UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The meter of Guthlac B: a generative model Mines, Rachel 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1994-0547.pdf [ 2.67MB ]
JSON: 831-1.0087402.json
JSON-LD: 831-1.0087402-ld.json
RDF/XML (Pretty): 831-1.0087402-rdf.xml
RDF/JSON: 831-1.0087402-rdf.json
Turtle: 831-1.0087402-turtle.txt
N-Triples: 831-1.0087402-rdf-ntriples.txt
Original Record: 831-1.0087402-source.json
Full Text

Full Text

The Meter of Guthiac B: A Generative ModelbyRachel MinesB.A., The University of British Columbia, 1992A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMASTER OF ARTSinTHE FACULTY OF GRADUATE STUDIESDepartment of EnglishWe accept this thesis as conformingtoAhe required standardThe University of British ColumbiaJune 1994© Rachel Mines, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.__________________________Department of________________The University of British ColumbiaVancouver, CanadaDate Jk(1_1 i+ iH’1DE-6 (2/88)11AbstractThe approach to Old English (OE) poetic meter traditionally taken is todescribe the meter in terms of a list of foot or. verse (half-line) types. It has beensuggested that this approach, however, is open to criticism on several pointsFirst, a list of metrical types is unconstrained in that there is no principled reasonwhy other members may not be added to the list. Second, such a theory includesno constraints on substitutions; any metrical type may always be substituted forany other. A description in the form of a list therefore cannot rule out unmetricallines (Halle and Keyser “Iambic Pentameter”).This thesis proposes a model of OE poetic meter based on Hanson andKiparsky’s parametric theory of universal meter. Hanson and Kiparsky argue thatthe constituents relevant to meter are not arbitrary or conventional, such as a listof foot or verse types, but are just those that are also relevant to language. Theypropose that all pOetic meters are comprised of binary feet, which, like thephonological constituents defining prominence in language, consist of a strong(S) member which is the head, or prominent position, and a weak (W) memberwhich is an unprominent position. Structure parameters establish headedness(either SW or WS) and the number of feet in a line, A position parameter definesthe maximal amount of prosodic material that may occupy a given metricalposition in terms of phonological constituency: mora (,t), syllable (G), foot (0), orword (2.). Prominence rules define first, whether S positions must containprominent constituents and/or whether W positions must contain unprominentconstituents; and second, whether prominence is defined by weight, strength, orstress (“Best of all Possible Verse”).The model I have proposed for OE defines the meter in terms of a fixednumber of binary left-headed (SW) feet together with constraints on both S andW positions: S positions must contain stressed syllables, further defined as the111heads of prosodic words; and W may contain the heads of prosodic words only ifthey are prosodically weak. No metrical position may contain more than aminimal word (?.min).ivTABLE OF CONTENTSAbstract iiTable of Contents ivAcknowledgements vIntroduction 1Chapter 1 What is a generative metrical theory? 4Chapter 2 Background to studies in OE meter 13Chapter 3 OE metrical phonology 41Chapter 4 A generative model 55Chapter 5 Special licenses and functional constraintsRelaxation of prominence constraints oninitial positions 84Empty metrical positions 86Extrametricality 88Chapter 6 Overgeneration and rare verse typesOvergeneration 95Metrical complexity 96Overlap 98Nonexisting and rare verse types 103Chapter 7 Alliteration 106Chapter 8 Hypermetrical verses 123Conclusion 127Works cited 135Appendix 1 Rules summary 141Appendix 2 Sample scansion 143VAcknowledgementsI would like first of all to aknowledge the members of my committee: GernotWieland for his academic and moral support; Kristin Hanson, for encouraging meto get the details exactly right, whether I wanted to or not; and StephenPartridge, for his willingness to put his time and energy into a subject rather farafield of his usual areas of research.I owe a debt of gratitude to the University of British Columbia for thefinancial support which made it possible to give my full-time attention to thisproject.I owe a further debt of gratitude to Paloma Housing Co-op and the federal coop housing program (now sadly defunct), which have made it possible for me topursue an academic career while housing myself and my daughter safely andaffordably. I would also like to thank my friends at Paloma, especially Leah,Margaret, Elaine, and Carmen, for sticking with me during the production of thisthesis with humour and patience, no matter how boring and one-tracked I got;special thanks to Diane Fans for her consultative role in the “Greenhouse Effect”(discussed on 77-78). Thanks also to my daughter Sarah, for getting her owndinners (and sometimes mine) and keeping the stereo turned down while I was inthe throes of creation; and to my cat Max, for being warm, furry, and exceedinglystupid whenever I needed a break from it all.This thesis was to a large extent produced on a caffeine high generated by theexcellent cappuchinos produced by the proprietors of the Calabria coffee bar onCommercial Drive, Vancouver.1IntroductionIn 1987, Geoffrey Russom published Old English Meter and LinguisticTheory,1 a study which situates Eduard Sievers’s earlier descriptive model of OldEnglish (OE) poetic meter within a framework based on linguistic principles.More specifically, Russorn replaces Sievers’s list of five metrical types, whichrepresent the various patterns of stressed and unstressed syllables occurring in OEhalf-lines (or verses), with a list of verse-types which have certain features incommon: each verse is composed of two feet, and each foot is derived from thestress pattern of an OE word. Russom’s model is thus a theoretical improvementover Sievers’s in that he replaces a list of metrical types which has no apparentmotivation (in that there is no principled reason given as to why only these typesand no others appear in the poetry) with a list of metrical types based uponcertain phonological properties - the stress patterns of words - of the OE language.What I propose to do in this paper is to suggest a number of ways thatRussom’s reanalysis of Sievers may be improved upon in terms of bothdescription and theory.First, the model which Russom presents is not a single or uniform meter, but alist of metrical subtypes, each verse of which, as mentioned above, conforms tothe stress pattern of two OE words. But this analysis forces Russom into a numberof inconsistencies. Unstressed prefixes (such as ge-, on- etc.), for example, must bedefined as words in order to allow a foot boundary to fall between a prefix and itsstem; while in other cases a phrase composed of two major-category words (suchas an adjective + noun) must be treated as though it were a single word in orderto allow it to occupy a foot. In the latter case, syntactic rules must be invoked inorder to determine the placement of the foot boundary; the two words which1Subsequently referred to as OEM2form a syntactic unit are treated as if they were a single word and may thereforeoccupy a metrical foot. However, syntactic criteria never play a role in the case ofa prefix + stem unit, which may not appear as a foot. Thus Russom’s definition ofwhat constitutes an OE word, which is crucial to his theory, forces him to adoptad-hoc rules in order to describe the placement of foot boundaries. I propose toavoid these problems of definition by bypassing Russom’s “word-stress” level toshow that OE metrical foot patterns are not derived directly from the stresspatterns of words. Instead, I shall argue, both the stress patterns of OE words andconstraints on OE meter are governed by the phonological rules which assignprominence (or stress) in language. Furthermore, with the elimination ofRussom’s “word stress” level, which necessarily generates a list of metricalsubtypes, it becomes possible to reduce the various metrical patterns proposed byboth Sievers and Russom to a single, consistent, pattern.Secondly, rules of Russom’s metrical model are language-specific and notgeneralizable to other metrical systems. Hanson and Kiparsky have recentlysuggested, however, that rules of poetic meter, like rules governing generativegrammar, are anchored in universal principles. They propose a parametrictheory of poetic meter, based on phonological principles, from which, theyargue, meters optimal in terms of a given language’s phonology are derived. Allpoetic meters, that is to say, appear to have certain structural features incommon, being based on universal principles of phonology, just as all humanlanguages are composed of syllables that are themselves arranged into higherorder structures such as prosodic feet and words. If Hanson and Kiparsky arecorrect in their claim, OE meter is exactly like other meters in that its rules makereference to the same structures that determine phonological prominence inlanguage. A key difference, therefore, between my model and Russom’s is that inRussom’s theory, OE words define the meter, whereas the theory about to be3presented here is consistent with Hanson and Kiparsky’s proposed universalmetrics in that a single underlying metrical pattern, together with rulesconstraining the placement of prosodic constituents on metrical positions,regulates the appearance and placement of words.In sum, then, I would like to propose two kinds of improvement to Russom’stheory: first, descriptive, in that Russom’s theory can be made more internallyconsistent; and second, theoretical, in that his theory can be made to conformmore closely with other metrical theories, thus situating OE meter within theframework of universal metrics.4Chapter 1What is a generative metrical theory?The intent of this thesis is to place Old English (OE) poetic meter into theframework of generative metrical theory first proposed in the works of OttoJespersen, Halle and Keyser (“Iambic”), and Paul Kiparsky (“Stress”, “Rhythmic”).These theorists worked primarily with iambic pentameter, seeking to replace thetraditional description of this meter - five feet of alternating unstressed andstressed syllables varied by the occasional substitute foot - with a descriptionbased on generative principles: a description, that is, built on the notion that anunderlying abstract metrical pattern, together with rules for matching thisabstract pattern with the poetic language, will generate a metrical line.Geoffrey Russom (Old English Meter) takes a generative approach to OE poeticmeter in his attempt to replace the descriptions of Sievers, Pope, and othertraditional OE metrists2with a model based on linguistic principles. Russom’smodel, however, has some descriptive and theoretical shortcomings. First, thereare problems with his definition of the OE word, which leads to inconsistenciesin his rules for the placement of foot boundaries. Second, Russom proposes notone meter for OE, but a list of metrical subtypes, which is in contradiction to theassumptions of generative metrical theory (Halle and Keyser “Iambic” 222,Hanson and Kiparsky 2-3). Third, Russom’s metrical rules for OE are notgeneralizable to other languages, whereas it has recently been proposed that rulesof poetic meter are based on universal principles (Hanson and Kiparsky).Therefore my intent is to reanalyze his model in order to develop a model of OEmeter which more accurately reflects the findings of recent work in generativemetrics,2Sievers, Pope, Bliss, Cable, and Creed are the best known of these.5But first I would like to spend some time on the question of just what agenerative metrical theory is. After all, Sievers’s descriptive theory has been thepredominant model of OE meter for the past hundred years, and a number ofalternative accounts have also won their followings. In what way are thesealready established traditions less than satisfactory? Why should they beimproved upon, how can they be improved upon, and what, if any, poetic orlinguistic principles should lie behind such an attempt? If the sheer multiplicityof efforts to improve on Sievers’s principles is any indication, the answers to thesequestions have not yet reached consensus. Perhaps there never will be aconsensus, but I believe that recent advances in our knowledge of linguistics,particularly metrical phonology, will allow us to at least come closer than wehave in the past to some more satisfying answers.Why do we perceive poetry as poetry? What differentiates poetry from theeveryday use of language? I shall consider two possible answers to this question.The first possibility is that people are taught by their culture what theconventions of poetry are; formal features of poetry are not objective facts of thetext but are entirely a product of their interpretation (Fish). The secondpossibility is that poetic devices such as rhyme and meter exist independently ofinterpretation. If this is the case, the question follows: what are the structuralunderpinnings of the formal features of poetry? Why do certain features ofpoetry (such as rhyme and alliteration) exist, while others (such as a rulerequiring every third word of a line to contain the same number of sounds) donot? Paul Kiparsky argues, first, that formal features of poetry do existindependently of interpretation; and, secondly, that these formal features derivefrom the innate capacity of human beings to understand and produce language(“On Theory”, “Role”).6There are problems with the idea that all formal features of poetry are to beidentified with cultural convention, although, like language in general, there arecertain aspects of poetry which may be explained in this way. For example, justas the choice of formal or informal diction and syntax may depend on a speaker’ssocial context or a writer’s intended audience, a poet may similarly choose towrite in either rhyming couplets or blank verse. And certainly the language inwhich poetry is composed is determined by the audience for whom it isproduced and the poet who produces it. Urdu poetry, untranslated, is not likelyto be much appreciated by English speakers with no knowledge of this particularlanguage. But these are fairly trivial examples (though perhaps not withoutinterest in their own right). More to the point, there is a strong case for the claimthat certain formal features of poetry, such as rhyme, alliteration, meter, and soforth, are constrained by the same linguistic rules that govern all language. Wereit otherwise, as Kiparsky suggests, we might expect a great variety of poetic ruleswhich do not in fact exist (“Role” 12).There are, for example, no schools of poetry which require identity in thethird sound of every word. Why should this be so? The rule is simple enough,and, as Kiparsky (13) puts it, a visiting Martian, whose brain may processlanguage quite differently from ours, would perhaps find such a rule quite aslogical as our rules for rhyme and alliteration. The difference between such anon-occurring rule and an actually existing rule is that human languageprocesses do not count sounds; the relevant phonological processes for all humanlanguages are processes that take account of sounds only when they are arrangedinto particular structures, such as syllables, feet, and words. OE alliteration, forexample, is not a process that involves the first sound in a word (even thoughthis might be a convenient shorthand way of describing it). Rather it is a morecomplex phenomenon which depends on syllable structure, and the repetition7not of word-initial sounds, but of initial constituents of word-initial (or, moreaccurately, stem-initial) syllable onsets.3Roman Jakobson, one of the founders of the Prague School of linguistics, wasamong the earlier scholars to recognize the possibility that poetic forms areconstrained by the same rules of phonology and syntax that constrain ordinarylanguage. Says Jakobson: “Poetics deals with problems of verbal structure.Since linguistics is the global science of verbal structure, poetics may be regardedas an integral part of linguistics” (350). He notes further that the poetic functionis not confined to poetry itself, but is a fundamental component of all language.We may say Joan and Margery, for instance, rather than Margery and Joan; notbecause we prefer Joan to her sister, but because it “sounds smoother” (357). Inother words, according to Jakobson, we are aware at some level of the rhythm oflanguage, even if we cannot enunciate the reasons for our linguistic preferences -in this case, the preference for roughly similar intervals between stresses.But if the poetic function is present in all language, what differentiates poetryfrom prose? Jakobson suggests that the essential difference between the two liesin the poet’s arrangement of constituents (such as syllables, word stresses, orsyntactic constructions) into units of measure which are in some way equivalentto each other. The speaker or writer of prose, on the other hand, follows no suchconstraints in the ordering of constituents.According to Jakobson, there are two basic modes of arranging components oflanguage: selection and combination. Selection refers to the choice of element: forexample, the choice of one word out of a number of synonyms. Arrangement hasto do with the order in which these elements are placed. Says Jakobson: “Thepoetic function projects the principle of equivalence from the axis of selection into the3For a more complete account of OE alliteration, see Chapter 7.8axis of combination” (358)4. When a speaker or writer of prose composes asentence designed to communicate a particular point, she or he selects wordsfrom a number of equivalents or synonyms and combines these constituents intovarious orders. For example, the sentences: “I am writing this paper” and “Thisthesis is being composed by Rachel Mines” communicate roughly the sameinformation, despite differences in word selection and word order. A poet, onthe other hand, arranges the words she or he has selected into equivalentmeasures. Words, that is, are combined into equivalent units, such as parallelsyntactic structures (as found in modes of poetry that are based on syntacticparallelism) or poetic feet (as in meter). It is this repetition of equivalent abstractstructures that differentiates poetry from prose.Kiparsky, building on Jakobson’s ideas, speaks of poetry as involving not onlyrecurrence or repetition of abstract structures, but of abstract structures whichare matched with equivalent linguistic units, that is, “certain patterns . . . whichare filled by linguistic (syntactic and phonological) elements” (“Role” 12). Theseabstract patterns, together with the “sames,” or equivalent grammatical units -sounds, syntactic patterns, word stresses - with which they are matched, are thebasic building blocks of poetry. Therefore a theory of poetry, according toKiparsky, must address the following two questions: What patterns are relevant?and What linguistic sames are relevant? (13).Although such abstract patterns as the five recurring feet of iambicpentameter or the alliterative patterns of OE verse are not obligatory elements ofordinary language, Kiparsky claims that the linguistic “sames” which fill theseabstract patterns are just those which are relevant in grammar (13). Formal41n Saussurian terminology, the paradigmatic axis and the syntagmatic axis,respectively.9features of poetry are not arbitrary or conventional, that is, but are a result of howlanguage itself is structured. It follows from this claim that there must beuniversal principles of poetry, just as there are linguistic universals which holdtrue across all languages. This makes possible the attempt to uncovermeaningful, grammar-based rules for meters in general, and for OE meter inparticular.But assuming that meter consists of a repeating, underlying abstract pattern,as Jakobson and Kiparsky claim, the situation remains that there is not always aone-to-one correspondence between poetic language and meter. First, asJakobson points out, a given line of poetry is subject to variation in how it isdelivered. Jakobson maintains, however, that meter is independent of anyparticular form of delivery. For example, the meter of the first line ofShakespeare’s sonnet 29:When in disgrace with fortune and men’s eyesremains unchanged whether or not one chooses to stress the initial word. Agiven line of poetry may lend itself to various scansions, yet its meter remainsconstant. “[Mjeter - or in more explicit terms, verse design - underlies thestructure of any single line. . . . The verse shape of a poem remains completelyindependent of its variable delivery” (364, 367). If this is so, it presents severalproblems for the traditional theories of OE meter, particularly those of John C.Pope and his successor, Robert Creed. I shall discuss Pope’s theory in Chapter 2.Secondly, not every line written in iambic pentameter consists of exactly tensyllables, arranged into five feet of unvarying unstressed and stressed syllables.Language-level mismatches with the underlying metrical pattern, or verse shape,as Jakobson puts it, are not random, however, or due to poetic license, butconstrained by very precise sorts of rules. Otto Jespersen was the first scholar toformalize such constraints on poetic practice as found in iambic pentameter.10Allowable deviations from the basic metrical form of five feet of alternatingunstressed and stressed syllables are traditionally accounted for by the idea ofsubstitute feet: a poet may substitute, say, an occasional trochee, spondee, ordactyl for an iambic foot. “Once a metrical pattern has been implied in a poem,we can say that variations in the rhythm occur through the introduction ofsubstitute feet which here and there replace certain of the base feet” (Fussell 33).Jespersen notes, however, that this traditional account is deficient on severalgrounds. First of all, it cannot explain why, while the substitution of trocheesinto iambic verse is acceptable, the insertion of an iamb into trochaic verseproduces an unmetrical line (73). Secondly, it cannot account for thedistribution of trochaic substitutions in iambic verse. Trochaic inversion, or thesubstitution of a trochee for an iamb, cannot be solely due to poetic license, saysJespersen, since different poets writing in iambic meter at different times, andeven in different languages - he cites German and Danish examples, as well asEnglish- follow very nearly the same rules: such substitutions occur far morefrequently in the first foot than in the third and fourth, and only very rarely inthe second (73). Were these substitutions merely at the whim of the poet, whomay insert a trochee “here and there,” to quote Fussell, one might expect a moreeven distribution. Jespersen concludes that there must be some rule to accountfor the placement of trochees in iambic verse, and proposes that trochaicinversion can take place only after a natural pause or syntactic break, oftensignalled by punctuation (81). Syntactic breaks most frequently occur linefinally, may occur toward the end of the line, but only very rarely occur after thefirst foot. Trochaic inversion, then, is not an arbitrary departure from somemetrical norm, but is licensed by linguistic features of the text.Halle and Keyser (“Iambic”) concur that the traditional account of iambicpentameter has a major shortcoming in that the theory of substitutions cannot11account for the pattern of placement of trochaic inversions. But moreimportantly, they claim, the traditional account cannot even adequatelydifferentiate between metrical and unmetrical lines, or, to put it in strongerterms, between poetry and prose. Halle and Keyser point out that an unmetricalline such asOde to the West Wind by Percy Bysshe Shelleyis perfectly acceptable according to the standard theory of substitutions, which isunconstrained as to which or how many substitutions may appear in a given line;but to judge such a line as metrical is an undesirable consequence of the standardaccount (221). The problem with the standard theory, they point out (222), isthat allowable substitutions for feet of the base meter are dealt with in terms of alist: trochees, spondees, and so on; and such a list is in principle unconstrainednot only as to which or how many substitutions may appear, but in that there isno reason why other items- words beginning with “w,” words with exactly threephonemes, empty feet, and so forth - may not be added to it. Since such listscannot be constrained except by convention, and since poetic meter, as I hope Ihave made clear, cannot be explained by convention, such a list can serve adescriptive function only; it may provide some indication of what is found in thepoetry under consideration, but it has no principled basis on which to makepredictions about what is or is not metrical. This problem of the insufficiency oflists- their inability to encompass generalized principles of language and poeticform - will feature in my discussion of Sievers’s theory, which is presented as a listof five acceptable metrical “types,” in Chapter 2.If a list of foot or verse types, such as spondees, trochees, etc. in iambicpentameter, and Sievers’s five types, in the case of OE poetry, is insufficient for anadequate theory of meter, what is sufficient? Halle and Keyser propose that ametrical theory, like that of generative grammar, which models the rules12defining linguistic competence, should consist of two parts: first, the abstractmetrical pattern which underlies any given line of verse; and second, rules whichgovern ways that the abstract pattern may be realized in the linguistic materialthat makes up that line (“Iambic” 223). In accordance with Kiparsky’s suggestionthat meter involves the matching of an abstract pattern with linguistic “sames,”or equivalent grammatical units, I shall propose such a model for OE poetry inChapter 4. But first, in Chapter 2, I shall examine a few of the more importanttraditional approaches to OE meter. Chapter 3 will consist of a brief discussion ofOE metrical phonology.13Chapter 2Background to studies in OE meterIs OE verse metrical?The fact that this question has been asked, and by more than one scholar ofOE, is testament to the fact that OE verse seems to work on principles verydifferent from those of more familiar Modern English meters. The absence of aneasily (to our ears) recoverable underlying meter, such as that of iambicpentameter, together with the fact that familiar line-ending devices such asrhyme and punctuation are likewise absent, help to contribute to this idea. Atthe same time, some of the parallels between structural features of OE poetry andthose of OE prose, such as the fact that both poetry and prose tend to dividenaturally into two-stress verses or phrases, have led several theorists to considerthe possibility that there is actually very little difference between the two.This idea that OE poetry is a type of rhythmic prose was proposed by JamesRouth in 1923, who suggested that Sievers’s five metrical types “represent asimple, rudimentary, instinctive, and even primitive form of musical, or at leastrhythmical, expression” (429). Anglo-Saxon poetry, he claims, is nothing otherthan prose which has been rhythmically adapted, by the placement of stresses atregular intervals, to the requirements of song or chant. This idea was laterexpounded in more detail by Marjorie Daunt, who in 1946 again raised thequestion of a close connection between the structures and the rhythms of proseand poetry. Pointing out that the labels of Sievers’s five types5 are related to theirfrequency of occurrence in the poetry (A being the most frequent, E the least),5For a description of the five types, see (2.le).14Daunt claims that A appears the most frequently6 not because it happens torepresent a trochaic rhythm or meter, but because “it is the shape of nouns andadjectives grouped together, and nouns and adjectives occur most frequently inthe spoken language” (291). It is not surprising, she claims, that in a languagewith word stress on the initial syllable and many disyllabic nouns and adjectives,the trochaic pattern / x / x will appear often. Daunt’s analysis of 200 half-lines,taken at random from Beowuif, provides some support for a kind of relationshipbetween verse-type and grammatical category: A-type verses, she claims, tend tobe composed of nouns and adjectives, B-type verses tend to end withmonosyllabic verbs, C-type verses tend to be composed of prepositional groups orclauses, and so forth. Daunt concludes: “These groups or patterns are the shapethey are because the language itself is that shape and not because the poetarranged them” (293). An analysis of two brief passages of OE prose, which shefinds metrical according to Sievers’s five types when divided into phrases, lendsat least some support to this argument; some of the prose phrases she examinesscan well as poetry, although others do not.7Thomas Cable, like Daunt, examines the structure of OE prose as compared tothat of poetry, but comes to quite the opposite conclusion. Citing AngusMcIntosh’s 1949 study of Wulfstan’s prose, he notes two striking features aboutthe analysis; first, “the way in which the prose divides naturally into two-stressphrases”; and secondly, the fact that almost 50% of these phrases may be scanned6Sievers type A is represented by /x/x, with / indicating a stressed and x anunstressed syllable.‘ For example, while Ohthere sa?de is an acceptable A-verse, buton on iwumstoTwum is not an acceptable C-verse, as Daunt claims (295). It is more like an Atype with polysyllabic anacrusis. (Anacrusis is an extra syllable or group ofsyllables that sometimes appear before the initial stress in a verse; it will bediscussed further in Chapter 5.) Verses with polysyllabic anacrusis, whileprobably not unmetrical, are extremely rare in the poetry.15as x I x / x, or, in Sieversian terminology, as A with anacrusis (Meter 39).Anacrusis, however, is a very limited feature in Beowuif, occurring, by Cable’sestimate, in at most only 125 half-lines, that is, about two percent of all verses(37). According to Cable, the avoidance in poetry of this pattern, which is socommon in prose, is a constraint imposed not by morphology or by syntax, butby meter (43).Another argument for the metricality of OE poetry is one from alliteration.Almost all OE metrists point out the close relationship between alliteration andmeter, even though the nature of that relationship has sometimes been a matterof dispute. Russom argues that alliteration is an integral part of the metricalpattern itself, and cannot be explained without reference to the underlyingmeter (OEM, “New Kind”), as we shall see below. In fact, one metrist, DavidHoover, argues that alliteration is the meter, and that the rhythmic patterns ofOE poetry are insignificant in themselves.I shall take as my starting point the premise that OE poetry is metrical(though I do not necessarily agree that either the avoidance of anacrusis or thealliterative patterns of the verse are in themselves sufficient to establishmetricality); and if this is the case, it ought to be possible to formalize the rulesfor that meter. But while I disagree with the conclusions of Routh and Daunt,their observations are nonetheless of great value, for they indicate that themetrist must pay close attention not only to the stresses within OE words, but tothe stress patterns that also result when words are grouped together into largerunits such as phrases and verses.Some preliminary observations on OE word stress and alliterationIn order to clarify the following discussion, I shall describe, in very generalterms, a few of the generally held assumptions about word stress and alliteration16in OE. A more detailed account of OE phonology follows in Chapter 3;alliteration will be dealt with at greater length in Chapter 7.OE almost invariably places primary stress on the initial (or only) syllable ofthe word stem, the only exceptions being some prefixed nouns and adjectiveswhich have primary stress on the prefix. Otherwise, prefixes are unstressed.Syllables traditionally scanned as bearing secondary stress fall into two categories:first, a heavy stressed syllable which is immediately preceded by a heavy syllablebearing primary stress, such as the medial syllables of lto/jitan, earfride,dreo4ne; second, the initial (or only) syllable of the stem of the second lexicalelement of a compound, such as those in woruld-e, mon-drvh ten, feorh-geJ.Syllables generally considered to be unstressed include, as mentioned above,most prefixes; suffixes, including inflectional and derivational endings such as-fl and -weard; and function words, both mono- and disyllabic.Alliteration, or the repetition of stem-initial sounds (very roughly speaking,and excepting words with stressed prefixes, which alliterate on the prefix), acts tobind together the two verses, or half-lines, that make up an OE poetic line.Alliteration is dependent on word stress in that only stressed syllables mayalliterate; these are generally syllables bearing primary stress, or, more rarely,syllables which head the second lexical elements of compounds, which havesubordinated stress. Unstressed syllables, whether prefixes or function words,may not alliterate; or rather, although they may begin with the sound thathappens to form the alliterative pattern of the line, the identity of sounds isdisregarded. Alliteration on a preposition such as tU, for example, does not“count” in a line in which the alliterating element is [t]. The next section, as wellas Chapter 7, will discuss alliteration in more detail.17Traditional theories of OE meterMost traditional approaches to OE meter fall into one of two camps: theSieversian school and that of John C. Pope. In very different ways, these twometrists have provided insights into OE metrical theory which have been crucialto not only the traditional theories which have built on their work (such as Bliss’sand Creed’s), but to generative theories as well, including Russom’s and my own.It is impossible to discuss OE metrical theory without acknowledging thepioneering work of Eduard Sievers. Although his theory, first published in 1893as Altgermanische Metrik, has not been received without criticism, it seemslikely that his “Five Types,” as simple, descriptive labels for the rhythmicpatterns into which OE half-lines fall, are here to stay, even among those whofind fault with his analysis.Based on his extensive statistical survey of Old Germanic alliterative poetry,which of course includes OE verse, Sievers offers the following generalizations:8(2.1) a. Each long line consists of two verses or half-lines which areconnected together by alliteration. Each verse “must be agrammatical unit, i.e. it must contain a free separable clause”(279).b. The standard verse consists of four, occasionally five, segments.Two of these segments, known as rises (symbolized by I), areusually syllables which bear primary stress; more rarely theymay bear strong secondary stress. As a rule, rises are heavysyllables;9 however a rise may consist of a light stressed8A11 references to and quotations of Sievers’s work are taken from Gawaina D.Luster’s translation of H. Paull’s Grundriss der germanischen Philologie, 11.2(Strassburg, 1905), pp. 1-38.9Sievers actually calls these long syllables, as do many other metrists. Tobetter reflect modern linguistic terminology, and to avoid confusion between18syllable together with a following light or heavy unstressedsyllable. This metrical equivalence between a light syllablefollowed by an unstressed syllable on one hand, and a singleheavy stressed syllable, on the other, is called resolution (271).c. Segments carrying weaker stress (dips or falls) are usuallyunstressed syllables; however they may bear a secondary stress(symbolized \) (271). Syllables with secondary stress aregenerally heavy, but may be light if they are immediatelypreceded by a rise (272). One or several consecutive unstressedsyllables (symbolized x) may function as a single dip.d. Rises and dips combine together into metrical feet of one, two, orthree members. A foot with one member consists of a rise (/);one with two members consists of a rise and dip in either order(/ x or x /); one with three members consists of a rise,secondary rise, and dip; or rise, dip, and secondary rise (I \ xor /x\)(272).e. Metrical feet combine in the following five patterns (For purposesof illustration, I have included verses from Beowulf’°corresponding to each type):A / x I / x gomban gyldan ‘to give tribute’ (1 la)B x / I x I on flëãm gewand ‘turned in flight’ (lOOlb)C x I I I x gefëän habban ‘to have joy’ (274Db)long syllables and long vowels, I shall refer to syllables which contain a longvowel or are closed with a consonant as heavy. Syllables containing a short voweland which are not closed with a consonant are light.‘°Pill quotations from Beowulf in this paper are from Klaeber’s edition. I havehyphenated compound words as aids to scansion.19Dl I I \ x wis we1-ungen ‘wise [and] honored (1927a)D2 / I I x \ Fyrst forô gew5t ‘time passed’ (210a)E I \ x I I glo-mannes gyd ‘singer’s song’ (1160a)/ x \ I I mo4or-bed strêd ‘spread a murderous bed’(243 6b)Sievers does not acknowledge a / x I x / pattern, since two congruent unstressedsyllables always count as a single dip (273).In addition to these above basic patterns there are a number of sub-varieties:Type A with a secondary stress in place of one or both dips, for example, or typeA3, which has no alliterating rise in the first foot. D patterns may be extended bythe addition of an unstressed syllable immediately following the first rise. InType C it is not uncommon for the second rise to fall on a light rather than aheavy syllable (274-5).Two half-lines are bound together with alliteration, or the repetition of steminitial sounds, to form a line. Only stressed words alliterate; unstressed wordssuch as prepositions, conjunctions, and the like, are not involved in thealliterative pattern of the line. Consonants alliterate whether they precede avowel or another consonant (e.g. helm, ‘helmet’ alliterates with hla7ord, ‘lord’),with the exception of st, sp, and Sc, which alliterate only as clusters. All vowelsalliterate with each other (e.g. andsaca, ‘enemy’, alliterates with ellen, ‘courage’).The first rise of the second half-line (or off-verse) alliterates obligatorily; thesecond rise of the off-verse may not alliterate. The first half-line (or on-verse) mayhave one or two alliterating stressed syllables, or supports. If there is only onesupport, alliteration falls on the stronger rise, which is almost always the first; anexception being type A3, which has no alliterating stress in the first foot (276-7).Difficulties with Sievers’s theory fall into two general areas, which I shalldiscuss in turn. The first problem has to do with the relationship between20metrical positions (rises, secondary rises, and dips) and the linguistic materialwhich makes up the line; correspondence rules, or rules constraining matchingbetween linguistic units (stressed or unstressed syllables, for example) andmetrical positions are either applied inconsistently or are lacking altogether. Thesecond problem, which arises from the first, has to do with the insufficiency oflists as descriptions of what is allowable in a meter.Dips, first of all, may contain one or more unstressed syllables, as Sieverspoints out (272):(2.2) a. Qft Scyld Scftg ‘often Scyld Scefing’ (Beo. 4a)x / / x (C)b. et hTë on b healfa ‘that he on both sides’ (Beo. 1305a)x / /x(C)Verse-final dips, however, must contain a monosyllable. The constructed versebelow is therefore unmetrical in Sievers’s system:(2.3) *grt gst-hligne ‘[hel greeted the holy one’11I I \ x (Dl?)Some dips may be “intensified;” that is, they may contain a syllable withsecondary stress rather than an unstressed syllable or syllables (Sievers 273).However, these intensified dips are almost entirely confined to A-types. Sieversdoes not admit intensified B- and C-types like those in (2.4b-c):(2.4) a. fëbndes ft-i ‘enemy’s track’ 2289a)lxixb. ws gld-möd secg ‘he was a cheerful man’x I x /(B?)c, *h ws secg gked-mdx / / x (C?)11Asterisks will be used throughout this paper to indicate unmetrical verses.An asterisk preceding a given scansion indicates that the scansion is wrong,though the verse may be metrical.21Some dips may contain a syllable of primary stress, such as those in (2.5a-b)below; others, however, may not, such as the constructed unmetrical example in(2.5c), which shows a syllable of primary stress in the final dip of a C-type:(2.5) a. seofon niht swuncon ‘worked for seven nights’ (Beo. 517a)12I x I x (A)b. sc-holt ufan grg ‘grey-tipped spear’ (Beo. 330a)I x / x (A)c. *se lëof mon Md ‘the beloved man asked’x I / x (C?)The problem is not in itself that a dip may contain one or more unstressedsyllables, or even syllables of primary or secondary stress. This is a situation alsofound in Shakespeare’s iambic pentameter, in which a W position may containone or more unstressed syllables, a stressed monosyllable, or a strong syllable justin case it is line- or phrase- initial:(2.6) The expense of spirit in a waste of shameW S WSWSW S W SSvage, extreme, rude, cruel, not to trust (Son. 129)W SW S W SW SW SThe problem in Sievers’s system is that some dips in some verse types maycontain something other than a single unstressed syllable, with no generalprinciple or principles given that might account for or constrain this. What isneeded is some sort of rule analogous to the principle that a W position in iambicpentameter may not contain a strong syllable (Kiparsky “Rhythmic” 195, Hansonand Kiparsky 6); some sort of generalized statement as to what sorts of linguisticmaterial a dip may or may not contain, and under what conditions. Lacking12Note that the resolvable sequence of a light stressed syllable plus unstressedsyllable in seofon counts as a single rise as noted in (2. ib). I shall discuss thephonology of OE resolution in Chapter 3; in general, following Sievers andRussom, I shall treat resolvable sequences and heavy stressed syllables alike as asingle phonological unit,22such a rule, there is no reason why any dip in any verse type may not contain asyllable of any stress, with the result that unmetrical verses such as those in (2.3),(2.4b-c), and (2.5c) are predicted by the theory.Unlike dips, which behave differently depending on their environment,secondary rises behave quite consistently in that any secondary rise in any versetype may contain the following: an unstressed syllable (though not a sequence ofthese); a syllable of secondary stress; or a syllable of primary stress. I shallillustrate these respectively with a Di-type pattern:(2.7) a. wl liçdon ‘they well pleased’ (Beo. 639b)//\xb. fond man-ynes ‘mankind’s enemy’ (Beo. 164b)I / \xc. heard hr çjmen ‘the brave one [has] come here’ 376a)/ / \xThe situation is not, then, that secondary rises behave inconsistently; it is thatthey do not appear to be constrained by any principles whatsoever; or, if they are,these principles are not stated as part of Sievers’s theory. In fact, the necessity forthe existence of secondary rises in Sievers’s system, at least in some verse types,has come under question by some metrists, including Moulton, Bliss (Metre),and Russom (OEM).Rises generally contain syllables of primary stress, though they quitefrequently contain syllables with secondary stress, whether they are the heads ofthe subordinated words in compounds, as in (2.8a-b) below, or not (as in (2.8c)):(2.8) a. middel-hthm ‘at midnight’ (Beo. 2833a)/ x / x (A)b. geond ysne middan-geard ‘throughout this world’ (Beo. 1771b)x / x / (B)c. swylce gigfltas ‘such giants’ (Beo. 113a)x //x(C)23Rises may also contain unstressed syllables, which may be either affixes orfunction words (both of which are considered to be unstressed in Sievers’s theoryand which generally occupy dips). Rises in C-types, for example, are notuncommonly occupied by unstressed syllables, as in (2.9a) below; and the initialrise in Sievers’s A3 type has an unstressed syllable, which does not share in thealliteration, as in (2.9b):(2.9) a. ic ow wtjge ‘I will lead you’ (Beo. 292b)x //xb. w ealle ‘that we entirely’ (Beo. 94 Ia)/ x /xRises, then, like dips, may contain either stressed or unstressed syllables.Again, this is not necessarily a problem in itself; the same type of situation occursin Shakespeare’s iambic pentameter, where a S position may contain a stressed oran unstressed syllable:(2.10) When disgrace with fortune men’s eyes (Son. 29)W SW S W S W S W SHowever, with the lack of any sort of generalized constraints on either dips,secondary rises, or rises in Sievers’s system - if any position, that is, may contain aconstituent bearing any degree of stress - the system collapses as a theory. Likethe theory of substitutions in iambic pentameter, such a system cannot rule outunmetrical verses; it cannot distinguish between poetry and prose.Even in regard tO the metrical verses which appear in poetry, Sievers’s systemruns into problems, since, as we have seen above, many verses do not exactlymatch one of the five types. Given a verse that is not a very good match, how arewe to decide what type to match it to? Consider, for example, the following:(2.11) geolo-rand t gtioe ‘yellow shield to battle’ (Beo. 438a)Is this an E-type with an extra dip in the final position? An A-type with twosyllables (one which happens to be stressed) in the medial dip? Or is it24unmetrical? The result is the proliferation of metrical types and subtypes in aneffort to accommodate such anomalous (and some not so anomalous) verses. AnA-type with two syllables in its medial dip simply gets added to the list as/ x x I x; an A-type with stressed syllables in its dips likewise gets added asI \ I \; a B-type with two syllables rather than one in its first dip likewise getsadded to the list as x x / I x I. There is no reason why such a list cannot beextended indefinitely to include, let us say, as an extreme example, a sub-type ofA with a resolved first rise, an internal dip of four syllables (one disyllabic andtwo monosyllabic function words), two syllables of anacrusis, and a final dip of aclosed syllable, with the alliteration being on “w.” The logical (and ridiculous)result of indefinitely extending a list of sub-types in this way would be that everyhalf-line would form its own category. This of course would defeat the purpose oftrying to abstract a metrical pattern in the first place. Although (I hope) nometrist would go quite so far, I think I have only slightly exaggerated thepossibilities; Bliss’s listing of metrical types and subtypes, which is based onSievers, runs to 213 members, while Pope admits 107 sub-types of A, 58 of B, 39of C, 58 of D, and 17 of E: one for every 23 verses of Beowuif.Finally, the question arises as to why Sievers’s five types (which, let me add,despite the theoretical flaws discussed above, still capture important descriptivegeneralizations) should exist at all. As Cable puts it, “[t]he obvious question toask is why Old English meter should consist of exactly the patterns that Sieverspresents and no others” (Meter 84), The theoretical issue at stake here is the factthat lists such as Sievers’s have no adequate principles of exclusion, no rules bywhich additions to the set of members may be screened out. Since Sievers allowsa variety of feet and a variety of ways in which these feet may be combined intohalf-lines (not to mention the variety of ways in which syllables may be matchedto metrical positions), there is no reason in principle why metrical feet such as25those in (2.12 a-b) or a combination of feet into verses such as those in (2.12 c-d)may not be added to the set:(2.12) a. \/ b. x\c. x/I//x d. /x\IxThere is probably some perfectly good reason why these patterns do not occur;one hypothesis might be that the stress rules applying to words or phrases of thelanguage do not allow them. But in the absence of an explicitly stated constraintas part of the theory, the exclusion of these patterns from Sievers’s list seemssimply arbitrary.In sum, Sievers’s system has the following problems: first, correspondencerules licensing the matching of prosodic constituents with metrical positions areinconsistent or nonexistent; and second, the abstract metrical patternsthemselves are unmotivated, or appear unmotivated, in the absence of explicitlystated constraints. This is not to say that Sievers’s system is useless; far from it.He has in fact abstracted and formalized to a great extent the astonishing varietyof rhythmic patterns existing in OE poetry and thereby provided an invaluablestarting point for other metrists, even if perhaps on the basis of his intuitive feelfor the language rather than through the consistent application of rules. WhatRussom contributes, almost 100 years after Sievers, is a linguistic rationale for theexistence of the five types, together with rules which constrain the relationshipbetween metrical patterns and prosodic constituents. My intent is to carry theprocess somewhat further by bringing Russom’s theory into line with universalgenerative metrical theory.A very different approach to OE meter is that taken by John C. Pope. Pope’stheory depends on “the adoption of two isochronous, quadruple measures as thefoundation for the rhythm of each normal verse and on the free substitution ofquantitative equivalents, including [musical] rests” (x). Each verse consists of26two measures; each measure contains four quarter-notes or their equivalent, suchas two half-notes. The first note of each measure receives a major stress. However,in many B- and C-type verses, which begin with an unstressed syllable, this majorstress falls on a pause or musical rest in order to allow the first measure to occupythe same length of time as the second, which will then contain the two mainstresses of the half-line. He gives the following example (39):(2.13) egsode eorlas, syoan rest wearô CBeo. 6)irttr ri .r ri riNote that a rest replaces the initial stress of the first measure of the second verse(or off-verse). These rests occur in about 30% of all half-lines (89).Pope argues that these initial rests were not necessarily silent rests, but duringrecitation, at least at the beginning of a poem or after any significant pause, musthave been filled in by the sound of some rhythmic accompaniment such as thestroke of a harp, since without such accompaniment an initial rest would not beperceived as such by an audience (90). Therefore the harp becomes an essentialpart of Pope’s theory as an external method of regulating the beat.But the concept of the missing beat or rest is rather problematic, since Pope’ssolution to the problem of light or inadequately filled measures is intimatelyrelated to performance.. While it is generally accepted, based on evidence fromthe poetry itself, that the harp was a common accompaniment to poetry, thereare problems with claiming that it was essential to it. First of all, we have noassurance that poetry was always recited to the harp, which, if we accept Pope’scase, would be a necessary assumption; in fact, at least in the case of Cedmon inthe cow-stall, it seems highly unlikely. Secondly, if the harp were essential to anappreciation of the poetry, rather than being a pleasing adjunct to it, it seemsodd that OE poems were written as they were, across the manuscript page, withno indication of where these silent rests or harp-strokes should go. But finally, if27we accept that a non-linguistic element is necessary for a poem’s interpretation,one may as well abandon the idea that poetry, or at least OE poetry, necessarilyhas any kind of structure, linguistic or non-linguistic, at all; because then anykind of non-linguistic element, not only a pause or harp, could conceivably bethe organizing structure of the meter.The most telling objection to Pope is, however, one to the very foundation ofhis theory, isochrony itself. Pope’s system demands that the two measures, ordivisions of a half-line be temporally equal; but, as Cable (“Meter” 15), Taglicht(342-3), Hoover (3-4), Bliss (107), and others have pointed out, the assumption ofisochrony as a musical principle during the time in which OE poetry wascomposed is unwarranted. Silver-Beck argues further that isochrony is a featureof modern Western musical practice, and that to impose it on OE poetry isanachronistic. Since, she claims, there is no evidence that isochrony is anythinglike a universal principle, and much evidence that it is not, we cannot assumethat it was an organizing principle in OE verse.However, Pope has made some valuable contributions to metrical theory,despite these flaws in his arguments. In some cases it is quite possible toreconcile his claims with those of generative metrical theory, and here hisperceptions may be incorporated very nicely. Most important is his claim thatmeasures must start with a strong beat. In metrical theory, this amounts to aclaim that feet are left-headed, that is, that the strongest stress falls at the leftedge of the foot; this is the mirror image of the right-headed or WS foot found iniambic meters. One may paraphrase Pope as claiming that OE feet are SW, anintuition that seems very plausible, given the trochaic stress pattern of many OEwords. Pope therefore rearranges the foot boundaries in Sievers B- and C-types sothat the first rise in these half-lines heads the second foot:28(2.14) a. SieversB: x/ I x/ xIIx/b. Sievers C: x I I / x — x I I \ x.This placement of foot boundaries is more consistent than Sievers’s in that iteliminates a situation whereby both right-headed and left-headed feet may occurin the same meter (and sometimes, in the case of C-types, in the same verse), Onthe other hand, however, it results in unbalanced verses in which the first footnow has only one constituent while the second has three. Pope explains thisimbalance by postulating a verse-initial rest. However, the short foot too has itscounterpart in metrical theory, which allows for unfilled metrical positions. Theconcepts of left-headed feet and empty metrical positions are both central to myown theory, and will be discussed in Chapters 4 and 5.Generative theories of OE meterGenerative theories, or theories in which a metrical line is generated by rulesapplied to an underlying metrical pattern, are not new to OE metrical studies.The two earliest date back about 70 years, to those of James Routh in 1923 and W.Greg in 1925. Both argue that the fundamental OE verse pattern is:(2.15) x/x/xA dip (x) may be suppressed (Greg 12) or replaced with a pause (Routh 431) toyield each of the five Sievers types.Somewhat similar is Bliss’s theory of displacement, in which the five Sieverstypes, he claims (108), are generated by displacing either forward or backwardone or both stresses of the underlying pattern:(2.16) I x (x) / x (in which (x) is an optional unstressed syllable)None of these theories, however, suggest any linguistic rationale - syntacticfactors, or principles of word or phrasal stress, for example - which may accountfor either suppression or displacement of stresses; so while these generative29theories are not without interest, especially in view of their early dates, I will notdiscuss them further.’3Geoffrey Russom argues that OE half-line patterns are not due to any sort ofpoetic convention, but are derived from the stress patterns of OE words. Theseword-derived metrical patterns also determine the alliterative patterns of the lineas a whole. The strength of his generative theory is that he thus provides a solidlink between poetic meter, including alliteration as an integral component, andcertain linguistic features of OE. Four theoretical principles outline the barebones of his theory (OEM 2)14:(2.17) a. Foot patterns correspond to native OE word patterns.b. The verse consists of two feet.c. Alliterative patterns correspond to OE stress patterns. A metricalrule that mimics the OE compound stress ruledetermines the location of alliterating syllables.d. The line consists of two adjacent verses with an acceptablealliterative pattern.Rather than approaching the meter as traditional metrists do, with referenceto half-lines containing one or two primary stresses and some variable number ofweaker stresses, Russom’s theory makes use of the word boundary as the crucialfeature of the meter. Therefore, he points out, it is important to carefully definewhat is meant by an OE word. Russom defines the OE word as the following (11):(2.18) a. All stressed simplexes count as words.13More recent generative theories of OE meter, which I unfortunately do nothave room to discuss here, include the work of Keyser, Halle and Keyser (EnglishStress), Hoover, Huettner, and Cable (English).‘4A11 quotations of and references to Russom in this chapter are from OEMunless otherwise noted. For the sake of clarity and consistency, any of Russom’srules reproduced in this chapter are in the form of direct quotes, with theexception of any explanatory footnotes I have appended.30b. Unstressed prefixes count as “function words.”c. A compound may count as one word or as two.Abstract metrical foot patterns are derived from the stress patterns of OEwords (12). OE words generate three possible metrical positions: S, s, and x. The Smetrical position is generated by a syllable which is heavy: that is, one which hasa long vowel, a short vowel closed by a consonant, or a resolvable sequence (asdefined by Sievers in (2. ib)). The s position is generated by the subordinated rootsyllable of the second constituent of a compound word. The x position isgenerated from unstressed inflectional syllables and function words, which arealso considered to be unstressed. Note that Russom defines unstressed prefixes asfunction words by (2.1 8b), justifying this on the grounds that if a prefix were anintegral part of the word it adjoined to, it would acquire stress, since OE alwaysstresses the initial syllable of a word (8).The abstract metrical positions S, s, and x may be combined into nine footpatterns: x, S, xx, Sx, Ss, Sxx, Ssx, Sxs, and Sxxs. All OL words, according toRussom, correspond to one of these nine patterns. He lists the following possiblecorrespondences (13):(2.19) Feet Corresponding wordsx ond, ‘and’; ge-, prefixS gd, ‘good’; tilu, ‘loyal’xx oe, ‘or’; ofer-, prefixSx dryhten, ‘lord’; Iolode, ‘he suffered’Ss s-mann, ‘sailor’; megen-wudu ‘power-wood’,spearSxx bealdode, ‘he encouraged’; gryrelicu, ‘terrible’Ssx s-mannes, ‘sailor’s’; sigor-adig, ‘blessed withvictory’31Sxs middan-geard, ‘middle earth’; inwit-searo,‘malicious cunning’Sxxs sibbe-ge-driht, ‘band of kinsmen’This word-foot correspondence allows for a variety of foot patterns, butconstrains them within certain limits. For instance, as Russom points out, thereare no foot patterns like xxx, Sxxxs, etc. because there are no words in OE withthese stress patterns; there are no foot patterns such as xS, xxS, etc., becauseunstressed prefixes are defined as function words by (2. 18b), not as prefixes per se(14). Since feet with rising stress are thereby not allowed, Sievers types B and Cmust be analyzed as having an initial x foot followed by a three-foot part, as Popealso claims.The foot patterns given above represent idealized, abstract metrical patterns.To the extent that actual feet and verses deviate from the underlying patterns,what Russom (following Kiparsky “Rhythmic” 194) calls “mismatches” arecreated. Russom proposes the following labelling mismatch rules that constrainand account for differences between surface language and the abstract meter (15):(2.20) Labelling mismatch rules:a. A syllable with primary stress may occupy an S position or (undercertain conditions) an s position.b. A syllable with zero stress must occupy an x position.c. A syllable with secondary stress may occupy an s position or(under certain conditions) an S position.In accordance with (2.20a), syllables with primary stress normally occupy Spositions:(2.21) ur ran ‘to proceed further’ 254a)S xISx32But sometimes a situation arises in which two words with primary stress mayoccupy a single foot; this happens when the two words form a close syntacticunit. In these cases, according to Russom, the phrase “mimics the structure of acompound” (1 7), and the syllable of primary stress in the second word mayoccupy a s position. For example, Russom (45) scans Beo. 35b as:(2.22) on bearm cipes ‘in the ship’s hold’xl S s xIn accordance with (2.20c), syllables with secondary stress normally occupy spositions:(2.23) gu-ijnc gold-wlanc ‘a warrior decked with gold’ (Beo. 1881a)S sIS sHowever, in cases in which a compound word takes up an entire verse, the firstsyllable of the second element of the compound may occupy a S rather than a sposition. For example, Russom (26) scans Beo. 504b as:(2.24) middan-geardes ‘middle-earth’s’S xiS xBracketing mismatches arise when the word boundaries in a given verse donot correspond with the foot boundaries of the underlying meter. Like labellingmismatches of the type shown in (2.22) above, bracketing mismatches may resultwhen a word group rather than an individual word occupies a foot. Russomproposes the following bracketing mismatch rules to account for these verses (16):(2.25) Bracketing mismatch rules:a. Every foot boundary must coincide with a word boundary.b. In verses with three .or more stressed words, the stressed words areassigned to feet in accordance with their syntacticconstituency.According to rule (2.25a), a foot boundary may not fall in the middle of a word(though note that a foot boundary may fall between an unstressed prefix and its33stem, since unstressed prefixes are defined as function words by (2. 18b)). A B-type verse, for example, cannot be scanned as in (2.26a) below in Russom’ssystem; the foot boundary must fall after the first word and the verse must bescanned as in (2.26b):(2.26) a. on ancre-ftest ‘securely anchored’ (Beo. 303a)Six Sb. on ancre-ftestxi Sx sRule (2.25b) allows a foot in verses with more than two stressed words tocomprise two words just in case they form a syntactic unit. Consider, forexample, the second foot of a verse such as:(2.27) brim blöde fäh ‘sea stained with blood’ (Bçç 1594a)S ISxsThe second foot of the verse in (2.27) may comprise two words rather than onebecause blöde fah forms a phrase. The foot boundary may not fall after brimbljde, because these two words do not form a syntactic constituent (16).According to Russom, a phrase such as blJde fah corresponds to the pattern of acompound word (17) and may therefore occupy a foot, which may otherwisecontain only one word by (2.17a).Finally, since OE words do not have patterns such as xxxxS or Sxxxxx, someprovision must be made for strings of function words which may appear eitherverse-initially or medially. Russom proposes that unstressed extrametricalwords, which are regarded as lying outside the meter, may appear before eitherfoot (20):(2.28) Extrametrical words may appear before either foot.A list of 25 possible patterns result from pairing foot patterns into half-linepatterns. With nine possible foot patterns, there should in theory be 9 x 9, or 81patterns; but not all imaginable half-line patterns actually occur. 18 possible34pairings with x or xx in the second foot are eliminated because, according toRussom, OE half-lines do not end in proclitics (26)15. The pairing SxxIS iseliminated because it overlaps the foot pattern Sxxs. If, says Russom, SxxlS werean allowable verse pattern, a word like sibbegedriht could occupy either a SxxISverse or an Sxxs foot. “The result would be extreme confusion about the numberof feet” (27) in a verse. He therefore proposes a general constraint on foot patterns(26):(2.29) Foot patterns may not overlap verse patterns.Most other possible foot pairings are eliminated by the following rules (29):(2.30) a. A short foot must be paired with a long foot.b. Only one foot may be long.Here a short foot is defined as one which is shorter than the “standard” ornormative trochaic foot pattern Sx; a long foot is one which has three or fourmetrical positions. The outcome of these rules is that half-lines will have no lessthan four and no more than five metrical positions.Russom’s rule for alliteration is a metrical rule which corresponds to the stresssubordination rule which operates in OE compounds. The OE Compound StressRule (OECSR), just like the Modern English Compound Stress Rule (CSR), is abinary operation which assigns prominence to the first lexical constituent,subordinating the second constituent, which therefore receives a lesser degree ofstress. Russom, following Liberman and Prince, represents this by the followingtree structure (68):S W(2.31) s - mannes15A prociftic is a function word which cannot stand on its own, but which“leans on” a following lexical word. Examples in English are articles, possessivepronouns, prepositions, etc.35in which the first lexical element is labelled S, or strong, and the second W, orweak. Russom’s rule for metrical compounding works on the same principle (71):(2.32) When two constituents containing S positions appear within thesame metrical domain, label the first constituent strong and thesecond constituent weak.To illustrate his rules for alliteration, Russom gives the following example of aline made up of two simple Sx I Sx verses (71):stroil wèakstroiëak stroi’eak(2.33) Sx Sx Sx, SxOnce constituents are labelled, the following rules for alliteration apply (73):(2.34) a. The strongest two metrical positions within the line must containalliterating syllables.b. A weak constituent of a weak constituent may not contain analliterating syllable.c. No alliterating syllable may occupy an x position.d. Otherwise, alliteration is optional.In example (2.33) above, the strongest two positions in the line, according toRussom (72), are the first and the third; each must contain an alliterating syllableby (2.34a). The fourth position, being a weak constituent of a weak constituent,may not alliterate by (2.34b); this neatly captures Sievers’s generalization that thesecond stress of the second or off-verse never shares in the alliteration. Thesecond position alliterates optionally.Russom’s theory fulfills many of the requirements of a generative model: heprovides an abstract metrical pattern, which he derives from the stress patterns ofOE words, together with correspondence rules which constrain how these36abstract patterns may themselves be instantiated in the poetic language.Although his foot patterns are presented in the form of a list, it is a list withinbuilt constraints, since only items with stress patterns compatible with OEwords are admissible. Other imaginable foot patterns are thereby ruled out.Nevertheless, some problems arise with regard to Russom’s model. I shall firstdiscuss a few descriptive problems relating to his list of 25 allowable versepatterns and some inconsistencies caused by his definition of the OE word; thenthe more serious theoretical problem concerning radical differences betweenRussorn’s theory and generative metrical theory.Russom eliminates, correctly, I think, verse types ending with x or xx, such asSxx I x, Sx I xx, etc., from his list of 25 allowable foot-pairings. But his reason fordoing so is insufficient. OE half-lines, he argues, do not end in proclitics; andfunction words appearing verse-finally “almost always acquire a stress thatprevents their root syllable from occupying x positions” (26). This explanation asit stands is not entirely accurate. For one thing, OE half-lines quite often do endin function words, including proclitics: pronouns, the copula, thedemonstrative, the adverb ã, possessive pronouns, and prepositions. A fewexamples from Beowulf:(2.35) a. Scedelandum in ‘in Scedeland’ (Beo. 19b)b. hläford ]Tnne ‘your lord’ (ço. 267b)c. Em eafera wes ‘to them a son was’ (Beo. iZa)I am not disagreeing with Russom that these verse-final function words arestressed. But what causes them to acquire this stress? Russom (53) suggests thata determiner removed from its normal proclitic position becomes stressed andmay therefore occupy a S position by (2.20a):(2.36) a. grund-wong hone ‘the bottom’ (Beo. 2588a)Ss I Sx37b. mgas ra ‘of the kinsmen’ (Beo. 1015b)Sx I SxBut displacement of a function word from its normal position does notexplain every instance of such a word acquiring stress. While displacement mayexplain the examples in (Z.35a-b) and (2.36), for example, it does not account for(2.35c) or for the examples in (2.37) below, which have normal word order:(2.37) a. ic 1is gid be pë ‘I [told] this tale to you’ (Beo. 1723b)b. ã ho onfunden wes ‘after she was discovered’ (Beo. 1293b)6c. wealdan môston ‘they could control’ (Beo. 2038b)In the absence of a phonological rule which specifies the circumstances underwhich function words may acquire stress and therefore occupy S positions, theabsence of foot-pairs ending in x or xx from Russom’s list of allowable metricalverse patterns remains unexplained. I shall suggest such a rule in Chapter 3.But the basic problem with Russom’s theory is that his principle (2.1 7a), thatfoot patterns correspond with OE word patterns, forces him into a number ofinconsistencies.First, he is forced (in (2.18b)) to define unstressed prefixes as words’7 in orderto allow such a prefix to occupy a foot apart from its stem; in other words, to ruleout foot patterns with rising stress: xS or xxS, for example. If feet like these wereto be included in Russom’s inventory of foot types as reproduced in (2.19),unmetrical verses, as Russom points out, would result (23):(2.38) *gegaf güo-rincxSIS s16Mitchell notes that the normal OE word order for periphrastic verbconstructions in subordinate clauses is main verb followed by auxiliary (967).‘7This definition is unorthodox to say the least. There is little doubt that anunstressed prefix in OE forms a syntactic unit with the stem to which it isadjoined, even though, as Russom points out, this was not true of Gothic, anearlier form of Germanic, which had detachable prefixes (8).38Russom rules out verses like (2.38) above (an unmetrical C-type with anintensified final dip; see (2.4c)) on the basis that the unstressed prefix is a word.Gegaf as two words, cannot occupy a single foot by (2.1 7a); and therefore theprefix must occupy a foot separate from its stem, which produces an unmetricalx I Sss scansion of this verse (23).But this definition of unstressed prefixes as words, into which Russom isforced in order to explain the absence of certain unmetrical verses like (2.38),leads him into further inconsistencies. If a word like gegaf counts as two wordsand so cannot occupy a foot by principle (2.1 7a), why may a compound likegu]-rinc occupy a foot? On what basis is a compound more “wordlike” than aprefix + stem?More problematic yet, Russom does allow two lexical words to occupy a foot,in apparent violation of (2.1 7a), just in case they form a syntactic unit. See, forexample, (2.22) and (2.27); and consider the following additional examples (85;scansions are Russom’s):(2.39) a. secg weorce gefeh ‘the man rejoiced in his work’ (Beo. 1569b)S I S x xsb. hond rond gefëng ‘his hand grasped his shield’ (Beo. 2609b)S IS xsc. hoim heolfre wëoll ‘water was turbulent with blood’ (Beo. 2138a)SI S x 5In cases like these, in which a foot is occupied by two lexical words which form aphrase, the phrase, according to Russom, “mimics the structure of a compound”(17) and may therefore occupy a single foot as though it were a compound. Theclaim that not only compounds, but phrases, are more “wordlike” than prefixedwords are seems highly suspicious.That Russom’s metrical patterns are derived from the stress patterns of OEwords not only leads to inconsistencies within his theory, but also results in39incompatibilities between his theory and generative metrical theory. First,Russom’s abstract metrical patterns incorporate three levels of relativeprominence: S, s, and x. Generative metrical theory, on the other hand,recognizes only a binary distinction, generally symbolized, as we shall see inChapter 4, S (for strong) and W (for weak).18 The reason for this is not arbitrary,but is because, as I have argued (following Kiparsky “Role,” “On Theory”) inChapter 1, the structures relevant to poetic meter are the same as those relevantto linguistic phenomena such as word stress. Meter, according to Hanson andKiparsky, is “a stylization of prosodic properties inherent in language” (“Best” 2).As we shall see in Chapter 3, rules that assign stress in words are predicated on abinary rather than a ternary distinction between levels of relative stress; andmeter, as a stylization of this binary distinction, is therefore also binary.Secondly, instead of one metrical pattern, as there is in the generative analysisof iambic pentameter, Russom has many. “We seem to be dealing not with asingle meter but with a range of allowable submeters or ‘verse types.’ . . . TheBeowuif poet provides variety by switching from one metrical pattern toanother” (“Word” 387). This statement is in flat contradiction to generativemetrical theory, which, as we have seen in Chapter 1, disallows variation in theunderlying meter, and accounts for rhythmic variation on the language levelthrough correspondence rules.Variation in the underlying metrical pattern and a ternary distinction inlevels of prominence are both the result of Russom’s metrical patterns beingderived from words. Because the rhythmic patterns of words are varied, themeter must therefore likewise vary. Because OE words have at least three levels of18See, for example, Halle and Keyser (“Iambic”), Kiparsky (“Rhythm”,“Sprung”), Hanson and Kiparsky (“Best”), Prince (“Metrical”).40stress - primary, secondary, and unstress- this must be reflected in any metricalpattern which they generate.But what principles underlie the stress patterns of words? How are the stresspatterns of words themselves generated? I would like to suggest that the word-patterns ( S, Sx and the like) which are the basis of Russom’s theory are notthemselves the meter, but a redundant level between the underlying meter andthe surface prosodic (i.e. phonological) level. If generative theory is indeeduniversal, it ought to be possible to assign prosodic constituents to metricalpositions with direct reference to the rules which assign prominence in language,without an intervening “word stress” level.If Russom is right in his arguments for variation in the underlying meter anda ternary distinction in levels of prominence, OE meter is an exception to therules of generative metrical theory. But if principles of generative meter, likelinguistic principles, have universal application, they ought to apply equallywell to OE meter.Despite the problems and inconsistencies I have outlined above, Russom’stheory has a great deal to recommend it, Unlike traditional theories, Russomposits a metrical unit - the word - which provides a sort of abstract templatewhich actual words and word groups in a verse must conform to. Although, as Ihave argued, I do not think his metrical level is abstract enough, in this hismodel is a step beyond traditional theories, which in general define the meteronly in terms of a list of stress patterns which have little or no principled basis inlinguistic phenomena. Furthermore, he integrates the fact of alliteration into hismetrical theory in a logical and consistent way. It would be most satisfyingindeed to adapt Russom’s theory, since despite its flaws, it captures someimportant insights into OE meter, even more closely to the framework ofuniversal generative metrics.41Chapter 3VOE Metrical PhonologyI have argued in Chapter 1, following Kiparsky (“On Theory;” “Role”) andHanson and Kiparsky, that meter derives from the matching of grammaticalconstituents with an abstract metrical pattern. Sievers’s theory of OE meter, asdiscussed in Chapter 2, fails on this account for two reasons: V first, because rulesconstraining matching are either inconsistent or absent; and second, because hismetrical patterns are presented in the form of a list with no generalizedconstraints stated as to the reasons why other members may or may not be addedto that list. Problems arise also with Russom’s theory, as we have seen, in that thephonological constituent, “word,” which is matched to a metrical foot, isinconsistently defined. At one extreme, an unstressed prefix, such as ge-, isdefined as a word and so may occupy a foot; at the other extreme a phrase, suchas bearm scipes, ‘ship’s hold,’ is likewise defined as a word and may occupy a foot;however a prefixed word such as gegaf, ‘gave,’ is defined as two words and musttherefore occupy two feet. In Chapter 4, I shall establish the underlying metricalstructure for OE and the rules constraining the matching of phonologicalconstituents with abstract metrical positions. In the present chapter, I shalldefine the phonological level.I shall assume, following Liberman and Prince, that language has a metrical19structure which involves comparative prominence; stress, that is, is a relative19The word “metrical” may be used in several different senses in regard tophonology and poetic meter. The term “metrical phonology” is used to refer tophonological theory in which phonological constituents are represented in ahierarchical manner (Crystal 218), and so I shall occasionally use it in this chapterin references to the phonological structure of language. Whenever it is necessaryto make clear distinctions between poetic language and underlying meter, Ishall, following Hanson and Kiparsky (“Best”) use the word “prosodic” to refer tothe former and “metrical” to refer to the latter.42phenomenon in which grammatical constituents (such as syllables, words, etc.)have prominence only in relation to sister constituents. Liberman and Princeargue that relative prominence is best represented graphically by trees whoseterminal nodes are labelled S (for strong) and W (for weak):(3.1)S WblackbirdThe tree diagram in (3.1) indicates that the constituent black has a greaterdegree of prominence than the constituent bird; in other words, black is strongrelative to bird. This difference in prominence is the basis of our perception thatthe first lexical element of this compound has a greater degree of stress than thesecond.There are two important points about tree diagrams that must be noted. First,as Liberman and Prince point out, prominence is relative; therefore, inrepresenting prominence by means of tree diagrams, the labels S and W can havemeaning only in relation to each other (256). S must always be paired with W,and vice versa; and neither may appear in isolation. The following strings aretherefore meaningless:*S *W *S S *WW(3.2) a. black b. the c. blackboard d. and theThe second point to take note of is that, since prominence is defined as arelationship between sister constituents, branching is always binary. The triplecompound law degree requirement, for example, may not be represented as in(3.3a), but must instead be diagrammed as in (3.3b)20:20Note that bracketing reflects syntactic constituency. The words law anddegree, that is to say, rather than degree and requirement are grouped togetherbecause the former grouping, and not the latter, comprises a lower-ordercompound.43*sw(3.3) a. law degree requirement b. law degree requirementLiberman and Prince note that relative prominence is preserved underembedding. This means that the relationship between, for instance, theconstituents law and degree in (3.3b) above remains unaffected by the labelling, orindeed the presence, of the constituent requirement.Kiparsky (“Rhythmic”) and Hanson (“Resolution”), assuming that theproperties of poetry are derived from the properties of language, adopt Libermanand Prince’s tree diagrams in their representations of the metrical properties ofverse. In the following discussion, I shall likewise adopt Liberman and Prince’snotation 21Principles of syllable structureThe following is a summary of the discussion of syllable structure in Hanson(Resolution 6-7), which itself is built on the theoretical work of Liberman andPrince, Hayes (Metrical), Zec, and others; and several analyses of OE, mostimportantly Dresher and Lahiri.The determinant of stress in OE is syllable weight; that is, whether a syllable islight or heavy. Weight is determined by syllabic constituents called moras(depicted as j.i). Every syllable (depicted as ) contains either one or two moras;one mora makes a syllable light and two moras make a syllable heavy. The21For an alternative account of the metrical properties of language and verse,see Hayes (“Prosodic”). Hayes argues that a better way of depicting prosodicrelations is by means of a metrical grid. Hanson (Resolution 10), however, notesthat trees are able to encode strength relations above the level of the foot, whilegrids represent only the relationships between syllables. For a detailedcomparison of the relative merits of tree and grid representations, see Hogg andMcCully.44leftmost or only mora is the head of the syllable, that is, the syllable’s strongestconstituent. The moraic head of the syllable in turn contains as its head thevocalic nucleus, or sonority peak, of the syllable, together with any consonantalonset the syllable may have. If the syllable is heavy, as in (3.4 b-c) below, anyvocalic or consonantal segments following the peak belong to the second mora(Hanson Resolution 6-7).As shown below in (3.4a), a syllable containing V (a single short vowel) ismonomoraic, i.e. light, while a syllable as in (3.4 b-c) containing either VV (a longvowel or diphthong)22or VC is bimoraic, i.e. heavy.23 This equivalence in weightbetween VV and VC captures the generalization that a syllable containing a longvowel has the same phonological weight as one containing a short vowel which isclosed by a consonant:a a aA A1-tt/1(3.4) a. (C)V b. (C)V V c. (C)VC[bij [bi] [bit]In accordance with the Maximal Onset Principle, medial consonants in OEpolysyllabic words are assumed to belong to the following rather than to thepreceding syllable to the extent that this does not violate syllable structure rules(Suphi 196). For example, syllable boundaries fall as in (3.5), rather than as in(3.6), since ift, nd, nr and ng are not acceptable syllable-initial sequences in OE220E also has short (monomoraic) diphthongs, which result from “breaking”of short vowels in certain phonological environments. See Hutcheson (46) andLass (172-4).23VVC and VCC are “superheavy,” but the metrical phonology of both OE andPresent Day English (PDE) make no distinction between heavy and superheavysyllables. Additional segments are assumed to be adjoined to the second moraand do not make an already heavy syllable heavier (Kristin Hanson, personalcommunication). See below for a brief discussion of adjunction.45(note that these sequences never appear word-initially, which I assume isevidence for this):(3.5) a. orf te b. ston dan c. han ra d. gon gan(3.6) a. *frj rfte b.* sto ndan c. *ha nra d. *go nganOE foot tvpologvSyllables are parsed into higher-order constituents called prosodic feet(depicted as 0), of which one syllable is the head: that is, its strong or onlyconstituent. The syllable that heads a prosodic foot is stressed. Hayes, borrowingthe terminology of classical prosody, proposes three universal foot types intowhich syllables of various languages may be classified: the syllabic trochee, themoraic trochee, and the iamb (“Revised” 279). The foot type which OE constructsis the moraic trochee (Hanson “Resolution” 2).The ordinary moraic trochee has two moras. These moras may come from oneheavy syllable, as in (3.7a) below; or two consecutive light syllables, in which casethe first is strong, as in (3.7b):0a as awA(3.7) a. tj.t b. tAs discussed briefly in Chapter 2, all words in OE (disregarding words withstressed prefixes, which will be discussed below) have primary stress on the initialsyllable of their stem. If the stem-initial syllable is light and is immediatelyfollowed by a heavy syllable, a “resolved” moraic trochee, in which the W nodebranches, is constructed over the pair of syllables24:24For a more detailed discussion of the phonology of resolution in OE, seeHanson (“Resolution”).460-as owlÀ(3.8) i ttRecall that Sievers, as summarized in (2.lb), points out that a rise usuallyconsists of either a heavy stressed syllable or a light stressed syllable together witha following light or heavy unstressed syllable. This long-noted equivalence thusis seen to have a phonological basis, as the moraic trochee, which defines OEstress, comprises just these three structures.OE word stress ruleBecause all words in OE, except those with stressed prefixes, have primarystress on the initial (or only) syllable of their stem, it follows that moraic trocheesare constructed starting at the left edge of the word (the prosodic word, which inall cases consists of one or more prosodic feet, is depicted as ?):0 0 0a as ow as CYWI_ti_I. I_ti_t i_tI_tlI(3.9) a. man b. lu fu c. cy ningSome words, like those in the examples in (3.10) below, consist of more thanone moraic trochee. In this case, the two leftmost feet form a SW pair (theminimal word, or min). The head of the strong foot has primary stress; the headof the weak foot receives secondary stress:A AOsøwø øs 0w 0 Os OwIii II I I Ia a a a a a a a aAAAi_tJt I_ti_I. I_LI_I. I_ti_I. i_ti_I. 41 I_ti_I. i_4I i_I.(3.10) a. ha hg ran b. wal den des c. w den de47As we shall see in Chapter 4, min is the largest phonological constituent thatmay occupy a metrical position in OE.Marginal destressingSometimes a foot is created over a word-final inflectional syllable, as in (3. lOab) above. Since OE affixes are unstressed, McCully and Hogg propose a rule ofMarginal Destressing which operates at the right edges of OE words, deleting afinal W non-lexical25foot (327):0 0 0as aw a —> as cw atA iAAt 4L 4t Ji 4L J4L(3.11) cy nm gas cy mn gasAs we shall see below, Marginal Destressing does not apply if the W foot, inwords with stressed prefixes, happens to be the stem of a word.Note that secondary stress is thus predicted on derivational endings such as -lrc, -weard, -ing, -end, etc., only if they are immediately followed by anothersyllable, as in (3.1 2a). When these endings are word-final, they are destressed bythe Marginal Destressing Rule as in (3.12b)26:25! assume that by “nonlexical” in this context, McCully and Hogg mean thatMarginal Destressing applies only to constituents which are not the stems ofwords, such as derivational and inflectional endings. Russom points out,however, that the second elements of “semantically lexicalized compounds” suchas hlaford, as well as personal names such as Beowu!f, Guthiac, and the like, aregenerally assumed to be destressed (156-7, notes 5-6).26For a more complete discussion of stress in OE derivational endings, seeMcCully and Hogg (330-3 1).48øs 0w 0a a a a aA AA A AJ-w ,_wt !.t!.L !.t,_t J4t(3.12) a. h lend es b. h lendSyllable adjunctionNote that in example (3.1 Oc) above, a final syllable is left stranded, orunattached to a foot, after initial parsing is complete. Stranded syllables areassumed to be attached by a rule of stray-syllable adjunction (SSA). This involvescreating a new foot node under which the existing foot node and the syllablenode of the stray syllable are both subordinated:/\Os 0w 0as aw a as aw a/\AI tAttI.Lt It I 4L .I(3.13) a. w den de b. cy nm gaSyllables that are stray as a result of Marginal Destressing, as in (3.10), (3.11), and(3.12), are also adjoined as above.2727Stray-adjunction in fact applies to constituents other than syllables.Presumably long words such as mennisclicness, ‘humanity,’ which contains threefeet after Marginal Destressing, would have a final stray foot adjoined to 2min;however words like this seem to be uncommon in poetry (at least there are nonein Guthiac B).49ResyllabificationAs discussed in (3.5), syllables are constructed according to the Maximal OnsetPrinciple; that is, their onsets are maximized to the extent that this does notviolate syllable structure rules. In accordance with this principle, the onsets ofsome non-initial syllables in OE will contain a consonant cluster such as st, sn, tr,dr, and so forth. Note that these sequences are quite acceptable word-initially inOE; I therefore assume that they are always acceptable syllable-initially. When aword-medial syllable containing such a consonant cluster in its onset isimmediately preceded by a word-initial light syllable, the two syllables shouldtherefore be parsed together as a resolved moraic trochee:0 0 0Aas aw as aw as awI IA IAt 14L J1 4t t tIJ.(3.14) a. bro snung b. bi tran c. fa stenBut evidence from meter suggest that these sequences do not behave likeordinary resolved sequences, which have only one consonant in the onset of thesecond syllable. The final rise (in Sieversian terms) of a B- or E-type verse, forexample, may contain a word like wera or sefan, but never a word like brosnung,bitran, or fsten. I shall therefore assume that when a consonant cluster occupiesthe onset of a syllable preceded by a light word-initial syllable, the first consonantis resyllabified as the coda of the preceding syllable, thus rendering it heavy;28and that the resyllabified constituent is the one relevant to poetic meter:28Angelika Lutz, in her study of Anglo-Saxon scribes’ word-divisions at theends of manuscript lines, found that clusters such as st, sn, tr, etc. were usuallysyllabified C-C if the preceding vowel was short and stressed, whereas they wereusually syllabified -CC if the preceding vowel was either a) long, or b) short andunstressed (202). This appears to reflect a native speaker’s intuition that if a50(3.15) a. bro snung — bros nungb. hi tran - bit ranc. fa sten — fs tenCompound StressLike Present Day English (PDE), OE places primary stress on the head of thefirst lexical element of a compound word while subordinating stress on thesecond. This accounts for the falling or trochaic stress pattern of compounds inboth PDE and OE. The OE Compound Stress Rule (OECSR) is a rule that assignsprominence to the first lexical element of a compound word. The rule is thesame as that operating in PDE; according to Halle and Keyser (English 95), there isno reason to think the CSR has changed since the OE period:4L J4L J4L J4L(3.16) a. gked-mod b. swegi-wuidreThe head of the weak word of a compound is traditionally scanned in the poetryas having secondary stress.syllable is stressed, it is also heavy. Hanson (ieolution 23-27) discusses a similarphenomenon in PDE, in which the onset of a stressless syllable is resyllabifiedinto the coda of a preceding light stressed syllable. Therefore all (or almost all)stressed syllables in PDE are heavy in their surface structure. Resyllabification inOE is a point deserving of more study, since, as pointed about above,determining the syllable boundaries of words like brosnung and bitran will haveconsequences for resolution in the poetry.51PrefixesThe rules we have been discussing so far apply only to the stems of words plusany derivational ‘or inflectional endings they may have. But many OE words alsocontain prefixes, which may be either stressed or unstressed.Prefixes which attach to OE nouns and adjectives bear primary word stress.Following McCully and Hogg (323) and Suphi (182), I shall assume that in thesecases a foot is constructed over both the prefix and the stem of such a word.29The foot over the prefix bears primary stress and so is labelled S. I shall also followMcCully and Hogg in assuming that Marginal Destressing does not apply toword stems (we shall see some evidence for this in (4.40-41) below); therefore theW foot is retained over the stem:A. A.øs 0w Os 0wI I I Ia a a aA A A A1.tJL 4L JtI(3.17) a. . and giet b. un rtThe head of the weak foot is generally (though not always) scanned as bearingsecondary stress; see Russom (OEM 69-70) for discussion. Russom treats wordswith stressed prefixes like those in (3.17) above as compounds; however, as I shalldiscuss in more detail in Chapter 4,. words with stressed prefixes behavedifferently in meter than do compounds, on one hand, and simple words, on theother.3°29Although there is general agreement that prefixes in nouns and adjectivesare stressed, the mechanism by which they come to be so is not yet fullyunderstood. For a different formulation of rules applying to stressed prefixes, seeHalle and Keyser (English 90-93).30Unlike compounds, words with stressed prefixes may appear on S positions;see (4.5-6). Unlike simple words, they may not appear on W positions; see (4.40-41).52The prefix ge-; which is always unstressed, is an exception to the rule thatprefixes which attach to nouns are stressed; be- and for- exhibit variable usage,sometimes appearing stressed and sometimes unstressed (Halle and KeyserEnglish 95).Prefixes which attach to verbs, participles, and most adverbs are almost alwaysunstressed. The rules by which unstressed prefixes are attached to their stems arenot yet fully understood. According to McCully and Hogg (324-27), unstressedprefixes are simply adjoined to their stems at some stage after the prosodicstructure has been formed (note that even heavy unstressed prefixes, like those in(3. 18b-c) below are not footed):0 1 1/o /0/ø /0 /ø/1 \ I\ / I //‘\a a a a a a a a a as awI A A AAA A 1’ II tJI t 4L 14L J4L 4I. 4L .t t L(3.18) a. be bo dan b. f57 sed c. to deg d. be fo ranBut according to Suphi (182-3), unstressed prefixes are not adjoined at thelexical stage at all, but, like function words, receive their metrical structure at thepost-lexical stage. I shall leave the question open.Stress in function wordsNonlexical, or function words, are treated differently in meter than content orlexical words. More precisely, while the stress properties of content words areobligatorily respected by metrical rules, those of function words may be ignored(which is captured in the generalization regarding OE meter, as discussed inChapter 2, that function words, even disyllables such as cefter, ofer, and the like,53are unstressed). Hanson suggests that the differing behaviour of function wordsand lexical words in meter is because function words are not assigned stress bythe rules discussed above, but are generally agreed to receive stress postlexically,that is, according to rules which operate across word boundaries, taking phrasalstructure into account. Since function words do not receive lexical stress, it isassumed that they may be treated very differently from lexical words in meter(Resolution 27); while the rules of lexical stress assignment are obligatorilyrespected in meter, rules of post-lexical stress assignment are respected onlyoptionally. Various schools of poetry, as well as individual poets, may choosewhether to respect the stress patterns of function words or not.31Zec and Inkelas suggest that function words in PDE are assigned postlexicalstress under two conditions. First, disyllables may receive stress by a rule whichbuilds a binary foot over them. This accounts for stress on words such as under,over, beneath, among, and so forth (8). Second, a monosyllabic function word mayreceive stress just in case it appears phrase-finally (10). In support of their secondargument they note that, in spoken English, many function words which appearwith a reduced vowel in their normal sentence-position (to the left of their head,that is, the word to which they are subordinate) do not reduce when they appearin phrase-final position (5). For example:(3.19) a. The cat is (s) outside.b. I don’t know where the cat is (*s).c. I will (1) go if you will (*1).31Hanson and Kiparsky, for example, discuss Finnish lyrical style, in whichstressed nonlexical words are constrained by the same metrical rules thatconstrain lexical words. This metrical practice is in contrast with that followed inFinnish ballad style, in which any stress assigned to nonlexical words may beignored at the poet’s option (“Best” 29-30).54Since in PDE only an unstressed vowel may reduce, Zec and Inkelas suggest thatthe failure of function words to reduce phrase-finally is due to a rule whichassigns post-lexical stress to these words when no prosodic host appears to theirright; that is, when they are phrase-final (10):(3.20) Phrase-Final Stress Rule: Build a foot on a final phonological wordwhich has no metrical structure.There is evidence that the Phrase-Final Stress Rule applies not only to PDE, buthas relevance to OE poetry as well. I would like to suggest that a function word,whether mono- or disyllabic, is footed and is treated by the meter as though itwere a prosodic word just in case it appears phrase-finally. On the other hand,any stress properties of non-final function words, including disyllables, appearto be metrically irrelevant; that is, OE meter seems to be insensitive to Zec andInkelas’s rule that builds a foot on a non-final disyllable. This metricalinsensitivity to the stress patterns of non-final disyllables probably accounts forthe fact that most metrical theories of OE consider these words as beingunstressed; Russom, for example, treats a disyllable such as efter as a xx foot (13),even though the word was probably pronounced as it is today, with initial stress.Why OE meter is sensitive to one post-lexical rule concerning function words andnot another is a question which I shall not attempt to answer here (although itrequires answering). The consequences, however, of the meter’s differenttreatments of non-final and final function words will be taken up in Chapter 4.55Chapter 4.A Generative Model.Guthiac B is found in the Exeter Book, a manuscript dating from the secondhalf of the tenth century. While there is no proof that, as some have argued, thepoem was composed by Cynewuif, an Anglian who wrote sometime during theperiod 750-850, the poem is roughly contemporary with these dates, or mayperhaps have been composed somewhat later (Fulk 402). The poem appears onfol. 44b-52b, immediately preceded by Guthiac A, which is almost certainly by adifferent author of an earlier period (Roberts “Metrical” 119; Fulk 401).Guthiac, a member of the royal Mercian family, was born in 673. He lived as asoldier until the age of 24, when reflections upon the deaths of his nobleancestors led him into the monastery at Repton. In 699, inspired by stories of thedesert fathers, he withdrew into the Lincoinshire fens at Crowland to live a life ofsolitary contemplation. There he died in 714, and was confirmed a saint thefollowing year.Guthiac’s biographer, Felix of Crowland, composed his Vita sancti Guthlaci inthe middle of the eighth century, probably between 730 and 749. Guthiac Bshows more dependence on Felix than Guthlac A (Bradley 249), and itscomposition may have followed that of Guthiac A. This relative chronology issupported by internal evidence of the poems; for Guthiac A describes the saint’sstruggles with the devils as taking place “in ussera tida tTman” (7534)32, ‘in theperiod of our memory’; while Guthiac B refers to the authority of books, one ofwhich was possibly Felix’s Lji: “Us secgao bc” (878), ‘books tell us’ (Fulk 402).332M1 quotes from Guthiac are from The Guthlac Poems of the Exeter Book,edited by Jane Roberts. As aids to scansion, I have added macrons to indicatelong vowels and hyphens to separate lexical elements of compound words.33For further discussion of metrical, stylistic, and dialectal aspects of datingthese poems, see Roberts (“Metrical”) and Fulk (399-402).56Thematically, Guthiac A has much in common with Beowuif in its martialimagery and its depiction of heroic courage, although Guthiac’s battles, unlikeBeowuif’s, are more metaphorical than literal; his fights with evil beings arecarried out with words rather than with swords. Stylistically (even if somewhatimpressionistically) though, Guthiac B has more affinities with Beowuif, at leastin its greater number of compound words (Roberts estimates a ratio ofcompounds in Guthiac A to Guthlac B of 2:5 (“Metrical” 118, note 139)).Compound words provide a more secure starting place than phrases do for thepurposes of metrical analysis because their stress patterns in OE are betterunderstood than those of phrases; and therefore I have chosen to base my studyon Guthiac B.Guthiac B is 561 lines long, thus comprising 1122 verses or half-lines. Theedition chosen for this study is The Guthiac Poems of the Exeter Book, edited byJane Roberts, which is a conservative edition. Division of the manuscript lines,punctuation, and most capitalization have been added by the editor.The next part of this study consists of a close examination of the metricalstructure of this poem. My intent is to incorporate Russom’s metrical rules into agenerative model of the type which Halle and Keyser (“Iambic”) and Kiparsky(“Stress,” “Rhythmic,” “Sprung”) have constructed for iambic pentameter. Sucha model will consist of an abstract metrical pattern consisting of S and W metricalpositions together with correspondence rules licensing ways in which prosodicmaterial may fill these positions. In this way I hope to reduce Russom’s“allowable submeters or ‘verse types” (“Word” 387) to a single meter, thusbringing his theory more closely into line with tenets of generative metricaltheory.57Framework for a universal generative metricsAs I have mentioned in Chapter 2, Russom’s metrical theory, since it is basedon the stress patterns of OE words, is language-specific. Hanson and Kiparskyargue, however, that the forms of rules of generative metrical theory, like thoseof universal grammar which govern the prosodic constituents of language(syllables, feet, words, etc.), are universal. They propose a theory of universalgenerative metrics which comprises a set of parameters from which meters arederived in accordance with their optimality in terms of a given language’sphonology. The main points of their theory are paraphrased as follows (2-3):(4.1) a. The basic constituents of lines are metrical feet.b. Metrical feet are binary; that is, each has two metrical positions.c. A prominent position, labelled S, is the head of the foot; a non-prominent position is labelled W.d. Two parameters, structure parameters and realization parameters, arefixed.Structure parameters are those which determine the abstract metrical pattern.They establish heacledness and number of feet.Structure parameter settings for OEOE meter differs from most other meters in that its line is invariably dividedby a caesura into two cola (generally referred to by OE metrists as verses) of twofeet each. This division of the line follows naturally if we postulate the existenceof a higher-order relationship between feet which groups them into largerconstituents: Universal generative theory as proposed in Hanson and Kiparsky isinadequate to this task, since it makes no reference to higher-order relationsbetween feet. However Kristin Hanson suggests, following Prince (and see alsoHayes “Prosodic” 256, Kiparsky “Rhythmic” 229-30, and Youmans 347), thatthere is a universal generalization that feet are grouped into cola which are58themselves labelled in the same direction as feet and grouped into lines (personalcommunication). Assuming this higher-order relation between feet, thestructure parameter settings for OE are as follows:(4.2) a. Feet are left-headed (SW).b. Each line contains four feet.c. Each colon contains two feet.The metrical pattern of a line may thus be schematized: S W S W S W S WvvvvSW SWS WRussom (OEM 25) states that: “the meter [of Beowuif] changes unpredictablyfrom verse to verse.” In my proposed model of OE meter, however, the basestructure of the meter remains constant, in accordance with the structureparameters of universal metrics. The great variety of rhythmic effects which arecharacteristic of OE verse, and which tend to be conflated with the meter itself,are a result of matching not only syllables (o), or even prosodic feet (0), but arange of units of linguistic structure with the metrical positions.Realization parameters, or correspondence rules, have to do with the way inwhich prosodic features of the language are manifested in the meter. Hanson andKiparsky propose three realization parameters: position, prominence site, andprominence type. Position refers to the maximal amount of prosodic materialthat may be contained in a metrical position. The position parameter may be setat the mora, syllable, foot, or word. The prominence site parameter determineswhether it is S or W positions that are constrained in order to set up a binaryopposition: that is, whether constituents in S positions are required to beprominent; or whether constituents in W positions are required to beunprominent (or perhaps both). Finally, prominence type specifies whatdetermines prominence. Prominence type may be set at weight (whether a59syllable is heavy or light); strength (whether a prosodic constituent is strong orweak in relation to a sister constituent); or stress (3).Realization parameter settings for OE(4.3) The position parameter is set at the minimal word (2min).This position parameter setting captures Russom’s generalization that OEmeter is word-based, and not syllable-based, which has been the assumptionbehind most traditional theories of OE. There is a difference, however, in thatwhereas Russom proposes a correspondence between words and metrical feet,that is, that a metrical foot contains exactly one OE word, I am proposing acorrespondence between words and metrical positions: that is, that a metricalposition contains maximally a prosodic word.In terms of prominence, OE constrains both S and W positions. In this regard,contrary to those who claim the meter is lax or unregulated, it is actually stricterthan many modern meters, which constrain only one position.34(4.4) a. For prominence site S, prominence type is set at stress.35b. For prominence site W, prominence type is set at strength.In other words, S must contain material which is prominent (prominencedefined as stress), while W may contain only material which is unprominent(prominence defined as strength).36 This curious asymmetry in prominencetype, in which S must contain stressed material while W must contain weak (butnot necessarily unstressed) material, seems to be one of the reasons that the34Prominence parameters for Shakespeare’s iambic pentameter, for example,constrain only W positions (Kiparsky “Rhythmic” 195; Hanson and Kiparsky 6).35According to the choices as outlined in Hanson and Kiparsky. However,stress alone is not a sufficient condition; this rule will be revised in (4.19).36That W positions in OE may contain only prosodically weak material followsfrom a suggestion of Kristin Hanson (personal communication).60rhythmic patterns of OE verses are so variable, and why, as a result, theunderlying meter is so difficult for modern ears, trained in a very differenttradition, to determine.CompletenessAccording to Hanson and Kiparsky, the realization parameters position,prominence site, and prominence type are set according to the guiding principle ofCompleteness, “which requires parameters to be set in such a way that thelanguage’s core vocabulary can be used. . . the realization parameters are set so asto maximize the accommodation of the canonical word types of the language”(5).This principle has interesting implications for Russom’s theory, in whichmetrical foot patterns correspond to OE word patterns, and actual words andphrases within the poetic line correspond to or mimic these prototypical wordpatterns. Within the theory I am proposing here, however, the fundamentalpoint is not that words establish or set the meter, but that the meter allows thewords- a bottom-up rather than top-down process, so to speak.In the sections below, I shaildiscuss the properties of OE meter which have ledto the realization parameter settings as proposed in (4.3) and (4.4),Constituents in S positionsThe largest constituent that appears in a S position is the minimal word(2min). ?min may be broadly defined as the class of structures which is parseableas the SW pair of feet resulting from application of the OE Word Stress Rule (see(3.10) and (3.17) for examples). Any affixes are adjoined to this primitiveconstituent at a later stage of derivation, as shown in (3.13) and (3.18), and so donot count as part of min. Thus the largest member of the class of 2min in OE is a61structure consisting of a pair of prosodic feet, the leftmost of which is strong.Such a pair of feet may occupy a S position:(4.5) a. ne et onbid long ‘nor that interval long’ (904b)37SW S Wb. r ombeht-oegn ‘messenger servant’ (1146a)38S(W)S Wc. eal innanweard ‘completely inward’ (1320a)S(S WNote, however, that compound words, even those consisting of only twosyllables (for example wrcec-sTh, ‘exile’), may not count as ?min, even thoughtheir stress patterns may be identical to those in the examples above. So while a2min such as onbid may occupy a single S metrical position as in (4.5a) above, acompound, which consists of not one but two lexical words, may not occupy thesame position:(4.6) *ne Jet wrc-siô long ‘nor that exile long’SW S WThat ?min may occupy a single metrical position while a compound may notaccounts for the fact that most approaches to OE scansion do not admit foottypes containing three stressed syllables. Sievers and Bliss, for example, list nofoot types like */\/ ; neither does Russom include a foot type like *S5S. Russomaccounts for feet like edwrt-lif ‘life of shame’ (Beo. 2891b) by assuming destressingon the medial syllable in a multiple compound, which allows such a compoundto occupy a Sxs foot (OEM 70). On the analysis presented here, the assumption ofdestressing is unnecessary. Edwtt, like onbid, is not a compound, but a stressedprefix (ed) + stem (wit); that is a SW pair of feet, or ?min. It may therefore occupya single metrical position, whereas a compound, which consists of two lexical37A11 examples following are from Guthiac B unless otherwise noted.38Note that a W position is empty. Empty positions will be discussed in (5.3-6) below.62words, may not. Compound words must always occupy at least two metricalpositions:(4.7) wrec-sTh wpan ‘to bewail exile’ (1074a)S WSWThe largest member of the class of 2min, then, is the primitive structure,resulting from application of the OE Word Stress Rule, which consists of a pair ofprosodic feet, the leftmost of which is strong. This left-headed pair of feetcomprise a prosodic word which may or may not have one or more unstressedsyllables adjoined to it. But only the pair of feet which comprise min, and notany adjoined syllables, may occupy a single metrical position:(4.8) b. *ãres uncües word ‘words of the unknown messenger’SW S WThe fact that a metrical position may contain a constituent as large as thelargest simplex word in OE lends support to Russom’s claim that the word, andnot the syllable or even the prosodic foot, is .the basis for OE meter.As we have seen, ?min has the following characteristics: 1) it comprisesmaximally one prosodic word; 2) it comprises maximally two feet; 3) if itbranches, it is left-headed. Any constituent which does not violate theseparameters, that is, which is parseable as though it were 2min, may occupy asingle metrical position. Consider, for example, the following:I I Ia a a a a as awa(4.9) a. hal ge b. fo we re c. meo lii desConstituents such as those in (4.9) above, which consist of a single foot plusunstressed syllables adjoined rightwardly, and which violate none of theparameters defining Xmin, may appear on a S position:63(4.10) a. ymb et hlge htis ‘around the holy house’ (1310a)SW SWb. wron fowere ‘there were four then’ (1 134a)SW S Wc. shte sãwel-hüs ‘sought the house of the soul’ (1 141a)SW S Wd. meahtig meotudes egn ‘mighty servant of the lord’ (1243a)SW S WA wrinkle appears, however, when such a constituent is immediately followedby a function word or an unstressed prefix which is not itself the sole occupant ofa metrical position:(4.11) Hum ic swTôe ne ]earf ‘however I don’t very much need’ (1356b)SW S WHow should ne be parsed, according to our definition of ?min? Should ne occupythe preceding S position or the following W position? If it occupies W, not onlywill W contain two words, but the constituent which it contains will not beparseable as min, which may not have a 5 foot as its rightmost element. But if neis allowed to occupy 5, then the constituent in S, while left-headed, will stillconsist of two words.The generalization seems to be that function words, which receive stresspostlexically, behave differently in OE meter than lexical words do, as discussedin Chapter 3. Words like ne, or even disyllabic function words like ofer are notassigned a word node or even a foot node by the rules of the lexical phonology, incontradistinction to lexical words; this, it has been suggested, allows them tobehave differently from lexical words in meter (Hanson Resolution 27).39 Anyprosodic structure assigned postlexically to non-final function words is ignored39An exception is made just in case a function word appear phrase-finally, inwhich case it is footed by postlexical stress assignment rules, as discussed inChapter 3. These words will be discussed further below.64by the metrical rules; therefore these words are simply parsed together withwhatever foot precedes them, and treated by the meter in the same way as aninflectional ending adjoined to its stem. SwThe ne, then, is treated like a left-headed structure which comprises one foot plus rightwardly adjoiningunstressed syllables. Like fowere in (4.lOb), swThe ne is parseable as min and maytherefore occupy a single metrical position:(4.12) HumicswThenepearfSW S WA similar situation arises when a lexical word is immediately followed by aword containing an unstressed prefix:(4.13) in m &an gefan ‘in the eternal reward’ (1186b)SW S WA prefixed word like gefëan is not parseable as min because, as we have seen in(3.18), unstressed prefixes are adjoined to their stem at some stage after initialparsing has taken place, that is, a stage following the construction of ?min.4° Aword with an unstressed prefix can therefore only be interpreted as a derivedword, whereas sequences such as swiôe ne and ëcan ge- may be interpreted asconstituents which are parseable as words. An unstressed prefix must thereforepattern together rhythmically with the word preceding it, and not with its stem;note that this captures Russom’s insight that unstressed prefixes seem metricallyseparable. Since ge- (or any unstressed prefix), like a function word, is not footed,it may be parsed together with a preceding word without violating theconstraints on 2min. That is, any word boundary between a lexical word and anunstressed syllable to its right may be disregarded; the two may be parsed401f they are adjoined at the lexical level at all; see Suphi 182-3.65together and may appear on a single metrical position, just as if they comprised asingle prosodic word.41Finally, the class of constituents parseable as min includes the minimal foot,or 0mm - the smallest possible prosodic word in OE. The minimal footcomprises a single stressed syllable (4.14a) or a resolvable sequence (4.14b):(4.14) a. beorht in brostum ‘bright in [his] heart’ (843a)S WS Wb. weras 7 idesa ‘men and women’ (1232b)4S W SWAs discussed above, an unstressed syllable immediately following 0mm,which does not itself constitute a metrical position, is grouped together with0mm as a single constituent:(4.15) a. nes he forht se ah ‘he was not afraid however’ (961b)SW S Wb. Ijeah his lrc 7 gst ‘though his body and soul’ (967b)S WS Wc. t me sãr gehr5n ‘that pain reached me’ (1027b)SW S WProminence constraints on S positionsAs discussed above, constituents in S positions are required to have more thana certain degree of prominence, which may be set at weight, strength, or stress.In OE, the prominence setting for S positions is set at stress. Were prominence tobe set at strength, a S position could contain only a strong syllable, that is, asyllable that is strong in relation to a sister constituent: the head of a polysyllabic41For a discussion of a similar (although more unusual) process in Tennyson’smeter, in which resolution appears to operate across a word boundary, seeHanson (Resolution 154).42The symbol “7” is used by OE scribes to indicate ond, ‘and.’66word, for example. But S positions in OE may contain lexical monosyllables,which are stressed, but not necessarily strong in relation to another constituent:(4.16) a. Dëaô nakcte ‘death approached’ (1 139b)S(S Wb. fiftVnu gear ‘fifteen years’ (936a)SW S(W)Therefore strength is not the prominence type setting for S in OE. Weight isruled out by the fact that heavy unstressed syllables never appear in S positions:(4.17) a. hfl Gü1äc wearô ‘how Guthiac became’ (879a)*S WS Wb. ëadig Engle ‘blessed among the English’ (880a)*5W 5Scansions such as those in (4.17) above violate prominence constraints on W. Wpositions, according to (4.4b) (and which we will discuss in more detail below),may contain only prosodically weak material; this constraint is violated in theunderlined W positions. Therefore, of the choices as outlined in Hanson andKiparsky, the prominence type in OE must be set not at strength or at weight,but at stress.However, the requirement of stress alone in S is not sufficient. Consider thetwo following scansions, both of the same verse:AøsøwI IcY cY(4.18) a. nyd-cosjngum ‘with painful trials’ (1153b)S WS WAøsøwab. nyd- costingumS (W) S W67The second lexical element of the above compound, costingum, has two prosodicfeet, the strong foot dominating cos and the weak foot tin (both of which areheavy syllables, and therefore footed). Since the syllable tin has secondary stress,being a W prosodic foot, it could appear on a S position if stress were the onlydetermining factor. However, scanning the verse as in (4.18a) violatesprominence constraints on the underlined W position, which now contains astrong foot. In fact, a constituent occupying S must bear primary stress, that is,be word-initial; and the verse must instead be scanned as in (4.18b). Let us restate(4.4a) as:(4.19) A S position must contain the head of a prosodic word.A prosodic word (?), as shown in (3.9), is a word which receives its prosodicstructure by the rules of the lexical phonology, and whose prosodic structure istherefore assumed to be obligatorily respected by meter. Of course any suchconstituent may not violate the position parameter setting by comprising morethan 2min,The heads of simple lexical words may occupy S positions:(4.20) a. ifra yne ‘race of men’ (864a)SW S Wb. flsce bifggen ‘surrounded by flesh’ (994a)SW SWThe heads of the second lexical elements of compound words, which havesubordinated stress, are the heads of prosodic words, and may therefore occupy Spositions. For example:(4.21) a. helle-gna ‘servant of Hell’ (1069b)SW SWb. rëonig-jdum ‘sad at heart’ (1096a)SW SWc. gst-genum ‘with spiritual mysteries’ (1084b)S WS W68Finally, phrase-final function words are footed by the Phrase-Final Stress Ruleas given in (3.20). I would like to suggest that these words, although assignedstress post-lexically rather than lexically, behave exactly like prosodic words inthat the Phrase-Final Stress Rule assigns to them a structure which, like that ofprosodic words, is respected by the metrical rules. Therefore, like prosodicwords, their heads may occupy S positions. For example:(4.22) a. heortan mTnre ‘of my heart’ (1205b)SW SWb. onsne wearô ‘became visible’ (1254b)SW S(W)Russom, as we have seen in Chapter 2, disallows verse types that have x or xxas the last foot, arguing that verse-final function words acquire stress (OEM 26). Iwould like to suggest that this stress is assigned by the Phrase-Final Stress Rule,and not the displacement of function words from their normal position; since, aswe have seen in (2.37), sometimes that normal position in OE is phrase-final.The definition of S as given in (4.19) is compatible with Sievers’s rule (given in2. lb) that a rise bears a primary or strong secondary stress. It is also compatiblewith Russom’s rules (given in (2.20)) that a syllable with primary stress generallyoccupies a S position, while a syllable with secondary stress may likewise occupya S position just in case it is the head of the second lexical element of a wholeverse compound such as middan-geardes (Sx I Sx) (OEM 159). But (4.19) provides amore precise definition of what may occupy S than either Sievers or Russom do,in that it draws a formal distinction between the properties of the heads ofsecond lexical elements of compounds and those of the heads of weak prosodicfeet, both of which are traditionally treated as bearing secondary stress; (4.19)allows the former, but not the latter, to occupy a S position.69We have seen so far that a S position must contain the head of a prosodicword; and in Sievers A, D, and E-types this is always the case:(4.23) a. in brostum ‘bliss in heart’ (954a)SWS W(A)b. mône möd-socne ‘sad [and] heartsick (1261a)S W S W (Dl)c. i-spel tO ‘sad story too true’ (1343b)S W S(W)(E)But consider the examples of A3, B, and C below:(4.24) a. pm ic georne ‘whom I eagerly’ (1084a)S W S W (A3)b. 7 let word ãcwe ‘and said those words’ (1347b)SW S W (B)c. hyge-sorge ‘that you sad in heart’ (1205a)SWS W(C)In the examples above, initial S contains a non-final function word, the prosodicproperties of which, as discussed in Chapter 3, are not relevant tO meter.Therefore these words do not count as prosodic words, in apparent violation of(4.19). What has happened?The fact that an unstressed constituent may appear in verse-initial S seems tobe due to a special metrical license which allows prominence constraints to berelaxed in initial positions, according to the principle that left edges of majormetrical or prosodic units may be lax, while right edges must be strict (Hansonand Kiparsky 7, 24; Hayes “Prosodic” 247). Relaxation of verse-initial prominenceconstraints will be discussed in more detail in Chapter 5.Prominence constraints on W positionsWe have seen so far that there are two generalizations that can be maderegarding constituents in S positions. First, a S position may contain at most a70minimal word (2.min). Second, S must contain the head of a prosodic word,whether its stress is assigned lexically (as in content words) or post-lexically (as inphrase-final function-words).A W position, like a S position, may be realized in a number of ways. Butbefore embarking on an enumeration of the possibilities, I would like to proposethe generalization that the prominence constraint applying to W positions isthat the heads of words appearing in W must be prosodically weak. That is, Wpositions may freely contain material which is unstressed; but they may alsocontain stressed constituents as long as these constituents are subordinated toanother stress. The following discussion will, I hope, make this clear.W positions in Guthiac B contain unstressed syllables of lexical words,whether those syllables are prefixes or final unstressed syllables:(4.25) a. cynn ‘of the race of men’ (821b)SW S Wb. läc ‘made an offering’ (hub)SW SWc. ltc leahi ‘body’s frailty’ (1072a)SW SWW positions also contain verse-medial function words, the prosodic propertiesof which, whether mono- or disyllabic, as discussed above, are disregarded by themeter (4.26a-b). A function word may occur on W together with a followingunstressed prefix (4.26c):(4.26) a. word Z wisdom ‘words and wisdom’ (113 ha)S WS Wb. brëme after burgum ‘famous in the cities’ (883a)S W SWc. milde 7 gemet-ftest ‘mild and modest’ (1 107a)S WS W71However, while verse-medial W may contain a string of nonlexical syllables, suchstrings never appear on verse-final W. I would like to suggest that this is becausefunction words appearing verse-finally are footed by the Phrase-Final Stress Ruleas given in (3.20).Although the exact relationship of phrases to verses has not yet beenadequately formalized and needs further work,43 it is generally agreed that OEverse boundaries are usually congruent with phrase boundaries (Sievers 279).More specifically, function words do not appear verse-finally if they are not alsophrase final:(4.27) srum geswenced; ne h I sorge wg (1137)* W S W SW S (W)‘troubled with pain; he no I sorrow felt’Although I cannot go into the process of textual editing of OE manuscriptshere,44 editorial convention places the caesura before the start of the line-medialphrase:(4.28) sarum geswenced; I ne he sorge wg (1137)Though function words do not appear verse-finally if they are not also phrasefinal, phrase-final function words do appear verse-finally, as seen in (2.35), (2.36),(2.3 7), and (4.22) above. Therefore the Phrase-Final Stress Rule as applied to OEpoetry amounts to the claim that verse-final function words, because they arealso phrase-final, are footed by rules of postlexical stress assignment and so aretreated as prosodic words. Since phrase-final function words, unlike otherfunction words, have a prosodic structure which is respected by the rules of the43For discussion, see Mitchell (989-90) and Huettner (20-21). See also mydiscussion of OE phrasal stress below.44For a detailed explication of one editor’s reconstruction of lines and versesfrom OE manuscript lines (which do not indicate where such breaks should beplaced), see Creed (Reconstructing).72meter, they may not violate the position parameter of min; and therefore onlyone function word may appear on a verse-final metrical position.A W position is not restricted to containing unstressed material, but may alsocontain a stressed constituent as long as it is prosodically subordinate to astronger stress. In words which have two prosodic feet, a W position may containthe syllable with secondary stress which is the head of the W foot:(4.29) a. onji wynlfc ‘happy arrangement’ (824b)SW SWb. Ne bo u un ‘do not be unhappy’ (1064a)S W SWA prosodic word may appear in a W position, as long as it does not exceed theposition maximum (min), of course, and as long as its head is prosodically weak:that is, the word must be subordinate in stress to another word. Typical of aprosodically weak word appearing in a W position is the second element of acompound:(4.30) a. meaht 7 mund-byrd ‘might and protection’ (881a)S W S •Wb. ce ail-mihtig ‘eternal almighty one’ (930a)SWS Wc. iit-sThes georn ‘eager for the journey’ (1267b)SW S(W)Note that, in (4.30b-c) above, the fact that the head of a prosodic word in W isstrong compared to a weak syllable in the same word is irrelevant to its beingallowed on a W position. What is important is the fact that the word itself is theweak lexical constituent of its compound. This is in distinct contrast to thesituation in Shakespeare’s iambic pentameter, in which a syllable which is strongwithin its word may not normally appear in a W position. In OE, it seems thatonly strength relations between constituents at a level higher than the syllabledetermine whether the head of that constituent may appear on a W or a S73position. It is a strong foot or strong word within a compound - not a strongsyllable - which may not appear on a W position.Finally, OE meter appears to be sensitive to phrasal stress, allowing only theheads of weak words in phrases on W positions. The metrical behaviour ofphrases is best explained on the assumption that OE phrases, like compounds,and unlike phrases in PDE, have trochaic stress. Although we can have no directknowledge of stress patterns in OE phrases, some evidence has been put forwardin support of this. Since this evidence has been the focus of debate, I shall spellin out in some detail.Several metrists have argued for trochaic stress in OE phrases. Joan Maling,assuming that alliteration falls on the most strongly stressed element in a verse,suggests that the fact that the second element of a phrase never alliterates unlessthe first does, just as the second element of a compound never alliterates unlessthe first does, is because the same rule of stress subordination operates withinboth compounds and phrases. “Most of the alliterative patterns of Beowuif arecorrectly predicted if we allow the COMPOUND RULE to cycle on the nodes NP,VP, and S [noun phrase, verb phrase, and sentence]” (382).Thomas Cable likewise points out that of the two stressed elements incompounds, the first is always heavier, since it always alliterates. He, like Maling,argues that the same stress pattern - a falling or trochaic one - exists within bothcompounds and phrases (Meter 66). There is some reason to believe then, thatunlike the Nuclear Stress Rule (NSR) of PDE, which places stronger stress on thefinal element of a phrase, a rule of OE, which I shall call the Trochaic Stress Rule(TSR), places the strongest stress on the initial phrasal element. This seems to beborne out by metrical evidence in the light of the theory presented in this paper.Let me first point out, however, that there may be certain risks involved in theattempt to abstract linguistic stress patterns (particularly in constituents larger74than the prosodic word) from poetic meter. Shakespeare’s iambic pentameter, forexample, tells us nothing about the stress patterns of phrases versus those ofcompounds; both compounds and phrases may have their strongest syllable oneither a S or a W position, although SW positioning is favoured in compoundsand WS in phrases, in general conformity with the CSR and the NSR respectively(Kiparsky “Rhythmic” 19):SW(4.31) a. Let them say more than like of hearsay well (Son. 21)W S W S W SW SW SS Wb. And do whate’r thou wilt, swift-footed Time (Son. 19)WS WS W S W SW Sc. And nothing ‘gainst Time’s scythe can make defence (Son. 12)W S W S W S W S WSW Sd. And see the brave day sunk in hideous night (Son. 12)W SW S W S WS W SNote the labelling mismatch in (4.31b), which has the strongest constituent of acompound in a W position; and that in (4.3 ld), which has the strong word of aphrase likewise in W.Unlike OE meter, however, Shakespeare’s meter is sensitive to the strengthrelations within prosodic words; roughly speaking, as mentioned above, asyllable which is strong within its word may not appear in a W position (unless itis line- or phrase-initial), while syllables which are strong within compounds orphrases may. OE, on the other hand, does not disallow a strong syllable of apolysyllabic word in a W position; such syllables may occupy W if the word inwhich they appear is itself subordinate to another word, as we have seen in(4.30b-c). It might be expected then, that, in order to maintain a binary75opposition between S and W metrical positions - that is, that S positions mustcontain material which is prominent and that W must contain material which isunprominent - OE may be sensitive to the strength relations between prosodicwords, whether these words occur in compounds or in phrases. An examinationof how phrases behave, or, perhaps more importantly, do not behave in metermay shed some light upon their stress patterns, and, in turn, allow us toformulate a rule concerning their behaviour with regard to W metrical positions.If a W position may contain the head of a prosodic word only if it isprosodically weak, and if the NSR operated in OE, subordinating stress on thefirst word of a phrase, one might expect that phrases could pattern in the meterlike clitic groups45 do, with their first word on W. For example, a hypotheticalverse such as (4.32a) below could conceivably pattern like the existing verse in(4.32b) (phrase and clitic group are enclosed in square brackets):w S(4.32) a. *tdra [guman bn-fatj ‘the man’s body weakens’S W SWwb. t5drao [jj bän-ftetj ‘this body weakens’ (1265b)S WS WThe verse in (4.32a) corresponds roughly to a Sievers Dl with a syllable ofsecondary stress in the final dip. There are no labelling mismatches in that eachW position contains the head of a weak prosodic word. But verses like this do notexist in the poetry. Their absence may be explained on the assumption thatphrases have trochaic stress, and that the head of a word which is strong withinits phrase may not appear on W; that is, the first word of a phrase must occupy S:45A clitic group is defined by Hayes as “a single content word together with allcontiguous grammatical words in the same syntactic constituent” (“Prosodic”207). Clitic groups include sequences such as determiner + noun, preposition +object, auxiliary + verb, and so forth.76S W(4.33) * tVdrao [guman bn-ft]SW S WWhen scanned as above, this unmetrical verse is now shown to be unmetricalfor two reasons: first, because the second W position contains two prosodicwords, violating position rule (4.3), that a metrical position may contain at mostXmin; and second, that W contains the head of a word which is strong within itscompound.I would like to suggest that, in fact, the first word of a phrase may neveroccupy a W position, as it does in (4.32a), but instead must occupy S. The secondword of a phrase, on the other hand, may occupy W:(4.34) a. geongum [gocor fj ‘to the young man a sad heart’ (1048a)SW S Wb. hreer [innan swearc] ‘heart darkened inwardly’ (1052b)SW S WConversely, the second words of sequences that are not phrases never occupyW; in other words, sequences of prosodic words which are not phrases do notoccupy metrical feet. If they do, unmetrical verses like the following may result:(4.35) *[ghdmd secgl Gode ‘the joyful man to God’SW S WThe example above corresponds to an unmetrical Sievers E-type with a syllableof primary stress in the dip. If the second word (Gode) of the underlined sequencewere scanned as being in S rather than in W, the reason why this verse isunmetrical becomes apparent: like the unmetrical verse in (4.33) above, violationof the position parameter results:(4.36) *[g1d..mod secgj GodeS W S(77This suggests that a metrical foot in OE contains maximally a phrase.46I would like to suggest that this pattern, that the first word of a phrase mustoccupy 5, and that the second word of a phrase may occupy W; and, conversely,that a SW metrical foot contains maximally a phrase, may be explained on theassumption that OE phrases have trochaic stress.Russom argues, however, that while a rule analogous to the CSR determinesthe location of alliterating syllables (see (2.19c) and (2.32)), this rule applies at themetrical level, not at the prosodic level, and therefore should not necessarilyapply to the actual linguistic material of phrases (“New Kind” 438). I findRussom’s argument for the NSR in OE incompatible with his claim that a verse isinterpreted as two metrical feet, i.e. two OE words. Russom suggest thatalliteration plays a key role in the interpretation of verses consisting of three fullystressed words, pointing out that these verses invariably have double alliteration:(4.37) sweord swãte fh ‘sword stained with blood’ (Beo. 1286a)“Alliteration on swate . . . seems to act as a principle of cohesion, rendering whatwould otherwise constitute two feet equivalent to one” (443). Alliteration, thatis, forces - or at least encourages - an audience to perceive phrases such as swãtefcm as one word rather than two.I do not find this claim of Russom’s very plausible. Consider the examplesbelow:(4.38) a. Good I greenhouseb. Good I green house46Following Hayes, I shall assume that the structure relevant to OE meter isthe phonological rather than the syntactic phrase (“Prosodic” 205). Assuming,like Hayes, that the phonological phrase in PDE consists of an (optional) modifier+ head with a maximum of one clitic group adjoined rightwardly (for exampleëah his lfc 7 gsi. ‘though his life and spirit’ (967b)), and assuming further thatthis definition extends to OE, metrical feet in GuthB do not contain more thanone phrase. But more work obviously needs to be done in this area.78Speakers of PDE will place the strongest stress on the second element of thephrase green house, in accordance with the NSR, and on the first element of thecompound greenhouse, in accordance with the CSR. (If the NSR operated in OE,presumably OE speakers would do the same.) Speakers of PDE interpret goodgreenhouse as consisting of two words. But I see no reason to believe that anyspeaker of PDE would interpret the phrase good green house likewise as two words,regardless of the alliterative pattern. Good green house can only be interpreted asthree words, and I cannot imagine that a hypothetical speaker of OE, if the NSRhad also operated in that language, would interpret it otherwise. Since,according to (2.1 7a), word patterns correspond to foot patterns in Russom’stheory, three words correspond to three feet, which violates (2.17b); and since itis unlikely that alliteration has the cohesive force necessary to forceinterpretation as two feet, it is far more plausible to assume that phrases in OE,like compounds, have a trochaic stress pattern, and that the double alliterationin examples such as (4.37) has some other explanation.Let us assume, then that the TSR assigns greater prominence to the firstelement in an OE phrase, labelling it S. Since the second phrasal element hassubordinated stress, its head is labelled W and may therefore appear on a Wposition:(4.39) a. byrelade brd geong ‘the young bride poured out’ (870a)SW S Wb. in ôisse wonnan niht ‘in this dark night’ (1028b)SW S Wc. nolde fteder engla ‘the father of angels didn’t want’ (945b)SW S WA test case for the claim that W may contain the heads of words only if theyare prosodically weak is the minimal word (2min). These words, which contain79both a strong and a weak foot, should behave differently in the meter than wordsin which the only strength relationship is between syllables. In fact, this is thecase, at least in Guthiac B. A minimal word, which contains a foot which isstrong in relation to another foot, may appear in a S, but not in a W position:Aøsøw(4.40) a. ne ]t onbid long ‘nor the interval long’ (904b)SW S WAøsøwb. än ombeht-]egn ‘one servant’ (l000a)S(W) S Wbut not:AøsOw(4.41) *ne et long onbid 47S WS WTo summarize, W positions may contain unstressed syllables, whether thesebe syllables adjoined to feet in prosodic words, or unstressed function words. Wmay contain stressed syllables under several conditions. First, W may contain thehead of a weak foot which is subordinated to a strong foot within a prosodicword. Second, W may contain a prosodic word, as long as it is subordinate toanother word in the same compound or phrase. We may therefore restate theprominence constraint on W, replacing (4.4b) with the following:(4.42) A W position may contain the head of a prosodic word only if it isprosodically weak.47Note that the fact that a W position may not contain a word with a stressedprefix provides independent evidence for McCully and Hogg’s claim that onlynonlexical W feet are deleted by their Marginal Destressing Rule (327), asdiscussed in Chapter 3.80Foot boundariesAs discussed above in (2.25), Russom’s theory requires bracketing rules inorder to account for the placement of foot boundaries in some B and C-typeverses and in verses containing three or more stressed words. I would like tosuggest, however, that his bracketing rules are unnecessary, since the placementof foot boundaries follows naturally from the rule that W may contain onlyprosodically weak material.While OE metrists generally accept that the half-line maybe divided into twofeet, where the foot boundaries lie has sometimes been a matter of dispute.Sievers, for example, divides his B- and C-verses after the second position. Thiscreates an iambic foot initially in each: x / I x I and x I I I x, respectively.Pope, on the other hand, as shown above in (2.14), argues that the secondmeasure of Sievers types B and C begins with the first stress, leaving the firstmeasure to consist of only unstressed syllables, or even a single unstressedsyllable, plus an initial rest in some verses so that the two measures will both takean equal amount of time to pronounce (57). Although Pope’s theory of isochronyis probably unwarranted, as discussed in Chapter 2, this division of verse-typesbeginning with unstressed syllables is not unreasonable. As Russom points out,B-type verses in Beowuif never consist of two iambic words, while C-type versesonly rarely consist of an iambic followed by a trochaic word (OEM 18); this is alsotrue of Guthlac B. Therefore Russom, like Pope, argues that the foot boundaryin these verses should be placed immediately before the first stress. Pope’s andRussom’s foot divisions in these cases appear to be the correct ones, but not, Ibelieve, for the reasons they give. Rule (4.42) states that the heads of words in Wmust be prosodically weak. This rule alone is sufficient to rule out rising stress,and accords with Pope’s and Russom’s observations. The placement of the foot81boundary follows naturally. For example, verses such as the following cannot bescanned:wI I00(4.43) efter niht-glme ‘after night-darkness’ (943a)*S WSWsince the word niht is strong within its compound and therefore may not appearin W. The verse must instead be scanned with the foot boundary preceding thestrongest stress:s wJ I00(4.44) efter niht-glmeSW S WA related difficulty arises when a verse consists of three fully stressed words.In this case, the problem is determining into which foot two of the three lexicalstresses should be placed. Russom (OEM 16) and Keyser (338-39) suggest thatsyntax is the determining factor: those two words which form a syntactic unitgroup together into one foot. Keyser enumerates four possible groupings:(4,45) a. Adverb + verbHt ,ã I in beran (Beo. 2152a)(he) commanded then in to be broughtb. Noun inflected in the instrumental + verbFlöd I blöde wol (Beo. 1422a)Flood I welled with bloodc. Adjective or numeral + nounBëagas ond brad gold (Beo. 3105a)rings and I thick goldTwelf wintra I tid (Beo. 147a)twelve winter’s time82d. Noun + noun inflected in the genitiveSwutol I sang scopes (Beo. 90a)sweet I song of the scopBoth Russom, as discussed above re (2.27), and Keyser point out that in thesecases, syntactic criteria play a role in determining the location of alliteratingsyllables, since phrases like those described above are treated by the poet asthough they were compound words: alliteration, for example, falls on the first orthe first and second elements of the phrase- never on the second element alone.However, the theory presented here, unlike Keyser’s and Russom’s, does notneed a specific rule of foot-assignment according to syntax. Rule (4.42) states thatthe head of a word in a W position must be prosodically weak. Therefore thefollowing scansion is automatically ruled out:Phrase, ?s 2w1 1 I00 0(4.46) Dagas forô scridun ‘days went forth’ (969b)S W SWsince foro,, as the strong constituent of the phrase, may not occupy a W position.It must instead be scanned:Phrases wI I I0 0 0(4.47) Dagas fore scridunS(W)S WFoot assignment, then, is not determined directly by syntax, but follows fromRule (4.42). In fact, assignment of foot boundaries according to syntax leads toinconsistencies in Russom’s theory, as discussed above in (2.38-39). If two wordsof a phrase pattern together as a metrical foot, why must a word like gegaf, which83consists of an unstressed prefix + stem (and which must surely be a syntactic unitjust as much as a phrase is) occupy two feet? Why should syntax be thedetermining factor only in the cases in which a verse contains three lexicalwords? A claim that foot boundaries are determined not by syntax, but byphonology, eliminates the need for ad hoc rules that treat a phrase, but not aprefixed word or clitic group, as a syntactic unit. Phrases cannot pattern likeclitic groups, or like words with unstressed prefixes, not because of bracketingrules that are determined by syntax, but due to the phonological rule that Wpositions may not contain strong constituents.The determination of foot boundaries by phonological rules rather than bysyntax fits in well with generative metrical theory, which defines meter in termsof abstract phonological constituents such as syllables, prosodic feet, etc. Forexample, because words with stressed prefixes like onbid and compound wordslike s7e-man differ in their phonological makeup (the former comprising oneprosodic word and the latter two), they are treated differently in meter: onbidmay occupy a single metrical position while s-man may not (see (4.5-6)). Note,however, that although these words are treated differently in meter, they behavethe same- as single words - syntactically. Says Hayes: “I would like to suggestthat metrical rules NEVER refer to syntactic bracketing, only to prosodicbracketing... . meter is essentially a phonological phenomenon” (“Prosodic” 224).Summary of realization parameter settings for OEAs we have seen in the course of the preceding discussion, a metrical positioncontains at most a minimal word (7min). This is a class of left-headed structureswhich comprises maximally a) one word and b) two feet. A S position mustcontain the head of a prosodic word. A W position may contain the head of aprosodic word only if it is prosodically weak.84Chapter 5. Special licenses and functional constraintsThis chapter will discuss several features of OE meter which are licensed notby the metrical rules particular to OE as laid out in Chapter 4, but by generalpoetic principles that seem to be potentially available to all meters. Includedamong these general principles are: relaxation of prominence constraints oninitial positions, empty positions, and the appearance of extrametricalconstituents.Relaxation of prominence constraints on initial positionsAs briefly mentioned above concerning example (4.24), metrical theory allowsprominence constraints to be relaxed on an initial position (Hanson and Kiparsky7, 24; Hayes “Prosodic” 247). For example, Shakespeare’s iambic pentameterdisallows a strong syllable (like the stressed syllable of a disyllabic word such astiger) in a W metrical position. But this constraint is not uncommonly relaxed atthe beginning of a syntactic or metrical unit, that is, line-, clause-, or phrase-initially, as discussed in Chapter 1 •48 In OE, relaxation of prominence constraintsis allowed verse-initially. Since the verse-initial metrical position is always 5, andthe prominence rule is that S must contain the head of a prosodic word,relaxation of constraints means that verse-initial S need not do so. Thereforeverse-initial S may contain an unstressed prefix, the head of a function word, or,since any prosodic structure that non-final function words may have is not takenaccount of by the metrical rules of OE, any number of function words.Relaxation of constraints on S in OE seems to be motivated by linguisticrequirements: the need for unstressed prefixes, function words, or strings offunction words in line-initial position. “A meter flexible enough for epic481n iambic pentameter, this is traditionally known as trochaic inversion.85storytelling must of course allow for the long strings of function words that occurin a variety of Old English syntactic constructions” (Russom OEM 33). Relaxationof constraints on S allows these verse-initial unstressed syllables to exist withinthe meter:(5.1).Z se hälga song ‘and the holy song’ (1323b)SW S WSince any stress assigned postlexically to non-final function words isdisregarded by the meter, initial S may also contain a string of function words,Any number of. unstressed syllables are not parseable as a foot; in theory,therefore, an indefinite number of such function words may appear on a singlemetrical position.49 In practice, of course, the number is limited by the syntax ofthe language, and probably by stylistic factors. The poet of Guthiac B rarely hasmore than two, occasionally three, unstressed syllables on any given metricalposition:(5.2) ac h on m jde ‘but he on the land’ (831a)S W SWNote that on occasion the relaxation of constraints on initial S will result inambiguous scansions. In the above example, S may contain ac and W he’ on ldm;alternatively, S may contain ac he’ and W on pam, and so forth. This is not aproblem. The point of generative metrics, unlike that of traditional OE metrics,is not to rule out all but one possible scansion, but rather to describe whatdistinguishes metrical lines in contradistinction to unmetrical lines (or verses).Ambiguous scansions may be tolerated, as long as the lines to which they applydo not violate any metrical rules,5° Since a metrical position in OE may contain a49See Hanson’s discussions of such sequences in the metrical practices ofHopkins (Resolution 148) and Tennyson (155-56).50Just as they are in Shakespeare’s iambic pentameter, in which a resolvablesequence such as delicate may be scanned either delicate or delicate (Hanson,“Prosodic”). S W S W86number of unstressed syllables, it makes no difference whether these syllables inthe line above are assigned to S or W. But for the sake of consistency indiagramming, I shall arbitrarily assume that the first of a verse-initial string ofunstressed syllables is in a S position and the rest are in W.Empty metrical positionsIn Chapter 4 we discussed the various types of constituents that may appearon a given metrical position: the minimal word, together with the class ofstructures parseable as the minimal word, unstressed syllables, etc. A metricalposition, as we have seen in (4.5b-c), (4.16), and elsewhere, may also be occupiedby nothing; that is, it may be empty.An empty position may be the result of one of two factors. The first iscatalexis, or the absence of a peripheral weak constituent: for example, the emptyinitial W position sometimes found in iambic meters, or the empty final Wposition which is quite common in trochaic or dactylic meters. The secondreason an empty position may appear stems from the minimum realization of theposition parameter (which, recall, is set at 2min for OE meter), which definesonly the upper bound of what may occupy a metrical position.Although verse-final W positions in OE may be empty as a result of syllablecatalexis as in (5.3a) below, catalexis cannot explain empty W verse-medially, asin(5.3b):(5.3) a. eard-wTca cyst ‘best of earthly dwellings’ (853b)S WS(a. step stal-gongum ‘advanced with stealthy paces’ (1 140a)S(S WI shall therefore assume that empty positions in OE result from the minimumrealization of the position parameter. Because S positions must contain the headof a prosodic word by (4.19), only W positions may be empty.87Although the empty W position resulting from minimum realization of theposition parameter is not a common feature in the meters studied by generativemetrists, it is not unknown. Kiparsky’s study of Hopkins’s sprung rhythm showsthat Hopkins freely allows empty W positions (“Sprung” 311).51 This parallel useof empty W in both OE and sprung rhythm is probably one of the factors leadingsome critics to claim that there are stylistic affinities between the two.52In OE, empty W together with relaxation of constraints on initial S appearsin some B and C-type verses:(5.4) a. on fen-ttd ‘in evening time’ (lZl5a)S(W) S Wb. in sin-dramum ‘in eternal joy’ (839b)S(S WAn unusual variant of this pattern appears when it appears that the soleoccupant of the first foot is an unstressed prefix:(5.5) a. a sanian ‘to grow weak’ (1175a)S(W) S Wb. a cennedne ‘born’ (1361a)S(W)S WAccording to Russom (OEM 36), verses like the above occur about 50 times inBeowuif; there are five unambiguous examples in Guthiac B.53Scansion of these examples as given in (5.5) is somewhat troublesome, since anempty position appears between the stem and prefix of a word. Although this isnot a problem for Russom, who regards unstressed prefixes as function words51Kiparsky also reports two instances of empty S positions in Hopkins’s sprungrhythm, but this seems to be a highly marked usage.52For a more detailed comparison of sprung rhythm and OE meter, seeStephenson.53They are: 1324a, 1361a, 1175a, 1177a, and 1252a. Another three possibleexamples, 1 128a, 1 133b, and 1207b may be scanned as normal A-type verses withan initial extrametrical syllable if resolution on S is suspended.88OEM 8), it might be best to regard these unusual verses as occasional one-footverses with an initial extrametrical syllable, scanning them as:(5.6) a. (a) sanianSWb. (a) cennedneS WThe question naturally arises as to whether verses such as those in (5.4) shouldalso be considered as one-foot verses with an initial extrametrical syllable. Iconsider this unlikely, however, for reasons which I shall discuss below.ExtrametricalityExtrametrical syllables are syllables that lie outside the meter; that is, they arenot licensed by the rules which assign prosodic material to metrical positions.They have their counterpart in phonological rules which allow peripheral weakconstituents to be disregarded in parsing. In meter, extrametrical syllables mayappear immediately before or immediately after metrical constituents such asfeet, lines, etc.Iambic meters often allow an optional extrametrical unstressed syllable to fallafter a S position; that is, immediately following a WS foot and before a syntacticbreak (Kiparsky “Rhythmic” 231, Hanson and Kiparsky 25). Conversely, trochaicmeters allow an optional extrametrical syllable to fall between a syntactic breakand a S position - that is, immediately preceding an initial SW foot (Kiparsky“Rhythmic” 232). Traditional accounts of OE meter call this phenomenon, inwhich an optional unstressed syllable appears before the first stress in a verse,anacrusis. Unstressed syllables in anacrusis are occasionally found preceding theinitial stressed syllable in Sievers A-and D-types.5454Russom postulates that anacrusis does not appear before E-types because inthese cases the first foot of such an E-type would tend to be interpreted as a89Russom takes a much broader approach to extrametricality than thetraditional accounts of OE meter do. Because his theory is based on acorrespondence between words and metrical feet, with foot patterns beingderived from word patterns, he is forced to treat unstressed syllables which donot fit within his derived metrical patterns as extrametrical. For example,because no OE word corresponds to the pattern Sxxxx, he must regard theparenthesized elements in:(5.7) egnas (syndon ge-) wre ‘thanes are united’ 1230a)Sx (xx x) Sxas lying outside the meter (OEM 19), scanning the verse above as Sx I Sx.Russom’s rule for extrametricality, given above in (2.28), states that extrametricalunstressed syllables may appear before either foot in a verse. This is a much looserinterpretation than that of Sievers’s account of OE meter, which considers one ormore unstressed syllables indifferently as a dip, or non-stressbearing segment(272). Since the generative model of OE meter which I am outlining on thesepages always allows a sequence of unstressed syllables to occupy a single metricalposition, Russom’s rule that extrametrical syllables may appear before either footis unnecessary. In example (5.7) above, for example, syndon ge- may occupy asingle W position, and there is no need to consider any of these syllablesextrametrical. In this regard, the generative theory proposed here is more in linewith that of Sievers and other traditional theorists than it is with Russom.Anacrusis is a very limited feature of OE poetry. The number of A-types withanacrusis in Beowuif has been calculated differently by various metrists; thenormal C-type verse (see the discussion of overlap in Chapter 6 below). “A versepattern (x) Ssx I S would have a false sense of closure at the verse-medialboundary, creating confusion about the number of feet” (OEM 34). A possibleexception to the avoidance of anacrusis in E-type verses is GuthB 1317a.90highest count of A-types given is 125 (Cable Meter 33), though Bliss lists only 27instances of A with anacrusis and 28 of D with anacrusis (Metre 127, Table III).The syllable in anacrusis is almost always an unstressed prefix or the negativeparticle ne (Cable Meter 35, Duncan 16). As discussed in Chapter 2, theavoidance of anacrusis, according to Cable (Meter 43), seems to be a major part ofthe poet’s craft, since almost 50% of OE prose phrases are introduced by one ormore unstressed syllables.Anacrusis is likewise a very limited feature of Guthlac B, occurring in only 24verses: 13 A-type verses, 10 D-types, and one E-type. Examples are:(5.8) a. fsed on for-sTh ‘impelled on the journey’ (939a)SW SWb. IstOd stronglice ‘withstood strongly’ (903a)S(S WWith only four exceptions (840a, 922a, 1291 a, and 131 7a), the syllable precedingthe first S position is an unstressed prefix. Anacrusis in Guthlac B ismonosyllabic with only two exceptions:(5.9) a. et iãm halgan 1owon ‘from the holy servant’ (922a)SW SWb. swã se burg-stede ws ‘so the citadel was’ (1317a)SW S(W)Extrametricalitv in Sievers B- and C-versesIt may be argued that the initial unstressed syllables of Sievers types A3, B or Care, like the unstressed syllables preceding some A- D-, and (rarely) E-type verses,extrametrical. This is a point which deserves consideration, especially in view ofthe fact that, unless one wishes to postulate an empty W position word-medially,an unstressed prefix may in any case stand extrametrically before a single foot insome B- and C-type verses, as discussed re (5.5-6). I shall, however maintain that,with the possible exception of verses like those discussed in (5.5), the initial91unstressed syllables of A3, B- and C-types are not extrametrical, but result fromthe relaxation of prominence constraints on initial S positions.Assuming for the moment that the initial light syllables of A3, B- and C-typeswere indeed extrametrical, one would have to postulate the existence of twodistinct types of extrametricality: that shown in these verses, and that shown inA-, D, and E-type verses, which I shall continue to refer to, for the sake of clarity,as anacrusis.First, as Cable (Meter 35) and Duncan (16), point out, the syllable in anacrusisin A- and D-types (they do not discuss the very rare instances in E) is almostwithout exception a monosyllabic unstressed prefix or the negative particle ne.While the initial unstressed material in B and C does sometimes consist of amonosyllabic unstressed prefix (though the first foot of an A3 verse never consistsof only one syllable), far more often it consists of function words or strings offunction words of a type which only very rarely appears in anacrusis:prepositions, demonstratives, pronouns, complementizers, auxiliary verbs, etc.Sometimes, as we shall see in Chapter 7, the first foot of a B or C verse evencontains a finite verb which does not participate in alliteration. Finite verbsnever appear in anacrusis in A, D, or E-types; at least there are none in Guthiac B.The generalization seems to be that syllables in anacrusis almost always form atight syntactic unit with the following word. Note that no material mayintervene between an unstressed prefix and its stem, nor between adverbial neand the following verb (in fact, ne plus a following verb beginning with a vowel,Ii, or w is generally contracted, producing nis from ne is, nes from ne wtes, and soforth).Secondly, anacrusis consistently appears in the first half-line, or on-verse inGuthiac B; while B and C verses appear as both on- and off-verses (A3 verses, onthe other hand, almost always appear in the on-verse in Guthiac B). Third,92anacrusis is rare, as we have seen above: 24 verses in Guthiac B, or about 2%; whileB and C verses are ubiquitous, comprising about 25% of Guthiac B’s verses. (Thisfinal point, while it is not in itself crucial to establishing a difference in kindbetween anacrusis and the first feet of B- and C-types, in combination with theother factors, is highly suggestive.)Therefore, in form (monosyllabic prefix or ne vs. polysyllabic function wordsor even finite verbs), distribution (on-verse vs. either on- or off-verse), andfrequency (rare vs. common), syllables in anacrusis are almost always quitedistinct from the unstressed syllables preceding the first stress in Sievers B- andC-type verses. In only one OE text, the metrical Psalms of the Paris Psalter, doespolysyllabic anacrusis before A, D, and E-types occur with any frequency (Bethel34). Even in this text, which is highly unusual, anacrusis occurs fairly seldom, inonly 6.8% of full lines (33), and almost without exception occurs in the on-verse(35).As well as these differences in form, frequency, and distribution, to considerthe initial function words in A3, B, and C as extrametrical would mean that theseverses would consist of only two or three metrical positions:(5.10) a. nU ka gearwe const ‘now you readily know’ (1045b)SW S(W)b. 7 his sefan trymman ‘and prepare his mind’ (11 16b)S(S Wc. 7 a rendu ‘and the messages’ (1296a)SWd. nis me earfeôe ‘it is no hardship to me’ (1065b)SWe. et ]ü gesecge ‘that you tell’ (1179a)SW93Without the initial unstressed syllables, verses like those in (5.lOa, c-d) wouldhave only three syllables; some, like (5.lOc-e), would consist of only one foot; andsome, like (5.1 Oe) would have only two syllables. In Chapter 6 we shall discussthe absence of two-and three-syllable verses which might result from an empty Wposition in terms of an overlap constraint which serves to minimize confusionbetween verses and feet by requiring that verses be interpretable as two feet. Itmay be argued that the overlap constraint is at work in the examples above byrequiring extrametrical syllables which serve to mimic a foot, thus allowing asingle foot to be interpretable as two. This seems unlikely, however, for if thehead of a metrical foot is always required to contain a stressed syllable, that is, ifrelaxation of prominence constraints on initial S were not allowed, extrametricalunstressed syllables are unlikely to be interpretable as a foot.Finally, if hypermetrical verses are regarded as having three feet (as I shallargue in Chapter 8), it is not unreasonable to suppose that OE has one-foot verses;and I have already allowed that B- and C-type verses preceded by only anunstressed prefix may be better regarded as occasional one-foot verses. But interms of distribution, hypermetrical verses are normally set aside in clusters andnot intermingled with normal (that is, two-foot) verses. This may be for stylisticeffect, or, alternatively, to aid the listener’s recognition of them as having morethan the usual number of feet. But the putative one-foot verses discussed aboveare mixed in randomly with normal verses. A3 verses, it is true, do typicallyappear in the on-verse; but even these verses are not clustered near each other,but occur widely spaced throughout the poem. Now just because verses that arelonger than the norm tend to cluster together, it does not necessarily follow thatverses that are shorter than the norm should behave in the same manner. It doessuggest, however, that putative one-foot verses behave more like two-foot versesthan otherwise. I suggest that these verses are two-foot verses.94To summarize, then, the initial unstressed material in A3, B-, and C-typeverses is probably not extrametrical, but arguably results from the relaxation ofprominence constraints on initial S. Unstressed syllables preceding A-, D-, and(very rarely) E-types are, on the other hand, extrametrical; extrametrical syllables,as discussed above, may in trochaic meters fall between a syntactic break and a Sposition. Extrametrical syllables in OE are almost without exceptionmonosyllabic unstressed prefixes or ne; they almost always occur line-initially.Since there exists the occasional exception to both of these tendencies, however, Ishall not state them as rules.95Chapter 6. Overgeneration and rare verse typesOvergenerationAlthough the metrical rules and special licenses as given in Chapters 4 and 5adequately describe all of the lines in Guthiac B (excepting hypermetrical verses,which I shall discuss in Chapter 8), they also vastly overgenerate, predicting linesthat do not actually occur, not in Guthiac B, nor, that I am aware of, anywhereelse in the corpus of OE poetry.The position parameter is one of the culprits here. Remember that themaximum size of a constituent which may occupy a metrical position is the?min, or minimal word. This means that words such as onbid (two feet) as well aswords containing less prosodic material, for example cynnes (one foot plus anadjoined syllable), may occupy a single position. But there is no metricalconstraint, and can be no metrical constraint, that prohibits such words fromoccupying every metrical position in a verse. As an example, the rules proposedhere allow lines such as:(6.1) *ende.dgor middan-geardes ‘final day of middle-earth’SW S Win which a foot plus adjoined material occupies every metrical position.The second culprit is the empty W position. If W positions are allowed toremain unfilled, the theory predicts that verses such as the following example,which has two empty W positions, will occur:(6.2) wts word ‘wise words’S (W) S (W)What prevents such verses from occurring freely in the poetry? Twointerrelated arguments can be made. First is an argument from complexity;second is an argument which Russom proposes, an argument which derives from96the fact that OE poetry was oral, meant to be listened to rather than read. He callshis argument the overlap constraint. I shall take these arguments in order.Metrical ComplexityAccording to Russom, the prototypical OE verse is SxISx, or, in Sieversianterms, the A-type: Ix I Ix. “This pattern, which corresponds to Sievers’s type Al,is an obvious candidate for the [metrical] norm, since it has by far the highestrelative frequency” (OEM 28). This statement fits in well with the generativetheory presented here, which has an abstract underlying SW I SW metricalpattern. If the trochaic Ix I /x is the most neutral expression of the underlyingmeter, this suggests that the normative correspondence between metricalpositions and linguistic units is more on the order of 0mm, the minimal foot (astressed syllable or a resolvable sequence), or the syllable rather than, say, a footplus adjoined material. That is, while the position parameter (4.3) allows a givenmetrical position to contain as much material as two feet provided they are in thesame prosodic word, and while this is indeed sometimes the case, this is notnecessarily the most prototypical or neutral expression of the correspondencebetween metrical positions and linguistic units. I would like to suggest thatwhen a metrical position contains a constituent larger than the prosodic foot, theresult is a more complex, though of course quite legitimate, realization of theunderlying metrical pattern.Halle and Keyser suggest that the more complex a given line is, the moredifficulty it poses for the reader, whose task it is to interpret lines as beinglegitimate realizations of the underlying meter. Even a line which is perfectlymetrical (in that it breaks none of the rules of the meter) may nevertheless be socomplex that any reader would be hard pressed to discern the underlyingpattern. Such highly complex lines, while metrical, will be disfavoured in meter,97just as grammatical but highly complex sentences will be disfavoured in speechor writing.55 It would be highly unlikely, for example, to find a line of iambicpentameter with a resolved sequence in every S position or a stressedmonosyllable in every W position, even though the metrical rules maytheoretically allow these possibilities. According to Halle and Keyser: “If it isgranted that the complexity of a line is directly related to the difficulty that theline in question poses for the reader, and if one further supposes that poetsnormally do not wish to turn their poems into difficult crossword puzzles theartistry of which cannot be appreciated without laborious pencil and papercalculation, then it is not unreasonable to assume further that there is an upperbound on the complexity that a given poet would ever wish to impose on hislines” (“Iambic” 233-34).According to Youmans, the reader (or listener) measures lines of verse againstan abstract metrical prototype. Whether or not a given line is judged as metricalis determined by how closely it conforms to this prototype (341). The distinctionbetween metrical and unmetrical lines is not, Youmans suggests, always a clear-cut one. He therefore proposes that there is a “Platonic” component to meterwhich allows for degrees of metricality, or for a fuzzy rather than a well definedset of metrical lines. This he contrasts with the “Aristotelian” approach ofgenerative metrics, in which metricality is an all-or-nothing affair and a givenline either is or is not metrical according to some particular standard. Youmansproposes that metricality is a relative phenomenon, and that metrical rules“must define degrees of metricality rather than clear-cut distinctions betweenmetrical and unmetrical lines” (342).55For example, consider a grammatical but multiply embedded sentence suchas “This is the cheese that the rat that the cat that the dog chased hunted ate.”98There is, I think, some merit to Youmans’s point of view, although I wouldnot advocate abandoning the Aristotelian perspective altogether. The purpose ofthis present paper, and of generative metrics in general, is to describe theAristotelian aspect of meter. Now that we have established the structure andrealization parameters for OE, it seems clear that verses which violate theseparameters by having an unstressed syllable in non-initial S, by having a strongconstituent in W, or by having more than min in any position, are indeedunmetrical by any standard. But the Platonic aspect obviously plays anenormous part in OE meter, and needs to be further explored. There seems to bea “fuzzy” area in OE meter in which complex verses predicted by the meter donot actually occur: not because they are unmetrical according to the rulesproposed in this paper, but because, perhaps, they would result in the listener’sbeing unable to easily discern the underlying metrical pattern. Russom, with hisoverlap constraint (given above in (2.29), is the first metrist to recognize anddiscuss in detail this Platonic component in his study of OE meter.OverlapFollowing Russom’s suggestion that the metrical pattern Sx I Sx (in Sievers’snotation /x I Ix) is the prototypical or normative realization of OE meter (OEM28), I would like to suggest that the rhythmic pattern /x I /x is the most neutralor least complex realization of the underlying SW I SW meter. Each position haslinguistic material in it; there are no empty positions. Each S contains the head ofa prosodic word, and each W contains material consisting of an unstressed affixor function word. The verse is therefore easily discernible as two feet, and isvirtually transparent to the underlying meter.Empty metrical positions and relaxation of constraints on initial S (which maytend to invite interpretation of two feet as one), as well as filling a metrical99position with more than 0mm (which may tend to invite interpretation of onefoot as two), are all more complex realizations of the underlying form, andrequire more analytical effort by the listener. So verses like:(6.3) a. *se man ‘the man’S(W) S (W)b. *gd man ‘good man’S(W)S (W)c. *endedogor middan-geardes ‘final day of middle-earth’S W S Wdo not occur in the poetry, not because they violate any metrical rule, butbecause, I would like to suggest, they are so complex that the average Anglo-Saxon listener (who does not even have the advantage of pencil, paper, andcrossword-solving experience) would be virtually unable to recover theunderlying meter. The problem with the verses in (6.3a-b) is that the rhythmicpattern of each more closely approximates that of a foot than that of a normativeverse. A verse like (6.3c), on the other hand, is more likely to be interpreted astwo verses. Both of these forms of complexity - that arising from empty positionsand that arising from “overstuffing” a verse- are subsumed by what Russom callsoverlap.Overlap, according to Russom, results when the stress pattern of a given footis identical to that of a possible verse, which results in the possibility that alistener might become confused about where the foot and verse boundaries lie(OM 26). For example, consider this verse from Guthiac B:(6.4) ô se al-mihtiga ‘then the Almighty’ (950b)SWS W100This verse does not violate any metrical constraints as outlined in Chapter 4, butit is the only verse of its kind in the poem.56 The question naturally arises: ifverses like this are metrical according to our theory, why are they so rare?The answer seems to be that the second foot contains material which is itselfinterpretable as two feet:(6.5)- mihtigaS(W) S WThe rhythmic pattern of the second foot of (6.4), that is, corresponds to a SieversDl verse type, and verses with the same rhythmical pattern are not uncommonin Guthlac B or OE poetry in general. A typical Dl verse is:(6.6) gst - haligne ‘man holy in spirit’ (1 149a)S (W)S WPresumably, an Anglo-Saxon audience presented with the verse in (4.55) wouldtherefore be quite likely to interpret it as consisting of three rather than two feet:(6.7) ô se el - mihtiga*SWS(W)S WTherefore verses which would tend to be interpreted this way are generallyavoided by the poet.57Even when OE poetry was written down, the poetic texts included no visualcues, such as punctuation or line endings, to indicate the boundaries of prosodicunits, such as phrases or sentences, or metrical units such as lines or verses(O’Keefe 1-2). Therefore it would have been very important that such prosodicand metrical facts be easily recoverable from the poetic language itself (21), It is561t is unmetrical according to Sievers’s classification system, being a C-typewith two syllables in the final dip: xx! I \xx. I am assuming that cel- has primarystress, since it bears the alliteration of the verse.57Another interpretation of Guthiac B 950b might be that the verse consists oftwo feet with two syllables of anacrusis. This interpretation is unlikely, since, asdiscussed in Chapter 5, the OE poet almost always limits anacrusis to the onverse. Furthermore, anacrusis typically involves only unstressed prefixes or thenegative particle ne.101likely that one element of prime importance in recovering such metrical factsfrom either text or recitation would be to maintain a clear distinction betweenfeet and verses. Since OE verses consist of only two feet, and since some feetactually contain more syllables than some verses do, there is a real potential forconfusion between the two. Constraints must be provided to allow a listener todisambiguate feet from verses; otherwise some feet which are longer and heavierthan the norm may be perceived as constituting an entire verse. Russom,therefore, postulates that an overlap constraint is necessary to allow a listener todisambiguate feet from verses (OEM 26):(6.8) Foot patterns may not overlap verse patterns.What Russom means by this rule is that, for instance, since SxISx is anallowable verse pattern (analogous to Sievers A), the meter does not allow a footpattern of the form Sxsx (26).In fact, Russom’s overlap constraint may too strict. He invokes it specificallyto rule out the “light E” verse SxxIS which overlaps the foot pattern Sxxs (26).However, the “light E” does actually exist, although it is rare:(6.9) Adame geaf ‘gave to Adam’ (869b)S WS(W)Since verses such as those in (6.4) and (6.9) do actually exist, as well as the veryoccasional three-syllable verse (although there are none in Guthiac B), it might bebest to regard the overlap constraint as a general tendency rather than a strictprohibition, and so I shall reword it as:(6.10) Maximize the distinction between foot patterns and verse patterns.This rule means that the rhythmic pattern of normal verses, which resultsfrom matching prosodic material with metrical positions, should beinterpretable as two feet. Since hypothetical verses such as that in (6.3c) andactual verses such as in (6.4) are easily interpreted as four and three feet,102respectively, they violate the overlap constraint; and therefore verses such asthese will occur very rarely, if at all.58The overlap constraint, as Russom suggests, may also account for the lack ofthree-syllable verses resulting from empty W positions, which could be easilyinterpreted as comprising a single foot (OEM 29). For example, consider thefollowing hypothetical verses:(6.11) a. *dryhten bad ‘the lord experienced’S W S(W)b. *Men cunnon ‘men knew’(S Wc. *wis word ‘wise word’S(W)S (W)Respectively, these hypothetical verses would have similar rhythmic patterns tothe following feet:(6.12) a. on a gocran trd ‘in the sad time’ (976b)b. onne so orag cvmeô ‘when the time comes’ (1350b)c, lof mon lofum ‘beloved man to the beloved’ (1 164a)In fact, it is tempting to speculate that the overlap constraint, whichmaximizes the rhythmic differences between foot patterns and verse patterns,ensuring that a verse is interpreted as two feet, is the basis of Russom’s claim thatOE foot patterns correspond to OE word patterns. Can it be that one of thePlatonic constraints on OE meter is that the listener be able to recover a pair ofwords, or, more accurately, phonological words, from a given verse? If so, thislends support to Youmans’s claim that there are two components to meter, andincorporates Russom’s claims within this framework. The Aristotelian orgenerative component defines the parameters of the meter according to rules.58Hypermetrical verses, which, I shall argue, do actually have three feet, willbe discussed in Chapter 8.103The Platonic component, in OE, at least, would be a functional constraint thatgoverns the interpretation of the resulting verse as two phonological words.59 Ifthis is so (and I am offering it only as a suggestion), then OE poetry, far frombeing lax and unregulated, is far more complex and sophisticated than hashitherto been suspected.Nonexistent and rare verse typesSeveral verse types predicted by the generative theory presented here are sorare that they are generally considered unmetrical. Consider, for example:(6.13) a. frorig 7 fero-wrig ‘cold and soul-weary’ (1157a)/xx I /\xb. dor-möd on dëgle ‘bold man in darkness’ (952a)/\x I /xc. morjor-bed strêd (Beo. 2436b)/x\I/d. *glaw.mod hyge-gomor (prudent one, sad in spirit)The metricality of verses such as (6. 13a), in Sieversian terms a Dl verseexpanded by two unstressed syllables in the first foot, has been a matter of somedebate. Russom disallows such verses on the basis that there are no verses inBeowuif in which each foot is made up of a single word. Actually, there is at leastone such verse:(6.14) eahtodan eorlscipe ‘praised his nobility’ (Beo. 3173a)59Which would perhaps make the position parameter setting of ?min rathercounterintuitive; if min were to occupy every metrical position in a verse, theverse would tend to be interpreted as four words. In fact, as I have pointed out,the more normative constituent on a metrical position is the syllable or the foot(0mm). Unambiguous examples of ?min occupying a single metrical positionare not very common.104Since it does not matter to generative theory, which is not based on a word-footequivalence, whether or not a foot is comprised of one or more words, and sincethere are five examples of verses with the same rhythmic pattern in Guthiac B,6° Isee no reason to consider verses like that in (6.13a) unmetrical. Fulk also accepts“the expansion of type D*. . . (though this type is rare)”(224).Verses like that in (6.1 3b) are considered unmetrical both by traditionaltheorists and by Russom.61 Again, since the generative theory presented in thispaper predicts this type, and since Guthlac B has seven examples, I see noproblem with considering verses like these metrical, although, like the examplesdiscussed just above, they are quite rare.62On the other hand, Guthiac B has no examples of verses such as that in(6. 13c), and even Beowulf has just the one quoted. And I could find no examplesin either poem of verses like that in (6.13d). Russom discusses the absence ofthese two patterns as due to their extreme complexity (OEM 30-31), which maybe the case. However, this analysis, I think, begs the question: Why are verseslike those in (6.13) so complex that they rarely (and in some cases never) appear,while other verses, which seem to be just as complex - for example, B or C-typeverses with an empty W and relaxation of constraints on initial S - appear quiteoften?One answer may be that there are several levels of complexity. We havealready discussed the overlap constraint, which rules out verses which areinterpretable as having fewer or more than two feet. Still, there are verses whichdo not seem to be interpretable as having the wrong number of feet, such as those60They are: 1157a, 1172a, 1275b, 1284a, and 1331a.61For a full discussion, see Fulk (153-68). Russom (OEM 31) notes that “thereare no reversed half-line patterns such as Ssx I62The examples are: 952a, 993a, 1102a, 1219a, 1244a, 1331b, and 1357a.105in (6.13), which are either very rare or nonexistent. There must be some otherfactor or factors constraining these verses. Russom suggests that verses like(6. 13c) are avoided because a listener would tend to misinterpret the footboundary as following the second, rather than the third syllable (OEM 30). Thissounds plausible, but it does not explain the non-occurrence of verses such as(6. 13d). It would be quite satisfying to come up with some generalizations thatwould explain the existence of some complex types of verse and the nonexistence of others. Having separated out the Aristotelian from the Platoniccomponents of OE meter, that is, it now seems that there are a number ofcomponents to the Platonic aspect. But that study is beyond the scope of thispaper.106Chapter 7. AlliterationAs briefly mentioned in Chapter 1, alliteration in OE poetry has itscounterpart in phonological rules constraining ordinary language. Kurylowicz(112) and Kiparsky (“Role” 19-20) point out that the alliterating constituents ofwords in North and West Germanic meters (which include OE meter) areidentical to the elements which are repeated according to the rules forreduplication of initial constituents in the preterites of some Gothic verbs. Theserules include: (1) repetition of the initial consonant sound; (2) treatment of theclusters st- sp- and sk- as units; and (3) identity of vowel sounds.In terms of syllable structure, the syllable onset of the head63 of a lexical wordis involved in both Gothic reduplication and OE alliteration. The first segment ofthe onset, or the first two segments in the case of st- sp- and sk-, in which the sappears to be extrametrical64is the constituent which is repeated. In syllableswhich have no onset, that is, which are vowel-initial, the empty onset isrepeated; hence all vowels may alliterate with each other. Due to phonologicalchanges in the history of OE, the cluster sk- appears as sc-. The two allophonesspelled g ([g] and [j]) alliterate with each other, as do the two allophones spelled c([kJ and []). For the sake of simplicity, in the following discussion I shall refer toa syllable of primary stress in which the onset contains an initial nonextrametrical segment identical to at least one other such segment in the line asan “alliterating syllable.”63The head of a word, as discussed in Chapter 3, is its strongest or onlysyllable. In OE, the head is always either the initial syllable of the stem or, lesscommonly, a stressed prefix. Unstressed prefixes are never involved inalliteration.64Kristin Hanson, Class lecture in Stylistic variation; October 1991.Extrametrical constituents in phonology are peripheral constituents which aredisregarded in parsing. For a discussion of extrametricality in PDE, see Hogg andMcCully (106-24).107Russom explains alliteration as a function of the metrical hierarchy, arguingthat a rule of metrical subordination, analogous to the Compound Stress Rule,determines the location of alliterating syllables (OEM 67). His argument, thoughbased on metrical rather than prosodic structure, ultimately derives fromKurylowicz, who notes that “from the rhythmical point of view the Germanichemistich [half-line] is a kind of COMPOUND. . . i.e. a rhythmical unit of ahigher order than the ordinary compound word” (119). The OECSR, as we haveseen in (3.16), is a rule which labels the first lexical constituent of a compoundword S, subordinating the stress on the second lexical constituent. Russomapplies an analogous rule to the metrical structure of an OE line (71):(7.1) When two constituents containing S positions appear within thesame metrical domain, label the first constituent strong and thesecond constituent weak.Note that this rule applies only to S positions, that is, positions occupied bysyllables bearing primary stresses (or secondary stresses in the case of whole-versecompounds such as middan-geardes (SxISx)). Like the OECSR, which createscompounds only from lexical words, function words (symbolized x in Russom’snotation) are irrelevant to rule (7.1), The metrical compounding rule as given in(7.1) above also has no effect on s positions, since they always have subordinatedstress (O.FLM 73). A s position is always labelled W.Russom (71) gives the following example of a line made up of two Sx I Sx (orSievers A) verses:str ng weakstro’eak strc’”eak(7.2) Sx Sx Sx SxNote that in each metrical domain- foot, verse, and line- the first constituentcontaining a S position is labelled strong and the second weak. Russom then108proposes the following rules, which, when applied to the labelled positions,account for the distribution of alliterating syllables (73):(7.3) a. The strongest two metrical positions within the line must containalliterating syllables.b. A weak constituent of a weak constituent may not contain analliterating syllable.c. No alliterating syllable may occupy an x position.d. Otherwise, alliteration is optional.In example (7.2) above, the head of each verse alliterates by (7.3a). The fourthposition may not contain an alliterating syllable by (7.3b). The second positionmay contain an alliterating syllable by (7.3d).Russom’s rules for alliteration in Sievers B- and C-types, however, areinconsistent with the rules for universal metrics as discussed in (4.1-2). Russomgives, for instance, the following example of a C-type first half-line, or on-verse(73):strong (on-verse) weak (off-verse not shown)weaongstrong weak(7,4) x I S sxNote that the second level from the top- the foot level- is labelled WS whileall other levels are labelled SW, As noted above re (4.2), however, Kristin Hanson,following Prince, points out that every level on the metrical hierarchy is assumedto be labelled in the same direction, which, in the case of OE, is SW (personalcommunication). There is no mechanism in the model I am proposing here thatallows metrical labelling to invert to WS just in case an initial position containsunstressed rather than stressed material. Therefore, unlike Russom, and more in109keeping with the assumptions of traditional metrists, I shall argue that rules foralliteration make reference not to the metrical level, but to the prosodic level.Meter and alliteration, I would like to suggest, are independent structures, eachmaking reference to the same linguistic constituents in different ways.Russom’s rules for alliteration as given in (7.3) may be easily adapted to themodel I am proposing in this paper, with the proviso that they apply not at themetrical level but at the prosodic or language level. But before discussing theserules and the amendments I shall propose, let us provide a formal definition of analliterating syllable:(7.5) An alliterating syllable is defined as the head of a prosodic word theonset of which contains an initial non-extrametrical segmentidentical to at least one other such segment in the line.Note that function words, since they are not prosodic words, may not contain analliterating syllable unless they are footed by the Phrase-Final Stress Rule andthereby count as prosodic words. That is, even if the onset of a function wordhappens to contain the same constituent which takes part in the alliteration ofthe line, it does not count as an alliterating constituent. By the same token,unstressed prefixes are ignored by all rules concerning alliteration, since they tooare not the heads of prosodic words. For example:(7.6) a. rest tre jdese [7] hëo AdameS W SW SW SW‘first to the woman and she to Adam’ (983)b. worulde lifes. ôã [ws] wop 7 hafS W SW S W S W‘of life in the world. Then was mourning and wailing’ (1047)c. dig on lne Qndcwis [ã]geafS W SW S W S(W)‘[man] blessed in courage gave an answer’ (1026)110d. t ho is jãn-fat j2eorge [bi]fsteS W S W S W SW‘that she should commit this body to the grave’ (1193)In the examples above, constituents involved in the alliterative pattern of theline are underlined. The elements in square brackets, since they are not heads ofprosodic words, do not constitute part of the alliterative pattern; the fact thatthey happen to have onsets identical to those involved in the alliteration of theline is irrelevant.Having defined what constitutes an alliterating syllable, it remains toimplement the labelling rule which will determine the distribution ofalliterating syllables. Recall that the OE Compound Stress Rule (OECSR), asshown in (3.16), labels the first element of a compound word S. In order toextend this rule to fit the higher-order compound, as Kurylowicz puts it, of theOE verse - or, I would like to suggest, the OE line - I propose a rule similar to theOECSR, which I shall call the Compound Alliteration Rule (CAR):(7.7) Within the domain of the line, for any pair of sister nodesdominating prosodic words, the leftmost is S.Just as the OECSR applies within the domain of the word, and the TrochaicStress Rule (TSR) applies within the domain of the phrase, the CAR takes as itsdomain the higher-order constituent of the line. Note, however, that the CAR isnot a prosodic rule of the same order as the CSR and the TSR. The poetic line isnot necessarily a prosodic unit as is the utterance or sentence, but is anabstraction, since a clause or even a sentence boundary may occur at the caesura;in fact, enjambment is a favoured stylistic device among OE poets:(7.8) fger 7 gefalrc. I Fder ws kenned (825)‘fair and joyous. To the Father was born’111If the CAR were a prosodic rule, one would have to assume that the on-versewould always be stronger than the off-verse. Since a sentence boundary mayintervene, as in (7.8) above, this is highly unlikely.I would like to suggest, however, that the CAR is a rule which mimics theoperation of a prosodic rule like the CSR in that it treats the line as though it wereprosodically a sentence, in much the same way that metrical rules as discussed in(4.14-16) treat sequences like swThe ne as though they were rhythmically words.There is some slight independent evidence that can be adduced in favour of thisargument: of the approximately 440 verbs in Guthlac B, fully half are in line-finalposition, and only about one-third appear in the on-verse. It may be argued thatthis tendency to distribute verbs toward the end of the line is a reflex of thetendency for OE verbs to appear finally in subordinate clauses.Rule (7.7), then, is a rule which takes the entire poetic line as its domain,treating it in the same way that the OECSR treats a compound, or the TSR treats aphrase. The term sister nodes refers to nodes on the same level of the prosodichierarchy, that is, of a tree diagram of the sort discussed at the beginning ofChapter 3. To illustrate using the example already discussed in (3.3), the triplecompound law degree requirement is parsed:52s XwI 1 I(7.9) law degree requirementNote that each prosodic word is labelled as such by the notation . (For thepurpose of this discussion, I shall omit labelling at the foot, syllable, and moralevels.) The words law and degree are sisters, since they are both elements of thecompound law degree. The node of this lower-level compound is in turn a strongsister to the node dominating requirement.112Extending Rule (7.7) to a line of OE poetry, consider the following:AA?s w w ?s1 2 3 4 5(7.10) rëot-hord gnornao, st hine f53seo ‘body mourns, spirit hastens’S W SW S WSW (1266)I have numbered the nodes dominating prosodic words for ease of reference.Nodes 1 and 2 in the example above dominate prosodic words within acompound; they are sister nodes and so labelled S and W respectively by theOECSR. Nodes over greot-hord and gnornaô are sisters within the domain of thephrase. Nodes 4 and 5 are likewise sisters; 4, the leftmost, is labelled S. The nodeover the on-verse is itself labelled 5, since it is sister to that over the off-versewithin the domain of the line.The following rules then determine the location of alliterating syllables:(7.11) a. The head of each verse must contain an alliterating syllable.b. A weak constituent of a weak constituent may not contain analliterating syllable.The head of the verse is the strongest or only constituent in it. The strongestconstituent may be defined as what Liberman and Prince term the designatedterminal element (DTE). The DTE is that constituent which is dominated only byS-nodes below the level of the root, or unlabelled topmost node of the treediagram (259); an example is the word law in the triple compound given in (7.9),The head, or DTE, of the verse is therefore that word which is dominated by noW nodes below the level of the verse.According to Rule (7.1 la), the head of each verse must contain an alliteratingsyllable. In example (7.10) we see this is the case; the words greot and gcst - the113head of the S verse and that of the W verse respectively- do indeed containalliterating constituents.Note that the word f57seo is dominated by two W nodes: one within thedomain of the verse, one within the domain of the line. According to Rule(7.1 ib), a weak constituent of a weak constituent may not contain an alliteratingsyllable. Since the head of only one prosodic word in any given off-verse is neverdominated by two W nodes, Russom points out: “We do not need a special rulestating that the second half-line contains only one alliterating syllable” (OEM 75).The word gnornao, since it is dominated by only one W node, and thereforedoes not violate (7.1 ib), may contain an optional alliterating syllable. In example(7.10), this is the case. Here we can see that the reason the head of the secondprosodic word in a verse never alliterates unless the first one does is not due toarbitrariness or convention, but because (7.11) states that the head of thestrongest word in a verse must alliterate, whereas that of its weak sister alliteratesonly optionally.The rules for alliteration given in (7.11) apply in precisely the same mannerwhen a given verse is a more complex realization of the meter in that a metricalposition contains more than 0mm. For example, consider the following SieversD2-type, in which the S of the second foot contains a foot plus an adjoinedunstressed syllable (the word yrfe):5(7.12) dges yrfe-stOl ‘blessed home’ (1319a)SW SWThe head of ëadges, the strongest word in the verse, alliterates obligatorily by(7.11 a). The word stJl is dominated by a W node of a W node, and therefore maynot alliterate by (7.1 ib). The word yrtè, since it does not contravene the terms of114(7.1 ib), contains an optional alliterating syllable. When a D-type verse appearsin the off-verse, its second word is dominated by two W nodes and therefore maynot contain an alliterating syllable:(7.13) Jst ealle well ‘perform entirely well’ (1171b)S(W)S WWhen an on-verse with constraints relaxed on initial S appears, only the nodecontaining the head of a prosodic word is labelled, since the CAR, like theOECSR, does not apply to the heads of function words. Because an on-verse is bydefinition the leftmost verse when paired with its sister verse, its node is alwayslabelled S:(7.14) a. ac his if genm ‘but his wife took’ (846a)SW S WSb. ealra arymma jirym ‘majesty of all majesties’ (1 103a)SW S W/\,s2wc. et git mösten ‘that you always may’ (1371a)SWS WAAs A.wd. on gein-omld ‘in a world of turmoil’ (85 7a)SWS W115(7, 14a-b) are Sievers B-types, (c-d) are C-types. Note that in (b) and (d) an optionalalliterating syllable appears in the W position, giving these verses doublealliteration. Optional alliteration on W, while permissible, appears to be lessfavoured (at least by the poet of Guthiac B) than optional alliteration on S. Of 226A-type verses appearing in the on-verse, 145, or 64%, have double, alliteration -that is, an optional alliterating syllable on the second S position. Of 179combined B- and C-types appearing in the on-verse, only 31, or 17%, contain analliterating syllable on W.B- and C- type verses appearing in the off- (or weak) verse, may not have twoalliterating syllables, according to rule (7.1 lb):(7.15) a. ws se 1ohta g1m ‘the beam of light was’ (1289b)SW S Wwb. be m twëonum ‘between two seas’ (1359b)S(S WIn the above examples, only the first prosodic word of the verse may contain analliterating syllable, since the W word is dominated by a W node at the level ofthe verse.Sievers E-types occurring in the on-verse have two.locations in which anoptional alliterating syllable may appear. For example:S,s?w ?w(7.16) g-hengest rc ‘water-horse [ship] drove forth’ (1329a)S W S(116The word wg, which is dominated only by S-nodes, is the head of the verse andmust contain an alliterating syllable. Wrc, which is dominated by only one Wnode, may (and, in this example, does) contain an optional alliterating syllable.But the word hengest is likewise dominated by only one W node. Since this wordis not a W constituent of a W constituent, it may contain an optional alliteratingsyllable. Although Guthlac B has no examples of an alliterating syllable in thisposition, Russom (77) gives several examples from Beowulf, including:5s &w(7.17) yn-nedum weath ‘gobbled in great gulps’ (743 a)S W S(W)Like all verse-types appearing in the off-verse, E-types occurring in the off-verse may contain only one alliterating syllable by rule (7.1 ib), since only the firstprosodic word is never dominated by two W nodes:ws 2..w ?w(7.18) feorh-hord on1ac ‘unlocked the life-hoard’ (1 144b)S W S(W)Russom’s rules as adapted in (7.7) and (7.11) adequately capture almost everyinstance of alliteration in Guthlac B. The poem contains only four examples inwhich these rules fail to correctly predict the location of alliterating syllables.First, line 1234 contains no alliteration whatsoever. Since alliteration is anobligatory feature of Germanic verse, we may assume that this failure ofalliteration is a product of error in transmission, and no more need be said in thiscase.1171034a is a possible exception to rule (7.lla) if one scans the prefix Un- ashaving stress. The result is a Sievers A-type:2LW(7.19) unfrt laces ‘ready for battle’ (1034a)S WSWSince unket is an adjective, and since, as discussed in Chapter 3, adjectivalprefixes are stressed, scanning the verse as in (7.19) violates rule (7.1 la), since thehead of the verse, unket, does not contain an alliterating syllable, which isdefined as the head of a prosodic word by (7.5). A stressed prefix, recall, is thehead of the word; therefore alliteration should be on a vowel if 1034a is scannedas above. However, according to Kendall, stress on un- is variable; sometimes theprefix appears to be stressed, and sometimes not (48). If un- is not stressed, thisverse might then be scanned as a Sievers C-type:As w(7.20) (un)let lacesSWThis example is one of the unusual verses discussed in (5.5), an occasional one-foot verse in which an empty position would intervene between a prefix and itsstem if the prefix were not extrametrical. It is also one of only 6 C-type on-verseswith double alliteration in Guthlac B (out of a total of 89 C-type on-verses). Nomatter which way it is scanned, then, with a stressed or an unstressed prefix, thisparticular line is very unusual.Finally, two verses in Guthiac B show alliteration in a W constituent of a Wconstituent, violating Rule (7.1 lb). Both are D2-type verses:118S2s sw(7.21) a. rc in gewd ‘pain went inward’ (1028a)(W)S WSb. gst sThes georn ‘spirit eager for the journey’ (1045a)S(W)S WIn gewod and sipes geom form syntactic units and are therefore sisterconstituents (note that the word in in (7.21a) is an adverb, not a preposition, andtherefore, I shall assume, counts as a prosodic word; but see Hanson’s discussionof function words in Resolution 29-36). The W word in each phrase should notcontain an alliterating syllable by (7.1 ib).Verse-initial finite verbsFinite verbs form a class of words which sometimes appear to violate the rulesfor alliteration when they appear verse-initially. For example:w2L5 2w(7.22) a. fonde his rnon-dryhten ‘then he found his lord’ (1007b)S W S Wb. Wäst fl, rëo-dryhten ‘do you know, lord’ (1021b)swS wc. wät his inc-giefan ‘[he] knew his treasure-giver’ (1352b)SwS w119In the examples above, the alliterating constituents have been underlined. Notethat the strongest word in each example, as shown in (7.22a), appears to be aninitial finite verb the head of which does not, however, share in the alliterativepattern of the line, in violation of (7.1 la). Leaving aside the copula, auxiliaryverbs, and semi-auxiliaries such as wille, ‘want’ and ongon ‘began,’ which arecompleted by infinitives, there are 14 examples of non-alliterating fully lexicalverse-initial finite verbs in Guthiac B. All appear in Sievers B- and C-types.65It is difficult to say for certain why the head of a prosodic word, and only amember of this one class of prosodic words, sometimes stands as an exception tothe rules for alliteration. Many OE metrists assume that finite verbs which appearclause-initially are unemphatic and thus bear a lesser degree of stress than whenthey appear later in the clause (Russom OEM 101). There are a number ofproblems with this argument, not the least of which is that we do not knowenough about normal word order in OE to state this as an incontrovertible fact.66And even assuming, for the sake of argument, that clause-initial finite verbs areunemphatic, a problem remains in that just because a verb appears early in theverse does not necessarily mean it is likewise early in its clause. For example(alliterating constituents are underlined):(7.23) Him se dga wer ãgeaf2ndsware (1163)‘The blessed man gave him an answer’The verb Jgeaf, though verse-initial, actually appears rather late in its clause, andtherefore ought to alliterate, since it is not, by this argument, unemphatic (notethe initial syllable is an unstressed prefix). So even if some finite verbs appearing65They are: 920a, 1007b, 1021b, 1108b, 1147b, 1158b, 1163b, 1224b, 1293b,1294b, 1302b, 1327b, 1344a, and 1352b.66For discussion of verb-first clauses in OE prose, see Mitchell 969-78.120early in their clause are unemphatic and hence do not alliterate, this argumentdoes not account for all cases of non-alliterating verse-initial finite verbs.Spencer Cosmos, in an argument which looks to pragmatic function ratherthan to syntax, suggests that lexical finite verbs behave exactly as do otherstressed words such as nouns and adjectives; that is, they normally share in thealliterative pattern of the line, whether they appear early or late in the verse (311-12).67 However, some finite lexical verbs may sometimes bear a low degree ofstress, depending not on syntax but on their communicative function in thesentence (313). When a verb has “low communicative dynamism,” that is, whenit contributes little in the way of semantic information to the utterance, its stressmay be reduced and it is perhaps for this reason that it may be passed over by therules which determine the distribution of alliterating syllables (313-14).In determining whether a verb has low communicative dynamism, contextmust be taken into account, for the same verb may carry meaningful informationon one occurrence and not on another. I cannot at this time carry out a detailedanalysis of the pragmatic factors influencing the alliteration or lack of it on finiteverbs in Guthiac B; nor can I evaluate the relative merits of Cosmos’s argumentand the argument from syntax. But a cursory examination of the 14 examples ofnonalliterating verbs in Guthlac B do seem to support Cosmos’s point. Consider,for example, the following:(7.24) Wäst O £reO-dryhten,sWS whü ëos dle scyle nde gesettan? (1021b-1022)S WS W SWSW‘Do you know, lord, how this illness must come to an end?’67hi support of Cosmos’s argument, about 30 verse-initial finite verbs inGuthlac B share in the alliterative pattern of the line, as opposed to the 14 thatdo not. See, for example, 870a, 906a, 908a, 928b, 1015a, 1140a, 1265b, 1270b.121In the example quoted above, the devoted servant has found his master,Guthiac, stricken with a severe illness. He asks him if he knows how his illnesswill end - whether he will live or die. The intent of his question is apparently notto determine the state of Guthiac’s knowledge (WJst 111 hi7, ‘Do you know how’),but is rather to ascertain the consequences of his illness (Wãst 1ü Lfl, ‘Do youknow how?’). In this example, the verb wast seems to be unstressed because itcarries little meaningful information; it could even be left out and the essence ofthe utterance would remain (‘How will this illness end?’).I shall tentatively assume, then, that Cosmos’s explanation for non-alliteration in some finite verbs is correct. I shall also assume that the heads ofthese verbs are stressed by the rules of lexical phonology, and they are thereforeconstrained by the metrical rules of OE just as any other prosodic words are; atleast in Guthlac B, the heads of these verbs all arguably occupy S positions. Buttheir behaviour in regard to the rules for alliteration, which apply to prosodicand not to metrical structure, is anomalous; in particular, finite verbs aresometimes not treated as prosodic words by rule (7.5). In these cases they may,like function words, escape labelling:s w(7.25) Cwöm a frëorig-fero ‘then the sad-hearted one came’ (1344a)S WS WNote that if the finite verb in the example above were to be labelled S by (7.7), theconstituent fer3 would contain an alliterating syllable in violation of (7.1 ib):///S ?w(7.26) Cwöm frorig-feroS WS W122As noted above, I have so far assumed during this discussion that verse-initialfinite verbs are stressed by rules of lexical phonology. However, this assumptionis based only on the 14 examples in Guthiac B of nonalliterating initial finiteverbs; and since this is a very small sample, I shall not argue too strenuously forthis assumption. It is, in fact, quite possible that some finite verbs are not stressedby the rules of lexical phonology in OE, just as have and be are not stressed in PDEeven when they are functioning as main verbs (Hanson, Resolution 33). Were thisthe case, one might occasionally expect to see such verbs, like function words,which receive their stress post-lexically, appearing on W positions. This does notseem to be the case in Guthlac B; however, there is at least one possible examplein Beowuif (note alliteration is on n):(7.27) Dä corn non dges ‘then came the ninth hour’ (1600a)SW S WObviously more work needs to be done in this area, first, to ascertain whetheror not some finite verbs receive stress post-lexically rather than lexically, and ifso, under what circumstances. A further step would be to determine theconsequences for alliteration. For the moment, I shall leave the question open.123Chapter 8. Hypermetrical versesHypermetrical verses are long verses which occasionally appear in the poetrysingly, but more often are set apart in clusters. By my count, 19 such versesappear in Guthiac B: a pair at line 1110, five between 1160b and 1162b, threebetween 1294a and 1295a, and six between 1301a and 1303b. An additional 3isolated hypermetrical lines appear at 1 158b, 1225a, and 1060a.68Russom scans hypermetrical verses as having two feet: the first a normal foot,which typically has the pattern Sx or Sxx (in the on-verse) or xx (in the off-verse);the second a long or four-position foot, most often Sxsx (OEM 60). Note this longfoot overlaps the normal verse pattern Sx I Sx. Thus he scans a verse such as2996aas:(8.1) mon on middan-gearde ‘man on middle-earth’Sx I SxsxThis way of scanning hypermetrical verses, with a normal first foot and a longsecond foot, allows Russom to preserve his generalization that all OE versesconsist of two feet. However, he is then forced to modify his overlap constraint(given in (2.29) and (6.8)) in a rather forced and artificial way, since he must takeaccount of the fact that the second foot of a hypermetrical verse (and only thesecond foot of a hypermetrical verse) must in fact overlap a normal verse pattern(60):(8.2) The second foot of a hypermetrical verse overlaps a normal versepattern with an S position in the first foot.This rule is rather odd from a theoretical viewpoint. Why should a word likemiddan-gearde, which must otherwise occupy two Sx feet, be allowed to occupy68For the status of 1060a as hypermetrical, see Roberts (“Metrical” 100).124one foot just in case it is preceded by another foot? And why are there nohypermetrical verses with patterns like:(8.3) a. *Sx I xSsxb. *SxlxSxsin which the second foot corresponds to Sievers B and C, respectively? Whymust the first element of the second foot be S rather than x?It seems to me much more logical to regard hypermetrical verses as havingthree feet69, thereby avoiding an apparently ad-hoc modification to the overlapconstraint, even at the expense of being forced to add a special category to thestructure parameters as given in (4.2c):(8.4) Structure parameter settings for hypermetrical lines:a. Each line contains six feet.b. Each colon contains three feet.The abstract metrical pattern for a hypermetrical verse may be schematized:(8.5) SW SW SW- — -.--I shall not at this point, since Guthlac B contains so few hypermetrical verses,make any general claims about their higher-order metrical structure.Hypermetrical verses seem to have structure parameter settings (with theexception of (4.2c)) and realization parameter settings as laid out in Chapter 4.That is, they behave exactly like normal verses, save that they contain threerather than two feet. For example (constituents in S positions are underlined):(8.6) a. dig jnes gemyndig ‘blessed [man] mindful of courage’ (1294a)SWS W S Wb. glad-mOd to geofona inum ‘[he looked] gladly to the rewards ofS W S W S W gifts’ (1303a)69See also Hieatt, “Alliterative” and “New Theory.”125Hypermetrical verses appearing in the off-verse generally have prominenceconstraints relaxed on initial S. Therefore this position may contain anunstressed syllable or syllables:(8.7) a. Jjj he his Isna truwade ‘how he trusted his way’ (1 161b)S W SWS Wb. r on hine dëaô onsgde ‘before death prostrated him’ (11 62b)S W S WS WSince prominence constraints, as discussed re (5.1), may be relaxed only on verse-initial S, we do not need Russom’s rule that S of the second foot must contain astressed syllable. The fact that it always does so follows naturally from rule(4.19), that a S position must contain the head of a prosodic word.Russom bases his analysis of hypermetrical verses as having a second four-position foot Sxsx on the fact that in Beowuif this foot is often filled by a singlecompound word, such as middan-gearde. While this is not at all true of Guthlacwhich contains only one example of a four-syllable compound in ahypermetrical verse (1294b), it does seem that the final two feet, rather than thefirst two feet, tend to pattern together syntactically, for example, often formingan adjective+noun or verb+object phrase7O, This being the case, rules foralliteration as given in (7.7) and (7.11) apply:/N2s2w Xs 2Lw Xs(8.8) ô-mOd elan gyfle swylce h his gan ontyndeS W SW SW S W SW SW‘humble because of the noble food, likewise he opened his eyes’ (1301)70A few examples of the former are 11 lOa, 1 158b, I 162a, and 1225a; examplesof the latter include lllOb, 1161b, 1301b, 1303b. Other verses (for example,1294a, 1302b) contain genitive+adjective phrases, or verb+instrumental (1160b).126Alliterating syllables fall on the strongest words of each verse (aO - and êagan)by (7.1 la). 1Eelan, which is dominated by only one W node, contains anoptional alliterating syllable. Gyfle is dominated by two W nodes and may notcontain an alliterating syllable by (7.1 ib); likewise -tynde in the off-verse.The above discussion of hypermetrical verses is, I am aware, rather cursory andimpressionistic in nature. But I believe it serves to bear out the claims of ourmetrical system as ft relates to normal verses. Were hypermetrical verses tofollow significantly different metrical and alliterative rules from normal verses,this would tend to shed doubt on the theory as a whole. The fact that the samerules seem to hold for both types of verse lends, I think, some support to thetheory. Of course more research is needed before such a statement can be madewith complete confidence.127ConclusionThe traditional approach to OE meter taken by Sievers and Pope, amongothers, has been to describe the various rhythmic patterns present in OE verses inthe form of a list. This approach, like the traditional theory of substitutions iniambic pentameter, is open to criticism on several points. First, as Halle andKeyser have pointed out, a list of metrical types is unconstrained in that there isno reason why other members may not be added to the list (“Iambic”). Second,such a theory includes no constraints on substitutions; any type may always besubstituted for any other. For these reasons a description in the form of a listcannot rule out unmetrical lines. Halle and Keyser therefore propose a generativemetrical theory, in which meter is defined as consisting of two parts: anunderlying, abstract meter consisting of weak and strong metrical positions, anda set of correspondence rules constraining ways in which this underlying patternmay be instantiated in the surface- or language-level of the poetic line. Matchingof metrical positions to phonological constituents such as syllables, according tothe rules for the meter, generates a metrical line.Sievers’s “Five Types” theory of OE meter, as we have seen, suffers fromdrawbacks stemming from its failure to encompass general principles relating tothe phonological structure of OE. First, the theory is presented in the form of alist, with no criteria stated, whether on phonological or other grounds, as to whyother members may or may not be added to his lists of either foot or verse types;nor are there any generalized constraints on the pairing of feet into verses.Secondly, although metrical positions may be matched with various prosodicconstituents, such as stressed and unstressed syllables, matching rules fail togeneralize across foot and verse types, so that, for instance, a rise in the first footof an A-type may contain an unstressed syllable, though this is not possible in thesecond foot, or in the first feet of other types.128Pope’s theory, which is derived from musical theory, is based on the principlethat each verse contains two isochronous measures, each in turn containing fourquarter-notes or their equivalents, such as two half-notes. As in music, the firstnote of each measure receives a major stress. Therefore Pope shifts the footboundaries of Sievers’s B- and C-types in order to allow the second measure ofeach to begin with a stress, thereby eliminating Sievers’s rising or iambic feet.Although the establishment of consistently left-headed foot patterns is atheoretical improvement over Sievers’s mix of iambic and trochaic feet, Pope’sassumption of isochrony creates its own difficulties. The most serious of theseobjections is that verse-initial rests in some B- and C-types, which Pope proposesso that first feet may take the same amount of time to pronounce as second feet,must be filled in during performance by a harp-stroke in order to regulate thebeat. Kiparsky, however, has argued that since we can imagine a great variety ofpoetic rules which never occur, and since occurring poetic rules make referenceto grammatical constituents such as syllables, rules of poetry must beconstrained by the same rules that constrain language (“Role”); and therefore wemust reject the harp as an organizing principle of OE meter.Several of Pope’s ideas, however, are not incompatible with generativemetrical theory. Like Russom, I have followed Pope in assuming that OE feet areheaded by a prominent position. Unlike Russom, I have also adopted Pope’sprinciple that a metrical position may be empty, although I have applied the ruledifferently: in my theory, only weak positions may be empty, and emptypositions may occur in either foot, I explain the initial unstressed syllables inSievers B- and C- types by a general poetic principle that allows prominenceconstraints to be relaxed on initial metrical positions.Russom’s theory is an improvement over both Sievers’s and Pope’s in that hereplaces Sievers’s list of verse types with a list based on phonological principles:129each verse is comprised of two feet, and each foot is derived from the stresspattern of exactly one OE word (OEM). Therefore his list, unlike Sievers’s, ismotivated by linguistic features of the OE language, and is constrained by thesefeatures in that metrical types which do not correspond to two OE words will notappear. But Russom’s definition of the OE word, which is crucial to his theory,leads him into a number of inconsistencies. First, he is forced to defineunstressed prefixes as words in order to explain why they may never appeartogether with their stem as a foot. But he also must define both compounds andphrases as words in order to explain why two words may sometimes occupy asingle foot. In the case of phrases, Russom allows syntactic constituency todetermine the placement of foot boundaries, so that two words which form asyntactic unit are defined as a unit that mimics the structure of a word, and maytherefore occupy a foot; but syntactic rules are applied inconsistently in that theymay not be invoked to allow a prefix and its stem to occupy the same foot.In addition to these descriptive problems, Russom’s word/footcorrespondence gives rise to theoretical inconsistencies in that his modelincorporates metrical positions corresponding to the three levels of stress in OEwords (S, s, and x), whereas generative metrical theory is predicated on a binarydistinction between metrical positions (S and W). Finally, Rüssom proposes notone metrical pattern, but a list of 25 metrical subtypes. Generative metricaltheory, as we have seen, however, proposes a uniform underlying meter, withdifferences between the surface rhythm and the underlying meter beingaccounted for by rules matching prosodic units to metrical positions.Since both internal and external inconsistencies in Russom’s theory stem fromthe fact that his metrical patterns are derived from words, I have bypassed his“word stress” level in order to make direct reference to the phonological ruleswhich assign prominence in language. I have suggested, following Kiparsky, and130Hanson and Kiparsky, that both the stress patterns of OE words and constraintson OE meter are governed by the same linguistic principles.The metrical model which I have proposed in these pages is based on Hansonand Kiparsky’s parametric theory of universal meter. Hanson and Kiparsky claimthat all poetic meters are comprised of binary feet, which, like the phonologicalconstituents defining prominence in language, contain a strong (S) memberwhich is the head, or prominent position, and a weak (W) member which is anunprominent position. A position parameter defines the maximal amount ofprosodic material that may occupy a given position in terms of phonologicalconstituency: mora (Es), syllable (a), foot (0), or word (k). Prominence rules definefirst, whether S positions must contain prominent constituents and/or whetherW positions must contain unprominent constituents; and second, whetherprominence is defined by weight, strength, or stress.The model I have proposed for OE therefore defines the meter in terms of afixed number of binary left-headed feet together with constraints on both S andW positions: S positions must contain stressed syllables, further defined as theheads of prosodic words; and W may contain the heads of prosodic words only ifthey are prosodically weak. No metrical position may contain more than aminimal word (?min).The position parameter setting of min captures Russom’s generalization thatOE meter is word-based, as opposed to syllable-based, which has been theassumption behind most theories of OE meter, including Sievers’s and Pope’s.But having metrical rules making direct reference to phonological structurerather than to an intervening “word stress” level avoids the internalinconsistencies of Russom’s model. The work that, in his theory, is done bydefinitions of what may constitute a word (unstressed prefix, simple word,compound, or phrase - but not an unstressed prefix plus its stem) together with131bracketing rules which apply syntactic criteria inconsistently to some structuresbut not others, is done in my theory by matching phonological constituents witha single, uniform, underlying meter.As well as eliminating Russom’s internal inconsistencies, my model avoids thetheoretical inconsistencies of Russom’s system. Russom’s model, both because itincorporates a three-way, rather than a binary, prominence distinction in theunderlying meter, and because it is presented in the form of a list of metricalsubtypes rather than as a single meter, is incompatible with generative metricaltheory. Again, these inconsistencies arise from the fact that his metrical types arederived from the stress patterns of OE words. Because OE words have at leastthree levels of stress - primary, secondary, and unstress - any metrical patternsthat are derived from them must likewise incorporate three levels of prominence.Because OE words have a number of rhythmic patterns, any metrical patternswhich they generate will likewise be varied. Because the basis of OE meter is theOE word, Russom’s theory is necessarily language-specific and cannot begeneralized to other meters (for example, it is not the case that iambic metersallow only words with iambic stress patterns to appear in the poetry).Elimination of Russom’s “word stress” level in order to make direct reference tothe phonological constituents of language makes it possible to integrate OEmetrical theory with universal metrical theory.The assumption that OE meter is based on left-headed binary feet which arematched with phonological constituents gives us a new tool with which toexplore certain ideas about the language that have so far been rather opaque toanalysis. First of all, as discussed in Chapter 3, function words are not stressed byrules of the lexical phonology and are assumed therefore to behave differently inmeter than lexical words (Hanson, Resolution). While verse-initial or -medialmetrical positions in OE may be occupied by a string of function words, this is132not true of verse-final positions, which may contain only one function word. Inboth Sievers’s and Russom’s systems, such verse-final function words occupystrong positions (I, 5, or s); yet Sievers gives no reasons, and Russom givesinadequate reasons to explain why these words are treated differently in themeter than other function words are. I have assumed that Zec and Inkelas’sPhrase-Final Stress rule builds a foot over phrase-final function words, whichallows them to be treated in the meter as though they were prosodic words.Explaining the behaviour of function words as a product of their phonologicalstructure allows us to make predictions about their behaviour in meter, whichRussom’s explanation, that function words acquire stress when they are removedfrom their normal proclitic position, does not; namely, because not all phrase-final function words are in fact removed from a proclitic position, and Russomproposes no other mechanism by which some function words may acquire stress.Secondly, the assumption that an underlying binary meter is matched withphonological constituents allows us to test hypotheses regarding OE phrasalstress, which has been a matter of some debate. Phrasal stress in OE is generallyassumed to be trochaic by theorists such as Kurylowicz and Maling, who assumethat alliteration depends on prosodic stress. On the other hand, those whoassume that alliteration is based on the metrical level, such as Russom, argue thatphrasal stress in OE, like that in PDE, is governed by the Nuclear Stress Rule,which places greater stress on the final lexical element. I have demonstrated,however, first, that phrases never pattern in the meter like clitic groups do, withtheir first word on a W position; in fact, the first word of a phrase must occupy a Sposition. Secondly, while the second word of a phrase may occupy a W position,this is not true of a sequence of prosodic words which is not a phrase. Thisevidence leads me to conclude, independent of evidence from alliteration, that133stress in OE phrases is trochaic and that W positions may contain onlyprosodically weak constituents.If OE has trochaic stress in phrases, it is unnecessary to assume, as Russomdoes, that rules for alliteration make reference to the metrical level. In fact, as wehave seen, this assumption creates further inconsistencies within Russom’stheory, since he is then forced to assume that metrical labelling is notunidirectional at all levels of the metrical hierarchy, but may invert from SW toWS just in case the first foot of a verse contains no lexical stresses. Such a labellinginversion contradicts universal theory, which assumes labelling in the samedirection at all levels of the metrical hierarchy. However, with the assumption oftrochaic stress in phrases, this theoretical inconsistency is avoided. I havesuggested, following Russom (though applying the rule to the prosodic ratherthan to the metrical level), that a rule similar to the OE Compound Stress Rule,which takes the OE line as its domain, determines the location of alliteratingsyllables. This rule accounts for alliteration in all of the lines of Guthiac B, withonly a few exceptions.The rule that W positions may contain only prosodically weak constituentsalso drives the placement of foot boundaries, which in Russom’s system is, as wehave seen, a function of his (inconsistent) definition of the OE word, togetherwith a rule that appeals to syntactic structure on an inconsistent basis, that is,just in case a verse comprises three fully stressed words. This rule is additionallya theoretical improvement over Russom’s bracketing rules in that, as Hayes hasproposed (“Prosodic”), and as is assumed by universal theory, metrical rules donot make reference to syntactic constituency, but to phonological constituency.Finally, the rule that W positions may contain only prosodically weakinaterial provides a tool with which to compare the metrical behaviour oflexicalized compounds (such as hla7ord) with words with stressed prefixes (such134as onbid). As we have seen, words with stressed prefixes, which contain a SW pairof prosodic feet, may occupy S positions, in contrast to compounds, which mustoccupy at least two positions. Words with stressed prefixes may not, however, incontrast to simple words, occupy W positions, since they contain a strong foot.With this three-way distinction in mind, we may examine lexicalizedcompounds in order to observe their metrical behaviour and perhaps form someconclusions as to their degree of lexicalization; do they act more like compoundwords? Simple words? Or somewhere in between, like words with stressedprefixes?As we have seen, although the generative theory which I have proposed inthese pages accounts for all of the verses of Guthlac B, it also overgenerates,producing verses which are not found in the canon of OE poetry. I havediscussed this in terms of Youmans’s Aristotelian and Platonic aspects of meter,arguing that while generative theory accounts for the Aristotelian, or rule-basedaspects, a Platonic component, which takes overall complexity into account, alsoplays a large part in OE poetry, and needs to be further explored. Russom’soverlap rule, a functional constraint which allows a listener to di.sambiguate feetfrom verses, incorporates itself well into the theory I am proposing, and serves torule out a number of nonexistent verse types. However, questions still remain inthat some complex verse types appear quite often in the poetry, while othersappear rarely if at all. Determining why this is so is an area for further research.It is obvious that much more work needs to be done before OE meter is fullyunderstood (if it can be fully understood). I offer this paper in hopes that it willcontribute in some way to that understanding, and in respect and admiration forthe Anglo-Saxon poets, the grace and sophistication of their poetry, and themeter in which they chose to express it.135Works CitedBessinger, Jess B. and Stanley J. Kahn, eds. Essential Articles for the Study of OldEnglish Poetry. Hamden: Archon, 1968.Bethel, Patricia. “Anacrusis in the Psalms of the Paris Psalter.” NeuphilologischeMitteilungen 89 (1988): 33-43Bliss, Alan J. The Metre of Beowuif. Rev. ed. Oxford: Blackwell, 1967.Bradley, S.A.J., ed. Anglo Saxon Poetry. London: Dent, 1982.Cable, Thomas. The English Alliterative Tradition. Philadelphia: U ofPhiladelphia P, 1991.The Meter and Melody of Beowuif. Urbana: U of Illinois P, 1974.Cosmos, Spencer. “Kuhn’s Law and the Unstressed Verbs in Beowulf.” TexasStudies in Literature and Language 18 (1976): 306-28.Creed, Robert P.”A New Approach to the Rhythm of Beowulf.” PMLA 81 (1966):23-33.Reconstructing the Rhythm of Beowuif. Columbia: U of Missouri P, 1990.Crystal, David. A Dictionary of Linguistics and Phonetics. 3rd. ed. Oxford:Balckwell, 1991.Daunt, Marjorie. “Old English Verse and English Speech-Rhythm.” Bessinger289-304.Dresher, B. Elan and Aditi Lahiri. “The Germanic Foot: Metrical Coherence inOld English.” Linguistic Inquiry 22 (1991): 251-86.Duncan, Edwin. Stress, Meter, and Alliteration in Old English Poetry. Diss. U ofTexas at Austin, 1985.Fish, Stanley. Is there a Text in this Class?: The Authority of InterpretiveCommunities. Cambridge: Harvard UP, 1980.Freeman, Donald C., ed. Essays in Modern Stvlistics. London: Methuen, 1981.136Fulk, R. D. A History of Old English Meter. Philadelphia: U of Philadelphia P,1992.Fussell, Paul. Poetic Meter and Poetic Form. Rev. ed. New York: Random, 1979.Greg, W. W. “The ‘Five Types’ in Anglo-Saxon Verse.” Modern Language Review20(1925):12-17.Halle, Morris and Samuel J. Keyser. English Stress: Its Form, its Growth, and itsRole in Verse. New York: Harper, 1971.Halle, Morris and Samuel J. Keyser. “The Iambic Pentameter.” Versification:Major Language Types. Ed. W.K. Wimsatt. New York: MLA, 1972. Rpt. inEssays in Modern Stylistics. Ed. Donald C. Freeman. London: Methuen, 1981.21 7-37.Hanson, Kristin. “Prosodic Constituents in Poetic Meter.” Proceedings of the13th Annual Meeting of the West Coast Conference on Formal Linguistics.Stanford: CSLI Publications, 1994 (forthcoming).“Resolution: Evidence from Modern English Meters.” Proceedings of theNorth East Linguistics Society 23. Ed. A Schafer. GSLA, U of Massachussets atAmherst, 1993. 159-73.Resolution in Modern Meters. Diss. Stanford U, 1992.Hanson, Kristin, and Paul Kiparsky. “The Best of All Possible Verse.” ms. January24, 1994.Hayes, Bruce. Metrical Stress Theory: Principles and Case Studies. UCLA, ms.January 1991.“The Prosodic Hierarchy in Meter.” Kiparsky and Youmans 201-60.“A Revised Parametric Metrical Theory.” Proceedings of the North EastLinguistics Society 17. Ed. Joyce McDonough and Bernadette Plunkett. Vol. 1.GSLA, U of Massachussets at Amherst, 1987.137Hieatt, Constance B. “Alliterative Patterns in the Hypermetric Lines of OldEnglish Verse.” Modern Philology 71(1973): 237-42.“A New Theory of Triple Rhythm in the Hypermetric Lines of Old EnglishVerse.” Modern Philology 67 (1969-70): 1-8.Hogg, Richard, and C. B. McCully. Metrical Phonology: A Coursebook.Cambridge: Cambridge UP, 1987.Hoover, David L. A New Theory of Old English Meter. New York: Lang, 1985.Huettner, Alison K. “A New Look at Old English Metrics.” Kansas WorkingPapers in Linguistics 14 (1989): 20-5 6.Hutcheson, B.R. “Quantity in Old English Poetry: A Classical Comparison.” inGeardagum 12 (1991): 44-53.Jakobson, Roman. “Linguistics and Poetics.” Style in Language. Ed. ThomasSebok. Cambridge: MIT, 1960. 350-77.Jespersen, Otto. “Notes on Metre.” Linguistica. Copengagen: Levin andMunksgaard, 1933. 71-90.Kendall, Calvin B. “The Prefix ‘un’ and the Metrical Grammar of Beowuif.”Anglo Saxon England 10 (1982): 39-52.Keyser, S. J. “Old English Prosody.” College English 30 (1969): 33 1-56.Kiparsky, Paul “On Theory and Interpretation.” The Linguistics of Writing:Arguments Between Language and Literature. Ed. Nigel Fabb et al. New York:Methuen, 1987.“The Rhythmic Structure of English Verse.” Linguistic Inquiry 8 (1977): 189-247.“The Role of Linguistics in a Theory of Poetry.” Daedalus 102 (1973): 23 1-244.Rpt. in Freeman 9-23.“Sprung Rhythm.” Kiparsky and Youmans 305-40.138“Stress, Syntax, and Meter.” Language 51(1975): 576-616. Rpt. in Freeman225-72.Kiparsky, Paul, and Gilbert Youmans, eds. Rhythm and Meter. Vol. 1 ofPhonetics and Phonology. San Diego: Academic Press, 1989.Klaeber, Fr., ed. Beowulf and the Fight at Finnsburg. 3rd. ed. Lexington: Heath,1950.Kurylowicz, Jerzy. “Linguistic Fundamentals of the Meter of Beowuif.”Linguistics and Literature, Sociolinguistics and Applied Linguistics. Vol. 4 ofLinguistic and Literary Studies in Honor of Archibald A. Hill. Ed. MohammadA. Jazayery, Edgar C. Polomé, and Werner Winter. The Hague: Mouton, 1979.111-19.Lass, ROger. “Quantity, Resolution, and Syllable Geometry.” Folia LinguisticaHistorica 4 (1983): 15 1-80.Liberman, Mark, and Alan Prince. “On Stress and Linguistic Rhythm.”Linguistic Inquiry 8 (1977) : 249-336.Lutz, Angelika. “The Syllabic Basis of Word Division in Old EnglishManuscripts.” English Studies 67 (1986): 193-210.Maling, Joan M. “Sentence Stress in Old English.” Linguistic Inquiry 2 (1971):3 79-99.McCully, C. B., and R. M. Hogg. “An Account of Old English Stress.” Journal ofLinguistics 26 (1990): 315-39.Mitchell, Bruce. 1985. Old English Syntax. Vol. 2. Oxford: Clarendon, 1985.Moulton, William G. “Secondary Stress in Germanic Alliterative Verse.” Studiesin Descriptive and Historical Linguistics: Festschrift for Winfred P. Lehman.Ed. Paul J. Hopper. Amsterdam: Benjamins, 1977. 393-404.O’Keeffe, Katherine O’Brien. Visible Song: Transitional Literacy in Old EnglishVerse. Cambridge: Cambridge UP, 1990.139Pope, John Collins. The Rhythm of Beowulf. Rev. ed. New Haven: Yale UP, 1966.Prince, Alan. “Metrical Forms.” Kiparsky and Youmans, 45-80.Roberts, Jane, ed. The Guthiac Poems of the Exeter Book. Oxford: Clarendon,1979.“A Metrical Examination of the Poems Guthlac A and Guthlac B.”Proceedings of the Royal Irish Academy 71 Sec. C (1971) : 1-137.Routh, James. “Anglo-Saxon Meter.” Modern Philology 21 (1924): 429-34.Russom, Geoffrey. Old English Meter and Linguistic Theory. Cambridge:Cambridge UP, 1987.“A New Kind of Metrical Evidence in Old English Poetry.” Papers from theFifth International Conference on English Historical Linguistics. Ed. SylviaAdamson et al. Current Issues in Linguistic Theory 65. Amsterdam:Benjamins, 1990. 435-57.“Word and Foot in Beowuif.” Style 21(1987): 387-99.Shakespeare, William. The Complete Works of William Shakespeare. London:Spring, 1958.Sievers, Eduard. “Old Germanic Metrics and Old English Metrics.” Trans.Gawaina D. Luster. Bessinger 267-88.Silver-Beck, Barbara L. “The Case Against The Rhythm of Beowuif.”Neuphilologische Mitteilungen 77 (1976): 510-25.Stephenson, Edward A. “Hopkins’ ‘Sprung Rhythm’ and the Rhythm ofBeowuif.” Victorian Poetry 19 (1981): 97-116.Suphi, Menekse. “Old English Stress Assignment.” Lingua 75 (1988): 171-202.Taglicht, Josef. “Beowuif and Old English Verse Rhythm.” Review of EnglishStudies 12 (1961): 341-51.Youmans, Gilbert. “Milton’s Meter.” Kiparsky and Youmans 341-79.140Zec, Draga, and Sharon Inkelas. “Phonological Phrasing and the Reduction ofFunction Words.” ms. Stanford U, Jan. 10 1988.141Appendix 1To summarize the discussion of Chapters 4 and 5, the following rules capturethe essential facts about OE meter:I. Structure parameter settings(4.2a) Feet are left-headed (SW).a. Normal lines(4.2b) Each line contains four feet.(4.2c) Each colon contains two feet.b. Hvpermetrical lines(8.4a) Each line contains six feet.(8.4b) Each colon contains three feet.II. Realization parameter settingsa. Position(4.3) The position parameter is set at the minimal word (min).b. Prominence rules(4.19) A S position must contain the head of a prosodic word.(4.42) A W position may contain the head of a prosodic word only ifit is prosodically weak.III. Overlap constraint(6.10) Maximize the distinction between foot patterns and versepatterns.IV. Alliteration(7.7) Within the domain of the line, for any pair of sister nodesdominating prosodic words, the leftmost is S.(7.1 la) The head of each verse must contain an alliterating syllable.142(7.1 ib) A weak constituent of a weak constituent may not contain analliterating syllable.V. Special licenses(1) Prominence constraints may be relaxed on an initial position.(2) A W position may be empty.(3) Extrametrical unstressed syllables may appear following asyntactic break and preceding a S position.143Appendix 2: Sample scansionThe following is a sample scansion, according to the rules laid out in Chapters3-7, of the first 105 verses of Guthiac B. Constituents occupying S positions areunderlined. The Sievers type of each verse appears in the right-hand column.Asterisks indicate a type which is “expanded” by additional unstressed materialin the first dip; the plus sign indicates that the verse is prefixed by a syllable inanacruisis.Note that ambiguous scansions are not uncommon in the generative modelproposed here (see note 49). A word like lofesta, for example may be scannedwith either its first syllable (0mm) or its first two syllables (2min) in S:den festa or den lëofesta ‘beloved lord’SW SW SW SWI shall therefore adopt the convention, in keeping with Russom’s claim thatOE meter is word-based, of respecting word boundaries (and thereby preferringthe second of the two possible scansions above) to the extent that this does notresult in unnecessarily creating empty W positions. I would not, for instance,scan eThden in the example above as occupying S while leaving the following Wempty, even though it is a possible scansion, since it is always more normative fora metrical position to contain some amount of linguistic material.Sievers types 2w 2s819 DIET IS WIDE CUD WEra cnëorissum B; DlSWS W S(W) S W144S w/ /\2w w820 folcum gerge, te fryma God A*; BS WSW sw_s W/ Aw821 ne restan e1da ynes C; ASW SW SWSWAs As As Aw822 çf Iere c1nestan, cyning i-mihtig, C; DlS W S W S (W)S WS wAs Aw As Aw823 foldan geworhte. wes fruma nTwe A*; CS WSW SW S WS wAs Aw As Aw824 elda tüdres, st1 wynlic, A; ASwS w S W S WS wAs Aw As Aw825 fteger 7 gef1ic. Fder ws ãçned A*; A*S WSW S WSW145S w- 2s 2w 2s w826 Adam rest urh ëst Godes A; CSWSW S(W)S W/\2,s w 2w827 nëorxna-wong j him nnges ws B; BS(W) S W SW S WS w• /2s w 2s828 i1an s?n welan brosnung A; CS WSW S(W) S WS wAs2w 2s 2w829 ifs lyre lices hryre B; BS(W) S W S(W) S WS w/2s 2w As w830 drames dryre daôes cyme B; BS(W) S W S(W) S WAAs As w831 ac h on mide iI[gan ste A3; AS W SW SW SW146S wA A.2s w s w832 ealra leahtra 1as, iQflge tan B; ASW S W SWSWS wAs 2w s ?w833 nrwra gena; he orfte A*; CS WSW SWS WA As w834 lifes ne lissa in m 1ohtan ham A*; BSWSW SW. S WS wA2s 2w 2s 2w835 urh e1da tid ende geidan B; A*S (W) S W S WS W836 ,ç efter £ste to m frestan A3; CSW SW SW SW2s 2w 2w ?s 2w837 heofon-rices gefan hweorfan stan E; AS W S (‘N) S W S W147S/s w As ?w838 leomu ç somud Z gst Dl; BS (W)S W S(W)S WS wAAs As839 7 er siH,an in sin-drëamum B; CSW S W S(W)S WS wAs Aw As840 to widan fëore wunian mOstun A+; ASW SW S WS WS w841 dryhtne on gejh5e Jtan dëaôe forô A*; BS W.SW SW• S WS wAs w As842 g hy halges word healdan idun, B; ASW S W SW SWS wAs ?w As w843 beorht in brëostum, 7 his bebodu 1stan, A; CS WS W S WS W148S wAS AS 2w844 efnan on 1e; i to r arat A*; BS WSW SWS WAAS AS 2w845 hy waldendes ji1an 1sten; C; ASW S W SW SWS wAS 2w AS w846 his wif genOm ymes irum B; ASW SW SWSWS wAS AS Aw847 blëde forbodene 7 of bëame ahnop A*; BSWSW SW S WAAS AS Aw848 wstm biweredne ofer word Godes, A; CSW SW SW S WS wAS Aw AS ?w849 fldor-cyninges, 7 hyre were sealde A; CSW SW SW S W149wA?s 2w s 2w 2w850 urh dofles searo daô-berende yfl B; ES(W) S W S W S(W)S ww851 oa -hTwan swvlte getah. C; BSWS W 5(W) S WwS2s 2w w852 Si1an se e1 genge wearô A3; ES WSW S W S(W)ww 2s2w?w853 Adame 7 uan, eard-wica cyst, A*; ES WSW S W S(W)S w/NAs As 2w854 beorht, oden, Z hyra bearnum swã, A; BS WSW SW S W/\As Aw855 eaferum fter, , hy on uncVoôu, A; CS WSW SW SW150S wAs w856 scomum sctidende, scofene wurdon Dl; AS (W) S W SW SWS w/As As 2w857 n geju-woru1d; weorces ongjjjdon C; A*SWS W S WSWS wAs )w As Aw858 dëopra fitna dëaôes cwealm A; BS WSW S(W) S WAs As859 hy unsnvttrurn gefremedon; C; ASW S W SWS WS wAs Aw As860 jre yfl-wrce jjcjan sceoldon C; ASWS W SW SWS wAs As861 mego 7 mcgas morjres onyjdon, A; A*S WSW S WSWS_ _AAs Aw Aw As Aw862 -scy1dge gym, ]urh gst-gdä1, E; BS W S(W) S(W) SW151w/Xs 2w s w863 dopra fiina. Daô in gerong A; D2S W SW S (W) S WS w2s 2w864 ftra yne, fgond rixade A; DlSW S W S (W)S WS wXs 2w As Aw865 geond middan-geard. 4nig monna ws B; BS(W) S W SW S WS w2s2w As 2w866 of äm ig-tudre jian fre C; ASW SW SWSWSAs Aw As Aw867 Godes willan es georn gvnn-wtsed E; CS W S (W) S(W) S WS w/\As Aw As868 jh bibtigen mge ne bitran drync B; BSW S W SW S W152S wXs 2w- As869 jne fyrn Athme geaf, B; ESWS W SW S(W)w//As As A,w As870 byrelade brVd geong: jj him bm gescöd A*; BSW S W SW SW871 n m doran ham. BSW S W


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items