3.1 Middle English long high vowels+r

Lass (1999b, 109) discusses several facts that point to schwa-epenthesis as a late-ME phenomenon, or one that started in the very early period of EMoE. From the fifteenth century onwards spellings like hyar/hyer for hire start appearing in the Cely Papers, dated to around 1475 (Wells 1982, 214; Dobson 1957, §218, etc.) for ME ⟨-īr⟩. Hart (1569) has ⟨-er⟩ transcriptions for fire, mire, dear, here (discussed below). Cooper (1687) has hire = higher, tire = ty her. Nares (1784) says that power, hour are disyllabic (Lass 2006), etc. OE words in ⟨-ūr⟩ regularly have awə/ɑə/ɑː (< awər < awr) in SSBE. The spelling of these is either (unpredictably) ⟨our⟩ or ⟨ower⟩. Of course, originally disyllabic words have ⟨ower⟩ or something similar, see (1). The same holds for words having OE iː or Old French (OFr) words that developed such a vowel in ME. All these words predictably have ɑjə/ɑə/ɑː before r in SSBE, see (2).

There are at least four ideas on how the Great Vowel Shift (GVS) affected the ME long high monophthongs, and all four have one thing in common: the presence of postvocalic j/w. According to view 1, ME iː, uː is found as EMoE əj, əw, a view espoused by many traditional textbooks (Kökeritz 1953; Dobson 1957; Strang 1970; Görlach 1978, 1991; Cercignani 1981; Minkova 2014; Crystal 2016). View 2 has it that ME iː, uː changed into EMoE ej, ow (Sweet 1900, §721; Jones 1909; Chomsky & Halle 1968; Wolfe 1973; Lass 1999b). View 3 (Kristó 2015, 192) holds that the ME long high monophthongs were first diphthongised to ij, uw. To this can be added Sweet's (1900, §721) view, who calls the change from ii, uu to ij, uw (and later by divergence) to ei, ow (still later to əj, öu and finally ai, au) cleaving (i.e., the gradual increase in sonority of the first half of the diphthong, or a type of dissimilation). The least current view holds that ME iː, uː reached the EMoE əj, ow stage in two steps, first becoming ɨi, ɨu, then əj, əw (Stockwell 1961), summarised in (3). Whatever the beginning, the two ME vowels result in the modern wide diphthongs aj, aw (with the quality of the first elements greatly varying from accent to accent).

There are very good reasons, rooted in the contemporary orthoepical descriptions of the time, to believe that view 2 is the correct one (cf. Lass 1989), at least for EMoE. The EMoE diphthongs ej, ow never merged with any other ME diphthongs in EMoE (at least not in the predecessor of SSBE) with əj (< ME uj poison, after the u > ə change in some dialects) or ɛj/ej/eː (< ME aj, day/wey), or with oː (< ME ɔː, no) or ɔw (< ME āw, know, blow). Whatever the reality was, all attempts analyse the result of the first phase of the GVS as involving diphthongisation (of an original high long monophthong).

However, we will see below that diphthongisation does not have to be assumed at all, as there were already j/w-final diphthongs in the system, at least phonetically so. Even if there was no epenthesis before the GVS set in, now there is, as shown by examples like hyer/hyar, tower with ejər, owər. There are further examples: mire (see above) or wire (< OE wīr), which were also spelt mier and wier in ME, sour as sower and shire as shier, chier. This is to be expected if we bear in mind the well-grounded assertion that English spelling has never been rationalised (Brengelman 1980). Spelling may not be reliable, but phonology is. We have evidence for breaking in (late) ME for the j/w-final diphthongs (coming from OE long monophthongs), continued into EMoE and the later periods, but not always orthographically visibly.

The next question is whether there remained other diphthongs after ME ‘long i’ and ‘long u’ were diphthongised. The words briar is lumped together with the words in (1) and (2), as in Lass (1999b, 109). However, the word is problematic if it is placed in the same group with fire and tower, as it did not contain a long high monophthong in ME. The OE word is (Anglian) brēr, brǣr. The ME ī (or ii according to some) and ē (ee) before r show a special development, a merger with ī (continued as ij in late ME, and later as ej in EMoE, as in hide), something that does not occur before the other consonants, and, more importantly, is not found for the short vowels i and e, as words like ferm, ster, lerke are not found merged with i. Rather, e is lowered to a or ɑ before r: farm, larke, a reaction (in all likelihood) to pharyngeal r.

Let us propose that ME brer was brejr. With breaking we have brejər, spelt brier, briar, briyer, breir. As discussed earlier, words with OE ī (as well as words from OFr or Latin) like wire, desire, attire, tire all show breaking with ijr > ijər. There is uncertainty about the quality of the short vowels in ME (cf. Stockwell & Minkova 1990), but it is conceivable that the merger of ijər and ejər was possible because ijər was (phonetically) ɪjər, with a centralised (non-peripheral) vowel close to e, which was raised and merged with it. Whatever the phonetic content, examples show that ejər merged with ijər. In addition to briar, we have further examples (in words of either OE or OFr origin) for this merger in (4), from Jespersen (1961, §3.125) and MED.⁵ The merger must only have occurred in some parts of England, as it did not affect all the available inputs (OE dēore > dear, dēor > deer, **dire, OFr chēre > cheer ‘face’, **chire).

Not only do we have examples for the merger of ejər and ijər as ijər (traditionally: ē, ī > ī), there is also the (occasional) confusion of ē for ɛ̄ (and the other way around) before r. What follows is a reinterpretation of Jespersen (1961, §3.244). This is another special development of the two non-high front vowels before r. Spelling (late as it is, starting to show a distinction in the orthography in the 16th century between EmoE ī (ME ē) and ē (ME ɛ̄) as ⟨ee, ie ei⟩ and ⟨ea, eCe⟩, respectively) shows ⟨ea⟩, where ⟨ee⟩ would be expected for OE ēo or ē: while steer, leer, beer are regular (in showing OE ēo), dreary, dear, weary show (less regularly) a lower ‘long e’ (for OE drēorig, dēor, wērig). ⟨ea⟩ is regular in ear, sear, near, tear, all with OE ēa, and also in spear (OE spere with ME open syllable lengthening). There are also indications that fear (OE fǣr ǣ) and year (gēar ǣ) are found with ME ē. French words like appear, rear, clear, mere, etc. also have ē (although some of them may have had ME ɛ̄). It is impossible to give a full appraisal of the developments before r here, but ⟨ea⟩ in dreary looks orthographically similar to ⟨ia⟩ in briar. It is possible that ME ɛ̄, similarly to ē, may have been diphthongal, i.e., ɛj. Before r we should like to find breaking: ME ere ‘ear’ ɛjr > ɛjər. The closer version of (ej) had a similar development before r (as discussed for friar ejr > ejər). It is conceivable that there was epenthesis producing another possible locus for merger: ere > ɛjər < friar. This merger vied with that of ejər > ijr < ijr. In this conflict of mergers, there is little wonder that the result of the merger was unpredictable: dear (OE ēo) shows a form presumed to have had ɛjər once, merging with year where ɛjər is historically ‘regular’ (OE ēa), while steer underwent no merger with ɛjər, showing the regular continuation of OE ēo, ME ejər. In addition, some OE words with ēo chose yet another path, merging with ij (desire with ME ī, tire with OE ēo). The only merger we do not encounter is the long front low vowel merging with the long front high vowel: OE ēa/ǣ before r is not found as late ME ij (traditonally ī), but OE ēo/ē (> ME ē) before r is found as ME ijər (briar), ɛjər (dear), or unchanged ejər (steer). As ejər represents the middle ground for the two merging tendencies, we are not surprised to find examples of OE ēa/ǣ (ME ɛ̄, or rather ɛj) before r merging with ejər (fear with ME ē), see (5) for a possible schematic representation of these competing mergers.

Admittedly, the diphthongal representations for what established wisdom (e.g. Smith & Horobin 2002) has long monophthongs in ME (ī, ē, ɛ̄) may be unusual, but we should bear in mind that ME did have Vjər (ijər, ejər, ɛjər, ?æjər) sequences from OE and OFr bi-syllabic sources that do not appear to stir much sensation. In addition, ij/ej developed out of i/e before ç. An epenthetic schwa in jr sequences in (5) would not have been out of the ordinary, see (6).

SSBE, as a modern reference dialect, shows the (lexicalised) continuation of a number of co-existing lineages, hence the absence of any Lautgesetz in (5) in the Neo-Grammarian sense of the term. From the point of view of SSBE, which word remains in (5b) or (5c) is haphazard: whereas the selected list of words in (5a) continue their lines undisturbed into Modern English, dear, dreary, fear, ear in (5c) now seem to be present continuations of the changes in (5b), just like deer (the ‘regular’ continuation of the OE vowel), indicating that whichever merger was found in whichever source in the past, there must have been a parallel lineage where such a merger did not happen (and this is the one inherited by SSBE, an amalgam of many parallel lineages). Words that remain in the line of development represented by (5c), spelt with ⟨ea⟩, are tear (v), wear, bear (having ME ɛ̄ from OE e in open syllables).

OE (as well as Scandinavian and French) words with a short a in open syllables in ME develop aː. None of these words came to be spelt with ⟨ea⟩, showing that they were monophthongal (care, ware, snare, shake, naked, able, male, vapour, etc.). As a result of a long-protracted series of changes, most words with ɛː (from ɛjər) in SSBE are derived from this source (words with ⟨ea⟩, pronounced ɛː, are in the minority, but this may just be an accident, see (5)). Judging by ɛː in SSBE, ME aː must have developed a diphthongal pronunciation to be affected by r along the following lines at some point: care kaːr > kajr > kæjər > kɛjər > kejər, from which kɛə (with smoothing, laxing and monophthnogisation) > kɛː straightforwardly follows, cf. also Jespersen (1961, §13.321), who says that ME aː and æj do not become ej before r, but rather (in his transcription) ɛ^.ər, which starts with a low, front long vowel gliding on to the next vowel, found as such in the 18th century. Whatever the phonetic reality, it seems once a low long front vowel becomes raised/fronted, it develops into a diphthong (aː > ?æj/ɛj), which has nothing to do with r (as it is also found in non-pre-r environments, as in late, make). Maybe a back low long vowel when raised also becomes diphthongised.⁷ If the reaction of ME to French ē (as in frēr) was diphthongisation, ME was then similar to SSBE, where the same would happen to any non-low long vowel arriving with a loanword: iː > ij (petit pətɪj), uː > uw (voodoo vuwduw), eː > ɛj (liberté lɪbətɛj), oː > əw (mauve məwv), as opposed to suave ɑː.

3.2 Middle English back vowels+r

Not surprisingly, similarly to front long (or rather diphthongal) vowels, there is also confusion in the back vowels+r sequences (cf. Jespersen 1961, §13.351; Lass 1999b, 111, among many others). One is tempted to say that r has been a great leveller of contrast (poor, pour, force, course, north all have oː today). As Lass (1999b, 111) says ME uːr, oːr, ɔːr is not a good predictor of its nuclear vowel in EMoE. Additionally, the picture is further complicated by ‘short o’+r (short, north), which will ultimately lead to another source of ‘long o’ before r. If ME uːr is not followed by a consonant, we usually have the expected EMoE owər. In ME poor we see that oːr became uːr (as traditionally understood) due to GVS, giving the expected (now dated) ʊə in the predecessor of SSBE. From the point of view of SSBE, ME ɔːr will ultimately merge with ME aw (lore = law) as oː, but lore still had oə/ɔə for Jones (1918), at a time when law was ɔː (but not oə/ɔə). Generally, however, as Lass (1999b, 111) contends, back vowels before r have “messy histories”. In addition, EMoE also inherits from ME the continuation of OE āw (as in know) and ōw (as in glow). The EMoE pronunciation (Hart 1569, 1570) seems to be ow. Our purpose is not to disentangle the impossible, but try to approach the problem from the point of view of breaking, a phonological indicator of the presence of diphthongs.

Some of the orthoepists of EMoE offer invaluable information on the pronunciation of the time. Hart (1569, 1570), as analysed along different lines by Jespersen (1907), Danielsson (1955, 1963), Lass (1980) and many others, discusses all the vowels of the time. There is little doubt that the continuation of ME iː (ij) is ej (cf. Lass 1980, 1989; Starčević 2022), rather than əj as most standard textbooks have it (e.g. Algeo 2010, 144). Hart's ‘long e’ ([e^.] in Jespersen's standardisation) is the continuation of ME ɛː, his ‘long o’ ([o^.]) is generally the continuation of ME ɔː (as in go, note). What is of interest for the analysis is that Hart also shows a “vowel of mixed quality” (= schwa), as Jespersen (1907, 30) puts it, in words where such schwas are not justified historically, shown in (7), transcribed in a more modern fashion in (8).

Hart's phonetic rendition for fire, mire, pure and sure is not surprising (they show breaking in jr, wr). Hart could not have written ⟨mejər⟩ or ⟨pjuwər⟩ for disambiguation as for him j and i, as well as w and u, are the same entity, which makes it difficult for him to render into his new orthography words like world, would (wu) or ye (jij, which, at an abstract level, is an ‘overlong i’, i.e., iii). Hart also has pou^.er, pouër ‘power’, which was bi-syllabic (similarly to piuer ‘pure’), as well as hie^.l ‘heal’ with epenthetic shwa before ɫ. Jespersen (1907) does not recognise the importance of schwas in these words. What is more interesting for us are words that (traditional analyses assume) show EMoE monophthongal oː (oar from ME oor ɔ̄r) and iː (dear, here showing EMoE iː, a more closed version of ME ɛ̄, not uncommon in Hart). A word like e^.er shows a disyllabic pronunciation of perhaps ever, alongside other historically disyllabic words like viuer ‘viewer’, deier ‘dyer’, beier ?‘buyer’, heier ‘higher’, pouer/pou^.er ‘power’, ple^.er ‘player’. Whether ⟨er⟩ in the words in (8) shows what it does, a sequence of a schwa+r (piuer pjuwər) or a syllabic r (pjuwr̩) has no bearing on the argument: these shwas are historically ‘unjustified’, as traditionally understood.⁸

Hart is not infallible though: there are no schwas shown in aspi^.rd, api^.r, tʃi^.r, tʃe^.r, e^.r ‘ear’, pre^.r ‘prayer’, mo^.r, fo^.r, befo^.r, ro^.r, so^.r, etc. in words with etymologically identical vowels to those shown in (7). Perhaps for Hart the schwas in these words were so entrenched in his dialect that he glossed over the intrusive schwas in words like api^.r (rather than expected api^.er), after all, everyone would have had appear as əpijər, or, for lack of anything more phonological, one can simply blame the absent ⟨e⟩ on the printer's shop and negligent typesetters. The reason the words in (8) are shown with j/w makes sense if we look at Hart's transcriptions of the short vowels before r: whatever quality they have (and it seems i, e and u are still kept apart qualitatively), there is never an intrusive schwa between a short vowel and r, as expected, in lerner (for er), afirmd (ir), murðr (ur), former (or), etc. We may conclude that his attempts to show an epenthetic schwa in o^.er ‘oar’ is justified by the impossibility of the ‘long o’ having been a sequence of two ‘short o's’: oar is thus /owr/, rather than /oor/, pronounced owər). A caveat is in order: one must be careful in admitting a diphthongal pronunciation into the analysis. MacMahon (1998, 459) points out that the first account of a diphthongal pronunciation of oː comes from the Scottish orthoepist William Smith in 1795 (cf. also Wells 1982, 210). While this may be true, accounts such as this one are not grounded in phonology. Phonological proof must remain what it is: phonological.

3.3 How many diphthongs did Hart really have?

Hart's Orthographie (1569) and Methode (1570) were analysed by Jespersen (1907) and, more extensively, by Danielsson (1955, 1963). There is no doubt that Hart changed some of his conceptions regarding the representation of the front diphthongs in his Methode. In 1569, Hart had the same representation for ME ɛ̄ and ME ej/aj, namely ⟨e^.⟩, see (9).

Words in (9a) and (9b) have the same vowels, if we follow Hart's transcription. This cannot be correct, as the two lineages do not merge: words in (9a) have ij today, those in (9b) have ɛj (Jespersen's [e^.i]). This leads us to the impossible conclusion that Hart is describing a merger here, which was ‘undone’ after his time exactly along the correct historical lines, so there are no unexpected crossovers: deal being dɛjəɫ, May being mij, for example (cf. also Lass 1980, 87). We must agree with Jespersen that Hart got something wrong. He is somehow unable to correctly disambiguate the vowel in sea from that in say, whatever the vowels were.

As far as ‘O’ is concerned, Hart does expressly (if not successfully) differentiate between a vowel consisting of two ‘short o's’ and one of a ‘short o’ followed by a ‘long o’. Whatever he meant by this, he wrote o^.u in o^.un ‘own’, so^.ul ‘soul’, but he generally has o^. (or just a short o) in gro^., kno^., kno^.n, tʃarko^.l, etc. Charcoal probably shows a vocalised ‘dark l’, but grow, know, known, own all have ME ɔ̄w (< OE āw or ōw). Also note that ⟨ou⟩ is systematically used for the diphthongised reflex of ME uː (in traditional accounts), as in hous, mous, out, sou ‘pig’. Hart is obviously struggling trying to capture the set of oppositions of his time: own never merged with clown, but it did with words that continue ME ɔː as oː at a later point in EMoE (cloak = own). There were also those words that had ⟨u^.⟩ (ME oː), as in tu^. ‘two’, tu^.k ‘took’, su^.þ ‘sooth’, see (10).

The pressing question is why Hart has ⟨e^.⟩ in words like day and obey. Jespersen argues that Hart did not know how to deal with diphthongs in 1569, just as he wrote gro^. for grow (although it had a diphthong), he wrote de^. and obe^. for day and obey. In 1570, he corrected his view on the back diphthong (now writing o^.u in soul and own), but not on the front one in day (cf. Lass 1980, however, who discusses a possible say–sea merger). Perhaps a new (third) edition would have had e^.i for day. There are a few isolated examples in Hart that show what he may have contemplated upon: in 1569, we have ste^.id for ‘stayed’ (also ste^.d) and ste^.iþ for stayeth (all meant as monosyllables). There are other tell-tale signs of a diphthongal pronunciation in the isolated aehtþ ‘eighth’ (with ⟨ae⟩ from Latin), also with ⟨ai⟩ in auluai(s) ‘always’, generally written aulue^.z, tʃamberlain (in these examples, the supposed diphthong occurs unstressed, and perhaps it is the spelling conventions that crept back). It is only in painter where we find ⟨ai⟩ stressed (but painted is pe^.nted). If these ⟨ai⟩’s were not just errors, Hart's ⟨e^.⟩ was intended to show a diphthong with a low first element (aj perhaps, or æj, as Jespersen argues). Descriptions from Hart's contemporaries are difficult to interpret and often show an artificial difference in the pronunciation of diphthongs in, let us say, paint and day, something based on spelling.

Hart's descriptions of his ‘long a’, that is ⟨a^.⟩, in words like gates, rare, gave leave little doubt that the quality of this vowel was low front (or central, undiphthongised, cf. Jespersen 1961, §8.13; Lass 1989), as do his descriptions on the continuations of ME iː and uː (or rather, as discussed, ij and uw), which are shown as ⟨ei⟩ and ⟨ou⟩. In a more modern transcription, they must be ej and ow (as in leif ‘life’, -lei ‘-ly’, bound, found). These are never shown as (or mistaken for whatever sound Hart had before his mental eyes for) ⟨e^.⟩ (day, deal) or ⟨o^.⟩ (grow, home), meaning they must have been sufficiently distinct from each other to preclude confusion. Jespersen (somewhat hazily) describes ej and ow as ‘fast’ diphthongs, the ones in day (?æj or ɛj) and know (?ɔw) as ‘slow’ diphthongs. The diphthong in day will ultimately merge with the vowel of name, that of know with cloak. Jespersen explains the difference between the two by saying that the ‘fast’ diphthongs had a short first element and a more rapid upward movement, probably their first element was also closer.⁹ This probably translates as ej (my) vs æj (day) and ow (house) vs ɔw (know). No member of these pairs merged with the other member of the pair (day ≠ my, know ≠ house), see (11).

Jespersen's difference between short and long first elements of these diphthongs would usher in a much-contested difference between short and long diphthongs (cf. Lass 1980), something we have no evidence for, especially if a difference in height of the first elements (or the relative difference between the starting and the end point of the diphthongs) can explain the difference.

3.4 Diphthongs with back first vowels

Hart must have had at least 4 diphthongs. But did he have more than 4? We saw that he showed an epenthetic schwa in oar ⟨o^.er⟩. The data are scant, but it is worth not discarding the idea that ⟨e⟩ in ⟨o^.er⟩ was more than just a typographical error. We have seen examples for epenthetic schwa on several occasions: these keep appearing in what must be jr/wr sequences. He also had an epenthetic schwa in pure and sure. If Hart showed a schwa in oar, let us assume there was a schwa because there was a diphthong there (and consequently an epenthetic schwa before r). This gives us a proliferation of these entities: if pure, sure has juw, house has ow, own has ɔw, what does cloak and home have (Hart's ⟨o^.⟩)? Perhaps the diphthong in own had a more open first element (ɒw) for him. This diphthong originates in ME ɔw (OE āw/ōw) in traditional accounts. If for a change we think of diphthongs as simply sequences of vowel+j/w (which is possible for modern English, see Szigetvári 2016), the lowering of the first element of Hart's ɔw may be regarded as a possibility (ME short monophthongal ɔ was also lowered to ɒ, as in soft, cot, etc., as witnessed by Elizabeth I's spelling of stop as stap). It's also possible that Hart's ɒw diphthong simply continues the ME vowel unchanged (see further below). Let us see this new back territory of vowels reinterpreted in (12).

Of these four diphthongs, it is ⟨u^.⟩ and ⟨o^.⟩ that are not recognised as diphthongs by Hart. Of these two, uw is a (very) narrow diphthong. Recognising its diphthongal quality took a long time, and even then it disappeared from sight for decades in the 20th century, so this oversight on Hart's part can be forgiven. The interesting question is how Hart failed to recognise the diphthongal quality of ⟨o^.⟩, as set against ⟨ou⟩, which he pinpoints with great accuracy and does not confuse with other back vowels. As far as the vowels were concerned, Hart worked only with the 5 letters of the alphabet (modified here and there with diacritics), and these were woefully inadequate to capture all the oppositions. Hart obviously found it difficult to capture and discuss the various degree of openness (observing the degrees of openness is notoriously difficulty, much more so than pinpointing the points of obstruction in consonants). Perhaps Hart simply did not have the requisite amount of vowel letters to work with (or perhaps the knowledge): the u's, o's and dots were simply inadequate to show all the oppositions.

We may say that Hart was also aided in his recognition of the EMoE ow diphthong by the traditional orthography: ow was perfectly represented by the digraph ⟨ou, ow⟩ (house), whereas ⟨o^.⟩ was only sometimes shown with the digraph ⟨oa⟩ (moat) and even then the second half of the digraph did not correspond to an a sound in ⟨o^.⟩. The diphthong ej was also represented (if not perfectly mnemonically) by ⟨i, y⟩ (in a free graphic position: fire, lyre, cipher, mitre). It is also possible that in words like day/obey Hart did not think it necessary to show the diphthongal vowel as ⟨e^.i⟩ because conventional spelling had ⟨ay/ey⟩ anyway (despite Hart's wish to liberate himself from the conventions and use his own orthography consistently). Of course, as discussed, it remains a mystery why the vowel in words like deal, each was shown to be identical to those of day, obey. No reformer can be perfect.

A later change will take away the ow (cow) diphthong by centralising its first element, producing əw (and even later aw). ɒw (know) and ɔw (home, cloak) will eventually merge and raise to ow (traditionally taken to be oː, to be diphthongised to ow, see below). There is no comparable merger to the meet–meat merger in the back high and mid-high region: boat was not merged with boot.

3.5 Diphthongs with front first vowels

The next interesting question is how many diphthongs with first elements as front vowels Hart had. We see from Heart that words with ME ɛ̄ (like read, stead, streaks) already have a ‘long i’ alongside words with ME ē, where ‘long i’ is expected (need, see, teeth). We also have dier ‘dear’ and hier ‘here’, where ⟨e⟩ show breaking before r. Given that epenthesis happens in jr clusters, dier and read must have the same vowels, which we will represent with a diphthong ij: Hart's ri^.d ‘read’ and diër ‘dear’ show the same vowel, which happens to be followed by schwa in dier as a result of epenthesis.

There is also Hart's ⟨ei⟩ for ME ī, found as ej in EMoE, as expected. The appearance of the epenthetic vowel is expected, as shown by words like feier ‘fire’, meier ‘mire’. There is also the diphthong æj (or aj, cf. Lass 1980) in words like day, obey that Hart had difficulties representing (see (9)), but which he (probably) intended to show as ⟨e^.i⟩ (i.e., a diphthongal vowel on a par with ⟨o^.w⟩), but never managed to introduce the symbol for. There was also the back (central?) long vowel in words like name, blame, etc. (“with wyde opening the mouth, as when a man yauneth”, depending on how one interprets the act of yawning). The question is what the vowel was in words with ME ɛ̄ that had not yet been raised to EMoE ij. They are shown as ⟨e^.⟩ by Hart, but were already unstable as testified by the ongoing early merger with ij (read).

The question now is what may the diphthong in words like bear, deal, death, fear, Jesus have been, where we see Hart's ‘long e’, that is ⟨e^.⟩ (ME ɛ̄). Although there are no straightforward examples for epenthesis before r in words where modern spelling has ⟨ea⟩ (bear is be^.r), let us here entertain the possibility that ‘long e’ was in fact a diphthong. If a word like bear had a diphthong and an epenthetic vowel, Hart would have struggled to show this orthographically: ?be^.er ‘bear (v)’ (⟨e^.e⟩ is only found in ple^.er(s) ‘player(s)’ where it ⟨e⟩ is part of the agentive suffix, and e^.er ‘?ever’ where ⟨^.⟩ may simply show the omission of ⟨v⟩), ?beier (he would have had difficulties showing a diphthongal ‘ey’ which is different to the continuation of ME ‘long i’). Let us suppose, unorthodox as it is, that it was ɛj, wedged between ej and æj (two diphthongs that seem justified in Hart).

Given the constraints of the rather confined vocalic space in which this diphthong was wedged, Hart would have had no other orthographic means but show it as a ‘long e’ (⟨e^.⟩). ɛj would have been impossible to pinpoint accurately as a diphthong (similarly to ⟨i^.⟩ ij in need) and describe it using the loose scientific terminology of the time. From the further developments we see that words with ⟨e^.⟩ (those that continue ME ɛ̄) will more readily merge with ij (meet) than æj (days, daze) in the South of England. A handful of Hart's ɛj's will ultimately merge with æj (the well-known examples include steak, great, wear, bear, yea). In the long run, ɛj will merge with ij (meet, meat) after ej centralised to əj (fine), emptying the position for æj to move in (which was ej for Sweet, later still ɛj, day, great), with which the fronted aː will eventually merge (daze, days, great; Wells' long mid merger). This chain of events is parallel to that of the no–know merger in the back (note that Hart still has aw for ME aw, it had not yet monophthongised to ɒː, as in sau ‘saw’, law ‘law’, tauht ‘taught’), see (13).

Hart recorded all the oppositions right using his orthographic conventions. No opposition was neglected or glossed over (if we let him get away with the slip-up involving ⟨e^.i⟩). In Hart's times, there is still no trace of the systematic raising of ME ɛː, ɔː or aː (EMoE ɛj, ɔw or aː as analysed here), used as evidence for the claim that the GVS cannot be seen as a uniform process affecting all the long ME monophthongs in one fell swoop (cf. Stockwell & Minkova 1988; Lass 1989, 1992). As we saw above, it has been a bit of a mystery why Hart conflated the continuations of ME ɛː (sea) and aj (say) as ⟨e^.⟩, which Jespersen put down to Hart's inability to deal with the front diphthongs, and Lass (1980, section 4) to a possible say–sea merger that characterises some North and West Midland dialects. The evidence is complex and not conclusive, but viewed from the point of (13) perhaps Hart found it difficult to differentiate between ɛj (sea) and æj (say), or explain what the difference was. It seems more probable that Hart found it difficult to differentiate between two (very) similar diphthongs than (as traditionally claimed) between a long monophthong (ɛː) and a diphthong (ɛj/æj). Admittedly, (even) in Hart's time the space where these 4 front diphthongs cohabitated was crowded and would be even more so after aː (name) began to be fronted (and diphthongised). In contrast, there were only 4 back diphthongs. With the centring of ow only 3 will remain (and later 2 after the no–know merger). This may explain why there was no push for a boot–boat merger, as opposed to the front half which (even after the centring of ej) had 4 remaining diphthongs (meet, meat, days, daze), leading to the meet–meat merger and a system of 3 diphthongs (meet/meat, days, daze), and ultimately 2 after the days–daze merger, producing a more balanced system of 2 front and 2 back non-low diphthongs: ij (meet, meat), ej (days, daze), uw (boot), ow (no, know), see next section. These will later be complemented by aj (< əj, mine), aw (< əw, cow) and ɔj (boy).

5.1 Parallel lineages

One of the consequences of the analysis is that the comparison of the various accents of English becomes slightly more entangled at first sight. Wells (1982, 210) calls long mid diphthonging one of the most recent innovations in the predecessor of Received Pronunciation (and generally in the south of England), setting it apart from the north. It follows that the monophthongal pronunciation (eː, oː) was the original one (after the days–daze and no–know mergers), from which the diphthongs can be derived in those accents that have them. It also served as one of the reference points in the categorisation of accents, known as face-goat diphthonging. The wide diphthongs (whatever their phonetic realisation today) are considered part of all varieties of English, given that they form a shared inheritance continued (in some form) from ME iː, uː and oj/uj. Of course, Wells (1982) does not have a comparable rule in the form of ‘long high diphthonging’ that would take care of ij and uw in SSBE. If it was recognised, this last diphthonging rule would be historically the latest one in the long series of diphthongisations. However, the data we have gathered for epenthesis show that j/w-final diphthongs must always have been part of English, starting from at least ME. The question now is how the accents with eː and oː end up with long monophthongs if the mid diphthongs have never been extinguished under the long monophthongs.

The explanation may lie in parallel lineages: OE had both monophthongal long vowels and diphthongs, which were, respectively, continued as long monophthongs and diphthongs in ME. j/w-final diphthongs show the result of vocalisation, a process whereby OE palatalised ɣ was reinterpreted as j, and ɣ as w. Additionally, there was postvocalic w, adding to the pool of new diphthongs in ME (cf. Sweet 1900, §806; Lass 1992, 50 for a full inventory), to which we must also add OFr diphthongs ending in j/w. As Minkova (2014, 266) says, the history of ME long monophthongs cannot be separated from the history of pre-existing diphthongs in the system and that “allomorphic clouds merged towards either a diphthongal or a monophthongal realisation” (265).¹⁰ By allomorphic Minkova probably means lexical free variants, similarly to the competing pronunciations of either, room, economic, fulcrum, etc. The difference between Modern English free variants in either, for example, is that in Modern English aj and ij are phonemes. This is where the difference lies: in ME there was no phonemic opposition between ij and iː, uw and uː, ej and eː, etc. Our analysis claims that the system of opposition rested on either ij vs uw vs ej, etc. or iː vs uː vs eː, etc. (see next section).

5.2 uː/iː, uw/ij > uw/ij merger in the South?

But what was this merger exactly? Given the origin of some the vowels, it is plausible that ME had both uː and uw at the same time, sometimes even in the same word, or possibly one variant was preferred in one phonological position (word-final, for example), the other in another position (similarly to what we can observe in SSBE for the broken vowels, cf. Lindsey 2019, chapter 13). The vowels of OE sū suː and sugu suɣu ‘sow (female swine)’ would be continued as ME uː and uw(ə): soue, suw, suwe. Let us see the data in (16) for diphthongs with front first elements first.

The details of the diphthongisation processes, as well as the inputs, are complex and manifold (cf. Minkova 2014, 206f; Brinton 2017 and especially Maiwald 2017 and references therein), but not immediately relevant. The point is that these new diphthongs¹¹ of ME coexisted with (presumably) long monophthongs continued from OE: līf ‘life’ (ME life, leif), sǣ ‘sea’ (ME se, sei), sēon ‘see’ (ME sēn, sein, sen), etc. Although it is notoriously difficult to assess what spellings with ⟨iy⟩ or ⟨ei⟩ for expected monophthongal ⟨i⟩ or ⟨e⟩ in ME (or any other period) stand for (cf. Minkova 2015), there is every likelihood that these long vowels were diphthongal (irrespective of their OE monophthongal origin, as traditionally assumed). ME ⟨ou, ow⟩ may be the perfect representation of ME uw (with which uː had previously merged). As Minkova (2014, 257) says, for iː (of life) and ij (of bridle), the perceptual difference between the two would have been minimal and hence impossible to maintain in the long run (the same goes for uː and uw). Possibly, such a minimal difference never existed in the first place: our claim is that ME words like līf were reinterpreted as lijf, based on brīdel brijdəl. Similarly, fire fiːr is now fijr and thus a pharyngeal onglide can form in jr (> jɑ̯r), merging it with ijə observed in words like sijər (< OE siger) ‘glutton’. Put simply, the option between iː and ij was tilted towards ij in this (southern) variety of ME, a situation which seems to have been the reverse of what happened in another (typically northern) variety, where bridle was reinterpreted as briːdəl, based on liːf. Here the merger worked in the opposite direction, merging the ME j/w-final diphthongs with the long monophthongs. We have also seen examples for OFr ej adopted in ME as ej (freyer, friar), adding to the pool of ME diphthongs.¹² We will go as far as to say that ME friar had both a diphthongal (frijər) and monophthongal (friːr) pronunciation, but not in the same variety of ME. The same holds for fire (< OE fīr, fȳr), which is either fijər or fiːr, but not **fijr once the stage for prevocalisation was set (cf. the very same situation in Modern English fête pronounced as either fɛjt or feːt, depending on the variety). Let us see some data for diphthongs with back first elements in (17).

The diphthongs o(ː)w and ɔ(ː)w, as well as æ(ː)j and e(ː)j will merge into back and front diphthongs, respectively. In addition, any contrastive length difference that may be postulated based on OE was lost and length was reinterpreted as a feature of the whole diphthong, rather than its first (more sonorous) part (the vowel), cf. Maiwald 2017, 63; Minkova 2014, 208, who says that metrically the two kinds of diphthongs behave identically. Lass (1992, 50) reconstructs these as ɔu and ai. In addition to the examples in (17), southern ME continues OE monophthongs: cū ‘cow’ (ME cow uː), stōd ‘stood’ (ME stood oː), āc ‘oak’ (ME ook ɔː), Old Norse lān ‘loan’ (ME lōn ɔː). OE ō before r in ME probably shows an epenthetic schwa, as in OE mōr ‘moor’ > ME mōre, mour, maure, not usually found before consonants other than r (as in the continuation of OE mōna > mōn(e)), indicating the possibility that ME oː was diphthongal.

All in all, the vocalisations seen in (17) cater for an environment in ME where long monophthongs and diphthongs have identical (or very similar) stressed peaks. No system would be able to maintain the phonological opposition between iː vs ij, for example, for too long. In such a system, either the long monophthongal or the diphthongal pronunciation is expected to win out. In the south of England, judging by breaking, the diphthongs won out. Additionally, (traditional) co-articulatory processes (ongliding to ɫ in words like tawlk ‘talk’ and waulk ‘walk’ producing au from a) muddy the picture even more.

Minkova (2014, 265) concedes that the pure long monophthongs and the diphthongs coexist and alternate historically. Alternation is probably meant in the ‘pandialectal’ fashion of a certain phonologically ‘long’ vowel to appear as either monophthongal or diphthongal depending on the variety. This scenario produces a rather different kind of southern ME, one that that merged its long vowels with its diphthongs, see next section for a tentative suggestion that needs further research.

5.3 An alternative Middle English vowel inventory

In this scenario, the GVS applies to (some of) the diphthongs, not the long monophthongs, as these did not exist in the diphthonging dialects. In these dialects, ME ij, uw, ej and ow changed to EMoE ej, ow, ij and uw, a change which did not affect standalone ME i, u, e and o.

Mixed diphthongs (with alternate front and back halves) like aw, ɛw, iw, uj, oj were also part of the inventory. These would have balanced out the front and the back diphthongs in (18): not only was there ij (life), but also iw (rule), ɛj (dēth) and ɛw (few), etc. The short monophthongs in the system were i, e, a, o, u (cf. Lass 1980, 1992) or perhaps ɪ, ɛ, a, ɔ, ʊ (Stockwell & Minkova 1990; Murray 2012, 257f) and ə/i, which was only found unstressed. In the present analysis, southern ME may only have had one long monophthong, and even that one was ultimately diphthongised: aː (sane, naked).

This ‘abundance’ of diphthongs is not an impossibility. Ritt (2012) says that “the diphthongisations illustrate that Middle English saw an elaboration of the vowel inventory”, although his analysis is traditional in the sense that it retains both long vowels and diphthongs. The system can be likened to early 20th century standard English described by Sweet (1900, §690): it had 5 short monophthongs (if we take stressed schwa (Sweet's ɐ) and unstressed schwa to be the same phoneme), 11 (w-, j- and ə-final) diphthongs and only 3 long monophthongs (ɑː father, ɔː law and əː nurse). Of course, there were a few other (substandard) diphthongs, too, such as ew (bell), ɪw (mill), the result of l-vocalisation. The modern balance between the long monophthongs (5 or 6, depending on whether the ɵː of pure is counted) and the diphthongs (7 in a conservative count) is the result of loss of schwa-final diphthongs to long monophthongs in SSBE (Lindsey 2019). There are still 6 short monophthongs,¹³ just like in ME.

It may be objected that the ME diphthongs in (18) were minimally distinct, but it is possible for a system to maintain such differences: most modern varieties of English have a minimal phonetic difference between e and ɛ, or ɛ and a (met vs mat). In systems such as General American, usually analysed as having a tense-lax opposition, there is a stepwise contrast of i (beat) vs ɪ (bit) vs e (bait) vs ɛ (bet) vs a (bat), and u (pool) vs ʊ (pull) vs o (pole) vs ɔ (short), ɑ (Paul), all of which stand in contrast to ə (cut) (cf. Durand 2005; Backley 2011; Pöchtrager 2020, etc.).

The favouring of diphthongal pronunciations has influenced the shape of the modern English vowel system, Minkova concludes. The vocalic space is also crowded, something that was reduced in EMoE and later by the meet–meat, daze–days and no–know mergers. The crowdedness of space is also indicated by the day–way (OE weg) and know–grow mergers in the south which already affected ME (in the North grow and knaw remain distinct). ME aw (law) did not merge with ɒw (grow, know). Lass (1992, 50) reconstructs ɔw for grow/know, ai for day, ā daze. Even ā will ultimately be diphthongised, connected to its raising in the front. Ours is an experimental take on ME diphthongs: lessening the novelty of the approach by showing, for example, the diphthong in ōk as ɔ^w or o^w to imply it escaped being analysed as a diphthong because it was less of a diphthong as the one in know (shown as ow), which everyone recognises as such, would be dodging the bullet. In the same vein, claiming that oak had a short diphthong and grow a long one (or the other way around) would be resting on an unsubstantiated difference and go against uniformitarianism (cf. Lass 1980; White 2004). The consequences remain to be explored in the future. What's more, there would be no phonological proof for this behaviour.

Southern ME gives way to southern EMoE of the kind described by Hart in the 16th century. Hart was not a southerner by descent, but he held a post in the Chester Herald in their London office indicating that he may not have described provincial/non-southern features (Jespersen 1907). Danielsson (1963, 271) thinks Hart is a Londoner, although there is much uncertainty about his background. Given the above discussion, it does not seem far-fetched to say that the diphthongal bias with which we approached Hart's pronunciation may not be far from the truth. Jespersen seems to have been right in his claim that there never was a merging of the mid vowels in favour of the long monophthongs (ɛj, eː > eː, ow, oː > oː) only to be re-diphthongised in the 19th century (eː > ej, oː > ow).

Mid-vowel diphthongisation is an unnecessary construct. What we have always had is variation on a common theme: diphthong vs monophthong. What we have on our hands is various accents choosing either the monophthongal or the diphthongal pronunciation over the other. If an accent chooses to be monophthongal for the narrow diphthongs, adopting a diphthongal pronunciation with a ‘loanword’ from a diphthongising accent is not possible. If an accent, like the predecessor of SSBE, which developed in EMoE and early MoE times both diphthongs and long monophthongs through a long chain of development (involving r-loss, compensatory lengthening, etc.), chooses to adopt a loan from a monophthonging accent, this loan may be lexicalised with a long monophthong (but have a spelling that invokes the ‘wrong’, diphthongal pronunciation). This may have happened to the unique example of broad (and its derivatives), having oː, rather than ow (> əw) in SSBE: **brəwd (< OE brād, cf. German breit, cf. Wright 1905 for a list of dialectal pronunciations). If we are right, brəwd was lost, or rather supplanted by its monophthongal cognate coming from a monophthonging variety. A now lost groat ‘a medieval coin’ with oː (probably from Dutch groot oː ‘big’) was like broad, it seems, cf. Jespersen (1961, 315) for a different interpretation. Its pronunciation was analogically reinterpreted on the basis of spelling (cf. Michaelis & Jones 1913, v–vi, for examples). Brooch and broach (both brəwtʃ) all come from OFr broche with open syllable lengthening in ME (ɔw), with broach showing the expected spelling in MoE, and brooch showing the spelling of oː in post-ME times in an accent that was monophthonging.

We may conclude that SSBE əw in stone continues ME ɔw (with EMoE and later interludes) and is not a ‘prestige’ innovation of the predecessor of SSBE (apart from the quality change affecting the first element), while (monophthonging) northern English stoːn continues ME stɔːn. If diphthongs have always been part of the diphthonging accents of English, the conclusion is that breaking in jr/wr-clusters must also have been an ongoing process starting (at least) in ME in those varieties where long vowels merged with diphthongs.¹⁴