Help talk:IPA/English/Archive 2

Archive 1

Archive 2

Archive 3

Archive 4

Archive 5

annoying gibberish

A moment ago I was seeking to find out the consensus pronunciation of "fungi", which I have variously heard pronounced "fun-gee", "fung-guy", or "funj-eye". I went to the proper page and find this nonsense in lieu of such readily understandable phonetic breakdowns as those I offered above. Anyone with the patience to learn this code, much less debate its finer points on this page, probably is a bit of a linguist and knows how to pronounce more words than the rest of us, who in turn(assuming my reaction will not be uncommon) will not benefit from your efforts... —Preceding unsigned comment added by 24.164.147.147 (talk) 06:09, 1 April 2008 (UTC)

If the reader is familiar with the IPA, he will likely find IPA pronunciations much clearer than the sort of respellings you mention above. If he's not familiar with the IPA, this page is here to help him understand it. If he neither knows the IPA at all nor wishes to learn anything about it, then this page will not help him. It still benefits the other two classes of users, however. Also, according to Wikipedia:Manual of Style (pronunciation), you are free to add respellings of the sort you mention above in addition to IPA transcriptions. Spacepotato (talk) 22:29, 1 April 2008 (UTC)

Although I totally sympathize with that feeling of "WTF?" upon seeing a bunch of IPA (especially if you try to sit down and read the main IPA article as a linguistic layperson), I must point out that *this particular implementation* of IPA, as shown on this help page, is just as simple as AHD's transcription system (which is my personal favorite). It is equally broad and phonemic. It has a comparable number of symbols. Just like with AHD's, a layperson can learn it, and then be familiar with it, once he/she has made the initial ramp-up of getting acquainted with it. — ¾-10 01:13, 2 April 2008 (UTC)

And then you're good for just about anywhere in the world. — kwami (talk) 02:18, 2 April 2008 (UTC)

Syllabification

"Three kinds of justification have been offered for the syllable. First, the syllable is a natural domain for the statement of many phonotactic constraints. Second, phonological rules are often more simply and insightfully expressed if they explicitly refer to the syllable. Finally, several phonological processes are best interpreted to ensure that the string of phonological segments is parsable into syllables." (Kenstowicz 1994)

If you're going to play at phonology, would it kill you to crack open a basic book on the subject? RandomCritic (talk) 12:55, 6 April 2008 (UTC)

Your quote from Kenstowicz is a complete non sequitur as a response to Kwami's comment, though. What are you getting at? —Angr ^{If you've written a quality article...} 13:00, 6 April 2008 (UTC)

That syllabification does matter (in English too!) and that if you, or he, bothered to study some phonology, you could find that out.

You're reputed to be the bright one in this crowd. Dazzle me with your proof of English "ambisyllabicity". RandomCritic (talk) 13:03, 6 April 2008 (UTC)

He didn't say syllabification doesn't "matter" in English, he said it is not distinctive (as in phonologically distinctive, making minimal pairs) in English, and indeed it isn't. Kenstowicz does not claim anything to the contrary. As for ambisyllabicity in English, read Daniel Kahn's dissertation and Clements and Keyser's CV Phonology for starters. —Angr ^{If you've written a quality article...} 13:38, 6 April 2008 (UTC)

Incorrect. While there may not be many precise minimal pairs in English, there are sufficient near-minimal pairs to show that syllabicity cannot always be predicted from the surface form. I personally distinguish [mɪˈstɛjk] "error" from [mɪsˈtʰɛjk] "take (understand) erroneously". Even those who do not make the distinction will still have [mɪsˈtʰɹiːt], [mɪsˈtʰajm], etc. Under such circumstances, it is impossible to conclude that syllable divisions do "not need to be indicated" in English. In fact they are frequently fundamental environments for determining phonetic outcomes. RandomCritic (talk) 12:19, 7 April 2008 (UTC)

That's a good point. Personally, I hear the difference as primarily one of vowel reduction: /mɨsteɪk/ vs. /mɪsteɪk/. I only get aspiration with emphatic prosody, motivated by the need to distinguish near homonyms, but you may be right here: this is almost like an internal word boundary. However, this kind of thing is only going to occur in morphologically complex examples, and would be relevant to very few entries in Wikipedia, where of course we can highlight the distinction. (I suppose you could argue it's not the syllable boundary which is distinctive, but the morpheme boundary.) It's very much like every vs. memory vs. mammary, where the actual numbers of syllables vary, or similar arguments that phonemes are not real / not adequate to completely describe English. However, the concept of 'phoneme' is good enough for a dictionary pronunciation guide, regardless of the merit of such disputes.

In simple lexical entries, indicating syllable boundaries has often lead to edit wars by people who hear things differently, which is why I'd prefer to avoid them, except perhaps in special cases such as mis-take. Can you give an example where non-predictable syllable boundaries are relevant to Classical names, which was your original contention? AFAIK, that depends on Latin syllabicity, which is not problematic except for /sC/ and /sCL/, and in those cases both /.sC/ and /s.C/ pronunciations are used, so it is still entirely predictable from the segments. kwami (talk) 18:32, 7 April 2008 (UTC)

I thought of a couple more examples before lunch today: "He was [əˈpʰɹaj.zɪŋ] the bureau about the progress of the [əpˈɹaj.zɪŋ]" and "Formerly employed by the railroads, the [ɛksˈpʰɔɹ.tɚ] had gone into the shipping trade and become an [ɛkˈspɔɹ.tɚ]. One can amuse oneself for an entire afternoon finding other examples with "ex-".

Differences in realization of the vowels are due to differences in the syllabification, rather than the other way around, creating (e.g.) open syllables instead of closed ones. RandomCritic (talk) 01:09, 8 April 2008 (UTC)

Although the syllabification results in the differing vowel and consonant allophony, what we pick up on is the allophony so that the contrasts given so far are not minimal contrasts. Opportunities for contrasts solely on syllabification with nonce words may be enlightening: ex-ray vs eck-sray, off-rum vs awe-from, mosh-room vs maw-shroom, etc. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 01:27, 8 April 2008 (UTC)

The above comment does not make sense. Of course minimal pairs contrast! If the implication is that the changes in sound quality are one thing, and the syllable structure is another, so that there are two different contrasts, that is an error. The surface contrasts are in (e.g.) vowel values; the underlying contrasts are in syllabification and stress. The examples given are indeed different in pronunciation -- but not because their segmental phonology is different. Obviously, in a language where metrical concerns (syllabification and stress) are so prominent as English, there are (much of the time) going to be phonetic variations dependent upon those metrical concerns. That is exactly the point, and is why indicating syllable structure is important. That is why I chose precisely the examples I did: because they obviously manifest superficial phonetic contrasts, while having phonemic structures that are or might be (depending on dialect) identical except for the syllable structure. Quod erat demonstrandum.RandomCritic (talk) 22:40, 13 April 2008 (UTC)

Sure, but all are cases where we're emphasizing a morphological boundary in a nonce word for the purpose of disambiguation. Elementary phonology texts often comment that the stress patterns of compound words differ from those of other words; this is a similar phenomenon, and a minor issue that can be brought up on the rare occasion it is relevant. Hardly reason, in my opinion, to mark syllable boundaries for every word, especially when people then start arguing about them.

It's hard for me to comment on RC's examples, because in my dialect they differ in stress. But per the OED, and as Ƶ§œš¹ said, the first pair is /əpraɪzɪŋ/ vs. /ʌpraɪzɪŋ/. kwami (talk)

Denying the existence of data isn't a viable strategy. The OED represents one dialect. There certainly are dialects -- including, I would guess, those spoken by present interlocutors -- in which the [ə] vs. [ʌ] contrast is nonexistent. It is at any rate certainly not the case that the aspiration of the /p/ is dependent upon the quality of the preceding vowel. RandomCritic (talk) 22:40, 13 April 2008 (UTC)

Would you cut it out with the straw men? It's really annoying. I didn't deny any evidence, I merely stated that your example was hard for me to judge, because in my speech there's something different going on: The stress differs, and the aspiration depends on the stress. kwami (talk) 00:05, 14 April 2008 (UTC)

Let's get back to the RC's original point: He had complained about not having syllable marks in Classical names, such as those for surface features of Mars, claiming that they were necessary for correct pronunciation. He has yet to give us a single example. kwami (talk) 05:57, 8 April 2008 (UTC)

I'll return to the main point, which is otherwise likely to be lost in a maze of irrelevant minutiæ. Angr's question about minimal pairs depending on syllabification is not really on point (even though I have shown that such pairs exist) because the existence of minimal pairs is a matter of chance. What is relevant is that syllable structure conditions a host of phonetic phænomena, which could not be predicted without it.

I list some of the metrical-dependent phænomena existing in English dialects:

• r-deletion. In non-rhotic dialects, /r/ is deleted (often with compensatory lengthening, or other vowel changes) in a syllabic coda. [N.b., for those unfamiliar with the terminology, every syllable can be considered to consist of an onset, a nucleus, and a coda. The nucleus is (usually) the vowel at the heart of the syllable; the onset is what precedes it; the coda is what follows.] The rule is /r/ -> Ø / __C₀$ (read: /r/ is deleted before any number of consonants (including none) before a syllable boundary). The rule is not /r/ -> Ø / V___ or even /r/ -> Ø / V___C!!! If that were the case, then /r/ would be deleted, in the first case, even in words like /stɑrɪ/, or in the second case, would fail to be deleted in words like /fɑr/. What the deleted /r/s in words like /fɑr/, /fɑrm/ and /ˈpɑr.tɪ/ have in common is that they are all in the syllabic coda. Contrast /stɑrɪ/, which is necessarily syllabified /stɑ.rɪ/ and not /stɑr.ɪ/, where the /r/ is retained because it is in the onset.
• aspiration. This one is well known and needs little elaboration: voiceless stops are aspirated at the beginning of the onset of a stressed syllable. This explains the difference in aspiration in [dɪstʰɛjst] vs. [dɪstɝb]; the first is /dɪs.ˈtest/, the second /dɪ.ˈstʌrb/.
• glottalization. /t/ is, in many dialects, converted to [ʔ] in coda position. Thus we have for /awtredʒ/ the realization [æwʔɹɛjdʒ], and likewise /awtredʒəs/ -> [æwʔɹɛjdʒəs]. But compare /mætrɛs/ -> [mætɹ̥əs] and /ətroʃəs/ -> [ətɹ̥ɔwʃəs] and you see that in the same V__r environment, the /t/ is not glottalized. Why not? That's right, because it's not in coda position. Without a representation of syllabification, there is no way to predict where glottalization would occur. But if you represent the phonemic structure thus: /ˈawt.redʒ/ vs./ˈmæ.trɛs/, /awt.ˈre.dʒəs/ vs. /ə.ˈtro.ʃəs/, it's plain as punch.
• affrication: In many dialects, /t/ and /d/ undergo affrication to [tʃ] and [dʒ] before /j/ -- but only when both are in the same syllable. (I.e., C[-sonorant][-continuant][+coronal][-nasal] -> [-anterior][+strident]/$__j, usually with consequent deletion of the /j/; in many dialects this is limited to unstressed syllables.) So /ˈstæ.tju/ can be realized as [ˈstætʃu] but /ˈbot.jɑrd/ will never be **[ˈbɔwtʃɑrd].
• affrication 2, rhotic boogaloo: In some dialects, affrication occurs also between coronal oral stops and /r/. Thus /tren/ becomes [tʃɹ̥ɛjn] (and /strit/ may become [ʃtʃɹ̥iːt]) and /dræg/ is [dʒɹæg]. But once again, the /tr/ or /dr/ must be in the onset. /ˈmæ.trɛs/ may become [ˈmæ.tʃɹ̥əs] and /tɔ.drɪ/ -> [tɒːdʒɹi], but /ˈawt.redʒ/ is never **[ˈæwtʃɹ̥ɛjdʒ] and /ˈhɛd.rɛst/ is never [ˈhɛdʒɹɛst]. Again, these distinctions cannot be deduced from the phonemic transcription unless there is some representation of syllabic divisions.
• liquid devoicing. As shown in several examples above, /r/ loses voice following a voiceless stop in the same syllable. The same is generally true of /l/, as in /æ.ˈplaj/ (voiceless) vs. /ʌp.ˈlɪf.tɪŋ/ (voiced), /ˈtrɪ.plɛt/ (voiceless) vs. /ˈʍɪp.læʃ/ (voiced).
• vowel reduction. Vowels are often reduced to a greater extent, or with greater frequency, in closed syllables rather than in open ones. Compare /æn.ˈtɛ.nə/ -> [ænˈtʰɛnə] vs. /æ.ˈstɒ.nɪʃ/ -> [əˈstɑnɪʃ]. In both cases the underlying initial vowel is followed by two consonants, but in the first case its full value is maintained, and in the second it is (in some dialects) reduced.
• glottal insertion. A glottal stop is in some dialects inserted before an internal syllable beginning with a vowel, e.g. /ʌp.ˈɛnd/ -> [əp.ˈʔɛnd] "upend"; contrast /æ.ˈpɛnd/ -> [ə.ˈpʰɛnd] "append"; likewise /ɛks.ˈæk.tər/ -> [ɛks.ˈʔæk.tɚ] "ex-actor" and /ɛg.ˈzæk.tər/ -> [ɪg.ˈzæk.tɚ] "exacter".

These are not rare or bizarre occurrences. They are intrinsic parts of an English language whose phonemic structure, prior to the application of phonological rules, includes syllable-structure and stress. Nor is the basis for syllabification a mystery. In general, as the examples above show, words in English are syllabified according to the principle of maximal onset -- in layman's terms, you pack as much into the onset of the syllable as you could start a word with. So the example eck-sray [ɛk.srɛj] is not even possible; [sr] is not a potential onset to an English word. The exceptions are cases where (1) there are multiple licit onsets possible (general constraints on possible onsets and codas always apply; e.g. "extra" cannot be syllabified either **/ˈɛ.kstrə/ or **/ˈɛkstr.ə/); (2) there is a perception that a word is a compound is composed of one or more elements that could stand alone as words. This perception may not correspond to historical facts; e.g., "outrage" is not actually a compound of "out" and "rage", but is treated as such, while mistake is a compound of "mis-" and "take", but not all speakers feel it to be so. Sometimes there is reanalysis of a compound; for instance, "painstaking", originally a simple compound "pains+taking" /ˈpenz.ˌte.kɪŋ/ has been frequently reanalyzed as "pain+staking" /ˈpen.ˌste.kɪŋ/. The surface pronunciations differ accordingly in the aspiration of the /t/; but the difference is not the result of the difference between /s/ and /z/ -- the sibilant is underspecified for voice, and derives its voice from whether it is perceived to be in the coda of the first syllable, in which case it agrees with /n/ in voice, or in the onset of the second syllable, in which case it agrees with /t/. Both voicing and aspiration result from the underlying syllable structure, and not vice versa.

Syllable structure is therefore part of the lexical information available about a word ("lexical" referring not to print dictionaries, but to the stored information, not derivable from other information, that each speaker has about his or her language.) There is a good deal of unity, even cross-dialectally, about what English syllable structure is. However, in determining syllable boundaries, recourse should be had to linguistic facts that derived from particular syllable structures (like those listed above), and in absence of those, to the general principles that those facts reveal (e.g., maximization of the onset). Dictionaries are notably bad with syllables, mostly because they have other issues (e.g. spelling) to worry about, or because, due to an old bad lexicographic habit, they have become accustomed to manipulating syllable boundaries to suggest particular vowel qualities. The usual culprit is the use of closed syllable structure to imply "short vowels", e.g. "ˈtrip-lət" for /trɪ.plɛt/, using the closed syllable of "trip" to remind the reader of the lax vowel /ɪ/ -- though the devoicing of the /l/ indicates that the onset of the second syllable is really /pl/. While it's true that lax vowels are or were preferred in closed syllables (particularly in words of Latin or Latinate origin), and they generally don't appear in open monosyllables, there is absolutely no bar to the lax vowels æ, ɛ, ɪ etc. appearing as the nuclei of stressed open syllables internally. Such syllables are in fact very common in English.

RandomCritic (talk) 22:40, 13 April 2008 (UTC)

Not so sure about 'as plain as punch'. R-deletion is syllable final—fine, but in all of your rhotic examples the syllable boundary is predictable. You may wish to posit a phonemic syllable boundary, but that's a theoretical claim. I can posit other mechanisms for many of your other observations. For example, in outrage vs. mattress the glottalization can be chalked up to the following unreduced vowel (which is phonemic regardless of syllabification); in statue vs. boat yard, apply vs. up-lifting, tawdry vs. head-rest, up-end vs. append, and ex-actor vs. exacter you are once again relying on morphological boundaries; in antenna vs. astonish the syllabic boundary is predictable; in pains.taking vs. pain.staking it is a theoretical claim that the voicing depends on syllabification rather than vice versa [and anyway, couldn't I use this example to argue that the difference is allophonic?]. The examples I find most convincing are distaste vs. disturb and triplet vs. whiplash, though of course they can both be argued to be due to morphological boundaries. Your point about syllabification depending on perceived morphology is a good one; we're getting into Joan Bybee here. There is a lot of "subphonemic" detail such as aspiration in people's lexicons, which is one reason why generative phonology and the concept of an absolute phoneme do not work. Regardless, I'd like to see a good reference demonstrating how syllable boundaries are phonemic, else we're veering into OR. In practice, however, this isn't much of an issue. In many of your examples, the placement of the stress mark forces the issue, and for the rest, if it's actually needed, we can always add syllabification marks. As you admit, most of the time they would not be needed even if we bought all your arguments, so it would be still redundant to add them most of the time. Nearly all the time, in fact, because the kinds of examples you've given us are characteristic of daily vocabulary, which we're not likely to transcribe in an encyclopedia anyway. Once again, this came up because you claim that marking syllable boundaries is required to properly indicate the pronunciation of Classical names. Once again, I ask you: Can you give us an example of a Classical name where the pronunciation cannot be properly conveyed without syllabification? —kwami (talk) 00:05, 14 April 2008 (UTC)

Also, back to one of Ƶ§œš¹'s points, there's the problem of ambisyllabic consonants. This isn't necessarily phonetic detail; many phonological approaches require determining ambisyllabicity before assigning aspiration, for example. (Syllable-initial plosives are aspirated except when ambisyllabic, as in happy; you might want to put stress into the equation, but that just makes things more complicated for no good reason, and Kahn for example argues that positing ambisyllabicity is preferable to trying to work without it. Okay, Kahn's rather old, but since this is also intuitive to native speakers, we do get into edit wars over this.) —kwami (talk) 00:22, 14 April 2008 (UTC)

(edit conflict) My point above was that your examples don't show that we as speakers make phonemic contrasts between /V.CCV/ and /VC.CV/. Your further examples are a lot more compelling (though we do have words sith /sr/ in the coda onset: Sri Lanka and Sranan) and, upon further reflection, the logic I was trying to use above would imply that we don't make phonemic contrasts between /t/ and /d/ because plato and play-doh are homonyms. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]}

Detailed critique

Introduction

I haven't had the time previously to follow up on my comments; I don't have unlimited time to play around on Wikipedia, and I would have been perfectly content to allow this page to take whatever form its editors chose if it were not for the fact that one of its editors, in an excess of misguided zeal (at the very least) has chosen to repeatedly engage in an edit war against certain articles which I created, in an effort to impose his own vision of a phonetic transcription on those articles. Although he repeatedly cites as the basis for his edits the Wikipedia:Manual of Style, neither that page nor Wikipedia:Manual of Style (pronunciation) provides any justification for his alterations. I am therefore presenting my response on this page, since it is the only one that actually encodes his scheme, and because this is not really a matter of "style", but one of substance, engaging the facts of language, which are not alterable by editorial fiat.

As the readers of this page will doubtless recall, not long ago I presented a number of basic objections to the entire concept of creating a "Wikistandard" for the IPA representation of English. Those objections remain cogent, and I do not retreat from them. However, if one accepts that some such "standard" is going to be presented, the scheme presented in this Help page is, frankly, not at all good, for reasons gone into below; and it is made far, far worse by being both misrepresented and misapplied.

The responses to my critique of this page -- admittedly one that, due to time constraints, I could not present in detail -- were frankly extremely disappointing. I don't mean that with regard to the outcome, but rather with regard to the quality of the arguments presented and the attempted rebuttals. I realize that few if any of the contributors to this page have ever taken advanced linguistics courses, much less taught them; so I will try to match the argument to the audience. Nonetheless, I cannot teach even an elementary course in phonology (and much beside) in one posting, and I can only suggest that if you find the arguments below hard to follow, that you ask me about it on my talk page -- surely a much better approach than pretending that the arguments here made do not exist or can be dismissed by handwaving. Failure to understand an argument does not prove that the argument is irrelevant.

It will doubtless be claimed, as it already has been, that I am "alone" in my criticisms, and that for that reason they should be disregarded. I do not know if I am "alone" or not among the four or so editors of this page; I dare say a couple of them will object simply out of an aversion to seeing their pet project criticized. I should point out, however, that several of the better-informed comments above make points similar or identical to the ones I am going to make. This is a reflection of the common state of linguistic study today. Those arguments were, unfairly, ignored; but I really do not see why advances in linguistic science should pass this page by. It is no longer 1950 or even 1980.

Cyclical argument

As far as the representation of the scheme goes, there is a vicious cycle of arguments, each one false on its own merits, but each of which provides a refuge from the refutation of the preceding argument. It goes somewhat as follows:

(1): The phonetic scheme is stated to be "pan-dialectal", "dialect-neutral", or some similar phrase, meaning that it is renderable into any of the several dialects into which English has separated over the last several centuries, or at least into the "major" ones. (It is not clear who decides what a "major dialect" is, or why, or on what grounds: linguistic? political? -- but of that more later.)

(1a): It is responded, that the symbols, as pronounced, actually represent a very particular and peculiar dialect of English (one which, to the best of my knowledge, does not actually exist).

(2): It is argued, as if in refutation of (1a) above, that the symbols are "phonemic", and not phonetic, and that, therefore, although a straightforward phonetic interpretation of the symbols, as used in the International Phonetic Alphabet might seem odd, in their use as phonemic symbols, they might represent a variety of phones.

(2a): It is responded, that in the present scheme, for various reasons (to be detailed below), both the symbols used, and the way they are applied, are characteristic of a phonetic and not a phonemic representation: that is, the symbolism encodes sub-phonemic, conditioned variation which one does not expect in a representation of phonemes. It is also pointed out that there is an inherent and ultimately unresolvable tension, or rather contradiction, between a phonemic representation, and a "pan-dialectal" representation -- or even a multidialectal one, except perhaps for very closely related dialects. It may be noted that in its divergences from phonemic simplicity, the scheme also diverges from its "pan-dialectal" pretensions!

(3): It is said that whether or not it is truly phonemic, the scheme is "standard" -- an appeal to authority, nominally backed up by certain dictionaries.

(3a): It is pointed out that the scheme is in fact not standard, nor backed by any particular authority (nor is the quality of the supposed authority considered), but is a bizarre, hybrid concoction unique to Wikipedia.

(1): It is then said that such hybridism is necessary to be "pan-dialectal" -- and so we return to our starting point.

In fact, all of the claims made, for the scheme being "pan-dialectal", "phonemic", or "standard" are utterly false. With regard to the first, it would be extremely difficult and perhaps impossible to create a single notation for English that would cover every existing dialect, some of which are very divergent. Even for the most similar dialects, there are notable complications which have not really been considered: for instance, changes in the positioning of stress, the deletion or epenthesis of phones.

With regard to the second, phonemic analyses are performed on particular dialects. Phonemic analyses look for the basic structures that "underlie" an observed group of phones (i.e., actually produced speech sounds). The idea is that several different sounds that are produced in speech may reflect an "ideal" sound that possesses a unity, in terms of the overall sound-structure (aka the "phonology") of the language, but which is pronounced slightly, or even very, different in different phonetic environments, e.g: at the beginning of a syllable, at the end of a syllable, word-initially, word-finally, after a consonant, between vowels, in a stressed syllable, in an unstressed syllable, in an open syllable, in a closed syllable, et cetera. English is particularly prone to variations, both of consonants and vowels, related to syllable structure and stress.

This variation is quite different from the variation produced by dialectal splits. Dialectal variation is historical; it is produced by alterations of sound which appear a language over time. These possess their own regularity, and in some cases there is an overlap between a diachronic (historical) and a synchronic (of one time) analysis; for instance, an originally phonetic difference may become phonemic over time, due to other alterations in the phonological structure of a language, but can still be largely accounted for by a distributional analysis.

As for "standards", it should be pointed out that dictionary-makers, though sometimes aided by linguists, generally do not follow linguistic norms. They present a brief and sometimes tendentious snapshot of the language, influenced by rather parochial and indeed idiosyncratic considerations. Often somewhat bizarre pronunciations, of very limited usage (temporally or geographically) get put into print as "standard". The dictionary pronunciations are typically neither phonetic nor phonemic; they are "phonemic" to the extent that more or less common phonetic variations are ignored (a notorious example is the general omission, in American dictionaries, of the "flapping" of intervocalic /t/ and /d/, as in "catty" and "caddy", though for many speakers this actually effects a neutralization which renders the determination of the underlying sound, without reference to either historical or morphological data, impossible.

Claimed and actual sources

In this connection it is worth saying something about the sources of the scheme used in the Help:Pronunciation page. It has been claimed that this is the "Random House Dictionary", with slight variations, though, very inappropriately, this information is not included on the page itself, which of course makes checking the claim rather difficult! The claim, however, is thoroughly false.

Consulting the "Random House Dictionary" to see who was responsible for the supposed "phonemic" system on the Help:Pronunciation page, I discovered that in fact this is not the Random House system at all. Random House uses its own idiosyncratic system, with non-IPA symbols, and which is -- if not completely phonemic -- at least more phonemic than the Help page scheme.

The introduction to the Random House Dictionary does describe its system in terms of IPA. However, for its various symbols, it (quite properly) offers several different phonetic interpretations, and does not call any of them "phonemic", and does not pretend that its list is exhaustive. There is therefore no authoritative way of converting Random House's system into IPA.

It may be objected that such a conversion occurs at [1], where you can click on a link to get the "IPA pronunciation" for an entry "based on the Random House Unabridged Dictionary".

Who is responsible for the conversions, I do not know. They appear to me to have been done in a mechanical and, I think, not very intelligent manner. However, the following things need to be noted about them:

First, despite the mistake of using slanted brackets, these IPA representations are in fact phonetic, not phonemic; they use one phonetic selection out of the many offered in the actual Random House Dictionary.

Second, the Help:Pronunciation scheme is still not identical to the dictionary.com version. For instance, the dictionary.com representation of "chorus" is either kɔrəs or koʊrəs. At Wikipedia, however, under the present régime we are supposedly to be forced to write it kɔərəs, a bizarre, unintelligible spelling, without the slightest justification: phonetic, phonemic, historical, or "standardized". This is the kind of thing that I have been complaining about, with Wikipedia creating its own artificial pseudo-standard. It's unacceptable to just make something up and then try to pass it off as if it were based on some kind of competent linguistic authority.

Problems in creating a standard

Linguists who write about English phonology don't have recourse to any ready-made "standard". Rather, they are aware of several different approaches to phonemically representing English, and they will use the one most suitable to their particular purpose (which does not invalidate others) modified by their own research and analysis. For this reason, in several different books and articles by equally competent linguists, you may find a different phonemic representation in each one. Some of this is fairly cosmetic;, but other variations may reflect a different analysis of the number, underlying quality, or relations of the various phonemes in the dialect under consideration.

Considering this, and the other reasons given above, it would be a very bold person who would consider trying to manufacture a "standard" representation of English for Wikipedia or any other publication. Such a person would have to be skilled in phonological analysis, aware of the history of English pronunciation for at least the past half-millennium, and be acquainted with a representative variety of English dialects.

I know perfectly well that none of the editors that have worked on this scheme have that kind of background, and it does show. I do have that kind of background, though I wouldn't claim to be familiar with above a dozen or so dialects of English spoken in the United States, Canada, Ireland, the U.K., Australia, New Zealand, and elsewhere.

I do not think that manufacturing a standard is a particularly desireable or necessary thing to do. I think the best thing to do would be to admit that a variety of IPA transcriptions is always going to be present, and suggest ways to get from the IPA to the spoken pronunciation. I definitely would not want to come up with a scheme and then find that some editor was using it as an excuse to bring everything on Wikipedia into conformity with it; that reflects not only a misunderstanding of the nature of transcription, but is also simply rude and shortsighted. Who really wants to be responsible for changing every pronunciation on Wikipedia every time an alteration is made in the pronunciation key?

However, I am also aware that I'm better aware of the nature of the problem than most of the editors who have already worked on it, and so, with all proper diffidence owed to the its intrinsic insolubility, let me offer the following suggestions, first on methodology, then on the particular application of those principles.

A Better "Standard"

I. Phonology

It is indeed desirable to keep a practical system of transcription as phonemic as possible, for several reasons. First, many phonetic variations in pronunciation, though audible enough, are simply not salient to the ear of a native speaker. The differences between aspirated or unaspirated consonants, between nasal and non-nasal vowels, or between palatalized and non-palatalized velars are both thoroughly predictable and not even noticed by most speakers.

But even more relevant to the practical side of things is the fact that phonemic notation is just simpler. For the ordinary reader, learning several new symbols is already a strain on the memory. Reducing that strain, and rendering the the system relatively painless and intuitive, are laudable goals. So fewer symbols should be used; and those symbols should be, where possible, familiar ones (for this reason, and this reason alone, is it possible to tolerate /r/ where on purely phonological grounds we should have /ɹ/ -- there is no other good reason). Excess and unnecessary symbols should be pruned.

II. Dialects

On the other hand, covering as many dialects as possible -- at least of variant pronunciations that can be applied to written Standard English -- is also desirable. The introduction to the Help page says that it does not cover certain distinctions which occur in Australian English and "Scottish English" (I suppose SSE, Scottish Standard English, is meant). Why not? The question is merely one of providing sufficient phonemes to cover all cases. In the case of SSE, at least, no new phonemes need be employed at all. Australian and SSE are certainly major English dialects by any intelligible use of the term.

III. Synthesis

There are thus two opposed tendencies at work: toward simplicity, in the case of phonemic notation, and toward complication, in the case of covering dialect variation. The appropriate course to take is to start with a phonemic analysis of as many dialects as are to be covered; allow the phonemes to overlap as much as possible; and where there is a discrepancy, due to either split or (more commonly) merger, use both.

The question then arises: what happens when two dialects differ with respect to the quality of a phoneme that they otherwise agree on the unity of? The answer must be "choose one" -- though the choice may be conditioned by factors such as the availability of sufficient distinctive symbols (a resort to minute diacritics is to be avoided if possible), and perhaps even by the history of the sounds in question.

For instance: English has a large number of low vowel sounds. The actual values of these vowels varies by dialect. In GA the vowels in "hat", "bath", "hot", and "caught" are either [æ], [æ], [ɑ] and [ɒ], or else the last two are merged as either [ɑ] or[ɒ]. In SSE the first two are [a], the last two [ɔ]. In RP, however, they are [æ], [ɑː], [ɒ] and [ɔː]. RP has more phonemes -- 4 rather than 3 or 2 -- and it might seem that this is therefore the model to be followed.

However, historically, the 18th-century ancestor of both GA and RP (and many other English dialects) used a different four-vowel system, for which the values (and this is well-attested in contemporary phonetic notations and commentaries) were [æ], [aː], [ɑ] and [ɒː]; RP has backed and/or raised the last three vowels. There is as much ground for using pre-dialectal [æ], [aː], [ɑ] and [ɒː] to represent the four phonemes as there is for using dialectal [æ], [ɑː], [ɒ] and [ɔː], except for one detail. Pre-dialectal English also had a phoneme [ɔ], quite distinct in quality as well as length from [ɒː]. In fact, it mostly appeared as an unstressed variant of [oː]; but it also (more rarely) appeared in stressed position, as in [ɔnlɪ], contrasting with [loːnlɪ]. Dialects still exist today in which the vowels in "only" and "lonely" are distinct, and if this were an essential detail to keep, there would be a case for using all of the symbols æ a ɑ ɒ ɔ. It could be argued that this ɔ is too marginal to make a fuss about, but the point is that our choice of symbol can be forced by the number of symbols we choose to pack into the system.

Application

Now, with regard to the actual application of this methodology, I refer first to the concepts of fewer symbols and more familiar symbols.

YES! — [Unsigned comment added by 140.233.137.169 (talk • contribs).]

Long vowels and diphthongs

I would start by abolishing the diacritic [ː], except for extreme cases where the sounds can be distinguished no other way. I don't deny that length distinctions do exist in English, certainly in RP; however, they are not phonemically distinctive; there is no word where there is a distinction in meaning dependent upon the contrast between [iː] and [i] for instance, and whether one says [siːt] or [sit], the word will always be recognized as "seat" and never confounded with "sit". Moreover, the length is largely dependent on stress; the /i/ in [iˈvɪns] "evince" is certainly shorter than the /i/ in [rɪˈviːl] "reveal", but it would be bad phonology to consider them as two phonemes. Linguists generally reserve the use of the diacritic [ː] for cases where (as in Finnish) minimal pairs distinguished by contrastive length can be found. Discounting the dialectal cases where we have compensatory lengthening for a deleted /r/ (irrelevant in a quasi-phonemic notation where the r's will be marked anyway), such contrasts don't exist in English.

Second, I would simplify the diphthongs now marked as [eɪ] and [oʊ]. For RP and especially GA, these notations are traditional but inexact. In pre-dialectal English and in many modern dialects, their values were and are [eː] and [oː]. In line with my suggestion about abolishing the ː, I would use /e/ and /o/, symbols which are used today by some linguists. It also solves the problem of dealing with the off-glide deletion which frequently occurs (e.g. before liquids, or in unstressed syllables) and can create ugly and counter-intuitive sequences like /moʊr/ for "more", where /mor/ is visually unexceptionable. Phonologically, it's more than slightly arguable that the process is one of off-glide insertion, rather than off-glide deletion.

[aɪ], [aʊ], and [ɔɪ] are today used indifferently with [ai], [au], [ɔi] or, more commonly, [aj], [aw], [ɔw]. I prefer the last, as it avoids the suggestion that the final vowel of the diphthong is or could be syllabic; otherwise it might be hard to distinguish the pronunciation of "dais" and "dice". The value of the first vowel is highly variable among the dialects, e.g. [aj] being articulated anywhere from [ɛj] to [ɒj]; as [a] is a reasonably central vowel, and traditional, it does no harm to continue using it. If it were not in direct conflict with the standard usage of IPA, I would suggest using [y] for [j] both as a consonant and in diphthongs, as it is more recognizable. Some do. However, I think we are stuck with [j].

"Rhotics"

The use of [ɪər], [ɛər], [ɔər], [ʊər] is an offence against both simplicity and phonology -- it is, in fact, the most glaringly obvious evidence that the Help:Pronunciation scheme is not by any stretch of the imagination phonemic. As everyone knows, these are nothing more than dialectal phonetic reflexes of /ir/, /er/, /or/, /ur/ -- this is true both in phonological analysis and historically. In RP, there is a case (not, I think, a very good one, but a case nonetheless) to be made that [ɪə], [ɛə], and [ʊə] are phonemic, because of the rather dubious status of syllable-final /r/ in RP. Nonetheless, any "pan-dialectal" representation must necessarily be "rhotic", because otherwise many important distinctions preserved in many dialects would be lost (the distinction between "pass" and "parse", or "fort" and "fought", for instance). As long as the r's are present in the representation, the [ə] is totally superfluous; in RP the [ə] substitutes for the original /r/, it does not exist beside it. And if the idea is to hybridize the RP and GA pronunciations (which is not an acceptable phonological procedure, though invoked as justification on the Talk page) then by the same logic, we aught to see /pɑ ːrs/ and /fɔːrt/. If the argument is simply that in some' dialects the [ə] appears, even before /r/, the same is true of /l/, but it is not suggested that "seal" be written /sɪəl/ or /siəl/.

The use of [ɔər], besides being faulty in itself, suffers from encoding a dialectal merger. The aforementioned editor who keeps altering my transcriptions apparently cannot distinguish the first syllables in "torus" and "Taurus", and so indiscriminately represents both barbarically as /tɔərəs/ [sic]. In fact, in the current system, there is no good way to distinguish the two, although the distinction is old and still maintained in many dialects. According to my suggestions, the two would be distinguished as /torʌs/ and /tɔrʌs/. I should point out, before anyone does it, that to condemn as parochial dialectalism the preservation of a distinction which it is the entire point of a "pan-dialectal" system to preserve is to undercut the entire concept of this page. The fact that a distinction may not exist in your personal dialect does not mean it is unimportant. In light of my care for these distinctions, it is quite an Orwellian experience to see it proclaimed that "He prefers to go at it the other way, and indicate only those vowels which are distinct in his speech, subsuming all dialects into his. That violates our commitment to neutrality." This is precisely the reverse of the truth; it is the current Help page that represents a non-neutral, dialectal form of English, even if the dialect is an artificial one.

If, for the lax vowels, we use /ɪ ɛ æ ɒ ʌ/ then there is no bar to using /ɪr ɛr ær ɒr ʌr/ for the "rhoticized" variants, either over a syllable boundary (where the vowel quality is generally preserved) or within a single syllable (where the vowel quality is altered to a greater or lesser degree in some dialects). Those who merge /ɪr ɛr ʌr/ will have no difficulty in pronouncing them all as [ɝ] (if that is indeed their pronunciation). There is no need to postulate a phoneme /ɜ/. I realize that /ær/ for the ar in "card" would be a little surprising, although some dialects (e.g. in parts of Ireland and Canada) use exactly that pronunciation. However, if /ɑ/ is to be used for the "a" in "calm" or (British) "bath", then /kɑrd/ should be acceptable. /kɒrn/ for "corn" should not be too shocking to anyone's senses. At any rate, it does not do to suggest that in all dialects the vowels in the last syllables of "centaur" and "mentor" are identical, when they are not -- neither phonologically, historically, nor in the pronunciations of many dialects.

Other simplifications

For the reduced vowels, there is absolutely no need for a phoneme */ɨ/. This is purely and simply identical to /ɪ/. If [ə], or a variant thereof, appears in its place in any word, that is simply the result of phonological rules -- namely, the centralization of open unstressed /ɪ/.

In the actual application of the current scheme, /ə/ generally appears far too much. Usually it appears as the replacement for some other vowel in an unstressed syllable; application can be made to other dialects for the original value, though frequently that can also be determined from spelling. It notably should be distinguished from /ʌ/. There is a difference in pronunciation, in at least some dialects, between the final syllables of "shogun" and "slogan". This difference is obscured if they are both transcribed ending in /gən/. Once again, just because this distinction may not be made in your personal dialect does not mean it is unimportant.

Classical names

The matter of the dispute between myself and the aforementioned editor focuses, not on ordinary English words but on Anglicized Latin/Latinized Greek names. These names are to some extent extraordinary in terms of their phonology and their relation to the rest of English. They are not, as some believe, pronounced haphazardly, but are based on a system, to some extent artificial but founded upon medieval French pronunciations of Latin, that evolved in English schools, reaching its present form in the 18th century, and continuing almost unchanged down to the early 20th century. The rules relating the spelling to the phonemic structure are very exact and only rarely admit of variation. For many of the names in question, some of which are extremely abstruse, no dictionary or encylopedia will provide a "pronunciation key"; recourse must necessarily be had to the fundamental principles of the system.

As this system has been in desuetude since the early 20th century -- a little less than 100 years -- it is not terribly well known (as a system), though it is recorded and can be researched and compared (with large agreement) with those words and names whose traditional pronunciation remains well-known. Unfortunately, some people tend to substitute their "best guess" for the correct pronunciation, often influenced by other approaches for pronouncing Latin. The system is quite complex, though rigorous, involving an assessment of stress, syllable boundaries, neighboring sounds, and position within a word, and it would take an essay longer than the one written here to explain it. I happen to have studied it, using original Anglo-Latin texts, with pronunciations phonetically indicated, dating from as far back as the time of Shakespeare. My would-be "corrector" has not; which is one of the things that makes dealing with his zeal so irritating, as he does not merely restrict himself to altering the symbolism of the pronunciations arrived at with such care, but also deletes the (crucially important) syllable boundaries and randomly changes the value of vowels, mostly in the direction of slurring distinctions in words that were intended to be spoken with great clarity and exactness. I confess that I have been very short with him, and with some of the other editors taking his side; I regret any offense caused, but I trust the cause of the shortness has been made quite apparent.

Conclusion

In conclusion, I would suggest the incorporation of all of the above suggestions into this page. They may suffer from certain imperfections, but they are as nothing compared to the imperfections codified in this page as it stands. I do not believe that I, or any other editor on Wikipedia, is under the slightest obligation to adhere to a "standard" so flawed and faulty; and I cannot see the slightest reason to try to impose such a standard on them. RandomCritic (talk) 12:08, 6 April 2008 (UTC)

Discussion

Ƶ§œš¹

There are a few things I'd like to point out.

• From what I understand, RandomCritic is correct that the schwas preceding /r/ are conditioned variants in rhotic dialects. The present scheme uses the schwas so that there is visible similarity for speakers of non-rhotic dialects (that is, a choice in regards to the POV presentation). I for one wouldn't mind seeing these schwas gone.
• Randomcritic says "Usually [schwa] appears as the replacement for some other vowel in an unstressed syllable; application can be made to other dialects for the original value, though frequently that can also be determined from spelling." If I understand this correctly, this means that about would be transcribed as /eˈbawt/ rather than /əbawt/. My understanding of vowel alternations between vowel alternations such as with vagina/vaginal (sorry, this is the only word that comes to mind whenever I think about this) are allomorphic, not allophonic. Such a representation would be, thus, not phonemic but morphemic. I would rather see us use schwas where dictionaries use them than prompt editors (who may only have a dictionary understanding of pronunciation) to guess at the underlying vowel.
• There seems to be varying interpretations of English stress. Our article on English phonology points to two. The primary, secondary, and unstressed. and another scheme with one degree of stress but with unstressed vowels being reduced or not reduced. There's also a scheme that postulates four levels, but I'm not familiar with it.
• A minor point: Randomcritic says "Even for the most similar dialects, there are notable complications which have not really been considered: for instance, changes in the positioning of stress, the deletion or epenthesis of phones." I'm pretty sure that the number of such words is few and the number of words like this that we would actually need to transcribe (and not confined to List of words of disputed pronunciation) is even fewer. In such instances, we can give both.
• While <ɜ> is unnecessesary, we might want to include it since most systems do so (right?). If we don't, then I would rather see the vowel of perk transcribed as /pərk/ than /pʌrk/.

I'd want to check the value in SSE. It may be /pɛrk/, which would then need to be distinguished from other values neutralized as ɜ or ɝ. RandomCritic (talk) 01:21, 8 April 2008 (UTC)

• My understanding is that words like happy have a neutralization between /iː/ and /ɪ/. While there are dialects that pronounce this as [ɪ], in RP and GA this is [i] and is shorter than the long /iː/. If we don't have the <ː> then we'll need to transcribe this unstressed vowel as /ɪ/ so that readers don't confuse it with an unstressed long /iː/.

— Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 20:49, 6 April 2008 (UTC)

I don't have a problem with either representation -- and by the way, variation in the value of final -y between /iː/ and /ɪ/ is at least four centuries old! If there are dialects which contrast, say, the final vowel in Galilee with that in happily, then of course it should be /ɪ/. RandomCritic (talk) 01:17, 8 April 2008 (UTC)

kwami

First of all, RC, I'm glad to see you engaging in constructive criticism here. I understand you may not have had time to do so earlier.

Pardon me if I misrepresent you below, but you're using impersonal passives for both argument and counterargument, so I might get them mixed up.

1. "Dialect neutral"/"pan-dialectal": True, this does not accommodate all dialects, so your term "multidialectial" is better. It covers the dialects you're likely to see in a dictionary: RP, GA, Oz (except for bad/lad), NZ, Canada. Scottish English does not fit, and no-one has worked out whether South African English, Indian English, or Jamaican English fit. As for "who decided" which dialects were included, simple: The people who worked on it. If no-one here knows a particular dialect, or has a good reference for it, of course they cannot incorporate it into the scheme. If you can improve IPA chart for English and expand it to additional dialects, that would be much appreciated.

2. "Phonemic": You are of course correct that this is not phonemic, nor is it phonetic. It is a generalization of the phonemic inventories of several dialects. I don't know of a word that describes it, though it's common enough in US dictionaries, such as Webster's New Twentieth Century Dictionary of the English Language.

3. "Standard": This only means Wikipedia standard, as established by people working on the MoS. Certainly you can understand that we want to follow a single system, so that our readers aren't put off any more than the have to be. People are constantly whining about us using the IPA at all; having a different scheme for every editor is a disaster, and the results are often not intelligible to readers who don't speak the dialect of the editor. That's why I have several times invited you to join in on deciding what the standard should be.

4. Sources: That's clear enough, though there may be a factual error: Dictionary.com Unabridged is based on the Random House Unabridged Dictionary; it may not be same publisher.

If you wish to change <ɔər> to <oʊr>, fine. We only have it that way through analogy with <ɛər>, <ɪər>, <ʊər>. Personally, I would prefer all of them to be analogous to the non-rhotic vowels: <oʊr>, <eɪr>, <iːr>, <uːr>, but I was voted down by RPers who found that counter-intuitive. The change can be automated if we decide to follow you on this.

I. Phonology: I agree.

II. Dialects: Ideally, I'd want it as you do. However, there is a practical problem here: If we include the Oz bad/lad split, we'll end up misrepresenting most words. That's because few dictionaries indicate this distinction, and therefore few editors will make it when transcribing. We'll end up with a key that claims we distinguish æː from æ, but compliance will be sporadic at best. Same problem with SSE and the fur/fir/fern merger. I don't see any option but to exclude distinctions that will end up being ignored regardless. It would be different if we had control over who adds the pronunciations. We have the same problem with Americans ignoring RP distinctions, and Brits ignoring rhotic vowels (often even when they're phonemic in RP), but we have enough editors of both that this can be easily corrected. This brings us to an additional consideration, accessibility.

Accessibility: When making choices as to which distinctions to transcribe, we should consider whether enough editors control the distinction, or enough sources indicate it, to make its inclusion practical in a collaborative effort like Wikipedia.

Otherwise, I agree with you that simplicity and familiarity should be deciding factors in which symbols to use.

III. Synthesis: I agree.

Application (lede): I agree. However, based on your principle of "more familiar symbols", I oppose the reprentations <aw, aj, oj>. As you've noted, <j> is already counter-intuitive to English speakers, which is worse than being merely unfamiliar like <aʊ, aɪ, ɔɪ>. IMO extending <j> to diphthongs makes the situation even worse. <aw> isn't as bad, but due to pronunciation respellings, many people are familiar with it as /ɔː/.

Long vowels: Again, a practical problem. There are two major uses of the IPA vowels <a, e, i, o, u>, as "long" [ɑː, eɪ, iː, əʊ, uː], and as "short" [æ, ɛ, ɪ, ɔ, ʊ]. Both traditions already found in thousands of Wikipedia articles, often not linked with an IPA template, so we can't even track them down. Therefore I think as a practical measure we should avoid the simple vowels <a, e, i, o, u>, in order to make it clear that those transcriptions do not follow the key. This is a variant of my accessibility concern, though accessibility on the part of the reader rather than the editor.

(P.S. An additional benefit to avoiding <a, e, i, o, u> is that we will easily be able to automate the conversion if we ever decide to change the vowel symbols — say from <oʊ> to <əʊ> or <oː> or some such. Once we reduce the vowels to <a, e, i, o, u> we can never go back.)

Schwas: This has come up before and needs resolution. However, there is a phonemic distinction between full and reduced ɪ, which is sometimes represented as "stress", much as the difference between /ə/ and /ʌ/ is sometimes chalked up to stress. If we do not use ɨ (or some other symbol), how do we make this distinction? Or do we simply ignore it as some dictionaries do?

Classical names: You might be surprised to find that I agree with you here as well. I see that you've made a fantastic start in your sandbox at a description of the method of anglicizing Latin/Latinized Greek names, and would welcome that being made into a full article, replacing the more elementary description I wrote up. I've made only a cursory study of this, have not internalized all the rules, and I know I've made errors of transcription—though your rants on whether the <x> of Thelxinoe should be /z/ or /ks/, when there are no parallels to draw from, your continued spouting off about it long after I'd conceded the point, and your repeated wild accusations of "vandalism" have been entirely inappropriate to civil discourse, so I hope you can understand why I've been curt with you as well. I've asked you several times to show where I've misrepresented vowels rather than merely changed their transcription, but you've refused.

Also, please show where decisions in this system on which vowels to employ depend on non-predictable syllable boundaries. I have yet to see an example. Representing syllable boundaries in English is problematic, so I'd rather avoid it. (I know you've argued that theoretically it is not problematic, but that's a theoretical claim, and counter-intuitive to many English speakers.)

Conclusion: With the exception of the practical problems I mentioned (you are more concerned with theory, whereas I've spent more time trying to achieve some sort of consistency on Wikipedia), and syllable boundaries (which I'm willing to reconsider), I support the incorporation of your suggestions into this page. Perhaps you'd care to draw up a list of which symbols you'd like to see used, and we can all discuss their relative merits, as was originally done to create this scheme.

I propose that when debating this, we restrict ourselves to the arguments presented from this point forward, and not resurrect old arguments which have already been conceded as straw men, which has characterized much of the debate until now. kwami (talk) 21:32, 6 April 2008 (UTC)

Kwami, you say that you've been "voted down by RPers" and you've made allusions to past discussions about Wikipedia's use of IPA. Do you think you could point us to a few of these discussions? I think doing so can inform us of past consensi so that we're not acting alone. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 21:47, 6 April 2008 (UTC)

Woodstone started a discussion here, which involved him, me, and JHJ. (Note that the tables are not in their original form, but were modified as the discussion progressed.) This is one of the later discussions; if I remember correctly, much of the early stuff was on a page that's since been deleted or turned into a redirect. I can't find it off hand. — kwami (talk) 22:30, 6 April 2008 (UTC)

Add R-colored vowel ɚ to the table

ɚ is missing from the table. Help:IPA says that it is pronounced as the "er" in runner. 128.111.207.145 (talk) 21:40, 7 April 2008 (UTC)

It's in the footnotes. kwami (talk) 21:50, 7 April 2008 (UTC)

Is there value in this at all?

Kwami has made a good and thorough, if a bit long winded critique of the IPA. I'm clad someone is looking critically at the subject.

The big problem with the IPA is that everyone must learn to recognize a whole new alphabet of characters before they have any clue how to pronounce a word. I've never met anyone obsessive enough to have memorized these unfamiliar characters on their own. So I find the value of such an alphabet in the context of wikipedia to be almost nil. Wouldn't it be more expedient to render pronunciations is plain speech that normal speakers can clearly understand. For the very few, this "culturally unbiased" system is probably better, but for the rest of us the IPA is absolutely useless.

Wikipedia should be as accurate and deeply rooted as possible, but we have to keep in mind that this is an encyclopedia for common reference and as a starting point for further research. With that in mind it seems the secondary focus, after accuracy, should be clarity to the normal reader. We shouldn't be compelled to learn (or exhaustively reference) an entire alphabet of characters just to figure out how to say W.E.B. Du Bois' name.—Preceding unsigned comment added by 140.233.137.169 (talk • contribs) 06:02, 10 April 2008

User:Randomcritic was critiqueing our use of the IPA (not the IPA itself). While there are a number of characters that have to be learned for the IPA, many of those used in English are either the same used in spelling or are fairly intuitive (and we are working on making our English IPA convention easier for users). Also, and most importantly for your example, the pronunciation of English names and words can be given in both IPA what we call Pronunciation respelling for English. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 19:26, 10 April 2008 (UTC)

links

Another question, do we want the link through "pronounced" or "IPA:", or should the transcription itself be linked to this page? Personally, I think the latter would be much more elegant. We could format it so that there is no underline to interfere with the letters. kwami (talk) 17:25, 11 April 2008 (UTC)

Also, should we have a {{respell}} template the way we have IPA templates, to standardize transcription, or do we want to allow more diversity in representation than that? kwami (talk) 17:34, 11 April 2008 (UTC)

I think "pronounced" is the best option so that those completely unfamiliar with the IPA know what it is they're seeing between brackets.

The point of the respelling key is to allow people to use their instincts on indicating and understanding pronunciation based on their own literacy. I don't think we'd need to outline a standard unless there have been disputes on exactly how to respell a pronunciation. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 17:58, 11 April 2008 (UTC)

Yeah, we already have a standard pronunciation guide, namely this one. There's no point in trying to standardize pro-nun-see-AY-shun GIDEZ; that would defeat their purpose. —Angr 18:16, 11 April 2008 (UTC)

I see you picked up on my idea. The main argument in favour of linking the transcription itself instead of the word 'pronounced' is that the latter link could easily be understood to lead to a page explaining what the word 'pronounced' means. However, I think we should allow editors to decide for themselves on a case-by-case basis, just as we do today with {{pronounced}} vs. {{IPA2}}.

As for respellings, I think they should always be self-explanatory. There is little point in having two transcription systems either of which must be learnt. If no self-explanatory respelling is possible, add a footnote on the pronunciation instead, like I did at Athenry.

As for standardising how self-explanatory respellings are done, I see no point in that.

Timeineurope (talk) 18:30, 11 April 2008 (UTC)

Okay, just a thought. (We'd worked on a standardized key back in the days when so many people were complaining about the IPA that it looked as though we might go another route.)

Timeineurope, I do like your way of linking. Ƶ§œš¹, what if we just charged {{pronEng}} from displaying "pronounced /abc/" to "pronounced /abc/", so that the word "pronounced" was still in there? Then {{IPA-en}} could go from "IPA: /abc/" to a less jarring "/abc/". I believe the reason we linked from "IPA:" in the first place was to avoid underlining the transcription, but that's no longer an issue. kwami (talk) 18:46, 11 April 2008 (UTC)

Genre

Comment originally posted at User talk:Kwamikagami:
That genre can be pronounced with both [ʒ] and [dʒ] doesn't mean that the phoneme /ʒ/ can ever be realised as [dʒ]. Some people won't have [ʒ] in certain positions, but that's not because they realise the phoneme /ʒ/ as [dʒ] in such positions – it's because they use the phoneme /dʒ/ instead. While it makes sense to say that German Rat and Rad, both [ˈʁaːt], are /ˈraːt/ and /ˈraːd/, respectively, because other forms of the lexeme Rad, such as Räder, have [d], I don't find that it makes sense to say that journal and genre – when both are pronounced with [dʒ] – have /dʒ/ and /ʒ/, respectively, because there is no word related to genre that has [ʒ]. Timeineurope (talk) 19:02, 14 April 2008 (UTC)

Comment originally posted at User talk:Kwamikagami:
Change the wording, then. This is the same issue as with /ʍ/. The point is that if someone who does not have initial /ʒ/ (the vast majority of the monolingual population) looks up "genre" in the IPA key that you keep reverting to, they won't know how to pronounce the word. That takes precedence over any philosophical quibbles you may have. kwami (talk) 19:44, 14 April 2008 (UTC)

I changed the wording to "substituted" for both /ʍ/ and initial /ʒ/. kwami (talk) 19:56, 14 April 2008 (UTC)

Comment originally posted at User talk:Kwamikagami in response to the first of Kwami's two comments above; the second had not been posted on his talk page:
No, it's not the same issue. While /ʍ/ and /w/ are separate phonemes only for some people, /ʒ/ and /dʒ/ are separate phonemes for everyone, and so /ʒ/ can't double for /dʒ/ the way /ʍ/ can double for /w/. The IPA key was fine with beige as an example; you were the one who introduced genre as an example of /ʒ/. Since you have now provided a reference that it is in fact pronounced either with /ʒ/ or with /dʒ/, it was obviously a bad example of a word with /ʒ/ and you should never have substituted it for beige. The several dictionaries I consulted indicated that genre was always pronounced with /ʒ/ and so I had no qualms letting it stay as an example of /ʒ/. If those dictionaries were right, everyone would have initial /ʒ/ and so the problem of someone not having initial /ʒ/ not knowing how to pronounce the word wouldn't arise. As it turns out, genre can be pronounced in two different ways (and I have edited the article Genre accordingly), but get this: genre can be pronounced in two different ways not because /ʒ/ is ever realised as [dʒ], but because genre can be pronounced both with /ʒ/ and with /dʒ/. There's no need for anything taking precedence over something else, as it is perfectly possible to combine the necessary level of accuracy with accommodating those monolingual speakers. As long as the false statement about /ʒ/ ever being pronounced as anything else than [ʒ] is removed, everything's fine. The problem for the monoglots would only be that a word, genre, that's pronounced with /ʒ/ by some and /dʒ/ by others, was falsely said to be pronounced with /ʒ/ by everyone. Again, I only let genre stay as an example of a word pronounced with /ʒ/ because several dictionaries say that's the way it's pronounced. Had I known that it can also be pronounced with /dʒ/, I wouldn't have let it stay as an example of a word with /ʒ/. As for philosophy, that doesn't come into it. This is linguistics, it's a science. Linguistic statements can be right or wrong, and yours was wrong, so it had to be removed. Timeineurope (talk) 02:10, 15 April 2008 (UTC)

Some people have initial /ʒ/, most don't. In initial position, it's exactly parallel to /ʍ/, which some people have, and most don't. We need the example, or the chart will be incomplete. kwami (talk) 05:43, 15 April 2008 (UTC)

BTW, we both have 4 reverts in 12 hours. I suggest we let this go until others have their say. kwami (talk) 06:06, 15 April 2008 (UTC)

Two things:

• The main difference between the wh situation and that of genre is that the former is a phonemic merger (or, if you'd like, a consonant cluster reduction) while the latter is a replacement of one phoneme with another. The variation between affricate genre and fricative genre is similar to that of garage, (and possibly consierge). I can't think of any other words with a potentially initial /ʒ/ so I don't know if we can justifiably make a generalization about initial /ʒ/. Maybe this is spelling pronunciation, or maybe it's analogy.
• The point of this guide is to illustrate phonemes with examples. While it would be nice to be exhaustive in indicating the sound's spelling or syllable distribution, we can't really do so without getting straining the ease of the system (which is what separates it from English phonology). beige and vision illustrate the sound quickly and clearly for most speakers of English. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 07:31, 15 April 2008 (UTC)

Some of the ones beginning with <ja> are jabot, jacamar, Jacquerie, jaguarundi, jalousie, jardiniere. Some of these are listed with both pronunciations, some with one or the other depending on the dictionary. However, even the ones listed with only /ʒ/, so that TE would object to including /dʒ/ as an alternate transcription in the entry, will be pronounced with /dʒ/ by most English speakers.

If an entry states that a word begins with /ʒ/, and a reader who doesn't control initial /ʒ/ comes here for help, they won't be able to pronounce the word from the key unless we include this detail.

This isn't a slippery slope. There are only two phonemes which don't occur initially in English: /ʒ/ for many speakers, and /ŋ/ for practically all, so we don't need to do this for any other phoneme. kwami (talk) 08:01, 15 April 2008 (UTC)

I wasn't trying to make a slippery slope argument; each of our footnotes and nuances should be necessary rather than simply informative. I hadn't quite thought out the process for /dʒ/-only speakers but you're right, such speakers may initially be confused about the distinction between the g in gin and the g in genre but the present explanation that the two are different for some speakers will then clarify the issue. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 08:44, 15 April 2008 (UTC)

I guess I'm not following. How do we get that point across, if we only use examples such as vision and beige where these speakers do have /ʒ/? kwami (talk) 08:49, 15 April 2008 (UTC)

Sorry. The footnote that says "When /ʒ/ occurs at the beginning of a word, many people substitute /dʒ/" is appropriate. That's what I meant by "you're right." — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 09:07, 15 April 2008 (UTC)

First, let me reiterate that we can't use genre as an example since it can be pronounced in two ways. All examples have to be perfectly unambiguous. Aeusoes1, there's no more potential for confusion here than with /ɑː/, /ɔː/ and /ɒ/, all pronounced [ɑː] by some speakers.

The fact that some words can be pronounced with both initial /ʒ/ and initial /dʒ/ is a property of those individual words and has got zero to do with the phoneme /ʒ/. There are no words that have initial /ʒ/ where 'many people substitute /dʒ/', like the footnote claims. People who pronounce genre as /ˈdʒɑːnrə/ don't substitute /dʒ/ for /ʒ/ any more than when they pronounce journal with /dʒ/. They simply pronounce genre with /dʒ/, no substitution involved.

If there are words that are pronounced with /dʒ/ by 'most English speakers' even though no trace of such a pronunciation can be found in any dictionary, I would have no objection to including the pronunciation with /dʒ/ in the relevant article if a reliable source could be found.

Finally, I want to ask Kwami: What percentage of speakers of American English do you think pronounce Zsa Zsa Gabor as Ja Ja?

Timeineurope (talk) 16:02, 15 April 2008 (UTC)

TE, Your argument is backwards. No one is claiming that anyone ever substitutes /dʒ/ with /ʒ/. It's just the opposite. And the footnote is worded to be vague about how many people, how often, and for which words so that it could be construed as the property of those individual words.

Dictionary.com lists only the fricative pronunciation for genre (as well as jacquerie) while it gives two for jabot, jaguarundi, jalousie, and jardiniere. Jacamar is given in the affricate pronunciation. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 17:27, 15 April 2008 (UTC)

You misunderstand me. I'm saying that it's being claimed that many people substitute /dʒ/ for /ʒ/, not with /ʒ/. That claim is false, so it has got to go. Timeineurope (talk) 17:37, 15 April 2008 (UTC)

You said "People who pronounce genre as /ˈdʒɑːnrə/ don't substitute /dʒ/ for /ʒ/ any more than when they pronounce journal with /dʒ/." I took this to mean that, since people don't ever pronounce journal with a /ʒ/ that something other than what we're claiming is going on. What are you trying to say? — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 17:41, 15 April 2008 (UTC)

I'm saying that some people pronounce genre with /ʒ/ and some pronounce it with /dʒ/, and those who pronounce it with /dʒ/ don't substitute /ʒ/ with /dʒ/, just like those who pronounce it with /ʒ/ don't substitute /dʒ/ with /ʒ/. And it's not true that the footnote is 'vague about [...] for which words', it indicates that it's for all words. Timeineurope (talk) 12:57, 16 April 2008 (UTC)

And I'm saying this is part of a general substitution of /dʒ/ for /ʒ/ by people who don't control the latter in initial position, even though they may hear others use /ʒ/. What do you hear people say, if not /dʒ/? Or do they simply not use these words at all? kwami (talk) 17:41, 16 April 2008 (UTC)

Of course I agree that people say /dʒ/, but that doesn't mean they are substituting /ʒ/ with /dʒ/ every time they say /dʒ/, as your footnote claims. Any variation in the pronunciation of individual words can be adequately covered in the individual articles. If, for a given word, it's impossible to find a source for a pronunciation with /dʒ/, then I see no reason why we should include such a pronunciation. Timeineurope (talk) 20:29, 16 April 2008 (UTC)

Webster's and the OED, on the other hand, have only the fricative for jacamar. Webster's has only the fricative for jaguarundi, but only the affricate for jalousie, whereas the OED has only the fricative for jalousie, jabot, and jardiniere, so which we get is really the luck of the draw. I'm guessing that many words consistently listed with only the fricative, such as jacquerie, tend to only be used by people who are familiar with French.

TE, you're right about Zsa Zsa. I asked my mom and got /zɑːzɑː/.

All the more reason to remove the footnote, then. Timeineurope (talk) 12:57, 16 April 2008 (UTC)

It's obviously a foreign word to my mom. /zɑːzɑː/ might be a spelling pronunciation.

Look, this is simple: Your footnote implies that there are people who pronounce Zsa Zsa (which must be an English word, since it's not Hungarian) with /dʒ/. No source has been offered for this claim. I don't need a consensus to remove an unsourced statement. Timeineurope (talk) 20:29, 16 April 2008 (UTC)

When I gave her /ʒɑːʒɑː/ she could say it, but was somewhat uncomfortable with it, and really exaggerated the enunciation, the way she does when she tries to pronounce Spanish words correctly. She commented on what an odd sound that is. I got her to say Jacques Cousteau with the fricative, but as soon as she repeated it, it came out /dʒɑːk/. kwami (talk) 18:04, 15 April 2008 (UTC)

I think also relevant is the special attention introductory French textbooks give to this sound. They don't just say that French g and j are pronounced like the g in beige, but go on in some detail on how to practice putting that sound at the beginning of a word such as Jacques. This is clearly not natural to a great many English speakers. One of the errors beginners make is substituting [dʒ] for [ʒ] at the beginnings of French words. kwami (talk) 18:26, 15 April 2008 (UTC)

If it comes to that, English speakers sometimes have difficulty making sounds in positions where they do occur in English. When I was learning Welsh, I was astonished to observe how many English speakers had difficulty pronouncing /ð/ at the beginnings of words, even though it occurs word-initially in English. (My working hypothesis is that English speakers without training can only pronounce /ð/ at the beginning of unstressed function words but have to really work to learn to put it at the beginning of stressed lexical words in phrases like y ddynes [ə 'ðənes] or yn dda [ən 'ða:].) —Angr 21:53, 15 April 2008 (UTC)

TE, you don't have any consensus to remove the footnote. Let's see if we can get people to agree before we start another edit war. kwami (talk) 17:41, 16 April 2008 (UTC)

I'm getting really confused here. Not all of the following can be true:

1. There are words where everybody says [dʒ]

2. There are words where some people say [dʒ] and other people say [ʒ]

a. Speakers who pronounce [dʒ] are doing so because of a phonological or diachronic change that turns/turned word-initial [ʒ] into [dʒ]

b. Such speakers are unable to pronounce [ʒ] in the syllable onset

c. ??!

3. There are words where everybody says [ʒ]

We know 1 and 2 for certain. I'm pretty sure 3 is false. I'm putting my money on 2a being the explanation for 2 and 2b being probably an oversimplification. I have no idea what TE is arguing but let's call it 2c. What is 2c, TE? — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 19:21, 16 April 2008 (UTC)

Speakers who pronounce /dʒ/ aren't doing so because they are substituting /ʒ/ with /dʒ/ every time they say /dʒ/, as the footnote claims. Many words have more than one pronunciation, but that, and why, is not a subject we cover at Help:Pronunciation. Timeineurope (talk) 20:29, 16 April 2008 (UTC)

Please address my question and stop edit warring. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 20:55, 16 April 2008 (UTC)

I have never expressed an opinion on why there are words where some people say [dʒ] and other people say [ʒ], I don't even have one, so I have no answer to your question. What the footnote is currently saying is that words like genre are pronounced with /ʒ/. They're not; they're pronounced either with /ʒ/ or with /dʒ/. The footnote then goes on to say that when people pronounce those words with /dʒ/, they are performing a substitution of /ʒ/ with /dʒ/. I'm saying they're not; they pronounce genre as /ˈdʒɑːnrə/ because that's how that word is stored in the lexicon of their idiolect. If you want to state something to the effect that they are performing a substitution every time they say a word that gets pronounced with initial /ʒ/ by some people, you need to source that. See below for further criticism. Timeineurope (talk) 00:37, 17 April 2008 (UTC)

You can make the same argument for /ʍ/. Use whichever wording you like. kwami (talk) 01:37, 17 April 2008 (UTC)

I suggest something slightly different: English never had initial /ʒ/, except for some people when pronouncing foreign borrowings with [ʒ], rather like final [x] in words such as Bach in those dialects which don't otherwise have /x/. But the reason for (2) is not important here, we just need to let people know it happens.

The article currently gives genre as an example of a word with /ʒ/, even though it can also be pronounced with /dʒ/. Every example in the key to a phonetic alphabet needs to be perfectly unambiguous.

But they aren't unambiguous. What is either /ʍ/ or /w/, analogous to genre. kwami (talk) 01:37, 17 April 2008 (UTC)

It's not analogous. In our system, /ʍ/ is whatever sound people use in what, be it [ʍ] or [w], while /ʒ/ is always [ʒ] and /dʒ/ is always [dʒ]. Timeineurope (talk) 17:35, 17 April 2008 (UTC)

In our system, with the footnote, initial /ʒ/ is whatever sound people use in genre, be it [ʒ] or [dʒ]. kwami (talk) 18:58, 17 April 2008 (UTC)

That's a bad idea, among other things because it would only apply to initial /ʒ/. It would be too confusing for readers to have to interpret the same symbol in different ways according to where in the transcription the symbol is used. Also, the footnote is plainly incorrect, as no source has been offered for a pronunciation of Jacquerie with initial /dʒ/. That word is clearly not pronounced in French, with [ʁ] and so on, when used in English sentences, and so its pronunciation has to be written between slashes. Timeineurope (talk) 17:06, 20 April 2008 (UTC)

The article is currently implying that some speakers pronounce jacquerie and Zsa Zsa with /dʒ/. No source has been offered in support of this claim. In fact, several sources have been consulted about the pronunciation of jacquerie and every one of them says it's pronounced with /ʒ/, period.

Zsa Zsa is not an English name. Foreign names often violate English phonotactics—Vladimir, for example. (English does not allow initial /vl/.) The OED lists Jacquerie as an unassimilated loan. It states,

"In a few French words, distinctly recognized as alien, j has the French sound (ʒ), as in déjeuner, jeu d'esprit."

The same situation applies to /y/, /ø/, and nasal vowels. Better to cover it here, than to rely on an unknown number of articles. kwami (talk) 01:37, 17 April 2008 (UTC)

Zsa Zsa is not Hungarian, so it must be English, and as an English word it has an English pronunciation, namely /ˈʒɑːˌʒɑː/. As long as we give the pronunciation of Zsa Zsa between slashes, it's relevant to the footnote.

English does actually allow initial /vl/, as in the completely English word vlog.

As for Jacquerie, it's an English word if it's used in English sentences without code-switching. For it not to be an English word, people would have to code-switch to French every time they used the word in a sentence. Again, as long as we give a pronunciation between slashes, it's relevant to the footnote. Timeineurope (talk) 17:35, 17 April 2008 (UTC)

Actually, Zsa Zsa *is* Hungarian. It's baby talk for Sari.

No, the Hungarian is apparently Zsázsa. Zsa Zsa is English.

I agree the slashes would be problematic if we used them, but we don't.

We should, however. Zsa Zsa has an established English pronunciation. Established English pronunciations go between slashes. We'll want to give the pronunciation of Zsa Zsa generally considered correct in the United States, and that's /ˈʒɑːˌʒɑː/. We can't put ˈʒɑːˌʒɑː between square brackets, because it's not a Hungarian pronunciation. It's an English pronunciation, and as such it goes between slashes. Then the footnote implies that some people pronounce it /ˈdʒɑːˌʒɑː/, and that appears not to be true.

Remember that this key is for all words whose pronunciation is given between slashes, regardless of their degree of Englishness. Timeineurope (talk) 17:06, 20 April 2008 (UTC)

By your logic, English has nasal vowels, because people say vin blanc. That is what "unassimilated loan" means. It's a basic concept in languistics; many, perhaps most, languages have such sounds. French, for example, has /ŋ/ in English loans, for which many people substitute /ɲ/. Japanese has English loans which violate Japanese phonotactics. Etc. etc. etc. Nothing unusual about this. kwami (talk) 18:58, 17 April 2008 (UTC)

We don't 'need to let people know it happens', especially since, per the available sources, it doesn't. As I wrote above: Any variation in the pronunciation of individual words can be adequately covered in the individual articles. If, for a given word, it's impossible to find a source for a pronunciation with /dʒ/, then I see no reason why we should include such a pronunciation. Timeineurope (talk) 00:37, 17 April 2008 (UTC)

TE, we do cover the subject of different people pronouncing words differently, for example with /ʍ/ and final /i/. If you don't like "substitute", change the wording. That isn't a reason to delete the footnote. And you can repeat yourself twenty times; if we don't understand what you mean the first time, we're not going to understand the twentieth, unless you explain what you mean. kwami (talk) 21:50, 16 April 2008 (UTC)

The footnote concerning final /i/ is about transcription conventions, not pronunciation. While we do note that some people pronounce both /ʍ/ and /w/ as [w], we don't note that some people pronounce all of /ɒ/, /ɔː/ and /ɑː/ as [ɑː], and no need is felt. Timeineurope (talk) 00:37, 17 April 2008 (UTC)

I think a footnote is necessary because this phoneme has a bit of quirkiness in English. Here's the main problem with the "beginning of a word" argument: if it were so that English speakers cannot or do not pronounce word-initial /ʒ/ then Zsa Zsa would be /dʒɑːʒɑː/. And think about homage. It has a variety of pronunciations ranging from foreigny /oʊˈmɑːʒ/ to Englishy /ˈhɒmɪdʒ/. I suspect that, rather than being the result of a complicated diachronic change resulting in a phonological or phonemic process, the situation with this word is the same for genre and other words beginning with /ʒ/:

i. Borrowed words from French indicate /ʒ/ orthographically with the same letters we use for /dʒ/. There is no spelling rule in English that a certain letter or digraph indicates /ʒ/ and not /dʒ/ word-initially.

ii. Time or distance legitimizes the /dʒ/ pronunciation; we have no way to determine if this is the process of innovation from children (i.e. diachronic change) or simple spelling pronunciation.

We could reword it to say "a number of words are cited as having two pronunciations with either /ʒ/ or /dʒ/." It's possible that we could add "speakers tend to have difficulty with word-initial /ʒ/" but that goes too far into OR speculation. Putting emphasis on citation also restricts the examples of alternate pronunciations to those given two pronunciations in one or more dictionaries. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 04:12, 17 April 2008 (UTC)

It should be easy enough to find a source supporting the claim English speakers have trouble with initial /ʒ/. I remember trying to learn to do it in high school after reading in a French textbook that it doesn't occur in English. kwami (talk) 04:25, 17 April 2008 (UTC)

[2]? See page 62. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 04:49, 17 April 2008 (UTC)

This reference says the opposite—it says that English speakers have no trouble with word-initial or syllable-initial [ʒ]. Spacepotato (talk) 19:25, 17 April 2008 (UTC)

(unindent)All right, I've edited the footnote accordingly. I'm sure it could do with some rewording but we have a source (Kahn 1976) that says:

"Now note that although there are no native words with initial /ž/, the slow-speech syllabication of pleasure appears to be [ple.žər], not [plež.ər], just as pressure is [pre.šər], not [preš.ər]. Thus there can be no restriction against syllable-initial /ž/. Although it might be supposed that there exist a constraint against word-initial /ž/, independent of the syllable-structure constraints, it seems more likely that a simple gap is involved, i.e., that it is accidental that there are no words with initial /ž/. In support of this analysis, note that the ordinary speaker of English pronounces the initial /ž/ of the Russian name Ženya with no apparent difficulty, while initial [ŋ] is accessible only to the student of phonetics."

Thus are inclined not to say that this phenomenon has anything to do with phonotactic constraints. — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 19:57, 20 April 2008 (UTC)

Since according to this analysis genre is not an unambiguous example of /ʒ/, I've removed it from the key. As for readers who wish to find the pronunciation of "genre" itself, there is no issue as the article correctly indicates that it can be pronounced either /ˈʒɑːnrə/ or /ˈdʒɑːnrə/. Spacepotato (talk) 20:27, 20 April 2008 (UTC)

But it isn't just an issue of that one word. All assimilated loans starting with /ʒ/ are pronounced /dʒ/ by a large portion of the population, except maybe reduplicated Zsa Zsa. Yes, you can get people to pronounce /ʒ/ if they concentrate, which you cannot do with initial /ŋ/, but then you can also get people to pronounce final /h/, or initial /ts/, and both violate phonotactic constraints in English. The point is that once you leave them alone, they won't pronounce it that way. Garage is idiosyncratic: it varies in a way that beige does not, but genre varies like every other /ʒ/-initial word (except Zsa Zsa, with its spelling pronunciation /ˈzɑːzɑː/), so it's completely typical. kwami (talk) 22:44, 20 April 2008 (UTC)

This is the part of the discussion where you ought to cite sources. Heck genre (with two pronunciations) has been in the language for almost 250 years while gelatin (with one) for 200. What's to say that elitists haven't reintroduced the /ʒ/ pronunciation of the former in order to make it sound Frenchier? — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 22:59, 20 April 2008 (UTC)

That could be! Whatever the historical order, those sounds alternate in initial position. The exceptions are a proper name and words which the OED marks as unassimilated loans. Elementary French textbooks introduce English-speaking students to a new idea: that a word can start with /ʒ/. (Sorry, no, I don't have the ref. That was a long time ago.) kwami (talk) 23:49, 20 April 2008 (UTC)

I hope we can find a ref that goes into this more. Kahn (1976) really mentioned this more in passing. By the way, final [h] and initial [ts] are about as difficult for English speakers as initial [ŋ] (at least, from what I can tell). — Ƶ§œš¹ _{[aɪm ˈfɻɛ̃ⁿdˡi]} 01:04, 21 April 2008 (UTC)