammatical contexts listed above, namely, syntactic environment (main vs. subordinate clause); tense (present vs. non-present tense); identity of subject (first, second, third person); verb type (e.g. mental vs. activity); lexical support (e.g. adverbial modification). The search yielded KWICs (Key words in context), the basic data for coding in the different lexico-grammatical contexts and subsequent frequency analysis. The construction identified as the progressive consists of auxiliary verb be plus a present participial verb form (V + ing). However, not all instances of the sequence be + ing were counted as progressive verb forms. I counted progressive constructions that occur with the perfect and modal verbs. I did not include be + ing constructions with ergative functions, illustrated by the expressions in (3a) and (3b). In these examples, the verb phrase conveys a force that is arguably more passive than progressive.

Meanings and uses of the progressive construction 143 (3) a.


Nay more, when the event is past dispute, even then we are willing to be deceived, and the Poet, if he contrives it with appearance of truth, has all the audience of his Party; at least during the time his Play is acting: (John Dryden, 1668, An Essay of Dramatick Poesie. [jdessOOl]) Lady Wishfort: 0 Daughter, Daughter, 'tis plain thou hast inherited thy Mother's prudence. Mrs. Fainall: Thank Mr. Mirabell, a Cautious Friend, to whose advice all is owing. (Congreve, Way of the World, 1700. [ccdraOOl])

In (3a), the verb act has an inanimate subject with the semantic role of patient rather than agent, giving rise to a passival meaning (Denison 1993). In this example, the be + ing construction has aspectual force, but only because of the adverbial, during. Similarly, the verb owe in (3b) is governed by an inanimate subject, namely, Mirabell's advice, and so its thrust is passival. Passival expressions such as that in (3a) disappeared in the course of the nineteenth century with the rise of passive progressive phrases such as his play is being acted (Pratt and Denison 2000). In addition, there are occasionally constructions that seem equivocal, such as be wanting. Dryden's reflexive use of the adjectival expression, to be wanting, is less usual than the example in (4b), in which Defoe adopts be wanting as with an inanimate subject, nothing. (4) a.


But let the world witness for me, that I have been often wanting to myself in that particular. (John Dryden, A Discourse Concerning the Origin and Progress of Satire, Aug. 18, 1692. [jdess002]) But you may assure yourself (my Lord) nothing shall be wanting to represent either yourself or your affaires to your Lordships greatest advantage, and I hint by the way that no man is Fitter to move in such a case than the Duke of Newcastle, whom your Lordship mentiond. (Defoe to David Erskine, Earl of Buchan. 29 May 1711. [DefoeLET])

Example (5) contains a be + ing construction that is not a progressive at all. (5) Quaer. Whether Mr. Collier, in the pit, will not think it had been more becoming his Character to have invited and exhorted them to it? (Congreve, Amendments of Mr. Collier's False & Imperfect Citations, etc. 1695. [ccess002])

144 Susan Μ. Fitzmaurice

In Present day English, Congreve's be + ing construction might be better construed as 'will not think that it might have been more becoming to his character.' The be + ing construction, in addition to colligating in a pluperfect modal-less subjunctive, is more properly identified as a nonfmite clause governed by copula be. Somewhat clearer examples of the very common use of the non-finite complement clause appear in (6a) and much more clearly, with a definite article identifying the non-finite clause as a noun phrase, in (6b). (6) a.


The first Circumstance, as he ingenuously confessed to me (while we were in the Coach together) which helped to disabuse him, was [seeing King Charles I on Horseback at CharingCrosslnp; (Addison, The Freeholder, No. XLVII, Friday, June 1, 1716. [aess021]) So that whether this War were prudently begun or not, it is plain, that the true Spring or Motive of it, was [the aggrandizing a particular Family]NP, and in short, a War of the General and the Ministry, and not of the Prince or People; since those very Persons were against it when they knew the Power, and consequently the Profit, would be in other Hands. (Jonathan Swift, The Conduct of the Allies, &c.~ 1711. [asess005])

Finally, I also omitted from consideration, the expression, 'be going to' where it is used as a semi-modal with future time reference, as in Lady Mary Wortley Montagu's example below (7). (7) I am not now going to declaim in Favor of Slavery, but I would fix a signification to the words free people that has not yet entered into the wise heads of any party (Lady Mary Wortley Montagu, Nonsense of Common-Sense, No. 6, January 24, 1738 [mmess006]) There are also several limitations as regards the sampling of texts for analysis. Because the corpus can collect examples only from as many registers or genres that make up each writer's repertoire, and because some repertoires are represented by a range of genres whereas others are indicated by just one or two genres, the corpus is obviously variable in terms of text numbers and size. For this reason, I have not conducted extended comparative frequency analysis or statistical tests for significance on the registers that are represented by very few texts, for example, fiction and drama.

Meanings and uses of the progressive



5. The progressive construction in the NEET corpus 5.1. Frequency and distribution across registers A rapid survey of the search results indicates that the progressive construction is readily available to speakers in a number of contexts and registers. Figure 1 reports the normed total frequencies of the progressives across register regardless of lexico-grammatical context. The raw total number of progressives is 980 occurrences in 1,353,837 words. This figure can be reported as a normed total of 7.24 occurrences per 10,000 words or 724 occurrences per million.






Figure 1. Distribution of progressive constructions across register

The progressive occurs most frequently in prose drama, more than 11 times per 10,000 words, followed by fiction, at about 9 times per 10,000 words, and only then by letters, at about 7 per 10,000 words, and essays, at 6 occurrences in 10,000 words. This finding differs from Smitterberg's finding that nineteenth-century letters exhibit a greater frequency of the progres-

146 Susan Μ. Fitzmaurice sive than other genres. The contrast between our findings should perhaps not be very surprising if examined in the light of Biber and Finegan's (1989, 1997) demonstrations that the genres with which historical corpus linguists work are themselves historically volatile. For instance, in their survey of the linguistic shape of five genres over time, they observed that the familiar letter in the eighteenth century is a more "literate" register than it appears to be in either the seventeenth or the nineteenth centuries. As earlier treatments of the progressive have indicated, the progressive construction seems to appear more frequently in registers that have situational characteristics that mark them as more oral than literate. In our period, the prose comedy is possibly the register that exhibits features that place it closest to the oral end of the oral-literate pole whereas the essay is arguably the furthest away from the oral end. 5.1.1.

Grammatical Context

The following analysis presents the breakdown of these figures by context and by register. Figure 2 represents the distribution of progressives according to two key syntactic contexts for analysis - main and subordinate present and past tense clauses in different registers in the corpus. The distribution of the construction across registers is more complicated than Figure 1 suggests. In particular, the distribution of the progressive in main and subordinate clauses differs according to whether it occurs in the present or past tense. For example, although drama seems to exhibit the highest frequency of the construction overall, present tense main clause use appears to account for only part of this frequency, at the rate of 4 occurrences per 10,000 words. Letters and essays share a tendency to prefer present tense to past tense use, in both main and subordinate clause contexts, a tendency that is made clear by a quick inspection of the raw figures in Figure 2. However, in the light of low frequencies overall, this preference should be regarded as notional at best. In contrast, past tense use in both main and subordinate clauses claims the lion's share of the progressive's occurrence in fiction. Indeed, the most common grammatical context for the progressive in this whole period appears to be the past tense subordinate clause - a grammatical context that is least likely to host a non-aspectual, figurative progressive. Table 3 presents the raw numbers with the percentages they represent that inform the normed frequency summary in Figure 2.

Meanings and uses of the progressive construction


Subordinate past

Subordinate pres

Β Essays D Letters E5 Fiction Β Drama

Main past

Main pres




Frequencies per 10,000 words

Figure 2. Progressives by clause type and tense Table 3. Distribution of progressives by tense and clause type: Raw numbers and percentages

Essays Letters Fiction Drama

Main clause present 65 (29%) 127 (36%) 20 (8%) 76 (53%)

Main clause past 32 (14%) 24 (7%) 76 (30%) 30 (21%)

Subordinate clause present 78 (35%) 143 (40%) 17 (6%) 21 (15%)

Subordinate clause past 49 (22%) 62 (17%) 143 (56%) 17 (11%)

Total Ν 224 (100%) 356 (100%) 256 (100%) 144 (100%)

If we consider the relation of subject identity to main or subordinate clause occurrence, the picture's texture alters somewhat. I examined the distribution of first person and third person subjects in main and subordinate clause contexts respectively (Figures 3a, 3b). Note that the frequencies counted are very low - ranging between virtually zero and just 3 oc-

148 Susan Μ. Fitzmaurice currences per 10,000 words. Tables 4a and 4b provide the raw numbers that inform the graphic representation of the normed frequencies in Figures 3a and 3b.

Frequency per 10,000 words

Figure 3a. First person singular (I) subject + progressive across register and clause type Table 4a. First person singular (I) subject + progressive across register and clause type (raw numbers and percentages) Clause and tense main clause present main clause past subordinate clause present subordinate clause past Total Ν

Drama 19 (47.5%) 15 (37.5%) 3 (7.5%) 3 (12.5%) 40

Fiction 9 (13%) 21 (31%) 5 (8%) 32 (48%) 67

Letters 68 (45%) 12 (8%) 59 (39%) 11 (8%) 150

Essays 18 (30%) 0 (0%) 25 (41%) 9 (29%) 61

Meanings and uses of the progressive construction


Frequency per 10,000 words

Figure 3b. Third person singular subject + progressive across register and clause type

Table 4b. Third person singular subject + progressive across register and clause type (raw numbers and percentages) Clause and tense main clause present main clause past subordinate clause present subordinate clause past Total Ν

Drama 24 (59%) 8 (20%) 6 (15%) 3 (6%) 41

Fiction 3 (3%) 41 (34%) 4 (3%) 73 (60%) 121

Letters 25 (27%) 8 (9%) 32 (34%) 28 (30%) 93

Essays 25 (34%) 10 (13%) 17 (23%) 22 (30%) 74

Perhaps surprisingly, first person singular subjects occur almost as often as third person singular subjects with the progressive, 9.67 times per 10,000 words compared with 10.1 times per 10,000 words. Perhaps unsurprisingly, the distribution of these progressives is uneven across registers. For instance, drama, letters and fiction favor first person progressives at a comparable level of frequency (2.78, 2.92, and 2.39 per 10,000 words respectively). (For raw numbers and percentages, please see Tables 4a and 4b.)

150 Susan Μ. Fitzmaurice

Preference for the construction varies according to clause type and tense. So although the first person progressive occurs more frequently in main and subordinate present tense clauses than in past tense clauses in the letters corpus (84% of the first person progressives in letters), the construction occurs much more frequently in past tense main and subordinate clauses than in any other context in fiction (79% of the first person progressives in fiction). In contrast, drama appears to select first person progressives in present and past tense main clauses in preference to subordinate clauses (85% of the time), regardless of tense. Main clause contexts are also salient for third person subjects with the construction in drama, for third person progressives occur more frequently in present tense main clauses than in any other context in this register (59%). Indeed, proportionally, there are more third person present tense main clause progressives in drama than in any other register. The register that selects third person over first person as preferred subject for the progressive is fiction. Third person past tense progressives predominate in main clauses (at the rate of 1.46 occurrences per 10,000 words, or 34%), and more notably in subordinate clauses (7.6 occurrences per 10,000 words, or 60%). This picture of grammatical contexts and subject preference for the progressive in the eighteenth century is congruent with the picture drawn from an earlier study of progressives in drama (Wright 1994). However, the picture presented here is one that has more detail and texture, as we examine lexical contexts in addition to the progressive's grammatical frame.


Lexical context

By 1700, speakers choose the progressive construction with a range of verb types. The most frequent type of verb is activity verbs, which occur with the progressive between 49% in essays and 60% in fiction, confirming Ryden's and Ziegeler's observations regarding the dynamic focus of the progressive. The occurrence of mental verbs - including private and cognitive verbs - ranges between 9% in letters and 18% in drama. Communication verbs - verbs of speaking and writing - account for between 13% in fiction, and 23% in essays. See Table 5 and Figure 4.

Meanings and uses of the progressive construction 151 Table 5. Verbs in collocation with the progressive across registers

Register Letters Fiction Essays Drama


































6 (3%)




















What is interesting is the range of different verbs deployed with the progressive. Thus even though activity verbs account for the majority of the progressive constructions, there are 148 different verb types that select the construction. Progressive constructions with the most common activity verbs, go, come, do, and make, amount to 24% of the progressives in letters, 15.7% of the progressives in fiction, 7% of the progressives in essays, and 19.4% of the progressives in drama. However, among the activity verbs are much less common verbs such as assault (fiction), demolish (letters), march and meddle (essays), and make love (drama). Their deployment with the progressive is exemplified in (8a-c): (8) a.



It may be here considered that our Mind is Such a buisy thing that it will never Stand Neutre, but is medling and Interesting it self upon all Occasions. (Matthew Prior, Heads for a Treatise upon Learning. [mpessOOl]) What-a-Pox does he put it off for? Does he think our Horse is not marching up at the same Time? But let us see what he says further. (Richard Steele, Tatler - No. 178, Saturday, 27 May to Tuesday, 30 May 1710. [rsess007]) CELEMENA. Whatever he is, I prefer him to a Fop-Mr. Sneaksby, you are not apt to be jealous, I hope-Mr. Toywell is making Love to me-How do you approve of it? SNEAK. Making Love to you-ugh! (Eliza Haywood, A Wife to be Let, 1723. [hdraOOl])

In (8a) the progressive occurs in a conjoined verb phrase with meddle and interest, two quite different verb types. Of course, though meddle is clearly

152 Susan Μ. Fitzmaurice an activity verb, its conjunction with a mental verb seems to invite an interpretation that allows the inference of a particular attitude on the part of the (mental) subject, the brain. The progressive in (8b) occurs in a subordinate clause governed by an interrogative mental verb phrase. The expression is rhetorically more appropriate to conversation or to drama than to the essay, though Steele's polemical perspective turns the Tatler essay into an interesting kind of political speech. The last example, (8c), is from Eliza Haywood's comedy, A Wife to be Let (1723), and it illustrates one of the typical uses to which the progressive is put in this genre, namely, main clause, present tense use. In Steele's and Haywood's hands, the choice of verb with the construction conveys a vigorous, and humorous sense of language use respectively.

ES Occurrence Β Mental • Facilitative • Existence 0 Communicative a Activity






Figure 4. Verb classes across registers Common communication verbs like write, speak, talk and tell amount to 12.5% of the verbs in the essays, and half that proportion, between 5-6% of all the verbs in letters, fiction and drama. Writers deploy the progressive with these verbs in letters to make observations about the nature of their interactions, to reflect upon the relationship being conducted in the epistolary medium, and to encourage their interlocutors to reply in similar spirit. The examples taken from the letters of Addison, Pope, and Congreve illustrate this attention to the conduct of epistolary conversation:

Meanings and uses of the progressive construction 153 (9) a.

If you have not yet written to [the Lord Lieutenant] I think you shoud and if at any time you woud recommend any person or affaire to him I am sure there is no person in Ireland whom he would be more desirous of obliging. You will Excuse my talking to you in this manner but I flatter my-self that I am talking with an Intimate friend which is a title I shall be always ambitious of on my side if you will give me the honour of it. (Addison to William Conolly, October 26, 1714 [aletl44])

Addison first offers Conolly some friendly, if presumptuous advice regarding the manner in which he might approach Addison's own boss, Lord Wharton, then Lord Lieutenant of Ireland, for favors. He then seems to register the possibility that his gesture might be interpreted as presumptuous and so begins to apologize to his addressee. But before he does so completely, he offers as his excuse, his own sense of the intimacy of the relationship between the correspondents. The present tense progressive verb phrase in the ίΑαί-complement clause follows a gerundive with the same communication verb, my talking to you, and works to underline his consciousness of the communicative act in which he is engaging as he reflects on his attitude towards the addressee. In (9b), Pope adopts the same expression as he interrupts his intimate confession of his desire to see his addressee's innermost soul to speculate that the object of his desire might not be acknowledged to exist, just in case his addressee has converted to Mohammedism. Under the cover of humor, Pope seizes the opportunity to express considerable epistolary intimacy with his addressee, Lady Mary Wortley Montagu: (9) b.


Without offence to your modesty be it spoken, I have a burning desire to see your Soul stark naked, for I am confident 'tis the prettiest kind of white Soul, in the universe-But I forget whom / am talking to, you may possibly by this time Believe according to the Prophet, that you have none. (Pope to Lady Mary Wortley Montagu, September 1, 1718, [bplet059]) They put me in mind Tho at a different time of year) of the Roman Saturnalia, when all the Scum, and Rabble, and Slaves of Rome, by a kind of Annual and limited Manumission, were suffered to make Abominable Mirth, and Profane the Days of Jubilee, with Vile Buffoonry, by Authority. But I forget that I am

154 Susan Μ. Fitzmaurice writing a Post Letter, and run into length like a Poet in a Dedication, when he forgets his Patron to talk of himself. (William Congreve to John Dennis, [cclet053]) In (9c), William Congreve uses the progressive with the verb write in the same kind of frame as Pope does, I forget (that/whom) I am writing/talking, in order to comment humorously on his own epistolary loquacity. His comparison of his epistolary self-indulgence with the inattention to his patron of a poet writing a dedication seems perfectly designed to entertain his addressee, the critic John Dennis. Again, the choice of the communicative verb draws attention to the subject's self-awareness as a correspondent and to his consciousness of the act of writing a letter. Mental verbs occur with the progressive 18% of the time in drama, and just under that in fiction and essays (14% and 16% respectively). Interestingly, in letters and essays the progressive selects communication verbs more frequently (20% and 23% respectively) than mental verbs (9% and 16% respectively), whereas drama seems to prefer the progressive with mental verbs more often (18%) than with communication verbs (14%). The most common mental verb in all registers is think, though even with verbs like consider and examine, this common set makes up less than 1% of the verbs in letters and fiction. These verbs amount to more than 3% of the verbs in essays and 6% of the verbs in drama. Two verbs, muse and wonder, occur three times each in fiction, in the contexts illustrated below. (10) a.



I had a great mind to bring it home if I could, for I had often been musing whether it might not be possible to get a kid or two, and so raise a breed of tame goats, which might supply me when my powder and shot should all be spent. (Defoe, Robinson Crusoe, 1720. [ddficOOl]) While the Marquess and Don Fabritio, were wondering at, and lamenting the Misfortune of her loss, Hippolito came towards Don Fabio and interceded for his Son, since the Lady perhaps had withdrawn her self, out of an aversion to the Match. (Congreve, incognita, 1695. [ccficOOl]) ... and taking him by the hand, [I] laughed at him, and pointing to the kid which I had killed, beckoned to him to run and fetch it, which he did; and while he was wondering, and looking to see how the creature was killed, I loaded my gun again; (Defoe, Robinson Crusoe, 1720. [ddficOOl])

Meanings and uses of the progressive construction 155 Defoe deploys the verb muse with first person singular subject in a main clause with pluperfect and temporal adverb often. This verb expresses the experiencing subject's assessment of the continual consideration of a question or possibility. In contrast, the verb wonder in (10b) and (10c) is used to capture the narrator's inference about the nature of a third subject's overt and expressed reaction to an event, based upon observation. Indeed, because the assessment of the subject's mental processes can only be guessed at, the verb selected describes those processes in terms of how they are observed as communicated to the narrator. In both of these cases, the verb could be construed, not in the same way as the subjective experience of thinking, but rather as the intersubjective, expressive process of marveling at some event. Most of the time, the writers in NEET do not modify or amplify the progressive construction by using lexical adverbial support. Out of the 980 progressives counted in the sample, only 190 are modified by adverbs, almost 20%. Further, this number is unevenly distributed across the registers, with almost 40% of all the adverb-supported progressives occurring in letters, about 30% in essays, 23% in fiction, and the tiny remainder, a paltry 7%, occurring in drama. However, if we examine the proportion of constructions that are supported by adverbs within each register, the picture looks a little different. For example, 27% of all the progressives in essays are supported by adverbs, 23% are supported by adverbs in the letters, 21% in fiction and 12% in drama. This contrast suggests that we might expect to find particularly subjectively colored or emotive progressives uses in the essays register in comparison with the other registers. In addition, a range of adverbs and adverbials are available to speakers for lexical support (see Figure 5 and Table 6, which reports the proportional occurrence of adverb type across registers). Although it is possible to categorize these adverbs into the standard categories of time, degree, manner and place adverbs, speakers adopt particularly marked versions in order to convey their attitude towards what they are talking about. For example, although I have categorized adverbs like eternally, perpetually and constantly as temporal adverbs, they participate strongly in the construction of powerful speaker stance as they are used to convey far more than the situation warrants.

156 Susan Μ. Fitzmaurice Table 6. Proportional occurrence of adverb type across registers Adverb type time degree manner place

Letters N=75 44% 51% 4% 1%


Fiction N=44 48% 50% 2% 0%



Essays N=56 72% 18% 5% 5%

Drama N=15 20% 73% 7% 0%



Figure 5. Adverb choice across register

The downtoner just (N=22) and the temporal deictic now (N=ll) account for nearly 43% of the occurrences of adverbs in the letters. All the writers adopt now as a modifier of the aspectual progressive, as illustrated in the quotations below, selected from the letters subcorpus:

Meanings and uses of the progressive construction 157 (11) a.



I have yet a Mother of great age & infirmitys, whose last precarious days of life I am now attending, with such a solemn pious kind of officiousness, as a melancholy Recluse watches the last risings & fallings of a dying Taper. (Pope to Lady Mary Wortley Montagu, 1718. [bplet052]) I am now reasoning as if I had bin against complying with what is usual. So far from it that I was willing to part with double to what is customary, and he woud surely accept of it if he did not prefer the other. (Edward Wortley to Lady Mary Pierrepont, 1712. [ewlet027]) We the Corporation are just now drinking your Lordships health a Welcome first Service. (Richard Steele to Sir Andrew Hume of Kimmerghame, Nov. 1717. [rslet067])

The quotations in (11a) and ( l i b ) illustrate the use of now as a temporal adverb which specifies and restricts the activity being reported to the writer's here and now. In (11a), Pope indicates his attention to his dying mother. The progressive occurs in a relative clause with first person singular subject to express the temporary nature of Pope's care of his mother. In contrast, Edward Wortley's language expresses a shift in his attitude towards the whole process of negotiating marriage terms with Lady Mary's father, Lord Kingston. The choice of the mental verb reason with the proximal deictic suggests a degree of self-consciousness on the part of the speaker, and it is tempting to ascribe an experiential function to the construction in this first person present tense context. Example (11c) combines the deictic with what seems to be an emphatic, just, in a main clause to suggest the writer's desire to impress the spontaneity and immediacy of the toast upon the addressee. Always, a key pragmatic strengthened occurs 6 times, to make up 8 % of occurrences in the letters. Just (N=12) and now (N=3) account for 34 % of the progressive-adverb constructions in fiction, and 18% of the constructions in essays (now = 8, just = 2 out of 56). In essays, always occurs 5 times - 9%. This leaves considerable space for expressive adverbs such as perpetually (which occurs 6 times in essays - nearly 11% of the total number of adverbs counted). The use of such expressive adverbial support is illustrated in (12a-c): (12) a. All our modern Writers follow this System- We are always mistaking Knavery for Wisdom, and Luxury for Politeness. - I shall

158 Susan Μ. Fitzmaurice be told that I forget the Churches are still open, and that we have Weekly Sermons full of Piety and Morality; but I am afraid they have no farther Influence on our Manners, than the Exercises of the Militia have in teaching the English Youth the Art of War. (Lady Mary Wortley Montagu, Nonsense of Common-Sense, No. 6, January 24, 1738. [mmess006]) b. At the same time I intend to do justice upon our Neighbours, Inhabitants of the upper Parnassus; who taking advantage of the rising Ground, are perpetually throwing down Rubbish, Dirt, and Stones upon us, never suffering us to live in Peace: (Alexander Pope, chapter 1, Martinus Scriblerus His Treatise of the Art of Sinking in Poetry. 1714. [bpess012]) c. Behold a fourth, in much and deep conversation with himself, biting his thumbs at proper junctures, his countenance checkered with business and design, sometimes walking very fast, with his eyes nailed to a paper that he holds in his hands; a great saver of time, somewhat thick of hearing, very short of sight, but more of memory; a man ever in haste, a great hatcher and breeder of business, and excellent at the famous art of whispering nothing; a huge idolator of monosyllables and procrastination, so ready to give his word to everybody, that he never keeps it; one that has forgot the common meaning of words, but an admirable retainer of the sound; extremely subject to the looseness, for his occasions are perpetually calling him away. (Alexander Swift, A Digression Concerning Madness, 1704. [asess009]) What is interesting about the quotations in (12a) through (12c) is that they are all extracted from satirical essays. The function of these essays is to point to the folly of attributing value and respect to hack writers (12a), to petty poets (12b), and to madmen (and moderns) (12c). The effect of the lexical support appears to be to provide pragmatic strengthening for the expression, enabling the writer to overstate a case in order to cause the reader to treat the proposition with suspicion. One function of satire is to present a case whose absurdity leads the reader to interrogate the basis for the portrait drawn. Shifting a habitual activity (call) into an iterative event as in (12c) serves to misrepresent the activity described, and thus draws attention to its oddity. The use of always allows the writer to attribute excess to any activity being reported, and thus similarly allows the interpretation of unwarranted or unruly behaviour (12b). The use of

Meanings and uses of the progressive construction


perpetually in (12c) invites the inference that there is other and more meaning to be attached to the claim that the subject has no control of his language.


The use of the progressive in the Network of Eighteenth-century English Texts

6.1. The distribution of the progressive across the generations in NEET In this section, let us briefly examine whether, and if so, to what extent, the three generations of writers differ from one another with respect to their use of the progressive construction. Because the generations differ considerably with respect to the amount and range of writing they produce, this picture should not be construed as truly representative of the periods that the writers belong to. Figure 6 indicates the overall frequencies for each generation for each register.






Figure 6. Mean frequencies across registers and generation

It is important to regard the picture in Figure 6 with some skepticism, not least because Generations 1 and 3 have far fewer texts and far fewer writers than Generation 2. For instance, although the progressive appears

160 Susan Μ. Fitzmaurice

to be more prevalent in Generation 1 's use than in that of either of the later generations, this may simply reflect the fact that it captures the use of two people only, in the case of two registers (letters and essays) of one person, namely Dryden. In contrast, the range of use that informs the means for each register for Generation 2 averages the contributions of two or three people at the least (in the cases of drama and fiction respectively), and nine (for essays) and eleven (in the case of letters) at the most. Generation 3, like Generation 1, is sparsely populated, and this uneven contribution may well give the impression of more frequent use than a more representative sample might do.7 However, this picture does allow us to infer that register is a more salient variable in distinguishing the use of the progressive than generation might be. For example, essays are quite distinct from the other registers with means ranging between 2.5 and 2.9 occurrences per 10,000 words. Similarly, although there is more variation across generations within the letters and drama registers than there is in the essays register, there is clearly less variation across generations than across registers.

6.1. The distribution of the progressive within Generation 2 in NEET The possibility that the network ties that select the writers in NEET may play some role in marking uses of the progressive may be of more interest than the use of the progressive across the generations. Because Generation 2 is the center of the networks that define NEET, let us consider the variation in this group. A quick study of the range of use by individual and across register within Generation 2 might allow us to make inferences about the extent to which belonging to a cohort of colleagues and friends might influence individuals' linguistic practices. As it happens, there is considerable variation in individual preferences for the progressive in Generation 2. Table 7 summarizes individual levels of use across registers. The range of frequency across essays for this generation is between 1 and 9 occurrences per 10,000 words, and for letters, the range is almost double this - between 1 and 17 occurrences per 10,000 words. In contrast, fiction and drama - both registers with comparatively few exponents - exhibit smaller ranges of frequency. Fiction has a range between 7 and 16 occurrences per 10,000 words and drama has the narrowest range of all, between 6 and 9 occurrences per 10,000 words.

Meanings and uses of the progressive construction 161 Table 7. Level of individual selection of the progressive: Generation 2 Essays



Daniel Defoe Sarah Churchill George Stepney Matthew Prior Mary Astell

9.16 1.85

17.28 5.52 7.11 12.95


Jonathan Swift

7.89 0.95


Susannah Wesley Delariviere Manley

4.13 1.73


1.60 7.06 1.47

William Congreve


Edward Wortley Joseph Addison

4.93 2.76



9.38 16.01

8.01 8.55 Richard Steele Note: a gap indicates that the individual has no score for that register



The figures in Table 7 suggest that Generation 2 comprises people who eschew the progressive on the one hand, and individuals who appear to select the progressive on the other. For example, Susanna Wesley and Mary Astell show very conservative levels of use, regardless of register. Wesley simply does not use the progressive very much at all (0.95 and 1.47 occurrences per 10,000 words in essays and letters respectively); Astell is very consistent in her very scant use of the construction (1.73 and 1.60 per 10,000 words in essays and letters). Sarah Churchill shows a similarly conservative level of use in her memoir, but her letters exhibit a tendency to select the construction more frequently than a number of her peers. It is worth noting that these three women have no connections with each other or with the central network defined by Addison's relationships. At the other end of the frequency scale, Defoe exhibits a greater tendency to deploy the progressive in both essays and letters than anybody else in the generation. Indeed, he is the most profligate user of the construction out of the whole group in letters (mostly to Robert Harley). In contrast, he uses the progressive proportionally least frequently in fiction, here represented by his (1720) novel, Robinson Crusoe,8 Interestingly, as the women whose scores are reported above are outside of the network, so Defoe is also outside of the network. These four people have several traits in common. Firstly, none of them has any direct or lasting connection with the figures

162 Susan Μ. Fitzmaurice

in the central network. Secondly, their educational and personal experiences differ greatly from those of the people in the central group. None of the women would have been educated at school, but tutored privately in very different circumstances. Daniel Defoe, as a non-conformist, attended an academy for dissenters rather than pursuing first a grammar school, and then an Oxford or Cambridge education. Let us turn to the preferences exhibited by the central group in the network. The social networks analysis sociolinguistics literature suggests that the stronger the ties that people contract among them, and the more contexts in which they encounter one another, the more likely it is that these social affinities will influence people's linguistic practices (Fitzmaurice 2000b; Milroy 1987; Milroy and Milroy 1985; Tieken 2000). In other words, just as people's personal and professional connections influence and shape one another's tastes in art and literature, so these connections are likely to shape their language use. Indeed, Addison's closest associates exhibit comparable levels of use of the progressive, suggestive evidence that the stronger the ties and the more dense the connections among people, the more likely it is for these people to use language in similar ways to one another. The range of frequency for letters in Addison's network is 10 occurrences per 10,000 words, with Wortley at the bottom end (2.76 occurrences per 10,000 words) and Prior at the top end (12.95 occurrences per 10,000 words). For essays, the range is much narrower at 6 occurrences per 10,000 words, with Congreve at the bottom end (2.44 occurrences per 10,000 words), and Swift at the top end (7.89 occurrences per 10,000 words). Addison himself follows the middle course - his own use is moderate relative to his cohort with 5.44 occurrences per 10,000 words for essays and 7.88 occurrences per 10,000 words for letters. Richard Steele, Addison's closest professional associate and long-lived friend, George Stepney, his early associate and diplomatic colleague, and Jonathan Swift, collaborator and friend, all exhibit a very similar level of progressive use in their letters, between 7 and 8.55 occurrences per 10,000 words. There are three figures in Addison's network whose use falls outside this central range; Edward Wortley and William Congreve demonstrate slightly lower preferences for the progressive in their letters than Addison does, and Matthew Prior exhibits a comparative penchant for the construction in his letters. There is no similar, identifiable central tendency in the network members' use of the progressive in essays. Although the numbers reported in Table 7 might suggest linguistic affinities among the people in Addison's cohort, they offer little help in

Meanings and uses of the progressive construction 163 identifying the possible sources of what might turn out to be linguistic affinity. Of course, what might be construed as individual preferences and network effects may be dampened by the more robust effects of register. Only two individuals seem to exhibit idiolectal patterns instead of registerbased patterns: Jonathan Swift appears to select the progressive about as often in essays (7.89 per 10,000 words) as in letters, and Richard Steele's selection shows considerable comparability across registers (8.01, 8.55, and 9.40 per 10,000 words in essays, letters, and drama). Other writers show much more dramatic variation depending upon the register. Accordingly, virtually all the figures in Generation 2 show lower levels of use in essays than letters.

7. Aspectual and experiential uses of the progressive In this final section, let us consider the extent to which the experiential progressive participates in the uses of the construction by the NEET writers in the registers that they produce. I suggested above that the context in which the experiential function of the progressive was most likely to be active was in present-tense main clause contexts with first person singular subject and private or mental verbs. To ascertain how susceptible the progressive is to appropriation as an indicator of subjective expression in different registers, I examined the collocation patterns used for the frequency analyses in section 5 above. Specifically, I examined all the present-tense, first-person singular, main-clause progressives with mental verbs and adverbial support in order to discover whether a pragmatic analysis would yield the inference of subjective expression. I also looked at the same contexts without adverbial support, to see whether subjective expression survives without the pragmatic strengthening that lexical adverbs routinely provide. The total number of first person main clause progressives is 114, almost 11% of the total. Ofthat number, nearly half, 54, exhibit adverbial support. Of those 114 progressives, only 14 collocate with mental verbs, including think, consider, recollect, pity, perplex, reason, wish and examine. There are other verbs that would not ordinarily be categorized as mental verbs, but which the writers seem to appropriate to communicate mental states or processes, such as trouble and drudge (one of Dryden's favorite ways to describe his literary labours). The numbers indicate that the lexico-

164 Susan Μ. Fitzmaurice grammatical contexts that might routinely yield experiential uses of the progressive are negligible. However, it is worth examining exactly what these progressives mean before dismissing the construction as a candidate for recruitment to subjective pragmatic use. Because Jonathan Swift and Richard Steele exhibit what appear to be idiolectal preferences for the progressive in their letters and in their essays (see Table 7 above), it is worth considering the extent and range of functions that they assign the progressive. Swift deploys the progressive in constructions with the first person singular pronoun subject in main clauses more frequently (10 first person singular main clause occurrences) than everybody in the corpus except for Defoe (13 first person singular main clause occurrences in letters), followed closely by Steele (8 first person singular main clause occurrences). Let's examine some of their uses in an effort to ascertain the progressive's functions. The examples in (13a) and (13b) provide some sense of the ways in which Steele deploys the progressive in the brief, intimate notes that he was in the habit of having hand delivered to his wife throughout the day. In (13a), Steele appears to be trying to make up after a quarrel over his constantly impecunious state and his persistent failure to keep any money that he happened to lay his hands on. The progressive occurs only once in this excerpt, but it appears in a distinctive grammatical context - in the complement of an imperative that is the apodosis of a conditional 'if-then' construction. Further, it occurs with what at first seems to be an activity verb, pursue, but which can only really be construed figuratively with the abstract object our Mutuall Good. The sentence containing the progressive clause - his plea to his wife to believe that he is invested in working in their best interests - is really a preamble to the meat of Steele's letter. And the meat consists of his confession that despite the fact that he expects to secure some income, much of that income will be paid out immediately to his creditors. In this context, the choice of the progressive as a complement of the imperative believe seems intended to persuade his addressee that he is utterly sincere in his efforts to ensure the financial security of his family. In pragmatic terms, he is overstating his ability to do this, and so the effect is that he appears to say more and other than is warranted in the sitiation, predictably engendering his wife's impatience and beginning a new quarrel. (13) a. I wish I knew how to Court you into Good-Humour, for Two or Three Quarrells more will dispatch Me quite. If you have any Love for Me beleive / am always pursuing our Mutuall Good. Pray con-

Meanings and uses of the progressive construction 165 sider that all My little fortune is to [be] settled this month and that I have inadvertently made my self Liable to Impatient People who take all advantages. (Richard Steele to Prue Steele, 7 June 1708) b. It is a strange thing because you are handsome, that you won't behave your self with the obedience that people of Worse features do, but that I must be always g/ving You an account of every trifle, and minute of my time. I send this to tell you I am waiting to be sent for again when my Lord Wharton is stirring. (Richard Steele to Prue Steele, December 1708) The complete letter in (13b) consists of three progressives, two first person singular progressives in subordinate clauses and one main clause present tense third person progressive. The example is striking for the frequency with which the construction occurs in such a small stretch of text, and for the variation in its use in this example. Steele first complains about his wife's constant need to know his whereabouts and activities every minute of the day, and then he complies with her wishes by telling her what he is doing as he writes this note. To construct the complaint, he pays his wife a back-handed compliment which contrasts the pleasantness of her demeanour and person with her impatient and meddling behavior. To express the extent to which he feels put upon by her demands, he deploys a modal, must, to supply the sense of the obligation he feels is imposed upon him, and an intensifying adverb, always, to supply the sense of persistence, in collocation with an apparent verb of activity, give, in the progressive, to supply the sense of iteration. The components of obligation, persistence and iteration that are built into the progressive construction are mentioned in the object NP in the lexical compound, every trifle and every minute of my time. Thus Steele uses the progressive as part of his rhetorical strategy for expressing his attitude towards his addressee and what he is talking about. The other progressives in the letter occur in the information he judges his wife to be interested in, and serve to describe the duration of the activity of waiting on the one hand (/ am waiting), and the momentariness of the expected arrival of Wharton (is stirring). These instances together underline the use of the construction as a means of conveying the writer's attitude both towards his addressee and to what he is writing about. (Note that examples (Id), (2d), (8b), and (11c) illustrate further the range of Steele's deployment of the construction.) Finally, let us examine how Jonathan Swift utilizes the progressive construction in his letters. (Note that examples (lb, e) and (12c) illustrate his use of the construction in the essay register.)

166 Susan Μ. Fitzmaurice (14) a. My letter you saw to Lord Bolingbroke has shown you the Situation I am in, and the Company I keep: If I do not forget some of the Contents. But I am now returning to the noble Scene of Dublin in to the Grande Monde, for fearing of burying my parts to Signalise my self among Curates and Vicars, and correct all Corruption crept in relating to the weight of Bread and Butter through those Dominions where I govern. Mr Lewis sent me an Account of Dr Arbuthnett's Illness which is a very sensible Affliction to me, who by living so long 'out of the World have lost that hardness of Heart contracted by years and generali Conversation. I am daily loosing Friends, and neither seeking nor getting others. O, if the World had but a dozen Arbuthnetts in it I would burn my Travells but however he is not without Fault. There is a passage in Bede highly commending the Piety and learning of the Irish in that Age, where after abundance of praises he overthrows them all by lamenting that, Alas, they kept Easter at a wrong time of the Year. So our Doctor has every Quality and virtue that can make a man amiable or usefull, but alas he hath a sort of Slouch in his Walk. (Swift to Pope, [aslet054.txt]) Interestingly, the tone of Swift's letter to Pope, written from Ireland in the 1730s, resembles the ironic, self-deprecating one that marks some of his satirical essays. In this excerpt, the progressive occurs with the first person singular subject in a main clause with adverbial support. Although Swift chooses the activity verb return, he does so for a particular effect; he is reporting his return from the country to Dublin at the same time as he is describing his departure from obscurity into the comparative social exposure that being in Dublin affords him. Thus there are twin effects of the material and concrete on the one hand, and of the social and psychological on the other. The lexical and rhetorical structure of the comment that follows the progressive cements the experiential, strongly stanced color of Swift's report of his comeback to the world of the capital. He deploys hyperbole to describe his fear of being buried in obscurity among the lower clerical orders in the country and his ambition to attend to the accuracy of weights and measures in the land. Here he mixes lexical understatement (bread and butter) with overstatement (corruption, dominions) to cast a humorous light on his report. In the next paragraph, Swift assuages the impression of his evident distress at receiving the news

Meanings and uses of the progressive construction 167 of Arbuthnot's illness by adopting a tone that gently mocks, and thus displaces, the possibility of his sounding self-pitying. Thus his lament of the loss of friends is expressed somewhat comically in a first person main clause progressive modified by a temporal adverb, daily, which offers the impression of a man actively losing friends each day. Again, although the verb is an activity verb, it also conveys a figurative, emotive overtone, as if the subject is somehow culpable for the disappearance of his friends. The sense of culpability is underlined by the conjunction of progressives that expresses the observation that he neither seeks nor gains new friends. The sentence that follows, a pledge prefaced by the dramatic interjection, O, but rapidly undercut by the observation that despite Arbuthnot's worth, his friend is not without fault, further reinforces the mock lament quality of Swift's observations. The effect is extended by his anecdote regarding the venerable Bede's qualified appreciation of the Irish, a comic form of selfreference that would resonate with his friends. (14) b.

I know no People so ill used by your Men of Business as their intimate Friends. About a fortnight after Mr Addison had received the Letter you were pleased to send me, he first told me of it with an Air of Recollection, and after ten days further, of Grace, thought fitt to give it me, so you know where to fix the whole Blame that it was no sooner Acknoledged. Tis a delicate Expedient you Prisoners have of Diverting your selves in an Enemyes Country, for which other men would be hanged. I am considering whether there be no way of disturbing your quiet by writing some dark Matter that may give the French Court a Jealosy of you, I suppose, Mon. Chamillard or some of his Commissioners must have this Letter interpreted to them before it comes to your Hands; (Jonathan Swift to Robert Hunter, March 22, 1708-9. [aslet010.txt])

In the excerpt from a letter to Robert Hunter in (14b), Swift deploys the bare progressive with the first person in a main clause with a mental verb, consider. He acknowledges a letter received after delayed delivery by their friend Addison, and proceeds to tease Hunter, an English diplomat at the French court, by characterizing him as a prisoner in enemy territory. And he shares with his friend a possible plan for repaying him in kind for sending the letter via Addison. Specifically he teases Hunter with the prospect of sending a mischievous letter to him via the French Commissioners. Swift hints at this mischief with the impression of careful

168 Susan Μ. Fitzmaurice

deliberation, expressed in the present progressive, I am considering, which governs the nominal headed by whether. The lexis and syntax of logic and reasoning provides a cover for the joke about the hazards of trusting intimate friends. These two sets of examples illustrate the kind of functions and effects that the progressive could be invested with in rather different kinds of texts. Steele's intimate, impatient notes to his wife contrast with Swift's literary, self-mocking letters to Pope, but they both deploy first person main clause progressive constructions for particular rhetorical and pragmatic effects, both subjective and aspectual. In sum, the main clause, first person subject frame does seem to accommodate the communication of the speaker's subjective experience. The evidence that lexically unsupported uses of the construction can fulfill this function suggests that adverbial support or selection of a mental verb may not be necessary for the progressive to have this function. Lexical context thus appears much more malleable than grammatical context, but this is a question for further investigation. Additionally, the brief qualitative analyses of a number of quotations offered in section 5 above indicated that the contexts examined in (13)(14) above are not the only uses from which expressive or experiential force can be inferred. Indeed, these comments suggest that the expressive force conveyed by first person singular present tense main clause progressives may be transferred to other contexts, so that it may be inferred from use with plural subjects or third person subjects, in subordinate as well as main clauses, and indeed with communication and activity verbs as well as mental verbs. A close reading of examples such as those in section 5 above suggests a sense of the degree to which the progressive might be understood as fulfilling functions other than mapping the internal aspectual structure of events and activities.

8. Concluding remarks This study both builds on and departs from my earlier work on the history of the experiential or subjective progressive construction by situating the study of its frequency and function within a broader context, and by adopting a test for the differentiation among experiential/subjective progressives and aspectual progressives. The context consists of a survey of the progressive's overall distribution - both experiential and aspectual functions - across multiple registers. The baseline frequency counts facili-

Meanings and uses of the progressive construction


täte the beginnings of an assessment of the extent to which the nonaspectual, non truth-conditional progressive occurs relative to the aspectual construction. Although I continue to invoke five grammatical contexts as those likely to trigger the experiential function of the progressive, I now rely upon the figurative or metaphorical function of the lexical verb in colligation with the progressive as the indicator of subjective meaning. The results of this study provide some suggestive implications for future research on the subject. The frequency analyses reported in Section 5 indicate that the progressive does not seem to prefer any single lexico-grammatical context above all others in the four registers examined. The closest that we might come to a statement regarding a preferred context for the progressive is that in three of the four registers (excluding drama), the construction tends to occur most often with activity verbs in past tense subordinate clauses. In the other register - drama - main clauses accommodate the progressive, notably in present tense with first person subjects and activity verbs. This observation supports the inference that the preferred function for the progressive construction is aspectual, satisfying criteria like duration, temporariness, iteration or framing of an event. And indeed the figures display this tendency. However, close analysis of first person present tense main clause progressives, as illustrated in section 7, suggests that the construction may be recruited for the pragmatic function of subjective expression. The question of whether this expressive function might be regarded more as a feature of a writer's idiolect than of the language of the period is, however, one that we must leave for investigation on another occasion.

170 Susan Μ. Fitzmaurice

Appendix Table A.I. Text sampling by date per writer by register Figure John Dryden Aphra Behn Daniel Defoe Sarah Churchill George Stepney Matthew Prior Mary Astell Jonathan Swift Susannah Wesley Manley, Delariviere William Congreve Edward Wortley Joseph Addison Richard Steele Alexander Pope Mary Montagu Eliza Haywood

Essays 1668-1691

Letters 1653-1700

1701-1717 1745

1707-1729 1700-1727 1694-1698 1694-1720 1693-1694 1691-1731 1702-1722

1700-1714 1695 1704-1729 1709-1741 1709-1711 1700-1705 1715-1716 1709-1714 1704-1737 1713-1738

1694-1725 1710-1757 1694-1709 1702-1722 1713-1718 1710-1792

Fiction 1683-1688 1719

1709 1695

Drama 1673 1688


1705 1756 1719

1734 1723

Notes 1. This study both builds upon and departs from earlier work (Wright 1986, 1994, 1995), which argued that the progressive underwent a process of subjectification in grammaticalisation in the course of the seventeenth century. 2. For details about the structure and rationale for NEET, see Fitzmaurice (2002, 2003b). 3. In contrast to the central network's key figures, Defoe's lower-class, nonconformist background and lack of university training hindered the access he desired to those more powerful people who shared common backgrounds with their clients. Unlike Richard Steele, whose connections won him a Membership of Parliament and a knighthood, Defoe's efforts did not gain him a position that gave him comparable status and political power. Instead, when preferment came to him, it was in the form of a job as an itinerant, political worker for Robert Harley. At the opposite end of the social scale was Sarah Churchill, Duchess of Marlborough. She was Queen Anne's primary confidante until 1705, when her quarrel with Anne led to her replacement by Abigail Masham, a relative of

Meanings and uses of the progressive construction


Robert Harley. Matthew Prior had attempted to win her notice when he wrote a poem in praise of Marlborough in 1704, but the duchess returned his letter to him unopened. Prior endorsed the letter thus: "Mem: dam She sent back the letter unopen'd and said she was sure yt Mr Prior write but what he would, He could not wish well to Her and her family" (Prior Papers [Marquess of Bath] vol. 13: 55). Susannah Wesley, by virtue of her very different background as the outspoken wife of an Anglican vicar in the provinces, was not connected at all with any of the circle. Daniel Defoe and Susannah Wesley remain outside the group at large, proving to be the only true outsiders in social terms. 4. For example, the essays register includes Wortley Montagu's Spectator essays from 1714, and Pope's contributions to The Spectator and to The Guardian from 1713. 5. It might be argued - with more data in support - that the progressive's use increases from Generation 2 to Generation 3. For example, if we examine the range of use within Generation 2, it is clear that some writers use the expression very rarely. For example, Susanna Wesley and Mary Astell are very conservative in their use (1.47 per 10,000 words and 1.6 per 10,000 words respectively), in contrast with Daniel Defoe and Matthew Prior (17.28 and 12.95 per 10,000 words respectively). Indeed, a quick examination indicates that women in the middle generation are far more conservative in their selection of the progressive than men in the same generation are, regardless of the register. However, this observation should not be taken as a claim that women are more conservative overall than men. 6. This latter figure may well be an artifact of the sample size and scope. The corpus includes the entire novel, somewhat skewing the sample of fiction for the corpus as a whole.

References Amaud, Rene 1982 On the progress of the progressive in the private correspondence of famous British people (1800-80). In Papers from the Second Scandinavian Symposium on Syntactic Variation, Sven Jacobson (ed.), 83-94. Stockholm: Almqvist & Wikseil. 1998 The development of the progressive in 19th century English: a quantitative survey. Language Variation and Change 10: 123-152. Biber, Douglas, and Edward Finegan 1989 Drift and the evolution of English style: A history of three genres, Language 65: 487-517. 1997 Diachronic relations among speech-based and written registers in English. In To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, Terttu Nevalainen and

172 Susan Μ. Fitzmaurice Leena Kahlas-Tarkka (eds.), 253-275. (Memoires de la Societe Neophilo-logique de Helsinki 52.) Helsinki: Societe Neophilologique. Denison, David 1993 English Historical Syntax: Verbal Constructions. London/New York: Longman. Fitzmaurice, Susan 2000a The Spectator, the politics of social networks, and language standardisation in eighteenth-century England. In The Development of Standard English, 1300-1800, Laura Wright (ed.), 195-218. Cambridge: Cambridge University Press. 2000b Coalitions and the investigation of social influence in linguistic history, European Journal of English Studies 4 (3): 265-276. 2002 Politeness and modal meaning in the construction of humiliative discourse in an early eighteenth-century network of patron-client relationships. English Language and Linguistics 6(2): 1-27 2003a The textual resolution of structural ambiguity in eighteenth-century English: A corpus linguistic study of patterns of negation. In Variation and Varieties of Language: Corpus Approaches, Randi Reppen, Susan Fitzmaurice, and Douglas Biber (eds.), 227-247. Amsterdam: Rodopi. 2003b The grammar of stance in early eighteenth-century English epistolary language. In Corpus Analysis: Language Structure and Language Use, Charles Meyer and Pepi Leistyna (eds.), 107-131. Amsterdam: Rodopi. Grundy, Isobel 1996 Lady Mary Wortley Montagu: Romance Writings. Oxford: Clarendon. Huddleston, Rodney, and Geoffrey Pullum 2002 Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Milroy, James, and Lesley Milroy 1985 Linguistic change, social network and speaker innovation. Journal of Linguistics 21: 339-384. Milroy, Lesley 1987 Language and Social Networks. 2nd ed. Oxford: Blackwell. Pratt, Lynda, and David Denison 2000 The language of the Southey-Coleridge circle. Language Sciences 22: 401-22. Ryden, Mats 1997 On the panchronic core meaning of the English progressive. In To Explain the Present: Studies in the Changing English Language in

Meanings and uses of the progressive construction


Honour of Matti Rissanen, Terttu Nevalainen and Leena KahlasTarkka (eds.), 419-429. (Memoires de la Societe Neophilologique de Helsinki 52.) Helsinki: Societe Neophilologique. Smith, Nicholas ms. Ever moving on? The progressive in recent British English. Smitterberg, Erik 2000 The progressive form and genre variation during the nineteenth century. In Generative Theory and Corpus Studies: A dialogue from 10 ICEHL, Ricardo Bermudez-Otero, David Denison, Richard Hogg and C.B. McCully (eds.), 283-297. Berlin/New York: Mouton de Gruyter. Strang, Barbara 1982 Some aspects of the history of the be + ing construction. In Language Form and Linguistic Variation, John Anderson (ed.), 427-474. Amsterdam: John Benjamins. Tieken-Boon van Ostade, Ingrid 2000 Social network analysis and the history of English. European Journal of English Studies 4.3: 211-216. Wright, Susan 1986 Tense, aspect and text: Processes of grammaticalisation in the history of the English auxiliary. Ph.D. diss., Cambridge University. 1994 The mystery of the modal progressive. In Studies in Early Modern English, Dieter Kastovsky (ed.), 467-485. Berlin/New York: Mouton de Gruyter. 1995 Subjectivity and experiential syntax. In Subjectivity and Subjectivisation: Linguistic Perspectives, Dieter Stein and Susan Wright (eds.), 151-172. Cambridge: Cambridge University Press. Ziegeler, Debra 1999 Agentivity and the history of the English Progressive. Transactions of the Philological Society 97(1): 51-101.

Investigating the expressive progressive: On Susan M. Fitzmaurice's "The meanings and uses of the progressive construction in an early eighteenthcentury English network"1 Erik Smitterberg

In her article, Susan M. Fitzmaurice focuses on progressive constructions that have experiential and/or subjective functions in addition to, or instead of, aspectual functions (I will refer to such progressives as "experiential/subjective" henceforth). In doing so, she addresses a central issue regarding the progressive in English, viz. what the functions of the construction are. This issue is of special relevance from a diachronic perspective, as there is evidence to suggest that the functions of the progressive have not remained constant in diachrony. Most importantly, many scholars claim that the Old English progressive's main function was emphatic rather than aspectual (see, for instance, Ryden 1997), whereas there is broad agreement that the most frequent function of the late Modern English progressive was aspectual, as acknowledged by Fitzmaurice. Despite the quantitative shift from chiefly emphatic to chiefly aspectual functions, most researchers seem to be of the opinion that not all occurrences of the progressive can be given a solely aspectual interpretation even in Present-day English. As there is no difference in the syntactic make-up of aspectual and experiential/subjective progressives, it is relevant to ask how language users distinguish these two types. Fitzmaurice suggests that a number of linguistic features, alone and in combination, may help to trigger an experiential/subjective reading. Her suggestion is of great interest, as "[t]he appeal to pragmatic competence or intuition in ascribing to different occurrences of the progressive an experiential or subjective function is hard to account for at best and it is unreliable at worst" (Fitzmaurice, this volume). Finding linguistic features that tend to frame experiential/subjective progressives would make studies of these progressive constructions less dependent on close readings, which always contain an element of subjectivity that decreases the comparability of different classificatory frameworks.

176 Erik Smitterberg However, even if such linguistic features can be found, it is not advisable to study the functions of the progressive wholly independently of the overall frequency of the construction. To begin with, a comparatively high frequency of progressives in a given text will lead to a high incidence of experiential/subjective progressives, provided that the percentage of progressives with experiential/subjective connotations remains constant across texts. But as studies like Fitzmaurice (this volume) and Smitterberg (2002) indicate, genres where the progressive construction is comparatively frequent may also have a higher percentage of experiential/subjective progressives than other genres: this factor further increases cross-genre differences in the frequency of experiential/subjective progressives. It is thus appropriate that Fitzmaurice begins the presentation of her results by discussing the overall frequency of the progressive in her corpus (NEET, the Network of Eighteenth-century English Texts). As the parameter of genre is important to the occurrence of the progressive, Fitzmaurice's inclusion of four genres - drama, essays, fiction, and letters - in the investigation is of great value. Indeed, as in previous studies, there are considerable cross-genre differences in the overall frequency of the progressive: drama exhibits the highest frequency of progressives and essays the lowest, with fiction and letters covering a middle ground. In contrast, the overall frequency of the progressive does not appear to change across the time span covered by NEET, although there are considerable differences between individual writers in this respect. However, as Fitzmaurice points out, characteristics of the corpus make-up may have influenced these results. The NEET corpus was designed primarily to reflect the language of members of the network around Joseph Addison. This choice of design is justified, but it means that individual writers do not contribute the same amount of text to the corpus; nor is the period and genre coverage balanced regarding either the number of words or the number of writers sampled. These factors may skew the results. Nevertheless, Fitzmaurice's finding that the genre parameter "is a more salient variable in distinguishing the use of the progressive than generation might be" enables interesting comparisons with previous research. Biber and Finegan (1997) found that cross-genre differences in the occurrence (and co-occurrence) of linguistic features have become more marked during the Modern English period; their analysis does not include the progressive, but Fitzmaurice's findings, together with results reported on in Smitterberg (2000), indicate that this feature follows the same pattern.

Investigating the expressive progressive


In an analysis of idiolectal differences regarding the frequency of the progressive, Fitzmaurice makes full use of the fact that the corpus consists chiefly, but not solely, of texts produced by members of the same network. Her analysis centers on the possibility that their belonging to the network around Addison may have influenced the network members' use of the progressive. The investigation shows that the progressive is used sparingly by three of the women whose texts are sampled - Susanna Wesley, Mary Astell, and Sarah Churchill, none of whom are members of the central network. Daniel Defoe, another outsider, also stands out linguistically from his contemporaries inside the network: Defoe uses the progressive more in letters and essays, but less in fiction, compared with network members. However, several extralinguistic parameters may have influenced these results, which makes it difficult to estimate what impact the network itself has had on the occurrence of the progressive in NEET: Fitzmaurice points to differences in "educational and personal experiences." Given the comparative scarcity of progressives in the women's writing, the gender parameter is of special relevance. Interestingly, studies of nineteenth-century English have indicated that women use the progressive more than men do (Arnaud 1998; Smitterberg 2002); this pattern would imply that the progressive is the innovative form in a change from below, and that its use is not subject to overt sociocultural evaluation (see Labov 2001: 292-293). More research is needed to ascertain whether the different picture afforded by Fitzmaurice's investigation suggests that gender differences in the use of the progressive were reversed between the eighteenth and nineteenth centuries, or whether other factors (e.g., network membership and educational and socioeconomic status) are important. As for Defoe, Fitzmaurice emphasizes the social difference between him and the members of the network, which opens up the possibility that we are dealing with socioeconomically conditioned variation. In further studies, it would be worthwhile including additional writers outside the network around Addison, with the aim of balancing the variables of genre, network, gender, and socioeconomic status in the corpus make-up: such a study would be able to answer some of the further research questions Fitzmaurice's innovative method raises. In this respect, it is helpful that the central period covered by NEET is fairly narrow (c. 40 years). First, this makes it more feasible to assume that two writers with similar backgrounds and occupations have the same socioeconomic status, as the socioeconomic system itself will not have changed dramatically during the period covered by the corpus. Secondly, as the frequency of the progressive has been in-

178 Erik Smitterberg creasing throughout the Modern English period, a relatively short period span increases the reliability of a procedure where the period is treated as a synchronic whole. I now turn to Fitzmaurice's treatment of experiential/subjective progressives, and specifically to her analysis of linguistic features that may cooccur with these progressive constructions. The two linguistic features Fitzmaurice first focuses on are clause type and tense: Fitzmaurice considers that main clauses and the present tense are more likely environments for experiential/subjective progressives than subordinate clauses and the past tense. Based on the four genres investigated, Fitzmaurice draws the conclusion that "the most common grammatical context for the progressive in this whole period appears to be the past tense subordinate clause - a grammatical context that is least likely to host a non-aspectual, figurative progressive." However, a close look at the figures that she presents (Table 3) suggests that main-clause, present-tense progressives, which would be much more likely to carry experiential/subjective readings, are roughly as common as subordinate-clause, past-tense progressives in the four genres taken together (29% vs. 28% of the total number of progressives, 31% vs. 27% as a mean of the percentages in the individual genres; the latter figures compensate for the fact that the different raw frequencies of progressives in the four genres influence the totals to differing degrees). In addition, the speech of the period may have provided even more frequent opportunities for experiential/subjective progressives to appear. In some respects, we can hypothesize about the spoken language of the past by extrapolating from the language of informal, speech-related written genres (see Rissanen 1986: 98). As Fitzmaurice (this volume) suggests, in the early 1700s "the prose comedy is possibly the register that exhibits features that place it closest to the oral end of the oral-literate pole whereas the essay is arguably the furthest away from the oral end." Compared with drama, essays (along with the other two genres included, fiction and letters) contain a smaller percentage of main-clause, present-tense progressives that would facilitate an experiential/subjective reading. However, we cannot assume that all linguistic characteristics of drama texts are even more pronounced in the spoken language of the period. First, we do not know how well the two specific parameters of tense and clause type would reflect the hypothesized general pattern of speech being an extrapolation of speech-related writing. Secondly, "spoken language" is an umbrella term that would correspond to several different written genres: oral narrative, for instance, may well be characterized by different linguistic

Investigating the expressive progressive


features than, say, a discussion of current events. Nevertheless, the prevalence of main-clause, present-tense progressives in drama implies that the experiential/subjective progressive was relatively common in eighteenthcentury speech, and that it may have spread gradually to written genres. When Fitzmaurice adds another linguistic feature, i.e., identity of subject, to the analysis, there are some changes to the picture. In Fitzmaurice's study, first-person singular subjects are considered more likely than thirdperson singular subjects to host experiential/subjective progressives. If only progressives with first-person singular subjects are considered, drama and letters are now comparable as regards their proportion of main-clause, present-tense progressives (47.5% vs. 45%, as against 13% for fiction and 30% for essays). As recognized by Fitzmaurice, the low frequencies preclude definite conclusions, but it seems that drama and letters provide comparatively numerous opportunities for experiential/subjective progressives to occur. As we shall see below, nineteenth-century English displays similar trends. The final two linguistic features Fitzmaurice considers are lexical: verb class, and lexical support in the form of adverbs, particularly pragmatic strengtheners such as always, constantly, and perpetually. These two features are not quantified in combination with those of clause type and tense, but it is significant that drama displays the highest percentage of progressives of mental verbs (letters, somewhat surprisingly, contain few such progressives), which may invite experiential/subjective readings. Expressive adverbs such as always co-occur with progressives in letters and essays, and may also help to allow a subjective inference. It seems, then, that drama and, to a lesser extent, letters are the genres that provide the most promising contexts for the occurrence of experiential/subjective progressives. Fitzmaurice finds that few progressives in her material meet all five criteria that may help to prompt an experiential/subjective reading. However, a closer examination of some relevant examples indicates that not all criteria need to be present for an experiential/subjective reading to arise: indeed, the experiential/subjective progressive may be moving out of this fairly narrow linguistic frame. This, in turn, raises new questions of identification (Fitzmaurice argues that verb choice is an important criterion). Fitzmaurice's findings in this respect are similar to what I found in a similar investigation of nineteenth-century English (Smitterberg 2002). Based on Fitzmaurice's pioneering work in the field (Wright 1994), I selected four linguistic features that may make an experiential/subjective

180 Erik Smitterberg reading of a progressive more likely: occurrence in a present-tense verb phrase, occurrence in a main clause, occurrence in a stative situation, and having a first or second-person subject. 2 1 then classified the progressives in CONCE (A Corpus of Nineteenth-Century English), a one-million-word corpus of British English covering the period 1800-1899, on each of these four parameters. The database program was instructed to select all progressives that met three or four of the four criteria, in order for me to obtain a rough quantitative picture of the potential occurrence of experiential/subjective progressives. My decision to draw the line at three criteria thus accords well with Fitzmaurice's statements concerning the occurrence of experiential/subjective progressives in her material that do not meet all criteria. Fitzmaurice's material and mine do not cover the same period; moreover, the genres overlap only partially. In addition, as the frequency of the progressive continued to rise in the eighteenth and nineteenth centuries, CONCE contains 2,440 progressives as against 980 in NEET, the corpus used by Fitzmaurice, in spite of the fact that CONCE is c. 350,000 words smaller than NEET. These differences notwithstanding, the results of the investigations match one another well. In my study, drama has by far the highest proportion of progressives that meet three or four of the criteria (30% of all progressive constructions in this genre), followed by letters (20%). Of the other genres in the corpus, 8% of the progressives in parliamentary debates, fiction, and trial proceedings meet the same criteria. Texts belonging to the two written expository genres in the corpus, science and history, contain few progressives that meet three or four criteria: only 4% of the progressives in science were selected, and in history not a single example was found. It thus appears that potentially experiential/subjective readings occur chiefly in genres that provide opportunities for people - real or imaginary to express themselves in a spoken or speech-like manner (it is noteworthy that, in fiction in CONCE, the percentage of progressives selected was considerably higher in dialogue than in narrative passages). The seemingly more pronounced tendency for private letters to host experiential/subjective progressives in CONCE than in NEET may be due to a factor Fitzmaurice (this volume, referring to Biber and Finegan 1989, 1997) herself points to in connection with her statements on the frequency of the progressive construction. As private letters have been found to be closer to the "oral" end of the "oral'V'literate" continuum in the nineteenth than in the eighteenth century, and as the progressive is characteristic of "oral"

Investigating the expressive progressive


texts such as conversation (see, for instance, Biber et al. 1999), it should not surprise us that nineteenth-century familiar letters yield both a higher frequency of the progressive construction overall and a higher proportion of potentially experiential/subjective progressives. In sum, Fitzmaurice's study succeeds because it combines quantitative insights from corpus linguistics with detailed readings of individual examples. It is hoped that further studies of Modern English will continue to increase our knowledge of the development of the progressive as a host for subjectivity in texts.

Notes 1. 2.

I am grateful to Meija Kytö and Terry Walker for their comments on a draft version of this response. There are a few differences between Fitzmaurice's criteria and mine. First, I used Quirk et al.'s (1985) concept of situation types rather than looking at lexical verbs in isolation, disregarding any effect the selection of a progressive over a non-progressive may itself have had on the situation. Secondly, I included both first and second-person subjects, with both singular and plural number, as my hypothesis was that speech-like dialogic turn-taking, characterized by the occurrence of second-person pronouns, would promote the occurrence of experiential/subjective progressives. Thirdly, I chose to investigate one type of adverbial support - modification by adverbials such as always separately, and thus did not include that feature in the analysis.

References Amaud, Rene 1998 The development of the progressive in 19th century English: A quantitative survey. Language Variation and Change 10: 123-152. Biber, Douglas, and Edward Finegan 1989 Drift and the evolution of English style: A history of three genres. Language 65: 487-517. 1997 Diachronic relations among speech-based and written registers in English. In To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, Terttu Nevalainen, and Leena Kahlas-Tarkka (eds.), 253-275. (Memoires de la Societe Neophilo-logique de Helsinki 52.) Helsinki: Societe Neophilologique.

182 Erik Smitterberg Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 Longman Grammar of Spoken and Written English. Harlow: Pearson. Fitzmaurice, Susan M. this vol. The meanings and uses of the progressive construction in an early eighteenth-century English network. Labov, William 2001 Principles of Linguistic Change. Vol. II: Social Factors. (Language in Society 29.) Maiden/Oxford: Blackwell. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London/New York: Longman. Rissanen, Matti 1986 Variation and the study of English historical syntax. In Diversity and Diachrony, David Sankoff (ed.), 97-109. (Current Issues in Linguistic Theory 53.) Amsterdam/Philadelphia: John Benjamins. Ryden, Mats 1997 On the panchronic core meaning of the English progressive. In To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, Terttu Nevalainen, and Leena KahlasTarkka (eds.), 419-429. (Memoires de la Societe Neophilologique de Helsinki 52.) Helsinki: Societe Neophilologique. Smitterberg, Erik 2000 The progressive form and genre variation during the nineteenth century. In Generative Theory and Corpus Studies: A Dialogue from 10 ICEHL, Ricardo Bermudez-Otero, David Denison, Richard M. Hogg, and C. B. McCully (eds.), 283-297. (Topics in English Linguistics 31.) Berlin/New York: Mouton de Gruyter. 2002 The progressive in 19th-century English: A process of integration. Ph.D. Thesis, Department of English, Uppsala University. Wright, Susan 1994 The mystery of the modal progressive. In Studies in Early Modern English, Dieter Kastovsky (ed.), 467-485. (Topics in English Linguistics 13.) Berlin/New York: Mouton de Gruyter.

A brief response Susan M. Fitzmaurice

Erik Smitterberg's response to my essay (Smitterberg, this volume) raises several matters worth immediate comment, and others that are worth highlighting for future attention and discussion. There are three particular matters that I want to attend to here. The first has to do with the framework of analysis used for diagnosing and indeed interpreting experiential/subjective progressives; the second has to do with the role of privileged grammatical contexts in the semanticization of the experiential/subjective construction; and the third concerns the role of quantitative measures in predicting the occurrence of experiential/ subjective progressives in any register. I think that consideration of these matters will advance our conversations about the best and most well motivated research methods in the domain of historical pragmatics using corpus linguistics techniques and database sources. In his response, Smitterberg observes that our respective studies of the experiential/ subjective progressive in eighteenth- and nineteenth-century English (Fitzmaurice this volume; Smitterberg 2002), indicate that the five linguistic contexts that I originally argued (Wright 1994) may accommodate an experiential/subjective reading of a progressive do not have to be present for such a reading to be warranted. Indeed, it may be helpful in refining how we approach the study of the subjectification of constructions such as the progressive if we distinguish more clearly between the use of linguistic features as potential contexts of occurrence, and their use as tests for a particular reading. It is clear that the test of an experiential/subjective progressive is not whether it occurs in a present-tense form of a mental or private verb with a first person subject in a main clause; such a context has been demonstrated not to guarantee that the progressive will be experiential or subjective. If we accept that experiential/subjective progressives are not truth-conditional or primarily aspectual, the best independent indicator that a progressive has a subjective meaning is if the verb colligating in the progressive is figurative. In other words, the only subjective progressives will be those in construction with a verb with metaphorical meaning; often the verb might be understood as having shifted its semantic category, for instance from concrete to abstract, or from activity to mental. The reliance

184 Susan Μ, Fitzmaurice

on such a test clearly presents a challenge for historical linguists who depend upon text corpora for the automated investigation of large datasets because a test for function unfortunately cannot be sought and coded automatically in the same way that formal grammatical features can be in a corpus-based study. Because of its functional nature, the test can be applied only in a procedure that involves reading all tokens and then assigning a literal or denotative sense to each one. The second matter that I think warrants some discussion is somewhat related. This concerns the question why first person contexts might be considered to privilege the instantiation of the subjective progressive, even though we encounter examples with third person subjects, such as those illustrated below in (l)-(5). (1) She is always seeing Apparitions, and hearing Death-Watches {Spectator 7: 1. 34) In pragmatic terms, the expressive function of the subjective progressive would seem more likely to be deployed by a speaker about him- or herself than about a third person. One reason is that as with many non-truth conditional or non-factive expressions, the reference of the activity in the experiential progressive cannot be attributed with any confidence to a person whose attitude or mental state cannot be accessed by the speaker. Instead of being an intersubjective expression - one that can be used to capture the interlocutor's attitude towards a topic - the subjective progressive will always express the speaker's own attitude towards what is being talked about, whether that is another person or an event. However, it is evident that the first person subject does not privilege the attribution of experiential/subjective meaning to the progressive in the eighteenth-century corpus because the progressive occurs with experiential/subjective force with third person subjects too. (An empirical investigation remains to be done to determine the proportion of first person experiential/subjective progressives to third person instances in the NEET essays corpus, for instance.) Indeed, because we expect the experiential/subjective force of the progressive to be more easily accommodated by first person contexts than any other, it seems reasonable to suggest that its occurrence in third person contexts, without adverbial support, might be taken as a positive indicator of the extent to which the experiential/subjective function may have consolidated or semanticized in the progressive (Fitzmaurice 2002). In other words, the increasing occurrence of recognizably experiential/ subjective progressives

A brief response 185 in contexts other than first person, present tense, main clause adverbially supported ones could be construed as indexing the semanticization of subjectivity in constructions like the progressive.1 As illustration of the potential robustness of experiential/subjective progressives with third person subjects, I select some examples from the essays register of the NEET corpus, from which I drew the first person instances in section 7 of Fitzmaurice (this volume). In each example, the speaker deploys the combination of figurative verb meaning and progressive to express his own anger, contempt, amusement, or sense of superiority, respectively towards the subject or person that is the topic of the proposition: (2) Blessed doings is this indeed! Are the wounds of Europe to be thus healed? Will the fatal breaches of nations be stopped this way? When the Christian interest is bleeding to death, is this the haste her doctors make to stop the blood? When the Protestant interest is on fire in the world, is this the dispatch they make to quench and throw water upon it? (Defoe, A Pun upon Peace Makers, April 12, 1712. [ddess015.txt]) (3) Wherefore considering with no small Grief, how many promising Genius's of this Age are wandering (as I may say) in the dark without a Guide, I have undertaken this arduous but necessary Task, to lead them as it were by the hand, and step by step, the gentle downhill way to the Bathos; (Alexander Pope, Martinus Scriblerus: His Treatise of the Art of Sinking in Poetry - Chapter 1, 1714. [pbess012.txt]) (4) With these, and other such Passages, (says my Author) the poor Gentleman grew distracted, and was breaking his Brains Day and Night to understand and unravel their Sense. (Steele, Tatler - No. 178, Saturday, 27 May to Tuesday, 30 May 1710. [rsess007.txt]) (5) There was, I remember, a little French Marquis who was often pleased to rally him unmercifully upon Beef and Pudding, of which our Countryman would dispatch a Pound or two with great Alacrity, while his Antagonist was pidling at a Mushroom, or the Haunch of a Frog. I could perceive the Lady was pleased with what I said, and we parted very good Friends, by vertue of a Maxim I always observe, Never to contradict or reason with a sprightly Female. (Steele, Guardian, No. 34 - Monday, April 2, 1713-1714. [rsess020.txt])

186 Susan Μ. Fitzmaurice In the first example (2), Defoe personifies "the Christian interest" in politics, lamenting that the ongoing peace-conference of English, German and French plenipotentiaries in Utrecht is showing no signs of ending war in Europe. The complaint frame sets up the verb bleed for appropriation as a metaphorical verb, and the result is the emotive force conveyed as righteous indignation. In the second (3), Pope satirizes the relationship of the modern poets with the muses and their desire to reach Parnassus, presenting them as "wandering" museless and clueless, in the dark. In this example, he recruits the physical verb of action (albeit devoid of volition) to capture the intellectual aimlessness that plagues his subjects. In (4), Steele discusses Cervantes' tragicomic character, Don Quixote, with gentle mockery, conveying his own amusement at Cervantes' representation of his character's struggles with books through the metaphorical verb break, with the progressive. Incidentally, the use of a modifying temporal adverb, "Day and Night," adds to the sense of hopelessness with which Don Quixote is presented working to understand the "charming sentences" in the books on knighthood with which he hopes to educate himself in the arts of war. The last example (5), also Steele's, reports his apparently gallant treatment of a group of young women at afternoon tea, despite their woeful lack of any judgment of what constitutes a true gentleman. As an example of his supposed accomplishment as a conversationalist, he cites his stereotypes of the beef-eating, robust English gentleman compared with the picky, mincing French gentleman. The "little French marquis" is merely toying in a desultory fashion with his mushroom or frogs' legs while the Englishman attends heartily to his roast beef. My choice of the examples in (1) through (5) above of third person progressives with experiential/subjective force from the register of essays raises a final question. This concerns the extent to which the frequency with which the aspectual progressive occurs can be used as a way to predict the occurrence of the experiential/subjective progressive. Smitterberg infers from my results that the registers that appear to accommodate the progressive most frequently will also be those in which the experiential/subjective function of the progressive is most likely to appear. Although initially persuasive, this suggestion is, I think, problematic. If the determining test for a non-aspectual, experiential progressive is not a set of formal grammatical contexts but a reading test for metaphorical meaning, then it may be the case that the register that is most salient for the experiential/subjective progressive is the one that will be most conducive, in topical, pragmatic, and emotive terms, to vigorous speaker self-expression

A brief response


on the one hand, or to the vigorous expression by a speaker of emotive, critical or approving evaluation of what he or she is talking about on the other.2 I used the register of essays in the NEET corpus to illustrate the occurrence and function of experiential/subjective first person progressives in section 7 in Fitzmaurice (this volume). Clearly, despite the fact that in my study the essays exhibit the smallest frequencies of the aspectual progressive, this fact about low frequency does not appear to preclude the occurrence of the experiential/subjective progressive in a number of contexts. In light of these observations, and the questions that have emerged, let the conversation continue!

Notes 1.


I am grateful to the editors of the volume for raising the question of whether the experiential/subjective function of the progressive is semanticized or become salient in first person subject contexts or in third person subjects. This is a question that needs proper empirical investigation, but on the basis of the theoretical literature (e.g., Traugott 2000; Traugott and Dasher 2002), it would seem that an increasing occurrence of third person experiential/subjective progressives might well index the spread from privileging first person contexts to third person contexts. Smitterberg suggests that my operationalization of subjective meaning as nonaspectual meaning is too strong because it excludes the possibility that the progressive form of a metaphorical verb could be construed as having both aspectual and subjective functions. This would be a particularly pertinent objection if we were to understand the examples in (2), (3), and (5) for instance, to have aspectual force. However, if it is the case that the progressive form is not obligatory to express aspect in the verb in the language of the period, so the room for interpretation is quite wide. In light of this observation, it might be important to consider the relative weight attributed to various conditioning factors as we advance the empirical investigation of subjectivity in the history of the language.

References Fitzmaurice, Susan M. 2002 The modal meanings of the progressive revisited and revised: a historical pragmatic study. Paper presented at the Second Conference on Studies in the History of the English Language, Seattle, March, 2002.

188 Susan Μ. Fitzmaurice this vol.

The meanings and uses of the progressive construction in an early eighteenth-century English network. Smitterberg, Erik 2002 The progressive in 1 ^-century English: A process of integration. Ph.D. thesis, Department of English, Uppsala. this vol. Investigating the expressive progressive: On Susan M. Fitzmaurice's "The meanings and uses of the progressive construction in an early eighteenth-century English network." Traugott, Elizabeth 2000 From etymology to historical pragmatics. Paper presented at SHEL1, UCLA May 27, 2000. Traugott, Elizabeth, and Richard Dasher 2002 Regularity in Semantic Change. Cambrige: Cambridge University Press. Wright, Susan 1994 The mystery of the modal progressive. In Studies in Early Modern English, Dieter Kastovsky (ed.), 467-485. Berlin/New York: Mouton de Gruyter.

Modal use across registers and time Douglas Biber

1. Introduction Linguists over the past two decades have become increasingly interested in the linguistic devices used to convey personal feelings and attitudes (e.g., Coates 1983; Labov 1984; Chafe 1986; Ochs 1989; Biber and Finegan 1988, 1989; Biber et. al 1999: Ch. 12; Hunston and Thompson 2000; Conrad and Biber 2000; Precht 2001). These investigations of personal expression have been conducted with a variety of complementary methodologies, ranging from detailed descriptions of a single text sample to empirical investigations of general patterns in large computer-based corpora. Although most of the studies listed above have had a synchronic focus, there have also been several related studies taking a diachronic perspective. For the most part, these diachronic studies have focused on a single linguistic system: modality. Some of these studies document the patterns of use in a particular historical period. For example, Fitzmaurice (2002) describes the use of modal verbs in eighteenth-century British English (BrE) letters, and Kytö (1991) describes the use of modal verbs in a range of seventeentheighteenth-century American English (AmE) registers. Other studies document historical patterns of change in the use of modals. Biber, Conrad, and Reppen (1998: Ch. 8) describe a general decline in the use of modals, accompanied by a rise in the use of semi-modals, over the past three centuries. Krug (2000) provides a much more detailed description of the increasing use of semi-modals (especially (have) got to, have to, and want to) in AmE and BrE. Myhill (1995, 1997) focuses on meaning differences among the 'obligation' modals {got to, have to, must, should, ought), documenting historical changes in the functions of these modal verbs in nineteenth- and twentieth-century AmE drama. Finally, Leech (forthcoming) adopts a shorter-term perspective, showing how there have been dramatic changes in the use of modals and semi-modals over the past 30 years. For example, while modal verbs are generally on the decline, selected modals (e.g., can and will) and selected semi-modals (e.g., be going to in AmE) are used with increasing frequencies.

190 Douglas Biber

Both Krug and Leech include some discussion of register differences in their analyses. For example, Krug (2000: 76-83) documents the rise in the use of have to and have got to in both drama and fiction over the past four centuries. In this case, the patterns of change are similar in these two registers, although the increases in use have been somewhat more pronounced in drama. In the present paper, I also adopt a comparative register perspective. The study complements previous studies in that it adopts a relatively long term historical perspective (1650-1990), compares the patterns of use across four major registers (drama, personal letters, newspaper prose, and academic prose), and considers the full set of modal and semi-modal verbs. I employ a corpus-based approach, based on analysis of the ARCHER Corpus, to track historical change in the use of modal and semi-modal verbs across registers, showing how particular registers are innovative in different ways. Section 2 of the paper introduces the modal/semi-modal verbs considered here and surveys their synchronic patterns of use, as a baseline for the historical analyses. Section 3 briefly introduces ARCHER, the corpus used for analysis. Then, in Section 4,1 turn to the historical analysis of these verbs. Finally, in Section 5,1 summarize the findings and discuss implications relative to the motivations underlying the patterns of change.

2. A survey of modals and semi-modals in English: Synchronic patterns of use The synchronic description of modal verbs used here is adapted from the Longman Grammar of Spoken and Written English (LGSWE; Biber, et al. 1999 [especially Section 6.6]). The LGSWE describes the major synchronic patterns of use for modal verbs, employing a corpus-based analysis of texts from four registers: conversation, fiction, newspaper language, and academic prose. The analyses were carried out on the Longman Spoken and Written English (LSWE) Corpus, which contains c. 40 million words of text overall, with c. 4-5 million words from each of these four registers. Modal and semi-modal verbs are used to express many different kinds of meaning. In the present analysis, though, only three major subcategories are distinguished:1

Modal use across registers and time 191

1.) permission/possibility/(ability) modals (can, could, may, might); 2.) obligation/logical necessity modals (must, should)·, semi-modals (have to, got to, ought to, better)', 3.) volition/prediction modals (will, would, shall)·, semi-modal (be going to). Overall, modal verbs are considerably more common in conversation than in the written registers. It might be expected that semi-modal verbs predominate in conversation, while the core modal verbs would be more common in writing. However, Figure 1 shows that this is not the case: rather, both semi-modals and the core modal verbs are more common in conversation than in the written registers. 30000

Figure 1. Frequency of modals and semi-modals in four registers (based on LGSWE-. 486, Fig. 6.9)

Figure 2 shows that this same pattern holds for almost all individual modals and semi-modals. The core modals can, could, will, and would are all considerably more common in conversation than in academic prose, as are the semi-modals have to, (have) got to, (had) better, and be going to.

192 Douglas Biber

Only the core modal verb may, and to a lesser extent must, are notably more common in academic prose than in conversation.


Figure 2. Comparison of individual modal and semi-modal verbs in conversation and academic prose (based on LGSWE: 489, Table 6.6)

These findings reflect the patterns of use in BrE conversation, but as Figure 3 shows, there are some noteworthy differences between AmE and BrE. The obligation/necessity modals must, better, and got to are more common in BrE conversation, while the related semi-modal have to is more common in AmE. Among the volition/prediction modals, the core modal verb will is more common in BrE conversation, while the semi-modal be going to is preferred in AmE. (The dialect differences between AmE and BrE were not considered to be relevant for academic prose, due to the leveling effects of editing, and the fact that authors on both sides of the Atlantic work with publishers on the opposite side.)

Modal use across registers and time 193 6000 - Τ Γ

a BrE Conv • AmE Conv





Modal verb

Figure 3. Comparison of necessity and prediction modal verbs in BrE and AmE conversation (based on LGSWE: 488, Fig. 6.20) The modal verbs are notoriously productive in expressing a range of meanings (see, e.g., Coates 1983; Leech 1987; Palmer 2001), and it turns out that different registers tend to rely on particular meanings. In the present analysis, two major meaning domains are distinguished: personal meanings and logical meanings. For modals like can and may, the personal meanings express permission, while the logical meanings express possibility. For example:2 Personal/Permission


(1) Oh that's so cool. You can say anything you want. (Conversation, AmE) (2) Well, you may have a peek. (Conversation, AmE)

194 Douglas Biber

Logical/possibility meanings (3) See a product can be an object, it can be a service, it can be an activity, it can be a person, place, organization, or even idea. (Spoken lecture, AmE) (4) I don't react to poison ivy, but ... You may be lucky. (Conversation, AmE) In addition, can is used to express meanings related to ability, as in: (5) So this is what I'm explaining is that I'm - I can be really supportive ... (Conversation, AmE) For modals like must and should, the personal meanings express obligation, while the logical meanings describe a situation that necessarily exists. For example:

Personal/obligation meanings (6) Once you are denied at level one, you must go to level two to get it dismissed. (Conversation, AmE) (7) We should make one to show to the kids. (Conversation, AmE)

Logical/necessity meanings (8) He said there was no air, no motion or nothing and they were way up there [...] So he said Lord this is it, I must be dead. (Conversation, AmE) (9) Well let's see, tonight should be much more active I think. (Conversation, AmE) In general, personal meanings are more common in conversation than in the informational written registers like academic prose; the logical meanings are preferred in academic prose. Figure 4 shows that per-

Modal use across registers and time 195 sonal/permission meanings are much less common than logical/possibility meanings or ability meanings for the modals can, could, may, and might. However, when permission meanings are expressed, they are most likely to be found in conversation. For example, the modal can is used with all three meanings in conversation, but the permission meanings of this modal are extremely rare in academic prose. Interestingly, may is rarely used with a permission meaning, even in conversation, while it is extremely common with a possibility meaning in academic prose.

Ξ Personal·· permission • Logicalpossibility 0 Ability










Academic prose

Figure 4. Frequency of personal and logical meanings for permission/possibility modals (based on LGSWE: 491, Fig. 6.12) Figure 5 shows a few new wrinkles on this general pattern. The obligation/necessity modals should and have to follow the expected pattern with the personal/obligation meanings being especially prevalent in conversation. However, it turns out that academic prose also uses the personal meanings for these modals to a greater extent than the logical meanings. Even more surprising are the patterns of use for the modal must: conversation avoids the personal/obligation meaning of this modal, and instead prefers the logical/necessity meaning when must is used. Conversational speakers express their logical conclusions with must, as in:

196 Douglas Biber

(10) Mark must not have told him that I was coming. (Conversation, AmE) (11) I don't have it here. It must be in my desk over at the other building. (Conversation, AmE) In contrast, academic writers use must for persuasive purposes, conveying a sense of personal obligation, as in: (12) To break the bonds of injustice toward women, the church must accept its call to action. (Academic prose) (13) Although his work has been briefly noted by Edgerton (1944) and by Chao (1940), it has not received the attention it deserves among Chinese specialists. I must confess to having failed to check his views until quite recently. (Academic prose)

Ξ Personalobligation

• Logicalnecessity




SHOULD Conversation





Academic prose

Figure 5. Frequency of personal and logical meanings for obligation/necessity modals (based on LGSWE: 494, Fig. 6.13)

Modal use across registers and time 197 In sum, three major patterns emerge from these analyses: (1) There are overall differences across registers in the extent to which modals and semimodals are used at all: more common in conversation than in news or academic prose; (2) Particular modals are used to differing extents: can, will, and would are especially common in conversation, while may is especially common in academic prose; (3) There are striking differences in the preferred meanings of individual modals across registers. Building on this brief description of modals and semi-modals in present-day English, I turn now to the historical patterns.

3. Overview of the ARCHER corpus The historical study of modal verbs here is based on analysis of the ARCHER corpus (Biber, Finegan, and Atkinson 1994). ARCHER - A Representative Corpus of Historical English Registers - was designed for a specific research agenda: to analyze historical change in the range of written and speech-based registers of English from 1650 to the present. The general design goal for the corpus has thus been to represent a wide range of register variation, sampled systematically across texts from the last three and a half centuries. The overall structure of the corpus comprises ten major register categories, sampled in 50 year periods from 1650 to the present, as summarized in Table 1. Altogether, the complete corpus includes 1,037 texts and c. 1.7 million words. Among the written registers, the corpus includes personal styles of communication (journals/diaries and personal letters), prose fiction, popular exposition represented by news reportage, and specialist expository registers, represented by legal opinions, medical prose, and scientific prose. The corpus similarly includes several different kinds of speech-based registers: dialogue in drama and dialogue in fiction as reflections of casual face-to-face conversation, and sermons as a reflection of planned monologue styles. Registers are represented by 10 texts per 50-year period, in most cases chosen using random selection techniques (with available bibliographies serving as sampling frames). American English registers are sampled for only one 50-year period per century. Biber, Finegan, and Atkinson (1994) and Biber, Conrad, and Reppen (1998: Methodology Box 2) provide more details about the design, sampling and compilation of ARCHER.

198 Douglas Biber Table /. Overview of ARCHER (A Representative Corpus of Historical English Registers) Total size: Time-span covered: Dialects covered: Registers:

1,037 texts; c. 1.7 million words. 1650-1990, divided into 50 year periods British (all periods) and American (1 period per century) 7 Written Categories: journals/diaries, personal letters, fiction prose, news reportage, legal opinions, medical prose, scientific prose 3 Speech-based Categories: drama, fiction dialogue, sermons

Target Sampling:

10 texts, at least 2,000 words, per genre (and dialect) in each period.

A full sampling for a register includes 100 texts: 1650-1699, British: 10 texts 1700-1749, British: 10 texts 1750-1799, British: 10 texts 1750-1799, American: 10 texts 1800-1849, British: 10 texts 1850-1899, British: 10 texts 1850-1899, American: 10 texts 1900-1949, British: 10 texts 1950-1990, British: 10 texts 1950-1990, American: 10 texts

Four registers from ARCHER were chosen for the present analysis: drama, personal letters, newspaper reportage, and medical prose. Table 2 summarizes the composition of this sub-corpus by register category. These four registers cover much of the range of register variation available in ARCHER, and they allow easy comparisons to the findings from the LGSWE (summarized in Section 2 above): drama provides an indication of the characteristics of face-to-face conversation in earlier periods, while newspaper reportage and medical prose correspond closely to news and academic prose in the LSWE Corpus. Personal letters were included because of their importance in previous studies of stance (e.g., Biber and Finegan 1989; Besnier 1989; Fitzmaurice 2002, 2003). Similar to conversation, letter writers have highly involved purposes. In fact, these previous studies have shown that personal letters can be even more marked for stance than conversation, apparently because they are less facethreatening (because of the physical separation of participants).

Modal use across registers and time 199 Table 2. Breakdown of registers used for the present study Register drama (only 5 texts from 18 Λ century American) letters (more than 10 texts per period; most texts shorter than 1,000 words) newspaper reportage medical prose (no 18th century American)

Number of texts 95 275

100 90

In general, the analyses of modals focus on BrE texts, because they have a more complete representation in ARCHER (see Table 1 above). However, selected patterns of change are reported for AmE when they are especially noteworthy.

4. Historical change in the use of modals and semi-modals 4.1. General distribution of modals and semi-modals Figure 6 shows that modals have undergone a marked decline in all registers over the last 50-100 years. Specific modal verbs affected by this decline include may, must, should, will, and shall (see discussion below). These patterns agree generally with the even more recent patterns of change documented by Leech (forthcoming). In contrast to the decline for modal verbs, Figures 7 and 8 show that there has been an increase in the use of semi-modal verbs. This increase is stronger in AmE than BrE, but it has been restricted primarily to drama and letters in both dialects. Thus, the decline in modal use for newspaper and academic prose shown in Figure 6 is not offset by a corresponding increase in semi-modal use in those registers, suggesting that the two trends (decreasing modal use vs. increasing semi-modal use) are at least partially independent.

200 Douglas Biber In addition, it is important to note that semi-modals are still considerably less common than core modal verbs. (Note that Figures 7 and 8 have a 0 - 5 scale, while Figure 6 has a 0 - 3 0 scale.)

- Drama Letters — Newspapers —Medical prose


Figure 6. Historical change in the frequency of modal verbs in BrE

For example, semi-modals occur only c. 3 times per 1,000 words in BrE drama from the most recent period (Figure 7), while modal verbs are used about 22 times per 1,000 words in the same register and period (Figure 6). These patterns of use for modal and semi-modal verbs should be considered relative to the full range of linguistic devices used to express stance in English, including stance adverbials (e.g., possibly) and stance complement clauses (e.g., want to-clause, hope ίΑαί-clause); I return to this point in Section 5 below.

Modal use across registers and time


5 4.5 4 3.5 3 • - Drama • Letters


— Newspapers


— M e d i c a l prose






Figure 7. Historical change in the frequency of semi-modal verbs in BrE

Έ ο Ϊ

- Drama •m - Lettens -A—Newspapers


8 2* =3 ?

— · — Medical prose


u- 1.5





Figure 8. Historical change in the frequency of semi-modal verbs in AmE

202 Douglas Biber

Text sample 1 illustrates the heavy reliance on modal verbs in eighteenth-century personal letters, while Sample 2 illustrates the reliance on semi-modal verbs in twentieth-century drama: Text Sample 1: 18th-century personal letter (Addison to Swift, 1710.) [modal verbs are marked in bold] I have run so much in debt with you that I don't know how to excuse my self and therefore shall throw my self wholly upon your good nature and promise if you will pardon what is passed to be more punctual with you for the future. [...] I shall not trouble you with any occurrences here because I hope to have the pleasure of talking over all affaires with you very suddenly. We hope to be at HolyHead by the 30th Instant. Lady Wharton stays in England. I suppose you know that I have obeyed yours and the Bishop of Clogher's commands in relation to Mr Smith, for I desired Mr Dawson to acquaint you with it. I must beg my most Humble Duty to the Bishop of Clogher. I heartily long to eat a dish of Bacon and Beans in the best company in the world. Mr Steele and I often drink your Health. I am forced to give myself Airs of a punctual Correspondce with you in discourse with your friends at St James's Coffee-house, who are always asking me Questions about you when they have a mind to pay their Court to me, if I may use so magnificent a Phrase. [...] Text Sample 2: 20,h-century drama (Parker, Dorothy and d'Usseau, Arnaud, 1954. Ladies of the corridor.) [semi-modal verbs are marked in bold] {=m CASEY} Hey, listen to that rain! {=m HARRY} Yeah, I hate to go out in it. The one night in the week I pick to go home, and it's got to come down in buckets. {=m CASEY} A big week, huh? {=m HARRY} One of the biggest. The town's full of it. The way the dames are handing it to you, all you have to do is reach out and take it. Well, if they want to offer it to you on a platter with parsley around it, what are you going to do?

Modal use across registers and time 203 The appendix summarizes the patterns of change for individual modal and semi-modal verbs. These results must be interpreted with caution, because the samples for the individual cells (combinations of register/ dialect/period) are quite small. However, some general patterns are apparent here. For example, there seems to have been a general increase in the use of the modal can and a decrease in may, especially in personal letters. Interestingly, may appears to be decreasing in use over the past 50 years in both newspaper prose and medical prose. The modal would has increased in use in drama and newspaper prose, but it has decreased in use over time in medical prose. The modal shall has decreased in use in all registers, most notably in AmE. Among the semi-modals, have to shows the most dramatic increase in use, especially in drama and personal letters. These historical shifts in frequency of use have been accompanied by meaning shifts as well (see also Myhill 1995, 1997). Two case studies are briefly considered in the following sections, focusing on the patterns of change in personal letters: can vs. may, and must vs. have to. 4.2. Historical change in the meanings of individual modals and semimodals: can vs. may, and must vs. have to The case studies in this section focus on the meanings of modals and semimodals in personal letters. Two case studies are presented here: can vs. may, and must vs. have to. These forms have undergone notable shifts in their distributions, described in the last section (e.g., can has increased in frequency while may has decreased). Those changes can be better understood by consideration of the typical functions of each verb. Although the members of each pair are potentially synonymous, a closer consideration of their typical meanings shows that they have undergone functional as well as distributional shifts over the last three centuries. This section focuses primarily on the typical meanings of these verbs in personal letters, but I also provide a brief discussion of must vs. have to in newspaper prose. 4.2.1. Can vi. may In the seventeenth- and eighteenth-century letters from ARCHER, the modal can is used mostly to express meanings of personal ability, as in:

204 Douglas Biber

(14) This is all I can say to excuse my inclosure of such papers ... (Temple 1667) (15) I am here entertained wth the prettiest variety of snow-prospects that you can Imagine ... (Addison 1702) (16) ... and I can easily believe, what I am told, that the decorations and habits cost the Emperor thirty thousand pounds sterling. (Lady Montagu 1716) (17) All I can offer in return for the favour which I ask is many, many thanks ... (Boswell 1764) Although the frequency of can has increased over time, the personal ability meaning continues to predominate: (18) He can make himself a hero if he'll do that. (Truman 1950) (19) ... but I am hurrying so Mary and I can get off on a walk (Hemingway 1961) In addition, the typical uses of can have been extended to include a kind of impersonal ability, as in: (20) ... the consequences can last a week. I'm sure somebody meant to be kind but kindness can kill. (MacLeish 1980) (Possibility meanings are also attested for can in ARCHER letters, but they are comparatively rare.) The patterns of use for may are quite different from those for can: may has decreased in frequency over time in letters, and it has undergone a shift in typical meanings. In seventeenth- and eighteenth-century letters, may was used commonly with possibility meanings: (21) ... but I shall not be forward to make tryall of any medison upon the report of a confident mountebank nor though others may have found releife by i t . . . (Conway 1664)

Modal use across registers and time 205 (22) ... if you leave off writing French, you may perhaps neglect that grammatical purity, and accurate orthography, which, in other languages, you excel in ... (Chesterfield 1747) (23) You may remember, Madam, that I expressed my affection for that dish in the strongest manner ... (Boswell 1764) (24) For my own part, I am well in mind and body, busy with my books, (which may perhaps produce something next year, either to tire or amuse the world,)... (Gibbon 1774) While can normally expressed meanings of ability (rather than permission), the modal may was often ambiguous with ability or permission meanings: (25) I begg of you to let Mrs. Warren send me word when I may attend you. (Steele 1707) (26) You may now make merry and kill your fatted calf... (Rush 1765) (27) If you and Kitty will keep the house, I think I shall like it best. Kitty may carry on the trade for herself, keeping her own stock apart, and laying aside any money that she receives for any of the goods which her dear good Mistress has left behind her. I do not see, if this scheme be followed, any need of appraising the books. My Mother's debts, dear Mother! I suppose I may pay with little difficulty, and the little trade, may go silently forward. I fancy Kitty can do nothing better, and I shall not want to put her out of a house where she has lived so long and with so much virtue. I am very sorry that she is ill, and earnestly hope that she will soon recover; let her know that I have the highest value for her, and would do any thing for her advantage. Let her think of this proposal. I do not see any likelier method by which she may pass the remaining part of her life in quietness and competence. (Samuel Johnson 1759)

206 Douglas Biber In contrast, may has become rare in twentieth-century letters, and it is almost always used for possibility meanings: (28) But I want you to think about the fact that my appointment may reflect upon you and your administration. (Truman 1950) (29) You know, perhaps when things straighten out here, I may come over - with the two kids. (Miller 1951) (30) I may be getting proofs of my new Cupid & Psyche story in June. (Lewis 1956) (31) Of course, as you hint, I may not get it - and maybe Norman Mailer will. (Miller 1978) Thus, the decrease in frequency for may seems to be associated with a restriction in meaning (i.e., the dramatic decline in ability/permission meanings noted in Section 2 above). May rather than can has remained the preferred choice to express possibility meanings in letters, but can has become the normal choice to express ability meanings. Neither modal is commonly used in letters to express permission meanings. (Of course, the modals could and might need to be considered to provide a complete picture of these shifts.)


Must vs. have to

Changes in the meanings of the modal must and the semi-modal have to have also interacted in interesting ways. In terms of frequency, must has remained relatively constant across these periods in personal letters. However, must has expanded in its range of typical meanings. In seventeenth- and eighteenth-century letters, must almost always expressed meanings of personal obligation, as in: (32) ... from hence I must goe into Northamptonshire to my Lady R. (Osborne 1654) (33) We must admire before we love. (Pope 1705)

Modal use across registers and time


(34) ... for every heart must lean to somebody, and I have nobody but you (Johnson 1759) (35) ... and after a few turns of this busy stage of action, we must depart hence and be no more. (Rush 1765) (36) I must now beg of you to try to amuse Mrs. Jackson and prevent her from fretting. (Jackson 1797) In contrast, must in the twentieth century has extended its typical meanings to include both logical necessity and personal obligation meanings: Logical necessity meanings (37) You must have had a grand time at old Lady Astor's party. (Truman 1951) (38) ... you'll have a grand time and it must be fun planning for it. (Rhys 1952) (39) I must have seemed decrepit! (Miller 1960) (40) I know his death must have hurt you, too ... (MacLeish 1965) Personal obligation meanings (41) Anyhow, you must get an evening dress. (Nicolson 1950) (42) As soon as I finish the Rhodes book which I hope to do by the end of the year I must face the idea seriously. (Durrell 1951) (43) Now I must see Max - first a long journey - and I must pack a bit. (Rhys 1963) Although the semi-modal have to has had marginal occurrence in English for over 1,000 years (see Krug 2000: 74), it has only become commonly attested in the past few centuries. Have to is first attested in the late eighteenth century in ARCHER letters, and it has shown a marked increase in frequency across the following periods. Across all periods in

208 Douglas Biber

ARCHER letters, have to has been used mostly to express meanings of personal obligation: In late 18th-century letters (44) I have to acknowledge the kind care of both my Friends in the conveyance of Letters. (Abagail Adams 1775) (45) ... we find but little Fruit, except Huckel berries, and live in our Camp, as retired as we used to do on Lake Erie - Labouring Hands in this Country can scarcely be had at any rate; my estimate was twenty; but I have to wade slowly thro' with six ... (Ellicott 1791) In 20,h-century letters (46) Tomorrow I have to break the bad news to Louis Johnson. (Truman 1950) (47) ... the elaborate joint holidays he had planned for us in the summer will probably have to be cancelled. (Lewis 1956) (48) This poor bugger will have to wait ten years now with his book gathering dust! (Durrell 1979) In sum, have to has been encroaching on the semantic domain of must in personal letters, being used more frequently to express meanings of personal obligation. At the same time, must has increased in its use with logical necessity meanings, and as a result, this core modal has remained relatively constant in frequency of use across periods. The patterns of change for must in newspaper prose are somewhat different. In seventeenth- and eighteenth-century newspapers, must was generally rare, but it was already coming to be used for both personal obligation and logical necessity meanings: (Personal) obligation meanings in 18 th-century news (49) ... we again discovered the 15 Turkish Ships, who had only made a show of returning to Barbary, and had divided themselves into two Squadrons at the mouth of the Gulph by which we must of necessity pass ... (1697)

Modal use across registers and time 209 (50) As the House soon must be cleared, the said Goods will be sold under prime Cost. (1762) (51) The countenance which this procession received in every quarter of the town, while it marks the spirit and wishes of its citizens, must serve as an additional motive ... (1788) Logical necessity meanings in 18th-century


(52) ... there is no doubt but he will prevent this Scheme, by forcing the Enemy to an Action, the Success of which, it is generally believed, must prove favourable to him. (1744) (53) It is to be observed, that if the Parma had rose five Inches higher, it would have over-topp'd all the Dykes, and then the Damage must have been immense. (1752) By the twentieth century, must had increased in frequency in newspaper prose, but it came to be restricted primarily to obligation meanings: (54) ... we must also recognise that in South Vietnam there are those who are prepared to resist the violent extension of Communism in their country... (1967) (55) The union led by Mr Moss Evans must take the main responsibility for wrecking Labour's plans to keep inflation down to single figures. (1979) (56) Many countries sent an instant reply telling the Americans that they must ensure that their own airlines became responsible ... (1989) In contrast, the semi-modal have to is extremely rare in newspaper prose. It is interesting to compare these patterns of use in ARCHER to the patterns described by Leech (forthcoming) in a comparison of the Brown and LOB corpora (1961) to the Frown and FLOB corpora (c. 1991). Leech's study documents a dramatic decrease in the use of must over the 30 year period from 1961 to 1991, in both BrE and AmE (considering all written registers together). While similar patterns are found here for some registers in ARCHER (note especially the dramtic decrease in BrE medical

210 Douglas Biber prose and AmE drama), other cases are more difficult to interpret. In part, this difference might be due to sampling differences in the two studies: ARCHER combines texts from 1950-1990 into a single period, which would obscure recent changes in use over the past 30 years.

5. Conclusion The present study has analyzed modal and semi-modal verbs across historical periods, comparing the patterns of use across registers (and, to a lesser extent, dialects). The findings are complex, with few global generalizations. For example, modals have been decreasing in frequency in all registers, but semi-modals have shown noteworthy increases only in drama and personal letters. Similarly, the patterns of change for individual verbs often differ across registers. The modal can is one of the few verbs to have increased in use across registers. Would has also increased in frequency in drama, letters, and news, but medical prose shows a systematic decrease in use for this verb. The modals may and must were used with increasing frequencies in medical prose until the last 50 years, when they both showed sharp decreases in frequency. In contrast, must in personal letters has remained relatively constant in frequency across periods, and it has actually increased in newspaper prose. The present paper considers only one aspect of this larger shift in the use of stance features: modal and semi-modal verbs. The patterns of change documented here should be considered relative to the full range of linguistic devices used to express stance in English. For example, forms like need to and want to are sometimes regarded as semi-modals and could be compared to the set of verbs considered in the present study. However, there is actually a large set of verbs and adjectives that control a to-clause to express modal-like meanings. For example, verbs such as (would) like to, wish to, and hope to express a kind of intention or weak volition, similar to the modals might or would. Adjectives like possible to or likely to express possibility (similar to the modals can, could, might, and may), while competent to, fit to, and inclined to express specific meanings related to ability or intention. In addition, /Aaf-clauses (e.g., appear that, fact that) and stance adverbs (e.g., apparently, certainly) are also commonly used to express similar kinds of stance meanings related to modality.

Modal use across registers and time 211

Biber (forthcoming) examines historical patterns of change for this wider set of grammatical devices used to express stance. Of these, only modal verbs have undergone a decrease in use. In contrast, semi-modals, stance adverbials, and stance complement clause constructions have all increased in use across historical periods. The patterns documented in that study argue against a simple grammatical reorganization (with modal verbs being replaced by semi-modals). Rather, those findings suggest that stance meanings are being expressed more commonly overall, indicating a general shift in cultural norms: speakers and writers are apparently more willing to express stance in recent periods than in earlier historical periods. These patterns require much more detailed study, considering historical patterns in the typical meanings of individual modal verbs in addition to changes in frequency.

212 Douglas Biber

Appendix: Frequency of individual modal/semi-modal verbs across registers and periods (rates per 1,000 words of text) Table A.I. Frequencies for the permission/possibility modals: can, could, may, might



#of texts






1700-99 1800-99 1950-90

5 10 10

1.7 3.5 5.0

2.3 2.5 1.8

1.0 1.2 0.9

0.6 0.5 0.7


1650-99 1700-99 1800-99 1900-49 1950-90

10 18 20 11 11

4.1 3.1 3.3 5.1 3.7

0.4 1.4 2.3 3.4 1.4

1.9 2.0 2.0 1.5 1.5

0.5 0.4 0.9 1.1 1.1


1700-99 1800-99 1950-90

29 28 31

2.4 3.0 5.1

1.9 2.3 2.8

3.3 1.0 1.4

0.6 1.1 0.3


1650-99 1700-99 1800-99 1900-49 1950-90

24 64 52 28 28

2.5 3.0 2.0 3.7 4.4

1.8 1.8 1.2 1.7 1.8

3.0 2.7 2.0 1.3 0.9

0.8 1.0 0.7 0.5 0.7


1700-99 1800-99 1950-90

10 10 10

0.5 0.4 0.7

0.3 0.6 0.5

1.9 1.9 0.8

0.2 0.2 0.6

BrE BrE BrE BrE BrE Medical prose AmE AmE

1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.7 0.3 0.7 0.4 1.0

0.7 0.6 0.7 0.9 1.8

1.1 3.2 1.6 2.7 1.1

0.8 0.2 0.1 0.6 0.4

1800-99 1950-90

10 10

0.8 1.0

0.9 0.6

2.1 2.3

0.6 0.2


1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.7 1.0 0.6 1.3 1.0

2.1 1.6 1.5 0.7 0.5

1.2 0.8 2.5 4.4 2.8

0.8 1.1 0.7 0.3 0.3



Modal use across registers and time Table A.2.

Frequencies for the obligation/logical necessity modals (must, and the volition/prediction modals (will, would, shall)





#of texts







1700-99 1800-99 1950-90

5 10 10

1.3 2.8 0.9

2.3 1.8 1.2

2.7 6.5 4.2

3.1 3.4 3.4

1.0 2.5 0.6


1650-99 1700-99 1800-99 1900-49 1950-90

10 18 20 11 11

3.1 2.6 3.2 2.6 3.2

1.0 1.5 2.2 1.0 1.3

4.9 5.6 4.7 4.4 3.1

1.4 2.1 2.6 2.6 5.2

1.9 3.6 3.1 1.8 1.0


1700-99 1800-99 1950-90

29 28 31

1.1 1.6 1.3

2.2 2.2 0.8

9.0 7.0 4.6

2.8 2.5 4.6

2.7 1.9 0.4


1650-99 1700-99 1800-99 1900-49 1950-90

24 64 52 28 28

1.2 1.3 1.4 2.2 1.2

1.9 2.1 2.0 1.8 1.1

5.4 5.3 5.2 4.4 3.6

1.9 3.1 2.2 2.4 2.4

3.5 2.5 2.4 2.1 1.7


1700-99 1800-99 1950-90

10 10 10

0.5 0.3 0.6

0.6 1.2 0.6

3.4 3.5 4.1

0.6 1.3 3.9

0.9 0.7 0.0

1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.5 0.6 0.8 1.5 1.3

0.2 0.8 0.9 1.7 1.1

3.6 2.9 3.0 6.0 3.8

0.8 0.7 1.5 2.4 2.8

0.8 0.3 0.6 0.0 0.1

1800-99 1950-90

10 10

0.2 0.1

1.3 0.1

1.2 0.1

1.4 0.2

0.0 0.0

1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.6 0.6 0.5 1.5 0.1

0.1 0.6 1.0 0.9 0.9

0.9 1.2 0.7 1.2 0.2

2.5 1.5 1.6 1.0 0.6

0.4 0.4 0.3 0.1 0.0




BrE BrE BrE BrE BrE Medical prose AmE AmE BrE BrE BrE BrE BrE



Table Α.3.


Frequencies for semi-modals: have to, got to, be going to



U of texts





1700-99 1800-99 1950-90

5 10 10

0.2 0.9 1.2

0.0 0.3 0.4

0.2 1.0 2.4


1650-99 1700-99 1800-99 1900-49 1950-90

10 18 20 11 11

0.0 0.2 0.6 1.0 1.5

0.0 0.2 0.0 0.5 0.4

0.0 0.1 0.0 0.8 0.9


1700-99 1800-99 1950-90

29 28 31

0.3 0.7 1.3

0.0 0.0 0.1

0.1 0.0 0.6


1650-99 1700-99 1800-99 1900-49 1950-90

24 64 52 28 28

0.0 0.2 0.8 1.5 2.1

0.0 0.0 0.0 0.1 0.1

0.0 0.1 0.4 0.4 0.3


1700-99 1800-99 1950-90

10 10 10

0.3 0.5 0.3

0.0 0.0 0.0

0.1 0.2 0.0


1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.0 0.1 0.2 0.1 0.4

0.0 0.0 0.0 0.0 0.0

0.0 0.1 0.0 0.1 0.3


1800-99 1950-90

10 10

0.1 0.1

0.0 0.0

0.1 0.0


1650-99 1700-99 1800-99 1900-49 1950-90

10 20 20 10 10

0.0 0.3 0.2 0.2 0.2

0.0 0.0 0.0 0.0 0.0

0.0 0.0 0.0 0.1 0.0




Medical prose

Notes 1. The verbs need and dare have marginal modal status; for example, they can pattern like other modal verbs in interrogative and negative clauses. However, this

Modal use across registers and time 215 use is rare and restricted to BrE; see Biber et al. (1999: 163-164, 217-218). These verbs are not included in the present analysis. 2. All synchronic examples in the paper are taken from the Longman Corpus Network.

References Besnier, Niko 1989 Literacy and feelings: The encoding of affect in Nukulaelae letters. Text 9: 69-91. Biber, Douglas forth Historical patterns for the grammatical marking of stance: A crossregister comparison. Journal of Historical Pragmatics. Biber, Douglas, Susan Conrad, and Randi Reppen 1998 Corpus Linguistics: Investigating Language Structure and Use. Cambridge: Cambridge University Press. Biber, Douglas, and Edward Finegan 1988 Adverbial stance types in English. Discourse Processes 11: 1-34. 1989 Styles of stance in English: Lexical and grammatical marking of evidentiality and affect. Text 9: 93-124. Biber, Douglas, Edward Finegan, and Dwight Atkinson 1994 ARCHER and its challenges: Compiling and exploring a representative corpus of historical English registers. In Creating and Using English Language Corpora, Udo Fries, Gunnel Tottie and Peter Schneider (eds.), 1-14. Amsterdam: Rodopi. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad, and Edward Finegan 1999 The Longman Grammar of Spoken and Written English. London: Longman. Chafe, Wallace L 1986 Evidentiality in English conversation and academic writing. In Evidentiality: The Linguistic Coding of Epistemology, Wallace L. Chafe and Johanna Nichols (eds.), 261-272. Norwood, N.J.: Ablex. Coates, Jennifer 1983 The Semantics of the Modal Auxiliaries. London: Croom Helm. Conrad, Susan, and Douglas Biber 2000 Adverbial marking of stance in speech and writing. In Evaluation in Text, Susan Hunston and Geoffrey Thompson, (eds.), 56-73. Oxford: Oxford University Press. Fitzmaurice, Susan 2002 Politeness and modal meaning in the construction of humiliative discourse in an early eighteenth-century network of patron-client relationships. English Language and Linguistics 6: 239-266.

216 Douglas Biber 2003

The grammar of stance in early eighteenth-century English epistolary language. In Corpus Analysis: Language Structure and Language Use, Charles Meyer and Pepi Leistyna (eds.), 107-131. Amsterdam: Rodopi. Hunston, Susan, and Geoffrey Thompson (eds.) 2000 Evaluation in Text. Oxford: Oxford University Press. Krug, Manfred G. 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Kytö, Meija 1991 Variation and Diachrony, with Early American English in Focus. Frankfurt am Main: Peter Lang. Labov, William 1984 Intensity. In Meaning, Form, and Use in Context: Linguistic Applications, Deborah Schiffrin (ed.), 43-70. Washington, D.C.: Georgetown University Press. Leech, Geoffrey 1987 Meaning and the English Verb. London: Longman. forth. The English modal auxiliaries 1961-1991: Modality on the move. In Modality in Contemporary English, Roberta Facchinetti, Manfred Krug, and Frank Palmer (eds.). Berlin/New York: Mouton de Gruyter. Myhill, John 1995 Change and continuity in the functions of the American English modals. Linguistics 33: 157-211. 1997 Should and ought: The rise of individually oriented modality in American English. English Language and Linguistics 1: 3-23. Ochs, Elinor (ed.). 1989 The Pragmatics of Affect. Special issue of Text, Vol 9. Palmer, F. R. 2001 Mood and Modality. Cambridge: Cambridge University Press. Precht, Kristen 2001 Patterns of Stance in English. Ph.D. diss., Department of English, Northern Arizona University.

The need for good texts: The case of Henry Machyn's Day Book, 1550-1563 Richard W. Bailey

"Good texts make good histories," as the neighbor in Robert Frost's "Mending Wall" might have said on an occasion like this one. It is a commonplace that the Oxford English Dictionary could not have been made as good as it was without Edward Arber's little reprint series or F. J. Fumivall's many "societies" devoted to creating editions of both published and unpublished work. Most of these are far from modern standards of editing, but without them the record of English would be poorer. Similarly the great difficulty the Middle English Dictionary had in getting going was not so much the Great Depression or World War Π (though these were important factors) but the difficulty in dating, ascertaining, and even locating Middle English texts.1 We cannot study the history of the English language without well-understood ideas about the reliability of the material that makes up the data for our history writing. Such a statement should be both obvious and true. Alas! Recently I reviewed a collection of essays devoted to the "development of Standard English," and one of the arguments found in this work of several authors is that "standardization" took place earlier than is generally believed. In making this argument, one writer quoted a text he alleged to have been composed in 1586 and first published in 1674. (In fact, it was written after 1605, as one can see by internal evidence, and published in 1614, as one can see by looking at the book; see Wright 2000: 43). The reason for these difficulties is partly because the historian used the edition of 1870, a choice that seems indefensible given that there is an excellent edition, published in 1984, based on the manuscript. While perhaps adequate to the standards of 1870, the edition chosen is useless for linguistic history because so much of it is "modernized," thus giving the misleading impression that "modern" English appeared much sooner than it did. Thinking about this essay drew my attention back to the well-known essay by Norman Davis: "The Text of Margaret Paston's Letters" (1949). In it Davis sketched the history of the letters' publication and the credence

218 Richard W. Bailey

given to her "unconsciously phonetic spellings." "Some far-reaching theories concerning the date and the course of the 'great vowel shift' [Davis wrote] were founded on these and similar spellings elsewhere, especially by Zachrisson, whose conclusions were largely accepted by Wyld, Jordan, and others" (13). Severe criticisms of Zachrisson's theories laid out in his monograph of 1913 led him to recant, and he was forced to admit that "many of the supposedly significant 'occasional spellings' were in fact no more than editorial misreading" (14). This act of contrition Zachrisson performed in 1927. The problem here was that the mistakes had by this time been lodged in the historical narratives, and, though purged of some errors, H. C. Wyld's third edition of A History of Modern Colloquial English, as Davis pointed out, "still contained] serious inaccuracies" (14). He then detailed some of the problems to be found in the unreliable editions of Margaret Paston's letters, and it was hardly a surprise that he would soon embark on his authoritative edition of the entire Paston correspondence. All that follows herefrom is the story of Henry Machyn's Day Book and its misuse and use at the hands of the linguistic historians.2 In his forthcoming entry on Machyn for the new Dictionary of National Biography, Ian Mortimer gives all that is known with certainty. The "chronicler," as Mortimer accurately styles him, was born the Wednesday after Whitson, probably in 1497, though he is inconsistent about the year. Admitted to the company of Merchant Taylors by 1530, he seems to have prospered in providing regalia for funerals: pennants, palls, gowns, and other costumes. In 1550, he began to keep a chronicle in which he recorded the funerals of prominent persons, but he soon noted even more sensational happenings. In 1563, in midsummer, he recorded an outbreak of the plague which may have carried him off. He was buried in the parish where he had served as clerk: Holy Trinity the Less, a church destroyed, not to be replaced, in the Great Fire of 1666. Mortimer believes, reasonably enough, that Machyn gained a taste for chronicle-making through his duties as parish clerk. One hundred and sixty-two octavo leaves of the Day Book found their way into the collection of that wonderful antiquary Robert Cotton, and in due course the text was discovered by the historian John Strype who printed long sections of it in his historical works. Strype (1643-1737) did not regard it as useful to give a literal transcript, and in his books allowed the publisher to employ the conventions of the day: abundant punctuation, capital letters for major nouns, italic type for emphasis, and so on. When

Henry Machyn 's Day Book 219 these works were reissued in the early nineteenth century, the publisher simply substituted new preferences for old and thus moved the text farther from the source. Thanks to Strype, however, the Day Book survived, and Strype's anthologies of historical documents including much of it made it "one of the most important sources for the history of mid-sixteenth-century English politics and religious reform" (Mortimer 2002: 981). Text 1: Eighteenth-century Strype. On the 14th of September Three were set on the Pillory for playing with false Dice; and for deceiving honest Men by that Means. The 17th of the same, came forth a Proclamation, That all Vagabonds and Loiterers, as well English, as all manner of Strangers, having no Masters, should avoid the City, and the Suburbs forthwith, upon great Pains enjoyned by Law: And that none that kept publick Houses should give Entertainments to any serving Men, unless they brought Testimonials under their Masters Hands. On the 20th, were two Men drawn on Hurdles unto Tyburn, to Execution, for Coining of naughty Money, and deceiving the Queen's Subjects therewith. (Strype 1733, vol. 3: 199). Text 2: Nineteenth-century Strype. On the 14th of September three were set on the pillory for playing with false dice, and for deceiving honest men by that means. The 17th of the same came forth a proclamation, that all vagabonds and loiterers, as well English as all manner of strangers, having no masters, should avoid the city and the suburbs forthwith, upon great pains enjoined by law and that none that kept public houses should give entertainment to any serving men, unless they brought testimonials under their master's hands. On the 20th were two men drawn on hurdles unto Tyburn, to execution, for coining of naughty money, and deceiving the Queen's subjects therewith. (Stripe 1832, vol. 3.1: 316)

220 Richard W. Bailey

Α major event in the survival of the chronicle was the fire at Ashburnham House in 1731 and the near destruction of Robert Cotton's library. Only with the heroic efforts of the firefighters did the Beowulf manuscript survive, and, in the same bookcase, Machyn's Day Book was severely damaged and substantial portions of the top and edges of the paper reduced to ashes. The pages were gathered and stored in a box where they remained for nearly a century. In 1829, Frederick Madden at the British Library examined the fragments and arranged them in order by date. Each page was then carefully mounted in a paper frame cut just smaller than the outline of the burned pages and the recto pasted to it. The result is that a small portion of the verso of each page is obscured by the frame, but the technique helped to preserve the manuscript from further damage. In 1848 was published the sole printed edition of the whole Day Book, edited for the Camden Society by the John Gough Nichols, and called by him and others since "Machyn's Diary.'''' Nichols declared that text missing because of the fire had been restored by reference to Strype or by conjecture, and these insertions, he asserted, were clearly marked.3 He gives no real explanation of his editorial policy, but his text does not resemble the quotations found in either edition of Strype and is obviously more archaic. To the innocent eye, it would seem that Nichols had removed Strype's "modernizing" and given an edition close to the manuscript. Such an impression is, however, quite wrong, though it took eighty years for the truth to emerge. In the 1930s, Axel Wijk made a careful comparison between the manuscript and Nichols' edition. Nothing of historical importance needed to be corrected, Wijk wrote, but "on the other hand, Mr. Nichols' rendering of M.'s orthography must be deemed extremely careless. His edition simply bristles with errors and inaccuracies ..."(2). Text 3: Nichols edition. The xiiij day of September was iij sett in the pelere for playhyng with falsse dysse and deseyffeng honest men in playng; and the same day was ij wypyd a-bowt London, [after] a care-hars, for lotheryng, and as wacabondes wher they taken. The xvij day of September was a proclamasyon that all vacabonds and lotherus, boyth Englys men and all maner of

Henry Machyti 's Day Book 221 strangers, that have no master, shuld avoyd the cete and the subarbes a-pon gret payn. The xx day of September was ij men dran of ij hyrdles unto Tyburne and un-to hangyng, the ij for qwynnyng of noythy money, and deseyvyng of the quen('s) subjects; the one dwelt in London sum tym. (Nichols 1848: 68-69)

Text 4. Michigan transcript. {Italics show supplied material.) The xiiij day was of septeber was iij sett on ye pellory ffor playhyng w' ffalsse dysse & deseyffeng honest me« in playng & ye sam day was ij wypyd a bowt london after a carehars ffor lotheryng & as wacabonds wher they taken The xvij day of September was a proclamosyon y' all vacabonds & lotheras boyth englys me & all maner ο / strangers y'haue no mast- shuld a voyd ye cete & ye subarbes a pon gret payn The xx day of septeber was ij me drau of ij hyod les vnto tyburne & iiij to hangyng ye ij for qwynnyng of noythy money & deseyvyng of ye quen subjects ye one dwelt in london sum tym (f35v) Before that declaration had been made, unfortunately, the chronicle had become a major document in the history of English, and its "evidence" had, like that based upon the nineteenth-century editing of the Margaret Paston's letters, been taken as secure.5 Quoted forty times in the OED, Nichols' Diary did not rise to the attention of the early twentieth-century linguistic historians like Karl Luick.4 When, in 1906, Wyld began to write books about English, the Diary had not yet come to his attention but he celebrated its excellence in the first edition of A History of Modern Colloquial English in 1920. His enthusiasm for Machyn led him into rhapsodizing about its excellence: "Machyn's work is a priceless monument of the English of the Middle Class Londoner with no particular education or refinement" (1936: 141). Wyld's commentary made the Diary a key document in the histories of sixteenth-century English. Not careful in selecting extracts from Nichols' flawed edition, Wyld nonetheless gave prominence to a neglected

222 Richard W. Bailey

document, and extracts have since become a specimen for textbook writers and a sample from it appears in the Helsinki Corpus.6 Wyld viewed the Diary as a text that "lets us into more secrets of contemporary speech than does any other work of the period" (1936: 146). Unfortunately, he was betrayed by Nichols' faulty edition. It was left for Wijk, however, to point out just how far Nichols ranged from the manuscript: abbreviations silently expanded, spellings regularized, insertions made from Strype or by conjecture not properly set apart. Especially misleading was Nichols practice in supplying words. Wijk pointed out a large class of defects: "while Mr Nichols sometimes gives the supplied readings in modern orthography, he often employs M.'s normal sp. or more occasionally some other kind of archaic sp." (2). Table 1. Versions of the Day Book Michigan Transcript


Wyld. 1

Wyld. 3

abrod sadyll a ffor mysse forten fflynge on in to fforest

abrod, abrod abrod sadylle Sadylle Sadylle a-fore afore afore mysse-forten myssefortyn myssefortyn flyng flyng flyng, on vn vn into into in-to forest, forest forest # requiem requiem requiem corsse corsse; corsse, corsse, in to in-to into into offesers offesers offerers offerers gentellwomen gentell-women gentill-vomen gentill-vomen a grett agrett a grett a grett # (abbat's) abbots abbots darce Darce Dacre Dacre whent wher whent wher The pound sign indicates a place in the manuscript destroyed by fire. Some of these mistakes reflect mere carelessness: vn for on, for instance; offerers for offesers·, wher for whent. Others like gentill-vomen were designed to portray Machyn as a Cockney. Displayed here are the first and third editions of Wyld's history.

Wijk was not gentle in his treatment of Wyld's interpretation of the usage of the Diary. Aside from his downright inaccuracies, Wyld presented

Henry Machyn 's Day Book


unrepresentative spellings to support his view of how London English must have sounded even though these are sometimes fictitious or rare variants of frequent forms.7 So, for instance, Wyld asserted that her "occurs at least once" for 'their', ignoring the hundreds of times that Machyn uses ther for 'their'. 8 At the end of a list of sins of omission and commission, Wijk stated his conclusion: "Wyld has in fact been far wide of the mark, and his phonological statements regarding M. are largely wrong because founded on an erroneous supposition" (12). So much for Wyld. Wijk did a remarkable service to scholarship in his engagement with the Diary, and, having done so much, it is surprising that he did not publish an edition.9 He did prepare a nearly exhaustive word index to the whole (omitting only the most frequent words and not printing the proper nouns he had collected since he believed their phonetic value was limited). Most important for the story, he worked from the manuscript and not from Nichols' defective edition. As he immersed himself in the text, he became persuaded that Machyn was not a "Cockney" (as Wyld had supposed). "The linguistic evidence," he wrote, "is so ample and unambiguous...that no doubt can reasonably be entertained regarding M.'s northerly extraction" (15). Unfortunately Machyn studies seem to be plagued by doubts, and what one scholar asserts so firmly will, it seems, be undermined by the comments of a successor. Thus in an essay published in 1963, R. M. Wilson attempted to demolish Wijk's argument that Machyn was a northerner, declaring that Wijk had not made the case for rejecting the "earlier view" that the chronicle was "written in a middle-class London dialect of the period" (216). Influenced perhaps by Davis, he asserted that "occasional spellings may or may not be significant; no use can be made of them because we can never be certain that their explanation is linguistic and not psychological" (206). Wijk had argued that the use of - s in the inflection of third-person plural verbs was an argument for northern origin, but Wilson was unpersuaded: "These give no support at this date to the presumed northern localization" (215). He found similar fault with Wijk's phonological features, arguing that London English of the day was "mixed" or that the numbers are not large enough "to be statistically significant" (206). But perhaps the most telling of Wilson's criticisms is that Wijk worked backward from nineteenth-century sources such as the dialect material gathered by, among others, A. J. Ellis and Joseph Wright.

224 Richard W. Bailey Heralding what may be a new civility in Machyn studies, Derek Britton returned to the question of origins in an essay published in 2000. Wilson, he wrote, "failed to appreciate that the set of features listed by Wijk could in no way occur in a text written south of the northern parts of the 'North Midland' area" (571). Though the word failure may seem harsh, it is a marked improvement in tone compared to Wijk (in his treatment of Wyld) and to Wilson (in treating Wijk). Britton recognized that the method of looking backward from nineteenth-century dialect evidence presented uncertainties and problems, so he made use of the Linguistic Atlas of Late Mediaeval English (LALME) to search for a fit between Machyn's usage and the scribal traditions of the fourteenth-century. In an earlier study using this approach, Anne Curzan drew tracing-paper isoglosses for particular features and stacked the sheets to locate the small area in the West Riding of Yorkshire enclosed by them all (see Bailey and Curzan 1997). Britton did the same thing, but he also drew upon the expertise of his colleague at Edinburgh University, Keith Williamson, for a computer algorithm comparing the Day Book with the LALME patterns and producing a fit for the intersection of them all (see Williamson 2000).10 Britton's discussion of the Day Book is an excellent and carefully argued essay with valuable methodological insights. He rightly pays tribute to Wijk's book - "a fine piece of scholarship, and in a sense a pioneering one" (573) - and he sets his paper-and-pencil results against Williamson's computer generated findings (which used a slightly different list of Machyn's usages). They settle upon an area south and west of Wakefield, though differing slightly in their localizations. The two focal areas are about twelve miles apart in the West Riding of Yorkshire. Even more recently, Ian Mortimer has brought the expertise of a historian to the problem of Machyn's origins. It was Wijk who reported a memorial plaque in the church of St. James Garlickhithe, London; it described Machyn as "Taylor and Clerke of the parishe and Clerke of Trinity the less." Unfortunately, Mortimer wrote, "this discovery was not made by a historian but a student of linguistics" (985), and Wijk failed to take the next steps that a historian would have immediately pursued. A search of records in the Guildhall produced Machyn's will, dated November 8, 1563, and prepared in the anticipation of his death. Further inquiry into the bequests and the names of family members traced Machyn's family to Hoby, a village northeast of Leicester (Mortimer 2002: 989). But Mortimer sensibly pointed out that Machyn must have arrived in

Henry Machyn 's Day Book


London by 1519, a quarter century before he began to write in his Day Book, and it is entirely reasonable to presume that his writing (and indeed speech) may have changed over this long time. This argument brings us to the most recent work on the Day Book as a key text for the history of English, though it is certainly not to the last word. Of course these are only the linguistic uses of the text, and I have said nothing about the other cultural practices where Machyn's testimony has been found valuable. A search of the citation indexes turned up references to the Day Book in work on undertaking, Morris dancing, musical composition, and "Butchers and Fishmongers: Their Historical Contribution to London Festivity" (Billington 1990). Here is a document of enormous value to the historical record. We believe that it is time that a reliable (and verifiable) edition be produced, and my colleagues and I at Michigan are preparing one that will incorporate in one electronic format: an image of the manuscript page; a meticulous transcript of it (with additions supplied from Strype); and a modernization to make the Day Book more accessible to non-specialist readers. We hope this work will be available in 2004 through the University of Michigan Press and the American Council of Learned Societies.

Notes 1.




In his impressive history of the MED, Michael Adams describes the importance of John Trevisa's translation of On the Nature of Things, a text that provides 2.5% of the quotations in the Dictionary. Trevisa's work is, as he writes, an invaluable encyclopedia of Middle English thought. Though the OED had drawn on an incomplete printed edition, the MED editors saw that this text was of great importance, and they employed a "rotograph" photostat until, and even after, the text appeared in a published edition in 1975. (See Adams 2002: 110 n. 23). Mortimer (2002) argues that it is misleading to call this compilation of entries a Diary (as has long been the case); Machyn himself mentioned it in his will as "my Chronacle" (quoted in Mortimer 2002: 986). Because Strype had made such extensive extracts, Nichols first located the passage and recorded the citation; these are still to be seen as faint pencil markings distributed through the manuscript. Apparently the first historian of the language to use information from the Day Book was Ludwig Diehl (1906) whose incompetence is carefully documented byWijk (8-11).

226 Richard W. Bailey 5.

Advisors commenting on my paper for this volume have asked that I point out some of the differences between the three published editions and the transcript (texts 1—4). The printed editions simply omit portions of the text (e. g., "wypyd a bowt London after a carehars" is omitted from both editions of Strype) or add sections not found in the manuscript (e.g., "Testimonials under their Masters Hands" in Strype). Grammatical features are "modernized" (e.g., "were two Men drawn on hurdles unto Tyburn" in place of "was ii me dran of ij hoyd/es vnto tyburn"). Spellings are either "modernized" (e.g., false or falsse for ms.jfalsse) or, as elsewhere in Nichols, deliberately archaized (e. g.,farre for ms. fare). Wijk (1937) lists a great many of the misreadings and misinterpretations that make Nichols edition so unreliable (222-228). As these brief extracts show, however, Nichols made a great advance on Strype in textual accuracy. As will shortly be seen, Nichols suffered the misfortune of having his text misused by some of his successors. 6. Wyld carried over the representation of the Day Book from the first edition (1920: 147) to the third (1936: 147). See also Fisher and Bomstein (1974: 219-222); Görlach (1991: 147-148); and Cusack (1998: 165-167). 7. Wyld regarded Machyn as "an Elizabethan Cockney" and so alleged that his usage resembled that of Victorian Cockneys. Wyld reported that Machyn spelled welvet 'velvet', werger 'verger', vornan 'woman', and vomen 'women' (143). Here are the actual distributions of these spellings in the Day Book: velvet (3), welvet (0); verger (1), werger (0); woman (54), vornan (3); women (123), vornan (11). 8. Nichols glosses "her plasse" as 'their place—the Draper's Hall' (141) on the assumption that the two-year anniversary of the death of "good master Lewyyn, yrnmonger" was celebrated there. Here is the text in question: The sam day at after non was ye ij yere myne of good Μ lewyn yrmonger & at ys durge was all ye leverey ye fifurst Μ altherman drap & aft- to her plasse & they had a kake & a bone a pesse by syd ye pryche & all comers & wyne he nowgh for all comers (Michigan transcript F74r; Nichols 141) Wijk pointed out that the cakes, buns, and wine served after the liturgy might well have been provided at the widow's place, and thus the Day Book should be taken at face value. Wyld did not offer this example in his first edition (see p. 145) but had added it by the third (145). This is a crucial point in suggesting the survival of ME here 'their'. Later in his History, Wyld wrote: "An undoubted example of her in late colloquial use is pointed out by Mr. Orton of Merton College, in Machyn 141, - 'and after to her plasse, and they, &c"' (328).

Henry Machyn 's Day Book


Britton has a different view; he suggests that Machyn might simply have forgotten to write the in ther as he occasionally forgot to write the in the (584). My point, however, is that Wyld uses this dubious example to assert that third-person plural her was in use in mid-sixteenth century London. Despite many uses of the third-person plural in the Day Book, this is the only one that lends itself to such a view. 9. In his review of Wijk's book, Mosse (1939: 374) hoped that he would publish an edition. After the war, Wijk became an enthusiastic advocate of English spelling reform; he did not, in print at least, return to the Day Book. 10. Britton lists the linguistic traits he studied but does not explain precisely which ones were used by Williamson.

References Adams, Michael 2002 Phantom dictionaries: The Middle English dictionary before Kurath. Dictionarie 23: 95-114. Bailey, Richard W., and Anne Curzan 1997 The Diary of Henry Machyn: An electronic text. In Tracing the Trail of Time: Proceedings from the Second Diachronic Corpora Workshop, Raymond Hickey, et al. (eds.), 25-31. Amsterdam: Rodopi. Billington, Sandra 1990 Butchers and fishmongers: Their historical contribution to London festivity. Folklore 101: 97-103. Britton, Derek 2000 Henry Machyn, Axel Wijk and the case of the wrong riding: The south-west Yorkshire character of the language of Machyn's Diary. Neuphilologische Mitteilungen 101: 571-96. Cusack, Bridget 1998 Everyday English 1500-1700: A Reader. Ann Arbor: The University of Michigan Press. Davis, Norman 1949 The text of Margaret Paston's letters. Medium Aevum 18: 12-28. Diehl, Ludwig 1906 Englische Schreibung und Aussprache im Zeitalter Shakespeares, nach Briefen und Tagebüchern. Anglia 29: 193-204. Fisher, John H., and Diane Bornstein 1974 In Forme of Speche is Chaunge: Readings in the History of the English Language. Englewood Cliffs, NJ: Prentice-Hall. Görlach, Manfred 1991 Introduction to Early Modern English. Cambridge: Cambridge University Press.

228 Richard W. Bailey Mcintosh, Angus, Μ. L. Samuels, and Michael Benskin 1986 A Linguistic Atlas of Late Mediaeval English. 4 vols. Aberdeen: Aberdeen University Press. Mortimer, Ian 2002 Tudor chronicler or sixteenth-century diarist?: Henry Machyn and the nature of his manuscript. Sixteenth-Century Journal 33: 981— 998. forth. Machyn, Henry. In Dictionary of National Biography. 2nd ed. Oxford: Oxford University Press. Mosse, F. 1939 Review of Wijk (1937). Etudes anglaises 3: 373-374. Nichols, John Gough 1968 The Diary of Henry Machyn, Citizen and Merchant-Taylor of London. Camden Society, o. s. 42. New York: AMS Press. Original edition, [London: J. B. Nichols and Son], 1848. Strype, John 1721-33 Ecclesiastical Memorials. 3 vols. London: John Wyat. 1820-40 [Historical and Biographical Works]. Oxford: Clarendon Press. (Ecclesiasstical Memorials appeared as vols. 14-16 in 1822.) Wijk, Axel 1937 The Orthography and Pronunciation of Henry Machyn, the London Diarist: A Study of the South-East Yorkshire Dialect of the Early 16th Century. Uppsala: Appelbergs Boktryckeriaktiebolag. Williamson, Keith 2000 Changing spaces: Linguistic relationships and the dialect continuum. In Placing Middle English in Context, Irma Taavitsainen (ed.), 141179. Berlin/New York: Mouton de Gruyter. Wilson, R. M. 1963 The orthography and provenance of Henry Machyn. In Early English and Norse Studies Presented to Hugh Smith in Honour of his Sixtieth Birthday, Arthur Brown and Peter Foote (eds.), 202-216. London: Menthuen. Wright, Laura, (ed.) 2000 The Development of Standard English, 1300-1800: Theories, Descriptions, Conflicts. Cambridge: Cambridge University Press. Wyld, Henry Cecil Kennedy 1907 The Historical Study of the Mother Tongue: An Introduction to Philological Method. London: John Murray. Original edition, 1906. 1920 A History of Modern Colloquial English. 1st ed. London: T. Fisher Unwin, Ltd. 1936 A History of Modern Colloquial English. 3rd ed. Oxford: Basil Blackwell.

The perils of firsts: Dating Rawlinson MS Poet. 108 and tracing the development of monolingual English lexicons Ian Lancashire

1. Introduction There are two preconditions for finding the first monolingual English lexicon: knowing how lexicographic works were classified generically in the Early Modern English period, and having scholarly editions of them. Robert Cawdrey's Table Alphabeticall (1604), which gives wordequivalents for just under 2500 headwords, is generally named as the first of its kind, but Cawdrey drew on Edmund Coote's English Schoole-maister (1596), and Jürgen Schäfer traced the English-only lexicon even further back to hard-word glossaries appended to books translated into English from the late fifteenth century. For contemporaries like rhetorician Richard Sherry, bilingual lexicons served English too, or so he said, half a century earlier, of Sir Thomas Elyot's Latin-English Dictionary. John Palsgrave, who chose English headwords for his bilingual English-French Lesclarcissement (1530), believed that he also served his native tongue. If they can be trusted, Richard Pynson's publication of the medieval Promptorium Parvulorum, an English-Latin lexicon (1499), qualifies as a vernacular reference work. Like bilingual dictionaries, topical or encyclopedic works such as herbals and terminologies serving law and medicine are not easily distinguished from the monolingual lexicon. To define a thing, logically, and to define the word denoting the thing, lexically, may seem hair-splitting, but monolingual English lexicography itself emerged slowly and uncertainly from the hands of writers for whom what we term a dictionary would have been strange. With generic indeterminacy, there are many firsts, all overlapping variants of text-types consisting of explanatory entries. The traditional way to date a text is in a full scholarly edition that brings to bear diverse evidence from sources, authorship, transmission, and topical allusions on the twin problems of when that text was composed,

230 Ian Lancashire and when published. Printed, microfilm, and digital facsimile series of many early lexicons are now widely available, but only two period lexicons out of hundreds have been accorded a full scholarly edition: Laurence Nowell's manuscript dictionary of Old English and Mark Ridley's manuscript glossary of Slavic. The following case studies illustrate the perils of assigning firsts in historical lexicography when these two preconditions are unmet. The supposed first monolingual lexicon, a fragment in Bodleian Rawlinson MS Poet. 108, dates well after Cawdrey's Table Alphabetical!. Although no scholarly edition, my transcription of the Rawlinson fragment (appended to this article) brings to light textual details that make necessary that redating. Similarly, by marginally expanding our generic boundaries for the "lexicon," John Rastell's neglected Exposiciones terminorum legum anglorum (1523) emerges as a first. Its lexicon has English headwords and explanations but also translates, on opposing pages, a parallel lexicon entirely in French. This early legal vocabulary, the antecedent of the Les Termes de la Ley, a work popular throughout the entire Early Modern English period, complicates making any clear-cut generic division among bilingual, monolingual, and encyclopedic lexicons.

2. The Rawlinson lexicon The Bodleian Library summary catalogue dates MS Rawlinson Poet. 108, a commonplace book, as having been "written in about A.D. 1570," information evidently taken from a note on the manuscript contributed by Philip Bliss to The British Bibliographer in 1812.1 Its first known owner, Eliner Gunter, was sister of Edward Gunter of Lincoln's Inn, London.2 This thin quarto records, near the beginning and in a well-formed Elizabethan secretary hand, two orations by a Mr. Pownde of Lincoln's Inn, the first in Shrovetide 1565 (i.e., 1566) and the second in June 1566. This must be Thomas Pound (15387-1616?) of Beaumonds, Farlington, Hampshire. The first mask to which he contributed celebrated the February 1566 marriage of Henry Wriothesley, earl of Southampton, Pound's cousin, to Mary Browne, daughter of Anthony Browne, Viscount Montagu; and the second mask, on July 1, the marriage of another of Pound's cousins, Frances Radcliffe, sister of the earl of Sussex, to a Mr. Mildmay of Lincoln's Inn. By 1570, Pound had left the comfortable life of a lawyercourtier to become an imprisoned recusant (Chambers 1923, ΙΠ: 468-469).

The development of monolingual English lexicons 231 It is not known whether Rawlinson Poet. 108 was his own manuscript, but the association of its first known owner with Lincoln's Inn, and the contents of the last two pages, which concern "Costs in the escheker of my ladye abbesse of Ambresburye for maykinge quitte of here fraunches in the Contie of Wyltes," a county that neighbors Thomas Pound's, support the possibility. Most of the rest of the manuscript, fols. 45r-81v, however, has nothing to do with court, nobility, or law. It holds what Noel Osselton described as the first (fragmentary) monolingual English "dictionary," in the dating of which he follows the Bodleian summary catalogue by assigning it to "the last quarter of the sixteenth century" (Osselton 1986: 175), specifically between 1589, when the glossarian's source, John Rider's Bibliotheca Scholastica, was published, and 1600. This could predate the first general-purpose English-only lexicon, Edmund Coote's hard-word glossary in The English Schoole-maister (1596). Manuscript commonplace books, unlike printed books, cannot routinely be dated by the first items in them. A new look at what has come to be known as the Rawlinson dictionary gives good grounds for dating the last important item in the manuscript, the dictionary, not between 1589 and 1600, but well into the seventeenth century, at least forty years after someone began the manuscript by writing out Thomas Pound's orations. The dictionary's alphabetic headings are copied from Francis Holyoke's revision of John Rider's English-Latin lexicon published sometime after 1611 and possibly as late as the Restoration. Although interesting for its dialectal forms, and as evidence that some in south-west England had never heard of Robert Cawdrey, John Bullokar, and perhaps Henry Cockeram and Thomas Blount, whose monolingual English lexicons had been printed numerous times from 1604 to 1656, the Rawlinson dictionary turns out not to be an early sign that English, as Richard Foster Jones argues, was on the verge of triumph. 3 Noel Osselton thought that the glossary-maker had cannibalized a manuscript table with alphabetic headers previously drawn up, sometime before 1589-1600, for making what Philip Bliss rightly calls "an index to some book" (618). I have appended the first printed transcription of the Rawlinson dictionary below 4 because only when the jumbled, six-part document is disassembled does the method of its compilation emerge. A closer reading shows that the glossarian did not model his work on a book index. The document has six sections. By differentiating them, we can establish that the glossarian created the alphabetic table. The first section

232 Ian


is the main glossary (fols. 45r-81v), holding both word entries and index entries, each in a different distinctive hand. The remaining five sections consist of (ii) another set of headings for the letter A (fols. 63r-64r); (iii) about thirty blank pages (fols. 64v-79r); (iv) a page of Latin headwords and comments, beginning "Abbas vnde & quide - eius officium et potestas - impedimenta" (fol. 79v; the source for these entries is unknown); (v) a second complete set of alphabetical headings, from A.b. to Ζ Ο; and (vi) a continuation of letter A (fols. 80-8 lr). The most striking anomalies here are the two sets of alphabetical headings, and the two continuations of the letter A. Looking more closely at the document will show why they occur. The main glossary, on fols. 45r-62r, about 35 pages, holds some 185 alphabetical headings from "A. b." to "Z O" under which one-part or twopart word-entries - that is, solitary headwords, or headwords followed by a single-word gloss or equivalent - extend from "abandon" (45r) to "worshiping" (62r). There are only 209 full two-part entries, ones with both headword and gloss, and they end at the headword "base" (46r), but the glossarian gives 444 entries in all (headword-only, and headwordequivalent), not counting a dozen canceled lines. The task of listing new headwords was abandoned at the term "baskett" (46r). Twice the compilerglossarian nested two or more alphabetized word-lists inside an alphabetic heading. Under A. b., for example, a well-alphabetized sequence (twenty headwords from "abandon" to "absurdity") is followed by a horizontal line and a second list of nine more headwords, from "abbot" to "abuse." Supplementary lists also characterize the holdings of A. c. and A. P. This part of the glossary shows someone in process of compilation. After the packed word-entries for A and Β run out, most pages with alphabetic headings are near-empty, but each page generally offers a few word entries. These 99 single headwords sprinkled through the remainder of the alphabet, in fact, tell us something important about the glossarian's method of composition. With one exception ("divorce - forsake," at fol. 49r), the single, unglossed headwords after the letter Β are not new words. Rather, they are copied from the equivalent or gloss used in two-part entries under A and B. The glossarian must have aimed at inflating his glossary by recycling the keywords in post-headword explanations as additional headwords. Henry Cockeram had produced a comparable reversible lexicon in 1623. Blended with the word-entries in the first-section glossary is a partial index of a book. Osselton did not identify the work, but the title is written in the manuscript at the end of the first section on fol. 62r. This part of the

The development of monolingual English lexicons 233 alphabetic table, having almost no word-entries, might easily be overlooked. The title is The Queens Closet Opened. Incomparable Secrets in Physick, Chirurgery, Preserving. Candying, and Cookery by W. M. (Nathaniel Brook, 1655). The book-indexer's entries are either simple, with a headword and one or more page numbers, or complex, with a headword and sub-headings, each with page numbers. The first book-index entries occur just after "baskett" on fol. 46r: "Balsom. 49.95.123" and "lucatello 5.179" refer to discussions of balsam and of Lucatello's book of the same name, Balsam, on pages 49, 95, 123, and 179 of The Queens Closet Opened. A check of other book-index entries confirms this source for the rest of the index entries. At first glance, it is unclear whether the glossary or the book-index came first. Osselton thought the glossarian re-used a grid originally marked up with alphabetical headings by the book-indexer and argued that we owe our hard-word glossaries in English to an opportunistic, novel application of indexing. There are two places where a book-index entry trespasses on or is written over a glossary entry. On fol. 57r, one word ("prevent") from the last sub-entry under the main index entry "Pocks" ("to prevent. 140."), overwrites and partly obscures the glossarian's headword "power," which he seems to have added after finishing the eleventh full entry, "able power" (45r). On fol. 61v, under the book-index heading "water of Life," the indexer had to enter the subentry ". Doct Butbers. 293" on a slant to avoid overwriting the glossarian's word "wandering", which was added after the gloss in the earlier full word-entry, "astray - wander" (45v). The glossary headword "watch-word" also forced the book-indexer to drop the page number ("296.") of the next sub-entry ("Mint.") below the line. Only by transcribing the whole document do these details draw attention to themselves and become meaningful. Section (ii) of the "Rawlinson dictionary" (63r-64r) removes any doubt that someone other than the glossarian, someone who possessed different handwriting, re-used the largely blank pages created and then abandoned by the glossarian to compile the book-index. This section holds twenty alphabetical headings from "A b" to "A X y z" under which the bookindexer has written entries for the headwords "Ach," "Agye," "Almond milke," and "Apoplex." The only reason that the book-indexer would have created a separate section for the first letter of the alphabet would have been because the glossarian had already filled up the pages devoted to "A." If the book-indexer had created the original alphabetical grid, the second section would have been superfluous.

234 Ian Lancashire

The fifth and sixth sections, fols. 80r-81r, further illuminate the glossarian's method of compilation. Short, only three pages long, they contain another series of alphabetical headings - from "A" to "Χ. Υ. Z." and then a continuation of "A". Under the "A" heading of section (v) is a roughly alphabetized list of 72 headwords, from "abbot" to "ambling pace," and a sprinkling of 14 cancelled headwords can be found under headings for D, F, H, L, Ο, P, S, and T. Under the "A" heading of the continuation (section vi), there are 45 headwords from "ambushment" to "apron." The hand of the glossarian wrote sections (i), (v), and (vi). For example, the capital "N" in their alphabetical headings has a reversed crossbar (e.g., "Ε H" at fol. 50r and "H" at fol. 80v). The headwords in these sections can be found interspersed in groups within the main glossary in the first section, where they stand without equivalents or glosses. It thus appears that the glossarian folded many unglossed headwords listed in the first section, the main glossary, into the last two sections. For this reason, the first section must predate the last two sections. Why the glossarian compiled these subsequent lists is not clear. Perhaps they collect incomplete glossing work from "abbot" to "apron" into a working area where glossing was to be delegated to an assistant or to be separately researched. Let me summarize. The "Rawlinson dictionary" is a partial draft of word-entries from A.b. to Β. Α., and of a sprinkling of words taken from the glosses in those word-entries and distributed through the complete grid of alphabetic headings (up to Ζ Ο), as well as supplementary lists of unglossed "A"-headwords found at the end of the manuscript. All three lists were compiled by the same hand. The glossarian, for unknown reasons, abandoned his task not far into the letter B. Afterwards, someone wishing to make an index of the 1655 book, The Queens Closet Opened, re-used the glossary's alphabetical headings, most of which were followed by blank spaces. The book-indexer, however, had no room to insert his AB entries into the glossary and so had to make another (the second) section, just following the main glossary, to hold A and Β index headwords. This index was also abandoned unfinished.

The development of monolingual English lexicons 235 3. Date By understanding the sequencing of sections in the Rawlinson document (that sections (i), (v), and (vi) at least predated section (ii)), we can see that the book-index postdates the glossary. Further, the book's date of publication, 1655, can have little bearing on the date of composition of the glossary. It must have been written between June 1566 (the date of Mr. Pound's second oration) and possibly not long after 1655 (the earliest date that the book-indexer could have started writing in the glossarian's grid). The handwriting of this glossarian is roman or italic in character (without secretary forms, including h) and looks to date from the seventeenth century. Fortunately, palaeography is confirmed by two other types of evidence: the glossarian's main source for his word-entries, and usage information about selected headwords. Osselton correctly identified the glossarian's probable source as John Rider's Bibliotheca Scholastica (1589). The over ninety headwords from "abandon" to "admonition," for example, were available in or had variants suggested by Rider's first edition, but there are a few exceptions, such as the words "abaleanate" (a variant of "ab-alienate"), "able," "abrogate," "academy," and "adiure." These exceptions can be found in the third edition of 1612, expanded by Francis Holyoke, but not in the second edition of 1606. The earliest date for the composition of the Rawlinson glossary, then, must be 1612. Because the third edition of Rider's dictionary was republished afterwards in 1617, 1624, 1626, 1627, 1633, and 1640, a much later date for the compilation of the Rawlinson glossary is possible. Holyoke's Rider was a popular reference work. What lexical evidence there is, and it is hardly conclusive, points to a mid-century date. Three of the glossarian's headwords under B. A. imply a provenance in the west country about 1640-1650. One headword, "stow ball" (46r), first occurs in the OED in 1634. It cites John Aubrey's explanation, in his Nat. Hist. Wilts. (1686): "Stobball-play is peculiar to North Wilts, North Gloucestershire, and a little part of Somerset near Bath." The OED records the second datable headword, "bannings. - fish," first in 1655 as a west-country term for the hackle or stickleback ("hackle" 2). The OED refers to the spelling of the third headword, "baize," as "rare before 1800" and notes it first in Pepys' diary in 1667. Another headword, which I transcribe as "baskett," following Osselton, may be "baskoll," meaning 'bascule', as in "bascule-bridge," which the OED records first in 1678. At least three words in the Rawlinson glossary, then, were first used,

236 Ian Lancashire according to the OED, in 1634, 1655, and 1667 respectively, long after 1612, when Holyoke first issued the version of Rider's dictionary that the glossarian used. Two of these terms belong to the west country. It is perhaps relevant that the manuscript's closing reference to costs of the Abbess of Ambresbury, Wiltshire, also places the manuscript in this part of England after the partial glossary and index were produced. What have we learned, after all this? Only the disappointing news that a fragmentary glossary, supposed the first of its kind, postdates Edmund Coote and Robert Cawdrey, and possibly John Bullokar, Henry Cockeram, and Thomas Blount? Far from being just a footnote to the history of English dictionaries, the Rawlinson document shows that vernacular glossarians routinely worked from bilingual Latin-English dictionaries and sometimes swelled the number of their English headwords by converting terms in the explanatory part of their word-entries into new English headwords. By transcribing the complete Rawlinson document, and by determining the interrelationships of all sections in it, we gain more information on how Early Modern English lexicographers worked. If the Rawlinson glossarian lived in the west country (as some word-forms suggest), perhaps he did not realize, until his compiling of entries for A-B was already well advanced, that London booksellers were already selling at least the small lexicon by Cawdrey. That being so, to many English, a monolingual vernacular lexicon would still have been a bright new idea for some years after 1612.


Three lexical streams in Early Modern English

Far from holding the first English monolingual dictionary, MS Rawlinson Poet. 108 has a derivative and late glossary. Its misdating can in part be put down to the disordered construction of Early Modern English manuscript commonplace books, to the difficulty of the hands in which they were written, and to the neglect of primary manuscripts that document Early Modern English vocabulary. The great bibliography by R. C. Alston treats printed books exclusively. Although a decade ago he prepared a brief typed handlist of some manuscripts in major British libraries, now available in the British Library Manuscript Reading Room (Alston 1991), few researchers have delved into archives for the manuscript records that medievalists, for example, have used successfully to reconstruct their period's vocabulary.

The development of monolingual English lexicons 237 The misdating of the Rawlinson dictionary is consistent with a scholarly view that the English only became lexically self-interested after 1581, when Richard Mulcaster published the first large English word-list, which has 8,000 headwords, including function words. Yet Edmund Coote's English Schoole-maister, published in 1596 with a glossary of 1368 hard-words, is small in comparison to English-language reference works produced under Henry VIII. Throughout the sixteenth century the English did homage to other languages by building huge bilingual lexicons, but when their headwords were English, these reference works served English as much as the foreign tongue. The first published bilingual Latin dictionary alphabetized by English headwords was Richard Pynson's edition of Promptorium Parvulorum (1499). The greatest early reference book on English was Lesclarcissement (1530), a French grammar by the Henrician courtier John Palsgrave. He anchored his work in very large English-French lexical tables. Even Sir Thomas Elyot's bilingual LatinEnglish lexicon, completed at Henry VIII's request and published in 1538 as the first self-proclaimed "dictionary" in England, was thought by Richard Sherry in 1550 to be an English lexicon suitable for "searchinge oute the copye of oure language in all kynde of wordes and phrases" (A3r, my italics). These and other great bilingual dictionaries signal English linguistic nationalism. Monolingual printed glossaries, surveyed excellently by Jürgen Schäfer from as early as 1475, are the second stream combining to form the seventeenth-century monolingual lexicons by Robert Cawdrey, John Bullokar, Henry Cockeram, and Thomas Blount. Such hard-word lists were aids to reading individual treatises with relatively narrow topical vocabularies. Words imported into English from other languages in books on astronomy, botany, history, law, logic, mathematics, medicine, and religious controversy had to be translated into easier synonyms for readers. Publishers readily supplied lists of such "terms of art" to fill gatherings with unused pages. These English-only lexicons, Schäfer implies, arose directly from bilingual dictionaries. Their lexicographical model translated headwords into equivalents: bilingual lexicons translated, into English, foreign words that were not in English; monolingual glossaries translated so-called hard-words that had just been absorbed into English and were as yet unfamiliar to many people. A third type of English-language reference work that emerged in the 1520s as an offshoot of Henrician nationalism may be termed the lexical encyclopedia. Sometimes this form gave what might be definitions as

238 Ian Lancashire understood in logic. Logical definitions, as lexicons in the Early Modern English period document them, explicitly claim to describe things, although they appear also to document words denoting those things (Lancashire 2002). The distinction between a lexical definition of a word and a logical definition of a thing denoted by that same word is not easy for many to recognize even today. How much more difficult it would have been in the Early Modern period, when the modern English worddefinition barely existed. Yet anyone familiar with the lexicographer's trade knows that modern lexical definitions provide more than wordequivalents, inflections, and etymology. Often our word-definitions today will give genus and differentiae, the old logical definition of a thing having become the referential definition of a word. While Schäfer astutely observes that "[tjhroughout the sixteenth and early seventeenth centuries the border between a monolingual dictionary explaining words and an encyclopaedia remained unclear" (I, 6), he and others have focused on hard-word glossaries to the neglect of quasi-encyclopedic technical books. These lexical encyclopedias, I suggest, are the third source of monolingual English lexicons printed from 1596 to 1656.

5. The first printed herbal lexicon Consider the first printed herbal, issued from the shop of Rychard Banckes on March 25, 1525/26. Entitled a newe mater / the whiche sheweth and treateth of ye vertues & proprytes of herbes / the whiche is called an Herball, Banckes' little compendium was republished 22 more times up to 1567. It was indeed new in its time. The grete herball, issued by Peter Treveris four months later on July 27, 1526, overshadowed it in size, but not in influence: only four more editions of Treveris' book survive, the last in 1561. Schäfer lists neither the Banckes nor the Treveris herbals as monolingual glossaries, although he does admit five pages from the latter, a section headed "The exposicyon of the wordes obscure and euyll knownen" (Ddlr-Dd3v; I, 21). Banckes' 206 entries, listed in alphabetical order, each give a single-paragraph description of an herb under a centered heading. The work might be called a lexical encyclopedia; its entries designate herbs both as things and then as names or words. Typically, a description begins by stating that "men do call" the herb denoted by the heading (or that the herb is "called") one or more names. Consider, for example, the description under the first entry, "Agnus castus":

The development of monolingual English lexicons 239 %Agnus castus. THis herbe Agnus castus/that me« do call Tutesayne / & otherwyse Parke leues. This herbe hathe leues somdele reed lyke vnto ye leues of Orage. And this herbe hathe senowes on his leues as hath Plantayne / and it hathe yelowe floures and bereth blacke berys/and it groweth in dry woodes. The vertue of this herbe is/it wyll kepe men & women chast. For as Discolidion & Placens do say / this herbe is called Agnus castus/for the knowlege and the vse of this herbe maketh men chaste. Also this herbe wyll open the poores of a man and let out wycked humours and spyrytes of his body. This herbe destroyeth the moysture of mannes sede. Also the same Auctour saythe that yf this be sothen with Fenell sede in Aysell / it is good to destroye the Dropsy. Also yf this herbe be sothen with Smalage and Sage in salte water / and afterwarde ye hynder parte of mannes heed be well wasshed therwith it heleth it & vnbyndeth an euyll that is called Bitarge. Also this herbe destroyeth the fowle lust of lechery/and it be dronken / or yf it be borne aboute hym. Therfore somtyme they do ete it rosted / bycause it shall kepe thera chaste . For yf this herbe be eten rawe / it wyll engendre heed ache. This herbe is good to defye the hardnes and stoppynge of the mylte. Also a playster of this herbe / is good to do away the ache of a mannes heed that is engendred of wycked humours. This herbe is dry. After identifying the herb in the heading, and then naming it at the beginning of the entry, Banckes' herbal proceeds to the plant's appearance and medicinal powers. Agnus castus can be recognized from its red leaves, yellow flowers, and black berries, and it grows in dry woods. The herb has many curative "virtues." It keeps one chaste, makes one sweat out bad humours, dries out male sperm, ameliorates dropsy and bitarge, causes headaches (if eaten uncooked), takes them away (if applied in a plaster), and leads a woman to lactate. Nothing is said about the meaning or etymology of the words "agnus" and "castus" in the explanation. Banckes' entries lack most lexical functions, yet the book resembles a reversed lexicon. The names come second, in the post-lemmatic segment, after a label for the thing those names denote. This inverts the sequence typical of mod-

240 lan Lancashire ern dictionaries, in which the name or names precede an explanation that refers to the logical definition of the thing {genus, differentiae). Banckes' herbal might be characterized as bilingual except that nowhere does the anonymous translator characterize the centered names of these herbs as either English or Latin. The headwords or head-phrases appear in the same black-letter type as the rest of the book, which is written in English. Although most heading names appear to be Latin, one-eighth (25 of them), mainly in letters Μ and after, are English: colewortes, iris, morell or nyght shadowe, mastyke, motherworte, maces, piper, pympemell, rybworte, reed netle, rosemary, smalage, saxfrage, sothernwood, stychworte, scamony, sene, selondyne, scabyose, town cresses, vyolet, veruayne, water cresses, wylde neppe, and wormwood. As well, the OED characterizes some apparently Latin headwords as English. Nearly a third of the 27 headings under A (agnus castus, absinthium, aristologia [aristolochia], ameos, alleluia, asterion, auricula, and allium) have main OED wordentries. Banckes' headings, then, are multilingual. By making no distinction between Latinate and English words (either by type font or labelling), Banckes implicitly treats the headings as if they were new, hard-tounderstand additions to English vocabulary. Typically, the post-lemmatic segments of his entries give the easier native names for the herb. The Banckes herbal resembles many hard-word glossaries identified by Schäfer. Although the 1525/26 herbal treats hard-words in one subject area only, includes encyclopedic content, and is named "herbal," it has a reasonable claim to serve as a lexical reference work. Banckes' entries explain the senses of words, the things to which those words point to as signs. They inform contemporary readers about their own current language. Of course, there were many earlier reference books on herbs that afforded him models, sources, and cribs (Callery 1978; Wellisch 1978).

6. John Rastell, the first Early Modern English lexicographer Schäfer recognizes another very early monolingual glossary, published one year after The grete Herball: it is another "Exposicion of old wordis" (fols. 85r-87r), this time from one section in a re-edition of John Rastell's abridgement of English statutes, published on December 22, 1527 (Schäfer 1989,1: 21). These five pages hold a glossary of "old wordis," the first one of which is "Sok." Rastell took this "exposition" of old terms from an ancient glossary surviving in dozens of early manuscripts.5 However,

The development of monolingual English lexicons 241 Schäfer overlooks the rest of Rastell's book, which consists of about 350 entries on topics from "Abiuracyon" to "wytnes." Each entry consists of a centered single-word or phrasal heading in large black-letter type, and a prose description in smaller black-letter type under it. The first edition of Rastell's abridgement of statutes, published on October 25, 1519, has just 145 such entries, and some have a lexical-encyclopedic form. For example, the heading "Ambydexter" has the explanation: "Ambidexters y' is he y' takyth money on bothe partes shall be put in no Iurre/And euery on before whom y* he passyth hath power to inquire of his defaute & to determyn hit/The.v.yere of E.iii.capitulo.x." (A2v). Other entries collect information related to the headword rather than explain it. Rastell opens his entry on "Admyrall," for instance, by specifying penalties to be suffered by mariners who, retained by the king, leave his service without permission. The lexical-encyclopedic entries in this work, however, inspired Rastell to publish, five years later, what is another first in English dictionaries. His Exposiciones terminorum legum anglorum ... The exposicions of ye termys of ye law of england (1523) consists of about 170 lexical-encyclopedic entries in two parallel columns, the first an italicized entry in French, and the second a black-letter English translation of it (cf. Cowley 1932; Graham 1954). 6 Rastell's title explains his purpose as plainly lexical. The preface rewords that purpose as being "to declare and to expown certeyn obscur and derke termys consernyng the lawis of thys realme" (A2). The glossary on the "exposition" of old words that Schäfer excerpts from the 1527 edition of the abridgment of the statutes, a different book, differs from Rastell's 1523 book in only one way: the age of the vocabulary. Here is Rastell's first English entry: ABiuracion is where one that hath committyd murder or felony fleith to ony church or other place pri uilegyd for the sauegard of hys lyff / and ther be fore the coroner makyth suche con fession which may make a sufficient indite ment of felony then ye coroner shall make hym to forswere y e realme & shall assyne to hym to what port he shall go & shall swere hym that he go not owt of the hye way / & that he shall not abyde at the port / yf he may haue good passage / but one flod & an ebb / and yf he can not haue passage y* he shall go euery day during xl. days in

242 lan Lancashire to the see to hys kneys but yf sych a felon that abiuryth go owt of the hey way and fleyth to a nother place / yf he be takyn he shalbe brought before the luge / and ther shall haue iugement to be hanged / But yf he whiche so prayeth the priuilege for xl. dayes. and euery man may geue hym mete & drynke but yf any geue hym sustenaunce after .xl. dayes all though it be his wyfe : suche geuynge is felony/ Also he that doth abiure shalbe delyuerd from one constable to another / & from one frau/j ches to another tyll y' he com to his porte / & yf the constable wyll not receyue him he shalbe greuously amercyd. ^ Loke in the tretyse De abiuratione coronatorum. Rastell gives an extended logical definition beginning "ABiuracyon is ..." He also outlines what "shall" be done to those abjuring the realm. Linguistic attributes of the noun, such as etymology or equivalents in other languages, are absent. The explanatory sentences even use related verbal forms, "abiuryth" and "abiure," instead of the noun. Yet Rastell's entry gives information about meaning found in, as well as neglected by, the OED. Its definition, "an oath taken to leave it [a town] for ever" ("abjuration," 4), lacks the rider that anyone may abjure the realm and yet stay in England simply by taking sanctuary in a church. The search for firsts in early English lexicography is partly a semantic problem. We are looking for a genre that emerged only gradually from the mutual influence of overlapping text-types like the bilingual lexicon, the encyclopedia, the herbal, and the hard-word glossary. Sir Thomas Elyot used the term "dictionary" first in 1538 for a bilingual lexicon.7 Even as late as 1623, when Henry Cockeram published The English Dictionarie: or an Interpreter of hard English Words, the word retained that meaning. Cockeram did not write lexical definitions but rather lexical equivalents; he turned hard words into easy, and easy words into hard. If the first dictionaries in England often did translation, then any early lexicon that translates from an English headword, or into an English explanation, was in the eyes of contemporaries an "English" dictionary. A Henrician might well have thought the (1499) Promptorium Parvulorum to be more English than even Sherry thought Elyot's lexicon to have been. John Rastell's bilingual Exposiciones terminorum legum anglorum (1523), consisting of a

The development of monolingual English lexicons


French entry in one column, and the English translation of it in a parallel column, also gives translations, but they are complete English lexicalencyclopedic entries. Playwright, coroner, printer-publisher specializing in law text books, merchant-adventurer, one of Cardinal Wolsey's lawyers, son-in-law of Sir Thomas More, and the first public stage-owner in England, John Rastell is an English-language pioneer deserving of further study.8 His book has a reasonable claim to be the first printed stand-alone monolingual English lexicon, and John Rastell to be our first native lexicographer. On the other hand, Rastell took the generic name for his legal lexicon, the expositions of terms, from an ancient glossary of antiquated law words, the "Exposicio vocabulorum," that he later copied into his (1527) abridgement. To this work, so titled and found very widely in medieval and sixteenth-century manuscripts, can then be traced the monolingual English hard-word glossary that became the dictionaries we use today. If so, is this ancient manuscript item not then the true first of its kind? Perhaps it is, at least until we investigate, in turn, its sources in medieval learning.

Appendix: Transcription of "' [fol. 45r; col. 1] A. b. abandon -- forsake abaleanate -Abase -- few« humble, abash. abate -- diminish abreuiate -- abstract abbridge abhomination -- offence abhoure -- loath abide -- -- tarry abide -- suffer, abiect -- -- forsake abiure -- forsweare able -- power, abolish -- destroy -- abound -- plenty -- abrogate -- destroy -- absolue -- pardon

Rawlinson Dictionary" -- absolutnesse -- peril*] fection -- absurdity -- foolish[*]nesse -- abbot -- abbey -- ablatiue case abortiue abricot appl absence abstinence abstract abuse abstract A. c. academy -- vniversity accnt accent

244 Ian Lancashire

accesse vnto -accomplish -- perfection Accoumpt account ootimation accorde -- agreement account - estimation. Accumulate -Accusation accustome -- vse ace -- dice Ache . achieue -- perfection acknowledgment confes[*]sion Acorne Aquaintance Aquittance Acre Act[illegible] act -- decree statute action -- deed Action in the law Actiuity adamant αton addition accent acceptance accidens accompany accoumpt accuse acorne act actiuity action [col. 2] A. d. Adamant - stonn. Addition Adder addicting -- employing adhaerent -- partaker Adiectiue

adioyning [illegible] adiourne -- delay adiunct adiure -- coniure Administration admiral of the [*]seas admiration Admission admonition adoption adonation -- a worship[*]ing adoune aduancement -- prefer[*]ment aduantage -- gaine aduenture -- enterpuise aduenturing -- trying [*] [?] proue aduenture chaunoe fortune aduerse aduersary aduersity aduertisment -- admoni[*]tion aduise -- councell Adultery aduocate aduowson admonition A. f affability -- courtesy affaires -- business Affection affectation -- -- curi[*]osity affiancing -- betrothing affiance - faith affinity -- kindred. Affirmation Affliction affluence -- plenty

The development of monolingual English lexicons


Aide Aime Aire

Alderman alder -- tree alembick -- stillatory ale alehouse -- inne aliant -- strainger alienation -- forsaking alegation -- proofe of [*]matter allegory alley alliance -- kindred allot -- appoint allowanc allowance allurement -- entisment alusion almanack -- calender almes house of almes -- houoe [*]hospital almond aloes alpes alpha alphabet altar alteration -- chainge altitud -- height allum.

Α. Κ

A .Μ

ake - - ache akorne -- acorne

Ample -- feaue ambage -- circumstance Ambassage Amber ambiguity -- doubtfulnes Ambition Ambling pace ambrey -- cupboord ambusment amend -- repaire amend -- correct amerse -- fine Amesse -- robe amity -- frindship

[col. 3] A. G Age Agent -- action Agrauate agility -- actiuity Aglet agnaile -- corne agnition -- kowledg agony -- anguish agreement ague agreement a 1 [illegible] Α. Η I.

[col. 4] A . L Alabaster alacrity -- mirth alarum - watch-word alay -Albe Alchymy

246 Ian Lancashire

Amphitheatre amphibology [fol. 45v; col. 1] A. N. Analogy analosys anarchy an anatomy anatomy out. ancestors anchor -- anker ancientness -- antiquity ancients -- ensigne ancle andiron anuil -- anuile angell anger angle -- rod angle - corner anguish animaduersion -- counts sell aniseed anker annals -- cronicles anihilation -- frustrate tion [illegible] anniuersary annotations -- notes annoyance -- loathing annoynting -- oyntment annuity answeare antheme -- song antichrist antik-work antidote antiquity antipathy anuile auprile

antiquity au[?]er anguish A. P. Ape Aphorisme aple -- apple apocalyps -- vision apocripha -apology -- defence Apoplexy apostasy -- forsaking Apostle apostume -- impostume apothogmc apothegme -- sentence apothecary appaire -- deminish appale -- discorage Apparrell apparition -- vision appeach -- accuse appealing appeare appease -- pacify appendix appendent appetite applause aple apple oth eye -- eye applycation appointing appose -- examine apprehension apprentise approach approuement -- inlarging approbation appurtenance [col. 2] Apron aptness - fitness

The development of monolingual English lexicons


appoynt apple [inverted] Art Artechoke artery -- sinow Article Artificer artillery -- gun [/inverted] applo appoint A. R. aray -- apparell aray -- order arbitrator iudge arbitrement iudgment Arbour Archangel1 - herbs Archange11 Archbishopp Arch deacon Arch-duke Arch. Archer - - shooter Arch -- heretick arch prist architect. -- builder Argument arising Arithmetick Arke Arme Armes Armyc Armour army arquebre -- pistoll arrainge -- accuse Arras -- cloth Are st arriue -- approach arrogancy -- pride


Arsnick -- vid supra [col. 3] A. S. Ascend. Asiuile -- impulse aschamed -- bashfull[*]ness. Ash -- tree Aoctrco Asties Ash Wednesday askeng -- demaund aokew aoquint Aspe aspiration -- breathing aspire -- ambition assay -- proue Assaile -- assault asse Assembly assent -- agreement assentation -- flattery assertion -- affirmation assiduity -asseige -- beseige assigne -- appoint Assise assistance -- aide associat -- accompany assoile -- pardon assure-affirme warrant Aswage astonish -- feare astray -- wander Astrolabe -- instru[*]ment Astrology Astronomy.



atchieue -- perfection

248 Ian Lancashire

Atheiste attach -- arrest attaine -- obtaine attainture -- corruption attempt -- -- enterprise attendance -- seruice. Attentiuenesse attenuat -- diminish attire -- Antick -- ap[*]parrell attonement -- agreement Attoorney attrap -- trapping attribute -- impute

Awake award --

iudgmennt defend

blow awe -- feare Α. X Axe Axell-tree axiome -- aphorisme Azure -- color. [fol. 46r; col. 1]

[col. 4] B. A. Α. V. auaile -- profit auarice -- coueteousness audacity -- boldnesse audience -- attention Auditor . - Keep [*]Reciptes auenge - reuenge averr -- affirme augment -- [illegible] Augue auguration -- conniue August Aunt Auoid -eschew -- I loath empty auouch -- affirme authority authority, autumne - haruest awe -- feare Awake award -- iudgment award a blow -- defend A. W.

babe, baby - child. Batcheler - vnmarr. Batcheler of art. B. Baccus -- god put back -- repulse Keep back -- reteine giue-backe -- recoile goe back - returne. Back sliding -- apostasy Backe. Back-bonn. backbite - slaunder. [illegible] bacon badge - armes. Badger - - gray badger, for corne. bagg bagge.-- bagagc pipe. [*]pipe. bagg pudding - pudding, bag & baggage bay. for shipps to lie [*] in bay tree. baile. -- maineprise bailife. -- or ser[*]geant baite . for fish baize . - frise.

The development of monolingual English lexicons 249

Ba bakeing . er -- bee house -- pan. balad. baoao balasse for [*]ships Baldness. bale -- baile Balk - furrow. Ball. -- hand foote ball, stow ball. ball of eye -- eye. ballance Balme band of souldiers band to bind, band hat band, bane -- poison banneret bang -banish. banke . hillock, banker, for money, banne -- curse bannings. - fish, banner -- banner - - en [*]signe. banquerupt banquett baptisement barber Barbarian Barbarisme barenesse -- nakednesse bargane. barge. barke like a dogg bark of a tree bark or little shipe barly barme or yest. barne barnaile for horse baron bar --or bolt

barne or Couer barre where yt plead barrell barren. barrier barrester barrow [col. 2] barter -- chaunge base - musicke. baleneshe bashfulnesse. basolike baskett Balsom. 49. 95. 123 lucatello 5. 17 9 Back strengthened 66. 134. 168. 184. For whiles & heats 94. 95. for wasting 191. to strengthen 288. 289. B.E Belly great & hard. 155. ache. 184 Bezorston. 87

[col. 3] [col. 4] beseige Β. I Bile. 89. to breal to ripen. 112. 153 Biting of dog or serrp 131. 153. 154. 154.

[fol. 46v; col. 1]

250 Ian Lancashire Β L (.) purged. Βloud. 10. 161. preserved 291. Bleeding, stayd. 163 spetting. 234. Bladder. 237. (1) clen sed.

66. m [fol. 47r; col. 1] Bruises. 36. 38. 90. 104. 122. 156. 184. breath

[col. 2] Β V Β. Ο Burn or ocald.—149. [col. 3] [col. 2] Bone Ache. 72. 156. Bone broke. 104. Bone out of ioynt. 153. B. R Breath. Stopt. 5. 76. 78. stinking. 53. 111. 178 276. Breast sore. 85. Hard or swoln by Bruise. 91 sore. 109. Ague. 110. sore. 117. by cold or festring by milk. 147. Braine. weak. 86. com= forted. 178. 235. 236. 236. moyst. 236. weak. [*]278. 279. Breast, ague 165.—165. 167. 167. [col. 4] Broth -- chyna: 34 restoratiue. 42. 50. Hearbs for br. 65.

Burning, or scalding 60. 77. 78. 104. 149 162 builder businesse C A [col. 3] [col. 4] Canker. 36. 111. 153. [fol. 47v; col. 1] C. Ε C. Η [col. 2] Childrens Teeth. 47. Nauell out.109. dead ch dd. 160. brought to life. 159. wormes. 176. thrush. 36

The development of monolingual English lexicons Saye. 235. Lauender flowers. 236. Marjeram. 236 Piony flowers 237

Choler purged. 176. 177. cooled. 233. [col. 3] chaing

[col. 4]

C. I.


[col. 4]

Cordiall. 2. 4. 7. 11 14. 18. 24. 27. 86. 235. 237. 243. 244. Sr W R .274. lady Mulletts. 27 5. doc stee. 275. 280. c o m e s . 104. 146.

C. L clyster for wind 30. 100. Fox feauer. 132. [fol. 47r; col. 1] C. Ο

Cordiall. 2 83 285. 286. 288. 290. 291. 293.

Cold. 27. 164. 306. [fol. 48v; col. 1] Cooling=Julepe 61. 128. 177. 233. [col. 2] Μ [col. 3] Ν Conoumplio Conserues Roses. 207. 208. Violets. 233. Roses. 233. Borage flowers 234. Rosemary flow= ers. 234. Betony 235.

confession conniue counsel1 courtesy corne connect corner corruption coueteousnes Cough. 55. 56. 63. 100. 164. 299. - - o f Lungs. 59. 62. 258. Cramp. 141 [col. 2] C. R

252 Ian Lancashire

[col. 2] [col. 3] D. 0 cronicles dog mangy. 12 6. C. V


[col. 4]


cupboard curiosity

D. R. Draw.iron. 153.

C Y Draw Humors. 184 D. A [col. 5]

[col. 3]

I. V

Dropsie. 11. 53. 276. 295.

[fol. 4 9r; col. 1]

D. V


[col. 4]

Deafnes. 45. 105. 162.


[col. 2]


[col. 3]

[fol. 50r; col.

demaund dyseued delay deade destroy


[col. 4]




diuorce -- forsake

[col. 2]

discorage died diminish


[fol. 4 9v; col. 1]


Ε Β Ε C.


The development of monolingual English lexicons 253 Elicampan Paste. 258.

[col. 3] Ε Υ

[col. 3] Ε Μ Emplaister. lead. 183. deminum. 188. Mellilot. 188 Diapalma. 189. Oxicrocium. 189. Paracelsi. 150. emty Ε Μ

Eyes. 18-r- 28. 32-·- 147. that cheunie. 18. 32. 110. 181. 288. weake. 130. 147. Filme. 148 Pine well. 171. 172. 173. red. 171. 173. [*]181. aking. 171. Hot rhume. 172 Pearle. 17 3. good. 184.

[col. 4]


enlarge ensigne entisment enterprise


Ε Ρ Ε Q Ε R. [fol. 5Ov; col. 1] Ε S eschew estimation

Face 18-.—l·^ swelled 13. made faire. 53. [*]115. 291. rednes. 53. 54. [*]173. 180. sweating. 55. Bruised. 144. Scurfe. 173. heat. 173. 173. [*]180. Falling sick: 88. 142 237 .

Ε I Fainting. 133. Ε V Face made faire. 180. [col. 2] [col. 4] Ε X faith Expel 1


254 Ian Lancashire F Ε Freckles. 146. Fellon. 48. 102. 184 Feauer. 61. 98. 123. 128. 132. 133. pestilent. 25.

frustration frindship [col. 2]

[fol. 51r; col. 1]



Fundamt cured 184.

F I G A [col. 2] [col. 3] Fistula. 70. 79. 117. [*]153.


Fish by Angle. 108.


fitnesse find

[col. 4]

F L [col. 3] Flos vnguent. 40. Fluxe bloudy. 106. 108. 184. Flux. 13 0.

G I G L [fol. 52r; col. 1] G Μ G Ο


Gold-colour. 114. Gold letters. 114.


[col. 2]

[col. 4]

Gout. 77. cold: 156.

fortune foolishnesses forsweare forsake

G R.

[fol. 51v; col. 1] F R

Green sicknes. 69.85. [col. 3] G V

The development

of monolingual

English lexicons


gan hospitall G Y Η V [col. 4] Humours, grosse. 70 Η A Hum. 287. Haire to take of 55. To grow. 100. 270. To be black. 271.

Humble [fol. 53r; col. 1]

[fol. 52v; col. 1] I A haruest Η Ε Head light. 72. 235. 237. meagrim. 119. 235 Ach. 120. Broke out. 148. Rhumatick -- 235. 236. cold. 235. 236. Heart fainting. 86. 133. 289. 291. passion. 98. 287. 289. sick. 142 cooled. 234. 291 comfortod 234 [col. 2] height. Η I [col. 3]

Iaundies. 73. 74. yellow. 74. 105. I D I Ε [col. 2] I F I G I L I Μ Impostume. 89.(·) to break, to ripen. [*]112. 153. to break: [*]184. impulse impostume imploying [col. 3]



[col. 4]

[col. 4]

256 Ian Lancashire inne

Leg. old. sore. 111. 117.

I 0 Lead plaister. 183. Ioynt: felon. 102. fol. 54r; col. 1] [fol. 53v; col. 1] L I g. I R Life: water. 16. Iron from casting. 113. I

Licoris juice. 47.


light head. 72 Itch. 77. [col. 2] Κ A Liuer. 10 ( ) obstructed. 268. 295. good for it. 92 To purge it. 93. to coole. 184. to warme. 236.

Κ Ε La kents powder 274. [col. 2] Kidnies. vlcer. 50. swoln. 56. To clense. β-5-τ- 83. 237. To coole.


Kings euill. 117. Kindred Κ Μ Kouwledg [col. 3] L A [col. 4] D Ε

[col. 3] L Ο Longs. 3. 10. 14. 22 34. 47. 123. 186. 299. 278. 279. 291. 295. Loosenes. 80. 135. (se Flure) 143. Loosen: pro. 159. 233. loathing [col. 4] L V Μ A [fol. 54v; col. 1]

The development

of monolingual



[col. 2]



NAKED nakod & not aohamod. Com—2-r 2-5-T-

Melancholy. 4. 23. 177. 289. 291. 293.

Nauell out. 109. Measles. 29. 137. Memory. 291 Meagrom 119. 235 Mead. 284 [col. 3] m

Μ Ε [col. 3]

]/[ I Nipples healed. 174. 175. to get nipples. 175.

Μ I 14 0 Mi Ike. 35.(.) increa sed. 113.

Noli me Tangere. 52

Mirabilis aqua. 290

Nose. red. 53. 54. bleeding. 163.

mirth notes [col. 4] [col. 4] Μ 0 Μ V Morphew. 146. 0 A Mother rising. 63. suffocation. 236. [fol. 55r; col. 1]

Ο Β obtaine [fol. 55v; col. 1]

Mouth. 36. cankett tt 111. Mosten it 132. Rhume. se Throat.

0 C


0 D

[col. 2]


258 Ian Lancashire offence

3 t Io: wort. 131.190'

[col. 2]


0 I

[fol. 56r; col. 1]

Oyle o£ Swallowes 182.

Palsey. 6. 92. 77. 291. shaking. 276.

Ο Κ Ο Li

Paracelsus pluis iter. 150


{col. 2]

[col. 3] Ο Μ

pardon pacify partaker

0 Ρ


0 R

Percle. Magistry. 269.

order [col. 4] Ο S. Ο Τ Ο V Ο W

[col. 3] Perfume: 24. perfection Ρ Η Phlegm. 44. 57. 142. [col. 4]

Oyntmt green. 97. 158. Ρ I Ο χ Piles. 36. 43. 101. Oximel composition. 67 Oyl of Excester. 75. [*]156. of Eggs.77. of Mustard seed. 77 of Fennell. 78 of Rue. 78. camomill. 78.

pistoll. Ρ L Plague 1. 17. 24. 25. 30. 31. 33. 39. [*]40. 67

The development of monolingual English lexicons Ά priservatiue eg agst it -- 1. 17. [*]33. 39. 40. 67. 78. 107. [*]lis water 3T-& 277. 279. after infection. [*]24. 25. 30. 31. 105. [*]107 142. [fol. 56v; col. 1] perfection pardon plenty Ρ 0

profit pride. proof of matter preferment Ρ S [fol. 57v; col. 1] Ρ V Put away - - foraake Purging. 2. 68. 161.279. Purge. 52. 102. 161. 161. 280./ Purples. 82

[col. 2]

[col. 2]

Powder curing, 19: Gascony. 187. Lady Rents. 274. Docto r Atkins. 74 Doct Mores. 178.


Pocks sm. 29. 136. not to be seen. 135. 137. 138. To dry yem. 136. in Throat. 137. Eyes. 139. struck in. 139. 29. 136. to out yem. 140. 140. to prevent. 140. power Ρ R

R A [col. 3] R Ε rcmoue

Reines, running. 184. clensed. 237. [col. 4] [fol. 57v; col. 1] R H. Rheume. 47. 57. 58. 76.

[col. 3] [col. 4]


Reuenge Repaire


260 Ian Lancashire R I

5 Η

Rickets. 126. 127.

ohamo. both naked

[col. 2]

6 not aohamod. Gon 2.25.


Shingles. 153.

[col. 3]

[fol. 58v; col. 1]



Rupture. 12 9. 145.

[col. 2]


Sinew-ache. 72. weak. 86. 235. hurt. 153. contracted. 276.

Saffron = water. 18.

Siluer lefcteres. 114.

Salue. 40. green. 10

si now

S' Anth: fire. 153.


[fol. 58r; col. 1]

[col. 3]



Scalding or burning 60. 77. 78. 104. [*]149. 162 .

Sleepe procured. 60. 101. 277.

Scuruy. 149. 185

[col. 4]

[col. 4]

[col. 2]




S 0

Sear cloth. 38. 157.

Sowning. 86. 133. Ά Babe. 159.

[col. 3] [fol. 59r; col. 1] service sentence


[col. 4]

[col. 2]

The development of monolingual English lexicons

S Ρ Spirit of Castoreum. 20. Spleen, good for it. 66 123. to warme. 236. to d e n s e . 268. & Melancholy. 164 165. 291. hardnes. 236. Sprain. 122. 131. Spettle bloudy or corrupt. 234. [col. 3] S Q [col. 4] S Τ Stich. 65. 112. [fol. 5 9ν; C O l . 1] Stomack.—5-·—6-»—1-5-.—3-5-r —Pain.—4& — C o l d . 27, 15. 57. —wind Ston in k. 7. 50. 99. 235. 276. Ston in bl. 8. 276. Ston. 21. 33. 46. 52. 59. 84. 94. 163. 166. 167. 168. 170. & gra= veil. 185. Stomack. stuffed. 5. 63. 161. 235. 236. [*]268. 295.


-- digestion helpt. 6. 15. 35. 178. 235. 235. 236. 236. 243. [*]244. 289. 291. -- cold. 15. 35. 57. 100 235. 236. 236. 236. 243. 244 pained. 48. 121. -- windy. 35. 57. 77. 178. 236. 236. 236. 243. -- to gett Appotite. 61. 188. 235. 236. 236. 236. 243. 289. [*] 291 -- full. 65. -- rheumetick. 65. 100. 235. 235. 236. 236. 236. 243. 291. - - T o clear & com= fort. 12 8. 12 9. 161. 235. 235. 236. 236. 236. 244. [*]279. 286. 289. to -- cooled 234. 243. Straine. 122. Strangury. & ston 168. 170. 191. [col. 2] Sowning 86.— Stung. 154. strainger stillatory statute S V Surfet. 292. [col. 3]

262 lan Lancashire suffer

Throat. 44. 51. sore Rhewm. 57. 76. cannot Swallow 51. cooled. 233.

[col. 4] S W Swelling. 36. 104. 184. or Ach. 108. 157.

[col. 3] for a Timpany. wa by water. 65.

s y Τ I Syrup of Turnip 9. Citronpills. 9. 28 limons. 28. 221. Learmaines. 23. Ale. 95. cloue jnly flow 205. Hysope. 266. Citrons. 221. Elder berryes 267. Raspberryes. 284

Tinclure of Am bergreese. 24. [*]216. [col. 4] Τ 0 Tong speech les. 70 72. Τ R

[fol. 60 r; col. 1]

[fol. 60v; col. 1]



Taste gained. 61.



[col. 2]

[col. 2]




Tetter. 22. 174

[col. 3]



come. 47.

to d e n s e . 273. Τ Η Thrensh. 36.

V A V Ε veines. windy. 122. [col. 4]

The development of monolingual English lexicons 263 V I

Vsquabah. 208. 28.

[fol. 61 r; col. 1]





vuula. 52.



V Vi

[fol. 61v; col. 1]

Vncombe. 81. 81. Vnguent Album. 189 Dialthece. 189. Populium. 189. nervinum. 189. tulice. 189. busilicon. 189 Balsum. 190. pomalum. 212.

[col. 2] [col. 3] vniuersity vniucrfity V Ο vomitt pro.57. contra. 133. V Ρ vrine. 8. 14 bloudy. not held. 91. 135. stopt. 99 sharp. 162

[col. 4] V S

warts. 145.

[col. 2] water of Li£e. 16 . of Saffron. 18. . do: Stephens. 87. [*]21. . Orenge 206. . Cherry 222. 282. [*]283. . Cordial. Sr to R 274. . Cordiall. La: Mal= lets. 275. . Doct Stephens. 275. . Plague. 277. 181 . Poppy. 277. . Lemon. 285. . Tyme. 287. . Snayle. 288. 294 . Clary. 289 . Doct Momfords. . Mirabilis. 290. . Surfett. 292. . Doct Butbers. 293. . Hint. 296.

wandering warrant. watch-word watch wordd

264 Ian Lancashire

W Ε weknes. 42. 2 88. wen. 144. [col. 3] W Η [col. 4] W I w i f e . mado to bo a holp to m a m — G o n mann fleoh & bono«—Con« 2.23. mano oolfc. v.24. [fol. 62r; col. 1] winde. 30. 28. 44. 63.= collick. 63. 99. [*]105. 160. A Clyster. 30. [*]160.

woman wth ch: 46. 121. uuu • in labor. 86. 119. [*]130 • after throws. 118. • Sowning after. 118 • To avoyd after =bur then. 1129. 159. 160. • Loosenes. 143. wormes. 76. 89. 89 116. 176. 276. wound. 122. 131. 153. 179. Bleeding. 163. Healed at distance 181. • woman to conceiue • 168. 184. 235. 276. • courses provokd 184 wombe clensed. 23 5. warmed. 236. worshiping

[col. 2]

[col. 3]



w o m a n . Gen.—2-.—te-r mano holp./ vndo facta Gon»—2t—22/ q S — & quaro vocatur woman. Gon ι—2-r —namod by Adam, ibid.—& chapt—3.20.

Y A Υ Ε [col. 4] Υ Ο

Wound. 38. 90. 112. gren wound. 103. pricking of a Thorn. 103. drink. 115.

Ζ Ε Ζ 0 A Table to the Queens closet

The development of monolingual English lexicons Opened. 1655.

[fol. 63v; col. 1]

[fol. 62ν; col. 1]



Almond miIke. 82. 83. 83. 108.

iudgment [fol. 63r; col. 1]

A m

A b

[col. 2]

A c


Ach. 38. 153. 156. 184. by cold. 72. scialica. 101. 156. or swelling. 108. 184.

A ο [col. 3] A ρ Apoplex. 236.

[col. 2]

A q

A d

[col. 4]

A e


[col. 3]


A f

[fol. 64r; col. 1]

A g


Agye. 31. 57. 127. 101. quartain. 52. tertian. 161. in any part. 165» 167.

A V [col. 2] A W A X y z.

[col. 4] [fol. 64v-79r blank] A j [fol. 79r] A k [inverted]

266 Ian Lancashire

Abbas vnde & quidem -[*]eius offi= cium et potestas -- im [*]pedimenta Albatista Abdjcicatio -[·.·.] [/inverted] [fol. 80r; col. 1] A abbot. ο abbey ablatiue-case abortiue abricot-apple absence abstinance abstract abuse. c accent. acceptance accident. accompany accoumpt accusation ache acorne aquaintance aquittance acre action oth law actiuity d adamant ston addition adder adiectiue adiunct administration admirall seas

admiration admission admonition adoption aduers aduersary aduersity adulery [?] aduocate aduowson


affection affirnation [sic] affliction g age agrauation aglet aguement ague h i [illegible] aide aiune [?] aied [?] 1 alabaster albe alchymy alderman alder-tree ale allegory alley allowance allusion almes almond aloes alpes alpha alphabet altar allum m ambassage amber

The development of monolingual English ambition ambling pace [col. 2] Β [col. 3] C [col. 4] D diminioh destruction [col. 5] Ε [col. 6] F foroaking forowcaring fooliohnooo [fol. 80v; col. 1] G [col. 2] Η humble [col. 3] I [col. 4] Κ

pardon parfcction [col. 2] [col. 3] Q [col. 4] R [col. 5] S auffcring. [col. 6]


268 Ian Lancashire

[col. 7] Τ

tarrying [col. 8] V

[fol. 81v; col. 1] W


[col. 2] X. Y. Z.

[col. 3] A

ambushment amesse robe amphitheatre amphilology

angle rod anguish aniseed anker anniuersary annuity answeare anticrist antidote antick -- work antidote antiquity antipathy anuled


analogy analosys anauely anatomy anatomy ad [?] ancestors ancle andiron angell anger

ape aphourisme apology aposed apothecary apparrell appendix appendant appetite applause apple application appointing apprehension apprentise approach approbation appurtenance apron -q -R

Notes 1.

For the catalogue entry, see Bodleian Library [1953, III: 304]; and for a reproduction of the manuscript, Bodleian Library [1988: reel 9], The research reported here was supported by a grant from the Social Sciences and Humanities Research Council of Canada. I would like to thank the anonymous referees of my article, and the editors, for their valuable criticisms.

The development of monolingual English lexicons



These are perhaps the Elinor born on April 12, 1579, and the Edward born on Jan. 1, 1565, to John Gunter and Alice Keeblewhite, of Blewbury, Berkshire. The last known owner of this manuscript was William Oldisworth (16801734), an antiquarian and collector. 3. Jones (1953: 169) argues that, by the last quarter of the sixteenth century, "No longer was the vernacular only a practical instrument, the efficacy of which depended upon simple clarity and humble plainness; it was, instead, a free medium of expression, in which brave new words and elaborate figures could puzzle or displease whom they would." In my opinion, Jones underestimates achievements of the English language under the early Tudors. 4. This transcription uses old spelling, expands abbreviations (in italics), normalizes word spacing, and places editorial comments (such as noting of illegible letters or words) within square brackets. Cancellations are uniformly represented by striking through. The indexer's work, written in a larger, heavier script than the glossarian's, is set off in bold. Soft hyphens introduced into this transcription by the formatting of this article are noted by a [*]. I am grateful to Duke Humphrey's Library in the Bodleian Library for access to the original manuscript. The transcription will form part of Lexicons of Early Modern English (LEME), an online database to be published jointly by the University of Toronto Press and the University of Toronto Library. LEME builds on the Early Modern English Dictionaries Database (1996-1999). 5. Rastell's "Exposition" (fols. 85r-87r) is adopted from a medieval list found in many fourteenth-, fifteenth-, and sixteenth-century manuscripts, usually entitled "Exposicio vocabulorum." See, for example, British Library Arundel MS 221, fols. 12v-13r, with charters of Sibton Abbey, Suffolk. Baker and Ringrose itemize dozens of copies from most major libraries (1996). 6. Graham (1954: 22) attributes to John Rastell the "establishment of systematic English lexicography." 7. The first OED citation for the word dictionary, citing that of Peter Bercharius, is to Pilgrymage of perfeccyon (1526). 8. Information about his life and career can be found in Geritz and Laine (1983). Salmon (1989) describes Rastell's work in fostering literacy in English.

References Alston, Robin Carfrae (comp.) 1991 Handlist of unpublished finding aids to the London Collections of the British Library. The British Library. Baker, John Hamilton, and Jane S. Ringrose 1996 A Catalogue of English Legal Manuscripts in Cambridge University Library. Woodbridge: Boydell Press.

270 Ian Lancashire Banckes, Richard A newe mater / the whiche sheweth and treateth of ye vertues & proprytes of herbes / the whiche is called an Herball. London: Rycharde Banckes. Bliss, Philip (P. B.) 1812 Some account of a manuscript in Dr. Rawlinson's collection in the Bodleian Library. In The British Bibliographer, Sir Egerton Brydges, K. J., and Joseph Haslewood (comps.) Vol. II, 609-18. London: T. Bensley for R. Triphook. Blount, Thomas 1656 Glossographia. London: Thomas Newcomb for Humphrey Moseley and George Sawbridge. Bodleian Library, Oxford 1953 A Summary Catalogue of Western Manuscripts in the Bodleian Library at Oxford. Oxford: Clarendon Press. 1988 British Literary Manuscripts from the Bodleian Library, Oxford. Series One: Part One, The English Renaissance, ca. 1500-C.1700. Harvester Microform. Bullokar, John 1616 An English Expositor. London: John Legatt. Callery, B. G. 1978 Ancients, moderns and reliable friends: stated sources of some English herbals. Journal of the Society of the Bibliography of Natural History 8 (4): 4 3 5 ^ 4 4 . Cawdrey, Robert 1604 A Table Alphabetical!. London: J. Roberts for E. Weaver. Chambers, Edmund Kerchever 1923 The Elizabethan Stage. Oxford: Oxford University Press. Cockeram, Henry 1623 English Dictionarie: or, an interpreter of hard English words. London: Eliot's Court Press for N. Butter. Coote, Edmund 1596 The English Schoole-maister. London: Widow Orwin for Ralph Jackson and Robert Dexter. Cowley, John D. 1932 A Bibliography of Abridgments, Digests, Dictionaries and Indexes of English Law to the Year 1800. London: Seiden Society. Elyot, Sir John 1538 The Dictionary of syr Thomas Eliot. London: T. Bertheleti. Galfridus Grammaticus 1499 Promptorium Parvulorum. London: Richard Pynson. Geritz, Albert J., and Amos Lee Laine 1983 John Rastell. Boston: Twayne. 1525/26

The development of monolingual English lexicons


Graham, Howard Jay 1954 The Rastells and the printed English law book of the Renaissance. Law Library Journal 47: 6-25. Jones, Richard Foster 1953 The Triumph of the English Language: A Survey of Opinions Concerning the Vernacular from the Introduction of Printing to the Restoration. Stanford: Stanford University Press. Lancashire, Ian 1996-99 Early Modern English Dictionaries Database. University of Toronto: Computers in the Humanities and Social Sciences. URL:

'Dumb significants' and Early Modern English definition. In Literacy, 2002 Narrative and Culture, Jens Brockmeier, Min Wang, and David R. Olson (eds.), 131-154. Richmond, Surrey: Curzon. M„ W. The Queens Closet Opened. Incomparable Secrets in Physick, 1655 Chirurgery, Preserving. Candying, and Cookery. London: Nathaniel Brook. Mulcaster, Richard 1582 The first part of the elementarie. London: Thomas Vautroullier. Nowell, Laurence 1952 Vocabularium Saxonicum. Albert H. Marckwardt (ed.). Ann Arbor: University of Michigan Press. Osselton, Ν. Ε. 1986 The first English dictionary? A sixteenth-century compiler at work. In The History of Lexicography, R. R. K. Hartmann (ed.), 175-184. Amsterdam: J. Benjamins. Palsgrave, John 1530 Lesclarcissement de la langue francoyse. London: R. Pynson and J. Haukyns. Rastell, John 1523 Exposiciones /ermmorum legum anglomm. London: John Rastell. 1527 Thestatutes. London: John Rastell. Rider, John 1589 Bibliotheca Scholastica. Oxford: Joseph Barnes. 1612 Riders Dictionarie Corrected, and with the Addition of Aboue Five Hvndred Words enriched, Francis Holyoke (comp.) 3rd ed. Oxford: Joseph Barnes. Ridley, Mark 1996 A Dictionarie of the Vulgar Russe Tongue. Gerald Stone (ed.) Köln: Böhlau.

272 Ian Lancashire Salmon, Vivian 1989 John Rastell and the normalization of early sixteenth-century orthography. In Essays on English Language in Honour of Bertil Sundby, Leiv Egil Breivik, Arnoldus Hille, and Stig Johansson (eds.), 289-299. Oslo: Novus Forlag. Schäfer, Jürgen 1989 Early Modern English Lexicography. 2 vols. Oxford: Clarendon Press. Sherry, Richard 1550 A treatise of schemes & tropes. London: John Day. Treveris, Peter 1526 The grete herball. London: Peter Treveris. Wellisch, Η. Η. 1978 Early multilingual and multiscript indexes in herbals. Indexer 11: 81-104.

Section 3 Constraint-based studies

Introduction: Constraint-based studies Anne Curzan and Kimberly Emmons

The constraint-based studies in this section have two distinct foci: constraints on early English meter and optimal theoretic constraints on early English syntax and Proto-Germanic as well as Early Modern English phonology. Working with evidence from early written texts, the papers in this section continue the conversation about the intersection of philological and linguistic analysis in historical English linguistics. In the lead article of the section, "The evolution of Middle English alliterative meter," Geoffrey Russom, an expert on early English meter, extends his important work on word-foot theory to argue that, as it moved into the Middle English period, Old English meter profoundly changed through reanalysis, rather than through re-invention. His analysis stresses the continuous evolution of English alliterative meter in the period. Continuing a scholarly conversation among such scholars as Hoyt Duggan, Thorlac Turville-Petre, and Thomas Cable, Russom proposes four rules of Middle English alliterative meter and demonstrates that the two critical constraints, which are traditionally viewed as innovative to Middle English, can already be found as strong tendencies in particular Old English verse types. He relates these changes in verse form to contemporaneous structural changes in the language, including the loss of inflectional endings and the decline in productivity of compounding, specifically in this case of poetic compounding as a verse craft. Russom turns to Old Saxon for evidence of a possible transitional stage in this evolutionary process. Robert Fulk, whose work on Old English meter is foundational to the field, responds to Russom's analysis with additional evidence for a continuous tradition. Fulk focuses particularly on the central position and role often attributed to /Elfric's work, arguing that his "celebrity" is an unjustified result of the lack of identifiable authors in the period: the few authors that we can identify become central figures in part because we know their names. It is something of a "leap of faith" to derive popular alliterative verse in Middle English from a highly learned tradition from Old English. Fulk's discussion treats critical questions in English historical linguistics, such as the limits of the textual record given the centrality of orality to

276 Anne Curzan and Kimberly Emmons

early English culture. In Russom's response, he agrees to disagree about the value of maintaining strict boundaries between verse and prose during the period in, for example, Kline's verse-like works, and it clearly remains a scholarly question open to further exploration. XingZhong Li, in "A central metrical prototype for English iambic tertrameter verse: Evidence from Chaucer's octosyllabic lines," argues that Chaucer's octosyllabic lines are metrically footed. As a result, all English iambic tetrameter can be assumed to share the same central prototype, which addresses language-meter interactions. To explain patterns and variation in those patterns, Li employs a theory of gradient metrical saliency. Through a systematic sampling of Chaucer's octosyllabic lines and detailed statistical analysis, Li examines the lines gradient conformity to abstract prototypes of meter. He concludes that metricality is related to well-formedness with gradient, violable principles. While only a side note, Li mentions that his analysis is consistent with Optimality Theory (OT), and in the subsequent two papers, OT takes center stage as an explanatory framework for two historical changes, one syntactic and one phonological. Brady Clark tackles the complex topic of shifts between Old and Middle English clause structure in "Early English clause structure change in a stochastic optimality theory setting." Stochastic optimality theory, which offers a new approach to gradual syntactic change, allows for gradual change with variation in intermediate stages. In this way, Clark's analysis will be of interest both to theoretical linguists focused on OT and historical sociolinguists interested in 'S'-shaped language change. Clark argues that a theory of constraints and an understanding of the reranking of those constraints explains the shift in tendencies to use particular clause structures as well as the absence of one theoretically possible clause structure, the reverse brace construction. Clark's analysis, which works with compiled evidence from early English texts, demonstrates what it can look like to bring philological data and cutting-edge linguistic theory into conversation. And his results open the possibility of sociolinguistic analysis as well. Olga Petrova, in "The role of perceptual contrast in Verner's Law," returns to the long-standing question of the mechanisms involved in Verner's Law (VL), which resulted in voiced allophones in Proto-Germanic of voiceless fricatives in post-atonic environments in Proto-Indo-European. Petrova makes the very interesting move of equating this phonological change to what she calls "seventeenth century VL," which also results in voiced fricatives, in this case in Romance borrowings. With the data from these two historical changes, as well as from cross-linguistic studies by

Section 3: Introduction


other scholars, Petrova suggests that stress shift may have induced fricative voicing - as opposed to the traditional view that VL occurred before the shift in stress. The stress shift caused a perceptual interpretation of these medial fricatives as voiced. Citing her earlier work on Grimm's Law, Petrova argues that both sound changes involve strategies to avoid phonological merger. In a move of interest to historical linguists and theoretical linguists alike, Petrova examines the data within an optimality theoretic perspective, arguing that this historical data provides more evidence for the interaction of faithfulness and dispersion, which must be integrated. The analysis relates a pre-Germanic sound change to a similar sound change several millennia later in the history of English and applies current insights from OT to a long-standing scholarly conversation about sound change in and before - the history of English.

The evolution of Middle English alliterative meter Geoffrey Russom

1.1. Rules for the Middle English b-verse In the mid-1980s, Thomas Cable and Hoyt Duggan, working independently, discovered two important rules of Middle English alliterative meter (Cable 1988; Duggan 1988). These rules apply to the closing verse of the alliterative long line, called the b-verse. No rules of such strictness have been discovered for the opening verse of the line, called the a-verse. According to the first rule, a b-verse must contain a long dip, defined as a group of two or more unstressed syllables. I will call this the long-dip requirementThe second rule, which I will call the long-dip restriction, permits only one long dip in a b-verse. The fact that the last verse of the line is more strictly regulated can be attributed to a universal principle of closure (Hayes 1983: 373), which requires stricter adherence to metrical norms at the end of a metrical unit. In addition to complying with rules for long dips, the b-verse must have exactly two stresses and must end with exactly one unstressed syllable. I will refer to these rules as numbered and formulated below: Rule 1. Rule 2. Rule 3. Rule 4.

A b-verse must contain a long dip (long dip requirement). A b-verse must not contain more than one long dip (long dip restriction). A b-verse must contain exactly two stressed syllables. A b-verse must end with a trochaic constituent.2

1.2. Departures from Old English metrical practice Rules 1 and 2 are clearly innovations, yielding b-verses significantly different from those favored by Old English poets. Compare the following two Middle English examples from Sir Gawain and the Green Knight with some prototypical examples from Beowulf:

280 Geoffrey Russom (1)

and teldes bigynnes and dwelling-ACC.PL begin-3.SG 'and sets up dwellings' (SGGK 1 lb)



vnder Krystes seluen under Christ-GEN.SG. self-ACC.SG 'under Christ himself (SGGK 51b)



beorhte / frcetwe bright-ACC.PL treasure-ACC.PL 'bright treasures' (Beo 214b)

Sx/Sx (A)


on / morgentld x/Sxs (B) in morningtime-ACC.SG 'in the morning' (Beo 484b)


in / geardagum in yoreday-DAT.PL 'in days of yore' (Beo lb)

x/Ssx (C)


feond / mancynnes enemy-NOM.SG mankind-GEN.SG 'enemy of mankind' (Beo 164b)

S/Ssx (D)


beahhorda / weard ringhoard-GEN. SG guardian-NOM.SG 'guardian of the ring-hoard' (Beo 921b)

Ssx/S (E)


(ond) his / ellenweorc and his valordeed-ACC.PL 'and his brave deeds' (Beo 3173b)

x/Sxs (B)

In scansions to the right of the examples, "x" represents a weak metrical position normally occupied by an unstressed syllable. "S" represents a strong position normally occupied by a syllable with primary stress. Lower-case "s" represents a position normally occupied by a syllable with subordinate stress. The slash employed in items (3-8) notates the boundary between the two feet of an Old English verse. In items (3-8), the tradi-

The evolution of Middle English alliterative meter 281 tional verse-type designations of Sievers (1893) appear between parentheses following the scansions. The verse patterns represented by examples (1) and (2), which are common in Sir Gawain, have extremely low frequency in Beowulf. Example (3), a natural syntactic constituent of two trochaic words, represents the favored realization of the favored verse pattern in Beowulf, but lacks the long dip required in Sir Gawain. Examples (4-7) represent the remaining Old English verse types identified by Sievers. These examples have a trisyllabic compound or derivative in their twoword realizations, which are highly favored in Beowulf but violate rule 1, like example (3). To fill the longer foot in verses like (4-7), the Beowulf poet frequently employs a type of compound that is rare in prose.4 In Middle English alliterative verse, this type of compound is no longer employed and ordinary compounds appear infrequently. Old English meter allows for addition of extrametrical unstressed words, as in (8), a type Β variant with the extrametrical word in parentheses. This verse conforms to rules 1 and 2, but comparable examples are rare in Sir Gawain due to restricted employment of compounds. Turville-Petre (1977: 16-17) argues that Middle English alliterative meter developed independently during the later Middle Ages after a complete break with the Old English tradition. Duggan (1988: 145) finds such a hypothesis difficult to reconcile with widespread adherence to rules 1 and 2. If the common metrical tradition had been lost, we would expect a wider variety of alliterative verse forms in the Middle English period. Here I will argue that Old English meter was indeed profoundly changed, but through reanalysis rather than reinvention.

2.1. The word-foot theory of Old English meter To explain the distribution of unstressed syllables in Old English meter, I will employ the word-foot theory (Russom 1987, 1998), which consists of two fundamental principles, expressed below as equations: Principle 1. Corollary:

Foot Pattern = Word Pattern Feet corresponding to common word patterns are easiest to identify during real-time scansion, and the normative foot has the paradigmatic word pattern (trochaic).

282 Geoffrey Russom Principle 2. Corollary:

Alliterative Binding Rule = Compound Stress Rule Since linguistic compounding is a binary operation joining two lower-level constituents into one higherlevel constituent, alliterative binding of feet into verses and of verses into lines will proceed in a binary fashion, yielding verses of two feet and lines of two verses.

Principles 1 and 2 define an Old English verse as a metrical constituent with two word feet, where each word foot has the phonological pattern of a native Old English word. Word feet can be projected from all native words, including unstressed words and compounds as well as stressed simplexes. Within this theoretical framework, the basic realization or paradigm of a verse type is a pair of words. Items (3-7) are paradigmatic in this sense. Two kinds of deviation from type paradigms are allowed, subject to constraints that ensure recoverability of the basic two-word pattern (Russom 1998: 52-4, 87-96). First, extrametrical unstressed words may be added, as in (8). Secondly, a polysyllabic word may be replaced by a word group with the same stress contour, as in the following examples from Beowulf. (9)

feond on fiend-NOM.SG. in 'a fiend in hell' (101b)


helle hell-DAT.SG


on / hreon in rough 'in a bad mood' (1307b)

mode mood-DAT.SG


s&gon folc / to folk-NOM.PL on look-3.SG.PRET 'the people looked on' (1422b)

Sx/Sx (A)

xJSsx (C)

S/Ssx (D)

In (9), an Sx foot is realized as a word group rather than as a bisyllabic simplex. Items (10) and (11) illustrate realization of an Ssx foot as a word group. Such deviations from a type paradigm add to the metrical complexity of a verse and may impose conspicuous restrictions on its employment in the second half of the line, as we shall see. A verse may also be complex because its metrical pattern is inherently deviant. The pattern Sx/Sx, represented by item (3), has normative status because it consists of two feet with

The evolution of Middle English alliterative meter 283 the normative Old English word pattern (see Principle 1). Other verse types are complex to the extent that they deviate from the norm of two stresses and four metrical positions established by the Sx/Sx pattern, which is subcategorized by Sievers as type Al. Relative to a type Al norm, the type D and Ε verses represented by (6) and (7), which have three stresses, are abnormally heavy.

2.2. Constraints on unstressed syllables in the word-foot theory Fundamental principles of Old English meter impose a number of constraints on unstressed syllables. We turn attention first to constraints on realization of a light foot projected from an unstressed function word. This type of foot is most often used to offset the weight of a heavy compound foot in type Β or C, as in items (4) and (5). The light foot gives these types normal weight, and they are employed with high frequency, as expected.5 Unstressed prefixes provide interesting borderline cases for a word-foot theory. In Old English, these closely bound proclitics originate as function words. Most of them still look like prepositions and they still behave like words from a phonological point of view. If unstressed prefixes shared a word-level domain with the stressed words to which they are bound, for example, phonological rules with detectable consequences would apply to them; but such rules do not in fact apply (Russom 1987: 8-9). From a syntactic point of view, on the other hand, unstressed prefixes are word parts, and in this respect they make problematic word feet. As it turns out, the Beowulf poet does employ these problematic word feet, but in an extremely restricted way, as the borderline status of unstressed prefixes would lead us to expect. (12)

ä-/bredwade kill-3.SG.PRET 'he killed' (2619b)

x/Sxx (C)


(ful oft) ge-/beotedon full often boast-3.PL.PRET 'very often they boasted' (480a)

x/Sxx (C)

The Beowulf poet occasionally uses an isolated unstressed prefix for a light foot, as in (12), but clearly prefers not to do so. Variants like (12)

284 Geoffrey Russom

have low frequency and restricted placement (Russom 1998: 46-47). Normally, a prefix bound to a stressed word in the second foot is supported by extrametrical unstressed words that create a long dip. In (13), the extrametrical words seem to be added primarily for metrical reasons. As far as I can tell, nothing in the narrative context makes it pertinent to specify that the Danes boasted "very often." Adjoining extrametrical words to a light foot makes it more conspicuous, ensuring that it will be recognized by the audience as one of exactly two feet mandated by the rules of verse form. This is important because in certain special cases, as we shall see, unstressed syllables before the first stress of the verse must not be interpreted as light feet. The distribution of long dips is restricted in part by Principle 1, which requires every foot to conform to the pattern of an Old English word. Although extrametrical words can sometimes be added before a foot, adding them inside a foot would disrupt the word pattern. Creation of a long versefinal dip with extrametrical words is blocked by a principle of syntactic integrity, according to which the verse must be filled by a natural syntactic constituent (Russom 1998: 30). Any Old English function word added to the end of a constituent would acquire phrasal stress and could not be extrametrical for that reason.6 Thus in types Β and C, Old English metrical constraints, interacting with purely linguistic constraints, favor verses with one and only one long dip.

2.3. Unstressed syllables in Old English types B, C, and A3 The tendency towards a single long dip operates most strongly in the most common Β and C subtypes. Below are representative examples for type B, with items (14) and (15) repeating items (4) and (8): (14)

on / morgentld x/Sxs (83x) in momingtime-ACC.SG 'in the morning' (484b )


(ond) his / ellenweorc and his valordeed-ACC.PL 'and his brave deeds' (3173b)

x/Sxs (752x)

The evolution of Middle English alliterative meter 285 (16)

on / feonda geweald into fiend-GEN.PL power-ACC.SG 'into the power of fiends' (808a)

x/Sxxs (15x)


(scolde) his / aldorgedäl x/Sxxs (158x) must-3.SG.PRET his lifedivision-NOM.SG 'his loss of life was destined' (805b)

A strong tendency to support the light foot with extrametrical syllables can be observed in type B. Examples like (14) and (16), with the minimum number of syllables, account for only 98 of the 1,008 type Β verses in Beowulf. In the longer Β subtype x/Sxxs, which has an xx group as part of its metrical pattern, support for the light foot with extrametrical syllables will violate Middle English rule 2, the long-dip restriction. The x/Sxxs subtype has relatively low frequency, however, with a total of 173 instances, as compared with 835 total instances for the subtype x/Sxs, in which no violation of rule 2 can occur. The x/Sxs subtype has higher frequency because its Sxs foot corresponds to a more familiar word pattern and adds less to verse complexity (see Principle 1). Minimal realizations of the x/Sxs pattern, represented by (14), violate the long-dip requirement (rule 1), but these are very much in the minority (83:752). The situation is similar in type C, which also has a subtype with an xx group in the second foot of its metrical pattern (x/Sxx). (18)

in / geardagum in yoreday-DAT.PL 'in days of yore' (lb)

x/Ssx (260x)


(pcet) wees / god cyning x/Ssx (840x) that be-3.SG.PRET good-NOM.SG king-NOM.SG 'that was a good king' (1 lb)


swä / rixode thus rule-3.SG.PRET 'thus he prevailed' (144a)

x/Sxx (3x)


(ic) eow / wisige I you-DAT.PL guide-l.SG Ί will guide you' (292b)

x/Sxx (19x)

286 Geoffrey Russom In Beowulf, there are 260 violations of rule 1 like (18), which repeats item (5), and 19 violations of rule 2 like (21), for a total of 279 violations, as compared with 843 type C verses that comply with both Middle English rules. Rule violations predominate only in the rare x/Sxx subtype, which has lower frequency because the Sxx foot is generated from an uncommon word pattern attested in certain forms of weak class Π verbs.7 In the predominant subtype x/Ssx, violations of rule 1 are considerably outnumbered by verses that conform (260:840). The remaining Old English verse patterns with a light foot are classified by Sievers under type A3: (22)

poet se / mcera that the famous-NOM.SG 'that the famous one (wished)' (2587a)


(efne swä) hwylcum / even so which-DAT.SG 'to whichever of men' (3057a)


(poet he) pone/ grundwong that he the seabottom-ACC.SG 'that he [saw] the sea-bottom' (2770a)

xx/Sx (8x)

manna xx/Sx (185x) man-GEN.PL

xx/Ss (23x)

In all but eight of the 216 type A3 verses, the light foot is supported by extrametrical syllables. The xx/Sx pattern, with just one stressed syllable, contravenes Middle English rule 3, which requires two stresses in a bverse.8 This pattern is also abnormally light by Old English standards and is confined to the a-verse by the principle of closure, which restricts metrical complexity in the b-verse. Exclusion of type A3 from the Middle English b-verse follows as a matter of course from the hypothesis that the meter evolved continuously, without any radical break.

2.4. Unstressed syllables in Old English types Al, A2, D, and Ε We turn now to verse types beginning with an S position rather than an χ position. In these types, violations of rule 1 are more common but violations of rule 2 are extremely rare. The pattern of highest frequency is

The evolution of Middle English alliterative meter


Sx/Sx, Sievers's type Al. About 32% of the verses in Beowulf have this pattern. Since type Al has no inherent complexity, it allows for a variety of deviations from its two-word paradigm, and there are several cases to consider. Also classified under type Al by Sievers is a low-frequency variant with an xx group in the first foot, which I include below: (25)

sinces / bryttan treasure-GEN. SG distributor-ACC.SG 'distributor of rings' (1922b)

Sx/Sx (1317x)


sunnan / (ond) mötian sun-ACC.SG and moon-ACC.SG 'the sun and moon' (94b)

Sx/Sx (64 lx)


[ge-Jsfgan (cet) / scecce defeat-INF at combat-DAT.SG 'to defeat in combat' (2659a)

Sx/Sx (38x)


fne not

ge-Jfeah he / rejoice-3.SG.PRET he (päre) fiehöe the feud-GEN.SG 'he did not rejoice about that feud' (109a)


preatedon / pearle oppress-3.PL.PRET severely 'they oppressed severely' (560a)


brüc penden / (pü) enjoy-IMP while you-SG 'enjoy while you may' (1177b)


[on-Jßh pissum / receive-IMP this 'receive this cup' (1169a)

Sx/Sx (3x)

Sxx/Sx (72x)

möte Sxx/Sx (4x) may-2.SG.SUB

fülle cup-DAT.SG

Sxx/Sx (lx)

In Beowulf, about one-third of the Sx/Sx verses have extrametrical syllables before the second foot, satisfying the long-dip requirement (rule 1). All verses with the Sxx/Sx pattern satisfy rule 1, since they have an xx

288 Geoffrey Russom

group as part of the metrical pattern. Addition of extrametrical material is allowed under certain conditions at the beginning of type Al verses. In examples (27), (28), and (31), extrametrical unstressed syllables at the beginning of the verse are placed between square brackets. The very low frequency of verses like these indicates that they deviate very significantly from the two-word paradigm. There are only 42 instances among the 2,076 type Al verses in Beowulf. The exceptional placement of the bracketed constituents is indicated by the traditional term anacrusis, which is not applied to the extrametrical words that appear so frequently at the beginning of a type A3, B, or C verse. In most cases, anacrusis consists of a closely bound proclitic, usually an unstressed prefix or a negative preverb. Use of independent function words in anacrusis is kept to a minimum. This constraint on anacrusis stands in complementary relation to the constraint on the light foot, which is realized as an isolated prefix only in exceptional cases like (12). Constraints on the distribution of isolated prefixes and independent function words minimize confusion about the number of feet in the verse.9 The maximum extent of anacrusis is determined by the maximum size of a prefix. Since a few bisyllabic prefixes are present in Old English, bisyllabic anacrusis is a possibility, though a marginal one. Item (28) is one of just three comparable instances, and these three are the only type Al verses in Beowulf that violate rule 2, the long dip restriction. The anacrusis in (28) is particularly complex, with a negative preverb and a monosyllabic prefix substituted for a bisyllabic prefix. The two remaining patterns of normal weight, classified by Sievers under type D, have an xx group in the second foot: (32)

stig / wisode path-NOM.SG direct-3.SG.PRET 'the path led the way' (320b)

S/Sxx (105x)


oftost / wisode often-SUPERL direct-3.SG.PRET 'most often he led the way' (1663b)

Sx/Sxx (28x)

The xx group in this type satisfies rule 1. Since no examples in Beowulf have bisyllabic anacrusis, there are no violations of rule 2. The patterns classified by Sievers under type A2 are heavy, exceeding the norm of two stressed syllables established by type Al. 10 Such patterns

The evolution of Middle English alliterative meter 289 have one or two Ss feet occupied by a bisyllabic compound or equivalent word group: (34)

heardecg / togen Ss/Sx (245x) hardedge-NOM.SG draw-PRET.PARTICIPLE 'a hard-edged [sword] drawn' (1288b)


güdrinc / goldwlanc battleman-NOM. SG goldproud-NOM.SG 'gold-adorned warrior' (1881a)

Ss/Ss (33x)


Grendles / güdcroeft Grendel-GEN.SG warcraft-NOM.SG 'Grendel's war-skill' (127a)



Tsig / (ond) ütfüs Sx/Ss (39x) icy-NOM.SG and outyearning-NOM.SG 'icy and yearning to set out' (33a)


[he he

ge-Jfeng pa/ take-3.SG.PRET then fetelhilt ringhilt-ACC.PL 'he took the ring-adorned hilts' (1563a)

Sx/Ss (lx) 11

Heavy verses are inherently complex and deviation from their two-word paradigms is accordingly restricted. Use of extrametrical words is infrequent in heavy verses. Since A2 subtypes have no xx group in their minimal realizations, most A2 verses have no long dip. In Beowulf there are only 39 instances like (37), with a long dip verse-internally. Example (38) is the only instance with a long dip verse-initially, in anacrusis. Thus rule 1 is satisfied by only 40 of the 349 type A2 verses in the poem. In this type, violation of rule 2 would require bisyllabic anacrusis and one or two extrametrical words before the second foot, an unlikely combination when employment of extrametrical constituents is restricted. No such violation appears in Beowulf. Long dips are also rare among the remaining heavy verses of normal length, which include variants of types Da (S/Ssx), Db (S/Sxs), and Ε (Ssx/S):

290 Geoffrey Russom (39)

feond / mancynnes enemy-NOM.SG mankind-GEN.SG 'enemy of mankind' (164b)

S/Ssx (343x)


deorc / (ofer) dryhtgumum dark-NOM.SG over troopman-DAT.PL 'dark over the men of the troop' (1790a)

S/Ssx (5x)


här / hilderinc hoary battleman 'old warrior' (1307a)

S/Sxs (137x)


fofer-Jfleon / Jotes trem S/Sxs (lx) overflee-INF foot-GEN.SG space-ACC.SG 'to flee a footstep' (2525a)


häpenra / hyht heathen-GEN.PL hope-NOM.SG 'hope of the heathens' (179a)

Ssx/S (40lx)


wincernes / (ge-)weald winehall-GEN.SG control-ACC.SG 'control of the wine-hall' (654a)

Ssx/S (42x)

Only 48 of the 929 verses in this set have the long dip required by rule 1. Variants like (40) and (44) have a long verse-internal dip created by employment of extrametrical words before the second foot. In (42), the long dip is created by bisyllabic anacrusis. No verses in this set have bisyllabic anacrusis and a long verse-internal dip, so there are no violations of rule 2. The remaining verse types are both long and heavy. These include expanded type Da (Sx/Ssx), type Db with a long second foot (S/Sxxs), and expanded type Db (Sx/Sxs, Sx/Sxxs): (45)

märe / mearcstapa Sx/Ssx (133x) famous-NOM. SG marchstepper-NOM. SG 'notorious border-stalker' (103a)


wongas / (ond) wicstede field-NOM.PL and homestead-NOM.PL 'fields and homesteads' (2462a)

Sx/Ssx (4x)

The evolution of Middle English alliterative meter






S/Sxxs (19x)

see-INF sibtroop-ACC.SG 'to see the kindred band' (387a) (48)




Sx/Sxs (33x)

person-GEN.PL landwork-ACC.SG 'the people's stronghold' (938a) (49)






Sx/Sxs ( l x )

fall-3.SG.PRET then 'he fell then, weary' (1543a)






grief-NOM.SG earl-GEN.PL 'grief to each earl' (1420a)


[of er-] swam overswim-3.SG.PRET sioleöa



Sx/Sxxs (2x)

then bigong

Sx/Sxxs (lx)

water-GEN.PL expanse-ACC.SG 'then he swam over the expanse of the waters' (2367a) Within this set of 193 long, heavy verses, rule 1 is satisfied in 27 instances and rule 2 is satisfied in 192 instances. Rule 1 is satisfied in (47) and (50) by virtue of an xx group in the basic pattern. In (46), rule 1 is satisfied by an extrametrical word internal to the verse. In (49), rule 1 is satisfied by a bisyllabic prefix in anacrusis. Example (51), with an xx group and bisyllabic anacrusis, shows a rare violation of rule 2 in a heavy verse.

2.5. Summary of Old English constraints Generalizing over the whole array of verse patterns, we observe that long dips are common in verse-initial or verse-internal position but seldom occur in both positions. Rule 2 is already established as a strong tendency during the Old English period. Rule 1 is established as a strong tendency in Old English verse types beginning with a light foot. In verse types begin-

292 Geoffrey Russom ning with an S position, instances complying with rule 1 are less common but comprise a substantial minority in subtypes with normal weight, which include the predominant type Al. The most deviant types, from a Middle English perspective, are the heavy patterns, which have a very low frequency of compliance with rule 1.

3.1. Decline of compounding and increased use of function words A scenario taking us from the meter of Beowulf to the meter of Sir Gawain can be constructed by reference to just two of the familiar linguistic changes that transformed Old English into Middle English. The first change, mentioned briefly above, was a decline in the productivity of compounding and its metrical reflection, loss of poetic compounding as a technique of versecraft. This eliminated two-word realizations of types A2, B, C, D, and E, obscuring their word-foot structure. Substitution of a word group for a compound had always been restricted in the heavy A2, D, and Ε patterns, which allow only limited deviation from two-word paradigms, as we have observed. Loss of the two-word realizations would have left a small, anomalous remnant of heavy variants; and these would soon be excluded from the b-verse by the principle of closure, a universal tendency persisting throughout the transition. As a result, rule 1 violations, which are characteristic of heavy types, would occur much less frequently in the b-verse. The deviant type A3, with just one stress, had always been excluded from the b-verse by Old English poets. Exclusion of heavy types from the b-verse would remove all remaining exceptions to rule 3, which requires exactly two stresses. A second familiar historical development, decay of the inflectional system, made it more difficult to omit unstressed words for metrical purposes. A traditional Old English poet could use instrumental inflection of the adjective, for example, to obtain two-word type Al paradigms like title werode 'with a small troop'. This useful archaic construction can still be found in a relatively late example of traditional versecraft, The Battle of Brunanburh, at verse 34a (Dobbie 1942: 18). In jElfric's less traditional late verse, however, the old formula is modernized to mid lytlum werode, with dative inflection and the preposition mid expressing the concept 'with'. 12 If it appeared in Beowulf ^Elfric's modernized formula would arouse editorial suspicion as a type Al variant with peculiar anacrusis.

The evolution of Middle English alliterative meter


3.2. Verse-initial dips in a late Old English poem Variants with nontraditional anacrusis have become commonplace in very late verse of an otherwise traditional character, notably The Battle of Maldon, an incomplete poem of which 648 verses survive.13 The following examples from Maldon show independent function words appearing as anacruses in type Al: (52)

[his] ealdre / (ge-)lästan his lord-ACC.SG support-INF 'to support his lord' (1 lb)


[he] lihte/ φα mid) he alight-3 .SG.PRET then amid 'he alighted then among the people' (23a)


[mid] gafole / (for-)gyldon with tribute-DAT.SG purchase-INF 'to pay off with tribute' (32b)


[and] georne / (ge-)säwon and readily see-3.PL.PRET 'and they readily saw' (84b)


[se] eorl wees / (pe) the earl-NOM.SG was that-INST 'the earl was happier because ofthat' (146b)


[pe] ähte / (his) which own-3.SG.PRET his 'which his lord had owned' (189b)

leodon person-DAT.PL

blidra happier-NOM.SG

hläford lord-NOM.SG

Here we find anacruses consisting of a possessive pronoun (52), a personal pronoun (53), a preposition (54), a conjunction (55), a definite article (56), and a relative pronoun (57). Most of these examples are b-verses. All of them have a long verse-internal dip and none of them has bisyllabic anacrusis, so there are no violations of rule 1 or rule 2. Seven additional verses with nonprefixal anacrusis can be found in the relatively short Maldon·.

294 Geoffrey Russom

68b (preposition), 96b (preposition), 136a (personal pronoun), 182a (conjunction), 185a (personal pronoun), 193a (conjunction), and 202b (definite article). Except for 202b, which lacks a long verse-internal dip, these also satisfy rules 1 and 2. In the more than 6,000 verses of Beowulf, by contrast, there are perhaps nine similar anacruses (25a, 93b, 107a, 666b, 1223b, 1248a, 1549a, 1987a, 2447a).

3.3. Verse-initial dips in an Old Saxon poem Just a few hundred lines of late traditional verse like Maldon have survived, complicating analysis of some important developments. It would be interesting, for example, to see how longer nonprefixal anacruses were managed in type A l , but at the stage of metrical evolution represented by Maldon such anacruses were apparently still rare. Fortunately, the more than 12,000 surviving verses in the cognate Old Saxon tradition provide useful analogues. Because Old Saxon had already developed characteristics of later English, including stricter constraints on omission of function words, Old Saxon meter illustrates developments expected for Old English meter after Maldon (Robinson 1992: 134-35). Consider the following variants of type Al from the Old Saxon Heliand: (58)

[an them is ] uuärun / in those his true-DAT.PL 'in those true words of his' (3939a)

uuordun word-DAT.PL


mildi / (uuas he mild-NOM.SG was he an is) in his 'he was well disposed to them'


fgi-J lestead / (an thesumu) perform-3 .SG.PRET in this 'performed in this life' (1626a)

im they-DAT.PL mode mind-DAT. SG (1259a) liohte light-DAT.SG

In (58), the bracketed anacrusis extends beyond the Old English limit of two syllables and includes independent function words that would not be

The evolution of Middle English alliterative



employed for anacrusis by the Beowulf poet. The Heliand contains 78 type Al verses like (58), more than one for every hundred lines. The complexity of long Old Saxon anacruses is mitigated by keeping the main part of the verse as simple as possible. In (58), for example, the main part, uuärun / uuordun, is a paradigmatic two-word realization of the Sx/Sx pattern. Among the 78 Al variants with a long anacrusis, 44 (56%) have this kind of main part. There are perhaps 11 Al variants with a long anacrusis and a main part of significant complexity (2-3a, 410a, 496a, 592a, 1662a, 3481a, 4974a, 4993a, 5088a, 5091a, and 5485a). Several of these are doubtful examples (Russom 1998: 149). The Heliand also has an elevated frequency of extrametrical words occupying the preferred traditional location in type Al, before the second foot. When a verse has this kind of complexity, anacrusis is normally avoided, as in (59), or abides by traditional restrictions, as in (60), where the anacrusis is a monosyllabic prefix. Attempts in the Heliand to manage an increasing number of function words bring us significantly closer to the Middle English rules for long dips. Long dips occur much more frequently in Old Saxon type Al, and expansion of one dip has a conspicuous inhibiting effect on expansion of other dips within the same verse. As we have observed, the Beowulf poet's most highly favored location for extrametrical words is before the first stress in types beginning with a light foot, which are represented in the bverse by types Β and C. In the Heliand, these types almost always contain an extrametrical word, further increasing the frequency of verses with one and only one long dip.14

3.4. Effects of metrical closure in a late Old English verse translation An important tendency in the evolution of alliterative meter is illustrated by Psalm 118 in the Paris Psalter, a set of Old English verse translations that is notorious for metrical irregularity (Krapp 1932: xvii). With 1,070 verses, Psalm 118, the longest poem in the set, is significantly longer than Maldon. Psalm 118 shows a dramatic change in the distribution of verses like beorhte/ frcetwe (item 3), the ideal, two-word realization of the normative Al pattern Sx/Sx. Throughout the Old English period, such two-word paradigms have high frequency in the b-verse due to the principle of closure, which favors assignment of the least complex verses to the closing half of the line. In Beowulf, 65% of the two-word type Al paradigms are bverses; in Psalm 118, the frequency jumps to 94%. The motive for this

296 Geoffrey Russom spectacular redistribution is to make room in the a-verse for complex new variants that are totally unsuited to the b-verse. Consider the following examples, which have no analogues in Beowulf. (61)

(on pinre) mild-fheortnesse in your-SG mildheartedness-DAT.SG 'in your benevolence' (159.3a)


(pine) sööfceste / your-SG truthfast-ACC.PL 'your trustworthy deeds' (83.4a)


(on pmum) egesan / in your-SG fear-DAT.SG 'first in fear of you' (161.4a)


weorc (6x) work-ACC.PL

ärest first


These peculiar verses are difficult to scan within systems designed for traditional poetry. I have scanned (61) as type Da (S/Ssx), (62) as type Ε (Ssx/S), and (63) as type Al (Sx/Sx). On this analysis, the 25 verses like (61-63) have nontraditional anacrusis and comprise all examples of nontraditional anacrusis in the poem. Although constraints on verse patterns are breaking down in Psalm 118, the poet is surprisingly sensitive to metrical complexity in variants like (61-63), placing all 25 examples in the averse, the appropriate location for verses that deviate markedly from the norm. I have found no variant of comparable complexity in the b-verse. When variants avoided in traditional poems appear in Psalm 118, they are situated in the a-verse.15

3.5. Effects of metrical closure in a late Old English heroic poem The Battle of Maldon employs traditional diction in a heroic narrative, differing markedly in these respects from Psalm 118. Yet the Maldon poet uses exactly the same strategy to make room in the a-verse for new complex variants. In Maldon, the frequency of two-word type Al in the b-verse is 96%, as compared with 94% for Psalm 118 and 65% for Beowulf. If this strategy was widely adopted, as its employment in two quite different poems suggests, it is easy to understand why verse rhythm is so free in the Middle English a-verse. As the Old English metrical system evolved, the

The evolution of Middle English alliterative meter 297 principle of closure would concentrate new complex variants in the first half of the line. Concentration of type Al in the second half of the line would enhance the frequency of b-verses ending with a trochaic constituent, which conform to Middle English rule 4.

3.6. Summary of metrical developments By the early Middle English period, the low frequency of compounds and the high frequency of unstressed function words would utterly obscure the two-word structure of Old English verse types. At this point, the value of two-word realizations would no longer be apparent, and idiomatic function words would be added to them, preferentially at first in locations where extrametrical words had been most common. Adding more function words in traditional locations would elevate an already high frequency of verses with a single long dip. When Middle English poets began to perceive such verses as normative, the principle of closure would concentrate them in the b-verse. In the final stage, increasingly strong tendencies would be reanalyzed by careful poets as rules - that is, as rules 1-4.

4.1. Cable's problematic formulation of the radical-break hypothesis Cable (1991: 42) argues that the meter of Sir Gawaiti developed out of something like yElfric's rhythmical prose, which allows for very free employment of unstressed words. In Old Saxon poetry, expansion of one dip has an inhibiting effect on expansion of others, but no such inhibition is apparent in rhythmical prose. The first fifty lines of /Elfnc's Natale Sancti Oswaldi, for example, include twelve b-verses with two long dips.16 On Cable's hypothesis, the Middle English rule system represents a "reestablishment" of control over unstressed syllables and must be characterized as a "surprising development" (Cable 1991: 51). On the hypothesis presented here, the Middle English system is exactly what we would expect. This account mandates exclusion of ^Elfric from the line of development that leads to Sir Gawain,17 Rhythmical prose illustrates what can happen when traditional rules are suspended, but does not provide a starting point for development of rules 1-4.

298 Geoffrey Russom 4.2. Problematic external evidence for the radical-break hypothesis The argument that poems like Sir Gawain follow a radical break with tradition seems to have been aimed in the first instance against an assumption that alliterative poets preserved Anglo-Saxon national feeling and correct Old English meter throughout the Middle English period. 18 Such an assumption certainly underestimates the extent to which metrical practice was transformed; and as for the supposed link between poetry and nationbuilding, Turville-Petre provides a wealth of evidence showing that alliterative poems in strict traditional meter were excluded from the cultural mainstream (1977: 35-36 and passim). Alliterative poems that entered the mainstream, from Lajamon's Brut to the late fourteenth-century Pierce the Ploughman's Creed, did so by abandoning the devices of traditional composition (Turville-Petre 1977: 13, 31). Given such active hostility to traditional devices within the literary establishment, it seems superfluous to propose additional explanations for the fact that no manuscripts containing strict traditional verse happen to survive from the period when French literature was overwhelmingly dominant.19 The external evidence seems consistent with an unbroken but evolving alliterative tradition that provided entertainment for moderately prosperous households in the North and West of England. 20 As Turville-Petre observes (1977: 1-2, 36-37), several poets represent their work as exactly this form of entertainment. In Sir Gawain, the author announces that the story he is about to narrate has a long history of alliterative performance in the land and that he has learned it by ear (lines 30-36). This passage may be a literary artifice inviting an audience of readers to participate imaginatively, but if so it presupposes that the kind of performance described was a familiar one. Turville-Petre's observations can be explained independently of his radical-break hypothesis.

4.3. Sudden death of the meter as evidence for traditional continuity The alliterative tradition went into surprisingly rapid decline after a century of major artistic achievement and significant regional popularity. First-rate poets seem to have abandoned the tradition by the early fifteenth century; a century later, its vestiges were restricted to a heartland in the Cheshire area.21 Some insight into this conundrum can be obtained from parallel developments in Greek and Indo-Iranian meters. At first these me-

The evolution of Middle English alliterative meter 299 ters seem to have been "natural," in that they were based on delimitation of word and phrase boundaries. This stage is attested by Avestan poetry. Then word and phrase boundaries were suppressed and the verse pattern evolved into an artificially ordered sequence of syllables with contrasting prominence, as in Homeric poetry (Kurylowicz 1970: 422). The word-foot theory represents Old English meter as "natural," in that the foot is projected from the word.22 By the fourteenth century, however, the meter had lost this direct link to language, as we have seen. Homeric meter could spread to Rome because the culture it embodied had high international prestige and because its single verse pattern was relatively easy to learn. Middle English alliterative meter had far less prestige outside its native region and learning it would have posed two major problems, since the verse patterns had remained quite diverse while becoming more artificial. It is difficult to imagine how anything like this meter could have been invented or learned independently of tradition. The hypothesis of continuity seems to be required by the very nature of rules 1-4.

5. Envoi An alliterative tradition that persists through major language changes should provide new kinds of evidence for research on the history of English. Since the early days of generative phonology, Paul Kiparsky has emphasized the value of such traditions for diachronic studies, drawing particular attention to the evidence of metrical variants (Kiparsky 1968, 1972). Recently, Putter and Stokes (2001) have shown that the Gawain poet employed a variety of metrical variants, represented with surprising precision in manuscript spellings, to comply with rules 1, 2, and 4. Such variants shed unique light on a complex sociolinguistic milieu within which English syllable structure was undergoing drastic changes. Deeper understanding of these and other important changes can be expected as we trace the remarkable evolution of alliterative meter.23

300 Geoffrey Russom

Notes 1.



4. 5.




Duggan (2001: 497) finds that the a-verse must also have a long dip, but rejects Cable's requirement of two long dips or three stressed syllables (Cable 1991: 86). This is the practical effect of requiring the last stress to be followed by exactly one unstressed syllable. The trochaic constituent is normally a simplex word, though on rare occasions it is the secondary constituent of a compound. Duggan (2001: 481), who surveys a wide variety of alliterative poems, treats the final unstressed syllable as optional, contra Cable (1991). For evidence that rule 4 applies strictly in some poems, see Putter and Stokes (2001). Sir Gawain is cited from Tolkien and Gordon (1967). Unless otherwise identified, Old English examples are from Beowulf, which is cited from Klaeber (1950). The remaining Old English examples are cited from Krapp (1932) and Dobbie (1942), with macrons added to indicate vowel length. The Old Saxon Heliand is cited from Behaghel (1984). When a verse cannot be translated comprehensibly in isolation, information is added between square brackets. For discussion of poetic compounds, see Russom (1987: 92-97). In Beowulf, type Β patterns (x/Sxs, xx/Sxs) have a frequency of 13% and type C patterns (x/Ssx, xx/Ssx) have a frequency of 17%. The corresponding heavy patterns have much lower frequencies, with 2% for type Db (S/Sxs, Sx/Sxs) and 6% for type Da (S/Ssx, Sx/Ssx). See further Russom (1998: 37-43). Statistics for verse types in Beowulf are derived from an electronic database. Researchers interested in obtaining a copy of this database should write to the author, currently at the Department of English, Box 1852, Brown University, Providence, RI 02912. With regard to phrase-final stress, see Campbell (1959: §§97-99). Old Norse does make regular use of enclitics (Russom 1998: 17) and there is evidence for limited employment of enclitics in Old English (Russom 1987: 112), but these constituents do not provide extrametrical syllables verse-finally. Although an enclitic may appear without phrase-final stress at the boundary between verses, it counts as part of the word to which it is attached, and must accordingly occupy an χ position in the second foot of the verse preceding the boundary, as consistent with rules applying to this verse. Such an enclitic contrasts sharply with a proclitic at the boundary between verses, which must conform to the rules that govern the verse following the boundary. Sievers (1898: §78.5) posits a medial stress in weak class II forms to protect his two-stress hypothesis from counterexamples, but this expedient is unacceptable for a variety of reasons. See Russom (1995: 154-156), Russom (2001:45-47). The xx/Sx analysis of type A3 follows Bliss (1958: §10-11). Sievers hypothesized that one of the function words in an A3 verse provided a light second stress, but there is no independent evidence for this hypothesis and it

The evolution of Middle English alliterative meter 301





13. 14. 15.

16. 17.


fails to account for significant metrical differences between type A3 and type A l . See Bliss (1958: §§67-69). Selective placement of verses within the long line normally indicates the correct scansion for an isolated verse-initial prefix. It is almost always safe to scan such a prefix as an anacrusis in the opening half of the line and as a light foot in the closing half. When the prefix is scanned wrongly, the verse will appear to have one foot or three feet, contrary to Principle 2. See further Russom (1998: 45-59). Evidence for this view of anacrusis is established independently of the word-foot theory. For the initial observations, see Cable (1974: 35). For more recent empirical discussion, see Suzuki (1995). For reasons given in Russom (2001), counts of heavy verses now include A2 examples in which the third stressed constituent is a word-final secondary root in a lexicalized compound such as a proper name. Reclassification of these examples as type Al would have no significant consequences for the argument. In this verse, the constituent fetel- is subject to resolution, which assigns a short stressed syllable and a following unstressed syllable to the same metrical position (Russom 1995). Resolution is also assumed in scansions for examples (51), (54), and (62). In Natale Sancti Oswaldi regis et martyris, verse 15a (Skeat 1900: 126). Similar use of mid occurs in verse 66b of this poem, mid blidum mode 'with a happy mind'. The archaic instrumental construction bilde mode is employed in more traditional poems, e.g. Dream of the Rood 122b, Christ 280a, and Guthlac 608b. Brunanburh and Maldon were probably composed not long after the battles to which they refer, which took place in CE 937 and 991, respectively. For complete verse lists see Hofmann (1991). Additional variants of extreme complexity include Psalm 118 6.2a, 27.1a, 50.3a, 51.3a, 69.1a, 75.3a, 92.1a, 96.3a, 114.3a, 127.1a, 127.3a, 154.1a, 154.1a, and 160.2a. At lines 5, 12, 14, 15, 19, 23, 26, 27, 41, 45, 48, and 49. jElfric's line of development has its own intrinsic interest and may include Middle English alliterative poems in nontraditional meter (Oakden 1935: 6, 211). On the break with tradition in rhythmical prose and verse translations, see Griffith (1991). The term "rhythmical prose" protects jElfric appropriately from charges that he was simply unable to conform to traditional rules; but if we deny outright that jElfric was a poet, it seems impossible to account for his regular employment of syntactically coherent half-lines, precisely indicated in some works by manuscript pointing; and for his very frequent binding of these halflines into lines by sound echoes of an alliterative or rhymelike character. For our present purposes, of course, the crucial feature of rhythmical prose is absence of traditional constraints on verse types. Turville-Petre (1977: 15) cites this crudely nationalistic hypothesis from an introductory chapter by Chambers (Hitchcock and Chambers 1932: xlv-

302 Geoffrey Russom



21. 22. 23.

clxxiv). The relevant passage is on p. lxvii, where Chambers maintains that alliterative verse emerged suddenly from oral tradition, "correct, vigorous, and bearing with it a whole tide of national feeling," to form "a link between Old England and Modern England." The observation that a poem is composed in strict traditional meter does not presuppose artistic superiority to nontraditional alliterative poems, whose meters are insufficiently well understood for comparison. Here it is unnecessary to consider the relative importance of orality and literacy in transmission of traditional constraints. The question of continuity is the crucial one. Complete alliterative poems are always found in unpretentious manuscripts, indicating that their intended audience was not the high nobility; hence, according to Turville-Petre (1977: 46), "We must be prepared to include here a wide range of social classes with different interests." See Turville-Petre (1977: 122-123). Kurylowicz (1970: 426) places Germanic alliterative meters in the "natural" class with Avestan. I owe thanks for useful advice to Donka Minkova, Ad Putter, and Jacqueline Haring Russom.

References Behaghel, Otto (ed.) 1984 Heliand und Genesis. 9th ed. Revised by Burkhard Taeger. Tübingen: Niemeyer. Bliss, Alan J. 1967 The Metre of Beowulf. Revised ed. Oxford: Blackwell. Cable, Thomas 1974 The Meter and Melody of Beowulf. Urbana: University of Illinois Press. 1988 Middle English meter and its theoretical implications. The Yearbook of Langland Studies 2: 47-69. 1991 The English Alliterative Tradition. Philadelphia: University of Pennsylvania Press. Campbell, Alistair 1959 Old English Grammar. London: Oxford University Press. Dobbie, Elliott Van Kirk (ed.) 1942 The Anglo-Saxon Minor Poems. New York: Columbia University Press. Duggan, Hoyt N. 1988 Final -e and the rhythmic structure of the b-verse in Middle English alliterative poetry. Modern Philology 86: 119-145.

The evolution of Middle English alliterative meter 303 2001

Some aspects of a-verse rhythms in Middle English alliterative poetry. In Speaking Images: Essays in Honor of V. A. Kolve, Robert F. Yeager and Charlotte C. Morse (eds.), 479-503. Asheville: Pegasus Press. Griffith, M. S. 1991 Poetic language and the Paris Psalter: the decay of the Old English tradition. Anglo-Saxon England 20: 167-186. Hayes, Bruce 1983 A grid-based theory of English meter. Linguistic Inquiry 14: 357393. Hitchcock, Elsie Vaughan, and Raymond W. Chambers 1932 The Life and Death of Sir Thomas Moore, Knight, Sometymes Lord High Chancellor of England. (Early English Text Society, Original Series, 186.) Oxford: Oxford University Press. Hofmann, Dietrich 1991 Die Versstrukturen der altsächsischen Stabreimgedichte Heliand und Genesis. Vol. 2. Heidelberg: Winter. Kiparsky, Paul 1968 Metrics and morphophonemics in the Kalevala. In Linguistics and Literary Style, Donald C. Freeman (ed.), 165-181. New York: Holt, Rinehart and Winston. 1972 Metrics and morphophonemics in the Rigveda. In Contributions to Generative Phonology, Michael Brame (ed.), 171-200. Austin: University of Texas Press. Klaeber, Frederick (ed.) 1950 Beowulf and the Fight at Finnsburg. 3rd ed. Lexington, Massachusetts: D. C. Heath. Krapp, George Philip (ed.) 1932 The Paris Psalter and the Meters of Boethius. New York: Columbia University Press. Kurylowicz, Jerzy 1970 The quantitative meter of Indo-European. In Indo-European and the Indo-Europeans, George Cardona, Henry M. Hoenigswald, and Alfred Senn (eds.), 421-430. Philadelphia: University of Pennsylvania Press. Oakden, James P. 1935 Alliterative Poetry in Middle English. Vol. 2. Manchester: Manchester University Press. Putter, Ad, and Myra Stokes 2000 Spelling, grammar and metre in the works of the Gawain-poet. Parergon 18: 77-95.

304 Geoffrey Russom Robinson, Orrin W. 1992 Old English and its Closest Relatives: A Survey of the Earliest Germanic Languages. Stanford: Stanford University Press. Russom, Geoffrey 1987 Old English Meter and Linguistic Theory. Cambridge: Cambridge University Press. 1995 Constraints on resolution in Beowulf. In Prosody and Poetics in the Early Middle Ages: Essays in Honour of C. B. Hieatt, M. Jane Toswell (ed.), 147-163. Toronto: University of Toronto Press. 1998 Beowulf and Old Germanic Metre. Cambridge: Cambridge University Press. 2001 Metrical evidence for subordinate stress in Old English. Journal of Germanic Linguistics 13: 39-64. Sievers, Eduard 1893 Altgermanische Metrik. Halle: Niemeyer. Skeat, Walter W. 1900 JElfric's Lives of Saints. Vol. 2. London: Kegan Paul, Trench, Trübner. Suzuki, Seiichi 1995 Anacrusis in the meter of Beowulf. Studies in Philology 92: 141-163. Tolkien, John Ronald Ruel, and Eric V. Gordon 1967 Sir Gawain and the Green Knight. 2nd ed. Revised by Norman Davis. Oxford: Clarendon. Turville-Petre, Thorlac 1977 The Alliterative Revival. Cambridge: Brewer.

Old English poetry and the alliterative revival: On Geoffrey Russom's "The evolution of Middle English alliterative meter" Robert D. Fulk

The idea that Old English poetry simply died out and that Middle English alliterative poetry should instead be a lineal descendant of/Elfric's alliterative style seems to have won the day among medievalists, and without a genuine battle. Yet Russom is surely right to raise a warning. Though it has been adopted, modified, and elaborated by metrists and linguists as astute as Cable (1991: 41-65) and Minkova (2003: 14-15), not to speak of the perceptive literary scholars who seem universally to accept it, the reasoning on which the idea was first constructed is rather improbable.1 Just on general principles the idea ought to seem questionable: yElfric is a celebrity for the reason that his name is attached to a rather large body of vernacular literature from a tradition in which authors' names are rarely discoverable. Is it then just a coincidence that the most prolific of the three writers of late Old English whose names we happen to know should have been the fountainhead of the later alliterative tradition? Or is this simply the same sort of historical reductivism that prompted Anglo-Saxonists of a century ago to ascribe such stylistically diverse poems as The Phoenix, the riddles of the Exeter Book, and Wulf and Eadwacer to Cynewulf for no better reason than that we happen to know his name? The notion that the homilist Wulfstan (the only other identifiable vernacular homilist of the Old English period) should have been co-founder with ^Elfric of this alliterative tradition only highlights the absurdity of singling out those whose names we happen to know, especially given that Wulfstan's connection to alliterative composition is so very tenuous: alliteration is not the commonest of his varied rhetorical devices, and his rhetorical periods do not have anything like the regularity of structure of yElfric's verselike compositions. And although it is not impossible that yElfric should have exerted such a considerable influence on succeeding generations - his homilies were widely employed, we know, because they were copied and adapted as late as the thirteenth century in an impressive number of manuscripts, as Blake (1992:

306 Robert D. Fulk 512-513) observes in support of his hypothesis - it does seem something of a leap of faith to derive the form of the popular alliterative verse of the later Middle Ages from a learned and literate source of several centuries earlier. The extensive copying of vElfric's work even after the Conquest in fact raises a particular doubt, for it is remarkable that although his compositions provoked admiration, no one in the Old English period seems to have attempted to imitate them - so what should have prompted later poets to do so? And the recopying of jElfrician texts after the Conquest does not actually lend a great deal of support to the hypothesis that vEIfric sparked a poetic revolution, since it is chiefly his homilies in ordinaiy prose that were recopied rather than his alliterative lives of saints: for example, a popular homily like the one for Pentecost in the first series of Catholic Homilies is found in 17 manuscripts (see Ker 1957: 512-513), while the much-admired passio of St. Oswald is found in just three (see Hill 1996: 247). The derivation of the tradition from iElfric is not in fact based on any broad consideration of general probabilities but simply on supposed similarity of form. Yet the similarity seems to me largely imaginary, deriving only from the observation that both yElfric and the poems of the Alliterative Revival distribute unstressed syllables more freely than classical Old English verse does, and the schemes of stress and alliteration are different. Yet we now know that poems resembling The Wars of Alexander, like Beowulf, and unlike ^ l i n e ' s alliterative prose, have a fairly rigorous metrical structure in which unstressed syllables are arranged on a principled basis. Is it really necessary to appeal to the example of ^ l f r i c to explain the looser treatment of unstressed syllables in early Middle English alliterative compositions? Or can we assume that it would have resulted anyway from the decay of inflectional endings? Surely the latter is the case, since we see the beginning of the process already in late Old English poetry. More important, the focus on metrical structure alone leads to gross misunderstandings, since the difference between prose and poetry in Old English is defined not just by metrical form but also by the deployment of poetic diction, which is generally lacking in Kline's work. Poetic diction, however, much of it inherited directly from Old English verse, does characterize works of the Alliterative Revival. The editors of the standard critical edition of Sir Gawain and the Green Knight illustrate the point in regard to one semantic field in that poem:

Old English poetry and the alliterative revival 307 the well-known group of synonyms for 'man' which supplement the ordinary mon, kniit, noble, prince—burn, freke, gome, hathel, lede, renk, schalk, segge, wyie. Of these renk, though it may be directly from ΟΝ., has a close cognate in OE. riwc; all the others descend from OE. (hathel evidently a blend of hcelep and cepele), and all the OE. words are found either exclusively or predominantly in poetry. (Tolkien and Gordon 1967: 139) Such items of vocabulary can be explained only as derived from the Old English poetic tradition, and unless one is willing to suppose they result from the study of written Old English at a much later date - an assumption that I think few would now entertain, and which in any case conflicts with the evidence of words like renk and hathel, whose altered form better accords with popular transmission than with careful study of a dead language - they must come from a continuous oral tradition of poetry that differed markedly from Kline's compositions. Oakden (1930-1935: II, 113-401) has compiled a massive corpus of data pointing to a continuous, if largely unrecorded, tradition in alliterative versemaking between the Old English and later Middle English periods, given similarities in nominal compounding, poetic vocabulary, formulaic alliterative phrases, and certain stylistic traits, such as the substantive use of adjectives.2 Even the formal evidence for jElfric's influence is questionable, since the testimony of alliteration is rather negative: while it is very common to find three alliterating staves in an Old English verse pair, and three is the norm in, for example, Sir Gawain and the Green Knight, this sort of alliteration is infrequent in yElfric. Scholars are understandably reluctant to posit a large body of lost literature: hence, for example, the preference of Salter (1988: 56) for the jElfrician derivation even of La3amon's verse. But reluctance on this score is a reflex of the study of written texts. It would indeed be undesirable to posit a large body of lost writings for which there is no evidence. Oral literature, however, is a different matter, for we must assume that even during the Old English period, when monastic conditions were conducive to the recording of alliterative verse, the amount actually written down was minuscule in comparison to the amount in oral circulation. One of the admirable things about the word-foot theory is how elegant its explanations for metrical phenomena so often are, explanations that convince in part by their cohesiveness within a framework that ties meter firmly to the facts of historical Germanic prosody and morphology. In the present instance, though, the arguments make sense even without the supporting framework of the word-foot theory. Whether or not there was in actuality a general decline in the production of compounds in early Middle

308 Robert D. Fulk English, certainly compounds grow infrequent in most of the verse of the later Old English period, rendering verses of types C, D, and Ε relatively rare. The remaining types A and B, because they are more regular and canonical than the new types found in late Old English compositions, tend to be restricted increasingly to the off-verse, giving an initial approximation of what alliterative verse structure was to be like three hundred years later. Russom's demonstration that the first fifty lines of Kline's account of St. Oswald run counter to this trend - they evince a significant number of offverses with more than one protracted drop - is striking, and it made me wonder whether the development of the form could be traced to any extent in early Middle English alliterative verse. To my surprise and great interest, the evidence that I found was mostly negative. Four of the 24 offverses of The Grave (ca. 1150-1200) contain more than one protracted drop (i.e., 16.7%), and at least 12 of the first 50 complete lines of the Worcester fragments of The Soul's Address to the Body (ca. 1200) do the same (i.e., 24%). The figure for the first 50 complete lines of Lajamon's Brut (ca. 1200-1220, Caligula MS) is perhaps as high as 8 (i.e., 16%).3 These figures are not arrestingly different from the one Russom arrives at for iElfric (i.e., 24%), which is all the more surprising given that the early Middle English period is the one in which the differentiation of on-verse and off-verse should have been greatest, with the most canonical verse types strongly preferring the off-verse. Could it be that these three compositions are not actually attempts at verse but at "rhythmical alliteration" like ^ l f r i c ' s ? This is hardly plausible, given Oakden's findings about the lexical features that these works share with Old English poetry. And the Worcester fragments and Lajamon's Brut are measurably different from jElfric inasmuch as they incorporate rhyme in alternation with alliteration. They are also rhythmically not nearly so loose as the alliterative compositions of the Katherine Group (see Millett 1982: liii, with references, but cf. Cable 1991: 63). In any case, La3amon himself thought of his composition as verse, for he calls it loft-song (36). There must be another explanation. The question whether compositions like The Grave and The Soul's Address to the Body were actually intended as verse is in actuality a profoundly unfashionable one, for scholars of the transitional and Middle English alliterative traditions, with some uniformity, at least profess allegiance to the notion that alliterative compositions of the period should not be classified as "verse" or "prose" but as manifestations of an alliterative mode that reveals itself at various points on a continuum from the more verselike to the more prose-like. Hence Blake's noncommittal term "rhythmical

Old English poetry and the alliterative



alliteration" employed above, intended to encompass all kinds of alliterative texts. I suppose this notion has provoked no very memorable dissent because it seems to capture the compositional variability from one text to the next that is apparent to everyone. Yet what may seem a commonsensical idea in this instance is nothing of the kind, since Blake's notion of doing away with the distinction between verse and prose is logically incongruous with his claim that Old English verse died out and ^Elfric became the model for alliterative composition. If the distinction between prose and verse is too crude when one's purpose is to trace Middle English alliterative verse to jElfric, the difference cannot logically be reinstated for the purpose of killing off Old English poetry. That is clearly tendentious. And of course the purpose of inventing the term "rhythmical alliteration" is intimately connected to the aim of reconceiving /Elfric's role. This should be clear: if we were obliged to refer to his alliterative compositions as prose, the reasonableness of assuming the death of a verse tradition of long standing in order to derive Sir Gawain and the Green Knight from a method of prose composition imitated by no one in the Old English period might not seem so obvious. Another difficulty raised by obliterating the distinction between prose and verse is that it would seem to imply that /Elfric himself would not have had reason to differentiate what he composed from verse. There was in fact a time when it was debated which classification better describes jElfric's Lives of Saints (see, e.g., Luick 1905: 142-143; Kuhn 1973), though Russom (see his n. 17) may be the only one now willing to keep the question open. Indeed, the arguments against classifying this as verse have not generally been good: the strongest argument mustered by Lipp (1969: 695) is that scholars would not like to see it called this. It is in fact impossible to prove that jElfric did not think he was writing verse, but certain considerations hint that he did not. In his Latin preface to the Lives of Saints (ed. Wilcox 1994: 120), jfilfric says he has translated the Latin simplici et aperta locutione 'with simple and plain phrases', and this does not suggest verse composition. In his letter to Sigeweard ("On the Old and New Testament"), he refers to his alliterative renderings of the books of Esther and Judith as having been translated on ure wisan 'after my manner' (see Crawford 1922: 48), using the phrase twice, seemingly in reference to his alliterative style. If he had considered his translations poetic, surely he would have used the phrase on leodwison 'in verse', which is how, in a note (ed. Wilcox 1994: 113) in the second series of Catholic Homilies, he describes a certain life of St. Thomas (now lost), translated gefyrn 'long

310 Robert D. Fulk ago', whose existence he says makes it unnecessary for him to translate the work. The evidence is hardly conclusive, but it suggests that ^Elfric did not think of himself as a poet. If that is the case, the modern consensus that we ought to conflate alliterative prose and verse into a single category would appear to be a mistake. Blake (1969: 120) perhaps anticipates this objection when he says, "It might be helpful if modern scholars adopted this name [i.e., 'rhythmical alliteration'], not because it accurately reflects medieval classifications, but because our own classification of prose and poetry has become so rigid." The perception itself I think is mistaken: certainly there is now a wider array of terminology in this area, and thus wider possibilities for classification, than there was in early medieval times, when the distinction between prose and verse, codified in the monastic Latin curriculum, was of the first importance (see Lapidge 1996: 1-5). Yet even if the perception were correct, the argument would evince the same sort of illogic that we have seen characterizes this line of analysis throughout. For if our purpose is to understand the evolution of poetic form as it really happened, surely we should at least attempt to understand the phenomenon on the basis of categories that could actually have been current in medieval times. The alternative is to be doomed to solipsism, ever recrafting our terminology so as to ensure that we perceive only a Middle Ages of our own fashioning. Though the oral nature of most early English vernacular literature has long been acknowledged, we have never, most of us, really come to terms with this recognition, since literate habits render the supposition of a huge body of unrecorded texts a scholarly embarrassment. Linguists, too, are sure to squirm at arguments predicated on unavailable data. But if an enormous body of oral alliterative verse did exist throughout the medieval period - and logic tells us that it almost certainly did - it is impossible to know whether compositions like The Grave and The Soul's Address to the Body, being literate productions, are not very unlike oral works to which we will never have access. This may explain the unruliness of the data. But if so, it raises profound questions about explanation in linguistics, most particularly whether the aim of historical linguistics should be to explain the data available or to analyze texts of earlier periods from a realistic historical perspective - that is, whether the primary allegiance of historical linguistics should be to linguistics or to history.

Old English poetry and the alliterative revival


Notes 1.



The idea is generally said to have originated with Blake (1969: 120): "Classical Old English poetry disappeared about the time of the Conquest or earlier. But works continued to be written in rhythmical alliteration following the tradition popularized by jElfnc and Wulfstan." The use of the word "written" in this passage explains much about the misconceptions at work in this hypothesis. Astonishingly, Hanna (1995: 43 and n. 3), in support of the assertion that "Middle English verse displays no continuity with Old English poetry," claims, "Oakden's stylistic study (2: 113—402) shows conclusively the break in the tradition." The situation is the reverse: Oakden says that his findings about compounds are "proof of continuity despite change" (166); the poems of the Alliterative Revival contain many more strictly poetic and archaic words than other types of Middle English texts, many inherited from Old English (183— 88); and reflexes of Old English alliterative formulas, while far commoner in Lasamon's Brut than in the later alliterative poems, nonetheless indicate continuity of tradition (264). Worthwhile comparison with The Proverbs of Alfred is not feasible because off-verses with a single lift are so common in that text. The relevant lines in the other poems are the following: Grave: 3, 10, 15, 19; Soul's Address: A 3, 11, 23,40,41,43, 44, 45, Β 1,4, 5; Brut: 3, 13, 26, 27, 29, 44, 46 (?), 52.

References Blake, Norman 1969 Rhythmical alliteration. MP 67: 118-24. 1992 The literary language. In The Cambridge History of the English Language, II: 1066-1476, Norman Blake (ed.), 500-541. Cambridge: Cambridge University Press. Cable, Thomas 1991 The English Alliterative Tradition. Philadelphia: University of Pennsylvania Press. Crawford, S. J. (ed.) 1922 The Old English Version of the Heptateuch. EETS o.s. 160. London: Oxford University Press. Hanna, Ralph, III 1995 Defining Middle English alliterative poetry. In The Endless Knot: Essays on Old and Middle English in Honor of Marie Boroff, M. Teresa Tavormina and R. F. Yeager (eds.), 43-64. Cambridge: Brewer.

312 Robert D. Fulk Hill, Joyce 1996

Ker, N. R. 1957

The dissemination of jElfric's Lives of Saints: A preliminary survey. In Holy Men and Holy Women: Old English Prose Saints' Lives and Their Contexts, Paul E. Szarmach (ed.), 235-259. Albany: State University of New York Press.

Catalogue of Manuscripts Containing Anglo-Saxon. Oxford: Clarendon. Kuhn, Sherman 1973 Was JElfric a poet? PQ 52: 643-662. Lapidge, Michael 1996 Anglo-Latin Literature. In his Anglo-Latin Literature 600-899, 1 35. London: Hambledon. Lipp, Frances Randall 1969 iElfric's Old English Prose Style. SP 66: 689-718. Luick, Karl 1905 Englische Metrik: Geschichte der heimischen Versarten. In Grundriss der germanischen Philologie, Hermann Paul (ed.), 2nd ed., IIb, 141-160. Strassburg: Trübner. Millett, Bella (ed.) 1982 Hali Meidhad. EETS o.s. 284. London: Oxford University Press. Minkova, Donka 2003 Alliteration and Sound Change in Early English. Cambridge: Cambridge University Press. Oakden, J. P. 1930-35 Alliterative Poetry in Middle English. 2 vols. Manchester: Manchester University Press. Salter, Elizabeth 1988 English and International: Studies in the Literature, Art and Patronage of Medieval England, Derek Pearsall and Nicolette Zeeman (eds.). Cambridge: Cambridge University Press. Tolkien, J. R. R., and Ε. V. Gordon (eds.) 1967 Sir Gawain and the Green Knight. 2nd ed. Revised by Norman Davis. Oxford: Clarendon. Wilcox, Jonathan (ed.) 1994 /Elfric's Prefaces. Durham Medieval Texts 9. Durham: Department of Medieval Studies.

A brief response Geoffrey Russom

Fulk provides much interesting evidence and a convergent line of argument for my claim that Middle English alliterative meter evolved straightforwardly within a continuous tradition. We also agree that jElfric's rhetorical prose stands outside this tradition. What I would like to consider here is whether rhetorical prose is a kind of poetry. I agree with Fulk that this question is "profoundly unfashionable." I may indeed be "the only one now willing to keep the question open," but would like to maintain a minority stance in hopes of provoking more substantive responses like Fulk's: the first attempt, I think, to tackle the question in a serious (if unfashionable) way. As Fulk points out, j©fric did not employ the usual Old English terms for "poetry" to describe his work. This is indeed significant, as Fulk claims, "if our purpose is to understand the evolution of poetic form as it really happened." jElfric's remarks suggest that he neither had nor sought any influence within the poetic culture of his day. Such evidence does not exclude the possibility that ^Elfric was a poet, however. Old English terms for "poetry" were normally applied to works like Cynewulf s Elene or, in scholarly usage, to works in established Latin meters. If j®fric used such terms to describe his homily on the death of St. Oswold, he would leave himself open to easy objections, since the homily does not conform to any familiar standards of versecraft. vfilfric's departure from established forms was almost certainly deliberate and quite justifiable in the context of Christian preaching. Traditional alliterative poems made crucial use of syntax that had become distinctly archaic by the end of the tenth century, along with esoteric figures of speech from secular aristocratic culture. For a preacher addressing a general Old English audience, such poetic impedimenta might well be abandoned as distracting; and composition in Latin would have missed the point entirely. For me the question that arises is whether, in his efforts to find language that would move his audience, j®lfric created non-traditional poetry or heavily ornamented prose. If he invented a new verse form, no precise word for it would have been available to him, and on ure wisan ("after my manner") might have been as

314 Geoffrey Russom close as he could come. The phrase simplici et aperta locutione certainly signals a rejection of traditional versecraft, but Wordsworth described his poetry in similar terms. Whether iElfric was a poet is a question that arises most sharply within a universalist framework. One much-discussed candidate for universal status is a maximum average of nine words for the poetic line. This constraint makes the special powers of rehearsal memory available for what the tenth-century Indian metrist Abhinavagupta called "savoring" (Hogan 1997: 241-44). Lines of jElfric's Oswold narrative, as represented in modern editions, fall well within the poetic limit. Rosenberg (1970) prints sermons of American folk-preachers as verse when they fall into phrases of comparable length. If we abstract away from accidents of time and place, K l i n e ' s status as a poet becomes a problem of genuine interest.

References Hogan, Patrick Colm 1997 Literary universale. Poetics Today 18: 223-49. Rosenberg, Bruce A. 1970 The Art of the American Folk-Preacher. New York: Oxford University Press.

A central metrical prototype for English iambic tetrameter verse: Evidence from Chaucer's octosyllabic lines1 Xingzhong Li

Nat that I wilne, for maistrye, Here art poetical be shewed, But for the rym ys lyght and lewed, Yit make hyt sumwhat agreable, Though som vers fayle in a sillable; And that I do no diligence To shewe craft, but ο sentence. Geoffrey Chaucer The House of Fame (1094-1100)

1. A prototype for English iambic tetrameter verse Adopting the notation of Liberman and Prince (1977) and following Youmans (1989: 346-347), Li (1995: 159-160) proposes (1) below as the prototype for Chaucer's iambic tetrameter lines of verse. Assuming that (1) applies to a wider range of iambic tetrameter verse by other poets, this paper further proposes .(1) as the central metrical prototype for the centuries-old English iambic tetrameter verse in general. As current metrical literature stands, pre-Chaucerian (Minkova 1996) and postChaucerian iambic lines of verse (Tarlinskaja 1976, among others) in early Middle English, Middle English, and Modern English are all foot-based while Chaucerian metrics is unfooted (Halle and Keyser 1966, 1971; Cable 2002). This paper provides statistical evidence from Chaucer's octosyllabic lines and claims that his lines are indeed metrically footed. If this is correct, then all English iambic tetrameter verse can be assumed to share the same central prototype as in (1), with branching nodes labeled S for more prominent verse constituents and W for less prominent ones.

316 Xingzhong Li

(1) Prototypical iambic tetrameter line: Line:



Metrical Feet:





/ \ S W/ \ S

Metrical positions:


Syllable [±stress]:

-s 1

+s 2

-s 3

+s 4

/ \

/ \S


-s 5

+s 6


-s 7

/\ +s 8

Metrical saliency in iambic tetrameter

In addition to the controversial question of whether Chaucerian syllablecounting verse is foot-based or unfooted, there is another equally controversial issue, which is whether iambic tetrameter in general and Chaucer's verse in particular is gradiently or categorically well-formed (or ill-formed for that matter). A quantitative study of Chaucer's octo-syllabic lines suggests a metrically gradient nature, encoded in (1) and elaborated below. Generative metrists all agree that metrical studies should address linguistic form as interacted with metrical form rather than the linguistic form alone (Halle and Keyser 1971: 140; Hayes and Kaun 1996: 246-247). The prototype (1) encodes several such types of language-meter interactions: between metrical positions and linguistic syllables, between feet and iambic words or clitic phrases, between hemistichs and major phrases, and between metrical lines and clauses or sentences. Of course, actual lines of verse can deviate markedly from the prototype, but the more closely they conform to it, the more "regular" they are perceived to be. More importantly, the prototype (1) defines four metrical saliency principles contingent upon the gradient prominence and well-formedness of metrical constituency, corresponding to metrical positions, feet, hemistichs, and metrical boundaries.

English iambic tetrameter



2.1. The gradient saliency in metrical positions The prototype (1) suggests that the eight metrical positions differ from one another in their metrical saliency. This can be explained by analyzing the vertical dominance relations in it. If we treat unstressed [-s] syllables in all odd-numbered metrical positions as W, as Youmans (1983, 1989, 1996), Li (1995), Youmans and Li (2002) do with their iambic pentameter prototype, (1) defines four degrees of relative stress: W for [-s] syllables in all oddnumbered positions; 2S for stressed [+s] syllables in positions two and six; 3S in position four; and 4S in position eight. Prototypically then, the strongest stress is on the final syllable in the line, the next strongest on the fourth syllable, the still next strongest on the second and sixth syllables, and the weakest on all the odd-numbered syllables, summarized in (2) as follows: (2) The prototypical saliency in the metrical positions: [1W 2S] [1W 3S] [1W 2S] [1W 4S] 1 2 3 4 5 6 7 8 Since there are cases in which two or more positions possess the same degree of saliency, we have to collapse those cases if we are to rank the saliencies in all the positions, as treated in (3): (3) The ranked saliency in the eight metrical positions (provisional): W [1 3 5 7] «

2S 3S 4S [2 6] « [4] « [8]

This rank represents a quantum leap and significant departure from the understanding that all metrical positions in a line of verse have equal metrical saliency or weight. If we could differentiate collapsed positions further to sensitize them for tension metric, (3) would be perhaps more useful. This new goal can be achieved by counting individual sequences of nodes in (1) all the way down from the top, treating [-s] as W and [+s] as S. For example, the first [+s] syllable in position two is first dominated, from the top, by W (the first hemistich), then by another W (the first weak foot), and then by S (the first strong metrical position) and is thus assigned a

318 Xingzhong Li saliency sequence of [WWSS], while the seventh syllable, which is [-s], is assigned a sequence of [SSWW], being first dominated, from top, by S (the second hemistich), then by another S (the fourth foot, which is strong), and then by W (a weak position). Then, why should we count from top down but not bottom up? The answer lies in this. If we count from bottom up, all saliency sequences in the first hemistich would end with W, and all those in the second hemistich with S. Consequently, the [+] syllable in position two would be assigned a sequence of [SSWW] while the [-s] syllable in position seven a sequence of [WWSS], Sequences derived this way would lead to an inaccurate saliency discrimination because, given the strong closure principle in both phonology and metrics, [WWSS] is more salient than [SSWW], but in fact position seven is in no way more salient than position two. The top-down counting eliminates the mis-ranking because sequences in all weak positions will always end with W whereas sequences in all strong positions always with S. This way of counting yields eight, instead offour, degrees of relative saliency, presented in (4): (4) The prototypical saliency in the metrical positions: [WWWW] [WWSS] [WSWW] [WSSS] [SWWW] [SWSS] [SSWW]








[ssss] 8

Three comments are needed to interpret the sequences. First, the more Ss a sequence has, the more salient it is considered to be. Thus, [WSSS] in position four is considered to be more salient than [WWSS] in position two. Second, when two sequences show equal numbers of Ws and Ss, such as in positions four and six, where each has 1W and 3S, the position of S in the sequence determines the degree of saliency. The closer the S is to the right edge of a sequence, the more salient that sequence is. Thus, [WSSS] is considered to be more salient than [SWSS]. Similarly, [WWSS] in position two is more salient than [SSWW] in position seven, and [WSWW] in position three more salient than [SWWW] in position five. Position one, assigned [WWWW], is obviously the least salient, while position eight, assigned [SSSS], is the most salient. The result is a fine-grained rank of metrical saliency in all the eight positions, given in (5):

English iambic tetrameter verse


(5) The ranked prototypical saliency in the eight metrical positions (final): [WWWW] « [ s w w w ] « [ w s w w ] « [ s s w w ] « [ w w s s ] « [ s w s s ] 1 5 3 7 2 6 «[wsss]«[ssss] 4 8 While (3) claims that the saliency of positions two and six is the same, (5) ranks position six as more salient than position two. While (3) ranks all weak positions the same, (5) finely discriminates them. It is generally understood that the more salient a metrical constituent is, the more constrained it becomes, the lower the frequency of metrical deviations that would occur to it, and the more disruptive the deviations would be. The same principle holds true for Chaucer's iambic tetrameter, as the statistics in the third section show.

2.2. The gradient saliency in the metrical feet The prototype (1) embodies not only a gradient saliency in metrical positions but also a gradient saliency in metrical feet. The foot saliency is the sum of, but not the difference between, the saliency sequences of the two metrical positions in a given foot, as is presented in (6): (6) The prototypical saliency in the metrical feet: [wwww+wwss] [wsww+wsss] [swww+swss] [ssww+ssss] Fl F2 F3 F4 The Ws and Ss in a given foot (F) cannot cancel each other, for doing so would derive four identical, indiscriminative sequences of [WWSS] for the feet. This is because the cancellation respects no sequential ordering of the Ws and Ss. We can use the same principles we used to interpret (4) to interpret (6): The more Ss a foot has and the closer they are to the right edge of a foot boundary, the more salient that foot is considered to be. Thus, the four feet must be ranked as in (7):



(7) The ranked prototypical saliency in the metrical feet: Fl « F3 « F2 « F4 That is, the fourth foot is the most salient, and the second foot is the next most salient, followed by the third foot; the first foot is the least salient. Predictably then, metrical deviations would likely occur the most frequently in the first foot, the next most frequently in the third foot; they would occur the least frequently in the fourth foot and the next least frequently in the second foot.

2.3. The gradient saliency in the hemistichs The prototype (1) also defines a ranked salience principle with regard to hemistichs. There is no need to add up all the saliency sequences in each hemistich for determination of its saliency; (1) has made it clear that the second hemistich, labeled S, is more salient than the first hemistich, labeled W, as ranked in (8): (8) The ranked prototypical saliency in the hemistichs: Hl « H2 Predictably, metrical deviations in the first hemistich would be more frequent, but would be less rhythmically disruptive if at all, than those in the second hemistich.

2.4. The gradient saliency in the metrical boundaries Both traditional and generative metrical principles observe caesuras, their locations and their roles in a line of verse. However, do caesuras possess the feature of gradient saliency? How does (1) define the issue? Specifically, (1) describes four types of metrical boundaries or caesuras: line, hemistich, foot, and position. Prototypically, a line has one endstopped boundary (LB), two hemistich boundaries (HB1, HB2), four foot boundaries (FBI, FB2, FB3, FB4), and eight syllable boundaries (SB1

English iambic tetrameter



through SB8), totaling fifteen boundaries distributed on a strictly layered hierarchy. In this hierarchy, each of the odd-numbered SBs locally and optionally requires a one-syllable siring before it and gives no heed to how the rest of a line is divided; so does each of the even-numbered SBs, which, however, differ from the odd-numbered SBs in that they coincide with FB, FB and HB, or FB and HB and LB. Imagine that any given boundary at any given hierarchical level only locally requires a string of a given number of syllables that goes before it and pays no attention to how the rest of a line is further divided. The prototype in (1) would generate a great deal of lines with no two being subdivided alike, to which may be added a variety of new division patterns if enjambment is considered. A thorough classification of line-division patterns could be best achieved using the OT approach, but given the space limit, I opt for the most obvious and choose as representatives LB from the group of {LB, HB2, FB4, SB8}, HB1 from the group of {HB1, FB2, SB4}, FB3 from the group of {FB3, SB6}, FBI from {FBI, SB2}, and rank them along with the four odd-numbered SBs. As (1) indicates, LB is more salient than HB1 insofar as it is the most expected, and HB1 is more salient than any FB insofar as it is the most obvious line-internal caesura. FB3, being closer to the end of a line and embedded into a more prominet hemistich, is more salient than FBI and, as such, should occur less frequently than FBI does. Given the symmetrical structure of (1), SB3 (3//5 pattern) should be compared with its mirror image of SB5 (5//3 pattern). Since SB3 is embedded into a more prominent foot (i.e., the second foot) while SB5 into a less prominent foot (i.e., the third foot), SB3 should be considered more salient than SB5 and, therefore, occur less frequently than SB5 does. By the same token, SB7 (7//1), being embedded into the fourth foot, the most prominent foot, is more salient than its mirror image of SB1 (1//7), which is embedded into the least salient foot and, therefore, should occur less frequently than SB1 does. If the discussion is correct, the saliency rank of primary boundaries in an iambic tetrameter line should take the shape of (9): (9) The ranked saliency in the primary metrical boundaries (caesuras): SB1 «

SB7 « SB5 « SB3 « FBI « FB3 « HB1 «


And the occurrences of these primary metrical boundaries should be ranked, prototypically, as in (10):

322 Xingzhong Li (10) The ranked frequencies of the primary metrical boundaries: SB7 «

SB1 «

SB3 «

SB5 « FB3 « FBI «

HB1 «


To reify (10), let us take "//" for a primary phonological pause or a punctuation mark. We then obtain the eight types of division patterns, listed in (11): (11) Eight types of line-internal and line-final primary caesuras:

a. b. c. d. e. f.









X X / / XX








xx// x//x





g· h.










(after LB: (after HB1 (after FBI: (after FB3: (after SB5: (after SB3: (after SB 1: (after SB7:

8//0 division) 4//4 division) 2//6 division) 6//2 division) 5//3 division) 3//5 division) 1//7 division) 7//1 division)

Most preferred •


Least preferred To sum up, the prototype in (1) encodes four principles of ranked gradient metrical saliency, as presented in (5), (7), (8), and (9). How is (1), along with its derived metrical saliency principles, borne out by statistical evidence from actual lines of verse? When deviations occur, what patterns of unconformity do they exhibit?

3. Statistical evidence from Chaucer's octo-syllabic lines 3.1.

Corpus of data

I adopted a "systematic sampling" method in setting up my corpus of data for analysis because systematic samples "are easier to obtain than random samples" and because "like random samples, systematic samples are also representative samples. ... [T]he findings based on the [systematic] sample are not biased and ... [are] generalizable ... to the parent population" (Hopkins, Glass, and Hopkins 1987: 117).

English iambic tetrameter verse 323 Given that Chaucer wrote a total of some 11,260 octo-syllabic lines of verse, I used The Riverside Chaucer (Benson 1987) and systematically selected every tenth line from them, yielding a total of 1,126 lines as the corpus of data. One anonymous reviewer questioned the sampling technique, asking if it would systematically sample only certain parts of Chaucer's regular metrical groupings such as stanzas, couplets, or the like and, therefore, skew the analysis. That is not a concern, though. Of the total sampled lines, only 17 are from stanzas in "Sir Thopas" in The Canterbury Tales (as "Thop" below) and "Anelida and Arcite" (as "Anel" later). Since none of the stanzas has just ten lines each for the method to take the tenths, the sampling technique actually picks up lines from different places in the stanzas. All of Chaucer's other iambic tetrameter lines are written in couplets, including The Romaunt of the Rose: Fragments A, B, C (as RomA, RomB, RomC below), The Book of the Duchess (BD), and The House of Fame (HF). It seems then that the sampling method does tend to choose the second line in a couplet. However, the two successive lines in a couplet consist of "the same meter" (Murfin and Ray 1997: 62). Given the metrical sameness, the sampling technique would not skew the analysis. Additionally, one might think that the second line in a couplet is more likely to end with a sense of closure than does the first line in the same couplet and thus bring about some sort of difference. Insofar as completing a rhyme scheme is concerned, such a thought makes sense. However, Chaucer sometimes did just the opposite by ending the first line in a couplet with a punctuation mark, but not the second line, as exhibited in six of the seven couplets in (12), for instance, where rhyming pairs are italicized (emphasis mine): (12)


Our first mater is good to kepe. So whan I saw I might not slepe Til now late this other night, Upon my bed I sat up right And bad oon reche me a book, A romaunce, and he it me tok To rede and drive the night away; For me thoughte it better play Then playe either at ches or tables. And in this bok were written fables That Clerkes had in olde tyme,

324 Xingzhong Li And other poetes, put in rime To rede and for to be in minde, While men loved the lawe of kinde. (BD 43-56) Thus, the sampling technique does not result in skewed analysis.


Interactions between position saliency and stress patterning

The saliency principle in (5) claims that the stress patterning and its variations occur in a predictable manner. Two sets of evidence support this claim. The first set is from my own sample of 1,126 iambic tetrameter lines of verse by Chaucer. The second set is from Tarlinskaja (1976). She scanned samples that ranged from 308-1200 lines of iambic tetramater verse of 14 post-Chaucerian poets, all during the seventeenth through nineteenth centuries (Milton, Marvell, Butler, Swift, Cowper, Scott, Southey, Wordsworth, Coleridge, Byron, Shelley, Browning, Tennyson, and Rossetti) and then constructed stress profiles for each poet (1976: 260). For nearly every poet in her sample, the highest incidence of stressed syllables is in the eighth position, the next highest is in the fourth position, and the lowest incidence of stressed syllables is in the second position. Conversely, in W metrical positions for almost every poet, the lowest incidence of stressed syllables is in the seventh position, and the next lowest is in the fifth position. As a sharp contrast, the first position has the highest incidence of stressed syllables. For the sake of comparison, Table 1 below presents my computation of the average stress profile for Chaucer in my sample and for all the 14 poets in Tarlinskaja's. Table I. Percentages of stressed and unstressed syllables in each metrical position in the sampled lines of Chaucer's iambic tetrameter in Li (1995: 181) and in the sampled lines of 14 poets' iambic tetrameter in Tarlinskaja (1979: 260) Position Chaucer Average (%) 14 poets Average (%)

1 6.7

2 83.0

3 4.4

4 91.6

5 5.2

6 85.6

7 0.8

8 98.6









English iambic tetrameter verse


A ranking of the percentage averages from the lowest to the highest yields the following results, as opposed to the prototypical ranking in (5): (13) A position-stress conformity comparison between (5) and Table 1: 1« 5« 3« 7« 2« 6« 4« 8 7« 3« 5« 1 « 2« 6« 4« 8

(5) (Table 1: Chaucer)

1« 5« 3« 7« 2« 6« 4« 8 7 « 5 « 3 « 1« 2 « 6 « 4 « 8

(5) (Table 1: 14 poets)

In (13), deviations are underlined. For both Chaucer and the 14 poets, no unconformity occurs in the prototypical even-numbered positions. In this regard, their verse is alike, which shows that metrical deviations tend to occur in less salient metrical constituents. The unconformity by the 14 poets is the switch between positions seven and one, which merely confirms the obvious: At the beginning of an English iambic line, there is the tendency for trochaic inversion. Chaucer's unconformities include the one-seven and three-five switches, suggesting that for him, there is the tendency for trochaic inversions not only at the beginning of a line but also at the beginning of the second hemistich, and they occur more frequently at the beginning of a line than at the beginning of the second hemistich. Table 2, derived from Table 1, places emphasis on trochees and how they are distributed at the level of metrical foot in the verse by Chaucer and the fourteen post-Chaucerian poets: Table 2. Percentages of distribution patterns of trochees in each metrical foot in the sampled iambic tetrameter lines of Chaucer and 14 poets Foot Chaucer

1 6.7%

2 4.4%

3 5.2%

4 0.8%

14 poets





Table 2 illustrates three points of theoretical interest. First, stress patterning deviations from (1) occur significantly less often in Chaucer than in the aggregated verse by the 14 post-Chaucerian English poets, showing metrical idiosyncrasies by individual poets in different periods. Second, when deviations such as trochaic inversions do occur, they tend to

326 Xingzhong Li

fall in similar places in a line for all 15 poets. In this connection, their verse is alike. Third, trochaic inversions occur in exactly the same gradient way as the foot saliency principle (7) predicts for Chaucer and only slightly less so for the 14 poets, as shown in (14): (14) A trochee-foot conformity comparison between (7) and Table 2: Fl « F3 « F2 « F4 Fl « F3 « F2 « F4


Fl « F3 « F2 « F4 Fl « F2 « F3 « F4


(Table 2: Chaucer)

(Table 2: 14 poets)

As Tables 1 and 2 illustrate, deviations from (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 the 4//4 lines for Chaucer and the second foot in the 2//6 lines for the 14 poets). Conversely, deviations are least common at the end of the line (the fourth foot) and next least at the end of the first hemistich for both Chaucer and the 14 poets. Hence, Tarlinskaja's statistics and the statistics in this project parallel closely, and both studies lend strong support to the "loose beginning and strict ending" principle as presented in Youmans (1989: 360-364), Li (1995: 181-185), Youmans and Li (2002: 156-165), Hayes and MacEachern (1998), and Fabb (2002: 173-175).


Interactions between boundary saliency and line subdivisions

The metrical prototype for iambic tetrameter in (1) is also well supported statistically in terms of line-internal and line-final primary divisions (indicated by "//"). To determine primary linguistic pauses or subdivisions, I used line-internal punctuation, if any, to guide my analysis: (15) a. b. c. d. e. f. g-

1//7 2//6 3//5 4//4 5//3 6//2 7//1

Loo, // to the Hous of Fame yonder, And ye, // me to endite / and ryme The gladdest, // and the moste at reste. To slen your frend, // and namely me, Where that he cometh, // overall, This is moche my desir, // shortly. But he can cherles daunten, // he,

HF 1070 HF 520 BD 1280 Anel 260 RomB 3260 RomC 6970 RomA 880

English iambic tetrameter verse


However, most of the sampled lines do not contain line-internal punctuation marks. I then applied Hayes's strictly layered prosodic hierarchy involving "the Utterance, the Intonational Phrase [I-phrase], the Phonological Phrase [P-phrase], the Clitic Group [C-group], and the [phonological] Word" (1989: 206) that Hayes restated essentially from Selkirk (1980, 1981a, 1981b) and Nespor and Vogel (1982, 1983). In Hayes's definition, the domain of phonological words parallels individual terminal words in syntactic trees. That is, morphological suffixes and contractions are included within phonological words. Thus, "visited" and "wouldn't" are single phonological words, whereas "visit # it" is two (206-207). A C-group includes one and only one content word (such as a noun, main verb, adjective, or adverb). This content word may be preceded or followed by one or more clitic words (minor category words such as articles, prepositions, conjunctions, pronouns, and auxiliary and modal verbs). The content word serves as the host of the group, as for example in the C-groups in [ c he kept it] [ c in a large] [ c jar] (208). The P-phrase is formed from one or more C-groups, and it coincides with syntactic phrases defined by the X-bar system: N " (noun phrase), V " (verb phrase), A " (adjective phrase), and so on (211-218). The N " [P [c a Chinese] [ c poem] ] is an example, containing two C-groups. The I-Phrase is "a concatenation of one or more P-phrases," so, for example, [i This is the cat] [, that caught the rat] [i that stole the cheese] contains three I-phrases (219). A prosodic hierarchy of this sort, the C-group and the P-phrase in particular, can help clarify the phonological phrasing above the level of syllables in metrical trees and determine line-internal divisions when no punctuation is available. For example: (16) a.


b. c. d.

2//6 3//5 AHA

e. 5//3 f. 6//2 g·




(No examples found, unless there is a punctuation mark, as in 15a.) [c Overal] // [ c I entremete me] RomC 6840 [c That duelleth] // [P [ c in a cave][ c of stoon] ] HF 70 [P [ c Of all] [ c this world] ] // [P [ c I have] [ c the cure] ] RomC 7680 [ P [ c Nat longe] [ c tyme] ] // [ c to endure] BD 20 [ρ [ c But men] [ c myght axe me] ] // [P [ c why] [ c soo] ] BD 30 [ρ [c Ne callen you] / [ c his freend] ] // [ c also] RomC 7650 [c Amonges these apostlis] // [ c newe], RomC 6270

328 Xingzhong Li However, this prosodic hierarchy does not account for all of Chaucer's prosodic phrasings. So Youmans, in his (1996) article on Chaucer's iambic pentameter, extends the C-group to include some minor category words, which are normally subordinate to a clitic host but are sometimes blocked from cliticization by movement transformations (194). He cites the following examples to show how minor category words can be promoted to clitic hosts, reproduced here as (17): (17) a. [ c That time] [ c of year] [ c thou mayst] [ c in me] [ c behold] (boldstyle mine for emphasis) Shake.Son.73 b. [ c Thou mayst behold] [ c that time] [ c of year in me] The movement transformations have the effect of promoting [thou mayst] and [in me] as new C-groups, with "mayst" and "me" being new clitic hosts respectively, and thus creating two more line-internal phonological boundaries. I have found this C-group extension to be true in Chaucer's iambic tetrameter, too. In particular, Chaucer tends to promote personal pronouns and auxiliary verbs to clitic hosts, creating new subdivisions within a line: (18) a.

2//6 3//5

b. 4//4 2116

[ c Whan I] // [ c her blisful] [ c songes] / [ c herde], RomA 500 [ c Whan I herde] // [ c her blisful] [ c songes], [ c Therfore] [ c y kan] // [ c hem now] [ c beleve], HF 990 [ c Therfore] 11 [ c y kan beleve hem] [ c now].

Run-on lines occur in 203 (18%) of the 1,126 lines, and Chaucer consistently ends them all with attested phonological boundaries: iambic words (21%) as in (19a), proclitic phrases (45%) as in (19b), and other types (34%) as in (19c). (19) a.


b. c.

m u m 1/3//3/1

[ c Portraied] [ c without] // [ c and wel] [ c entailled] RomA 140 [ c Amonge al this] // [ c I fond] [ c a tale] BD 60 [ c Thoo] [ c atte last] // [ c aspyed] [ c y] HF 1320

English iambic tetrameter verse


In (19c), the pronoun "y" is promoted to a clitic host, and nowhere else in the sample did I find a single case of enjambment that separates a determiner from its head noun or a periphrastic auxiliary (excluding modals) from its head verb. Thus, Chaucer seems to avoid enjambment such as that in (20), employed by Wordsworth, where the romantic poet seems to suggest that just as the lines are bound together, so are "my days": (20) And I could wish my days to be Bound each to each by natural piety. ("My Heart Leaps Up" 8-9) There is one line in my data whose primary boundary is ambiguous: (21) Duchesse, // ne countesse, // ne chasteleyn,

RomB 3740

So, using as guides punctuation marks, Hayes's prosodic hierarchy, and Youmans's C-group extension, I analyzed all the sampled lines of Chaucer to determine primary line-internal and line-final pauses, and I present the results in Table 3 below: Table 3. Distributions of primary line-internal and line-final pauses in 1,126 lines of Chaucer's iambic tetrameter (excluding the line in (21)) After the position of Primary subdivisions Percentages

























As Table 3 indicates, Chaucer distributes his primary phonological pauses in close conformity to (10), a prototypical ranking of occurrences of the primary metrical boundaries, as shown in (22): (22)

A conformity comparison between the prototypically ranked frequencies of primary metrical boundaries in (10) and Chaucer's actual primary phonological pauses in Table 3: SB7 « SB1 « SB3 « SB5 « FB3 « FBI « HB1 « LB S B 1 « SB7 « SB3 « SB5 « FB3 « FB1 « HB 1 « LB

(10) (Table 3)

330 Xingzhong


As is shown in (22), Chaucer systematically observes (10) except for the underlined SB1 and SB7, which differ by only 0.1%, an insignificant statistic. Specifically, in addition to LB, by far the most common location for primary subdivision is HB1 (after the fourth position), resulting in prototypical 4//4 lines. The next and the third most common subdivisions are FBI (after the second position) and FB3 (after the sixth position) respectively. Lines with 2//6 or 6//2 subdivisions reinforce iambic foot boundaries, and they differ from 4//4 lines only in the location of the major caesura. Both 2//6 and 6//2 lines shift the prototypical hemistich boundary to a subsidiary foot boundary. Typically, these lines are subdivided further. For example, lines divided with a 2//6 pattern tend to isolate the final (and most prominent) foot, usually as proclitic phrases, resulting in 2//4/2 lines as in (15b); whereas 6//2 lines tend to divide 4/2//2, isolating the third (now the second most prominent) foot as either a proclitic phrase or as iambic word, as in (16g). Thus, 2//6 lines are often subdivided into either 2//4/2 or 2//2/4 lines, and their mirror image of 6//2 lines into either 2/4//2 or 4/2//2. Such subdivisions affect a line's metrical boundaries, and consequently, a line's prototypical stress profile and the saliency of individual feet. Hence, these variants require prototypes of their own for the purpose of computing metrical tension (for details that bear on this issue, see Li [1995: 168-177]). Put together lines of 4//4, 2//6, and 6//2 subdivisions, their number reaches 896, taking up 79.6% of the total sampled lines, demonstrating that (1) is statistically normative with respect to phrase boundaries and that Chaucer's octosyllabic lines are a foot-based sequence of [WS] [WS] [WS] [WS] rather than an unfooted sequence of alternatingly unstressed and stressed syllables WS WS WS WS.


Syllables and metrical positions

A syllable-counting meter counts syllables into metrical positions in a line. There is evidence that Chaucer was conscious of syllable count in his metrical composition, for he acknowledged that "som vers fayle in a sillable," and that he did "no diligence / To shewe craft, but ο sentence" (HF 1098-1100). Normally, each metrical position should match one syllable, but Chaucer allows certain kinds of variation by occasionally filling a position with no or more than one syllable, implying that it is not syllables but some other metrical principles that come into play.

English iambic tetrameter verse 331

In the 1,126 sampled lines, I identified a total of 89 missing syllables, each occurring only once in a given line (7.9%), and I present a few examples below: (23) a. 0 Peyntid in the cristall there. b. I was ryght yong, 0 soth to say,

RomA 1600 BD 1090

There are two syllables that seem to be missing in ambiguous positions: (24) a. The righte weye 0 ben goon. b. The righte weye ben 0 goon.

(?) (?)

(25) a. Anoon-ryght 01 wente ner; (?) b. Anoon-0 ryght, I wente ner; (?)

RomB 4000

BD 450

I counted (24a) because Chaucer, unlike Wordsworth, does not split periphrastic auxiliaries from head verbs in his clitic groups, and (25a) because it is morphologically preferred. Table 4 below presents the percentage distributions of the 89 lines with missing syllables in different metrical positions: Table 4. Missing syllables and their distributions Position Occurrences Percentages

1 76 85.4

2 0 0

3 0 0

4 1 1.1

5 11 12.4

6 1 1.1

7 0 0

8 0 0

Obviously, the majority of the unfilled positions occur line-initially, and hence can be accounted for by headless lines. The rest occur in the fifth position, so there are headless hemistichs. Omissions occurring in position five are immediately after the caesura in what would otherwise be prototypical 4//4 lines. Hence, headless hemistichs may be explained as an extension of headlessness to the second hemistichs. Such facts illustrate a significant Chaucerian metrical principle as in (26). (26) Headlessness principle: Metrical W positions in Chaucerian iambic tetrameter verse are occasionally realized as 0 at the beginning of a line and more occasionally at the beginning of the second hemistich, but almost never elsewhere.

332 Xingzhong Li

While some of Chaucer's lines "fayle in a sillable," others show the opposite tendency: they include one or more extra weak syllables. Excluding all the lines containing possible contractions, elisions, resolutions, and the word-final -e, I found 55 lines that contain extrametrical syllables in the sample, and I give their percentage distributions in Table 5: Table 5. Extrametrical syllables and their distributions Foot Boundaries Occurrences Percentages

1 2 3.6

2 14 25.5

3 0 0

4 39 70.9

The majority (70.9%) of extrametrical syllables occur at the end of the line, about one fourth of them (25.5%) at the end of the first hemistich, and only about 3.6% of them at the end of the first foot - rare, but permissible: (27) a.

Ye that ageyns youre love mistakith, RomA 1540 1 2 3 4 5 6 7 8 χ


My wyt ys foly, my day ys nyght, 1 2 3 4x 5 6 7 8

BD 610


My lady, that is so fair and bryght! 1 2 x 3 4 5 6 7 8

BD 1180

No lines include extrametrical syllables at the end of the third foot. Thus, headlessness and extrametricality are mirror-image principles in Chaucer's iambic verse: metrical W positions are omitted most often at the beginning of a line and next most often at the beginning of the second hemistich; conversely, metrical W positions are added most often at the end of the line and next most often at the end of the first hemistich. Taken together, these tendencies lend strong support to (1) that the metrical structure of Chaucer's octo-syllabic verse is dipodic. Like other iambic poets, Chaucer occasionally includes two unstressed (or weakly stressed) syllables in a metrical W position:

English iambic tetrameter verse

(28) a.



[W W S] And mat in the myd poynt of the checker 1 2 3 4 5 6 7 8 χ

BD 660

[SW S] And though they die, they sette not a lek. 1 2 3 4 5 6 7 8

RomB 5730

[W S S] That so swetely smelleth in my nose. 1 2 3 4 5 6 7 8

RomB 3660


Of the 35 anapests (only 3.1% of the sampled lines) I found, 27 (77.1%) appear as the "light" [WWS] pattern, as in (28a). By contrast, the "heavy" [SWS] pattern in (28b) rarely occurs in Chaucer's tetrameter - j u s t once in the 1,126 lines. It is in RomB. Slightly more common is the "heavier" [WSS] pattern, as in (28c) - seven out of the thirty five, and all occur exclusively in RomB. This fact suggests that RomB follows a different accentual principle and, therefore, seems to have been translated by a different poet other than Chaucer. Such metrical evidence corroborates the claim by some Chaucerian scholars that RomB is not Chaucer's work (Benson 1987: 686). Obviously, then, Chaucer's stress norm for anapests is WWS, and his underlying metrical pattern for iambic tetrameter remains duple rather than triple even if he occasionally permits split positions. But this is just what (1) predicts. 4. Evaluating the prototype in (1) and other metrical rules Metrical rules, including prototype (1), can be evaluated by syntactic inversions (Youmans 1983, 1989, 1996; Li 1995; Youmans and Li 2002) because they "reinforce rhyme schemes and/or increase metrical regularity" (Youmans and Li 2002: 162). For example, (29)

Whan that Aprill with his shoures soote ... * soote shoures

A.GP. 1


Before me stont, clad in asure, *Stont before me,...

Anel 330



In (29), an adjective-noun sequence transforms into a noun-adjective one to enforce rhyme (soote : roote); whereas in (30) a verb-complement sequence transforms into a complement-verb one to regularize rhythm. Syntactic inversions like those in (29) and (30) comprise the overwhelming majority of occurrences in Chaucer's iambic tetrameter. In this study, I found that some 80% (i.e. 345) of the 431 inversions identified shift a rhyme to the end of a line, and that the remaining 20% (86 cases) affect rhythm. Conversely, no syntactic inversions shift a rhyme away from the end of the line because line-final rhyme is a categorical requirement in Chaucer's verse. More significantly, I found a perfect correlation between syntactic inversions and strongly normative generative metrical rules - the Stress Maximum Principle (SMP) (Halle and Keyser 1966, 1971) and the Monosyllabic Word Constraint (MWC) (Kiparsky 1977) - as Table 6 illustrates. Table 6. Syntactic inversions (SI) with no effect on rhyme in Chaucer

Verse sample Octosyllabic lines

Total lines 1126

SI with no effect on rhyme 86

SI preventing violations of the SMP or MWC 58 (67.4%)

SI Causing violations of the SMP or MWC 0

This correlation is close to that in Youmans and Li (2002: 162): of the 20,465 iambic pentameter lines by Chaucer, Shakespeare, and Milton they analyzed, they found 3,397 syntactic inversions with no effect on rhyme and a correlation between those syntactic inversions and the SMP and/or MWC at 59%. Hence, "like the correlation between syntactic inversions and rhyme schemes, the correlation between syntactic inversions and strong normative metrical rules is almost perfect, and the statistical probability that this correlation occurs by chance is nearly zero" (Youmans and Li 2002: 163). Below, I give a few examples to illustrate how syntactic inversions prevent the violations of the SMP and the MWC. Briefly, the SMP prohibits a sequence of a fully stressed syllable flanked by two unstressed syllables in the same syntactic constituent from occupying odd-numbered metrical positions. The MWC requires all stresses, if mapped to weak

English iambic tetrameter verse


metrical positions, to be monosyllabic words unless a phrase boundary immediately precedes these stresses. (31)

And some she graunted the contraire w s w s w w w s *And she graunted some ... w w s w s



(graunted violates both the SMP and the MWC)

That may me hele; but that is don. w w w s *That may hele me ... w w s w

BD 40

(hele violates the SMP only)

With floures feie, faire under fete, w SWS With fele floures ... w s s w

HF 1540

BD 400

(floures violates the MWC only)

All inversions that can be accounted for by the SMP and/or the MWC can also be accounted for by (1) because all those inversions increase metrical regularity and make the containing lines more closely conform to it. However, there are 28 syntactic inversions (32.6% of 86) that are constrained by neither the SMP nor the MWC, a fact that shows the two metrical rules are insufficient by themselves to account for Chaucer's syntactic inversions. For example, (34)

For yong she was, and hewed bright, RomA 1030 For she was yong,... (permitted by both SMP and MWC)

Hence, we are forced to search for additional metrical principles. Earlier in this paper, I have shown that Chaucer tends to take advantage of movement transformations to augment phonological boundaries, clitic phrases in particular, so as to prevent minor category words from attaching themselves to their clitic hosts and promote them to clitic hosts. Thus, in (34), the C-group [ c for she was yong] is promoted to two C-groups - [ c for



yong] [ c she was] - each of which coincides with an iambic foot, tilting the line strongly toward (1). Of the 28 "cruxes," ten are accounted for in this way. Another effect of syntactic inversions is to shift metrical irregularities from more salient feet to less salient ones, thus reducing metrical tension and conforming more closely to the ranked foot saliency principle (7). For example, (35) a. Such game fonde they in her hod. s s w s w w w s

HF 1810

b. They fonde such game ... (permitted by both SMP and MWC) w s s s w The verb-object sequence in (35b) transforms to an object-verb sequence in (35a) to shift a spondee from the second foot to the first foot, as shown in (36): (36) Moving metrical deviations forward for reduction of metrical complexity Fl « 4

F3 «

F2 « I


Consequently, the line conforms more closely to the ranked foot saliency in (7). Of the remaining 18 unexplained cruxes, nine fall into this category. Still another effect of syntactic inversions in Chaucer is to bring irregular stress patterns into closer conformity with the stress pattern predicted by (1), as in (37) below, though there are only four such cases: (37) a.


But men this thenken evermore, w s w s w s w s

RomB 4840

But men thenken this ... (permitted by both SMP and MWC) w s s w w

The paraphrase in (37b) does not violate the SMP; nor does it violate the MWC, since the underlined trochee is at the onset of the verb phrase

English iambic tetrameter verse 337 men NP] [VP thenken this. Such lines are disfavored, but not prohibited, by the tension metric of both rules, as well as by (1). We are now left with five cruxes, like those given in (38) and (39): (38) a.


(39) a.


But he can cherles / daunten, // he, s w s w

RomA 880

... daunten cherles, // he s w s w Stode I to loken or to poure, s w

RomA 1640

I stode... w s

Obviously, neither the SMP or the MWC, nor tension metrics in general can explain inversions like the one in (38a), because the verse line and its paraphrase are nearly identical for metrical purposes. However, the ranked occurrences of primary metrical boundary principle in (10) can account for the inversion. By fronting cherles, the inversion creates a slightly more salient C-group boundary after it and shifts the boundary from after position seven - the least preferred boundary - to after position five (a more preferred boundary), thus increasing the metricality of the line, as (40) shows: (40)

A metricality-enhancing boundary movement: SB7 «

SB1 « SB3 «

« FB3 « FBI « HB1 «


Although inversions like the one in (39a) have negative metrical effect (only three cases found), all of them occur phrase initially, where trochaic inversions are the most common, as shown in Table 2. In sum, of the 86 syntactic inversions that affect rhythm, 58 (67.4%) of them can be explained by either the SMP or the MWC, but they can also be well explained by (1). Of the 28 inversions that cannot be explained by the SMP or the MWC, 25 can be explained in terms of (1) and its derived saliency principles. Thus, (1) can explain about 96.5% of the 86 inversions

338 Xingzhong Li that affect rhythm. These inversions either increase the saliency of higherlevel metrical boundaries (such as foot and hemistich boundaries) or shift metrical irregularities from more salient to less salient feet. Only three inversions have negative effects, but they result in minor, and permissible, phrase-initial inversions. Hence, the syntactic inversion test strongly supports (1), and it does so more strongly than it does the SMP or the MWC.

5. Conclusion In general, all metrical rules fall into two major classes: either categorical or gradient. Rules such as the SMP and the MWC are more concerned with defining an exact boundary between metrical and unmetrical lines. Recent work in this tradition, such as Hanson and Kiparsky (1996 and 1997), continues to treat rules such as the MWC as categorical constraints rather than as gradient rules - despite counterexamples to the MWC cited in Youmans (1983, 1989, 1996) and Li (1995). By contrast, gradient rules focus on discovering abstract prototypes of different kinds of meter and examining a given line's gradient conformity to the prototype to which it belongs for determining its relative degree of metricality. The prototype (1) is such a frequency-motivated, hierarchically structured, foot-based, saliency-ranked prototype and is thus consistent with Optimality Theory (Prince and Smolensky 1993). Statistical evidence from Chaucer's octosyllabic lines - stress patterning, line-final divisions, subdivisions within the lines, missing syllables, extrametrical syllables, and syntactic inversions - all strongly tend to corroborate the prototype in (1). It portrays metricality not as a categorical concept but as one of relative well-formedness with gradient, violable principles, as is also suggested in Hayes and Kaun (1996), Golston (1998), and Hayes (2000). More significantly, it prompts us to understand Chaucer's octosyllabic lines as gradience-based iambic tetrameter verse, a long metrical tradition precursored by poems such as the first half lines in Ormulum (Minkova 1996), developed and popularized by Chaucer, and carried down to us by many post-Chaucerian iambic poets.

English iambic tetrameter verse 339

Notes 1.

I would like to thank Gilbert Youmans, whose Prototype Theory, notably in his 1989, 1996 articles and in his co-authored 2002 article with the current author on English iambic pentameter, forms the theoretical basis of this article on English iambic tetrameter. I would also like to thank two anonymous reviewers, whose critical commentary and thoughtful suggestions greatly benefited this article. None, however, is responsible for errors and shortcomings.

References Benson, Larry D. (ed.) 1987 The Riverside Chaucer, 3rd ed. Boston: Houghton Mifflin Company. Cable, Thomas 2002 Issues for a new history of English prosody. In Studies in the History of the English Language: A Millennial Perspective, Donka Minkova and Robert Stockwell (eds.), 125-151. Berlin/New York: Mouton de Gruyter. Fabb, Nigel 2002 Language and Literary Structure: The Linguistic Analysis of Form in Verse and Narrative. Cambridge: Cambridge University Press. Golston, Chris 1998 Constraint-based metrics. Natural Language and Linguistic Theory 16: 719-770. Halle, Morris, and Samuel Jay Keyser 1966 Chaucer and the study of prosody. College English 28:187-219. 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. 1997 The nature of verse and its consequences for the mixed form. In Prosimetrum: Cross-cultural Perspectives on Narrative in Prose and Verse, Joseph Harris and Karl Reichi (eds.), 17-44. Cambridge: Brewer. Hayes, Bruce 1989 The prosodic hierarchy in meter. In Phonetics and Phonology: Rhythm and Meter I, Paul Kiparsky and Gilbert Youmans (eds.), 201-260. San Diego: Academic Press. 2000 Gradient well-formedness in Optimality Theory. In Optimality Theory: Phonology, Syntax, and Acquisition, Joost Dekkers, Frank

340 Xingzhong Li van der Leeuw, and Jeroen van de Weijer (eds.), 88-120. New York: Oxford University Press. Hayes, Bruce, and Abigail Kaun 1996 The role of phonological phrasing in sung and chanted verse. The Linguistic Review 13: 243-303. Hayes, Bruce, and Margaret MacEachern 1998 Quatrain form in English folk verse. Language 74: 473-507. Hopkins, Kenneth D., Gene V. Glass, and B. R. Hopkins 1987 Basic Statistics for the Behavioral Sciences. 2nd ed. Englewood Cliffs, New Jersey: Prentice Hall. Kiparsky, Paul 1975 Stress, syntax, and meter. Language 51: 576-616. 1977 The rhythmic structure of English verse. Linguistic Inquiry 8: 189— 247. Li, Xingzhong 1995 Chaucer's meters. Ph.D. diss., Department of English, University of Missouri, Columbia. Liberman, Mark, and Alan Prince 1977 On stress and linguistic rhythm. Linguistic Inquiry 8: 249-336. Minkova, Donka 1996 Non-primary stress in early Middle English accentual-syllabic verse. In English Historical Metrics, C. B. McCully and J. J. Anderson (eds.), 95-119. Cambridge: Cambridge University Press. Murfin, Ross, and Supryia M. Ray 1997 The Bedford Glossary of Critical and Literary Terms. Boston: Bedford Books. Nespor, Marina, and Irene Vogel 1982 Prosodic domains of external sandhi rules. In The Structure of Phonological Representations, Part I., Harry van der Hulst and Norval Smith (eds.), 225-255. Dordrecht: Foris Publications. 1983 Prosodic structure above the word. In Prosody: Models and Measurements, Anne Cutler and D. Robert Ladd (eds.), 123-140. Berlin: Springer-Verlag. Prince, Alan and Paul Smolensky 1993 Optimality theory: Constraint interaction in generative grammar. Unpublished ms. Rutgers University and University of Colorado, Boulder. Selkirk, Elisabeth O. 1980 Prosodic domains in phonology: Sanskrit revisited. In Juncture: A Collection of Original Papers, Mark Aronoff and Mary-Louise Kean (eds.), 107-129. Saratoga, Calif.: Anma Libri. 1981a On prosodic structure and its relation to syntactic structure. In Nordic Prosody II, Thorstein Fretheim (ed.), 111-140. Trondheim: Tapir.

English iambic tetrameter verse 341 1981b

On the nature of phonological representation. In The Cognitive Representation of Speech, Terry Myers, John Laver, and John Anderson (eds.), 379-388. Amsterdam: North-Holland Publishing Co. Tarlinskaja, Marina 1976 English Verse: Theory and History. The Hague: Mouton. Wordsworth, William 1967 My heart leaps up when I behold. In English Romantic Writers, David Perkins (ed.), 279. San Diego: Harcourt Brace Jovanovich. Youmans, Gilbert 1983 Generative tests for generative meter. Language 59: 67-92. 1989 Milton's meter. In Phonetics and Phonology: Rhythm and Meter I, Paul Kiparsky and Gilbert Youmans (eds.), 341-379. San Diego: Academic Press. 1996 Reconsidering Chaucer's prosody. In English Historical Metrics, C. B. McCully and J. J. Anderson (eds.), 185-209. Cambridge: Cambridge University Press. Youmans, Gilbert and Xingzhong Li 2002 Chaucer: Folk poet or litterateur? In Studies in the History of the English Language: A Millennial Perspective, Donka Minkova and Robert Stockwell (eds.), 153-175. Berlin/New York: Mouton de Gruyter.

Early English clause structure change in a stochastic optimality theory setting Brady Z. Clark

1. Introduction Early English clause structure variation and change has typically been modeled in terms of competing grammars (see Fuss and Trips 2002; Kroch 1989; Kroch and Taylor 2001; Pintzuk 1999).' In this paper, I argue that there are empirical and conceptual reasons to prefer a stochastic optimality theory (Boersma 1998; Boersma and Hayes 2001) account over the competing grammars account. In Section 2, I discuss basic facts about clause structure variation and change in Old English and Middle English. In Section 3, I provide a stochastic optimality theory account of these facts. I end with a discussion of some advantages of the stochastic optimality theory account over the competing grammars account. 2. Clause structure variation in early English The standard account of clause structure change in early English (van Kemenade 1987; Lightfoot 1991) is that, for clauses with a finite semiauxiliary verb and a non-finite verb, Old English clauses were right-headed and Middle English clauses were left-headed. Recent work by Allen (2000), Kroch and Taylor (2001), and Pintzuk (1999) has shown that individual texts in Old English and early Middle English exhibited the three structures listed in (la-c). The structure in (Id) is unattested (although see van Bergen 2003: 56-57 for some possible counterexamples). I call the right-headed structure in (la) the "all-final" construction - as in,you God's commandment keep will. I call the left-headed structure in (lb) the "allmedial" construction - as in, you will keep God's commandment. I call the structure in (lc) the "brace" construction - as in, you will God's commandment keep. I call the unattested structure in (Id), the "reverse brace" construction - as in, you keep God's commandment will.

344 Brady Ζ. Clark (1) a. All-final (you God's commandment keep will): S



b. All-medial (you will keep God's


IP / \ XP I' / \ I VP V-inf


c. Brace construction (you will God's commandment


IP / \ xp r / \ I VP YP


d. *Reverse brace construction (you keep God's commandment IP (unattested)




Early English clause structure change 345 Note that I assume that the appropriate structure for the order Vinf-Vfil1 is an S, rather than an IP (cf. Pintzuk 1999), as in (la). Kiparsky (1996: 165-166) discusses evidence that all-final orders do not have a distinct right-headed I(nfl) projection; see also Zwart (1997: 52-57). For example, the inseparability of verb clusters and particle-verb combinations is difficult to explain on the assumption that the finite verb appears in the head of a right-headed IP in OV languages (Zwart 1997). Van Gelderen (1993) makes a similar point for Old English. More generally, the methodological approach assumed here is that syntactic categories like I are presumed to exist in a language only if they are motivated by direct evidence (Dalrymple 2001: 59); see also Cinque (2002: 3 4). In Section 3.1,1 show that the (proto-)theory of phrase structure proposed in this paper derives the absence of head-final IP structures. Broadly speaking, in late Old English, the all-final construction dominates in subordinate clauses, while the all-medial and the brace construction appear at lower frequencies (Pintzuk 1999). In early Middle English, the all-medial construction dominates in subordinate clauses, while the all-final and the brace construction appear at lower frequencies (Kroch and Taylor 2001). In the next section, I present representative textual evidence for the coexisting structures in (la-c). The examples are drawn from Fischer et al. (2000), Kroch and Taylor (2001), and Pintzuk (1996, 1999). The evidence I give for the different structures is from different texts, but Pintzuk (1999) and Kroch and Taylor (2001) provide frequency distributions that show that all three structures coexist within the texts they analyzed. Subordinate clauses are used to illustrate the different structures because orders like all-final are vanishingly small in early Middle English main clauses (Kroch and Taylor 2001). For the purposes of this paper, I remain agnostic about what explains the frequency distributions discussed below. Two recent discussions of this issue include Trips (2002) and Kiparsky (1996). Trips (2002) argues that the low frequency of all-final orders in early Middle English is a consequence of contact with Scandinavian. Kiparsky (1996) accounts for the OV to VO shift in terms of an interaction between an underlying preference for uniform directionality of head-complement relations (the "efficient" cause) and the attrition of evidence for OV structures (the "enabling" cause).


Evidence for structures (1 a-c) in late Old English

Examples (3)-(5) provide representative textual examples of structures (la-c) in late Old English.

346 Brady Ζ. Clark 2.1.1. Evidence for all-final in late Old English Evidence for the all-final structure in (la) comes from clauses in which a non-finite main verb follows its complement and finite semi-auxiliary verb follows the main verb (Pintzuk 1995; Pintzuk 1999). In (3), the non-finite verb forwiernan 'prevent' follows its complements and the finite verb mehte 'could' is preceded by more than one heavy constituent. (3)

him peer se gionga cyningpees oferfcereldes forwiernan mehte him there the young king the crossing prevent could '...the young king could prevent him from crossing there'. (c894, Orosius 44.19-20 [SOURCE: Pintzuk (1996: 245)])

2.1.2. Evidence for all-medial in late Old English Evidence for the all-medial structure in (lb) comes from clauses in which a finite semi-auxiliary verb is in second position and a particle, object pronoun, or monosyllabic adverb follows a non-finite main verb (Pintzuk 1995, 1999). In (4), the finite semi-auxiliary verb wolde 'would' is in second position. The non-finite verb adrcefan 'drive' is followed by the particle ut 'out'. (4)

he wolde adrcefan ut anne cepeling he would drive out a prince '...he would drive out a prince...' (clOOO, ChronB(T) 82.18-19 [SOURCE : Pintzuk (1999: 104)])

2.1.3. Evidence for the brace construction in late Old English Evidence for the brace construction structure in (lc) comes from clauses in which a finite semi-auxiliary verb is in second position and a nonquantificational object precedes a non-finite main verb (Kroch and Taylor 2001). In (5), the non-quantificational N P p a s y n f u l l a n sawle 'the sinful soul' follows the finite verb mceg 'can' and precedes the non-finite main verb geliffcestan 'endow-with-life'.

Early English clause structure change



He mcegpa synfullan sowie purh his gife geliffcestan he can the sinful soul through his gift endow-with-life 'He can endow the sinful soul with life through his grace' (c991, jElfric's Homilies I, 33.496.30 [SOURCE: Fischer et al. (2000: 143)])

2.1.4. Summary In sum, there is evidence for three different clause structures in late Old English, presented in (6). The reverse brace construction in (6d) is unattested. (6)


a. if you God's commandment keep will [all-final] {α, β} > β

364 Brady Ζ. Clark Further, stochastic optimality theory predicts the 'S'-curve in certain situations - a property also held by the competing grammars model proposed in Yang (2000). If the difference in ranking of two constraints which are crucial to the choice between two outputs A and Β are changing linearly, then the proportion of output A is given by the logistic curve (Bresnan, Dingare, and Manning 2001: 28). However, if the constraints are not changing linearly, then the 'S'-curve is not predicted. While the 'S'-curve has been shown to hold for some types of change (e.g., the rise of rfo-support in English; see Kroch 1989), there are possible exceptions. For example, while the general trend in late Old English and early Middle English is toward Infl-medial (constructions where the finite verb is in medial position), the increase of Infl-medial was not a simple progression where the Infl-medial structure increased from one point on the curve to the next; see Pintzuk (1999: 236) and Allen (2000: 9). Finally, long periods of stability are possible in the stochastic optimality theory approach to syntactic change; i.e., there is no internal pressure e.g., in language acquisition—for stochastic aspects of the ranking to disappear over time (cf. Kroch 1994). That is, stochastic optimality theory, as presented here, does not differentiate between long and short periods of stability. This seems to fit with the facts. The coexistence of two forms can last for a long period of time: e.g., in the case of OV and VO order in early English. Alternatively, two forms can coexist for only a brief period: e.g., "the brief development and demise of the 'egressive' aspectual verbs stint and fin (meaning approximately 'leave off V-ing', 'stop V-ing')" (Hopper and Traugott 1993: 36). To sum up, the stochastic optimality theory approach provides a new framework in which to view syntactic change, yielding empirical and conceptual gains over some previous approaches. Future work will need to determine how this new model of change fits into a more complete picture of change, acquisition, and use.

Appendix: The Gradual Learning Algorithm Starting from an initial state in which the constraints have the same ranking value (selected here arbitrarily as 100), the algorithm compares the output of the current grammar for the assumed underlying form of each learning datum (input-output pairs which the learner assumes to be correct). If the form generated by the gram-

Early English clause structure change


mar is identical to the learning datum, then nothing happens. However, if there is a mismatch, the Gradual Learning Algorithm adjusts the grammar: (34)

a. For any constraint which disfavors the given output, the ranking value of the constraint is moved downward along the scale by a small quantity. b. For any constraint which disfavors the learner's own optimal candidate, the ranking value of the constraint is moved upward along the scale by a small quantity.

The numerical quantity by which the algorithm adjusts the constraints' ranking values is called plasticity (Boersma and Hayes 2001: 52). Figures A.l and A.2 illustrate a mismatch and then a learning step. /Sub], Obj, Complex Tense/ V all-medial (learning datum) brace-construction (learner's output)

Spine-R **t *


*I *



Figure A.I. A mismatch between the learner's form and the adult form /Subj, Obj, Complex Tense/


V all-medial (learning datum) brace-construction (learner's output)


** *

*i *

PG fabar would have merged with the HL pitch contour of 'broOar if the stress shift in 'fabar (destressing of the second syllable) had not been perceptually reanalyzed as pitch perturbation caused by fricative voicing (cf. Garrett and Hale 1993), possibly, ultimately perceived as HL (high - superlow) since tone depression after a voiced obstruent, or tonogenesis, is wellattested.3 Conversely, because an acute pitch contour (HL, as in 1bro6ar) belongs to the class of phonetic shapes that can be robustly approximated (Dogil 2000), it was preserved relatively intact (with the Η tone possibly further elevated to the point of superhigh, in an attempt to avoid the pitch contour merger). This should have resulted in a perceivable difference between HL in 'brodar and HL in 'fabar so that the perceptual contrast between two pitch contours was maintained. The non-serial essence of Optimality Theory is ideally suited for capturing the phenomenon of perceptual contrast preservation like the one in VL. In Petrova (2000), I suggest that Grimm's Law (GL) is best explained within an Optimality Theoretic (OT) framework which integrates constraints that enforce input/output faithfulness with those which enforce output/output contrast (comparative, or dispersion constraints).4 Here I suggest that Verner's Law presents another piece of evidence for the interaction of faithfulness (McCarthy and Prince 1995; Kirchner 1997; Beckman 1999; and others) and dispersion (Benua 1995; Burzio 1997, 1999; Flemming 1995; Padgett 1997; and others), which are complementary in accounting for language change. Whereas dispersion is responsible for enforcing contrast itself, faithfulness is in charge of the adequate segmental composition of the contrast. Suppressed faithfulness obtains when output/ouput contrast dominates faithfulness via markedness, the Optimality Theoretic (OT) hierarchy configuration, which is responsible for both Grimm's Law and Verner's Law. In the case of Verner's Law, the requirement to maintain a HL-LH pitch contour contrast, via markedness constraints, outranks input/output laryngeal faithfulness so that a pitch contour contrast is maintained, whereas laryngeal faithfulness is not observed. First, the relevant data will be presented. Then I will make a case for the abovementioned interpretation of VL, emphasizing the advantages of the proposed approach over the recent pre-OT analyses. Next, I will outline

Perceptual contrast in Verner's Law 373 the integrated faithfulness/dispersion OT framework, which makes it possible to handle the phenomenon of merger avoidance. Finally, I will introduce the relevant OT constraints and present the analysis of VL (including exceptions) in the proposed framework.


Verner's Law: Data

2.1. Prosodic changes There have been a number of different proposals concerning the nature of word stress in PIE and PG, specifically at the time of Verner's Law (see Rooth 1974; Page 1996). A traditional assumption has been that PIE had a pitch-accent system and Germanic had so called dynamic stress. Several scholars (including Verner himself) argued for a mixed system for PG. It has been shown in recent studies that pitch is crucial to the phonetic implementation of both pitch accent and dynamic stress (Ohala 1977; Laver 1994). Namely, dynamic stress systems were shown to involve pitch, duration and intensity, with pitch playing a dominant role (see Page 1996: 11 based on Laver 1994: 515-516). Page maintains that stressed syllables in languages with dynamic stress can often be characterized as being of relatively high or relatively low pitch. Thus Lass (1987) suggests a relatively high pitch for stressed syllables in British English, and a relatively low pitch on stressed syllables in Danish. While the phonetic value attached to the tone in stress languages depends on a variety of factors, there seems to be a tendency to associate stress with the pitch peaks. Thus Hirose (1997: 135) presents a chart that relates the peak of activity of the cricothyroid muscle (responsible for elongation and stretching of the vocal folds) to the stress placement in English, concluding that "...CT [crychothyroid] activation occurs slightly ahead of the pitch peak associated with stressed syllable "(134, emphasis mine). Likewise, in the pitch perception experiments with nonsense words (Pierrehumbert 1979), as reported by Moore (1997: 557), "The stressed syllables ... were associated with peaks in the f0 contour." There is a large overlap in the phonetic implementation of pitch and voicing in obstruents. Thus, according to Gandour and Maddieson (1976: 185), the physical correlate of pitch is fundamental frequency (the rate of vocal cord vibration), which depends on the tension of the vocal cords and the pressure drop across them. The same factors condition voicing in ob-



struents. This fact opens up the possibility of an interaction of pitch in vowels and voicing in obstruents — the possibility is borne out in the phenomenon of tonogenesis in Asian languages (whereby obstruent voicing is reinterpreted as pitch distinctions). The reverse influence (that of pitch on voicing in consonants) has been recently proposed as an account of Verner's Law (see discussion further in the paper). As for the placement of stress, in Proto-Indo-European, in which some morphemes were inherently (or lexically) accented, stress was placed on the left-most accented syllable, or, in the absence of accented vowels, on the left-most syllable, unless additional stress rules operational in affixes intervene (Kiparsky and Halle 1977; Halle 1997; Dogil 2000).5 (1)

accented stem Skt ma'rut-e 'wind' (sg.dat) 'svasr-e 'sister' (sg.dat) (Halle: 1997: 291).

unaccented stem 'duhitar 'daughter' (sg.voc.) 'pad- 'foot' (sg.voc.)

As is obvious from (1), stress coincides with inherent accent in lexically accented 'wind' and 'sister', whereas it falls on the left-most syllable in unaccented words, in the absence of additional stress rules. Lexical accent is lost in Germanic, so that the Germanic stress rule requires that stress fall on the root-initial syllable. (2)PG .'sunu 'word-urn ge-'word-en on-giet-an


'son' ( 'word' ( 'become' 'understand' (Lass 1992: 85)

Sound changes

At the time of the Germanic repositioning of accent to the root-initial syllable6, voiceless fricatives developed voiced fricative allophones in what has been referred to as post-atonic environments. (3) PG 'broQar 'brother' 'seyo 'sow' 'texun 'ten'

cf. fdbar 'father' (Rastorguyeva 1983:40) cf. ndzo 'nose' cf. swai'vro 'mother-in-law' (Garrett and Hale 1993:3).

Perceptual contrast in Verner's Law 375 There exist a number of exceptions to Verner's Law to be accounted for. First, although verbal prefixes never bore a stress,7 the first consonant of the root (which was thus in a position for voicing) was never voiced. (4) Early PG *gi - hafan » Old Saxon (OS) gi- haQan 'have' (Prokosch 1939: 65-66). Whereas voicing applies to root-final ' f , it fails to apply to root-initial ' h \ Second, Verner's Law did not apply after stressless vowels if the target fricative was followed by a voiceless obstruent (i.e., in clusters /ft/, /xt/, /fs/, /xs/): (5) PIE * nep'tii cf. Gmc. *nefti * ok'to *ax'tau

'little girl' (not nevtii, or nevdii) 'eight' (Noyer 1993: 3).

Finally, Gothic exhibits a number of forms (especially in verbal paradigms) which violate Verner's Law8: Infin Past sg Past pi (6) OE ceosan ceas curort curon Got kiusan kaus kusum (Rastorguyeva 1983: 41; Lass 1992: 39)

Past Participle coren9 kusans' choose'

In Middle English, the Germanic Stress Rule is augmented by the Romance (quantity-sensitive) Stress Rule, which dictates that stress must fall on the last heavy syllable in the word within a final three-syllable window. In section 4.3 of this paper, I propose that it is the stress shift (this time, HL>LH) that accompanies the incorporation of the Romance words into English, which accounts for the comeback of Verner's Law in the seventeenth century. (ME po'ssess [s] » ENE po'ssess [z], also cf. NE an'xiety [gz]/ 'anxious [kj]; luxurious [gz]/ 'luxury [kj]) (Dobson 1968: 941). As in PG VL, root-initial fricatives are immune to voicing: [s] remains in resource, research, beside. The questions to be answered are: What is the mechanism of voicing? Why do only fricatives voice? Why is there no voicing root-initially or in obstruent clusters? And last, but not least: Why does VL occur twice, and at the specific times it does? While the majority of the questions above have been tackled with varying degrees of success at different times by

376 Olga Petrova

different scholars, to my knowledge, the last question has never been seriously addressed.


Verner's Law: Pre-OT accounts

Recent accounts of Vemer's Law vary. The seemingly obvious explanation that fricative voicing is conditioned by the following stressed vowel rather than the preceding unstressed vowel appears untenable (since the word-initial fricatives fail to voice), whereas word final fricatives do voice (PIE *ayos 'iron' > PG ayaz > OHG er in Holsinger, Iverson and Salmons 2000: 4; PIE *wiiro-s > PG wiraz > O.Icelandic ver-r in Noyer 1992: 3 based on Bennett 1968: 220).10 On the other hand, the nature of the relationship between the consonant and the preceding unstressed vowel (which conditions voicing) proves hard to explain. Verner (1877) himself assumed that in early Germanic, the pattern of syllabification was (C)VC.V, so that the incidence of the lack of the preceding stress was really upon a tautosyllabic segment." Such a syllabification, however, is hardly plausible (see dicussion in Collinge 1985: 208220). In accounting for the mechanism of voicing, recent pre-OT analyses proceed from two major assumptions: (1) that pitch is crucial to the phonetic implementation of both pitch accent (in PIE) and dynamic stress (in PG) (Ohalal 977; Laver 1994: 515-516; Page 1996: 11); (2) that pitch and voicing are implemented with identical laryngeal gestures (high pitch and voiceless obstruents with stiff vocal folds, low pitch and voiced obstruents with slack vocal folds), which allows for identical laryngeal features for both, whatever they are ([+/-stiff vocal folds], as in Halle and Stevens 1971; Calabrese and Halle 1998; [stiff/slack vocal folds], as in Noyer 1992; [Register], as in Yip 1996). Identical features for tone and voicing in obstruents were first proposed by Halle and Stevens in 1971, and subsequently elaborated in a number of publications (see, in particular, Iverson 1983; McCarthy 1988; Crystal 1989: 126; Halle 1992, 1995; Halle and Stevens 1991; Keyser and Stevens 1994; Stevens 1994; Kenstowicz 1994: 39-41; Clements and Hume 1995: 270; Steriade 1995: 155-156, 172). The motivation for the proposed features is as follows. First, Halle and Stevens (1971) observe that the prime factor that sets vocal folds into vibration is airflow due to the subglottal/supraglottal pressure differential. The pressure drop is much higher for

Perceptual contrast in Verner's Law


sonorants (including vowels) than for obstruents since the latter involve an oral obstruction which traps the supraglottal air, thus increasing the supraglottal pressure. Therefore, it involves greater effort to make the vocal folds vibrate in obstruents than in sonorants. Second, Halle and Stevens (1971) note that the vocal folds' vibration can be inhibited by increasing their stiffness. Crucially, this strategy of voice prevention only works when the pressure drop is small (i.e., in obstruents), whereas with a large pressure drop (i.e., in vowels and sonorants), increasing the stiffness of the vocal folds results in an increase in the rate of vibration, i.e., the fundamental frequency of the sound. Thus vocal fold stiffening has a dual function: it prevents voicing in obstruents, on the one hand, and raises pitch in vowels, on the other (also, see the discussion in Maddieson [1997: 627-629]). In the most recent version of Halle and Stevens' (1991) analysis, the feature [stiff vf] serves to distinguish between voiceless ([+stiff vf]) and voiced ([stiff vf]), and the pitch contrasts in vowels (with [+stiff] for high pitch, and [-stiff] for low pitch), while [+/-slack] serves to distinguish four degrees of pitch in vowels and sonorant consonants. Identical laryngeal features open the possibility of certain laryngeal interactions between obstruents and sonorants, attested in a variety of forms, including the phenomenon of tonogenesis in Sino-Tibetan (Noyer 1992, based on Haudricourt 1954; Kingston and Solnit 1988), whereby voiceless obstruents condition high tone, and voiced obstruents condition low tone. Nevertheless, some scholars have preferred [voice] (Lombardi 1991, 1995a, 1995b, 1995c, 1996, 1999; Rice 1992), choosing the traditional designations Η and L for high and low tone, respectively. For example, Bradshaw (2000), in her analysis of consonant-tone interactions in Suma and Makaa (the languages of Central Africa) resorts to a "synthetic" feature [L/voice], The stiff/slack distinction proposed by Halle and Stevens has been criticized not only because it allows for unattested crosslinguistic patterns (Lombardi 1991: 23), but primarily because, while consonants influence tone, tone, allegedly, does not influence consonants (Hyman 1976; Anderson 1978: 166).12 However, the proposed lack of mutual consonant/tone influence is contested in Maddieson (1976), who demonstrates that low pitch causes voicing of adjacent consonants (Jingpho dialect of Korean; Tankhur Naga, a Tibeto-Burman language; Igbo; Hausa) whereas rising pitch causes glottal closure (Jeh, a Mon-Khmer language; Vietnamese; Latvian, Danish). In particular, Maddieson (1976: 148) observes:


OlgaPetrova As for pitch-bearing segments, the principal active means of pitch control is the adjustment of vocal cord tension. It seems difficult to argue that differences of vocal cord tension do not extend to adjacent segments, given the extensive co-articulation that typically occurs in speech. Differences in vocal cord tension could be expected to have effects on the speed and degree of adjustment of glottal aperture, and on the timing of the cessation or beginning of vocal cord vibration (i.e., on voice onset and offset time).

According to Maddieson (1976), the influence of tone on an adjacent obstruent spreads both right and left. Thus, final voiceless stops voice in certain low-tone environments in Jingpho (rightward spreading). By contrast, glottal fricatives appear to be partially voiced before low tones in the Ohuhu dialect of Igbo (leftward spreading). Thus, Maddieson (1976) examines the data from Luce (1965) regarding the two dialects of Riang, a language in the Palaungic division of Khmer-Nicobarese, and suggests that in the history of these dialects "... a falling [tonal] contour provides a context favoring the interpretation of a preceding stop as voiceless" (133). The effect of low tone on voicing is exemplified in Maddieson (1976) with the data from Jingpho, a Tibeto-Burman language, "... when these segments [final stops] appear voiced, it is the result of the low tone environment, voiceless stops do not alternate with voiced segments after high tone" (134). While Hyman (1976) asserts that "low tone is accompanied by breathiness in Jingpho" and attributes voicing to this feature, Maddieson (1976), quoting Maran (1971), objects, contending that there is no evidence that breathiness is at play in this case. In his view, "[t]he best explanation of Jingpo facts, archaic and synchronic, is that there is, or was, a rule that voiced the final voiceless stops in certain low-tone environments" (135). In a more recent publication, Yip (1995), discussing tone in East Asian languages, supplies additional evidence for the mutual influence between vocalic pitch and voicing in obstruents. Yip (1995: 484-485) follows Bao (1990) and Duanmu (1990), who identify Register (High or Low) and obstruent voicing as one and the same thing, proposing the laryngeal feature model, in which the pitch node depends on the Register. The consequence of such feature geometry is that obstruents should block the spreading of incompatible register values, or assimilate in voicing to the spreading value. Both predictions are borne out in different East Asian Languages (Yip 1995: 485-487). In particular, Yip (1996: 485) presents an example from Wuyi, where the spreading of high registered tones is accompanied by onset devoicing.

Perceptual contrast in Verner's Law


While Yip (1996) does point out that the influence of tone on obstruent voicing is rare, Maddieson (1976) believes this is due to the bias in linguistic literature, rather than due to the real state of affairs: "cases where consonants influence tone are overrepresented, particularly in historical phonology, and cases where tone influences consonants are underrepresented" (131). In an important development, recent pre-OT analyses of a subset of the data under consideration (such as Verner's Law) also attribute accentrelated voicing to the effect of vowel pitch on adjacent segments (Noyer 1992; Page 1992; Calabrese and Halle 1998; Holsinger, Iverson, and Salmons 2000). Thus, for example, according to Calabrese and Halle (1998) VL involved spreading [-stiff vf] from the unstressed vowel to an immediately following continuant. Conversely, Garrett and Hale (1993), Page (1996), Holsinger, Iverson and Salmons (2000) assume that VL involved devoicing of previously voiced fricatives (due to intervocalic lenition) in a stressed (i.e., high pitch) context. Noyer (1992) and Holsinger, Iverson and Salmons (2000), in addition, assume that VL targeted only fricatives because the latter were laryngeally empty at that time.13 However, resorting to the device of autosegmental feature spreading, whatever features it would involve, would be redundant and incongruous with a perceptually-based OT account of Verner's Law, proposed in this paper.14 Halle and Stevens' characterization of obstruent voicing and L tone as [-stiff vocal folds], is now generally regarded as an oversimplification. Speakers control phonation and pitch through a complex interaction of a number of articulatory gestures: adjustment of sub-glottal pressure, glottal abduction/adduction, larynx raising/lowering, adjustment of vocal fold length, and supralaryngeal cavity adjustment, as well as adjustment of vocal fold stiffness (see e.g., Silverman 1996). I suggest that it is auditory, rather than just articulatory features, which are at play in Verner's Law, and the ones that motivate the VL changes are likely to be the perceptual ones. Flemming (1995) suggests that the auditory and articulatory features, while overlapping for some phenomena, aren't just redundant versions of existing features, but play different roles for many phonological processes. In particular, there is ample evidence that there are auditory features like [±raised pitch], as well as the VOT features that mark voicing, and it appears plausible that changes in pitch associated with stress marking could lead to a reanalysis of the fricatives as voiced.



According to Nooteboom (1997: 644), "[h]uman pitch perception is, for signals with clearly defined periodicity, remarkably accurate, the differential threshold (just noticeable difference) being in the order of 0.3 to 0.5 per cent." Specifically, Fujimara and Erickson (1997: 110) maintain that "... in certain situations, the perceptual identity of the consonantal feature (such as voiced/voiceless) may be determined by the local f0 contour." The pre-OT analysis that comes closest to such a perceptual account of VL is that of Hale and Garrett (1993). The authors propose to reconstruct for IE the phonetic realization of accent previously reconstructed for Vedic Sanskrit, whereby the highest pitch in a word occurs at the beginning of the syllable immediately following the syllable with the lexical accent. Under the assumption that PG inherited this pitch distribution, the fricatives unaffected by Verner's Law (remaining voiceless) would be those immediately followed by the highest pitch in a word. Further, Hale and Garrett show that the schematic pitch contours following voiceless fricatives (unaffected by Verner's Law) vs. voiced (affected) fricatives are similar to those associated with voiceless vs. voiced obstruents, respectively (see the diagrams below, reproduced from Hale and Garret (1993: 8-9). after [-voice]

after [+voice]

Figure 1. Schematic obstruent voicing effects on F0



Figure 2. Schmatic reconstructed pre-Proto-Germanic pitch contours

Perceptual contrast in Ferner's Law


According to Hale and Garrett (1993: 9), "That similarity suggests that the unusual property of Verner's Law might have originated when a pitch contour originally determined by accent was reanalyzed as the one determined by voicing." Hale and Garrett claim that lenition affects consonants in the words with both pitch contours, such as 'texun and na'so, yielding /tetfun/ and /nazo/. Subsequently, when the accent system become opaque, the fricative in /tevun/ is reanalyzed as voiceless. Crucially, what distinguishes the interpretation advanced in this paper from all the previous accounts of Verner's Law is the proposal that, like Grimm's Law, Verner's Law is also motivated by the requirement to maintain a surface perceptual contrast. In the proposed OT-based analysis, the shifting of stress to the root-initial syllable in 'faQar induced the perceptual reinterpretation of the pitch contour contrast between fa'Qar and 'broQar due to vowel f0 to that determined by voicing (y"abar versus 'broQar, with the perceived lower tone after δ), crucially maintaining the contrast between the two pitch contours. As will be shown in section 4, the major advantage of the proposed perceptual account of voicing is that it allows an answer to the question why fricative voicing happened at the two specific points in time (PG VL and the VL of the seventeenth century) since a similar explanation holds in the account of VL of the seventeenth century, whereby stress shift induces perceptual interpretation of fricatives as voiced. This question does not appear to be adequately answered in the alternative solution. This alternative suggestion, advanced in a number of pre-OT publications mentioned above, as well as by one of the anonymous reviewers of this paper, is that VL is simply intervocalic lenition, and that voicing is not induced in 'fabar, but, rather, inhibited in 'broQar by the preceding super-high tone. This alternative will be addressed further in this paper, after the OT analysis.


Verner's Law as a manifestation of contrast maintenance: An OT analysis


Theoretical framework

In Petrova (2000), I argue that Grimm's Law results from specific configuration of the integrated dispersion/faithfulness constraint hierarchy conducive to the shift of opposition. Here I suggest that Verner's Law presents

382 Olga Petrova another piece of evidence for the interaction of faithfulness and dispersion. The solution lies in integrating faithfulness and dispersion. In a dispersionbased approach, language change can be viewed as the reranking of the constraints regulating perceptual distance among the segments, both with respect to the constraints stabilizing the number of available contrasting forms and with respect to the constraints prohibiting certain featural cooccurrences (markedness constraints). In a faithfulness-based framework, language change can be regarded as the reranking of the featural faithfulness constraints relative to the featural markedness constraints. As a result, two overlapping constraint subhierarchies obtain: systemic, in which comparative (dispersion) constraints interact with markedness constraints; and non-systemic, in which faithfulness constraints interact with the same markedness constraints. The overlap between the two subhierarchies (manifested through the markedness constraints) allows for the integration of the two into a single hierarchy, with markedness constraints serving as an interface between the two relatively independent modules in OT grammar. The alternative constraint rankings within the integrated hierarchy of systemic and non-systemic constraints yield alternative hierarchy configurations, each of which is indicative of a specific state of featural contrastiveness (Petrova 2000, 2001). When both the constraint enforcing output/output contrast and the one enforcing faithhfulness outrank the relevant markedness constraint, contrastiveness obtains; with the reverse ranking, the relevant feature is noncontrastive. When faithfulness outranks output/output contrast via markedness, dispersion is suppressed; conversely, suppressed faithfulness obtains, whereby output/ouput contrast dominates faithfulness via markedness (7). (7) holds for Verner's Law. Under the pressure of the highest ranked GERMANIC STRESS (a shorthand for the interaction of a number of prosodic constraints [Prince and Smolensky 1993; Beckman 1999], which enforces root-initial stress, HL and LH pitch contours are in danger of completely merging. Such total merger is avoided by constraint configuration as in (7), whereby a requirement to maintain PITCH CONTOUR CONTRAST dominates a positional faithfulness (Beckman 1999) constraint IDENT IO PRESONORANT SPREAD GLOTTIS (which enforces spread glottis faithfulness in presonorant position [Petrova, Plapp, Ringen and Szentgyorgyi 2000] via a markedness constraint *VOICED FRICATIVE (cf. Alderete's [1997] * VOICE OBSTR).



in Verner's Law


(7) OUTPUT/OUTPUT PITCH CONTOUR CONTRAST I •VOICED FRICATIVE I IDENT10 PRESON SPREAD GLOTTIS Voicing enforced by the resulting constraint configuration partially modifies pitch contour in one of the members in a contrasting pair, thus preventing a total pitch contour merger. My analysis resonates with Fox's (1999) interpretation of Middle Chinese tonogenesis precisely as an example of conservation of contrast, whereby the loss of the Middle Chinese voiced / voiceless contrast in stops led to "compensatory" register split in Modern Chinese. The proposed interpretation is also in keeping with abovementioned Garrett and Hale's (1993: 10) suggestion that Verner's Law occurred when the innovated Germanic system of initial stress caused the inherited accentual system to become opaque, which created a problem for the learner and induced the reanalysis of a pitch contour as a contour determined by voicing.


Verner's Law in Proto-Germanic

In general, cross-linguistic data indicate that the greater is the distance between any two classes of segments in the sonority hierarchy, the less likely laryngeal feature sharing between them is (see Ito, Mester and Padgett's [1995] family of constraints NO LINK). From the above, it follows that sharing laryngeal features between obstuents and vowels is far less common than sharing within the class of obstruents or within the class of sonorants. It also follows that vowels and fricatives are likelier to share laryngeal features than vowels and stops since, in terms of sonority, fricatives are closer to vowels than stops are. This conclusion is supported by experimental data. For instance, Garrett and Hale (1993: 9) observe that a tone-to-voice interaction (as in VL) is rare (also, see Yip 1996) and primarily affects fricatives because pitch affects voicing judgments only where a primary cue, voice onset time, is ambiguous, which is rarely the case with presonorant stops, but is true for fricatives, which lack a burst at release (Haggard, Ambler and Callow 1970; Fujimara 1971; Haggard, Summerfield and Roberts 1981; Abramson and Lisker 1985; Whalen, Abramson, Lisker and Mody 1990). The latter generalization can be expressed in OT as the following universally ranked subset of the set of NO LINK con-

384 Olga Petrova straints: N O LINK LAR s , o p .VOcoid » N O LINK LARfticative -vocoid· The constraints (adapted from Ito, Mester and Padgett's [1995: 600]), are formulated in (8) below. (8)

NO LINK LARstop-vocoid: stops and vocoids must not share laryngeal features

NO LINK LARfriCatjves-vocoid· fricatives and vocoids must not share laryngeal features In the analysis, it is suggested that the latter markedness constraint was outranked by a dispersion constraint enforcing pitch contour contrast to yield a voiced fricative as a by-product of pitch readjustment.15 Stops were immune to voicing under Verner's Law since N O LINKstop.vocoid was ranked above the constraint enforcing pitch contour contrast. Notice that the immunity of stops also indicates the high-ranking of the manner-sensitive positional faithfulness constraint IDENT PRESON LAR s t 0 p, which prohibits laryngeal modifications in presonorant stops (for justification, see Petrova 2001: 98). Since both in PG and in the seventeenth century VL coincides with stress shift, it is stress shift that may have triggered fricative voicing. In the case of PG, the obliteration of lexical accent and the shift of stress to the root-initial syllable in PG entailed the obliteration of contrast between alternative accentual patterns (distinctions) that were contrastive in PIE. The shift of stress to a root-initial position was enforced by the reranking of prosodic constraints.16 PIE STRESS represents a ranked set of constraints, the interaction of which yields the following (here, considerably simplified) accentuation pattern: lexically accented syllable are stressed; in the absence of a lexical accent, the left-most syllable is stressed. (9) PIE STRESS: an accented syllable is stressed. In an unaccented word (in absence of lexical stress), the left most syllable is stressed. The above is a shorthand for a set of ranked prosodic constraints which are formulated below. (10) HEAD-MAX (McCarthy 1995): If α is a prosodic head and α 6 Domain (/), then f (a) is a prosodic head.



in Verner's Law


In essence, HEAD MAX requires that lexical accent be preserved, i.e., stress should fall on a lexically accented vowel. (11)

ALIGN L (Ft, PrWd): The left edge of the foot must be aligned with the left edge of the prosodic word.


ALIGN L (Rt, PrWd): The left edge of the root must be aligned with the left edge of the prosodic word.

These constraints belongs to the group of ALIGN constraints, which require that edges of phonological and/or morphological units coincide. In effect, ALIGN L (Ft, PrWd) enforces left-most stress in a Prosodic Word. (13) Foot Form: Trochaic (-') The general pattern of the PIE stress placement is yielded by the following two rankings: (14) PIE STRESS ( a ) HEAD-MAX »

ALIGN L ( F t , P r W d ) , ALIGN L ( R t , P r W d ) ,

which signifies that a lexically accented syllable is stressed even if is it is not left-most in the output. (b) HEAD-MAX » Foot Form, which means that a lexically accented syllable is stressed even if the trochaic foot is degenerate. The crucial ranking within PIE STRESS (HEAD MAX » ALIGN L (Ft, PrWd)) in effect signifies that preserving lexical accent is more important than leftmost stress. The advent of the GERMANIC STRESS RULE, which requires that stress fall on the left-most syllable, was thus manifested in the reversal of the ranking between the HEAD MAX and ALIGN L constraints. With HEAD MAX outranked by the two ALIGN L constraints, the requirement to stress the leftmost root syllable would override the need to preserve lexical accent. The major ranking responsible for Germanic stress is given in (15). ( 1 5 ) GERMANIC STRESS ALIGN L ( F t , P r W d ) , ALIGN L ( R t , P r W d ) »


386 OlgaPetrova Under the pressure of the highest-ranked GERMANIC STRESS, HL and LH pitch contours were in danger of completely merging, which was circumvented by voicing. Specifically, the shifted HL< LH pitch contour of PG fa Oar < PIE fa'bar would have merged with the HL pitch contour of 'brofar had it not been for the the low tone (the auditory feature [-raised tone]), which, while shifting to the root-initial syllable in 'fa Oar, induced the perceptual «interpretation of the pitch contour contrast due to vowel f0 to that determined by voicing fabar, crucially maintaining the contrast between two pitch contours (with relatively low, or super-low, tone after the voiced segment in 'fabar (HL)): HL versus HL. As is demonstrated in section 4.3 of this paper, Verner's Law of the seventeenth century can also be interpreted as a strategy to avoid the perceptual merger of two pitch contours. The dispersion constraint enforcing mulated in (16) below.


is for-

(16) PITCH CONTOUR CONTRAST: Maintain a 2-way pitch contour contrast. The constraint in (16) crucially refers to a contrast in two contours rather than in feature specification. A number of recent publications support the plausibility of this constraint. Thus, Yip (2000), discussing the interaction between tone and prominence in a number of tone languages, proposes a sequential markedness constraint PROMINENCE-TONE MATCH, which refers to the tone contour. Likewise, Akinlabi and Liberman (2000: 1) propose that "a successful account of tonal phonology in terms of universal constraints requires enriching tonal representations to include some simple kinds of structures, which organize tones in somewhat the same way that segments, syllables and feet organize non-tonal features." Specifically, Akinlabi and Liberman (2000) propose a tonal unit consisting of paired HL or LH tones, which function phonetically much like pitch accents in other languages. They further argue that such units, long postulated as underlying elements in accentual systems, also play a crucial role in tonal phonology more generally. I suggest that, as in Grimm's Law, the total merger of the two alternative pitch contours is avoided by a constraint configuration as in (7), elaborated in (17) below, whereby the requirement to maintain a PITCH CONTOUR CONTRAST dominates the positional faithfulness constraints IDENT SPREAD GLOTTIS and IDENT VOICE via markedness constraints NO LINK

Perceptual contrast in Verner's Law


LARfncative -vocoid and *VOICED FRICATIVE. Since stops are immune to voicing under Verner's Law, NO LINKst0p.V0C0id must outrank PITCH CONTOUR CONTRAST.

(17) n NO LINK L A R . s t 0 p . v 0 C 0 i d = = = = — = = = = = NO LINK LAR,top.vocoid I I PITCH CONTOUR CONTRAST | I I NO LINK LAR_fricative•vocoid = = = = = NO LINK LARfHcative vocoid I I »VOICED FRICATIVE -VOICED FRICATIVE I IDENT PRESON VOICE I IDENT PRESON SG

The integrated hierarchy configuration in Verner's Law is thus indicative of suppressed faithfulness, whereby output contrast dominates faithfulness via markedness. Clearly, although contrast preservation in Verner's Law is somewhat different from that in Grimm's Law, the two cases are identical in essence: Grimm's Law involves preserving a three-way VOT contrast, with a different segmental composition (Petrova 2000, 2001); Verner's Law involves preserving the pitch contour contrast with a different segmental composition. Voicing enforced by the resulting constraint configuration partially modifies the pitch contour in one of the members in a contrasting pair, thus preventing a total pitch contour merger. Why did 'brodar escape laryngeal changes? According to Dogil (2000), relative phonetic invariance may determine phonological stability; i.e., in the face of inevitable phonetic variance, the more perceptually robust a contrast is, the more stable it will be over time, or, in Dogil's terms, "sound laws tend to preserve [relatively] invariant phonetic shapes." As an example, Dogil (2000: 7) uses experimental phonetics to approximate the diachronic changes in Lithuanian pitch accent, concluding that the acute pitch contour (i.e., HL, as in brodar, 'texun) is the only one that is preserved intact precisely because it belongs to the class of phonetic shapes that are perceptually salient and,



therefore, can be robustly approximated. Dogil's observation is encoded in this analysis in the following faithfulness constraint: (18) IDENT IO HL PITCH CONTOUR: An output correspondent of an input sequence of segments with a HL pitch contour must have HL pitch contour. IDENT HL PITCH CONTOUR ensures that voicing does not apply in words that have HL pitch contour in the input (such as /texsgun! or /broffgar/): as a result, under the pressure to maintain PITCH CONTOUR CONTRAST, the other member of the pitch contour opposition has to voice. Finally, like Noyer (1992) and Calabrese and Halle (1998), I assume that fricatives in clusters fail to voice under Verner's Law due to the dominance of voicing assimilation (devoicing), in the proposed analysis enforced by SHARE LARYNGEAL, w h i c h c r u c i a l l y o u t r a n k s PITCH CONTOUR CONTRAST. F i r s t , PITCH CONTOUR CONTRAST »

NO LINK L A R ^ . ™ ^ . T h i s r a n k i n g

signifies that, under the requirement to maintain pitch contour contrast, fricatives voice, the respective output violating NO LINK LARfriC-Vocoid(since [-raised pitch] is perceived as shared between a vowel and a preceding f r i c a t i v e ) . S e c o n d , PITCH CONTOUR CONTRAST »


which means that voicing occurs although a spread glottis voiceless fricative surfaces as a non-spread glottis voiced fricative, whereby laryngeal faithfulness is compromised. 18 The interaction of the relevant constraints for VL in PG is given in (19)—(21), where output candidates are pairs of words: in (19) /"broG^ar/ (with a HL input pitch contour specification) 19 contrasts with /f g a'9 s g ar/ (LH in input); in (20), with /i'tsgan/ (also LH, but with intervocalic stop); and in (21), with /nef g, t sg i/ (LH with an obstruent cluster). Voiceless fricatives in (19-21) are spread glottis as a result of Act Π of GL (see Petrova 2001: 114).


contrast in Verner's Law


(19) Inputs: /f g a'0 sg ar/ ['f s aöar] 'father'; /'bhro0sgar/ ['bhro6sgar] 'brother' /•bhroeS8ar/

Germ Stres

/f g a'0 s g ar/

Shar Lar

No Link Pitch Ident HI No Link *Voi Ident Preso Contour Pitch Lar Fricat-Voicid fric SprGl Lar, Contr Contour StopVocoid (VI)

l.'bro0' 8 ar 'f8a&ar < = i

2.'bro0' ar




* * *

f'a^ar 3.'bro0S8ar



W a r 4.'broÖar






In (19) the winning candidate set 1 contains a marked voiced fricative, which results from the perceptual reanalysis in 'fgabar, with the shifted stress. As a result, (1) violates NO LlNKfricatjve.vocoid as well as * VOICE FRIC. Notice also a double violation of IDENT PRESON SG by (1): not only does the murmured stop in bhroQar surface as plain voiced, but the spread glottis 0 ig surfaces as a non-spread glottis δ due to a high-ranked constraint *V0ICE SPREAD GLOTTIS, which reflects the default non-spread-glottis status of voiced fricatives.20 The competing input candidate sets in (19) incur more serious violations. A nearly fully faithful candidate set (2) exhibits no accent shift, consequently violating GERM STRESS, which is crucially ranked above NO LlNKfncative-vocoid· Candidate set (3), with a shifted accent, but with a spread glottis fricative in 'fgaQsgar, complies with GERM STRESS, but fails to comply with the high-ranked PITCH CONTOUR CONTRAST: since, in the absence of "compensatory" fricative voicing, the pitch contours between 'fgaQsgar and 'broQ5gar completely merge. Finally, candidate set 4 exhibits fricative voicing in 'brobsar (the other member of the pitch contour opposition, which has HL contour in the input). As a result, this candidate set succeeds in maintaining a pitch contour contrast, but violates IDENT HL PITCH CONTOUR, which is ranked above NO L l N K ^ t ^ . vocoid and which enforces preferential faithfulness to an input string HL pitch contour.

390 Olga Petrova

(20) Inputs: / " b W W [ΤΛΟΘ^Γ] 'brother'; / i W fitan] 'eat' /'bWar/ /i'tan/

Germ Stres

Ident HI No Link *Voi Ident Shar Lar No Link Pitch Preso Contour Pitch Lar ^ fric Lar S U * . SprGl Contr Contour Vokid Vocoid (VI)

l.'bro0!8ar PG *'wiraz is of a different origin. While the details of the analysis of word-final voicing are outside the scope of this paper, below I just highlight its essence. Namely, I suggest that, although both intervocalic voicing in Verner's Law and voicing at word edges involve perceptual sharing of [-raised pitch], the two processes are triggered by different requirements. While voicing in Verner's Law occurs under the pressure to maintain pitch contour contrast, voicing at word edges is the instantiation of voicing under low stress, exactly parallel to the word-final and word-initial fricative voicing in function words in Late Middle English (e.g., is [is] > [iz]; the [0e] > [be]), whereby shortened fricatives under low stress are perceived as voiced (Berg 1995). There are independent phonetic and extra-linguistic arguments for treating voicing in Verner's Law and word-final voicing



separately. First of all, intervocalic voicing is phonetically justified, whereas voicing at word edges is cross-linguistically rare and physiologically unnatural (Berg 1995; Kirchner 1998: 56-57). Second, unlike voicing in Verner's Law, both in PG and in ME, voicing at word edges primarily involves a closed class of functional items (Bennett 1968: 222; Berg 1995: 188), whereas fricatives at word edges in full lexical items are much more resistant to voicing, even when they occur in low stress environment.26 Third, chronologically, the two kinds of voicing are attested to have occurred independently. For example, in ME, voicing at word edges in functional items preceded "Verner's Law of the seventeenth century" by roughly two centuries. As a whole, I argue that low stress voicing reveals the effect of the dispersion requirement to maintain the minimal inviolable perceptual space allotted to each segment in a viable contrast. In particular, it is shown how the changes in the values of non-contrastive phonetic features (e.g., duration) that contribute to a phonological contrast affected the space allotted to each segment in a relevant contrast, ultimately bringing about phonological consequences. In conclusion, the analysis of Verner's Law demonstrated the interaction of the dispersion constraints on contrast with faithfulness, whereby the dispersion requirement to maintain a surface contrast, via markedness, outranked input/output faithfulness, creating the hierarchy configuration conducive to a sound shift. My account entails the alternative timing of Verner's Law relative to accent shift, whereby accent shift is claimed to induce voicing rather than voicing just being the function of PIE accent placement. This approach allows for a unified analysis of the two instantiations of Verner's Law: in Proto-Germanic and in Early New English.

Notes 1.


I would like to thank Jill Beckman and Catherine Ringen, as well as the two anonymous reviewers, for invaluable comments on this paper. Any possible errors or inaccuracies are entirely mine. Thus, Jespersen (1933) referred to fricative voicing in the seventeenth century as "Verner's Law in Early New English." Admittedly, some scholars (including one of the anonymous reviewers of this paper) deny the seventeenthcentury fricative voicing the status of a sound change, assuming that French words were voiced yet in French and were borrowed as such. See the discussion further in the paper.

Perceptual contrast in Verner's Law 3.


5. 6.



9. 10.



13. 14.


An anonymous review has expressed a legitimate objection that the accentual contrast is likely to be displaced to nearer phonetic categories, so the more extreme H->L differential appears more plausible in HL > HL than in LH > HL. In either case, the goal of preserving the pitch contour contrast would be achieved. Flemming's (1995) Dispersion Theory is formulated as an alternative (rather than a supplement) to input/output faithfulness. In subsequent research, Flemming (1998) even claims that faithfulness is generally redundant. Conversely, proponents of faithfulness reject the need for dispersion in an account of change (Boersma 1998). In Petrova (2001), I present a number of arguments for the relevance of both dispersion and faithfulness, and for the advantage of an integrated dispersion/faithfulness approach to language change. All the stress rules are presented here in their basic (considerably simplified) form. The dating of Vemer's Law proposed here reflects the proposed scenario of Vemers's Law, whereas tradionally Verner's Law is assumed to have occurred before the stress shifted. While in Germanic languages a formerly purely lexical stress shifted to the first syllable of the root, it could also fall on the nominal or adjectival prefixes; however, those were reinterpreted as full lexical items rather than prefixes. By contrast, verbal prefixes, which retained their prefix status, were never stressed (see Lass 1992: 85-86; Hogg 1992: 48-49). The operation of Verner's Law is especially prominent in the Germanic verbal paradigm, characterized by an uneven stress pattern. Thus, PIE infinitive and past tense singular forms bear word-initial stress (thus escaping voicing), whereas past tense plural and past participle forms bear word-final stress, hence word-medial fricatives in those forms are voiced under Verner's Law. [r] in the last two forms derives from [z] as a result of rhotacism. Lerchner (1971:108) accounts for the failure of the word-initial fricatives to undergo Verner's Law by crediting the word-initial position with a special intensity to match its greater information rating (see Collinge 1985: 207). Cf. the discussion of root-initial faithfulness in Beckman (1999), and also in section 4.3 of this paper. This opinion is shared by Abrahams (1949). The tautosyllabicity of the unstressed vowel and the following target obstruent could then fall under the general case of lenition in an unstressed syllable. As discussed in Jessen (1998: 129-130), the full range of evidence in support of Halle and Stevens's proposal (collected through electromyographic studies measuring the activity of the cricothyroid muscle) is not conclusive. A more detailed overview of the recent accounts is presented in Petrova (2001: 172-179). I very much appreciate the helpful suggestions of the two anonymous reviewers regarding the analysis proposed in this paper.

400 Olga Petrova 15. The anonymous reviewer suggested that the reliance on the hierarchy of C-V laryngeal interaction is incongruous in an account which otherwise has aspirations towards deeper functional explanation. The reviewer suggests that a better account would involve dispersion principle of laryngeal contrast maintenance: voicing was already contrastive in stops in Proto-Germanic; stops resisted the Verner's Law voicing that accompanied stress shift. Fricatives, however, could be voiced without neutralizing a contrast. Admittedly, the proposed account of stop/fricative asymmetry does not fit well into the dispersion -based explanation. However, such an account appears problematic for several theoretic and practical reasons, including, among others, the difficulty in contrasting the forms containing both stops and fricatives (as in (20)), as well as the fact that it would significantly complicate the analysis of Gothic exceptions to Verner's Law (as in (22)). I leave this issue for further research. 16. I will only outline the interaction of prosodic constraints here without presenting any tableaux since prosodic constraints in OT have been discussed in detail in numerous OT publications, beginning with the ground-breaking classic manuscript of Prince and Smolensky (1993), and are beyond the scope of this paper. 17. In the diagram a dispersion hierarchy is on the left, and a faithfulness hierarchy is on the right. The markedness constraints appear twice (i.e., once in each hierarchy) for ease of reference. Double lines stretching between the markedness constraints in the two hierarchies, therefore, indicate that the markedness constraints are shared by the hierarchies, and are in each case, in effect, a single constraint serving as an interface between dispersion and faithfulness. 18. Notice also that PITCH CONTOUR CONTR must outrank IDENT PITCH CONTOUR, the latter not included in the hierarchies. This ranking signifies that the requirement to maintain pitch contour contrast is observed although a pitch contour is slightly modified. 19. Although, unlike words such as */'texsgun/ 'ten', which were lexically HL in PIE, words like /b\o9 sg ar/ are claimed not to have lexical accent in PIE (Halle 1997), I assume an input specified as HL due to Lexicon Optimization in early PG, whereby the output of one generation served as an input for another (PIE /bro0S8ar/ > ['bro9sgar] >PG /'bro9sgar/ (e.g., see discussion in Holt 1997: 22). 20. *VOICE SPREAD GLOTTIS has jurisdiction over a superset of segments prohibited by *MURMUR, banning any voice spread glottis segments (see discussion in section 2.1.5 in Petrova 2001). This constraint is not included in the hierarchy since in this case it is irrelevant for the issues being discussed. 21. An earlier account by Boer (1924: 123-124), who also denied the existence of any synchronic stage at which voicing was accent-related, rests on the assumption of co-existence of pitch and accent. More specifically, Boer assumed that Early PG *fa'Qer (derived from PIE *pa'ter) would become *'faber, with primary stress on the first syllable and higher pitch on the second. Notice that his account is compatible with the account proposed here, whereby shift of stress

Perceptual contrast in Venter's Law


occurs simultaneously with voicing, with the output differing in pitch contour from the items with non-shifted stress. 22. Under certain conditions, prefixes bore stress in PIE (cf. Greek perf. Ind. 1 sg. active 'le-loip-a) (Lass 1992: 85). In Germanic, stress shifted rightwards to the root syllable in such cases, which, according to the present analysis, should have induced voicing. The lack of voicing in those cases must have been due to the high-ranking of positional faithfulness constraint IDENT VOICE Rt. initiai(cf- Beckman 1999: 54), which enforced laryngeal faithfulness in a psycholinguistically salient root-initial position. The evidence for this constraint is especially convincing in later Old English and in "Vemer's Law of the seventeenth century" (see Petrova 2001: 251-257). 23. Consider also relatively recent instances of folk etymology, e.g., the French C'est fini with the word-final stress was reanalyzed in English as Symphony, with the root-initial stress. 24. For details of the constraint ROMANCE STRESS, see Petrova (2001: 251252). 25. The given analysis would not explain voicing in some polysyllabic words, such as an'xiety vs. 'anxious; lu'xurious vs. 'luxury. For discussion, see Petrova (2001:256-257). 26. Few examples of voicing in full notional words include: PIE* aios (Skt *'ayas) > Goth aiz (Roberge 1983: 147); ME ac'tif > NE 'active (Bennett 1968: 220). The second case, however, involves a stress shift, so word final voicing in this case might be due to the high rank of PITCH CONTOUR CONTRAST, akin to intervocalic voicing in Verner's Law.

References Abrahams, H. 1949 Etudes Phonetiques sur les Tendances Evolutives des Occlusives Germaniques. Aarhus: Universitetsforlaget. Abramson, Arthur, and Leigh Lisker 1985 Relative power of cues: F0 shift versus voice timing. In Phonetic Linguistics: Essays in Honor of Peter Ladefoged, Victoria Fromkin (ed.), 25-33. Orlando: Academic Press. Akinlabi, Akinbiyi, and Mark Liberman 2000 Tonal complexes. Paper presented at the 31st Conference of the North Eastern Linguistics Society, October 7, 2000. Washington, DC. Alderete, John 1997 Dissimilation as local conjunction. North East Linguistic Society 27: 17-32.

402 Olga Petrova Anderson, Stephen R. 1978 Tone features. In Tone: A Linguistic Survey, Victoria Fromkin (ed.), 133-176. New York: Academic Press. Bao, Zhi-ming 1990 On the nature of tone, Ph.D. diss., MIT. Beckman, Jill N. 1999 Positional Faithfulness. An Optimality Theoretic Treatment of Phonological Asymmetries. New York/London: Garland Publishing. Bennett, William H. 1968 The operation and relative chronology of Vemer's Law, Language 44: 219-223. Boer, Richard Constant 1924 Oergermaansch Handboek. Haarlem: Willink and Zoon. Benua, Laura 1995 Identity effects in morphological truncation. In University of Massachusetts Occasional Papers in Linguistics. 18 papers in Optimality Theory, Jill Beckman, Suzanne Urbanczyk, and Laura Walsh (eds.), 77-136. Amherst, MA: Graduate Linguistic Student Association. Berg, Thomas 1995 Word-final voicing in the history of English: A case of unnatural phonology? English Studies 2: 185-201. Boersma, Paul 1989 Modelling the distribution of consonant inventories by taking a functionalist approach to sound change. Proceedings of the Institute of Phonetic Sciences of the University of Amsterdam 13: 107-123. 1998 Functional Phonology. Formalizing the Interaction between Articulatory and Perceptual Drives. Netherlands Graduate School of Linguistics. The Hague: Holland Academic Graphics. Bradshaw, Mary M. 2000 Consonant-tone interaction as downstep. Paper presented at the 74th Annual Meeting of the Linguistic Society of America, January 7, 2000. Chicago, Illinois. Burzio, Luigi 1997 Cycles, non-derived-environment blocking, and correspondence. Ms., The Johns Hopkins University. 1999 Surface-to-surface morphology: When your representations turn into constraints. Paper presented at the 1999 Maryland Mayfest. University of Maryland, College Park. Calabrese, Andrea, and Morris Halle 1998 Grimm's and Vemer's Law: A new perspective. In Mir Curad. Studies in honor of Calvert Watkins, J. Jasanoff, H. C. Melchert, and L. Oliver

Perceptual contrast in Verner's Law


(eds.), 47-62. Innsbruck: Institut der Sprachwissenschaft der Universität Innsbruck. Clements, George N., and Elizabeth V. Hume 1995 The internal organization of speech sounds. In Handbook of Phonological Theory, John Goldsmith (ed.), 114-174. Cambridge, MA: Blackwell. Collinge, Neville E. 1985 The Laws of Indo-European. Philadelphia, PA: John Benjamins Publishing Company. Crystal, David 1989 A Dictionary of Linguistics and Phonetics. Cambridge, MA: Basil Blackwell. Dobson, Eric J. 1968 English Pronunciation, 1500-1700 Vol. 2. Oxford: Clarendon Press. Dogil, Grzegorz 2000 Phonetic invariance and phonological stability: Lithuanian pitch accents. Paper presented at the University of Iowa Linguistics Colloquium, February 24, 2000. Iowa City, Iowa. 1 Duanmu, San 1990 A formal study of syllable, tone, stress and domain in Chinese Languages, Ph.D. diss., MIT. Flemming, Edward 1995 Auditory representations in phonology, Ph.D. diss. University of California in Los Angeles. 1998 Some arguments against using underlying representations in OT phonology. Handout from the paper presented at the University of California Phonology Workshop, March 3,1998. Los Angeles, CA. Fox, Anthony 1999 Conservation of contrast. Paper presented at the 14th International Conference on Historical Linguistics, August 1999. Vancouver, BC. Fujimara, Osama 1971 Remarks on stop consonants - Synthesis experiments and acoustic cues. In Form and Substance: Phonetic and Linguistic Papers Presented to Eli Fischer-Jorgensen, L.L.Hammerich, Roman Jacobson, Eberhard Zwirner (eds.), 221-232. Copenhagen: Akademisk Forlag. Fujimara, Osama, and Donna Erickson 1997 Acoustic phonetics. In Τhe Handbook of Phonetic Sciences, William J. Hardcastle and John Laver (eds.), 65-115. Cambridge, MA: Blackwell Publishers. Gandour, Jack, and Ian Maddieson 1976 Measuring larynx movement in Standard Thai using the cricothyrometer. Studies on Production and Perception of Tones. UCLA Working Papers in Phonetics 33: 162-191.

404 Olga Petrova Garrett, Andrew, and Mark Haie 1993 The phonetics and phonology of Grimm's and Verner's Law. Paper presented at the Annual Meeting of the Linguistic Society of America. Haggard, Mark, Stephen Ambler, and Mo Callow 1970 Pitch as a Voicing Cue. The Journal of the Acoustical Society of America 47: 613-617. Haggard, Mark, Quentin Summerfield, and Martin Roberts 1981 Psychoacoustical and cultural determinants of phoneme boundaries: Evidence from trading F0 cues in the voiced-voiceless distinction. Journal of Phonetics 9: 49-62. Halle, Morris 1992 Phonological features. In International Encyclopedia of Linguistics. Volume 3, W. Bright (ed.), 207-212. New York/Oxford: Oxford University Press. 1995 Feature geometry and feature spreading. Linguistic Inquiry 26 (1): 1 46. Halle, Morris, and Kenneth Stevens 1971 A note on laryngeal features, In Quarterly Progress Report 101, 198212. Cambridge, MA: Research Laboratory of Electronics, MIT. 1991 Knowledge of language and the sounds of speech. In Music Language, Speech, and Brain, J. Sundberg, L. Nord, & R. Carlson (eds.), 1-19. London: MacMillan. Haudricourt, Andre-Georges 1954 De l'origine des tons en Vietnamien, Journal Asiatique 242: 68-82. Hirose, Hajime 1997 Investigating laryngeal physiology. In The Handbook of Phonetic Sciences, William I. Hardcastle and John Laver (eds.), 114-136. Cambridge, MA: Blackwell Publishers. Hogg, Richard M. 1992 Grammar of Old English Vol. 1. Cambridge, MA: Cambridge University Press. Holt, Eric 1997 The Role of the listener in the historical phonology of Spanish and Portuguese: An Optimality-Theoretic account, Ph.D. diss. Georgetown University. Rutgers Optimality Archive # 278-0898. Holsinger, David, Gregory Iverson, and Joseph Salmons 2000 Verner's Law is passive voicing. Paper presented at the 6th Annual Conference of Germanic Lingustics, April 2000. Milwaukee, WI. Hyman, Larry 1976 On some controversial questions in the study of consonant types and tone. In Studies on Production and Perception of Tones. UCLA Working Papers in Phonetics 33: 90-98.

Perceptual contrast in Verner's Law


Ito, Junko, Armin Mester, and Jaye Padgett 1995 Licensing and underspecification in Optimality Theory. Linguistic Inquiry 26 (4): 571-613. Iverson, Gregory K. 1983 Korean "s." Journal of Phonetics 11: 191-200. Jespersen, Otto 1933 Voiced and voiceless fricatives in English, In Linguistica. Selected papers in English, French and German, Otto Jespersen (ed.), 346-383. London: George Allen and Unwin. Jessen, Michael 1998 Phonetics and Phonology of Tense and Lax Obstruents in German. Amsterdam: John Benjamins. Kenstowicz, Michael 1994 Phonology in Generative Grammar. Cambridge, MA: Blackwell. Keyser, Samuel Jay, and Kenneth N. Stevens 1994 Feature Geometry and the Vocal Tract. Phonology 11: 207-236. Kingston, John, and D. Solnit 1988 The tones of consonants. Ms., Cornell University and University of Michigan, Ann Arbor. Kiparsky, Paul, and Morris Halle 1977 Towards a reconstruction of the Indo-European accent. In Studies in Stress and Accent, SCOPIL Vol. 4, Larry Hyman (ed.), 209-238. Los Angeles: USC. Kirchner, Robert 1997 Contrastiveness and faithfulness, Phonology 14 (1): 83—112. 1998 An effort-based approach to consonant lenition, Ph.D. diss., University of California, Los Angeles. Rutgers Optimality Archive #276-0898. Lass, Roger 1987 The Shape of English. Structure and History. London: J.M. Dent. 1992 Phonology and morphology. In The Cambridge History of the English Language. Vol. 2, Norman Blake (ed.), 23-155. New York: Cambridge University Press. Laver, John 1994 Principles of Phonetics. Cambridge: Cambridge University Press. Lerchner, Gorrhard 1971 Zur II. Lautverschiebung im Rheinisch-Westmitteldeutschen: Diachronische und Diatopische Untersuchungen. Mitteldeutsche Studien 30: 103-119. Lombardi, Linda 1991 Laryngeal Features and Laryngeal Neutralization, Ph.D. diss. University of Massachusetts, Amherst. 1995a Laryngeal neutralization and alignment. In University of Massachusetts Occasional Papers in Linguistics. 18 Papers in Optimality Theory, Jill



Beckman, Suzanne Urbanczyk, and Laura Walsh (eds.), 225-248. Amherst, MA: Graduate Linguistic Student Association. 1995b Laryngeal features and privativity. The Linguistic Review 12: 35-59. 1995c Laryngeal neutralization and syllable well-formedness. Natural Language and Linguistic Theory 13: 39-74. 1996 Postiexical rules and the status of privative features. Phonology 13: 1 38. 1999 Positional faithfulness and voicing assimilation in Optimality Theory. Natural Language and Linguistic Theory 17(1): 276-302. Maddieson, Ian 1976 A further note on tone and consonants. Studies on Production and Perception of Tones. UCLA Working Papers in Phonetics 33: 131159. McCarthy, John 1988 Feature Geometry and Dependency: A Review. Phonetica 43: 84-108. 1995 Extensions of faithfulness: Rotuman revisited, Ms., University of Massachusetts, Amherst. Rutgers Optimality Archive #110-0000. McCarthy, John, and Alan S. Prince 1995 Faithfulness and reduplicative identity. In University of Massachusetts Occasional Papers in Linguistics. 18 Papers in Optimality Theory, Jill Beckman, Suzanne Urbanczyk, and Laura Walsh (eds.), 249-384. Amherst, MA: Graduate Linguistic Student Association. Moore, Brian 1997 Auditory processing related to speech perception. In The Handbook of Phonetic Sciences, William J. Hardcastle and John Laver (eds.), 539565. Cambridge, MA: Blackwell Publishers. Nooteboom, Sieb 1997 Prosody of Speech: Melody and Rhythm. In The Handbook of Phonetic Sciences, William J. Hardcastle and John Laver (eds.), 640673. Cambridge, MA: Blackwell Publishers. Noyer, Rolph 1992 Vemer's Law and underspecification theory. Ms., MIT. Presented at the 66th Annual meeting of the Linguistic Society of America. PA: Philadelphia. Ohala, John J. 1977 The physiology of stress. In Studies in Stress and Accent. (Southern California Occasional Papers in Linguistics 4.) Larry Hyman (ed.), 145-168. Los Angeles: University of Southern California. Padgett, Jaye 1997 Candidates as systems: Saussure lives! Paper presented at the Hopkins Optimality Workshop, May 12,1997.

Perceptual contrast in Verner's Law


Page, Richard 1996 Verner's Law. Paper presented at the 2nd Annual Conference of Germanic Linguistics, April 26-28,1996. Madison, Wisconsin. Petrova, Olga 2000 Grimm's Law in Optimality Theory. In Proceeding of the Eleventh Annual UCLA Indo-European Conference, (Los Angeles, June 4-5, 1999). Journal of Indo-European Studies Monograph Series No.35, Karlene Jones-Bley, Martin E.Huld, Angela Delia Volpe (eds.), 45-67. Institute for the Study of Man: Washington, D.C. 2001 The evolution of the English obstruent inventory: Evidence for the dispersion/faithfulness interaction in OT. Ph.D. diss., Department of Linguistics, University of Iowa. Petrova, Olga, Rosemary Plapp, Catherine Ringen, and Szilard Szentgyorgyi 2000 Constraints on voice: An OT typology. Paper presented at the 74th Annual Meeting of the Linguistic Society of America, January 8, 2000. Chicago, Illinois. Prince, Alan, and Paul Smolensky 1993 Optimality Theory: Constraints interaction in generative grammar. Ms., Rutgers University, New Brunswick, and University of Colorado, Boulder. RUCCS-TR-2. Prokosch, E. 1939 A Comparative Germanic Grammar. University of Pennsylvania, PA: Linguistic Society of America. Rastorguyeva, Tatiana 1983 A History of English. Moscow: Vyssaja skola. Rice, Keren 1992 A reexamination of the feature [sonorant]: The status of sonorant obstruents. Language 69 (2): 308-344. Roberge, Paul T. 1983 Those Gothic spirants again. In Indogermanische Forschungen 88, Wolfgang Schmid (ed.), 109-155. Berlin/New York: Walter de Gruyter. Rooth, Eric 1974 Das Vernersche Gesetz in Forschung und Lehre 1875-1975. Lund: Gleerup. Rubach, Jerzy, and G.E. Booij 1990 Edge of constituent effects in Polish. Natural Language and Linguistic Theory 8: 427-463. Steriade, Donca 1995 Underspecification and markedness. In Handbook of Phonological Theory, J. Goldsmith (ed.), 114-174. Cambridge, MA: Blackwell.

408 Olga Petrova Stevens, Kenneth Ν. 1994 Phonetic evidence for hierarchies of features. In Phonological structure and Phonetic Form. Papers in Laboratory Phonology III, Patricia A. Keating (ed.), 242-258. Cambridge: Cambridge University Press. Silverman, Daniel 1996 Phonology at the interface of morphology and phonetics: Root-final laryngeals in Chong, Korean, and Sanskrit. Journal of East Asian Linguistics, 5 (3): 301-322. Verner, Karl 1875 Eine Ausnahme der ersten Lautverschiebung. Zeitschrift fur Vergleichende Sprachforschung 23: 97-130. Whalen, D.H, Arthur Abramson, Leigh Lisker, and Maria Mody 1990 Gradient effects of fundamental frequency on stop consonant voicing judgments. Phonetica 47: 36-49. Yip, Moira 1996 Tone in East Asian languages. In The Handbook of Phonological Theory, J. Goldsmith, (ed.), 476-494. Cambridge, MA: Blackwell. 2000 The complex interaction of tone and prominence. Paper presented at the 31s' Conference of the North Eastern Linguistics Society, October 7, 2000. Washington, DC.

Section 4 Dialectology

Introduction: Dialectology Anne Curzan and Kimberly Emmons

Research in historical English dialectology often calls attention to gaps in available linguistic data in the record of the history of English, either geographical or chronological. Scholars of earlier periods of English must confront the limits of the available textual record and speculate around and in the gaps: dialect areas with few if any texts or texts concentrated in particular genres; classes of speakers underrepresented in written texts; time periods with relatively fewer texts written in English. Dialectologists focused on current English dialects, with an almost overwhelming richness of available technologies for capturing dialect data, can afford to focus on ever-more local dialect areas to fill ever-more local gaps (i.e., specifying within broader generalizations). For all scholars of English dialects, however, exploring the evidentiary record involves balancing the theoretical and methodological concerns that are part of the ongoing and rich conversation about the history of the English language. Each of the essays included in this section addresses the issue of source material in a different way: Michael Montgomery and Connie Eble introduce new textual evidence for the pen/pin vowel merger in Southern American English; Betty Phillips discusses contemporary recordings of the caught/cot merger in West Central Indiana speech; and Richard Hogg explores the dialectal nature of negative contraction in Old and Middle English using a variety of published sources. Individually, these essays contribute to ongoing conversations about the history of American and British dialects; together, they provide a framework for attending to the collection and analysis of new and existing data sources. In their essay, "Historical perspectives on the pen/pin merger in Southern American English," Montgomery and Eble explore the early history of this characteristic SAE vowel merger in a study that will be of interest to researchers of Southern varieties of Egnlish and African American English, as well as to those interested in dialect contact more generally. Using three corpora of North Carolina "colloquial documents," Montgomery and Eble locate the occurrence of the pen/pin merger almost a century earlier than previous scholars have done. Admitting that "written data may disguise

412 Anne Curzan and Kimberly Emmons pronunciation," Montgomery and Eble nevertheless argue that the "occasional spellings" they analyze provide ample justification for future research into written records. Montgomery and Eble further suggest that such additional resources can enhance scholarship on SAE, including the "possibility of an African American role in [the] development" of the merger. In response to this claim, Guy Bailey, in "Digging up the roots of Southern American English," argues that more careful, quantitative comparisons of the North Carolina data are necessary. At the same time, Bailey applauds Montgomery and Eble's analysis for providing "much needed additional evidence on early African American speech." The addition of these kinds of historical text resources allows new understandings of the historical development of Southern dialect features. Drawing on his own scholarship on SAE, Bailey offers further, admittedly speculative, evidence for the AAVE origins of the merger. Together, Montgomery and Eble and Bailey offer important new data on SAE, and they articulate a compelling case for further research into early documentary sources of dialect writing. Betty Phillips, in "Vowel merger in West Central Indiana: A naughty, knotty project," concentrates on a different and much more widespread American vowel merger, the caught/cot merger. While this merger has spread throughout much of the United States, Phillips centers her analysis on a community in Indiana where the vowels are still variable. Phillips' study, which originated as a class project, reports on six semesters worth of interviews conducted by her students. While her findings generally confirm previous scholarly work - that the distinction between caught and cot is still maintained for older speakers, that the distinction is generally dying out, and that females are leading the change - Phillips' essay contributes an additional pedagogical dimension to current scholarship. Including student reflections on the class project, Phillips argues that her students demonstrate their acquisition of knowledge about the processes of vowel merger and about the variations within dialects themselves. Phillips' students thus had the opportunity to participate in ongoing scholarly research in dialectology: learning about research design and implementation, producing valuable intellectual data, and developing more sophisticated understandings of dialect variation. In the final essay in this section, "The spread of negative contraction in early English," Richard Hogg focuses on the progression of one particular linguistic change through dialects of Old and Middle English. Access to dialect variation in medieval English depends entirely on the appearance of assumed dialectal forms in localizable written texts - an assumption as

Section 4: Introduction


necessary and potentially tenuous for a study of nineteenth-century American English such as Montgomery and Eble's as it is for a study of medieval English such as Hogg's. Combining data from several sources, including the work of Samuel Levin, Yoko Iyeiri, and The Linguistic Atlas of Late Medieval English, as well as close examination of several medieval texts, Hogg argues that negative contraction first became categorical in the region surrounding Gloucester and spread through mechanisms of dialect contact. Interestingly, the spread of negative contraction may have been inhibited, at least in written texts, by factors such as the robust Mercian literary tradition. By late Middle English, however, evidence suggests that negative contraction is recessive. Hogg's inclusion of patterns of geographic movement by speakers in these periods echoes Bailey's and Montgomery and Eble's discussion of migration patterns of SAE speakers and the effect of population movements, and the resulting dialect contact, on language change. The essays collected in this section represent a broad range of recent work in dialectology. Each provides rich theoretical and methodological insights; each addresses current debates; each opens new avenues for investigation.

Historical perspectives on the pen/pin merger in Southern American English Michael Montgomery and Connie Eble

1. Introduction The merger of front lax vowels I I I and I I I before /n/ and /m/ in stressed syllables, producing the homophony of pinlpen and Jim/gem, is a wellknown feature in the American South today, found in the speech of blacks and whites and in nearly all social and sub-regional varieties within the region.1 In innovative research that analyzes orthographic and phonetic data, Vivian Brown (1990, 1991, 1993) has presented evidence that the merger as a phonological process may have arisen in the latter half of the nineteenth century and become a sound change moving toward completion only for speakers born after 1880 (Brown 1990: 39-40). Her claim has prompted Guy Bailey (1997) to argue that the merger, like many other hallmark features of Southern American English pronunciation and grammar, is a post-Civil War development. This ambitious hypothesis is potentially self-fulfilling because it is largely based retrospectively on data collected in the twentieth century from people born after the midnineteenth century. In this paper we explore the earlier history of the merger by using manuscript evidence and push back the time line on this feature a century by employing three corpora of documents from the American South dating as early as 1761. At the end of the paper we also examine two other contentions of Brown (1990): that the merger spread rapidly after the Civil War due to urbanization of the region; and that in the South the merger has not had an ethnic dimension, specifically, that evidence is lacking that it arose from African American influence on white speech. The data that we bring to bear are particularly well suited for this effort, in that they include two corpora of letters from white plantation overseers and from African American soldiers who served in the Civil War. Though not robust, our evidence suggests not only that the merger before nasals began earlier, but that African American English may have been partly responsible for it.

416 Michael Montgomery and Connie Eble

Brown's case is based on four sets of data: 1. responses to a written questionnaire completed between 1915 and 1922 by 223 elderly Tennessee Civil War veterans born in the second quarter of the nineteenth century (Dyer et al. 1985); 2. 152 Linguistic Atlas of the Middle and South Atlantic States (LAMSAS) interviews from North Carolina conducted between 1934 and 1937 using a standard questionnaire (McDavid et al. 1983);2 3. 159 Linguistic Atlas of the Gulf States (LAGS) interviews from Tennessee conducted in the early 1970s using a standard questionnaire (Pederson et al. 1981);3 4. twenty-five loosely structured interviews recorded with native Southerners (fourteen from Tennessee, six from Alabama, etc.) born between 1845 and 1960. Brown recognizes that some of her sources do not provide substantial evidence of speech prior to 1880. For example, only two of the LAGS speakers were born before that date, the earliest one in 1871. (The lack of records of Tennessee speech from the third quarter of the nineteenth century to compare directly to the Civil War veterans surveys led her to employ LAMSAS records from the earlier period from North Carolina, a state from which many early settlers of Tennessee came.) Also, although five of the native Southern interviewees were born before 1875, four of these were African Americans from rural Alabama or Texas. For these reasons, she concentrates on tracking progress of the merger since 1880 (about the time she argues it became actualized) rather than attempting to account in detail for its earlier stages. Much about the history of the merger of III and III before nasals in American English remains unclear, including its origin, its possible antecedents in the British Isles, its earlier development, and reasons for its spread. This paper contributes to the earlier picture of the merger by investigating "meaningful occasional spellings" (Stephenson 1967: 37) in three corpora of manuscripts from the late-eighteenth through the midnineteenth century:

Perspectives on the pen/pin merger 417 1. letters, legal documents, and church records from North Carolina from 1761 to 1800; 2. letters written by white plantation overseers from across the American South - North Carolina to Louisiana - between 1811 and 1876 (but primarily from the 1830s to the 1850s); and 3. letters written by African Americans (mainly Civil War veterans) from throughout the eastern half of the country between 1830 and 1868. The manuscripts analyzed in this paper represent language either contemporaneous with the Tennessee Civil War Veterans Survey (i.e., our African American writers) or one or more generations earlier than that source (i.e., our plantation overseer and North Carolina writers). Collectively our documents are among the most colloquial ones available from Southerners in the antebellum period and represent an appropriate testing ground for tracking the pen/pin merger. They add the perspective of a century to the existing picture of the merger and support more generally the use of manuscript evidence in reconstructing features of earlier American English.

2. Earlier history of variation between III and 111 The raising of III to h i before nasals has a long history in English, going back to the development of short (or lax) e before nasal plus consonant in Germanic, which yielded in Old English (OE), for example, the present system vocalism of third class strong verbs, as in bindan (from *bendanan). Middle English (ME) saw the raising of Iii to h i before velar nasal plus consonant, as in ME link from OE hlence, ME string from OE streng, ME wing from Old Norse veng, ME hinge from OE henge, and ME ink from Old French enque. The most familiar examples are think from OE pencan and England!English from Englaland/Englisc.4 Most of these forms are now and were in Middle English spelled with the letter i, except of course for England!English (Wright 1924: 41). According to Jespersen (1909: 65), the "short /i/" (i.e., [i]) pronunciation for England!English is mentioned by many sixteenth- and seventeenth-century orthoepists and appears to be the general pronunciation by that time. In A History of Modern Colloquial English, Wyld (1956: 222) reports from sixteenth- and seventeenth-century writings the spelling i instead of e before alveolar nasal plus consonant combinations, e.g., n+d (atinding);

418 Michael Montgomery and Connie Eble n+t (Gintlemen, rintes); n+s (sincible). (The first three of these forms are indeed found in our eighteenth-century documents.) If spelling evidence reflects pronunciation, the raising of III to 111 was not confined to environments before nasals in the Early Modern period, as the i spelling also showed up before s, as in opprision/opprissing (Queen Elizabeth I) and requist, and "more rarely" before /, as in rebillion, will 'well', and till 'tell' (Wyld 1956: 223). In The English Dialect Grammar (1905), Wright reports in nineteenth-century dialects of the British Isles short i before nasals in hen, men, pen, end, went, twenty, bench, and others (54-55). Before s, he reports it in nest and chest (51) and before I in seldom, self, sell, tell, else, and other words (52). Traditional accounts like those of Jespersen and Wright do not tie the development of /£/ to 111 solely or primarily to a following nasal environment subsequent to the raising of "short e" in words like string in Middle English. A further indication of alternation between /£/ and 111 in speech over several centuries is Wyld's positing of a simultaneous shift in the opposite direction, a lowering of 111 to III also by the time of Early Modern English. Says Wyld, In documents of all kinds, public and private, during the fifteenth century and in the successive centuries until the eighteenth, there are numerous examples of e written for original i. It cannot be doubted that these spellings reflect an actual tendency in pronunciation since late in the eighteenth century Edmonston censures 'tell' for 'till', and 'sense' for 'since', &c, as London vulgarisms. Whatever may have been the history of the introduction of these forms in London and Court English, there is no doubt that from the middle of the sixteenth century or so, on to the first third of the eighteenth century at any rate, they were current in circles whose speech, however much we may now take exception to this or that feature, was certainly not the vulgar speech of the day. (226-227)

Among Wyld's examples in stressed syllables are cheldren, well, lemeted, deficulte, sence 'since', prevelegys, untel, hether, and many others (228229). Wyld accepts the lowering of h i to IZl as a real change and suggests tentatively that it arose in the East Midlands, spreading to London by the fifteenth century; he is reluctant to associate the change with a lower class way of speaking (227). By the end of the eighteenth century, the choice between Iii and 111 before nasals and also other segments was associated in England with "correctness." John Walker's influential and prescriptive A Critical

Perspectives on the pen/pin merger


Pronouncing Dictionary and Expositor of the English Language was popular in many editions and printings on both sides of the Atlantic during the last quarter of the eighteenth and first quarter of the nineteenth century. An 1823 version recommends about the letter e, "This letter falls into an irregular sound, but still a sound which is its nearest relation, in the words, England, yes, and pretty, where the e is heard like short i. Vulgar speakers are guilty of the same irregularity in engine, as if written ingine, but this cannot be too carefully avoided" (23). Spelling and pronunciation of "short e" and "short i" have fluctuated even more in unstressed syllables than in stressed ones. Philological works like those of Jespersen and Wright indicate that in prefixes, suffixes, and other unstressed syllables, the raising of III to hi took place in words like exist, duchess, and elegant and the lowering of III to either III or /9/ in words like possible and evil.

3. Colonial American period American English in the eighteenth century exhibited considerable variation between the spellings i and e in the same word, as is shown in our documents from North Carolina. Sometimes this spelling variation represented historic forms still vying for prestige (cf. Walker's comment earlier). At other times, orthographic forms represented either multiple standard pronunciations or pronunciations no longer reflected in standard spellings (e.g., been by that time was in America generally pronounced /bin/). Evidence of variable pronunciation comes from Benjamin Franklin, who transcribed get and friend as git and frind in his proposed phonetic notation (Krapp 1925:11 99; see also Wise 1948). Ann Louise Sen (1974: 42) finds variation between i and e spellings common in eighteenth-century New Jersey manuscripts regardless of the author's social class. Relying overwhelmingly on data from seventeenth-century New England, Krapp argued that the two vowels often fluctuated and commented specifically on their merger in prenasal contexts: "the pronunciation of e before η as /, that is men, ten, tennis, as min, tin, tinnis, prevalent in Georgia, Alabama, and other regions of the South, is but a survival of a [general] Colonial pronunciation that probably passed current on all levels of society" (1925, Π: 34). While considering an Irish source to be possible, Krapp thought the pronunciation was "so old and so general in American speech as to make it unnecessary to call in the aid of the Irish immigrant to explain its presence" (Π: 97).

420 Michael Montgomery and Connie Eble To Krapp's own data, which was not from the South, we add a corpus of North Carolina documents assembled with the aid of records compiled by the late Edward Stephenson (1958). Contrary to Brown, who states that Iii and III "are presumed to have been brought to this continent as stable variants in the early 17th century" (1990: iii), our data and other evidence from the American colonial period cited by Wyld, Krapp, and Stephenson, indicate that the two vowels came from the British Isles as highly unstable variants.



Our method of presentation is to identify two types of meaningful occasional spellings (phonetic and inverse) that reveal apparent variation between III and hi.5 (This is in contrast to Brown, who considers only phonetic spellings in her written evidence from Civil War veterans surveys.) Because we are looking for earlier evidence of the current-day merger of these vowels before nasals, we group these types of spellings in Tables 1, 2, and 4, respectively, as involving variation in a) stressed prenasal environments; b) unstressed prenasal environments; or c) stressed non-prenasal environments. (Spellings that occur in unstressed, nonprenasal environments are not considered, e.g., an alternation of the letters a, e, and i in the second syllable of knowledge.) The stressed prenasal environment is of primary interest (category (a)), but the other two categories are pertinent in at least two ways. First, if fluctuation occurred largely or exclusively in unstressed prenasal environments, where vowels are typically reduced or neutralized as a general phonetic process in American English, this would support the Brown/Bailey hypothesis of the penlpin merger being a post-Civil-War development. Such a pattern might suggest that the merger spread from unstressed prenasal environments to stressed ones. Second, the degree to which the vowels vary in non-prenasal environments may indicate the extent to which variation between /£/ and hi was a general phonetic tendency rather than a phonological process. Variation affecting many forms and in different environments would suggest such a tendency. However, if only a few forms and environments are involved, this suggests either lexical variation or neutralization. One must always be aware that the nature of written data and the influence of present-day standards may complicate analysis of earlier manuscript evidence. Written data may disguise pronunciation.

Perspectives on the pen/pin merger


Categorization of forms with prenasal vowels is not always certain for stress; for example, words like entire may have received initial stress by past generations of Southerners, as they do today. Also, spellings that today would be classified as "occasional" may well have been conventional in an earlier era. Those spellings that are not standard today but that according to the OED were known to be current in published writing in the eighteenth century are identified with an asterisk in our tables. (These cluster in category (b), unstressed prenasal environments, e.g., imployment in section 2a of Table 1.)

5. Early North Carolina data Relying on Stephenson (1958), we have compiled a corpus of sixty lateeighteenth-century North Carolina documents from the Southern Historical Collection at the University of North Carolina at Chapel Hill. Some are legal documents or church minutes, others are letters. Most of the letters are from members of the eastern North Carolina educated, landed class; one writer, for example, was State Treasurer and a member of the University of North Carolina's Board of Trustees. However much formal education they might have had, these writers display a broad range of variation in the spellings at issue, as shown in Table 1. In stressed, prenasal environments, 37 different forms suggest the vowel merger (ginneral, offince, trimbling)\ in unstressed prenasal environments, 17 occur (imbrace, incumberance, indeaver)\ and in stressed non-prenasal environments, 79 occur (chist, kittle, requist). Although our tables cite how often forms that depart from modern conventional spellings occur, these are raw rather than relative frequencies. Because it is impossible to determine (or estimate) the precise extent to which a writer's spelling may disguise his or her pronunciation, it is not meaningful to calculate quantitative variation in terms of ratios of unconventional to conventional spellings. The distribution of forms in eighteenth-century North Carolina by environment suggests that the vowel merger was no more likely to occur in stressed prenasal environments than elsewhere. Individual writers could have been an exception to this overall picture, but since only one or two writers produced more than one document, there is insufficient data to profile individuals here. Nevertheless, several of the early writers did display forms in all three categories of environments.

422 Michael Montgomery and Connie Eble Table I. 18th-century North Carolina data (based in part on Edward Stephenson 1958: 194-207) (items marked * were current spellings in the 18th c. according to the OED) (figures refer to numbers of tokens in the corpus) 1) stressed prenasal environments: a) phonetic spellings: agin (again) 1, atining (attending) 1, bint 1, expinces 2, finces 2, frinch 1, frind (friend) 5, ginneral 1, gintlemen 2, hince 1, mind 1, mintioned 1, ofFince 1, offinders 1, primeses 3, primises 10+, primisis 1, pro tim 1, recomminding 1, rint 1, sind 1, tin 1, trimbling 1, twinty 2, vinesen 1, Widnesday 1 b) inverse spellings: contenue 1, enterest 1, flents 2, hendirance 1, lennin (linen) 1, menerals 1, openion 1, semple 1, sence (since) 3, sens 2, venegar 1 2) unstressed prenasal environments: a) phonetic spellings: imbrace 3, *imployment 1, incorrage 1, incouragement 1, incombrances 1, incumberance 4, indeaver 1, indevour 2, ineble (enable) 1, *ingage 1, ingaiged 1, *injoy 4, intails 1, *intirely 3, "intitled 1 b) inverse spellings: entarogated 1, sensere 1 3) stressed non-prenasal environments: a) phonetic spellings: chist 8, derictions 2, diligates 2, dilligates 1, distitute 1, domistick 1, filt 1, flish 1, frish 1, git 10+, gitin 1, giting 1, gitters 1, gitting 2, hild 1, hilth (health) 1, intilengence 1, intiligence 1, kittle 4, kittles 2, midow (meadow) 1, midows (meadows) 1, mit 1, mittle (metal) 1, mysilf 1, niver 1, nixt 2, pistoll (pestle) 1, predistination 1, prisbatary 2, ridgments (regiments) 1, rigiment 1, requist 1, requisted 1, shirriff 1, shirriffs 1, sildom 1, spilling 1, vigatables 1, whither (whether) 5, writched 1, yisterday 1, yit 3 b) inverse spellings: admession 1, besiness 1, chezels (chisels) 1, crebbs (cribs) 1, debelitated 1, deffrant 1, delevered 2, delevery 1, descipplien (discipline) 1, dessolution 1, destrict 2, detch 1, exhebit 1, indeferent 1, leving rume 1, levith (liveth) 1, mettens 1, preveledges 1, prevelidge 1, prevelidges 1, prevelig 1, previledge 3, previledges 7, rever 2, revits (rivets) 1, scesars (scissors) 1, selk 2, selver 1, setuate 2, submet 1, tel 2, tell 3, untel 2, untell 3, vermelion 1, wetness 2 The numbers in our table must be interpreted cautiously. The North Carolina documents, like those in the other two corpora, were selected because they exhibit one or more deviations from established modern-day pronunciation or grammar. Not every eighteenth-century manuscript from the Southern Historical Collection that we have read has been tabulated only those of linguistic interest. Furthermore, a "typical" letter of twenty lines seldom has more than five or six forms of linguistic interest. The

Perspectives on the pen/pin merger 423 authors of legal records and church minutes in the North Carolina corpus are anonymous, and about the background of writers who produced letters in all three corpora little is usually known but the name of the author and the location from which he or she wrote. The North Carolina writers almost certainly include some who were born in the British Isles and emigrated to America.

6. Plantation overseer data Our second corpus contains 536 letters from 50 plantation overseers born between the late-eighteenth and the mid-nineteenth century. (For further information on the corpus, see Schneider and Montgomery 2001.) The letters were written by semi-literate, working-class white men reporting to absentee owners of plantations. Typically they dealt with the progress of work, the state of crops, the weather, diseases among slaves, and so on. Their style is neutral, factual, business-like, with little attention to the mode of expression. Documents in this corpus and the African American corpus rarely have punctuation of any kind, and capitalization is highly inconsistent. Compared with writers in our North Carolina corpus, the overseers (most of whom wrote from plantations in North or South Carolina) display a markedly different inventory of forms, as seen in Table 2. Seventeen different forms evidence the vowel merger in stressed prenasal environments (agin, fince, Sinate), and 23 in unstressed prenasal environments (imbrace, indeavor). However, in stressed non-prenasal environments one finds only git and its derivatives (nearly 200 times), a striking lack of variation compared to our eighteenth-century corpus. Because the overseers were, from internal and external evidence, considerably less literate than the North Carolina writers and because the overseer corpus contains nearly ten times as many documents, we should reasonably expect much more variation than is found, especially in stressed non-prenasal environments, where 80 different forms exhibit occasional spellings in the eighteenth century. This disproportion appears to present strong evidence that fluctuation between III and h i in stressed non-prenasal environments in working-class white speech was highly conditioned by lexical form by the 1830s. One might speculate that the disproportion of occasional spellings between stressed prenasal and stressed non-prenasal environments reflects increased consciousness of standard spelling practices, but this is contradicted by many other types of aberrant spellings,

424 Michael Montgomery and Connie Eble

both phonetic and non-phonetic, that are found in the letters but are not pertinent to discuss here. Table 2. Plantation overseer data (1817-1865) (items marked * were current spellings in the 18th or 19th century according to the OED) (figures refer to numbers of tokens in the corpus) South Atlantic (North Carolina, South Carolina, Georgia) 1) stressed prenasal environments: a) phonetic spellings: agin (again) 4, aginst 1, enny (any) 2, eny (any) 15, frind (friend) 3, ginerly (generally) 1, intered (entered) 1, meny (many) 1, Sinate 2, sind 3, sint 6, thin (then) 1 b) inverse spellings: cence (since) 1, menut 1, Prence 1, *sence (since) 14, tember 1 2) unstressed prenasal environments: a) phonetic spellings: *disincouragd 1, imbrace 1, *imploy 4, imployer 1, indavour 1, indeavour 3, *ingage 3, *injoy 4, *intire 1, *intirely 1, intirley 1, intirly 1, Recomindations 1 b) inverse spelling: *entend 1 3) stressed non-prenasal environments: a) phonetic spellings: git 75, giting/gitting 24, gits 4 Mississippi Valley (Mississippi, Louisiana) 1) stressed prenasal environments: a) phonetic spellings: attintion 1, entinded (intended) 1, enny (any) 10, eny (any) 14, fince 1, frind (friend) 3, ginerly (generally) 2, meny (many) 3 b) inverse spellings: cence (since) 10, contenue 1, hem 1, *sence (since) 4 2) unstressed prenasal environments: a) phonetic spellings: imbaresment 1, imbrace 1, imploide (employed) 1, imploier 1, *imploy 3, indeavour 1, *ingage 2, ingagement 3, *injoy 6, *intire 1, *intirely 6 b) inverse spellings: *enduce 1, entinded (intended) 1 3) stressed non-prenasal environments: a) phonetic spellings: git 24, gitin/gitting 14, gits 2 Alabama 1) stressed prenasal environments: a) phonetic spellings: ginerl/ginerly (general/generally) 10 2) unstressed prenasal environments: a) phonetic spellings: indeaver 1, *intire 1 3) stressed non-prenasal environments: a) phonetic spellings: git 43, gitin/giting 4, gits 1

Perspectives on the pen/pin merger 425 Table 3 is a subset of the writers in Table 2. It compiles the forms of three individual overseers, a separation not meaningful for the other two corpora, because they include, with very few exceptions, only single documents from given individuals. Several of the overseers were engaged by a plantation owner for a number of years, with the result that series of their letters accumulated in the papers of a plantation family. Table 3. Sample individual overseers from the South Atlantic (figures refer to numbers of tokens in the corpus) Alexander Carter (North Carolina) 1) stressed prenasal environments: a) phonetic spellings: agin (again) 1, ginerly (generally) 1, sind 3, sint 2, thin (then) 1 2) unstressed prenasal environments: a) phonetic spellings: "imploy 2 b) inverse spelling: *entend 1 3) stressed non-prenasal environments: a) phonetic spellings: git 4, gitting 1, gits 1 Hugh Macaulay (South Carolina) 1) stressed prenasal environments: a) phonetic spellings: sint 4 2) unstressed prenasal environments: a) phonetic spellings: *disincouragd 1, indavour 1, *ingage 1, intirley/intirly 3 3) stressed non-prenasal environments: a) phonetic spellings: git 26, giting/gitting 2 Β W Allen (South Carolina) 1) stressed prenasal environments: a) phonetic spellings: agin (again) 4, aginst 1, frind (friend) 1 b) inverse spellings: Prence 1, *sence (since) 3 2) unstressed prenasal environments: a) phonetic spellings: *injoy 1 3) stressed non-prenasal environments: a) phonetic spellings: gitting 3

426 Michael Montgomery and Connie Eble

7. African American data Heretofore documents written by nineteenth-century African Americans have not been used for phonological analysis; by contrast, their morphological evidence has been used as a testing ground for the creolist hypothesis (Montgomery et al. 1993; Montgomery 1999; Kautzsch 2002, etc.). To our knowledge, this is the first study to examine phonological variation in nineteenth-century African American manuscripts. (For further information on the documents in this corpus, their authors, and the circumstances under which they were composed, see Montgomery et al. 1993.) In this corpus of 164 letters written mainly to the Freedmen's Bureau in the 1860s, we find 13 different forms evidencing the vowel merger in stressed prenasal environments and 18 in unstressed prenasal environments, as shown in Table 4 (which divides the letters into six regions on the basis of the state from which they were written). Occasional spellings in the first two categories resemble those from overseers' letters in their number and their identity. For example, note frind, ginerally, and ingage in the South Atlantic region in Table 2 and the same forms in the South Atlantic region in Table 4. In contrast, in the African American letters we find 23 forms with i rather than e in stressed non-prenasal environments, among which git and its derivative gitin occur only seven times. This is quite a different inventory from the overseers. Table 4. Freedmen's Bureau letters (1860s) (items marked * were current spellings in the 18th or 19th century according to the OED) (figures refer to numbers of tokens in the corpus) Northeast (Maine, Massachusetts, New York) 1) stressed prenasal environments: a) phonetic spellings: eny (any) 3, iny (any) 1 2) unstressed prenasal environments: a) phonetic spelling: *inlisted 4 b) inverse spelling: senserely 1 3) stressed non-prenasal environments: a) phonetic spellings: git 1, rigerment 3, siveral 1 b) inverse spelling: egnorent 2

Perspectives on the pen/pin merger


Table 4. Freedmen's Bureau letters (1860s) Continued Mid Atlantic (Pennsylvania, Delaware, Maryland, District of Columbia) 1) stressed prenasal environments: a) phonetic spellings: eney (any) 6, enney (any) 1, meney (many) 2 b) inverse spelling: *sence (since) 4 2) unstressed prenasal environments: a) phonetic spellings: *inlisted 2, *intire 1 3) stressed non-prenasal environments: a) phonetic spelling: rigmat (regiment) 1 b) inverse spellings: afflection 1, inlesmet (enlistment) 1, untell 2 Middle West (Ohio, Indiana, Michigan, Illinois, Iowa) 1) stressed prenasal environments: a) phonetic spelling: enything 2 b) inverse spelling: *sence (since) 6 2) unstressed prenasal environments: a) phonetic spellings: *inlist 1, ingagement 3 Border States (Tennessee, Missouri, Kentucky) 1) stressed prenasal environments: a) phonetic spellings: ennythin (anything) 1, eny (any) 5, menny (many) 1 b) inverse spellings: cence (since) 1, sen (since) 1, *sence (since) 2 2) unstressed prenasal environments: a) phonetic spellings: imbrace 1, imploye 5, inlis (enlist) 1, *inlisted 3 3) stressed non-prenasal environments: a) phonetic spellings: git 1, yit 1 b) inverse spellings: presen (prison) 1, spent 1, tell 1 South Atlantic (Virginia, North Carolina, South Carolina, Georgia) 1) stressed prenasal environments: a) phonetic spellings: eney (any) 3, enny (any) 3, eny (any) 16, frind (friend) 2, ginearl (general) 1, ginerel (general) 3, ginerl (general) 3, inny (any) 1, menny (many) 2, meny (many) 2, pinsel (pencil) 1, whin 1 b) inverse spellings: *sence (since) 4, sense (since) 1 2) unstressed prenasal environments: a) phonetic spelling: ingaged 1, inlested (enlisted) 2, intier (entire) 1 b) inverse spellings: *enform 1, enterseed (intercede) 1 3) stressed non-prenasal environments: a) phonetic spellings: git 3, gitin 1, hilpe 1, respictfully 1, till 1 b) inverse spellings: enlested 2, inlested (enlisted) 2, untell 1

428 Michael Montgomery and Connie Eble Table 4. Freedmen's Bureau letters (1860s) Continued Mississippi Valley (Mississippi, Arkansas, Louisiana, Texas) 1) stressed prenasal environments: a) phonetic spelling: eney (any) 2, enny (any) 1, eny (any) 1, ingenears (engineers) 1, meny (many) 1 b) inverse spelling: *sence (since) 3 2) unstressed prenasal environments: a) phonetic spelling: imbrace 1, ingenears (engineers) 1, *injoy 1, inlected (enlisted) 1, inlested (enlisted) 1, b) inverse spellings: egnorence 1, *enform 1, enspection 1, entention 1 3) stressed non-prenasal environments: a) phonetic spellings: arristed 1, fideral 1, git 3, ribs (Rebs), rigment (regiment) 1 b) inverse spellings: inlected (enlisted) 1, inlested (enlisted) 1, retten (written) 1

Because most of these letters were from Civil War soldiers and veterans and because African American soldiers were recruited from many different parts of the country, including sections of the South that were early occupied by Union troops, it is possible to group writers according to region and to display the occasional spellings accordingly, as is done in Table 4.

8. Discussion Our three corpora do not offer unambiguous evidence for the merger of I I I and I I I before /n/ and /m/ in stressed syllables in the late-eighteenth- and early-nineteenth-century American South. Each corpus has a notable lack of occasional spellings for such common words as them, men, and then. A possible explanation for this is that even barely literate writers were familiar with the spelled forms of extremely common words. Whin for when occurs only once, in the African American corpus; tin for ten occurs only once, in the eighteenth-century North Carolina corpus. In the overseers corpus then occurs 116 times vs. thin once, pen 39 times but pin not at all. Countless other forms that would potentially show the merger do not vary from the standard spelling in our three corpora. Such negative evidence is consistent with the claim of Brown and Bailey that the occurrence of the merger in stressed syllables was only incidental in the

Perspectives on the pen/pin merger


South before the Civil War. Although the merger may have been underway in earlier periods, convincing evidence of it is not yet at hand. Merger of the vowels in unstressed syllables is common in all the documents that we have looked at in this study and today is reflected in such standard alternate spellings as ensure vs. insure. There can be little doubt of the prevalence of the merger in such environments today and in centuries past. At least one writer (Sen 1974) has argued that h i spread to stressed syllables from unstressed syllables in eighteenth-century New Jersey speech. The data in our documents do not support this hypothesis for the South. They suggest that her hypothesis is implausible or, at the very least, that its evidence is not relevant to the question at hand (indeed, the merger in stressed syllables is apparently undocumented from New Jersey by any other scholar). To the contrary, our North Carolina data support the view of widespread fluctuation between the vowels in all three types of environment in the eighteenth century. The distribution of occasional spellings in that corpus suggests variation that had not yet developed into a phonological process confined to prenasal environments. We find the greatest range of forms in all three types of environments in that earlier corpus. Still, a closer look at our data raises a number of questions that must be answered before we draw conclusions with confidence about the merger of I I I and I I I in the South before the late-nineteenth century. First, the evidence for this feature, meager as it may be, is much greater than that for nearly all other features involving pronunciation in the documents at hand. For many hallmark Southern features, such as variation in the /ai/ diphthong, the documents of all three corpora are completely silent. It is unlikely that all of the several dozen features identified by Bailey (1997) were later developments, given the fact that he has few sources dating from before 1880, when he claims the features began spreading vigorously. The non-occurrence that he notes may merely reflect the tendency of written documents to obscure pertinent evidence. Second, although the range of forms suggesting merger in the plantation overseer and African American corpora is much narrower than in the North Carolina documents, and common forms in which we might expect the merger most likely to be in evidence do not show it, most forms that do show variation involve prenasal environments. This is true for both the overseer letters and the African American letters if we consider the spelling e rather than standard a in forms like eny and meny as indications

430 Michael Montgomery and Connie Eble of raising.6 Thus, there may be more evidence of the merger than at first glance. To account for the abrupt increase of the merger in the late-nineteenth century, that is, its actuation as a phonological process, Brown hypothesizes that the South urbanized rapidly after the Civil War and that merger of vowels accelerated through improved transportation and urban growth. The merger was a "rural, insular" pattern at mid-century (Brown 1990: 79), but within two generations it was the non-merger that had become "a relic, insular feature confined to rural areas ... by the beginning of the twentieth century" (1990: 81). We believe that Brown's account raises far more questions than it answers. Few of Brown's 223 Civil War veterans were from towns, so she does not compare data from rural and urban veterans, and we do not know the level of occurrence of the merger in towns - or indeed whether it could be found at a higher rate there much earlier. How could it migrate to towns and then recede in the countryside so quickly? Brown's account seems to presuppose that rural nineteenthcentury Southerners had either a different origin from their urban counterparts or a significantly different history, or that rural speakers had linguistic features not shared with towns. Most especially, Brown's hypothesis leaves unanswered questions about how and when the merger originated in rural areas, in particular whether and what kind of language contact might have produced it then. An intriguing possibility that emerges from our data is that the increasing merger in white Southern speech might have resulted from contact with African Americans. Table 5 indicates that the percentage of letters displaying possible evidence of the merger in prenasal environments (the column _/N) in African American letters varied across the six American regions, ranging from 24% in the Northeast to 46% in the Border States. In marked contrast and more importantly, the letters show virtually no variation between and in non-prenasal environments (column _/-N) when considered in this way. For all intents and purposes, the only evidence for the merger among African Americans appears before nasals, particularly if we accept eny and meny as spelled forms indicating the tendency to merge vowels in prenasal environments.

Perspectives on the pen/pin merger 431 Table 5. Merger of vowels in Freedmen's Bureau letters (figures refer to numbers of letters; Ν = nasal) Total Northeast Mid Atlantic Middle West Border States South Atlantic Mississippi Valley






3 6 4 9 20 6

1 2 0 1 1 3

1 2 0 2 1 3

12 21 6 12 36 12

4/17 8/31 4/10 11/24 21/58 9/24

24 26 40 46 36 38

Brown (1990: 30-40) argues that the incidence of the vowel merger among neither Tennessee Civil War veterans nor North Carolina LAMSAS informants shows correlation with ethnicity and that it "does not seem to have originated in either Black or White speech" (Brown 1990: 74). However, the validity of this finding is limited by the nature of her indirect evidence. For white Civil War veterans she posits "ethnic differences" by dividing respondents into two groups based on whether they or their family had ever owned slaves. She then assumes that only respondents who had owned slaves had contact with African American speech. Using this criterion, she finds that non-slave-holding veterans in Tennessee had less evidence of the merger (actually, neither group shows much evidence of the merger - 10% and 14.8%, respectively). One might reasonably expect some, if not many, of the non-slave-holding group to have had contact with African American speech and always to have had such contact. Choosing LAMSAS records from North Carolina to analyze because many of its informants were born around the middle of the nineteenth century and because many settlers of Tennessee came from that state, Brown also finds less evidence of the merger among African Americans of North Carolina than among whites. However, only seven of the 152 LAMSAS North Carolina interviews were conducted with African Americans, in the eastern half of the state, a long way from the western areas from which Tennessee was settled in the eighteenth century. For LAGS interviews from Tennessee, she reports (1990: 73) no ethnic difference. However, that evidence for African Americans is rather later than ours (in LAGS the oldest African American speaker from Tennessee was born in 1894).

432 Michael Montgomery and Connie Eble An examination of our three corpora of manuscripts calls for reconsideration of the claims of other scholars with regard to the origin of the pen/pin merger in the American South. The possibility of an African American role in its development is admittedly speculative, but it needs to be investigated further. How and when could it have arisen in African American speech? Realistically speaking, evidence from linguistic projects conducted well into the twentieth century can probably only suggest that the merger arose after the Civil War, because no one interviewed by them was born before about 1850. Thus, manuscript evidence is crucial to explore further the possibility that development of the merger was a postCivil War phenomenon. Although such writings are rare, identifying them and analyzing evidence from their spelling are the next logical steps in tracking the development of this feature and others so characteristic of Southern pronunciation today.

Notes 1.

2. 3. 4. 5.

We acknowledge our reliance on the painstaking records of Edward Stephenson and are grateful to William A. Kretzschmar Jr. of the University of Georgia for making them available to us. Brown classifies a LAMSAS speaker as having the merger if he or she has hi in ten or twenty. Brown classifies a LAGS speaker as having the merger if he or she has 111 in ten, twenty, or pen. Spellings in i- (especially inglis) were also common in English and Scots at one time. The OED records them from the fourteenth to the sixteenth century. The best discussion of the use of spelling to reconstruct the pronunciation of English from manuscript evidence is Stephenson (1967). We follow Stephenson's classification. An occasional spelling is "any departure from the conventional spelling of a word" (37). Excluding accidental miswritings that give no information about pronunciation, potentially meaningful occasional spellings are of two kinds, phonetic and inverse. A phonetic spelling is "one in which the writer has substituted for the conventional spelling a spelling based on some familiar correspondence of symbol to sound" (39), e.g., ginneral for general. An inverse spelling is "one in which the writer has substituted for the conventional spelling of a word a spelling based on the analogy of some other word containing an orthographic fossil, perhaps etymologically justified but no longer symbolizing a sound in the writer's dialect... The imitative introduction of an orthographic fossil into the spelling of a word where it is not traditional usually means that the writer has created or employed an unhistorical spelling, although the fossil may be historical in the model the writer is imitating" (40),

Perspectives on the pen/pin merger



e.g., kneed for need, wright for rite. Examples in the latter category may also be characterized as orthographic hypercorrections. Thus, phonetic spellings are those which directly and transparently reflect pronunciation, based on our knowledge (historical or contemporary) of actual pronunciation, e.g., since for sense. On the other hand, inverse spellings reflect a degree of confusion on the part of the writer about sound/letter correspondence. They reflect pronunciation indirectly, e.g., sence for since and usually represent orthographic hypercorrections. The tendency for /ae/ to raise before nasals has been documented by investigators of African American English more recently. For example, Bridget Anderson (personal communication) notes the tendency among African Americans in Detroit in fieldwork conducted in 2001.

References Bailey, Guy 1997 When did Southern American English begin? In Englishes around the World I, Edgar W. Schneider (ed.), 255-275. Amsterdam: Benjamins. Brown, Vivian 1990 The social and linguistic history of a merger: /I/ and /E/ before nasals in southern American English. Ph.D. diss., Texas A & Μ University. 1991 Evolution of the merger of /£/ and hi in Tennessee. American Speech 66: 303-315. 1993 Linking data from oral and written sources to trace and characterize a sound change in progress: The merger of /£/ and 111 before nasals. In Proceedings of the International Congress of Dialectology, Volume 3: Regional Variation, Colloquial and Standard Language, Wolfgang Viereck (ed.), 80-94. Stuttgart: Steiner. Dyer, Gustavus, John Trotwood Moore, Colleen M. Elliott, and Louise Armstrong Moxley (eds.) 1985 Tennessee Civil War Veterans Questionnaire. 5 vols. Easley, SC: Southern Historical Press. Jespersen, Otto 1909 A Modern English Grammar on Historical Principles, Part I: Sounds and Spelling. Heidelberg: Carl Winter's. Kautzsch, Alexander 2002 The Historical Evolution of Earlier African American English: An Empirical Comparison of Early Sources. Berlin/New York: Mouton de Gruyter. Krapp, George Philip 1925 English Language in America. 2 vols. New York: Ungar.

434 Michael Montgomery and Connie Eble McDavid, Raven I., Jr., William A. Kretzschmar, Jr., and Gail A. Hankins (eds.) 1982-86 North Carolina records. Linguistic Atlas of the Middle and South Atlantic States and Affiliated Projects: Basic Materials. MSS on Cultural Anthropology 69: 368. Chicago: Joseph Regenstein Library, University of Chicago. Montgomery, Michael 1999 Sierra Leone settler English: Another exported variety of African American English. English World-Wide 20: 1-34. Montgomery, Michael, Janet M. Fuller, and Sharon DeMarse 1993 "The black men has wives and sweet harts [and third person plural -s] jest like the white men": Evidence for verbal -s from written documents on nineteenth-century African American speech. Language Variation and Change 5: 335-354. Pederson, Lee Α., Susan Leas, Guy H. Bailey, and Marvin Bassett (eds.) 1981 Linguistic Atlas of the Gulf States: The Basic Materials. Ann Arbor, MI: University Microfilms. Schneider, Edgar, and Michael Montgomery 2001 On the trail of early nonstandard grammar: An electronic corpus of Southern U.S. antebellum overseers' letters. American Speech 79: 388-409. Sen, Ann Louise F. 1974 Dialect variation in early American English. Journal of English Linguistics 8: 41-47. Stephenson, Edward A. 1958 Early North Carolina pronunciation. Ph.D. diss., University of North Carolina at Chapel Hill. 1967 On the interpretation of occasional spellings. Publication of the American Dialect Society 48: 33-50. Walker, John 1823 A Critical Pronouncing Dictionary and Expositor of the English Language. 3rd ed. New York: Collins and Hannay. Wise, C. M. 1948 Benjamin Franklin as a phonetician. Speech Monographs 15: 94120.

Wright, Joseph 1905 The English dialect grammar. Oxford: Henry Froude. 1924 An Elementary Historical New English Grammar. London: H. Milford. Wyld, Henry C. 1956 A History of Modern Colloquial English. 3rd ed., with additions. Oxford: Blackwell.

Digging up the roots of Southern American English: On Michael Montgomery and Connie Eble's "Historical perspectives on the pen/pin merger in Southern American English" Guy Bailey

1. Introduction The origins of Southern American English (SAE), like those of its relative African American Vernacular English (AAVE), have long been of interest to linguists and nonlinguists alike.1 As McMillan and Montgomery's comprehensive bibliography indicates, the body of literature on the origins of SAE is both vast and varied. Unlike research on the roots of AAVE, however, it has largely been uncontroversial - at least until the last decade of the twentieth century. Until that time, most researchers simply assumed that the distinctive features of SAE had been around since the colonial period and that those features had their roots in the British regional varieties that were brought across the Atlantic during the original settlement of the South. During the late 1980s and early 1990s, though, our research group, based first at Texas A&M University and then at Oklahoma State, discovered that a number of the hallmarks of SAE were far more frequent and widespread in the third quarter of the twentieth century than they had been a century earlier.2 Such features as the Southern Shift, monophthongal /ai/, the quasi-modal fixin to, and even the pronoun yall were either rare or nonexistent in data from the oldest Southerners we examined (born between 1850 and 1900) but became progressively more frequent and widespread after that time. Based on the research group's analysis of the distribution of a number of hallmarks of SAE in both real and apparent time, I concluded that what we think of now as Southern American English is largely a fairly recent development that reflects the expansion of some older features that had been relatively uncommon earlier in the South along with the emergence and rapid expansion of a number of linguistic innovations.3

436 Guy Bailey What made these conclusions possible was a number of new and "rediscovered" data sources on SAE that became available during the late 1980s and early 1990s. Among these were:






the Linguistic Atlas of the Gulf States (LAGS), completed in the early 1980s and published between 1986 and 1993; a set of mechanical recordings made during the 1930s and 1940s with former slaves born between 1844 and 1865, a source which had lain fallow for half a century; statewide random-sample surveys of Texas and Oklahoma completed in 1989 and 1991, the former of which allowed for real time comparisons with LAGS; extensive longitudinal research in the east-central Texas community of Springville with black and white informants born between 1894 and 1996; interviews with of a number of descendents of ex-Confederates who had settled in Brazil after the Civil War, the oldest of whom speak a "sister dialect" of SAE that provides evidence for comparative reconstruction; the Tennessee Civil War Veteran's Questionnaires·, scattered mechanical recordings with a few white Southerners born in the middle of the nineteenth century that were discovered in various archives around the South.

Taken together, these resources provided extensive new data on the last 150 years of SAE. The convergence of data from sources as diverse as these (and as Brown showed, the convergence was remarkable) enhances our confidence in conclusions based on them. Research on SAE continues to uncover exciting new sources of data. "Historical Perspectives on the pen/pin Merger In Southern American English," by Michael Montgomery and Connie Eble, introduces us to several of theses new sources. The letters from the plantation overseers pushes the evidence on early SAE back a generation or so, and the data from the Freedman's Bureau letters comprises much needed additional evidence on early African American speech. The data from the eighteenth century is even more interesting since it allows for inferences about the speech of some of the earliest residents of the South. Montgomery and Eble's application of data from these sources to the pen/pin merger

Digging up the roots of SAE 437 provides some striking insights into the general instability of III and Ν in eighteenth- and nineteenth-century SAE and strong confirmation of Vivian Brown's (1991) contention that the merger of /Zl and IV before nasals in SAE was largely a late nineteenth- and early twentieth-century development. While Montgomery and Eble note the limitations of these sources (and of written documents in general), they also demonstrate the range of uses to which the new sources can be put and the kinds of inferences that can be made from them. At least two of those inferences deserve a closer look. Montgomery and Eble suggest that (1) the pen/pin merger, contra Brown, may have begun in African American speech and then spread to white speech; and (2) the social context that Brown suggests might have triggered its spread (i.e., the post-1880 process of urbanization) is unlikely to have done so. Although Brown found no evidence of ethnic differences in the distribution of the merger in any of the corpora she examined and found little evidence of it in the mechanically recorded evidence from former slaves, the notion that the pen/pin merger may have originated in AAVE is intriguing. The data that Montgomery and Eble present on this point, however, are difficult to assess because of an inconsistency in their method. In the discussion of data from eighteenth- and nineteenth-century whites, they do not give percentages of occurrence/nonoccurrence, arguing that "because it is impossible to determine (or estimate) the precise extent to which a writer's spelling may disguise his or her pronunciation, it is not meaningful to calculate quantitative variation in terms of ratios of unconventional to conventional spellings." The point is well taken, but Montgomery and Eble then proceed to make the case for an African American origin for the pen/pin merger based on "the percentage of letters displaying possible evidence of the merger in prenasal en-vironments." Without similar percentages from whites, we simply cannot know whether or not the African American origins theory has any merit. The case for an African American origin, nevertheless, can be made indirectly based on acoustic analyses of the vowel systems of African Americans born during the middle and last half of the nineteenth century. Thomas and Bailey (1998) and Thomas (2001) present acoustic evidence of a late nineteenth-century development in the AAVE vowel system - the raising of Ixl - that may have triggered the raising of III before nasals. 4 Two aspects of the raising of /ae/ are particularly interesting. First, it apparently was raised before nasals initially, moving /as/ into the Iii range in that environment by the last quarter of the nineteenth century. Second,

438 Guy Bailey while /«/-raising also occurs in white varieties of SAE, it does so only before nasals - and seemingly only after it occurred in A A V E . It may be that /«/-raising before nasals in late nineteenth-century S A E triggered the subsequent raising of /£/ before nasals to avoid a merger with /as/; however, as III before nasals moved upward, it merged with /I/ because /V had nowhere to go. The scant acoustic evidence we have suggests that the upward movement of /ae/ before nasals may have begun in A A V E , and hence that the pen/pin merger may have begun in A A V E . All of this, though, is simply speculation based on what seems to make sense; we have no direct evidence for the scenario outlined above. Without a careful, quantitative comparison of data from early and mid nineteenth-century blacks and whites, no firm conclusions about the possible A A V E origins of the pen/pin merger are possible.

Census Year Figure 1. Growth of the urban population in the American South

Digging up the roots of SAE


Montgomery and Eble's argument that urbanization is an unlikely social trigger for the merger raises an excellent point: the kind of evidence needed to confirm the hypothesis with any certainty simply does not exist. Their argument also suggests a common misunderstanding of the nature of urbanization in the late nineteenth-century South, however. In the last quarter of the nineteenth century and first decade of the twentieth century, urbanization in the South was not so much the movement of people to towns and cities as the creation of towns and cities.5 Figures 1-3 illustrate this process.

Cities Figure 2.


The growth of towns and cities in the South, 1880-1910 (cities = communities with populations greater than 25,000; towns = communities with populations greater than 5000) (Source: Frank 1999)

As Figure 1 shows, in 1860, only about 10% of the Southern population lived in urban areas (communities of 2500 or more), and most of this 10% lived in just four cities with populations greater than 25,000 and 33 towns with populations greater than 5,000 (see Figure 2). After 1880, the urban population began to increase rapidly, not so much as people began to migrate to existing urban areas, but as new towns and cities began to emerge throughout the South (see Figure 3). The process of town building

440 Guy Bailey

was one of the two most important social processes in the late nineteenth century South and led not only to increased mobility among Southerners, but also to significant dialect contact among a myriad of local vernaculars brought together in the newly emerging villages, towns, and cities.6 Dialect contact, of course, is exactly the kind of social context that leads to language change.7 Given what we know about the actuation of language change, it seems reasonable to assume that the pen/pin merger most likely emerged in the context of the dialect contact brought about by the development of villages, towns, and cities throughout the late nineteenthcentury South, the first stage in the century-long urbanization of the region. While the sociolinguistic evidence needed to confirm this scenario does not exist, the close parallel between the rise of the merger (shown in Figure 4) with the growth of the urban population in the South (Figure 1) and with the emergence of towns and cities (Figure 3) provides support for the scenario.


Figure 3. Number of villages crossing the urban threshold (reaching populations of 2500) between 1880 and 1910 (Source: Ayers 1992)

Digging up the roots of SAE 441 The data summarized in Figure 4 suggests one other point that should be made about Montgomery and Eble's conclusions. Although evidence from sources such as the ones they mine is crucial for outlining the shape of early SAE, it is not necessarily crucial to the argument that hallmarks of SAE such as the pen/pin merger, monophthongal, and the Southern Shift are recent developments. The linguistic atlas data alone in Figure 4 show clearly that the pen/pin merger occurred among only a minority of Southerners before 1900 but was predominant in the South by the middle of the twentieth century. Regardless of the linguistic situation in the early

100 -TN Civil War Vet Questionnaires 90

LAMSAS - North Carolina • LAGS — Tennessee

ä u





s xt ΐ


50 ε

OS 40 30 20

10 1820-1839






Dates of Birth of Respondents

Figure 4. The evolution of the pen/pin merger in Tennessee (Source: Brown 1991) nineteenth or late eighteenth centuries, the pen/pin merger became a general feature of SAE only after 1900. The same is true of monophthongal /ai/. The acoustic analyses of vowel systems of mid-nineteenth-century Southerners in Thomas and Bailey (1998) and Thomas (2001) all show strong diphthongs for /ai/ (as does the analysis of one of the descendents of ex-Confederates in Brazil). By the last quarter of the nineteenth century, the analyses begin to show weakening of the glides before voiced sounds,

442 Guy Bailey and by the early part of the twentieth century monophthongs begin to appear, although they compete with full and weakened glides. By the middle of the twentieth century most Southerners analyzed in Thomas (2001) have monophthongs in voiced environments, and some have monophthongs in voiceless environments as well. 8 Clearly, then, regardless of its presence or absence in the early nineteenth- or late eighteenthcentury South, monophthongal /ai/ did not become a generalized feature of SAE until the second quarter of the twentieth century. The conclusions in Bailey (1997) are based on just this kind of data.' The evidence that present-day SAE is primarily a consequence of recent linguistic developments is substantial and in my view conclusive. However, the linguistic innovations and expansions of the last 125 years were clearly developments of something that came before them. While those innovations and expansions can be documented and tracked largely through existing data, understanding why they occurred requires that we understand the linguistic context in which they arose - that we have access to sources of data that drive our knowledge base back to the early nineteenth and late eighteenth centuries. The evidence presented in "Historical perspectives on the pen/pin merger In Southern American English" does just that. As researchers like Montgomery and Eble uncover and analyze these kinds of rich documentary sources, we have a unique opportunity not only to understand the factors that gave rise to SAE, but also to gain real insight into the actuation of linguistic change.

Notes 1. 2.




See McMillan and Montgomery (1989) for a long list of works on the history of SAE. Of particular interest is Brooks (1937, 1985). Members of that group included at one time or another Lisa Abney, Cynthia Bernstein, Vivian Brown, Patricia Cukor-Avila, Sonia Davenport, Margie Dyer, Grace Kerr, Garry Ross, Lorie Sand, Clyde Smith, Jr., Erik Thomas, Jan Tillery, and Tom Wikle. These conclusions, of course, have an interesting parallel in our conclusions about the development of AAVE. See Bailey and Maynor (1987) and Bailey (1993). The raising of /ae/ in AAVE differs from that of Northern dialects in that it is not part of a chain shift. See Bailey and Thomas (1998) and Thomas (2001) for vowel plots that clearly show the progress of /as/-raising. See Bailey and Cukor-Avila (forthcoming) for an extensive discussion of the process of village, town, and city building. This process was a consequence in part of the emergence of farm tenancy and of the development of general

Digging up the roots of SAE


7. 8.



stores to serve the tenant population and in part of the emergence of industry in the South. The work of Walt Wolfram and his research team demonstrates the kind of extensive local variation that can occur among insular communities that are located only a few miles from each other. See Trudgill (1986), and see Bailey, Wikle, Tillery, and Sand (1996) for an example. The representativeness of mechanically recorded data from the mid-nineteenth century is sometimes questioned. Clearly, that data is not representative in the way that a modern linguistic survey might be, but then neither are written documents. Thomas (2001) analyzes some 20 Southerners born before 1900, nine of them bom before 1880. What gives confidence in the results of these analyses is that they fit nicely with results from the last decade of the nineteenth and first decade of the twentieth centuries, time periods for which there is extensive data. See Tillery and Bailey (2003) for additional evidence of the same type.

References Ayers, Edward L. 1992 The Promise of the New South: Life after Reconstruction. Oxford: Oxford University Press. Bailey, Guy 1993 A perspective on African-American English. In American Dialect Research: An Anthology Celebrating the 100th Anniversary of the American Dialect Society, Dennis Preston (ed.), 287-318. Amsterdam: John Benjamins. 1997 When did Southern American English begin? In Old Englishes and Beyond: Studies in Honour of Manfred Gorlach, Edgar Schneider (ed.), 255-275. Amsterdam: John Benjamins. Bailey, Guy, and Patricia Cukor-Avila forth. The Evolution of a Vernacular: The Development of African American Vernacular English Since J850. Cambridge: Cambridge University Press. Bailey, Guy, and Natalie Maynor 1987 Decreolization? Language and Society 16: 449-473. Bailey, Guy, and Erik Thomas 1998 Some aspects of AAVE phonology. In The Structure of African American Vernacular English, Salikoko Mufwene, John Rickford, Guy Bailey, and John Baugh (eds.), 85-109. London: Routledge.

444 Guy Bailey Bailey, Guy, Tom Wikle, Jan Tillery, and Lori Sand 1996 The linguistic consequences of catastrophic events: An example from the Southwest. In Sociolinguistic Variation: Data, Theory, and Analysis, Jennifer Arnold et al. (eds.), 435-451. Stanford: CSLI. Brooks, Cleanth 1937 The English language in the South. In A Southern Treasury of Life and Literature, Stark Young, (ed.), 350-358. New York: Scribner's. 1985 The Language of the American South. (Mercer University Lamar Memorial Lectures, no. 28.) Athens: University of Georgia Press. McMillan, James B., and Michael Montgomery 1989 Annotated Bibliography of Southern American English. 2nd ed. Tuscaloosa: University of Alabama Press. Thomas, Erik R. 2001 An Acoustic Analysis of Vowel Variation in New World English. PADS, 85. Durham: Duke University Press. Thomas, Erik, and Guy Bailey 1998 Parallels between vowel subsystems of African American Vernacular English and Caribbean anglophone Creoles. Journal of Pidgin and Creole Languages 13: 267-296. Trudgill, Peter 1986 Dialects in Contact. London: Blackwell. Tillery, Jan, and Guy Bailey 2003 Urbanization and the evolution of Southern American English. In English in the Southern United States, Stephen J. Nagel and Sara L. Sanders (eds.), 159-172. Cambridge: Cambridge University Press.

A brief response Michael Montgomery and Connie Eble

The scholarship of Guy Bailey and his colleagues has set high standards for the profession and added immensely to our knowledge of the development of Southern American English. Like all good research, it provides hypotheses which other researchers can explore and test. Our paper endeavors to do exactly this by examining the merger of front lax vowels before nasals in data one to three generations earlier than previously presented. We find intriguing, but admittedly limited, evidence for the merger. This suggests that the account advanced by Vivian Brown for this feature, and more broadly by Bailey himself for the development of SAE, is overstated and potentially self-fulfilling because it is based largely on recordings and other data collected in the twentieth century from people born after the mid-nineteenth century. We argue that the use of earlier written documents is crucial to understanding developments (other sources can also be valuable). While some hallmark features of SAE no doubt spread rapidly in the latenineteenth century, which ones are they? Are the sources used by Bailey et al. sufficient to determine their actuation in each case? There is nothing incompatible about identifying a British or Irish root of a feature while showing that it spread a century or more after arriving. Does establishing the latter justify labeling a feature as a "linguistic innovation"? This unnecessary, if not false, juxtaposition in part represents the motivation for our study. Bailey wonders if our idea of possible African American influence "has any merit" if we do not analyze letters from our overseer and Freedmen's Bureau corpora in exactly the same way. In fact, Tables 2 and 3 indicate that we have extracted forms from letters in the same manner, i.e., by identifying the raw frequency of items reflecting possible vowel merger. For reasons we state, we do not believe it meaningful to calculate quantitative variation in ratios of unconventional to conventional spellings in our documents. Brown's research using the Tennessee Civil War Veterans' Questionnaires took the same approach. It is true, on the other hand, that we have not summarized data from overseers comparable to

446 Michael Montgomery and Connie Eble Table 4 for African American writers. This is because the two corpora are fundamentally different in some ways and do not permit the same type of generalization. For example, the overseer corpus is limited and unbalanced geographically and has, unlike the African American corpus, few single letters from individuals. Nonetheless, if we were to follow the same methodology rigidly, using a random selection of letters from multipleletter writers, the resulting percentages for the South Atlantic and Mississippi Valley overseers would be lower than for African Americans in the same regions. Bailey also suggests that we and other scholars have misunderstood the nature of the "urbanization" of the post-bellum South, i.e., that it reflected the growth of towns rather than simple migration off the countryside to work in newly developing industries. We note that the latter view seems to be the one of Brown, who characterizes the vowel merger as a "rural, insular" feature that expanded with the urbanization of the South after about 1880. If not from off the land, from where else would newly urban Southerners have predominantly come? If they were mainly the children of town residents, would towns have grown so rapidly? On these matters we look forward to Guy Bailey's forthcoming book.

Vowel merger in west central Indiana: A naughty, knotty project Betty S. Phillips



The merger of the h! and lal phonemes in words such as naughty and knotty in American English is a widespread sound change whose beginnings predate the twentieth century (Kurath and McDavid 1961: 31-35, 62-65; Lance 1994: 356). Much of the United States has been affected by this merger, one area of which apparently spread from western Pennsylvania through central Ohio and Indiana, yet Ohio and Indiana speakers in some communities still vary in the extent to which they merge the two vowels. For example, according to Thomas (2001:94-95), two males from Johnstown, Ohio, exhibit different behaviors one merges the vowels, the other does not - even though their birthdates are only one year apart (1921 versus 1922). Similarly The Phonological Atlas of North America (Labov, Ash, and Böberg 2002) finds variable pronunciations and perceptions in much of Indiana. The variability of this merger is also readily apparent in classes on the history of the English language at Indiana State University, located in Terre Haute. Most of the students at Indiana State come from west central Indiana, and often the majority in a class express difficulty perceiving the difference between the two phonemes and insist that the pair caught and cot are pronounced identically. Since mergers are not infrequent in the history of English and the process by which they occur is often difficult to conceptualize, the investigation of the h l - l a l merger became a class project, the methodology and results of which forms the basis of this article. The aim of this article, therefore, is twofold: first, to report on a small portion of that project over several years (1998-2001); and, secondly, to demonstrate the value of incorporating such a project into a course on the history of the English language.

448 Betty S. Phillips 2.


The students in the class, usually juniors and seniors, interviewed and recorded at least three people who had lived in Indiana preferably all of their lives but at least since their fifth birthday. After gathering background information about them, including age, education level, occupation, and places they had lived, students let the subjects silently read through and then read aloud a short paragraph devised by the original class and later modified only slightly to include the words naughty and knotty. For this reason, those words are not available from all of the speakers. To lessen the phonetic variables, the potential homonyms were limited to words containing vowels followed by the alveolars /η, 1, t/ {Dawn-Don, Pauley-Polly, naught-knot/not, naughty-knotty, taught/taut-tot, wrought-rot), plus homonyms of two words which retained an original /l/ in their spelling {palms-poms, stalk-stock). The final paragraph is included in example (1). (1) Don Smith taught History in my junior high school. His face was often wrought with worry that we did not understand what he was telling us. He would stand behind his knotty pine lectern or else stalk around the room rubbing his palms against his thighs. I think the time of year that was the worst was when he discussed the stock market. My stomach would tie into a knot. One student, Pauley, was always trying to find reasons for getting tests cancelled, but it was all for naught. Two other students in the class, Dawn and Polly, once brought their pom poms to class, to take to a pep session for "Toys for Tots." Don was not pleased and said they were being naughty. He pinned the poms to the bottom of the window blinds, raised them, pulling the rope taut, and left them there to rot. The students were instructed not to tell the interviewees ahead of time what the specific object of the investigation was. They were admonished, however, to assure the interviewees that there was no right or wrong way to pronounce any of the words in the paragraph.1 The people interviewed were then asked to do a self-report, indicating whether they thought the pairs or triplets being investigated were homophones or not. To emphasize again the non-prescriptive nature of the task, the instructions included the following: "Homophones are words that are spelled differently but pronounced the same, for example, knight and night. Do you consider the following pairs/triplets homophones? (If they are pronounced the same, write SAME; if they are pronounced differently, write DIFFERENT. For triplets, give specifics.)"

Vowel merger in west central Indiana 449 The students were also instructed to listen to their tapes, to fill in a check list of which words were pronounced with which vowel, and to compare their findings with those on the TELSUR (Telephone Survey) website of the Phonological Atlas of North America (Labov, Ash, and Böberg 2002) closest to the location of their informants. Although students collected data from all over Indiana, only those from Terre Haute are included in the data below. The TELSUR data on Terre Haute is given in Table 1. Table I.

TELSUR Informant TS228, 30-year-old female nurse from Terre Haute and TS229, 19-year-old male student (Labov et al. 2002)

Male's Female's Male's Female's Production Production Perception Perception same before nasals close same close before Ixl close same same same close before Dd same close distinct before III close same same same overall close close close close *Characteristic pairs: Don~Dawn, cot~caught, sock~talk, collar-taller Environment*

When the students handed in their projects, we discussed their experiences and compared the self-reports of the age groups for the largest group of informants, those from Terre Haute or nearby.


Linguistic Results

After six semesters, the students had provided enough data to lend some insight into the merger. The data reported on below are from white informants from Terre Haute, excluding those who had not fulfilled the requirement of having lived there all of their lives at least since their fifth birthdays. Since the project covered three years, the ages were adjusted, adding, for example, three years to the earliest informants, those interviewed in the fall of 1998. Three groups emerged whose results could be fruitfully compared: 10 younger women (ages 18-24), 10 younger men (ages 18-24), and 10 older women (ages 47-53). There was only one older male informant from Terre Haute in the age range 47-53, so older males were not included in the study. Only one of the younger women, an instrument technician at a hospital, listed her occupation as something other than student. Of the young men, six listed their occupation as student, one worked in a pizza shop, two were waiters, and one left the

450 Betty S. Phillips question unanswered. Of the older women, two were teachers, and the others listed their occupations as administrative assistant, coordinator of youth ministry, custodian, homemaker, management, nurse, secretary, and shop owner. One of the teachers had attained a master's degree, but the other respondents ranged from high school graduate (2) to college graduate (1), with two leaving the question unanswered. More detailed social factors were not controlled for.

3.1. Younger versus older females A comparison of the younger and the older females reveals the path of the merger in apparent time. The first set of data, shown in Table 2 below, are the self-reports. In this, as in all subsequent tables, taut and taught are reported as or (in the auditory analysis tables) pronounced the same unless specifically noted otherwise. The word not is included in the self-report tables because it was invariably reported as being pronounced the same as knot, it is not included in the auditory analysis tables, because often speakers failed to give it sufficient prominence. Table 2. Self reports of "same," Terre Haute, IN, females 18-24-year-olds 50% Dawn-Don Pauley-Polly 80% wrought-rot 50% taught/taut-tot 70% naught-knot/not 70% naughty-knotty* 63% palms-poms 20% stalk-stock 10% *of 8 younger informants, of 6 older informants

47-53-year-olds 10% 20% 50% 20% 30% 33% 10% 10%

Although self-reports are notorious for being unreliable, in this case it is nonetheless interesting that the respondents, having been told that these so differently spelled forms could be pronounced differently, nevertheless report many of them as the same. Table 2 shows that for almost all of the pairs/ triplets, the younger speakers are more likely than the older speakers to report them as homophones. The exceptions are wrought/rot and stalk/stock, for which the groups have equal percentages. Younger speakers often report a phonetic [1] in palm and stalk, and occasionally they do actually produce an [1],

Vowel merger in west central Indiana 451 so the difference they report may not be the same difference that the older speakers are reporting. In order to compare the actual pronunciations, I took the 20 recordings and sectioned out the minimal pairs using Signalyze software. I then listened to them side by side to determine if they sounded "the same" or "different". If I could not decide after multiple listenings, then I labeled them as "close." Actual phonetic manifestations varied from [A] to [D] to [D] to diphthongized variants. Certainly, many Terre Haute speakers produce a rounded variant as the merged vowel. For instance, my first encounter with the merger in Terre Haute was hearing a now 40-year-old refer to her brother as Dawn; and years went by before I learned that her friend Pauley spelled her name P-o-l-l-y. The results of the auditory analysis are shown in Table 3. From this table, it does appear that younger speakers not only are more likely to report the investigated pairs as homophones, but that they also in fact are more likely to produce the pairs as homophones. Table 3. Auditory analysis, Terre Haute, IN, females 18-24--year-olds 47-53· -year-olds Same Close Same Close Dawn-Don 70% 20% 30% 0 Pauley-Polly 90% 10% 50% 10% wrought-rot 40% 10% 30% 0 taught/taut-tot* 56% 10% 0 20% naught-knot 90% 0 70% 0 naughty-knotty* * 50% 10% 17% 0 palms-poms 20% 40% 60% 0 stalk-stock 40% 0 60% 20% * For taught/taut-tot one younger female speaker's pronunciations were indeterminate; she pronounced taut as taunt and taught very quickly. **Of 8 younger informants; of 6 older informants

Table 3 reveals something else as well. When one looks at which words and which environments most encourage the merger, if synchronic variation mirrors diachrony, the results disagree with Bailey (1973: 19), who says that this shift began before IM + vowel (e.g., naughty), then extended to other environments involving a following alveolar (e.g., caught, dawn). According to Table 3, the environment before /t/ produces both the most and the least merger. It depends on the word pair in question. The pair naught/knot has the highest percentage of merger for both younger and older female speakers (90% and 70% respectively), with the merger of naughty and knotty far behind (50%

452 Betty S. Phillips and 17%, respectively). The environment before /η/ and l\l - Dawn/Don, Pauley/Polly - shows more frequent merger than the other words ending in /t/: Dawn/Don (70% and 30% for younger and older speakers, respectively), Pauley/Polly (90% and 50%), taught/taut/tot (50% and 20%), wrought/rot (40% and 30%). And the results in Table 3 conflict to some extent with the Phonological Atlas of North America, which says that, "[throughout the studies of this merger, it has been observed that it occurs more rapidly before nasals.... There are 107 such cases. But there are only 28 ... representing the reverse situation, where the merger is more advanced before /t/" (Labov, Ash, and Böberg 2002). Actually, if one reads it carefully, this quotation says that quite a few studies, over 25%, show the merger as more advanced before /t/. Our results show a mixture, with the individual words being perhaps as important as the specific phonetic environment. With such a small sample group encompassing different grammatical classes, however, one hesitates to appeal to lexical diffusion (see, for example, the discussion on lexical diffusion and word class in Phillips 2001: 128-134). One can only speculate, therefore, on what might contribute to the difference between some word pairs; for example, naught is a far less frequent, less familiar term than is naughty·, wrought is also infrequent and relatively unfamiliar but its initial Irl may exert some influence.

3.2. Males compared to females When young male speakers are added to the data (Tables 4 and 5), the results support the merger as a sound change that is led by females. According to the males' self-reports, shown in Table 4, except for the words containing etymological l\l, the males are either in agreement with both female groups (i.e., in the case of wrought-rot) or are intermediate between the two groups. Table 4. Self reports of "same," Terre Haute, IN, including young males 18-24 Females 18-24 Males 47-53 Females Dawn-Don 50% 20% 10% Pauley-Polly 80% 40% 20% wrought-rot 50% 50% 50% taught/taut-tot 70% 30%* 20% naught-knot/not 70% 50% 30% naughty-knotty 63% 40% 33% palms-poms 20% 0 10% stalk-stock 10% 0 10% *Two who reported "different" wrote "taught different, tot and taut same."

Vowel merger in west central Indiana 453

In Table 5, the auditory analysis of the males has been added, with very similar results, although not as consistently. Still, the males overall show fewer mergers than the females in their age group, which suggests that the females are leading this sound change. This finding replicates that of Thomas (2001: 92-102), who found that among speakers from Johnstown, Ohio, and its vicinity all the females analyzed (born 1900, 1910, 1933, 1956, 1959, 1963, 1981, 1982, 1982) had merged h i and /a/, excepting perhaps one, born 1935, about whom it is noted, "Though she distinguished lal and bl in minimal pairs, it is not clear that she did in the reading passage" (96). The younger males also merged the vowels (born 1981, 1982, 1982). But of the males born before 1960, more kept the vowels distinct (bom 1917, 1922, 1950, 1956) than merged them (born 1921, 1959). Table 5. Auditory analysis, Terre Haute, IN, including males 18-24 Males 47-53 Females 18-24 Females Same Same Close Same Close Close 0 Dawn-Don 70% 20% 20% 0 30% 50% 10% Pauley-Polly 90% 10% 67% 0* wrought-rot 30% 0 40% 10% 67% 0 0 taught/taut-tot* 56% 10% 10% 10% 20% naught-knot 70% 0 90% 0 44% 0 naughty-knotty* * 50% 10% 0 17% 0 25% palms-poms 40% 60% 0 20% 80% 10% stalk-stock 20% 40% 30% 20% 60% 0 *One male inverted the pronunciations: Pauley [a], Polly [d] """Indeterminate pairs, due to recording quality or misreadings, included for wrought-rot 1 male, for taught/taut-tot 1 younger female, for naught-knot 1 male, and for naughty-knotty 2 males.

3.3. Perception After several semesters, a very basic perception experiment was added to the project. It consisted of having the subjects listen to a tape on which I had recorded a list of words containing either the rounded or the unrounded vowel - taught or tot, for example - and having the subjects circle which word of the pair on their response sheet they thought they heard. If they were unsure, they were to circle both words. Since the respondents often clearly assumed that they would not hear the same word twice, thus nullifying the second of the

454 Betty S. Phillips pair, only the results of the first of a pair is reported below in Table 6, namely the words Pauley, naught, wrought, Don, knotty, tot, stock, and poms, pronounced with their traditional vowels. And only responses that indicated a loss of the distinction in perception are indicated. The respondents are the same as those reported on above but are of necessity limited to those who participated in the experiment after the perception task was added: 3 younger females, labeled a-c; 7 younger males, labeled d-j; and 4 older females, labeled k-n. This labeling allows one to see how some specific informants had particular difficulty with the task (f, k, and 1, for example) and how many informants had no difficulty (c, i, j, n). Note that even among the oldest group of informants, the distinction was not uniformly maintained in perception. Table 6. Perception difficulties Read



Polly indecisive knot indecisive rot indecisive

naught wrought

Don knotty tot stock poms


Dawn indecisive naughty indecisive taught indecisive stalk indecisive palms indecisive

Three 18-24 Females: a-c b

Seven 18-24 Males: d-j

Four 47-53 Females: k-n



f k e,f a a,b

1 f d, h

k, 1

?>. f >l a,b

1 g d,h f h

k, 1, m

Pedagogical results

Students who participated in the above project predictably learned a lot about research design, implementation, and the drawing of conclusions. Class discussions included concerns over the shortcomings of each part of the project, including such issues as what other speaker variables might be

Vowel merger in west central Indiana 455

significant, how many informants constitute an adequate sample, whether the results mirror natural speech, and what other factors might influence the results. These of course are the same questions they should ask of all research, whether in historical linguistics or in other fields. In addition, two of my course goals are (1) for the students to appreciate the systematic nature of language change as well as the variation within that systematicity, and (2) for them to understand how dialects develop, thus deepening their appreciation of the diversity within Modern English. By incorporating into the course a project such as the one described above, not only do students gain an understanding of how other mergers in the history of English could have occurred, but they see how a diachronic change affects synchronic variation in the language around them. Their own written project summaries often reflect this understanding (examples (2)-(6)). (2) I chose a mother and daughter to speak. The mother tended to have a very obvious distinction between the lal and bl. The daughter had less of a distinction. I think the reason for this is that the mother was raised in northern Indiana ... (3) I was surprised at some of my findings because I had never listened closely enough to know how differently my friends and relatives spoke from each other and from me. (4) It's funny how two friends who live just miles apart in the same county have such differences in the pronunciation of these words. (5) My sister and I spoke more similarly than our parents.... My sister and I pronounced more of the words on the list the same than my parents. (6) My brother and mom could only distinguish between a few words with these sounds and made no real distinctions when they articulated words with these sounds. My dad, on the other hand, could hear the difference in all the words played for him and made the distinctions a few times himself when he read the passage. One student in particular summed up another important, yet unstated, goal of the course, especially important for students who are preparing to become high school English teachers; see example (7).

456 Betty S. Phillips

(7) My mother has always been a real stickler about correct pronunciations, and I have always somewhat looked down on people who can't pronounce words correctly. But this project... has allowed me to see how varied pronunciations can be. And that change is due to change in dialect...not intellect. So I think I have lost some of my bias against people who don't pronounce things the same way I do. Overall, the benefits of the project may be summed up by a Confucian proverb often quoted by proponents of experiential learning: "Tell me and I forget. Show me and I remember. Let me do and I understand."

5. Conclusion This investigation of the merger of b l and lal in Terre Haute, Indiana, reveals that the distinction still maintained in production by many older speakers is being gradually lost by the younger generation. The findings are consistent with those of Böberg and Strassel (1995) for Cincinnati; Thomas (1996,2001) for Johnstown, Ohio; and Rueppel (1996, as reported in Böberg and Strassel 2000) for St. Louis. Similar to findings in Thomas (2001: 92-102), it seems that the females are leading this change. According to Böberg and Strassel (2000: 108), "One important region of North America that has largely escaped detailed study is the interior North Midland, the territory extending from Pittsburgh to St. Louis, between the Inland North and the Ohio River." Perhaps this modest investigation into the speech of Terre Haute will encourage others to investigate Indiana speech more thoroughly - and to include such projects in courses in the history of the English language no matter where they are taught. Note 1.

The phrase "fraught with worry" would probably have been more idiomatic than the "wrought with worry" in the paragraph. But to my knowledge no one ever commented about this phrase.

References Bailey, Charles-James 1973 Variation and Linguistic Theory. Arlington, VA: Center for Applied Linguistics.

Vowel merger in west central Indiana 457 Böberg, Charles, and Stephanie M. Strassel 1995 Phonological change in Cincinnati. In Proceedings of the 19th Annual Penn Linguistics Colloquium, Rajesh Bhatt, Susan Garrett, Chung-Hye Han, and Roumyana Izvorski (eds), 25-35. (University of Pennsylvania Working Papers in Linguistics 2.2.) Philadelphia: University of Pennsylvania. Böberg, Charles, and Stephanie M. Strassel 2000 Short-a in Cincinnati. Journal of English Linguistics 28: 108-126. Kurath, Hans, and Raven I. McDavid, Jr. 1961 The Pronunciation of English in the Atlantic States. Ann Arbor: University of Michigan Press. Labov, William, Sharon Ash, and Charles Böberg 2002 Phonological Atlas of North America. Linguistics Lab, University of Pennsylvania. 18 July 2002. Available: . Lance, Donald 1994 Variation in American English. In American Pronunciation. 12th ed., expanded, John Kenyon. Donald M. Lance and Stewart A. Kingsbury (eds.), 345-373. Ann Arbor: George Wahr. Phillips, Betty 2001 Lexical diffusion, lexical frequency, and lexical analysis. In Frequency and the Emergence of Linguistic Structure, Joan Bybee and Paul Hopper (eds.), 123-136. Amsterdam: John Benjamins. Rueppel, Rebecca 1996 A study of change in the speech of St. Louis, Missouri. Unpublished manuscript. Thomas, Erik R. 1996 A comparison of variation patterns of variables among sixth graders in an Ohio community. In Focus on the USA, Edgar W. Schneider (ed.), 148-168. Amsterdam: John Benjamins. 2001 An Acoustic Analysis of Vowel Variation in New World English. (PADS 85.) Durham, NC: Duke University Press.

The spread of negative contraction in early English Richard M. Hogg

One of the most obvious features of Old and Middle English syntax is the phenomenon known as negative contraction.1 As far as Old English is concerned, the process is one where the negator ne combines with a following verb: in Old English the relevant verbs are beon 'be', habban 'have', willan 'want', witan 'know' and, albeit less frequently, ägan 'own'. Thus in negative contraction we obtain forms such as: nis 'is not', ncebbe 'has not', nelle 'does not want', nyton 'don't know'. Similar facts hold for Middle English until the time, largely in the fifteenth century, that the construction disappears. It is well known that this construction is subject to various controls. Firstly, the process of contraction only occurs when there is an immediately following verb which has either a zero onset or initial fhf or /w/.2 Secondly, contraction only occurs if the following verb is beon, habban, willan, or the preterite-present verbs ägan and witan. Strangely, there is no sign in either Old or Middle English of negative contraction with weordan 'be, become', despite the fact that this verb clearly shares syntactic features with beon and has onset /w/. Levin (1958: 493) points out that nertha as the negated form of wertha 'become' does occur in Old Frisian, so its absence from Old and Middle English remains unexplained. However the main purpose of this paper is to examine a further feature of negative contraction, rather than either of the ones above, namely the dialectal status of the construction. There are three existing studies which discuss this matter, namely Levin (1958) and Iyeiri (1992, 2001), which, together, present most of the necessary material. This current work attempts, whilst respecting that body of work, to suggest that there must remain some uncertainty as to the most plausible history of the construction. The obvious place to start is with Levin's work, because it is earlier and because it deals with both Old and Middle English, whereas Iyeiri deals only with Middle English and to a greater extent with verse rather than prose. Examining Levin's study of Old English, we find that he deals with the following texts:

460 Richard Μ. Hogg


West Saxon (1): Cura Pastoralis, Orosius, and Late West Saxon Gospels (Corpus ms); West Saxon (2): jElfric, Wulfstan, Late West Saxon Gospels (Hatton ms); Mercian (1): Rushworth' and Vespasian·, Mercian (2): Peterborough Chronicle·, Northumbrian: Lindisfarne Gospels and Durham Ritual.

In Tables 1 and 2 below, I have followed Levin's chronological classification for West Saxon and Mercian, both of which he sorts into preand post-1000 material, but see Levin's paper for details for which I have no space here. The results of Levin's study are as follows: Table I. Negative contraction in West Saxon West Saxon Cura Pastoralis Orosius Gospels: Corpus jElfric Wulfstan Gospels: Hatton

+ contract 82 145 79 477 281 78

- contract 3 3 3 4 14 4

% 96% 98% 96% 99% 95% 95%

- contract 33 43 23 28

% 73% 61% 63% 69%

Table 2. Negative contraction in Anglian Anglian Vespasian Lindisfarne Rushworth1 Peterborough

+ contract 88 66 39 61

Although there are some problems with these results, problems of which Levin is fully aware and including his decision to take The Peterborough Chronicle as an example of Old rather than Middle English, the overall conclusion is unavoidable: uncontracted forms are almost completely restricted to Anglian dialects, and furthermore this state of affairs appears to be persistent. This is confirmed by data collected by Levin from a variety of fourteenth- and fifteenth-century texts (Table 3). To the above we can add two significant elements noted by Levin. The first of these is

Spread of negative contraction 461 that the material attributed to Wulfstan actually contains homilies not written by him, and all of the uncontracted forms come in manuscripts which should not, or are unlikely to be, directly associated with him. The second feature is that the overall results do not make it clear, as Levin's paper itself does, that the uncontracted forms in The Peterborough Chronicle are exclusively from the later or interpolated sections of the Chronicle and these sections are quite clearly based on an East Midlands dialect. Table 3. Middle English negative contraction (Levin, 1958)


DAT· Snrt cm. IJOO (B. 1-1000) Owl *ni Ik* 13th NilhtingeU Cent Vn *ni Utt WtlJ ijth Cent AncmΜ RbeU 13th (pp. i - » o ) Cent. St. Uergent 13th Cent 13th Cent Si. Jettane ijth Cent SnUi Ward* 13th Cent. BMUtiSktd 13th Cent. Robert of Glon- cs. cater'· Ckren- 1300 id* (pp. I-IOO) / « » f t *f Arim* 14th Μ* Cent St. KaUuritu



DIA- COM- tmooNerr LECT NACNS TSACNO SWM B*tti*ry none 40 S SWM SWM




There is much to admire in Levin's work, yet from the point of view of dialectology there are, at least, some issues which are inadequately addressed. In order to demonstrate this, let me take two short quotations from the paper:

462 Richard Μ. Hogg

The significant fact is that in West Saxon the usage almost entirely favors contraction, whereas in Anglian uncontracted forms are freely employed. (Levin 1958: 495) In Middle English the area of contraction comprises the Southern and West Midland dialects (these two dialects in Middle English carry on the West Saxon literary and linguistic traditions); on the other side of this isogloss are texts of East Midland and Northern provenence. (Levin 1958: 498) At first sight this looks unexceptional, and it certainly equates with Tables 1-3 above. It does not, however, equate with anything else we know. In particular it should be noted that Levin assigns Vespasian to Mercian Anglian, as we might expect, but he assigns the AB Middle English dialect texts, such as Ancrene Riwle and Hali Meidhad to his southern, i.e., nonAnglian, group. Yet, at least in terms of Second Fronting, but in other respects too (see Hogg 1997 for further shared features which extend to Middle English as well as Old English), there is a set of important isoglosses which serve to demarcate the west midlands as a dialectally distinctive area. In order to substantiate his argument, therefore, it would have been necessary for Levin to provide more evidence for his claim that although Vespasian is Anglian, a text such as Ancrene Riwle is not. The only comment is the one quoted above that the west midlands dialects, like West Saxon ones, "carry on the West Saxon literary and linguistic traditions." But this notably fails to explore the consequence of an older Mercian literary tradition as discussed by Vleeskruyer (1953). It is also clear that the style of the AB texts owes as much to the rhetoric of contemporary Latin writers as it does to Old English forms, see Bethurum (1935) and Millett (1982: lii-lvi) amongst others. It seems, therefore, that, in terms of the Vespasian and Ancrene Riwle, Levin's hypothesis does not provide an explanation for the radical divergences between these two texts. Leaving these difficulties aside for a moment, there is still much to be discerned from Levin's figures. For example, in the texts before 1200, not only can it be seen that there is a distinct difference between West Saxon texts and the others, but it is noticeable that there are fewer uncontracted forms in Vespasian (73%) than in Lindisfarne (61%) and Rushworth' (63%) and that The Peterborough Chronicle (69%) might be held to be closer to the west midlands Vespasian than the northern Lindisfarne, although this is ultimately undecidable. When we move to Levin's later texts from the East Midlands and the North, they appear to divide into three main sub-groups: (i) minority uncontracted - Ormulum; (ii) 50-50 - Vices and Virtues; (iii) mostly uncontracted - Bestiary, Genesis and Exodus,

Spread of negative contraction 463 Body and Soul; (iv) the remainder, including texts without uncontracted forms. Perhaps the most reasonable conclusion to draw from his figures is that there was a general tendency in these northern and eastern areas for the preference for uncontracted forms to become more and more common, to the point, indeed, of universality. If that were all that there was to say, then I would be in the same position as Levin. In particular I would not have given any explanation of why Vespasian should be so different from the Middle English AB texts. In order to attempt some explanation I want to broaden the investigation to take in two further texts, namely the Kentish Sermons, Laud ms. 471 (see Bennett and Smithers 1968: 213-222) and St Chad (see Vleeskruyer 1953). Neither of these texts is sufficiently rich in examples to provide any measure of certainty, but they still remain of considerable interest and importance. The latter text provides an intermediate point between Vespasian and the AB dialect, whilst the former repairs an omission in Levin's study since it did not include any Kentish texts. Presumably one reason why Levin did not include any early Kentish texts is because they simply have too little information about the Old English period, with only a rare scattering of forms which fail to convince. Nevertheless, when we turn to the early Middle English Kentish Sermons, the results are perhaps somewhat surprising. For what we find is that, in the chosen sample, there are 7 uncontracted forms against 2 contracted forms. Even if the sample is rather small, it points quite clearly in favor of what Levin calls the East Midland and Northern category and sharply away from the Southern and West Midland category. Although the sample is small, it is of some interest that an extract from Kyng Alisaunder (see Bennett and Smithers 1968: 28-39) which may originate from London, although it shows predominantly contracted forms has a minority of uncontracted ones, see also Iyeiri (2001: 172). Particularly interesting is the contrast between both texts and the Owl and the Nightingale, see Table 3. This all suggests that Levin's classification of a category South and West Midland needs to be recast as a line which starts immediately to the west of London so that at least Kent, and perhaps London, falls into an eastern category. Turning now to St Chad, what we find there is a single uncontracted form alongside about half-a-dozen contracted forms. This rather confirms what Levin reports about the AB dialect, but in doing so it rather contradicts what he reports about Vespasian, as I have already noted. It also appears to be confirmed by contracted forms used by the Worcester Tremulous Hand, and not merely in his manuscript of Kline's Grammar. What conclusion should we draw? I think that there is more than one conclusion that might be available to us, but unfortunately they are not wholly compatible with each other. Firstly, given the real possibility that

464 Richard Μ. Hogg Vespasian might well show a dialect of north-east Worcestershire (see Kitson 1990), we might be tempted to suggest that the isogloss for contracted forms cuts south-east to north-west through Worcestershire. A second possibility is that Vespasian does indeed directly reflect the dialect of Lichfield. However, even if that is correct, it is not clear that it solves anything, because the relationships among Vespasian, the early glossaries, St Chad, and the AB dialect texts are not changed by such random geographical relocation. There are, I think, two better ways in which we can attempt a solution. To a large extent both ways take as their starting point the situation, which is quite clear, that pertains in Old English. Firstly, although it is indisputable that the available Old English material shows both that West Saxon was a "contracted" area and that Anglian, in relative terms, was an "uncontracted" area, there is no reason why we should then go on to assume that the same is true of even the earliest Middle English or even, for that matter, the transitional language of the twelfth century, but see in particular the comments in Fulk (1992: esp. 147-149). Thus there is no reason for not suggesting, at least at present, that although contraction originated in the West Saxon area, it then spread north-westwards, after the ninth century but before the time of a text such as St Chad, in other words, before, say, 1100. This explanation would allow us to account for some of the difficulties, and it could well be right. But note an unexpected consequence. It necessarily involves the creation of an isogloss which cuts Mercia in two, one area, which we can call West Mercia, and a second, which we may as well call East Mercia, if only for the sake of convenience. Alternatively the line of Watling Street might serve as an iconic starting point, see Figure 1. This leads me to the second point. If we turn back to Table 2, it is at least plausible to suggest that Vespasian has a significantly higher percentage of contracted forms than Lindisfarne and Rushworth'. Now there is always an area close to any given isogloss which shows extreme variability. Could Vespasian be an example of such variability with respect to contraction?

Spread of negative contraction 465

Figure I. Roads in Anglo-Saxon England (after Hill, 1981) Quite obviously it is rather difficult, perhaps even vainglorious, to attempt any more detailed analysis of the Old English data. Nevertheless it is worth looking at a further text not examined by Levin. This text is The Salisbury Psalter (PsGlK, see Sisam and Sisam 1959). The interest and advantage of this text is that, like Vespasian it is an interlinear psalter gloss; it is also, of course, an interlinear gloss similar in type to Rushworth' and Lindisfarne. The following remarks by its editors are relevant (Sisam and Sisam 1959: 28): This language is interesting chiefly for the evidence it provides of Southwestern English at the beginning of the twelfth century. The gloss was probably copied at Shaftesbury ... and it is basically late West Saxon ... it bridges the gap between late West Saxon and the South-Western dialects of early Middle English.

466 Richard Μ. Hogg The fact that the gloss is late places it just a little earlier than St Chad, and since it is localizable to Shaftesbury, probably at the Benedictine nunnery there, 3 means that we are dealing with a text from the very center of historical Wessex, only a handful of miles from Wilton, which is, at the least, to be closely associated with Alfred - see Kitson (1993), also Kitson (1997: fn. 2). The reasonable expectation, therefore, would be that Salisbury would, overwhelmingly, show negative contraction, perhaps even to the exclusion of uncontracted forms, just as all the Old West Saxon texts in Table 1 do, and as the further expansion of contracted forms in early Middle English demonstrates. I have examined all the contracted and uncontracted forms of negative + is, wees and wceron in the text. In all, I collected 50 uncontracted forms and 7 contracted forms. In other words there are more than seven times as many uncontracted forms as contracted ones, with 88% uncontracted. What non-dialectal explanation can be offered? The mere fact that Salisbury is an interlinear gloss is irrelevant, for so is Vespasian. On the other hand, it is notable that the usual, although not exclusive, form of the negative in Salisbury is na rather than the much more common ne. It could be argued, therefore, that contraction does not take place because na is more stressed than ne. This is plausible, although it does not explain the presence of 7 contracted forms. It is worth, therefore, comparing Salisbury with the Winchester-produced tenth-century Regius Psalter (PsGID, see Roeder 1904), for that gloss, like Salisbury, has a strong preference for na over ne. The relevant figures are: uncontracted 9, contracted 44. In other words, the percentage of uncontracted forms is 17%. There can be all sorts of explanations for this discrepancy, such as the unwillingness of the Salisbury scribe, unfamiliar with Latin (see Sisam and Sisam 1959: 17-18), to depart from the isolated Latin non. But such exceptions can only be taken so far. If we were generous, then we might suggest that the "real" proportion of uncontracted forms in Salisbury was nearer to 50% than 88%. But that itself is an unwarranted assumption. There seems to be no alternative to the view that this text has a significantly high proportion of uncontracted forms and that this is entirely out of line with our expectations. That the Regius Psalter has 17% of uncontracted forms is also of some significance. The figure is low in comparison with Salisbury, but it is high compared with the figures found by Levin in what may be regarded as canonical West Saxon texts. It would seem that there were competing varieties even in the areas generally regarded under the single rubric of West Saxon. In order to take all the above issues forward, we need to examine the work done in Iyeiri (1992, 2001), together, as we shall see, with the data

Spread of negative contraction


presented in LALME (Mcintosh, Samuels and Benskin 1986). The one disadvantage of Iyeiri's work, from our point of view, is that she is dealing primarily with verse texts. However, this does not invalidate her results, even if it means that the range of texts is thereby limited. As we saw above, negative contraction is, both in Old English and early Middle English, a syntactic feature which appears to be dialectally controlled. Iconically, at least, there seems to be a persistent split between texts to the south and west of Watling Street, which favor contraction, and texts to the north and east, which generally show uncontracted forms. Obviously, therefore, the first question concerns the degree to which Iyeiri's results match those from Levin (1958). Here is a very brief account of the texts used by Iyeiri:4 Table 4.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

Texts analyzed by Iyeiri (1992)

Poema Morale (C 13b)5 The Owl and the Nightingale (Cotton Caligula A ix, C 13b)6 King Horn (Cambridge U.L., Gg.iv.4, C13-14, West Berkshire) Havelok (Bodley, Laud Misc. 108, ca. 1300, West Norfolk) South English Legendary (Cambridge, CCC 145, CI4a, West Berkshire) English Metrical Homilies (Edinburgh, Royal College of Physicians, C13b-14a, Yorkshire) Genesis and Exodus (Cambridge CCC 444, CI4a, West Norfolk) William ofShoreham (BL Additional 17376, C 14a, Kent) Cursor Mundi (Cotton Vespasian A iii, CI4a, West Yorkshire) Sir Ferumbras (Bodley, Ashmole 33, C14b, Devon) Confessio Amantis (Bodley, Fairfax 3, C14b, Kent/Suffolk)7 Handlyng Synne (Bodley 415, ca 1400, Hertfordshire) Kyng Alisaunder (Bodley, Laud 622, ca 1400, Essex) Gawain and the Green Knight (BL, Cotton Nero A x, ca 1400, Cheshire) A lliterative Morte A rthure (Lincoln, Cathedral 91, C14a, Lincolnshire) Alexander and Dindimus (Bodely, Bodley 264, ca 1450, Gloucestershire) Destruction of Troy (Glasgow, Hunterian 338, ca 1450, Lancashire) Stanzaic Mörte Arthure (BL, Harley 2252, C14b, Rutland)

The results of Iyeiri's study are as follows:

468 Richard Μ. Hogg Table 5.

Contraction in Middle English verse (Iyeiri 1992)

Poema Morale Owl and Nightingale King Horn Havelok South English Legendary English Metrical Homilies Genesis and Exodus William of Shoreham Cursor Mundi Sir Ferumbras Confessio Amantis Handlyng Synne Kyng Alisaunder Gawayne and the Green Knight Alliterative Morte Arthure Alexander and Dindimus Destruction of Troy Stanzaic Morte Arthur

+ contract 95 102 31 3 626 0 3 153 0 113 38 4 135 12 1 6 1 33

- contract 1 0 2 56 3 17 27 5 52 43 22 46 59 0 11 9 7 12

% contract^ 99% 100% 94% 5% 99% 0% 10% 97% 0% 72% 63% 8% 70% 100% 8% 40% 12% 73%

If we classify the above results into categories, then we find the following groups: Table 6.

Contraction by frequency based on Iyeiri (1992)

+ contract Owl&N Gawayn Ρ Morale SE Legend Shoreham King Horn

% + contract Stanzaic MA Sir Ferumbas Kyng A Confessio A

% - contract Gen & Ex Destr Troy Alex & Dind

- contract Ε Metrical Η Cursor Μ Havelok Η Synne Allit MA

Referring back to Table 4, it can be noted that of the fully contracted texts, two are placed in Berkshire, one is Kentish, perhaps two from near Herefordshire and one from Cheshire. In contrast, of the fully uncontracted texts two are from Yorkshire, one from Lincolnshire, one from Norfolk, and one from Hertfordshire. The last of these, Handlyng Synne, must cause obvious concern, but I shall return to the issue below. The mostly

Spread of negative contraction 469 contracted texts include one from Essex, one with both Kent and Suffolk origins although probably written in the London area, one from Rutland, and one from Devon. Let me, for a few moments leave aside the mostly uncontracted texts, each of which poses special questions. In principle, however, Iyeiri's results both confirm Levin's earlier work and nevertheless present a rather different picture of some parts of the country. It is, of course, true that her results do show that negative contraction indeed remained as a dialectal variable until as late as the midfifteenth century.8 There is one practical difficulty as we approach this topic, in that the maps provided by LALME only cover the southern part of the area, but even so we can make considerable progress. Iyeiri (1992: 219) sees the central area of negative contraction (in the south) as being Worcestershire, Warwickshire, Gloucestershire, Oxfordshire, and Essex, at least in the case of nam. The final county in this list must look quite aberrant. It neither fits Levin's view of a north-east south-west divide, nor the suggestion that Watling Street might be regarded as an iconic divide of the two contrasting areas.9 Yet Iyeiri's characterization of the central area of contracted forms in the south itself seems somewhat at odds both with what her results show and the dot maps of LALME. Even in her later book (Iyeiri 2001: 159), she sees a similar divide between a western and an eastern area. One explanation of some of the problems which arise is one which Iyeiri herself suggests, namely that, especially in the later texts, there is a chronological issue involved. As is mentioned in footnote 7, the system of verbal negation itself was changing everywhere and preverbal ne was being lost. This would have an effect on both contracted areas and uncontracted areas, albeit unequally. In areas where contraction was the norm, the loss of preverbal ne would not necessarily have any immediate effect, for the contraction might have been preserved idiomatically. But it is also the case that Iyeiri's figures do not tell the whole story. Take, for example, the following line from Gawayne: Bot Arthure wolde not ete til al were serued Although there are good metrical reasons why neither nolde nor ne wolde would be appropriate here, the absence of any preverbal negation is, more than anything else, a sign of the process of syntactic change; that is to say, it tells us only that verbal negation without preceding ne has become entirely possible, even if such negation is still only a rival to negation with preceding ne. This means, alas, that the example tells us very little of further value.

470 Richard Μ. Hogg

If the contemporary changes in verbal negation make it difficult to assess the dialectal status of the relevant forms, this suggests that the only possible solution is to look at the dialectal distribution of the forms themselves, in order to see whether or not there might be some internal clues. The search for these clues, of course, can only proceed by the use of the dot maps of LALME, and, as I have already indicated, the maps LALME provides are, unfortunately, only for the southern part of the country. 10 Nevertheless, there is much to be gleaned, in principle as well as practice. One issue of principle, for example, is that there are a variety of different contracted forms, such as ne + am, ne + have, ne + will, ne + wite, and it is not necessarily the case that any of these forms follows exactly the same pattern as any of the others (indeed, it is highly unlikely). Consider, therefore, the following dot map which shows the location of all southern examples of contraction before all forms of be:


• >

Figure 2. ne+be (LALME)

The first point to note is that I have added isoglosses to the original LALME map. It must first be noted that although this might appear to be no more than a clarification, if it is, then it is a dangerous one. It is always the case that any isogloss is merely a single interpretation of the data and it is always possible to draw any given isogloss differently. 11 I have attempted in the isoglossing in Figure 2, and also elsewhere, to include with any given isogloss items which are not separated by more than two or three

Spread of negative contraction 471 contrasting locations. But it is readily observable that in the eastern area of Figure 2 this has not been followed, since that area includes many locations which do not show the contraction. In such cases there is a choice to be made. Leaving aside other factors, to which I shall return shortly, the basic decision is whether to be restrictive or free. Under the first option we strictly limit the isoglosses so that within them only the locations with the relevant feature appear within the boundaries of the isogloss. Under the second we allow the isogloss to encompass every instance of the feature even though that will mean that many locations not having the feature will nevertheless be contained within the isogloss. Fortunately this decision is not between polar opposites but, rather, demonstrates the continuum along which any given decision must be placed. In the end the best placing of isoglosses is the placing which is most helpful in the relevant context. There is a further point to be noted here, one that is particularly acute in historical studies. In Figure 2, the interpretation of locations without negative contraction is always ambiguous. Of course this could imply that contraction was absent from such locations; but it can also always mean no more than that the textual material is insufficient to provide examples. This situation is far less likely to arise in present-day studies, where we have live informants who can be questioned in one way or another in order to elicit data (or its absence). Therefore the absence of negative contraction could reflect no more than the paucity of data.12 If we acknowledge that possibility and combine it with my comments on restrictive and free isoglossing as discussed in the previous paragraph, this leads to the conclusion that there should be a bias in favour of the latter type of isoglossing, since it takes into account the possibility that some locations may accidentally not furnish the necessary material. Having cleared away these procedural issues, we can now turn our attention to actual distributions displayed in Figure 2. The most striking result, obvious even without the benefit of isoglosses, is that there seem to be two quite distinct areas of contraction, one in the south-west especially around the Severn and the Gloucestershire Avon, and another in the southeast, centered around the Thames Estuary. This in itself should make us take stock, for it is a result which is quite clearly different from that obtained in Levin (1958), as is noted by Iyeiri (1992). Does this mean that Levin, quite simply, was wrong? Or does it show a change over time? Or most probably of all, that Levin's results are somewhat flawed because of his relatively small sampling? Whatever the case, it remains rather strange that there appear to be two separate areas of negative contraction. Helpfully, Iyeiri (1992) lists the counties where contraction seems to be found, but although her listings confirm that there are distinct western and

472 Richard Μ. Hogg

eastern areas of contraction, the difficulty of explaining this situation remains unsolved. But before I go any further, it is worth looking at another LALME dot map involving negative contraction, namely the one for the forms will. It will be immediately noticeable that there are some minor differences between this map and the map in Figure 2. As I noted above, some of these differences may be due simply to lack of evidence, and therefore it would be unsafe to draw too many radical conclusions. Yet there are points to be made. We can note, firstly, that there is one quite large area in this second map where contracted forms are absent, namely Kent. Conversely, Figure 3 has a significant number of forms in the area of Bedfordshire and north Buckinghamshire, whilst in Figure 2 there are only two isolated examples. There also appear to be rather more examples of ne+will and of ne+am in the far south-west.

Figure 3. ne+will


Let us return to Iyeiri's analysis and discussion, concentrating, but not exclusively, on the texts which show variability. Amongst texts which have only minimal variation, Handlyng Synne is remarkable, for it stems from an area where contraction is regular, but is itself almost always uncontracted. Iyeiri (1992: 207) suggests that this might be because the original author,

Spread of negative contraction 473 Robert Manning, lived in Bourne, Lincolnshire. However, such a supposition is not entirely convincing, because it would imply that there were other features, perhaps on a quite wide range, which also betrayed similar northern features. I shall discuss other possibilities later, but let us now move on to the texts which are closest to Handlyng Synne, in that they are predominantly, but not exclusively, uncontracted. As far as Genesis and Exodus is concerned, given that the manuscript is probably from West Norfolk and hence at the very edge of the north-eastern limits of contraction, it presumably shows the expected results. The Destruction of Troy, as Iyeiri (1992: 215) argues, is likely to come from an area at the north-western fringe of contraction and is in any case so late as to offer only minimal information. Alexander and Dindimus, on the other hand, is associated with Gloucestershire, generally recognized as a core area of contracted forms. This is a puzzle. However it is also the case that this is almost the latest of the texts used by Iyeiri, hence of the period when the construction was being lost. The relatively few examples are reasonably well balanced between contracted and uncontracted forms and overall it is most probable that what we are witnessing is the loss of the construction. But it is also worth recalling the discussion of the Old English material we encountered, in the shape of the Salisbury Psalter, a text which had predominating forms which were at odds with the surrounding area. The small group of texts which show mainly, but not exclusively, contracted forms may, to some extent, be rather like the case of The Destruction of Troy above. That is to say, they partly appear to have originated in fringe areas; that, at least, is possible in the cases of Sir Ferumbras (Devonshire) and the Stanzaic Morte Arthur. That is less likely, however, in the cases of the Confessio Amantis (Kent/Suffolk) and, even more seriously, Kyng Alisaunder, which generally has typical London and Essex characteristics. As Iyeiri (1992: 209) notes, this is exactly the area which forms the center of the eastern concentration of contracted forms. Iyeiri is unable to explain this situation fully, although her remarks on the relative failure of contraction with forms of have quite clearly point to a differential outcome for different verbs. It seems, therefore, that although we can observe that there is a longstanding dialectal pattern to negative contraction, the details of the patterning are much less clear. This would seem to imply that we can go no further with this particular investigation. But there is an alternative approach which, I believe, may be more appropriate. Let us start off with the suggestion that the predominant area for contraction was approximately the area south and west of Watling Street. That, it has to be said, is no longer tenable, for the simple reason that the LALME maps quite clearly demonstrate that areas such as Essex and perhaps Suffolk are also areas of

474 Richard Μ. Hogg contraction. Thus in the east of the country negative contraction extended rather further north than has been generally assumed. But if we look at the evidence presented by Levin (1958), what we find is that the situation in Old English (OE) was rather different, apparently, from that which obtained in Middle English (ME), at least before 1400. For in OE the heartland of negative contraction appears to have been Wessex. Given the early ME evidence about the AB dialect, then, as I suggested earlier, it seems reasonable to suppose the contraction quickly spread in a north-westerly direction. Yet, even with the caveats I have already expressed concerning Kentish, see in. 8, and Iyeiri (1992), this may not be quite the right picture. If we compare the LALME maps with the supposed distribution in OE, then it might be more appropriate to suppose that the original area of contraction was already slightly to the north-west of Wessex, in particular that it was most frequent and earliest in an area centred around Gloucestershire. The principal reason for saying this is that this is the area which shows the densest concentration of contraction in the ME period. A secondary reason is that this would help to explain why the AB dialect shows far more contraction than is seen in a text such as the Vespasian Psalter. The implication would be that contraction fanned out from the south-west midlands, and its quicker spread through Wessex and, indeed, much of the south-east, as opposed to its relatively slower spread in the west midlands proper could quite easily be explained by the dominance of the Mercian literary tradition which I have discussed elsewhere. At this point it is important to note that in OE the phenomenon of contraction is never categorical except in the Wessex heartland. And the most interesting feature of these tables, perhaps, is not so much the contrast between West Saxon and Anglian texts, nor the virtually categorical situation in West Saxon; rather it is surely that every Anglian text appears also to favor contracted forms. The difference is simply that in Anglian such forms are not categorical. At this distance in time any explanation of this can be little more than speculative. There could be any number of reasons for the variations. They might be due to stylistic preferences, although in this respect the range of textual material makes this unpromising; there might be phonological reasons, concerning in particular unstressed vowels, but the evidence is lacking. More useful are the types of syntactic context investigated in, for example, Blockley (1988, 1990), although note the reservations of Fulk (1992: 151). Perhaps the best that we can do is suggest that negative contraction was, when it first appeared, a random choice throughout the country. Then, as I have said, it became categorical in the south-west. At first this must have had no effect on other parts of the country, where random variation continued.

Spread of negative contraction


Then, however, it seems probable that almost everywhere else too random variation was given up in exchange for categorical structure. But the loss was not unidirectional. Thus in areas such as the west and northwest midlands negative contraction became the norm; but in the east and, especially, the north-east, negative contraction was entirely lost. At first sight this looks like a classical neogrammarian split. However, there is good reason for supposing that, even if the account I have given is the most likely interpretation of the data, that account in no way supports a neogrammarian split. Why do I say this? After all, we have seen that it is generally agreed that negative contraction is one of the few, if not the only, clear instance of a syntactic dialect feature in early English. But I don't want to dispute that. The issue is quite different. For the central issue is one which I have already highlighted as central to this work, namely the inadequacy of a neogrammarian typology. The crucial feature of split is the assumption that there is a clear-cut binary division and that such a division was not previously present. But in the present case this is demonstrably false. If, for example, we compare, say, The Owl and the Nightingale with the Ancrene Wisse, then it is true that both texts fall firmly on the same side of the binary divide. But if we step back in time and compare, for example, jElfric and the Vespasian Psalter, only the former consistently shows contraction. Yet the differences in location between the latter pair are not significantly different from the differences in the former pair. The explanation for this was given earlier, namely that contraction spread north-westwards over time. As a result the categorical divide which places geographical areas such as that to which Vespasian Psalter belongs into the group of dialects with negative contraction had simply not fully emerged at the time of that text. Such an explanation is commonplace and very widely accepted, but it is much less often noted that it is incompatible with the once-and-for-all neogrammarian theory of split. It is equally instructive to compare the data presented above for a cluster of relatively early East Midlands texts, namely Peterborough Chronicle, Ormulum, Havelok and Genesis and Exodus, the Alliterative Morte Arthure and the Stanzaic Morte Arthure. The following table, derived directly from those above, shows the percentage of contracted forms in each of these texts:

476 Richard Μ. Hogg

Table 7. Negative contraction in the East Midlands Text Peterborough Chronicle Ormulum Havelok Genesis & Exodus Allit. Morte Arthure Stanzaic Morte Arthure

Location Northants, Lines. Norfolk Norfolk Lines Rutland

% contracted 69% 66%

5% 10% 8%


Any attempt to discuss these texts in terms of a binary split seems destined to fail. For example, although these texts are displayed in roughly their chronological order, it is not obvious that chronology plays much of a role. The only possible exception to this is the last of them, the Stanzaic Morte Arthure. It could be argued that it shows decline in the contraction at a time when pre-verbal ne is in any case being lost. But otherwise the situation looks rather messy. For example, that both the Peterbrough Chronicle and Ormulum have a fair degree of contraction sits oddly with the virtual lack of contraction in the two Norfolk texts, especially when, in the case of Havelok, there remains the possibility that it too stems from Lincolnshire (see Iyeiri 2001: 11). Unfortunately, this is one case where the LALME evidence is not helpful, for LALME provides no maps of the relevant area of Lincolnshire, which is central to the issue (see again note 10). It may be true that at first sight it is the lack of contraction in the above Norfolk texts which is problematic. However, we can turn the question around and ask why is it that Lincolnshire has so many contracted forms, at least in the case of the Ormulum. I have no immediate answer to these questions, but there is a critical methodological issue here, which returns us to an earlier point. It may be recalled that although we accepted that such contraction was dominant in the West Saxon area in the Old English period, we also noted that one text, undoubtedly originating from Shaftesbury, namely the Salisbury Psalter, had very little contraction, in contrast to almost every other West Saxon text we examined. We also noted the possibility of some other texts in the same area having a significant minority of uncontracted forms, and that this leads to the possibility of local pockets of variation, which speak against the unbridled use of Stammbaumtheorie in the understanding of closelyknit dialect variation. It may, after all, be the case that both the Salisbury Psalter in Wiltshire and Ormulum in Lincolnshire merely show local variations which contrast with the normal forms of the region. It can be argued that this is no explanation of anything, but this is the opposite of the truth. It opens out the possibility of genuine progress. If we

Spread of negative contraction 477 look back at the data and take that together with the LALME maps, a new possibility can be envisaged. It is indeed the data provided by LALME which is crucial. As can be seen from the two maps I have presented, and also from the further maps drawn in the Atlas, by far the densest concentration of contracted forms is around Gloucester and, more generally, the lower valley of the Severn and its estuary. Is it not, therefore, plausible to suggest that this is a long-standing situation? In other words, we might suggest that this area, roughly speaking the area associated with the Hwicce, is where negative contraction first became categorical. To claim this is not entirely fanciful, although it is undoubtedly speculative. We can mount a variety of arguments in its favor. Firstly there is a kind of general argument which is derived from the uniformitarian hypothesis that the likelihood of any linguistic state of affairs has always been roughly the same as it is now (see Lass 1997). Here, of course, "now" must be taken as referring to, say, 1400, before the loss of negative contraction was complete. And all that I am suggesting is if in 1400 the greatest concentration of negative contraction was in Gloucestershire and the lower Severn, it is likely that this was the case at earlier periods too. Although this may be speculative, it is only by such speculation that we can hope to arrive at probable explanation of the rise and spread of negative contraction. The proof of the pudding must lie in its eating. In other words, what can we buy with such a suggestion? If we return to Old English, it is plain that at a very early stage negative contraction becomes categorical in most of the West Saxon area. Is there a plausible explanation for that? Why did contraction not remain as a variable feature there, as elsewhere? Perhaps the place to start is with the West Saxon literary tradition. And this is in essence a tenth-century innovation with Alfred and his teachers as slightly earlier precursors. From what we know of Alfred's teachers, in essence the comments in the preface to the Cura Pastoralis, they came from quite diverse origins, but above all from areas to the west and north-west of heartland Wessex. It is, therefore, quite plausible to suggest that forms from west and north-west could strongly influence the precursors of the literary tradition and then the establishment of that tradition itself. After this stage it is known that West Saxon forms became increasingly prevalent down the Thames valley and the London area. It is, therefore, eminently reasonable to see the extension of negative contraction to the London area as the result of a normal sociolinguistic event, simply another case arising from dialect contact. Why have I found it necessary to make explicit what others might have thought obvious? The categorical status in West Saxon is not at issue and could obviously be explained in more ways than one, but only a closer comparison of the Old English and the Middle English data points out a

478 Richard Μ. Hogg

serious issue which is glossed over by Levin. And that is a genuinely fundamental problem. It will be recalled that for Old English Levin contrasts West Saxon and Anglian. On the other hand, for Middle English he contrasts the South and the West Midlands with the East Midlands and the North. At first sight this seems straightforward, and wholly acceptable. But this is far from the case. For in the Old English situation the West Midlands is seen as one of the foci of Anglian, the area from which texts such as the Vespasian Psalter originates. But in the Middle English period such a location becomes, in Levin's terms, part of the south and west. This demonstrates the critical problem. For whereas Vespasian Psalter shows typical Anglian Old English forms, the later West Midlands texts all show categorical contraction. How can this be? Levin does not address this issue at all. He slips from Old English into Middle English without observing that his terminology changes too. But let us now suppose that it was indeed the case that the contraction first became categorical in the area of the Hwicce. If it then spread quite quickly to West Saxon and along the Thames valley, why did it not do the same up the Severn? The movement up the Severn seems certainly to have been later. I have already mentioned the West Saxon literary tradition, but it should not be forgotten that there also existed, and at an earlier date, a Mercian literary tradition, of which the Vespasian Psalter is clearly an example, see Vleeskruyer (1953). At the time when negative contraction became standard further south, I would suggest, the Mercian literary tradition would have been sufficiently strong to resist the process becoming categorical rather than optional. But the later spread of West Saxon features would have been less inhibited because of the twin features of the generalization of West Saxon over large areas of the country and the lessening role of the Mercian literary tradition. Admittedly there is every reason for supposing that the tradition survived well into the Middle English period, but at the same time such generalizing features as contraction would have been particularly prone to spread, especially given that they had been, previously, available as optional elements. So there is an explanation of why West Midlands texts of the earlier Old English period might have behaved differently from texts from much the same area in early Middle English, an argument which Levin fails to note. If the arguments I have presented in this chapter are worth pursuing further, then the data which have been collected for later dates in Iyeiri (1992, 2001) and LALME point to a gradual decline in the use of contraction - that is to say, in the fourteenth century contracted forms are becoming less frequent, for a variety of different reasons, including structural ones. In other words, negative contraction is recessive. And if, as LALME indicates, the principal area in which it is being lost first is the area

Spread of negative contraction


of the south-central midlands, this would be in itself noteworthy. The fact that, for example, the use of ne in Wycliffite texts (see here Jack 1978: 5 8 59) is always rare, being found in only about 5% of possible cases, is a structural change which has the immediate effect of causing a decline in contraction. Furthermore, Iyeiri (2001: 172-173) shows that in The Canterbury Tales there are significantly fewer cases of contraction in the prose tales than in the verse, which rather leads to the same conclusion as above, namely that the construction is in decline. Indeed, Iyeiri also shows that by this time the construction is becoming restricted to only examples involving the copular verb. Again a sign of recession.

Notes 1.

I am grateful to Meg Laing, Yoko Iyeiri and Johanna Wood for their comments on an early version to this paper, and doubtless if I had heeded their comments more this version would have been improved too. My thanks too go to two anonymous referees. 2. Theoretically we would expect the same condition to apply to initial /j/, but there are no relevant examples. 3. This may mean that the scribe was a woman. This is presumably irrelevant here, but given the rarity of female scribes it remains of interest. It is also noteworthy that the editors of the text refer to the scribe as 'he' throughout. 4. The brief notes for each text are taken from Iyeiri (1992) supplemented by LALME and Laing (1993). 5. For this text only Iyieri uses three manuscripts: Lambeth Palace 487 (North Herefordshire/Shropshire - see Laing 1993: 111); Cambridge, Trinity College 335, London; Bodley, Digby 4. This is very unfortunate, although it might not to be significant in the current rather narrow context, see further below. Meg Laing has kindly provided me with details of contraction in the various texts of the poem, which, regrettably would require more space to discuss here than is available. For further discussion of different versions of the poem, see Laing (1992). 6. The language of the manscript is difficult to place dialectally. The original poem may be associated with Surrey, but this version appears south-western, whilst the other version, in Oxford, Jesus 29, comes from near the Herefordshire/Gloucestershire border. For further remarks on this poem, see Laing (1998). 7. See Samuels and Smith (1981) for a convincing discussion of this text, its authorship and location.

480 Richard Μ. Hogg 8.

By then, of course, negative conti action was fast disappearing everywhere as the structure of verbal negation changed with the loss of the particle ne, see further below. 9. Note also Levin's comment (1958: 498 fn. 22) that Kentish agrees with the East Midlands. 10. There are good reasons why this should be, for generally the northern areas, as one would expect, rarely show contracted forms, a fact which is magnified by the generally later date of northern texts. No doubt a small amount of material is available there, but there are limits to what is feasible and practical. 11. An excellent discussion of the issues raised by isoglossing can be found in Francis (1983). 12. In theory, of course, it is possible to diminish the difficulty by a close examination of each suspect text. If the text shows clear examples of, in this case, uncontracted forms, this helps us to infer that the text is in an area which lacks the feature. However the task is large, and the problems I mentioned above in a text such as Gawayne slill remain.

References Bennett, Jack. A. W., and Geoffrey. V. Smithers 1968 Early Middle English Verse and Prose. 2nd ed. Oxford: Clarendon Press. Bethurum, Dorothy 1935 The connection between the Katherine Group and Old English prose. Journal of English and Germanic Philology 34: 553-564. Blockley, Mary 1988 Constraints on negative contraction with the finite verb and the syntax of Old English poetry. Studies in Philology 85: 428^450. 1990 Uncontracted negation as a clue to sentence structure in Old English verse. Journal of English and Germanic Philology 89: 475-490. Francis, W. Nelson 1983 Dialectology: An Introduction. Harlow, Essex: Longman. Fulk, Robert D. 1992 A History of Old English Meter. Philadelphia: University of Pennsylvania Press. Hill, David 1981 An Atlas of Anglo-Saxon England. Oxford: Basil Blackwell. Hogg, Richard M. 1997 Using the future to predict the past: Old English dialectology in the light of Middle English place-names. In Studies in Middle English Linguistics, J. Fisiak (ed.), 207-220. Berlin/New York: Mouton de Gruyter.

Spread of negative contraction 481 Iyeiri, Yoko 1992 Negative constructions in selected Middle English verse texts. Ph.D. Diss., University of St. Andrews. 2001 Negative Constructions in Middle English. Fukuoka: Kyushu University Press. Jack, George B. 1978 Negation in later Middle English prose. Archivum Linguisticum 9: 58-72. Kitson, Peter 1990 On Old English nouns of more than one gender. English Studies 71: 185-221. Kitson, Peter 1993 Geographical variation in Old English prepositions and the location of iElfric's and other literary dialects. English Studies 74: 1-50. Kitson, Peter 1997 When did Middle English begin? Later than you think! In Studies in Middle English Linguistics, J. Fisiak (ed.), 221-269. Berlin/New York: Mouton de Gruyter. Laing, Margaret 1992 A linguistic atlas of Early Middle English: the value of texts surviving in more than one version. In History of Englishes: New Methods and Interpretations in Historical Lingustics, Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.), 566581. Berlin/New York: Mouton de Gruyter. 1993 Catalogue of Sources for a Linguistic Atlas of Early Medieval English. Cambridge: D. S. Brewer. 1998 Raising a stink in The Owl and the Nightingale: A new reading at line 115. Notes & Queries 243: 276-284. Lass, Roger 1997 Historical Linguistics and Language Change. Cambridge: Cambridge University Press. Levin, Samuel R. 1958 Negative contraction: an Old and Middle English dialect criterion. Journal of English and Germanic Philology 57: 492-501. Mcintosh, Angus, Michael L. Samuels, and Michael Benskin 1986 A Linguistic Atlas of Late Medieval English. Aberdeen: Aberdeen University Press. Millett, Bella 1982 Hali Meidhad, (Early English Text Society, 284.) London: Oxford University Press.

482 Richard Μ. Hogg Roeder, Fritz 1904

Der altenglische Regius-Psalter, (Studien zur englischen Philologie, 18.) Halle. Samuels, Michael L., and Jeremy J. Smith 1981 The language of Gower. Neuphilologische Mitteilungen 82: 294-304. Sisam, Celia, and Kenneth Sisam 1959 The Salisbury Psalter, (Early English Text Society, o.s. 242.) London: Oxford University Press. Vleeskruyer, Rudolf 1953 The Life of St. Chad: an Old English homily. Amsterdam: NorthHolland.

Name index

Abbott, Evelyn, 75 Abrahams, H., 399 Abramson, Arthur, 383 Adams, Michael, 87, 111, 225 Addison, Joseph, 133, 136-140,144, 152-153, 161, 177, 202, 204 Akinlabi, Akinbiyi, 386 Alderete, John, 382 Algeo, John, 117 Allen, Cynthia, 343, 358, 360, 364 Alston, Robin Carfrae, 236 Ambler, Stephen, 383 Anderson, James, 31, 132 Anderson, Stephen R., 377 Amaud, Rene, 141,177 Ash, Sharon, 447,449, 452 Astell, Mary, 136-137, 140, 161, 177 Atkinson, Dwight, 197 Bailey, Charles-James, 415,428429, 442,445,446, 451 Bailey, Guy, 416,442,443, 445^46 Bailey, Richard W„ 36, 224 Banckes, Richard, 238, 239 Bao, Zhi-ming, 378 Baron, Dennis, 111 Barritt, Clyde W., 59, 70 Bassett, Marvin, 416 Bause, Josef, 89 Beckman, Jill N., 372, 382, 393, 399 Behaghel, Otto, 89 Behn, Aphra, 136-137 Benjamin, Walter, 31 Bennett, Jack A. W., 463 Bennett, William H , 376, 397, 398 Benskin, Michael, 23,467 Benson, Larry D., 323, 333

ßenua, Laura, 372 Berg, Thomas, 397, 398 Besnier, Niko, 198 Bethurum, Dorothy, 462 Biber, Douglas, 146, 176, 180, 189, 190, 197, 198,211,215 Billington, Sandra, 224 Blake, Norman, 305, 308, 310, 311 Bliss, Alan J., 300-301 Bliss, Phillip, 230 Blockley, Mary, 474 Blount, Thomas, 237 Böberg, Charles, 447,449, 452, 456 Boer, Richard C., 400 Boersma, Paul, 34, 343, 353, 358360, 399 Booij, G. E., 393 Bomstein, Diane, 226 Boswell, James, 7, 204, 205 Bradshaw, Mary M., 377 Brandreth, E. L., 103 Bresnan, Joan, 355, 358, 363-364 Bright, James, 57 Brinton, Laurel, 16 Britton, Derek, 224, 227 Brocket, John T., 102 Brandum-Nielsen, Johannes, 94 Brook, Nathaniel, 233 Brooks, Cleanth, 442 Brown, Vivian, 415-416, 428, 431, 437,445 Browne, Edward, 230 Browne, Mary, 240 Brunner, Karl, 31, 39 Bullokar, John, 231, 236 Bunt, G. Η. V., 39 Burzio, Luigi, 372 Butters, Ronald R., 111

484 Name index Cable, Thomas, 279, 297, 300, 301, 305,315 Calabrese, Andrea, 371, 376, 379, 388 Callery, B. G., 240 Callow, Mo, 383 Campbell, Alistair, 300 Catford, John C., 62, 65 Cawdrey, Robert, 229, 231, 236 Cerquiglini, Bernard, 36 Chafe, Wallace L„ 189 Chambers, Edmund Kerchever, 230 Chambers, J.K., 47, 51 Chambers, Raymond W„ 22, 301302 Chance, F., 100, 103, 104-105 Chase, Alston H., 75-76 Chirrey, Deborah, 47, 51 Churchill, Sarah, 136-138, 140, 161, 177 Cinque, Guglielmo, 345 Clankie, Shawn M , 111,115, 120 Clark, Brady Z., 358 Clark, Cecily, 8, 12 Clements, George N., 376 Coates, Jennifer, 189, 193 Coetzee, Anna E., 89 Cohen, Gerald L., 91 Collinge, Neville E., 399 Congreve, William, 136-140, 143, 152, 154, 161 Conolly, William, 153 Conrad, Susan, 189, 197 Cockeram, Henry, 231, 232, 236, 242 Coote, Edmund , 229, 236 Cotgrave, Randle, 101 Cotton, Robert, 219-220 Cowley, John D., 241 Crawford, S. J., 309 Crook, Eugene, 34 Crystal, David, 376 Cukor-Avila, Patricia, 442

Curzan, Anne, 224 Cusack, Bridget, 226 Dasher, Richard, 187 Daunt, Marjorie, 22, 59, 79 Davis, Norman, 217, 223 De Bont, A. P., 89 Defoe, Daniel, 131, 136-140, 143, 154-155, 161, 164, 170-171, 177, 185-186 Dekkers, Joost, 353 DeMarse, Sharon, 426 Denison, David, 143 Desmet, Piet, 36 Dickens, Charles, 99 Dingare, Shipra, 358, 363-364 Dobbie, Elliott Van Kirk, 292, 300 Dobson, Eric, 26-27, 375, 393 Dogil, Grzegorz, 374, 387 DrydenJohn, 132-134, 136-140, 143, 158 Dryer, Matthew S., 354 Duanmu, San, 378 Duggan, HoytN., 20, 279, 280, 300 Dyer, Gustavus, 416 Eble, Connie, 436-437,439 Elliott, Colleen M., 416 Ellis, Alexander J., 37, 223 Erickson, Donna, 380 Fabb, Nigel, 326 Fee, Margery, 16 Finegan, Edward, 146, 176, 180, 189, 197, 198 Fischer, Olga, 347 Fisher, John H., 226 Fisiak, Jacek, 9 Fitzmaurice, Susan M., 135, 137, 139, 162, 175-180, 184, 189, 198 Fleischman, S., 12-13, 36-37 Flemming, Edward, 372, 379, 399

Name index 485 Forby, Robert, 100-103 Fox, Anthony, 383 Frank, Roberta, 10-11 Franklin, Benjamin, 429 Fujimara, Osama, 380, 383 Fulk, Robert D„ 38, 313,464,474 Fuller, Janet M., 426 Fuss, Eric, 343, 355 Gandour, Jack, 373 Garrett, Andrew, 374, 379, 380-381, 383,391,395 Geritz, Albert J., 269 Gill, Alexander, 28 Gimson, A.C., 14, 28,51 Giner, M. F. Gracia-Bermejo, 15 Glass, Gene V., 322 Goodfellow, Robin, 102 Golston, Chris, 338 Gordon, Ε. V., 300, 307 Görlach, Manfred, 226 Graham, Howard Jay, 241, 269 Grandgent, C. H., 15 Greene, David, 73-74 Griffith, R. H., 90 Grimshaw, Jane, 349 Grose, Francis, 93 Grundy, Isobel, 139 Gunter, Edward, 230 Haggard, Mark, 383 Hale, Mark, 374, 379, 380-381, 383, 391,395 Halle, Morris, 315-316, 334, 371, 374, 376, 377, 379, 388 Halliwell, James O., 102 Hankins, Gail Α., 416 Hanna, Ralph, III, 311 Hanson, Kristin, 338 Harmon, William, 117 Harms, Robert, 67 Harris, David Payne, 25, 39 Harris, John, 50-51

Harvie, Dawn, 48 Haudricourt, Andre-Georges, 377 Hawkins, John Α., 354 Hayes, Bruce, 279, 316, 326, 327, 329, 338, 343, 358-360 Haywood, Eliza, 136, 138, 151, 152 Hellquist, Elof, 97 Hickey, Raymond, 40 Hill, Joyce, 306 Hirose, Hajime, 373 Hitchcock, Elsie V., 301 Hogan, Patrick Colm, 314 Hogg, Richard M„ 9, 37, 67, 392, 399, 462 Holman, C. Hugh, 117 Holoyoke, Francis, 235, 236 Holsinger, David, 371, 376, 379, 395 Holt, Eric, 44 Hopkins, B. R., 322 Hopkins, Kenneth D., 322 Hopper, Paul, 364 Huddleston, Rodney, 132 Hughes, John P., 103 Hume, Elizabeth V., 376 Hunston, Susan, 189 Hyman, Larry, 377, 378 Ito, Junko, 383, 384 Iverson, Gregory, 371, 376, 379, 395 Iyeiri, Yoko, 459, 463, 466-^69, 471-474,478,479 Jack, George B., 479 Jamieson, John, 100, 101, 104 Jesperson, Otto, 14, 26, 28, 51, 393, 398,417,418 Jessen, Michael, 399 Johnson, Samuel, 102, 205 Jones, Richard Foster, 233, 269 Jordan, Richard, 34 Kaun, Abigail, 316, 338

486 Name index Kautzsch, Alexander, 426 Keene, Derek, 49 Kenstowicz, Michael, 376 Ker, N. R., 306 Kerswill, Paul, 48 Keyser, Samuel Jay, 315-316, 334, 376 Kingston, John, 377 Kiparsky, Paul, 74, 75, 77, 82, 299, 334, 338, 345, 355, 363, 374 Kirchner, Robert, 372, 398 Kitson, Peter, 464,466 Klaeber, Frederick, 300 Kluge, Friedrich, 89, 90, 96 Koerner, Κ., 8-9, 36 Koziol, Herbert, 87 Krapp, George Philip, 295, 300,419, 420 Kretzschmar, William Α., Jr., 416 Kristensson, Gillis, 38 Kroch, Anthony S., 343, 346-348, 354, 357, 364 Kroeger, Paul, 355 Krug, Manfred G., 189, 190, 207 Kurath, Hans, 15, 447 Kurylowicz, Jerzy, 299, 302 Kytö, Meija, 189 Labov, William, 49-51, 177, 189, 447, 449, 452 Ladefoged, Peter, 62-65 Ladusaw, William, 37 Laine, Arnos Lee, 269 Laing, Margaret, 479 Laker, Stephen, 34, 38 Lancashire, Ian, 238 Lance, Donald, 447 Landau, Sidney, 111, 114-119, 122 Lapidge, Michael, 310 Lass, Roger, 9, 26-27, 49, 50, 61, 70-73, 373, 374, 392, 393, 399, 477 Laver, John, 62, 64-65, 373, 376

Leas, Susan, 416 Leech, Geoffrey, 189, 193, 199 Lerchner, Gotthard, 399 Lerer, Seth, 41 Levin, Samuel R., 459-460, 462463,465-467,469,471,474 Li, Xingzhong, 315, 317, 324, 326, 330,333-334,338 Liberman, Anatoly, 104 Liberman, Mark, 315, 386 Lightfoot, David, 343 Lipp, Frances Randall, 309 Lisker, Leigh, 383 Lombard, Daan, 63 Lombardi, Linda, 377 Luick, Karl, 25, 34, 39, 309 Lutz, Angelika, 31-32 MacEachem, Margaret, 326 MacEoin, Gearid, 66 Machyn, Henry, 26, 217-228 Mackay, Charles, 91,98 MacMahon, Michael, 27, 37,40 Madden, Frederick, 220 Maddieson, Ian, 37, 62-65, 373, 377, 378, 379, 395, 396 Mallinson, Graham, 69 Mandel, Jerome, 104 Manley, Delavariviere, 136-137, 161 Manning, Christopher D., 358, 363364 Mansfield, Ε. D., 75 Marchand, Hans, 87 Matthews, William, 26 Maynor, Natalie, 442 McCarthy, J. Thomas, 121 McCarthy, John, 372, 376, 384 McCone, Kim, 73, 74 McDavid, Raven I., Jr., 15, 416, 447 Mcintosh, Angus, 23,467 McMillan, James Β., 86, 435, 442 McSparran, Frances, 38

Name index 487 Mester, Armin, 383, 384 Millett, Bella, 308,462 Milroy James, 33,48, 50-51,162 Milroy, Lesley, 50, 55,162 Minkova, Donka, 12, 32, 37,40,47, 48-51,57, 305,315 Mitchell, Bruce, 8-9 Mody, Maria, 383 Montagu, Charles, 132 Montagu, Mary, 131-132, 136-140, 144, 153, 157, 204 Montgomery, Marshall, 16, 40 Montgomery, Michael, 15, 423,426, 435,436-437, 439,442 Moore, Brian, 373 Moore, John Trotwood, 416 Morimoto, Yukiko, 351 Mortimer, Ian, 218-219, 224, 225 Mosse, F., 227 Moxley, Louise Armstrong, 416 Mugglestone, Lynda, 48 Mulcaster, Richard, 237 Murfin, Ross, 323 Murray, James A. H., 92-93 Myhill, John, 189, 203 Nathan, Geoffrey S., 8, 11, 36 Nespor, Marina, 327 Nevalainen, Terttu, 49 Nichols, John Gough, 220, 221, 222, 225, 226 Nooteboom, Sieb, 380 Nowell, Laurence, 230 Noyer, Rolph, 371, 375, 376, 377, 379, 388, 391, 395 Oakden, J. P , 307, 308 Ochs, Elinor, 189 Offord, Μ. Y., 38 Ogilvie, John, 102 Ohala, John J., 373, 376 Onions C. T., 93 Orchard, Andy, 36, 37

Osselton, Ν. Ε., 231,232, 233 Padgett, Jaye, 372, 383, 384 Page, Richard, 371, 373, 376, 379, 395 Palmer, F. R., 193 Palsgrave, John, 237 Pederson, Lee Α., 416 Petrova, Olga, 371, 372, 381, 382, 384, 387, 388, 393, 397, 399, 401 Pharr, Clyde, 76 Phillips, Betty, 452 Phillips, Henry, Jr., 75-76 Pintzuk, Susan, 343, 345-346, 354, 357-358, 360, 363-364 Plapp, Rosemary, 382 Ponelis, Fritz, 70 Pope, Alexander, 134, 136-138, 140, 152-154, 157-158, 166, 185-186, 206 Pound, Thomas, 230, 231, 235 Pratt, Lynda, 143 Precht, Kristen, 189 Prince, Alan, 315, 338, 372, 382, 400 Prior, Mathew, 136-137, 151, 161 Procosch, E., 375 Pullum, Geoffrey, 37, 132 Putter, Ad, 299, 300 Pyles, Thomas, 117 Pynson, Richard, 229, 237 Quirk, Randolph, 181 Radcliffe, Frances, 231 Rastell, John, 242, 243, 269 Rastorguyeva, Tatiana, 374 Ray, John, 98, 100-101 Ray, Supryia M., 323 Reppen, Randi, 189, 197 Rice, Keren, 377 Richardson, Charles, 102 Rider, John, 231

488 Name index Ringen, Catherine, 382 Rissanen, Matti, 177 Roberts, Martin, 383 Robinson, Orrin W., 294 Rooth, Eric, 373 Rosenberg, Bruce Α., 314 Rubach, Jerzy, 393 Ruprecht, Knecht, 102 Russom, Geoffrey, 281-284, 295, 300, 301, 305, 308 Ryden, Mats, 134, 140, 150, 175 Saenger, Paul, 36 Salmon, Vivian, 269 Salmons, Joseph, 371, 376, 379, 395 Salter, Elizabeth, 305 Samuels, M. L„ 23,467, 479 Sand, Lori, 443 Schaefer, Ursula, 36 Schäfer, Jürgen, 238, 240 Schneider, Edgar, 423 Schröder, Heinrich, 88, 89, 96 Scott Charles P. G., 92 Seebold, Elmar, 90, 96 Selkirk, Elisabeth, 327 Sells, Peter, 349, 351, 355 Sen, Ann Louise F., 419,429 Shuy, Roger, 115 Sievers, Eduard, 25, 31, 39, 281, 287,300 Silverman, Daniel, 379, 398 Sisam, Celia, 465-466 Sisam, Kenneth, 465-466 Skeat, Walter W., 94, 100-103 Skinner, Stephen, 102 Smith, Jeremy J., 479 Smithers, Geoffrey V., 463 Smitterberg, Erik, 134, 141, 145, 176, 177, 179, 183, 187 Smolensky, Paul, 338, 382,400 Smyth, Herber W., 76 Smythe Palmer, Α., 99 Solnit, D., 377

Spitzer, Leo, 89, 97, 103 Stanley, Eric G., 69 Steele Richard, 131-134, 136-140, 151-152,157, 161, 164-165, 185,202 Stephenson, Edward Α., 416, 422, 432 Stepney, George, 132-133, 136— 137, 161 Steriade, Donca, 34, 376 Stevens, Kenneth, 371, 376, 377 Stockwell, Robert P., 59, 70, 74 Stokes, Myra, 299, 300 Strang, Barbara, 141 Strassel, Stephanie M., 456 Strype, John, 218, 220, 222, 225, 226 Stuart-Smith, Jane, 47, 51 Summerfield, Quentin, 383 Suzuki, Seiichi, 31, 301 Swift, Jonathan 14, 131-132, 136138, 144, 161, 164, 166-167, 202 Szentgyorgyi, Szilard, 382 Tarlinskaja, Marina, 315, 324 Taylor, Ann, 343, 346-348, 354, 357 Thomas, Erik R., 437,441, 442, 443, 447, 453, 456 Thompson, Geoffrey, 189 Thome, David, 70 Thumeysen, Rudolf, 67 Tieken-Boon van Ostade, Ingrid, 162 Tillery, Jan, 443 Todd, H. J., 102 Tolkien, J. R. R„ 300, 307 Tottie, Gunnel, 48 Traugott, Elizabeth C., 187, 364 Trips, Carola, 343, 345, 355 Trudgill, Peter, 15, 48

Name index 489 Turville-Petre, Thorlac, 20, 281, 298, 301, 302 Tusser, Thomas, 102 van Bergen, Linda, 343 van de Weijer, Jeroen, 353 van Gelderen, Elly, 345, 363 van Kemenade, Ans, 343 Vennemann, Theo, 31 Vemer, Karl, 371 Vikner, Sten, 349 Vleeskruyer, Rudolf, 462-463, 478 Vogel, Irene, 327 Walker John, 30, 419 Webster, Noah, 98 Weekley, Ernest, 95 Wellisch, Η. Η., 240 Wells, J. C., 14-15 Werner, H„ 9-10 Wesley, Charles, 140 Wesley, John, 137 Wesley, Susannah, 136-137, 140, 161, 177 Whalen, D. H„ 383 White, David L., 60 Wierzbicka, Anna, 119 Wijk, Axel, 220, 222-224,225

Wikle, Tom, 443 Wilbraham, Roger, 101 Wilcox, Jonathan, 309 Williams, Ann, 48 Wilson, R. M., 223, 224 Winters, Margaret E., 8, 11, 36 Wise, C. M , 419 Wolfram, Walt, 443 Wortley, Edward, 136-138, 157, 161 Wright, Joseph, 37, 91, 93-94, 100, 223,417 Wright, Laura, 217 Wright, Susan, 134, 141-142, 150, 179, 183 Wriothesley, Henry, 230 Wyld, H. C., 26, 86, 99, 218, 222223, 226,417,418, 420 Yang, Charles D., 364 Yip, Moira, 376, 378, 379, 383, 386, 395-396 Youmans, Gilbert, 315, 317, 326, 328,329, 333-334, 338 Ziegeler, Debra, 140, 150 Zwart, C. Jan Wouter, 345, 363

Subject index

ablaut, 88 accent four-mora accent, 76-78 three-mora accent, 77-78 recessive accent, 76-77 activity verb, 136, 150 adverb durative adverb, 133 intensifying adverb, 165 monosyllabic adverb, 346 temporal adverb, 133-134, 155, 157,167, 186 jElfric, 292, 297, 301, 307-311, 313-314,460, 475 African American English, 48,415, 431-432, 433,435, 436-438, 442 African languages, 65 Afrikaans, 70-73, 78, 80, 89 allegro form, 92 alliteration, 13-14, 32, 103 alliterative binding rule, 282 alliterative data, 35 alliterative evidence, 27,47,49 alliterative gibberish, 103 alliterative prose, 310 alliterative verse, 12-13, 279302, 305-311,313-314 Middle English alliteration, 1822 Old English alliteration, 16-17 rhythmical alliteration, 308-310 alliterative meter alliterative revival, 305-310 development of Middle English alliterative meter, 279-302, 305-311,313-314

Middle English a-verse, 279, 296, 300 Middle English b-verse, 279, 286, 292-293, 295-297 Old English metrical constraints, 284 alphabetic heading, 232-234 American English, 17, 86-87, 115, 189, 192-193, 199, 201,203, 209, 415-433, 4 3 5 ^ 4 3 , 445-446,447-456, see also Southern American English anacrusis, 288-289, 291-292, 294296 anapest, 333 Ancient Greek, 70, 74-78, 87, 91 Anglian dialect, 460 apocope, 94 apodosis, 164 ARCHER corpus, 190, 197-199, 203, 207-209 aspectual force, 143 aspectual meaning, 142 aspiration, 15-16 attrition, 47, 51 Avestan, 299 Battle of Brunanburh, 292 Battle of Maldon, 293-296, 301 Beowulf, 220, 279-281, 283-289, 292, 294-296, 300 Bibliotheca Scholastica, 231, 235 bisyllabic anacrusis, 288-291, 293 blend, 91 borrowing, 25, 29, 88, 91 British English, 35, 136, 141, 180, 189, 192-193, 199, 201, 209-210, 373

492 Subject index broadening, 112 caesura, 330-331 Caligula, 18-19, 308 Camden Society, 220 Canadian English, 16, 51 Catholic Homilies, 306, 309 caught/cot merger, 447-456 auditory analysis, 451,453 Celtic, 91 Chaucer, 10, 22, 94, 315, 316, 319, 328, 332 Canterbury Tales, 323, 479 Chaucerian verse, 315-338 House of Fame, 323 Chinese, 383 circumflex, 76-77 Civil War, 415—417,420, 428-430, 432,438 clause structure in early English, 343-370 all-final construction, 343-344, 346-347, 352, 355, 357, 360362 all-medial construction, 343-346, 348, 355-356, 360-362 brace construction, 343-346, 348, 352, 356, 360-362 object-verb order, 348, 364 reverse brace construction, 343344, 353-354, 357, 363 verb-object order, 364 clitic clitic group, 327 clitic host, 328-329, 335 clitic phrase, 316, 335 enclitics, 300 proclitic, 283, 288 proclitic phrase, 328, 330 cliticization, 328 closure principle, 318 coarticulation, 33, 65 Cockney, 223

coinage, 95-96, 98, 100 collocation, 13, 35 communication verb, 150-159, 168 compound stress rule, 282 compound, 105, 297 trisyllabic compound, 281 compounding decline of, 292-294 CONCE corpus, 141, 180 connotation, 104 consonant cluster, 47, 61 cluster development, 21 cluster reduction, 19-20, 31 cluster simplification, 32-34 constraint alignment constraint, 353 categorical constraint, 338 constraint dominance, 359 constraint ranking, 349, 352, 358-359, 364, 383-394, 398 constraint set, 363 constraint violation, 352 Head-L constraint, 349-358, 361-363 *I constraint, 351-354, 357-358, 361-363 markedness constraint, 349, 351, 355 monosyllabic word constraint, 334-338 on unstressed syllable, 283 relative ranking of constraints, 351 *S constraint, 351, 354, 357358,361-363 Spine-R constraint, 349-358, 361-363 contextualization, 12 copyist, 13 Cynewulf, 305, 313 Danish, 93, 373, 377 DARE, 91, 119

Subject index Daunt Hypothesis, 59 declension, 75, 77 deictic proximal deictic, 157 temporal deictic, 156 denotative sense, 184 diachrony, 57, 451 diachronic, 10, 68, 111, 118, 141, 175, 189 diachronic change, 120, 455 diacope, 87 dialect contact, 440 dialectal variable, 469 dialectology, 89,461 dictionary, see also lexicon bilingual dictionary, 229 Latin-English dictionary, 236 differentiae, 238, 240 dip verse-initial dip, 293-295 verse-internal dip, 294 diphthong bimoraic diphthong, 60, 67, 7273 long diphthong, 70 monomoraic diphthong, 60, 72 overlong diphthong, 70 short diphthong, 58-81 trimoraic diphthong, 70 directionality, 31, 35 dominance relation, 317 Don Juan, 91 i/o-support, 364 double articulation, 65 doublet, 94 downtoner, 156 durativity, 132, 140 Dutch, 93, 96 Early Modern English, 26-31,4951, 93, 229-230, 236-240, 418 Eastern Tamang, 396


egress, 62-64 elision, 332 empiricism, 11, 57 enclitic, see clitic English linguistic nationalism, 237 English Schoole Maister, 229, 231 English word formation, 92 Estonian, 68 etymology, 12, 85, 87-91, 93-94, 96-103, 238-239, 242 euphemism, 87 evaluation point, 358-359, 361 Exeter Book, 305 experiential force, 186 extended form, 88-105 extrametrical syllable, 285-288, 332,338 extrametrical word, 281-282, 289291, 295, 297 figurative meaning, 134 figurative usage, 117 finite verb, 345, 346-348, 357 FLOB corpus, 141, 211 folk etymology, 102 foot, see metrical foot foot saliency principle, 326 ranked foot saliency principle, 335 Freedman's Bureau, 426,431,436, 445 French, 92-93, 95-97, 101, 105, 241,243,298, 393 fringe area, 473 function word, 237, 283-284, 288, 293-295, 297 functional head, 351 Gaelic, 98-99 Gawain, see Sir Gawain generative linguistics, 9

494 Subject index generative metrics generative metrical principle, 320 generative metrical rule, 334 generative metrist, 316 generative phonology, 299 generative phonologist, 75 genericide, 112, 119-120,121 genericness, 111, 115-120 generic reference, 111 generic words, 111, 112, 114 genre, 146, 176, 180, 242 genus, 238, 250 German, 88-90, 92, 96, 99 German word formation, 89 Germanic, 21, 68, 87, 94, 97, 374, 395 Germanic languages, 87, 393, 399 Germanic prosody, 307 Proto-Germanic, 371-376, 380, 384, 388, 392-394, 398,400 Germanic Stress Rule, 374, 375, 385, 392 Glorious Revolution, 139 glossary head-word glossary, 242 monolingual glossary, 240 Gothic, 87, 375, 391-392,400 Government and Binding, 85 Gradual Learning Algorithm, 360362, 364-366 Great Fire, 218 Great Vowel Shift, 218 Greek, 82, 298 Greek metrics, 78 Grimm's Law, 371, 372, 381, 386, 387, 395 harmonic bounding, 352-355 Hausa, 387 A-dropping, 33, 48 Head Law, 31 headlessness principle, 331

headword, 229-230, 233-235, 237, 240, 242 Heliand, 294-295 Helsinki Corpus, 222 hemistich, 316-318, 321, 325-326, 332 headless hemistich, 331 hemistich boundary, 330, 338 historical linguistics, 9, 49, 86, 112, 184,310, 455 historical phonetics, 85 historical pragmatics, 12, 183 historical sociolinguistics, 48-49, 51 Homeric verse, 75, 299 homonym, 448 homophone, 26-27, 72, 448, 451 homorganic, 62-63, 64, 65, 79-80 hyperbole, 133, 166 iamb iambic foot boundary, 330 iambic pentameter, 317, 328, 334 iambic tetrameter 315-338 iambic word, 316, 328, 330 ICEHL, 8-9 Icelandic, 68-70 ICHL, 9 iconicity, 89 Igbo, 377, 378 incorporating languages, 87 Indo-European, 86 Proto-Indo-European, 371, 373, 374-376, 384, 385, 395, 397 401 Indo-Iranian meter, 298 infixation, 86, 88 inflection, 238 inflectional system decay of, 292, 306 ingress, 62-63 homorganic ingress, 64 intonational phrase, 327

Subject index 495 inverse spelling, 421,424-428 Irish, 91 isogloss, 462, 464,470-471 Italian, 97, 99 Japanese, 95 Jeh, 377 Jingpho, 377, 378, 396 Kentish, 360 Kit Cat Club, 138 LAGS, 416,431,432,438 LALME, 23, 224, 467, 469-470, 472-474,476,478,479 LAMSAS, 416, 431, 432 language acquisition, 364 language contact, 48 Lanham Act, 113 Latin, 87, 232, 240, 309,466 Latvian, 377 Lasamon, 17-19, 20-22, 39, 298, 307, 308 LEME, 269 lexical encyclopedia, 237-238 lexical innovation, 119 lexico-grammatical context, 135, 142, 164 lexicographer, 111, 115, 120, 236, 238, 243 lexicography, 85, 112, 115, 118, 229 English lexicography, 229 historical lexicography, 230 lexicology, 111 lexicon bilingual lexicon, 229, 237, 242 encyclopedic lexicon, 230 monolingual English lexicon, 229-243 monolingual lexicon, 230, 237 reversed lexicon, 232, 239 lexicosemantic change, 112 LGSWE corpus, 100

linguist, 9, 11,59, 111 linguistic archeology, 8 linguistic innovation, 445 Linguistic Society of America, 9 linguistics, 9-11, 13, 16, 35, 47, 57, 81,86,310 literacizer, 59 literacy, 29-30, 35 Lithuanian, 387 Lives of Saints, 309 LOB corpus, 141, 209 London English, 223 long dip, 279, 281, 284, 289, 291, 295, 297 long dip requirement, 287 long dip restriction, 279, 285 Low German, 93-94, 96,105 LSWE corpus, 190 main clause, 146-150, 152, 163, 165, 168, 178-179, 183 Makaa, 377 markedness, 31 medieval literary culture, 7 mental verb, 150-159, 163, 168, 179, 183 merger, 17-18, 22-23, 25, 27-28, 31,35,47-48, 50, 55,415432,437, 455-446, 449, 451-452, 456, see also vowel merger meta-analysis, 92 metaphor, 113, 117 metathesis, 97, 101 metathetic alternation, 67 metonymy, 117 metrical boundary, 316, 329-330, 338 metrical foot, 280, 283, 284-287, 288-290, 295,315-316, 319-320, 332, 338, see also alliterative meter, iamb, trochee

496 Subject index foot boundary, 319 foot pattern, 281 heavy compound foot, 283 light foot, 283-286, 288, 291, 295 weak foot, 317 metrical quantity, 74 metrical saliency principles, 316— 322 metrical tree, 316, 327 metrics, 75, 318 Middle Ages, 12, 306, 310 Middle Dutch, 98-99 Middle English, 20-26, 32, 34, 104, 279, 285-286, 292, 295-296, 297-299, 305, 307,315,343, 345,357, 371,397, 398,417, 459,460, 462-468, 474, 477478 Early Middle English, 17-19, 307, 347-348, 354, 360-361 Middle English syntax, 459, see also clause structure Middle High German, 89 minimal pairs, 71, 75,451, 453 modal, 142, 165, 189-197, 199-211 logical meaning of modal, 193196 modal force, 134 personal meaning of modal, 193196 modality, 189, 210 Modern Dutch, 99 Modern English, 14, 27, 32, 60, 99, 104, 112, 176, 178, 238,355, 455 monophthong, 60, 69, 442 mora, 60-62, 69 non-moraic, 63, 69 phonemic mora, 79 morpheme boundary, 72, 79 morphology, 85-86, 88 morpho-syntax, 48, 49

movement transformation, 328, 335 native speaker, 7, 12-13, 58, 119, 121 NEET corpus, 134-140, 145, 155, 159-160,163, 176-177,180, 184-185,187 negative contraction in Early English, 459-479 Neogrammarian split, 475 neologism, 97 network effect, 163 network tie 160 neutralization, 72,420 New York English, 51, 55, 72 non-finite clause, 144 non-finite verb, 343, 346-347, 357 non-projective category, 355 non-volitional verb, 133 Northern Welsh, 70-73, 78, 80 off-verse, 308 Old English meter 279-284, 285287, 299, see also alliterative meter Old English, 14, 16-17, 20, 27-28, 32, 34-35, 47, 58-59, 64-67, 70-71,78-81,92, 97, 104, 175,280-283,286, 291-292, 294-295, 298, 305-309, 343, 345, 348, 352, 357, 360-361, 375,417,460, 462,465,467, 473-474, 477—478 Old English clause, 343, see also clause structure Old English syntax, 8-9, 459 Old French, 417 Old Frisian, 459 Old High German, 376 Old Icelandic, 69, 87, 375, 397 Old Irish, 66-67, 73-74, 79, 81 Old Norse, 417 Old Saxon, 294-295, 297, 375, 392

Subject index Old Saxon meter, 294-295, 297 onomastics, 8 onomatopoeia, 29 onset, 31-33, 61 on-verse, 308 Optimality Theory, 35, 85, 321, 338, 349,358, 372-373,379,381 Stochastic Optimality Theory, 343, 349, 358, 360, 362-364 orality, 7, 12-13 orthoepist, 26-27, 51,417 orthoepistic evidence, 26, 28, 50-51 Otho, 17, 19 oxytone, 75, 77 palatalization, 65-67, 68 palatalized consonants, 58, 6667, 73 pen/pin merger, 415-432, 435-438, 440-442 acoustic analysis, 437—438,441442 perception experiment, 453-454 periphrastic auxiliary, 329, 331 Peterborough Chronicle, 8,460461,476 philologist, 7-9, 57, 111 philology, 7-13, 14, 35, 47, 57-58, 81, 85-86, 111 English philology, 81 linguistic philology, 9, 33 new philology, 40-41 phonemic contrast, 33-34, 73 phonemic transcription, 58 phonemicization, 67 phonetic economy, 95 phonetic spelling, 218,421,424-428 phonetic transcription, 57 phonetic variable, 448 phonetics, 62-63, 85 Phonological Atlas of North America, 15, 447,449, 452 phonological boundary, 328, 335


phonological change, 51 phonological distinction, 51 phonological pause, 322, 329 phonological phrase, 327 phonological reconstruction, 7 phonological variation, 47 phonological word, 327 phonology, 13, 31,48, 62-63, 85, 318 phrasal stress, 284 phrase boundary, 299, 330, 335 Piers Plowman, 21-22, 38-39, 9799 pluperfect, 144 post-lemmatic segment, 239-240 pragmatic competence, 141, 175 pragmatic strengthener, 133, 179 pragmatics, 111 prefix, 87-88, 284,419 bisyllabic prefix, 288, 291 monosyllabic prefix, 288, 295 unstressed prefix, 283 prenasal environment, 419, 429430,437 stressed prenasal environment, 420-428 stressed non-prenasal environment, 420-428 unstressed prenasal environment, 420-428 preverb, 288 principle of closure, 279, 292, 295, 297 principle of syntactic integrity, 284 proclitic, see clitic progressive, 131-132 as foregrounding device, 141 as framing device, 11 aspectual progressive, 140, 141, 163-168, 175, 183, 186 bare progressive, 167 collocational context of progressive, 135

498 Subject index experiential progressive, 135, 141, 142, 163-168, 175, 176, 179, 183-184, 187 expressive progressive, 134, 175 grammatical context of progressive, 146-150 lexical context, 150-159 present progressive, 132 subjective progressive, 140 Proto-Germanic, see Germanic Proto-Indo-European, see IndoEuropean proto theory of clause structure, 349 protracted drop, 308 quasi modal, 435 radical-break hypothesis, 297-298 ranking distance, 360 ranking value, 358, 361 Rawlinson Lexicon, 229-243 Received Pronunciation, (RP), 14, 37 referential definition, 238 reflex, 7, 20, 105 register, 14, 134-136, 139, 144-145, 155-156, 159-161, 178, 183, 185, 189-190, 193,198-200, 210 expository register, 197 high register, 50 register variation, 197-198 speech-based register, 197 written register, 191, 194, 197, 209 Restoration, 231 rhetoric rhetorical effect, 168 rhetorical strategy, 165 rhetorical structure, 166 rhythm, 13 Riang, 378 Romance Stress Rule, 371, 375, 392393 Romance suffixes, 88

Russian, 82 S-curve, 364 saliency in hemistich, 320 in metrical boundary, 320-321 in metrical foot, 319 in metrical position, 317-318 saliency principle, 324, 337 Scandinavian languages, 97 Scandinavian, 91, 345 Scots, 28, 71-73, 78-80, 93, 101 Scottish English, 14 secondary articulation, 65-68, 7980 secondarily articulated consonants, 73 secondary meaning, 114 semanticization, 183, 185 semantics, 85, 105, 111 semi-modals, 189-197, 199-211 SePedi, 63-64, 79 Sino-Tibetan, 377 Sir Gawain and the Green Knight, 20,21,279-281,292, 299, 306, 309 Slavic, 393 social variation, 85 sociolinguist, 12, 47 sociolinguistics, 47,49-50 sociolinguistic variants, 49-50 sociolinguistic variation, 51 variationist sociolinguistics, 48 sonority, 13, 31-32 sonority hierarchy, 383 sound symbolism, 86 Southern American English, 415, 435-436, 438,442, 445, see also pen/pin merger Southern Historical Collection, 423 Southern Shift, 435,441 Spanish, 89, 93, 98-99, 101 split, 28

Subject index 499 spondee, 336 stance adverbial, 200,211 standardization, 217 stative verb, 133-134, 142 Stochastic Optimality Theory, see Optimality Theory stress maximum principle, 334-338 stress profile, 324, 330 structuralism, 9, 85 stylistics, 85 sub-entry, 233 subordinate clause, 146-150, 165, 178, 345 substitution, 34 substratum, 21 suffix, 93-94, 101,419 Suma, 377 Swedish, 91, 98 Swiss, 99 syllable closed syllable, 69 English syllable structure, 299 open syllable, 69 stressed closed syllable, 68 stressed open syllable, 68 stressed syllable, 68, 286, 288, 324, 334,415,419, 428-429 syllable boundary, 60 syllable with primary stress, 280 syllable with subordinate stress, 280 unstressed syllable, 280, 281, 284, 286, 297, 306, 332, 334, 419, 429 synchrony, 57 synchronic variation, 455 synchronic, 68, 111, 118, 178, 189-190, 451 syncope, 5 synecdoche, 113, 117, 118, 121 synonym, 93 syntactic inversion, 333-335, 338 syntax, 8, 85, 87-88, 111

syntactic change, 358, 364, 469, see also clause structure syntactic constituent, 334 syntactic phrase, 327 Table Alphabeticall, 229-230 Tankhur Naga, 377 taxonomic approach, 9 TELSUR, 37,449 Z/iaZ-complement clause, 153 time depth, 51 tmesis, 87 tonogenesis, 372, 377, 383 trademark, 111-121 trochee, 325, 336 trochaic, 281 trochaic constituent, 279, 297 trochaic inversion, 325-326, 337 trope, 117 uniformitarian principle, 51,477 unitary phoneme, 61-63, 65, 80 un-merger, 27-30, 50, 55 unstressed word, 282, 292 variation, 47 variationist sociolinguistics, see sociolinguistics velarization, 65 phonemic velarization, 79 phonetic velarization, 79 velarized consonants, 66, 74 labio-velarized consonants, 7374, 78 Vemer's Law, 371^401 Vietnamese, 377 vowel front lax vowel, 415,445 prenasal vowel, 421 rounded vowel, 453 vowel merger, 415-432,435-438, 440-442, 447-446, see also pen/pin merger, caught/cot merger

500 Subject index Wars of Alexander, 306 Welsh, 91 Whig writers, 137 word formation, 85-88, 94, 96 word-foot structure, 292 word-foot theory 281-283, 299, 307, 313

Wulfstan, 305, 460, 461 Wuyi, 378 X-bar, 327