13 Conclusion: Bimoraism in Medieval English and Norse

The data for prosody from medieval English and Norse is fully as messy and varied as anyone might expect. Disentangling phonological developments from purely morphological reworkings, building pictures of vowel deletions from the scanty records of runestones, finding the right generalisations across diverse and varied manuscripts, discerning phonological structure in verse forms – these philological and linguistic challenges can often be met, but there is always a sense in which theoretical phonological conclusions are partial and provisional.

Still, when taken together, the evidence of vowel reductions, morphophonemic alternations, and metrical resolution adds up to a fairly consistent conclusion: that bimoraic feet played a significant and enduring role from prehistory through into the later Middle Ages both in English and in North Germanic. In some instances, such as early Old English, this bimoraic foot must very specifically be the bimoraic trochee, a foot consisting of ideally two moras, which in the simplest cases come from one heavy syllable or from two light syllables. In other contexts, such as early Middle English, it is hard to conclusively rule out the possibility of some other kind of bimoraic foot, such as the extended Germanic foot (essentially a bimoraic trochee plus an optional extra weak syllable) – though since the bimoraic trochee is sometimes needed, and always sufficient, it seems safest to assume that this typologically well-supported foot type was the primary basis for all kinds of bimoraism in earlier Germanic.

This basic continuity of bimoraism is, however, only part of the story. Here at the end of this book, it seems worth taking a step back and attempting to provide a synthesis of the prosodic history of these languages from (at least) Proto-Germanic through the end of the Middle Ages.

There are three pieces of evidence for bimoraic trochees in Proto-Germanic. The first is that in the reconstructed language, as in all the older attested Germanic languages, there is a strict minimal-word requirement: every prosodic word must have at least two moras (Kuryłowicz 1949: 38; Russom 1998: 15–16; Fikkert, Dresher & Lahiri 2006: 128; Goering 2016b: 281–282). One interpretation of this is that every prosodic word must contain at least one foot, which is (minimally) bimoraic (McCarthy & Prince 1996: 6–7). Examples of prosodically minimal content words reconstructible for Proto-Germanic include *k^waþ ‘said’, *snau ‘snowed’, and *skipaⁿ ‘ship’, and if the final *-z is extrametrical, also *kū〈z〉 ‘cow’, *wini〈z〉 ‘friend’, etc. The only words shorter than this that can be reconstructed are unstressed grammatical words such as *ni ‘not’, *sa ‘that (masc.nom.sg)’, and *bi ‘beside, around’. Such words were probably clitics rather than full prosodic words, and when they did occur with independent stress they were probably lengthened: compare frequent Gothic ni ‘not’ with lengthened nei [niː] (attested in Corinthians II 3:8 and Skeireins 1:5), or Old English big [biː] alongside the more usual short be.¹

Garrett (1999) warns that this form of evidence is not strong, since many languages show minimal-word requirements that are slightly different from their minimal-foot requirements. One example is the Uto-Aztecan language Cahuilla (discussed in §4.5.1.2), where content words must end in at least a short vowel plus any single consonant (e.g. net ‘ceremonial chieftain’), even when these consonants do not count as moraic for foot structure; only coda [ʔ] contributes a mora in this language (Garrett 1999: §2.1). That is, there are minimal words smaller than minimal feet, at least in some languages. It is also worth remembering the warning of Bermúdez-Otero (2018: 3) that a language may allow a minimal word of the shape LL, but this does not necessarily provide evidence for a resolved bimoraic trochee (§2.5).

In general, minimal-word requirements may stem rather from phonetic pressures to ensure that content words usually meet a certain minimum absolute length, which may not have much to do with feet in any given language (Garrett 1999: §4). Though none of Garrett’s criticisms apply specifically to Germanic, where the assumed minimal-foot and minimal-word requirements align exactly – the generalisation that a minimal prosodic word must have at least one full foot seems plausible for these languages – it is probably best not to put undue weight on this sort of evidence as the starting point in any argument.²

Similar qualifiers and cautions also apply to the second possible source of evidence for the early bimoraic trochee: shortenings of overheavy syllables in pre-Proto-Germanic.³ There are two different kinds of shortenings, which may not have occurred particularly close to each other in time. One is known as Osthoff’s law,⁴ and involves the shortening of a long vowel when followed by a sonorant consonant in the same syllable (Kroonen 2013: xxiv–xxvi; Ringe 2017: 94–96; Fulk 2018: 55). Classic examples of this shortening include:

(209) *wēn-tos⁵ > *wen-tos > Proto-Germanic *windaz ‘wind’

(210) *(t)pḗr-snah₂⁶ > Proto-Germanic *fersnō ‘heel’

The other type of shortening is consonantal, specifically the simplification of geminate consonants to singletons after long vowels (Kroonen 2013: xl–xli; Ringe 2017: 106; Fulk 2018: 116):

(211) *h₁ēd-tós > *ēssós > Proto-Germanic *ēsaz ‘carrion’

(212) *weit-nós > *hweittós > Proto-Germanic *h^wītaz ‘white’

Such changes reflect a general preference for bimoraic syllables, and similar developments recur throughout the history of many languages, include later Germanic. In English, words such as enwintre ‘yearling’ (from *ān-wintrī) point to a perhaps relatively early shortening that was at least partly conditioned by syllable weight (Luick 1921: 186–188; Hogg 2011: 207–208),⁷ and shortening in closed syllables became very widespread in later Old English (§8.1). Such shortenings do not provide detailed evidence for bimoraic trochees specifically, but especially in a system that retained contrastive vowel quantity, they do suggest that the bimoraic syllable remained optimal, and point to an aversion towards trimoraic syllables. However, as with minimal-word requirements, it seems very likely that general phonetic pressures towards regular syllable lengths (on a phonetic rather than phonological basis) could also be at work here.

The best source evidence for phonologised foot structure in Proto-Germanic is the third: Sievers’ law. This is an alternation involving historical *j and *ij in Proto-Germanic suffixes. No Germanic language as attested retains it as an exceptionless living alternation, but its outcomes are apparent in all branches of the group. The best account of the law is that of Kiparsky (1998), who presents an explanation specifically for Gothic, but whose general principles are easily adapted to apply to Proto-Germanic.⁸ On a general Germanic level, Sievers’ law involves the variation between simple *-j-and a longer variant *-ij-. The details vary from language to language, but in general the reflexes of the two are kept distinct, as can be seen by a comparison of j-stem nominals. As table 13.1 shows, words with a base L or HL before the suffix tend to show reflexes of simple *-j-, while those with bases of H or LL show reflexes of longer *-ij-. The testimony of some forms on the table is particularly weak; these are given in square brackets and discussed below.⁹

Gothic, Old English, and the first Norse column mostly show the characteristic developments of Proto-Germanic *-ja-versus *-ija-: in Gothic, this is reflected by-ji-versus-ei-, in Old English by the presence of consonant gemination and no vocalic reflex versus no gemination and an-e suffix, and in Norse by nothing versus-i-. It is important to note that in the (relatively late) Norse sources, this distribution is no longer phonological, and there are words – chiefly proper names such as Hymir, the eponymous giant of Hymiskviða – that show the ‘heavy’ stem-i-even after light stems (Noreen 1970: 258), but the historical conditioning by weight is nonetheless apparent in most of the lexicon. The position before a retained vowel element in Norse (the pre-V column) is also diagnostic in most words: shorter *-jV-shows retention of the-j-, while longer *-ijV-does not (except after velars). A very similar pattern is also found, in all these languages, for j-stem adjectives and for class I weak verbs.

The essential distribution of forms should by now be very familiar. Just as with high-vowel deletion (chapter 4) and ie-reduction (chapter 6), one kind of form occurs after light stems, and another after heavy stems and, importantly, light disyllables. With Proto-Germanic *j, forms such as *niþ-jaz are prosodically unobjectionable without further adjustments, with the initial syllable forming a bimoraic foot on its own.¹⁰ A form such as *hirðjaz, however, would, if syllabified *hirð-jaz, begin with an overheavy syllable, or else would be divided as *hir-ðjaz, with an unacceptable syllable onset (Kiparsky 1998: 351). The solution – Sievers’ law – is to insert an *i, yielding *hir-ði-jaz with a nicely bimoraic initial syllable.

That Sievers’ law is about the weight of feet rather than syllables is shown by the application of the same process in words such as *raginijaz. The evidence here is chiefly from East and West Germanic: Norse medial syncope (§10.3.2) means that this type no longer readily survives in that language, as LL-base words such as *harisijaz (> hersir) became indistinguishable from plain heavy stems. Given that this syncope postdates the main operation of Sievers’ law by a very long time, they probably do testify to the regular outcome for LL bases, but their collapse with the H bases introduces some uncertainty on this point. I have also not included any examples of LH bases. Such forms exist later in Old English (§4.5.1.2) and Gothic,¹¹ but I can find none that I would confidently reconstruct to Proto-Germanic. Most later examples of such bases in any word-class (not just ja-stems) are derived formations such as cyning ‘king’ or færeld ‘journey’, or univerbated compound words such as woruld ‘world’. If LH words really do only enter the languages after Proto-Germanic proper, that might, perhaps, reflect a stricter avoidance of overheavy feet at that early stage.

Beyond the very robust evidence for Sievers’ law applying after monosyllabic and LL stems, it seems to have also operated after non-initial feet. This is an important sign that the foot in question really was a bimoraic trochee, and not some other variant such as the Germanic foot. The evidence of HL stems in particular is important here, though there are some complications with the data. You may have noticed the blank cell in this row in table 13.1, where I could not find a word in Norse that examplified the outcome-r from *-jaz after an HL stem. That is, I couldn’t identify a masculine ja-stem of this shape in Norse. That doesn’t mean that Norse entirely lacks evidence on this point, however. The genitive Fjǫrgynjar rather than ^xFjǫrgynar groups this word with light-stemmed feminine jō-stems such as ben ‘wound’, whose genitive is benjar – contrast these with a heavy feminine jō-stem such as hildr, genitive hildar. Note also that hildr, like other heavy jō-stems, has acquired an-r ending in the nominative, which is lacking in Fjǫrgyn. While this was difficult to include in tabular form, such evidence as Norse provides for HL stems aligns with the much more robust evidence of Old English.

Unfortunately the evidence of Gothic is not of much value for HL stems. The only potentially relevant forms in that language happen to all be genitive singulars of neuter ja-stems. This is purely an accident of attestation: there are potentially relevant feminine jō-stems such as lauhmuni* ‘lightning’ and verbs such as swōgatjan* ‘sigh’ and glitmunjan* ‘shine’, but these happen to all be attested only in inflectional forms where Gothic does not reflect Sievers’ law distinctions. It is worth emphasising that though apparently diagnostic forms such as ^xswōgateiþ and ^xglitmuneis have a history in the scholarly literature, these are all ghost-words, and are not found in the Gothic corpus (Kiparsky 1998: 353; Goering 2021c: 149–150).

The reason why the neuters are a problem is simple: there is a tendency, in this category specifically, to generalise the ending-jis at the expense of-eis, regardless of weight. This means that where masculine heavy ja-stems have consistent genitives such as hairdeis (identical to the nominative), neuter genitives may be either like trausteis ‘covenant’ or like reikjis ‘dominion’; some words are even attested with both variants (e.g. andbahteis, andbahtjis ‘service in office’; see further Mossé 1942: 92; Kiparsky 2000; Braune 2004b: 94). Since this regularisation is morphologically limited, it doesn’t really obscure the operation of Sievers’ law in the language as a whole, but it does mean that a form such as fairgunjis is ambiguous. It does show the outcome one might expect given the Norse and, especially, Old English evidence, but in principle the ending-jis could have displaced *-eis in this word, just as it did in reikjis and andbahtjis.

This leaves Old English (along with its close relatives such as Old Saxon) to provide the strongest evidence that HL stems behave in parallel to L stems. The presence of an old *j is here, as in all West Germanic languages, signalled through consonant gemination: old sequences of *-Cj-became *-CC(j)-, a change which did not affect *-Cij-(Goblirsch 2018: 41–56). In Old English, light j-stem nouns generally show this gemination, which is most consistently reflected intervocalically before case endings. Relevant examples of HL stems, given here in the dative, include: ānette ‘solitude’, bærnette ‘arson’, brygenne ‘burial’, byrþenne ‘burden’, fæstenne ‘fortress’, hæftenne ‘captivity’, hengenne ‘hanging’, nyrwette ‘narrowness’, rēwette ‘rowing’, and þēowette ‘slavery’ (Dahl 1938: 74–81; Barrack 1998: 159–161).

Taken together with the much slighter evidence from Norse, it looks like Proto-Germanic HL stems behaved just like L stems. This implies a prosodic structure of sequential bimoraic trochees as the basis for Sievers’ law. It is worth noting that Proto-Indo-European may have had some form of Sievers’ law (Byrd 2010a, 2010b: 116–147; Barber 2013: 377–388), and if the Germanic process is a continuation of this, it might be that the bimoraic trochee should be seen as the basic prosodic unit as far back as the histories of English and Norse can be reconstructed.

The evidence of Sievers’ law in particular is most easily explained if Proto-Germanic made use of the bimoraic trochee. The foot formation rules would seem to be simply:

I would not be surprised if the inflectional final *-z, at least, were extrametrical, as it probably was in Early Runic (§10.2.2) and more clearly was (as-r >-r) in later Norse (§12.2.2), but reconstructing the precise details of extrametricality for Proto-Germanic proper is difficult due to lack of evidence.

There seems to be a rather strong avoidance of overheavy feet, as evidenced by both Sievers’ law and the shortenings of overheavy syllables. Proto-Germanic did, however, contain some overheavy syllables. Some of these are monosyllables, such as *rīkz ‘ruler’ or *kaust ‘you chose’, where liberal application of extrametricality could get rid of the unwanted moras, but word-final extrametricality will not help in words such as *berhta-‘bright’, *wurhtē ‘made’, *þāⁿhtē ‘thought’,¹² *aihtiz ‘property’, or *þurftiz ‘need’. Since these all involve *r and/or *h (i.e. [x]), perhaps these consonants could be considered as optionally extrametrical (non-moraic) even word-medially, though it is not clear to me how to test or investigate this possibility further.¹³

A final unanswered (and maybe unanswerable) question: how were LH sequences such as *gebō ‘gift’, gumô ‘man’, or *kuningaz ‘king’ (if this word existed that early) footed in Proto-Germanic (compare Dresher & Lahiri 1991, 257; Riad 1992, 100; Schulte 2004)? Was there a sufficiently strong overheavy licence already present that allowed the footing *(ge-bō), etc., even if this wasn’t prosodically optimal? Or were light, ‘degenerate’ feet more acceptable, leading to *(ge)(-bō) and the like?

Comparing the developments discussed in chapters 4 and 10 suggests a broad parallelism between the earlier prosodic changes in both English and Norse. Specifically, both languages underwent extensive vowel losses, which significantly increased the number of heavy syllables. Words such as *gastiz ‘guest’ and *druhtinai ‘war-leader (dat.sg)’ were reduced to Old English gest, drihtne and Norse gestr, dróttni, all with overheavy initial syllables. In Old English, this all took place prehistorically, but in the runic records of North Germanic, it may be possible to discern a point where overheavy syllables became more tolerated, and forms such as mannz (presumably mænnz or the like) ‘men’, from *manniz, seem to be acceptable by around the year 700 at the very latest (§10.2.1).

In Old English, phonological and metrical data suggest that this increased acceptance of overheavy syllables was not unrestricted. The vowel deletions that produced the early Old English paradigm of hēafud ‘head’ were shaped in part by the desire to avoid overheavy feet: a nominative plural *(hæ⁀u)(-βu-du) was fine, but a dative plural ^x(hæ⁀u)(-βu-dum) was not, precisely because the final two syllables would together have had three moras. The medial syllable was instead unfooted, and eventually deleted, giving the historical hēafdum. But this form, even as it was motivated by the avoidance of a non-initial overheavy foot, created a new overheavy initial syllable (§4.5.1). This suggests that in Old English, at least, a fairly specific overheavy licence had emerged, which allowed overheavy initial feet when necessary, but which did not extend to medial feet, where strict bimoraism continued to hold. This view is reinforced by the metrical patterning of Beowulf (chapter 5), which allowed the resolution of bimoraic words such as wine ‘friend (nom.sg)’ with no restrictions, but blocked the resolution of overheavy LH sequences such as cyning ‘king’ when they occurred in the non-initial position of a word, or (as a result of metrical cohesion) followed a metrically stressed heavy syllable. Overheavy single final syllables were also apparently tolerated, and at least early on in Old English carried secondary stress – an unusual feature for final feet, which otherwise were seemingly extrametrical for the purposes of stress assignment.

The Norse evidence is much less fine-grained for this period, but given what is found in later periods, a similar conditional tolerance of overheavy syllables probably pertained there as well. Both the overheavy licence on initial syllables and the acceptance of apparently overheavy final, stressed syllables are attested to in later poetry, and there seems to be no reason these features would not already be present by the earlier Viking Age.

The first really radical prosodic innovation in either English or Norse was the second round of vowel losses that affected North Germanic in the later Viking Age (§10.3). Like English, Norse had at first avoided syncope in light disyllables, which formed nice bimoraic trochees: sunur ‘son’ is found into the 9th century. But sometime after 800, the unstressed syllables of such words were lost in Norse, resulting in the classical Norse sonr, etc. If the final-r remained extrametrical, then such words were not overheavy, but they now formed their bimoraic trochees in a rather different way, with both moras coming from the same syllable. These syncopes represented a significant step towards the alignment of the syllable and the foot, though they did not involve a full shift towards a syllabic trochee foot type.

Metrical evidence from West Norse suggests that surviving light-disyllabic sequences such as synir ‘sons’ continued to be treated as resolved in verse (chapter 11); evidence of vowel-balance alternations in Old Swedish and Norwegian suggest that the bimoraic trochee persisted generally across much of mainland Norse (§9.5, Riad 1992: ch. 4). But despite the peristence of resolution in its basic form, both metrically and phonologically, its behaviour in verse became significantly limited in several important ways. One is the lack of anything like Kaluza’s law: when resolution is suspended, this is (in almost all cases) an automatic consequence of a preceding heavy, stressed syllable, and affects LL and LH sequences equally. This suggests that while a foot of the shape (LL) remained possible, it was a less optimal formation that was limited to verse-initial position (or its metrical equivalent). The one main exception to the usual rules about resolution concerns words such as konungum ‘kings (dat.pl)’, which resist resolution no matter the metrical context (§11.2). This contrasts with the resolution that can and does take place both in disyllables such as konungr, and the rare LL-initial words such as svaraði ‘answered’. It seems that resolution is a less optimal process in Norse: it is tolerable when the result is a precisely bimoraic word-initial foot, or when it allows the whole word to fit into a single foot (even if this is overheavy). But when the result would be an overheavy foot at the start of a longer word, such as in a hypothetical ^x(ko-nun)(-gum), resolution is not allowed.

Further evidence for the restricted nature of resolution in Norse comes from the ‘constrained position’: the fourth position of a half-line of fornyrðislag or dróttkvætt (§12.1) – a position which also sees a virtual ban on overheavy nominals (Craigie’s law; §12.2). The evidence of Craigie’s law also suggests that final inflectional-r was extrametrical, as it probably was in Early Runic (as *-z), and may well have been in Proto-Germanic.

In English, the story during this same period is largely one of continuity. While the details of northern and eastern dialects are hard to recover, it seems that the bimoraic trochee persisted in the South and West past the year 1200, and in Kentish past 1300. This provided the prosodic context for the variable reduction of ie to i in many dialects (chapter 6), and is reflected in the metrical resolution attested in Laȝamon’s Brut and the Moral Ode (chapter 7). Unstressed vowels also held on for a fairly long time. There were many reductions: already in prehistory, by the Ingvaeonic stage, certain unstressed vowels had merged (§4.3); unstressed vowel quantity was given up in the 8th and 9th centuries (Dahl 1938: 186–191, Fulk 1992: 386–389); and a full reduction to schwa was achieved even in the most conservative areas by the 13th century (Kitson 1997). Still, throughout all this, the metrical structure of a word such as sunu, later sune, was preserved, and it was only over the course of central and later Middle English that such syllables were finally lost entirely.

Schwa loss finally brought Middle English roughly to the point that Norse had reached some centuries earlier, and shortly after (or perhaps concurrently with) this change came a further prosodic innovation: open-syllable lengthening (chapter 8). This change above all moved English strongly in the direction of fully eliminating stressed light syllables, which would have meant achieving a full alignment of syllable weight and stress. Such a system would have been a variety of syllabic trochee, in which bimoraism was not necessarily irrelevant (a requirement for all stressed syllables to be precisely bimoraic was at this point being approached), but in which the foot was not structured around moraic groups. Weight would merely follow from stress.

Such a system did not develop in English – or at least not in the southern and western dialects that provide the best evidence for Middle English prosody. Against the general alignment of syllable structure and vowel quantity was the process of trisyllabic shortening, which created alternations such as boody ‘body’ and bodies ‘bodies’ (whence also bodice). This process was most likely very significantly reinforced by the influx of French loans such as vanity. The changing structure of the lexicon allowed the widespread generalisation of words such as body, saddle, and water in what would become standard varieties of English. This can, in many ways, be seen as a striking reversal of course, away from a straightforward trajectory towards a syllabic trochee, and back towards a bimoraic trochee. Since languages are not, of course, working towards set goals – they are not teleological – such back-and-forth shifts and apparent changes of course are not surprising.

North Germanic took a different course. In many varieties, open-syllable lengthening, together with closed-syllable shortening, did see a full shift towards a syllabic trochee foot, the abandonment of contrastive vowel length, and a simple system where stressed syllables tended to have uniform weight.¹⁴ This is exemplified in table 13.2, whose data is taken from the handbooks referenced in note 14; the lengthening in-VC forms may generally suggest final-consonant extrametricality. In this kind of system, there is no place for the characteristic equivalence of ĹL = H ́ that had characterised earlier Germanic prosody for so long.

I would like to end by going beyond the narrow evidence of prosody in English and Norse, and reflecting briefly on what the prosodic developments in this book might tell us about prosodic change more generally – or at least what these developments might look like in the light of assumptions that currently seem reasonable at this point in the history of phonological research. Diachronically, foot structure is presumably transmitted the same way any other element of a language is: by learners being exposed to the linguistic material around them, and (shaped by any biases in cognitive processing) extracting generalisations from what they hear. Variations in the material each learner hears – along with the effects of phonetic variation of various sources – introduce instability, allowing for the possibility of new generalisations. Contact between linguistic varieties can also produce new variants among speakers of any age. Once multiple linguistic forms exist in a speech community (or even a single speaker’s habits), they can compete and interact in all the complicated ways that we should expect. The exact developments may run along certain expected lines, but there is no ‘goal’ (as Goblirsch 2018: 32, 69 puts it) to disparate prosodic changes taking place over long periods of time – only, at best, pertinacious prosodic frameworks that are successfully transmitted over time, and general influences on what pathways of change may be more or less likely from any given point.

When it comes to the bimoraic trochee, I have argued at length that it was remarkably persistent and robust in the history of earlier Germanic, through medieval English and Norse, before either giving way to a syllabic trochee, or having this change barely averted (or even undone) under the influence of language contact and borrowed words. This large-view telling, however, obscures the many changes that clearly took place even when the bimoraic trochee was maintained – a dynamic of pertinacity, or ‘same pattern, different output realisation’ (Dresher & Lahiri 2005: 75). Even when the foot type as such didn’t change, and continued to be parsed from left to right, and stressed on the first foot, this does not mean there were not significant prosodic changes during these spans of time. Most of these changes were essentially ways to accommodate other linguistic developments, especially vowel losses. Even if Proto-Germanic dispreferred overheavy syllables and feet, this did not stop later speakers of Germanic languages from dropping large numbers of unstressed syllables, and in so doing adding very significantly to the count of overheavy feet in their languages. These reductions could sometimes be influenced by foot structure – compare the protection of high vowels within feet in Old English and earlier Old Norse alike – but some losses took place anyway, and were presumably driven by other factors. The variable tolerances of overheavy feet evident in Old English and Norse prosodies can be seen as foot-based reactions to non-foot-based linguistic changes.

Some of these changes presented greater challenges to the bimoraic trochee than others. The loss of unstressed vowels in words such as sunur in Norse and open-syllable lengthening in Middle English can both be described within the moraic structure of the bimoraic trochee, but feet are not likely to have played a role in driving either change. The former most likely took place for the same reason that vowel reductions and deletions take place anywhere (presumably ultimately connected with the phonetic reduction of unstressed syllables), while the phonetic basis of the latter is evident in how the first general lengthening in English only affected the phonetically longer non-high vowels:¹⁵ some combination of the gradually reduced functional load of vowel quantity (reducing the phonological resistence to such a major upheaval), the phonetic length of lower vowels, contact with other linguistic varieties, and the ever-present role of random chance in language change led to the general lengthening of non-high vowels in (many) open syllables.

The foot-structure responses to these changes was never a given. In Norse, the presence of enough stressed light syllables, and perhaps the effects of cues towards foot structure that are no longer recoverable, prompted learners to continue to parse konungr as a single foot, even though the change of sunur to sunr could have led them to take the syllable as the core prosodic unit and generalise a syllabic trochee. Later on, however, when open-syllable lengthening began to spread to North Germanic, learners tipped the other way: they failed to replicate the bimoraic trochee, and instead generalised a syllable-based system. In English, this change in foot type did not establish itself, but I see no reason why it could not have, especially if the influence of French loans had been somewhat smaller. Learners of southern English in, say, the 14th century would have been confronted with a very messy phonetic reality, and there was no guarantee that a new generation would resolve contradictory pressures (or reconstruct prosodic generalisations) in the same way that previous ones had.

That the bimoraic trochee persisted through these changes and permutations may suggest that it is a particularly easy type of prosodic structure for the human brain to latch on to. If Hayes (1995: 71–74) is right in identifying three and only three foot types possible in spoken human language, then the recurrance of the bimoraic trochee is very easily explained. It is not that these three feet are somehow hardwired into the our cognitive capacity, but that the things that are (arguably) hardwired in – the tendency to find syllables in speech, and to organise syllables hierarchically into prosodic groupings – lead to only a few possible (or at least likely) outcomes when confronted with the physiological and acoustic realities of spoken language. The basic phonetic fact that some syllables can be longer in duration than others is perhaps a sufficient basis for the recurring role of the mora in human languages, and any language that phonetically maintains a variety of syllable lengths not obviously derivative from stress is open to having such variation phonologised into moraic feet.

From this perspective, the history of the bimoraic trochee in Germanic takes on a slightly different appearance. The general status of most coda consonants as moraic put the onus of syllable weight distinctions largely on the vowels in Germanic: in contrast to a language such as Cahuilla where-kaʔ-is a heavy syllable but-ʔiš-is light, in older Germanic light syllables could only be open syllables with short vowels. The extensive maintenance of vowel length contrasts throughout the word in prehistoric Germanic, along with an inherited lexicon filled with many stressed syllables of varying lengths, would give learners ample resources to constantly recreate the bimoraic trochee as long as these linguistic features persisted. The periodic shocks of vowel reductions, often going hand-in-hand with morphosyntactic shifts, as well as other quantity readjustments, together increasingly reduced the evidential basis for continuing the bimoraic trochee, until it was either abandoned (as in North Germanic), or the lexicon was sufficiently changed in a way that reestablished the potential basis for the foot type (as in English). In every case – whether during the uneventful periods of little relevant change, or during the major transitions – learners were doing the best they could to extract a reasonably consistent prosodic system (a foot type and pattern of footing across the word, rules for stress, and any necessary tolerances or allowances) from the inevitably messy and phonetically inconsistent sounds they heard around them, and which they themselves were coming to produce.