Six
Traditional Phraseology in Beowulf and Old English Poetry

The history of scholarship on formulaic phraseology in Old English poetry is long and complex, stretching from the German Higher Criticism of the nineteenth century to the present, and so I shall make no attempt to summarize its evolution here,^[1] choosing instead to proceed directly to an analysis of the phraseology. As in the preceding. chapters on ancient Greek and Serbo-Croatian epic diction, the object will be an understanding of the phraseology on its own tradition-dependent terms, that is, on the basis of traditional rules particular to Old English poetry. Once again, reference will be made to the findings of chapter 3 on comparative prosody, and distinctions will be drawn among the three phraseologies involved.^[2]

From Prosody to Formulaic Structure

In lieu of attempting to reshape an existing model in order to represent fully the complex structure of Old English poetic phraseology, let us follow the practice established in earlier chapters and begin by asking what we might reasonably expect to find in the way of formulaic structure. In other words, given the idiosyncratic prosody of Beowulf and other Anglo-Saxon poetry, what kind of diction is possible and likely?

We recall that, while the ancient Greek and Serbo-Croatian metrical

lines—the hexameter and deseterac , respectively—show at least some similarity in their outward make-up and inner morphology, the Old English alliterative line diverges widely from these two comparands. Not only is the Anglo-Saxon line a stress-based rather than a quantitative meter, but it also varies significantly in syllable count, has no caesura and therefore no colonic structure, and exhibits nothing of the Indo-European tendency toward what we have called "right justification." Under such variant conditions there is, we discovered, no reason to expect the colonic formulaic morphology typical of Homer and the guslar . Indeed, this important dissimilarity was what led to our outright rejection of formula-density tests of Old English verse that use the colonic formula as the phraseological model.

Instead of the encapsulated phrase that can vary only in strictly defined ways under the constraints of syllable count, colon configuration, and other features inherited from the symbiosis of Indo-European meter and phraseology, we encounter in Old English poetry a phrase that in its very inscrutability reflects the prosody that supports it. It is perhaps not an accident that no issue in Old English scholarship has been more and longer contested than the definition of the meter, in part because this Germanic line does not fit any familiar, institutionalized Greco-Roman model. And with the understanding of the prosody in a state of uncertainty, the "metrical conditions" of Parry's concept of the formula could not be adequately and fairly set. If investigators adhered too stubbornly to the colonic formula, or searched for too tidily synchronic a model of its morphology, it is at least in part because they had no way to confront the highly idiosyncratic prosodic underpinnings of Old English poetic diction.

In chapter 3 it was demonstrated that most metrical systems proposed for Old English poetry agree in large part that the chief prosodic features of the line are stress emphasis and alliteration. Beyond these regularly occurring characteristics of the meter, we found that the poet used a limited set of stress-patterns; whether we refer to these patterns as Sievers's Five Types or employ the revisions of Pope, Greed, or Cable, the overall picture is approximately the same: four stress maxima (SMs) per line, with a varying number of secondary and minimal stresses. The possibility for variation is greatly enhanced by the further factors of resolution (more than one syllable under a heavy stress) and ramification (multiple minimally stressed syllables), while the restricted set of permitted half-line types counterbalances the tendency toward multiformity. Whatever metrical theory we may select to tabulate line-types and explain their interrelationships, these are the bare metrical facts one must confront in attempting a faithful account of formulaic structure in Old English verse.

A few additional facts can be derived from these first premises, and they will aid us in pursuing the assortment of traditional structures we find in Beowulf . First, from the prior scholarship on prosody and from chapter 3 we can readily

see that the alliterative line is a hybrid prosody—that is, it reveals both half-line (or verse) and whole-line levels of metrical organization.^[3] At the half-line level, as demonstrated earlier, only a limited number of stress patterns are permitted. Furthermore, all of these patterns, however defined, can occur in either the first or the second verse, so that there is an interchangeability among half-line metrical patterns that is restricted only by the prosodic rule that second half-lines (or b-verses) cannot bear double alliteration. Again unlike the Homeric hexameter and Serbo-Croatian deseterac , then, the Anglo-Saxon alliterative line is a balanced, symmetrical unit with essentially interchangeable halves.

At the same time, because alliteration binds the two half-lines together and creates a larger unit at another level, the alliterative line also shows a whole-line metrical structure. Although the individual verses are largely interchangeable, this agreement of initial sounds defines the composite structure as an entity in its own right. And, as we saw in chapter 3, there is evidence that the Beowulf poet superimposed on this whole-line variety a set of three favored line-types or -patterns which further organized the meter and his phraseology. Although the corpus of comparable poetry in Old English (that is, epic poetry that could serve a comparative role under the criterion of genre-dependence) is far too small to permit determination of whether this kind of metrical formula was a general phenomenon^[4] or not, the fact of the increased metrical conservatism constitutes further evidence for whole-line metrical organization.

To summarize, then, we should begin our search for the formulaic structure of Beowulf by realizing that the Old English alliterative line consists not of a colon-based, quantitative meter with some expression of right justification,^[5] but rather of a stress-based meter which figures itself forth in a limited series of multiforms governed by morphological rules. And while we note the importance of the half-line unit in these respects, we should also remember that the ways in which these multiforms combine under the aegis of binding alliteration reveal that the poet also composes in whole lines. The truest explanation of the compositional process, at least from a metrical perspective, is as a two-level or hybrid process. If we choose to reduce it to only the one or the other (probably to concentration on the half-line, as has usually been done), we may simplify the task of description, but we shall at the same time

remove the rich complexity inherent in the process and sabotage any attempt at a faithful account of formulaic structure.

One last feature, uniquely a characteristic of the Old English line, needs to be brought to light before we proceed to an overall description of the kinds of traditional patterns permitted under the alliterative prosody. This feature stems from the balance between verses in any line, and for that matter among all verses. Given such balance, or interchangeability—so different from the "four and six" cola of the deseterac or the complex collection of unequal cola in the Homeric hexameter—it is inherently unlikely that lines will be end-stopped as frequently in Old English as in the other two poetries. That is, the balanced line structure may be construed as actually encouraging enjambement; more often than not, that enjambement may well be of the "unnecessary" variety that carries over into the next line to add to an already syntactically complete utterance,^[6] but there is nothing metrical to discourage the use of "necessary" enjambement as well. Indeed, since one verse is, with the exception of the alliterative constraint and metrical formula, just like any other, a phrase may be completed at mid-line or end-line as the poet wishes.^[7] Or he may "sort" his traditional patterns over more than a single line, since there is little metrical resistance to doing so. We shall soon see in more detail how this feature affects not only enjambement but also traditional patterns larger than a single line, but for the moment it is perhaps enough to observe that the typically Anglo-Saxon poetic figure of variation (the accrual of appositives to a noun or verb to form a paratactic string of poetic synonyms)^[8] can be traced to this metrical feature.

The Spectrum of Traditional Phraseology in Beowulf

With these tradition-dependent features of Old English prosody in mind, we should be able to predict the shape of traditional expression to be encountered in Beowulf . First, because there is a definite half-line structure to the alliterative line, there should arise a collection of verse formulas and formulaic systems . To the extent that the half-line is a functional unit in the act of composition, these "classic" phrases, analogous but not directly comparable to colonic formulas in ancient Greek and Serbo-Croatian epic, should constitute some reasonably extensive part of the poet's and tradition's wordhoard. Of course, there will be qualifications, such as the matter of idiolect versus traditional dialect as

illustrated for the guslar in chapter 5,^[9] but prior research has shown that a significant number of such verse units actually does exist and that they arc functional.

But here the synchronic model of formulaic structure will no doubt require further articulation from a diachronic point of view—for, as we have seen with other traditional phraseologies, not every "formula" or "system" in a given poetic tradition will answer precisely the same definition. To put it more tellingly, not all of the verse-length patterns found in Beowulf and Old English poetry are likely to be exactly equivalent in terms of semantic, rhythmic, or syntactic complexity. If Anglo-Saxon phraseology is as multi-leveled as ancient Greek and Serbo-Croatian, then some half-line patterns will be able to be defined rigorously using all of these three criteria, some will depend more on one or another of the measurements, and some will be mere predispositions that nonetheless recur with regularity and must be recognized as traditional. We may wish to reserve the classical term formula only for those units that answer a rather strict definition, but we shall still need in some fashion to explain and to take account of those obviously traditional elements that do hot fit the imposed definition. In short, we may expect, if prior experience is any guide, a whole spectrum of formulaic diction that has evolved differentially (and was no doubt continuing to do so)^[10] under the tradition-dependent rules of Old English verse-making.

In addition to verse-length units, we should expect to find whole-line formulas and systems ; the whole-line dimension of the alliterative meter, both the linkage made by alliteration between verses and the whole-line metrical formulas, would promote the formation and maintenance of such longer units. Once again, however, we shall have to make allowance for the probability that not all such longer patterns will readily conform to a single synchronic definition. No less than the verse-length units, whole-line patterns are likely to reveal a spectrum of forms, from absolutely constant verbatim repeats such as lines of introduction, through systemic lines in which one verse is regularly associated with, for example, a given alliterating word in the other verse, and on to weaker associations involving metrical predispositions. Positing both half- and whole-line phraseological units does not amount to a contradictory hypothesis; rather it recognizes the hybrid character of the meter and anticipates its reflection in the poetic diction.

Given the tradition-dependent nature of Old English prosody, and especially its emphasis on stressed elements and relative inattention to the unstressed syllables that surround these metrical peaks, it is not difficult to understand

why recurrent single words play such an important stylistic role in this poetry. If the meter is such that it foregrounds these words—that is, the root syllables of alliterating and other stressed elements—then these are also the words that are likely to be constant as the poet uses and re-uses his traditional diction. Indeed, this is no more than, for example, Fry (1967b, 203) has pointed out by defining his "formulaic system" in terms of "the identical relative placement of two elements, one a variable word or element of a compound usually supplying the alliteration, and the other a constant word or element of a compound, with approximately the same distribution of unstressed elements."

But can single words or root syllables also function as traditional elements outside the verse or whole-line formula? It would seem logical that, especially because the Anglo-Saxon metrical filter is so relatively permeable, roots of single words could form traditional complexes of their own, with their half- or whole-line contexts assembled, as it were, at the moment of composition—although under the traditional rules of poetic composition. Two possibilities, with some middle ground in between them, emerge: first, an associative "bundle" of roots might form around some concept or event and bring along with it the formulas and systems in which the words in question are most often found; or, second, the associative package might supersede verse and linear formulaic structure, just as enjambement is in effect encouraged by the balanced structure of the line. If such morphemic bundles can occur in a phraseology as colonic and relatively tightly bound as ancient Greek,^[11] then how much more likely are they to surface in Old English poetry? In fact, to look ahead just slightly, it may be that we can trace the origins of the well-known Anglo-Saxon principle of "verbal echo"^[12] in this ultimately prosodic feature of emphasis on roots of single words.

This in turn means that we must hold open the possibility of formulaic structures not only lesser than but also greater than a single verse or line . There may be such things as three- or even four-verse patterns which are not multi-line "runs" so much as integral structures—that is, they are not concatenations of unitary elements but rather elements in and of themselves. Albert Lord (1986b) has suggested that such longer patterns exist in Anglo-Saxon poetry, and the balanced, additive prosody would seem to favor their formation. If they do exist, then we shall have another very clear example of the tradition-dependent character of Old English verse.

To these working hypotheses we must add a caveat. If analysis of a sample from the Beowulf text turns up even a portion of what might be expected on the basis of the alliterative prosody, and especially if—as in ancient Greek and Serbo-Croatian epic—not all members in a certain category can be considered equivalent in form and function, then we shall need a further explanation

that somehow rationalizes the variety. That is, if the synchronic model of the half-line system cannot account for the richness of Beowulf 's traditional diction, then another, more fundamental solution must be sought. But first let us see what the poem itself has to tell us.

Grendel's Approach to Heorot and Traditional Structure

The ritual gesture for formulaic analysis customarily includes either an exhaustive analysis of the entire text or an apology for the chosen sampling and a claim that the sample to be analyzed is representative of the whole. Since, however, we seek not to reduce the complexity of traditional diction in Beowulf to any one model or to "prove" its orality, but rather to discover the kinds of traditional structures that underlie the poem and give it meaning, we shall be content with looking closely at a sample of about thirty lines: the famous tripartite passage involving Grendel's approach to Heorot (702b-30a).^[13] Yet at the same time, we shall not be content merely to find some evidence of half-line systems, or of any other single type or form of phraseological structure; instead we shall attempt to uncover whatever sort of structure seems to be operative in each verse and line.

First I present the passage in the original Anglo-Saxon, followed by a quite literal translation into modern English:^[14]

Com on wanre niht
scriðan sceadugenga. Sceotend swæfon,
pa pæt hornreced healdan scoldon,
ealle buton anum. pæt wæs yldum cup,              705
pæt hie ne moste, pa Metod nolde,
se s[c]ynscapa under sceadu bregdan;—
ac he wæccende wrapum on andan
bad bolgenmod beadwa gepinges.

Da com of more under misthleopum               710
Grendel gongan, Godes yrre bær;
mynte se manscaða manna cynnes
sumne besyrwan in sele pam hean.
Wod under wolcnum to pæs pe he winreced,
goldsele gumena gearwost wisse               715
fættum fahne. Ne wæs pæet forma sioð,
pæet he Hropgares ham gesohte;
næfre he on aldordagum ær ne sipoðan
heardran hæle, healoðegnas fand!

Com pa to recede rinc sioðian              720
dreamum bedæled. Duru sona onarn
fyrbendum fæst, sypðan he hire folmum (æthr)an;
onbræed pa bealohydig, ða (he ge)bolgen wæs,
recedes mupan. Rape æfter pon
on fagne flor feond treddode,              725
eode yrremod; him of eagum stod
ligge gelicost leoht unfæger.
Geseah he in recede rinca manige,
swefan sibbegedriht samod ætgædere,
magorinca heap.              730

In the dark night
the shadow-goer came stalking. The warriors slept,
those whose duty it was to guard the horned building,
all but one. It was known to men that,              705
if the Ruler did not wish it, the injurer
could not draw them into the shadows;—
but [Beowulf], awake and fiercely angry,
awaited the battle's result, enraged in heart.

Then, out of the moor, under the misty cliffs              710
came Grendel, he bore the wrath of God;
the wicked ravager intended to trap
one of the men in that high hall.
He advanced under the clouds until he could most readily see
the wine-building, the gold-hall of men              715
adorned with plates. Nor was that the first time
that he sought Hrothgar's home;
 [but] never in earlier days, either before or since,
did he find stronger hero[es], hall-thanes!

Then to the building the warrior came journeying,              720
the one bereft of joys. The door, held fast by fire-forged bonds,
immediately sprang open after he touched it with his hands;
then the baleful-minded one swung open that building's mouth,
when he was enraged. Following this, the fiend
quickly trod over the decorated floor,              725
went along angry in spirit: from his eyes
there stood out a horrible light, most like fire.
In the budding he saw many a warrior,
the band of kinsmen sleeping all together,
a troop of young warriors.              730

Analysis

In what follows I shall consider selected examples of the traditional structure of this passage, proceeding half-line by half-line but remaining aware that

our examination of alliterative prosody has suggested that patterns both larger and smaller than the single verse are to be expected. Once the analysis is complete, it will be possible to propose a single, overarching theory for traditional phraseological structure that will explain the idiosyncratic form of Old English diction.

702b. Com on wanre niht ("Came in the dark night"). Comparative evidence from other Old English poems^[15] may at first sight seem to encourage one of two hypotheses about the traditional structure of this phrase:

# 1: Com [x], where x is loosely defined
#2: Com [x], where x is defined as a prepositional phrase

With the added proviso that the stable core of the phrase may include the ubiquitous adverb pa , we find numerous examples of the first formulaic system and twelve instances of the second, more limited kind of phrase, including line 702b. But while either of these two descriptions adequately symbolizes some of the connections between the half-line under examination and apparently related verses elsewhere in the poetic corpus, neither of them seems rigorous enough—that is, basic enough—to explain line 702b as a traditional element. What of the rest of the phrase? And what of the relative placement of com and the prepositional phrase?

Turning then to onwanre niht , we have no more satisfying result. The closest phrase, on sweartre niht (Chr 872a; also translated as "in the dark night"), occupies an entire half-line by itself, and the only other combination of the root wan with niht in a prepositional phrase likewise fills a verse on its own (Glc 1028b). In fact, this problem of varying "size" or "extent" furnishes us with an early glimpse of a general problem: how are we to explain phrases which serve as both half-lines and parts of half-lines in different contexts?^[16]

If the system Com [x] in either of its descriptions and the system on [x] niht cannot completely account for line 702b, we might consider further analysis of the verb cuman in order to determine whether there could be some semantic basis for the formula in an "essential idea." Once again, however, the search yields no programmatic result, since other forms of cuman behave very differently from the preterite third singular com .^[17] In addition to providing

evidence that the Com system, if that is what this phrase really is, does not allow much morphological variation, this difference in behavior tells us something significant about formulaic dynamics in Old English. It tells us that the metrical form s (a single stress) is somehow importantly different from s s^[*] (a single stress plus an unstressed syllable). If we recall the conclusions of chapter 3, we shall remember that the primary aspects of consistency in the Old English line are alliteration and stress patterns. While the extra syllable in comon would mean nothing as far as the absolute syllable count of the line is concerned, the addition of even a single unstressed syllable can interrupt the stress pattern and cause it to change. In effect, such a shift of stress pattern would mean that Parry's criterion of "under the same metrical conditions" would be violated.

But while we do not find this or other such Old English systems exhibiting much morphological variety, we can detect a definite patterning having to do with word-type placement.^[18] The monosyllabic preterite third-singular verb tends toward either the very beginning or the very end of the half-line; if it occurs at the beginning, as it does in 702b, it almost never alliterates.^[19] This phenomenon, chiefly a metrical requirement because it involves a word-type rather than a single word or semantic grouping, bears no necessary connection with what follows in the half-line, and we find a great variety of complements, from the prepositional phrase as we have it in the present example, through adverbs, to subjects of the verb, and so on. And the phrase on wanre niht , itself highly variable in that it can occupy an entire verse or part of one, is combined with the central idea, com , to comprise the first in a "rhetorical" series of three "comings."^[20]

From a traditional perspective, then, the most formulaic aspect of com seems to be its placement as a particular word-type at the beginning of the verse. The poet also has at his command a phrase we may profitably construe as on [x] niht , as long as we do not demand that it be a classically defined verse formula. Diachronically, the predisposition toward placing com first and in a non-alliterating position is accompanied by the existence of a phrase that can either fill a verse by itself or complement that placement and word-type. From a synchronic point of view, we may see the poet substituting wanre to match the alliteration of the preceding half-line. In short, although we cannot define classical systems that satisfactorily describe the structure and dynamics of 702b according to the usual notion of formulaic structure, there is no doubt that this half-line is the product of traditional processes.

703b. sceotend swæfon ("the warriors slept"). The second half of the line has

enough correspondences to other diction in the corpus to permit two hypotheses regarding its formulaic system:

#1: sceotend [x]
#2: [] swæfon^[21]

The second of these possibilities will better serve our needs, since its essential idea ([x] slept) is more focused than that of the first posited system (the warriors [x]). In addition, it turns out that the very act of sleeping in the Beowulfian context is extremely important and rife with overtones—the "sleeping-feasting" theme has been recognized by a number of commentators as a traditional pattern,^[22] and so we are further justified in locating the traditional core of the phrase in swæfon .

But what does this mean for the rest of the line, since swæfon is the only word that does not alliterate in the entire line (the alliteration being in sc , or [š ])? For one thing, this situation should warn us away from making any necessary and exclusive connection between the alliterative staves and the fundamental traditional meaning of the phrase; swæfon bears the fourth stress in this line and does not participate in the "sc" series, but it is clearly the most significant word thematically. From another point of view, we may note what amounts to the converse of that observation: that is, the poet manages to harmonize all of the "non-essential" words in the same alliteration. From the core idea of "[x] slept," expressed in the second verse, he constructs a whole-line increment without the benefit of another formula or system. Is this first half-line then an example of "nonce" phraseology^[23] —of diction created for the moment with no life of its own outside this circumstance?

It would be hard to reconcile purely nonce diction with a traditional perspective. And indeed there is no reason to try to do so, for 703a, while it may have no (surviving) formulaic relatives in the sense conveyed by the classical definition, does participate in a pattern larger than the single half-line, and even than the whole line—that is, in a cluster of roots that recur together^[24] and that collectively embody the idea of the onset of night, shadows, and the ending rhythm of the diurnal cycle. To make the point, I reproduce both another occurrence from Beowulf and a section of the passage under consideration:

The density of the italicized words and roots of words that recur even in this small sampling argues strongly for a cluster of morphs, a group of elements associated with a particular traditional idea—the onset of darkness and all of the terror and ravaging that it poetically connotes. Just as comparative prosody would predict, the key elements in this association are stressed roots (whether words or elements of compounds), and the particular shape they take in a given instance is (again as predicted) only very loosely controlled. The fact of their co-occurrence is not to be attributed to half-line formulaic structure, which is at a loss to explain certain phrases. But the cluster of morphs, which as indicated above is a structure consonant with the tradition-dependent character of Old English prosody, accounts both for the striking resemblances between the two passages as a whole and for elements as apparently untraditional as the two hapax legomena sceadugenga and scaduhelma , in this case an association further focused by their mutual alliterative collocation with scriðan . Although we could not imagine this sort of structure occurring in ancient Greek or Serbo-Croatian epic, the cluster of roots—a sorting of traditionally associated morphs in metrically and formulaically permissible patterns—is in fact a common feature of traditional structure in Beowulf .^[25]

Diachronically, then, we can ascribe the first half of line 703 and the association with scriðan to a cluster of roots which also occurs elsewhere in Beowulf and which bears certain connotations. These connotations are catalyzed by the juxtaposition of the "sleeping" idea, which has definite negative implications in each of the three monster-fights in Beowulf . The actual phraseology owes something to what we may style the "[] swæfon " system, but this is a half-line structure very much subject to the larger design of the cluster, as the traditional character of the hapax legomenon sceadugenga also illustrates. The poet has thus summoned the core idea of "sleeping and feasting" and linked this important mythic and thematic complex with the associative cluster connoting the onset of night and its perils.

704b. healdan scoldon ("had to guard"). The phrase healdan scoldon is a highly traditional phrase whose structure is best appreciated on a number of levels. If we posit a very generic system, [] [sculan/willan ], where x represents an in-

finitive beating the alliteration, we find a great many examples of second-verse phrases. In this case I believe it is helpful to postulate such a genetic level of structure,^[26] even if it has no true essential idea, because the syntax is so widespread and lends itself to so many different sorts of formulaic systems. In addition, a glimpse of different levels of structure reminds us that the synchronic model of the half-line system, summoned to explain all diction, has to be supplemented to reflect the morphological complexity and diachronic nature of the phraseology.^[27]

Within this large group, there are thirteen additional instances in which the infinitive is healdan , so we may identify a more focused system, healdan [sculan/willan , finite form], to compare with the system described above.^[28] This second-verse pattern has clearly become a traditional element in the diction, and serves a useful compositional purpose in widely different situations with a large selection of half-line partners. But while the system in 704b obviously has a diachronic identity, in both the generic and more focused forms, the same cannot be said for 704a, which merely fills a permitted metrical type. We may thus conceive of the poet as working toward a stable core in the second half-line and adjusting the first verse, with the alliterating element horn- providing the synchronic solution to the problem of whole-line structure.^[29]

706b. pa Metod nolde ("if the Ruler did not wish it"). At first sight this phrase seems to be a classically defined formula, with a verbatim repetition at Beowulf line 967b. But we may also posit a system, pa/pæt [] nolde , where x is the subject of nolde and the alliterative stave for the half-line. The first possibility is restricted to Beowulf and thus seems, on the basis of available evidence, to be an idiolectal formula, while the more genetic system also finds expression in two phrases from The Battle of Maldon —pæt se eorl nolde (6a) and pæt se cniht nolde (9b)—and thus is arguably a tradition-wide system.

But once again we gain a full understanding of the phrase only by taking the whole-line pattern into account as well. Consider these two lines from Beowulf :

Clearly, the Beowulf poet is employing a whole-line pattern, with just the sort of consistency and also the kind of variability we predicted on the basis of the Anglo-Saxon alliterative meter. The first half-line shows traditional placement of the single stressed item in an a-verse configuration,^[30] and the line is filled out by the much more stable formula in the b-verse. In each case the b-verse, though informational, is also a bridge to the next line (and next alliterative pattern), and the bridging is accomplished by use of a ready-made formula. Again in each case, the poet will go on to specify the very general action of this line in a subsequent infinitive dependent on moste or mihte . One of the aspects of the utility of 706a, then, is its non-specific lead-in to a specific following action.

To summarize, this line exhibits half-line prosody and phrase patterns (placement of moste/mihte and the b-verse formula), as well as whole-line prosody and phrase patterns (collocation and the whole-line system). In effect, the entire construction reaches forward to the next line, since moste/mihte requires an infinitive to complete its sense. On the basis of available data, the phrase pa Metod nolde seems to be an idiolectal reflex of a more generic, tradition-wide formulaic system, while the whole-line pattern is the Beowulf poet's own compositional element.^[31]

707a. se s[c]ynscapa ("the injurer"). This half-line, whether construed as the manuscript synscaða or as above,^[32] has three formulaic relatives in four additional occurrences of the main word:

At least we can define a system—[x] s(c)ynscað- , where x represents one or more function words—that could theoretically underlie all of these phrases. But in ascribing the half-lines to this system, we are doing little more than conferring formulaic character on a single word. It would be better, and truer to Old English diction as we have examined it, to recognize in this regularity not a substitutable system but rather consistent placement of a word-type () at the end of a verse, with flexibility in what precedes the word. The C-Type metrical pattern institutionalizes that flexibility in a particular way, a way that in turn allows for grammatical and syntactic adaptability,

as the three phrases cited above demonstrate. The traditional, diachronic character of the verse lies in the placement of s[c]ynscapa and its recurrences (all but once)^[33] in the C-Type metrical pattern; synchronically these features make available to the composing poet a good deal of grammatical and syntactic latitude in the opening syllables of the verse.

A further level of structure then derives from this consistent word-type placement in verse-final position. We can describe two formulaic systems involving the same root:

#1: [x] []-sceaða-
#2: [] []-sceaða-

where x is defined as one or more function words and y is a compounding element bearing the alliteration, and a is an adjective that alliterates with y . System #1 can occur in either half-line and demonstrates great syntactic variability because of the many possibilities for function words (thus a direct object, relative clause, prepositional phrase, and other elements are observed). System #2 is restricted to the first half-line because of the double alliteration it involves, but it in turn yields a number of related phrases, such as

The first of the two systems takes a C-Type metrical configuration and occurs twenty-seven times in the corpus, whereas the second is usually a DI-Type in its sixteen instances;^[34] more significantly, there are no occurrences of compounded sceaða that do not fall into one of these two patterns. At the root of both systems is the word-type placement at verse-end.

708a. ac he wæccende ("but he, watching"). The most well-defined system that could be posited for this and ostensibly related lines is [x] [y] wæccende , where x is a function word and y a pronoun subject.^[35] Once again, this phrase seems as much an example of a traditional rule—namely, that the metrically most

complex elements seek verse-end except in E-Type arrangements—as the product of a formulaic system.^[36]

708b. wrapum on andan ("fiercely angry"). Here the system is [] on andan , where x is defined as the alliterating element and can be either an adverb (including a dative adverb) or a noun in the dative plural. In the line as a whole we may see the wæccende phrase as a response to the narrative surroundings, which at this point are emphasizing that all of the men slept, with the sole exception of Beowulf. Matched to that wæccende is wrapum , the substitutable word in the second-verse system. While its halves are perhaps less closely linked than some of the other half-line pairs we have examined, the line as a whole does follow the favored metrical formula and provides an alliterative bridge to the next line, to which it is also linked semantically.

709a. bad bolgenmod ("he awaited, enraged in heart"). The best hypothesis for a half-line system would be [] bolgenmod , where x is either a verb (alliterating) or function words.^[37] But since this configuration lacks a clear essential idea ("[x] enraged in heart" is hardly a well-defined concept) as well as consistent alliteration and syntax, it is difficult to see how it could have been a functional element in the poet's and tradition's working idiom. What these few phrases do have in common is the placement of the metrically extensive and complex element bolgenmod , which, like wæccende and s[c]ynscaéa in preceding lines, occurs at verse-end, this time however in a D₂ pattern.^[38]

As for bad , it behaves similarly to com as described above (line 702b). That is, this word-type—a third-singular preterite verb—seeks either the beginning or the end of a verse. In the case of bad , most instances occur at verse-end (23 of 28). Comparative analyses of swæf and bær , both examples of the same word-type, confirm this explanation and show that if such a word-type is used as an alliterating item, it must take a D configuration (in the first half-line) or a B configuration (in either verse). If non-alliterating, it will occupy the final position in an E pattern in the second half-line. These patterns of occurrence are thus due not to the formulaic system involving a particular word, or for that matter even to the behavior of a given single word, but rather to the behavior of a word-type.

Even if we cannot demonstrate a clear debt to any one formulaic system, then, line 709a certainly merits being considered a traditional element. Not only is the anger signaled by the root bolgen- conventionally associated with the hero's fight against the monster in Beowulf ,^[39] but, more importantly,

traditional rules of word-type placement underlie the construction of the verse.

709b. beadwa geæinges ("the battle's result"). A number of factors converge to form this half-line, among them the system [] geæinges , where x is a noun in the genitive plural (though translated here as singular) beating the alliteration. But this is not the only aspect of structure in this case; we must also cite an alliterative collocation with bolgen (cf. Bwf 1539) and, perhaps even more suggestively, an association with the verb bidan ("wait") that seems not to depend on alliteration. The comparative evidence makes the latter point:

While the passage from Elene links the roots in contiguous verses, in the Beowulf passage they occur in separate lines. What apparently underlies all three cases, then, is a sequence bidan ... [] gepinges , which the poet may or may not sort into a single line. And although we find a fairly well focused formulaic system in 709b, we miss the complexity of the overall traditional structure if we end the investigation without taking account of the collocative association with bolgen and the larger pattern involving bidan .

710a. Da com of more ("Then came out of the moor"). Like 702b, this half-line answers rules of word-type placement better than the hypothesis of an underlying formulaic system. One difference between the two phrases, however, is the prepositional complement that follows com ; in the former example, that phrase could and did stand by itself as a half-line, while in 710a of more is not long enough metrically to do so.

710b. under misthleopum ("under the misty cliffs"). A system such as under [] would have no useful meaning or function, so we must posit [x] misthleop- or, more genetically,

[x] []-hleop- ,

where x is defined as a preposition (usually under ) and y is a compounding word that supplies the alliteration. To this pattern we may add the collocation between mor- and mist- as further evidence for traditional structure.^[40] The variability of the system in the second half-line allows the poet to fill out the opening verse, itself primarily a product of traditional placement rules, with a properly alliterating element. It should also be pointed out that the second-verse system is conventionally associated in the poetic tradition with

danger and death, that is, with actions performed in liminal areas; this conceptual link helps to account for the system's presence and function in this passage.^[41] Once again, then, a line exhibits both half-line and whole-line structure.

711a. Grendel gongan ("Grendel going"). After 710b, which serves in effect as an alliterative bridge to this line (but which in the process brings forth some traditional associations, as we have seen), 711a completes the pattern begun with com by supplying the dependent infinitive, just as scriðan did in 703a. Indeed, we shall encounter the same pattern again with the Com of 720a and the infinitive siðian of 720b. This com ... [infinitive] sequence is a core phrase which, like the bidan ... [] gepinges sequence examined above, is not necessarily expressed in a simple whole-line form. Nor is it a unique phrase or one peculiar to Beowulf , as the following quotations illustrate:

This traditional pattern thus furnishes another example of a reasonably common phenomenon in Old English poetic diction: an association of lexical and syntactic items that is functionally independent of single-verse or single-line prosody. It may be realized within one verse or line, but its coherence as a traditional element is not dependent on such a realization. But this is nothing more than our study of the alliterative prosody would lead us to expect—a sequence or pattern of words larger than the single line.

The importance of this pattern becomes obvious when, after examination, we find that (1) no other cases of verse-initial Grendel exist (so a system based on that element is out of the question) and (2) any system involving gangan as the constant element would be of such generic definition as to be of little use in composition. Nonetheless, the poet does manage to complete the "pending" com with gangan , to match that completion with Grendel, and to do so in a metrical A-verse.

712a. mynte se manscaða ("the wicked ravager intended"). There are only three instances of the particular form mynte in the poetic corpus; all of them occur in Beowulf , and the other forms of the same verb, being different word-

types, behave quite differently. Within this small sample, we can posit a system mynte [], where x represents the alliterating subject of the clause, on the basis of mynte se mæra (Bwf 762a). The second part of 712a, however, can also serve as a whole verse by itself, on the model of Bwf 707a, se s[c]ynscapa , and so the situation turns out to be analogous to 702b, in which on wanre niht served to fill out the Com phrase even though metrically it could stand by itself. From the point of view of larger patterns, we should also note that mynte seeks completion in a dependent infinitive and, as we shall see, 712b provides an alliterative bridge that leads to that infinitive. Synchronically, then, the poet has matched the mynte pattern, which involves a verse-initial alliterating item and eventual completion in an infinitive, with the same formulaic system that produces 707a.

712b. manna cynnes ("of mankind"). This is a classically defined formula, a phrase that recurs nineteen times without variation in other Old English poems. As such, it provides a phraseologically sturdy bridge from the essential and incipient action of mynte in 712a to its completion in 713a. While not a "tag," this phrase—perhaps best understood as a single word^[42] —offers the poet a metrical element that is highly adaptable contextually and therefore useful in any number of compositional situations. As the mynte . . . besyrwan idea takes shape over two lines, this formula serves as a stable configuration against which to balance the alliteration.

713b. in sele pam hean ("in that high hall"). As opposed to the first half-line, this verse seems to be an idiolectal formulaic system (virtually a formula), that is, a phrase used exclusively by the Beowulf poet. Its pattern is [in/on/to] sele pam hean ,^[43] and it occurs three additional times in the poem and nowhere else. Like manna cynnes in the preceding line, this phrase serves as a functional, adaptable element that fills out the alliteration and, while not adding necessary or crucially important action to the passage, specifies the locale of that action. AS in other lines considered above, the best way to understand 713b is to consider it in the context of the two-line sequence 712-13, wherein the poet solves the compositional problem of metrically and traditionally rendering the idea mynte . . . besyrwan (itself a traditional syntactic pattern) by an assortment

of formulas, formulaic systems, alliterative bridges, and placement rules. This spectrum of diction, rather than a simple series of half-line systems, constitutes his traditional phraseology.

714a. Wod under wolcnum ("He advanced under the clouds"). Moving on to this much-discussed verse, we must first decide whether in the system postulated as underlying it and its relatives, [] under wolcnum , we should limit the alliterating element x to a verb form (as does Magoun [1953] 1968, 109) or also allow the substitution of adjectives and nouns and even the inclusion of semantically similar but formulaically unrelated phrases (as does Riedinger

1985, 297-304).^[44] In order to preserve the integrity of the phraseological unit, I would advocate drawing the line at the formulaically related phrases. But there is another question to ask as well: in short, just what is the usefulness of the system assumed to underlie 714a? Is the essential idea merely "[something] under the clouds"? And if so, how does it compare as a compositional element with phrases like [] besyrwan (e.g., 713a), where a definite idea is presented in a relatively well focused way?

Perhaps it would be better to begin by pointing out that what under wolcnum represents at a minimal level is a consistent occurrence of a phrase in the same metrical position: at verse-end in an A-Type pattern. The fact of its consistency as the core "word" in this phrase allows the tremendous semantic and syntactic flexibility that the multiform as a whole enjoys. If the pattern is employed in the opening half-line, as happens thirty-six of fifty times in the corpus, the initial element will almost always alliterate (thirty-four times) and the second half-line will of course join in the same alliteration in w .^[45] We may add to these specific observations the much more general fact that the word-type represented by under wolcnum —defined as [prep] [], where x is a noun that bears the alliteration and is the object of the preposition—seeks consistent placement either as a complement like of more (710a) to fill out a verse after a certain other kind of word-type or, if it is metrically extensive enough, as a verse in itself.

This pattern differs from the Com phrases in that it includes double alliteration, and therefore alliteration on the initial element. Its fundamental shape and morphology, although immediately attributable to a formulaic system, are best understood as deriving from traditional rules for placement of word-types. Like some other phrases discussed above, under wolcnum seeks verse-end and, from a synchronic point of view, opens up the substitutable

first position in the opening half-line and sets the alliteration for the line as a whole. Since, as the comparative data indicate, this core recurrence has produced a pattern to which so many different metrical solutions have accrued over time, the pattern becomes a very useful compositional device, and also one heavily laden with associative meaning.^[46]

716b. Ne wæs pæt form sið ("Nor was that the first time"). With the onset of the next unit of thought we encounter a true formulaic system, [x] forms sið , where x is defined as a group of function words, usually including some form of the verb "to be," and the half-line takes a B-Type metrical configuration. This phrase answers the classical definition of the system precisely, with its stable core and adaptable first section. Because of this adaptability, it is available synchronically for syntactic adjustment; for example, it can begin a new sentence or paratactically continue an already-started one.

Apart from the four additional occurrences of this phrase (three of them in Beowulf ) we also find a semantically related formula, forman siðe ("on the first time or occasion"), which when compared to the system underlying 716b well illustrates the tradition-dependent character of Old English poetic diction. The second phrase, forman siðe , takes a different metrical form (an A-Type), never varies, and lacks the special syntactic flexibility of the system. The traditional word-type rule under which forms sið seeks verse-end (the word-type s s ^[*][*] ) cannot apply to forman sið because of its different stress pattern (s s ^[*] s s ^[*] ). Thus traditional rules, based on metrical strictures, are seen to be more fundamental than the lexical elements that make up an expression.

717b. ham gesohte ("he sought the home"). Evidence from the poetic corpus might lead us to consider this phrase a formula, since there are two exact repetitions:

But if we choose this route, we miss the complexity and truly traditional character of 717b. For of the thirty-three instances of the verb form gesohte , fully thirty occur in A-Type metrical patterns of the sort [] gesohte , where x represents a noun bearing the alliteration; furthermore, seventeen of these thirty have x as the direct object of gesohte . Or we may turn to the possibility ham [x], where x represents the verb taking ham as a direct object, a well-attested pattern that can involve infinitives as well as finite verb forms.

All of these levels of structure are, however, built on the foundation of traditional word-type placement rules, here chiefly the stricture that calls for

verbs prefixed by ge- , and thus of the generic form ge -ROOT^[*] + ending^[*] , to tend strongly toward the end of an A-Type metrical pattern. The element to which such a word-type is joined is, of course, a stressed monosyllable or the equivalent^[47] —very often, as we have seen, the direct object of that verb form. Thus the stressed monosyllable or its equivalent comes to bear the alliteration in what is exclusively a second-verse phrase pattern. From a synchronic viewpoint, this and phrases like it provide the poet with a flexible way of handling prefixed verbs, which otherwise, especially in the first verse, can cause rather unwieldy metrical configurations. What is more, this half-line offers a good example of why the "formulaic system" concept is often quite useful and descriptive, but at the same time fundamentally inadequate to the task of demonstrating all aspects of the traditional character of a phrase. Whatever system we choose to nominate as the mold for 717b, it will conceal as much as it reveals about the essential structure of this and related half-lines; for an accurate measure of that structure we must return to the most basic traditional rules and derive levels of patterning from that point.^[48]

718a. næfre he on aldordagum ("never he in earlier days"). The fact that aldordagum is a hapax legomenon in the Old English poetic corpus may obscure the fact that this half-line belongs to one of the richest, most generative (and therefore most useful and echoic) systems in the poetry. We may define the core of the system as [x] []-dagum , where x is defined as a preposition and y is a compounding element that bears the alliteration.^[49] This phrase, one of the first to be studied as a compositional element,^[50] proves extremely plastic: it can be employed as a half-line by itself in a wide variety of different situations (in the form schematized above), or it can "add on" function words before the preposition and thus take on another whole selection of syntactic functions. Not only does it recur in both a "core" formation filling a verse and as the stable end of a longer phrase (as in 718a), but it also can take on special functions in either identity. As a whole-verse pattern, the system is perhaps most familiar to us as part of the heroic proem sequence common to a number of Anglo-Saxon poems; for example,

As part of an enlarged pattern, we can cite the following idiolectal usage:

In a real sense the [prep.][]-dagum phrase acts as a "word" in the diction, open to additive morphology but not to internal re-arrangement. Few phrases in Old English poetry are as self-contained as this one.^[51]

As integral a unit as this system is, however, we should also point out that traditional rules do supersede, or rather govern, its formation. In cases involving the element [y]-daguma that do not fit this formulaic system, we observe that the element still regularly seeks verse-end, in these latter cases almost always to constitute a variation on the DI-Type pattern (e.g., Gen 1072a: frod fyrndagum ). What we have, then, is a word-type, specifically [y]-dagum , as the verse-final element, which then takes on two separate identities in the tradition—one as the system underlying 718a and so many additional verses and the other underlying a DI-Type verse where the "word" customarily acts as a dative complement.

What the poet manages to accomplish in 718a, from this perspective, is to state the condition "never" (næfre ) in traditional formulaic language, using an expandable system built on the core [prep.][]-dagum and a compounding element that, while it yields a hapax legomenon, matches the alliteration of the second-verse formula.

718b. ær ne sipðan ("before or since"). Although this half-line recurs only once in the corpus, we can safely call it traditional. For besides satisfying the minimum criteria for a formula, it is related to a whole system of diction using the same words with slight variations in form and meaning. We have, for example, eighteen occurrences of the patterns sið [x] ær and ær [x] sið , where in both cases x stands for and/ond, ne ("nor"), or oppe ("or"). The difference, of course, lies in word-type: while 718b (and Chr 39b) fill a whole verse, the two metrically shorter patterns cited above must comprise only part of a half-line,^[52] most often a B-Type in place of the A-Type in the half-line under consideration.

720a. Com pa to recede ("Came then to the building"). This phrase shows the same general structure as 702b and 710a, as discussed above, although it should be pointed out that within this traditional pattern the poet is able to vary the complements to the action of com in a way that increases the tension of the approach: Grendel is seen first "in the dark night," then advancing "out of the moor," and finally coming "to the building."

720b. rinc siðian ("the warrior journeying"). Here the poet completes the com . . . [infinitive] pattern within a single line by taking advantage of the alliterative collocation between reced and rinc (cf. Bwf 412 and 728). Although we might posit for this half-line the system [] siðan , where x is defined as an alliterating noun with the semantic value of "man, warrior," some other analogs illustrate the more fundamental structure of the whole line:

The pattern com . . . siðian , a more focused version of the com . . . [infinitive] phrase we have encountered twice before, provides the backbone for line 720 and these three additional examples on a whole-line basis, and each verse is adjusted for alliteration with complements—in the case of 720, with a collocative pair.

721a. dreamum bedæled ("bereft of joys"). At first sight we might interpret this half-line as simply an alliterative bridge, a filler that harmonizes syntactically with 720a and allows the poet to proceed on to the next important action. To be sure, it is a highly formulaic phrase, a variation on the system [] bedæled , where x is a noun bearing the alliteration and designating some positive communal value.^[53] In fact, it may well be best to consider dreamum bedæled a true formula on its own account, since the essential idea of loss of social context is so powerfully and memorably expressed.

But it is a measure of the poet's singular art that, far from being the slave of his traditional diction, he molds that diction toward aesthetic ends. While 721a is clearly meant to serve as a bridge to Duru (and thus to Grendel's bursting into Heorot) in the second verse, dreamum bedæled is hardly a throw-

away; by metonymic reference to the wordhoard, the poet deftly brings out another aspect of Grendel—that of Exile.^[54]

721b. Duru sona onarn ("The door immediately sprang open"). There is only one other instance of onarn in the poetic corpus, and it turns out to be an exact repetition of this phrase (And 999b). On this basis we can certainly call 721b a formula, but stopping the analysis at that point will obscure the traditional depth of the phrase and therefore its metonymic meaning. To pursue the matter in proper perspective, we must quote the passage from Andreas in which the formula appears (996b-1003; quoted from Brooks 1961):

Da se halga gebæd
bilwytne fæder, breostgehygdum
herede on hehðo heofoncyninges god,
dryhten dem[de]. Duru sons onarn
purh han[d]hrine  haliges gastes,
ond pær in eode , elnes gemyndig,
hæle  hildedeor. Hæðene  swæfon ,
dreore druncne, deaæwang rudon.

Then the holy one prayed to
the gentle Father, in his inmost thoughts
praised the Heaven-king's goodness on high,
glorified the Lord.  The door immediately sprang open
through the hand-touch  of the holy spirit,
and in there he went , mindful of valor,
the battle-brave hero . The heathens slept ,
drunk in blood, they reddened the death-place.

The italicized words and elements of compounds also occur as a group or cluster in the passage under examination from Beowulf . In addition, we may note that this cluster has a narrative basis in the action of a "warrior" invading a sanctuary where his foes are sleeping and wreaking havoc among them. In the case of Beowulf , that warrior is Grendel the Exile; in Andreas it is St. Andrew bringing revenge on the Mermedonians. Thus this duster, a traditional unit larger than the verse or line, assists in the composition of this section of the poem through the agency of associated words, and we must likewise take account of the traditional meaning of the cluster in reading both passages.^[55]

722a. fyrbendum fæst ("[held] fast by fire-forged bonds"). Although fyrbend proves a hapax legomenon in the poetic corpus, we can describe a common system underlying this verse, []-bendum fæst , where x represents a compounding element bearing the alliteration. This compound takes a wide variety of forms, from the straightforward irenbendum ("iron bonds," Bwf 774b and 998b) to the more abstract hygebendum ("mind-bonds," Bwf 1878b). It occurs once in Guthlac and six times in Beowulf , with a total of seven different compounding elements. And yet this is not the end of the story, for in addition to these instances we find two related structures. First, the compound [x]-bendum can, if the initial element is metrically extensive enough, occupy a whole verse and not just part of one; thus irenbendum can either join fæst in the system under discussion (Bwf 998b) or stand by itself (Bwf 774b). Second, there are numerous compounds, many of them closely synonymous to [x]-bend- , that behave in very similar ways, some of them even participating in analogous systems with fæst . Here are two examples:

With these and other examples, two further perspectives emerge. While this system apparently has a life of its own in the poetic tradition (chiefly in Beowulf ), its first clement may be employed alone—outside the context of the formulaic system. As we saw above in the case of certain prepositional phrases that acted as complements to the com series, some traditional elements do have this ability to play different versificational roles, depending on the particular situation. This sort of multiple function is impossible in symbioses of prosody and phraseology such as those associated with the Homeric Greek hexameter and Serbo-Croatian deseterac . Interestingly, however, as we shall see in chapter 10, this twofold function is paralleled at the narrative level by story-patterns that can serve either as simplex sequences for an entire poem or as additive elements that can be combined to form longer and more complex poems.

Another insight on the compositional process is afforded by the behavior of words outside of the specific []-bendum system which nonetheless match this element in word-type. The similarity in formulaic morphology and the consistent behavior of fæst as a monosyllabic adjective seeking verse-end in this and many other situations^[56] lead us to understand that the system observed in its many manifestations is also, and fundamentally, a reflection of traditional rules. The system is a highly focused structure built on these rules, a pattern that has apparently proven useful and has developed its own identity in the

poetry. While it would be misleading not to recognize 722a as a formulaic system, it would be equally wrong not to place this pattern in the context of the rules under which it has come into prominence and under which it continues to function.^[57]

722b. sypðan he hire folmum [æth]ran ("after he touched it with his hands"). This is perhaps the most idiosyncratic verse in the passage under examination, for there is nothing in the poetic corpus that could be summoned as evidence for any kind of system. We may, however, juxtapose to that lack of apparent structure two compositional features. The first is specific to this passage and to the cluster that helps to define its structure and meaning: the inclusion of -hran ("touched"), one of the words that we noted was also associated with St. Andrew's forced entry, and folmum ("with hands"), whose synonym hand(- ) was also employed in the Andreas scene (handhrine , "hand-touch"; 1000a). The second feature is more general and typical of many verses in addition to 722b. This amounts to the simple fact that "acephalic" metrical types—that is, Sievers B and C—are much more open to variation than are other types because of their lack of a stressed element in initial position. If one or more function words can occupy first position in a metrical type, with no real restriction on semantics or grammar, then such types will of course exhibit wide variation even if the following stressed elements show regular occurrence. This is an elementary but very significant characteristic; what is more, it provides yet another reason why all half-line phrases cannot be judged equal and why, as a result, we must expect a spectrum of traditional phraseological forms.^[58]

To sum up, the central idea of these few lines can be abstracted as "The door sprang open at Grendel's touch," which is expressed by 721b and 722b using a formula and duster-words traditionally appropriate to the poetic task. Line 722a, a relatively common formulaic system with its roots in traditional rules, acts as an alliterative bridge, not absolutely necessary to the expression of the central action but, thanks to the poet's skill, an artful detail that amplifies Grendel's entry. Thus the sequence 721b-22 contains a formula, a system, a half-line we cannot ascribe to a system, and part of a traditional cluster, and all phraseological and narrative units are meshed to form a finely crafted and

resonant description; such is the rich inheritance of tradition and such is the art of the Beowulf poet.

723b. ða (he ge)bolgen wæs ("since he was enraged"). This verse is clearly a formula, as four verbatim or very nearly verbatim usages illustrate (Gen 54b; Bwf 1539b, 2220b, 2550b). In each case the formula is associated with a major battle, whether with God's fight against Satan and his company in Genesis , Beowulf's battle against Grendel's mother, the dragon's anger over the stolen cup, or Beowulf's fight with the dragon. In other words, this phrase seems to be associated, particularly in Beowulf , with the idea and expression of battle; and we would not be far wrong if we considered it thematically associated with the monster-fight scene in Beowulf , as a metonymic signal of the magnitude and intensity of the battle. To be gebolgen in this way is to possess battle fury, and this phrase encodes that special anger.

When we add to these observations the evidence for a collocation between the root of the verb in 723a and the highly charged word gebolgen (not to say its ritual expression in this formula),^[59] we perceive another example of a half-line formula which nevertheless has traditional associations outside the single verse. The overall picture then becomes one of the fundamental idea (-bræd... ða he gebolgen wæs , "drew or swung open... since he was enraged"), with the []-hydig compound filling out the first verse under the aegis of traditional rules. Although one half-line is a true formula and the other not demonstrably formulaic, the line as a whole is once again thoroughly traditional.

724a. recedes mupan ("the building's mouth"). Since this verse is metaphorical, denoting as it does the door of the hall, neither the general pattern recedes [x] nor the literal pattern [] muð can serve a systemic function; the first possibilitity is too generic and lacks an essential idea, while the second would need further definition of muð in the special metaphorical sense.^[60] Thus only two half-lines in the poetic corpus present themselves as possible comparands:

The first phrase describes the opening of Noah's ark, and the second serves as a variation on duru in the preceding half-line. Both phrases could in a sense be said to participate in a formulaic system with 724a were it not for the fact

that they do not match it metrically. The verse from Genesis also uses a synonym for reced rather than the word itself.

Still, we may be able to show more of a connection between the Maxims and Beowulf phrases than is immediately apparent. Both verses amount to virtually the same expression connoting "door" and, apart from metrical differences and the word rum , are virtually identical. Consider for a moment the poet's essential idea, in this case one that is highly focused, metaphorical, and ostensibly traditional, and the continuing synchronic problem of metrical versification that he faces. Both phrases are successful renderings of that idea in that they conform to a permitted metrical configuration, and both use the core phrase recedes muð- . The major difference between the two is that Mx2 37a assumes a strong masculine declension for the main word (nom. mup ) while Bwf 724a assumes a weak masculine declension (nom. muða ); further evidence for both possibilities in the corpus indicates what seems to be a true ambivalence in the lexicon and therefore in the diction. For the Beowulf poet, the weak form yields a metrical verse in the accusative singular, recedes mupan (724a), while the strong form would result in a short, unmetrical half-line: *recedes mup . This is precisely the problem faced and solved by the Maxims poet, who understands mup as a strong form and adds the adjective rum to the core phrase recedes mup to give a metrical phrase, rum recedes mup .^[61] Indeed, although these two examples cannot be shown to proceed out of the same formulaic system, they are without doubt related traditional phrases. And the example of their less-than-obvious relation sheds further light on the morphology of Old English poetic diction.

726a. code yrremod ("[he] went angry in spirit"). Two patterns present themselves as possible systems: [x] []-mod,^[62] where x is a verb and y is a compounding element usually bearing the alliteration (x also alliterates in the more common first-verse pattern), and eode [], where z is an alliterating adjective modifying the implied subject of eode . If we arc willing to term the first possibility a bona fide system in spite of its lack of semantic focus and essential idea, there arc numerous examples in the poetic corpus to support interpreting 726a as a product of that system. Meanwhile, phrases like eode unforht (Exo 335a) and eode ellenrof (Bwf 358a) can be drawn on to support the second hypothesis.

Since the first pattern is so generic, and since eode participates in so many different kinds of phrases (so that it seems not to follow a specific traditional rule), the second hypothesis is a better description of the system underlying 726a. But we should note that []-mod , in obeying the traditional rule oft he

metrically more complex element seeking verse-end, constitutes the more fundamental structure of the phrase; in effect, this word placement creates the sequence eode [].

726b. him of eagum stod ("from his eyes stood out"). Phrases like pæt he on botme stod (XSt 718b) or se ðe on greote stod (And 254b) may encourage positing a system [x] [prep^[*] phrase] stod , where x represents one or more function words and the prepositional phrase, with the object alliterating, is defined only grammatically. But since such a pattern is much too generic to serve the composing poet, we would do better to point out that stod follows the traditional rule of monosyllabic preterite verb seeking final position and to see eode yrremod as an alliterative bridge (and partial variation of the preceding half-line, feond treddode ) that allows the poet to proceed smoothly from one action to the next. And while the phrase of eagum and the idea it embodies may seem ornamental, in fact the detail of a mysterious light marks all three monster-fights in Beowulf .^[63] Both the traditional phraseological rule and the "deep structure" of the monster-fight theme, then, help to shape the second half-line.

727a. ligge gelicost ("most like fire"). Underlying this half-line is the common system [] gelicost , where x is a noun in the dative singular case bearing the alliteration and the phrase as a whole constitutes an institutionalized way to construct a simile. Ten other instances occur in the poetic corpus, in a variety of poems, so we may safely assume that this system is tradition-wide rather than idiolectal. To these general comments may be added the more specific observation that this phraseological element, like many others we have examined, exists both as a unit in itself and as part of a larger pattern. An example is furnished by lines drawn from the sea-voyaging passages in Andreas and Beowulf :^[64]

The Beowulfian version of the simile also provides an alliterative bridge to the leoht of the following half-line, and thus adds a synchronic function to its traditional, diachronic identity.

728a. Geseah he in recede ("In the building he saw"). Although we may posit a system Geseah [x], where x is a prepositional phrase, this pattern has no more validity as a compositional construct than does the similar pattern

initially proposed for the half-lines involving com . More to the point is the consistent behavior of the verb geseah , which in eighty occurrences in the poetic corpus seeks verse-initial position thirty times and verse-final position forty-nine times.^[65] This verb form can occupy either slot in either verse and very seldom takes part in the alliterative structure of the line. To put the same matter another way, this word—and word-type—can be and is used outside formulaic systems and the alliterative complications they regularly entail. In so employing this word-type, the poet is beginning a construction that, again like the Com pattern, demands closure in an infinitive;^[66] he will bridge the gap between Geseah and that infinitive with an alliterative collocation and a formulaic system.

728b. rinca manige ("many warriors"). The formulaic system is rinca [x ], where x is a noun or adjective that specifies the group denoted by the genitive plural rinca . The alliterative collocation consists of the pair rinc and reced , which the poet has matched twice before in Beowulf , at lines 412 and 720. We begin to see the whole-line pattern as well as a continuation to the next line through the central action of Geseah ... swefan . The infinitive swefan , of course, will participate in its own alliterative and formulaic series in addition to providing closure for Geseah .

729a. swefan sibbegedriht ("the band of kinsmen sleeping"). Just as the relatively independent word and word-type geseah took on an alliterative series with a built-in (that is, traditional) collocation in line 728, so the infinitive swefan not only completes the action of Geseah but also takes on an analogous series of its own. For while we find some support for positing a system [] sibbegedriht , where x is an infinitive, that same evidence (Bwf 387) also argues for a more extensive pattern:

seon sibbegdriht samod ælgædere
to see the band of kinsmen all together

To call this correspondence a half-line system overlooks both the second verse and the whole-line pattern. In addition, two other examples, both of them from outside Beowulf , demonstrate the central significance of the collocation (as opposed to possible systems) and the pan-traditional character of the alliterative link:

The line from Exodus shows that the relationship between the alliterating elements is not dependent on the first-verse system, while the example from Guthlac illustrates independence of the collocation from the second-verse formula.^[67]

729b. samod ætgædere ("all together"). This phrase is of course a common formula in the Old English tradition, occurring seventeen times in a variety of poems. Once again, then, we have an instance of a unit—this time a half-line formula—that has a life of its own in the tradition and yet can also combine with other phraseological elements to produce a larger pattern that has its own identity. In the other direction, we note that ætgædere always occurs in either. verse- or line-final position,^[68] as its metrically extensive word-type would lead us to expect. Thus the formula observed in 729b is erected on the basis of traditional rules.

To sum up, the central action begun in 728a by Geseah and completed by swefan in 729a is filled out in each line by an alliterative collocation, the latter of these also assuming the form of an idiolectal whole-line system in Beowulf . Evidence from outside Beowulf helps us to see what the poet's contribution has been to this series, and once more we encounter a hierarchy of structure, from a two-line series through collocations, a whole-line system, a formulaic system, a formula, and traditional rules. It is the poet's achievement to use the materials of his tradition to convey the central action with such power and resonance.

730a. magorinca heap ("a troop of young warriors"). A variation on rinca manige and sibbegedriht , this half-line may be assigned to a system [] heap , where x is an alliterating noun denoting "warriors, troops" in the genitive plural (partitive construction). This pattern can stand by itself if the noun is sufficiently extensive metrically and can form a part of another system if it is not. While not as well defined as many formulaic systems, the [] heap phrase does express a focused essential idea and proves useful in many different situations. It is also a more specific example of the general pattern of [partitive gen. pl.] [number/group noun] observed throughout the poetry.^[69]

Quantitative Analysis

Table 21 will serve as a reference point for the foregoing discussion, recording the status of each half-line in the passage (whether formula, formulaic system, or neither) and the traditional rules that apply to each verse (abbreviations are explained in the following discussion). Two caveats are necessary: first, my judgment of whether a phrase is a formula or a system is developed from the analysis presented selectively above and may not agree at all points with the definitions of other investigators;^[70] second, designating a phrase as belonging to one or another category of diction hardly explains its entire traditional identity and context. As a practical matter, if a phrase can be termed a formula on the basis of available evidence, I have not indicated in this table whether we can also call it formulaic.^[71] Nonetheless, it may be instructive to compare the figures we derive with the perspective offered by traditional rules.

In purely quantitative terms, the passage consists of 19.5 percent formulas, 51.8 percent formulaic systems, and 28.6 percent verses that cannot be satisfactorily explained as belonging to either category. As mentioned in chapters 1 and 2 and earlier in this chapter, I cannot accept the premise that such quantitative analysis determines the oral or written provenance of this or any other passage, first because of the criteria of tradition-dependence and text-dependence that are two of the guiding principles of my approach to the study of comparative oral and oral-derived epic. On the basis of the line-by-line investigation and its results, we may add another reason to question the density test: if all systems are not equal in complexity and usefulness, and further if the system hypothesis does not fully explain the traditional character of many of the half-lines, how then can we expect it to serve as an unambiguous measure of a given text's orality?

What should be emphasized about table 21 is the absolute consistency of traditional rules; not a single verse is left unexplained by this more fundamental level of structure, while the formula/formulaic system model leaves more than a quarter of the passage unexplained as traditional phraseology. We shall finish our discussion with a summary and codification of traditional rules, the foundation on which a viable aesthetics can in the future be erected.

Traditional Rules: A Summary

Earlier in this chapter we attempted to predict the kinds of phraseological structures that might be expected to form under the aegis of the idiosyncratic

(Table continued on next page)

Old English alliterative meter. As expected, our subsequent analysis revealed not one but a variety of types of diction: alongside the classical half-line phrase stand single words, whole-line patterns, multi-line patterns, collocations, clusters, and themes. Likewise, even within these different categories not all members are equivalent; some formulaic systems are more variable than others, some larger patterns more restrictive than others, and so on. Then too, we found a number of elements that had more than one compositional identity, as when a phrase could serve as a half-line unit by itself or join with another word or words to make up a verse. In most cases, even if we could realistically posit a formulaic system underlying a given phrase, simple reporting of that proposed system proved insufficient to a complete understanding of the phrase as an element of traditional phraseology. In short, the synchronic model of half-line substitution systems, while a valuable first step in assessing the character of Old English poetic diction, does not go far enough; the simplification it offers is purchased at the price of incomplete analysis and therefore of an unsound basis for aesthetic investigation. At the same time, the wide spectrum of traditional forms uncovered in the passage from Beowulf analyzed above, while fulfilling what was predicted about the diction, begs the question of how the poet and tradition could have managed to handle so many apparently diverse forms concurrently.

The first response to this question must be a general one, but one that is not often enough appreciated by those involved in the analysis and explanation of oral traditional phraseology. Most simply put, the diction we encounter in

Beowulf and Old English poetry—and in the oral epics of Homer and the Yugoslav guslar —is by its very nature not a set of substitution systems. Even if we were to hypothesize (quite unrealistically) that at some point in the distant past of a poetic tradition the Kunstsprache was made up entirely of equivalent elements, so that one formulaic system, for example, was always and everywhere equal to all others, the dimension of diachrony that separates us from that Ur-diction would demand a reinterpretation. Over time, some systems and other structures will develop differently from others, and this inherently uneven evolution must produce some structures that permit extensive variation, some that permit none, and many that fall somewhere in between these two poles. Likewise, the same evolution (and the natural-selection model may not be entirely inaccurate for the process) must yield structures of different "sizes" and complexity, with certain phraseological relationships "growing" from one size to another, perhaps retaining the original form, but perhaps not. And we must also introduce into this same equation the dimension of different singers versus the local tradition and the tradition as a whole—what were called idiolect, dialect, and language in relation to the Serbo-Croatian poems. Although we lack the background information and textual records to apply this kind of discrimination to Beowulf , we have in fact noticed some structures that seem, on the basis of available information, to be idiolectal.

What the dimension of diachrony indicates is thus another level of complexity in a diction, another reason to analyze carefully what is really there instead of settling for a convenient and simple model that partially explains a healthy percentage of the cases we encounter. Of course, evolution of traditional forms will be guided both by the conservatism of tradition in general and by the prosodic filter of the poetry in particular, but we should not expect the text we are interpreting to submit tamely to a synchronic analysis. The language of Beowulf is a highly complex and resonant instrument, one that carries with it enormous internal resources of associative meaning and yet is also pliable enough to respond to the poet's individual craft, one that harmonizes well with Yeats's notion of the poet as "both finger and clay."

While we cannot reduce the complexity of this spectrum of traditional phraseology and still hope to provide a firm foundation for aesthetic interpretation, we can nonetheless rationalize the diversity of many forms to a single set of traditional rules that, like the corresponding inventory in the poetic idioms of Serbo-Croatian and ancient Greek epic, are neither as situation-specific as those governing the formula and system nor as general as metrical constraints. Lying in between these two extremes and built, as indicated above, on the logic of the tradition-dependent prosody, traditional rules are inscribed in every half-line and line, whatever category of structure that half-line or line falls into. They serve both as regulators of incoming phraseology and as continuing supports for the morphology of that diction which the poet is accustomed to employing. In a word, these rules supersede the synchronic

dimension, for they are not phrase-specific or even category-specific; rather, they connect all phraseological patterns—no matter how small or large, how simple or elaborate—by serving as the group criteria of what is (and becomes) traditional.

The most far-reaching of these rules, word-type placement , furnishes a prime example. This set of constraints governs the recurrence of words not on the bases usually prescribed for the formula and system (those of lexicon and semantics "under the same metrical conditions") but on the basis of the word's metrical and grammatical type. Indeed, this is exactly what we would expect in a prosody that emphasizes not the syllable and the colon but rather the sequence of stresses. For example, we found during the earlier analysis that monosyllabic preterite verbs consistently occupied first or last position in a verse; this phenomenon was a result not of any one verb or (primarily) of any particular formulaic system, but of a rule that applies to a metrical-grammatical category of words. To summarize, I list below the various kinds of word-type placement encountered in the passage from Beowulf :^[72]

1. Monosyllabic preterite verbs are either verse-initial or verse-final

2. Metrically more extensive word-types seek verse-end (suspended by rule #1)

3. Attributive-specifier : any sequence involving a specifying word and a dependent attributive (partitive genitive, dative of respect, participle plus instrumental, etc.) will occur in the order attributive followed by specifier

4. Finite verbs in general seek verse-final position

5. Prepositional phrases seek verse-final position (subject to the priority of rules #1-4)

6. Single stressed words seek verse-final position (subject to the priority of rules #1-4)

7. Any sequence rules are subject to inversion in the first half-line if (a) the verse has double alliteration and (b) the inversion corresponds more closely to normal prose word-order

8. WTP at the level of the half-line or verse in general obeys the metrical laws of the alliterative line (as symbolized, e.g., in Sievers's Five Types)

9. WTP at the level of the whole line in general follows the predisposition of the metrical formulas described in chapter 3

These rules are arranged in approximate order of descending importance; that is, #2 is regularly superseded by #1, as in 719b (healpegnas fand ), and #4 by #2, as in 712a (mynte se manscaða ). Constraint #3 is a very powerful and general stricture, accounting for phrases as apparently diverse as 721a (dreamum bedæ>led ), 722a (fyrbendum fæst ), and 730a (magorinca heap ). Prepositional phrases are essentially treated as the single words they linguistically are and positioned

at the ends of units, unless another rule supervenes that localization. Note also that the first verse in any line shows special flexibility in allowing reversal of the usual patterning if conditions are right. Correspondingly, the second verse provides a regular locus for a non-alliterating fourth stressed element, and many second-verse phrases take advantage of this provision. Rules #8 and #9 apply at the prosodic level to virtually all of Beowulf ; even if no other constraint orders a given phraseological element, the verse meter and whole-line metrical formula will prescribe pattern in the involved diction.

Other rules, compositionally analogous to the second-level "focusing" features discussed in relation to Serbo-Croatian epic phraseology, also contribute to the poet's and tradition's structuring of the phrase and line:^[73]

1. Collocation (alliterating pair of words or roots)

2. Cluster (association of words outside of the alliterative constraint)

3. Theme (narrative structure)

4. Special "rhetorical" structures (e.g., litotes)^[74]

The Beowulf passage has given us examples of all these rules in action, from the quite common pairing of alliterating words through the cluster of morphs and on to the theme and rhetorical figures. These influences are much more telling on the actual phraseology than is usually recognized, especially because the conventional model assumes a simple inventory of formulas and systems for thematic units.^[75] These rules, just like the word-type placement rules, contribute both uniformity and variety to traditional phraseology: the most fundamental aspects of structure remain consistent, while the absolute verbal shape of any one instance—or even one set of instances—varies considerably.^[76]

Finally, the point needs to be made that the discovery of traditional rules and of their significance at the deepest levels of composition does not in any sense invalidate or call into question the permanently useful research done on the formula and system—for traditional rules do not supersede these structures; they simply rationalize them. Instead of settling for an approximation of phraseological patterning, a convenient explanation that has the

advantage of simplicity, we can now go further and explain what different formulas and systems have in common among themselves and even what they share with ostensibly non-formulaic (but still traditional) diction. We need not be burdened with conflicting definitions of the formula, none of which captures all of the half-lines in Beowulf and explains fully their traditional character; indeed, we can see that, given the tradition-dependent nature of Anglo-Saxon prosody, such a definition is an impossibility. Most importantly, by recognizing that formulas and systems have their roots in traditional rules, we are basing our understanding of the poetry on the idiosyncratic reality of the individual poetic tradition, as well as on a comparative perspective on oral epic poetry as a whole. By placing the broadly comparative concept of formulaic structure against the distinctly tradition-dependent background of traditional rules, in other words, we are creating a responsible interpretation of the diction of Beowulf , one that will serve faithfully as the foundation for later aesthetic investigation.