7 Metrical Resolution in Early Middle English

In the previous chapter I reviewed the evidence of ie-reduction in southern and western Middle English. It is natural to wonder whether this finds any echo in the poetry of the time, the way that the prosodic structure underlying Old English high-vowel loss is closely paralleled in verse. After a review of the poetic landscape of 12th-and 13th-century English-speaking Britain – a complex backdrop to the specific works under investigation – I take a look at two poems from the 12th century that show metrical resolution. The first is by Laȝamon,¹ the son of Leouenað, who wrote a a very long history of the Britons now usually called the Brut (§7.2). His rather messy and elusive metre, which mixes alliteration and rhyme, shows indirect but extensive evidence for metrical resolution, and the general equivalence of light disyllabic sequences (LX) with a single heavy syllable (H). The other poem is the Moral Ode (§7.3), whose evidence for resolution has been subject to some debate in recent years.

Old English poetry, at least as it survives, was entirely alliterative. Even the so-called Rhyming Poem in the Exeter Book still observes strict patterns of alliteration, with end-rhyme featuring as an additional further element. By the 12th century, however, vernacular poetic fashions were at least partly shifting towards models in Latin and French. This involved not only an increasing orientation towards rhyme on the part of many poets, but frequently the importation of new rhythmical forms. For example, the septenarius metre, very common in Latin, was employed in both the Moral Ode (§7.3) and the Ormulum.

Such forms of verse are usually relatively transparent, showing a regular metrical scheme constructed along the same lines as much later English poetry: iambic or trochaic feet, repeated in certain numbers and arrangements, and subject to familiar licences such as foot inversion, promotion or demotion of monosyllables, and elision. This is not to say that extracting prosodic information from such poems is problem-free. Especially with the Moral Ode, the textual situation and number of manuscript variants makes it difficult to fully unravel the poem’s history: it is not always clear which forms can be attributed to poets (whether the first composer, or metrically aware copyists), and which are due to scribal alteration done without any concern for verse form. Orrm’s long poem, the Ormulum, is textually more straightforward: the single surviving copy (perhaps the only medieval copy ever produced) is probably in Orrm’s own hand.

Much more difficult is the alliterative poetry of this period. The septenarius, octosyllabic verse, and other new poetic styles supplemented the alliterative tradition, but did not displace it. As discussed in §3.3, there is a substantial body of alliterative verse attested from the 14th century and later that forms a fairly coherent metrical corpus with its own distinctive rules and regularities, which have been increasingly decoded since the late 1980s. These rhythms are not those of Old English verse, but they are also not those of Latin or French poetry: they seem rather to represent the outcome of a long period of development from Old English verse along ‘internal’ lines (Russom 2004a,b; Fulk 2004).²

The exact nature of this development is hard to trace in detail, and has been the subject of some debate. Between the copying out of the great Old English poetic codices – MS Junius 11, the Vercelli Book, the Exeter Book, and the Beowulf manuscript – roughly around the turn of the millennium and the poems of the so-called Alliterative Revival in the 14th century, there is a relative dearth of surviving written alliterative poetry. ‘Relative dearth’ is not ‘complete absence’, and there are a fair number of poems found in this gap (Oakden 1968: 133–151; Weiskott 2016: 76, 175–182). Still most of these are relatively brief, in striking contrast to the earlier and later periods, and there is virtually nothing from the later 13th and earlier 14th centuries. This might simply be due to the loss of texts, as Weiskott (2016) argues, though Pascual (2017) replies with a vigorous defence of the possibility that a robust oral tradition carried alliterative verse through these periods. What is important here is that, as both Weiskott and Pascual emphatically agree, there must have been a continuous history of use and development of English alliterative verse (written or oral) during these centuries, but direct textual sources for studying this are, for whatever reason, relatively scarce, and generally short or fragmentary when they do survive.

The outstanding exception to this generalisation is the work of Laȝamon. At slightly over 16,000 lines in its longer version, his Brut, partly alliterating and partly rhyming, towers over not only other alliterative poems of his time in size, but is substantially longer than Beowulf, the longest poem in Old English (this is true even if one removes all of Laȝamon’s rhyming lines). Despite its many difficulties, it is one of the richest and most valuable sources for early Middle English to survive.

The title Brut for Laȝamon’s only known work is modern; in the manuscripts it is known as Hystoria Brutonum or Libri Brutonum. It exists in two manuscripts, which differ notably in many respects. The more conservative manuscript is Cotton Caligula A ix (Caligula), the other is Cotton Otho C xiii (Otho). The latter represents an abridgement and reworking of the text, with many passages condensed or trimmed (both intentionally by a redactor as well as accidentally by fire), and the specific wording of many lines changed, frequently modernising the diction (Dance 2003: 56–60). Like many others, I focus entirely on the Caligula manuscript here, though there is much to be gained by considering the two versions together (Cooper 2013).

Both manuscripts date closer to 1300 than 1200 (Otho is probably the later of the two), but Laȝamon most likely composed the Brut in the later 12th century, before 1216 at the very latest (Le Saux 1989: 1–10). In a prologue to his poem, he claims to have been a priest at a place called Ernleȝe, which is identified as Areley Kings, a village some miles north of Worcester. This places him in the West Midlands: broadly the same region as the AB texts discussed in the previous chapter, and their dialect is, though not identical with, closely related to Laȝamon’s.

The metrical system of the Brut has proven rather troublesome to pin down. The clearest feature is a familiar one from other forms of alliterative verse: long lines are divided into two halves – an on-verse and an off-verse – which are linked together variously by either alliteration (not always on the first stresses, in contrast to Old English verse) or end-rhyme (which can sometimes strike the modern ear as rather loose or forced). In the manuscripts, the half-line break is graphically indicated by a punctus elevatus, a mark resembling an upside-down semicolon, though I will simply use extra white space as I do for other metres:³

(118) Ān prēost wes on lēoden Laȝamon wes i·hōten

‘There was a priest among the people, he was called Laȝamon’ (Brut 1)

(119) On Italiȝe hēo cōmen to londe þar Rōme nou on stondeð

‘They came to land in Italy where Rome now stands’ (Brut 55)

For anyone interested in resolution, such matters of rhyme and alliteration are among the least interesting aspects of a metre, except as clues to stress (§6.4.3). The internal rhythmical organisation of these half-lines is the important thing. In this respect Laȝamon was for many years very ill-served in the scholarly literature. As Cornelius (2017: 82; cf. 177, n. 71) aptly puts it:

Indeed, the enduring legacy of Blake’s 1969 intervention is that subsequent studies of Lawman’s “prosody” have typically been studies of alliteration, assonance, and rhyme, leaving the metrical structure itself unexamined.

An important advance in understanding Laȝamon’s metre has been made by Yakovlev (2008: ch. 3), who, building on Hanna (1995), successfully applies ideas developed with regard to later Middle English alliterative metre (§3.3) to establish, if not a fully worked out metrical system, at least a set of clear regularities and trends that show Laȝamon’s metre to be much less chaotic and unregulated than it appeared to earlier generations of scholarship.

Not all mysteries are by any means solved, and Laȝamon’s verse is certainly not identical either to Old English verse or to the later Middle English systems. Yakovlev (2008: 208–210) instead treats Laȝamon’s rhythm as an ‘intermediate stage’ (Weiskott 2016: 73 uses the phrase ‘evolutionary missing link’) between these older and later phases. Strictly speaking, Pascual (2017: 257) is correct that this can’t be literally true: the use of rhyme in the Brut and the closely related Soul’s Address to the Body (Moffat 1987)⁴ mean that these works represent a slightly different trajectory than the one that led to the poems of the Alliterative Revival. Nonetheless, Laȝamon has incorporated metrical innovations that are also found in later Middle English verse, and Yakovlev’s success in viewing Laȝamon in light of what recent scholarship has uncovered about the workings of these later alliterative metres speaks for itself. Laȝamon may not be the direct forerunner of the Pearl Poet, but he and those like him were close cousins (poetically speaking) of those who were. Historical metrical trajectories aside, Yakovlev’s findings shed considerable light on Laȝamon’s metre in synchronic terms, and his approach provides a basis for investigating resolution.

Yakovlev (2008: 161–162) identifies 17 metrical principles at work in the Brut, including both abstract statements about verse structure, and specific points of linguistic-metrical correspondence, such as elision (which he does not find evidence for; point 10). Abstracting away from this detailed summary, a few points emerge in terms of what the basic building blocks of Laȝamon’s verse are, and how they are put together.

These basic elements are familiar from descriptions of both Old (§3.1) and later Middle English (§3.3) alliterative verse: a stressed unit is a lift (S), and a weak one a dip (w). Among dips, there is an important difference between a simple monosyllabic dip and a long dip with two or more syllables. The way these elements are put together is in general less obviously systematic than in poems such as Sir Gawain and the Green Knight, but especially in the off-verses there are a number of regularities, and very frequently the rhythms do correspond well to those of later Middle English. Take line 55b, for instance, extracted from (119) and metrically annotated:

(120) þar Rṓme nou on stóndeð

‘where Rome now stands’

This has the rhythm wSwwwSw. In more abstract terms, there is one short dip, a lift, a long dip, a second lift, and a final short dip. This checks all of the boxes for a classical Middle English off-verse: two lifts, one (and only one) long dip, and an ending in Sw.

As in later Middle English poetry, aside from the very final position, short dips of just one unstressed syllable seem more or less optional, included or omitted freely:

(121) ségge to·súmne

‘say together’ (Brut 32b)

The rhythm is SwwSw, but in structural terms this is essentially equivalent to (120), as it would be in later Middle English. The basic elements in both are the two stresses, the long dip, and the final weak syllable. Neither the extra short dip at the beginning nor the varying syllable count within the long dip seem to matter much.

The similarities with later Middle English verse discussed so far are striking, but Laȝamon includes a substantial minority of verses that do not conform to these rules. Take the following line:

(122) and þā máðmes of his lónd

‘and those treasures from his land’ (Brut 450b)

The rhythm here may be notated as wwSwwS. This still has two lifts, but the arrangement of the dips is very different. There are two long dips instead of just one, and the line ends with a stressed syllable instead of the standard trochee. It is clear that Laȝamon does not follow the classical Middle English rules (which, after all, may not have fully come into being yet), but this does not mean that there are no trends in his rhythm. In particular, the presence of two anomalies together – the two long dips and the monosyllabic ending – is intriguing, both as a potential means of understanding Laȝamon’s metre better in general, and as a way of identifying whether LX words resolve (to behave as monosyllables) or not (and behave as trochees).

In his study based on a 600-line sample of the Brut, Yakovlev (2008: 217–221, 252–260, 262) concluded that the poem does show a ‘metrical equivalence of the “short+any” sequence to the long syllable’. That is, he argued that Laȝamon made use of metrical resolution. To test whether this holds up in a larger sample, and to focus in more narrowly on resolution alone, I have taken a sample of Laȝamon’s poem, and parsed each off-verse into two components:⁵

Other shapes for final words occur too rarely to give much useful information, and in any case I am not sure what rhythm to assign to longer names such as Cassibellaune and Asclepidiot. For the initial, I ignore monosyllabic dips and take no note of the exact number of syllables in long dips, so that both (120) and (121) are notated as Sww for the initial, plus HX for the final word.

The idea behind this division is essentially to test how certain linguistic units (the final words, notated in terms of their linguistic structure: H, HL, etc.) relate to metrical patterns (the initials, notated as metrical abstractions: S, ww, etc.). The central question is how LX and LXX words behave. Do LX words tend to follow the same range of initials as H words (which would imply resolution, LX = metrical S), or the kinds of patterns associated with HX words (which would suggest non-resolution, LX = metrical Sw)? And if resolution is present in LX words, then LXX words might be expected to behave metrically like HX ones (both metrically Sw) and unlike HXX words (metrically Sww).

The various arrangements of lifts and long dips in the initial in my sample fell out into six broad types. The examples for each give the full half-line, with the final word separated from the initial by |. I have tried to pick examples that illustrate how extra weak syllables are reduced down to these patterns:

The selection of these categories is strongly influenced by the features that Yakovlev has found to be of metrical relevance. Off-verses with two stresses before the final word (so three, or occasionally more, in total) can vary in their internal rhythms, but as a group seem to be distinguished from the much more ‘standard’ type, with two stresses in total (Yakovlev 2008: 246–248). My distinction between types 1 and 2 aligns with the suggestion that the latter type might be more associated with monosyllabic final words (Yakovlev 2008: 244–246). Having made these divisions, the other types follow as the remaining possibilities on a similar level of abstraction.

I do not claim that these are metrical ‘types’ with any genuine realities. They are provisional classifications of parts of verses, whose final words have been cut off. Their value lies entirely in the possibility that some of these partial contours tend to be followed by monosyllables, others by disyllables: this is a contingent sorting of the data for the narrow purposes of testing for metrical resolution.

I should also add a caveat that will already be apparent to anyone who has looked at Laȝamon’s metre: there are many uncertainties in how to assign stress patterns. The following verses illustrate the two most typical problems:

(123) þat hē a·midde to·clǣf

‘so that he cleaved (the helm) apart’ (Brut 10688b)

(124) bā bi dæie and bi niht

‘both by day and by night’ (Brut 1976b)

In (123), the overall pattern seems to be wwwSwwS, which, taken at face value, should be parsed with an initial of category 3 and a final word H. However, Yakovlev (2008: 198–200) argues that Laȝamon has an optional prefix licence, whereby unstressed prefixes – such as the to in to·clǣf – can be metrically ignored if need be (compare appendix E.1). If prefixes are set aside, then the scansion is instead wwSwS, with a category 2 initial. Yakovlev (2008: 244) in fact applies the prefix licence to this exact verse. I am much more sceptical about the arguments for the prefix licence in Laȝamon, however, and provisionally count syllables where they are found in the manuscript.

A different kind of uncertainty stems from the difficulty in assigning stress or metrical ictus to many words. Is the bā ‘both’ in (124) stressed or not? The answer is not immediately obvious, and there are many other words which present similar uncertainties (also remember the stress shifts discussed in §6.4.3). Yakovlev (2008: 195–198) lays out some useful heuristics for identifying stress, but these results should be regarded as provisional, and in any case, in my sample bā does not occur frequently enough to apply Yakovlev’s tests.

To keep uncertainties to a minimum, the data presented in the following sections is based only on scansions I consider reasonably ‘secure’. This is really a gradient metric, but for present purposes this means setting aside verses where relevant vowel lengths (particularly in proper names) seemed too uncertain, where there were severe uncertainties about the stresses of words, or where there were significant and relevant textual problems (though I have accepted simple and straightforward emendations based on non-metrical grounds). Taken together, the various problems affect a relatively large number of verses, and what remains is a core of some 1,835 verses, which I list in appendix G.

Table 7.1 breaks down the range of types of initials found with final H words and with final LX. The first thing to note is that there is little in the way of absolutes. Either type of word can occur after almost any kind of initial. It is this kind of flexibility that has contributed to the long-standing impression of chaos in Laȝamon’s metre.

That said, there are clear trends, and final H and LX words are not distributed evenly across different kinds of initials – and moreover, they broadly share the same relative ordering of preferences. The clearest pattern is that both H-and LX-final words come most often in verses with a single further lift and two long dips (initial type 3). That is, the most typical kind of verse ending in these word-shapes is:

(122) and þā máðmes of his | lónd

‘and those treasures from his land’ (Brut 450b)

(125) þēr wēore fḗondes to | fĕole

‘there were too many enemies’ (Brut 645b)

For both types of word, the second most common pattern is type 2, a verse beginning with a long dip followed by a lift (with or without a short dip following this):

(126) þat wes a sélcuð | bĕarn

‘that was a marvellous child’ (Brut 142b)

(127) þe wes i kínges | stúde

‘who was in the king’s place’ (Brut 121b)

These two kinds of initials together account for 75 per cent of verses ending in H words such as lond, and 67.7 per cent of LX words such as fĕole. Type 1 and 4 initials constitute most of the remaining verses. In itself, none of this data speaks strongly against a general metrical equivalence of H and LX words in line-final positions, but table 7.1 on its own can demonstrate no more than that. What is needed is a contrast in behaviour between, on the one hand, H/LX words and, on the other, HX/LXX words.

The majority of off-verses in the Brut end in HX sequences. As explained in note 5, the data for this word-type in table 7.2 is based a scansion of 484 verses extracted from across the first 8,000 lines of the poem, 328 of which scan clearly enough that I am willing to accept them as the basis of analysis. I also include the data for HXX words such as lēuede ‘believed’, which form a notable contrast to HX and LXX endings.

Whether a verse ends in an HX word such as stondeð or an LXX word such as æðele, the preferences for the kind of initial that precedes it are strikingly similar.

Both word types are most frequently found in the most common type of off-verse in the Brut: type 1, a verse with a lift (perhaps preceded by a single unstressed word), followed by a dip, and concluding with a HX word or what seems to be its metrical equivalent, LXX. Typical examples are:

(120) þar Rṓme nou on | stóndeð

‘where Rome now stands’ (Brut 55b)

(128) his lḗode hine | hắteden

‘his people hated him’ (Brut 3506b)

While this is the most common type of initial for HXX finals as well, their preference is notably more muted, a moderate plurality rather than a slight majority.

The second most frequent type also involves a single preceding lift and one long dip, but in the other order (type 2):

(129) and of gṓde | lónde

‘and from a good country’ (Brut 2028b)

(130) þat hē wolde þār cástel | mắkian

‘that he wanted to make a castle there’ (Brut 826b)

Again, there is a clear contrast with HXX-final verses, which occur in this type at a considerably lower rate.

In a smaller minority of verses there is no obvious further stress beyond the final word, apparently of type 5:

(131) æfter þone | kínge

‘after the king’ (Brut 1026b)

(132) bēo swīðe | swĭ́kele

‘are very treacherous’ (Brut 7909b)⁶

This type is a smallish remainder for HX and LXX verses, contrasting both with H and LL verses (where apparently stressless initials are vanishingly rare) and with HXX verses, where such contours are remarkably common, in second place in terms of frequency:

(133) efter his | álderen

‘following his elders’ (Brut 3438b)

This point is a bit fuzzy, since it is possible that in at least some of these initials, a lower-stress word should be seen as metrically promoted. Still, it is striking that there should be a clear grouping of this kind of initial according to the final word.

The other possible initials are found, but generally only in smaller numbers before any of these word-shapes.

Taken together, the preceding subsections provide a strong basis for assuming resolution in Laȝamon. In §7.2.3.2, both final H and final LX showed a very strong association for verses with two long dips (type 3), a moderate dispreference for verses with a lift followed by a single long dip (type 1), and a strong dispreference for verses without any further stresses at all (type 5). As seen in §7.2.3.3, both HX and LXX words (which behave remarkably parallel) show a very different set of preferences: they most often occur after type 1 initials, are slightly more tolerant of following type 5, and show a marked dispreference for type 3.

In other words, H (lond) and LX (stude) tend to occur after one spectrum of verse contours (both are metrically S), while HX (stondeð) and LXX (makian) tend to occur in a distinct spread (both metrically Sw); HXX words (lēuede) have a different set of preferences yet (metrically Sww). These are not absolute, categorical rules (though it is possible that a better understanding of Laȝamon’s metre in the future might sharpen the trends), but notable preferences that are clearly visible when enough verses are surveyed.

These equivalencies strongly suggest that Laȝamon employed metrical resolution. If he did not, then we would expect a set of parallel behaviours that do not occur: that LX would behave similarly to HX, and that LXX would be like HXX. The data in table 7.1 shows the metrical behaviour of final S, whether this is linguistically H or LX; and similarly, the first two columns of table 7.2 show the metrical preferences of final Sw, whether this is filled by HX or LXX.

Strictly speaking, this data demonstrates resolution only for the final words of the line. Yakovlev (2008: 254–260) further argues that resolution takes place freely in any stressed position. Though I have not systematically annotated my data to fully test this claim, my strong impression is that Yakovlev’s argument holds up without qualification, and that medial lifts resolve as readily as final ones.

The data discussed so far seems to establish resolution as a clear feature of Laȝamon’s metrical system. But just what is this system, and how systematic will it prove in the end? Any answers to these questions will probably be reached step-by-step. Hanna (1995) and especially Yakovlev (2008) have already made the biggest initial leaps forward, and done much to clarify the relevant metrical units, and to make a start on cutting through the thornier problems of stress assignment that bedevil any attempt to scan the work of Leouenað’s son. I have tried to make a very focused and specific contribution, confirming Yakovlev’s suspicion of resolution in the Brut. Each step taken reduces the uncertainties a little more, and allows other patterns to be identified and analysed, and I am optimistic that this process can be carried on further.

I would emphasise that the approach taken here of dividing out very broad-brush categories of initials is designed to allow a fairly large amount of data relating to resolution to be gathered and evaluated in aggregate. I want to reiterate that I make no claims that this division captures essential metrical properties of Laȝamon’s verse (for instance, I do not suggest that he conceived of his metre as having a metrical break after the initial, and my use of | in metrical markups is purely for convenience in exposition). Some metrical regularities are so strong that they emerge even in this very rough categorisation: 76.1 per cent of Sw-final verses (HX and LXX) conform to the later Middle English off-verse pattern of, in total, two lifts, a single long dip, and a verse-final short dip. And among verses ending in S (H or LX), there is a striking preference for verses beginning with long dips, with 73.6 per cent of such endings following type 2 and 3 initials. This might imply some kind of principle of metrical compensation, where the lack of a verse-final dip is ‘made up’ for by beginning with a longer dip.⁷

But this way of approaching the metre cannot answer all the interesting questions, and there is a great deal of fuzziness in many areas. The patterns I have grouped together as initial type 2 (a long dip followed by a lift and an optional short dip, followed by the ending), for instance, are a robust minority for both S-and Sw-final off-verses, and some nearly exact parallels can be cited:

(134) þe on þan londe wes

‘who was in the land’ (Brut 65b)

(135) ⁊ fram þan londe hælde

‘and passed from the land’ (Brut 3048b)

Why should this kind of overlap be so relatively frequent, accounting for 25 per cent of S-final, and 21.6 per cent of Sw-final off-verses? Is this apparent overlap really genuine? Examples such as (134) and (135) are striking, but it may well be that a closer look at the average lengths of the first dips, the presence or absence of a medial short dip, the precise linguistic material used to fill the dips, the kinds of on-verses they pair with, or some other factor will reveal larger aggregate distinctions and trends that are invisible under the divisions adopted here. The work that remains to be done on Laȝamon’s metre is very substantial indeed, and goes far beyond the questions of resolution that I have concentrated on.

Laȝamon’s Brut may not be the only source of evidence for resolution in Middle English verse. Fulk (2002), building on Schipper (1910), has argued persuasively that the 12th-century Moral Ode (Poema Morale, Conduct of Life) employs metrical resolution. Here are the first four lines, adapted from the Lambeth text (Payne 2018: 505–506), with reference to Fulk (2002: 345, 2012a: 166):⁸

(136) Ich ém nū álder þén^e ich wés a wíntre ánd a lā́re.

Ich wélde mā́re þén^e ich déde;  mi wít āh t^ṓ9 bōn mā́re.

Wēl lóng^e ich hábbe chíld i·bṓn  a wórde ánd a dḗde;

Þā́h ich bṓ a wíntre ā́ld, tō ȝúng ich ém on rḗde.

I am older than I was in years and in learning. I have more control than I did; my understanding ought to be greater. Quite long I have been a child in word and deed; though I am old in years, I am too young in wisdom.

(Fulk 2002: 345)

I have attempted to explicitly convey the metrical form, the septenarius: each line is divided into to two verses (half-lines), elided e is written in superscript, syllables carrying metrical ictus are signalled with acutes, and long vowels are indicated with macrons. Underlining signals potential resolution. The rhythm is broadly iambic, with four feet in the first half-line, and three in the second (for seven feet in total, whence the name septenarius). Like most iambic verse, there is some leniency in the stress assignment of monosyllables and in the shape of early feet, such as the short foot beginning line 4.¹⁰ By contrast, the ends of half-lines are fairly strictly regulated (cf. §3.4.2): on-verses usually end in stressed monosyllables (wes, bōn, āld), and off-verses end in heavy trochees of the shape HX: that is, a heavy, stressed syllable followed by a weak, unstressed syllable (lāre, māre, dēde, rēde).

This rather rigid structure is usually, in the earlier parts of half-lines, filled by syllables with no reference to their weight. Take the following off-verse:

(137) þe hít for·ȝĕ́teð sṓne

‘who forgets it straightaway’ (Moral Ode 39b)

This scans with regularly alternating stresses, and both hít for and ȝĕ́teð count as sequences of two metrical positions, the difference in the weights of the (metrically) stressed syllables notwithstanding.

Matters are different at the ends of half-lines, where two interesting features related to syllable weight can be observed. Firstly, the stresses of the trochees that end the line overall are almost always heavy. Words such as ȝĕteð, dŭre ‘door’, or spĕken ‘speak’ are avoided in this position (Minkova 2016: 131–132). This provides good evidence for vowel quantities (open-syllable lengthening has not yet taken place) and at least some kind of stress-to-weight principle for the line-final stress. It may also provide some evidence for resolution, since occasionally words such as wunien ‘remain’ and sunien ‘avoid’ are found in the same position (Minkova 2016: 132).

The other interesting feature comes at the end of the first half-line, where Fulk (2002: 346–350) identifies evidence of resolution. The on-verse typically ends in a single monosyllable H, such as wes ‘was’ (line 1) or āld ‘old’ (line 4), but a not insignificant minority end in LX words such as dede ‘did’ (line 2) and stude ‘place’ (line 43). From the first 100 lines of the poem, Fulk finds 20 such examples that involve no serious complications of etymology or metrical context, and 14 more than might well show resolution, but which are more ambiguous (usually because of uncertainies in scansion). Certainly at first glance this looks like a weighty body of evidence for the equivalence of LX and H: that is, for resolution.

In a response to Fulk, Minkova (2016) objects to his identification of resolution on several grounds. Some of her arguments are aimed at making a couple of Fulk’s ‘unambiguous’ examples seem more ambiguous (Minkova 2016: 133–134), but the heart of her argument is that HX words such as āre ‘grace, mercy’ could also occur in the same position, at the end of the on-verse. Granting that Fulk’s evidence for the first 100 lines is indeed very striking, she nonetheless concludes that beyond this initial stretch:

[T]he rest of the poem suggests randomness in the choice of L+σ vs. H+σ [that is, between LX and HX] in the same segmental and metrical environments

(Minkova 2016: 135).

In the second 100 lines of the Digby manuscript of the poem, Minkova finds perhaps 15 examples of LX words in this position, at the end of the on-verse, compared with 19 HX words, such as:

(138) Þet hḗ ne múȝe þann^e bídde ṓre

‘that he cannot then pray for mercy’ (Moral Ode 130)¹¹

Fulk (2002: 350) did note the presence of HX words in this position, including line 130, and provided a comprehensive list of just seven examples from the entire poem, three of which he convincingly identified as linguistically or textually dubious. He maintained that such ‘counterexamples are not to be ignored, but their evidence is not of sufficient weight to vitiate the larger generalisation that can be drawn’, a conclusion that seemed very plausible given his count of just four unambiguous HX endings in the entire poem,¹² compared to 20 LX in the first 100 lines alone.

Minkova’s very different count of HX forms would seem to tell a different story, one that allows the endings of on-verses to be much less regulated than Fulk held, and so to provide much less evidence for resolution (Minkova 2016: 135–140). How can these discrepancies be accounted for? Assessment here is a somewhat complex task, particularly since the poem exists in a number of manuscripts, which show a moderate degree of non-trivial variation. Both Fulk and Minkova take Bodleian Library, Digby 4 (Digby: Zupitza 1878; Marcus 1934) as the basis of their analysis, but this is just one text, and by no means the most linguistically conservative. I checked all of these examples of half-line-final HX words identified by either Fulk or Minkova in the parallel-text edition of Payne (2018: 505–656), in order to better evaluate how representative the Digby text is of the metre of this poem more generally. For linguistic variants, an especially important comparison is Lambeth Palace Library 487 (Lambeth: Morris 1988; Payne 2018: 365–382): this manuscript is in a very conservative orthography, and often seems to better represent what the first poet is likely to have written.¹³ Since the various manuscripts have different numbers and selections of lines, I follow Payne’s synoptic line numbering, while also supplying the numbering used by Fulk or Minkova where needed to allow a clearer comparison with their studies.

In one case, the strongest of Minkova’s new examples, the difference may lie in variable choices of scansion:

(139) Swīnes brēd^e is swīþe swēte, swō is of wilde diere

‘Swine’s flesh is very sweet, as is (that) of wild deer’ (Moral Ode 152)

Minkova seems to scan the off-verse as swō ís of wílde díere, but elision to swṓ ⁱs of wílde díere also seems possible. This would allow the final syllable of swēte in the first half-line to serve as the upbeat to the second. That is, the line in full would scan:

(140) Swī́nes brḗd^e is swī́þe swḗt-|-e, swṓ ⁱs of wílde díere

‘Swine’s flesh is very sweet, as is (that) of wild deer’ (Moral Ode 152)

This kind of scansion, where a weak syllable on the final word of the on-verse fills the upbeat of the off-verse, is defended by Fulk (2002: 349), and is taken into account in his assessment of potential resolution. Still, Minkova’s apparent scansion is also plausible, and this remains one of the better candidates for an on-verse ending in Sw.

The remaining discrepancies are of two kinds: Minkova’s acceptance of linguistic variants found the Digby manuscript, but not shared widely among other versions and probably secondary; and words that are identified as HX, but which are better understood as LX. An example of the first type is line 262 (Minkova 249):

(141) Þḗr is uḗr, þet éure brénneð

‘There is fire which burns forever’ (Moral Ode 262a)

Here the Lambeth manuscript reads bernd, and Fulk has apparently accepted the general evidence for a syncopated verbal ending in this and similar lines.¹⁴

The second type concerns linguistic misidentifications. Minkova (2016: 135, n. 29) lists the words wele (233 [Minkova 224], 330 [316]), ibede (351 [337]), and wane (372 [357], 386 [370]) as examples of HX half-line endings, but etymologically all of these words have short vowels, and should still be wĕle ‘wealth’ (Old English wela), i·bĕde(n) ‘prayers’ (Old English ge·bedu(m)), and wăne ‘lack’ (Old English wana, wona), since open-syllable lengthening has clearly not yet affected the language of this poem. These examples serve rather as evidence for resolved LX half-line endings. A slightly different case is uoluelð (324 [311]), which is presumably an accented heavy syllable with an unstressed prefix, uol·uélð ‘fulfils’ (Egerton fulð). This counts as a monoysllabic verse ending H, not HX.

Linguistic variants such as brenneð and hesne might be taken as a sign that the Digby copyist was sensitive to scansion, but relatively tolerant of HX forms at the end of the first half-line. If so, the analysis of Minkova (2016: 137–140) could stand as an assessment of this one copyist’s ‘metrical mind’ (though it may also be that this scribe was not so concerned with the metre). The further assertion that ‘the poet, the copyists, and the audience of the poem … were of one metrical mind’ (Minkova 2016: 137) is, however, not a supportable conclusion. In the family of the Moral Ode versions more generally, HX sequences were strongly avoided. The only really secure example remains line 130 (example 138), with ā́re or ṓre, with swēte in line 152 (139) being not implausible as a second example. To these, betere in 403 could probably be added: the variant bet would scan well as H instead, but all the manuscripts that have this line have betere, and the off-verse (beginning with weak þan in Trinity) has no room to absorb the final unstressed syllable. All the other possible instances HX at the end of the first half-line listed by Fulk (2002: 350) and Minkova (2016: 135, n. 29) seem either certainly incorrect, or suspect for one reason or another. Though it does not seem implausible that some of the merely suspect lines do reflect poetic intent – that of either the original poet or a redactor – the overall number of examples of final HX is very small even if viewed generously, and the word-shape was clearly generally avoided.

This forms a sharp contrast with the usual endings of first half-lines: the majority conclude with monosyllables, with a significant minority ending in LX, many of which cannot be eliminated or discounted on linguistic, metrical, or textual grounds. Such LX endings robustly attested, not just in the first 100 lines, but throughout the poem. In the second 100 lines of Trinity (lines 104–207 of the synoptic text), for example, I find 13 examples that seem reasonably secure.¹⁵

A drop-off from 20 reasonably good examples to 13 seems well within reasonable expectations of variation in vocabulary choice, and words such as dure continue to conclude a notable minority of all on-verses throughout the poem. Despite the complications of textual and linguistic variants that are pervasive in any study of this poem, the conclusion of Fulk (2002), that LX sequences are resolved and equivalent to single heavy monosyllables in this position, seems to be correct. This poem should be added to Laȝamon’s Brut in providing evidence for resolution in Middle English verse.

Both Laȝamon’s Brut and the Moral Ode are probably from the 12th century (the Brut could be from the early 13th), and are among the earliest texts conventionally considered to be Middle English. That both show resolution in at least some metrical positions is an indicator that some dialects of Middle English retained the bimoraic trochee on a phonological level, ready to be reflected in verse. Explanations such as a poetic conventionality carried over from Old English are particularly unconvincing in the case of Laȝamon, whose metrical system is decidedly innovative from the perspective of Old English verse. And above all, the congruity of metrical resolution and the evidence of ie-reduction – both found in very much the same times and places – is striking. Both approaches, phonological and metrical, reinforce each other in pointing to the maintenance of the bimoraic trochee in at least southern and western dialects of early Middle English.

An informative contrast is provided by a very substantial poem of a similar date: the Ormulum, by Orrm, who composed a long and linguistically interesting, though perhaps not poetically inspired, religious poem in his East Anglian dialect.¹⁶ Orrm made use of the same septenarius metre found in the Moral Ode, but he thoroughly fails to supply the kind of evidence for resolution found in that poem. An illustrative example of words ending his on-verses (I randomly chose to take these from lines 9001–9012) are: flocc, hemm, i·noh, inn, ȝuw, and sloþ. LX equivalents are never used. The off-verses show the requirement to end in a heavy trochee:¹⁷  bōke, lāre, follȝhenn, pre͞ostess, Crīste, hāldenn. As these examples imply, this is always filled by HX, never LXX or LX. Orrm is not utterly insensitive to syllable weight, but it is metrically relevant only as a correlate of stress in the key line-final trochee. As noted in the previous chapter, Orrm also shows no evidence of weight-sensitive ie-reduction: neither his phonology nor his metre give us much information about foot structure.

As far as I know, there is no further evidence for resolution anywhere in Middle English verse. Minkova (1997: 443–444) argues convincingly against finding resolution in the Proverbs of Alfred, another early alliterative poem perhaps also composed in the 12th century. Resolution in later Middle English alliterative verse can be excluded with some confidence. As noted in §3.3, the off-verses of classical later alliterative poems must end in a trochee, a stressed syllable followed by a weak syllable. While these final stresses are often heavy, due to the proliferation of heavy syllables caused by open-syllable lengthening (§8.2), there are words that certainly or possibly remained short, but which can – in sharp contrast to Orrm’s heavy cadences – stand in the strictly regulated position of a line-final trochee:¹⁸

(142) þei áuntred hem ðíder

‘they ventured to there’ (Alexander A 230b)

(143) hym rúched in his sádel

‘he turned himself in his saddle’ (The Green Knight 303b)

(144) if nō wáste cóme

‘if no destruction should come’ (Wynnere and Wastoure 253b)

If resolution were a strong metrical principle, we might instead expect to find words such as haþelez ‘heroes’ being used in such positions. Much the same, mutatis mutandis, could be said of Middle English rhyming poets such as Chaucer and Gower, where resolution very clearly plays no metrical role.

Syllable stress alone (relative, sometimes metrically manipulated), and not syllable weight, appears to be the only linguistic feature to make it into the metrical set (§3.5) of at least most later Middle English poets. This kind of negative evidence does not say much about phonological structure one way or the other. In the Brut and Moral Ode, resolution provides positive evidence for foot structure, which when taken together with phonological data, paints a fairly consistent prosodic picture for southern and western dialects. The absence of metrical data for resolution is, however, not evidence of absence with regard to prosodic feet. Syllables have always played a significant role in the linguistic structure of English, and their increasing prominence in metre and insensitivity to weight could equally reflect either changing poetic conventions or a hypothetical shift towards the syllabic trochee foot. For later stages of Middle English, phonological developments alone will have to suffice as evidence for foot structure, and it is this that I turn to now.