6 Feet in Early Middle English: Ie-Reduction

The previous two chapters focused on the prosody of Old English, and specifically of the earliest Old English: the late prehistoric process of high-vowel deletion, and the early metrical phenomenon of Kaluza’s law.¹ Both of these sources of evidence paint a fairly consistent picture of early Old English foot structure, based around the bimoraic trochee, which can be tracked with some precision until the shortening of final long vowels (§4.3). After this change, the evidence for foot structure in Old English becomes significantly sparser. In general terms, the contrast in high-vowel deletion remains robust – retention in words such as scipu ‘ships’, loss in words such as word ‘words’ and werod ‘troops’ – and resolution persists as a key metrical feature even in poems such as The Battle of Brunanburh (composed sometime after 937), and almost certainly in The Battle of Maldon (composed after 991). However, the further details of both processes are more complex, and need to be considered against other potential morphological and poetic factors.²

In this chapter and the next, I leapfrog over the problems in interpreting the later Old English data, and focus instead on sources from the 12th and 13th centuries, a stage conventionally called early Middle English – though as I will argue, in prosody as in so much else there is no sharp break in many dialects, and in many ways the linguistic situation of this period is better considered together with Old rather than later Middle English. In this chapter, I examine the development of unstressed ie-sequences in dialects of the West Midlands, where a strong sensitivity to moraic structures seems to condition sound change – this can be interpreted straightforwardly as a reflecting a bimoraic trochee foot type. In the next chapter I turn to metrical evidence of resolution during the same time period.

The chief break between Old English and Middle English is not so much linguistic as philological. The written standard of Late West Saxon continued to be widely used and emulated – with varying degrees of success – but over the course of the 12th and early 13th centuries, a wider range of dialects and orthographic approaches begin to appear.³ Sources from the far North remain scanty until the 14th century, but a range of sources from southern and central England give a broad (if still rather spotty) impression of a variety of dialects. These include two very long poetic texts from the 12th century – Laȝamon’s Brut from the southwest Midlands and the Ormulum by Orrm (Orm) from the dialectally very different northeast Midlands – which I will return to in more detail in the following chapter.

For now, I focus especially on two substantial manuscripts of non-metrical works representing a West Midlands variety from the earlier 13th century.⁴ The first of these is Corpus Christi College, Cambridge 402, which contains a version of the Ancrene Wisse (Tolkien 1962; Millett 2006; Millett & Dance 2006), a work of guidance for women who had gone into religious seclusion. The other manuscript is Bodley 34, which contains a variety of works dealing with the lives of women saints and female religiosity: Þe Martyrdom of Sancte Katerine (d’Ardenne & Dobson 1981), Þe Liflade ant te Passiun of Seinte Margarete (Mack 1934), Þe Liflade ant te Passiun of Seinte Iuliene (d’Ardenne 1961), Epistel of Meidenhad (Millett 1982),⁵ and Sawles Warde (Wilson 1938).⁶  Ancrene Wisse has been referred to traditionally by the siglum A, and the Bodley texts by B – whence ‘this algebra of A and B’ for the two together, the ‘language (AB)’ of Tolkien (1929), or more simply just the AB dialect.

This variety of Middle English in general developed from a type of Old West Mercian, and can be loosely regarded as a successor to the language represented in the Vespasian Psalter (though for rather minor qualifications to this view, see Ball 1970). There are also other sources that reflect closely related dialects, which I point out as needed: the most notable of these are the so-called Wooing or Wohunge Group (Thompson 1958) and the Lambeth Homilies (Morris 1988). For a full survey of these and other texts from the general area, see Dance (2003: ch. 2).

In these early Middle English dialects, a prosodically conditioned sound change took place: the unstressed sequence *iə either remained unchanged or was reduced to i (Tolkien 1929; d’Ardenne 1961: 188–189). This sound sequence occurred in various words, but its development is especially clear in that large group of weak verbs traditionally labelled class II. In Old English, the present tense of these verbs showed two types of stem element, a shorter one in-a-(in the second-and third-persons singular, e.g. lōcað ‘looks’), and a disyllabic one in-ia-or-ie (in all other present forms, e.g. lōciað ‘they look’, lōcie ‘look (subj.sg)’). At some point, unstressed e and a collapsed into [ə], written <e>, so that ie and ia merged as [iə], <ie>.⁷

The further reduction of this early Middle English ie to i was sensitive to syllable weight, with reduction taking place after heavy syllables, but not after light ones. The following examples of developments from Old English to AB are representative:⁸

(105) clĕopiað ‘they call’ > AB clĕopieð (Ancrene Wisse P.158)⁹

(106) (ge·)lōciað ‘they look’ > AB lōkið (Ancrene Wisse 4.1258)

(107) ĕardiað ‘they dwell’ > AB ĕardið (Margarete 22.16)

For want of a snappier label, I call this process ie-reduction, which at least has the benefit of being relatively transparent. The result of this process is a set of morphophonemic distinctions between i and ie in class II weak verbs, which is witnessed virtually without exception in the AB texts (Tolkien 1929: 122–124; d’Ardenne 1961: 189, 234–235).

The basic contrast of light versus heavy class II weak verbs is already significant from a prosodic perspective, showing a clear sensitivity to syllable weight in this variety of early Middle English. The impression that this might have something to do with foot structure, and the parallelism with high-vowel deletion, is only strengthened by the behaviour of ‘light disyllables’, with two light syllables before the verbal formant:

(108) swĕotolian, sutelian ‘to make clear’ > AB sutelin (Iuliene 167, 543)¹⁰

Even though the *ie formant follows a light syllable, it reduces to i, indicating that this change depends on something more than just the weight of the immediately preceding syllable (Keyser & O’Neil 1985: 91–94). Just as with high-vowel deletion, it seems plausible that prosodic feet might be at work. This parallelism between the two processes is easy to see when the effects are laid out side by side:

In contrast to Old English high-vowel deletion, which has been discussed and analysed from a wide variety of theoretical perspectives and with reference to a wide range of data, ie-reduction has received relatively little linguistic attention. The main exception is Keyser & O’Neil (1985: ch. 5), who review much (though not all) of the relevant data, and offer a formal prosodic analysis of the phenomenon. Much of their discussion remains valuable, though many specifics also need reconsideration.

Specifically, Keyser & O’Neil (1985: 91) argue that the stressed syllables of words are part of a two-mora foot, consisting either of a single heavy syllable such as *(lō)-ki-eð or two light syllables such as *(su-te)-li-en or *(clĕo-pi)-eð. Within this foot framework, ie-reduction is, under their analysis, a ‘Weak Foot Drop’ rule: an e is deleted when it follows another vowel, which itself follows the edge of a foot. So the e’s of *(lō)-ki-eð and *(su-te)-li-en are deleted, since they separated from the edge of the foot by a single vowel, i. This correctly gives deletion in lōkið and sutelin, while keeping clĕopieð unchanged, since there the e immediately follows the end of a foot.

This ‘Weak Foot Drop’ rule in and of itself will probably not appear very elegant to linguists today. That a rule should make reference to a foot boundary plus a syllable’s space seems peculiar at best, and is not the sort of thing that prosodic rules generally seem to do. As I will show later, there are also empirical problems with this rule in longer verbs.

On the data examined so far, a simpler explanation of ie-reduction might be to see the change to i as the default, unconditioned outcome, rather than the result of a special rule. The question would then be why ie remains in words such as clĕopieð. I suggest the following framing:

Unstressed *ie becomes i, unless a foot boundary falls between the two vowels. This accounts well for all the word-types examined so far:

(105) *(clĕo-pi)-e〈ð〉 > clĕopieð

(106) *(lō)(-ki-e〈ð〉) > lōkið

(107) *(ĕar)(-di-e〈ð〉) > ĕardið

(108) *(su-te)(-li-e〈n〉) > sutelin

Yet even if Keyser & O’Neil’s ie-reduction rule needs some revision along these (or other) lines, their basic assumption that it is sensitive to bimoraic feet seems sound. The examples so far are given with simple bimoraic trochees;¹¹ I will consider the possibility of the Germanic foot in §6.5. For now, the important thing from the perspective of the history of English prosody is that some kind of bimoraic unit, which can encompass two light syllables if needed, provides the necessary context for any adequate rule of ie-reduction.

As should be expected, a close look at the data of ie-reductions reveals a certain number of wrinkles and complications. Most of these are not terribly significant, and have largely been treated well by Tolkien (1929) and Keyser & O’Neil (1985). For instance, the exact number of words of the sutelin-type, with light disyllabic bases, could be debated. Some, such as euenin ‘make even’ come from Old English monosyllables, in this case efnian, with an epenthetic vowel providing the second light syllable. A few even alternate between light disyllabic and heavy monosyllabic bases in the AB corpus, such as sunegin, sungin ‘sin’. It is worth emphasising that in all examples, the disyllabic forms predominate, and are often the only ones attested, and epenthetic vowels in relevant verbs are often attested already in Old English (for instance, AB openin ‘open’ is preceded by Old English openian, alongside ge·opnian). This issue is, however, not of any great importance, since there is a core of words such as sutelin or tĕoheði ‘tithe’¹² which are disyllabic as far back as their history can be reconstructed, and all LL sequences, old or new, behave the same.

More interesting, but also more problematic, are longer class II weak verbs. Keyser & O’Neil limited themselves to the types discussed so far: light monosyllables (with L bases, such as clĕopieð), heavy monosyllables (with H bases, such as ĕardið), and light disyllables (with LL bases, such as sutelin), shapes which do account for the overwhelming majority of the data in the AB texts. There are, nonetheless, a few words with longer bases including heavy syllables.¹³

The evidence of HH bases is relatively straightforward. This type includes verbs such as cnāwlēchin ‘acknowledge’ and herbarhin ‘lodge’. As these examples suggest, as might be expected under any theory of ie-reduction, they almost uniformly show the change of the verbal stem to i. Under the bimoraic trochee assumed in the previous section, the development might have been something such as *(her)(-bar)(-ȝi-e〈n〉) > herbarhin, with the foot-internal ie-sequence reducing according to the rule.

The only apparent exception to the reduction to i is the imperative plural ēadmōdieð ‘make humble’ (Ancrene Wisse 4.1440). This is plausibly explained by Tolkien (1929: 121, n. 2) as a deadjectival formation to ēadmōdi, so that the verb is effectively from *ēadmōdiieð.¹⁴ This complex and unique structure probably does not tell us much about either foot structure or the usual rules of ie-reduction.

Verbs with HL bases are in equal measure intriguing and puzzling. They have the potential to shed light on the details of how the medial syllables of longer words are footed, but the evidence they provide is difficult to interpret satisfactorily. Ie-reduction is seen in some verbs of this shape, such as hersumin ‘obey’, while others preserve the old sequence, as in ondswerien ‘answer’.¹⁵

Although the evidence pool is very small, it seems that the presence or absence of ie-reduction in such verbs is related to their morphological structure. Specifically, the two verbs that retain the ie are both morphologically complex. In the relatively well-attested ondswerien, the verb is derived from the noun ondswer ‘answer’, which is itself made up of the prefix ond ‘toward, in response, anti’ and the verbal root swer (independently, though less frequently, attested in the verb swerien ‘swear’).¹⁶ This derivation occurred early – ondswĕorian (with its spelling variants) is well attested in Old English – so the extent to which it would still have been regarded as morphologically complex in AB Middle English is unclear. Still, it is not implausible that it might have retained a prosodically relevant juncture between its elements at this date. Tolkien (1929: 118) even goes so far as to suppose a ‘strong secondary accent’ on the second syllable, ónd-swèrien, an idea to which I will return shortly.

A compound structure also seems plausible for the only other HL-verb without ie-reduction: gristbĕatien ‘gnash the teeth’. A relevant present-tense form is only attested once (Iuliene 671), and although its etymology is not transparent, it does seem to have a complex internal structure (Tolkien 1929: 125–126). A compound structure would, of course, also hold for ēadmŏdieð, if this is held to have a shortened medial vowel (Diensberg 1975: 210; cf. note 14 above).

There are two further HL-verbs in the AB corpus whose relevant inflectional forms are attested: hersumin ‘obey’ and fēðerin ‘load up, weigh down’. Neither of these has a potentially compound-like structure. The first of these is formed with the nominalising suffix-sum-, a derivational element showing significant vowel reduction (cf. German gehorsam). This, intriguingly, seems to generally show ie-reduction in the two attestations of its infinitive (Katerine 53, 127). There is a further instance of the plural found as hersumeð (Katerine 98), which must be an error for either *hersumieð or (more likely, given the clearly attested infinitive forms) *hersumið. Setting aside this erroneous form, it seems that this verb regularly underwent ie-reduction.

The same seems to be true of fēðerin, though here it is possible that the initial vowel was shortened at some point. The word etymologically comes from *foē ðrian, and shortening of the long vowel in an overheavy syllable *foē ð-is possible (Hogg 2011: 207). Shortening of the initial syllable of a trisyllabic word is also possible, either in *foē ðrian or later, with the epenthetic vowel, in forms such as the third-person singular *foē ðerað > *feðereð (Fikkert, Dresher & Lahiri 2006: 140). The Nero manuscript of Ancrene Wisse might suggest such a shortened form in its spelling of the past participle as iueððred, but there is, as far as I know, no clear evidence for potential shortening from the AB corpus. If shortening did take place, then the word belongs rather with the sutelin-group; if it retained its etymological length, it is parallel to hersumin in showing ie-reduction after an HL base.

This philological evidence is frustratingly slight, but as it stands the impression is that in HL-verbs, those with a more compound-like structure (ondswerien, gristbĕatien, just possibly ēadmō̆dieð) retain the old ie-sequence, while those with only one lexical-like element show reduction to i (hersumin, perhaps fē̆ðerin). The obvious follow-up question is what implications this potential distinction might have for either foot structure or the process of ie-reduction.

The basic problem for ie-reduction in the AB corpus is how to explain the contrast between the more compound-like ondswerien and the simplex hersumin. As an initial observation, it is clear that the Weak Foot Drop rule of Keyser & O’Neil (1985: 91) can’t explain hersumin. This would have originally been footed either as *(her)(-su-mi)-en or *(her)-su-mi-en, neither option putting the e in the necessary position for that rule to apply (one syllable removed from the closing bracket of a foot). This can be seen as the final nail in the coffin for the Weak Foot Drop rule.

Unfortunately, coming up with a precise explanation for these words is a harder matter. If the foot structure continued the bimoraic trochee of early Old English, then both *(ond)(-swe-ri)(-en) and *(her)(-su-mi)(-en) should have had the same pattern of feet in the period just before ie-reduction took place. Broadly speaking, it seems that the second foot of *hersumien was in some way ‘weaker’ than that of *ondswerien. The ie-reduction rule might then be described as:

Unstressed *ie becomes i, unless the two vowels are separated by the boundary of a ‘strong’ foot.

I can see two principle ways of distinguishing foot ‘strength’ in this context. One, following Tolkien’s suggestion that ‘strong secondary accent’ is the relevant feature (1929: 118), would be to assume that ie-reduction depends on stress assignment, and that in the normal course of things only initial feet are assigned stress (in contrast to Old English). For ondswerien, the compound-like structure would lead naturally lead to the stressing of the second foot regardless.¹⁷ The other option is to assume that the boundaries of non-initial feet should be treated differently from those of word-initial feet: that is, a ‘strong’ foot is one that is initial in the prosodic word.

While theoretically more straightforward, the evidence of Middle English metre seems to speak against the former option, that óndswèrien receives a secondary, medial stress, while hérsumin does not. Poetic evidence for secondary stress is not as straightforward in Middle English as it was in Old English, but there is some evidence from stress shifts. Take the following rhymes from Laȝamon’s Brut:

(109) Hēo færden mid ȝēapscipe ⁊ mid wisdṓme

 swā longe þat hēo tō Alamaine cṓmen

‘They travelled with cunning and wisdom for so long that they came to Alemannia’ (Brut 1379)

(110) ⁊ dōn him hersumnésse and þurh him singen másse

‘and do him obeisance and sing mass through him’ (Brut 14838)

These suggest at least some prominence on the italicised syllables, which rhyme with standardly stressed words. Laȝamon’s metre is loose and often ambiguous (§7.2.1), but similar things are also found in the Ormulum, where the metre involves monotonously rigid alternating stresses:

(111) ⁊ sóþ wissdómess léome

‘the true light of wisdom’ (Ormulum 6729)

(112) ⁊ óff galnésse skír ⁊ fré

‘and free and pure of lasciviousness’ (Ormulum 8015)

As observed by Yakovlev (2008: 232–234), these lines seem to attest to a genuine stress shift, and scan as if the derivational syllable not only were the primary stress of the word, but as if the remainder of the word were unstressed (this is probably as true for Laȝamon as for Orrm). While this was clearly done for metrical reasons, it seems likely that such shifts capitalised on the presence of secondary stresses on heavy suffixes such as-dōm and-nesse. This is not entirely certain – a stress shift that seems to fully demote a primary stress could theoretically promote a fully weak syllable – but it is probably safest to assume that medial feet are indeed assigned (secondary) stress, more or less as in Old English. If this is right, then tying ie-reduction to stress is probably not sound, since the same processes that would create a secondary stress in (gā́l)(-nès)-se should also do so in *(hér)(-sù-mi)-en.

This leaves a ‘strong foot’ as, descriptively, one that is initial within the (minimal) prosodic word. Reduction in (clĕo-pi)-eð is blocked because the ‘)’ boundary of the first foot is a strong one, while it can take place in *(her)(-su-mi)-en, because the ‘)’ of the medial foot is in some way weaker. This is still a somewhat impressionistic way of characterising the situation, but it is not clear to me how to arrive at a more detailed or concrete explanation of ie-reduction on the evidence currently available.

Dresher & Lahiri (2022: 41–44) argue that while early Old English did have a true bimoraic trochee, later Old English – and subsequently Middle and modern English – altered this system by allowing the core bimoraic foot to be optionally expanded with a further weak syllable (see §4.1). The resulting foot type is what they call the Germanic foot. If this foot type did develop, then ie-reduction should be explainable within this framework. This is how the main word-types discussed so far would, just prior to ie-reduction, be footed under this system, using | to mark the boundary between the main body of the foot and the extra ‘weak branch’:¹⁸

(113) ^*(clĕo-pi|-e〈ð〉) > clĕopieð

(114) ^*(ĕar|-di)(-eð) > ĕardið

(115) ^*(su-te|-li)(-en) > sutelin

(116) ^*(her|-su)(-mi-e〈n〉) > hersumin

(117) ^*(ond)(swe-ri|-e〈n〉) > ondswerien

The generalisation within this framework is that *ie becomes i unless the *e stands in the weak branch of a foot. This is perhaps somewhat uncomfortable – the weak branch should be a position associated with reduction and deletion, not retention – but this is not necessarily a fatal problem. Someone already committed to the Germanic foot as a model for Middle English would not find any strong reason to abandon it because of ie-reduction. On the other hand, it is hard to say that this model has any special advantages in describing the process either. I personally do not find the evidence of a shift to the Germanic foot convincing,¹⁹but it should be considered as a reasonable possibility, with the evidence of ie-reduction not being detailed or varied enough to be decisive on this point.

Class II weak verbs provide the bulk of the evidence for ie-reduction for several reasons: the group is large; it includes verbal bases of various shapes, allowing the prosodic contrasts to be traced with relative clarity; and the effects of morphological analogy seem to be very limited (at least in the AB corpus). Still, there are various isolated lexemes, such as Old English hlǣfdī(g)e ‘lady’, which generally seem to show the same patterns of ie-reduction as the weak verbs do.²⁰

There is also one important general class relevant to the process: adjectives in-i, from Old English-ig, such as bisi ‘busy, active’ and hāli ‘holy’. These would, like any adjective, add an inflectional-e to mark agreement as needed. So, for instance, the plural of bisi is bisie. By contrast, hāli usually has the plural form hāli (e.g. Meidenhad 2.24), with no ending, just as would be expected for a form affected by ie-reduction.

These adjectives are, however, more open to morphological pressures than the weak verbs are. As Keyser & O’Neil (1985: 90–92) note, plurals such as hālie are in fact attested (e.g. Ancrene Wisse 1.394; cf. also d’Ardenne 1961: 217–218):

To a limited but surprising extent, [ie-reduction] fails: that is, from time to time in AB we find forms like creftie [‘mighty’] and haalie [i.e. hālie] where we expect … crefti and haali, but never bisi and dusi [‘foolish’] where we expect bisie and dusie.

As they observe, this limitation of this variation to heavy stems only means that it is systematic, and ‘cannot be the result of a simple confusion’ (Keyser & O’Neil 1985: 98), but is clearly due to morphological pressures:

In the adjectives, however, e is all there is to the inflectional system. Lose it and there is none of the essential information about definiteness and/or plurality conveyed: if the adjectival e goes, the paradigm goes. It is, then, interesting and not at all surprising to find adjectival e reasserting itself despite Weak Foot Drop [i.e. ie-reduction] – though not in great proportion compared to the presence of the expected e-less forms. We understand this fairly insistent violation of Weak Foot Drop among adjectives whose syllabic metrical structure is that of [hāli] to be a particularly clear case of analogy – perhaps the clearest case that we know of the force of paradigmatic regularity imposing itself on the forms of a language.

(Keyser & O’Neil 1985: 98–99)

As with high-vowel deletion, the workings of morphological pressures do not seriously undermine the relevance of ie-reduction to foot structure, nor mean that it is in any way a purely morphological process.

Taking a step back from the AB texts, there is a spectrum of outcomes for ie-reduction that reflects both changes over time and variation in different areas (Tolkien 1929: 119–120; Goering 2021b: 483–487). One extreme of this spectrum is represented by the East Midlands dialect of Orrm, writing in the middle of the 12th century. His work is slightly older than the AB texts, but in a dialect that has innovated in ways that essentially obscure any weight-based developments. Most importantly, the large set of class II weak verbs has entirely replaced its stem formant with e [ə] (from a). Only in isolated lexemes such as laffdiȝ ‘lady’, from hlǣfdīge, is there a hint that some kind of ie-reduction took place here as well (Tolkien 1929: 119).

Among those texts that do retain some kind of ie or i formant in class II weak verbs, a sense of the range of outcomes can be gained by using The Linguistic Atlas of Early Middle English, or LAEME, which is organised by manuscript. LAEME does not represent a full corpus from the period, but includes large enough selections from a wide enough range of texts to give a useful impression of the linguistic landscape of the period.²¹ In table 6.1, I summarise the results for ie-deletion in class II weak verbs, sorting the outcome into four columns. The first column, ie, gives the number of forms that seem to retain their disyllabic character without deletion, including rarer variants such as ia, iȝe, ii, and ihi. The second column, i, shows the number of forms that have been reduced to simple i. Column e indicates forms that had ia or ie in Old English, but are found as a or (much more commonly) the reduced form e. That is, this column shows the number of Orrm-style forms in each stem-category for each text. The fourth column, ∅, is included for the sake of completeness, and gives the number of forms where no ending at all is provided; these are few in number and probably all scribal errors. For each column, both the absolute numbers and percentages (as proportions within the row) are given.²² The data is further broken down by word-shape within each manuscript. Only the basic word-shapes of monosyllabic heavy stems (H), monosyllabic light stems (L), and stems of two light syllables (LL) are included. I will discuss words of other shapes in §6.7.2, but they are rather uninformative and too infrequent to include on the table. The manuscripts are arranged in decreasing order of what percentage of heavy stems retain the archaic ie-form.

Table 6.1 shows two main lines of development, which partly go hand-in-hand. One is the emergence of the weight-based distinction in whether the ie-form is retained or not. Some of the earlier manuscripts, such as Worcester and the Caligula MS of Laȝamon’s Brut, show this process only incipiently, with a large majority even of heavy-stemmed verbs retaining the archaic ie-formant: these are forms such as clensien ‘cleanse’ from Worcester.²³ Presumably ie-reduction simply hadn’t affected the dialect of these works yet. Royal and Caius, by contrast, approach the AB corpus in terms of retaining ie in light-stemmed verbs, and replacing it with something else (either i or e) in heavy stems and light disyllables. This is not surprising, since in content and language, these manuscripts are clearly broadly related to the AB texts (for Royal in particular, see Tolkien 1929: 108; Jack 1991). The Otho text of Laȝamon’s Brut might also be reasonably put in this category. The distinction is not so stark in the remaining manuscripts, but there is nonetheless a very clear tendency for light stems to retain ie at much higher rates than other stem types (even Trinity B.14.39 reflects this trend, though ie-forms constitute only a minority even among the light stems). Weight-sensitive ie-reduction may not usually operate quite as clearly outside of the AB corpus and some very closely related texts, but it seems to be broadly reflected throughout the South and West of England in the 13th century. The alternations created in class II weak verbs partly survive still even in 14th century Kentish, to judge by Dan Michel’s A·yénbite of Inwyt.²⁴

The other trend, which increasingly obscures this weight-based alternation, is the spread of the formant e (or occasionally and archaically, a). In the most archaic texts, Worcester and Laȝamon C, the e is already present in historically ie-contexts as a minority outcome. Interestingly, although there is very little sign of ie-reduction to i in these manuscripts, the e-variants (though never especially common) are decidedly weighted towards heavy stems. Worcester thus has both fostrien and fostren ‘to rear (as a child)’, but not fostrin (this last being the usual form in AB), while words such as clĕopie are never reduced.

It could be that this points to a variant type of ie-reduction, to e rather than i, but if so, this is not clearly followed through in any of the manuscripts. Three (Trinity Homilies, Titus, Caius) do show a pattern in which e is the most common formant for heavy stems and ie for the light stems – e.g. heavy clansen ‘cleanse’, light makien ‘make’ – but even so, especially for Titus and Caius, a significant minority of heavy stems show i, such as offrin ‘offer’. No texts show a pattern of *ie > e without some trace of *ie > i.

Whether or not there was some phonological reduction directly to e, the usual trend is best explained by the phonological development being to i, with this then being prone to morphological replacement by e. Such a morphological spread would not be surprising. Before ie-reduction (but after *a > e), class II weak verbs had two stem variants: e in the second-and third-persons singular, and ie in the remainder of the present. With the introduction of i in the heavy stems, the situation was more complex morphologically, with three variants distributed partly by function and partly by weight. This is the case in the AB system:

In other verb types, such as the strong verbs and class I weak verbs, the situation was simpler. Perhaps significantly, the third-person singular and plural often had identical endings, with the Old English singular-eð and plural-að merging as-eð, [əθ]. This identity was particularly widespread among heavy-stemmed verbs, such as wurcheð ‘s/he does’ or ‘they do’. Light stems (I count here verbs such as tellen ‘tell’ with light-stem forms in the paradigm) were frequently distinguished by other kinds of alternations, such as nimeð ‘s/he takes’ versus nĕomeð ‘they take’, or teleð ‘s/he tells’ versus telleð ‘they tell’ (d’Ardenne 1961: 236, 244; Diensberg 1975: 158–159, 202).²⁵ It may be that the limited spread of e among heavy stems already in Worcester and Laȝamon C – before general ie-reduction – might have been on the analogy of such verbs, with the general pattern being to use-eð as an almost uniform third-person ending for heavy stems of all classes. The pressure to adopt this form would only have increased after the change of ie to i. Replacing a form such as lōkið with lōkeð would not only bring the plural marker in line with heavy-stemmed verbs of other classes, it would reduce the number of allomorphs in the class from three (ie, i, e) to two (ie, e). Once this e had become established in the heavy stems, it could be spread from there to the light stems as well.

Simplifying the situation slightly, the texts can be loosely arranged into a sequence representing subsequent stages of this development. The starting point (stage I) – not reflected as such in the LAEME texts – shows the sound-change developments of Old English ia and ie to ie, preserved regardless of weight. Worcester largely reflects this stage, but with the introduction of very limited e-spreading in the heavy stems. The AB texts show little sign of e-spreading, but do reflect the introduction of general ie-reduction to i among heavy stems. I distinguish these two parallel types of innovation as stages IIa and IIb.

From there, things develop in a more or less regular sequence. A third stage, represented by Royal, shows ie-reduction and the entrenched presence of e-spreading in about a quarter of heavy stems. Stage IV is seen in Caius, where the proportions of Royal are nearly reversed in the heavy stems: e is now the majority variant. In Titus, representing the next stage (V), the heavy stems remain more or less the same as in Caius, but e has begun to spread to the light stems in significant numbers, being found in about a quarter of cases. Among the texts included in my LAEME review, Trinity B.14.39 would seem to carry this process of e-generalisation furthest, favouring e for all class II weak verbs, but retaining a significant minority of ie-forms for light stems (stage VI). The logical final outcome, stage VII, would be the situation seen early on in Orrm, and eventually reached by later authors such as Chaucer (see below).

This development is shown more schematically in table 6.2. Stages I–II show the first introduction of limited e-spreading and general ie-reduction among the heavy stems, III–IV the increasing spread of e among the heavy stems, and V–VII the further spread of e to the light stems.

This sequence is relatively robust, even for texts on table 6.1 that might not seem to fit in at first glance. To judge simply by the numbers, Dan Michel’s Kentish looks out of sequence. The heavy-stemmed verbs are still at stage III, with i-forms predominating, but there are both e-and i-forms found among his light-stemmed verbs. However, the variation in this instance largely stems from a single verb, maki ‘make’, which seems to have developed a special paradigm: the Old English infinitive macian had become maki (or its graphic variant maky), with ie-reduction exceptionally applying (in contrast to clepie ‘call’, hatie ‘hate’, herie ‘praise’, zuerie ‘swear’), and the third-person plural was makeþ. Since a similar irregularity is seen in the more sparsely attested waki/waky ‘wake’, the final k may have conditioned some kind of special phonological development – though alternatively the innovation might be ascribed to frequency effects in maki, spreading by analogy to the rhyming waki. Either way, the bulk of class II weak verbs in the A·yénbite of Inwyt do belong at stage III, with the exceptions forming a well-defined group caused by a further, complicated innovation at least partly independent from the more general kinds of e-spreading seen in other texts.

A more significant qualification is needed for stage VII. As already mentioned, Orrm, writing in an East Midlands dialect in the 12th century, shows a uniform e in class II weak verbs, giving him indiscriminate infinitives such as the heavy clennsenn ‘cleanse’ and light clepenn ‘call’. This is, however, probably not the outcome of Orrm’s dialect having gone through all the previous stages, but rather due to direct influence on an earlier form of his dialect from Norse (Tolkien 1929: 120; Warner 2017). Norse originally had a stem vowel a throughout the inflection of this verb type, with no variant corresponding to Old English ia or ie at all (this being an innovation restricted to the Ingvaeonic languages; Cowgill 1959). So kalla ‘call’ was both the infinitive and third-person plural, with kallar being the third-person singular. In regions where close interaction between English and Norse took place, it would not be surprising for the very simple Norse inflection of this class to have been generally adopted, leading (with reduction of *a to ə) directly to Orrm’s system.

It is, however, not hard to find dialects that probably reached this last stage primarily through internal change. Late texts – including the works of such famous authors as Chaucer and Gower – show a uniform e in class II weak verbs: lōken ‘look’, clēpen ‘call’. It is not surprising that the old weight-based distinctions would have been abandoned by that point, since the effects of open-syllable lengthening meant that many formerly light-stemmed verbs had become heavy in key inflectional forms (§8.2). After the third-person singular lengthened from clepeþ ‘calls’ to clēpeþ, it would be less than obvious why this new clēpeþ would have an infinitive clĕpien,²⁶ while the superficially similar lōkeþ ‘looks’ had an infinitive lōken (reflecting e-spreading). If e hadn’t already fully won out before open-syllable lengthening took place, it surely would have very quickly afterwards. As a consequence, in the large corpus of later southern Middle English texts, the weight-sensitive alternation of ie and i (or e) had been entirely given up, stage VII being achieved without any necessary influence from Norse.

One point that unfortunately does not become clear from my LAEME survey is the prosodic status of HL bases such as ondswerien and hersumin. There is support for a distinction between the more and less compound-like types, with manuscripts such as Virtues & Vices and Royal agreeing precisely with the evidence of the AB corpus: reduction in hersumin, but not in onswerien or grispatien. Beyond this, there is little to say. There are signs of e-generalisation in the expected manuscripts, such as Titus, with four examples of onswerien to six of onsweren. In general, the number of relevant tokens is very small, meaning that the larger story of weight-based alternations and e-spreading must be told mainly through the much better-attested, shorter word-types.

Overall, the LAEME survey, summarised in table 6.1, and the seven-stage development outlined in 6.2 (§6.7.1) provide a broadly plausible scheme for how ie-sequences developed in most of southern and western English-speaking Britain. This broad overview may, of course, be complicated by a closer investigation of the philological particulars of each manuscript, but the general picture that emerges seems to fit well with what might be expected to happen on more general grounds. At some point in the 12th century (perhaps in the early 13th in some areas), a sweeping wave of weight-sensitive ie-reduction took place. This suggests that a foot type similar to the bimoraic trochee was present not only in the narrow area of the AB dialect, but was very widespread in English, at least outside the North and East where influence from Norse led to early replacement of the ie-formant by e. The exact effects of this on the stem vowels of class II weak verbs were significantly complicated by the morphological spread of e, and eventually the entire foundation of distinguishing light and heavy stems was undermined by open-syllable lengthening. Together, these changes finally resulted in the generalised e familiar from widely read authors of later Middle English such as Langland, Chaucer, and Gower.

The evidence of ie-reduction points to a conservatism in Middle English – or to be precise, in western and southern dialects of earlier Middle English – which has often been overlooked in the phonological literature. Murray (2000: 622), for instance, states plainly that ‘resolution was no longer a prosodic feature of the language’. The evidence reviewed in this chapter suggests that far from being inoperative, phonological resolution – the equivalence of H and LL – was robust and widespread, persisting until well into the 13th century, and at least in the case of Dan Michel’s Kentish, into the 14th.

The precise details of foot structure are hard to recover on the available evidence, but it is simply accounted for by the same bimoraic trochee proposed for early Old English. In principle, all the data discussed in this chapter could also be described within the framework of the Germanic foot, if such a foot type could have arisen as an innovation in the intervening centuries (§6.5). Even if this did happen, there would still be an essential point of prosodic continuity: the presence of a bimoraic unit, either as the strong branch of a larger foot or (as I prefer) as the entire foot. Either way, these feet were clearly formed much as in earlier stages of English: from left to right, and with the first foot of the word carrying the main stress. While it is not possible to be as precise as for Old English, the following principles probably apply:

That there should be a significant prosodic continuity, with the bimoraic trochee persisting (or at most being extended with an extra weak branch) through later Old English and into at least the earlier Middle English period, should not really be very surprising. A change would need a cause, and none of the major developments before the 13th century would seriously disrupt the core operation of bimoraism in medieval English. It is not until the more significant changes of later Middle English, discussed in chapter 8, that the potential for a major change in foot type might be found.

That the evidence for continuous bimoraism has been overlooked may be due at least partly to the unfortunate effects of scholarly periodisation. While virtually all comments on the divide between ‘Old’ and ‘Middle’ English are followed by a formulaic invocation of how arbitrary such a division is, the separation of dictionaries, grammars, online tools, and other philological apparatus into these two groups undoubtedly hinders research into these kinds of continuities, and likely influences how the change to ‘Middle’ English is conceptualised. It is virtually always treated as a sharper break than it really is, a tendency not mitigated by an excessive focus on Orrm as a prime representative of ‘early Middle English’ for linguists, despite his writings representing quite a new dialectal variety compared to earlier records of English. There is much to be said for alternative periodisations such as that passingly proposed by Tolkien (1983a: 195, n. 9), and being admirably experimented with in practice by Ringe’s ongoing History of English.²⁷

There are at least two further potential ramifications of finding the bimoraic trochee in early Middle English which need consideration. One is whether this foot structure finds any reflection in the metrical systems of early Middle English poetry, which might help corroborate and strengthen the phonological evidence. The other is how this prosodic framework fares in later Middle English: what implications does it have for our understanding of changes such as open-syllable lengthening and trisyllabic shortening? These two issues are, respectively, the subjects of the following pair of chapters.