Old English Newsletter


Back  |  Print


What are Old English Metrical Studies For?


Thomas A Bredehoft, University of Northern Colorado

I never expected to study Old English meter. Like most American Anglo-Saxonists whose training came in an English department, I originally saw my work as being concerned primarily with interpretive literary questions (with a subordinate focus on manuscript issues thrown in, mostly to give me an excuse to visit libraries in Britain). But, as things turned out, I kept finding that there were certain questions—literary, interpretive questions—that I simply could not answer for myself without spending some time seriously considering issues of meter. To give one familiar example, I wanted to know if the emendation of Beowulf 53b to "Bēow Scyldinga" was metrically "necessary" or not, as it has often been described. [1] In the context of my own work, I wanted to know if all of the passages printed as verse in Plummer's edition of the Anglo-Saxon Chronicle were truly verse or not. [2] Another question that might best be approached by metrical study is the relationship between the Ruthwell Cross verses and the Vercelli Book Dream of the Rood. Many similar questions of literary history, I think, demand at least some attention to developments in metrical form across the period. And while I don't necessarily think that everyone should spend as much time studying Old English meter as I have, it seems worthwhile at least to try to articulate the value of metrical studies.

The most important thing to say about Old English metrical studies is that virtually everything we know about Old English meter is a conclusion drawn from an inductive process. That is, modern scholars have derived the rules of Old English meter by studying a body of evidence and making hypotheses about what "rules" might have generated that evidence. As is well known, Old English manuscripts almost never make meaningful visual distinctions between Old English poetry and prose, so we can only rarely depend upon Anglo-Saxon opinions about which genre a given text belonged to: most often we are forced to judge for ourselves, and virtually the only way to do so is through an understanding of the principles of Old English meter. But because our understanding of Old English meter results from an inductive analytical process, we must first identify examples of Old English verse in order to have a body of evidence from which to derive our inductive conclusions. Metrical studies, therefore, are troublingly hampered (or invigorated?) by a methodological circularity, one which ought to encourage us to accept received opinions with a grain of salt. Shifting the focus of the data (from Beowulf, say, to the Metrical Psalms) has the potential to alter radically the metrical rules (i.e., conclusions) that make up our descriptions of Old English verse. [3]

Of course, we can always refuse to judge for ourselves, which boils down to letting others judge for us: and that, of course, is the best reason for having at least a passing familiarity with some of the questions and methods of Old English metrical study. In the remainder of this essay, I will briefly discuss some of the central questions of metrical inquiry, as well as its methods and forms. If sometimes I allow my own metrical prejudices to shine through a bit too clearly, I hope readers will remember that I will be happiest if they judge for themselves, rather than relying only on my opinions.


What Old English meter is not

Old English metrical study takes the following issues as its primary arena of interest: which half-lines (verses) are allowed and which are disallowed, how half-lines are linked to one another, and how such half-lines correspond to linguistic structures (usually understood as patterns of stress). Discussions of meter generally do not concern themselves much with formulas, formulaic themes (e.g., the "beasts of battle"), or individual words. Thus, to the degree that modern interpretive strategies focus on the latter sorts of topics, meter might seem to be of little value to literary readers.

A current trend in metrical studies, however, has begun to explore the relationship between meter and syntax (see, e.g., recent works by, among others, Dan Donoghue, Mary Blockley, Hal Momma, and Calvin Kendall), and it seems to me that there is much of interest in such work. [4] To take just one example, Blockley's work investigates (among other things) how (or if) we can tell the difference between a statement and a question. If metrical studies can help clarify such issues, then they have obvious relevance to literary interpretation.

In my opinion, though, literary interpreters have too rarely asked some of the key questions which must arise about the relationship between Old English meter and poetics. Metrical linking strategies such as double alliteration, cross alliteration, and rhyme (both internal rhyme and verse rhyme) have been far too poorly understood as metrical options for poets, and the possibility that these sorts of metrical options might have been used for poetic effects has only just begun to be explored. [5] But in the sense that metrical structures stand as the basic tools of Old English poetics, some knowledge of meter would seem to be of at least potential value for literary interpreters—unless we understand the formulaic tradition and its conventional expressions as the masters of poets, rather than the other way around.


What does Old English meter measure?

Roman Jakobson quotes from Gerard Manley Hopkins in suggesting that verse involves "wholly or partially repeating the same figure of sound." [6] In the case of Old English, one aspect of such repetition surely lies in alliteration, but for a surprisingly long time, scholars had little if any inkling of what else made an Old English poem poetic. Early critics and editors were often able to identify lines of poetry when they published poetic works (thus giving later scholars a data set to work from), but unable to articulate the metrical basis that underlay those lines. [7] Towards the end of the nineteenth century, however, the German scholar Eduard Sievers finally offered a descriptive account of Old English verse that generated enough consensus to be adopted by the community of Anglo-Saxon scholars at large. [8] In brief, Sievers suggested that the vast diversity of Old English verse types could be understood as deriving from five basic types, each conceptualized as the juxtaposition of two metrical "feet," made up of abstract patterns of stress. Marking stress ("/"), half-stress ("\") and lack of stress ("x") [9] allows the Sievers types to be briefly summarized (I use "|" to indicate the boundary between metrical feet): [10]

Type A   /x | /x

Type B   x/ | x/

Type C   x/ | / x

Type D   / | / \ x

Type E   / \ x |

Such an analysis, of course, suggests that, along with alliteration, stress patterns are at the very heart of what is measured in Old English meter, and at least since the work of Sievers, most metricists have suspected that there is such an additional linguistic "measure" at work in Old English verse (although David Hoover argued in 1985 for the essential primacy of alliteration in Old English). [11] Given the diversity of specific verse types in Old English, J. C. Pope's idea that Old English meter measured out time itself in an "isochronous" verse style seemed intuitively useful: the poetry had a "rhythm" which, while variable, had a fairly regular "beat" that could be expressed by Pope's musical notation. [12] Perhaps because (in the absence of audio recordings) isochrony cannot be demonstrated, Pope's work has been more influential for his revisions to Sievers's views of types B, C, and D than for its commitment to isochrony. Thomas Cable's English Alliterative Tradition suggested that Old English is essentially a "syllable counting" meter, although one which clearly "counts" varying numbers of syllables from one verse to another; in this sense, Cable's system differs relatively little from Sievers's. [13] Rick Russom's word-foot formalism essentially identifies the "counted" or "measured" entities as including exactly two "word stress contours" that combine according to a handful of pretty simple rules: in Russom's system, a "typical verse" is two words juxtaposed, and atypical verses can be understood as allowed variations on this norm. [14] Metrical linguists (or are they linguistic metricists?) have suggested that various aspects of meter make a count of "morae," linguistic constituents at a level smaller than the syllable, but I am not enough of a linguistic metricist to be much more specific on this topic.

As this summary should make clear, there is still surprisingly little consensus about what is measured by Old English meter; but each of these preceding positions shapes the kinds of questions we ask about meter and the ways we answer them and, in that sense, all of these perspectives appear to be viable and contribute to our ever-evolving understanding.



All descriptive metrical systems for classical Old English verse must account for the same verses, of course, and so they have a great deal in common, although the labels they place on specific verses may differ from one another. Ultimately, though, scansion (the process of labeling specific verses according to one system or another) is nothing more than a detailed kind of mark-up system, and as such scansion lies at the heart of all metrical descriptions.

In brief, Old English meter (like most English meters) depends upon a general and widespread poetic principle of matching metrical stress to linguistic stress (think of iambic pentameter, in which the alternating metrical stresses are generally, but not exclusively, matched up with items of high and low linguistic stress). [15] Here, it is sentence-level stress that is most operative: open class, lexical words (nouns, adjectives, verbs and many adverbs) have high linguistic stress, while closed-class, non-lexical words, prefixes, and particles generally have low linguistic stress. [16] Our identification of basic metrical forms relies powerfully on this understanding, and only a couple of additional observations are needed to effectively scan (or mark up) OE verse. First, the last word of every verse is always stressed, regardless of its "natural" class, and the "unmarked" position for unstressed sentence elements (excluding proclitics, unstressed elements like prepositions, prefixes or articles that are attached to a particular word) is before the final stress of the first verse of the clause (when placed elsewhere, such elements usually become marked and stressed). Given these rules, we can scan a passage of verse (Beowulf 755-61) [17] very straightforwardly (again using "/" "\" and "x" for primary stress, secondary stress, and unstress): [18]

Hyge wæs him hinfūs,    wolde on heolster flēon    / x x x / \    / x x / x /

sēcan dēofla gedræg;    ne wæs his drohtoð þǣr    / x / x / x    x x / x /

swylce hē on ealderdagum    ǣr gemētte.    x x x x / x \ x    / x / x

Gemunde þā se gōda    mǣg Higelāces    x / x x x / x    / / x \ x

ǣfensprǣce    uplang āstōd    / x \ x    / \ x /

ond him fæste wiðfēng;    fingras burston;    x x / x x /    / x / x

eoten wæs ūtweard,    eorl furþur stōp.    / x x / \    / / x /

All that remains is to match up this preliminary mark-up with your favorite verse-naming system, and the scansion is complete—but that, of course, is far from as simple as it might seem.

The first problem with moving from a mark-up like that given above to a final scansion is that such a mark-up yields a vast number of verse-types: only two of the fourteen verses given here, for example, exhibit identical stress patterns (757b and 760b), and even those two place the word-boundaries in different positions (and hence these verses are sometimes identified as belonging to different types). A scansion system that involves hundreds of specific verse types is of very little value for talking about trends or patterns or similarities among verses; what metricists do at this point is to attempt to determine (usually by employing some kind of statistical analysis) what sorts of specific types can be regarded as minor variations of a much smaller number of basic "umbrella" types.

As noted above, the most familiar and widely taught metrical system, that of Eduard Sievers, limits the "umbrella types" to five: a great deal of the appeal of Sieversian scansion lies in its explicit promise to bring a remarkable amount of simplifying order to the troublesome diversity of specific types. Does this make Sieversian scansion "right," however? Not necessarily. The "correctness" of a scansion system can never be proved, I suspect, although it can probably be disproved (for example, a system is "incorrect" if it systematically identifies verse types as "the same" when they clearly do not belong together, for one reason or another; in practice, though, even this criterion can be difficult to apply). But certain ways of grouping verses together and certain principles of scansion are so widely agreed-upon that we can take them as reliable givens.

The first is resolution, which allows us to treat a short stressed syllable and its immediate successor as counting the same in our scansion as a long stressed syllable alone. [19] In the passage above, resolution results in the following changes:

755a: Hyge wæs him hinfūs    / x x / \

757a: swylce hē on ealderdagum    x x x x / x \

758b: mǣg Higelāces    / / \ x

761a: eoten wæs ūtweard    / x / \

The evidence for resolution in classical Old English verse is fairly clear: in a large number of cases (such as the verse types given), long stressed syllables and sequences of "short-stressed plus unstressed" syllables appear in identical environments, essentially in free variation. In other cases, the only time three unstressed syllables appear together is when one follows a short stress, suggesting that the maximum number of unstressed syllables allowed is two, and the first of the three counts as part of the stress (or "lift"). [20] In short, most of the time, there is little or any benefit to assuming that resolution does not apply, and a great deal of benefit—considerable simplification of the proposed rule system—to assuming it does (although in some circumstances resolution is blocked or disallowed). [21]

Second, metricists regularly allow the parts of compounds to be counted as equivalent to separate elements with primary stress and, conversely, allow an element with primary stress to be "demoted" to metrical secondary stress. That is, in terms of scansion, it is often important that lingustically stressed words be matched with metrically stressed positions, but less important that the levels of stress match exactly. In the given passage, this leads to the following conventional scansions: [22]

759a: ǣfensprǣce    / x / x

761b: eorl furþur stōp     / \ x /

Metrically, then, 759a is considered to be the same type as 757b and 760b, and 761b can be considered the same as 759b.

Third, naturally unstressed elements can be promoted to full stress or naturally stressed elements can sometimes be treated as unstressed, although both processes are relatively rare—precisely for the reason that the meter seems to prefer a close match between linguistic stress and metrical stress. Too many mismatches would undermine the viability of the meter; if the metrical patterns are not normally manifest by speech rhythms they would be unintelligible (remember that, unlike iambic pentameter, the listener cannot expect a single underlying metrical 'beat'). In the passage in question, the double "d" alliteration in 756a, "secan deofla gedraeg," has often suggested to metricists that the infinitive "secan" here must be treated as unstressed, because otherwise it ought to be expected to alliterate. [23]

Taking these three conventional agreements into account, we arrive at a scansion as follows:

Hyge wæs him hinfūs,    wolde on heolster flēon    / x x / \    / x x / x /

sēcan dēofla gedræg;    ne wæs his drohtoð þǣr    x x / x /    x x x / x /

swylce hē on ealderdagum    ær gemētte.    x x x x / x /    / x / x

Gemunde þā se gōda    mǣg Higelāces    x / x x x / x    / / \ x

ǣfensprǣce    uplang āstōd    / x / x    / \ x /

ond him fæste wiðfēng;    fingras burston;    x x / x x /    / x / x

eoten wæs ūtweard,    eorl furþur stōp.    / x / \    / \ x /

As the preceding discussion suggests, these simplifications reveal additional similarities, but this still yields eleven different metrical types in only fourteen verses. To simplify things beyond this level of scansion, however, we must adopt a particular scansion system.

Before doing so, it is worthwhile to address the implications of choosing a system. First, why should we move beyond this level? The answer to that question may seem subjective but I hope will be obvious: scanning at this level still leaves us with dozens, if not hundreds, of different allowed forms, and most metricists (and I am one of them) prefer to believe that the large numbers of specific observed forms are simply the "output" of a system that has a much smaller number of conceptually distinct forms. From the point of view of description, as well as for the purposes of interpretation and reconstruction, a smaller number of verse-types is more 'elegant' and efficient, and more likely to reflect the realities of composition (so far as these will ever be known). For example, we simplify the system greatly if we see the following verses as minor variations of one basic type, the former differing from the latter by a single additional "extrametrical" syllable:

755a: Hyge wæs him hinfūs    / x | (x) / \

761a: eoten wæs ūtweard    / x | / \

In this notation, the vertical line has been added simply to point out that the verses both fall easily into two parts, and that, except for "him" in 755a, the verses are otherwise metrically identical in both the first part and the second. The similarities suggest that both verses can be said to belong to the same basic "type" of verse; they differ only in a single unstressed syllable. The major differences between scansion systems revolve almost entirely around how they treat such unstressed syllables (or more precisely, syllables that are generally perceived as unstressed). This being so, choosing among competing scansion systems is a surprisingly difficult task, as the choice requires attending to the very syllables that are frequently lumped together (often by metricists themselves) as an indistinguishable mass. [24]

It is at this point in the analysis, then, that metricists tend to depart from one another, each generally preferring his or her own system for his or her own reasons and ends. Sievers-Bliss [25] analysis, for example, takes the two-part structure noted above, based on the location of stressed syllables, as its central feature, and suggests the following combinations of beginnings and ends, with optional numbers of syllables in various places:

Type A    / x | / x

Type A3    x x | / x

Type B    x / | x /

Type C    x / | / x

Type D    / | / \ x or / | / x \

Type E    / \ x | /

Russom, echoing (in some ways) Pope's analysis of measures (as well as Bliss's "light" verses), replaces "/" with "S" and "\" with "s" to mark primary and secondary stress, giving: [26]

Type A     Sx|Sx

Type A3     xx|Sx

Type B     xx|Sxs

Type C     xx|Ssx

Type D     S|Ssx or S|Sxs

Type E     Ssx|S

Russom's system, with its clarification of foot-boundaries in types B and C, has surprising consequences, the most important of which, perhaps, is the consequence for alliteration. Types A, D, and E, which have two S feet, are generally marked for double alliteration in the a-line, while the "weak-onset" types B and C (having only one S foot each) take double alliteration only optionally. Since alliteration is such a major feature of Old English meter, to my mind Russom's system offers an explanation for something that Sievers' system cannot clearly explain, and thus it seems worth serious consideration. If you believe that the purpose of scansion is merely to provide the simplest description of the data, the Sievers scansion may be for you. If you believe that a scansion system should provide insight into the poetic system itself, Russom's system may be superior.

My own scansion system for classical verse modifies Russom's system to deal more effectively with the problem of finite verbs, which sometimes alliterate at the beginning of verses and sometimes do not. Identifying Old English "finite verb feet" or "s-feet" as including the following, "s" "sx," "xs" "xsx" "xxs" and "xxsx", my system allows the following basic types (examples from Beowulf):

Type A (ends with Sx or Ss):

    xA (e.g., xx|Sx)     Sievers A3     1175a: Mē man sægde

    sA (e.g., sx|Sx     Sievers A or A3     926a: stōd on stapole

    SA (e.g., Sx|Sx)     Sievers A     1282a: Grendles mōdor

Type B (ends with Sxs or Sxxs)

    xB (e.g., xx|Sxs)     Sievers B     135a: ac ymb āne niht

    sB (e.g., sx|Sxs)     Sievers B or D4     358a: ēode ellenrōf

    SB (e.g., S|Sxs)     Sievers D4     218a: flota fāmīgheals

Type C (ends with Sxx or Ssx)

    xC (e.g., xx|Ssx)     Sievers C     3a: hū ðā æþelingas

    sC (e.g., sx|Ssx)     Sievers C or D     2321a: Hæfde landwara

    SC (e.g., S|Ssx)     Sievers D     1641a: frome fyrdhwate

Type E

    Ssx/S     Sievers E     636a: fēondgrāpum fæst

Because the classical verses described by Sievers-Bliss, Russom, and me are all the same, there is a great deal of overlap in all three systems: which Sievers type my sA, sB, and sC types correspond to, for example, simply depends on whether or not the finite verbs alliterate: for Sievers and Bliss (and Russom), the presence or absence of such alliteration was often sufficient to distinguish between different types; in my system alliteration changes the character of the verse (including issues such as where in the clause it might appear), but not the scansion or the type.

I like to tell myself that my system is an improvement over Russom's because just as his system could better explain double alliteration requirements, my system deals more effectively with unstressed syllables, especially those before alliterating stresses in the first foot (i.e., syllables in what is called anacrusis). But as should be clear by now, my conviction that these little unstressed syllables are of crucial importance may well be no more than a reflection of my investment in my own system.

In the end, it is almost certainly worth repeating that scansions sytems—like all the results of metrical study itself—stand at the end of an inductive process, in which metrists simply try to describe the data before them. The data is complex, and its integrity is unknown—that is, while we know that the surviving manuscripts of Old English poetry were not always copied carefully or precisely, we don't always know exactly where or when a given scribe may have altered his text, changing its metrical details in ways we may not be able to detect—and the variety of scansion systems available serves as a reminder that bright and well-intentioned people often differ in opinion about the best ways to describe complex phenomena. At their best, scansions systems might indeed help give us insight into the reasons why one type of half-line is common and another rare, but it is equally important to remember that none of the scansion systems we use to describe Old English poetry was likely to have been used as a prescriptive system by Old English poets themselves—poets presumably employed an internalized and instinctive sense of metricality. The complicated functioning of resolution (if nothing else) should powerfully remind us how useless our own internal guidelines and instinctive poetic senses sometimes are in this field. Further, poets are notorious rule breakers, and the task of metricists is to describe the general system they made use of, not necessarily to account for every single verse.

Each scansion system, then, has its own particular strengths and weaknesses, and in the best-case scenario, the choice of a scansion system should be made on the basis of its strengths and weaknesses, not merely on a basis of popularity, traditionality, or old habit. So before we give up entirely on discussing scansion systems, it seems valuable to look briefly at how each of these systems deals with line 758a, a notable crux in Beowulf, to briefly consider the strengths and weaknesses of various systems.


Meter and editorial choice

Beowulf 758a, "Gemunde þā se gōda" is conventionally identified as "Type A with anacrusis"; [27] the following scansions ought to apply in the systems under discussion:

(x) / x x x | / x (Sievers)

x x x x x / x (Bliss)

(x) Sx | (xx) Sx (Russom)

xsx | (xx) Sx (Bredehoft)

Considered in isolation, there are no apparent problems with this verse: the problem arises when we look at 758b, "mǣg Higelāces," which ought to demand "m" alliteration in the a-line. For most metricists, then, the single alliteration on the finite verb gemunde is problematic—and I have chosen my words carefully in that formulation, so as not to favor any particular system. For Sievers and Russom, verses of this type ("Type A with anacrusis") demand double alliteration: the problem with the verse lies in the fact that "gōda" does not alliterate. [28] In Bliss's system, the alliteration on the finite verb is non-functional; he feels that gemunde is completely unstressed and thus that the verse as it stands fails to alliterate at all. Bliss's solution is to emend metri causa, replacing "gōda" with "mōdga" (21). [29] The verse then has appropriate (single) alliteration on the adjective. In my formulation, the problem with the verb is an "alliterative mismatch": Type sA generally alliterates on the A-foot, but here the alliteration falls (only) on the s-foot. Bliss's emendation (which itself has a long list of supporters) would indeed also resolve the problem(s) for me, Russom, and Sievers. But is it necessary that all anomalies need to be resolved? Only if we believe that every single verse must fit straightforwardly into the larger system. What is important about this discussion lies not in the identification of anomalies, however, but in how the different scanning systems conceptualize this kind of problem. Bliss's treatment of the verb as an "unstressed sentence particle" forces him to emend; Sievers and Russom, by treating the verse as Type A with anacrusis, are forced to note the alliterative problem (single alliteration where double alliteration is expected); my system forces me to also note the alliterative anomaly (alliteration only on the s-foot of an sA verse). But for Sievers and Russom, the anomaly is that the type itself demands double alliteration; for me the anomaly lies in where the alliteration falls. It is very much worth considering which kind of constraint the poet is likely to have felt (even if we conclude the poet chose not to be constrained by it). And it is precisely this kind of question that makes the choice of a metrical descriptive system important.


Classical Old English verse and Late Old English verse

Before I leave Beowulf 758a, however, let me explore how meter helps (or fails to help) us think about such anomalous verses in another way. For Sievers, Bliss, and Russom, the forms (and alliteration patterns) of Old English metrical verses rarely change. If we feel that Beowulf 758a ought to read "Gemunde þā se mōdga", we would probably explain the manuscript reading "gōda" by blaming a careless, inept, confused, or dozing scribe. Katherine O'Brien O'Keeffe, however, has powerfully argued that many scribes of Old English verse often seem to have functioned as active tradition-bearers, and that when there is evidence for the changes they made in copying poetic texts, the alterations often have the character of allowed and allowable, even formulaic, variations. She labels this sort of scribal activity "formulaic reading"; the existence of such a phenomenon should, perhaps, remind us to blame an anomalous verse line on scribal incompetence only as a last resort. [30]

In the case of Beowulf 758a, in fact, we might see the manuscript reading as a classic, even textbook, example of such formulaic reading. For a parallel example we might look at the following verse from the Old English Metrical Psalms (PPs 84.5.1a), "Ne wrec þū þīn yrre," where the b-line clearly indicates "w" alliteration. Here, too, then, we have alliteration on the a finite verb ("wrec") in the first foot of a similarly structured verse ("Type A with anacrusis" or "sA") in preference to the following fully stressed word. Conventionally, metricists have regarded the frequency of such departures from the metrical norms of Beowulf as evidence that the Psalms are metrically poor or weak—explicitly not up to the high standards of the Beowulf poet. Such a viewpoint, of course, makes implicit claims about which poems deserve to belong to the database from which we derive our rules.

An alternative viewpoint, however, would suggest that the Metrical Psalms belong not to the classical metrical tradition exemplified by Beowulf but to a different, perhaps later, tradition of Old English verse. [31] The Psalms are regularly dated to the late tenth century, and within the Psalms verses like "Ne wrec þu þin yrre" are relatively common and absolutely normal. So by the time the Beowulf manuscript was copied—around or shortly after the turn of the century—it is perfectly possible for us to imagine that a scribe might have written a verse like 758a "Gemunde þā se gōda," even if his imagined exemplar might have read something else. A scribe acquainted with the forms and habits of Psalms-style Old English verse might well have felt that what he wrote in the manuscript was a perfectly metrical verse with non-anomalous alliteration. If formulaic reading involves replacing readings in an exemplar with metrically viable alternatives, then at the time of the Beowulf manuscript's copying, those alternatives would have included such "late" Old English verse-patterns (including ones that diverge from the standards we reconstruct from Beowulf as a whole), and 758a suddenly becomes utterly non-problematic—for the scribe, at least, if not for the modern scholar.

As this example suggests, rethinking late Old English meter can even help us rethink some of the problem of earlier poems surviving in eleventh-century manuscripts. In the end, however, the metrical perspectives we take are not merely conclusions about constraints that might have been felt by poets: they also have the potential to constrain the ways in which we think about the viability of metrical alternatives. To the degree that the editorial decisions that lead to the texts we read and use have been constrained by their editors' thinking about Old English meter, it is of real value for literary interpreters to be able to think through or beyond those constraints. In the context of the verse from Beowulf, for example, we might ask: should an editor attempt to represent an authorial version of the poem, reconfiguring an "anomalous" verse in order to bring it into conformity with the general metrical patterns we find elsewhere in the poem, or should an editor accept line 758a as it stands, since it is grammatically and logically plausible, and may have been perceived as metrically acceptable by the scribes who produced the poem's only surviving witness? [32] But with a fuller consideration of the late Old English verse tradition, we have a larger and more interesting set of options to consider when we ask such questions.

It may be best to conceptualize Old English verse as a complex system of traditions which changed over time, probably varied across the social spectrum, and probably even countenanced different contemporary opinions about what constituted a "correct" verse or line of poetry. It is only by starting with such an understanding that we have much hope of sorting out the differing contexts of various Old English poems, and how those contexts affected their form and content; without such an understanding, the verse from the Psalms itself is simply one more metrical anomaly to be explained or emended away, or lumped together with a considerable number of other verses as evidence of the endemic metrical irregularity of the Metrical Psalms, further evidence of their poetic inferiority.


Prose-like verse or Verse-like prose?

The traditional perspective in Old English literary studies suggests that the corpus includes a number of powerful, technically competent poems like Beowulf, an uncertain amount of "poor" or "debased" or "irregular" poetry, some highly rhythmical and even alliterative prose, and even a fair amount of normal, unadorned prose. Into which category compositions like the Metrical Psalms fit seems to me to be a matter of at least some uncertainty: most scholars feel more comfortable at the ends of the spectrum, where they find either unadorned prose or the highly regular classical meter of Beowulf. But within Old English literature, some of the most interesting texts fall somewhere in the middle ground. This state of affairs, in my opinion, suggests how important it is that we should not refuse to ask the key questions: where should the boundary line between verse and prose be drawn? Where would the Anglo-Saxons have drawn that line? Would the line have been drawn at different places at different times?

The answers to such questions, in my opinion, are far from obvious, and they are made only more difficult by the rarity with which Anglo-Saxon authors indicated their own ideas about what genres they might have been working in. But in the end, the questions all appear (to me) to be metrical questions. If the key distinction between prose and verse is (as I believe, along with Hopkins and Jakobson) that verse is characterized by a repeated "figure of sound," then the question is very clearly a metrical one. The figures of sound involved in classical Old English verse are clear enough, even if metricists will continue to argue over the details. But the question I have recently attempted to ask about whether there might have been a second, perhaps subsequent, tradition of versifying in the late Old English period is an important one precisely because it imagines the possibility of a differing tradition. If there is indeed a different tradition, we should not expect all surviving Old English verse to be characterized by the traditional features of classical verse (formulas, formulaic themes, formulaic compounding, etc). And if that is the case, the only way we will be able to determine the existence or non-existence of a late Old English verse tradition, I think, will be by careful, metrical, analysis.

In the end, though I want to acknowledge again that detailed metrical study is not for everyone, the results of metrical study are, and have always been, of interest to editors and readers of Old English poetry and prose, at least at some level. To the degree that such readers rely upon the conclusions drawn by particular metricists (and their particular metrical theories), especially when these conclusions are used to label one text as "good" and another "bad," it behooves them to have at least some understanding of the metrical theories on which these value judgments are based. To the extent that readers care about literary history—the ordering of poems as early or late, the possibility of influence between one poem and another, the similarities or differences in style that group different poems together—it is necessary to have some grounding in the conclusions (i.e., "metrical rules") formulated by metricists, and the differing perspectives that generate those conclusions. And to the extent that they rely upon the integrity of the texts they read, readers must, whether they like it or not, have some informed opinion about Old English meter. I think it matters a great deal whether we have "Bēow Scyldinga" or "Bēowulf Scyldinga" in Beowulf 53b; and if readers hope to choose between these two alternatives, they must attend to questions of meter.



[1] Cf. A. J. Bliss, The Metre of Beowulf, rev. ed. (Oxford, 1962), 58.

[2] C. Plummer, ed., Two of the Saxon Chronicles Parallel (Oxford, 1892-99). One could say that in Textual Histories: Readings in the Anglo-Saxon Chronicle (Toronto, 2001), I attempted an essentially literary answer to the question posed by Plummer's lineations, by tracing literary themes through the Chronicle poems; in "The Boundaries between Verse and Prose in Old English Literature," in Joyce Tally Lionarons, ed., Old English Literature in its Manuscript Context (Morgantown, WV, 2004), 139-72, I attempted a "manuscript studies" answer; in Early English Metre (Toronto, 2005), I finally tried a metrical answer.

[3] The issue here is that Beowulf and the Metrical Psalms differ widely in a number of metrical details; historically, the metrical system used in Beowulf has been understood as definitive for Old English verse in general, and as a consequence, the Psalms have historically been seen as having a "very general metrical irregularity" (G. P. Krapp, ed., The Paris Psalter and the Meters of Boethius, ASPR V [New York, 1932], xvii). But the decision to place Beowulf at the center of our understandings of Old English meter deserves to be explicitly defended, partly because (from the perspective of the Psalms) we might alternatively interpret Beowulf as departing from the "standard" practice. In this sense, the choice of a data set for our inductive conclusions is of crucial importance, and the conventional choice to place Beowulf at the metrical heart of things is neither obvious nor innocent.

[4] D. Donoghue, Style in Old English Poetry: The Test of the Auxiliary (New Haven, CT, 1987); M. Blockley, Aspects of Old English Poetic Syntax: Where Clauses Begin (Urbana, IL, 2001); H. Momma, The Composition of Old English Poetry (Cambridge, 1997); and C. B. Kendall, The Metrical Grammar of Beowulf (Cambridge, 1991).

[5] Some definitions may be in order: double alliteration is the use of two alliterating stresses in the a-line; cross alliteration involves two different alliterators used in both half-lines (e.g., Beowulf 1: Hwæt, we Gar-Dena in geardagum, with both "g" alliteration and "d" alliteration); verse rhyme is rhyme that links the last stresses of two half-lines within a line.

[6] R. Jakobson, "Linguistics and Poetics," in Roman Jakobson, Language in Literature, Ed. Krystyna Pomorska and Stephen Rudy (Cambridge, MA, 1987), 72.

[7] George Ellis's 1805 Specimens of Ancient Poetry (London, 1805) said of Old English verse that "its mechanism and scheme of versification, notwithstanding all the pain which Hickes has employed in attempting to investigate them, are still completely inexplicable" (8). Cited in R. M. Liuzza, "Lost in Translation: Some Versions of Beowulf in the Nineteenth Century," English Studies 83 (2002), 284.

[8] E. Sievers, Altgermanische Metrik (Halle, 1893). The general acceptance of Sievers's arguments, of course, does not suggest that there have not been dissenters to Sievers's views; indeed, this essay largely concerns itself with those dissenters. But Sievers's explication of Old English verse was so powerful and persuasive that it remains central to most beginning discussions of Old English meter.

[9] It is probably worth clarifying what we mean by stress: in Old English (as in Modern English) two types of stress exist: word-based stressed and sentence-based stress. At the level of the word, every word may be fully stressed (which simply means that every word has a key acoustic peak, which is where the stress falls when a word is stressed). Multisyllabic words may have multiple stresses of varying intensity: in classical Old English verse, three stress levels are of particular significance: primary stress, which falls (in Old English) on the first root syllable of any word; secondary stress, which falls on any secondary elements of compounds; and "unstressed" which characterizes any remaining syllables. (There has been a long debate in OE metrical and phonological studies about the possible relevance of "tertiary stress," a stress level between secondary stress and unstress, but I do not believe it is, in fact, relevant.) In terms of sentence-level stress, many "function" words (e.g., pronouns, copular verbs, conjunctions, and so on) are generally unstressed, leaving "content" words with full stress (i.e., the full value of the "word-based" stress pattern). So, in the most common pronunciation of a modern English sentence like "I will ask the pickpocket," the "content words" are ask and pickpocket, and thus we would mark the syllables ask and pick- with primary stress; -pock- with secondary stress, and all other syllables as unstressed. See below for additional comments on sentence-level stress in OE verse.

[10] Sievers's system, of course, also notes a number of other features, such as where additional unstressed syllables are allowed, and so on, but these forms are so central to his system that Old English verse continues to frequently be characterized as having five basic types.

[11] D. L. Hoover, A New Theory of Old English Meter (New York, 1985).

[12] J. C. Pope, The Rhythm of Beowulf, rev. ed. (New Haven, CT, 1966).

[13] T. Cable, The English Alliterative Tradition (Philadelphia, 1991).

[14] G. Russom, Old English Meter and Linguistic Theory (Cambridge, 1987).

[15] Within the tradition of iambic pentameter versification, it may be worth noting that later poets tend to employ the tension between metrical stress and linguistic stress more and more frequently and powerfully as a poetic device: where Chaucer's iambic pentameter is highly regular, with relatively few disjunctions between poetic and linguistic stress, such disjunctions increase in the works of poets like Shakespeare and Milton. Old English verse operates more like Chaucerian iambic pentameter, in that the number and variety of differences between verse stress and speech rhythm are generally minimized.

[16] As the preceding passages indicate, any word in Old English can be stressed in poetry, depending on its placement in a verse. And as noted above, for words of more than one syllable, some additional rules apply: the primary stress of any word is on the root syllable (that is, prefixes and suffixes are unstressed) and compound words have secondary stress (or "half-stress") on their secondary elements.

[17] Cited from F. Klaeber, ed., Beowulf and the Fight at Finnsburh, 3rd ed. (Boston, 1950).

[18] I should note here that in Early English Metre, I hypothesize a different scansion system for late Old English verse, as opposed to 'classical' verse. In classical verse (which is the variety of Old English verse used in Beowulf and a good number of other poems), three stress levels are metrically operative, and we need to mark out secondary stresses. In late Old English verse (which includes the Metrical Psalms, many Chronicle poems and a number of other pieces), there are, on the other hand, only two levels of metrical stress. Thus, in practice, even the most basic scansion depends upon correctly identifying whether a passage is classical or late.

[19] In general, an Old English syllable is long if it has a long vowel, or if it is "closed" by a syllable-final consonant. In a two-syllable word like "hyge" (755a), the syllable boundary is considered to fall before the "g" as hy-ge; thus both syllables of this word are short. Like many other aspects of Old English phonology, syllable length is something that Modern English speakers have little intuitive feel for, but it is nonetheless a crucial phenomenon.

[20] Part of the difficulty of writing about (or explaining) resolution is that, here too, we have very little (if any) intuitive insight into how it might have operated for Anglo-Saxon poets and listeners. But to clarify the logic for resolution, consider verses like the following:

Beowulf 362a: ofer geofenes begang     xx/ | xx/

Beowulf 3068a: þurh hwæt his worulde gedāl     xxx/ | xx/

In Beowulf (and in classical Old English verse in general), these scan as B-type verses, with three syllables between the stresses—but such "trisyllabic dips" only occur with any frequency when (as here) the initial stress falls on a short syllable. The principle of resolution simply explains this distributional pattern by absorbing the first syllable of the trisyllabic dip into the first "lift" (or stress) of the verse, leaving a more typical two-syllable dip, with the resulting (Sieversian) scansions shown as above. So much of classical Old English verse-type distribution is clarified or simplified by resolution that it is widely accepted as a feature of classical scansion.

[21] Most accounts of classical Old English meter include some account of where resolution is and is not blocked or suspended. In general, it is often the case that resolution is suspended if employing it would result in a thoroughly unmetrical verse, e.g., one with fewer than four metrical (that is, counted) syllables or positions.

[22] I have given the conventional reading of 761b, which scans it as an E-type verse because type D4 is understood as being excluded from the b-line. I believe, however, that this conventional position is "wrong" (i.e., not as descriptively valuable as it might be), and that a more apposite scansion here would be / / x \, placing the verb ("stōp") and modifying adverb ("furþur") together as a syntactic unit separate from the noun.

[23] In the interests of full disclosure, I want again to confess that my own position on stress promotion and demotion (especially with finite verbs) differs somewhat from most other metricists, who treat finite verbs at the beginning of a clause (such as "wolde" in 755b) as always unstressed, or at least only stressed if they alliterate in the a-line. So most metricists, when "marking up" this passage, would have assigned "xx" to "wolde" from the very first step.

[24] Sievers's system simply counts unstressed syllables: differences in subtypes depend almost entirely on the number and position of half-stressed or unstressed syllables. For Bliss, the type and location of unstressed syllables leads to his theory of the verse-internal "caesura" and resulting "breath groups." For Russom, many (but not all) unstressed syllables are "extrametrical" (that is, they are literally not counted by the meter, which "counts" only the two metrical words or word-like stress structures). My scansion system departs from Russom's by allowing extrametrical syllables only in the medial position.

[25] With some hesitation, I treat Sievers and Bliss together here. Bliss himelf claimed his analysis was "a triumphant vindication of Sievers" (v), but he allows one-stress verses, and his theory of the caesura often uses a quite different dividing line between the parts of two-stress verses; as such the details of the two systems are really not especially compatible.

[26] Pope had noted that the final three positions of Sieversian types C (x / | / x) and D (/ | / \ x) were often filled identically, and this insight has generally been accepted, leading to Bliss's type d (x | / \ x), which replaces many Sievers C types, and Russom's x|Ssx (which explicitly parallels his S|Ssx).

[27] Anacrusis (because it deals with unstressed syllables) is one of the classic difficult topics in Old English scansion, defined in various ways by various metricists. Technically, anacrusis refers to unstressed syllables occurring before the beginning of the verse proper. From such a perspective, 758a has anacrusis only for the scansions given below in the Sievers and Russom systems. More generally, however, we might define anacrusis as including any unstressed syllables occurring before an alliterating stress in the first foot, in which case my scansion system would also allow 758a to be counted as featuring anacrusis. Once again, different systems deal with unstressed syllables in often surprisingly varied ways.

[28] Such verses "demand" double alliteration. Of course, what this means is that virtually all such verses exhibit double alliteration, and so when we fail to see it in this verse, we must note that it goes against the general trend. But when a "rule" (i.e., observed pattern) like this appears to be "broken," the traditional response has often boiled down to the assertion that the Beowulf poet would never have broken this rule. I hope the numerous assumptions embodied in that traditional response are clear.

[29] One should note the implications of this emendation, which explicitly locates the problem in 758a. But it is also possible that the problem (if there really is one) might lie in 758b. If "mǣg" were replaced, for example, by a word alliterating on "g", then the alliterative problem in 758a would literally disappear. Emending "gōda" to "mōdga" is attractive in part because it changes a minimal number of letters, but the very principle that says the "best" emendation is the one that minimally alters the manuscript text assumes a paradigm of scribal activity that may not apply. Cf. O'Brien O'Keeffe's very valuable notion of "formulaic reading," which would allow us to understand a quite radical alteration to 758b as a quite typical scribal intervention. See also the discussion below.

[30] Katherine O'Brien O'Keeffe, Visible Song: Transitional Literacy in Old English Verse (Cambridge, 1990).

[31] This viewpoint is argued more fully in my chapters on "Late Old English Verse" in Early English Metre.

[32] Klaeber and Dobbie let 758a stand; Bliss emends it.