4.1 Historical Metrics

This chapter surveys predictable developments in alliterative meter after Beowulf, which was probably composed around ad 700.¹ The most important evidence for these developments comes from The Battle of Maldon, a late Old English poem about one-tenth the size of Beowulf. Our language-based theory makes many predictions and testing them systematically can compensate for a smaller body of evidence. Suppose, for example, that the theory predicts an increase in frequency for a variant with an extrametrical word. If Beowulf provides one instance of the variant, one instance in Maldon will represent a tenfold increase in frequency. Since Maldon provides only one instance, however, the frequency rise might be due to chance. This kind of prediction has limited weight on its own, like the prediction that a flipped coin will land heads-up. Predicting that a coin will land heads-up nineteen times out of twenty is quite a different matter.² Making the correct prediction for a variety of cases can validate a theory even if each prediction has limited weight. An array of predictions is questionable if the researcher has ignored an important false prediction or has failed to make a prediction explicit enough for proper testing (has failed, in technical idiom, to make the prediction falsifiable). From a statistical point of view, however, what matters is the probability of chance occurrence for the whole array of successful predictions.

Poetic universals provide additional help in dealing with a small corpus. The principle of closure makes it possible to deduce the complexity of a verse pattern from its placement within the line. The universal principle of “interest” provides help of a different kind.³ We would expect Old English poets to use complex types at appropriate frequencies for metrical variety, with frequent return to the norm for metrical coherence. As Ruth Lehmann pointed out, the Beowulf poet does exactly that, pairing the optimal type A1 with a more complex verse pattern in a typical line.⁴ For maximum interest, a poet would observe a consistent interval between occurrences of a given metrical variant, distributing it evenly throughout the work. Such an effort could provide good coverage of the metrical possibilities in a rather small sample. As it turns out, Old English poets are remarkably adept at making the metrical rounds. Surprising though it may seem, we could deduce the essentials of Cynewulf’s meter from the shortest Old English poem bearing his runic signature: Fates of the Apostles, a work of 122 lines. In Cynewulf’s longer poems, as in Beowulf, the distribution of types and variants in a random sample of 100 lines comes quite close to their distribution within the entire work.⁵ With 324 lines, Maldon should provide a representative sample of metrical practice in the late tenth century.

The evidence of Maldon would be difficult to interpret if the poet no longer grasped the principles of word-foot meter. Several kinds of evidence show that these principles, though more difficult to implement, were not yet lost. Two-word type A1, the simplest realization of the simplest pattern, provides the most informative statistics. In Beowulf, two-word A1 variants like wordum mǣlde ‘spoke with words’ comprise 14 percent of total verses. In Maldon, the frequency is slightly under 12% (77/650). Comparable two-word phrases did not often occur by chance and both frequencies required effort to maintain (section 3.2). In Beowulf, 57 percent of the two-word A1 variants appear in the b-verse. In Maldon, the relative frequency of b-verses jumps to 96 percent (74/77). Placement in the closing half of the line shows that the Maldon poet still perceived two-word type A1 as the metrical optimum. The shift to the b-verse is attributable in part to complex new variants that demanded room in the a-verse, their only appropriate location. Concentration of two-word type A1 in the b-verse would also have communicated the metrical norm more emphatically to a late Old English audience faced, like the poet, with the cultural disruptions of language change.

As it became more difficult to avoid unstressed function words, we would expect a rise in the frequency of variants with extended dips. Variants of a formulaic line depicting cowardice, the binary opposite of heroism, can illustrate this trend.⁶

1. 1. (a) flugon on fæsten  ond feore burgon  (El 134)

      ‘fled to (that) safe place and saved (their) lives’

   2. (b) flugon on þæt fæsten  and hyra feore burgon  (Mald 194)

Example (1a) is from Cynewulf’s Elene, composed between 775 and 850.⁷ Example (1b) is from The Battle of Maldon, which narrates an event datable to ad 991 and was probably composed soon afterwards.⁸ Translation of (1a) into Modern English requires addition of the parenthesized function words. Item (1b) translates word for word.

4.2 Classification of Metrical Lifts, Metrical Dips, and Extrametrical Dips

Generalizing across verse types, we can classify metrical and extrametrical positions according to the prominence of the corresponding linguistic constituents. The positions are presented below in order of prominence, beginning with the most prominent. For concision, I refer to a stressed word as occupying a lift if its stressed syllable occupies a lift, whether or not the word also contains unstressed syllables that occupy x positions.

1. Level 1 The dominant lift corresponds to a syllable with primary word stress and prominent phrasal stress. It is the first lift in the verse and it subordinates following lifts (section 2.9). The dominant lift is most often occupied by a prominently stressed noun or adjective. Finite verbs with subordinate phrasal stress may occupy the dominant lift but this seldom occurs when there is a prominent noun or adjective in the same verse. When a verse contains two stressed words differing in prominence, the more prominent word normally stands first on the more prominent lift (Sievers’s rule of precedence).

2. Level 2 A subordinate lift corresponds to a syllable with subordinate stress. The second S position in a verse corresponds to a syllable with primary word stress and subordinate phrasal stress. The s position corresponds to a syllable with subordinate word stress. A subordinate lift provides an appropriate location for finite verbs and other constituents with weak but perceptible stress. The final lift of the b-verse, which requires stress but does not permit alliteration, is the optimal site for all finite verbs except finite forms of wesan ‘to be.’ In heavy verses with more than two lifts, the relative prominence of subordinate lifts can be indicated by subdividing level 2 into 2.1, 2.2, etc.

3. Level 3 A light foot corresponds to an unstressed function word. It is the most conspicuous and least heavily restricted kind of metrical dip. Though normally realized as an unstressed word, it can be realized as a constituent with weak but perceptible stress such as a finite verb. It is never realized as a prominent noun or adjective. In a handful of apparent exceptions, the stress on the noun or adjective is subordinated in a predicate nominative or predicate adjective construction.

4. Level 4 A foot-internal dip corresponds to one or more unstressed syllables in a stressed word and is normally realized with such syllables. Alternative realizations are typically prefixes or inconspicuous function words with zero stress. The internal dip of an Sx foot is occasionally realized as a finite form of wesan, the verb of highest frequency and lowest prominence. Realization of level-4 dips with other verbs is permissible but very rare, much rarer than with dips at level 3. No verb ever appears in the internal dip of an Sxs or Sxxs foot.

5. Level 5 An extrametrical dip before the second foot interrupts the metrical pattern and should be as inconspicuous as possible. It is realized most often as an unstressed prefix, sometimes as an isolated preposition, a pronoun, or some other function word with zero stress. In a few cases, the Beowulf poet realizes an extrametrical dip as a finite form of wesan.

6. Level 6 An extrametrical dip in anacrusis is subject to very strict constraints due to the risk of confusion between anacrusis and light feet. It is normally realized as an unstressed prefix in the a-verse. Non-prefixal realizations are function words with zero stress, typically prepositions. No verb of any kind may appear in anacrusis. The number of syllables in anacrusis is limited to two, the maximum number of syllables in an Old English prefix.

4.3 Poems Surveyed for Evidence of Metrical Evolution

Metrical change in the late Old English period can be illustrated in a group of five poems: Beowulf, Cynewulf’s Elene, The Battle of Brunanburh, Judith, and The Battle of Maldon.⁹ Elene provides a second long poem in strict traditional meter.¹⁰ Verse frequencies common to Beowulf and Elene represent standard metrical practice before the tenth century. Brunanburh, a small heroic poem of 73 lines, narrates an event datable to ad 937. Brunanburh still adheres well to the metrical style of Beowulf. Judith, a tenth-century poem of 349 lines, also avoids conspicuous departure from traditional norms, though as we shall see some innovations have emerged. The meter of Maldon is widely regarded as peculiar, especially with regard to employment of unstressed words. Maldon and Brunanburh have many important features in common. Both employ traditional formulaic language to narrate recent historical events. Significant metrical differences between these poems can hardly be due to chance. Maldon provides our most important evidence for metrical change during the late Old English period.

4.4 Metrical Responses to Reduction of Unstressed Vowels

Metrical changes in Maldon are attributable to a major sound change that occurred not long after the composition of Brunanburh. As Fulk observes, “by the late tenth century scribes regularly confuse all unstressed vowels, indicating the sort of vowel centralization that accompanies the general lowering of levels of stress.”¹¹ Given the heavy investment of energy in the first syllable of a Germanic word, any general diminution of that energy leaves non-initial syllables vulnerable to reduction and loss. The first step in English was reduction of unstressed inflectional vowels to schwa, a central vowel pronounced in the most relaxed articulatory position, which requires the least energy. The next step, completed in early Modern English, was outright loss of schwa in most phonetic environments (section 10.1).¹²

When Maldon was composed, many grammatical endings were no longer distinguishable from one another and the ambiguity was resolved by more frequent employment of function words.¹³ An ambiguous dative inflection, for example, would increasingly be clarified by a preposition like to or for. Ambiguous definite inflections would be clarified by a determiner equivalent to modern English the or that. A verb inflection no longer provided adequate identification of the subject. The Beowulf poet could omit subject pronouns to improve the meter but this option was becoming unworkable in the tenth century. Shortly after the battle of Maldon took place, Ælfric was modernizing poetic diction, for example by transforming lȳtle werode, a traditional formula, into mid lȳtlum werode (section 3.2). This extension of the formula replaced an archaic instrumental construction with a modern prepositional phrase. In an earlier poem, Ælfric’s version of the formula would arouse suspicion as a type A1 variant that never appears in Beowulf.¹⁴

As R. D. Fulk observes, “It is sometimes difficult not to portray late aberrations from classical norms as representing decline rather than simple change.”¹⁵ Reflexive dispraise of Maldon is surely inappropriate.¹⁶ This poem narrates a complex military engagement in an unusually detailed, knowledgeable, and well-organized fashion unparalleled in any other Old English poem, Beowulf included. In narrating the battle of Ravenswood, for example, the Beowulf poet represents only two combats, each involving only two combatants (lines 2961–81). On a smaller scale, the Maldon poet tinkers artfully with formulaic language when close proximity would create a repetitious effect, just like the Beowulf poet.¹⁷ Maldon differs from Beowulf primarily because language change made it impossible to maintain the same degree of metrical concision.

4.5 Evolution of Extrametrical Dips in Anacrusis

The word-foot theory identifies type A1 as the most favorable site for metrical experiments. With no inherent complexity, this type has maximum tolerance for departure from its two-word norm (section 3.4). We first consider placement of prefixes and function words in anacrusis before the first alliterating syllable of type A1, at level 6. In Beowulf, the frequency of anacrusis for this type is 2 percent (42/2,045). Of the 42 instances, 35 are a-verses (83 percent). In the a-verse, the Beowulf poet normally realizes anacrusis as an unstressed prefix, the least conspicuous kind of extrametrical constituent. The verses with non-prefixal anacrusis are represented in item (2).

1. 1. (a)

      (wið) ord ond / (wið) ecge (Beo 1549a)	(x)S:x/(x)Sx  (4X)
      ‘against spear and sword’

   2. (b)

      (Tō) lang ys / (tō) reccenne  (Beo 2093a)	?(x)S:x/(x)Ssx  (1X)
      ‘It is too lengthy to narrate’

   3. (c)

      (Hū) lomp ēow / (on) lāde (Beo 1987a)	(x)S:x/(x)Sx  (1X)
      ‘How did it go for you on the journey?’

   4. (d)

      (swā) guman ge- / frugnon  (Beo 666b)	(x)S:x/Sx  (4X)
      ‘as men had heard’

Most of the a-verse instances have a preposition in anacrusis, like item (2a). Prepositional anacrusis is a fairly short step from prefixal anacrusis. Most Old English prefixes are derived from prepositions and still look like them. The step from prefixal anacrusis is not much farther in (2b), which has a prepositional adverb in anacrusis. This verse is emended to type A1 in the standard edition. Item (2c) is the only a-verse example with a wholly non-prepositional element in anacrusis: hū, used here as an adverb. Anacrusis occurs less often in the b-verse but the relative frequency of non-prefixal anacrusis is higher: 57 percent (4/7), as compared with 17 percent for the a-verse (6/35). In all but one of the verses like (2d), the word in anacrusis is swā, used as a conjunction (cf. 93b and 1223b). In the remaining example (2247b), the word in anacrusis is nū ‘now, now that,’ also used as a conjunction. Minimizing non-prefixal anacrusis in the a-verse helps distinguish syllables in anacrusis from light feet.¹⁸

The frequency of anacrusis in Maldon has risen to 11 percent in type A1 (23/216), as compared with 2 percent in Beowulf. Constraints on non-prefixal anacrusis have also been weakened in the later poem.

1. 1. (a)

      (for)hēawen / (æt) hilde (Mald 223a)	(x)Sx/(x)Sx  (4X)
      ‘hewn down in battle’

   2. (b)

      (mid) gafole / (for)gyldon  (Mald 32b)	(x)Sx/(x)Sx  (4X)
      ‘to buy off with tribute’

   3. (c)

      (Tō) lang hit / (him) þūhte  (Mald 66b)	(x)S:x/(x)Sx  (2X)
      ‘it seemed too long to him’

   4. (d)

      (and) georne / (ge)sāwon  (Mald 84b; cf. 182a, 282b)	(x)Sx/(x)Sx  (4X)
      ‘and clearly saw’

Variants like (3a), with prefixal anacrusis, no longer predominate, accounting for just 17 percent of total instances (4/23). Variants like (3b), with prepositional anacrusis, are not uncommon in Beowulf, but more complex forms of anacrusis have achieved conspicuously higher frequency in Maldon. There is a pair of instances like (3c), with a prepositional adverb. The much longer Beowulf has a single instance. The coordinating conjunctions and ‘and’ and ac ‘but’ have very high frequency in Old English but are never used for anacrusis in Beowulf. The much shorter Maldon has three instances like (3d), with and, and another with ac (Mald 193a). Half of these are a-verses. Making relatively little use of prefixal anacrusis in the a-verse, the Maldon poet has no reason to concentrate non-prefixal anacrusis in the b-verse.

There are no analogues in Beowulf for the remaining instances of type A1 with anacrusis, represented in item (4).

1. 1. (a)

      (Hē) scēaf þā / (mid ðām) scylde (Mald 136a; cf. 23a, 185a)	(x)S:x/(xx)Sx
      ‘he shoved then with the shield’

   2. (b)

      (his) ealdre / (ge)lǣstan  (Mald 11b)	(x)Sx/(x)Sx
      ‘to assist his lord’

   3. (c)

      (Se) eorl wæs / (þē) blīþra  (Mald 146b; cf. 14a, 202b)	(x)S:x/(x)Sx
      ‘the earl was the happier’

   4. (d)

      (þe) āhte / (his) hlāford  (Mald 189b)	(x)Sx/(x)Sx
      ‘which his lord had owned’

   5. (e)

      (þæt) wǣre / (hit ūre) hlāford  (Mald 240b)	(x)Sx/(xxx)Sx
      ‘that it had been our lord’

A functioning inflectional system made it possible to leave pronouns and determiners unexpressed in the eighth century. These do appear in Beowulf, but not when they add unusual complexity to the verse. Maldon provides examples of anacrusis with a personal pronoun (4a), a possessive pronoun (4b), and a determiner (4c). Anacrusis also occurs with the relative pronoun þe ‘which, that’ (4d) and the subordinating conjunction þæt ‘that’ (4e). If the variants in (4) were as likely to appear in Beowulf, we would expect to find 80 instances rather than none.

Non-prefixal anacrusis is less common outside type A1. There are three instances in Maldon (items (5a–c) and two in Beowulf (5d–e).¹⁹ If such variants occurred with equal frequency in both poems, we would expect thirty instances in Beowulf.

1. 1. (a)

      (Hē) bræc þone / bordweall  (Mald 277a)	(x)S:xx/Ss
      ‘he broke the shield-wall’

   2. (b)

      (Se) flōd / ūt gewāt  (Mald 72a)	(x)S/S:x-s
      ‘the flood tide went out’

   3. (c)

      (Þā) flotan stōdon / gearowe  (Mald 72b)	(x)S:xx/Sx
      ‘the sailors stood ready’

   4. (d)

      (Hē ge)fēng þā / fetelhilt  (Beo 1563a)	(xx)S:x/Ss
      ‘He seized the linked hilt’

   5. (e)

      (æt) fōtum // (sæt) frēan / Scyldinga  (Beo 1166a)	(x)Sx//(x)S/Ssx
      ‘sat at the Scylding lord’s feet’

Item (5a) is a variant of type A2. A single A2 variant in the poem has prefixal anacrusis (Mald 138a). Item (5b) is a Db variant with the pattern S/Sxs. A single expanded Dbx variant has prefixal anacrusis (Mald 90a). Maldon contains no variants with prefixal anacrusis analogous to (5c), a type Ax variant with a determiner in anacrusis. Prefixal anacrusis is not favored in these variants from Maldon. Item (5d), a type A2 variant, is the only verse in Beowulf with a pronoun in anacrusis. There are 4 other A2 variants with prefixal anacrusis. Item (5e), a hypermetrical a-verse, has a preposition in anacrusis. Beowulf has twenty-two instances of prefixal anacrusis in type D but no instances of non-prefixal anacrusis. In heavy verses, which are inherently complex, the Beowulf poet tolerates the simplest kind of anacrusis but rarely employs the more complex kinds. The added complexity had to be tolerated in the era of Maldon because unstressed function words were more urgently required.

In Cynewulf’s Elene, anacrusis has an overall frequency of 2 percent (64/2,642). There are three instances of non-prefixal anacrusis in Elene and sixty-one instances of prefixal anacrusis. In the very small Brunanburh, anacrusis has the same overall frequency as in Beowulf, 1% (2/146). Both instances are of the simplest kind, type A1 with prefixal anacrusis. The overall frequency of anacrusis in Judith is 3% (18/696), higher than in Elene but lower than in Maldon. The three normal verses with anacrusis in Judith are unremarkable. All three are of type A1, two with prefixal anacrusis (Jud 196a, 258b) and one with prepositional anacrusis (Jud 116b). More conspicuous experiments with anacrusis occur in hypermetrical a-verses, which comprise an unprecedented 17 percent of total a-verses in Judith (64/348). The next highest frequencies for hypermetrical a-verses are 4 percent for Daniel (34/764) and 2 percent for Guthlac A (17/818).

The frequency of anacrusis in hypermetrical variants is high in Judith. Of sixty-four hypermetrical a-verses in the poem, 8 (13 percent) have prefixal anacrusis, like item (6a).

1. 1. (a)

      (ge)rēnode // rēadum / golde  (Jud 338a)	(x)Sxx//Sx/Sx (8X)
      ‘decorated with red gold’

   2. (b)

      (on) eorðan // un-/swǣslicne  (Jud 65a)	(x)Sx//S/Ssx (4X)
      ‘unpleasantly on earth’

   3. (c)

      (se) rīca // (on his) reste / middan  (Jud 68a)	(x)Sx//(xx)Sx/Sx (2X)
      ‘the mighty one amidst his bed’

   4. (d)

      (Þā wearð) yrre // ānmōd / cyning  (Dan 224a)	(xx)Sx//Ss/Sx
      ‘then the fierce king got angry’

   5. (e)

      #Yrre // (þā wearð) ānmōd / cyning

      Sx//(xx)Ss/Sx

Corresponding frequencies are 9 percent for Daniel (3/34) and 18 percent for Guthlac A (3/17). Judith has four other variants like item (6b), with prepositional anacrusis. These comprise 8 percent of the hypermetrical a-verses in the poem (5/64). The corresponding figure for Daniel is 6 percent (2/34) and there are no such variants in Guthlac A. Anacrusis is realized twice in Judith with determiners, as in (6c). These are no such realizations in Daniel or Guthlac A. As we have seen, the Maldon poet employs determiners in anacrusis but the Beowulf poet never does so in any verse type. In Judith and Maldon, clearly, the need for function words modifying the first stressed word of the verse was weakening traditional constraints and anacrusis was less sharply delineated from other kinds of dip. A unique instance of a verb in anacrusis occurs with a finite form of weorðan in item (6d). This seems unlikely to be authentic, however. The emendation in (6e) employs an archaic word order still employed by poets but increasingly vulnerable to modernization by scribes. From here on the symbol < # > will be used for hypothetical emendations of my own that regularize the meter.

4.6 Evolution of Extrametrical Dips in Verse-Medial Position

Extrametrical words occur most freely in the medial dip of type A1 when the first foot is realized as a trochaic word (section 3.4). There is a large sample of A1 variants like those in item (7). For clarity, a few verses that also have anacrusis are excluded from the sample.

1. 1. (a)

      ealle / (on)wōcon  (Beo 111b)	Sx/(x)Sx (151:228)
      ‘all awoke’

   2. (b)

      beornas / (on) blancum  (Beo 856a)	Sx/(x)Sx (76:12)
      ‘men on horses’

   3. (c)

      geongum / (ond) ealdum  (Beo 72a)	Sx/(x)Sx (46:12)
      ‘to young and old’

The least prominent extrametrical constituents have highest frequency in the medial dip (at level 5). As with extrametrical constituents in anacrusis (at level 6), prefixes outnumber prepositions, which in turn outnumber purely non-prepositional words such as pronouns, articles, and conjunctions. With over 300 instances in Beowulf, variants like (7a), with an extrametrical prefix, have about four times the frequency of those like (7b), with an extrametrical preposition. Variants like (7c), with a non-prepositional word between feet, have the lowest frequency. Distribution of variants like (7a) within the line shows that the extrametrical prefix adds little to complexity. The greater complexity of variants like (7b–c) creates a significant attraction to the a-verse.

In a poem one-tenth the size of Beowulf, we would expect about 30 instances like (7a), and in fact there are 26. Distribution of these variants in Maldon differs dramatically, however, with 23 of the 26 instances in the b-verse. As we have seen, two-word Sx/Sx, the simplest expression of type A1, has shifted to the b-verse in Maldon, freeing up room in the a-verse for the growing number of more complex variants. Type A1 variants like (7a) were evidently perceived as simple enough to participate in this shift. For variants like (7b), frequency and distribution are similar in the two poems, with 7 a-verses and 2 b-verses in Maldon, as compared with 79 a-verses and 11 b-verses in Beowulf. With 58 variants like (7c) in the larger poem, we would expect about 6 in Maldon, and in fact there are 5; but this smaller sample shows no bias toward the opening half of the line, with 2 a-verses and 3 b-verses. Extrametrical words that caused significant complexity in Beowulf were beginning to be perceived as routine and were less strictly regulated by the principle of closure.

In (8a–c), otherwise analogous to (7a–c), the first foot is realized as an S:x word group. The complexity added by the word group inhibits employment of extrametrical words in the medial dip.

1. 1. (a)

      Hete wæs / (on)hrēred  (Beo 2554a)	S:x/(x)Sx (6:9)
      ‘Hate was stirred up’

   2. (b)

      Flota wæs / (on) ȳðum  (Beo 210b)	S:x/(x)Sx (9:1)
      ‘The boat was on the waves’

   3. (c)

      Cwico wæs / (þā) gēna  (Beo 3093b)	S:x/(x)Sx (11:6)
      ‘He was still alive then’

   4. (d)

      Forð ðā / (ēode) Wīstān  (Mald 297b)	S:x/(xx)Ss (0:1)
      ‘Forth then went Wistan’

Beowulf contains only 15 total variants like (8a), with an extrametrical prefix between the feet, as compared with over 300 like (7a). Variants like (8b), with an extrametrical preposition, total 10, as compared with 90 like (7b). Variants like (8c), with a non-prepositional word, total 17, as compared with 59 like (7c). Restrictions on extrametrical words in variants like (8a–c) bring out sharp differences between Beowulf and Maldon. Maldon has 11 variants analogous to (8a), with 8 a-verses and 3 b-verses. If such variants occurred as freely in Beowulf, we would expect about 110 instances rather than the 15 actually found. There are 4 variants like (8b) in Maldon and half of them are b-verses. If Beowulf had the same frequency for this variant, we would expect 40 instances rather than 10. Maldon has 10 variants like (8c), of which 4 are b-verses. The corresponding figure for Beowulf would be 100 instances rather than 17. Item (8d), a type A2 variant from Maldon, merits special attention because it has a finite main verb as an extrametrical constituent between the feet. This might also be analyzed as type A1, since the compound name Wīstān has reduced stress on the second syllable and might scan as Sx rather than Ss. In Beowulf, main verbs never appear as extrametrical words in the medial dip of type A1 or A2 (at level 5). The only dips employed with significant frequency for main verbs in Beowulf are level-3 dips at the beginning of types A3, B, and C.

Late experimentation is more restricted in types A2, D, and E, which have significant inherent complexity. Differences between Beowulf and Maldon emerge when we consider variants with two extrametrical syllables in the medial dip. Here the number of instances is rather small and verse counts will include variants with extrametrical syllables in anacrusis.

1. 1. (a)

      Hyrte / (hyne) hordweard  (Beo 2593a)	Sx/(xx)Ss (5:1)
      ‘the hoard-guardian took heart’

   2. (b)

      deorc / (ofer) dryhtgumum  (Beo 1790a)	S/(xx)Ssx (3:0)
      ‘dark over the troop-men’

   3. (c)

      heals ealne / (ymbe)fēng  (Beo 2691b)	S:sx/(xx)S (0:1)
      ‘seized around his entire neck’

   4. (d)

      ealle / (būton) ānum  (Beo 705a)	Sx/(xx)Sx (46:9)
      ‘all but one’

   5. (e)

      stonc ðā / (æfter) stāne (Beo 2288a)	S:x/(xx)Sx (7:0)
      ‘sniffed then along the stone’

   6. (f)

      Secge / (ic þē tō) sōðe (Beo 590a)	Sx/(xxx)Sx (5:0)
      ‘I tell you in truth’

There are unambiguous instances in type A2 (9a) and type Da (9b). Scansion of (9b) as Sxx/Ssx is ruled out by the constraint against two long feet (section 3.3). Item (9c), the unique manuscript example in type E, has no value for metrical theory, since ymbefēng emends easily to ymbfeng, a form attested elsewhere in the poem.²⁰ With emendation to ymbfeng, (9c) scans as Ssx/(x)S, a common type E variant with a monosyllabic prefix before the second foot. The small group of examples like (9a–b) has higher frequency in the a-verse, as expected for complex variants. Item (9d) shows the distribution of somewhat simpler A1 variants with the first foot realized as a trochaic word. Several of these are b-verses. More complex variants like (9e) realize the first foot as an S:x word group. All of these are a-verses. Variants like (9f), with three extrametrical syllables between the feet, are also restricted to the a-verse.

Variants analogous to those in (9) have enhanced frequency in Maldon.

1. 1. (a)

      Sende / (ðā se) sǣrinc  (Mald 134a; cf. 297b)	Sx/(xx)Ss (1:1)
      ‘then the sea-warrior sent’

   2. (b)

      wyrcan / (þone) wīhagan  (Mald 102a; cf. 286a)	Sx/(xx)Ssx (2:0)
      ‘to form the battle-wall’

   3. (c)

      wācian / (æt þām) wīge (Mald 10a; cf. 28a)	Sxx/(xx)Sx (2:0)
      ‘to weaken at the battle’

   4. (d)

      bēgen / (þā ge)brōþru  (Mald 305a)	Sx/(xx)Sx (7:2)
      ‘both the brothers’

   5. (e)

      (Hē) scēaf þā / (mid ðām) scylde (Mald 136a)	(x)S:x/(xx)Sx (2:0)
      ‘he shoved then with the shield’

   6. (f)

      (þā) hwīle / (þe hē mid) handum  (Mald 14a; cf. 240b)	(x)Sx/(xxx)Sx (1:1)
      ‘while he with (his) hands (could wield weapons)’

Item (10a) is an A2 variant with the same pattern as (9a). With 2 examples in Maldon, we would expect 20 in Beowulf instead of 6. We would also expect 20 instances in Beowulf for variants like (10b), a Dax variant, and another 20 instances for variants like (10c), an Ax variant. In fact there are no instances of either variant in Beowulf. The complex A1 variants (10d–f) are analogous to (9d–f.) Given the number of instances for these variants in Maldon, Beowulf would be expected to yield 100 instances like (10d) rather than 57, 10 instances like (10e) rather than 7, and 20 instances like (10f) rather than 5. There are no parallels in Maldon for variants like (9b) or (9c), but since these total no more than four in Beowulf, we would expect less than one in Maldon.

Variants with extended medial dips seem to have developed quite late. They do not show comparable frequency rises in the other poems sampled. Table 1 summarizes verse counts for variants represented in items (8–10), adding counts for Elene, Brunanburh, and Judith.

The two most common A1 variants on the list show slightly enhanced frequency in Elene, which is a little more than two-fifths the size of Beowulf. For the first of these patterns, the frequency in Judith would correspond to 90 examples in Beowulf rather than the 42 examples actually attested. For the second A1 variant, however, Judith has only 4 examples, which would correspond to 40 in Beowulf, fewer than the 55 actually attested. Only Maldon shows such a consistent rise in frequency for extrametrical syllables in such a variety of verse-medial dips. This would be expected if Judith was somewhat earlier than Maldon and less strongly affected by loss of inflectional distinctions.

4.7 Evolution of Medial Dips in Hypermetrical Verses

Hypermetrical a-verses have two medial dips (section 3.13). In accord with the principle of closure, the second medial dip is more heavily restricted than the first. Dips are extended most frequently in the simplest hypermetrical pattern, Sx//Sx/Sx. This consists of an Sx foot followed by two-word type A1, the simplest normal variant. Variants with anacrusis are included in the counts.

1. 1. (a)

      dryhten // dugeða / waldend  (Jud 61a)	Sx//Sx/Sx (8X)
      ‘God, ruler of hosts’

   2. (b)

      fæste // (be) feaxe / sīnum  (Jud 99a)	Sx/(x)Sx/Sx (14X)
      ‘firmly by his hair’

   3. (c)

      (ā)gotene // gōda / (ge)hwylces  (Jud 32a)	(x)Sx//Sx/(x)Sx (1X)
      ‘emptied of every good’

   4. (d)

      wurpon // (hyra) wǣpen / (of)dūne  (Jud 290a)	Sx//(xx)Sx/(x)Sx (8X)
      ‘threw their weapons down’

The variants like (11a) have no extrametrical constituents in either medial dip. The variants like (11b) have an extrametrical syllable in the first medial dip. Item (11c) is the only instance with an extrametrical constituent in the second medial dip only. The variants like (11d) have extrametrical constituents in both dips. In the more tolerant first medial dip, an extrametrical constituent is typically an independent function word. There are no exceptions among the 24 variants like (11b) and (11d). Among these 24 variants there are 8 like (11d), with two extrametrical syllables in the first medial dip. In the second medial dip, an extrametrical constituent is always a prefix.

Hypermetrical a-verses can be constructed with initial foot patterns other than Sx. These more complex variants are less tolerant of extended medial dips. Variants with an initial Ss foot are illustrated in item (12).

1. 1. (a)

      swīðmōd // sinces / brytta  (Jud 30a)	Ss//Sx/Sx (8X)
      ‘bold-minded distributor of treasure’

   2. (b)

      mundbyrd // (æt ðām) mǣran / þēodne  (Jud 3a)	Ss//(xx)Sx/Sx (3X)
      ‘protection by the great lord’

   3. (c)

      torhtmōd // tīðe / (ge-) fremede  (Jud 6a)	Ss/Sx/(x)Sx (1X)
      ‘gave a gift, inclined to glory’

The most common such variant in Judith is the one free of extrametrical syllables, represented by (12a). Among the variants with the first medial dip extended, item (12b) has two extrametrical syllables and the others have just one. Item (12c) is the only instance with extension of the second medial dip. Here the extrametrical constituent is a prefix, as expected. There are no instances with both medial dips extended. Item (12) contrasts markedly with item (11), where variants with extended medial dips predominate.

In the remaining hypermetrical a-verse types, the medial dip is extended in one of sixteen instances (13d).

1. 1. (a)

      nēalǣhte // niht sēo / þȳstre  (Jud 34a)	Ssx//S:x/Sx (2X)
      ‘the dark night drew nigh’

   2. (b)

      (ge)rēnode // rēadum / golde  (Jud 338a)	(x)Sxx//Sx/Sx (3X)
      ‘decorated with red gold’

   3. (c)

      bealde // byrn-/wiggende  (Jud 17a)	Sx//S/Ssx (9X)
      ‘bold armored warriors’

   4. (d)

      ā tō // (ðām) æl-/mihtigan  (Jud 7a)	S:x//(x)S/Ssx (1X)
      ‘forever to the Almighty’

   5. (e)

      sweord ond // swātigne / helm  (Jud 337a)	S:x//Ssx/S (1X)
      ‘sword and bloody helmet’

   6. (f)

      (ne) feax // fȳre / (be)swǣled  (Dan 437a; cf. 225a)	(x)S//Sx/(x)Sx (2X)
      ‘nor (was their) hair burned by fire’

Items (13a–b) have trisyllabic initial feet followed by a type A1 variant. Items (13c–d) end with a light type Da variant. In (13d), the Da variant is realized as a compound word. Item (13e) ends with a two-word type E variant. Initial S feet are not common in hypermetrical a-verses. Unambiguous instances like (13f) show that they do occur, at least in some poems. Item (13d) might imaginably be scanned as S//(xx)S/Ssx and (13e) as S//(x)Ssx/S, with initial S feet and extrametrical words in the medial dip. The small size of the hypermetrical corpus can make it difficult to identify preferred scansions.

Table 4.2 summarizes the Judith poet’s employment of medial dips in the two most common hypermetrical a-verse patterns, adding counts from Beowulf, Elene, Daniel, and Guthlac A for comparison. There are no hypermetrical verses in Maldon. All extended medial dips are represented by a single parenthesized (x), without regard for the number of added syllables. Instances with anacrusis are included in the counts as before.

Hypermetrical patterns had always allowed for placement of extrametrical words in medial dips. Even the small samples from Beowulf and Elene include an instance with both dips extended. All samples include an instance of Sx//(x)Sx/Sx, the optimal type with its first medial dip extended. In Elene, Guthlac A, and Judith, this is the most common variant. In Beowulf and Daniel, the variant without extrametrical syllables predominates, accounting for about half of the sample. In comparison with Beowulf and Daniel, Judith expands the first medial dip more often in this hypermetrical type but not more often than Elene or Guthlac A. Overall, however, Judith expands medial dips more frequently than any of the other poems. Beowulf and Guthlac A make no use of the variant Sx//(x)Sx/(x)Sx, which has both medial dips extended. This variant has a frequency of 13 percent in Elene (1/8) and Daniel (3/23), as compared with 19 percent in Judith (8/43). Guthlac A and Elene make no use of the hypermetrical type Ss/Sx/Sx, which occurs twelve times in Judith, with extension of the first medial dip in three instances. Judith contains a single instance of Ss//(x)Sx/(x)Sx, the most complex variant in Table 4.2. The other two instances of this variant come from Beowulf and Daniel, which realize dips more cautiously in the most common hypermetrical type. Given surprises like this, we have no reason to suppose that a unique hypermetrical variant in Judith would have been unacceptable to the Beowulf poet. There were traditional precedents for extended medial dips in these verse types. What stands out in Judith is the frequency of extension. In this late poem, function words occupy level-5 locations where they had always been permissible but with the higher frequency demanded by language change.

Item (14) illustrates realization of medial dips in hypermetrical b-verses from Judith. The parenthesized (x) in initial position stands for any number of extrametrical syllables, which can be added freely to the initial dip. Parenthesized (x) positions in the more strictly regulated medial dip stand for a specific number of extrametrical syllables.

1. 1. (a)

      swylce ēac // sīde / byrnan  (Jud 337b)	(x)xx//Sx/Sx (46X)
      ‘broad mail-coats as well’

   2. (b)

      Hīe ðā on // reste / (ge)brōhton  (Jud 54b)	(x)xx//Sx/(x)Sx (13X)
      ‘then they brought her to bed’

   3. (c)

      swylce ēac // mēde / (on) heofonum  (Jud 343b)	(x)xx//Sx/(x)Sx (5X)
      ‘and reward in heaven as well’

   4. (d)

      swylce ēac // būnan / (ond) orcas  (Jud 18b; cf. 67b)	(x)xx//Sx/(x)Sx (2X)
      ‘cups and pitchers as well’

   5. (e)

      þe hyre // weorðmynde / geaf  (Jud 342b)	(x)xx//Ssx/S (1X)
      ‘who gave her honor’

Hypermetrical b-verses have a light initial foot with the pattern xx. The embedded normal verse is usually a two-word realization of type A1. The variant in (14a), with no medial dips extended, has the highest frequency. Realization of extended dips observes the usual order of frequencies in (14b–d), with prefixal extensions highest, followed by prepositional extensions and non-prepositional extensions. The variety of hypermetrical patterns is severely restricted in the closing half of the line. Among 67 instances in Judith, 66 have the optimal pattern (x)xx//Sx/Sx, like (14a–d). The only other instance, (14e), has the pattern (x)xx//Ssx/S, with two-word type E as the embedded normal verse. The principle of closure also inhibits departure from the optimal verse pattern in Beowulf, Elene, Daniel, and Guthlac B, which yield a total of 134 hypermetrical b-verses. All but 8 of these have the pattern (x)xx//Sx/Sx. There are no instances like (14e) in these poems, but a few other patterns occur.

1. 1. (a)

      þǣr hīe þæt // āglāc / drugon  (Dan 237b)	(x)xx//Ss/Sx (3X)
      ‘where they endured that misery’

   2. (b)

      Spræc ðā // ides / Scyldinga  (Beo 1168b)	(x)xx//S/Ssx (4X)
      ‘the lady of the Scyldings spoke then’

   3. (c)

      þā hē þyder // folc / samnode  (Dan 227b)	(x)xx//S/Sxx (1X)
      ‘then he gathered a troop to that place’

Elene has only the optimal pattern (x)xx//Sx/Sx. Daniel, Guthlac A, and Beowulf each have a single instance of the pattern in (15a). The other three instances like (15b) occur in Daniel, and the variant in (15c) occurs only in Daniel. Guthlac A stands out among Old English poems in employing a hypermetrical b-verse pattern seventeen times in the a-verse. All seventeen instances have the optimal pattern (x)xx//Sx/Sx. In eight instances with an extended medial dip, the extension is limited to one syllable and is always realized as an unstressed prefix. In the a-verse, the pattern xx//Sx/Sx acquires a kind of complexity from its unusual placement and other kinds of complexity are inhibited.

Table 4.3 summarizes the Judith poet’s employment of medial dips in hypermetrical b-verse types, adding counts from Beowulf, Elene, Daniel, and Guthlac A for comparison.

In the optimal pattern xx//Sx/Sx, all five poets extend the medial dip with significant frequency. The frequency of 30 percent for Judith stands near the middle of the range (22 to 44 percent). In the more complex types, none of the poets extend the medial dip. When we consider the kind of constituent used to extend the dip, Judith stands apart. In Beowulf, Daniel, and Guthlac A, the medial dip is extended only with prefixes. A preposition is used once for this purpose in Elene at 584b. In Judith, five variants out of twenty employ a preposition, as compared with one out of forty total instances in the other poems. Only Judith extends the medial dip with a conjunction (in item (14d)). Extension of the medial dip adds significant complexity to a hypermetrical b-verse. In the optimal type, extension normally consists of an unstressed prefix, the least conspicuous and least disruptive kind of extension. The prefix is also the constituent preferred for anacrusis, another kind of extension that causes complexity. In Judith, the prefixal bias has been weakened in both kinds of extension: anacrusis in the hypermetrical a-verse and extension of the medial dip in the hypermetrical b-verse.

4.8 Medial Dips in Variants with the First Foot Realized as a Word Group

Extrametrical constituents lie outside the metrical pattern proper (section 3.4). In a medial dip, accordingly, extrametrical constituents at level 5 can be distinguished from constituents that occupy an x position in the first foot, at level 4.

1. (16)

   Sōð is / (ge)cȳþed  (Beo 700b)	S:x/(x)Sx
   ‘the truth is known’

In item (16), unstressed is, the first word in the dip, fills the x position of the first foot and the ge- prefix is extrametrical. A constituent occupying an x position at level 4, inside the Sx foot, will be more prominent, on average, than a following extrametrical constituent at level 5. Non-prefixal function words are more likely to appear within an Sx foot than as extrametrical extensions of the medial dip.

Though less heavily restricted than extrametrical words, function words in an Sx foot are more heavily restricted than those occupying a level-3 dip in types A3, B, and C. Verse counts for Beowulf in item (17) include type A1 variants with and without extrametrical syllables in the medial dip.

1. 1. (a)

      forð on- / sendon  (Beo 45b)	S:x/Sx (25:121)
      ‘sent forth’

   2. (b)

      gold on / grēote (Beo 3167a)	S:x/Sx (92:9)
      ‘gold in the grit’

   3. (c)

      grim ond / grǣdig  (Beo 121a)	S:x/Sx (70:20)
      ‘grim and greedy’

   4. (d)

      cearu wæs / (ge)nīwod  (Beo 1303b)	S:x/(x)Sx (8:14)
      ‘care was renewed’

Unstressed prefixes are the preferred substitutes for inflectional syllables in Sx words, as shown by the high overall frequency of variants like (17a) and their high relative frequency in the b-verse. A similar variant with a preposition in the first foot, represented by (17b), occurs less often, with higher frequency in the a-verse. A variety of non-prepositional function words, taken together, have the lowest frequency and occur most often in a-verses like (17c). Other variants like (17d), with a monosyllabic auxiliary in the first foot, have sharply lower frequency. Auxiliary verbs have high frequency in level-3 dips but seldom appear in level-4 dips.

Table 4.4 summarizes employment of variants like (17a–d) in Judith and the other poems surveyed.

Frequencies on Table 4.4 show some differences attributable to language change. In the late Judith and Maldon, there are relatively fewer variants with prefixes and relatively more with prepositions. Late poets needed prepositions as replacements for ambiguous case endings. In the “other non-verbal” category, the frequency for the late poems is about half the frequency for Beowulf, but the frequency for Elene is equally low. The most conspicuous anomaly is employment of main verbs in Maldon on the x position of an Sx foot (at level 4).

The level-4 verbs in Maldon appear in two identical verses like (18a).

1. 1. (a)

      Wæl fēol / (on) eorþan  (Mald 126b, 303b)	S:x/(x)Sx
      ‘The slain fell onto the earth’

   2. (b)

      ac hē / eorðan gefēoll  (Beo 2834b)	xx/Sx:x-s
      ‘but he fell onto the earth’

   3. (c)

      hægl fēol / (on) eorþan  (Seaf 32b; cf. Rid29, 12b)	S:x/(x)Sx
      ‘hail fell onto the earth’

   4. (d)

      God wāt on / mec  (Beo 2650b)	S:s:x/S
      ‘God knows (this) about me’

   5. (e)

      (æt) fōtum / (sæt) frēan Scyldinga  (Beo 1166a)	(x)Sx//(x)S/Ssx
      ‘sat at the feet of the lord of the Scyldings’

The repeated formula has modern syntax, with the verb between the subject and a prepositional phrase. The noun object of the preposition is more prominent than the verb and less appropriate as an occupant of the dip. If the poet uses such constructions at all in type A1, a non-alliterating verb in the medial dip represents the minimum departure from traditional norms. In (18b), the Beowulf poet employs an archaic OV construction with the object of the preposition before the finite verb. Here the concept of motion toward is expressed by the accusative case ending on eorðan. The VO construction in (18a) expresses the same concept with the preposition on. Variants like (18a) were certainly in existence before Maldon. Similar variants like (18c) can be found in the Seafarer and the Old English Riddles. These poems are relatively hard to date but they might have been quite early and were composed well before Maldon in any case.²¹ The same syntax can be found in (18d), a type E verse from Beowulf. In (18d), however, the finite verb occupies an s position, which is more appropriate than an x position for the stressed syllable of a main verb. The only main verb occupying a medial dip in Beowulf is sæt in item (18e), a complex hypermetrical variant of the form (x)Sx//(x)S/Ssx. The Maldon poet was less reluctant to employ finite main verbs in a medial dip but had not lost contact with poetic tradition. Acceptable SVO variants simply provided a better fit with late Old English and achieved higher frequency for that reason.

Late Old English trends in verb placement also show up in type A1x (Sxx/Sx) and in the A2 subtype that was destined to survive (Sx/Ss).

1. 1. (a)

      hrēo wǣron / ȳþa  (Beo 548b)	S:xx/Sx
      ‘rough were the waves’

   2. (b)

      Wigan wǣron / blīðe  (El 246b)	S:xx/Sx
      ‘the warriors were happy’

   3. (c)

      Hæleð wǣron / yrre  (Jud 225b; cf. 195b)	S:xx/Sx
      ‘the warriors were angry’

   4. (d)

      Bogan wǣron / bysige  (Mald 110a)	S:xx/Sx
      ‘bows were busy’

   5. (e)

      (Þā) flotan stōdon / gearowe  (Mald 72b)	(x)S:xx/Sx
      ‘the sailors stood ready’

   6. (f)

      Strǣt wæs / stānfāh  (Beo 320a)	S:x/Ss (6X)
      ‘the street was stone-paved’

   7. (g)

      Frōd wæs / (se) fyrdrinc  (Mald 140a)	S:x/(x)Ss (4X)
      ‘wise was the troop-warrior’

   8. (h)

      Wund wearð / Wulfmǣr  (Mald 113a)	S:x/Ss
      ‘Wulfmær got wounded’

   9. (i)

      Hyge wæs / (him) hinfūs  (Beo 755a)	S:x/(x)Ss
      ‘his mind was eager to escape’

Items (19a–d) are A1x variants with the medial dip realized as a disyllabic form of the verb to be. Item (19a) is the only instance in Beowulf. The only instance in Elene is (19b). Judith has a pair of instances like (19c), which would correspond to twenty instances in Beowulf. Item (19d) is a comparable instance in Maldon. In (19e), the Maldon poet realizes the medial dip as stōdon, a verb of greater prominence than the verb to be. This is the only such A1x variant in the poems surveyed. The A1x variants from Elene, Judith, and Maldon all have modern word order. The single instance in Beowulf has an archaic construction now restricted to literary usage. Items (19f–g) are A2 variants with the medial dip realized as a monosyllabic form of the verb to be. The four instances of this variant in Maldon are equivalent to forty instances in Beowulf, but only six are attested in the earlier poem. What stands out in (19g) is extension of the medial dip with a determiner equivalent to the definite article the. This occurs in three of the four instances. In the remaining instance, item (19h), the determiner is absent for purely grammatical reasons. Proper names normally refer to a definite individual and do not normally require definite articles. In (19f), a determiner could have been used before strǣt but the Beowulf poet omits it, avoiding non-prepositional anacrusis. In variants like (19f), the Beowulf poet never places a determiner in anacrusis and the only instance with extension of the medial dip is (19i). Here the extension is him, employed as an archaic dative of possession. The verse translates literally as “the mind to him was eager to escape.” Departures from norms in archaic variants established precedents for later variants with SVO syntax.

The SOV and SVO constructions in (19a–i) perform important metrical work. Reversal of the chosen word orders would yield metrical patterns that are much more complex or unmetrical, with the wrong word alliterating in some cases. All the poets surveyed here could employ SOV or SVO syntax as necessary to avoid unmetrical patterns. What increased was the frequency of acceptable SVO variants, a frequency that testifies to the greater strength of the SVO norm in the late Old English period. Such evidence for language change comes into focus only within a theory that defines metrical norms in concrete linguistic terms and links the distribution of a variant to departure from those norms.

4.9 Evolution of Dips Between Stresses in Compound Feet (Sxs and Sxxs)

The simplicity of the equation foot = word made it possible for native speakers to intuit the complete set of foot patterns. This equation would be compromised if any word pattern were denied foot status arbitrarily. The price to be paid for such high-level simplicity was feet corresponding to unusual words. Compound feet with unusual patterns had to be realized as single words often enough to establish their metrical identity, but this became more difficult as the productivity of compounding declined.

The Sxs foot corresponds to compound words of relatively low frequency with a monosyllabic dip between the stresses. This foot is realized directly 134 times in Beowulf as a compound like ellen-weorc ‘deed of valor’ or līf-gedāl ‘loss of life.’ Direct realizations are distributed quite evenly throughout the poem, highlighting the linguistic basis of the foot about once every twenty-five lines.²² Most often, however, the Sxs foot is realized with an Sx:s word group like flōdes ǣht ‘the grip of the flood’ (762X), which mimics the structure of compounds like ellen-weorc. Somewhat less common are realizations of the Sxs foot with an S:x-s group like dēað fornam ‘death took away’ (171X), which mimics the structure of compounds like līf-gedāl. Realization of the Sxs foot as a word group is most frequent in type B, where Sxs word groups bring the verse up to normal weight. Sxs word groups have much lower frequency in the heavy type Db, where they make the verse even heavier.

The Sxs foot is sometimes realized as an S:x:s word group bearing a less obvious resemblance to a compound.

1. 1. (a)

      ond his / cwēn mid him  (Beo 923b)	xx/S:x:s (11X)
      ‘and his queen with him’

   2. (b)

      bið se / slǣp tō fæst  (Beo 1742b)	xx/S:x:s (6X)
      ‘that sleep is too fast’

   3. (c)

      in / sele þām hēan  (Beo 713b)	x/S:x:s (26X)
      ‘in the high hall’

   4. (d)

      līcað / leng swā wēl  (Beo 1854a)	Sx/S:x:s (1X)
      ‘pleases the longer the better’

In (20a), the x position of the Sxs foot is occupied by a preposition. Word groups like cwēn mid him bear a certain resemblance to Sxs word groups like eft on-wōc, with the dip realized as a prefix derived from a preposition. In (20b), the medial dip is realized as a prepositional adverb. In (20c), it is realized as a non-prepositional function word. Item (20d) is the only type Db variant with an S:x:s word group in the second foot.

Table 4.5 provides counts from all the poems surveyed for realizations of the Sxs foot in type B.

In Maldon, the frequency of realizations with an Sxs compound is less than half that for the other poems surveyed. In the era of Maldon, language change had distanced the Sxs foot from its direct realization and type B had become more difficult to identify as a pattern with two word-feet.

Interpretation of counts in Table 4.5 can be checked against the relative frequency of a-verses, which would be expected to rise as type B became more complex. Table 6 provides the number of a-verses and b-verses for type B variants with Sxs compounds, Sx:s word groups, S:x-s word groups, and S:x:s word groups. Each set of verse counts is accompanied by the percentage of a-verses.

Brunanburh and Maldon have the same frequency of 100 percent for the Sxs variant represented on the top line. With that exception, Maldon has the highest percentage of a-verses in all four categories. In the era of Maldon, all realizations of the Sxs foot had greater complexity and types that contained this foot were more strongly attracted to the a-verse.

The Beowulf poet realizes the Sxxs foot with a variety of word groups in long type B variants.

1. 1. (a)

      on / fēonda geweald  (Beo 808a)	x/Sx:x-s (136X)
      ‘into the power of foes’

   2. (b)

      wæs þæt / gifeðe tō swīð  (Beo 3085b)	xx/Sx:x:s (18X)
      ‘that fate was too strong’

   3. (c)

      gē wið / fēond gē wið frēond  (Beo 1864a)	xx/S:x:x:s (6X)
      ‘both with friends and against foes’

   4. (d)

      swylc / Æschere wæs  (Beo 1329b)	xx/Sxx:s (6X)
      ‘such was Æschere’

   5. (e)

      Hē / bēot ne ālēh  (Beo 80a)	x/S:x:x-s (3X)
      ‘he did not forget his boast’

   6. (f)

      ōwer / feor oððe nēah  (Beo 2870a)	xx/S:xx:s (3X)
      ‘either far or near’

   7. (g)

      ðȳ hē þone / fēond ofercwōm  (Beo 1273a)	xxxx/S:xx-s (1X)
      ‘therefore he overcame the foe’

The variants with highest frequency are like (21a), in which the Sx:x-s word group imitates the structure of compounds like sibbe-gedriht. The variant with the next highest frequency, represented by (21b), also has an Sx word as the first constituent of the group but realizes the second x position as an independent unstressed word. Next in frequency are variants like (21c), a four-word realization of the Sxxs foot, and (21d), which realizes the first three positions of the foot as an Sxx word or an Ssx compound name with weakened secondary stress. In (21e), the foot is realized as an S:x word group followed by a monosyllabic word with a prefixal negative. Finally, there are a few instances with the internal xx sequence of the foot realized as a disyllabic function word (21f) or a disyllabic prefix (21g).

Table 7 provides counts for verses like (21a–g) in all the poems surveyed.

Every poem shows highest frequency for the variant represented by item (21a). Throughout the Old English period, the Sx:x-s word group was preferred to all others as the best substitute for an Sxxs compound. Elene and Brunanburh follow Beowulf in employing variants like (21b) with the second-highest frequency. Maldon stands apart in several ways that look toward Middle English alliterative meter. The Sxxs foot, with its long dip, has become more useful by the late Old English period. The 35 total instances for Maldon would correspond to 350 instances in Beowulf, but we find only 173. The frequency rise is particularly large for more complex variants like (21c–g). A total of 12 for these in Maldon would correspond to 120 in Beowulf, but there are only 19. As with anacrusis and medial dips, constraints on foot-internal dips were being relaxed in the late Old English period.

Table 4.8 shows the distribution of a-verses and b-verses for the realizations surveyed in Table 4.7, with the percentage of a-verses in parentheses.

Maldon shows a traditional frequency of a-verses for the preferred variant and elevated a-verse frequencies for all others except S:x:x-s, which occurs once in a b-verse. Like the Sxs foot, the Sxxs foot had increased in complexity, creating a stronger attraction to the a-verse.

In heavy Db and Dbx variants, which have great inherent complexity, the poet works harder to realize the second foot as a compound word. Item (22) provides verse counts for variants with an Sxs foot.

1. 1. (a)

      hār / hilderinc  (Beo 1307a)	S/Sxs (56X)
      ‘old battle-warrior’

   2. (b)

      sweord / swāte fāh  (Beo 1286a)	S/Sx:s (88X)
      ‘sword stained with blood’

   3. (c)

      Fyrst / forð gewāt  (Beo 210a)	S/S:x-s (33X)
      ‘the time went forth’

   4. (d)

      līcað / leng swā wēl  (Beo 1854a)	Sx/S:x:s (1X)
      ‘pleases the longer the better’

For type B variants with an Sxs foot, the frequency of direct realizations is 8 percent (see Table 4.5). For the Db variants in item (22), the frequency rises to 31 percent (56/178).

Table 4.9 compares the verse counts from Beowulf with counts from the other poems surveyed.

Like Beowulf, Brunanburh and Judith show a significantly enhanced frequency of direct Sxs realizations in the inherently complex Db and Dbx patterns (compare Table 4.5). The enhancement is less striking in Elene, with Sxs realizations barely ahead of S:x-s realizations. In Maldon, direct Sxs realizations fall well behind S:x-s realizations. Maldon’s frequency of direct realizations is even lower in type B, however, with 3 instances out of 102 (3 percent) as compared with 2 out of 16 in the Db variants (13 percent). The Maldon poet realized the Sxs foot as a compound more often in type Db, but direct realization of this foot had become more difficult to achieve in all verse types.

It was especially difficult to construct verses with an Sxxs foot realized as a compound word. Most instances occur in long type Db, which appears only twenty times in Beowulf (including a unique instance of expanded long Db). The inherent complexity of long type Db promotes realization of its compound foot in the simplest possible way.

1. 1. (a)

      sēon / sibbegedriht  (Beo 387a; cf. 729a)	S/Sxxs (4X)
      ‘to see the kindred band’

   2. (b) seah / (on) enta geweorc  (Beo 2717b; cf. 2774a)

      ‘looked on the giant-work’

   3. (c)

      Wīg / ealle fornam  (Beo 1080b)	S/Sx:x-s (15X)
      ‘war took away all’

   4. (d)

      hond / sweng ne oftēah  (Beo 1520b)	S/S:x:x-s (1X)
      ‘the hand did not withhold the stroke’

The four instances like (23a) include two with the compound entageweorc, which refers to a prehistoric kind of “cyclopean” stone construction. This is usually printed as a word group, as in (23b), but lack of alliteration on enta- in this verse shows that the following constituent had undergone subordination. Interpreted as a genitive compound, entageweorc has the expected semantic structure.²³ Direct realization of the Sxxs foot has a frequency of 20 percent in long type Db (4/20), as compared with less than 1 percent in long type B (1/174).

Table 4.10 adds counts for verses like (23a–d) in the other poems sampled, adding one realization not present in Beowulf.

Elene has only the variant most common in Beowulf. There are no long Db variants at all in Brunanburh or Judith. In Maldon, the two such variants add additional elements of complexity.

1. 1. (a)

      bæd / (þæt) beorna gehwylc  (Mald 257a)	S/(x)Sx:x-s
      ‘bade that each man (should)’

   2. (b)

      (ā)lȳfan / landes tō fela  (Mald 90a)	(x)Sx/Sx:x:s
      ‘allow too much land’

   3. (c)

      oncȳð / eorla gehwǣm  (Beo 1420a)	Sx/Sx:x-s
      ‘grief to each earl’

Item (24a) adds the complexity of an extrametrical word before the second foot. Only two of the twenty instances in Beowulf have an extrametrical word in this position (2367a and 2717a). If this complex variant had the same frequency in both poems, we would expect ten instances in Beowulf. In addition to its complex realization of the Sxxs foot, item (24b) adds other kinds of complexity. This variant has the pattern Sx/Sxxs, expanded long Db. A single variant in Maldon would correspond to ten instances in Beowulf but (24c) is the only such variant in the early poem. The Maldon variant also adds the complexity of anacrusis. The variants in the last two lines of Table 4.10 are the only ones that have independent function words in the foot-internal dip and do not imitate the structure of an Sxxs compound. Given one such variant in Maldon, we would expect about ten in Beowulf. The later poet exploits foot-internal dips more heavily in all verse types as locations for unstressed function words.

4.10 Evolution of Dips that Include Light Feet

Initial dips in types A3, B, and C contained light feet and had level-3 prominence. The number of extrametrical syllables was not strictly regulated in these dips and they stood ready to accommodate more function words. Maldon has an unusually high frequency of level-3 dips with more than three syllables.

1. 1. (a)

      Ne sceolon mē on þǣre / þēode (Mald 220a)	(type A3)
      ‘in that district they will not (need to rebuke) me’

   2. (b)

      ā hwīle þe hē / wǣpen mæge  (Mald 235b)	(type B)
      ‘while he may (hold) weapons’

   3. (c)

      Ne þurfon mē embe / Stūrmere  (Mald 249a)	(type C)
      ‘around Sturmere they will not need (to rebuke me)’

I scan these variants with an initial xx foot because the most prominent word in the dip has two syllables. The initial dip has seven syllables in (25a), five syllables in (25b), and six syllables in (25c).

Table 4.11 provides counts for initial dips with four or more syllables.

Maldon holds the record for the longest initial dip in types A3 and C. In type B, Maldon’s two initial dips with five syllables would correspond to twenty in Beowulf rather than the six actually attested. There are no apparent increases in frequency for Elene, Brunanburh, or Judith. Here as elsewhere we observe a sudden increase in the frequency of unstressed function words soon after catastrophic damage to the inflectional system.

4.11 Evolution of Verb Placement in the Poetic Line

Prehistoric SOV word order had become less regular in the era of Beowulf and its frequency continued to diminish as an SVO norm evolved. By the time of Maldon, we would expect verbs to appear less often verse-finally. Distinctions among Old English inflectional vowels had been lost a few decades before the events described in Maldon (section 4.4). The functions performed by inflections had to be clarified by other means. One was increasingly strict SVO order, which replaced case inflections as a way of distinguishing subjects from objects. In this section we will see how placement of finite verbs changed over time in Old English poetry.

Finite forms of wesan ‘to be’ provide a wealth of evidence for the prominence hierarchy (levels 1–6). The kind of lift or dip occupied by the finite verb and its level of prominence are provided to the right.

1. 1. (a)

      Hwæt / syndon gē  (Beo 237a)	Dominant lift (1)
      ‘what are you?’

   2. (b)

      sēlest / wǣre  (Beo 173b)	Subordinate lift (2)
      ‘would be best’

   3. (c)

      wǣron / æþelingas  (Beo 1804a)	Light foot (3)
      ‘the noblemen were (eager)’

   4. (d)

      sweord wæs / swātig  (Beo 1569a)	Foot-internal dip (4)
      ‘the sword was bloody’

   5. (e)

      Yrre / (wǣron) bēgen  (Beo 769b)	Extrametrical dip (5)
      ‘both were angry’

The low prominence of wesan makes its finite forms quite unsuitable for employment on the dominant lift. Item (26a) is the only example in Beowulf. In accord with the rule of precedence (section 4.2), the lift following syndon is occupied by a weakly stressed word rather than by a fully stressed noun or adjective. The weakly stressed word in this case is gē, a pronoun that occupies a level-2 lift with phrase-final stress. The suppletive form syndon has rather low frequency, appearing just five times in the poem, and is more prominent than wæs or wǣron, forms of extremely high frequency that derive from the same root as wesan. In (26b), wǣre acquires phrase-final stress and is well suited to the verse-final lift. This special stress is also detectable in Modern English, where it prevents contraction. We normally say she’s here in spoken English, for example, but always here she ís rather than *here she’s.

Finite forms of wesan occupy a subordinate lift 103 times in Beowulf. In item (26c), such a form occupies a level-three dip as a light foot. This verb has remarkably high frequency in level-three dips, with a total of 241 instances in the poem. No other verb in Beowulf has higher frequency at level 3 than at level 2, on a subordinate lift. Though less prominent than any other verb, wesan is more prominent than prefixes or independent function words like pronouns. As we drop to level 4, the level of foot-internal dips, occupation of such dips by finite wesan drops off sharply. There are twenty-nine instances like (26d), with finite wesan on an x position inside the foot. Only nine verses are like (26e), with finite wesan on an extrametrical position in the medial dip (at level 5). The Beowulf poet never places any kind of verb in anacrusis (at level 6).

Table 4.12 provides frequencies for finite wesan in all the poems surveyed.

All the poems employ finite wesan freely in level-3 dips, though the frequency is significantly lower for Elene than for Beowulf. Judith and Maldon have lower frequencies for verse-final placement (at level 2) and higher frequencies at level 4. The instances on the dominant lift in Judith and Maldon (at level 1) deserve special attention.

1. 1. (a)

      wǣron ætsomne  (Jud 255b)	Sx/(x)Sx
      ‘were together’

   2. (b)

      (þæt) wǣre / (hit ūre) hlāford  (Mald 240b)	(x)Sx/(xxx)Sx
      ‘that it was our lord’

Since the only comparable verse in Beowulf is (26a), (27a–b) represent a tenfold increase in frequency for the late poems. It is worth adding that wǣre and wǣron are high-frequency forms from the same root as wesan, unlike the suppletive form syndon that alliterates in Beowulf. Other unusual features of (27b) include a subordinating conjunction in anacrusis (þæt) and extension of the medial dip with three extrametrical syllables (hit ūre).

To double-check trends in Table 4.12, we need verbs more resistant to unstressed usage than wesan and we need to cast our net rather widely. The data set for this task includes all finite verbs that occupy a dip of any kind, even once, in any of the poems surveyed, and also occupy a lift at least once.²⁴

The closest grammatical relatives of Modern English be are auxiliary verbs like may, can, will, shall, do, and have. Though not as common as be, these auxiliaries are more common than main verbs and have lower prominence. Low prominence makes auxiliaries eligible for contraction with not. The form isn’t is paralleled by mayn’t, can’t, won’t, shan’t, don’t and haven’t. The verb want can be recognized as a main verb in part because it has no contracted form like *wantn’t. Old English ancestors of our modern auxiliaries are magan, cunnan, sculan, willan, don, and habban. Like its modern descendant have, habban can be used as a main verb with the meaning “possess.” Use as a main verb is also possible for Old English magan (“have power”), cunnan (“know”), willan (“want”), and don (“put on, behave”). Such main-verb usages will be treated separately.

Placement of auxiliaries in Beowulf is similar to placement of finite wesan except that instances on the subordinate lift (at level 2) outnumber instances serving as light feet (at level 3). The greater prominence of auxiliaries makes them slightly less suitable than wesan as occupants of x positions. Table 4.13 tallies all instances of finite auxiliaries from the poems surveyed. Percentages refer to totals at the right. The count includes both prefixed and unprefixed forms of these verbs. For the quasi-auxiliaries weorðan, lǣtan, mōtan, and onginnan, see Table 4.14.

Frequencies for verse-final placement decline from Beowulf to Maldon and frequencies for earlier placement rise. Judith trends away from Elene towards Maldon. Maldon places finite auxiliaries most freely in positions normally occupied by inflectional endings or prefixes, with a pair of instances in a foot-internal dip (at level 4), as compared with a single instance in Beowulf. If these unusual verses had the same frequency in both poems, we would expect twenty instances in Beowulf.

Other finite verbs that occupy a dip more than once in our sample have high frequency in Old English and fall for the most part into a few natural categories. Parenthesized counts show the proportion of stressed to unstressed usages in the poems surveyed.

1. 1 Finite forms of quasi-auxiliaries: lǣtan ‘cause’ (25:17), mōtan ‘be allowed’ (45:7), onginnan ‘begin’ (33:15), and weorðan ‘become’ (85:28).

2. 2 Finite verbs denoting speech acts: biddan ‘ask’ (10:7), cweðan ‘say’ (32:9), frignan ‘ask’ (18:3), secgan ‘say’ (22:6), sprecan ‘speak’ (20:3), and hātan ‘order’ (35:30).

3. 3 Finite verbs denoting acquisition of control over something, exercise of that control, transfer of it, or lack of it: āgan ‘own’ (8:11), beran ‘carry’ (37:2), main-verb dōn ‘put’ (5:3), findan ‘find’ (31:3), giefan ‘give’ (33:7), main-verb habban ‘possess’ (30:10), niman ‘take’ (36:5), scieppan ‘create’ (1:2), sellan ‘hand over’ (27:3), sendan ‘send’ (14:2), slēan ‘strike’ (14:2), and þurfan ‘need’ (19:5).

4. 4 Finite verbs locating an entity in space/time (as with the dative case) or tracking movement through space/time (as with the accusative case). These include finite forms of bīdan ‘abide’ (23:2), cuman ‘come’ (54:16), feallan ‘fall’ (14:2), gān ‘go’ (44:18), gewītan ‘go’ (23:14), hlēapan ‘leap’ (1:2), hweorfan ‘turn’ (17:3), licgan ‘lie’ (25:5), sittan ‘sit’ (28:2), standan ‘stand’ (35:3), and steppan ‘step’ (12:2).

5. 5 Finite verbs of mental life. These include finite forms of main-verb cunnan ‘know’ (25:2), hycgan ‘have in mind’ (7:2), hȳran ‘hear’ (30:7), sēon ‘see’ (38:7), wēnan ‘expect’ (20:3), witan ‘know’ (28:5), þencan ‘think’ (10:3), and þyncan ‘seem’ (13:4).

Table 4.14 shows the placement of verbs in categories 1–5, following the model of Table 4.13. The counts include prefixed and unprefixed forms.

Verbs in categories 1–5 occupy verse-final position with lower frequency in the two late poems. Maldon is remarkable for its employment of these verbs in foot-internal dips and as extrametrical syllables (at levels 4 and 5). The six examples in Maldon would correspond to sixty in Beowulf, but what we actually find is a single example. On Table 4.12, Maldon shows the highest frequency for auxiliaries on level-three positions and the lowest frequency for auxiliaries on level-two positions. On Table 4.13, however, Judith shows the most innovative frequencies for verbs in these positions, with Maldon in second place. The two late poets both innovate in response to language change but somewhat differently.

Table 4.15 provides frequencies for all remaining verbs in the data set, most of which occur just once in a dip: alecgan ‘lay down’ (6:1), cēosan ‘choose’ (6:1), cierran ‘turn’ (2:1), drincan ‘drink’ (2:1), fōn ‘grasp’ (25:1), gieldan ‘yield’ (12:1), healdan ‘hold’ (48:1), hliehhan ‘laugh’ (2:1), munan ‘remember’ (24:2), myntan ‘think’ (2:1), sēcan ‘seek’ (42:1), settan ‘set’ (11:1), unnan ‘grant’ (6:1), and wrecan ‘avenge’ (19:1).

Though somewhat more prominent than the verbs in Table 4.14, as indicated by their level-1 frequencies, most of these verbs belong to the same natural categories and might have been acceptable in the dip for that reason. To the control category we might add ālecgan, cēosan, fōn, gieldan, healdan, sēcan, settan, and unnan. Obvious additions to the mental life category are munan and myntan. The verbs hliehhan and wrecan are harder to distinguish from verbs that never appear in any kind of dip, which include the vast majority of main verbs.

Judith has the most innovative frequencies in Table 4.15. Maldon differs less sharply from Beowulf or Elene, though it has a somewhat higher frequency of unstressed usages. Although Maldon innovates most ambitiously in the least prominent metrical positions (at levels 4 and 5), Judith leads the way with the most prominent verbs. Maldon and Judith seem to represent distinct metrical dialects.²⁵ Such dialects result from discovery by individual poets of departures from norms acceptable to their audiences.

4.12 Evolution of Alliterative Patterns

In Beowulf, finite verbs sometimes alliterate alone before a prominently stressed word, violating Sievers’s law of precedence, but primarily when the verb is the first stressed word in its clause.²⁶ There are only two violations when the verb is clause-medial, with a stressed word of the same clause situated in a preceding verse.²⁷

1. 1. (a)

      lǣteð hworfan  (Beo 1728b)	Sx/Sx
      ‘lets wander’

   2. (b)

      wunode mid Finne  (Beo 1128b)	Sx/(x)Sx
      ‘dwelt with Finn’

In (28a), the weakly stressed quasi-auxiliary lǣteð alliterates before hworfan, a more prominently stressed infinitive. In (28b), a main verb alliterates before the more prominent object of a preposition. In these unusual cases, the poet uses modern word orders to obtain the required alliteration. Similar verses occur in Judith and Maldon.

1. 1. (a)

      þōhton tilian  (Jud 208b)	Sx/Sx
      ‘thought to supply’

   2. (b)

      drencte mid wīne  (Jud 29b; cf. 72b)	Sx/(x)Sx
      ‘drenched with wine’

   3. (c)

      hogode tō wīge  (Mald 128b)	Sx/(x)Sx
      ‘attended to warfare’

   4. (d)

      hē lēt him þā of handon  (Mald 7a)	(x)S:x/(xx)Sx
      ‘he let (a spear fly) then from his hands’

   5. (e)

      hē gehlēop þone eoh  (Mald 189a)	xx/S:xx:s
      ‘he mounted the horse’

Items (29a–b) from Judith are close syntactic parallels of (28a–b). Item (29c) from Maldon parallels (28b) and (29b). Like item (28a), (29d) has single alliteration on a finite form of lǣtan. Item (29e) translates word-for-word into modern English. Its SVO word order is achieved at the cost of pronominal anacrusis and alliteration on a clause-medial finite verb before a more prominent non-alliterating object. As examples (28a–b) show, the Beowulf poet could use modern word order to satisfy alliterative requirements. The later poets do so much more frequently, however. To the two instances in Judith we can add a third and even more innovative example: item (30b), discussed below. Given three instances in each of the late poems, we would expect about thirty in Beowulf rather than the single instance actually found.

Compounds normally alliterate in Beowulf. Only one verse in the poem has a non-alliterating compound after an alliterating finite verb: item (23b), repeated as (30a).

1. 1. (a)

      seah on entageweorc  (Beo 2717b)

      S/(x)Sxxs

   2. (b)

      salowigpāda sang hildelēoð  (Jud 211)	Sx/Sx // S/Sxs
      ‘the dark-coated (raven) sang a war-song’

The finite verb seah stands at the beginning of its clause in (30a). Alliteration on such a verb, though unusual, would not be unprecedented. On the other hand, the alliterating verb is altogether remarkable in (30b), a one-line sentence from Judith with SVO word order.²⁸ Here the verb stands in medial position, as in modern English, with the subject in the a-verse. Quite apart from its unusual alliteration, item (30b) has no syntactic parallel in Beowulf. There are no one-line clauses in Beowulf with a noun subject in the a-verse, the main verb in the b-verse, and a noun object following the verb. Absence of alliteration on the compound hildelēoð adds yet another unusual feature. SVO word order prevails in (30b) even though the result is spectacularly non-traditional. Such lines highlight the power of the innovating word order, providing linguistic evidence unobtainable from prose texts.

As Old English evolved into Middle English, the old intuitive guides to alliterative patterning became less and less helpful. In the era of SOV Germanic, when the final stress of grammatical constructions was weakest, poets were unlikely to forget that alliteration must occur on the first lift of each verse and must be avoided at the end of the line. When SVO order became common, the last stress of a grammatical construction was often the strongest one and alliterative patterns began to seem counterintuitive. As a poetic universal, the link between stress and alliteration would remain in force throughout the transition to SVO syntax. A line-final word would look more and more often as if it ought to alliterate. By the late Middle English period, alliteration on the line-final lift had become an acceptable departure from an archaic norm.²⁹

In Beowulf there is only one clear case of alliteration on a forbidden lift, item (31a).

1. 1. (a)

      þæt ic mid sweorde ofslōh  (Beo 574b)	xxx/Sx:x-s
      ‘that I slew with the sword’

   2. (b)

      hyre tōgēanes gān  (Jud 149b)	xxx/Sx:s
      ‘go towards her’

   3. (c)

      gifena in ðȳs ginnan grunde (Jud 2a)	Sx//(xx)Sx/Sx
      ‘of gifts in this wide earth’

   4. (d)

      sǣmen snelle (Mald 29b)	Ss/Sx
      ‘bold sailors’

   5. (e)

      hwæt þis folc segeð  (Mald 45b)	xxx/Sx
      ‘what this troop says’

   6. (f)

      sē wæs hāten Wulfstān  (Mald 75b)	xxxx/Ss
      ‘who was called Wulfstan’

   7. (g)

      Offa forhēawen  (Mald 288b)	xxx/Sx
      ‘Offa hewn down’

A handful of other apparent instances in Beowulf can be eliminated by “fairly certain emendation.”³⁰ Item (31b) from Judith is a similar instance with alliteration at the end of the line. Item (31c) is another non-traditional instance from Judith with alliteration on the last word of a hypermetrical a-verse. Items (31b–c) cannot be attributed to the sound change that differentiated palatal and velar g. Alliteration between these sounds is necessary for normal scansion elsewhere in Judith.³¹ Maldon provides another instance of double alliteration in the b-verse (31d). These instances from the late poems represent a tenfold increase in frequency relative to Beowulf.³² In (31e–g), the only alliteration occurs in the second of two prominently stressed words. These variants would be unacceptable as b-verses in Beowulf and equally unacceptable as a-verses. Given four instances of alliteration on a forbidden lift in Maldon, we would expect forty instances in Beowulf rather than a single instance. Here as elsewhere important trends will be missed if we simply count variants of a kind unattested in Beowulf. Trends toward Middle English meter show up best as changes in frequency, typically by an order of magnitude. Tracking frequencies also helps us deal with the power of formulaic conventions that resist language change. Such conventions perpetuate archaisms but with decreasing frequency over time.

Finite verbs of motion sometimes occupy a level-3 dip in Beowulf, but never when they alliterate. A remarkable exception appears in Judith.

1. (32)

   gewitan him wērigferhðe  (Jud 290b)	(x)xx//Sx/Sx
   ‘went away, weary in mind’

Item (32) must be a hypermetrical b-verse with the embedded type A1 verse realized as a compound. Normal scansion is possible only if alliterating -witan occupies the dip. If -witan were assigned to a lift (with obligatory resolution), the result would be (x)Sx//Sx/Sx, a hypermetrical pattern restricted to the a-verse. As SOV syntactic analogues for alliterative patterns were marginalized and traditional rules lost their link to intuition, such anomalies were likely to arise.

In Beowulf, word-initial sp-, st-, and sc- alliterate as clusters, with both consonants matching. Other kinds of cluster alliteration appear in late Old English poems, especially when the cluster begins with s.³³ Such a development would be expected as longer dips between alliterating syllables incorporated more words of intermediate prominence, words that sometimes alliterated. Since clusters have a lower probability of accidental matching, cluster alliteration highlights lifts more effectively in a more complex metrical environment. Cluster alliteration continued to extend its range in Middle English alliterative poetry.³⁴

4.13 A Gap in the Manuscript Record During the Early Middle English Period

After Maldon comes a gap of more than two centuries during which French and Latin works dominated manuscript culture while native English poetry fell out of favor.³⁵ Convergent lines of evidence show that poems in strict alliterative meter were excluded from the cultural mainstream throughout the Middle English period.³⁶ Alliterative poems that entered the mainstream, from Laȝamon’s Brut to Piers Plowman, did so by abandoning devices of traditional composition, with rhyme often replacing alliteration in the Brut, for example.³⁷ The influence of rhyme may be observable already in Maldon, line 271, where the half-lines are joined by perfect end-rhyme but the alliteration of s- with st- is non-traditional.³⁸ Turville-Petre argues that alliterative tradition was destroyed sometime after Maldon and reinvented during the later Middle English period. This hypothesis fails to explain why there are so many shared metrical features in the later poems.³⁹ The same problem confronts the hypothesis that the meter of sggk developed out of something like Ælfric’s alliterative compositions.⁴⁰

4.14 Metrical Change in Formulaic Traditions

The following chapters explain Middle English alliterative meter as the expected outcome of language change, building on a few initial assumptions.

1. 1 Alliterative poets inherited a formulaic technique of composition and worked strenuously to maintain traditional rules, modifying them only when language change made archaic requirements unsustainable.

2. 2 During the literary-historical gap, trends evident in late Old English meter continued in parallel with related linguistic developments.

3. 3 Universal constraints like the principle of closure guided metrical evolution, ensuring at all times that alliterative poetry could be scanned by intuition at the speed of performance.

4. 4 When old metrical rules were lost, new ones would be required for coherence. In accord with assumption 1, the new rules would be based on traditional norms that were still perceptible in a changed linguistic environment. Transforming these norms into rules would yield a more coherent set of verse patterns, facilitating inutitive scansion. As new rules were added, some late Old English trends might be pruned back, for example the trend toward extremely long dips in type C (section 4.10).

4.15 Some Rules of Middle English Alliterative Meter

At the other side of the literary-historical gap we find poems like Sir Gawain and the Green Knight (sggk). sggk is widely admired for vivid characters and intricate patterning of narrative detail. It seems safe to choose such a poem as a representative example of strict alliterative meter.

Until a few decades ago, researchers interested in sggk had nothing to draw on like the foundational work of Sievers and Bliss. Since the mid-1980s, however, constraints on Middle English meter have been identified and widely discussed.⁴¹ A strong consensus about rules for the b-verse has emerged, and some rules for the a-verse have been proposed in recent work.⁴² These findings are summarized in rules 1–4.

1. Rule 1 The b-verse must contain one and only one long dip with two or three metrical syllables.

2. Rule 2 The b-verse must contain two and only two syllables with metrically significant stress (i.e., two lifts).

3. Rule 3 The b-verse must have a strictly trochaic closure. Its final syllable must have zero stress and the immediately preceding syllable must be stressed.

4. Rule 4 The a-verse must contrast with the b-verse, for example in having two long dips or a non-trochaic closure. I will refer to rule 4 as the principle of asymmetry.⁴³

4.16 Parallel Developments in Cognate Germanic Meters

The evolution of Norse and Saxon alliterative meters can provide insight into what happened during the gap. Old Norse underwent widespread loss of unstressed syllables due to the influence of a very forceful primary stress. As a result, many verses in respected older poems were reduced to fewer than four metrical syllables.⁴⁴ Use of all such verses in fornyrðislag, which is much like the meter of Beowulf, would have created intolerable confusion about the number of feet in the verse. A short type S/Sx, for example, would have looked too much like a realization of the Ssx foot as an S:sx word group. Types that appeared to have traditional sanction were not discontinued, however. Instead they were incorporated into two new Eddic meters.⁴⁵ Short verses were accommodated in a new Eddic meter called ljóðaháttr that employed predictable placement as an aid to intuitive scansion. In ljóðaháttr, short verses occupied special positions in a well-defined stanzaic structure. Within a ljóðaháttr stanza, the audience could determine whether to interpret two words as the two feet of a small verse or as a realization of a compound foot. As English verse types became more difficult to scan, Middle English poets adopted a similar strategy, confining some verse types to the a-verse and others to the b-verse. This strategy built on traditional preferences for verse placement within the line.

Not every verse in prehistoric Norse poems had a linguistic structure allowing for contraction. Some verses retained their original length and were eventually too long for employment in fornyrðislag. For archaic long types another new meter was created: málaháttr. Some metrists have criticized málaháttr as an undisciplined form of fornyrðislag; but although more permissive in some respects, it is stricter in others.⁴⁶ In comparison with fornyrðislag, málaháttr had a far higher frequency of double alliteration in the opening half of the line, distinguishing the a-verse more sharply from the b-verse and creating a more salient line pattern. In fornyrðislag, a long dip was optional in types A, B, and C. In málaháttr, a long dip was obligatory in these types. Málaháttr verses were more sharply delineated by their syntax, with greatly enhanced frequency of one-verse clauses. In early Old English poetry, as in fornyrðislag, the verse was usually realized as a noun phrase, a verb phrase, a prepositional phrase, or some other sub-constituent of a clause; and sentences that filled several verses often began or ended in the middle of the line. More regular alignment of clause boundaries with line boundaries is already evident in the late Old English period. The frequency of end-stopped lines is about twice as high in Maldon as in earlier narrative poems.⁴⁷ In sggk, sentences that begin in the middle of the line are so rare that editors punctuate apparent instances as clauses (as for example in line 1487). One-verse clauses are much more common in sggk than in Beowulf.

Another Norse metrical innovation is dróttkvætt (‘court meter’). Like the hypermetrical a-verse in Beowulf, the dróttkvætt verse is constructed by adding a stressed word to a normal verse. In drottkvætt, the added word stands at the end of the verse. This closing word must always have the trochaic pattern Sx, the optimal word pattern in Old Norse as in Old English. Since dróttkvætt is notoriously strict in other respects, it is not surprising that the optimal word pattern is required in the foot most strictly governed by the principle of closure. As metrical rules were lost or weakened, a fixed trochaic closure would highlight the structure of the line as a coherent aesthetic unit progressing from metrical tension to metrical release (cf. section 1.12).

Old English hypermetrical b-verses must appear in clusters. This requirement distinguishes them from type A1 b-verses with anacrusis (section 3.13). Some variants in sggk look like hypermetrical b-verses but show no tendency to cluster. For insight into the resulting problems of scansion, we can turn to Old Saxon alliterative meter, in which some hypermetrical verses were clustered and others were not. Since Old Saxon was evolving more rapidly than Old English, we would expect the Old Saxon Heliand to resemble Middle English poems in some respects. As we shall see in Chapter 5, sggk and the Heliand use the same technique to distinguish type A1 b-verses from unclustered hypermetrical b-verses.

4.17 Major Developments in the Evolution of Middle English Meter

In following chapters we will see how Middle English poets repaired the damage inflicted on the meter by language change. We will trace the further evolution of the verse types introduced in Chapter 3, showing how each type was transformed by the consolidation of SVO syntax and by reorganization of the metrical system to create new kinds of metrical coherence. In broad outline, the evolution of Middle English meter proceeded as below.

1. 1 Type A1 continued to provide the most favorable site for metrical experiments. Its simpler variants had always favored the b-verse. Increasing frequency in the b-verse, already conspicuous in Maldon, continued in the gap as verses of unprecedented complexity crowded the a-verse. The simpler A1 variants were largely confined to the b-verse by the fourteenth century. A long, complex A1 variant confined to the a-verse in Beowulf, with disyllabic anacrusis and extension of the medial dip, became the most common variant in the Middle English a-verse. A1 verses of unprecedented length appeared.

2. 2 The shift of type B to the a-verse, already evident in Maldon, continued in the gap and went to completion. In sggk, type B was excluded from the b-verse.

3. 3 As the productivity of compounding declined, variants of type C with a word group in the Ssx foot began to look like optimal realizations. These variants had always favored the b-verse. By the fourteenth century, type C was vanishingly rare in the a-verse.

4. 4 Types A2a and E were eliminated due to unsustainable constraints on function words. Strict rules for placement of the major syntactic break were no longer essential in the surviving heavy types, but these rules survived as stylistic preferences.

5. 5 The a-verse had always been the preferred location for heavy types with two primary lifts followed by a secondary lift. A decline in the productivity of compounding made heavy types more difficult to scan by intuition, hence even less appropriate to the b-verse. By the fourteenth century, the surviving heavy types had been excluded from the b-verse.

6. 6 Restriction of metrical patterns to the a-verse or b-verse became an attractive possibility as the system continued along late Old English trajectories. At some point during the gap, alliterative poets realized that patterns characteristic of the b-verse tended to have a trochaic verse-final word, no more than one long dip, and no more than two metrically significant stresses. Regularization of these tendencies compensated for loss of some traditional rules.

7. 7 A decline in the productivity of compounding and a strengthening SVO norm destroyed the old system of metrical subordination, breaking the link between alliterative patterns and patterns of SOV phrasal stress. By the fourteenth century, the second lift in types B and C was no longer perceptibly subordinate and was reanalyzed as a primary lift. Marginal subtypes with a single lift were eliminated from the metrical system. These included type A3 (xx/Sx, xx/Ss) and the lightest C subtypes (x/Sxx, xx/Sxx). All surviving verse types now had two primary lifts, and a new metrical constant had emerged.⁴⁸

8. 8 Elimination of types E and A2a allowed for easy identification of primary lifts in the surviving heavy verse types. These were always the first two lifts. Alliteration on the secondary lift was now a permissible option.⁴⁹

9. 9 Alliteration was now obligatory on both primary lifts in all a-verses, enhancing asymmetry with the b-verse.

10. 10 The SVO norm exerted very strong influence by the fourteenth century. The most prominent stress in a two-stress English phrase was now normally the last one rather than the first one. The universal relation between stress and alliteration promoted optional alliteration on the final word in the b-verse. In Old English a-verses, alliteration had acted like compound stress within heavy feet, elevating the prominence of the first stress and subordinating any prominent stress on the s position (section 3.6). After word-foot principles had been lost, the smallest metrical pattern was the verse pattern and alliteration acted like phrasal stress. Alliteration on the first two lifts subordinated the third lift, paralleling Modern English three-stress phrases in which the third stress is weakest.⁵⁰ Since phrasal stress subordinates syllables less deeply than compound stress, a Middle English alliterative rule based on phrasal stress allowed for more alliterating syllables per verse.⁵¹

11. 11 Lengthening of vowels in open syllables eliminated most resolvable sequences.⁵² This had no effect on the inventory of verse types. Words that were resolvable in Old English simply standardized the value they had when unresolved. Basic metrical patterns were unaffected because resolved constituents were marked (exceptional) in relation to long stressed syllables.⁵³ In Beowulf, resolution created marked realizations of patterns normally realized with long vowels.

4.18 Summary

The word-foot theory distinguishes among several kinds of x position. Extrametrical x positions are normally occupied by unstressed prefixes. In word feet with an S position, x positions are normally occupied by unstressed syllables of stressed words. The x positions of light word feet are normally occupied by independent function words. Assigning a syllable of greater than normal prominence to an x position creates a metrically significant mismatch, inhibiting placement of finite verbs and independent function words at specific locations in the poetic line.

Given the emphasis on cultural preservation in formulaic poetry, a general increase in mismatch is most plausibly attributed to change in the poet’s language. Within the word-foot theory, the influence of a given language change is predicted to vary according to the level of metrical resistance. This prediction is fulfilled in a wide variety of cases. In the late Old English period, a strengthening SVO norm shifted finite verbs from their preferred location on a line-final lift into light verse-initial feet and, less often, into more resistant medial dips. Finite verbs were never shifted into verse-initial anacrusis dips, where resistance was strongest. Catastrophic damage to the inflectional system made independent function words more necessary and increased the average length of dips in the poetic line. As predicted, the preferred location for unusually long dips was verse-initially in types with a light initial foot, the location that had always been most tolerant of extrametrical syllables. Decline in the productivity of compounding had a significant effect on the foot patterns Sxs and Sxxs, which corresponded to compounds of restricted frequency. By the late Old English period, the x positions in these compound feet were occupied more often by independent unstressed words. Sxs and Sxxs feet became more difficult to associate with compound words, adding complexity to verse types that employed them and attracting these types to the a-verse. Some predictions of the word-foot theory involve small numbers of examples, but there are many such predictions and the cumulative evidence trends overwhelmingly in the same direction.