Reconstructing Alliterative Verse: The Pursuit of a Medieval Meter 1107154103, 9781107154100

The poetry we call "alliterative" is recorded in English from the seventh century until the sixteenth, and inc

403 22 1MB

English Pages X+224 [238] Year 2017

Report DMCA / Copyright


Polecaj historie

Reconstructing Alliterative Verse: The Pursuit of a Medieval Meter
 1107154103,  9781107154100

Table of contents :
List of Tables vi
Acknowledgments vii
Notes on the Text viii
List of Abbreviations ix
Introduction: What Was Alliterative Poetry? 1
1. An Unwritten Medieval Treatise 23
2. The Accentual Paradigm in Early English Metrics 44
3. The Origins of the Alliterative Revival 67
4. The Fourteenth-Century Meter 104
5. The End of Alliterative Verse 130
Epilogue: Edmund Spenser's Poetry Lesson 147
Notes 155
Bibliography 193
Index 213

Citation preview


The poetry we call “alliterative” is recorded in English from the seventh century until the sixteenth, and includes Cædmon’s Hymn, Beowulf, Sir Gawain and the Green Knight, and Piers Plowman. These are some of the most admired works of medieval English literature, and also among the most enigmatic. The formal practice of alliterative poets exceeded the conceptual grasp of medieval literary theory; theorists are still playing catch-up today. This book explains the distinctive nature of alliterative meter, explores its differences from subsequent accentual-syllabic forms, and advances a reformed understanding of medieval English literary history. The startling formal variety of Piers Plowman and other Middle English alliterative poems comes into sharper focus when viewed in diachronic perspective: the meter was in transition; to understand it, we need to know where it came from and where it was headed at the moment it died out. Ian Cornelius is Edward Surtz, S.J., Associate Professor in the Department of English at Loyola University Chicago. His other work includes essays on Boethius’ Consolation of Philosophy, the medieval disciplines of grammar and rhetoric, the English Rising of 1381, and Piers Plowman. He previously taught at Yale University.

c a mbr i dge stu di es i n m edi eva l lite rature General Editor Alastair Minnis, Yale University Editorial Board Zygmunt G. Barański, University of Cambridge Christopher C. Baswell, Barnard College and Columbia University John Burrow, University of Bristol Mary Carruthers, New York University Rita Copeland, University of Pennsylvania Roberta Frank, Yale University Simon Gaunt, King’s College, London Steven Kruger, City University of New York Nigel Palmer, University of Oxford Winthrop Wetherbee, Cornell University Jocelyn Wogan-Browne, Fordham University This series of critical books seeks to cover the whole area of literature written in the major medieval languages – the main European vernaculars, and medieval Latin and Greek – during the period c.1100–1500. Its chief aim is to publish and stimulate fresh scholarship and criticism on medieval literature, special emphasis being placed on understanding major works of poetry, prose, and drama in relation to the contemporary culture and learning which fostered them. Recent titles in the series Lee Manion Narrating the Crusades:  Loss and Recovery in Medieval and Early Modern English Literature Daniel Wakelin Scribal Correction and Literary Craft: English Manuscripts 1375–1510 Jon Whitman (ed.) Romance and History: Imagining Time from the Medieval to the Early Modern Period Virginie Greene Logical Fictions in Medieval Literature and Philosophy Michael Johnston and Michael Van Dussen (eds.) The Medieval Manuscript Book: Cultural Approaches Tim William Machan (ed.) Imagining Medieval English: Language Structures and Theories, 500–1500 Eric Weiskott English Alliterative Verse: Poetic Tradition and Literary History Sarah Elliott Novacich Shaping the Archive in Late Medieval England:  History, Poetry, and Performance Geoffrey Russom The Evolution of Verse Structure in Old and Middle English Poetry: From the Earliest Alliterative Poems to Iambic Pentameter A complete list of titles in the series can be found at the end of the volume.

R E C O N S T RU C T I N G A L L I T E R AT I V E   V E R S E The Pursuit of a Medieval Meter

IAN CORNELIUS Loyola University Chicago

University Printing House, Cambridge CB2 8BS, United Kingdom One Liberty Plaza, 20th Floor, New York, NY 10006, USA 477 Williamstown Road, Port Melbourne, VIC 3207, Australia 4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi – 110002, India 79 Anson Road, #06-04/06, Singapore 079906 Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence. Information on this title: DOI: 10.1017/9781316650516 © Ian Cornelius 2017 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2017 Printed in the United Kingdom by Clays, St Ives plc A catalogue record for this publication is available from the British Library. ISBN 978-1-107-15410-0 Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.


List of Tables Acknowledgments Notes on the Text List of Abbreviations

page vi vii viii ix

Introduction: What Was Alliterative Poetry?



An Unwritten Medieval Treatise



The Accentual Paradigm in Early English Metrics



The Origins of the Alliterative Revival



The Fourteenth-Century Meter



The End of Alliterative Verse


Epilogue: Edmund Spenser’s Poetry Lesson


Notes Bibliography Index

155 193 213



3.1 4.1 4.2 4.3 5.1 5.2

B-verses of Lawman’s Brut (ll. 10534–733 and 12346–91), Ælfric’s St. Edmund, and the alliterative lines of the Middle English Physiologus page 94 B-verses with a single categorically stressed word in Gawain, PPl.Creed, and PPl.Bx 111 Heavy verses in Gawain and PPl.Creed 117 The rhythmic structure of b-verses in Gawain and PPl.Creed 123 Spaced stresses in the b-verses of late Middle English poems 138 Two-lift a-verses in late Middle English poems 140



My studies of alliterative poetry and the Latin language arts began while a graduate student at the University of Pennsylvania. Research leading to the present book began during a year’s leave from duties at Yale University; I thank colleagues in the Department of English for their support and interest in my work. Draft versions have benefited from the intelligence of many readers, among whom I single out for special mention Ardis Butterfield, Thomas Cable, Ben Glaser, Eric Weiskott, and the anonymous readers for Cambridge University Press. A version of Chapter 2 was published in the Journal of English and Germanic Philology; it appears here in substantially revised form. Mary Raschko and Elizabeth Schirmer drew my attention to Piers the Plowman’s Creed and supplied helpful comments on a draft version of Chapter 4. Thomas Cable and Eric Weiskott generously supplied material in press. Georgina Boyle’s skillful copyediting improved my text immensely. For their elegant and attentive typesetting, I thank the production team, headed by Dawn Preston, at Out of House Publishing. I  am honored to see this book through press as a member of the Department of English at Loyola University Chicago. I thank my colleagues here for their confidence and warm welcome. My gratitude to Linda Bree and Alastair Minnis for their commitment to this book, and to Rita Copeland, Roberta Frank, and Ralph Hanna for their inspiration and friendship. No one mentioned here should be held responsible for errors, which are my responsibility alone. I dedicate this book to my teachers.


Notes on the Text

The abbreviations on pp. ix–x identify preferred editions of primary texts cited frequently. Full bibliographic references are provided at the end of this volume. Translations of Beowulf follow R. D. Fulk, ed. and trans., The “Beowulf ” Manuscript: Complete Texts; and “The Fight at Finnsburg,” Dumbarton Oaks Medieval Library 3 (Cambridge, MA: Harvard University Press, 2010). Other unattributed translations are mine. In quotations of alliterative verse, I  mark the caesura, or half-line boundary, with a raised point, regardless of the editorial or scribal punctuation: for example, “In a somer seson · whan soft was the sonne.” In scansions, a strong position may be marked by an acute accent, or bold type, or an “S.” The following are equivalent notations of metrical structure: In a sómer séson · whan sóft was the sónne In a somer seson · whan soft was the sonne xxSxSx · xSxxxSx In the third notation, “x” represents a weak syllable. Scansions are of words, not spellings; the word here spelled soft is assumed to be disyllabic, with an etymologically justified weak final syllable.



ASPR Beowulf Brut CT DIMEV




Anglo-Saxon Poetic Records, ed. Krapp and Dobbie Klaeber’s “Beowulf,” ed. Fulk, Bjork, and Niles Laȝamon: Brut, ed. Brook and Leslie The Canterbury Tales, in Benson, ed., The Riverside Chaucer The Digital Index of Middle English Verse (based on the Index of Middle English Verse [1943] and its Supplement [1965]), ed. Linne R.  Mooney, Daniel W.  Mosser, and Elizabeth Solopova, Early English Text Society Sir Gawain and the Green Knight, ed. Tolkien, Gordon, and Davis Grammatici latini, ed. Keil Journal of English and Germanic Philology Loeb Classical Library Middle English The Middle English Dictionary, ed. H. Kurath, S. M. Kuhn, and R. E. Lewis (Ann Arbor: University of Michigan Press, 1956–2001), online edn., med Notes and Queries Old English The Oxford English Dictionary, ed. J.  A.  H. Murray, H.  Bradley, W.  A. Craigie, and C.  T. Onions (Oxford: Clarendon Press, 1928; 2nd edn. prepared by J. A. Simpson and E.  S.  C. Weiner, 1989; 3rd edn. in progress), online edn., Old French Old Norse Piers Plowman: The B-Version Archetype (Bx), ed. Burrow and Turville-Petre ix

x PPl.Creed Rhet. Her. SEENET STC


List of Abbreviations Pierce the Ploughmans Crede, ed. Skeat Rhetorica ad Herennium, ed. and trans. Caplan Society for Early English and Norse Electronic Texts A Short-Title Catalogue of Books Printed in England, Scotland and Ireland and of English Books Printed Abroad 1475–1640, eds. A. W. Pollard and G. R. Redgrave, 2nd edn., rev. W. A. Jackson et al., 3 vols. (London, 1976–91) Scottish Text Society Yearbook of Langland Studies

Introduction: What Was Alliterative Poetry?

Alliterative poetry is first recorded in English from the late seventh century, which makes it the oldest poetry in this language. Surviving poems include several of the most admired works of medieval literature, including Cædmon’s Hymn, Beowulf, Sir Gawain and the Green Knight, and Piers Plowman. This is also a defunct poetry, having died out soon after the close of the period we call medieval, and it is a deeply mysterious poetry. It was christened “alliterative” in the eighteenth century, for the simple reason that it alliterates a lot. One wonders where this poetry came from, how it was organized, and why it died out. None of these questions has an easy answer. The question of origins is difficult because alliterative poetry is the earliest recorded in any Germanic language. The alliterative meters of Old English, Old High German, Old Saxon, and Old Norse are clearly cognate, but scholars have as yet been unable to push this history further back. Whereas the Germanic languages can be placed within the larger Indo-European family, no comparable genetic placement has been possible for the earliest Germanic verse forms.1 The verse forms of Common IndoEuropean are thought to have been syllabic and quantitative; Germanic verse has seemed unmoored from that archaic framework, distinguished from it by persistent alliterative patterning and especially by fluctuating syllable count. Though certainty is not possible, these peculiar features are usually attributed to the Germanic accentual system. At a pre-historic stage in the development of Germanic, accent became fixed on the initial syllable of words; the metrical forms then current among the Germanic peoples (whatever those may have been) were subsequently reorganized to take advantage of the new perceptual prominence of word onsets.2 That change is usually taken to have been comprehensive, such that it wiped out all traces of the prior metrical system. The earliest surviving verse in Italic and Celtic languages seems to have been affected in similar ways by the acquisition of initial stress accent in these languages, though metrical 1



reorganization was evidently less comprehensive in early Irish verse than in Latin or Germanic.3 To varying degrees, and with important differences in detail, accentual change yielded something we may call alliterative verse. Details are obscure. What is clear is that English alliterative verse differs profoundly from the oldest verse forms reconstructed for the IndoEuropean language family, and differs as well from the accentual syllabic forms that one tends to think of as natural in English. The organization of English alliterative verse has puzzled modern readers since at least the first printings of Piers Plowman and Old English biblical poems, in the mid sixteenth and mid seventeenth centuries, respectively. It has been the object of organized inquiry since the emergence of modern philological studies at the turn of the nineteenth century. Within this historical span, the past three decades have a special prominence. Research since 1985 has yielded achievements of theoretical formulation and empirical discovery rivaled only by the 1880s and 1890s, the decades that saw the publication of major studies by Eduard Sievers and Karl Luick.4 The principal objective of this research activity has been to reconstruct the meter of the surviving poems: to determine how the poetry is shaped at the level of the line. The other two questions posed above – where the poetry came from and why it died out  – have received less attention. In the final chapter of this book, I offer an account of the disappearance of alliterative poetry in English. The other end of the chronological spectrum probably remains beyond our grasp, at least for the present. There is, however, some reason to believe that the ground is being prepared for renewed efforts to reconstruct the pre-history of Germanic alliterative verse, and to clarify its relation to norms of verse-craft observed in other Indo-European traditions.5 At the beginning of the 1980s, just prior to the current flowering of research, one described the long lines of Sir Gawain and the Green Knight roughly as follows: there are four (or four “major”) stresses per line, and a variable number of unstressed syllables distributed around those stresses; the first three stresses are joined with alliteration. About Lawman’s Brut, one modestly shrugged and noted that irregular alliteration co-exists with irregular rhyme.6 About Beowulf and other Old English poetry, one could be more confident:  Eduard Sievers’s “Five-Type Theory” stood as a reasonably accurate description of the repertory of prosodic contours employed in half-lines of Old English poetry. The “Five Types” did little, however, to explain why poetry should be organized in this way. More generally, each of these chronological variants of English alliterative meter appeared as under-regulated, even erratic, by comparison with English



accentual-syllabic verse – and, indeed, by comparison with the other major traditions of premodern European poetry. Sievers’s Five Types at least permitted Old English poetry to be read as having four major stresses per line, and this undoubtedly encouraged a four-stress interpretation of Middle English verse. The defining features of English alliterative poetry would therefore be alliteration, a fixed count of stresses, and a “gabble of weaker syllables, now more, now fewer.”7 Absent a more articulate description – and faced with gaps in the historical record – many scholars have doubted whether Old English poetry, the Brut, and fourteenth-century alliterative poems belong to the same historical series. All this looks different today, at least in the specialist literature. In Old English metrics, research progress is most evident in monographs by Thomas Cable, R. D. Fulk, Geoffrey Russom, and Seiichi Suzuki – a series of creative re-engagements with Sieversian formalism, differing from one another in their specific machinery of historical and theoretical linguistics, and in the relative priority they assign to the individual components of Sievers’s theory.8 For Middle English poetry, one has lacked any comparable starting point. Descriptive typologies, Sieversian in inspiration, were proposed by Luick in 1889 and by J. P. Oakden in 1930, but those efforts failed to sustain research programs in the way that the Five-Type theory did for Old English.9 Nevertheless, it was as renewed essays in typological description, closely corresponding to Luick’s efforts a century earlier, that the current efflorescence of research on Middle English alliterative meter got its start. In the mid1980s, Hoyt N. Duggan and Thomas Cable showed that unstressed syllables in fourteenth-century alliterative verse are distributed in patterns far more regular than Oakden and later scholars had recognized, at least in the second half of the line.10 That rediscovery has proved immensely generative, serving as the point of departure for an array of further work, most notably by Russom, Nicolay Yakovlev, and a team at Bristol University consisting of Ad Putter, Judith Jefferson, and Myra Stokes.11 Though grounded in typological description of the second half of the line (the “b-verse”), this current work makes a comprehensive reassessment of fourteenth-century meter, with especially productive attention to matters of phonology, stress assignment, and historical genesis. The crowning achievement of this research trajectory is Yakovlev’s D.Phil. thesis, submitted at Oxford in 2008.12 In this paradigmshifting study, the main lines of recent research in Old and Middle English alliterative metrics are synthesized, extended with new empirical discoveries, and given a bracingly diachronic construction. In this rapidly expanding research field, the present book intends two contributions. First, I set the work of the past three decades in a longer



historical perspective. The details of disciplinary history – that is, the specific paths taken by scholarship over the past two centuries – place limits on what can be grasped and known. My first objective is therefore to demonstrate, by way of some disciplinary history, why this verse form has remained so mysterious for so long. My second objective is to render alliterative verse ever so slightly less mysterious. I offer descriptions of the Middle English incarnations of this meter, emphasizing matters of systematicity, historical development, and variant realization, and supported by empirical work presented here for the first time. Despite advances in the field of alliterative metrics, the perimeters of consensus remain narrow, and much of this book operates in contested areas. Among the claims defended here are the following: • The most elementary and widely cited description of alliterative meter – according to which lines have four major stresses – is a misapprehension, accurate neither of Old English verse nor of Middle English verse. • Lawman’s Brut occupies a central place in the development of English alliterative meter, whereas Ælfric’s rhythmical alliterating prose does not. • The general prosodic principle of closure – a key formulation of IndoEuropean comparative metrics – operates in both halves of the Middle English alliterative line and was a principal factor shaping the development of this meter between the Old English and Middle English periods. • Duggan’s and Cable’s typological description of the b-verse – the seminal discovery of the 1980s – is a theoretical formulation of the intermediate level, much like the Sieversian Five-Type theory: it may be derived from deeper phonological and metrical constraints, yielding a simpler and more powerful statement of the Middle English meter. • Alliterative verse died out in the fifteenth and sixteenth centuries because the meter became perceptually assimilated to the dominant accentualsyllabic system: it became impossible to hear alliterative verse for what it was, as a distinct metrical system. My argument for these claims is conducted in part by examination of poems underserved in research to date, especially Lawman’s Brut, Piers Plowman B, Piers the Plowman’s Creed, and alliterative poems in rhymed stanzas. Recent research has focused most productively on Gawain and a group of about a dozen other poems that share an approximately congruent metrical style; these poems are sometimes termed the “formal corpus” of Middle English alliterative verse.13 Preferential attention to this corpus has been justified as a research expedient; however, research progress places



us in a good position to now expand the field of inquiry and undertake comparative work. The remainder of this introduction consists in two sketches, corresponding to the two objectives laid out above. The first sketch is epistemological: I describe our contemporary distance from alliterative poetry and thus the conditions that frame our efforts to understand it. The second sketch is of the poetry itself:  I  present a baseline description of the late Middle English meter, as instanced by Piers Plowman. These two sketches are intended as points of departure for the explorations undertaken in later chapters. Chapter summaries are provided at the end. Poets in medieval Iceland composed treatises on the form and language of their vernacular poetry. These treatises are instructive, but they were not written for modern readers, and they omit much that would be of interest to us. Moreover, the Norse and English poetries differ in points of detail, precisely where one would want clarification from a contemporary critic. English poets left no instruction manuals, and the earliest substantive remarks on the form of an English alliterative poem are notably unhelpful. In a preface to his 1550 editions of Piers Plowman, Robert Crowley observed that lines “haue thre wordes at the least in euerye verse whiche beginne with some one letter”; he gave several examples, assured readers that, “This thinge noted, the miter shal be very pleasaunt to read,” and passed on to other topics.14 Langland’s first printer may well have understood more, but he did not have the inclination or vocabulary to express it. One must, then, confront the primary evidence of the poems themselves, as they are preserved in the documentary record: a family of metrical forms, spanning a long and inadequately documented period, discontinued at the end of the Middle Ages, and different from anything that succeeded it. These conditions of inquiry should be considered here at the outset. First, there is the simple fact that alliterative verse died out. The decades following the 1348–49 plague witnessed a general surge in production of English-language literature, and alliterative poetry benefited from that. By the middle of the fifteenth century, however, the form was recessive in southern England, displaced there by the Chaucer tradition. The literary tastes of the metropole were not immediately followed in the north and west, where new alliterative poems continued to be composed into the sixteenth century, and older ones continued to circulate. In Scotland the form remained viable until late in the sixteenth century, but alliterative verse had no chance of surviving the 1603 transplant of James Stuart’s court from Edinburgh to London.15 The deselection of alliterative verse was a complex and protracted process; it stretched over several centuries and did



not proceed in a linear fashion. Once deselection was complete, however, the meter became as inaccessible as a dead language. Among the accentual-syllabic forms that succeeded alliterative verse, the anapestic or triple meter bears a superficial resemblance. Robert Crowley’s line “If bokes may be bolde / to blame and reproue” instances a rhythm that one also encounters in some lines of Middle English alliterative verse.16 I consider the resemblance later in this introduction and again in Chapter 5: it provides some insight into how alliterative verse died out. The point to emphasize now, however, is that English accentual-syllabic meters and the metrical theories constructed to explain them supply no very reliable basis from which to reconstruct the workings of Gawain or Piers Plowman, much less Beowulf. When applied to alliterative verse, the theoretical constructions of generative metrics quickly reveal their accentualsyllabic biases.17 The differences between alliterative and accentual-syllabic meters in English constitute a second major challenge to reconstruction of the alliterative meters. A third challenge derives from the extraordinarily long life of alliterative verse in English. The English language underwent profound changes between the seventh century and the sixteenth. So, too, did the verse form. As a consequence, there are problems of periodization and typology. Should one speak of alliterative meter or meters? If the latter, where and how does one draw lines of division? Related to these problems of periodization and typology is another, concerning poetic language and the historical dimension of metrical systems. Alliterative meter was never fully in sync with itself; it had a way of distorting time, retaining linguistic forms after they had fallen out of the surrounding language, and retaining asystematic vestiges of its prior configurations. Each state of the meter was constituted from some combination of vestiges, innovations, and continuities. These problems are introduced at the end of Chapter 2, and pursued in each of the following chapters. I will distinguish three successive states of English alliterative meter, exemplified by Beowulf, Lawman’s Brut, and Gawain. Chronologically intermediate compositions show a blending of features, while poems in each state of the meter show considerable variety in metrical style. Some rhymed and late poems indicate that the meter was undergoing another transformation in the period when it was swallowed up into the accentual-syllabic tradition of English poetry. Finally, the material records of composition are distributed very unevenly, clustering in the periods 950–1050 and 1390–1475, with a thinner web of documentation extending earlier and later.18 The result is an especially acute case of an epistemological problem inherent to historical inquiry: to



distinguish the past from the surviving evidences of it. Continuities at the level of metrical system show that verse composition must have persisted across gaps in the manuscript record. (This will be argued in Chapter 3.) The two periods that left the richest documentation – c.1000 and the fifteenth century – were periods that saw an uptick in production and copying of literature in English generally. The result, however, is that we have unusually rich snapshots of the meter over these two stretches of its longer history. The surviving copies of poems composed between, perhaps, the eighth and tenth centuries, and again between the mid fourteenth century and the mid fifteenth, necessarily serve as anchors for interpretation of the less numerous documents that constellate the longer tradition. Here, then, is the object of inquiry: a lengthy and internally differentiated tradition of verse composition, poorly attested, defunct, and sharply different from the forms that came after it. Nevertheless, we know quite a bit. In the remainder of this introduction I offer a capsule description of the fourteenth-century meter, and situate this within the longer chain of developments. I relegate controversy and most supporting documentation to later chapters, for which this description will serve as a point of reference.19 Rules capture a practice from the outside, in pieces, and frozen in time. They miss its dynamic and interior unity. Although the difference may appear slight, we would do better to approach the fourteenth-century meter as a system of interdependent organizing principles: 1. Bipartite line structure. The poetry is composed of paired half-line units, shaped like and often corresponding to units of syntax. 2. Grammatical category. The metrical value of words is a function of their grammatical class. 3. Accentual contour. Metrical pattern is realized as an accentual contour, an arrangement of stressed and unstressed syllables. 4. Closure. Metrical pattern is defined most sharply at the right edge, or coda, of metrical units. 5. Segmental cueing. Alliteration sharpens the legibility of the meter by cueing metrical stresses. A discursive presentation is necessarily sequential, introducing the component parts one by one, yet the components function only in coordination, as elements of the completed system. Provided that this is understood, the point of entry into the system is not of great consequence, yet it would also be a mistake to imagine that the elements all have the same status. Of the five just listed, the first two – bipartite line structure and grammatical



word class – are foundational to alliterative verse throughout its long history. The third and fourth – accentual contour and closure – had important roles throughout, but operate most powerfully in the Middle English phase. The fifth element, alliteration, improves the legibility and coherence of the other four. It is best treated last. Alliterative poetry is built up from commensurate chunks of linguistic material. In Middle English verse, lines are typically end-stopped, which means that the basic formal unit is also a unit of sense and syntax. Lines are further divisible into two short verses, or half-lines. In many manuscript copies, the division between half-lines is punctuated by a point, slash, or punctus elevatus. This scribal practice was retained by Walter Skeat; some later editors of Middle English alliterative verse have marked the half-line boundary with a tabbed space, the typographical convention used in modern editions of Old English poetry. The punctuation supplied by scribes and editors is, in any case, only a cue to an inherent structure. The opening lines of Piers Plowman illustrate the centrality of bipartite line structure to the entire metrical system. I quote from J. A. Burrow and Thorlac TurvillePetre’s edition of the B archetype. The half-line boundary is punctuated with a raised point, the mark used for that purpose by the scribe of the base manuscript: In a somer seson · whan soft was the sonne I shope me in[to] shroudes · as i a shepe were In habite as an heremite · vnholy of workes Went wyde in þis world [·] wondres to here Ac on a May mornyng · on Maluerne hulles Me byfel a ferly · of fairy me thouȝte

(Bx Prol.1–6)

The first half-lines (“a-verses”) convey the core sense and narrative action. Indeed, a-verses in this passage make sense on their own: “In a summer season, I dressed myself in garments – in the habit of a hermit – [and] went far in the world. But on a May morning I  saw something special.” The second half-lines, or b-verses, contribute modifying material: subordinate clauses and adjectival and prepositional phrases. They enrich and complicate the sense of this passage, but they become meaningful only in combination with the a-verses.20 In these opening lines, Langland conforms to the traditional style of Middle English alliterative poetry: the a-verse is heavier and more substantive than the b-verse. Elsewhere in Piers Plowman, the trick of reading only a-verses will often fail. Consider, for example, the lines from later in the B Prologue, in which Langland introduces an angel who speaks a Latin warning to the King:



And sithen in þe eyre an hiegh · an angel of heuene Lowed to speke in latyn · for lewed men ne coude Iangle ne iugge · þat iustifie hem shulde But suffren & seruen (Bx Prol.128–31a)

In contrast with the opening lines, the b-verses here deliver syntactically essential material. The grammatical subject – an angel – is contained in the first b-verse of the sentence. In a second contrast with the opening lines, these are not invariably end-stopped. Line 129 ends in an auxiliary verb whose dependent infinitives are placed in the following a-verse. The next b-verse, “that iustifie hem shulde,” supplies an object-clause for the verbs jangle and jugge. The main line of thought resumes (and concludes) in 131a: “And then, up high in the air, an angel of heaven descended to speak in Latin so that uneducated people would not be able to dispute or challenge those who must govern them, but instead submit and serve.”21 The sentence rather unusually ends mid-line; 131b, not quoted above, is a syntactically independent introduction to the angel’s subsequent speech: “for-thi seyde þe angel / Sum Rex sum Princeps · neutrum fortasse deinceps” (Bx Prol.131b–132). One finds similarly intricate syntax in the confession of Wrath, later in the visio: I am wrath quod he · I was sum-tyme a frere And þe couentes Gardyner · for to graffe ympes On limitoures and listres · lesynges I ymped Tyl þei bere leues of low speche · lordes to plese And sithen þei blosmed obrode · in boure to here shriftes And now is fallen þer-of a frute · þat folke han wel leuere Schewen her schriftes to hem · þan shryue hem to her persones (Bx.5.138–44)

Here Langland’s handling of half-line units gives an erratic energy to his indictment of fraternal abuses. Wrath’s brusque self-identification and the accompanying inquit clause (“quod he”) compose a metrical unit with complete syntax: the confession threatens to conclude mid-line, before it properly begins. The inquit clause governs the remainder of the passage, but 138b opens a new sentence, which again ends mid-line, at the end of 140a:  “I was once a friar and gardener of the convent to graft scions onto mendicants and preachers.” The particle for in 139b provides a correct accentual pattern in this half-line (more on that shortly), but also directs us to read the half-line as a purpose clause (MED, “for, prep.,” 5b). Wrath was made convent gardener to perform arboreal surgery; he performed



this duty not on the convent’s fruit trees, but on its members themselves. The peculiar target of Wrath’s grafting is revealed only in 140a, in a prepositional phrase that must be construed closely with the preceding b-verse (to graft x on y), but which discloses, with characteristically Langlandian slyness, that we are now enveloped in metaphor. Having announced the metaphor by identifying one of its constituent elements, Wrath doubles back in anaphora to assign metaphorical value to the other principal element, the ympes that he grafted onto the rootstock-friars: those ympes were lies. Thus completed, the metaphor acquires a sort of vegetable life; in subsequent lines it grows into one of Langland’s brilliant “embryonic allegories,” drawing mischievously on stock elements of mid-fourteenth-century antifraternal satire.22 The fruit of Wrath’s grafting is not that lords’ daughters become pregnant (as antifraternal satire might lead one to expect), but that the sacrament of confession is corrupted: “folke han wel leuere / Schewen her schriftes to hem” – that is, to friars – “þan shryue hem to her persones.” The verb phrase is split across 143b and 144a, a sharp enjambment that, for the second time in this brief passage, vaults a key expository point into syntactic and metrical prominence. These three passages should suffice to show how the half-line functions as Langland’s basic unit of composition, and to show the expressive range that it afforded to him. The half-line is a properly “grammetrical” entity, fusing meter and grammar.23 As such, it is also the domain within which the next three principles of composition – grammatical word class, accentual contour, and closure – mesh with one another. The principle of grammatical word class means that metrical stresses (or “lifts,” or “strong metrical positions”) are normally supplied by the same classes of words that convey the core denotational sense of a passage. This feature is peculiar, and has not gone unchallenged in modern scholarship. It distinguishes alliterative meters from the accentual-syllabic meters that succeeded them in English. Nevertheless, the metrical function of word class in Middle English alliterative verse is supportable on historical grounds, folds neatly into an overall description of the meter, and may be schematized in broader linguistic terms. The basic distinction is between words that serve primarily to establish the surface structure of an utterance (that is, “grammatical” or “function words”), and words that convey its denotational sense (“lexical” or “content words”); most lifts are contributed by the latter category. The significance of this principle is immediately apparent if one compares the passages from Piers Plowman just quoted above with a passage of contemporary accentual-syllabic verse. Here I indicate the nucleus of syllables in strong positions with an acute accent:



And smále fóweles máken mélodýe That slépen ál the nýght with ópen ýe So príketh hém natúre in hír coráges (CT I.9–11)

These lines, from the opening of Chaucer’s Canterbury Tales, have five beats each, with even alternation of strong and weak metrical positions.24 In any given line, ordinary word-stresses of nouns, adjectives, and verbs may be expected to supply the anchoring beats of the line, but only those. To fill out the template, an accentual-syllabic meter does two things: it promotes function words (e.g., the pronouns hem and hir in line 11; also the prepositions in line 19, “Bifíl that ín that séson ón a dáy”) and it assigns a second beat to polysyllabic content words (e.g., mélodýe; also pílgrymáges in l. 12). Middle English alliterative meter is also anchored to the word-stresses of content or “open-class” words.25 Unlike Chaucer’s verse form, however, alliterative verse allows all other linguistic material to fall into strings of weak syllables between metrical stresses. Function or “closed-class” words do not usually supply a metrical stress. Meanwhile, content words generally supply only one metrical stress each, even when polysyllabic like melodye and pilgrymage. There are some problems with this terminology. The distinction between “open-class” and “closed-class” words is a linguistic one; so is the distinction between “content” and “function” words. There is room for negotiation between these linguistic designations and the meter’s use of them. Going forward, I will continue to employ the linguistic terms, but the distinction regards meter and runs as follows in Middle English alliterative verse: Strong metrical positions (“metrical stresses” or “lifts”) are supplied by the lexical stresses of nouns, derived adjectives, derived adverbs, infinitives, participles, and most finite verbs. Weak metrical positions (“dips”) are composed of any unstressed syllables in content words, and by articles, pronouns, prepositions, conjunctions, auxiliary verbs, the verbs have and be, grammaticalized verbs of speech (quod, said), simple adjectives (all, other, many, much, more, some, such, etc.), and simple adverbs (here, there, ever, never, full, etc.). Over short passages, grammatical category works reasonably well to assign lifts and dips. Consider again the opening lines of Piers Plowman, scanned below. An “S” designates a syllable forming a lift; an “x” designates a



syllable forming a dip or part of a dip. Scansions are of words, not spellings. Hence, 2a shope and 2b shepe are both monosyllabic: the former as first person preterit singular of a strong verb, the latter as derivative of OE scēap. In a sómer séson · whan sóft was the sónne I shópe me in[to] shróudes · as i a shépe wére In hábite as an héremíte · vnhóly of wórkes Wént wýde in þis wórld [·] wóndres to hére Ac on a Máy mórnyng · on Máluerne húlles Me byfél a férly · of fáiry me thóuȝte

[xxSxSx · xSxxxSx] [xSxxxSx · xxxSSx] [xSxxxxSSx · xSxxSx] [SSxxxS · SxxSx] [xxxSSx · xSxxSx] [xxSxSx · xSxxSx]

The metrical style of these opening lines is relaxed and transparent:  the syntactic and semantic essentials are contained in the a-verses, and the lifts are straightforwardly a function of grammatical category. The only deviations from the principle of grammatical category in my scansion of this opening passage are the lift I allocate to were (2b) and the second lift allocated to heremite (3a). Yet this treatment of were and heremite hints at the existence of other metrical principles, operating in coordination with category membership and to a certain degree qualifying its expression. The speech of the lunatik, later in the Prologue, shows these complexities in full bloom: Þanne lóked vp a lúnatík · a léne þing with-álle And knélyng to þe kýng · clérgealy he séyde Críst képe þe sire kýng · and þi kýngríche And léne þe léde þi lónde · so léute þe lóuye And for þi ríȝtful réwlyng · be réwarded in héuene And sithen in þe éyre an híegh · an ángel of héuene Lówed to spéke in látyn · for léwed men ne cóude Iángle ne iúgge · þat iústifie hem shúlde But súffren & séruen · for-thi séyde þe ángel

[xSxxxSxS · xSxxxSx] [xSxxxS · SxxxSx] [SSxxxS · xxSSx] [xSxxSxxS · xSxxSx] [xxxSxSx · xSxxxSx] [xxxxxSxS · xSxxSx] [SxxSxxSx · xSxxxSx] [SxxSx · xSxxxSx] [xSxxSx · xxSxSx]

(Bx Prol.123–31)

Violations of the principle of grammatical word class multiply in this second passage. My scansion allocates lifts to two auxiliary verbs (coude in 129b and shulde in 130b), an adverb (with-alle in 123b) and the speechverb seyde (124b), each at line-end. (Compare the lift on were in 2b above.) I have assigned a lift to seyde on one other occasion where that verb introduces speech, this time internal to 131b; and I have assigned two lifts each to the polysyllabic words lunatik (123a) and kyngriche (125b). (Compare my treatment of verse-final heremite in 3a.) Finally, I have demoted two nouns into dips in the b-verse:  þing in 123b; men in 129b.



These violations merely show that the principle of word class does not operate on its own. Its expression is constrained by other principles; until those other principles are introduced in their own right, their operation may be grasped only negatively, as exceptions, irregularities, or fuzziness in the metrical function of grammatical category. The principle of grammatical category predicts that words will always retain a single metrical function, regardless of context, yet this is not true. At the end of verse units, a word that does not bear categorical stress may nevertheless support a lift. The basic idea is that right boundaries should be crisp; this is termed the principle of closure.26 Readers may be best acquainted with the principle of closure as expressed in classical Latin dactylic hexameter and English iambic pentameter. In Latin dactylic hexameter, the first four feet may be resolved or not: dactyl and spondee occur in free variation. The last two feet are, however, almost invariable: the penultimate foot is almost always a dactyl, while the final foot is never a dactyl. Rhythmical variety gives way to rhythmical uniformity at the end of the line. In English iambic pentameter, “trochaic substitutions” and other deviations from weak–strong alternation are construable at the opening of lines as normal realizations of the meter, but as exceptional deviations at the close of the line.27 The close of the line is more uniform, and more tightly regulated, than the opening. The principle of closure has a whole cascade of effects in alliterative meter, which still await a synthetic theoretical treatment. The last lift in the b-verse may not be followed by more than a single weak syllable; a similar rule seems to apply at the close of the a-verse, as Yakovlev has argued.28 Hence the lifts assigned, in my scansion above, to 2b were, 123b with-alle, 124b seyde, 129b coude, and 130b shulde, and the second lifts assigned to 3a heremite, 123a lunatik, and 125b kyngriche. When verse-terminal, function words and final syllables of polysyllabic content words become metrically equivalent to independent content words. The a-verse may close either with a lift and zero dip, as in 123a and 124a (ending in lunatik and kyng, respectively), or with a lift followed by a monosyllabic dip, as in 1a, 2a, and 3a (ending in seson, shroudes, and heremite, respectively). The close of the b-verse was restricted to this second pattern: a lift followed by a monosyllabic dip. The fact that recent scholars of this meter have had so much to say about the rhythmic structure of the b-verse, while the a-verse has remained comparatively intractable, is likewise a consequence of the principle of closure: the rightward member of the bipartite line is more structured than its leftward counterpart. The a-verse exhibits great freedom in accentual



pattern, whereas most Middle English b-verses conform to one of just two basic shapes: Type 1: (x)Sx…xSx e.g., Prol.1b “whan sóft was the sónne” [xSxxxSx]; Prol.4b “wóndres to hére” [SxxSx]; Prol.123b “a léne þing with-álle” [xSxxxSx] Type 2: x…xS(x)Sx e.g., Prol.2b “as i a shépe wére” [xxxSSx]; 5.139b “for to gráffe ýmpes” [xxSxSx]; Prol.125b “and þi kýngríche” [xxSSx]

The b-verse should have two lifts. The second lift should be followed by a single unstressed syllable (a “short dip”). A string of two or more unstressed syllables (a “long dip”) should occur either between the two lifts or before the first. This is the standard description of the b-verse contours in recent scholarship; I will propose a small adjustment in Chapter 4. The contour type with a long medial dip is the more frequent one throughout Middle English alliterative verse. In Chapter 5 I explore the implications of that preference. These accentual contours were observed with impressive regularity in most Middle English alliterative poems. Lines lacking a final unstressed syllable are very rare in Piers Plowman, as are b-verses that lack a long dip. B-verses with long dips in both initial and medial position are, however, robustly attested in this poem. A possible example of that pattern is Bx.Prol.131b “for-thi seyde þe angel.” I  scanned that line xxSxSx above, with monosyllabic seyde, or, maybe better, with elision of the definite article, as þ’angel. Yet the scansion xxSxxSx is also possible. Secure examples of b-verses with two long dips in the Prologue are 160b “in þe cíte of lóndon,” 175b “to þis réson þei assénted,” 186b “þouȝ we cróupe vnder bénches,” and 203b “bi my cónseille be gréued.” B-verses with two long dips – in initial and medial position  – should be considered part of Langland’s metrical repertoire. This pattern sets him apart from other fourteenth-century alliterative poets, a point I return to at the end of Chapter 4. The accentual principle, in combination with the principle of closure, ensures that alliterative poems conform to a limited repertory of contours at the end of verse units. To achieve those contours, the meter often assigns a lift to a function word in verse-final position. A function word may also support a lift within the verse, in the absence of a content word. An example is Bx.Prol.161b “abóuten here nékkes,” where the preposition about supplies the first lift. There is a rudimentary accentual template in Middle English alliterative verse; one of the effects of the template is to elevate



linguistic material that would not be expected to support a lift on the basis of category membership alone. The template also works in the reverse direction, to demote words from grammatical classes that normally supply a lift. This is the reason for assigning the nouns thing and men to dips in Prol.123b and Prol.129b, respectively. In each case, demotion yields b-verses with ordinary accentual contour:  for example, Prol.123b “a léne þing with-álle.” Tension between word class and accentual contour is minimal when promoted function words are already prominent (because they carry emphatic or phrasal stress, or because they participate in the line’s alliterative scheme) and when demoted content words are monosyllabic or semantically weak. In the noun phrases lene þing and lewed men, the adjectives carry the greater denotational load and they alliterate, too; they probably carry the requisite lift in preference to their adjunct monosyllabic nouns. Semantic, morphological, and phrasal factors may contribute to the demotion of content words, but the main factor seems to be the accentual template – that is, the expectation for b-verses to fit into one of the two basic shapes. The a-verse had no such minimal repertory of shapes. Although this matter is far from being settled in the scholarship, it seems that the a-verse accentual contour was restricted only by a historically derived requirement for at least two lifts, and by a bar against a long final dip. A-verses may have three lifts, they may have more than one long dip, and they may end with a lift. Prol.126a exhibits all three features: “And léne þe léde þi lónde” (xSxxSxxS) (the final on londe is an artifact of the scribe’s writing system, without etymological support). Prol.7a exemplifies the verse type with two lifts and two long dips: “I was wéry forwándred” (xxSxxSx). The occurrence of contours like these in the Middle English a-verse is another expression of the principle of closure: the a-verse admits a range of patterns usually absent from the b-verse. Even when a-verses match the rhythmical patterning of b-verses, they typically differ from b-verses in the way that they coordinate prosody with segmental phonology. The final unstressed syllable in the b-verse usually contains schwa, the unstressable neutral vowel.29 This is the vowel we find in the final syllables of sonne (1b), were (2b), workes (3b), here (4b), hulles (5b), and so forth. By contrast, the final dip of the a-verse admits the full range of Middle English vowels, as illustrated by the verse-final words such as mornyng (5a), ferly (6a), wastours (22a), and liflode (30a, < OE līflād). A short medial dip in the b-verse is usually also filled with schwa, as in 5.139b “for to gráffe ýmpes” (the infinitive -e(n) inflection gives schwa in the nucleus of the medial dip) and Prol.54b “and here wénches áfter” (plural -es inflection



gives schwa in the same position), though Langland deviates from other fourteenth-century alliterative poets in this respect. In Chapter 4 I will propose that these schwas were, in fact, the mechanism that prevented b-verses from having more than a single long dip. The preferential allocation of schwa to non-initial short dips in the b-verse is one of the ways that the alliterative meter coordinates prosody with segmental phonology. The meter’s namesake alliteration likewise coordinates prosody with segmental phonology, and it, too, contributes to the overall asymmetry between a-verse and b-verse.30 In the normal pattern, alliteration coincides with metrical stress twice in the a-verse and on the first of the two lifts in the b-verse: this pattern is always schematized as aa/ax, where “a” represents an alliterating lift, “x” a lift without alliteration, and the slash represents the caesura. There were many variations. Langland’s alliterative patterning is famously irregular, and scholarship has tended to approach this variety by elaborating taxonomies of all patterns plausibly attributable to him. From the perspective adopted here, however, the taxonomic approach misapprehends surface effects as realia. Alliteration serves as a cue to lifts, helping to maintain the legibility of the whole metrical system. When there is a third lift-bearing word in the a-verse, that lift may also be cued by alliteration. The result is the “variant” pattern aaa/ax. Occasionally the line-final lift also alliterates, as in Bx.Prol.1, “In a somer seson · whan soft was the sonne.” The resulting aa/aa alliteration testifies to a shiftable balance of forces:  the principle that coordinates prosody and segmental phonology (cueing of stress by alliteration) extends itself at the expense of the usual asymmetry between a-verse and b-verse. Somewhat more interesting are the “deficient” alliterative patterns, ax/ax, xa/ax, and aa/xx.31 Lifts normally cued by alliteration are not. In a significant minority of lines, alliteration is displaced from a lift to a syllable adjacent to it. This occurs most often on the first lift in the b-verse, as Bx.Prol.22, “And wonnen that [þese] wastours · with glotonye destruyeth.” George Kane termed these shifts “modulation,” and that is a good name for them.32 The principle that coordinates prosody with segmental phonology is modulated, such that alliteration is displaced from the lift it would ordinarily cue. The five principles set out here are interdependent, organizing the poetic line through their systematic combination. In combination, they permit the scansions offered in the previous pages, and now is perhaps the time to acknowledge that those scansions are strange. Whereas one might reasonably expect an English meter to have a consistent number of strong positions in each verse unit, the a-verses of Piers Plowman and



other Middle English alliterative poems seem to admit variation in this parameter. Most a-verses have two lifts, but many others seem to have three. Second, whereas one might expect stresses to be evenly spaced, alliterative verse presents us with astonishing rhythmical variation: sometimes lifts adjoin one another; at other times they are separated by runs of four or more weak syllables. This rhythmical variety is beautiful, but it is also peculiar, at least when viewed from the perspective of English accentual-syllabic meters. Hence the doubts, for so long, about whether Middle English alliterative poetry had any properly metrical organization. We may lay those doubts aside: the fourteenth-century alliterative meter was a thing; it may be defined synchronically, as a system. The meter was, however, also a thing in transition:  its whatness was penetrated by the modalities of the no-longer and the not-yet. Synchronic description can only proceed so far. The peculiar configuration of the fourteenth-century meter comes into sharper focus in diachronic perspective: it was shaped by what it once was; simultaneously, it was bending towards what it might yet become (but was not yet). Clashing stress, variation in stress count, and rhythmical variety: Middle English alliterative verse owes these features to its Old English precursor. The rhythmical patterns schematized in Sieversian metrics are probably best considered epiphenomena, the outcomes of a metrical organization located at a deeper level. Most basically, the meter consisted in a fourposition frame and a binary division between strong and weak linguistic material. (This description is controversial, but it is based on the work of Cable and Yakovlev and draws support from Fulk’s analysis of secondary and tertiary stress; I set out the arguments in Chapter 2.) In the following example, from Beowulf, I place strong material in bold: 5a 5b 6a 6b 7a monegum mǣgþum · meodosetla oftēah · egsode eorlas · syððan ǣrest wearð · fēasceaft funden ⋁ | | | ⋁ | | | | | ⋁ | | ⋁ | | | | | | | 1 2 3 4 1 2 3* 4 1 2 3 4 1 2 3 4 1 2  3 4

[(He) deprived many peoples of mead-seats, terrified men, after he was first discovered destitute.]

A single weak syllable forms a metrical position on its own: an example is the dative -um inflection in monegum, which forms the second metrical position in verse 5a. Runs of consecutive weak syllables always group together:  an example is the two syllables of the adverb syððan, which together form the first metrical position in verse 6b. Under specific circumstances, consecutive syllables may also group together to form a single



strong position. An example is supplied by the first two syllables of monegum: here the root syllable is short, so it forms a strong metrical position in combination with the following syllable. This is termed resolution; with the exception of resolution, strong linguistic material always fills one metrical position per syllable. Thus, the first three syllables of 7a “fēasceaft funden” form three consecutive strong positions. The classification of linguistic material as strong or weak occurs at the level of the morpheme, as Yakovlev explains:33 Strong metrical positions are formed by (the long syllables or resolved sequences of ) roots, suffixes and stressed prefixes of open-class words [that is, nouns, adjectives, lexical verbs and derived adverbs], excluding finite lexical verbs; strong metrical positions are also formed by (the initial – except for unstressed prefix – long syllable or resolved sequences of ) any other word displaced from its normal syntactic position and/or standing verse-finally. Weak metrical positions are formed by (the syllables of ) inflections, unstressed prefixes, finite lexical verbs, and closed-class words [that is, articles, pronouns, auxiliary verbs, conjunctions, and prepositions].

Unstressed prefixes and the negative proclitic ne constitute a third class of morphological material, in that they may optionally be omitted from the count of metrical positions, as illustrated in the passage above by the verbal prefix of- in 5b. Yakovlev’s definition of strong and weak linguistic material combines into one statement the effects that I separated between the principles of grammatical word class and closure in my description of the late Middle English alliterative meter above: hence his caveat that function words contribute strong positions when placed in verse-final position. Provided that this is understood, the most important difference between Yakovlev’s statement for Old English verse and my earlier statement for Middle English verse is the priority of lexical stress in the latter. A word like wisdom supplies two strong positions in Old English verse, but typically only a single lift in Middle English verse. In the late Old English and early Middle English period, the alliterative meter gradually reorganized itself around the principle of stress accent. The b-verse came to be defined by a reduced set of accentual contours, construable as selective reflexes of the older system. Meanwhile, the a-verse remained relatively free from constraints on accentual contour:  indeed, the range of accentual patterning in a-verses expanded prodigiously as the four-position rule fell away. These transformations will be explained in later chapters. For now, the key point is the following: the extreme rhythmical variation we observe in Middle English verse is unusual in a meter ostensibly keyed to stress accent; that variation



occurs because the Middle English meter developed from an earlier one that was not stress-based. That is one temporal horizon. What lies over the other? Consider this passage: Þay boȝen bi bonkkez · þer boȝez ar bare Þay clomben bi clyffez · þer clengez þe colde (Gawain 2077–78)

[They went by banks where branches are bare. They climbed by cliffs where the cold clings.]

These lines are perfectly good Middle English alliterative verse. They are also emphatically anapestic. There is nothing to prevent a line of fourteenthcentury alliterative verse from conforming to anapestic rhythm, but anapestic rhythm exists for this meter as one possibility within a vast array of others. These two lines pop out against the background of Gawain. They give a little shock of regularity: precisely parallel syntax, precisely anapestic rhythm, and extra alliteration. The extra alliteration is, perhaps, a wink from the poet: he knows, and wants us to notice, that these are not ordinary lines. The poet is toying with us and with the rhythmical possibilities of his meter. A century-and-a-half later, Robert Crowley would select and generalize precisely this anapestic potential. Earlier in this chapter I quoted two lines from Crowley’s 1550 collection of didactic verse, One and Thyrtye Epigrammes. The poem was printed in short lines, but that presentation format is probably attributable to page size: the rhyme scheme and disposition of capitals point towards a long-line form. Accordingly, I present the poetry as paired half-lines, with my usual punctuation at the caesura: If bokes may be bolde · to blame and reproue The faultes of al menne · boeth hyghe and lowe As the Prophetes dyd · whom Gods spirite did moue Than blame not mine Autor · for right well I knowe Hys penne is not tempered · vayne doctrine to sowe (sig. A2r)

Whatever the line-to-line variations, the background rhythm is unmistakably anapestic. Crowley’s poetry shows us that he indeed understood more about alliterative meter than he indicated in the prefatory epistle to his editions of Piers Plowman. What Crowley understood, however, was just the possibility of running an anapestic transformation on that meter – that is, the possibility of selecting one of the incidental realizations of the alliterative meter and instituting it as a new norm.



Piers Plowman, Gawain, and other fourteenth-century poems are as distant from the anapestic future as they are from the morphological past, yet those two termini bring the fourteenth-century meter into focus. The meter is a system, but also a system in transition. Chapter  1, “An Unwritten Medieval Treatise,” asks why there are no medieval treatises on English alliterative verse, and what a treatise might have told us, had one been written and survived. To this end, I examine the thirteenth-century Icelandic treatises on Old Norse poetry, the remarks of Gerald of Wales on uses of alliteration in English and Welsh poetry, and the teachings of the classical and medieval disciplines of grammar and rhetoric. Gerald of Wales noticed alliteration in English poetry, but saw it only as an amplification of the ornamental device he had been taught to recognize in the Aeneid. Óláfr Þórðarson recognized that alliteration had a structural function in the poetry of Iceland, but struggled to express that perception within his chosen framework, Latin grammatical instruction. Óláfr’s uncle, Snorri Sturluson, composed an informative series of treatises on Old Norse poetry, but his account of poetic form lacks a concept of metrical stress. A medieval treatise on English poetry would have left many of our questions unanswered. Medieval literary theory lacked the requisite concepts. Those concepts have been a long time coming. Chapter  2, “The Accentual Paradigm in Early English Metrics,” argues that the main lines of metrical research over the past two centuries have underestimated the peculiarity of Old English meter and its difference from English accentualsyllabic forms. Since the beginning of the nineteenth century, study of Old English poetry has been guided by the notion that the meter was “accentual,” “strong-stress,” or “stress-based” (I use these terms interchangeably). In recent decades, however, scholars have become increasingly conscious of incongruities in stress-based theories of Old English meter:  there are smudges in the picture, and offending details that can be neither removed nor rationalized. Challenges to the “accentual paradigm” have been articulated by Cable, Fulk, and, most fully, by Yakovlev. This second chapter inquires how modern scholars first came to read Old English poetry as accentual, and why that research paradigm may now be replaced. Yakovlev has proposed a non-accentual theory of Old English meter. Simultaneously, he lays the groundwork for a new account of the genesis of the Middle English line, and thus for a reformed understanding of English literary history in the high and late Middle Ages. Chapter 3, “The Origins of the Alliterative Revival,” takes up the problem of continuity in the English alliterative tradition. From the time that



modern scholars first recognized Old English poetry as poetry, they also noticed its resemblance to the alliterative poetry of the fourteenth and fifteenth centuries. Both have alliteration; both lack end-rhyme; they share some distinctive vocabulary. These resemblances have been given several different interpretations. The basic positions were established already in the eighteenth century, in the writings of George Hickes, Thomas Warton, and Thomas Percy. There have been several revolutions of the merry-goround since then. By its vigorous return to the language of the texts, current research on alliterative meter enables a pervasive reinterpretation of surviving documents, and a more general reassessment of medieval English literary history. The balance of evidence, I argue, is that the surviving alliterative poems in Old and Middle English are the remnants of a continuous practice of verse composition in a continuously evolving verse form. Lawman’s Brut plays a key role in this narrative, and the fourteenth-century “alliterative revival” should be seen as rising out of a much longer history of verse composition, with tails of lower productivity before and after. Chapter 4, “The Fourteenth-Century Meter,” takes a closer look at the structure and diversity of unrhymed Middle English poems. Standard literary histories divide Middle English alliterative verse into poems of “high style” and poems of “plain style,” or into a “formal corpus” and an “informal corpus.” Research on alliterative meter has justifiably focused on the “formal corpus,” a group of about a dozen poems exhibiting considerable uniformity in language and meter. The “informal corpus” is a more miscellaneous category, yet it contains at least one robust sub-group, consisting of Piers Plowman and epigones. In Chapter 4 I describe the meter of Gawain, which I take as representative of the formal corpus; simultaneously, I explore the deviant metrical styles of two poems in the informal corpus, Piers Plowman B and Piers the Plowman’s Creed. In my account of the Gawain-meter, I  focus especially on the principle of closure, the articulation of the medial caesura, and the ranking of constraints in the second half of the line. I also suggest a reformulation of the b-verse rules established by Cable and Duggan. Comparison between the three poems shows what is distinctive about each and shines a light on metrical diversity, which remains an underexplored topic in the scholarship on alliterative poetry. In his line-work, as in other aspects of his poem, Langland was singularly adventurous. If fourteenth-century alliterative poems have seemed to constitute a literary movement distinct from other Middle English literature, that is perhaps only because alliterative verse subsequently died out. My last chapter, “The End of Alliterative Verse,” inquires how and why the English



language lost its alliterative meter. The inquiry begins in Scotland, in the court of James VI: in the 1580s, the Scottish monarch sponsored alliterative poetry at his Edinburgh court and also wrote about it in a treatise on poetics. James’s remarks have been taken as an insider’s view of alliterative verse; I argue instead that he viewed the meter through the distorting lens of accentual-syllabic verse, and that at least one poet at his court did the same thing. That Scottish courtier was not alone. From the fourteenth century to the sixteenth, poets repeatedly took alliterative verse as a model for something different, usually in combination with rhyme. In certain Harley lyrics and in some later poems in 13-line stanzas one can perceive the emergence of a new meter, shorn of complexity and reconstructed according to an anapestic template. This was one of the ways that English alliterative verse met its end: reduced to an anapestic swing, it was absorbed into the family of accentual-syllabic forms. An epilogue, “Edmund Spenser’s Poetry Lesson,” reads the ghosts of alliterative verse in Spenser’s Shepheardes Calender. By the 1570s, when Edmund Spenser set about inventing a retro look for new English poetry, alliterative verse had long since ceased to be a vital component of literary culture over most of the realm. Middle English alliterative verse could still be read, especially in the mid-century prints of Piers Plowman, but its metrical form had no chance of being understood. In the Shepheardes Calender Spenser wrote lines that recall the old alliterative line, but he also drew upon classical rhetoric to explain why this form belonged in the past.

Ch apter  1

An Unwritten Medieval Treatise

Consider clergie, defined by the Middle English Dictionary as “the clergy (as distinguished from the laity)” and “knowledge, learning; doctrine” (sense 1a, 3a). French in origin, the word began to appear in English texts around 1300. How would it be used in verse? Chaucer’s surviving poetry contains just one instance, in the Wife of Bath’s Prologue.1 There clergie is placed at line-end and rhymes with compaignye; from this placement, one infers that clergie retained its French shape, counting as three syllables and bearing stress on the penult, the last syllable containing a full vowel.2 The word also retained its French shape at the end of line 1570 in the Middle English alliterative poem Cleanness.3 If the b-verse “of pure clergye” is built like any number of others, then it scans xSxxSx: the weak e of pure (another French loan) and the initial syllable of clergie form the requisite long medial dip. The word clergie contributes three syllables, stressed on the middle one, just as in Chaucer’s usage. More difficult cases are, however, not hard to find. The poems that accompany Cleanness in manuscript contain only one other instance of clergie, and there the word alliterates: “And koyntyse of clergye · bi craftes wel lerned” (Gawain 2447).4 If the meter permits disjunction between stress and alliteration – which it might – then clergie may be construed to have the same accentuation in Gawain 2447 as in Cleanness 1570.5 If the meter permits a long dip at the end of the a-verse – which it might – then clergie may be construed in Gawain 2447 as retaining its trisyllabic shape while bearing stress on its initial, alliterating syllable. With initial stress, the word would soon lose its weak final -e and acquire the shape of our modern English clergy. That modern form may appear an attractive solution to Gawain 2447, for it would yield an a-verse with normal alliterative pattern and a normal final dip. Other evidence, however, tells against this interpretation. In Gawain, Patience, and Cleanness, the syllables contributed by historically and grammatically justified final -e almost always cooperate with the meter:  a reading that treats weak final -e as 23


An Unwritten Medieval Treatise

metrically relevant consistently matches expected metrical patterns, whereas a reading that treats this syllable as defunct often comes in a syllable short.6 That observation is justly taken as evidence that the language of these poems retained final -e, and thus one ought to assume that final -e is retained in Gawain 2447a. How, then, should one interpret the metrical structure of this half-line? A  Yakovlevian scansion assigns two lifts to clergie, scanning “And koyn-ty-se of cler-gy-e” (xSxxxSSx).7 That interpretation is difficult: it posits stress clash and three lifts, but it faithfully represents the poet’s conservative grammar of final -e and his avoidance of long dips in the verse coda. Other understandings of the poet’s language and meter will, however, inevitably yield other readings of this particular half-line. Ad Putter, Judith Jefferson, and Myra Stokes agree with Yakovlev that the Gawain-poet’s language retains final -e, yet they recognize no constraints on the coda of the a-verse and argue that prima facie three-lift verses are always reducible to two-lift variants.8 They would probably scan this verse “And koyn-ty-se of cler-gye” (xSxxSxx), with elision of -e in koyntyse before the following vowel.9 Hoyt N. Duggan accepts three-lift a-verses in principle, but would see no reason to posit a three-lift interpretation in this instance. He would presumably scan this verse xSxxSx, positing a phonology in which final -e is unsounded and plays no role in the meter. I could go on, but won’t. Among the meanings of ME clergie is “learning,” and anyone who has puzzled through alliterative poems and wondered about their form has probably wished for clergie direct from the source:  an indigenous ars poetica, unencumbered by the inferences and doubts that weigh so heavily on my previous paragraph. One would like to have things set out plainly by a native informant, who, perhaps, would divide the topic into what is required, what is permitted, and what is prohibited in the meter and its language. That request would not have been unrecognizable. Latin poetry was supported by a well-developed technical literature, transmitted in the medieval schools and particularly by the medieval discipline of grammar.10 Moreover, descriptive energies did not remain confined to the sphere of Latinity: between the twelfth century and the fourteenth, the Latin grammatical literature supplied concepts and models for description of poetic forms in two of the vernacular languages neighboring on English – Welsh and Old Norse.11 The treatises on Old Norse poetry rightly hold a central place in modern scholarship, as the only contemporary descriptions of the form and language of medieval Germanic poetry. The Háttatal (“list of verse forms”), written by Snorri Sturluson c.1221–25, supplies modern

An Unwritten Medieval Treatise


scholarship with the word stave (ON stafr) as a name for the alliteration linking (for example) the initial syllables of the first three nouns in Gawain 2447: “And koyntyse of clergye · bi craftes wel lerned.”12 Snorri’s treatises supply a number of other insights and hints about the structure of alliterative verse; however, an ars poetica written for Old Norse poetry could only ever be an approximate guide to the metrical form of English poetry. The two traditions were different in their details, precisely where one would want to have the opinion of a native informant. Someone could have written a treatise like Snorri’s for English alliterative poetry. As far as we can tell, no one ever did. This chapter inquires what a treatise might have told us about English alliterative verse, had one been written and survived. This is not the Borgesian exercise that it might appear to be:  my aim is not to compose the medieval treatise that medieval authors never did, but rather to emphasize how many questions a hypothetical medieval treatise on the meter of Beowulf (or the Brut, or Gawain) would have left unanswered. The metrical practice of alliterative poets exceeded the conceptual resources of medieval literary theory. The theorists are still playing catch-up today. Near the end of De arte metrica, after giving sufficient treatment of the more important meters (metris eminentioribus), Bede briefly mentions rhythmus, or accentual Latin verse. This, he says, is “scanned [examinata] not by a quantitative system, but by the number of syllables, according to the judgment of the ears, ut sunt carmina vulgarium poetarum.”13 That last clause was once taken to mean “like the songs of the vernacular poets,” and thus to affirm a basic similarity between non-quantitative Latin verse and English alliterative verse, at the very beginning of recorded history of the latter.14 Yet it is unlikely that vulgaris has the meaning here in Bede’s treatise that it will later have in Dante’s De vulgari eloquentia. Several lines below those just quoted, Bede contrasts vulgares poetae with docti poetae, and quotes two Latin hymns as products characteristic of the vulgares. The hymns are in accentual, non-quantitative meters.15 The distinction imposed by the word vulgares is therefore between two types of Latin poetry, represented as differing in sophistication and learning. Quantitative verse had become a creature of the schools in late antiquity: after vowel quantities were leveled in spoken Latin, the quantitative meters could be sustained only by meticulous grammatical instruction. Rhythmus, meanwhile, was the name given to the vulgar, or unschooled, Latin verse that could be written and enjoyed without reference to a fossilized phonology. Bede’s paragraph on this poetry, inclusive of the clause on carmina vulgarium


An Unwritten Medieval Treatise

poetarum, is taken directly from earlier treatises.16 There is no reason to think that his words refer to poetry in the vernacular. My question in this chapter is, however, slightly different: not whether Bede compared vernacular verse to accentual Latin verse (everything indicates he did not), but, instead, whether Latin verse and the vocabulary that grammarians employed in describing it would have been any help in conceptualization of English verse. Here, again, the answer seems to be “no, not much.” In an introduction to his treatise on elementary Latin grammar – the first such to be written in a vernacular language – Ælfric warned that “ars grammatica multis in locis non facile anglicae linguae capit interpretationem, sicut de pedibus uel metris” (many aspects of Latin grammar find no ready expression in the English language: for example, [the teaching] on feet and meters).17 We cannot be sure just how Ælfric conceived of this translation problem, for he did not attempt a treatment of these topics in his grammar. Yet the problem itself is readily recognizable, and not limited to quantitative metrics. Consider, for example, the treatment of accentuation in the Latin grammatical literature. The location of lexical accent is counted off from the end of the word, as befits analysis of this language.18 Medieval English calls for a totally different analytic framework:  the Old English accentual system is left-oriented, typically assigning word-stress to initial syllables; in Middle English the accentual system became more complex, and remained a poor match for the analytic tools on offer in Latin grammatical literature. Descriptions of Latin prosody, as found in late antique grammatical literature and – from the twelfth century – in treatises on rhythmus and cursus, would have been little help in making sense of what English poets were doing. If we turn from prosody to alliteration, we seem to be on surer footing:  the Icelandic treatises attended in detail to the alliterative pattern; the Middle English poets remarked on it; and alliteration has been central to modern perceptions of this poetry, from Robert Crowley forward. Moreover, discussions of alliteration (if not under that name) were available to medieval poets and commentators in the Latin grammaticorhetorical literature. For all these reasons, alliteration seems a good anchor for our inquiry into what a medieval ars poetica might have said about English alliterative verse. The topic has an added virtue, in that it will eventually return us to that more difficult, but more important, question of accentuation. Many of the texts treated in this chapter will be familiar, but it is only by setting the relevant materials out in array that one sees how little a medieval treatise would have answered our questions about alliterative verse.

Rum, Ram, Ruf, and All That


Rum, Ram, Ruf, and All That On the question of formal self-consciousness in the English alliterative tradition, one turns habitually to a trio of Middle English prologues. They are unhelpful; at most, they confirm that alliteration was perceived as distinctive of a certain type of versification. The clearest testimonium comes from a poet who never wrote alliterative verse. Geoffrey Chaucer has his Parson protest, “I am a Southren man; / I kan nat geeste ‘rum, ram, ruf,’ by lettre” (CT X.42–43). These words have been read as a literary prise de position and a statement on the status and circulation of alliterative poetry at the end of the fourteenth century. The lines have accordingly attracted a steady stream of commentary, aiming to decipher the force of the Parson’s self-affirmed geographic identity (“I am a Southren man”) in relation to compositional competence (“I kan nat geeste … by lettre” = “I am not able to versify alliteratively”).19 What is clear is that Chaucer made alliteration the distinguishing feature of a poetic mode; this fact is customarily leveraged in the interpretation of two other, more problematic, fourteenth-century testimonia. Early in Gawain, the narrating “I” affirms that his story is set down and fixed in the traditional way, “With lel lettres loken” (35). It is probably not possible to decide whether the truthfulness, accuracy, or faithfulness claimed by lel regards primarily the narrative record or its technical expression.20 Perhaps the ambiguity was intended. Finally, a line in the prologue to Winner and Waster has been interpreted as a third participant in the same referential system: the poet complains that patronage is now bestowed upon mere boys who “neuer wroghte thurgh witt · [thre] wordes togedire” (25).21 The emendation thre replaces the manuscript reading thies “these” and has been read as an allusion to the three alliterating stresses in a standard aa/ax line.22 One may doubt that interpretation: the emended line may accuse young rivals of alliterating incorrectly, or it may instead be a hyperbolic put-down for which the precise number of words (two, three, four, etc.) is a matter of indifference. The words werken, maken, werke, and maker recur insistently in surrounding lines of the prologue, where the poet protests loudly about compositional craft (“[to werken] thurgh witt”). One cannot be sure whether the emended line describes the poet’s compositional craft or merely denies that rivals possess it. Perhaps the more important point is this, however: even if we grant that the passages in Gawain and Winner refer specifically to alliteration, they reveal very little about verse technique. Nor should we expect technical description from these sources. They are, after all, literary prologues, not treatises manqués. As befits prologues, the


An Unwritten Medieval Treatise

passages are concerned with framing the reader’s or auditor’s perception of the subsequent literary compositions, not with explaining the mechanics of verse construction. The Gawain-poet and Winner-poet affirm that their chosen form of versification is an art, truthful in the broadest sense, and thus deserving of attention and sponsorship. To pursue this line of inquiry further, we need sources with a genuinely descriptive intention, and this means turning from prologues to treatises and their terms of art.

The Invention of “Alliteration”: A Postmedieval Excursus Terminological poverty is not unique to the Middle English prologues. On the contrary, it may be their most representative feature, for Latin technographic literature likewise lacked vocabulary for the forms and techniques of vernacular versification. The Latin word alliteratio goes back no further than the 1490s, when it was coined by the Italian humanist Jovianus Pontanus as an instrument for description of sound effects in Virgil’s Aeneid.23 This may be the appropriate moment at which to recall the subsequent history of the word. It was noticed at the beginning of the seventeenth century in Gerardus Vossius’s important and often-reprinted textbook of rhetoric.24 The word began appearing in English soon thereafter (for which, see the OED, s.v.). In 1656 Thomas Blount recorded the word in his Glossographia; or, A dictionary interpreting the hard words of whatsoever language, now used in our refined English tongue. He defined it as a figure of rhetoric “repeating and playing on the same letter.”25 The leap from classical rhetoric to English historical poetics occurred a hundred years later. Thomas Warton dedicated several pages to Piers Plowman in his Observations on the “Fairie Queene” of Spenser (1754), commented on the poem’s “alliteration,” and linked its “alliterative versification” to Old English poetry.26 The name “alliterative versification” was established for subsequent scholarship by Warton’s contemporary Thomas Percy, whose Reliques of Ancient English Poetry (1765) contained an “Essay on the Metre of Pierce the Plowman’s Visions.” There Percy set out to describe what he termed the “alliterative species of versification.”27 Beginning in the fourth edition of the Reliques, published in 1794, the essay itself was re-titled “On the Alliterative Metre, without Rhyme, in Pierce Plowman’s Visions.”28 We have been using that term ever since. In a narrowly terminological sense, “alliterative meter” was an eighteenth-century invention. I supply a fuller context for Percy’s terminological imposition in the next chapter of this book (the circumstances

Alliteration avant la Lettre


were not propitious). For my present argument, the important point is that technical vocabularies are made, not given, and they can remain slippery. Between Pontanus and Percy, there was a series of dislocations: from rhetoric to poetics; from the literature of ancient Rome to that of medieval England and Scandinavia; from incidental ornament to namesake. For Pontanus, alliteratio organized only a collection of sound effects, and it did not matter whether the repeated sound occurred at the beginning or end of words, or even in the middle. Thus conceived, alliteratio came very close to the type of word play that Roman rhetoric had named annominatio. Indeed, annominatio was the model on which Pontanus coined his new term, and the heading under which Vossius noticed the term in his handbook of rhetoric. The difference between the two terms was just one of typological conception: is the element repeated and modified basically a word (nomen) or a letter (littera)? After Pontanus, a whole series of semantic and discursive adjustments had to occur before his coinage could serve to designate a poetic form. Meantime, one could always fall back on phrasal description: sixteenthcentury English commentators wrote of running upon a letter, playing with the letter, hunting the letter, and coursing a letter; later and more scholarly writers on alliterative verse discussed literae plane similes and dictiones ab eadem initiali litera incipientes.29 The premodern disciplines of Latin grammar and rhetoric supplied a wider array of terminological and conceptual resources, particularly in the ubiquitous lists of figures and tropes. This vocabulary, however, was no better suited to Germanic poetry than alliteratio would be.

Alliteration avant la Lettre The standard term in the Latin grammatical tradition was parhomoeon, defined by Aelius Donatus as the figure that obtains “cum ab isdem litteris diuersa uerba sumuntur” (when different words are chosen from the same letters).30 That definition may be obscure, but Donatus’ example showed that the repeated letter should come at the beginning of words: “O Tite, tute, Tati, tibi tanta, tyranne, tulisti.” E.  H. Warmington rendered this  tongue-twister as “Thyself to thyself, Titus Tatius the tyrant, thou tookest those terrible troubles”; it derives from the Annales of the early Roman poet Ennius (239–169 bc) and it was already established in the textbook tradition when Donatus wrote in the mid fourth century ad.31 Ennius’ line would remain a favorite among schoolmasters throughout the Middle Ages and beyond. Isidore of Seville, writing about 625,


An Unwritten Medieval Treatise

retained Ennius’ line as an example of parhomoeon and improved upon Donatus’ definition: he defined this figure as “multitudo verborum ex una littera inchoantium” (a gathering of words beginning with one letter).32 Bede retained Donatus’ definition in De schematibus et tropis (c.710), but his examples  – drawn from the Bible rather than Roman poets  – again show that he considered parhomoeon to be a matter of repeated sounds at the beginnings of words.33 It seems that Donatus, Isidore, and Bede stand at the end of a long semantic development, for the word parhomoeon was not used by Greek authors in the precise sense it acquired in the Latin grammatical tradition. The Greek parómoios means “closely resembling” or “nearly equal”; Aristotle used the word paromoíōsis to name the placement of the same word or of similarly sounding words at the beginnings of paired cola, or the use of inflectional rhyme at the ends of paired cola.34 What mattered was not the specific type of lexical or phonic repetition, but instead the use of various kinds of repetition to highlight syntactic parallelism. At the turn of the common era, Publius Rutilius Lupus retained the generalized reference of the Greek word; his entry for parhomoeon begins “Hoc schema et homoeoteleuton et homoeoptoton fere non multum inter se distant” (There is scarcely much difference between this figure and homoeoteleuton and homoeoptoton).35 His example confirms that he considered alliteration and rhyme interchangeable. In the following centuries, the word underwent a semantic specification that testifies to the comparatively greater importance of alliteration in Latin than in Greek. In the second half of the third century, the grammarian Marius Plotius Sacerdos described parhomoeon as precisely the opposite of homoeoteleuton: the former concerns phonic similarity at the beginning of words, the latter at their ends.36 As an example of parhomoeon, Sacerdos supplied the phrase “casus Cassandra canebat” (Aeneid 3.183), which would join Ennius’ tongue-twister as a favorite of the schoolmasters. These examples sealed the meaning of parhomoeon for later grammarians. The word that had served in Greek rhetoric as a generic name for the various types of lexical and phonic assimilation came, in Latin grammar, to name precisely alliteration. The semantic development of parhomoeon in Latin grammatical literature in the first centuries ad presumably reflected differences between the literary language of Greek and Latin. Heavy alliteration in Old Latin is one of the reasons why scholars today think that language once had a fixed word-initial stress accent.37 When primary lexical stress subsequently moved rightwards, to its classical position, alliteration remained

Alliteration avant la Lettre


a prominent stylistic feature of Latin literature, and it always remained more important in that language than in Greek. By imposing a newly precise meaning on an existing technical term, Latin grammarians designated a stylistic feature with differential salience in the Latin language. Accommodation was, however, only one of several possible responses to this linguistic situation. The Rhetorica ad Herennium (written in Rome c.86–82 bc) noticed alliteration only to warn against over-doing it. This anonymous textbook would become the most important source for classical rhetorical teaching in the second half of the Middle Ages. Its notice of alliteration reads almost as an allergic reaction to the uses of alliteration in contemporary and earlier Latin literature. Some context is necessary. The remarks on alliteration occur in book four of the Rhetorica ad Herennium, where the topic is elocutio, or verbal execution. The author explains that an appropriate and finished style (elocutio commoda et perfecta) is distinguished by the three features elegantia, compositio, and dignitas (Rhet. Her. 4.12.17). The first, elegantia, consists in grammatical correctness (latinitas) and clarity (explanatio) (4.12.17). The third, dignitas, consists in decoration with figures of diction (verborum exornatio) and figures of thought (sententiarum exornatio) (4.13.18–4.56.69). The treatment of dignitas is by far the longest of the three subsections; it would have an enormous medieval circulation and influence, forming the basis for most treatments of the figures from Marbod of Rennes (active in the second half of the eleventh century) through the twelfth- and thirteenth-century arts of poetry and prose.38 Alliteration is noticed neither in the Herennium-author’s catalog of exornationes, nor the medieval catalogs dependent on it. Instead, notice of alliteration occurs in the briefer section on conpositio, or the arrangement of words in sentences. Instruction in this section is entirely negative, and includes the following warning against alliteration: Conpositio est verborum constructio quae facit omnes partes orationis aequabiliter perpolitas. Ea conservabitur … si vitabimus eiusdem litterae nimiam adsiduitatem, cui vitio versus hic erit exemplo – nam hic nihil prohibet in vitiis alienis exemplis uti: “O Tite, tute, Tati, tibi tanta, tyranne, tulisti”; et hic eiusdem poetae: “quoiquam quicquam quemquam, quemque quisque conveniat, neget.” (Rhet. Her. 4.12.18) [Conpositio is verbal arrangement that makes all parts of a work uniformly refined. It will be preserved … if we avoid excessive recurrence of the same letter, and this defect will be exemplified in the following verse (for nothing prevents us from using someone else’s words to illustrate defects): “Thyself


An Unwritten Medieval Treatise to thyself, Titus Tatius the tyrant, thou tookest those terrible troubles”; and this from the same poet: “… denies that anyone [owes] anything to anyone, whoever sues whomever.”]39

Geoffrey of Vinsauf reiterated this stricture around 1200 in his Poetria Nova, adding, however, that less insistent repetition could be a virtue.40 Alliteration would be censured in this way throughout the remainder of the Middle Ages and beyond; in the sixteenth century this textbook tradition, traceable to the Rhetorica ad Herennium, would shape Edmund Spenser’s censorious remarks about English alliteration, as I show in the epilogue to this book. Within the Rhetorica ad Herennium itself, the meaning of the warning may be inferred from the examples that illustrate it and from the company it keeps. The other devices censured as injurious to conpositio are hiatus (that is, collision of vowels at word boundaries), eiusdem verbi adsiduitas (lexical repetition), similiter cadens (inflectional rhyme), transiectio verborum (contorted word order), and longa verborum continuatio (long sentences). These are each devices of style. Even hiatus, almost universally spurned by Latin rhetoricians, was acknowledged by Quintilian to have its proper uses.41 All but hiatus and alliteration reappear, later in Rhetorica ad Herennium book four, in the catalog of exornationes.42 The appearance of these devices on both lists is consistent with the basic conceptual architecture of classical rhetoric. Figures are value-neutral per se; whether a given use is an asset (virtus) or a defect (vitium) is a judgment that can only be made contextually.43 The same devices that contribute dignitas may also ruin conpositio. For acts of excessive, injudicious, or infelicitous ambition in the field of stylistic ornamentation, rhetorical theory had the name cacozelon, “bad ambition.” In this way – that is, in theory – rhetoric’s conceptual architecture treats all stylistic devices as equal. In actuality, the historical dimension of style differentiates figures from one another. Of the six devices listed as detrimental to conpositio, the last two – transiectio verborum and longa continuatio – are features of the contemporary high style, with its rhythmical clausulae and long periods. The author is warning against overenthusiastic performance to contemporary stylistic norms. The three devices involving repetition – eiusdem verbi adsiduitas, eiusdem litterae adsiduitas, and similiter cadens – are traceable back to Aristotle’s devices for highlighting syntactic parallelism, but they are associated by the Herennium-author with an antiquated Latin style. That association is driven home by illustration of at least two of these three devices with verses from Ennius; the verses illustrating verbal repetition are not identified, but

Alliteration avant la Lettre


are thought to derive from Ennius as well.44 Verbal repetition, inflectional rhyme, and alliteration were used in Old Latin verse and prose as linking devices. Alliteration typically came in pairs and triplets, and served to link paratactic cola. The hymn to Mars recorded in Cato’s De agri cultura (c.160 bc) is an important example of this style.45 The elegiac couplet that Ennius is said to have composed for inscription on his tomb illustrates the archaic style adapted, with striking elegance, into a newly literary and Hellenic meter:46 nemo me lacrimis decoret nec funera fletu faxit. cur? uolito uiuos per ora uirum [Let no one honor me with tears, nor perform my funeral rites with weeping. Why? I fly, alive, through the mouths of men.]

There are three units of sense and syntax, the second and third marked off with alliterative rhymes. The by-form dacrumis would provide an alliterative rhyme for the first colon as well: an attractive emendation understandably rejected by editors. Alliteration on /f/ links the nouns at the end of the hexameter with the governing verb at the head of the pentameter: here traditional linking alliteration counter-balances the complexity introduced by enjambment. In the final unit, one may be tempted metri causa to read “uolito uiuo,” with two juxtaposed first-person singular verbs. That construction, which I do not recommend as an emendation, would employ alliteration in its most traditional way, in combination with Old Latin paratactic style. The reading “uolito uiuos” gives much the same effect. It is a beautiful epigram. In contrast with either the hymn to Mars or Ennius’ funerary couplet, the line picked up by the handbooks gives little indication why writers would ever want to use this device. “O Tite, tute, Tati, tibi tanta, tyranne, tulisti” demonstrates what parhomoeon is, but not what parhomoeon could do.47 It served well as a cautionary example. In summary, it seems that the absence of alliteration from the Herenniumauthor’s catalog of verbal ornaments responds to a precise historical conjuncture: Greek rhetoric had not identified alliteration as a device of style distinct from others, and Roman rhetoricians were ambivalent towards a device that they evidently associated with an outdated style. Later accounts remained fixed within this matrix. In one stream of medieval teaching, alliteration was merely registered as something that poets do. The other stream warned against alliteration, or warned against excessive use of it. Isidore combined the two streams when, quoting both the Ennian and Vergilian model lines in Etymologies 1, he praised Virgil’s greater temperance.48


An Unwritten Medieval Treatise

Grammar pointed out the bottle marked “parhomoeon”; rhetoric warned not to over-do it. This kind of teaching was bound to provoke some medieval schoolboys to acts of mischief. In the later ninth century Hucbald of Saint-Amand composed a poem of 136 hexameter lines, in which every word begins with the letter c.49 The poem is announced in its heading as an exercise in parhomoeon. This was not a promising framework for a description of English alliterative verse. In the absence of an alternative descriptive language, however, it was almost inevitable that some medieval scholars would frame their observations of alliterative verse in the terms supplied by Latin grammatico-rhetorical teaching.

Sort of Like Virgil: Gerald of Wales on Alliterative Verse In a chapter of the Description of Wales (1194) concerning “their sharp and subtle intellect,” Gerald of Wales described the peoples of western Britain as experts in forensic rhetoric (causis, actionibus, et foro civili) and folk song (cantilenis rhythmicis et dictamine).50 He praised the Welsh rhetoricians for omitting none of the canons of Roman rhetorical teaching and he praised their poets for exquisite use of the figures of speech and thought. The poets, he wrote, are like the Celtic bards of whom Lucan tells in the Pharsalia: hardened and pouring out song.51 In a passage long known to scholars of medieval English literature, Gerald remarks that vernacular poets use a lot of alliteration:52 Prae cunctis tamen rhetoricis exornationibus, annominatione magis utuntur; eaque praecipue specie, quae primas dictionum literas vel syllabas convenientia jungit. Adeo igitur hoc verborum ornatu duae nationes, Angli scilicet et Kambri, in omni sermone exquisito utuntur, ut nihil ab his eleganter dictum, nullum egregium, nullum nisi rude et agreste censeatur eloquium, si non schematis hujus lima plene fuerit expolitum. (Descriptio, pp. 187–88) [They use annominatio more than any other rhetorical ornament  – and especially that type of annominatio which links the initial letters or syllables of words together in conformity. These two peoples (that is, the English and Welsh) use this verbal ornament in all elevated discourse: so much so that they consider nothing to be well spoken, no saying to be excellent, and no expression to be anything other than rough and crude – unless it has been thoroughly polished with the file of this figure.]

We have seen that classical rhetoric did not recognize alliteration in its catalog of stylistic devices. The closest term available in the Rhetorica ad

Sort of Like Virgil


Herennium and the medieval catalogs dependent on it was the one used by Gerald in this passage:  annominatio, the figure of general word-play or punning.53 Gerald recognized that the term was less specific than he needed; thus his specification that the vernacular poets use a variant of annominatio that “links the initial letters or syllables of words together in conformity.” He then supplied examples in English and Welsh. I  set the Welsh illustrations aside as outside my competence.54 There are three English proverbs: God is togedere gamen and wisdom Ne halt nocht al sor isaid ne al sorghe atwite Betere is red thene rap and liste thene lither streingthe (Descriptio, p. 188)

[Play and wisdom are good together Neither voice every pain, nor reprove every injury Counsel is better than haste, and strategy better than wicked force]

The third of these proverbs is paralleled in Lawman’s roughly contemporary Brut, line 8590. Gerald’s version differs in metrical shape, but all three proverbs make for plausible early Middle English alliterative verse.55 My focus here, however, is not on the lines themselves but on Gerald’s commentary, for he next remarks that “the same ornamental device is often found in Latin writing”: “In Latino quoque haud dissimiliter eloquio, eandem exornationem frequens est invenire” (Descriptio, p. 188). As illustration he quotes part of Aeneid 3.183, “tales casus Cassandra canebat,” and an elegiac couplet spuriously attributed to Virgil: Dum dubitet natura marem faceretve puellam Natus es, O pulcher, pene puella, puer. [While nature hesitates, uncertain whether she should make a boy or a girl, you were born, O you beautiful boy, nearly a girl.]

I do not know where Gerald encountered this couplet.56 Aeneid 3.183 was, however, one of the stock illustrations of this figure, as we have seen. Gerald perceived alliteration as the salient feature of vernacular English and Welsh poetry, but also as an intensification of the sound effects that he had learned to recognize in Virgil’s poetry in school. The Description of Wales is not the sort of text to accommodate detailed investigation. It is a diverting collection of observations on the geography and inhabitants of a peripheral region, framed with persistent reference to the classical education that Gerald shared with his readers. Quotations from Roman poets are sprinkled liberally throughout; examples of Welsh


An Unwritten Medieval Treatise

wit are followed by Cicero’s jokes (Descriptio, 1.14; p. 192). Yet Gerald does identify one other way, beyond sheer intensity, in which the use of alliteration by Welsh and English poets differs from that of the Latin poets. This concerns the origin of their style, for he supposes that the English and Welsh owe their alliteration not to technical training (ars) but to unreflected habit: Nec ego tamen id crediderim, quod priores populi duo, tam diversi ab invicem et adversi, in hoc verborum ornatu ex arte conveniant, sed potius ex usu longo: qui, quia placuit solum, et facili similium ad similia transitu aures demulcet, per succedentia tempora inolevit. (Descriptio, p. 189) [Nor would I  have believed that two peoples so different from and even contrary to one another could converge on this verbal ornament through technical instruction (ex arte). More likely, it comes from long-standing custom (ex usu longo), which worked its way in over time, just because it was satisfying and because it caresses the ears with an easy movement from like to like.]

A quotation from the Rhetorica ad Herennium drives the point home: Sicut Tullius, in libro De Elocutione, de talibus qui usum habent et non artem loquitur, dicens “Ceteri, cum legunt orationes bonas, aut poemata, probant oratores et poetas; neque intelligunt quare commoti probent; quod eo scire non possunt, ubi sit, aut quid sit, quomodo factum sit id, quod eos maxime delectet.” (Descriptio, p. 189) [It’s like Cicero says in the book On Eloquence about the kind of people who have know-how (usum) but not technical knowledge (artem). He says: “Others, when they read good speeches or poems, admire them, but they do not understand why they are moved to admire. For they really do not know how that which most pleases them is made, or what it is, or where to place it” (Rhet. Her. 4.2.3).]

Usus here names a linguistic practice lacking any acquaintance with its own principles of organization: at best, unreflective usus may produce an effective speech and sense that it is good; it cannot explain how this happens and how to make it happen again in changed circumstances. This is a powerfully self-authorizing gesture for the theorist of rhetoric, but it becomes a trap in Gerald’s writing on vernacular poetry. His description of vernacular poetry can only proceed in the terms of grammatico-rhetorical ars, which he acknowledges as alien to the practice of the vernacular poets. Modern philology and linguistics would eventually step into that gap, elaborating an analysis of linguistic usus independent of rhetorical ars.

What Goes Without Saying in Icelandic Artes Poeticae


Gerald’s remarks have been interpreted to mean that alliteration was no more than an incidental ornament in twelfth-century English poetry, unconnected to verse design. “In Latin usage, and emphatically so in Gerald’s, the word [annominatio] refers to a rhetorical color,” Ralph Hanna observes, “that is, to a nonessential verbal embellishment, usually designed to confer ‘elegance’ on a statement. Gerald thus associates the English usage with an ornament, not an essential verse feature.”57 This is indeed what Gerald says, but one might wonder whether he would have been capable of saying anything else, once he had started to describe vernacular poetry in Latinate terms. Like Hanna, I am inclined to see alliteration as having a superficial relation to verse design in much of Middle English alliterative verse: it serves as an optional segmental cue to the location of metrical stresses. Yet Gerald of Wales cannot be regarded as a reliable witness on this question. Even if alliteration were an integral component of verse design in the poetry he heard, Gerald would, one expects, have perceived only an extreme proclivity for the grammatical figure that he learned to recognize in the Aeneid. The Descriptio Kambrie gives us a view of culture totally over-determined by that premodern “puberty rite,” the study of Latin.58

What Goes Without Saying in Icelandic Artes Poeticae Latin bias could be expected in any medieval treatise on vernacular poetry, but the Old Norse treatises show how much independence was possible. Between the mid twelfth century and the mid fourteenth, Icelanders wrote a collection of grammatical and poetic treatises that together constitute one of the fullest descriptions of a literary vernacular to be produced in medieval Western Europe.59 The treatises exhibit a considerable range in character and focus: from an exacting analysis of the vernacular sound system to a mischievous round-up of Nordic mythology. Two treatises correspond closely to our topic. The first is Snorri Sturluson’s Háttatal, which treats alliterative patterning, syllable count, rhyme, and stanza structure in Old Norse poetry.60 This is the treatise I signaled at the beginning of this chapter as giving modern scholars the term “stave”; it dates from about 1220 and is transmitted as the last of the treatises in Snorri’s Edda. The other treatise with direct relevance to our topic dates from about 1250 and was composed by Snorri’s nephew Óláfr Þórðarson.61 It consists of two parts. The second is a vernacular translation of Donatus’ catalog of figures and tropes in Ars maior, book III, replacing Donatus’ Latin illustrations with exemplary quotations from Old Norse poetry.


An Unwritten Medieval Treatise

This accordingly includes Donatus’ entry for parhomoeon. We must begin with Snorri’s Háttatal, which supplied Óláfr with some of his vernacular terminology. Early in the Háttatal Snorri divides poetics into three parts: setning, leyfi, and fyrirboðning, rendered by Anthony Faulkes as “rule,” “license,” and “prohibition.”62 Faulkes draws an apt comparison with Latin grammatical teaching, which was sometimes ordered into a pars praeceptiva, pars permissiva, and pars prohibitiva.63 The pars praeceptiva was the teaching on “parts of speech” and their morphology – the declensions of nouns, inflections of verbs, and so forth – as set forth in the first two books of Donatus’ Ars Maior. Grammatical figures, including metrically motivated phonological deformations of Latin words, were presented in the pars permissiva, which treated types of rule-breaking licensed to poets. For instance: lengthening of the first vowel in Italiam, so that the word may fit into a line of dactylic hexameter, as in Aeneid 1.2, Italiam fato profugus; or the omission of a prefix, as temno for contemno.64 Donatus’ treatise and its descendants contain many other examples of linguistic variation motivated by or employed in metrical composition. The comparison is useful but inexact. Snorri’s Háttatal is not a grammar, and so it necessarily assigns a different value to the grammatical tradition’s tripartite divisio materiae. Under setning, “rule,” Snorri put instruction regarding the correct count of lines per stanza and correct count of syllables per line, plus instruction on qualitative distinctions in meaning (málsgrein) and sound (hljóðsgrein). Málsgrein concerns poetic diction; hljóðsgrein concerns the disposition of internal rhymes and alliteration. Deviations in a prescribed count or disposition of alliterative sounds were forbidden (rangt, “wrong”), Snorri said, adding that a line with vocalic alliteration could, as a license (leyfi), contain pronouns, conjunctions, and prepositions that also begin with vowels, for these words would not contribute to the count of alliterating sounds. Snorri’s description of vernacular verse design is justly praised for its fullness, precision, and independence from Latin traditions. Yet there are important omissions, attributable to the fact that treatises tend to be written for publics already familiar with basic aspects of the craft they set out to teach. Precisely as a result of their pragmatic insertion, poetic and grammatical treatises omit information that would be useful to uninitiated and historically belated readers. Donatus’ grammatical treatises will not teach you to conjugate Latin verbs and Horace’s Ars poetica will not tell you how to write a metrically correct hexameter.65 Those deficiencies could be corrected in other treatises, but the deficiency

What Goes Without Saying in Icelandic Artes Poeticae


itself presumably did not exist for the original and intended users of the Latin treatises. Similarly, Snorri’s treatises are written for a public already initiated into the craft he set out to explain, even if less adept in it than the great masters of bygone times. Accordingly, his Háttatal rings the variations on alliterative patterning, syllable count, rhyme, and stanza structure; his Skáldskaparmál (“The language of poetry”) catalogs and deciphers a forest of specialized poetic words and kennings. Neither treatise presents a concept of metrical stress, a central component of all modern understandings of early Germanic meters.66 Snorri issues instructions for the count and placement of alliterative sounds, but not for the count and placement of stresses. It is true that the operations of metrical stress may be inferred from Snorri’s discussion: his list of words that do not contribute to the count of alliterative sounds in lines with vocalic alliteration (noted above, instancing the grammatical concept of license) consists entirely of pronouns, prepositions, and conjunctions – precisely those words that do not usually contribute a strong position in alliterative verse. The operations of metrical stress are clearly in play here, but its concept is not. The word stafir, “staves,” refers only to those strong positions marked with alliteration; Icelandic poetics had no term for an unalliterating strong position. A. V. C. Schmidt filled this terminological void when he designated the stressed syllables marked by alliteration “full staves” and those without alliteration “blank staves.”67 In Schmidt’s useful terminology, the initial syllables of koyntyse, clergye, and craftes are full staves in Gawain 2447, while the initial syllable of lerned, the last word in that line, is a blank stave. Non-alliterating strong positions are as much a part of Old Norse poetics as they are of Old English, but they receive no explicit notice from Snorri. Perhaps he did not recognize or think about stress contrasts in the metrical template of vernacular poetry. If he did, he evidently judged that this feature could go without saying. Suprasegmentals – the qualities of stress, pitch, and duration that differentiate the prominence of syllables in speech – are notoriously difficult to describe, and the ancient and medieval grammatical literature is notoriously unreliable in this area. The first part of Óláfr Þórðarson’s treatise contains a treatment of accent (hljóðsgrein) that is closely dependent on Priscian’s inaccurate treatment of Latin accent.68 (Priscian was, it seems, misled by Greek musical accent, which he and other Latin grammarians transposed into their descriptions of Latin.) Snorri managed to avoid distortions of this kind: his prologue to the Edda shows that he was keenly attuned to the ironies of cultural emulation. Yet he also gave a description of vernacular verse design that lacks what one might consider its central


An Unwritten Medieval Treatise

component – an account of prosody. Reconstruction of prosodic structure has been a modern project. The clergie transmitted in grammatical and poetic treatises is always both less and more than the opinion of a native informant. Essential aspects may be passed over in silence. Descriptions of one language (usually a more prestigious one) may be projected onto another. The very act of analysis could introduce distortions of its own. The bulk of the Háttatal consists of a poem of 102 stanzas, each in a different form or style, accompanied by a commentary on the unique features of each stanza. An earlier Icelandic poem and probable model for Snorri is called Háttalykill (“key to verse-forms”). The Latin grammatical tradition supplied precedents for treatises of this sort, under the titles clavis metrica (“metrical key”) or centimetrum (“one hundred meters”).69 That taxonomic drive sometimes expresses itself in Snorri’s treatise as a tendency to over-formalize verse practice – or, in other words, to confuse rule and license. Snorri states that fornyrðislag (“old story meter”; the chief form of the Eddic poetry) has only one stave in a-verses.70 This is indeed the majority pattern, but a-verses with two (full) staves are not uncommon, as also in Old English verse. Snorri split the form into two. As Geoffrey Russom writes, “The marked Eddic variants excluded from fornyrðislag by Snorri are not discarded but are assigned to a separate meter, Bálkar lag, where they become the rule rather than the exception.”71 In De centum metris, Servius had similarly split variant realizations into separate meters:  he designated dactylic hexameter with bucolic dieresis (e.g., Aeneid 1.33, “tantae molis erat Romanam || condere gentem,” where there is a word division after the fourth foot) as its own meter, separate from the heroic.72 Direct influence of Servius’ treatise cannot be demonstrated for Snorri’s Háttatal, and need not. The point is that the textbook genre runs the reification engine on overdrive, imposing differences of kind where verse practice has only a variant realization of a single kind. In a euhemerist prologue to his treatises on poetics, Snorri tells readers that those figures of Norse legend whom their ancestors revered as gods were in fact great men of the Trojan royal family.73 A young Trojan nobleman named Woden (“it is him that we call Odin”) left his native Turkey when he discovered, by prophecy, that he was destined to be remembered and venerated by the people of the North. Moreover, everything happened for the Asians (Æsir; Snorri here puns on the traditional name of the Norse gods) just as had been prophesied, thanks to the native Scandinavians’ boundless enthusiasm for the intricate social

What Goes Without Saying in Icelandic Artes Poeticae


system, unbroken political victories, and rich culture of the new arrivals. The Scandinavians even adopted the speech of the Asians as their own language (eigintunga).74 The first of Snorri’s treatises, titled Gylfaginning (“The tricking of Gylfi”), recounts how the Æsir cleverly abused the natives’ emulative enthusiasm and thereby secured for themselves the veneration that had been prophesied for them. With this fable, Snorri dignifies the ancient Icelandic language and mythology, providing it with a prestigious eastern Mediterranean provenance; simultaneously, Snorri’s fable disowns the Nordic mythology, setting that pagan heritage at a safe distance from the modern Christian creed. And it thematizes the quality of cultural emulation endemic to its genre. In a prologue to his translation of Donatus, Óláfr Þórðarson reprised his uncle’s Trojan narrative, now stripped of its wry reflexivity. Instead, Óláfr made the Troy story an analytic framework: In this book it may be clearly understood that the art of poetry (skáldskapr) which the Roman sages learnt in Athens in Greece and then transferred into the Latin language is the same art (ein listin) as the verse form of songs or poetry (ljóðaháttr eða skáldskapr) which Odin and other men of Asia brought hither northwards into the northern hemisphere; and they taught men this type of art (list) in their own language (á sína tungu), just as they had organized and learnt it in Asia itself, where beauty and power and knowledge were the greatest in the world.75

Óláfr sharpened the theme of translatio studii et imperii, but also gave the line of transmission two distinct branches. He retained Snorri’s story about Odin, but reintroduced the traditional line of translatio from Greek to Latin. Northern and southern branches share a common origin in the eastern Mediterranean, Óláfr affirms, and, in both, the original poetic art (list, skáldskapr) has been preserved across differences in language (tunga). This claim of cultural affiliation is both a thesis to be supported í þessi bók (“in this book”) and a pseudo-historical justification for the work of cultural synthesis that Óláfr undertook in producing his book, for his project was basically to establish equivalences between Donatus’ schemes and tropes and the devices of Iceland’s vernacular poetry. The result is that key elements of vernacular verse design, as identified by Snorri, are scattered among incidental ornaments and prosodic doublets. Kennings, internal rhyme (hending), and alliteration are identified with the grammatical figures metaphora, paronomasia, and parhomoeon, respectively. As an account of vernacular poetics, Óláfr’s treatise is totally disorganized by his Latin schema. Yet Óláfr was also a poet, and he was evidently aware of the lack of fit between systems. By replacing Donatus’ Latin illustration of parhomoeon


An Unwritten Medieval Treatise

with a vernacular one, he broke free of the Latin teaching tradition and assigned a positive function to the figure:76 Paranomeon er þat, ef mǫrg orð hafa einn upphafsstaf, sem hér: Sterkum stilli styrjar væni Þessi figura er mjǫk hǫfð í málssnildar list, er rethorica heitir, ok er hon upphaf til kveðandi þeirrar, er saman heldr norrœnum skáldskap, svá sem naglar halda skipi saman, er smiðr gerir, ok ferr sundrlaust ella borð frá borði. Svá heldr ok þessi figura saman kveðandi í skáldskap með stǫfum þeim er stuðlar heita ok hǫfuðstafir. [Parhomoeon is when many words have a single initial letter, as here: “There may be hope of battle for the strong ruler” This figure is much used in the art of eloquent speech, which is called rhetoric, and it is the foundation of that poetical effect that holds together Norse poetry, just as nails hold a ship together, which a [ship]wright makes, and [which] goes in loose order or plank from plank. So too this figure holds together the poetical effect in poetry by means of those staves which are called props and chief staves.]

The terms stafir, stuðlar, and hǫfuðstafir – “staves,” “props,” and “headstaves” – are borrowed by Óláfr from the Háttatal. Like Snorri, however, Óláfr lacks a technical language in which to state that the kind of alliteration that matters in Norse poetry is alliteration in a stressed position. He uses an analogy instead. The comparison of poets to mechanical craftsmen was well established in Norse literature, but Óláfr’s analogy is evidently unique in its formal specificity and argumentative aim.77 By means of analogy to ship-building, he emphasized that parhomoeon has a nonincidental, structural status in Icelandic poetry. The medieval artes poeticae indicate what counted as knowledge (clergie again) at a given time and place. By their shape, their contents, and their very existence, the treatises trace a historical line, one that divided the things that were desirable – or even possible – to put into discourse from the things that would be left unsaid. If we wish to know how this poetry is organized, we must turn from treatises to the poems themselves, and analyze their language with concepts often very different from those employed in the treatises. This conclusion, which I  operationalize in subsequent chapters, entails a parting of ways with the scholarly enterprises that go under the names “historical poetics” (or “historical prosody”) and “medieval literary theory” – research programs united in their call to read literary texts through the lens of indigenous, period theories of literature. The projects of “historical prosody” and “medieval literary theory” continue

What Goes Without Saying in Icelandic Artes Poeticae


to be fruitful, but they have limitations not reducible to the scarcity of surviving materials from early periods. If our objective is to understand past practices of poetic composition, indigenous theories of poetry can only get us so far, and have a questionable authority. In poetry as in other fields of culture, people don’t necessarily do what they say they are doing, and often do things they don’t say; practice has its own relative autonomy, whereas theoretical exposition is always partial. For the reconstruction of past literature, modern linguistic and philological description remains irreplaceable, and yields insights unavailable within any epistemological frame that could be claimed as medieval. One topic of the next chapter is the way that modern philology extricated itself from the grammatical tradition and established a new basis for study of English verse.

Ch apter  2

The Accentual Paradigm in Early English Metrics

In a moment of expository relaxation in A History of Old English Meter, R. D. Fulk briefly considers his topic from the perspective of the uninitiated: “the structure of early Germanic verse has always seemed so arcane to students of literature,” he writes, “that frequently nonlinguists have doubted even the most basic of Sievers’ conclusions”; “Modern English stress patterns do not seem different enough from Old English ones that a stress-based metrical system should be so difficult for the uninitiated to grasp.”1 I remember distinctly my own first encounter with Old English meter, and my consternation by it. I was an undergraduate student of literature and a nonlinguist, reading Beowulf in George Jack’s fine student edition. Jack’s word glosses and footnotes permitted basic access to the sense of the poem I was reading; simultaneously, they integrated this particular reading experience into any number of previous ones, similarly aided by glosses. If Jack’s glosses were more copious than I had encountered previously, the language of literary annotation was nevertheless familiar. I expected the same dialectic of alterity and familiarization from the account of verse form contained in my primer, Bruce Mitchell and Fred Robinson’s Guide to Old English. The form of Old English poems may well be unlike anything I had seen previously, but the language of metrical annotation ought to be familiar. A stress in Beowulf should, I thought, mean the same thing as a stress in later English verse, and my primer did little to disabuse me of that assumption. Phenomena of secondary stress and resolution should have been warning flags – indications that my received vocabulary of English metrical form would fail me in this case. These phenomena could, however, be subordinated, and presented as nuances within a stress-based metrical system. On this understanding, the Old English meter differed in pattern from anything that came after it in English, but not in its basic element. I scanned Beowulf as I would the Faerie Queene, with stressed syllables and unstressed syllables, reminding myself to be on guard against secondary stress and 44

The Accentual Paradigm in Early English Metrics


resolution. Sievers’s Five Types seemed a poor consolation for the bewildering variety. This is not a problem exclusive to the uninitiated: “Since before the time of Sievers,” Fulk writes, “the general assumption among metrists has been that the primary phonological correlate of ictus in Old English verse is stress.”2 That basic assumption causes problems, for it is not easily reconciled with other aspects of the meter: “It strains credibility to suppose that stress and length both played such pervasive roles … [T]he question arises whether the role of stress has been overestimated.”3 Fulk answers that question in the affirmative, with due caution, and moves on to other matters: his carefully delimited aim in A History of Old English Meter – basically, to establish the chronology of the surviving poems – prevented him from developing this point, which contains in nuce a new theory of Old English meter. Elaboration was left to Nicolay Yakovlev, who, by a different chain of reasoning, has also concluded that the role of stress in Old English meter has been overestimated.4 The present chapter is about the history and consequences of the overestimation of stress in Old English metrics. If Fulk and Yakovlev are correct that stress was less important to Old English meter than we have thought, we must make room for new conceptual possibilities: the meter will remain beyond our grasp until we have peeled back the accretions of stress-based or accentual thinking about it. In the present case, this means returning “before the time of Sievers,” to the opening decades of the nineteenth century. For it was then that moderns first determined that the poetry of the Anglo-Saxons was based on the patterning of stressed and unstressed syllables. Henceforth, the fall of accents would be analogized to the natural and energetic pulses of walking, finger tapping, and breathing. These pulses would eventually be seen to construct a minimal set of stress contours and the contours would, in turn, be taken to define the meter and guide study of it. I will call this the “accentual paradigm” of English meter. In doing so, I use the word “paradigm” in the sense given to it by Thomas Kuhn and succinctly expressed by the editors of the OED: “a conceptual or methodological model underlying the theories and practices of a science or discipline at a particular time.”5 The accentual paradigm has been a conceptual and methodological model underlying the discipline of English metrics since its nineteenth-century inception. As such, the paradigm’s force has extended beyond any of its individual instantiations; it has remained current even as individual theories and scansions have been revised and refuted. (I will use the two terms “stress-based” and “accentual” interchangeably, as often in the literature on English prosody.)


The Accentual Paradigm in Early English Metrics

Several decades ago, Jürgen Kühnel wrote a stimulating neo-Hegelian history of the scholarship on medieval Germanic meters.6 Kühnel’s accomplishment was to center disciplinary history on the ways that successive generations of researchers formulated problems. Research progress in the intervening decades makes it desirable to repeat Kühnel’s experiment, this time with a sharper focus on the emergence, consolidation, and subsequent elaborations of the accentual paradigm in early English metrics.7

The Quantitative Paradigm: George Hickes’s Thesaurus (1703–1705) Anglo-Saxon scribes wrote their vernacular poetry in continuous format, without line breaks, a presentation regime that created basic challenges for early scholars, and continues, on occasion, to do so today. The first modern scholar to recognize Old English poetry as poetry was probably Francis Junius, who, in 1650, undertook to edit the biblical paraphrases contained in the manuscript that now bears the shelfmark Junius 11 in the Bodleian Library.8 Junius recognized that he was dealing with poetry. In this, he was aided by his manuscript source, which is distinguished among the major codices of Old English poetry by its uniquely persistent metrical punctuation.9 The scribe wrote in continuous format, as always, but he frequently entered a raised point between half-lines, thus lending visual perspicuity to the basic unit of metrical composition. Junius extended this markup regime in his 1655 edition, adding some points omitted by the scribe. He similarly marked off half-line units in his transcription of poems in the prosimetrical version of the Old English Consolation of Philosophy; these transcriptions formed the basis for Christopher Rawlinson’s 1698 edition, where, for the first time, Old English poetry was lineated.10 Individual verses (that is, what we now recognize as half-line units) were set out by Rawlinson in columns, each on their own line. Once these texts were recognized as verse, the question naturally arose: verse of what kind? At the turn of the eighteenth century, George Hickes affirmed that the earliest English poetry was quantitative, constructed on the same principles as classical Greek and Latin verse: Imo non dubito, quin in Anglo-Saxonum carminibus, omnes illi pedes … & metrica ratio, prorsus ut in Pindaricis, perfecte explicari possent, si modo syllabarum quantitatem sciremus; cujus unius ignorantia obstat, quo minus Anglo-Saxonicae poesios secreta, qua metrica, quid si dicam, qua lyrica, aperire possimus.11

The Quantitative Paradigm


[Indeed, I do not doubt that all the feet in Anglo-Saxon poems … could be explained fully, and the metrical system as well – just as with the Pindarics – if only we knew the quantities of syllables. Ignorance of this one thing prevents us from uncovering the metrical and even (if I may say so) the lyrical secrets of Anglo-Saxon poetry.]

In Hickes’s estimation, all that lay between modern scholars and a complete knowledge of Old English meter was the identification of vowel quantities. No doubt, this position was motivated in part by extraneous and unanalyzed commitments: admiration of the Greek and Latin poetry on one hand, and undisguised distaste for modern English verse on the other. The chapter “De Poetica Anglo-Saxonum” is sprinkled with deprecations aimed at seventeenth-century English poetry, passages in which Hickes identified himself as a kindred spirit to the Elizabethan quantitative reformers.12 Yet, unlike that earlier generation of classicizers, Hickes held out no hope that quantities could be made a principle of modern English versification. On the contrary, his comparisons between modern and ancient English were in each case intended to show that the very features that distinguish Old English from the contemporary language also align it with the classical languages known to have quantitative systems of versification. Hickes was more dependent than Junius had been on graphic signaling of verse boundaries: where metrical punctuation was absent, he often failed to recognize a metrical boundary. Nevertheless, this much was clear: whereas modern English poetry tended to have lines with equal numbers of syllables, the verse units identified by Junius were syllabically unequal. Even if the syllabic values of double vowels and final -e were unknown, Hickes reasoned, those two sources of uncertainty could account for only a fraction of the observed variation. In the classical Greek and Latin verse, variations in syllable count were a function of a line’s foot-structure (for example, the alternation of hexameter and pentameter lines in Latin elegiac verse) and, second, of metrical resolution (that is, the possibility of putting two short syllables in the place of a long one). Hickes reasoned that the same factors must be responsible for observed variation in Old English. To this argument from syllable count, Hickes adduced three other inferential arguments that Old English poetry was probably quantitative. The Old English verse lacked end-rhyme; its syntax was characterized by bold inversions of word order; and its lexis, Hickes affirmed, was distinctively polysyllabic. By each of these criteria, Old English poetry grouped with the quantitative poetry of classical Greek and Latin, as against modern English poetry. Hickes also ventured a small number of scansions. He proposed to


The Accentual Paradigm in Early English Metrics

scan certain four- and six-syllable verses as spondaic. The following verse is an example. For the sake of fidelity to the state of knowledge in this early scholarship, I omit the macrons now used to distinguish long vowels; they will be reintroduced later in my exposition. (1) Þa man his riht tobræc when someone violated his right

Death of Edgar 23b, ASPR vi, p.23.13

Hickes did not indicate vowel lengths typographically and did not explain his scansions. Nevertheless, a tentative reconstruction of his reasoning is possible. Since he classified (1) as spondaic, he presumably assigned a long vowel to Þa (correctly, as it happens). According to Latin quantitative prosody, the third syllable of the line would be long by position: his is followed by an initial consonant. If Hickes believed Old English h had the prosodic function of other consonants, then the second syllable of the line could also be scanned long by position. This would be a departure from Latin prosody, but Hickes had earlier conjectured that h might have been pronounced harshly.14 The result is three spondees. A few pages after his list of spondaic lines, Hickes supplied another short list, this time of lines that he proposed to scan as adonics (that is, a dactyl followed by a spondee).15 An example is: (2) hæleða waldend ruler of men

Genesis 2139b, ASPR i, p.64.

Here Hickes must have assumed a long vowel in the first syllable of hæleða. We now know that to be incorrect. Regarding the length of the two subsequent vowels in this word, Hickes guessed correctly; they are short. Examples (1) and (2) will serve as recurring points of reference in my exposition below. For now, the important point is that Hickes recognized that such scansions were beyond his or anyone else’s present knowledge of the Anglo-Saxon language. Accordingly, the task he set himself was not to precipitate unknowns into knowns, but instead to gather and present evidence that the unknowns really existed as such – that is, that quantity was present and functional in the poetry. He urged that observed irregularities in syllable count would one day be recognizable as the consequence of as-yet undiscovered quantitative patterns. It was a reasoned hope. Yet it is also possible to hear a note of desperation, particularly in Hickes’s repeated comparisons of Old English verse to the Pindaric. Pindar’s odes served as a model of lyric exuberance in English poetry at the end of the seventeenth century, but his exceptionally complex metrical patterning was not yet understood.16 This fact bears significantly

The Paradigm Shift, 1765–1868


on Hickes’s claims for Old English poetry. The references to Pindar in the chapter “De Poetica Anglo-Saxonum” stand as a tacit acknowledgment that, if Old English poetry really had been quantitative, then it must have been composed of metra at least as complex and various as any such produced by Greco-Roman antiquity. Indeed, comparison to Pindar was almost the only way that Hickes could lend credibility to his claim that Old English verse was, despite its bewildering variety, also the expression of a metrical system.

The Paradigm Shift, 1765–1868 The Thesaurus was published between 1703 and 1705. By the second half of the eighteenth century, Hickes’s metrical classicism could be ignored. Then, at the beginning of the nineteenth century, the quantitative theory was swept definitively aside: reference to quantities was replaced by reference to “emphasis.” The concept was not new. George Gascoigne had set out rules for the patterning of “emphasis,” or accent, in his Certayne Notes of Instruction Concerning the Making of Verse or Rhyme in English (1575).17 In the eighteenth century, “emphasis” had become a key concept in discussion of English prosody, as a search in the Princeton Prosody Archive demonstrates. The concept had not, however, previously been extended to the vernacular poetry of the Anglo-Saxons. As such, it required definition from the ground up. “Emphasis,” wrote Joseph Bosworth in his Elements of Anglo-Saxon Grammar (1823), “is a perceptible stress of the voice laid upon a syllable, or a word.”18 Bosworth identified syllabic emphasis as “the superior energy with which at least, one syllable of a word is enunciated,” and he explained that the emphatic syllables must be separated by “remiss or feeble” ones in speech, for “[s]everal emphatic syllables cannot be conveniently enunciated in succession.” There is a tone of bold linguistic generalization in these pages. “It appears,” Bosworth wrote, “that in language emphasis and remission occur at certain intervals. On these depends rhythm, the vital principle both of speech and song.” In a footnote, he affirmed the conceptual break with classical frameworks: “The Greeks and Romans regulated their verse by the length of syllables … But the AngloSaxons modelled their verse by rhythm or metrical cadence.” Or, again: “in Saxon and in all modern languages of Gothic origin, [emphasis] holds the place of the Roman and Greek quantity.” In these sentences and in the account of Old English poetry that they introduce, Bosworth was communicating the newest philological


The Accentual Paradigm in Early English Metrics

scholarship. His most important sources were less than a decade old. John Grant’s 1813 Grammar of the English Language had described the operations of emphasis and rhythm in spoken language.19 In the same year, John Josias Conybeare had proposed that Old English poetry was organized not by quantity but “emphasis”: verses of Old English, Conybeare ventured, “will be found to consist for the most part of feet of two or three syllables each, having the emphasis on the first.”20 Conybeare would presumably have scanned (2) hæleða waldend as a three-syllable foot followed by a foot of two syllables (Sxx|Sx). The great Danish scholar Rasmus Rask, who was another of Bosworth’s authorities, had gone several steps further in his 1817 grammar of Old English.21 Rask had understood a fundamental point: the alliteration that mattered in early Germanic verse  – the alliteration described in Snorri Sturluson’s thirteenth-century ars poetica  – was located at the onset of stressed syllables.22 By seeing that alliteration and stress were linked, Rask was able to bring the Old English poetry to bear on the question of stressplacement, and – in a second move – bring this sharpened understanding of linguistic stress to bear on the question of Old English meter. Subsequent studies would give many more iterations to this feedback loop between study of language and study of meter. For Rask, the result of his linguistic inferences was twofold. He concluded that Old English verse had precisely two emphatic syllables per verse; he also made a rough division between the types of linguistic material that receive emphasis and those that do not. Where Conybeare had proposed to scan (3) secan and gesittan to seek and inhabit

Phoenix, 671a, ASPR iii, p. 112

with three feet and perfectly alternating rhythm (Sx|Sx|Sx), Rask objected that emphasis would fall only on the root syllables of major words.23 He therefore scanned the verse Sxxx|Sx. “[F]or those who wish not to compose A.S. verses, but merely to analyse such as they meet with, it is easy to determine the metre,” he affirmed; “The chief syllable in each word bears the accent. Compound words, consisting of two independent and, in themselves, significant words, are accented on the first.”24 In scholarship of the early nineteenth century one can sense the sparkle of discovery – of recognition that one has made a break into a new way of formulating a problem. Critical notices of Hickes were exemplary of this new spirit. Rask, the pioneering comparatist, scolded Hickes for having “possessed so little of the spirit of discovery.”25 Conybeare stated that the great and learned author of the Thesaurus “appears perhaps no where to so little advantage

The Paradigm Shift, 1765–1868


as in the pages which he has dedicated to this topic [that is, meter].”26 Yet, in estimating their accomplishment, these scholars looked as much to the intervening era as to Hickes himself, and their progress stood out sharply. During the second half of the eighteenth century, as Hickes’s classicism lost support, Old English poetry had come to appear hopelessly disorganized. In an essay accompanying his 1775 edition of Chaucer, Thomas Tyrwhitt had claimed to discern nothing resembling metrical organization in English prior to the Norman conquest.27 Tyrwhitt averred that the English owed meter, as such, to their Norman conquerors. George Ellis and Sharon Turner agreed. Ellis reported in 1801 that the “mechanism and scheme of versification [of Saxon poetry], notwithstanding all the pains which Hickes has employed in attempting to investigate them, are still completely inexplicable.”28 Turner, meanwhile, commended the unshaped roughness of Anglo-Saxon effusions. In this “rude and barbaric state” of English poetry, he discerned a true expression of the culture’s primitive energy.29 During this same interval, early Germanic poetry was christened “alliterative.” The historical irony should be apparent: early Germanic poetry acquired its Latinate denomination during precisely the interval when the Latin sciences of language were, in all other respects, losing their authority to expound it. Despite their opposition in all other respects, the outgoing quantitative paradigm and the emergent accentual one would at least agree that Thomas Percy’s “alliterative species of versification” designated a metrical system by a linguistic phenomenon that could not per se constitute or organize a meter. Hickes, one of Percy’s sources, had been clear about this: he recognized the alliteration in Old English poetry and considered it important, but as a rhetorical ornament, not an element of meter. He treated alliteration at the head of his section on “incidentals,” where it is followed by rhyme, metaphor, metonymy, apposition, and so forth, down the line of classical figures and tropes.30 Conybeare agreed. In the same paper in which he rejected Hickes’s quantitative theory, Conybeare stated that “alliteration (which indeed requires but a short notice) will be more conveniently treated of after we shall have ascertained the existence and nature of that metre of which it forms the chief ornament.”31 The importance accorded to alliteration in Percy’s “Essay on the Metre of Pierce the Plowman’s Visions” was therefore an aberration. It emerged during an interval when scholarship lacked a sure approach to early English poetry: the classical sciences of language had already lost their authority, yet the first articulations of a new paradigm still lay several decades in the future. Tyrwhitt’s confession that he could find no metrical organization in


The Accentual Paradigm in Early English Metrics

Old English poetry was one symptom of an epistemic interregnum. Percy’s christening of it as “alliterative” was another. For Conybeare, who looked back on this period from the standpoint of new conviction, the position that most needed answering was not Percy’s elevation of alliteration but Tyrwhitt’s skepticism. The unsupportable classicizing “extravagance” of Hickes had driven the esteemed editor of Chaucer into “the opposite extreme.”32 Between the one position and the other there was an unfortunate symmetry. The concept of “emphasis” or “accent” broke this binary deadlock by projecting a third position beyond both the thesis of classical quantities and the absolute skepticism engendered by it.33 Tyrwhitt had claimed he was “unable to discover any material distinction of the Saxon Poetry from Prose, except a greater pomp of diction, and a more stately kind of march.”34 That “more stately kind of march” would now be recognized as accentual rhythm. What had previously appeared as no principle at all was now apprehended as the expression of a different kind of metrical system – or, rather, as the same kind of metrical system employed in all later English poetry. Conybeare had brought the oldest English poetry into agreement with understandings of the modern language and its prosody. Edwin Guest’s genius was to perceive this new epistemic configuration and give it a material body. His two-volume History of English Rhythms (1838) made accentual rhythm the uniform organizing principle of English poetry, from Cædmon to the present.35 It remained for Walter Skeat to consolidate the previous half-century’s advances by supplying the new research paradigm with its own terminology: Nothing has more tended to obscure the rules and laws of English prosody, than the absurd and mischievously false terminology that has been made use of in discussing it. Whilst it is pretty clear that it is based on quite a different system from the Latin and Greek metres – on an accentual, that is, not on a temporal system – we have attempted to explain its peculiarities by terms borrowed from the Latin and Greek, such as trochees, dactyls, &c., and we make perpetual use of the words long and short.36

Skeat had recognized a signal ambiguity in previous scholarship. Apart from Guest, each of the pioneers of the accentual paradigm had retained the classical names of feet, even as they broke with the classical schema to which these names belonged. Skeat hoped to clear this residual confusion away once and for all. Maintaining that “the whole terminology of English prosody, if it is not to be misleading and fruitful in all kinds of errors, has yet to be invented,” he set himself the task of terminological invention.37 He proposed that one should speak not of long and short syllables, but

Elaborations and Challenges


loud and soft ones. He retained the concept of a metrical foot, but elaborated a new terminology in which the name of each foot instanced the stress contour it named. A single loud syllable would be called a “tone”; a loud followed by a single soft would be called a “tonic”; a loud followed by two soft syllables a “dominant”; and so forth.38 Skeat’s musicological terminology should not obscure his accomplishment, which was to crystallize a sequence of metrical theory begun 60 years previously. Whenever later prosodists have described English alliterative verse as “accentual” or “strong stress,” they simply report the state of knowledge encapsulated in this essay of 1868.39

Elaborations and Challenges The scansions adduced by Conybeare, Bosworth, Rask, Guest, and Skeat have long since ceased to be relevant. Yet the paradigm established by these scholars between 1814 and 1868 has a history that extends to the present. The most important features of the subsequent history may all be traced to the separation of medieval English into “Old” and “Middle.”40 For Skeat, the pre- and post-Conquest poetries were analyzable in precisely the same terms, composed of the same types of feet and organized by the same accentual principles. Soon afterwards, this unity would begin to pull apart; the accentual paradigm would develop along divergent trajectories in Old and Middle English metrics. Of these two trajectories, the one in Middle English metrics has proceeded more haltingly, but with fewer complications. At the end of the nineteenth century, Karl Luick described the accentual patterning of the late Middle English alliterative b-verse almost exactly as Thomas Cable and Hoyt N. Duggan would 90 years later. Fourteenth-century alliterative poets counted syllables in the second half of the line. There are two basic patterns: type 1 (x)Sx…xSx type 2 x…xS(x)Sx These two b-verse patterns are a central component of all recent descriptions of Middle English alliterative verse, including the one I made in the Introduction to this book. We can now place them within a history of thought about English meter. No less than Skeat’s “Essay on Alliterative Poetry,” the Cable–Duggan b-verse patterns grasp English meter as a function of the contrast between stress and unstress; the difference is a more precise template for arrangement of unstressed syllables, coupled with a


The Accentual Paradigm in Early English Metrics

rejection of Skeat’s metrical feet. With the work of Cable, Duggan, and subsequent researchers, we see a long-delayed elaboration of the accentual paradigm in Middle English alliterative metrics. Several factors, to be discussed at the end of this chapter, prevent one from characterizing this victory as absolute. In the field of Old English meter, the accentual paradigm has been more steadily productive, sustaining continuous research activity throughout the past two centuries, but also more persistently compromised. Irregularities and extraneous factors have proven unamenable to reduction; those irregularities present from the outset have been joined by others that have emerged only gradually. A history of this research activity would note the work of Karl Lachmann and his school in the mid nineteenth century, as Kühnel does, and the challenges posed to Sieversian metrics in the twentieth century by Andreas Heusler and John C. Pope. Rather than provide a chronological account, I simply sketch three salient problems in the accentual paradigm. The first is clashing stress. A basic premise of the accentual paradigm is that stressed syllables should either have differentiated prominence or be separated by an unstressed element. Nevertheless, scholarship of the nineteenth century quietly admitted the existence of verses that violate this rule. Conybeare scanned (4) tir welgade glory abounded

Riming Poem, 34b, ASPR iii, p. 167.

with stress on tir and the initial syllable of welgade.41 The chief problem, as Conybeare saw it, was that this verse and others like it has just one complete foot, preceded by what he called a “syllable extraordinary.” He proposed that the “emphasis [of the syllable extraordinary] might be so strongly marked, as to render it equivalent to two,” and Bosworth followed him in this.42 As the accentual system of Old English came to be better understood, scholars inevitably became more sharply conscious of clashing stress. Clashing stress followed, as a matter of course, from Skeat’s proposal to recognize stressed monosyllables (“tones”) in his system of scansion. At the end of the century, Sievers would classify verses like the one above as Type D, scanning S|Ssx, with secondary stress on the medial syllable of welgade.43 The Five-Type System formalized clashing stress as a central and prominent feature of Old English meter. Clashing stress has retained this status ever since, though not without provoking nagging doubts and ingenious efforts at obviation. Bosworth had been confident that stressed syllables

Elaborations and Challenges


must be separated. On the basis of vastly more sophisticated demonstrations, twentieth-century intonational phonology likewise affirms that the English language has a powerful preference for alternating accentual rhythm. “When two fully stressed syllables occur in succession,” Cable writes, “the rhythmical pattern runs counter to an idealized norm.”44 In Cable’s account, clashing stress should be understood as an artifact of the Sieversian abstraction, not a feature of the meter itself: the task of metrical theory is then to develop a more accurate abstract representation of the meter, one in which stress clash does not appear as such. To this end, Cable has proposed that Old English meter differentiated between four levels of stress and that it included compensatory temporal spacing – or pauses – as a metrical feature.45 These theoretical proposals respond to and thus flag a deep problem in the conceptual architecture of Old English metrics: stress clash is at once a result of research within the accentual paradigm and a violation of its most basic presuppositions. The problem, it should be noted, exists in Middle English alliterative verse as well. A  solution can only emerge after considering the other two challenges within Old English metrics. The second challenge is this:  quantity never fully disappeared. From the time of Jacob Grimm’s reconstructions of the phonology of medieval Germanic languages, it was evident that vowels in open syllables bearing primary word-stress were not invariably long.46 In the 1830s, John Mitchell Kemble championed Grimm’s reconstructions of vowel length, against the vociferous objections of the English Saxonists.47 The discrimination between short and long vowels had an immediate value in the fields of etymology and lexicography, where vocalic quantity served to distinguish words identical in spelling. Yet it was not immediately clear how, or whether, the reconstructed vowel system of medieval Germanic languages was relevant to verse design. Grimm himself had expressed doubt on this question. Skeat’s subsequent efforts to banish quantitative terminology were founded in the belief that quantity was entirely irrelevant to early English meter. Nevertheless, even prior to Skeat’s essay, the metrical effects of syllabic length were being noticed and described in Old High German and Old Norse.48 These early Germanic meters appeared to treat a short stressed syllable plus the following syllable as a unit equivalent to a long stressed syllable. The phenomenon was termed “resolution.” By the end of the nineteenth century, resolution would be identified in Old English verse as well: it was an integral component of Sievers’s system.49 So presented, metrical resolution in early Germanic verse differs sharply from the phenomenon of the same name in Greek and Latin quantitative verse, at


The Accentual Paradigm in Early English Metrics

least as presented in handbooks.50 Nor does resolution accord easily with stress-based conceptions of Old English meter.51 Skeat would have scanned (2) hæleða as a dominant (Sxx), which may be a good representation of stress contour. By the end of the nineteenth century, however, the root vowel was recognized to be short and the word was usually scanned with resolved stress, assigning the first two syllables to a single metrical position. It was clear that Old English poetry did not owe its metrical organization to accentual contour alone. Quantity had returned, forming a second irregularity within the accentual paradigm. The third irregularity is the metrical behavior of unstressed prefixes. Although this problem has been recognized only in the last several decades, its slow emergence may be traced across almost the entire history of modern metrical study, under the various names Maalfylding, complement, Auftakt, and anacrusis. We have seen that Rask understood verse rhythm to begin with the first stressed syllable; all syllables prior to the first stress could then be understood, without great difficulty, as a sort of extrametrical prologue.52 In a verse like (1) Þā man his riht tōbræc, Rask would have seen an extrametrical prologue of three syllables; the verse proper would begin with riht and would scan SxS. Within Sievers’s system, most verse-initial weak syllables were redefined as integral components of verse rhythm. Sievers Types B and C each begin with one or more unstressed syllables; example (1) is a Sievers Type B verse with a three-syllable initial dip (xxxSxS). Nevertheless, Sievers’s classificatory system left a residue of initial unstressed syllables that could not be accounted for in this way. For such syllables, Sievers retained both the name Auftakt and the rationalization that this linguistic material fell outside the metered portion of the utterance. This feature is illustrated by verses such as those below. I use “)” to set off syllables deemed anacrustic; the sequence “Sr” indicates resolved stress: (5) ārās þā se rīca   the powerful (man) arose then


Beowulf 399a

(6) genered wið nīðe redeemed from violence


Beowulf 827a

(7) ne frīn þū æfter sǣlum do not ask about happiness


Beowulf 1322a

Each of these verses belongs to Sievers Type A.  Each is understood in Sieversian metrics as having an anacrustic syllable. This was where research remained until it was noticed, first, that the syllables in anacrusis tended to be unstressed prefixes or the negative proclitic ne and, second, that these

A Second Paradigm Shift


morphemes exhibit similar behavior interior to the verse.53 This second point is perhaps illustrated most readily by verses traditionally classified as Type A with anacrusis, but which – unlike (5–7) above – have neither an unstressed prefix nor ne in initial position. An example is: (8) swā sǣ bebūgeð as the sea encompasses


Beowulf 1223b

If swā is regarded as anacrustic, this verse forms an exception to the morphological pattern illustrated by (5–7). Example (8) also contains an unstressed prefix, however: be-, between the two stresses. If the rule that operates in (5–7) is not “weak syllables preceding the first stress may optionally be excluded from the metrical count” but instead “unstressed prefixes (and ne) may optionally be excluded from the metrical count,” then (8) may be reclassified as Sievers Type C without clashing stress. Termed the “prefix license” by Yakovlev, this metrical feature is probably the greatest challenge to the accentual paradigm. “It is very hard,” Yakovlev observes, “to understand how a syllable may be completely omitted from the metrical count [in an accentual meter]: the prominence of the syllable may be low, but it will still be part of the intonational contour.”54 Yakovlev’s 2008 thesis may mark the point at which the accentual paradigm will have tipped over into another way of comprehending Old English meter. He calls the meter not “accentual” but “morphological.”55

A Second Paradigm Shift: Yakovlev’s Morphological Theory Stress clash, resolution, and the prefix license have the status of conceptual embarrassments for the accentual paradigm of English metrics: they are smudges in the picture, elements irreconcilable to the structure from which they nevertheless cannot be removed. Yakovlev’s theory does not remove the smudges; instead, it projects them into another conceptual space. The phenomena that always resisted rationalization within stressbased conceptions of Old English meter now reveal themselves as elements of another system. According to Yakovlev, the standard verse form of Beowulf owed its metricality neither to a fixed count of “major” stresses, nor to a minimal set of recurring accentual contours, but instead to a relatively simple count of four metrical positions per verse. Given this point of departure, the success or failure of Yakovlev’s theory rests largely on its ability to define a metrical position and show how the words and syllables that make up actual verses of Old English map onto a basic four-position frame. For this task he is


The Accentual Paradigm in Early English Metrics

able to draw on a deep file of prior scholarship. For, if his theory is directly contrary to the general assumption among metrists since before the time of Sievers, it is also a direct development of the single most important contribution to this modern tradition, that of Sievers himself. Here we must take a closer look at Sievers’s accomplishment. He demonstrated that most verses of Old English conform to one of five basic rhythmical patterns. Since their original presentation, the Five Types have always been illustrated by their simplest realizations (examples are from Beowulf): A B C D E

Sx|Sx: gomban gyldan, “to pay tribute” (11a) xS|xS: hīe wyrd forswēop, “events swept them” (477b) xS|Sx: oft Scyld Scēfing, “Scyld, son of Scef, often” (4a) S|Ssx: wīs wēlþungen, “wise, accomplished” (1927a) Ssx|S: glēomannes gyd, “musician’s song” (1160a)

The “Five-Type System” has been Sievers’s most influential legacy. It is the aspect of his theory that most nearly supports an understanding of the metrical system as accentual, stress-based, and thus homologous with later English meters. Yet the Five Types may not be the most important aspect of Sievers’s theory, or even the most reliable one. As Sievers also noted, the stress-based patterns he identified each unfold within a frame of four metrical positions. He termed these Glieder, or “members”; in the simplest realizations, each Glied is realized by a single syllable.56 This is the basis from which Yakovlev builds his theory. Prior to Yakovlev, metrists in the Sieversian tradition have always moved from the definition of four positions to definition of the stress-based rhythms or prosodic contours that form across them. Indeed, the great bulk of work within Sieversian metrics has been directed into cataloging and tabulating the various unique combinations of long and short, stressed, half-stressed, and unstressed syllables that occur in Old English verse – the Five Types and their forest of sub-types. Within this research program, the notion of the metrical Glied withdraws into the background; to the extent that it maintains a continuous presence, it is just a unit of verse possessed of a certain level of stress and which, in sequence with the other Glieder, forms the overall stress contour of the verse. Scansion aims to represent this contour. From here, it is just a short step to dispense with the metrical abstraction of Glieder altogether and treat the syllable as the definitive constituent. A.  J. Bliss, who took this step, identified 130 unique contours among the 6,342 non-extended verses of Beowulf.57 His alphanumeric notation is almost as intricate as the system of references to articles of the Summa theologica. (The first four verses of Beowulf are logged as types d3b,

A Second Paradigm Shift


d3a, 1D1, and 1A1a, respectively.) Against this taxonomic impulse, Cable has always urged the principle of metrical simplicity. In important contributions that form key precursors to Yakovlev’s theory, Cable emphasizes that the meter’s “general principle” is its frame of four positions; he also demonstrates that the Five Types are just the contours that may occur, provided a small number of additional restrictions, within this four-position frame.58 Yet even Cable turns from the definition of metrical positions to the definition of contours. By rejecting Sievers’s notion of metrical foot and challenging the notion of clashing stress, Cable has sought to bring the Five Types into closer agreement with intonational phonology. From one end of the Sieversian tradition to the other, researchers have interpreted their central task to be the correct description of stress-based prosodic contours. Yakovlev’s insight is that the stress contours may be epiphenomenal. Stress accent establishes the place of alliteration within the line, but it does not follow that the meter itself is stress-based, for alliteration may be only a regularized highlighting of stress peaks that are themselves the output of metrical organization located in a deeper linguistic layer. To explain how linguistic material maps onto the basic four-position frame, Yakovlev posits that the meter distinguished between three types of morpho-syllabic constituent.59 One class of morphemes contributes strong metrical positions, a second contributes either a strong or a weak metrical position, and the third, either a weak position or none. The third class consists of unstressed prefixes and ne: these morphemes may optionally go uncounted by the meter; when they are counted, they count as weak. The first and second classes are as one would expect from previous theories of Old English meter in the tradition of Sievers, except that Yakovlev dispenses with the distinction between primary, secondary, and tertiary stress. A morpheme with any of these three grades of stress forms a strong position; accordingly, the first class consists of “(the long syllables or resolved sequences of) roots, suffixes and stressed prefixes of open-class words, excluding finite lexical verbs.” The second class of morphemes consists of “inflections, unstressed prefixes, finite lexical verbs, and closed-class words.”60 Ambivalence in the function of the second morpheme class (“either a strong or a weak metrical position”) is due to well-recognized phenomena of contextual promotion. When placed at the right edge of the verse, the root syllable of a finite lexical verb or closed-class word is promoted to contribute a strong position.61 Even the -i- and -od- of class 2 weak verbs (e.g., þancian, pret. 3 sg. þancode) make position independently in the coda.62 A different kind of metrical promotion evidently occurs verseinitially, in the absence of a word from the first class in the first half of


The Accentual Paradigm in Early English Metrics

the verse.63 The promotability of morphemes – that is, the claim that an identical morpheme may contribute a weak position in one metrical context and a strong position in another – is simply carried over by Yakovlev from Sieversian metrics and should not be controversial. His innovation is instead to reduce the three (or four) stress grades into a binary distinction between strong and weak. Nor is Yakovlev’s simplification of metrical stress wholly unprecedented. It receives support from Fulk, who urges that researchers have not distinguished sharply enough between metrical and phonological stress in Old English. Primary, secondary, and tertiary stress are necessary and justified in description of Old English phonology; at the metrical level, however, “it appears to be possible to simplify the description and reduce the number of levels of stress required, perhaps even to two – that is, stress and no stress.”64 In Fulk’s assessment, “ictus at the tertiary level apparently amounts to syllable length” in combination with metrical context, and may therefore be described without reference to stress; meanwhile, “[t]he terms primary and secondary as applied to stress are convenient, but the distinction seems to have little to do with metrical realities,” for “secondary stress can be distinguished from primary on a purely positional basis: it is any full stress that immediately follows another full stress.”65 Fulk’s analysis here is based in the principle of simplicity: the distinction between primary and secondary stress is derivable from general principles of Old English phonology; accordingly, it does not need to be repeated in a description of the Old English meter, for which a simple binary distinction will suffice. Fulk does not pursue this line of thinking, but it supports his suggestion, quoted at the beginning of this chapter, that the role of stress in Old English meter has been overestimated. Indeed, some sort of theoretical adjustment quickly becomes necessary, for the simplification that Fulk and Yakovlev propose has the following unavoidable consequence: the number of metrical constituents designated as “stressed” increases massively. Consider, for example, the following two verses: (9) lēof lēodcyning beloved king of the people (10) glædman Hrōðgār gracious Hrothgar

Beowulf 54a Beowulf 367b

In a traditional Sieversian scansion, these verses are recognized as having two major stresses each:  (9) is registered as Type D, S|Ssx, with suspension of resolution in –cyning; (10) is registered as a variant of Type A, Sx|Sx, with secondary stress in the dips. In the simplified system hinted

A Second Paradigm Shift


at by Fulk and elaborated by Yakovlev, the scansion of (9) and (10) is, by contrast, SSSS in both cases. That improbable scansion – four consecutive “stresses” – is the direct consequence of treating primary, secondary, and tertiary stress as metrically undifferentiated: the tertiary stress of the suffix –ing, and the secondary stresses of the second elements in the compounds lēodcyning, glædman, and Hrōðgār are all counted as “stresses” in the simplified Fulk–Yakovlev system. Yet, the scansion SSSS is only unacceptable as long as we conceive of the Old English meter as stress-based, that is, a system keyed to contours of stress. The Sieversian scansions of (9) and (10) deliver reasonably accurate representations of the stress contours of these verses. The Yakovlevian scansion has a different representational target: it aims to represent metrical structure. The shift in representational target entails a shift in the meaning of scansion symbols; hence my scare quotes around “stresses” in the previous paragraph. In Yakovlev’s system, the symbols “S” and “x” should be taken to designate strong and weak metrical constituents, respectively, not stressed and unstressed syllables. Here we return to Sievers’s concept of the Glied, his most significant contribution to Old English metrics. Verses consist of four Glieder, or metrical constituents, in any combination of strong and weak. To this, one must add that weak syllables adjacent to one another always combine to form a single Glied. There are accordingly eight possible permutations of strong and weak Glieder: SxSx, xSxS, xSSx, SSSx, SSxS, SxSS, xSSS, and SSSS. Strong positions are always formed by a single syllable or its resolved equivalent. Weak positions in the first half of the verse are expandable, but those in the second half of the verse (that is, the third or the fourth position) are formed by a single syllable. Where a dip in the third or fourth position is polysyllabic, it is usually reducible by the prefix license.66 If the notation “(…)” is used for optional expansion of a dip, then the classificatory component of Yakovlev’s theory may be illustrated as follows.67 Examples are from Beowulf. Where I give two examples, the second has a prefix uncounted by the meter: I. II. III. IV. V. VI. VII. VIII.

Sx(…)Sx x(…)SxS x(…)SSx SSSx SSxS Sx(…)SS x(…)SSS SSSS

11a “gomban gyldan”; 1322a “ne frīn þū æfter sǣlum” 477b “hīe wyrd forswēop”; 74a “Ðā ic wīde gefrægn” 13b “þone God sende”; 1223b “swā sǣ bebūgeð” 1927a “wīs wēlþungen”; 2930a “ābrēot brimwīsan” 1160a “glēomannes gyd”; 5b “meodosetla oftēah” 17a “wuldres wealdend”; 217a “Gewāt þā ofer wǣgholm” 4a “oft Scyld Scēfing” 54a “lēof lēodcyning”


The Accentual Paradigm in Early English Metrics

These permutations cut diagonally across the Sieversian Five Types:  the members of Yakovlev’s first category are invariably Type A, but Type A verses with secondary stress belong to Yakovlev’s categories IV, VI, and VIII. The decisive factor is not the stress contour, but rather the quality of the linguistic material contributing each successive metrical position. The phenomena documented by accentual metrics are assimilated, transposed, and revalued  – in a word, aufgehoben – within a non-accentual theory of meter. Yakovlev lays out refinements to this system in his thesis, where he is careful to note that the accentual principle is not absent. The phenomena of alliterative patterning and contextual promotion show that “the accentual principle, with its attention to the general context of a phrase … is present in the background of the metrical system.”68 This fact has important implications for later stages of the alliterative meter. Another set of diachronic implications may follow from Yakovlev’s interpretation of persistent asystematic patterns, of which five-position verses (Sievers Types A* and D*) are the most significant. These he describes as the “historical residue” of a prior metrical configuration:  “a traditional metre,” he writes, “is hardly ever given opportunity to become completely cohesive. The average time span between major prosodic upheavals appears to be less than that required to eliminate any remains of the previous restructuring.”69 This observation and Yakovlev’s accompanying analysis may prove to be of considerable interest for metrical paleontology, that is, reconstruction of pre-historical states of Old English meter. I  limit myself, however, to two points at the close of this present chapter. One concerns historical evaluation of the accentual paradigm. The other concerns implications of Yakovlev’s analysis for later stages of the alliterative meter in English.

The Accentual Paradigm: Retrospect and Prospect Students are right to be perplexed. The Old English meter is unlike anything that has followed it in this language; the profound differences from later English meters could only ever be grasped incompletely, and confusedly, within the accentual paradigm, which did its best to ignore them. The accentual paradigm originated as an effort to theorize meter as a direct, unmediated expression of language. Stress accent was what English could justly claim as its own, and this, as the inner genius of the language, was the truth expressed in its meter. Thus stated, the accentual paradigm is a recognizably nineteenth-century

The Accentual Paradigm


formulation, exhibiting the structures of thought characteristic of that period:  the inner natures of things may be ontologically withdrawn, but the objects given to us in experience are nevertheless related to those inner natures as their expression.70 In an effort to grasp the genius of English prosody, scholarship of the early nineteenth century rejected the normative force of Greco-Roman metrics and attempted to comprehend all English verse, as such, under a uniform principle of accentual rhythm. Researchers sought to bracket, as inessential, both the immanent vagaries of history and the spectral exteriority of norms. Against those purifying moves, one should acknowledge that metrical systems are inherently normative and that history is inherently digressive. Meter is not tidied-up speech and it cannot be deduced from linguistic phonology alone, for it has its own aesthetic norms and is shaped by its own historical course. More specifically, the presence of stress accent in a language does not necessitate that its meter(s) be stress-based. Roman Jakobson’s justly famous formulation, to the effect that metrical prosody depends on linguistic prosody, does not allow us to deduce the former from the latter, for a system of versification only ever imposes its definitive equivalences upon a limited subset of the prosodic attributes available in the language. “If the system of versification is an unknown X and we are given nothing more than the prosodic features of the language,” Jakobson wrote, “we obtain an indeterminate equation, that is, the possibility that X could have more than one value”; though “the array of conceivable solutions” will always be limited, any actual system of versification represents only a single solution to an equation that might have been solved in another way, by fixing upon a different bundle of prosodic attributes.71 The case of classical Latin and its medieval reception may be instructive here. Though classical Latin probably had a stress accent, this did not prevent poets from writing in quantitative meters.72 When vocalic quantities were leveled in late antiquity, quantitative meters were sustained in the schools; simultaneously, new metrical systems were developed, keyed to the word-stresses that, though always present in the language, had played a secondary role in the classical quantitative meters. A  prosodic feature subordinate within the quantitative meters was now established as the basis of a new metrical system. Indeed, some of the new stressbased meters had their origin in the “misreading” of quantitative forms, attending to the patterning of stresses rather than the patterning of syllabic quantity.73 Following these medieval readers, it would be possible to make an accentual scansion of the Aeneid, marking the primary and,


The Accentual Paradigm in Early English Metrics

where relevant, secondary accents of each prosodic word in each line of Virgil’s poem:74 árma uirúmque cáno, tróiae qui prímus ab óris itáliam fáto prófugus làuiniáque uénit lítora.

This, it seems, is approximately what Bliss did for Beowulf: he produced an exacting record of a linguistic feature that is certainly present in the verse, and not without interest, but is epiphenomenal to the meter. In the Aeneid and other classical Latin verse, the organizing feature was syllabic length, matched to metrical positions designated by grammarians as feet; there was also some attention to word boundaries. In Beowulf and other Old English verse, the relevant features were syllabic length and morphological category membership, matched to metrical positions designated by Sievers as Glieder; there were also template-driven phenomena of metrical promotion, operative at key points in the line. The effects of stress contour may be perceived in various aspects of the Old English meter, and in Latin quantitative meters as well, but stress was only a subsidiary consideration in Latin meters of the classical period, and may not have been primary in Old English, either. The accentual germ  – whose presence in the Old English meter is already evident – gradually became the dominant principle. A morphological meter developed gradually into an accentual one, but traces of the morphological precursor persisted within the accentual successor, as embedded reflexes. The result is that some features of even the Middle English alliterative meter remain beyond the grasp of the accentual paradigm. Yakovlev’s historical perspective suggests solutions to several longstanding problems – or, better, it shows that the problems are problems in re, in the Middle English meter itself, and not merely in the modern understanding of it. The meter was negotiating its morphological inheritance. That inheritance will be investigated in the next three chapters of this book, but several points may be made at this juncture, as a caution against overextensions of English accentual-syllabic prosody. Recent work on the Middle English b-verse can seem to represent a historical vindication of the accentual paradigm: by defining meter in terms of stress contours, the accentual paradigm always implied that unstressed syllables ought somehow to be regulated, and Cable and Duggan showed that this was indeed the case in the second half of the Middle English line. That triumph might lead one to think that other principles of English accentual-syllabic metrics – the principle of a fixed stress count, for example, or the monosyllable

The Accentual Paradigm


stress rule – should be applicable to fourteenth-century alliterative verse. Yet the situation is considerably more complex. Consider, again, clashing stress. It is a rare and marked realization of accentual-syllabic meters, and runs counter to ordinary rhythms of speech articulation, but it occurs frequently in Middle English alliterative verse. It is, moreover, one of the normal realizations of the b-verse: (11) þer he bock radde where he read books


Brut 5b

(12) or of briȝte syluer


PPl.Bx Prol.168b.75

Why did an accentual meter make such persistent use of this difficult contour? Metricists have been right to flag the problem. For Yakovlev, however, the answer is simple: clashing stress is present within the Middle English meter as a reflex of its historical precursor.76 The morphological meter specified only arrangements of strong and weak elements, allowing the contour to take shape within those constraints. Strong positions frequently adjoined one another; when they did, the result was often stress clash. As the accentual principle became stronger in the early Middle English period, patterns with stress clash would come to appear increasingly marked. They nevertheless remained part of the rhythmical repertoire. Ambiguities in stress count may be explained along similar lines.77 The morphological meter permitted a minimum of two and maximum of four strong constituents per verse. Heavy verses – that is, those with three or four strong constituents – would usually have only two stress peaks, as is manifestly the case in (9) and (10). Nevertheless, it seems likely that a “strong constituent” came to be interpretable as a stress during the transitional period, thereby establishing three-lift verses as an acceptable realization of the meter in Middle English. This speculation is at odds with current understandings of Old English heavy verses and their fate, but the matter merits reassessment. It is well established that the frequencies of Sievers Types D and E are sharply reduced in late Old English verse; since Oakden, progressive disappearance of Types D and E from the metrical repertoire has been tied to the decline in compounding.78 Yet this line of reasoning may give too much credit to changes in poetic lexicon, and too little to the metrical system. Against the received view, one might suppose that Sievers Types D and E were deselected because they afforded no opportunity for a polysyllabic dip, which had become a requirement of the meter in the Middle English period. If the long dip requirement eliminated Sievers Types D and E from the repertoire, heavy verses (of which Types D and E are the most prominent representatives in the Sieversian notation) might be expected to


The Accentual Paradigm in Early English Metrics

survive in one form or another, for they were a central component of the metrical system. The relaxation of the four-position rule was decisive in this regard. The alliterative meter became more accommodating to unstressed syllables, which were in any case proliferating under the pressure of analytic syntax. The new infusion of unstressed syllables spaced out strong positions, while the shift from morphological to accentual organization meant that each strong position was increasingly supplied by the lexical stress of an independent word. The result was three-lift verses. Three-lift verses in Middle English are probably best understood as a reflex of an earlier configuration of the metrical system, one in which the meter did not count stresses, and permitted as many as four strong constituents per verse. In the transition from a morphological meter to an accentual one, the distinction between grammatical word classes remained largely intact: the Middle English meter derived metrical stresses, or lifts, from the same categories of words that formed strong positions in the Old English meter. What changed in the later period was just that the metrical system became keyed to the lexical stresses of these words. Polysyllabic nouns, adjectives, and participles (words that may have contributed more than one strong position in Old English) now typically contributed only a single metrical stress, on the syllable bearing primary lexical stress. This transition to an accentual system was a step towards the accentual-syllabic prosody of later English verse, but only a single step in that direction. The next steps will be traced in the next three chapters of this book, and their conclusion was the breakdown of the alliterative meter as an independent mode of versification in English. These are difficult problems, and my presentation of them is compressed. The first task, before offering a fuller reconstruction of changes in metrical system, is undoubtedly to establish that Old English poetry was indeed the historical progenitor of Middle English alliterative poetry, and that these two corpora belong to a continuous tradition of verse practice. This is where the next chapter begins.

Ch apter  3

The Origins of the Alliterative Revival

There is a standard story. The vernacular poetry of the Anglo-Saxons died out soon after the Norman Conquest, having succumbed to a one–two punch of linguistic change and socio-cultural displacement. The newly implanted Norman aristocracy offered no support to traditional English poetry, which had in any case become unworkable within contemporary forms of the English language. Thereafter, and for a period of about two centuries, literary composition in English was precisely exceptional: individual works came into being as exceptions to the cultural forces that tended to channel literary activity into French and Latin. Untethered from tradition, would-be writers of literature in English produced singular and implausible experiments, lacking either progenitor or progeny. To the extent that the pre-Conquest literature remained a generative presence for later writers of English, the tutelary spirit was Ælfric, Abbot of Eynsham, whose rhythmical alliterating style was echoed, it is said, in those postConquest compositions that may pass for poetry. For English literature was as yet without form, and darkness was on the face of the deep. The work of dividing poetry from prose, poetry from poetry – that work would fall to a later generation, who, in or about the time of the first great plague, spoke a new literary world into existence. Verse forms were perfected. The unformed matter of prose was coaxed into a new poetic line, and alliterative poetry was revivified in such literary treasures as Sir Gawain and the Green Knight, St. Erkwenwald, and The Morte Arthure. Distinguished by an artificial idiom and capacious narrative scope, by disciplined alliteration and prudent abstinence from rhyme, the poems of the “revival” have always seemed uncannily to recall the pre-Conquest poetry, from which they are separated by an unbridgeable interval of some three centuries. That, or approximately that, has been the standard story about postConquest alliterative poetry: Old English poetry died out; to the extent that the Middle English alliterative poetry had any lineal progenitor at all, it descends from Ælfric. Reviewing the recent scholarship, R. D. Fulk 67


The Origins of the Alliterative Revival

marvels that this narrative should have “won the day among medievalists, and without a genuine battle.”1 The present chapter offers a battle. I shall strive to overturn the view that has been standard since the 1970s, and establish instead that the surviving records of medieval English alliterative poetry are the remains of a continuous practice of composition in a continuously evolving verse form. Although the documentary record is very uneven, the metrical form of fourteenth-century alliterative poetry is, I will argue, traceable to the Old English meter. Taking up this fight, I  join several studies contemporary with or subsequent to Fulk’s remarks just quoted, principally Geoffrey Russom’s essay “The Evolution of Middle English Alliterative Meter,” Nicolay Yakovlev’s 2008 Oxford D.Phil. thesis, and Eric Weiskott’s English Alliterative Verse: Poetic Tradition and Literary History.2 From the time that modern scholars first recognized Old English poetry as poetry, they also noticed its resemblance to the alliterative poetry of the fourteenth and fifteenth centuries. The Old and Middle English poetries both have alliteration; they both lack end-rhyme; they share some distinctive vocabulary. For most of the historical period in which such questions have been formulable, resemblances between the two bodies of poetry have been interpreted as evidence that the later one derived in some way from the earlier. The precise mode of derivation has been a subject of speculation and disagreement. The thesis of a fourteenth-century “alliterative revival” was one attempt – or, really, a family of attempts – to formulate the obvious resemblances between the two poetries and the equally obvious differences between them. The metaphor of “revival” was, however, impressionistic, ambiguous, and an impediment to serious study. Does “revival” mean “returned to use” – that is, back in fashion – or “returned from death,” and what would revivification mean in this case? Nevertheless, the designation “alliterative revival” has proven difficult to shake. It hangs above Middle English poetry, directing would-be students to look back, beyond the Middle English poems, towards an earlier apparition of the alliterative poetry in this language. That instruction has on occasion been experienced as unwelcome, and not without reason. Beowulf and Gawain stem from astonishingly different cultural worlds: the relevance of these two poems to one another must rest, if at all, in an apprehension that they employ a poetic medium in some sense unitary, and mutually illuminating. Yet, the twentieth-century scholarship on the language and form of alliterative verse could justifiably be read as inconclusive on this count, and the governing metaphor of “revival” positively invited doubt. A realignment in research paradigms was probably inevitable. By affirming a radical

“Alliterative Revival”


break between the Old English and Middle English alliterative poetries, scholars of Middle English freed themselves to explore contemporary imbrications – that is, the poems as products of and participants in the culture and society of fourteenth- and fifteenth-century England. The cost of this realignment has been a certain amount of confusion regarding questions of continuity and innovation in poetic practice; relations between linguistic, social, and literary history; and interpretation of the surviving documentary record. By its vigorous return to the language of the texts, the current wave of research on alliterative meter offers an opportunity to frame the issues in new light. Much work remains to be done, but the arc of the tradition has come into sharper focus, enabling pervasive reinterpretation of surviving documents and a more general retelling of English literary history in the high and late Middle Ages.

“Alliterative Revival”: The History of a Problem The disciplinary history merits closer attention.3 Speculations about the relation between early and late instantiations of English alliterative verse began with George Hickes. His principal discussion occurs in chapter 21 of the Thesaurus, where the topic is Old English poetic language. Aided by Snorri Sturluson’s treatment of Old Norse, Hickes nevertheless commits errors and bad inferences at every level of analysis. That did not prevent him from articulating a problem that has remained central to subsequent discussions. Illustrating the persistence of Old English words for “man” in Piers Plowman, Hickes remarked Haec obiter ex Satyrographo nostro, cui Anglo-Saxonum poetae adeo familiares fuerunt, ut non solum eorum verbis versus scripsit, set tinnitum illum consonantem initialium apud eos literarum imitatus est, & nonnunquam etiam versus tantum non Saxonicè condidit.4 [These I draw in passing from our Satirist, to whom the Anglo-Saxon poets were so familiar that he not only wrote verse with their words, but also imitated that resonant agreement of initial letters in use among them and occasionally even composed a line very nearly in Saxon language.]

As support for his claim about Langland’s archaic language, Hickes quoted a line from Langland’s poem and transposed it into Old English wordforms. In Hickes’s judgment, Piers Plowman’s alliteration and its language were to be explained by the author’s direct access to, and immersion in, written copies of Old English poems. Hickes’s proposal finds no support:  Old English texts were annotated and glossed in the late Middle Ages, but not the poetry, and one can be confident that the poetry, if read,


The Origins of the Alliterative Revival

would not have been understood well enough to serve as a model for new compositions.5 Nevertheless, one may grant that written copies of earlier poems – perhaps early Middle English rather than Old English – constitute one possible source for Langland’s line. By the end of the eighteenth century, each of the other possibilities would be mooted. In Observations on the “Faerie Queene” of Spenser (1754), Thomas Warton reported Hickes’s judgment that “alliterative versification was drawn by Langland from the practice of the Saxon poets”; in the History of English Poetry (1774) he maintained that view, adding that he wished Langland had instead “avail[ed] himself of the rising and rapid improvements of the English language.”6 Warton, however, shared neither Hickes’s warm admiration for Langland’s poem, nor his care for the earlier literature: the actual historical derivation of alliterative verse mattered less to him than its archaistic profile in contemporary (that is, late medieval) literary culture. Later in the History of English Poetry, he was content to describe the form of Piers Plowman as a fourteenth-century invention, hesitating only over the question of attribution: should the title of “inventor” be accorded to Langland or to one of the other poets who shared his form, Warton wondered?7 In Thomas Percy’s view, the evidence of alliteration was sufficient to reject the position Warton would take in his History of English Poetry. Percy affirmed that Langland “will not be found to have invented any new mode of versification, as some have supposed, but only to have retained that of the old Saxon and Gothic poets.”8 Percy, however, differed from Hickes on the question of how that older poetry was transmitted to Langland. Hickes had pictured Langland as a brilliant and devoted reader of the Old English biblical poetry. Percy speculated instead that Langland and the Anglo-Saxon poets were linked in a continuous chain of poetic composition now mostly lost. Their shared “mode of versification,” Percy suggested, “was probably never wholly laid aside, but occasionally used at different intervals; tho’ the ravages of time will not suffer us now to produce a regular series of poems entirely written in it.”9 The situation envisaged by Percy in this passage is clarified in his “Essay on the Ancient English Minstrels,” also included in Reliques of Ancient English Poetry. There Percy urged the existence of a distinct “profession of oral itinerant Poet” in Anglo-Saxon England, and argued that the professional poets were a fixture in the houses of the English nobility down to the end of the sixteenth century.10 Although Percy himself did not make the argument, his Anglo-Saxon minstrels and their post-Conquest successors were presumably the people to whom he would credit the

“Alliterative Revival”


transmission of alliterative verse. The scarcity of records during these intervening centuries did not yet cause embarrassment. If it was not yet possible to “produce a regular series of poems” written in the alliterative meter, that failure could be accounted to the “ravages of time,” or to the medieval English minstrels’ always-already declining reputation. The neglect suffered by Percy’s poet-minstrels would naturally also decimate records of their compositions. Percy’s folio manuscript – the inspiration and principal source for his Reliques of Ancient English Poetry – contains an unrhymed alliterative poem commemorating the English victory against the Scots at the Battle of Flodden in 1513.11 As Percy recognized, this poem is an important witness, establishing that the “alliterative measure” had survived in some circles “so low as the sixteenth century.”12 Bede’s Ecclesiastical History places Cædmon in the seventh century. The outlines of a poetic tradition were being roughed out. In 1813, in the preface to his edition of Piers Plowman, Thomas Whitaker was able to offer the following historical sketch: The first specimen which we have been able to produce of alliterative poetry is of the sixth century [sic], the last which we know of belongs to the sixteenth. Such then having been the length of time through which this peculiar mode of versification prevailed among the northern nations, it is not improbable that some intermediate specimens of it between the Pseudo Cædmon and our author may have perished:  but, at all events, the man who could write the language of these Visions could have read the poetry of his Saxon ancestors with little difficulty.13

Like Percy, Whitaker recognized a lacuna in the historical record; like him, he attributed the absence of “intermediate specimens” to documentary loss. His affirmation of Langland’s linguistic abilities (a footnote cites Hickes) reads in this context like a hedge: even if we suppose that the intermediate specimens never existed, the integrity of the tradition is saved by the presumption that Langland could have learned his craft directly from the Old English poems. The documentary lacuna would impress itself more insistently as learning increased and “intermediate specimens” came to light. A complete edition of Lawman’s Brut appeared only in 1847, but the poem received notice in literary histories and anthologies before that.14 The presence of rhyme posed an interpretative difficulty. Edwin Guest remained uncertain whether the Brut belonged in the same series as Brunanburh and Piers Plowman, and persuaded himself that it did not matter. With or without the Brut, the documentary record contained an imposing gap. Guest was willing to suppose that linguistic change had driven the Old English meter


The Origins of the Alliterative Revival

to transform into that of the fourteenth-century poems, but he had reservations. Regarding the Piers Plowman meter, he wrote: [T]he sudden manner in which it seems to have started into existence, is by no means easy to account for. The year 1360 is the earliest date we can positively assign to any poem in this metre; and I know of none which we can, with any show of reason, suppose to have been written more than twenty or thirty years earlier. If we consider Layamon as an alliterative poet, here is a gap of nearly two centuries; and if we deny him that character, of more than two centuries and a half.15

These concerns would become a major preoccupation in the twentieth century, and Guest sketched out several of the possible solutions. Perhaps, he conjectured, the solution lay in literary geography: It is, I think, not improbable that alliterative rhythm may have yielded, in the south, to the more fashionable novelties of the twelfth and thirteenth centuries; and have kept its place in the north and west, till the success of Langland again made it one of our classical metres. This hypothesis would account for the blank, which breaks in upon the series of our alliterative poems; and must, if admitted, in some measure lessen our hopes of regaining what is lost.16

The alternative was to consider the meter of Piers Plowman “an invention of the fourteenth century,” as Warton had done.17 Guest inclined towards the first option. It is surprising that Walter Skeat felt no need to address this problem in his “Essay on Alliterative Poetry” (1868), for Skeat’s chronology differed from Guest’s only in his confident identification of Lawman’s Brut, the early Middle English St. Margaret, and The Lay Folks’ Catechism as alliterative poems.18 Perhaps Skeat’s knowledge of the manuscript record led him to consider the “blank” in the thirteenth and early fourteenth centuries unremarkable. In any case, Skeat’s handlist of post-Conquest poems, printed as an appendix to his essay, made the long silence and sudden burst of alliterative poetry readily apparent to subsequent students. In the following decade, Bernard ten Brink wrote of a fourteenth-century “Wiederaufblühen” or “neuer Aufschwung der Allitterationspoesie”; in the English edition (1889) the phrase was “the revival of alliterative poetry.”19 After 1900, the term became routine in English-language discussions: Saintsbury’s History of English Prosody (1906) and the Cambridge History of English Literature (vol. i, 1907)  each refer to “the alliterative revival.”20 Israel Gollancz argued in his edition of Winner and Waster that this poem was composed in the year 1352, thus “probably the earliest extant poem of the alliterative revival.”21 Gollancz also gave the “great revival of this archaic form of poetry” a place in a politico-cultural geography of

“Alliterative Revival”


medieval England.22 In a lecture on “The Middle Ages in the Lineage of English Poetry,” delivered in 1920, Gollancz asked his audience to imagine that the undeveloped western provinces of England were, in the fourteenth century, home to men “who held strongly to the older traditions of the English race,” and among whom “the English element predominated,” despite the Norman Conquest.23 These nativist formulations would soon be enshrined in James R. Hulbert’s argument that the “revival” poems were the cultural expression of northwest baronial opposition to royal government.24 Whereas London had “the gauds of rhyme and the harmonies of the French school of poetry,” the western marches had poets who “in respect of metre, manner, and spirit, for the most part harked back to the time before the Conquest.”25 Gollancz’s further assertion that poets of the revival were “something more than spiritual heirs of the older English poets” might seem especially perverse. By that topsy-turvy formulation, he nevertheless implies that there was “something more” linking these two poetries – not written copies, in his view, but rather a continuous practice of versification  – and that it ought to be susceptible to empirical treatment. “[T]he problem is a difficult one,” he acknowledged. J. P. Oakden’s Alliterative Poetry in Middle English (vol. i, 1930; vol. ii, 1935)  supplied detailed surveys of language and poetic form, supporting Gollancz’s construal of the literary history. A few years later, R. W. Chambers could cite both Oakden and Gollancz in support of casual remarks that obtained a surprising and disproportionate importance for later scholarship.26 Chambers and Gollancz were engaged in a nationalist project, for which “spiritual” identity figured in literary history as both premise and conclusion. Ontologized as a quasi-Platonic form, and triumphant over every external challenge, Englishness would remain true to itself through the ages. The concrete literary-historical problem nevertheless remained what it had been in previous treatments. Like their predecessors, Gollancz, Oakden, and Chambers recognized that Old and Middle English alliterative verse were different from one another; the difficulty always lay in establishing the meaning of the differences. Do observed differences represent an accommodation of one and the same metrical system to historical changes in linguistic medium? Or do they instead indicate the emergence of a fundamentally new and independent metrical system in the later period? Scholars of the interwar period decided in favor of the first option: they conceived of the Middle English unrhymed long line as a more or less direct descendant of the classical verse of the Anglo-Saxon period. “There can be few stranger things in the history of literature,” Chambers wrote, “than this sudden disappearance and reappearance of a school of poetry.”27


The Origins of the Alliterative Revival

Whatever Oakden’s claims, his scansions gave the impression of metrical disarray, not metrical continuity. Meanwhile, Chambers had seemed to claim that a practice of metrical composition had been sustained, for a period of two-and-a-half centuries, by nothing more substantive than national resolve. That was unacceptable. In the scholarship of the 1960s and 1970s, one finds a general rebellion against these procedures. If the resemblance between Old English and Middle English alliterative verse extended no further than alliterative patterning, shared vocabulary, and absence of end-rhyme, then those features ought to be explained by other means, without recourse to the dubious conjecture of an unattested tradition. Accordingly, N. F. Blake, Elizabeth Salter, Thorlac Turville-Petre, and David A. Lawton returned to the position first articulated by Thomas Warton. They proposed that the late Middle English unrhymed line was a fourteenth-century invention, not significantly older than its earliest manuscript attestations. They sought historically proximate origins for the form, and they explored resemblances between the aa/ax line and the formally various verse and prose products of the thirteenth and early fourteenth centuries, especially rhythmical alliterating prose.28 Most strikingly, they declined to hypothesize the existence of any literary activity not recorded in the surviving documentary record. As illustration of these procedures, one can consider again the poetic synonyms for “man,” first remarked upon by Hickes.29 In the Old English poetry, these are beorn, cempa, freca, guma, hæleð, leod, rinc, scealc, secg, and wiga. The corresponding forms in Middle English are burne, freke, gome, haþel, lede, renk, schalk, segge, and wyȝe. These words almost invariably participate in the alliterative pattern, which was their value:  they accommodate the story to the requirements of verse form. Battle, horses, movement, and other high-frequency narremes accrued similar word hoards, though the synonyms for “man” stand out for their consistency. The Middle English forms burne, gome, and lede have fairly wide distributions, though Chaucer did not use them. The remaining words are rare or unattested in Middle English outside of alliterative verse. Reviewing these showings, Turville-Petre concluded that they “perhaps demonstrate not the stability and purity of an unbroken alliterative tradition, but instead the resourcefulness of alliterative poets and their readiness to embrace words from every conceivable environment.”30 Indeed, the fourteenth-century alliterative poets are distinguished by their lexical acquisitiveness, evident especially in their Norse- and French-derived vocabulary. (These borrowings included a new poetic word for “man”: tulk < ON túlkr.) One could just imagine that shared vocabulary items were transmitted to the Brut

Final -e


and again to the late Middle English poems through non-literary usage or, alternatively, through written texts. In support of the non-literary channel, Turville-Petre pointed to the survival of freke and schalk in some modern dialects; meanwhile, Derek Pearsall noted that a copy of the Brut was available and in use in the first decades of the fourteenth century in the southwest Midlands.31 Other considerations count against these proposals. The synonyms for “man” belonged to a specialized poetic register in Old English; it is hard to imagine that substantially the same collection of specifically poetic words could have been reassembled in the later poetry, absent some sort of contact with the poetic practice that had gathered these words in the first place. Yet, the hypothesis of textual consultation offers no better fit with the evidence. The Middle English form haþel is usually explained as a blend of OE hæleð and æðele (“noble”), an unlikely result of textual consultation.32 The remaining Middle English words exhibit phonological developments typical of the transition from Old to Middle English. These showings point towards sustained use, not punctual recovery from the written record. In Marie Borroff’s words, “[t]he suggestion of continuity between the two traditions is irresistible.”33 An irresistible suggestion is nevertheless a suspect one. Pearsall urged that gaps in the historical record were likely an artifact of manuscript losses, but this, too, had a defensive air. Nothing could be more certain than the incompleteness of surviving records; saying so would nevertheless appear as special pleading unless or until previously overlooked evidence of continuity in verse practice could be gleaned from the surviving records themselves.

Final -e The first intimations of new evidence were announced in 1988 by Hoyt N.  Duggan, at the end of a game-changing article, “Final -e and the Rhythmic Structure of the B-Verse in Middle English Alliterative Poetry.” Duggan’s aim here was to integrate his metrical studies with the existing scholarship on Middle English final -e, for the metrical system that he and Thomas Cable had reconstructed for alliterative poems of the fourteenth and fifteenth centuries was not, in every respect, a good fit for the English language of that period, as then understood.34 In his 1988 article, Duggan argued that small anomalies in the metrical patterning of surviving Middle English poems pointed towards a long pre-history of versification in precisely this form.


The Origins of the Alliterative Revival

An extreme case, but a diagnostic one, concerned adjectives with disyllabic stems. In Old English, adjectives were inflected for gender, number, case, and definiteness, with separate paradigms (weak and strong) for definite and indefinite grammatical contexts. Gender and case distinctions were lost early on, but the singular/plural and definite/indefinite contrasts proved durable, and were expressed in Middle English by the presence or absence of an inflectional -e.35 In accentual-syllabic verse, inflectional -e remained viable into the early fifteenth century, at least in adjectives with monosyllabic stems.36 To cite a textbook example, Chaucer and his immediate followers distinguished between this olde man (definite usage, with inflectional -e) and an old man (indefinite usage, without inflection). Plurals regularly took inflectional -e whether definite or not. This account of adjectival inflection holds up well in southern dialects for monosyllabic adjectives with consonantal coda, like old, but did not appear to be generalizable to northern dialects or polysyllabic stems.37 Two bodies of evidence – spellings, and prosodic behaviors in accentual-syllabic meters – seemed to indicate that inflectional -e went defunct in disyllabic adjectives at an early date, perhaps by the turn of the thirteenth century. Whether employed in definite or indefinite grammatical contexts, and whether modifying singular or plural nouns, disyllabic adjectives such as berdlez, certain, gentill, Irish, menskful, selcouth, and witles would be expected to remain disyllabic, as would weak past participles such as hoked and selid. Nevertheless, fourteenth-century alliterative poems regularly employ these words and others of the same type – that is, weak or plural disyllabic adjectives with consonantal ending – in positions where one expects a long dip: “bot berdlez chylder” (Gawain 280b), “with certayn leteres” (Siege 85b), “and gentill knyghtes” (Morte Arthure 246b), “and Irische kynges” (Morte Arthure 4123b), “with menskful þingez” (Gawain 1809b), “selcouþe þynges” (Siege 1222b), “with selid lettirs” (Wars Alex. 1293b), “with hoked staues” (PPl-Bx.P.53b), and, illustrating definite inflection in direct address, “þou witles berne” (Wars Alex. 107b).38 Duggan interpreted such verses as having a short medial dip, on the well-founded assumption that inflectional -e was defunct, but he added that each of the verses would have been metrically regular, with inflectional -e, “had these poems been written further to the south and two centuries earlier.”39 The poems certainly cannot be dated two centuries earlier. How, then, to explain this surprising pattern of usages – especially given that the poems rigorously distinguish weak and plural adjectives from strong singulars? Weak and plural disyllabic adjectives are used almost exclusively where one expects a long dip, whereas strong singulars (which would lack an inflection)

Final -e


almost never form a long dip on their own.40 Instead, the poets made other uses of strong singulars, exemplified by such verses as Gawain 438b “þaȝ hedlez he were.” Duggan’s explanation was that “linguistic change prompted a change in the conditions of metricality.”41 Poets had been constructing b-verses like Morte Arthure 246b “and gentill knyghtes” prior to loss of inflectional -e, and they went on constructing b-verses in the same way after loss of the inflectional syllable. The syntactic frame in question was too useful to be abandoned, and too firmly entrenched in habits of versification; instead of changing the way they used disyllabic adjectives, poets introduced a new, narrowly delimited license: they “came gradually to take x/x/x to be metrical in this particular syntactic context.”42 (Duggan’s “/” represents a strong position, equivalent to my “S.”) In a later article, Duggan extended this line of argument to accommodate unexpected regularities in the distribution of adverbs ending in -ly.43 Adverbs with this morpheme almost never appear at the end of the line. The reason, Duggan proposed, was that the formulaic grammar of alliterative verse had become fixed already in the thirteenth century, when the disyllabic form -liche, then current, would have produced an unmetrical long dip at line-end. When forms in -liche gave way to forms in -ly, the poets barely took notice; they continued to avoid putting -ly adverbs at line-end. Reasoning thus, Duggan preserved received understandings of Middle English -e (he posited no extension to the life of this vowel), but at the cost of an audacious coup against received understandings of literary history. Against the view that Middle English alliterative verse was a fourteenth-century invention, Duggan brandished a collection of tiny metrical anomalies: b-verses like “and gentill knyghtes” testified to an otherwise unattested tradition of versification, stretching back to a time when those verses would have been metrically unexceptional. The surviving poems must, Duggan concluded, have been preceded by “a very long tradition of composition in essentially this form”: “Recent students have got it wrong in arguing that the surviving poetry of the revival represents a conscious new beginning.”44 The status of final -e was a matter of intense disagreement between Duggan and Cable in the 1980s and 1990s. There was, however, a telling symmetry between their respective positions, for they both found it expedient, at different points in their analyses, to introduce elements of pre-fourteenth-century phonology into their accounts of fourteenth- and fifteenth-century poems. Duggan introduced early phonology at the end of his study, once he found that certain verse-types were a syllable shorter


The Origins of the Alliterative Revival

than expected. Conversely, Cable introduced the earlier phonology at the beginning, in his initial scansions: he assumed that weak final -e was metrically relevant wherever justified by derivation or grammar.45 Accordingly, he would scan “and gentill knyghtes,” with a disyllabic medial dip (xSxxSx), yielding a perfectly ordinary accentual contour in the b-verse. Cable took this result as evidence that his initial phonological assumptions were in some sense true to the fourteenth-century poems, and he accordingly described the poems as written in a specifically conservative, elevated, or literary register, one that retained the traditional grammar of final -e.46 The b-verse “and gentill knyghtes” was rhythmically unexceptional, but the language of the poems as a whole was out of step with contemporary developments, unaffected by phenomena of syllable-reduction pervasive in other contemporary records. Subsequent research has confirmed Cable’s position.47 Words with grammatically or etymologically justified final -e behave in Middle English alliterative verse just as if they still retained that syllable, or retained a by-form with that syllable. Whether or not these final weak vowels were sounded in performance, the meter continued to count them.48 Indeed, the metrical grammar of final -e in Middle English alliterative verse, as now understood, corresponds closely to the grammar of this vowel in the poetry of Chaucer. A notable difference is that disyllabic adjectives go uninflected in Chaucer and other accentual-syllabic verse, whereas they usually preserve their inflections in alliterative verse. Here one should make two further observations. First, the consecutive weak syllables supplied by the inflected adjective gentille (scanning Sxx) would be useless to Chaucer. A Middle English poet writing in an alternating-stress template would need to contrive for elision of that inflectional syllable, or elect an au courant, e-less form. Poets regularly chose the second option, and there is no reason to think that choice particularly marked, for Middle English verse in alternating-stress meters exhibits a promiscuous mix of e-full and e-less forms from a very early date.49 Inflectionless disyllabic adjectives were simply part of that larger mix, governed by the principle of stress alternation. This line of reasoning is confirmed by a second observation. Though alternating-stress templates encourage poets and readers to omit inflections from disyllabic adjectives – which would be rhythmically irregular if inflected – these same templates could and did accommodate inflections on polysyllabic adjectives.50 An example is CT IX.119, “Therto he was the semelieste man,” where inflectional -e fills the last weak position in the line. Where an inflection contributes to stress alternation, poets used it to that end. The basic division

The Brut


is not between monosyllabic adjectives and all others, but between disyllabic adjectives and all others. Moreover, that binary division is peculiar to accentual-syllabic verse and wrongly generalized to the language as such. Existing descriptions of Middle English adjectival morphology have been skewed by uncritical interpretation of the data from accentual-syllabic verse, and should probably be revised. Though the tendency of the last few paragraphs is against Duggan’s views on final -e, these results should be seen to strengthen his attendant literary-historical inference. By the time the late Middle English alliterative meter first appears in the written record, that meter is already out of step with contemporary language. How might fourteenth-century poets have acquired their conservative register? As the putative building blocks of the fourteenth-century alliterative line, Blake, Salter, Turville-Petre, and Lawton had pointed to rhythmical alliterating prose and various forms of septenary and four-stress verse. I defer comment on prose until later in this chapter. It is clear, however, that Middle English accentual-syllabic verse forms would be unlikely vectors for a conservative grammar of final -e. The handling of -e in alternating-stress verse tends to be erratic, and has constituted a primary body of evidence for the attrition of that syllable in Middle English. Had these forms been the antecedents to the fourteenth-century alliterative line, it would be very difficult to see how the alliterative poets could have recovered a conservative grammar of final -e, and done so with such consistency. “The hypothesis of continuity,” writes Geoffrey Russom, “seems to be required by the very nature of [the] rules” posited by recent studies of this meter: “[i]t is difficult to imagine how anything like this meter could have been invented or learned independently of tradition.”51 It seems that surviving manuscripts give a very incomplete record of the historical practice of composition in alliterative verse. This, at least, is the implication of the metrical research.

The Brut If the fourteenth-century alliterative poems were preceded by (in Duggan’s words again) “a very long tradition of composition in essentially this form,” might the surviving alliterative poems from earlier centuries have a place in this story? Duggan thought not: he affirmed, without elaboration, that the fourteenth-century line “is not likely to have evolved directly from the classical [Old English] line nor from poems like Layamon’s Brut or the other verse written in early Middle English, at least from none that survives.”52 Cable examined the early Middle English verse in some


The Origins of the Alliterative Revival

detail, but abstained from diachronic speculation in The English Alliterative Tradition: his stated aim was only to offer a “series of synchronic analyses, arranged chronologically.”53 In any case, Cable’s results were not promising:  the Brut and other early Middle English verse continued to look nearly as confused as they had in Oakden’s treatment. Research had, in Harold Zimmerman’s apt formulation, reached an “impasse.”54 Details of verse design implied that the fourteenth-century meter had a longer history, but extant alliterative verse of the twelfth and thirteenth centuries seemed not to meet specifications. The break came in the form of renewed efforts at diachronic reconstruction, a project for which Cable’s synchronic snapshots prepared the way. In a key contribution, Russom showed how the b-verse contours of the fourteenth-century poems could have derived from a subset of the accentual contours employed in the Old English verse. In Russom’s reconstruction, the principle of closure would sift regularized reflexes of the Sievers Types A and C into the second half of the line, where subsequent generations of poets would reinterpret those patterns as the defining feature of their inherited meter.55 This was a brilliant insight, yet also incomplete, for Russom was not able to follow the trail of metrical evolution through the alliterative poetry of the twelfth and thirteenth centuries. Regarding these works, he said only that their “meters are insufficiently well understood for comparison.”56 In a perceptive reply to Russom’s essay, Fulk reported a small set of scansions from The Grave, The Soul’s Address to the Body, and Lawman’s Brut, finding in each case that they failed to support Russom’s elegant theory: “There must be another explanation,” he concluded.57 We must, then, turn to these poems, and especially to that great and greatly enigmatic monument of early Middle English poetry, Lawman’s Hystoria Brutonum. Lawman’s Brut is known to modern readers in two manuscripts of the later thirteenth century, British Library, Cotton MSS Caligula A.ix and Otho C.xiii.58 These manuscripts record sharply divergent texts nevertheless recognizable as versions of a single English poem, based principally on Wace’s mid-twelfth-century Anglo-Norman verse history of the English. The Caligula text runs to 16,095 lines in the EETS edition by G. L. Brook and R. F. Leslie. Otho differs from Caligula in omitting about 3,000 lines and much of Caligula’s traditional poetic diction. Caligula, which is always accorded priority in the scholarship, is presumably the text closer to the original state of Lawman’s poem. Otho has the appearance of being an intelligent and exacting redaction, aiming principally to cast the poem into more current language.

The Brut


The linguistic differences between Otho and Caligula were influentially leveraged by Eric G.  Stanley to argue that Lawman wrote in “an archaic, perhaps archaistic, idiom consciously, perhaps artificially, derived from Anglo-Saxon poetry”: the archaic diction and irregular spellings of the Caligula text are, Stanley contended, “ye olde signs,” expressing an ill-informed antiquarian enthusiasm.59 Published in 1969, “Laȝamon’s Antiquarian Sentiments” is easily the single article that has done most to guide subsequent understandings of the Brut; the essay accorded very well with the discontinuity thesis subsequently put forward by Blake, Salter, and Turville-Petre. Isolated from any viable poetic tradition in English, Lawman (one supposed) invented a new verse form and a nonce language through a daring act of historical emulation. That characterization has not gone unchallenged. Jane Roberts argues that the archaic element in the vocabulary of the Caligula text has been overstated.60 Mark Amodio emphasizes that Lawman’s poetic compounds are not fossilized; though the incidence of compounds in the Brut is much lower than in Old English poetry, the grammar of compounding remains vital and productive.61 Robert Millar introduces the useful concept of socio-literary register into analysis of conservative grammatical forms in the Brut, a move that parallels M. L. Samuels’ contribution to the study of final -e in Chaucer’s verse.62 These studies do not express a unitary vision of the Brut, but there is a common tendency among them. In a recent review, Françoise H. M. Le Saux remarks that, “It increasingly looks as if the language of the Caligula text, though clearly in an older, more elevated register than certain other thirteenth-century texts from the same area, was not necessarily contrivedly so.”63 Reassessments of Lawman’s language supply the context in which we may turn to his verse form, arguably the terra incognitissima of English historical prosody. In the introduction to his edition, Frederic Madden remarked that, “The structure of Laȝamon’s poem consists partly of lines in which the alliterative system of the Anglo-Saxons is preserved, and partly of couplets of unequal length rhiming together. Many couplets indeed occur, which have both of these forms, whilst others are often met with which possess neither.”64 Subsequent scholarship has not obviously improved upon that statement. The Brut is said to be “metrically a hybrid,” “a mixture,” “a representative of a transitional stage in English versification,” and “free verse [that] preserves the half-line structure of Old English verse.”65 At the opening of the twentieth century one sought to apply Sievers’s Five Types to the Brut.66 One debated whether half-lines have two stresses, as a Sieversian scansion would seem to require, or four.


The Origins of the Alliterative Revival

As models for a four-stress half-line, one pointed to Otfrid’s Old High German Krist, Wace’s Anglo-Norman octosyllabic Roman de Brut, accentual Latin hymnody, or an otherwise unrecorded tradition of “popular” Old English verse.67 In 1969 Blake swept these fruitless debates aside by proposing that Lawman’s most important forebear and formal model was not poetry at all, but rather Ælfric’s rhythmical alliterating prose.68 Angus McIntosh lent some prosodic evidence to Blake’s claim, in an essay that remains its most astute advocate.69 More recently, Thomas Bredehoft has constructed a literary lineage extending from Beowulf to Ælfric’s alliterating saints’ lives to Lawman’s Brut, all described as verse; the cost of including Ælfric in this lineage is a metrical theory that will designate almost any stretch of English as metrical poetry.70 Bredehoft does not renounce the poetry/prose binome. The de facto consequence of his metrical theory, however, is that he must defer to non-metrical criteria (alliteration, rhyme, and other elementary figures of sound) to discriminate poetry from prose. “Rhythmical alliteration” is what matters, just as Blake claimed. Indeed, the enduring legacy of Blake’s 1969 intervention is that subsequent studies of Lawman’s “prosody” have typically been studies of alliteration, assonance, and rhyme, leaving the metrical structure itself unexamined.71 Metrical structure does get a hearing from A. W. Glowka, whose “Prosodic decorum in Layamon’s Brut” is indicative of the difficulties in this area of research: Glowka reads Lawman’s verse wherever possible as an alternating-stress form; he reports that lines have anywhere between two and twelve beats.72 Another indication of difficulties may be found in the sentences that open and close Le Saux’s remarks on versification in her useful monograph: she begins with a positive assertion (“Laȝamon’s verse is a descendant of the Old English long line”), but ends two pages later with a doubtful conditional:  “If Laȝamon was indebted to Old English poetry …”73 This is not encouraging. From Edwin Guest forward, it has always been clear that the Brut bears at least a family resemblance to prior and subsequent alliterative poetry in English. In support of this minimal statement of affiliation, one may point to Lawman’s lexis (the presence in the Brut of some words rarely or never found outside alliterative poems) and his penchant for syntactical inversion (e.g., 1b “Laʒamon wes ihoten,” 2b “liðe him beo Drihten,” and 5b “þer he bock radde”). One might also sense that the background rhythmical patterns of the Brut – as illustrated by those same b-verses just cited – share something with a wider tradition of English alliterative poetry. Luick, Jakob Schipper, and Oakden each sensed this basic rhythmical affiliation and sought to express it in Sieversian scansion. Their efforts may

The Brut


be faulted for according an excess of reality to Sievers’s Five Types, but that is understandable: Luick and Schipper were attempting to generalize and extend a powerful new theory. In a 1995 essay that is an important precursor to Yakovlev’s work, Ralph Hanna made a strictly analogous assault on Lawman’s prosody: Hanna sought to demonstrate that the contours identified by Cable and Duggan in the fourteenth-century alliterative poems are already discernible in Lawman’s Brut, as the predominant rhythmical constituents of that poem’s b-verses.74 Hanna’s scansions disclose regularities of verse structure that would certainly have been apparent to Duggan, Cable, Russom, and Fulk; they also disclose irregularities that seem to support those scholars in their skepticism. A general solution will only be won by direct confrontation with Lawman’s irregularities. One may begin with Hanna’s 46-line sample. In the pages that follow I itemize the irregularities, then offer interpretations. Lines 12346–91 of the Caligula text have, in Hanna’s scansion, four lines with masculine ending, exemplified by 12357b “and al þi driht-liche uolc.” Though one may often advance alternative interpretations for individual lines, it cannot be denied that masculine endings occur with some frequency in the Brut, and are probably underrepresented in Hanna’s test passage and interpretation of it. Yakovlev would put the ratio of masculine endings as high as one in four.75 Since masculine endings are fully acceptable in the Old English meter (schematized as Sievers Types B and E), and since they are fully acceptable in at least Duggan’s theory of the fourteenthcentury meter, one might consider masculine endings in the Brut to be unremarkable. Yet, in the same article in which Duggan posited a very long tradition of composition in the Gawain-meter, he speculated that the earlier members of this tradition probably had uniformly feminine line endings. Duggan’s reasoning on this point is instructive, and runs closely parallel to his treatment of disyllabic adjectives. The fourteenth-century poets almost invariably end their lines with a word having a historically justified final weak syllable: for example, PPl.Bx Prol.1, sonne. This regularity is recognized by everyone who has studied the meter, and most metrists working today accept Cable’s interpretation of it: the fact that poets preferentially filled this line position with a word having a historically justified weak final syllable allows us to infer that these words indeed retained a weak final syllable, and that a feminine ending was understood by the poets to be a requirement of their form. Rejecting this inference, Duggan supposed instead that the fourteenth-century poets were acting out a metrical requirement fitted to an earlier stage of the language: “at some point in the tradition a final unstressed syllable had been a metrical requirement.”76


The Origins of the Alliterative Revival

When final -e was lost from the language, new generations of poets carried on composing as their predecessors had, placing the same traditional words in the same line positions, as if in a state of suspended animation. Much like Duggan’s account of the metrical grammar of disyllabic adjectives, his account of line-ending phenomena overestimates the formulaic element in alliterative poetry (how, for example, does one explain the assimilation of new loan-words to a prosodic structure presumed neutralized?), yet this line of reasoning also shows why Duggan could not recognize the Brutmeter as an antecedent to the fourteenth-century form. Where Duggan posited a prior metrical state with uniformly feminine endings, the Brut gives a promiscuous mix of masculine and feminine. Any reconstruction that would place the Brut in a line of development towards the Gawainmeter must accommodate this feature. More troubling than the masculine endings, however, are apparent long dips at line-end, for this is a pattern rigorously avoided in both the Old English meter and the fourteenth-century one. Hanna’s scansion registers three instances of this pattern: 12358b “i-cumen her forð-rihtes,” scanned with rihtes as a disyllabic dip, following primary stress on forð (xSxxSxx); 12378b “to Luces ure kaisere” (xSxxxSxx); and 12391b “þat þu ærst forhoȝedest” (xxSxSxx). There is an obvious solution in the first of these lines: treat forð-rihtes as phrasal, with stress thrown to the right, and scan the verse xSxxxSx. Nevertheless, these three verses may be taken as exemplary. Compounds and proparoxytone polysyllables appear at line-end with some frequency, creating what appear to be line-terminal disyllabic dips. Absent an interpretation of this linguistic material, it is hard to avoid the conclusion that the Brut does not observe the characteristic right-edge constraints of earlier and later alliterative poems. There are at least two more problems. The first regards the count and placement of non-final long dips. In Beowulf and associated poems, a dip is expandable only in the first two positions.77 Once one takes account of resolution and the prefix license, verses with two long dips are rare. An example of an exception is Beowulf 455b “Gǣð ā wyrd swā hīo scel” (“Things always go as they must”), in which the medial dip is formed by irreducible linguistic material, a conjunction and pronoun. This pattern is likewise rare in the b-verses of Gawain and affiliated poems: verses like Gawain 181b “vmbefoldes his schulderes,” with two long dips, are rare in the central corpus; most occurrences, in most poems, are justifiably regarded as unmetrical. By contrast with earlier and later practice, the Brut seems to have a significant number of b-verses like Beowulf 445b and Gawain 181b, in which initial and medial dips are both expanded.

The Brut


Hanna’s scansion records four such verses in the test passage: 12357b “and al þi driht-liche uolc,” 12368b “þat þu swa reh ært. iwurðen,” 12380b “and hine for lauerd icnawen,” and 12387b “þu scalt wursen vnder-fon.” To these, one should probably add 12354b “bi-uoren Arðure. þan leod-kinge,” 12371b “a þire aȝere hond,” and 12379b “þat he is king ouer þe.” It would seem, then, that Hanna’s sample passage of 46 lines contains as many as seven b-verses that violate expectation by having expanded dips in both initial and medial position. Seven offenders in 46 is a high ratio, and cannot be chalked up to a sampling error: it is easy to find verses with two long dips in the Brut. Indeed, Fulk pointed to precisely this feature in his assessment that early Middle English alliterative verse, including the Brut, fails to support Russom’s theory of metrical evolution.78 Finally, the last class of problem cases consists of b-verses interpretable as having three lifts. Hanna’s scansion of the sample passage records no instances of this pattern:  in 12377b “and word send to Rome” and four other verses like it, Hanna demotes an open-class word, yielding a normal two-lift contour. This is sensible. In the case of 12354b “bi-uoren Arðure. þan leod-kinge,” another b-verse that one might construe as three-lift, Hanna deletes Arðure, again yielding a normal two-lift contour (xxxxSSx), and one must grant that the word Arðure is suspect here. It does not appear in the corresponding Otho line, which reads “bi-voren þan kinge.” The phrase “bi-uoren Arðure” in Caligula 12354b is an exact repetition of the immediately preceding line, where that phrase forms the entire b-verse: it is possible that Arðure was intruded into 12354b by a copying error. Even the punctuation gives reason for doubt: the punctuation of Caligula shows that, somewhere in the stream of transmission, a scribe mistakenly construed this line to end at Arðure, for there are punctus (the usual end-of-line punctuation in both Otho and Caligula) entered after both Arðure and kinge. Yakovlev shows that the punctuation of the Caligula text is closely attentive to metrical form, and that instances of erroneous double punctuation in the stint of Caligula’s Hand B, as at 12354b, are often interpretable as a misfiring of metrical expectation: a scribe (probably not Caligula Hand B himself, but rather the scribe responsible for this portion of his exemplar) recognized bi-uoren Arðure as a well-formed b-verse, assumed that the following material would belong to the next a-verse, and entered a second punctus after þan leod-kinge when he recognized that these words in fact belong to the b-verse.79 Recognized as peculiar by a thirteenth-century scribe, the verse is also suspect from the perspective of textual criticism. This is all to say that Hanna’s treatment of potentially three-lift verses in his test passage is wholly justifiable.


The Origins of the Alliterative Revival

Nevertheless, my preference, following Yakovlev, is to base scansion on the Caligula text as transmitted, and to register in my scansion the distinctiveness of heavy verses. Line 12377b “and word send to Rome” and 8680b “þas stanes beoð græte. & longe” each have three open-class words; 6944b “hired-men heo luuieð for þi” has two open-class words (one a compound) and a lift-bearing function word in terminal position.80 If all but the most egregious instances of three-lift verses are classed from the outset as twolift, the result is a small residue that resists further analysis: what can one say about them, except that they are irregular? By contrast, a category of verses defined by their common lexical construction (either three openclass words or two open-class words and a verse-terminal function word) is sufficiently well defined to sustain analysis and sufficiently numerous to demand it. In a passage of 200 lines from the Caligula text, Yakovlev found that 22 percent of a-verses and 12.5 percent of b-verses have three-lift constructions.81 The density of three-lift verses varies somewhat from one stretch of the poem to another, but a range between 10 and 15 percent is probably a safe estimate for b-verses in the Caligula text. In Gawain, at least 28 percent of a-verses have three-lift constructions (I will increase that number in the next chapter of this book); by the same criteria, just 3 percent of b-verses are three-lift. Why should the b-verses of the Brut have so many more three-lift constructions than those of Gawain? At first glance, the Old English meter does not seem to offer any help with this question, for the standard Sieversian interpretation of that meter is that each well-formed verse has two major stresses. It is true that verses of Sievers Type D (SSsx) and E (SsxS) are conceptualized as having a secondary stress in addition to the two primary ones. It is true, too, that one occasionally meets a heavy verse interpretable as having three “major” stresses:  Beowulf 147a “twelf wintra tīd” (a period of twelve winters) supplies an example, composed of three separate open-class words. B. R. Hutcheson, who accorded “three-stress patterns” in Old English a higher degree of reality than most other scholars have done, found that they constituted less than 3 percent of verses in his corpus of ten poems.82 Moreover, the frequency of Sievers Types D and E drops off in late Old English poems, a feature that is always linked in the scholarship to the decline in poetic compounding.83 I  have already sought to undermine this explanatory edifice, at the conclusion of Chapter 2, but the consensus view should be emphasized. According to that view, Sievers Types D and E withdrew from the repertoire and would exert a correspondingly weak influence on the metrical configuration of an early Middle English successor.

The Brut


Old English hypermetric or extended lines supply another possible source for three-lift verses in the Brut and later Middle English alliterative verse. The systematic implications of the Old English hypermetric line are difficult to assess, however, since they are frequent in some poems, such as The Dream of the Rood and Judith, but quite rare in others.84 What is clear is that three-lift verses became a robust component of alliterative poetry from the twelfth century forward. Absent an explanation, it is hard to avoid the impression that such verses are prose-like, to be read as the degradation of poetry towards prose, or else as the reflexes of a prose antecedent. In sum, b-verses in the Brut may be seen to offend against metrical regularity in the following four features: 1. masculine endings, e.g., 12357b “and al þi driht-liche uolc” (xxxSxxS); 2. long final dips, e.g., 12391b “þat þu ærst for-hoȝedest” (xxSxSxx); 3. long dips in both initial and medial positions, e.g., 12357b again; also 12368b “þat þu swa reh ært iwurðen” (xxxSxxSx); 4. three-lift verses, e.g., 8680b “þas stanes beoð græte. & longe” (xSxxSxxSx). In each of these respects, Hanna’s scansion may be faulted as overly optimistic: a significant residue of Lawman’s b-verses exhibit contours rare or absent in b-verses of fourteenth-century poems. Yakovlev put the rate of prima facie nonconformity as high as one b-verse in five.85 His treatment does not wholly eliminate any of the offending prosodic features, but it yields a sharp reduction in the incidence of the second and third in the list above, and sets all four in a newly rationalized perspective. We can take the four features in order, beginning with the masculine endings, which are a problem only from the perspective of the fourteenthcentury meter. Duggan seems to have presumed that the fourteenthcentury meter developed from a purer and more regular version of itself, but this is unjustifiable. Masculine endings were a fully systematic expression of the Old English meter, so one might expect to find at least a residue of them in an intermediate metrical system. Russom’s principle of closure explains why the alliterative meter would eventually bar masculine endings from the b-verse, but there is no reason to posit a priori that the sifting process had run to completion by the time that the Brut and other early Middle English alliterative poems were composed. Yakovlev finds that the degree of differentiation between a-verse and b-verse in the Brut is, by several different measures, intermediate between Beowulf and Gawain.86 Long final dips, as in 12378b “to Luces ure kaisere” (xSxxxSxx) and 12391b “þat þu ærst for-hoȝedest” (xxSxSxx), would be a genuine violation


The Origins of the Alliterative Revival

of earlier and later meters, but Yakovlev shows that apparent instances are almost always reducible to a normal pattern. There are two mechanisms involved, exemplified by the two verses just quoted. The first of these verses, 12378b, may be reconciled with metrical expectation if we assume that stress is shiftable. Searching in the digitized text of the Corpus of Middle English Prose and Verse, I find 72 instances of kaisere in the Caligula Brut. In 48 cases, the word occurs at the end of a hemistich, and, in 21 of those cases, it participates in rhyme or assonance requiring non-initial stress: pairs include ære (adv.), beren (“bear,” v.), iferen (“companions”), ȝere (“year”), here (“army”), here (adv.), kare (“care”), spere (“spear”), þere (adv.), and weore (“were”). Stress shift of the sort required by kaisere in these pairings is supported by numerous other rhyme pairs in the Brut:  Yakovlev shows that the combination of stress shift and rhyme or assonance is very frequent in proper nouns and words of foreign origin.87 Stress shift is also supported in contemporary accentual-syllabic verse, the Poema morale and Ormulum.88 It seems that the transition from a morphological meter to an accentual one was accompanied by variable accentuation of some classes of polysyllabic words; this would become an enduring feature of English poetry. The second mechanism is backward looking. The word for-hoȝedest (< OE forhogian, “to scorn”) has an etymologically short root vowel in an open syllable. In Old English poetry this sequence would be expected to resolve, and Yakovlev shows that resolution continues to operate in strong metrical positions in the Brut, or at least in the last strong position of the line.89 Thus, -hoȝedest scans Srx (where “Sr” denotes a single strong position composed from a resolved sequence) and supplies a monosyllabic final dip to 12391. Similarly with line-final aðelest (10104), faȝerest (10701), for-lorene (10571), makeden (10677), sumeres (10703), and wateres (10723b). Each of these words had a short root vowel in Old English, and these etymological quantities seem still to be intact in Lawman’s language.90 Fulk has shown that resolution operates in a narrowly defined metrical context in the Poema morale (c.1170–90).91 Given this finding in an accentual-syllabic poem, it is not implausible that resolution should be active in an alliterative poem of approximately the same date. Incorporating both stress shift and resolution, the meter of the Brut is genuinely transitional. Resolution may also operate in non-final lifts, as Yakovlev argues, though here the evidence as I understand it is conflicting. In 12075b “and þat water wes liðe,” resolution reduces an anomalous long medial dip: if one assumes that resolution is operative in the first lift of the verse, it may be scanned xxSrxSx (=xxSxSx), a pattern well represented in lines

The Brut


without resolution and consistent with the overall diachronic reconstruction. In other cases, however, resolution yields a pattern rare in the Brut and difficult to reconcile with diachronic and systematic considerations. The frequent word aðele provides an example: when this word supplies the first lift of the b-verse, resolution, if functional, would often reduce the only long dip in the verse. Examples are 10551b “þas aðelen kinges” and 10562b “aðelest cunnes,” which would scan xSrxSx (=xSxSx) and SrxSx (=SxSx), respectively, with resolution. I count nearly 80 verses like these in the Caligula text, where resolution of aðele would yield a verse without a long dip; most are utterly formulaic phrases. The contour SxSx does occur without resolution in the Brut; an example is 9b “ærest ahten.” Yet this contour is very rare overall, accounting for perhaps 2 percent of the total, and its rarity accords well with available accounts of metrical change, including the one offered by Russom in his 2004 article. The first dip was always expandable in Sievers Types A, B, and C, and the overall proportion of polysyllabic dips could be expected to increase as the Old English inflectional system gave way to an analytic syntax, dependent on function words to establish basic grammatical relationships. A higher density of function words meant a higher ratio of expanded dips. Over time, a tendency was reanalyzed as a rule; what had been an option in the Old English meter became mandatory in its Middle English successor.92 True to expectation, the variant of Sievers Type A with expanded first dip is the most frequent pattern in b-verses of the Brut, occurring on average about once in every four verses: for example, 12351b “tuhte to-somne.” The comparative rarity of the unexpanded contour SxSx in verses without conditions for resolution makes resolution suspect in verses like “aðelest cunnes.” In other cases, resolution in the first lift would yield a verse having only three positions: for example, 14628b “þat he ouercume wæs” and 6938b “þat is an weoli godd” (OE weolig, “wealthy”). In my scansions of the Brut I have accordingly treated resolution as optional in non-final lifts, without being certain what criteria, beyond metrical pattern, may trigger the suspension of resolution in any given case. In discussion of resolution, we have turned from the final dip to medial dip, and we should now note that an offending long dip often contains an unstressed prefix. In Old English verse, prefixes and the negative particle ne seem to be excludable from the count of metrical syllables, and Yakovlev argues that these morphemes retain that peculiar metrical function in the Brut.93 I introduced the “prefix license” in Chapter 2, where it figured as central support for Yakovlev’s non-accentual interpretation of the Old English meter. It may be useful to review the argumentation.


The Origins of the Alliterative Revival

Verbal prefixes and ne appear with some regularity at the head of verses having the accentual contour xSxSx or xSx…xSx: for example, Beowulf 2659a “gesīgan æt sæcce” (“to sink in battle”) and Beowulf 1322a “ne frīn þū æfter sǣlum” (“do not ask about happiness”). If every syllable that contributes to the accentual contour contributes as well to the count of metrical positions, then these verses and numerous others like them violate the four-position rule. The verses may be recognized as normal, however, if we accept that unstressed prefixes and ne have a unique status in the meter, being optionally exempt from the count of metrical syllables. Likewise, where a third-position dip in Beowulf is polysyllabic, that dip often contains an unstressed prefix (or ne); the verse may be interpreted as regular if the prefix is exempt from the count of metrical syllables.94 An example is Beowulf 74a “Ðā ic wīde gefrægn” (“Then, widely, I have heard”), which scans xxSxpS (=xxSxS) with exemption of the verbal prefix ge-. The prefix license is a correspondence rule in the Old English meter – it shows how linguistic material is matched to metrical positions – and it seems to have remained vital in the Brut. In Hanna’s 46-line sample, three of seven b-verses with an unexpected long dip may be interpreted as regular by treating prefixes as exempt from the metrical count: 12368b “þat þu swa reh ært iwurðen,” 12380b “and hine for lauerd icnawen,” and 12387b “þu scalt wursen vnder-fon.” The prefix license probably also applies in 12391b “þat þu ærst for-hoȝedest,” where the spelling ærst may represent a disyllable. This analysis leaves four verses with irregular dip structure. Of these, 12357b “and al þi driht-liche uolc” and 12371b “a þire aȝere hond” fit a robustly attested asystematic pattern, in which the syllables in the medial dip belong to the first lift-bearing word.95 It is not clear why this pattern would have been acceptable to Lawman, but it apparently was. Line 12354b “bi-uoren Arðure þan leod-kinge” may have originated as a scribal error, but it scans as a three-lift verse. The only unpatterned exception seems to be 12379b “þat he is king ouer þe.” We turn, then, to the three-lift verses, the last category of apparent irregularities, and this brings us back to a question first mooted at the end of the previous chapter, namely the historical legacy of Old English heavy verses. On the assumption that Old English heavy verses (particularly Sievers Types D and E; also variants of the other types) were projected directly from compound word-stress and tethered to that vocabulary, metrists have assumed that heavy verses would disappear together with the decline in poetic compounds. As I have suggested, one may doubt that line of reasoning. A metrical template always has a certain degree of autonomy from the linguistic material coordinated to it. A change in the language may cause

The Brut


a template pattern to be discontinued, but it may also trigger a change in the correspondence rules of the meter. If items with compound stress became less frequent in the poetic language, the number of lightly stressed function words increased and eventually exploded the meter’s basic fourposition frame. Once that had happened, the strong positions of heavy verses could be spaced out and realized by independent words. The result, it seems, was three-lift verses in Middle English. The operative entity in this transformation was not Sievers Types D and E, but rather the heavy verse as such: that is, verse types with three or four strong constituents, represented in Yakovlev’s notation as SSSx, SSxS, SxSS, xSSS, and SSSS. Sievers Types D and E were the principal way that heavy verses were manifested in the Old English phase of the alliterative tradition; this particular manifestation was subsequently eliminated, but heavy verse types persisted in Middle English, now manifested as verses with three lifts. The prefix license and resolution derive from the Old English morphological meter, while three-lift verses are an accentual reflex of the morphological meter. In both cases, features of the older meter persisted – or were reinterpreted – within a later and basically accentual system. It remains to show that the Middle English b-verse contours could have been derived from the Old English meter. This involved just two basic innovations in the metrical system, both interpretable as accommodations to analytic syntax and the attendant increase in lightly stressed function words: 1. The expansion of an initial or medial dip – optional in the Old English meter – became mandatory. The result was the accentual patterns Sx… xSx, x…xSSx, and x…xSxS. The pattern xSx…xS seems never to have been common in the b-verse, possibly because a third-position dip was not expandable in Old English. 2. The four-position requirement fell away. The result was the accentual patterns xSx…xSx, x…xSxSx, and the various three-lift patterns. Precedent for this transformation was probably set by the metrically ambiguous status of prefixes and of syllables in resolution. The explosion of the four-position frame could have been fatal to the meter. The fact that it was not is due largely to the principle of closure, which pushed the most complex realizations to the a-verse, and thus maintained the legibility and coherence of the line. This process of sifting-out and reorganization was already in progress in the Brut, where three-lift verses are about twice as frequent in the a-verse as in the b-verse. Thus described, the Brut’s meter is astonishingly complex, but also stands in clear historical relation to both earlier and later alliterative poetry.


The Origins of the Alliterative Revival

The metrical theory presented in the foregoing pages is Yakovlev’s. My own engagement with the Brut has been predominantly in the mode of verification – that is, as a test to see whether I could duplicate Yakovlev’s scansion results. In Yakovlev’s core 200-line sample, Brut 10534–733, I find 19 robustly irregular verses, against his 14.96 In two other short samples, 6907–66 and 8671–720, I find that as many as a quarter of b-verses are irregular. There does seem to be some variation from one section of the poem to another; plainly, this is an area for further research. I close, however, by returning to the passage scanned by Fulk as a test of Russom’s theory of metrical development. In the first 50 full lines of the Brut, Fulk found eight b-verses with irregular dip structure:97 3b “at æðelen are chirechen.” 13b “þe mid heom weren on archen.” 26b “& fiede on boc-felle.” 27b “sette to-gadere” 29b “alcne æðele mon” 44b “þe was feondliche stor.” 46b “þe was mid him isund.” 52b “ȝeon þare wintrede sæ.”

These verses may now be analyzed as follows. Lines 3, 27, and 29 may be classed as regular with resolution of chirechen, to-gadere, and æðele, respectively. The compound boc-felle may support a lift on its second element, by stress shift, yielding a regular contour in 26b. 46b is a light verse, lacking a categorically stressed word to support the first lift; when such verses appear in earlier and later alliterative poems, most metrists assume that a function word is promoted. 44b and 52b exhibit the same asystematic pattern we encountered above in 12357b “and al þi driht-liche uolc.” These two verses join 13b as the only robustly irregular verses in this list. None of this analysis is self-explanatory, and most of it remains quite tentative, but it does seem that we are beginning to assemble the pieces.

Generalizations: The Alliterative Tradition, c.1000–1250 Even at the current state of elaboration, Yakovlev’s analysis of the Brut supplies a robust definitional criterion, enabling us to trace the alliterative tradition between the eleventh and fourteenth centuries. The rhythmical alliterating style of Ælfric’s Lives of the Saints is related to this verse tradition as, at most, a side-spur. If one applies Yakovlev’s scansion protocols to Ælfric’s Life of St. Edmund, one of the first things one might notice is that Ælfric’s text is deficient in three-lift verses.98 Though precise counts should



be treated with reserve, I find that approximately 10 percent of half-lines in St. Edmund have three-lift constructions. That figure is within the margin of error for b-verses of the Brut, but conspicuously low in comparison with the a-verses of that poem. Whereas three-lift constructions are about twice as frequent in the a-verses of the Brut, as compared with the b-verses of that poem, the Ælfrician half-lines are more nearly symmetrical. In the 263 lines of St. Edmund I count 32 three-lift a-verses and 27 three-lift b-verses. (The designations “a-verse” and “b-verse” are unsatisfactory in reference to Ælfric’s form, but I will use them for the sake of like-to-like comparison.) A perception that Ælfric wrote in paired two-stress units is, of course, one of the main reasons that scholars have construed his rhythmical compositions as verse, and, more narrowly, construed them as an intermediate step between the classical Old English verse and that of Lawman and later Middle English alliterative poets. Three-lift constructions in later verse are taken to represent a further loosening of Ælfrician precedent, but that line of reasoning may be precisely backward. If three-lift verses are granted a systematic status in Lawman’s meter, and derived from the heavy verse types in the classical Old English meter, then the dearth of three-lift constructions in Ælfric’s compositions suggests that his rhythmical style belongs to a developmental track distinct from the one that produced Lawman’s meter. Ælfric’s form and Lawman’s are alike in that they are both basically accentual, or stress-based (a morphological scansion of Ælfric’s Lives of the Saints yields chaos), but Lawman’s meter shows signs of morphological inheritance absent or muted in Ælfric’s form. Resolution is another case in point, for it is difficult to see how this traditional metrical device could appear in Lawman’s meter, if not as a morphological inheritance. If Ælfric’s rhythmical compositions indeed stand between Lawman and the classical Old English verse, one should expect resolution to operate in them with at least as much scope and regularity as in the Brut. Evidence for resolution is mixed, however. It reduces a long final dip on 14 occasions in St. Edmund, including 46b “tō ēadmunde cynincge” and 48b “cēne and sigefæst.”99 At least 26 b-verses in St. Edmund retain a long final dip.100 Moreover, resolution also creates new anomalies: in 33b “ēast mid his scipum” and five other b-verses, resolution of the final lift yields a three-position verse.101 More rarely, resolution yields the pattern xSx…xS, also disfavored in the Brut. Both categories of anomaly also occur outside of resolving contexts. The three-position verses are particularly noteworthy, for the four-position frame of classical Old English verse was retained in later verse as a minimum: verses with four, five, or six positions are acceptable, but not three. Ælfric’s half-lines diverge in


The Origins of the Alliterative Revival

Table 3.1. B-verses of Lawman’s Brut (ll. 10534–733 and 12346–91), Ælfric’s St. Edmund, and the alliterative lines of the Middle English Physiologus Brut count xSx…xSx x…xSxSx Sx…xSx x…xSxS x…xSSx Three-lift Subtotal of Brut types

Physiologus ratio

St. Edmund





65 19 68 14 24 30 220

0.26 0.08 0.28 0.06 0.10 0.12 0.89

38 35 29 8 17 17 144

0.23 0.21 0.18 0.05 0.10 0.10 0.88

41 33 32 25 20 27 178

0.16 0.13 0.12 0.10 0.08 0.10 0.68

Unexpanded Sievers types 3 xSxSx 1 xSx…xS 6 Long dips initial and medial 13 Final long dip 0 Three-position 3 Subtotal of types irregular 26 in Brut