Kökeritz Remodeled: The Problem and Promise of "Original Pronunciation"

When I began this post, it was as a simple diagnosis of what was wrong with David Crystal's ill-named "original pronunciation". It ended up taking the form of a series of loosely connected ruminations on this or that aspect of the reconstruction. At some point I intend to recombine it all into a more structured post. But I have many things to get done and only so much time.  

Talking about "Shakespearean pronunciation" is complicated by two factors:

(1) the linguistic situation of Shakespeare's time
(2) the linguistic attitudes of our time.

As David Crystal reminds us on his website, Shakespeare's London was in great flux, linguistically as much as anything else. To illustrate the problems of reconstructing "Elizabethan English" pronunciation, here is a brief and rough and stupifyingly simplistic history of the pronunciation of the words PALE, PAIL, DEAL, PEEL in London English over the course of two and a half centuries. For the benefit non-linguists, I've linked all my IPA transcriptions to audio files of me articulating the sounds. This chronology (mostly) takes after  Roger Lass in the Cambridge History of the English Language, Jeremy Smith's Sound Change and the History of English, and Dick Leith's Social History of English. There are uncertainties about what happened when, but I've tried to be vague enough to accommodate them.

14th century:
These words are pronounced /pa:l pail dɛ:l pe:l/

15th century:
Shakespeare's grandfather is born
PEEL has become /pi:l/
PALE is being, or has already been, raised to /pæ:l/ in innovative speech.

16th century:
Shakespeare is born (1564).

There is good evidence for at least three (actually probably four) different sociolects in the city at this point. They can very tentatively, generally and probably over-simplistically be identified with particular social groups.
— Sociolect 1 has merged PALE and PAIL into /pɛ:l/ while raising DEAL to /e:l/.
— Sociolect 2, has merged PAIL and DEAL into /ɛ:l/ while keeping PALE distinct as /pæ:l/.
— Sociolect 3 keeps them (mostly) distinct with /pæ:l pæil dɛ:l~de:l pi:l/ for PALE, PAIL, DEAL, PEEL.
17th century:
Shakespeare dies (1616). The three Englishes of London continue their developments in broadly similar but distinct directions. By 1640 at the latest:
— Sociolect 1 is on its way to raising DEAL to /di:l/, merging it with PEEL, and raising  PALE/PAIL to /pe:l/.
— Sociolect 2 has merged and raised PAIL, DEAL and PALE as /e:l/.
— Sociolect 3 has merged PAIL/PALE raised it to /pɛ:l/, while at the same time raising DEAL up to /de:l/, and keeping them both distinct from PEEL /pi:l/.
The times they were a-changing and the language did too. Between Shakespeare's birth and his death, just looking at these four vowels, the sounds of English rearranged themselves in different ways in three different accents all available to him in a single city. And not at a uniform rate, either. Cranky orthoepists give evidence that Sociolect 3 still has conservative stragglers resisting the PAIL/PALE merger for a century. What was Shakespeare's pronunciation? And when during his own lifetime shall we place "his pronunciation?"

Sociolect 3 seems to hold the day as the high English literary norm through most of the rest of the century, but remains in competition with Sociolect 2, and by the 18th century both have begun to give way to Sociolect 1. For a while the DEAL vowel retains two alternate pronunciations. The merged forms of Sociolect 2 and the residual highfalutin yet low-vowelling pressure of Sociolect 3 allows the /e:/ of DEAL to hang on in a few words (great, steak, break etc.) long enough to merge with the PAIL/PALE vowel instead of the PEEL vowel. (A few Englishes, such as the older West Country Newfoundland accent, have this in many other words like sea, beak and leak.)

There is thus no such thing as a single "Elizabethan" or "Shakespearean" pronunciation. The London of Shakespeare's day, no less than the London of our own day, was one of great linguistic variety. Pronouncing texts like a late 16th century Londoner requires asking first: what kind of Londoner exactly, and how far can you push the evidence?

Moreover, pronunciation used in performance or recitation is not necessarily how one speaks at home. (Eminem's normal speech does not display the pen/pin merger, but the rhymes and pronunciation he uses when he raps very much do.) Like other Renaissance English poets, Shakespeare could and did maximize his options for rhymes, as well as their expressive potential, by drawing on the different varieties of English available to him.

Sound-changes do not happen at a uniform rate among all speakers in all lexical items in which they are possible. Often (though not always) they happen in some words earlier than others. Which words those are may differ from speaker to speaker. (Click here for an account of precisely this phenomenon in my own pronunciation of English.) This has important implications for the use of rhyme to determine the chronology of mergers and vowel shifts: just because a vowel has shifted in a given word, it does not necessarily follow that all the other words in the same lexical class have shifted too. A word undergoing a shift will for a time retain two alternate pronunciations, and sometimes if the sound-change is interrupted it can end up shifting "back" and settling on its older form. If a poet only occasionally interrhymes two formerly distinct vowels, even assuming that rhyme can be taken to imply complete identity of syllable nuclei, this by itself cannot be taken to imply that the vowels have completely merged in the poet's dialect. If the interrhyming disproportionately occurs with the same small set of lexical items then one is doubly unjustified in suggesting that a full merger has occurred.

Ultimately, even though Early Modern London English is better documented than a lot of other historical Englishes, we still lack the level of evidence necessary to reproduce that level of granular detail from four and a half centuries ago. Any phonological reconstruction of a language in the middle of massive sound-change must content itself with approximate generalities.

That is the problem of Shakespeare's time.

***A Modern Myth***

The problem of our own time is a fungus of ideologies concerning Shakespeare's English and its relation to Modern English.

Many Anglophones like to feel that Shakespeare's English is "Our Language." It is part of many literate English-speaker's self-conception. This is why they so commonly overestimate how well they actually understand Shakespeare's language, and also why many who happily forgo the King James Bible for a more modern English Bible have continued to express horror and incredulity at the sacrilegious idea of literary translations of Shakespeare into modern English.

Numerous scholars for over a century have attempted to reconstruct "Shakespeare's" phonology. Most are less interesting for what they claim to show about Elizabethan England than what they do reveal about the scholars' own language ideologies. Few would attempt such a project without a certain investment in the history of English, or of English literature. Shakespeare must belong to English, and modern ideologies concerning "our" relationship to him as ModE speakers lie waiting in the wings. At least since Kökeritz, such projects have tended to sustain a belief that "Shakespeare's pronunciation differed considerably from Chaucer's, but differed only in small ways from ours." This historicization of ModE, which claims for it a relatable past and even a kind of determined canon, is a siren-song to which even great scholars are not immune.

Much discussion and description of what is now labeled OP has come from parties with axes to grind: whether it be the 19th century romantic fetishization of Shakespeare's English as exaltedly different, Kökeritz' vested aesthetic interest in a maximally "modern" Shakespearean English, or Crystal's pious faith in an OP that sounds equally relatable to all modern Englishes and which is also "no more difficult for an audience to understand than any modern regional accent." That is a rather innocuous way of putting it. I don't know whether the issue is audiences' sensibilities, the limits of the risk directors are willing to take, or a combination. The features apparently most objected to by directors, and often banished from Crystal's OP, are features strongly associated with parts of the British Isles outside England. Excuse seems to be found, when possible, to remove anything likely to sound too distinctively or overridingly Welsh, Scottish or Irish. Thus the insistence on unrounded STRUT vowels and the sidelining of a coronal trill for /r/. That both of these not only existed but were typical of late 16th century London speech at this time is a virtual certainty. Likewise a fullblown NURSE merger (which is in fact not attested in London speech before the 18th century) is anachronistically thrown in based on an argument that I have a hard time actually accepting that Crystal himself even believes.

Crystal gives an overview of the history of earlier work on the phonology of early modern English and "OP studies". I must say that although he has clearly read (or at least is capable of referencing) many of the major pertinent works of scholarship (Cercignani, Dobson, Kökeritz, Ellis etc.) he maintains positions that are mystifying in light of that scholarship, and are often just not at all supported by any of it. Worse still, some of his characterizations of that scholarship should strike anyone familiar with it as rather bizarre. For example, would he seriously have the reader believe that "the fundamental distinction between phonetics and phonology" is a "theoretical perspective...not strongly present in the work of Dobson"? I struggle to conceive of a meaning of "not strongly present" that could make this assertion even close to true. As someone who has read both volumes of Dobson's English Pronunciation: 1500-1700 cover to cover multiple times, as well as his edition of Robinson's phonetic transcriptions, and several of his individual articles, I can only say this is strikes me as a ridiculous claim to make in print.  There is plenty to take Dobson to task for, but this isn't one of them. Awareness of things like complementary distribution, surface realizations, and the distinction between a phoneme and a phone, is fundamental to Dobson's work (as it would be shocking for it not to be, in a work published mid-century). The most charitable explanation I can imagine is that Crystal read the first volume of Dobson's work, but only glanced at the second volume which actually presents a coherent discussion of phonology and phonetics rather, after the first volume had marshalled the sources on which that discussion was based.

And how can Crystal rightly describe Cercignani's work as "the latest and fullest attempt to review all the evidence of rhymes, puns, spellings, and metrics in the Shakespeare corpus" while completely failing to understand so many of Cercignani's points, and continuing to reproduce the errors he chided Kökeritz for? Some of Crystal's assertions about what is the "general consensus" of scholars are flat-out false or grossly misleading too. Crystal is quite capable of citing the secondary literature, but I am as yet without opinion as to how much of it he has actually read with any care. For a book published on this subject in 2016 to not take any account at all (not so much as a bibliographic notice) of the work of people like Gorlach and Lass on the subject is rather shocking. 

Finally, it's a pity that — once again — any and all recent work by researchers who don't write in English has been completely left out of account. Which is a shame since a lot of it is much better than anything Crystal has ever written on the subject. In fact, if you can read German and want to learn about early modern English pronunciation as a performer, you're better off completely ignoring anything Crystal ever wrote and just starting with Klaus Miehling's excellent Handbuch der frühneuenglischen Aussprache für Musiker (1500-1800) in two volumes before moving on to more specialized literature. Miehling's work has serious issues to be sure, and in its own way stands outside the mainstream on certain points (not always wrongly though, IMO). But it is a damn sight better at offering you what Crystal purports to be offering.

I have been unable to find a review of Crystal's dictionary by any scholar who actually is familiar with the evidence from the period. I suspect that a large part of the reason is that most actual historians of English pronunciation who work on the relevant material don't care at all about performances in reconstruction, and of those who do most don't care all that much. But some of it, I know from personal communication with a few who privately admit they find Crystal's reconstruction embarrassingly bad, is because many just see no reason to pick a fight like that and want to look on the positive side. "OP" isn't just a dubious reconstruction, but the cornerstone of a performance movement with considerable following. 

Fuck that. I'll pick the fight if only in a blog entry. Crystal's reconstruction and the explanatory material in his dictionary is just shot through with bad philology. Worse, it is presented in such a way that it can easily give the appearance of being well-argued and well-supported even to non-specialist linguists who haven't looked at the material themselves. It has the ability, in other words, to convince the naive reader that Crystal has done his homework and can be trusted — that no second-guessing is necessary. But it is. 


je: knɔu e:k ðat ɪn fɔrm ɔf spe:tʃ ɪs tʃɑundʒǝ
— Geoffrey Chaucer

Sing—and singing—remember 
Your song dies and changes 
—Carl Sandburg "Languages"

The aim of a scholar who sets about investigating Shakespeare's works from a phonological point of view can hardly be conceived as an attempt to reconstruct Shakespeare's own pronunciation. That this is now lost beyond recovery no one can ever hope to deny, for Shakespeare has left us neither his own views on spelling reform nor a treatise on pronunciation— let alone a detailed account of his own type of speech. Indeed, we do not even know how Shakespeare pronounced his own surname.
— Fausto Cercignani "Shakespeare's Works and Elizabethan Pronunciation"

A half truth is worse than a whole lie
אַ האַלבער אמת איז ערגער פון אַ גאַנצן ליגן
—Yiddish Proverb

***All Dumb and Glum, and Doesn't Look Like Luck***

David Crystal, reputed expert on the history of English, has defended his use of unrounded vowels in words like "cut" and even "doom" in OP with the point that rounded vowels "pushed the accent too much towards Irish, and – as a general principle – I find directors don’t want characters to associate too strongly with any one modern accent." Ultimately, as Crystal says, "the beauty of OP...is that it contains echoes of many modern accents but can be identified with none of them."

God forbid something so English as Shakespeare's plays be allowed to sound too Irish. 

The desire to preserve this "beauty of OP," an aesthetic of linguistic equidistance which validates the conception of Shakespeare's English as our "Mother Tongue", is in unresolvable tension with the professed aim of trying "to get as close as possible to the sound system that Shakespeare himself would have heard and used." In Crystal's case, one could be forgiven for thinking the quest for that beauty has led him to posit unrounded vowel (transcribed as /ɤ/ in Crystal's dictionary, but in practice rendered as /ʌ/ in OP performances) for Shakespeare's time not only in words like flood, cup, but even in words like doom.

Here is Crystal's section on the proposed /ɤ/ vowel, for which he gives keywords CUP, STUFF, DRUM (but other words such as DOOM, MOVE):
The quality is further back and closer than the equivalent vowel in RP, /ʌ/. Opinions vary as to how far back it would have been, with values proposed between [ə] and unrounded cardinal 7 [ɤ]. In my view, the latter is more likely, hence the choice of this symbol in the transcription. A u spelling is the norm for this vowel, and there are several instances where there is overlap with o, suggesting the back quality, as in sodaine / sudden, sommer / summer, Sonday / Sunday, dombe / dumbe, tombles / tumble. The emendation at Ham 3.3.18 of somnet to summit also reflects this quality. Contemporary writers reinforce this view, as in the quotation from Jonson (p. xx), where the o of love and prove is said to be ‘akin to u’, which in turn he describes as ‘thick and flat’ in such words as usAs both o and u were routinely used for rounded vowels, the question arises as to whether the vowel in these words was rounded, as in many parts of northern England today. The evidence is unclear: in the same section, Jonson describes o as being pronounced ‘with a round mouth’, but immediately adds that this ‘is a letter of much change, and uncertainty with us’. The spelling of slumber as slomber by Macmorris (H5 3.2.111) suggests a rounding that would be absent from the non-Scots form. And there are rhymes with unrounded front vowels that are also suggestive, such as shudder / adder, Sunday / array, us / guess, punish / languish. My view is that both unrounded and rounded variants were in use at the time (as they are today), but opting for the unrounded form as the default in this dictionary allows actors the choice of using the rounded variant if they want to differentiate a character. Certainly, if they were to replace all [ɤ] by [ʊ], it would result in an OP of a noticeably different auditory character (much closer to, say Yorkshire or Irish English in effect), as this vowel is very common, being used in some frequently occurring in words (must, us, under, the un- and sub- prefixes, etc.). On the other hand, they do not have to adopt such a noticeable lip-rounding as we hear in present-day regional accents, and I would not correct a slight degree of rounding, when working with a company.
There are a few problems here. The first is the reliance on Jonson. For reasons I cannot summon the strength of mind to fathom, Crystal accords a frankly weird prominence to Jonson's grammar in his dictionary. But the passages on orthography in Jonson's grammar are little more than plagiarisms and abridgements of other people's work (sometimes drafting off of sources that weren't even describing English to begin with.) In this case, he is piggybacking off of the Latin grammarian tradition, which in turn tended to conflate spelling and orthography. Terminology like "thick and flat" recycles (in translation) the Latin grammarians use of the words "pinguis" and "latus" and is useless as a characterization. Jonson uses these terms in English even more imprecisely, and less helpfully, than ancient grammarians did in Latin.

The variation between <o> and <u> in spelling is made far too much of here. While it is true that a "u spelling is the norm for this vowel" there is a range of words (off the top of my head: come, some, son, done, won and the adjectival ending -some) where the o spelling was and still is normal. Sun and son are two different words distinguished only by orthography, which were every bit as homophonous in Hamlet's pun then as they are today. The fact that <o> is often used in words where <u> was more customary really tells one nothing of great importance beyond the fact that the spelling was unstable in such words, (and actually for well known reasons having more to do with orthography, and absolutely nothing to do with pronunciation). Incidentally, <slomber> for <slumber> in Henry V occurs in Jamy's dialogue, not Macmorris'. But that doesn't hardly matter. What does matter is that <slomber>, though not the common spelling, is not at all an outlandish spelling variant. This spelling of this precise word also occurs in one of Queen Elizabeth's letters to James' I, in the Book of Common Prayer (1559), in Thomas Lodge's Rosalynde, multiple times in Spencer and in quite a few other places too numerous to bother looking up individually, often when the author has a fondness for old-fashioned orthography, unsurprisingly as this is an extremely common Middle English spelling. It's absolutely whacky that Crystal claims this as evidence against a rounded vowel in Southern English. Like, dude, seriously? 

In H5 <slomber> (like < theise> for these in the same passage) is not necessarily indicative of anything peculiar about Jamy's (or anybody's) pronunciation of that word. It looks rather like a case of eye-dialect, the use of a distinctive spelling not so as to suggest a particular, let alone peculiar, pronunciation, but to convey the writer's attitude toward the speaker, or indicate the "kind of person" a speaker is. <Gonna> and <wanna> are often used this way to represent the speech of uneducated persons in modern written English, even though the pronunciations they imply are common among all kinds of English-speakers in all but the most carefully enunciated, heavily monitored speech. Martyn F. Wakelin, in his anthology of Southwestern English dialect texts from the 16th-20th centuries, finds pervasive instances of this use of eye-dialect (to add a purely visual flavor even in texts genuinely meant to represent non-standard or regional language), which he terms "empty forms":
The writers use large numbers of what I am venturing to call 'empty spellings' or 'empty forms', forms such as iz 'is', lite 'light'....which add nothing in the slightest degree 'regional' to the  phonology of the text, but are presumably intended to heighten the effect of earthiness or rusticity (in early texts, however, it is sometimes difficult to distinguish them from merely older spellings).
<slomber>, if not such an "empty form", may indeed be a "merely older spelling." It is beyond unacceptable to dragoon this spelling into an implication that the Scots vowel in slumber sounded in any way peculiar to southerners, let alone in this specific way which actually mirrors a highly salient modern sociolinguistic variable and dialect stereotype.

Modern stereotypes are very much to the point, here. Crystal suggests the use of the rounded variant to "differentiate a character." The implication is that the rounded vowel might be treated as a sort of shibboleth in "OP" in a way that just so happens to mirror the attitudes and experiences of modern Southern Standard English speakers who respond to (and often make fun of) this vowel as a marked regionalism. We have hard evidence that educated men in London perceived words of this type to have a rounded vowel readily relatable to the rounded vowels of other European languages. John Florio's Italian-English dictionary of 1611, for example, equates the vowel of Dug, Stun with that of Italian amore. Even granting the plausibility of the statement that "both unrounded and rounded variants were in use at the time" there is zero reason to assume any kind of identity between their distribution of use "at the time" and how and where they are used today.

I'm quite not sure how array/sunday is at all relevant to the point, since this rhyme does not concern back vowels or rounded vowels of any kind.

In context in Love's Labor Lost, I'm not convinced that guess/us/thus is meant as a rhyme. It seems more likely that the rhyme is meant to be us/thus with the line-ending guess as a singleton breaking up a long series of couplets. And I genuinely doubt punish/languish in Pericles is actually intended as a rhyme at all, given the context. There's no arguing that shudder/adder is intended as a rhyme, though. And there are a handful more such forms. But they are rare, and no more common in Shakespeare than rhymes of the type hither/father which probably cannot be taken to imply anything about either vowel involved (pace Cercignani's suggestion that this rhyme might rely on variants with /ĕ/ in both words).

One must consider other rhyme evidence: love and prove were fuller rhymes then than they are today. Jonson indeed appears to describe these vowels as identical. But these words also rhyme with things like Strove, Jove for which Crystal satisfies the fetish for perfect rhyme by simply tossing in /ɤ/ as a variant pronunciation. Now the word Lover can rhyme with Over in this period, and none other than Jonson on the same page clearly implies that the vowels in these words was not the same. But /lʊvǝr/ and /o:vǝr/ make for a far closer rhyme than if one has /lɤvǝr/ for the former.

Given Shakespeare's habit of rhyming things like dull/pull and flood/food quite readily (even as he generally avoids crossrhyming the MOOD vowel with the reflexes of Middle English /ɛu ɪu/ in words like viewed, lewd, rude), if one takes Crystal's point about the importance of full-rhymes and the relative rarity of imperfect rhymes at face value, reconstructing a STRUT vowel extended to BLOOD and FLOOD is unjustified, and the tiny handful of mostly-debatable cases Crystal mentions here is dwarfed by the mass of rhymes of the dull/pull, flood/food and bud/understood types.

Moreover, even if, for the sake of argument, one allows for an unrounded STRUT vowel in the English underlying Ben Jonson's grammar, this in no way implies that it was commonplace in "Shakespeare's London." Jonson's grammar was published in 1640, two and a half decades after Shakespeare's death, and likely not written long before Jonson's death in 1637. By the 1640s-50s, all investigators from Eric Dobson to Roger Lass are agreed that lowered (and, in the view of most, unrounded) STRUT vowels are attested. But if one looks to the 1590s, evidence for unrounded STRUT vowels of any kind is rare and marginal (in Bellot's transcriptions, for example, there's no way to be sure whether the vowel in question is unrounded or merely lowered.) Consider the changes that took place within, say, Received Pronunciation from 1960 to ca. 1985 (i.e. within a single generation).

I suspect that Crystal gives us the transcription /ɤ/ and makes a point about the vowel being notably higher than that of modern cup, in part because <ɤ> is the IPA symbol for the unrounded counterpart to /o/ and (and traditionally this symbol is used in broad transcription to represent even higher vowels like [ɯ̽], which would pair nicely with /ʊ/.) This allows rhymes like dull/pull to be treated as rhymes that are only "differentiated by a single distinctive feature."

There really is little solid evidence for such an unrounded (as opposed to merely lowered) vowel in, say, STRUT existing in Southern English before Shakespeare's death (in fact, there is even less than Dobson supposed). What evidence we have suggests that, before about 1620, this was a marginalized realization. Thus a more "authentic" reconstruction might give the character of Henry V a rounded STRUT vowel which he consciously unrounds when going under the guise of "Harry Le Roy."

But Crystal has no other option, really, if he wants to both preserve rhymes like doom/come and avoid sounding "too Irish" for our delicate Anglo ears.

As Crystal wrote in 2005:
The thing about OP which makes it different from other non-RP performances – and the most fascinating thing about it – is the way it occupies a unique dialect space, resonating with several modern accents and yet at a distance from all of them.
All well and good, but there seems to be some motivated thinking (and no small amount of language-mysticism) going on here. Performing a reconstruction has the effect of distancing the text from Modern English, and it follows for artistic, and also ideological, reasons that that distance must be managed to productive ends. In practice, what actually seems to happen with many OP performances is a weird back-projection of modern notions of standard or correct language into the world of the Renaissance, a historicization of Modern Standard English that claims for it a relatable past and a determined canon of sorts.
In fact, one of the most noticeable features of the talkback sessions after the OP performances was the way people associated EME pronunciation with an accent they knew. Everyone felt at home with it, but for different reasons. The conclusion is obvious: no modern accent is identical with EME. All share some features, for the simple reason that we are talking about an accent (more precisely, a group of accents) which is the ancestor of the accents we hear in English today. And not just British English, but English all over the world. Captain John Smith and his settlers would arrive in Virginia in 1606, the year (we believe) that Shakespeare was writing Macbeth.
"All share some features, for the simple reason that we are talking about an accent (more precisely, a group of accents) which is the ancestor of the accents we hear in English today."

Dude, seriously?

Australian English is about two centuries younger as a phonological entity than Shakespeare's corpse. Neither Captain John Smith nor Sir John Rolfe was from London. No single chronological or regional variety of British English is directly ancestral to any of the many world Englishes. London continued to exert influence in port towns in America after the Revolution. This, combined with the effects of dialect leveling, make for a complicated history which I know Crystal is perfectly well aware of, but which sadly is not as sexy as the idea of Shakespeare's English giving birth to ours. (If anything, the ancestor of American English is not 16th century London English, but early 18th century London English.) I can't take seriously the term "British English" used to imply that any accent of Renaissance London English (let alone one with an unrounded CUT vowel) is the direct ancestor of Paul McCartney's Liverpool accent. I find this kind of mythmaking horribly irritating from anyone, but from someone with Crystal's standing it really frays my last nerve. I think linguists should be in the business of giving the general public factual information about language, not participating in the creation of something more like mythology.

***Arr Matey***

Speaking of delicate Anglo ears, it is also no accident that most OP productions don't use a trilled or tapped /r/ in initial or prevocalic position even though the sources strongly suggest that this was the case in at least some varieties of 16th century London English.

Crystal's discussion of Early Modern /r/ in his dictionary is unacceptably misleading and in some ways quite careless. More than any other part of his presentation of "OP" phonology, his discussion of /r/ not only makes it impossible to take seriously his professed aim to "get as close as possible to the sound system that Shakespeare himself would have heard" but also makes me wonder how carefully he read his secondary sources.

Quoth Crystal:
The exact phonetic quality of this sound is unclear. The descriptions of the sound by contemporary writers leave it open just how r is articulated, such as how far back the tongue curls. When used in front of a vowel, it would seem to have been the same sound as in RP, a postalveolar frictionless continuant, though there must have been a trilled variant (as today in some accents of Scotland and Wales), for Jonson describes r as a sound that ‘hurreth [vibrates] . . . with a trembling about the teeth’. But he then draws a contrast between r before and after vowels: ‘It is sounded firm in the beginning of the words, and more liquid in the middle and ends; as in rarer, riper.’ I interpret this to mean a continuant r, as in the West Country of England and much of America, but there is no way of knowing whether the focus of the articulation is post-alveolar or retroflex, so a great deal of variation will be heard in present-day OP productions.
On the contrary, the descriptions of /r/ from this period are hardly "unclear". A trilled [r] for Southern English is witnessed in the 17th century by several quite clear articulatory descriptions of it, including John Wallis (1653), John Wilkins (1668), William Holder (1669), Christopher Cooper (1685). Even Sir Isaac Newton (yes, that Sir Isaac Newton of gravity-discovering fame) in his phonetic notes (datable to the 1660s) describes the English /r/ as characterized by "the quavering or jarring of the toungs end against the fore parte of palate."

And Newton's description isn't even as good as it gets. Other 17th century phoneticians give far more detailed articulatory descriptions that afford no ambiguity. William Holder's Elements of Speech 1669) is a book primarily concerned with teaching deaf mutes to pronounce English, and in describing English /r/ gives a detailed and quite perceptive characterization of one defining feature of trills: directing airflow over an articulator so that it vibrates. Holder states that R is made
"...by a Pervious [=non-occlusive] Appulse [=obstruction] of the end of the Tongue, with its edge to the Goums, The Tongue being held in that posture, onely by the force of the … Muscles, and not resting any where upon the Teeth; except onely touching them loosely, so as to close the passage of Breath every where by the sides, and conduct it to the end of the Tongue. And this with a strong Impulse of Breath vocalized, so as to cause a trembling and vibration of the whole Tongue; which vibration being slow, does not tune the voice, but make it jarred; the Tongue not resting but […] agitated by strong impulse of Breath"
Crystal's purported perplexity as to "how far back the tongue curls" is — I think — the result of Dobson's misinterpretation of Wallis. Dobson was eager to justify his sense that Early Modern English /r/ was basically identical to that of conservative RP as he himself spoke it. To that end he seized on Wallis' mention of "tongue-curling" while also assuming ineptitude on the part of every single one of the 17th century sources (including Wallis) describing an apical trill. The whole matter would be laborious to discuss in detail, but suffice it to say that Wallis was describing the tongue curling back from the alveo-dental (where his D and N are articulated) to the alveolar region, or from the alveolar to the post-alveolar region. Both would characterize a trill, either alveolar (as in Italian) or postalveolar (as in Russian).

I'm glad Crystal at least mentions some evidence for a trilled initial /r/, and allows for the fact that there was indeed a trilled variant. Though what he chooses to mention is perplexing. Jonson's trill-description in particular has been shown to have little value on its own as evidence. As was first pointed out by Herford and Simpson (1952) in their edition of Ben Jonson's works, and as was re-stated both by Dobson and then by Cercignani, Jonson's entire description is actually plagiarized nearly verbatim from the French grammarian Petrus Ramus. Crystal has clearly read (or presents himself as having read) both Dobson and Cercignani, so it's strange that he ignores their point here. Admittedly, he is far from the only scholar to seemingly ignore the implications of Jonson's demonstrated plagiarism. The secondary literature is rife with specialists making unqualified use of Jonson's statement to support their theories about coda R-weakening. Of course, even granting Jonson's plagiarism, it does not necessarily follow that his description is worthless, if used to support conclusions inferred from other evidence. It is not impossible that Jonson based his description on Ramus because Ramus' characterization of /r/ could, in a rough sense, be used to describe English as he knew it. But that point should be argued.

Consistent with his aesthetic of an Early Modern-Enough English, Crystal domesticates the trill as an implicitly negligible "variant" and says without evidence that "when used in front of a vowel, it would seem to have the same sound as in RP." I can only infer that this claim (which is not even supported by Dobson) is meant to follow from Cercignani's overinterpretation of Robert Robinson's phonetic writings (1617). It does not follow, as Cercignani claimed, that a rhotic formed on "the upper gummes or outmost part of the roofe of the mouth" may be taken for a "post-alveolar frictionless continuant." This is in fact one place of articulation where trilling commonly occurs. Cercignani asserts, and offers no evidence, that an approximant is more likely. Robinson himself really offers no evidence either for or against a trill, and the only reason to entertain the idea of his English as possibly not containing a trill is that he gives no clear description of the tongue vibrating or trembling, whereas other sources overwhelmingly do.

To be fair to Crystal, one may commonly find assertions in 20th century literature to the effect that "the early occurrence of a frictionless or fricative r is indicated in final, preconsonantal, and even intervocalic position by Middle English developments such as ěr>ăr, and ẹ̄r>ę̄r" (this particular quote is from Cercignani p. 355). Ever since it was first made, the appeal of this argument in light of the subsequent history of English appears to have evaded a well-deserved death. Such retrospective explanations can have profound psychological effects, like losing one's virginity or the sight of Sean Connery's speedo in Zardoz: things were never quite the same afterward. Large theoretical edifices have been built on what this effect of ME /r/ implies about its articulation. But it is quite non-factual. The strict articulatory constraints for trills favor a lowered predorsum and mediodorsum. This can have a strong effect on preceding high vowels, lowering those that generally require raising of the tongue body. Cross-linguistically, such lowering can be readily observed before trills as much as before any other type of rhotic. In Central Swedish it occurs not only before retroflex /rC/, but also before an intervocalic trill and before trilled final [r:]. E.g. ära: /ɛːra/ → [æːra], ärt: /ert/ → [æʈː], öra /øːra/ → [œːra], dörr: /dœr/ → [dœ̞rː]. In most varieties of Dutch, vowel quantity and quality may be affected by following /r/ regardless of the realization of the rhotic

Virtually every available source for the 16th and 17th centuries with anything to say about English /r/ either implies or explicitly describes an apical trill (or at least non-approximant) for /r/ as normal. The few apparent exceptions are either hopelessly ambiguous (Robinson) or explicitly report trill-less speech as aberrant or a speech defect (Holder, Wilkens). Our best 18th century sources attest to continued trilled realizations among a minority of high-status London speakers. The alleged "difficulty" (Crystal 2016) of determining the nature of EModE /r/ is in part an artifact of scholarly unwillingness to believe what the actual sources say about the English known to them. It's also worth noting the complete absence of any mention of apical trills as a regionalism in this period. Many 16th and 17th century sources are eager to document and disparage various pronunciations as Scotticisms and Hibernisms, but Englishmen give no indication that Scottish, Irish, Welsh or Northern pronunciations of /r/ were at all peculiar to or distinctive before the 18th century. Even then the first reference by Defoe (1720) to an oddly pronounced Northern /r/ is to to uvular [ʀ]. From around the mid 1700s we have increasingly plentiful written testimony from Southern Englishmen as to the peculiar harshness of coda /r/ as pronounced by Irishmen, Welshmen, Scotsmen and Northern English speakers.

The whole matter of Early Modern English /r/ is in desperate need of an updated and detailed examination in light of what is now known about the behavior of rhotics. Which is why I am working on an article on it. I'll link to it here when it's published. I'll say this right now though:

Recent phonetic and sociophonetic research involving rhotics suggests that the various issues involving Early Modern London /r/ are readily explainable. Having an apical trill in your language basically lends itself to considerable instability in how your /r/ is realized in any given instance. Trills, especially when they do not contrast with another type of R-sound, tend to readily "devolve" into less complex, less energetically costly allophones. They thus provide a constant articulatory breeding ground for the creation of lenition forms whose co-variation among speakers is often conditioned as much socially as it is phonologically. This is particularly true of urban areas. Recent sociophonetic studies of /r/ variability for urban dialects of languages like Persian, Dutch, Arabic, Spanish, Polish all point in the same direction. If we assume that London English /r/ ca. 1600, had a trill as its highest strength grade, it would almost by definition have been prone to complex social and idiolectal variation in the distribution of lenition forms. Most speakers in Shakespeare's London probably had a trill in their speech, but how frequently they actually realized /r/ as a trill, how frequently they produced a tap, and how frequently (and in what social or phonological contexts) they used fricative or approximant lenition forms, will have varied considerably among speakers. In any case, the eventual loss of [r] as the highest strength grade of /r/ (which basically happens in the latter half of the 18th century) would have favored stabilization of its realizations.

Alright, enough about my research.

Having dropped Jonson's trill on the reader as a domesticable "variant", and having given no mention of the more worthy evidence, Crystal ignores the problem Cercignani and Dobson raise with Jonson as a witness, and interprets Jonson's "liquid" as referring to "a continuant r" occurring "before a vowel." The only question for Crystal, it seems, is whether it was "a post-alveolar or a retroflex approximant" i.e. more British or more American. There is zero direct support for this, and only weak circumstantial support. As with /r/ so with all else, the desire to preserve the "beauty of OP", relying on an aesthetic of linguistic equidistance which validates the conception of Shakespeare's English as our "Mother Tongue" is at daggers drawn with Crystal's professed aim "to get as close as possible to the sound system that Shakespeare himself would have heard."

Almost all modern rhotic dialects of English have an alveolar or rhetroflex approximant of some kind in words like round. So too do modern OP productions. The use of a trill or tap would inevitably sound too distractingly regional (and probably quite a few actors would have trouble with it.) It might even remind (Modern Southern English) speakers of the use of trills as a posh affectation, a fashion which until recently could commonly be heard on the English stage.

The fact that a trill was clearly not so regionally marked in Shakespeare's day turns out in practice to be of secondary importance in OP productions, where the general practice is to use whatever /r/ is present in the actors' native accents.


Crystal, to be fair, is not dogmatic. He has introduced various levels of variation in the OP performances he has helped stage. He understands that OP is a modern performance tool, not  a museum piece. This is probably why he has succeeded where many other OP experiments failed. But, again, there is a sleight of hand that I find more than irritating. I get the claim of "Shakespeare in his own authentic pronunciation" as a marketing tactic. At what point, though, does it become false advertising?

When I watch an OP performance, knowing what I know about the sources and issues involved, I can't help feeling like I'm being being asked to bask in the ostensible "authenticity" of it all, and at the same time being being subtly lied to. Like a man loudly bragging about his modesty, or a pathological liar rhapsodizing about how important it is to tell the truth, it just leaves an unsavory taste on my brain.

Look, I loved Mel Gibson in Braveheart, but I'd get real pissed if people advertized the movie as a documentary about Scottish history.

***Making Shakespeare Grate Again***

Looking in Crystal's dictionary, another thing that raised my eyebrows was that great and grate are transcribed as homophones (/grɛ:t/) whereas sate and seat are not. This is odd, especially since none other than seat and great rhymes in Shakespeare and one would have thought the desire to restore the euphony in Shakespeare's rhymes would've motivated at least a listing of a higher vowel as an alternate pronunciation of great, but the only rhyme-motivated alternate given for great is /gret/.

Here again it is hard to avoid the suspicion that modern standard forms of English are having a weirdly determinative effect on what Crystal would have OP sound like. As I have just described above, the real trajectory of the vowel in great (i.e. the DEAL vowel) was rather complicated. As late as the early 18th century, there were still speakers who pronounced great as if it were greet.  Transcribing great as having /ɛ:/ and thus merged anachronistically with grate (he also does this with brake/break) is to let modern standard pronunciation dictate the OP forms rather mechanically. (On the other hand speak is given two alternate pronunciations.)

Now, you've got to make things a bit simpler for the actors. Crystal's OP transcription generally has the (somewhat anachronistic) peel/deal merger:
These two types of word [DEAL/PEEL], phonologically distinct in Middle English, are not distinguished in this dictionary. It is not clear just how far a merger would have taken place by the end of the sixteenth century, or which words would have been affected. But there is a consensus that the gradual rising in this part of the vowel-space still had some way to go before reaching the present-day value of /i:/, which is shown in Gimson and derivative works as close to cardinal 2. In OP it seems likely to have been nearer to cardinal 1—and thus similar to the Modern French vowel in bébé. Transcriptionally, it could therefore be symbolized as /e:/—and this was the practice adopted in Crystal (2005). However, actors found this confusing, with the letter e also being used for the more open short vowel (see above); there was a persistent tendency to over-open the long vowel, so that sleep, for example, would be pronounced as /slɛ:p/, thus neutralizing the contrast between such pairs as meek and make. In the present dictionary I have accordingly kept the /i:/ symbol, so that in OP training it is necessary to remind practitioners of its more open character compared to RP.
Oh but it is clear that the merger had not fully taken place in most London Englishes by the end of the 16th century. Very few historical phonologists specializing in Early Modern English think otherwise today. The only thing I can imagine is that Crystal is relying on outdated research and giving scholars like Wyld far more credence than they deserve on this point. Rhymes of the type TEA/SAY continue into the early 18th century, and very few poets show anything like a full merger before then. Statistical analyses of rhyming habits tend to back this up as well. The DEAL vowel is indeed far more likely to rhyme with the PELL vowel or the PILL vowel than the PEEL vowel. None of the many sources we have for this period describe anything like /i:/ for the PEAL vowel. On the contrary, contemporary foreign observers tend to equate it with their language's /e:/ or /ɛ:/. The various early phoneticians trying to reform English spelling make it very clear that they perceive the PEEL and DEAL vowels as distinct.

More importantly for our purposes, Shakespeare generally does not interrhyme the two anymore than he interrhymes other similar (yet unarguably distinct) vowels. He is just as likely to rhyme the DEAL vowel with the PELL or the PALE/PAIL vowel (as in Macbeth/heath, bless/peace) as he is with the PEEL vowel (e.g. sea/thee, please/knees, beseech/teach.) Note that whereas he does interrhyme the PELL vowel with the DEAL vowel, he does not  generally interrhyme the PELL vowel with the PEEL vowel. This on its own, even were other evidence lacking, would strongly suggest that the two vowels were distinct for Shakespeare.

I am not sure what Crystal actually means when he says that "there is a consensus that the gradual rising in this part of the vowel-space still had some way to go before reaching the present-day value of /i:/." Seriously, as I read that claim I loudly erupted "what are you even on about, man!? Are you that ignorant, or just praying to sweet baby Jesus that I am?"

There is certainly a consensus that the DEAL vowel had not yet merged with the PEEL vowel (in most sociolects). But if he means that neither the PEEL nor the DEAL vowel had yet reached /i:/ then I am truly not sure what tree he is barking up. By most accounts, PEEL was already essentially /i:/ in this period. Historical phonologists of a great many methodological schools, from Donka Minkova to her cranky adversary Roger Lass are in agreement on this. (The most that some will allow anymore is that early on this may have actually been phonetically a bit lower for some speakers, a sound transcribable as [ɪ̟:] or [e̝:], but still distinct from /e:/.)

It is certainly possible for a language to lack an /i/ sound. Many Quechuan languages have only /ɪ/. Tehuelche has just the vowels /e a o/ with no close vowels. Adyghe and many Sepik languages have a vertical vowel system consisting only of /ɨ ə a/. But in languages like this, the lack of /i/ is paralleled by a lack of other close vowels, or is compensated for by allophonic variants that do surface as [i]. And it would be typologically anomalous in the extreme for a language whose vowel-grid contains /a: e: ɛ: ɔ: o: u:/ to lack /i:/ altogether. Usually when there is a gross asymmetry in the vowel grid, it is the back vowels that are lacking, and not the front vowels. 

That whole paragraph which Crystal devotes to this vowel frankly seems like it is soft-peddling its solution to two understandable problems.

First problem: as Crystal himself has admitted elsewhere, modern English-speaking actors cannot reliably be taught to pronounce three different vowel-heights for PEEL/DEAL/PALE within a reasonable amount of time. For a Modern English speaker, learning to correctly pronounce these three vowels of Elizabethan English would every bit as difficult as learning to correctly pronounce the three different vowels of Modern French pris, pré, près. I myself, as an accent coach for a stage performance employing Middle English, found it extremely difficult to get a single actor to reproduce an /i: e: ɛ:/ contrast consistently. Hell, just getting people to distinguish /ɛ:/ and /e:/ consistently is a trick. I could only imagine the problems that would be involved with getting a whole cast to do this. So that is completely understandable. But I see no reason not to be forthright about that in the dictionary. (For a case in point, listen to this recording of Ben Crystal, David's son, reading sonnet 116 in Crystal's OP. Even he, after lots of practice, still can't manage to keep the /ɛ:/ of shaken/taken distinct from the anachronistic /e:/ he is trying to pronounce in weeks/cheeks. He actually winds up pronouncing a higher vowel for the former!)

Second problem: a DEAL/PEEL contrast would actually make some of Shakespeare's rhymes (like beseech/teach etc.) rather less perfect than they would be if read aloud in a modern accent. (Some of these, though, like rhymes involving speech, probably just rest on a reflex of West-Saxon /ǣ/). That is really not a problem at all, unless you proceed (as Crystal seems to) from an ideological assumption that Shakespeare's rhymes must always be more euphonious in late 16th century pronunciation than in a Modern accent. On average this is certainly true, but a language with more vowel contrasts lends itself to somewhat more "imperfect" rhymes.

In a London where coexisting Englishes had related but perceptibly different mergers in progress, there is no reason at all to assume that the concept of "full" rhymes meant exactly the same thing to Shakespeare as it does to us. Shakespeare rhymes sea/plea, sea/thee, sea/play and it has long been obvious to scholars that multiple varieties of English are at work here. Nor is Shakespeare at all unusual in this. His contemporary George Chapman rhymes sea/way, sea/he, sea/plea all in a single work (though rhymes of the latter type are by far the most prevalent for him as they are for Shakespeare.) I can't believe I'm saying this, but one really ought not to make rhyme into an aesthetic fetish that blocks out other considerations.

But by giving us a merged PEEL/DEAL vowel, Crystal actually creates two completely unnecessary problems for himself. First, it limits the usefulness of his dictionary. Not all reconstructive actors or reenactors are alike. Some can and will go the extra mile for a vowel.  Those actors who want to try and give a performance with unmerged /i:/ and /e:/, but have not been cursed with the demonic drive that leads one to obsessively study Old and Middle English phonology, will be completely unable to figure out which words have /i:/ as in PEEL and which have /e:/ as in DEAL using this dictionary. They could usually look at the spelling, but the spelling is not a sure guide to a word's historical vowels, particularly with the loanwords that comprise the overwhelping majority of the English lexicon. For example, the following words had the DEAL vowel: these, complete, extreme, theme, scene, Jesus. 
Moreover, it makes certain rhymes less perfect than they otherwise would be. Shakespeare's rhymes like eats/gets, heat/sweat cannot accommodate a merged PEEL/DEAL vowel without positing a massive number of alternate forms. This is what Crystal does in many cases, but it's quite unnecessary. Variation there assuredly was, as I've said, but some of these supposed variants are suspect. Giving e.g. /hɛ:t/ alongside /hi:t/ for heat in order to justify the rhymes with get and sweat really seems like special pl[e:]ding.

***Does the Boy Buying a Buoy Say Bye?**

Speaking of mergers and lack thereof, Crystal reconstructs a single /ǝɪ/ for the vowel of SIGH and that of JOY/BOIL/LOIN. In his dictionary he says
The identity between the two diphthongs that are distinct in RP is an important source of puns in OP, such as voice / vice, lines / loins, boil / bile, and supported by such spellings as biles, byle, byles for boils (n) and the rhyme groin / swine. A few unexpected words take the same value, notably juice, rhyming with voice, which has OED spellings ioyce and joice. The central and higher quality of the opening element of the diphthong is critical here, and is one of the main auditory features of OP, in view of its use in several frequently appearing words, such as my, thy, by, like, time.
The first thing to note is that the spellings like <bile/byle> for boil in the sense "pustule" simply does not mean what Crystal implies. Boil in this sense is from (Anglian) Old English bīl, like Mile <- Mīl. The regular etymologically expected form would be *<bile>. The Middle English sources overwhelmingly show forms in <y> or <i> and only occasionally in <uy, oy>. The spellings <bile, byle> are the only ones found in Shakespeare's text for this word, and this is how the word is normally spelt in MSS in this period. Such spellings are never used for the verb boil (<- Old French buillir) which is spelled variously <boyle, boile, boyl>. The possibility that the spelling <byle/bile> is due to the peculiar history of this word, rather than evidence of homophony, is strengthened by the fact that we don't find such spellings as *<tye> for toy or the like. It is just unacceptable for Crystal to use spellings like <byle> in this manner.

One of the damningly unfounded assumptions that Crystal seems to have picked up from Kökeritz (despite apparently accepting Cercignani's arguments against it) is that puns may be taken to imply a vowel merger. But puns do not require complete phonological identity of individual words, let alone a full-blown merger in all eligible lexical items, in order to be effective. Puns are a complex phenomenon, and the perceptibility of a pun depends not only on phonological similarity or identity, but on a variety of syntactic, lexical, prosodic and even socio-cultural factors. Whether your culture accepts puns as a respectable rhetorical device for high artistry, or simply as cheap humor of dads and other groan men, has important implications for the contexts in which you expect and are likely to pick up on puns. (Renaissance English-speakers, like the Ancient Romans or Medieval Persians, held puns in considerably higher literary esteem than modern Britons or North Americans do.)

To a Modern English speaker, the pun in the film title Meet the Fockers is immediately obvious. Indeed, the MPAA almost gave the movie an R-rating just for the title, and was only convinced to give it a PG-13 rating when the filmmakers demonstrated that there were real American people with the surname Focker. We immediately recognize, or rather sense, the punniness of the title, and can infer that the movie is a comedy, even though Focker and Fucker are not, and never have been, homophones for the vast majority of Modern English speakers.

In any case, even granting that a given pun on loin/line or voice/vice is intended and was perceived as such by its Elizabethan audience, this does not by itself imply that the words were true homophones (let alone for all speakers in London) anymore than Focker/Fucker in modern English. All that can be said is that they sounded similar enough to be interpretable as puns by an audience used to puns in this context.

It is unclear whether the vowel of SIGH was at this point /ɛɪ/ or /ǝɪ/. I myself prefer to transcribe this vowel as /ɜɪ/. But let us allow /ǝɪ/ there for the sake of argument. The vowels of JOY and LOIN have different historical pedigrees, the former from Middle English /oi/ and the latter from Middle English /ui/. The ME /oi/ and /ui/ vowels develop differently.

What complicates matters is that ME /oi/ words seem to have had variants in /ui/, and all ME /ui/ words seem to have had variants in /oi/ as well. Later on in the century, some time after Shakespeare's death, we do have quite unambiguous evidence of a full merger for /ui/ words and some /oi/ words (leading to modern lexical doublets like rile/roil, heist/hoist) in at least some types of speech that were viewed as standard. But even then most original /oi/ words like <toy> were still distinct from the vowel of <tie> for many. Robinson has a fullblown merger to ME /oi/ (but no TILE/TOIL merger).

Later poets not uncommonly use rhymes like choice/device, and some of Shakespeare's contemporaries rhyme the reflex of ME /i:/ with that of ME /ui/. (Spenser, for example, has: destroy/Ispoil/beguile, join'd/mind, destroyed/cried. Samuel Daniel has while/toil and the like.) But even a rhyme like mile/toil does not necessarily entail a complete vowel merger at the time of composition. Mile/toil could still be a possible rhyme if the TILE vowel was a diphthong but TOIL vowel was a triphthong of some kind distinguished by a labial an onglide. I.e. TILE /tɜɪ̯l/ but TOIL /tu̯ɜɪ̯l/. See below on evidence for this.

There are other problems with the characterization of TOY/TIE, TROY/TRY, FOIL/FILE as complete homophones. One is that they just do not behave the same way when used as rhymes. For example, the final -y of words like remedy could rhyme either with that of TRY or with that of TREE. Rhymes of the type try/remedy are banally common in Renaissance English verse, and alternate freely in the same poet's work (and often in the same poem) with rhymes of the tree/remedy type, even though try/tree cannot rhyme with one another directly. Now, although the vowel of remedy is extremely commonly rhymed with that of try, it is very rarely rhymed with that of Troy (Marlowe has harmony/destroy followed by legacy/sky. I am unable to find other examples.) If you can have a rhyme remedy/try far more easily than remedy/Troy, it is unacceptable to assume that the two words were pronounced identically because of a few puns. It makes scarcely more sense than if one were to claim that the alternation between rhymes of the type try/remedy and tree/remedy meant that try and tree were homophones.

In fact, the TROY and TRY type words occasionally occur in positions relative to each other where a rhyme would normally be actively avoided. The following quatrain (Sonnet 58, lines 8-12) is taken from Thomas Watson's Tears of Fancie:

So have I found and now too deerely trie,
That pleasure doubleth paine and blisse annoy:
Yet will I twit my selfe of Surcuidrie,
As one that am unworthy to injoy

I see four possibilities here:
(1) that this ABAB pattern is simply based on spelling
(2) that the ABAB pattern is based on a convention established by literary precedent (as when 20th century poets use rhymes like memory/eye even though the vowels no longer sound anything alike)
(3) that the TRY/TROY merger was operating but still incomplete.
(4) that there was no merger yet at all, and that annoy/enjoy was indeed a closer rhyme than try/annoy. 

The former two are quite unlikely. The fourth makes a good deal more sense than the third.

Note also that English <oy> is sometimes equated with the French <oi, oy>. French orthography before the 19th century used <oi, oy> for two sounds, one corresponding to /ɛ/ (respelled as <ai> in modern orthography) and the other (still spelled <oi, oy>) to /wɛ/ (which gave way to the formerly stigmatized pronunciation /wa/ in the 19th century, though some 17th century sources seem to have a lexical distribution with /wa/ for some words, and /wɛ/ for others.) The two different values are found in <françois> with /ɛ/ as an adjective meaning "French" and with /wɛ/ as a male name. Claude Mauger  (1666), for example, equates the French <oi, oy> when pronounced /ɛ/ with the sound represented by English <ay>. He also equates the French <oy> /wɛ/ with the sound represented in English by the spelling <oy>. According to Mauger, the final vowel in the name François is to be pronounced with (something like) the same vowel heard in the English pronunciation of the word viceroy. This equation of French <oi> (=/wɛ/) is not limited to Mauger. It is first found in Alexander Barcley's French Grammar (1521), and continues for the next hundred years. Jacques Bellot's Familiar Dialogues, using French-based phonetic respellings to represent the sounds of English, respell point, joiner, boy, employ, joy as <poeint, ioeìner, bouê, imploey, ioè>. His only witness to the merger is <veiage> for voyage. Other descriptions of English <oy> in words like <boy>, <annoy> as being pronounced something like /wɛ:/ or /wɛɪ/ or /wǝɪ/ are found in the writings of Sir Thomas Smith (1552), Thomas Tonkins (ca. 1600), Charles Butler (1633) and Richard Hodges (1640s) among native sources. That these are not simply induced by traditional spelling is clear from the fact that a number of our sources note that some <oi> words are pronounced with /wɛɪ~wǝɪ/, others contain something in the range of /ɒɪ~ɔɪ/, and others vary between the two. 

We also find French <moi> equated with a made-up English word spelled moy by Pistol in Henry V. And then there's the following from King Richard II 5.3:

And if I were thy Nurse, thy tongue to teach,
Pardon should be the first word of thy speach.
I neuer long'd to heare a word till now:
Say Pardon, King, let pitty teach thee how.
The word is short: but not so short as sweet,
No word like Pardon, for Kings mouth's so meet.
Speake it in French, King, say "Pardonne-moy."
Dost thou teach pardon pardon to destroy?
Ah my sowre husband, my hard-hearted Lord,
That set's the word it selfe, against the word.
Speake Pardon, as 'tis currant in our Land,
The chopping French we do not vnderstand.
Thine eye begins to speake, set thy tongue there,
Or in thy pitteous heart, plant thou thine eare,
That hearing how our plaints and prayres do pearce,
Pitty may moue thee, Pardon to rehearse.

Here we have French moi rhyming with destroy (which had /ui/ in Middle English). Crystal's dictionary includes the pronunciation of French words found in the plays, and for moi his entry reads "Fr mwɛ, Eng məɪ" the latter apparently to supply the rhyme with destroy (since for other French words like Roi, doigt, foi, droit, doit, point, demoiselle he lists only /rwɛ, dwɛ, fwɛ, drwɛ, dwɛ, pwɛ̃, dǝmwɛzel/. The only way I can imagine to justify /mǝɪ/ for Fr. moi is as a fossilized survival of the pronunciation of Anglo-Norman <mei>. Otherwise, the only interpretation worth bothering over is that destroy was indeed pronounced something like /dɪstrwɜɪ/.

Dutch, French and German descriptions of words in the relevant lexical sets throughout the period make it very hard to seriously sustain the idea of a merged TRY/TROY vowel in Shakespeare's lifetime, let alone in all <oi> words. That a /ǝɪ/ pronunciation with no labial onglide was to become current in the later 1600s is quite clear, and it probably did have earlier currency in some sociolects, but it will not do to back-project it as Crystal does. Especially since Shakespeare in particular is distinctive, if anything, for how he avoids this kind of interrhyming. Whereas Spenser and others use rhymes of the type toil/compile (which may easily rest on a pronunciation /twɜɪl kʊmpɜɪl/ anyway), Shakespeare himself avoids them for the most part. The groin/swine rhyme that Crystal mentions above is the closest thing to a cross-rhyme of this type in the entire Shakespearian corpus, and it's far from clear-cut. Groin itself is an irregular development from earlier grine under the influence of loin. In fact, that line from the Quarto edition of Shakespeare's Venus and Adonis appears to be the earliest recorded instance of the modern spelling with <oi>. It can hardly be used as evidence for a thoroughgoing merger. Even if one accepts that the spelling implies pronunciation, swine has an onset cluster with /w/. The rhyme might rest on /swɜɪn/ and /grwɜɪn/ which would rhyme just fwine. 

***Of Whores and Hours***

Quoth Crystal:
The important point to note about this vowel is the lack of the diphthongal quality characteristic of RP, where it has a range of values running from [oʊ] to [əʊ] to [ɛʊ]. The pure vowel is widely used in present-day accents, such as those of the Celtic areas, and its frequency in English (in very common words such as go, know, so) makes it a noticeable feature of OP. Rhymes show its use as a variant in words that later would have more open vowels, such as one / throne, none / bone. Several words and prefixes spelled with or or our, shown in this dictionary with /ɔ:/, such as four, more, fore-, for- could also be sounded with a closer variant.
Here again I see the ghost of Kökeritz hovering behind the page. Positing /o:r/ for the reflex of ME /u:r/ is a problem. We have abundant evidence that in most varieties of English, including those current in London, ME /u:r o:r/ were kept largely distinct. By "largely distinct" I mean that they did not fully merge. There was some interchange between the two. Many words (even those where the fluctuation cannot have been inherited from Middle English) could clearly be pronounced two different ways by speakers who unarguably did not merge these two. 

This is one more point where puns lead Crystal, as they led Kökeritz, to make entirely too much of very little, despite his demonstrated familiarity with Cercignani's rebuttal of Kökeritz' eagerness to see puns anywhere and everywhere. The asserted phonetic identity between e.g. whore/hour/o'er is another instance of the problem. That words like POWER, HOUR probably had covarying pronunciations with and without a diphthong is likely. But it's an enormous leap from this fact to the fullblown PORE/POOR/POWER merger which Crystal reconstructs. In any case, to act as if any given instance of HOUR may be potentially taken as a homophone for WHORE and made to service a pun is quite unacceptable. Crystal is on this point even more inflexible than his predecessor. Even Kökeritz at least allows for variation between /ǝʊr/ and /o:r/ for the vowel of HOUR and POWER. But Crystal's transcriptions simply equate them as /o:/ with not a footnote's worth of wiggle-room. It is inadvisable to waste too many hours of one's day obsessing o'er whores. So to speak. 

Seriously, just read Cercignani's response to Kökeritz pulling this trick. Crystal really ought to have known better. He mentions Cercignani's work repeatedly. How is he acting like he didn't read it at all here? This is very irritating.

***Open your mouth and say <A>***

Often the sources Crystal cites in the introduction to his dictionary don't necessarily mean what he takes them to mean. For example, he cites Sir John Harrington's anecdote about how a gentlewoman mistook the French name Jacques for "Jakes" (the scabrous term for a privy.) Then since Jakes is rhymed with Makes by the same author, the logic goes, the pronunciation must be /dʒɛ:ks/. But this implies certain assumptions about how the vowel in makes is pronounced in the first place. It's not clear, to me anyway, that something more like /dʒæ:ks/ is not to be inferred instead.

(For what it's worth, I do think that /dʒɛ:ks/ and /dʒæ:ks/ could both be heard for some time.)

Many of Crystal's phonological arguments are just absurd. For example, he gives words like war, guard as /wɑ:ɹ, gɑ:ɹd/ with a back-vowel. He then claims that  this vowel
must have been a noticeable feature of OP as Jonson, among others, pays special attention to it, contrasting it with the normal use of a (‘pronounced less than the French à’): ‘when it comes before l, in the end of a syllabe, it obtaineth the full French sound, and is uttered with the mouth and tongue wide opened, the tongue bent back from the teeth’. He gives all, small, salt, calm among his examples.
This is, to put it mildly, a shamefully selective reading of Jonson. What Jonson actually says in full is
With us, in most words, is pronounced less than the French à : as in art, act, apple, ancient. But when it comes before L, in the end of a syllabe, it obtaineth the full French sound, and is uttered with the mouth and throat wide opened, the  tongue bent back from the teeth, as in all, small, gall, fall, tall, call.  So in all the syllabes where a consonant followeth the L, as in salt, malt, balm, calm. 
In other words, Jonson appears to hear the words art and apple as both containing the same kind of a-vowel. Furthermore, he finds this kind of a-vowel in art and act is perceptually different from that of all, small etc. Yet Crystal reconstructs the same /ɑ:/ for both small and art, and then gives a different vowel for words like act. This "OP" is quite clearly not the English that Jonson is describing. 

In any case, we have no indication, either from rhymes, foreign descriptions or native orthoepists, that the vowel in art was any lower than that of act until some time after Shakespeare's death. 

***Of Marjers and Murjers***

Crystal does something similarly unconvincing with /ɐ:/ for the vowel in bird, mercy, sir.
The open quality of this vowel is heard today in many regional accents, on both sides of the Atlantic, reflected in dialect-writing in such spellings as the exclamatory ‘marcy me . . . !’ The spelling evidence in the Folio is seen in the use of an a in such words as merchant / marchant, sterling / starling, German / Iarman, and rhymes such as serve / carve, stir / war. Phonetically, there is little difference between this quality and that of /ɑ:/ below, but I have kept the transcriptions distinct, to draw attention to the different phonological relationships with their present-day equivalents.
A few things spring to mind. First, I'm not sure that stir/war in Richard II is actually meant as a rhyme. Second, even granting the merger for the sake of argument, why would the result necessarily be /ɐ:/? Just because there are modern English dialects with pre-rhotic /ɐ/ for orthographic "e" does not mean that Elizabethan London English had the same kind of vowel. Cross-linguistically, we know that this kind of lowering before retroflex approximants can produce many different vowels in the vicinity of [a].  "Phonetically, there is little difference between this quality and that of /ɑ:/ below" says Crystal. Yet the sources describe the A-sound even before R as being rather different from whatever the vowel of ball is.

What is more likely, and more supported by both the rhyming practices of the day and the statements of orthoepists, is that the vowel of act and of art was indeed (more or less) the same vowel. The most straightforward inference is that this vowel was simply /a/ or something fo the kind. Certainly there's no real evidence of a phonemicization of /ɑ:/ in both words like "ball" and "art". 

Crystal, on the other hand, essentially uses /ɐ:/ to just put his OP through the NURSE-merger. That is, the merger of the vowels in EARTH, DIRT and TURN originally /ɛɹθ dɪɹt tʊɹn/. It seems clear to me that he accepts this merger for Shakespeare's English largely on the basis of broad statements by the likes of Wells, Nevalainen, Kökeritz etc. about how the merger "was accomplished by the 17th century." Crystal appears to have latched onto this for its apparent modernity and ignored the massive amount of better and more recent scholarship demonstrating that it was not so. His attempts to bolster his reconstruction of an a-like vowel rely on at least one gross misconstrual of evidence. 

The idea that the pronunciation reflected by "the use of an a in such words as merchant / marchant, sterling / starling, German / Iarman, and rhymes such as serve / carve, stir / war" has anything whatsoever to do with the three-way NURSE-merger, let alone with the quality of the vowel resulting from it, is just absurd. First, marchant actually reflects not an English sound change at all but a French one. The word came into English with /ar/ and the doublet in /er/ is probably due to the influence of Latin. Spellings such as Iarman reflect Late Middle English lowering rather than the Early Modern NURSE merger, as is made clear from the fact that it is only historical /ɛr/ words that turn up with an <ar> spelling. Barring a handful of (mostly) debatable exceptions, and some cases (like Crystal's stir/war) which may not be intended as rhymes at all, it is also only the /ɛr/ words which are optionally rhymable with /ar/ as in convert/art, and only the /ɛr/ words which are susceptible to rhymes like verse/pierce where the /ɛ:r/ of pierce was never NURSE-merged in the first place. This all involves, one way or another, a property peculiar to the /ɛr/ words, and the very existence of such a property implies something less than a fullblown three-way merger across the lexicon. Although we have common crossrhymes of historical /ɛr ar/, of /ɪr ʊr wɔr/, of /ɛr ɔr/ and of /ɛr ɪr/, we only very rarely have crossrhymes of the types /ʊr ar/ and /ʊr ɛr/ (e.g. Drayton's turn/earn) from the period. Even most of the seeming /ʊr ɛr/ or /ʊr ar/ rhymes, such as Shakespeare's quern/churn and farther/murther, have compelling explanations that suggest something other than a merger (e.g. here farther is in fact further <- OE furþor, and churn <- OE ċyrn would be expected to produce /ɪr/ in Anglian with a possible shift -> /ɛr/, which is indeed suggested by the spelling <cherne> in the First Folio.) That Crystal's OP reflects modern anglophone sensibilities on this as on other points is made plain by the fact that the merged /ɐ:r/ of Crystal's NURSE vowel is usually only given for those ME /ɛr/ words that happen not to survive with /a/ lowering in Modern Standard English. Thus starve is listed by Crystal as /stɑ:rv/ with the "normal" pre-rhotic a vowel, while deserve is listed as /dɪzɐ:ɹv/ with Crystal's NURSE vowel (even though, as Folio spellings like <sterve> suggest, there is no reason to think that these two words when rhymed had different vowels.)

This merger had not fully taken place in all London lects in Shakespeare's lifetime, and for many southern Englishes would not be complete before the 19th century. For long afterward what seems to be described by the sources is not a three-way merger, but a merger of the DIRT and TURN vowels, leaving the EARTH vowel distinct. The DIRT and TURN vowels begin to merge during Shakespeare's lifetime. At the same time, though, as Roger Lass has noted, there seems to be a competing lineage of English, in which the vowels of EARTH and DIRT merged into /ɛ/ while keeping the sound in TURN distinct. This allowed Shakespeare to interrhyme all three of them. A rhyme fir/fur would work for some speakers, and a rhyme learn/burn would do the same for others. These two merge-patterns seem to have interacted with one another in complex, unsystematic and idiolectal ways eventually.

A complete NURSE merger also causes rhyme problems. Among other things, it fails to explain why poets of the period quite commonly rhyme the NORTH vowel with the TURN vowel (e.g. return/morn) but not with the EARTH or the DIRT vowels, and it fails to account for rhymes such as pierce/verse.

Crystal, when confronted with the attested evidence of 16th-17th century orthoepists, has tended to suggest that they were influenced by spelling in the data they reported and that they should be "taken with a grain of salt". That they could be influenced by spelling is not in doubt. But the idea that, for example, the Londoner Robert Robinson's (1617) report of a three-way FIR-FUR-FERN distinction can be taken to rest purely on spelling should be absolutely preposterous to anyone who has actually read the man's writings. First, he reports /ur/ in expected places (from OE /yr/ or metathesized /ri/) where the orthogrphy would imply /ir/, and there's no way he could do that if he were not hearing a real distinction. Second, the man created an entirely new script deliberately unrelated to the Roman alphabet to get around the limitations of the latter. He was sometimes influenced by spelling (for which his transcriptions of Barnfield are full of evidence) but the non-spelling-based pronunciations he reports (such as /dl tl/ for orthographic <gl> and <cl>) are legion, and his distribution of /ur/ words simply cannot be explained as anything other than the result of a real pronunciation he heard and probably used himself. 

***Och, ya cannae drop yer velars sae easily***
"It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so."—Mark Twain
What there isn't evidence for is sometimes every bit as important as what there is evidence for. One example of this in Crystal's OP reconstruction will suffice. He reconstructs an /ɔ:/ for -aught and -ought words like wrought, sought, taught etc. Alright, fair enough. But consider what these words do and don't rhyme with. In Shakespeare, words of this kind only rhyme with each other (caught/thought, daughter/caught her etc.) and with a small handful of other words with a tautosyllabic consonant in the stressed syllable like oft/naught, after/daughter. A single passage in King Lear includes after/daughter/halter/caught her/slaughter as apparent rhymes with each other. And there is a possible (but not certain) rhyme of slaughter'd/butcher'd in Richard III.  (Crystal gives forms for halter and after in which the F and L are optionally not pronounced. Which is safe enough, as such forms did indeed exist. But oft cannot be thus explained.) This cannot be taken as evidence for a monophthong, however. Even in the King Lear passage, the five-way rhyme could easily rest on a diphthongal reflex of Middle English /au/. 

Now, Shakespeare rhymes things like note/pot, smote/not. But never once does he rhyme words like pot or note with -aught/-ought. The vowels /o:/ (which is the vowel of note in Crystal's OP) and /ɒ/ (the vowel of Crystal's pot) are less similar in every way than /o:/ and /ɔ:/. Why does Shakespeare rhyme note/pot without ever once using rhymes of the type caught/note caught/not? Assuming it isn't some freakish coincidence, the best explanation is that Crystal's OP has gotten some crucial feature of the -ught words quite wrong (or, rather, that he was quite wrong to follow Kökeritz yet again when it comes to this lexical set.)

Searching among poets born before 1600 I have not been able to turn up a single caught/not-type rhyme. Very occasionally, we do find early rhymes like notes/thoughts in a poet like Samuel Daniel, but we also find the -ught words rhymed in in other ways. George Chapman, in addition to paralleling Shakespeare's oft/naught by rhyming wrought/soft and brought/soft, also has growth/broughttaught/laugh'd and three instances of fought/out. Golding rhymes mought/about and laught/caught, laught/straught. We occasionally find fault rhymed with thought, brought (Lady Wroth) and nought (John Davies, John Standysh). Spenser has water/daughter. Drayton has slaughter/laughter, wrought her/laughter. John Donne, who was no slouch about rough rhymes, has no confirmed instances of crossrhyming -ught with any other type of word. Poets born in the 17th century do have rhymes like Vaughan's sought/not and Katherine Philips' wrought/knot, but poets of Shakespeare's generation seem to want to keep the -ught words in their own rhyme class for the most part, and when they do cross-rhyme them, it is generally not in the ways Crystal's reconstruction would lead one to expect.

There is a good reason for this. The pronunciation of words of this type seems to have varied in some quite drastic ways among speakers. Some apparently still had the velar fricative of Middle English here. Others will have had /f/ in many these words.

A later authority, Simon Daines (1640) tells us that <augh>

...sounds like Af for the most part, as in...daughter, laughter, which most of us pronounce dafter, lafter; except slaughter, which is slater, with A broad and full...the rest goe according to the tenure of the precedent rules, as caught, taught &c."

I should mention that the sound Daines is actually describing may not be as straightforward as it might seem. He could well be describing a sound similar but not identical to [f] such as [ɸ] or [ʍ]. The sheer perceptual similarity involved could be enough to license a rhyme with the word "after". Listen for your self: here is a recording of me pronouncing [daftɚ], [daɸtɚ], [daʍtɚ] in sequence. Still, the existence of genuine [af] variants is amply documented by other evidence.

Some others probably had a diphthong here. Robert Robinson (1617) describes a diphthong for thought and similar -ught words (but implies a monophthong in words like vault.)

Whether Crystal's /ɔ:/ existed at all for -ught words is not a sure thing. Speakers later on the 17th century seem to have had /ɒ:/ (merged with the reflex of ME /au/), and this value (or something like it) existed for some in the late 16th as well. It is just possible that the higher /ɔ:/ is attested by Bellot, but this is really not at all clear. 

But the diversity of realizations of -ught will have meant that one could not easily crossrhyme it with other types of words without risking a perceived "imperfect" rhyme for a great many speakers.

In any case, Crystal's simple reconstruction of /ɔ:/ for most such words is not the full story. His dictionary lists nought as rhyming with oft, but the variant pronunciation /nɒft/ implied by this rhyme is not given. Why on earth not? I cannot escape the impression that this is (once again) motivated by his aesthetic of linguistic equidistance, and has very little to do with an attempt to accurately represent what we know of London English ca. 1600.

***Oo look at Luke's yew and his ewe, w[ɪw] you?***

Three more of Crystal's vowels.
This rounded vowel seems to have had the same value as in conservative RP today (though it is now losing its rounding among young people). The only uncertainty is the extent to which it was used as an alternative in words with long [u:]. Rhymes such as tooth and doth, brood and blood, food and flood, and puns such as fool and full show that it was an option in some cases, but whether it should be applied to moon, afternoon, and others is an open question. Rhymes can be suggestive, such as boot / foot, but the direction of the rhyme is often unclear. The dictionary thus shows long and short vowels in these words, with the latter more likely in regional speech, as today.
This value seems identical with the one we have today in conservative RP accents (younger people tend to lose some of the lip-rounding), though—as noted above—several words that today have /u:/ could be shortened, such as fool. Spellings such as cooz and coosin (‘cousin’) show that oo could represent a short vowel as well as a long one.
The distinction between /u:/ and /ʊ/ (and later /ʌ/) is a vexing problem for anyone trying to reconstruct earlier Englishes because one cannot arrive at earlier forms simply by rewinding the sound-changes like an old style cassette tape.
While Crystal is willing to allow for shortened alternatives to modern canonically long vowels (as with /fʊl/ for fool) he for the most part does not allow for longer vowels in words that  today have short /ʊ/. He gives only /ʊ/ for Hook, Look, Book and others. He does give both /ʊ/ and /u:/ as possibilities for Nook. I again cannot escape the impression that he is letting the modern state of the language dictate his choices here. 

The other problem is the inclusion of NEW here. Assuming (acceptably) an early DEW—DUE merger for some Englishes in Shakespeare's London, the vowel of this word was a diphthong /nɪʊ̆/. Shakespeare and his contemporaries generally do not interrhyme words of the DEW—DUE type with words of the DO type. At least, they don't do it anymore than they interrhyme things like the HEAL and HEEL vowels. What's going on here, again, is that Crystal is letting a very small handful of rhymes like wooing/suing dictate the segmental inventory. 

***Do I rhyme or do eye-rhyme?***

On the matter of "eye-rhyme" (or rather approximate rhyme) Crystal cites George Puttenham's The Arte of English Poesie where the author rails against inexact rhyming. From this, he infers, that "It is the ear, not the eye, that is the theme of sixteenth-century writers."

"It is the ear, not the eye" indeed. But a few points emerge from this. First, the fact that George Puttenham rails against the "fowl faults" of the "bungler" who "falsifies his accent to serue his cadence" actually confirms — as Crystal admits — that inexact rhyme of whatever kind was in use as a poetic strategy and was widespread enough to annoy the likes of Puttenham. (In the earlier half of the sixteenth century, poets like Wyatt employed all manner of approximate rhymes: am/man, sprites/likes, loud/rood etc.) One must not mistake prescription for description.

Even if one assumes that the "agreeing sound" must exist in the ear or mouth of either poet or audience, this does not necessarily imply complete identity of syllable nuclei, let alone of syllable rime. In fact, it may not imply the same level of identity in all contexts.

In Shakespeare's corpus, inexact rhymes are not evenly distributed. Indubitably approximate rhymes (e.g. ship/split, sung/come) seem more frequent in certain contexts (such as the Gower chorus from Pericles.) This suggests that approximate rhyme could be used intentionally to generate particular effects. (In Gower's case, the intended effect might be antiquarian, evoking the somewhat freer rhyming practice of late 15th and early 16th century poetry. This would be most appropriate for a character meant "to sing a song that old was sung.")

The common assumption that inexact rhyme is to be understood primarily as "eye-rhyme" is the result of centuries of print culture. "Eye-rhyme" is beside the point. The question is: how similar do the vowels have to be in order to be acceptable as ear-rhyme? A cross-linguistic and cross-cultural comparison of poetic rhyme traditions throughout the world and history would show that this is not simply a question of phonetic, or even phonological, fact.

I might formulate some impressionistic rules of thumb. The more distinctive vowel contrasts there are in a given language, the more likely it is that different vowels will be cross-rhymed (whether occasionally or as normal practice). The more variation is deemed acceptable and normal among different speakers or listeners, the more likely it is that "compromise" rhymes (which are perfect for some speakers but not all) will be admissible. When and if this stops being true, "correct rhymes" frequently become a matter of tradition, and/or of adherence to a prescribed standard. (Thus the banishment of imperfect rhymes from good literary Dutch after the Rederijkers period.) When the standard itself admits some amount of variability, or is unable to eliminate that variability, phonologically inexact rhymes may be treated as full rhymes. Thus in traditional Italian versification, /ɔ/ and /o/ on the one hand, and /ɛ/ and /e/ on the other, are permitted to interrhyme (otherwise Dante wouldn't be able to enjoy rhyming amore/cuore on a semi-regular basis.) In German, höh/See and über/lieber are not considered inexact rhymes.

"An agreeing sound in the last silabes" does necessarily imply an agreeing sound in actual oral delivery. If (let us say) weather and breather are not perfect rhymes in my speech, but they are perfect rhymes in some other people's, that may well license me to use it. It does not necessarily follow that I myself in reading my own work would necessarily produce the pronunciation on which that rhyme rests. The "agreeing sound" may exist, but not in my mouth. The perfectness of a rhyme may rest on the reader's or listener's broad awareness of speech-types. This kind of "agreeing sound" would indeed be consistent with how poetic rhyming traditions often behave.

Crystal goes on to discuss the cases of inexact rhymes that even his OP can't get rid of and finds that many of them are cases "where the phonetic distinction is so slight that the rhymes might well have been perceived to be identical" and in a footnote says that "most so-called half-rhymes in Shakespeare are in fact differentiated by a single distinctive feature." Well then, suppose that two vowels only differ in prosodic length?

Suppose they are the indeed quite similar but not for all speakers? You can get around problems like Shakespeare's waste/cast by positing the (very likely) alternative pronunciation /wast/ for waste. Alternate byforms from Middle English can be invoked to explain rhymes like chat/gate (in Shakespeare), sake/back (in George Herbert), lamb/came (Spenser), a glass/place (George Herbert). 

What, on the other hand, is one to make of the fact that in some styles, he seems to want to keep the reflexes of Middle English /a:/ and of Middle English /ai/ distinct as rhymes? In Shakespeare's dramatic verse such cross-rhymes are strikingly rare. And in the entirety of his non-dramatic verse I find not a single secure crossrhyme of these two. All seeming exceptions — improperly taken by the likes of Kökeritz as evidence of a merged pronunciation — in fact rest on words like "gait" and "hair" which are known to have had monophthongal variants for other reasons. In A Lover's Complaint, (where came/aim/tame/maim/exclaim are meant to be rhymed ABABB) he seems to deliberately treat them as if they were separate rhymes in a single stanza. If these were merged in general in London at that point, then they must have been meant as "eye-rhymes" in some sense, a silent effect to be enjoyed by those familiar with older poetry. Of course, we know from the orthoepistic evidence (which Crystal prefers to dismiss as spelling-influenced in such regards) that a distinct diphthongal reflex of ME /ai/ survived in at least some people's living educated speech for another century. Some kind of peculiar effect is possibly intended here based on the fact that this stanza would seem like monorhyme in some but not all kinds of speech. 

The fact that Crystal just prescribes a merged monophthong here and doesn't mention the diphthong even in discussing sociolinguistic issues or other points touching on variation is bizarre.

*** Inconvenience: When Modern English Rhymes Work Better ***

One thing that Crystal's reconstruction tellingly avoides is the splitting of too many modern English lexical sets. While Crystal is willing to allow for mergers (tile/toil, poor/power etc.) that are absent from the modern standard, he is unwilling to undo mergers that are present in it. His unwillingness to reflect a three-way contrast between e.g. great/grate/greet is just one case in point.

The problem is that those early 17th-century speakers who recorded their own pronunciation directly simply do not back this up, and this makes nonsense of the myth that if you heard a 17th century English speaker read Shakespeare aloud, their rhyming would always be more perfect than what would obtain if the same text were read in a modern accent. This may be illustrated by reference to the idiolect of a single speaker who recorded his speech in detail: Robert Robinson, a younger contemporary of Shakespeare. The following are rhymes of Shakespeares' which would be perfect in modern accents but imperfect in Robinson's speech: beauty/duty, soul/whole, stir/spur, wooing/suing, beseech/teach, birds/herds, extremes/seems. 

We're talking about a poet in Elizabethan London, in contact with the kind of sociolectal variety you'd expect in a burgeoning urban center with a caste-stratified society. Why would such a man — over decades of literary activity — restrict himself to rhymes that could all be found in the mouth of a single individual reading their work? Find me a single English-language poet that has ever done so. And of course, as I have already pointed out, Shakespeare uses demonstrably inexact rhymes (come/sung, open/broken, father/hither), especially feminine rhymes.

Crystal's dictionary, of course, gives variant pronunciations. But, having contorted himself already into introducing features that are anachronistic (a three-way NURSE merger) or even typologically implausible (a MEAT-MEET merger at [e:]) the "variants" he often gives are themselves wonky. They are given — it seems — precisely so that the user of it can make every rhyme they encounter as perfect as possible. While there is evidence that Elizabethan readers might have made a few substitutions from outside their idiolectal range at line ends for rhyme's sake, it seems very unlikely that they actually did so this systematically.

Robert Robinson and Alexander Gil both transcribe their pronunciation of considerable amounts of verse. Robinson for his part only very occasionally opts for a clearly rhyme-driven realization, and is very often content to ignore rhyme even when his own idiolect could accomodate the requisite variation. His speech contained both diphthongal and monophtongal realizations of the word fair and yet in his transcription of the opening stanza of Barnfield's "Lady Pecunia" he gives the diphthongal form for fair and a monphthong for its rhyme heir, only to transcribe a monophthongal fair two lines later. Robinson occasionally even opts for variants that spoil the meter, which led Dobson to wonder if the man operated by sounding out each word rather than transcribing connected speech. But it's unlikely that he would have transcribed his pronunciation of verse this way had he thought it an "incorrect" way to read a poem.

Gil's transcriptions of Spenser do show more employment of variants at rhyme than Robinson, but he has limits. He does not drop the fricative in light for a rhyme with write. A few of the variants on which the rhymes rested were socially marked for Gil. Granted that others will not necessarily have had his attitude toward them, but what is worth noting is that for Gil a rhyme was felt justifiable and "correct" based on the existance of a supporting variant in the "communis dialectus" — no matter that it might be a variant whose existance he himself disapproved of. Even if "an agreeing sound in the last silabes" is required to justify rhyme, it does not follow that said sound must occur in the mouth of the reader all the time.

There is no implication that the normal way of reading rhymed verse aloud involved systematically and without exception articulating the appropriate variants in rhyme position. Some adjustment there clearly was. Some things that varied quite freely were probably especially up for grabs, like length of the vowel in the good/blood/mood lexical set or the diphthongal/monophthongal pronunciations of the final vowel in words like humánitỳ. But the precise degree of such adjustment must have varied considerably depending at least partly on performance context or social factors. Gil, for example, exercises his spleen at length against speakers who pronounce play identically to plea. One can hardly imagine that he would have permitted himself such a pronunciation when reading aloud poems by Drayton or Sidney (in which rhymes resting on such a variant are not uncommon). Individuals must have varied widely in the range and nature of rhyme-motivated variants they uttered in reading aloud. The fact that Shakespeare — with only two quite debatable exceptions — avoids rhymes of the buy/destroy type (whereas Daniel and Spenser freely employ them) might hint at the possibility that he personally didn't care for them much.

In some cases, though, the variation may have been so free that lack of phonetic identity wasn't even perceived. Do Americans without the CAUGHT-COT merger feel rhymes based on it to be faulty? I lack the merger, but interact with so many speakers who have it that just goes right past me.

And that isn't even the extent of the evidence making this a weird position to entertain. Both Robinson and Hodges transcribe phonetically their own verse compositions, showing that a poet's own rhymes (in Hodges' case forth/worth and in Robinson's what/that) might easily rest on pronunciations he himself did not use.

The point I'm making here is that the ideological centrality of rhyme to Crystal's enterprise has contributed to the ease with which the "OP" movement has embraced something that, judged purely as a reconstruction of linguistic reality, is quite unbelievable.

***More on rhymes***

For each word in his dictionary, Crystal gives a list of instances of where that word is used at rhyme in Shakespeare's corpus. This is a rather useful thing to have, and would be even more so if Crystal were not pronet to listing cases where there's no real reason to think a rhyme is intended. 

One rather egregious instance. Crystal would have marriage rhyme with charge at 3H6 4.1.33. This, I infer, at least partly motivates his reconstructing a form of the former word ending in secondarily stressed /ɑːdʒ/. The morpheme -age under secondary stress is elsewhere reconstructed in his dictionary with /ɛːdʒ/. In fact Crystal reconstructs /ɑː/ elsewhere only for a variant unrounded reflex of Middle English /au/ (which very likely was not a thing, but at least here he's actually taking a position based on something actual specialists like Luick have argued, though very few entertain the notion today), for a lengthened reflex of Middle English /a/ before tautosyllabic /r/ (in fact he quite perplexingly reconstructs pre-r lengthening as a generalized feature of "OP"), and in loans like Gonzalo and Gonzago

Looking at the actual lines in question one finds

And Warwick, doing what you gave in charge,
Is now dishonoured by this new marriage.

This putative couplet is not part of a series and exists in a passage of simple blank verse. There's no reason to think this is intended as a rhyme. The second of these two lines calls for a scansion "Is nów dishónourèd by thís new márriage" where a trisyllabic pronunciation of "marriage" with secondary stress would be hypermetrical. Granted, a disyllable with masculine rhyme on a secondarily stressed syllable would not be without parallel (cf. MND 5.1.129), but given the absence of real evidence for generalized r-loss in this period, the reasoning involved here just seems incredibly weird. 

Other cases of dubious or likely unintended rhymes which Crystal lists include dare:bear M 3.4.99-100 and stir:war R2 2.3.51

***Do You Even Read Non-Shakespeare***

The title of the Oxford Dictionary of Original Shakespearean Pronunciation stands in a long tradition of discovering Shakespearean essences. The typologically implausible phonology it contains is meant to be used in performing texts by Shakespeare, to produce a particular effect on audiences. Crystal's rhyme ideology (or, more precisely, the ideology he generates for himself from sources like Puttenham) combined with his utter collapse in the distinction between "Shakespearean English" and "Early 17th Century London English" leads to some blind spots he might have seen if he weren't treating Early Modern English effectively as if it were the language of a single human being named William.

The entry for the word nothing gives <no:tɪn, -ɪŋ> (quite preposterously) as the sole pronunciation. This is motivated by the fact that Shakespeare rhymes nothing/a-doting and by a pun on noting. The nothing/a-doting rhyme has a long history in discussions of "Shakespeare's pronunciation", going back to a time when scholars were rather more naive about certain things. It is a naivety expressed, perpetuated and endorsed by this entry. Puns, as I've already noted, do not need to rest on absolute phonetic identity, and it is inexcusably perverse to treat this as the only pronunciation of a common function word based on a single rhyme by a poet who is in any case quite willing to use such rhymes as open/broken.

There are, in fact, fairly obvious reasons why such a rhyme might be used even if it does not rest on phonetic identity. If the word was in fact normally /no:θɪŋ~nɔ:θɪŋ/ this would make it a word without many (or, perhaps, any) perfect rhymes. The fact that /θ/ was voiced intervocalically in Old English would see to that. So if one wished to rhyme on the word "nothing", one would need to compromise one way or the other. Rhyming /θ/ with /t/ would be one solution, but there are others. Had Crystal bothered to consider the rhyming of other poets, he would quickly have come upon them. Michael Drayton, for example, rhymes nothing/loathing.

The fact that every OP performance is populated by actors saying /no:tɪn/ for every instance of "nothing" (which, apart from this rhyme, is justified only by one shorthand spelling and occasional 15th century spellings <noting>) because of a set of beliefs concerning a single rhyme is a testament to what OP really is, and really is not. It is not a plausible reconstruction of any kind of English we have evidence for in Shakespeare's day (many of the developments it assumes appear to have been generalized only decades — and sometimes centuries — after Shakespeare's death). It's not even a plausible reconstruction of any kind of speech that ever existed  in English. It can be confidently stated that no human being ever spoke in this fashion before Crystal started coaching actors to do so

Rather it is an expression of a certain kind of modern Anglophone language ideology. I do not think that Crystal would have allowed himself to make assumptions about an entire speech community's realization of a word based on a single rhyme in a sonnet were it not for the way English-speakers are indoctrinated to think about Shakespeare.


There are lots of ways to interpret often-ambiguous data. But the way Crystal is interpreting it in particular is so peculiar, so at odds with what even the sources he uses seem to think they're hearing (and with the findings of a good many historical phonologists), that it is hard to simply attribute this to a difference of scholarly opinion. Many of the questionable parts of Crystal's OP have something in common: they help push the language in a more "modern mainstream" direction than would otherwise be the case. I cannot shake the suspicion that Crystal is manipulating his "OP" phonology into something that won't sound too off-puttingly weird. There is a bevy of aspects of the vowel system(s) that seem to have been unlike most or all modern Englishes, and which Crystal's OP effectively bleeds out.

In all likelihood "Shakespeare's English" would indeed be "more difficult for an audience to understand than any modern regional accent." But that wouldn't sell tickets. Just what would it take to teach modern actors to reliably pronounce it in a reasonable amount of time? What good would an OP-movement do if Shakespeare's phonology proved to be less congenial to us than his texts?

A more plausible reconstruction of the original pronunciation(s) used in reading dramatic and lyrical verse in Shakespeare's London probably would not give people what they prize in the mythical sounds of "OP". Not only is this "OP" realization demonstrably unlikely to reflect Elizabethan linguistic reality (even apart from the typological unlikelihood that a phonology like this could ever arise in a real human language) but it sustains itself based on a set of aesthetic priors that are about as remote from Elizabethan literary sensibilities as anything imaginable. I doubt "OP" phonology gets us any closer to Shakespeare than Tom Stoppard's Shakespeare in Love, but both are fictional works that take wild liberties with the truth in order to give the audience something they'll enjoy.

But I come to praise Crystal and not to bury him (well, ok, not just to bury him). First though, some lexical quibbles.


For daughter: a variant /daftǝr/ should given as well.
For stead, instead: a form /ɪnsti:d/ should be given alongside the more obvious and modern pronunciation.
For verdict: a form without the /k/ should be given.
For satire: a variant should be given with the same vowel as nature 
For fierce: a form /fɛ:rs/ should be given the same vowel as in pierce
For servile: a form /sɛrvɪl~sǝrvɪl/ with a non-tense vowel in the second syllable.
For inveigle: a variant in /ɪnve:gl~ɪnvɛ:gl/ (or in Crystal's transcription /ɪnvi:gǝl/) should be given with the SEA vowel
For edifice, prejudice, benefice, cowardice and similar words, a variant form ending in /ǝɪs/ should probably be given. (Rhymes like edifice/sacrificecowardice/despise and device/prejudice do not occur in Shakespeare, but they do in the work of his contemporaries, including Donne, Chapman and Elizabeth Cary.)
For yeoman a pronunciation /jɛmǝn/ or /ji:mǝn/ should also be given, per Jonson's grammar as well as later 17th and 18th century witnesses. 
For jeopardy, a pronunciation with /i:/ in the first syllable should also be given for the same reason.
For shire, a prounciation /ʃi:r/ ought to be given. 
For evil: alongside the etymologically expected /ɛvɪl/, the form with ME-lengthening /i:vl/ (attested from 1580 onward in Bullokar et al.) ancestral to the Modern English pronunciation should also be given. As, probably, should the form /ɪvl/ given by Hodges and Hart.  
For word, worm, work, world, worship, forth: a pronunciation with the FUR vowel at least alongside the NORTH vowel really should be given for all these words. In forth (as well as afford) also a variant with the same vowel as mourn.  
For Lucrece: a form with the DUE vowel in the first syllable should be given (per Robinson's transcription of this word specifically in reference to Shakespeare's poem)
For Saturn: a form with the MATE vowel in the first syllable should be given (per Robinson and Gil)
For spirit: a form /sperɪt/ reported by Hodges should probably be given as well. 
For build the pronunciation /bi:ld/ should be given.

Actually the absence of /sperɪt/ for "Spirit", /ɪvl/ for "evil" and /bi:ld/ for "build" in Crystal's dictionary is rather whacky. Given Crystal's ideology concerning rhymes with identical vocalic nuclei, it is curious that his dictionary only lists a form identical to the modern pronunciation. The former two are witnessed by Hodges and are completely etymologically explicable variants. Moreover all instances of the word build at rhyme in Shakespeare's works appear to rest on a form with a long vowel. It rhymes exclusively with the words field, yield and shield. The form /bi:ld/ is directly attested in 17th century descriptions of English pronunciation. I have no idea if this is an oversight, or if Crystal just didn't think audiences could handle it. Forms which Crystal would reconstruct as /bju:ld/ and /bǝɪld/ would also be worth including as variants, if for no other reason than that these are attested for the period (Alexander Gil in fact notes every one of these variants) and are implicated in 16th-17th century rhyming (though not Shakespeare's rhymes). Or at least, they would be worth including if Crystal's guiding aim really were, as he claims, "to get as close as possible to the sound system that Shakespeare himself would have heard". 

Crystal's indication of word-stress in those words where the accent has changed since the 16th century is generally good and reliable. Of those that I have checked, he usually gives the original form correctly: revénue, siníster, délectable, útensil, pretéxt, illústrate etc.

The following accentuations, however, should be given but aren't:
— retínue as in "But óther of your ínsolent retínue..." (attested securely through to the 18th century)
— súccessor as in "chalks súccessors their way, nor call'd upon."
— turmóil (n)  as in "and there I'll rest as, after much turmóil."
(Crystal gives the verbal form turmóilèd, but túrmoil for the noun. The sole instance of the noun in Shakespeare suggests turmóil at the very least as a variant. A contrast between túrmoil as a noun and turmóil as a verb is attested from the 18th century on, but matters were earlier different.)

Now then


Our funerals are never really for the dead, but the consolation of the living. No matter that the living often believe otherwise as an article of faith. So too, no matter what rituals of literary piety we may perform to preserve the soul of Shakespeare's Tongue in an imagined eternity, a modern performance of Renaissance drama is ultimately not about the Renaissance. Nor should it be. Historical appreciation and literary appreciation are not necessarily the same thing.

Which is why it hardly matters that many of the posited features of "OP" were in Shakespeare's own time probably quite marginal whereas others (the complete merger of all /ɔi ui/ words with ME /ī/) probably didn't exist at all. It hardly matters that it turns out to be practically impossible to train an entire cast of Modern English-speaking actors within a reasonable amount of time to reliably reproduce a three-way contrast between peel/deal/pale, or even that so many OP performances continue to employ a typologically implausible phonology with the DEAL/PEEL vowels merged as /e:/. 

Yes, Crystal's "OP" dictionary is a work of poor philology. The introduction, as I've discussed, routinely relies on unsound reasoning, misrepresents evidence, selectively quotes primary sources to make them seem as if they mean something other than what they say, and sometimes seems to just be straight-up mistaken about what specialists in the field really think. The actual lexical entries are almost worthlessly unreliable as a guide to early 17th century phonology. And yet for all that, it does bring "us" closer to the English of Shakespeare, if not all the way there. It makes a lot of Shakespeare's soundplay more aurally perceptible, if not always using the same phonemes he would have heard around him.

The primary value, justification and interest of the OP movement is not an accurate reproduction of late 16th century London English. Rather, it is the production of a new accent of Modern English that modern actors can be trained to pronounce, and that modern audiences can enjoy associating with Shakespeare. On this score, the modern OP movement has succeeded marvelously.

You may rightly wonder, then, why I don't just chill out and....

Image result for shh let people enjoy things

The answer is: because it's best for OP-proponents to be honest with themselves and others about what Crystalline OP is, and what it isn't, about what it does, what it doesn't do. The fetish of authenticity can only shortchange the accomplishment. Let OP shine for what it really is. You don't need to believe in Santa Clause to enjoy the magic of Christmas. Otherwise, assholes like me will just keep ruining the fun by calling attention to the man behind the curtain.

Now I've said all that, you may well ask: why then do I bother making recordings using reconstructions that aim for high-res accuracy?

Well, shit, why shouldn't I? It is fun and awesome. Do I need any more reason than that?

And, as audience responses suggest, Modern OP is also really fun. In fact, I see no reason why its use should be limited to the staging of Renaissance drama and the reading of Renaissance texts. Why not write new modern works, perhaps set during the Renaissance or reacting to Renaissance literature, to be performed in OP?

Here, let me try.

A Shakespearean Language Sonnet

Now doth my tung with Tyrant Tyme debate
In bloodie sport upon a ruined stage,
To second thy tired tongue, and lash the State
Whose centuries do beseige the famined page.
As ancient blades worne blunt in shocke with ages,
Thy lines which vaunted deathlesse at decay
Do fall with lesser moment in the pages
Turned by the powres which turne the world to-daye,
Unlesse thy voice unto my vice repare
And, steeld afresh, match Time with force which flows
In change unchanging, as this yeares fresh heire
Of last yeares rose still bears the sweet name Rose.
To keepe my word to thee, my wordes I break,
That though the tongue be mine, thy voice yet speake.

Can you spot the wordplay depending on 16th century semantics? How about the wordplay depending on OP? Here's a transcription into Crystalline OP to help:

nǝʊ dǝθ mɪ tɒŋ wɪθ tǝɪrǝnt tǝɪm dɪbɛ:t
ɪn blɤdǝɪ spɔ:ɹt ǝpɒn ǝ ru:ɪnd stɛ:dʒ
tǝ sɛkǝnd ðɪ tǝɪɹd tɒŋ ǝn laʃ ðǝ stɛ:t
u:z sɛntrǝɪz du: bɪsi:dʒ ðǝ famɪnd pɛ:dʒ
ǝz ɔ:nsɪǝnt blɛ:dz wɔ:ɹn blɤnt in ʃɒk wɪð ɛ:dʒǝz
ðɪ lǝɪnz ʍɪtʃ vɔ:ntǝd dɛθlɪs ǝt dɪkɛ:
du: fɑ:l wɪθ lɛsǝɹ mɒmǝnt ɪn ðǝ pɛ:dʒǝz
tɐ:ɹnd bɪ ðǝ po:ɹs ðat tɐ:ɹn ðǝ wɒɹld tǝdɛ:
ɤnlɛs ðɪ vǝɪs ɤntu mɪ vǝɪs rɪpɛ:ɹ
ǝn sti:ld afrɛʃ matʃ tǝɪm wɪθ fo:ɹs ʍɪtʃ flo:z
ɪn tʃɛ:ndʒ ɤntʃɛ:ndʒɪn ǝz ðɪs ji:ɹz frɛʃ ɛ:ɹ
ǝ last ji:ɹz ro:z stɪl bɛ:ɹz ðǝ swi:t nɛ:m ro:z.
tǝ ki:p mɪ wɔ:ɹd tǝ ði:, mɪ wɔ:ɹdz ǝɪ brɛ:k
ðat ðo: ðǝ tɒŋ bɪ mǝɪn, ðǝɪ vǝɪs jɪt spɛ:k


  1. Thank you for not sparing us the minute details!

  2. The o ~ u variation seems to be entirely next to n and m. Isn't this just another instance of o being used to avoid too many consecutive mınıms?

    Did you hear Fiona Hill testify in Trump's impeachment trial? Her FACE vowel is am[ɛ̞ː]zing.

    In German, höh/See and über/lieber are not considered inexact rhymes.

    Well, if they hadn't been grandfathered in, they wouldn't be considered rhymes at all; not "trying and failing", which is how Shakespeare's inexact rhymes would be judged, but "not even trying".

    The reason they have been grandfathered in is that half the German dialects have lost the rounding contrast, and had already lost it 500 years ago. Up until the mid-20th century, lots of people who wrote in German couldn't even pronounce front rounded vowels; for them, this type of rhyme was exact. Nowadays the permission for ö to rhyme with e and for ü to rhyme with i is taught as an arbitrary convention – "of course they don't rhyme, but we've always acted as if they did anyway".

    The biggest single reason for the historical prestige of the pronunciation of Standard German in Hannover is that Low German, as a whole, has kept the rounding contrast just like the spelling has. Upper Saxon, for instance, has not.

    1. That is sort of my point. Compromise rhymes which are perfect for some but not all speakers can sneak in and be felt to be true rhymes thanks to tradition long after the type of pronunciation on which they were based has disappeared. E.g. in English evil/devil, love/prove. By the same token, something that is a phonetic full rhyme but runs counter to traditions about what "really" rhymes with what may be offputting. Like all those non-rhotic speakers who lost their minds at Keats for rhyming higher/Thalia ( = /θǝlɑɪǝ/) calling it a "cockney" rhyme even though it worked just fine in their own speech.

      "they wouldn't be considered rhymes at all; not "trying and failing", which is how Shakespeare's inexact rhymes would be judged, but "not even trying"."

      I'm not sure that is necessarily true. It would depend on various sociophonetic factors and accidents of history probably. Vowels far less similar than that can be treated as full rhymes. In Arabic verse /ī/ and /ū/ can interrhyme, for example.

    2. I don't really think there's any actual limit on how much phonetic dissimilarity can be smoothed over by relying on precedent like that. Take a look at Chinese rhyming systems as they have developed over the past millennium or so. It's pretty wild.

    3. "The o ~ u variation seems to be entirely next to n and m. Isn't this just another instance of o being used to avoid too many consecutive mınıms?"

      Largely, yes. The only exception I can come up with off the top of my head is "sullen" which may be spelled "sollein" or "sollen" by e.g. Spencer.

  3. I forgot to mention that -ook words still have [uː] in (parts of?) northern England.

    1. Yeah, shortening of -ook words seems to have been a development of the 19th century in many places.

  4. I cannot possibly express the depth of my utter unconcern for your boobies. Goddamn bots gumming up my comments sections. It's like whack-a-mole.

  5. Hi, A.Z. I was wondering if you've ever found any other modern scholar than Crystal who discusses Early Modern English phonology. I agree with some of your qualms but have always come up empty looking through the 21st-century literature. Any help would be great, thanks!

    1. A miscellaneous collection of the kind of stuff out there:

      Bjurman, M. (1977). "The phonology of Jacques Bellot's Le maistre d'escole Anglois (1580): Together with readings of the anonymous editions of 1625, 1647, 1652, 1657, 1670, 1679, and 1695." Stockholm: Almqvist & Wiksell International.

      Danielsson, B. (1955, 1963) "John Hart's Works on English Orthography and Pronunciation 1551, 1569, 1570." Stockholm

      Dobson, E.J. (1968) English pronunciation 1500-1700. 2 vols. Oxford: Clarendon Press.

      Lass, Roger (1999), Phonology and Morphology. In: Roger Lass (ed.) The Cambridge History of the English Language, Volume 3: 1476–1776. 56–186. Cambridge: Cambridge University Press.

      Cercignani, Fausto (1981) Shakespeare's Works and Elizabethan Pronunciation, Oxford, University Press (Clarendon Press).

      Smith, Jeremy J. "Sound Change and the History of English" Oxford 2006

    2. If you want something from last year, here's a recent article by an independent scholar which is IMO a very mixed bag. Important insights but important misfires


    3. Thanks so much! I really appreciate that! I'll look into those.

      I've also been thinking about the possible homophony of /s, z/ that comes up for a lot of EME writers of the type "is/amiss," "paradise/wise," "alas/was," etc. (definitely appearing in Shakespeare, Wyatt, and Marlowe). How seriously do you take those kinds of rhymes?

      More and more thoughts keep popping up for me. If you have any way for me to privately send you my email address, I'd love to chat further (and if you're not comfortable with that, no problem!). Thanks for all the fascinating analysis.

    4. Words like "is" and "was" (also "this", "both", "of" and many other words) originally had voiced and unvoiced forms for the final consonant. Thus "this" was pronounced "thizz" when before a vowel or a voiced consonant. You can actually see this allomorphy directly represented in the mid-16th century phonetician John Hart who transcribes "ðiz bụk" and "bọð ðe" for "this book" and "both the" in his specimens of connected speech.

      Eventually many of these words got "frozen" in either their voiced or unvoiced allomorphs, and there seems to have been a fair amount of inconsistency in the process. For Alexander Gil in 1617, "waz" was the general form, but he does transcribe "was" before voiceless consonants occasionally.

      Probably for Shakespeare, Wyatt, and Marlowe you can assume that a voiceless consonant in "is" and "was" was available as a real option in pronunciation. When poets in the late 17th century do it, though, matters probably stand otherwise.

      "Paradise/wise" I think should be understood as an inexact rhyme in Shakespeare. It's worth noting that the Paradise/Wise rhyme in Shakespeare is in an intentionally-sloppy sonnet composed by a character in a comedy. The sacrifice/wise rhyme occurs in a chorus passage that also rhymes dumb/run, soon/doom, live/relieve.

      Rhymes were not always meant to be exact. Spenser in a very solemn poem can still rhyme see'st/seek'st.

      Feel free to email me. My email address, disguised to hide from irritating bots, is: poemsintranslation AT gmail døtt k0m

  6. This comment has been removed by the author.

  7. This comment has been removed by the author.

  8. "His only witness to the merger is <veiage> for voyage."
    That may well be a continuation of ME viage /viːˈaːdʒə/ rather than being evidence for the merger, given that viage is the usual Middle English form; forms with -oy- are nearly nonexistent in ME. Modern voyage (1527) would then be a adjustment to the form that ended up prevailing in standard French; it may have been originally only orthographic, with the pronunciation in /vɔɪ-/ being a later spelling pronunciation.

  9. I have been trained in the Crystals' version of "original pronunciation" and find your evidence about its shortcomings compelling. I'm trying to learn something more accurate (and more reflective of the diversity of the language at the time), and would love to discuss with you what your research shows. Is this something you're comfortable helping me with?

    If so, I would appreciate if you could get in touch. I'm craig b daniel at g mail.

  10. I would like to join others in thanking you for this long piece. Although I am not a specialist of early Modern English phonology, I have found some of Crystal´s ideas about Shakespeare´s original pronunciation quite bizarre. I was happy to realize I am not alone in thinking so. I like very much your transcription of that sonnet at the end. Thank you once again.

  11. Hi Alex, You put so much effort into deciphering and scorning DC's phonology, why have you never contacted him to ask about his choices? And the whole profiting from OP either through performance or publishing is neglible afaik. I worked with the Passion in Practice ensemble and though the performances we played were unique, they were limited and not necessarily money makers. best William S.